ims health rwe accesspoint 7 - november 2013

AccessPointNews, views and insights from leading experts in RWE and HEOR

IMS HEALTH ECONOMICS AND OUTCOMES RESEARCHOUTCOMES - Issue 1 Page 1

Carl de Moorconsiderschallenges ofimmortal time biasPage 34

Mike Nelsonvalidates trial-basedhealth economicanalysis using RWEPage 16

Nathalie Grandfilsexplores guidelineadherence with IMSDiabetes CohortPage 38

Key issues formarket accessin India

European pricecomparison

VOLUME 4, ISSUE 7NOVEMBER 2013

IMS REAL-WORLD EVIDENCE SOLUTIONS ANDHEALTH ECONOMICS & OUTCOMES RESEARCH

The promise of predictiveanalyticsA new crystal ball for healthcare?

Bias challengesin epidemiology

Designing outuncertainty

Adaptive techniquesimprove efficiency

The South-West testTechniques to interpret

cost-effectiveness ratios

AccessPointNews, views and insights from leading experts in RWE and HEOR

Interpreting cost-effectiveness resultsWhy ICERs in the South-West quadrant should not be ignored

page 43

Accounting for bias in epidemiologyApproaches for strengthening the validity of studies

pages 28 & 34

Innovative designs in observational studiesAdjusting for uncertainty with adaptive techniques

page 22

Fulfilling the promise of predictive analyticsWhy a major methodological shift is key

page 10

VOLUME 4, ISSUE 7NOVEMBER 2013

A roadmap for healthcare access in IndiaHow action across four key dimensions can reduce out-of-pocket spending

page 48

IMS REAL-WORLD EVIDENCE SOLUTIONS & HEOR

Contents

NEWS2 INTEgRATED EVIDENCE HUBS3 LINKED DATA COMPLETES PATIENT PATH6 DEEPER INSIgHTS FOR SAFETY7 IMS INSTITUTE 2017 OUTLOOK8 EVALUATINg PHARMACY INTERVENTIONS9 FORMALIZINg HTA IN LATIN AMERICA

INSIGHTS10 PREDICTIVE ANALYTICS

Fulfilling the potential16 IMS HEALTH SYMPOSIUM

Is RWE making a difference?22 OBSERVATIONAL METHODOLOgY

Innovative approaches to uncertainty28 CHALLENgES IN EPIDEMIOLOgY

Adjusting for bias and confoundingCorrecting immortal time bias

38 FRENCH DIABETES COHORTSupporting best practice

43 INTERPRETINg ICERs Issues in the SW quadrant48 HEALTHCARE ACCESS IN INDIA

A roadmap for action54 PHARMA PRICE COMPARISONS France impacts EU price gaps

PROJECT FOCUS60 VENOUS THROMBOEMBOLISM

RWE validates economic analysis63 ALZHEIMER'S DISEASE

Evaluating alternative Medicare data

IMS RWES & HEOR OVERVIEW66 ENABLINg YOUR REAL-WORLD SUCCESS

Solutions, locations and expertise

WelcomeThank you for reading the latest issue of AccessPoint – IMS Health’s periodic journal that explores issues shaping theHEOR, safety and RWE landscape. In this edition, we are focusedon innovation across several themes:

• New realities in generating data. The life science historicalmodel of ‘reactive data’ and dataset fragmentation needs tobe revisited. Through strategic planning and new databasetechniques, companies can now plan ahead for their evidenceneeds, creating proactive hubs that can increase the depth andbreadth of analytics, accelerate timelines and fully tap into thepower of RWE.

• Accelerating demand for evidence in many forms. A newEMA framework will shape drug safety monitoring in EU whileLatin American countries look to increase the importance ofHTAs. Market-level initiatives such as the UK’s CommunityPharmacy Future (CPF) indicate a vision for using RWE-generated insights to establish more strict guidance onhealthcare practice. And interest in using analytics to solvesome of the toughest global challenges from France to Indiareflects its boundless potential.

• Methodologies and technology are changing the deliveryof evidence and, potentially, the role of HEOR. We spotlightpredictive analytics, a newly developed dynamic data andanalysis platform, and adaptive designs as examples of driversincreasing the reliability and value of RWE. We also dedicatetime to reflecting on insights to ensure the best quality fromepidemiology and other studies.

• IMS Health continues to invest in enabling further RWEadvances. With over 230 multi-disciplinary experts focused onRWE solutions and HEOR globally, we continue to enhance ourRWE platform of scientifically-validated datasets throughstrategic data sourcing, sophisticated linkage and powerfulevidence technologies. Our new disease-focused datasets andpatented approaches to develop unprecedented tracking ofUS patients across their treatment journey, and standardizedDUS are a few profiled examples.

We at IMS Health are passionate about improving patientoutcomes and advancing healthcare at every level. I hope youenjoy AccessPoint and invite you to share your views with us.

“Leading companies are leveraging advances in evidence to turn the unknown into actionable insights.”

Jon ResnickVice President and general Manager Real-World Evidence SolutionsIMS [email protected]

AccessPoint is published twice yearly by the IMS Real-World Evidence (RWE) Solutions and Health Economics & Outcomes Research (HEOR) team. VOLUME 4, ISSUE 7. PUBLISHED NOVEMBER 2013.

IMS HEALTH 210 Pentonville Road, London N1 9JY, UK Tel: +44 (0) 20 3075 4800 • www.imshealth.com/[email protected]

©2013 IMS Health Incorporated and its affiliates. All rights reserved.Trademarks are registered in the United States and in various other countries.

ACCESSPOINT • VOLUME 4 ISSUE 7 PAgE 1

NEWS | INTEgRATED EVIDENCE HUBS

continued on next page

PAgE 2 IMS REAL-WORLD EVIDENCE SOLUTIONS & HEOR

Integrated evidence hubs drive timely, cost-efficient insights

Dynamic data and analysis platform enables value and evidence from the real world to meet growing post-launch requirements.

Achieving and maintaining market access anddriving competitive differentiation for newtherapeutic agents requires a comprehensiveapproach to meet post-launch evidencerequirements. A newly developed dynamic dataand analysis platform is providing a solution tothese continuous evidence demands. These platforms integrate data from a variety of existingand new real-world sources to enable assessment andmonitoring of treatment patterns, clinical outcomes andsafety, centered on disease areas of interest (see Figure 1 forthe range of platforms IMS Health is collaborating withclients to build and run). By integrating and linking, where relevant, patient-leveldata from multiple sources, the evidence hub overcomesthe limitations of assessing outcomes using a single datasource. This distinguishing feature allows specific businessand research questions to be addressed that can only beadequately answered by identifying and integrating datasources that fully reflect the real-world patient journey.

VALUE ALREADY BEING REALIZEDThe integrated platforms have delivered early andcontinuous value to franchise teams across a number ofdimensions, such as:

• Novel studies & external credibility: Execution ofobservational studies, for conference abstracts andpublications, centered on disease knowledge, patternsof care, clinical and safety outcomes, and comparativeeffectiveness (CE) research, with a recent chronic diseaseplatform leading to five publications within the firsteight months of platform creation.

• Internal validity & alignment: Monthly delivery of over20 analytical metrics (in an excel dashboard) on data tiedto key brand and market performance indicators, suchas evolving dynamics in treatment switching, line oftherapy, and discontinuation; relapse benchmarkingand population profiles.

• Timeliness: Ability to respond to payer requests foradditional CE evidence in less than two weeks, therebyavoiding costly market access delays.

•Cost efficiencies: Efficient hypothesis generation and testing.

WHERE TO NEXT?After the foundational data integration and platformcreation phase, the next question becomes how to scale thedata and analysis capabilities:

• Scaling data access through strategic sourcing:A core capability of these platforms is IMS Health’s abilityto strategically source deeper and richer clinical dataacross geographies to address evolving client needs andfill existing data gaps. Depending on the specific therapy area, this can include gaining access tostructured clinical data on disease severity, progression,or outcomes from specialty clinics to link to existingclaims based in the evidence hub. It can also includesourcing clinical endpoints and attributes fromspecialists or specialty clinics across a number ofgeographies to link to existing administrative datasets.

FIGURE 1: EXAMPLES OF LINkED DATA MARTS DEVELOPED BY IMS HEALTH

Hematology/Oncology

Nephrology

Neurology 1

Ophthalmology

Neurology 2

Oncology

Diabetes

Pharmacy claims

Medical claims

PharMetrics Plus

Electronic Medical Records

Laboratory data

Hospital charge data master

Mortality data, patient surveys


COMPREHENSIVE DISEASE RECORDS | NEWS

New linkage of patient data heralds significant disease-focused research potential withpreviously unavailable clinical detail on the patient continuum of care in the US.

IMS Comprehensive Disease Records complete US patient journey

As a committed partner in the life sciences, IMS Health has invested in creating disease-specificofferings that combine patient-level data across US settings of care for researchers focused onspecific therapy areas. Leveraging IMS’ patented, HIPAA-compliant matching technique, newComprehensive Disease Records make it possibleto track patients across their treatment journey inthe US, providing data for research questions thatpreviously could not be answered (Figure 1 overleaf). The move comes at a time of growing reliance on RWE formarket access, and consequent expansion of patient-leveldatasets and the technologies to extend their application.Today, more and deeper data than ever is available acrossthe treatment pathway, providing valuable clinicalinformation as patients interact with multiple touch pointsin different settings. While administrative claims serve as the thread connectinginformation on multiple services used, the clinical detail fromthe site of care is often lost. The details from each setting ofcare are then aggregated separately. For example, lab test

values are housed in a separate clinical warehouse from themedications filled at the pharmacy for a given patient.The fragmentation of data for a single patient persists dueto the aggregation of specific types of information byindividual data suppliers and the use of distinct data modelsthat are not interoperable. From a researcher’s perspectivethis has resulted in dependence on single data sources for agiven patient, preventing a comprehensive and detailedview of the patient continuum of care. Traditionally, the alternatives for the level of detail that willnow be achievable through Comprehensive DiseaseRecords have been chart reviews and prospective datacapture. Expensive and time consuming, with multiple siteagreements and contracts, these approaches are typicallylow yield in terms of patient numbers and their ability to benationally representative.Integrated delivery systems, comprising multiple datasetslinked together probabilistically, are a further option,offering the breadth of clinical information for a patient andthe coverage of settings of care. However, they are alsoexpensive and typically based on localized health plans. Inboth cases, limitations on access and use of the data can bea challenge.


INTEgRATED EVIDENCE HUBS - continued from previous page

For further informationon IMS integrated evidence hubs and examples ofrelated publications, please email Ian Bonzani [email protected]

• Scaling analytical capabilities through technologyand custom IMS analytical support model: IMS Healthhas dedicated a comprehensive HEOR/RWE team ofobservational data and research scientists to driveinnovative study and analysis throughput in a cost-effective way. In addition, the company continues toinvest in tools to help clients maximize their investmentsin RWE. The IMS Evidence 360 Interrogation Suiteincorporates searchable references for patient-level dataassets, cohort definition and reporting features tosupport seamless interaction with real-world data fromstudy conception to execution.

• External engagement: The successful curation of thisstrong foundational evidence platform paves the wayfor companies to create new opportunities for externalengagement through, eg, collaborative researchprojects with KOLs and disease networks, design ofdecision support tools for physicians, and operationalvalue-based payer pricing models.

SPECIALISTOFFICE

PHYSICIANOFFICE (GP)

• Age, gender, vitals• Date of diagnosis• Co-morbid condition, concomitant medications• Medications prescribed

Patient enters healthcare system

Patient exits healthcare system

• Medication !les (type, strength, days supply)• Treatment switch• Patient copay & cost

• Tumor staging• Biomarkers• Sites of metastasis• Performance status

• Date of mortality

• Reason for admission• Length of stay• Diagnosis, procedures, surgeries• Cost, charges

• Tests performed (HbA1c, Lipids, etc.)• Test results and interpretation

Within each setting, valuable clinical information is collected for a given patient

IMS Health has access to data across multiple settings of care which it can integrate at the patient level in a HIPAA-compliant manner

IMS Comprehensive Disease Records integrate data from di"erent sources for currently four disease areas

Patient treatment journey

STAND-ALONELABORATORY

PHARMACY

HOSPITAL

OVER-THE-COUNTER

MEDICATION

MORTALITYDATA

ADDITIONAL TREATMENTACUTE EVENTINITIAL TREATMENTINITIAL DIAGNOSIS DISEASE PROGRESSION PATIENT DEATH

T

TYPE 2 DIABETES ACUTE CORONARY SYNDROME BREAST CANCER NON SMALL-CELL LUNG CANCER

Oncology EMR

Mortality data

Health plan

claims EMR

Hospital charge master

Mortality data

Health plan

claims EMR Health

plan claims

Mortality data

Health plan

claims Oncology

EMR


Comprehensive Disease Records offer the same level ofclinical detail as integrated delivery systems, cover moresettings of care than prospective data capture, are nationallyrepresentative of the commercially insured population andbring the assurance of data that is drawn from the samepatients without usage or access restrictions. Multiple-useaccess permits cost savings over repeated purchases ofmultiple data assets, and pre-defined linkages createefficiency in data processing and spend.

MARKET-LEADING DATA & UNIQUE LINKAGEIMS Health is an established leader in patient-level dataofferings. Its broad suite of anonymized datasets spansmultiple care settings, with an unmatched volume of livescaptured from electronic medical records (EMR), hospitalchargemasters, and pharmacy and medical claimsdatabases. IMS LifeLink PharMetrics Plus™ is one of theindustry’s largest health claims data assets, comprisingadjudicated claims for more than 150 million uniqueenrollees across the US.

Strengthening the individual value of IMS Health dataassets is the company’s unique ability to seamlessly linkthem together using a patented and HIPAA-compliantlinked process. The company and its technology partnersdeploy proprietary encryption tools to de-identify patientdata at the source. De-identified patient data thenundergoes a deterministic matching process to be assignedto a unique and persistent IMS Patient ID. IMS Health is one of the few companies that is able to usedeterministic matching, employing actual patient-levelinformation to generate a match, thereby ensuring thatlinked records are truly for the same patient. By linkingdisparate data sources in this way, it can uniquely recreatethe original treatment journey while maintaining patientanonymity, to bring unparalleled insights into theirexperience across multiple channels. IMS Health linked data has already formed the basis of peer-reviewed articles in leading journals, and data martshave been developed for many of the top 20pharmaceutical manufacturers (Figure 2).

NEWS | COMPREHENSIVE DISEASE RECORDS

FIGURE 1: PATIENTS INTERACT WITH MULTIPLE SETTINGS OF CARE

Continued from previous page


IMS COMPREHENSIVE DISEASE RECORDS Comprehensive Disease Records comprise integrated,disease-specific, nationally representative datasets of rich,clinical patient information from multiple relevant caresettings linked deterministically. Initially focused on fourtherapy areas – Breast Cancer (BC), Acute Coronary Syndrome(ACS), Non Small-Cell Lung Cancer (NSCLC) and Type 2Diabetes (T2DM) – the offering brings significant gains overthe traditional data access model based on siloed datasets:

• Unmatched, comprehensive patient view across settingsof care for diseases

• greater volume of patients than is possible to obtainthrough chart reviews, prospective studies, local HMOsand other comparable detailed information feeds

• Unique variables not typically available in claims or otherclinical data (eg, mortality, HER2/ER/PR)

• New constructed variables specific to therapy areaconcerns

• Ability to answer disease-pertinent questions that couldnot be addressed prior to this type of data integration

Rich, relevant data sourcesEach dataset draws on the sources most relevant to thedisease in question. In the case of BC and NSCLC, forexample, information from oncology EMR and health planclaims yields important insights into tumor staging, markerstatus and other clinical characteristics which impacttreatment selection; the way in which care may differ by

stage, setting and physician; and long-term treatmentpatterns and resource use. In ACS, where hospitalization iscommon, hospital charge master information providesexceptional detail on inpatient medications, devices, andprocedures, which can then be linked to other treatmentand outcomes through EMR and claims data. And for T2DM,the linkage of detailed EMR with claims data enables, forexample, an understanding of the impact of treatmentagainst target outcomes, such as HbA1c levels, theimplications of diabetic complications, as well as insightsinto the patient’s overall health condition.

FUTURE EXPANSIONMarking a commitment to continuous innovation, IMS Healthhas plans to expand Comprehensive Disease Records toadditional therapy areas going forward. Work is alreadyunderway to refine and validate the standard data modelthrough clinical partnerships. The goal is to then upload themodel to a user-friendly data platform along with standardanalytics that can make the data accessible to a broad, non-technical audience. Launch of the first four disease areas isplanned for year-end 2013. Delivery options will includecustom studies, single or multiple use extracts.

For further information about IMS ComprehensiveDisease Records, please email Shibani Pokras [email protected]

COMPREHENSIVE DISEASE RECORDS | NEWS

FIGURE 2: EXAMPLES OF RECENT PEER-REVIEWED jOURNAL ARTICLES LEVERAGING IMS HEALTH LINkED DATA

• The disease burden of pertussis in adults 50 years old and older in the United States: A retrospective study

• Treatment patterns: Targeted therapies indicated for first-line management of metastatic renal cell carcinoma ina real-world setting

• Clinical burden of illness in patients with neuroendocrine tumors

• Cost of palliative radiation to the bone for patients with bone metastases secondary to breast or prostate cancer

• Risk of cardiovascular events in patients with neuroendocrine tumors• Risk of hepatic and gastrointestinal events in patients with neuroendocrine tumors

• Metastatic renal cell carcinoma: Patient characteristics and recent treatment patterns in real-world practice• Patient compliance with metastatic renal cell carcinoma treatments

• Risk of treatment failure after first-line tyrosine kinase inhibitors (TKI) therapy in patients with metastatic renalcell carcinoma

• Risk of anxiety/depression events in patients newly diagnosed with neuroendocrine tumors• Risk of osteoporosis/osteopenia events in patients newly diagnosed with neuroendocrine tumors

• Metastatic renal cell carcinoma: Patient characteristics, treatment patterns and schedule compliance in clinical practice

• Comparative antibiotic failure rates in the treatment of community-acquired pneumonia (CAP): Results from a claims analysis

TITLE


Enhanced, syndicated drug utilization studies enable more confident and efficientmonitoring of appropriate prescribing in Europe.

New approach brings deeper insights for improved risk minimization

At a time of heightened and increasing emphasison drug safety surveillance, conditional approvaland risk management, IMS Health has announcedthe development of enhanced drug utilizationstudies (DUS) in Europe. Recent legislation has established a clear regulatoryframework for drug safety monitoring in Europe, with thepublication of good Pharmacovigilance Practice (gVP) bythe EMA.1 Shifting the focus from the detection andelucidation of unknown risk to the control of known risk,the new regulation particularly stresses the importance ofassessing the effectiveness of risk minimization measures(RMMs). These assessments are now a prerequisite of riskmanagement plans (RMPs).

STANDARDIZED, SYNDICATED DUSBased on a standardized and syndicated approach,leveraging proprietary and external data sources across all European markets, these build on the company’strack record of successful collaboration with the EuropeanMedicines Agency (EMA) and ENCePP on protocols and studies.DUS provide simple metrics for monitoring appropriatedrug use and thus the implementation of RMMs – one ofthe key elements covered by the gVP mandate. Researchersat IMS Health have observed a near doubling of EMArequests for DUS every two years.2 Automated multi-country patient-level databases are used disproportionallymore in understanding patterns and monitoring ofappropriate prescribing. This is a consequence of theincreased availability of RWE data at the patient level (eg,databases based on EMR or claims as well as registries)coinciding with a focused and repeated search for similar,relatively basic data elements such as age, dosage andindication.

IMS Health is able to provide such data for most of the EU,using foremost proprietary and limited third-party dataassets. These approaches have been standardized withprotocols and SOPs validated by authorities in earlierstudies. This standardization also allows the developmentof syndicated approaches for multiple clients.

SIGNIFICANT BENEFITSEnhanced DUS bring major benefits for government andindustry alike, delivering deeper insights through largerpatient pools, enabling greater confidence in the quality ofplans and results, and serving as a foundation for otherstandard approaches, including PAES (Post-AuthorizationEfficacy Studies).

While reducing the burden on health authorities forapproval of plans across industry, they pave the way formanufacturers to gain from a more streamlined process ofapproval for regulatory plans as well as cost savings indeveloping real-world safety evidence. The standardizeddatabase approach offers higher speed, reduced cost andclear priori alignment between industry and regulators onthe chosen study method and data asset validity. The costadvantages are amplified when a syndicated approach ischosen. The syndicated approach also offers easiercomparison for authorities.

HERITAGE OF EXCELLENCEWith its longstanding experience in drug safetysurveillance, access to the most content-rich, validatedproprietary and third-party datasets, and extensive networkof relevant scientific affiliations, IMS Health is uniquelypositioned to design and execute DUS:

• Rigorous methods: Member of the ENCePP andcommitted to excellence in research through adherenceto the ENCePP guide on Methodological Standards.

NEWS | SAFETY MONITORINg

1 European Medicines Agency. good pharmacovigilance practices. 2013. Accessed 7 October, 2013 at:http://www.ema.europa.eu/ema/index.jsp?curl=pages/regulation/document_listing/document_listing_000345.jsp

2 Schroeder C, Keja A, Hughes B, Toussi M. Understanding patterns of monitoring of appropriate prescribing in Europe. Poster presented at ISPOR 16th Annual EuropeanCongress, Dublin, Ireland, 2-6 November, 2013

continued opposite, below


IMS INSTITUTE INSIgHTS | NEWS

For further information on IMS Drug Utilization Studies, please email Birgit Ehlken at [email protected]

• Unparalleled data: Leverages internal and externalpatient-level databases for DUS in a combined approachand is a partner of the EU-ADR network, which sharesdata from around 50 million patients in Europe.

• Proven credibility: IMS Health databases and protocolsare accepted and used for DUS by health authoritiessuch as EMA.

• Established expertise: Long history of working withsensitive data and proven expertise in such fields as datacuration, linkage and integration, and data qualitymanagement. IMS Health develops protocols for RMPsand uses templates issued by regulatory bodies forreporting.

Insights from the practical application of enhanced DUS willbe published in the coming months.

SAFETY MONITORINg - continued from previous page

Outlook reveals shifts in global use of medicines through 2017

IMS Institute for Healthcare Informatics quantifies the impact of trends and dynamicsdriving change in the healthcare market.

The IMS Institute for Healthcare Informaticscontinues to strengthen healthcare decisionmaking through research for use by governments,payers, academia, and the life sciences industry. In its latest report, The Global Use of Medicines: OutlookThrough 2017, Institute researchers, led by MichaelKleinrock, Director Research Development, offer aninformed and unbiased perspective on the driversimpacting pharmaceutical spending and use of medicinesover the next five years. In the extensive, complex andrapidly developing healthcare market, the report serves asa foundation for meaningful discussions around the value,cost and role of medicines during this period. Among itskey featured insights are:

• Annual global spending on medicines

• Balance of global pharmaceutical markets

• Top 20 therapeutic areas by market

• Changing contribution of generic medicines

• Relative importance of biologics

• Value of primary vs specialty care

• Impact of new medicines on disease treatment

• Relative access to innovation

• global spending growth and drivers

The analyses are based on IMS audits and include all typesof biopharmaceuticals, including biologics, OTC andtraditional medicines distributed and administered throughregulated delivery systems such as pharmacies, hospitals,clinics, physician offices, and mail order, where applicable.Spending figures are derived from IMS Market Prognosis™and are reported at ex-manufacturer estimated prices thatdo not reflect off-invoice discounts and rebates. IMSMIDAS™, Lifecycle™ R&D Focus, Lifecycle™ New ProductFocus, PharmaQuery™, Market Prognosis™ and TherapyForecaster™ were also used for assessing worldwidehealthcare markets, therapy class and product dynamicsand country-level pricing and reimbursement complexities. The IMS Institute for Healthcare Informatics is focused onproviding objective, relevant insights and research thataccelerate the understanding and innovation critical tosound decision making and improved patient care.

Further information on the report, published 19 November, 2013, can be accessed from the Institutewebsite at www.theimsinstitute.org, together withinformation on its extensive range of publications andresearch activities.


NEWS | PHARMACY SUPPORT SERVICES

Economic studies demonstrate impact of pharmacy medicines support services

IMS Health informs pharmacy initiative for vulnerable populations through service evaluation in England.

Health economic studies are supporting a patientinitiative by leading pharmacy multiples inEngland in association with the Department ofHealth. The initiative forms part of CommunityPharmacy Future (CPF), a project launched in2012 to design and evaluate new pathways forpharmacy services that can be rolled-outnationally to community pharmacies. Thesestudies, carried out by IMS Health, will serve todemonstrate the value of pharmacy interventionsfor both patients and the health system overall. The CPF project, including its initial implementation, isbeing jointly funded by the four largest pharmacy chains inEngland – Boots UK, the Co-operative Pharmacy, LloydsPharmacy and Rowlands Pharmacy. The services, initiallypiloted in the north of the country and now beingexpanded, are intended to offer patients practical supportfor achieving the best outcomes from their medicines. Theyhave been designed in full consultation with theDepartment of Health, PSNC, NHS Employers, local NHS,National clinical directors and gP representatives.

SUPPORTING VULNERABLE POPULATIONSThe current service provision focuses on two key vulnerablepatient populations: individuals with Chronic ObstructivePulmonary Disease (COPD) and those over the age of 65who are taking four or more medicines.

COPDCOPD is a progressive lung disease characterized by airflowobstruction. Caused predominantly by smoking, it has arelatively high prevalence, with nearly 900,000 diagnosedcases in England and Wales, and a further potential 2.7 million patients with undiagnosed COPD. Reducedhealthcare outcomes for patients are associated withsubstantial costs; direct costs to the NHS from COPD areestimated at around £900 million.

Elderly patientsSixty-five percent of all prescriptions in the UK aredispensed to elderly patients, who often have multiplemorbidities and are on several medications. Patients agedover 65 who use four or more medications represent ahigh-risk cohort, particularly susceptible to potentiallyinappropriate prescribing, imperfect medicationadherence, and an elevated risk of falls. The prescribing ofinappropriate treatment represents a substantial publichealth issue; adverse medication reactions and improperdisease management potentially result in a significantdisease burden and high healthcare costs. Total costs ofinappropriate medication prescribing have beenestimated to be £750 million in the UK. Managingmedication adherence in these patients can beparticularly challenging but has the potential to deliverimproved disease outcomes and reduced healthcare costs.

EVIDENCE-BASED APPROACH As health economists appointed to the project, IMS Healthis applying an evidence-based approach in evaluatingcomponents for inclusion in the services on the basis ofpatient need, best clinical practice, and potential togenerate healthcare benefits (to both patients and the NHS).Studies completed to date have revealed that pharmacysupport in these areas can drive significant direct andindirect cost savings and improved quality of life as well assubstantial improvements in medicines adherence, acutecare costs through improved disease management, andbetter patient outcomes over the long term.

For further information on the pharmacy initiative, pleaseemail Emma Bloomfield at [email protected]


HTA IN LATIN AMERICA | NEWS

ISPOR conference spotlights growing potential for RWE but signals challenges ahead.

Moving to value-based healthcare in Latin America

Today, more countries throughout Latin Americaare promoting a focus on real-world data, with evermore stringent evidence requirements to supportpricing, market access and formulary inclusion. As efforts to formalize the use of HTA continue to gathermomentum in Latin America, challenges are emerging forits local implementation as well as the success ofcollaborative initiatives to coordinate its adoption at theregional level. In September 2013, more than 740 expertsgathered in Buenos Aires to share their perspective on thesignificance of these developments at the ISPOR 4th LatinAmerica Conference. The conference topic of “Challenges for Health CareSystems in Latin America: Changing Models of HTA,Priority Setting, and Health Rights”paved the way for morethan 400 research presentations, many demonstrating theapplication of RWE in strengthening the basis for decisionmaking in the region.

PROGRESSING BEYOND THEORYIMS Health, a strong supporter of the event, contributed to theagenda on many levels, with a program led by the IMS HealthSymposium “Practical application of real-world evidence toHTA and decision making in Latin America: Moving beyondthe theory to practice.”

The highly experienced panel included guest presenter Dr Luis gonzález-Michaca, Health Economics, Market Accessand Pricing Senior Manager for the Latin America Region atJanssen, and experts from the IMS RWE Solutions & HEOR team,Dr Michael Nelson, Senior Principal, Julie Munakata, Principal,and Vanesa Leyva-Bravo, Senior Consultant. The session featured discussions on:

• Why RWE is becoming more important

• Specific challenges for RWE in Latin America

• RWE evolution in Latin America

• Implications from a commercial and payer perspective

• Added value of RWE to health economic modeling inLatin America

Proceedings of the symposium can be obtained fromAngelika Boucsein at [email protected]

RWE IN PRACTICE: ORIGINAL RESEARCHExperts from IMS Health also contributed to 11 posterpresentations during the ISPOR conference, showcasingexamples of projects, original research and analysesundertaken in the Latin America region leveraging RWE acrosskey disease areas (Figure 1).

ACCESS & COVERAGENovel pricing strategies to support sustainable access to medicine in Latin AmericaAccess to medicines index: Measuring how well countries provide access to medicinesguiding principles for effective private health insurance in emerging markets

CARDIOVASCULAR DISEASECo-morbidities associated to atrial fibrillation (AF) in selected Latin America countriesTreatment costs of ischemic stroke prevention and management in patients with atrial fibrillation (AF) in Latin America: Argentina, Brazil, Chile and Venezuela

Estimating the impact of statin therapy on direct and societal costs in Mexico: A comparison to SwedenTreatment patterns of atrial fibrillation (AF) in Latin America

DIAbETESDirect medical costs of treating diabetes-related complications in ArgentinaDirect costs of type 2 diabetes in Mexico from the public healthcare sector perspectiveDirect costs of type 2 diabetes from the Brazilian public healthcare sector perspective

ONCOLOGyCost/benefit analysis of first-line chronic lymphocytic leukemia (CLL)treatments in Mexico

FIgURE 1: IMS HEALTH POSTER CONTRIBUTIONS AT ISPOR 4TH LATIN AMERICA CONFERENCE

INSIGHTS | PREDICTIVE ANALYTICS

Predictive analytics promises to transform whole swathes of the healthcaresector, from early identification of at-risk individuals to reducing non-adherencethrough highly targeted, personalized interventions. But what exactly ispredictive analytics? Where will it see the greatest application? And how does it differ from conventional statistical analysis?

John Rigg, PHD, MPhil, BA is Senior Manager Advanced Analytics, RWE Solutions, IMS [email protected]

The authors

Ben Hughes, PHD, MBA, MRES, MSC is Senior Principal RWE Solutions, IMS [email protected]


A new crystal ball for healthcare?The digitization of patient, hospital, prescription, biological and other vast datastreams heralds a new era in the volume and complexity of healthcare information.As the data landscape continues to evolve so, too, must the tools employed tointerrogate the data if its true potential is to be realized. Cue predictive analytics.Predictive analytics is concerned with future or otherwise unknown events. Techniques in this field have progressedsignificantly in recent years, not least due to advances in computer processing, storage and retrieval. Increasingly, theyare now being applied to address problems in healthcare.

In 2011, the Heritage Provider Network offered a grand prize of $3 million in an open competition for the bestalgorithm at estimating the number of days a patient would be admitted to a hospital within the next year, usinghistorical claims data. The competition was intense, with advanced computing algorithms developed by data scientists,mathematicians, computer scientists, hedge fund managers and software engineers pitted against each other from 40 countries around the globe. Over 39,000 entries later, the prize was awarded in March 2013.

The aim of the sponsor, as articulated at the competition outset, captures the hopes of many for predictive analytics:“Once [the algorithm is] known, health care providers can develop new care plans and strategies to reach patientsbefore emergencies occur, thereby reducing the number of unnecessary hospitalizations. This will result in increasingthe health of patients while decreasing the cost of care. In short, a winning solution will change health care delivery aswe know it – from an emphasis on caring for the individual after they get sick, to a true health care system.”

The leading teams all used sophisticated analytical techniques known as machine learning algorithms. Machinelearning, a branch of artificial intelligence, specializes in developing algorithms that are highly effective at identifyingoften subtle or hidden patterns in large volumes of disparate data. These algorithms can be remarkably accurate atpredicting outcomes on new or unseen data. generalizations or predictions are often made from traditional statisticalanalysis based on observational or retrospective data. However, predictive analytics, incorporating machine learningalgorithms, can perform far better.

BROAD APPLICATIONIt is not just in the analysis of hospital admissions where predictive analytics is helping to drive healthcare solutions.Although its uptake is still embryonic, predictive analytics is expanding across the spectrum of healthcare withapplications such as: identification of patient risk factors; outcomes prediction; decision support; diagnosis; evaluationof treatment and pathways; clinical trial simulation; resource allocation; adherence; safety; product uptake; billoptimization; and fraud detection.

For example, there is a rapidly growing evidence base demonstrating the value of predictive analytics in diagnosisdecision support for physicians. Waljee, et al, for instance, reported in 2010 how a machine learning algorithmsubstantially out-performed standard metabolite tests in predicting the clinical response of patients with inflammatorybowel disease on thiopurines.1 This approach provides the potential for a low-cost, rapid alternative to metabolitemeasurements for monitoring thiopurine use.

PREDICTIVE ANALYTICS | INSIGHTS

The promise of predictive analytics





Another example is the EuResist project, a ‘real-life’ predictive analytics solution. EuResist is an internationalcollaboration designed to improve the treatment of HIV patients. A web interface allows physicians to specify patients'clinical and genomic data. The data is sent to the prediction engines and the combined response, displayed to thephysician, includes various suggested treatments and a prediction of their effect on the amount of HIV in the blood. In2009, the EuResist project was named as a Computerworld honors program laureate. This emphasizes the importanceof developing predictive analytics solutions as part of a real-time system.

An alternative approach is highlighted in a recent paper by Agneeswaran, et al, which shows how real-time machinelearning may be applied to aid physicians’ interpretation of electrocardiogram (ECg) reports for arrhythmia detection.2The physician submits the ECg reading on-line to a cloud-based machine learning algorithm which then supports theirdecision making by indicating which arrhythmia classification is most likely to apply.

PREDICTING THE IMPACT OF PREDICTIVE ANALYTICS FOR LIFE SCIENCES In a rapidly changing world, predicting the sectors where predictive analytics is likely to have greatest impact is itself achallenge perhaps best suited for a machine learning algorithm. Important ingredients in shaping its future willundoubtedly involve the regulatory environment, market conditions and the availability of high quality, integrateddata. However, commercial incentives are likely to be the most telling drivers, particularly in areas where predictiveanalytics can make the greatest business impact.

Successful applications by life science could either play pivotal roles in supporting external key stakeholders or internaldecision-making processes.

External applicationExternal application will work both on an aggregate-level, such as forecasting and aiding healthcare systems andinstitutions strategy, and at a micro-level, including specific choice in care management. Prediction at the micro-level issolutions tailored to the requirements of individual patients. For example, various pilots in risk prediction and caremanagement have demonstrated ~10-15% reduction in severe patient events. Other areas likely to see significantdemand for highly accurate prediction algorithms include diagnosis, non-adherence and resource utilization. For lifesciences in particular, a highly accurate prediction of adherence would allow stratified interventions in adherence orpatient programs to significantly improve effectiveness and return on investment. However, capturing value will hingeon delivery infrastructure and engagement with health systems.

Internal applicationCreating technology systems and necessary changes in business culture to action predictive analytics solutions in atimely, integrated and cost-effective manner may pose a greater challenge in many settings than creating thepredictive algorithms themselves. The second area where much greater accuracy is required for internal decisionmaking at a more aggregate level includes improving health economic and outcome models and their risk equations,or simulation for major decisions such as trial design. This is in addition to increasing the use of simulation in earlydiscovery, where various mathematical and simulation techniques are used.

Healthcare also needs to expand its skills profile to enable effective predictive analytics solutions. Data scientists,computer scientists and mathematicians all have skill sets likely to face increasing demand from within this sector.

CHALLENGESFor all the promise predictive analytics offers, it, too, faces challenges. Perhaps foremost is the lack of transparency ofmany machine learning algorithms, referred to as ‘black-box’. Creative visual representations and improved modeldiagnostics are helping end-users interpret output from machine learning algorithms. Nonetheless, opacity may tosome extent always be the price that must be paid for a more powerful predictive solution.



PREDICTIVE ANALYTICS VERSUS TRADITIONAL STATISTICAL ANALYSISIn 2001, the renowned Berkley statistician and predictive analytics advocate Leo Breiman argued: “The statisticalcommunity has been committed to the almost exclusive use of data models. This commitment has led to irrelevanttheory, questionable conclusions, and has kept statisticians from working on a large range of interesting currentproblems...If our goal as a field is to use data to solve problems, then we need to move away from exclusivedependence on data models and adopt a more diverse set of tools.” 3

Whilst not everyone would agree with this extreme stance, many would embrace a more empirically-driven focus tohelp address some challenges. So, how does predictive analytics differ from traditional statistical analysis usingobservational data? The following discussion may be of particular interest to those with direct experience of statisticalresearch. Disparities between the two approaches may at times reflect not so much a paradigm shift but differentpoints on the same continuum. Nonetheless, as shown in Figure 1, several differences characterize predictive analyticscompared to traditional statistical analysis:

• A research objective designed to maximize prediction accuracy rather than draw inferences about associationsbetween variables

• A scientific philosophy that embraces a data-driven inductive approach, rather than a hypothesis-driven deductive approach

• A choice of statistical model from a class of machine learning algorithms

• An analytical framework that uses separate samples for model development and evaluation/validation

• Evaluation metrics criteria focused on predictive accuracy of the model (eg, AUC - Area Under receiver Curve)

1. Research objectiveIn predictive analytics, the overarching objective is to maximize predictive accuracy. By contrast, variable ‘importance’ –the ability to draw inferences about the magnitude or significance of a particular variable (or variables) in relation to anoutcome – is typically the primary objective in standard statistical analysis.

FIgURE 1: KEY DISTINgUISHINg FEATURES OF PREDICTIVE ANALYTICS VS TRADITIONAL STATISTICAL ANALYSIS

Traditional statisticalanalysisPredictive analytics

Research objective

Scientific philosophy

Predominantmodeling techniques

Analytical framework

Evaluation metrics

• Maximize predictive accuracy

• Data-driven, inductiveapproach

• Random forests, neuralnetworks and model blending

• Separate samples for modeldevelopment and evaluation

• AUC/model-level predictionaccuracy

• Draw inferences on variableassociations

• Hypotheses-driven, deductiveapproach

• Regression, logistic regression,(sometimes Bayesian statistics)

• Single sample for modeldevelopment and evaluation

• R square, variable-level teststatistics (p value, t-test)




The approaches converge somewhat in that predictive analytics often attempts to quantify variable importance (asmeasured by the contribution of variables to the predictive power of a model). Moreover, traditional statistical analysesfrequently make generalizations – predictions. Nonetheless, differences in the primary research objective are a keydistinguishing feature.

2. Scientific philosophyConventional statistical analysis generally starts with a clear idea about variable associations of interest. It is ahypothesis-driven approach; a deductive method of scientific enquiry. Predictive analytics, on the other hand, usuallyinvolves fewer prior expectations about variable inclusion and associations. Algorithms are allowed a ‘free-reign’ to‘mine’ the data and find meaningful correlations. It is an inductive scientific philosophy. One notable commonalitybetween the approaches is the data exploration phase which takes place in many traditional statistical analyses.

3. Modeling techniquesUnlike regular statistical analysis, predictive analytics routinely use so-called machine learning algorithms. There is adiverse array of algorithms, some of the most popular of which are based on decision trees – schematic tree-shapeddiagrams showing how different combinations of variables (the branches of the trees) are associated with differentvalues of the target variable. Hundreds or thousands of trees are usually created, each on a subset of variables and/ordata, to form an ensemble decision tree model. Even if each tree is a poor predictor, combining the results from manytrees can still produce a highly accurate model.

Random Forest is a widely adopted ensemble decision-tree approach. A random subsample of observations is used togrow each tree and a random selection of input variables are used to determine each node (the combination of inputvariables that denote a fork in a branch or a branch end-point).

AdaBoost (Adaptive Boosting), is another popular ensemble decision-tree algorithm. Each tree places a higher weighton observations misclassified by previous trees, thereby placing extra emphasis on predicting ‘hard’ observations. Thefinal model is a weighted sum across all the trees.

Another strand of machine learning is inspired from advances in the mapping of biological neural networks. Theseartificial neural networks can provide powerful solutions for data characterized by highly nonlinear, interdependentassociations.

Some of the most accurate predictive modeling solutions involve combining output from separate learning algorithmsto produce a ‘blended’ model. These aim to capture the best features from different algorithms. This practice is in sharpcontrast to the use of a single model in standard statistical analysis.

The label ‘machine learning algorithm’ is potentially confusing. Logistic Regression, possibly the most widely usedmodeling technique in regular statistical analyses, is itself a powerful learning algorithm in that parameter precision isincreased with more data. Thus, machine learning tends to refer more loosely to ‘non-standard’ learning algorithmssuch as those mentioned above.

4. Analytical frameworkA defining feature of predictive analytics is the use of a separate sample for model development and modelvalidation/testing. This analytical framework helps overcome a phenomenon known as ‘overfitting’, where a model maydescribe accurately the data it is estimated (or trained) on, but has poor predictive accuracy on new or unseen data.

In order for predictive analytics to live up to its promise, there must be willingness to embrace a methodological shift in the approach to statistical analysis.



For instance, the analytical framework for predictive analytics may involve splitting the sample into three: a training,validation and test sample. A series of models is estimated on the training sample, including measures designed toreduce overfitting (a process known as regularization). Predictive accuracy of each model is assessed using thevalidation data and the preferred model identified. Finally, the test sample is used to assess the generalization/predictiveproperties of the chosen model. The test sample is necessary since a reliance on the validation data would violate theprinciple that data used in the final assessment of the model should not feature in model selection/training.

Not only does this analytical framework help minimize overfitting, it is a key component in producing good predictivesolutions. generalizations are often made from the results of traditional statistical analyses. given the biases associatedwith overfitting, such conclusions stand to benefit from adopting a similar analytical framework.

5. EvaluationEvaluation in predictive analytics is focused at the level of the model, rather than the typical variable-level focus instandard statistical analyses. Moreover, model performance is judged using metrics for prediction accuracy based onvalidation or test data (not the data the model was trained on).

A common metric in binary classification is the AUC. The curve, the Receiver Operating Characteristic, is a plot of theproportion of true positives classified by the model (True Positive Rate) on the vertical axis by the proportion of falsepositives classified by the model (False Positive Rate) on the horizontal axis, for different levels of the predictedprobability. It provides a measure of the discriminatory power of the model. The AUC neatly ties in with diagnosticdecision making since TPR is equivalent to benefits and FPR equivalent to costs. Thus, information about costs andbenefits can be explicitly built into the evaluation criteria and used to select the optimal model.

FULFILLING THE PROMISEHealthcare decision makers stand to benefit from application of predictive analytics solutions, from care managementto trial optimizations. These are powerful solutions that can make best use of the growing volume of rich data acrossthe healthcare sector. However, in order for predictive analytics to live up to its promise, there must be willingness toembrace a methodological shift in the approach to statistical analysis. Moreover, success will require applying thetechniques to the right business problems, where specific choices can be made as a result of the prediction, andbusiness processes are in place to execute those choices. While there is no playbook for this today, it will surely follow asevidence mounts highlighting the business impact of solutions based on predictive analytics •

Healthcare decision makers stand to benefit from application of predictive analyticssolutions, from care management to trial optimizations - making best use of thegrowing volume of rich data across the healthcare sector.

1 Waljee AK, Joyce JC, Wang S, Axena A, Hart M, Zhu J, Higgins PDR. Algorithms outperform metabolite tests in predicting response of patients with inflammatory boweldisease to thiopurines. Clinical gastroenterology and Hepatology, 2010; 8:143-50

2 Agneeswaran VS, Mukherjee J, gupta A, Tonpay P, Tiwari J, Agarwal N. Real-time analytics for the healthcare industry: Arrhythmia detection. Big data, Sept 2013; 1(3): 176-182. doi:10.1089/big.2013.0018.

3 Breiman L. Statistical modeling: The two cultures. Statistical Science, 2001; 16(3): 199-231

IMS HEALTH ECONOMICS AND OUTCOMES RESEARCH

INSIGHTS | IMS HEALTH SYMPOSIUM

Increased reliance on RWE has seen major advancements in data, technologiesand analytical approaches to further augment its value as a decision-makingtool. But is this heightened investment making a difference? First experienceswith enhanced RWE solutions were explored during an IMS Health Symposiumat the ISPOR 18th Annual International Meeting in New Orleans, revealingsome very tangible benefits to a growing range of healthcare stakeholders.

Jovan Willford, MBA is Principal RWE Solutions, IMS Health [email protected]

Michael Nelson, PHARMD is Senior Principal RWE Solutions & HEOR, IMS [email protected]

Frederic King, MBA is Global Payer Evidence Director, Payer & Real-World Evidence Group, [email protected]

The authors


OUTCOMES - Issue 1 Page 1

Filling the gap between basic research and decision makingThere has been a significant industry trend to incorporate RWE into strategic planningto support the value propositions for new and existing healthcare technologies. Today, multiple stakeholders are involved in the use of RWE, from early product development, market access planningand cost-effectiveness analysis to commercialization, HTA and clinical decision making between physicians andpatients. At the same time, new capabilities in data sourcing, technology, analytics and communication are enablingthe pursuit of better evidence to fill the gap between basic research, provider decisions and patient care. But as itprogresses from theory to practical application, how is RWE impacting decision making? How can its value be furtherextended to validate traditional research approaches? And how can it be used to support emerging trends such asoutcomes-based contracting?

CURRENT EVOLUTIONS IN RWEAt its most fundamental level, RWE allows life science organizations to engage external customers in an evidence-informed manner leveraging scientifically credible, patient-level outcomes data to improve the quality and value ofmedical treatments. Its pivotal role in powering retrospective data studies is well established, but there is also somenotable strategic movement in pockets of RWE innovation as manufacturers begin to mobilize on several fronts:

1. Organizational readiness for RWE: Building investments and structures to support and establish RWE as aplatform for innovation, including governance frameworks and cross-functional pull-through from pricing andmarket access to HEOR and commercial colleagues.

2. Strategic partnerships: Forging alliances and partnerships to solve data and access gaps and acceleratecollaborative initiatives.

3. big data to focused data: Creating evidence-based, franchised data hubs, integrating a wealth of disconnecteddatasets with the tools and technology to better manage and focus use of the data.

4. Analytic engine rooms: Breaking down organizational barriers by testing conventional ownership of RWE datainternally to ensure a skills base that understands how to better leverage integrated datasets and maximize its utilityacross functions.

5. Customer engagement: Bringing RWE tools together to engage more strategically with external stakeholders,across multiple mediums, including peer-reviewed journals and value-added services for payers and providers.

Beyond this, the evolution of RWE already has a wider range of implications for brand teams. Recent groundbreakingresearch from IMS Health, based on more than 50 interviews with clients including payers and HTA experts, identified100 products across countries and therapeutic areas where RWE has had a measurable influence on initial or ongoing market access (Figure 1).

IMS HEALTH SYMPOSIUM | INSIGHTS

The power of real-world evidence


There is some notable strategic movement in pockets of RWE innovation asmanufacturers begin to mobilize on several fronts.




These findings mark a clear shift away from a sprint to thefinish of product launch towards a process of continuousevidence management throughout the lifecycle. Thisplaces the industry firmly at the center of engaging withpayers, providers and physicians through evidence in avariety of ways – not just around retrospective studies butalso key account management, analytic-based servicesand value-based contracting (Figure 2).

GENERATING VALUE FOR PHARMAFor manufacturers, RWE has a role in supporting internaldecisions for investment around product developmentand commercialization, as well as formulation ofstrategies for external engagement with a range ofstakeholders, including payers to inform fundingdecisions, and providers and physicians to guide theirclinical practices and the way they deliver medicines topatients. Two recent case studies from AstraZenecademonstrate its practical impact in both of these areas.

Case study 1: Informing product development and launch strategy Payers take many different forms and their datarequirements for reimbursement dossiers vary in differentcountries. In the UK and Canada, for example, the focus ison cost-effectiveness modeling; in France and germany,on innovative value-based assessments; and in the US, onthe patient as payer and on traditional managed care andcontracting. Clinical evidence remains the platform todeliver the relevant clinical data, but there are gaps. Inorder to fill them, different types of projects across thecontinuum of RWE can be leveraged.

One recent example at AstraZeneca used RWE to informproduct development and launch readiness whenbuilding a payer strategy and assessing evidencerequirements over and above the randomized controlledtrial (RCT) data. The disease area in question was poorlydefined and managed largely by over-the-counter (OTC)medications. This created challenges for substantiatingclaims and for establishing the need to fund a newprescription product using traditional means of gatheringevidence. A systematic literature review revealed limitedavailable data to determine disease burden and economicand quality of life impact.

The absence of an ICD-9 code and current use of OTCs meant that very little claims data existed, making it difficult tocharacterize the disease by monitoring and tracking patients through retrospective data collection. Moreover, noprospective data collection studies were ongoing or planned.

To undertake a large, prospective burden of illness study would not only have been costly but resource intensive andtime consuming. Before deciding on such a significant investment, a two-step ‘fit for purpose’ approach was taken to fillthe evidence gaps.

FIGURE 2: PHARMA RWE CUSTOMER ENGAGEMENT OPPORTUNITIES

Num

ber o

f cas

e st

udie

s0

5

10

15

20

2522

23 34

911

16

21

10

Label Launch Access Ongoing Access Price Use

Source: IMS Health interviews with more than 50 clients including payers, health technology assessment (HTA) experts

IT USA UK SWE CA ES NL FR DE DK

25

20

22 21

15

10

5

0N

umbe

r of c

ase

stud

ies

11

16

SWEUKUSAIT

23 34

910

DKDEFRNLESCA

Label essccLaunch A

echnology assessments with morviewerte: IMS Health incourS

essccOngoing A eciPr eUs

tsxper) eAt (HT TAechnology assessmen, health ersyts including pae than 50 cliens with mor

e

, health

FIGURE 1: REAL-WORLD EVIDENCE MARkET IMPACT

1 2

3 4 5

Quality measures; Member pro�ling; Physician P4P metrics Population wellness; PRO tracking; Cost controls (Rx)

P4P Schemes; KAM; Franchise evidence Total cost of care; KAM; Quality measures Patient subpopulation; Patient journey tools; Population management design and track

1

2

3

4

5

RWE is the currencythat enablesthese discussions

PatientPayer Provider

Pharma

1

reyyePa

2

ttienaPviderorP

3

y

4 5

these discussionst enablestha

encrWE is the curR

Pharma

yenc

tion managemenopula Ption; Pt subpopulatiena P

Ae; Kost of carotal cTTotal c

M; FAchemes; K P4P S

ellness; PRtion wopula P

ember pres; My measurualit Q1

2

3

4

5

ackn and trt desigtion managemenools; ney tt jourtienation; P

esy measurualitM; Q

eanchise evidencrM; F

ols (Rx)tronost cing; CackO trellness; PR

icssician P4P metryho�ling; Pember pr

ics



The first stage involved examining RWE data through a cross-sectional survey pilot study, across four domains:

1. Feasibility: Establish the feasibility of finding patients and identifying providers willing to participate and involvesites in this type of study.

2. Clinical impact: Identify current treatments and their effectiveness, and characterize patients and demographics.

3. Economic impact: Determine whether patients present in clinics or emergency rooms, and the healthcare resourceutilization associated with the condition.

4. Quality of life: Ascertain quality of life impacts and how that would feed into establishing claims for payers aroundthe condition of the disease.

This approach provided signals across each of the domains which ultimately led to investment in a multinational, non-interventional study to understand the naturalistic distribution of the disease, the way in which patients are managed,and the clinical, economic and quality of life impact.

The data from the study will be leveraged across a range of deliverables for payers, providers and physicians to supportmarket access and reimbursement. It will form the basis of disease communications to further substantiate anddemonstrate a burden and unmet need, and as part of a pre-launch, unbranded campaign to establish with providersthe requirement for new treatment options. It will also feed into cost-effectiveness and budget impact models todemonstrate the size of the economic impact and baseline quality of life in different countries, as well as to build localpayer dossiers.

Case study 2: Engaging payers with evidence to drive decision making A second case study demonstrates the use of RWE to support discussions with payers, many of whom have beenattempting to lower prescription (Rx) drug costs through the use of formulary restrictions or incentives for prescribing apreferred medication. When the proton pump inhibitor (PPI) Nexium was excluded from the formulary of a large UScommercial health plan, AstraZeneca decided to evaluate the medical impact on patients with gastroesophageal refluxdisease (gERD) and erosive esophagitis, and those taking prescription NSAIDs. The approach was a retrospective claimsdatabase analysis, using the health plan’s own data, comparing patients who were switched to another Rx PPI followingformulary exclusion of Nexium, to patients who remained on the drug during the same time period (paying out-of-pocket). The goal was to compare these two cohorts to determine the impact on adjusted upper gI-specific adverseevents from a medical cost perspective and the implications of that for the drug budget.

More than 45,000 patients were deemed eligible for the study, of which about 26% continued on Nexium and 60%switched to another PPI. A comparison of these two cohorts over the 12-month period following the exclusion revealedthat, overall, patients who switched to other Rx PPIs had significantly higher average upper gI-specific medicalexpenditure and Rx PPI costs compared to patients who continued on Nexium – clearly validating the company’shypothesis that the drug should have been kept on formulary. Such was the power of this real-world finding that whenpresented with the data the commercial plan reversed its decision and placed Nexium back on formulary with preferredstatus – a compelling example of how RWE can be leveraged to inform decision making.

VALIDATING HEALTH ECONOMIC ANALYSISSignificant investment goes into the creation of health economic (HE) analysis for communicating the value of productsto decision makers at the time of regulatory approval and market access. Models are populated with available data(clinical trial, pricing, utilization, +/- RWE) to provide an indication of a new product’s likely economic impact on themarketplace. However, payers are increasingly seeking hard evidence beyond the estimates available from models –information that is only available from RWE based on product performance post launch. But how do these twocompare? What if the evidence in the real world is significantly different to the economic model or vice versa?

Case study 3: Validating a published economic analysisTo explore this issue further, researchers at IMS Health conducted a retrospective study to validate a well-publishedcost-effectiveness analysis using RWE from the PharMetrics Plus health plan claims database. The specific objectiveswere to generate clinical and cost inputs to repopulate the model and compare the updated model results with theoriginal model results and with the RWE evaluation of two therapies.




The analysis in question was for VTE prophylaxis in total knee and total hip replacement surgery over a one-yeartimeframe from a US payer perspective, with endpoints in the follow-up period focused on clinical event rates andcosts.1 The approach was to replicate the clinical populations used in the original models as closely as possible butwithout biasing the analysis. Propensity score methods were used to adjust for differences between the treatmentgroups. Some minor, but important, differences were noted between the RWE sample compared to the RCT:

• Commercially insured population with fewer females and lower average age

• VTE events were limited to inpatient encounters

• Follow-up period was longer, but the original model extrapolated to one-year costs

An important caveat to this approach is that not all models lend themselves to validation; certain product/diseaseattributes may not be conducive to leveraging RWE for health economic analysis and decision making, due to the nature of:

• Cost drivers (eg, direct costs versus indirect costs)

• Endpoint availability (eg, utilization data versus patient-reported outcomes)

• Timeframe (eg, short-term: asthma, COPD versus long-term: diabetes)

• Disease prevalence (eg, lack of sample size for rare conditions)

• Product uptake

• Payer factors (eg, enrollee turnover, population size)

ResultsEvent rates between the RCT data and RWE data were remarkably similar in the intervention group. They were somewhatdifferent for the comparator group, possibly due to disparities in definition and to real-world usage patterns, but theresults nevertheless supported the better performance of the new drug. Costs were also consistent, across the trial-basedmodel, PharMetrics Plus-based model and the direct reported results from the PharMetrics Plus data analysis.

There are two important takeaways from these findings:

1. The estimates in the original model occurred before the product was launched; those from RWE occurred afterlaunch, using real-world data, and essentially confirmed the estimated impact of the intervention. Side by side theseare compelling tools but when benchmarked against one another, HE analyses and RWE can provide a morecredible, powerful and harmonized message of anticipated economic impact.

2. The coupling of HE analyses and RWE has the potential to serve as the basis for pharma to engage in a moreoutcomes-centric relationship with payers – who are increasingly demanding not just the baseline cost of a productand the estimated impact but to contract based on the relative value of products in the market place. This is a globaltrend and one that will continue to grow. Already, early examples of outcomes-based contracting are emerging(Figure 3).

When benchmarked against one another, HE analyses and RWE can provide a morecredible, powerful and harmonized message of anticipated economic impact.

1 For a detailed discussion of the retrospective validation study, see Project Focus on page 60



The foundations thus exist for RWE to provide a post-launch comparison of economic impact as a basis for contracting.There are several important keys to optimizing the approach:

• Early synchronization of clinical programs to capture measurable endpoints in RWE to inform model design andensure consistency with post-marketing analyses.

• Rigorous and consistent definition of study populations and clear understanding of how post-launch populationsdiffer and how this impacts results.

• Ability of payers and pharma to provide data for evaluation, leverage HE methods where appropriate, conduct the analysis in a manner trusted by both parties, and enter into outcomes-based contracts which pass regulatory scrutiny.

• Application of good pharmacoeconomic principles to inform these analyses, including sample size, statistical as well as clinical significance.

CONCLUSIONAs the marketplace transitions from defining RWE to understanding exactly how it can be used to optimize internaland external decision making, examples are emerging to demonstrate the impact from a growing range of enhancedRWE applications. As the healthcare landscape continues to evolve to emphasize health outcomes in reimbursementdecisions and contracts, the future is set for the role of RWE alongside robust HE analyses as a critical component ofrelationships between providers and payers •

This article draws on presentations from the IMS Health Symposium “The Power of real-world evidence: Filling the gapbetween basic research and decision making” held during the ISPOR 18th Annual International Meeting in New Orleans,USA, in May 2013. Chair: Jovan Willford, MBA, Principal RWE Solutions, IMS Health. Speakers: Frederic King, MBA, Global Payer Evidence Director, Payer and Real-World Evidence Group, AstraZeneca, andMichael Nelson, PharmD, Senior Principal RWE Solutions & HEOR, IMS Health.

FIGURE 3: EXAMPLES OF OUTCOMES-BASED CONTRACTING ARE EMERGING

• Novartis used "no cure, no pay“ to boost uptake before new competitor entry

• Sanofi/Procter & gamble use risk-sharing deal to stave off both generic and branded competition

• Pfizer/Eisai is testing "integrated care" to assess the cost-effectiveness and impact of its dementiadrug, Aricept

• UCB received positive recommendation from NICE in the UK based on a patient access scheme

• Pfizer/Wyeth have a compliance support program with a german payer

• Merck & Co. and Cigna agreed to outcomes-related discount deal

• Serono and Cigna assessed percentage of hospitalizations and emergency room visits avoided

Assumptions made during trial design are intrinsic to its ultimate success butcan prove inaccurate on the basis of data that emerges as the study progresses.The ability to adjust for this uncertainty through the use of innovative adaptivedesigns can not only allow for greater efficiencies but also has growingpotential in observational analyses.

Sonia Rojas, MSC is Statistician RWE Solutions & HEOR, IMS [email protected]

The authors

Montse Roset, BSC is Director RWE Solutions & HEOR, IMS [email protected]

Núria Lara, MD, MSC is Senior Principal, RWE Solutions & HEOR, IMS [email protected]

Page 1 IMS HEALTH ECONOMICS AND OUTCOMES RESEARCH

INSIGHTS | OBSERVATIONAL METHODOLOgY


Adaptive techniques for improved efficiencyKey elements in a clinical trial, such as primary endpoint, clinically meaningfultreatment difference, and measure of variability, need to be specified during thedesign of the study. The accuracy of the assumptions considered at that stage willdetermine the success and validity of the study. The use of adaptive designs (ADs)allows uncertainly about assumptions to be addressed during the planning phase.

INNOVATIVE DESIGNS IN CLINICAL TRIALSADs were initially defined by a working group as a clinical study design that uses accumulating data to decide how tomodify aspects of the study as it continues, without undermining the validity and integrity of the trial.1 The same groupalso emphasized that ADs are not a solution for inadequate planning, but are meant to enhance study efficiency. The“guidance for Industry: Adaptive Design Clinical Trials for Drugs and Biologics”, developed by the FDA, defined AD as “astudy that includes a prospectively planned opportunity for modification of one or more specified aspects of the studydesign and hypotheses, based on analysis of data (usually interim data) from subjects in the study”.2

Characteristics and applicabilityADs allow several types of adaptation, including study sample size, study duration, treatment group allocation, numberof treatment arms, or endpoints assessed during the study. Each adaptation must ensure that the type I error rate iscontrolled and that the trial has a high probability of answering the research question of interest. In any case,adaptation rules must be clearly specified in advance in order to properly design the simulations.

Despite their suggested promise and the attention they have received in the literature, current acceptance and use ofADs in clinical trials are not aligned.

ADs are applicable to different phases of clinical trials, including learning stage designs, confirmatory stage designs,and adaptive seamless designs that seek to integrate multiple stages of clinical research into a single study.3

• Learning stage designs: In general, ADs are more accepted in the learning (exploratory) stages of clinical trials.2,4

Here, they allow researchers to learn and optimize based on accruing information related to dosing, exposure,differential participant response, response modifiers, or biomarker responses.2 Most common studies performed inthe exploratory stages include definition of safety dose or efficacy dose.

• Confirmatory adaptive designs: Here, ADs can be applied to several designs, including adaptive randomization,enrichment designs, sample size re-estimation and more sophisticated designs. Adaptive randomization varies theallocation of subjects to treatment groups based on accruing trial information,1,5,6 instead of fixing constantallocation probabilities in advance. Enrichment designs include several aspects of the study. Enrichment of a studypopulation makes it possible to ensure that participants in a trial are likely to demonstrate an effect fromtreatments,7 as well as allowing definition during the study of recruitment subsets of patients with a highertreatment response. Sample size re-estimations designs allow the parameter estimated to be updated during anongoing trial to be used to adjust sample size accordingly.8

Use of innovative designs inobservational studies

OUTCOMES - Issue 1 Page 1

OBSERVATIONAL METHODOLOgY | INSIGHTS





The primary concern when implementing ADs in clinical trials is their validity and integrity from a regulatoryperspective. FDA released a draft guidance for industry on adaptive design clinical trials.2 A similar draft was released byEMEA's Committee for Medicinal Products for Human Use (CHMP) in 20069 and a joint workshop was organized by theUS and EU regulators, academia and industry experts.10

EXTENDING THE USE OF INNOVATIVE DESIGNS TO OBSERVATIONAL STUDIESADs are more applicable to and used in clinical trials or interventional studies due to the possibility of modifying thestudy intervention, study groups or other study parameters. Observational studies are limited to observing patients interms of treatments and outcomes, and there is less room for influence on study design. Nevertheless, observationalstudies do require assumptions about the patients’ characteristics and about the distribution of the population that willbe enrolled during the study planning. Any discrepancy between the hypothesized and actual distribution of theenrolled population will affect the power of the study.

Study planning (interventionist of observational design) involves calculating the minimum sample size to achieve adesired level of statistical power. In interventionist studies or clinical trials, the specification of the effect of primaryinterest and the variance or overall outcome rate for continuous or binary outcomes, need to be defined. Often, thisestimation is based on limited available information which can lead to an inadequate sample size if the estimate is poor.An internal pilot design11 uses a revised nuisance parameter estimate at an interim stage of the study to adjust the final sample size. This approach selects a sample sufficiently large to achieve the desired power without usingunnecessary resources.

Internal pilot designAlthough extensions of the internal pilot (IP) design in clinical trials to the observational setting have beenconsidered,12 experience of using IP in observational studies is lacking. gurka, et al,12 assessed the use of IP inobservational studies using simulations. When planning an observational study, additional nuisance parameter arises,such as distribution of the subjects to the exposure groups. In clinical trials, the ratio between groups or arms is definedby the investigators. Nevertheless, in observational studies in which two exposure groups are compared, the ratiobetween groups is not necessarily predefined (Figures 1 and 2). If not (Figure 2), it follows the real distribution obtainedin clinical practice.

FIGURE 1: COHORT STUDY BEGINNING WITH EXPOSED AND NON-EXPOSED GROUPS

Exposed patients / Cohort 1

Disease No disease Disease No disease

Non-exposed patients / Cohort 2

Subjects are selected based on their exposure status and disease outcomes arecompared to a group of similar subjects who are unexposed

The use of adaptive designs allows uncertainly about assumptions to be addressedduring the planning phase.



gurka, et al,12 examined the application of an IP where using re-estimates of the exposure group distribution, and madeany necessary adjustments to the final planned sample size using power formulas that consider this distribution fixedand known. The research demonstrated the utility of IPs when designing an observational study, with the added sourceof randomness of membership among the exposure groups. The authors show that the IP is beneficial in these settingswith little deleterious effect on the validity of the final analysis, especially when the required sample size is large.

To date, these tools have been primarily used in the clinical trials setting. However, the utility of IP designs extends to awide range of studies. Furthermore, the implementation of IP designs may be more feasible outside of the regulatoryand logistical hurdles that arise in the clinical trials setting. Prospective cohort studies and cross-sectional studies oftenrequire large amounts of money and researcher-time to complete. Thus, ensuring optimal power should not be asecondary consideration during their design. Allowing for re-estimation of the proportion of subjects in each exposuregroup would be beneficial with respect to ensuring optimal power.

Further applicationsIn addition to the use of IP for sample size adjustments, other applications of ADs in observational studies should beassessed in order to optimize time, number of patients and the resources required for these studies. In any case, theapplicability of ADs in observational investigations needs to maintain the observational aspect of the study, theexternal validity of studies performed in terms of clinical practice, and ensure the statistical power of the study. ADs canprovide benefits in the design of observational studies including other assessments than sample size required. Some ofthese potential applications are:

• Selection of most relevant outcomes to be assessed: In exploratory observational studies, a high number ofoutcomes are normally included in order to describe or compare several outcomes in study patients. In longprospective studies, it would be possible to reduce the number of variables or outcomes to be assessed in follow-upvisits in order to reduce the workload and focus on main outcomes.

FIGURE 2: COHORT STUDY BEGINNING WITH A DEFINED POPULATION

Sampled patients

Targeted population

A defined population is selected before any of its members become exposed or before their exposures are identified

Disease No disease Disease No disease

Not randomly assigned


Exposed patients / Cohort 1 Non-exposed patients / Cohort 2



• Selection of most relevant subpopulations: In observational studies, a broad population of patients is usuallyincluded with possible differences in the outcome obtained from the study exposure. ADs can be helpful inidentifying subgroups of patients in which a higher benefit is expected, taking into account that external validitymust be maintained. Therefore they could be used to define subgroups of patients with higher benefits in terms ofeffectiveness (clinical practice), ie, disease stages, but not exclude those patients who are usually excluded fromclinical trials. The use of ADs in observational studies needs to maintain the external validity.

• Comparative observational studies: These usually include a group of patients receiving the drug or healthintervention of interest, and another group of patients receiving other treatments, usually ones received in clinicalpractice, and called a control group. In cohort studies that begin with exposed and non-exposed groups or studycohorts (Figure 1), the selection of the control group is of crucial importance in order to obtain a sample of patientsas similar as possible to the treatment group with respect to all other factors that may be related to the diseaseexcept for the exposure of interest. Several techniques to correct selection bias, including direct matching orpropensity score matching, have been applied. Nevertheless, the direct matching approach adds difficulties to therecruitment of patients. ADs could be used in order to assess differences in terms of confounding factors betweenstudy groups and to estimate their impact on outcomes.

STATISTICAL APPROACHES IN INNOVATIVE DESIGNS Two statistical methodologies are applicable to the design and analysis of studies: frequentist and Bayesian. Bayesianstatistics provide a formal mathematical method for combining prior information with current information at thedesign stage, during the conduct of the trial, and at the analysis stage.13,14,15,16 Then, Bayesian inference, used in ADs,derives the posterior probability as a consequence of two antecedents, a prior probability and a "likelihood function"derived from a probability model for the observed data. In using the entire posterior distribution of the parameter(s),the Bayesian theory approach accounts for uncertainty in the value of the parameters, and hence will correctly estimatethe variance of the predictive distribution to obtain un-biased estimators.

bayesian approachBayesian analyses use two types of information about the unknown parameters of interest:

1. Prior distribution, which represents the additional (external)

2. Available sample data, expressed formally by the likelihood function

The prior distribution used in the Bayesian approach is usually based on data from previous trials. The prior informationincluded needs to be carefully selected, encompass as many sources of good quality information as possible, and beincorporated into the analysis.15,17 The likelihood function is also an essential component of frequentist statistics. Theposterior distribution used to extract conclusion from Bayesian analysis is the product of the prior and the likelihoodfunction.15

Adaptive designs allow several types of adaptation, including study sample size,study duration, treatment group allocation, number of treatment arms, orendpoints assessed during the study.



FDA believes the Bayesian approach is well suited for surveillance purposes. It is possible, for example, to use theposterior distribution from a pre-market study as a prior distribution, to the extent that data from the clinical studyreflect how the device is used after approval.15 In the context of observational and effectiveness studies, priorinformation provided by clinical trials can be updated with preliminary data collected during the study execution inorder to obtain a more adjusted estimation of expected outcome, sample size required and other study design aspects.

Benefits derived from the use of Bayesian methods in the design of ADs are associated with additional efforts requiredduring the study implementation. The logistics of ADs require careful thought. During study design, ADs requirecomputer-based simulations. During the study itself, they require electronic data collection, due to the need forfrequent updates, and the data monitoring committee has to meet regularly and quickly to implement the adaptations.During the statistical analysis, Bayesian statistical approach is computationally and logistically complex, requiredspecialized statistical package to implement and analyze data and might not be practically feasible in all situations.

Based on the pros and cons of the applications of ADs in observational studies, it would be reasonable to assume thatthere is significant room for improvement in the design of observational studies, moving to more efficient alternatives •

1 gallo P, Chuang-Stein C, Dragalin V, gaydos B, Krams M, Pinheiro J. Adaptive designs in clinical drug development: An executive summary of the PhRMA working group. J Biopharm Stat, 2006; 16:275–283

2 US Food and Drug Administration. Draft guidance for Industry: Adaptive design clinical trials for drugs and biologics. Accessed 10 October, 2013 at:http://www.fda.gov/downloads/DrugsguidanceComplianceRegulatoryInformation/guidances/UCM201790.pdf

3 Kairalla J, Coffey C, Thomann M, Muller K. Adaptive trial designs: A review of barriers and opportunities. Trials, 2012; 13:145 4 Coffey CS, Kairalla JA. Adaptive clinical trials: Progress and challenges. Drugs R&D, 2008; 9:229–2425 Zhang L, Rosenburger W. Adaptive randomization in clinical trials. In: Design and Analysis of Experiments, Special Designs and Applications. Volume 3. Edited by

Hinkelmann K. Hoboken: John Wiley & Sons, 2012:251–2826 Rosenberger WF, Sverdlov O, Hu F. Adaptive Randomization for Clinical Trials. J Biopharm Stat, 2012; 22:719–7367 Temple R. Enrichment of clinical study populations. Clin Pharmacol Ther, 2010, 88:774–7788 Proschan MA. Sample size re-estimation in clinical trials. Biometrical J, 2009; 51:348–3579 Reflection paper on methodological issues in confirmatory clinical trials with flexible design and analysis plan. London, UK: EMEA; 2006. European Medicines Agency.

Report no: Doc. Ref. CHMP/EWP/2459/02.10 Corriol-Rohou S. Proceedings of the 2nd EMEA/EFPIA workshop on adaptive design in confirmatory clinical trials. London, United Kingdom: 2009. 2nd EMFA-EFPIA

workshop on clinical trials 2009: Introduction. Apr 02.11 Wittes J, Brittain E. The role of internal pilot studies in increasing the efficiency of clinical trials. Statistics in Medicine, 1990; 9:65–7212 gurka MJ, Coffey CS, gurka KK. Internal pilots for observational studies. Biometrical J, 2010, 5:590–60313 Spiegelhalter DJ, Myles JP, Jones DR, Abrams KR. Methods in health service research. An introduction to bayesian methods in health technology assessment. BMJ, 1999;

319:508–12 14 Spiegelhalter DJ, Myles JP, Jones DR, Abrams KR. Bayesian methods in health technology assessment: A review. Health Technol Assess, 2000; 4:1–13015 Food and Drug Administration. guidance for Industry and FDA Staff: guidance for the Use of Bayesian Statistics in Medical Device Clinical Trials, 2010. Accessed 10

October, 2013 at: http://www.fda.gov/medicaldevices/deviceregulationandguidance/guidancedocuments/ucm071072.htm16 Howard g, Coffey CS, Cutter gR. Is Bayesian analysis ready for use in phase III randomized clinical trials? Beware the sound of the sirens. Stroke, 2005; 36:1622–317 Sterne JA, Davey Smith g. Sifting the evidence – What's wrong with significance tests? BMJ, 2001; 322:226–31

Ensuring the validity and representativeness of epidemiological studies based on real-world data is essential to maximizing the value they deliver. Anticipating and managing their challenges is a pivotal part of this process.Here we consider the issue of internal validity and approaches to dealing with error, bias and confounding.

Massoud Toussi, MD, MSC, PHD, MBA is Principal and Medical Director RWE Solutions & HEOR, IMS [email protected]

The author

INSIGHTS | BIAS IN EPIDEMIOLOgICAL STUDIES


How to deal with random error, bias and confounding?The ultimate objective of an epidemiological study is to provide an estimate of a truevalue in the target population – known as accuracy of the estimation. In order to beaccurate, studies should be valid and precise. As many aspects of real-world studiescannot be controlled, it is important to be aware of different types of error that mayoccur in these studies and to know how to deal with them.

ERROR, BIAS, PRECISION AND VALIDITY An error is the difference between an observed or calculated value and a true value. It is often categorized as random orsystematic. A random error occurs by chance and represents the incapacity to produce similar results when theobservation is repeated several times. A systematic error or bias, on the other hand, occurs by design and represents adeviation of several independent estimates from the true value (Figure 1).

Systematic errors are often more difficult to discover than random errors. For example, repeated measurements of aperson’s blood pressure by an imprecise sphygmomanometer will produce a variety of estimates around the true bloodpressure. The wide distribution of results may trigger research for random error. Using a more precise device will reducethe random error and narrow the distribution of estimates. However, if the sphygmomanometer is not correctlycalibrated, it will generate estimates that are systematically deviated, for example higher by 5 mmHg, from the trueblood pressure. It is more challenging to discover the calibration error here because the true blood pressure of theindividual is not known (this is why we want to measure it). Moreover, when the device is precise, ie, generates similarresults in multiple measures, it drives less doubt towards the validity of its results (see Figure 2 on following page).

Enhancing the validity ofepidemiological studies

BIAS IN EPIDEMIOLOgICAL STUDIES | INSIGHTS


FIGURE 1: RANDOM AND SYSTEMATIC ERROR

A

Increased random error Increased systematic error

B C D




As seen in the example, precision is the opposite of random error and validity is the opposite of systematic error. It isexpressed by two means: 1. Internal validity, ie, validity of the results with regards to the source population

2. External validity, ie, validity of the results with regards to outside populations

Internal validity is generally considered to be necessary for external validity.

VIOLATIONS OF INTERNAL VALIDITY The internal validity of a study can be violated in a number of ways which can be classified into three key categories:selection bias, information bias and confounding.

Selection biasSelection bias is a systematic error induced by the selection of subjects who are different from those of the targetpopulation. It can be of various origins:

• Recruitment bias: Occurs when the probability that subjects enter into the study is related to one or more outcomes.This is particularly the case when subjects are recruited from healthcare institutions (hospitals, medical offices, etc).

• Incidence-prevalence, selective survival or Neyman bias: Occurs because asymptomatic patients and those whoexperience episodes of acute diseases are less likely to be enrolled in the study. For example, in a case-control studyevaluating the relationship between tobacco smoking and acute myocardial infarction (AMI) in which cases areinterviewed one week after the coronary attack, if smoker patients with AMI die more frequently before the end ofthe first week, alive cases will show a lower frequency of smoking.

• Self-selection bias: Occurs because the characteristics of volunteers who spontaneously participate in a study maybe different from those who decide not to participate.

FIGURE 2: VALIDITY AND PRECISION

A=High validity and low precision; B=Low validity and high precision; C=High validity and low precision; D=Low validity and low precision.The dotted line represents the true value. Situation B is difficult to discover when the expected true value is unknown.

A B

C D

Measure

Frequency

PRECISION

VALIDITY

High

High Low

Low



• berckson’s or admission bias: Occurs when both the exposure and the outcome affect selection, eg, when thestudy is conducted only among hospitalized patients.

• Healthy worker bias or effect: Occurs when study participants are selected from a population of active individuals.It is linked to the fact that in the workplace, individuals are able to hold a regular job and are therefore a priori inbetter health than the general population in which there are patients who are unable to hold a job.

• Immortal time bias: Occurs when cases identified through screening are less serious and evolve more slowly thanthose diagnosed in routine practice and who are symptomatic (see article on page 34 for a detailed review of thistype of bias).

• Losses/withdrawals to follow-up bias: Occurs when participants of a study drop out before completion and thuscannot be found later in the monitoring of the selected cohort. This is especially more frequent when theobservation period is long, and thus removal, death, etc, are more likely to occur.

• Attrition bias: Is due to differences between the initial and final groups, itself related to exclusion from the study ortreatment interruption.

• Prothopatic bias: Occurs when a particular treatment or exposure was started, stopped, or otherwise changedbecause of a manifestation caused by a disease or outcome.

• Non-response bias: Occurs when participants are different from non-participants.

• Publication bias: This bias concerns mainly systematic reviews and is closely related to selection bias. It occursbecause the studies with interesting findings (often significant or those which confirm existing prejudices) are morelikely to be reported or published than those which confirm the null hypothesis.

Information biasInformation bias is a systematic error that occurs when the measurement or the observation of a phenomenon isincorrect and leads to misclassifying subjects. It can take two different forms:

1. Differential misclassification

Differential misclassification arises when the probability of misclassification is not the same between two groups.Several types of differential misclassification exist:

• Detection bias: Occurs when the procedures for exposure assessment are not the same in cases and controls orwhen the procedures for assessment and follow-up of the disease are not the same between exposed and non-exposed subjects. For example, when the criteria used to identify subjects with disease are sensitive, but notspecific, the group of non-diseased subjects will contain some individuals with the disease. This is important indatabase studies in which algorithms used for detection of patients are sometimes not validated before theconduct of the study.

• Recall bias: Is due to the fact that subjects suffering from a disease (cases) may have better memory of their pastexposures than those who are not suffering (controls). For example, a cancer patient or a mother whose child isborn with malformations may have a very detailed memory of all the past events which could have led to theoutcome.

• Investigator or observer bias: Occurs when the prior knowledge of suspected exposure factors or pathologiesconcerned can influence the intensity of the research for exposure factors or lead to further investigation of thesubjects who are sick. Sometimes, this bias is caused by a conflict of interest arising from passionate beliefs held bythe investigators about the relationship between a cause and an effect. A particular situation is when the measureof an exposure, such as blood pressure, influences its value (apprehension bias).

• Underreporting bias: Is common with socially undesirable behaviors, such as alcohol consumption. It is alsocommon in studies using automated databases of electronic medical records, as physicians tend to recordabnormal values more often.




2. Non-differential misclassification

Non-differential misclassification occurs when the probability of error is the same for exposure and outcome groups. Itis often due to poor quality of instruments of measurement or study processes and may lead to decreased strength ofthe association between the exposure and the outcome. For binary outcomes, this bias moves the estimate toward thenull hypothesis.

ConfoundingConfounding arises when the estimate of a measure of association between an exposure and an outcome is distortedby the effect of one or several extraneous variables, called confounders, which affect both exposure and outcome.

ADDRESSING RANDOM ERROR, BIAS AND CONFOUNDING There are approaches and techniques for avoiding and managing the potential for random error, bias and confounding.They can be applied before, during and after epidemiological studies.

Random error: Increasing the number of individualsand observations before the study can help inreducing random error. Hence, the greater the samplesize, the more precise will be the study (Figure 3).During the study, random error can be reduced byuse of appropriate training of the study team andinvestigators, implementation of relevant standardoperating procedures and quality control. After thestudy, the statistical precision can be calculated inmost cases.

Selection bias: Randomization before the study isgenerally the best way to avoid selection bias. Afterthe study, it is difficult to compensate for itsoccurrence. However, it can be useful to try todemonstrate a potential selection bias by comparingparticipants with non-participants. Sampleadjustment can sometimes help in balancingsampling problems.

Publication bias: Specific measures for avoiding publication bias before a meta-analysis include the implementation ofresearch standards, pre-registration of protocols (as for randomized trials), and use of networking for data collectionabout published and unpublished studies. The European Network of Pharmacoepidemiology and Pharmacovigilance(ENCePP) and its e-register of studies are examples of such initiatives that can help to reduce problems related topublication bias. Funnel plots can help in displaying a potential publication bias in some circumstances during themeta analysis.

FIGURE 3: RELATIONSHIP BETWEEN STUDY SIZE AND RANDOM ERROR

Systematic error

Random error

Study size

Error

Being aware of the different types of bias and confounding during the designphase of a real-world study can help to avoid them as much as possible.



Information bias: To avoid information bias, the following can help before the start of the study: randomization,blinding process, use of diseased-controls in case-control studies, use of standardized measurement methods, precisedefinition of study processes (including the outcome measures and algorithms for disease definitions) andtransparency on conflict of interest. During the study, all possible effort should be made to minimize refusals, obtaininformation about lost-to-follow-up subjects, and multiply the measurements and quality controls. After the study, asfor selection bias, it is difficult to compensate for either differential or non-differential misclassification. However, whatcan help is to try to obtain at least partial information on the individuals who refused to participate or who are lost tofollow-up in order to analyze their characteristics and compare them to other participants. This will sometimes enablethe results to be corrected, or the extent of the bias to be estimated.

Confounding: The potential for confounding can be reduced before the study through the use of a control group,matching, randomization and careful selection of eligibility criteria, all of which can help in building more homogenousgroups. After the study, confounding can be dealt with in the statistical analysis through adjustment of results over theconfounder, a posteri matching, or the use of propensity scores.

CONCLUSIONWhile it is difficult to handle all the nuances of a real-life study, being aware of the different types of bias andconfounding during the design phase can help to avoid them as much as possible. Once the study has started, it is hardto control for bias. However, confounding can be controlled in the statistical analyses. After the end of the study, it isimportant to screen for potential biases that may have occurred, and to try to estimate their effect on the study results.One good way of doing this is to ask the following questions on the study results, as illustrated in Figure 4:

• Could the association be affected due to a selection or information bias?

• Could it be caused by chance?

• Could it be due to the effect of a third factor(confounding)?

• Could it be explained by a plausible causalrelationship? •

REFERENCES• Delgado-Rodríguez M, Llorca J. Bias. J Epidemiol Community Health. 2004 Aug 1; 58(8):635–41• Haynes RB, Ovid Technologies I. Clinical epidemiology: How to do clinical practice research. Philadelphia: Lippincott Williams & Wilkins; 2006 • Rothman KJ, greenland S, Lash TL. Modern epidemiology. Philadelphia: Wolters Kluwer / Lippincott Williams & Wilkins; 2008• Strom BL, Kimmel SE, Hennessy S. Textbook of pharmacoepidemiology. Chichester, West Sussex: John Wiley & Sons; 2013

FIGURE 4: QUESTIONS TO CONSIDER WHEN AN ASSOCIATION IS OBSERVED

Bias in selection ormeasurement YES

YES

ASSOCIATION

CAUSE

LIKELY

NO

NO

UNLIKELYChance

Confounding

Cause

EXPLANATION FINDING

Epidemiological studies offer growing potential to determine real-world drugeffectiveness and safety but their findings can be compromised by elements ofbias, especially in relation to the concept of immortal time. Understanding thestatistical approaches to correcting for this challenge in study design is key toensuring the validity of the insights derived.

Carl de Moor, PHD is Senior Principal Epidemiology, RWE Solutions, IMS [email protected]

The author



Ensuring relevant and valid study results Randomized controlled trials (RCTs) are considered the gold standard for evaluatingthe efficacy of medical interventions. However, they also may suffer from limitedgeneralizability because: the clinical trial setting does not reflect real-world practiceand decision making; the enrolled patients are homogeneous with respect todemographic and medical characteristics; and the trial size and duration oftenpreclude reliable detection of rare events.In recent years, it has become widely accepted that post-approval epidemiological studies are necessary forunderstanding real-world product effectiveness and safety. Study conduct has been facilitated by the growingavailability and use of computerized healthcare utilization (HCU) data, including medical claims and electronic medicalrecords (EMR). Epidemiological investigations have significant advantages over RCTs insofar as they better reflect routinecare; incorporate heterogeneous patient populations with diverse co-morbidities and concomitant medications; andmake use of larger sample sizes. However, without careful attention to study design, they also may suffer from biases thatcompromise internal validity, including patient selection bias, confounding bias, and immortal time bias.1

IMMORTAL TIME BIASImmortal time refers to a period of time in patient follow-up when the outcome of interest cannot occur.2 It arises inepidemiologic cohort studies when a patient must pass through a waiting period to receive a product or treatment andwhen that waiting period is erroneously considered as time exposed to the product or treatment. Immortal time wasfirst described in the literature by gail3 in the context of heart transplant studies that showed a significant survivalbenefit associated with transplantation. gail demonstrated that the survival benefit of transplantation could beexplained in part by the fact that the time the patients waited for a suitable donor was included in their survival time. Inother words, during the time the transplant patient waited for a suitable donor, the patient was “immortal” as deathcould not have occurred, or the patient would have been classified as a non-transplant patient.

Immortal time bias isdepicted graphically inFigure 1. Here, twotreatment groups (group A and group B) are shown with the samecohort entry point, butwith one group (group A)having a period ofimmortal time where the event of interestcannot occur.

In group A, measuring the time to the event of interest from the beginning of cohort entry creates an obvious biasrelative to the true time to event, often producing the appearance of a strong protective treatment effect where nonemay exist.

Challenges of immortal time bias in epidemiology

FIGURE 1: SCHEMATIC OF IMMORTAL TIME BIAS

GROUP A

GROUP B

Immortal time Time to event

Time to event

Cohort entry Start of exposure






Immortal time bias can occur in a number of different study designs. It arises during cohort formation from errors indefining the start of follow-up, potentially affecting studies regardless of whether cohort entry is based on a fixed date,a health event, product exposure, or combinations thereof.

Some examples that lead to immortal time bias include:

• Start of follow-up is defined by the fixed date, but exposure is defined by the number of prescriptions following afixed date (eg, the mean number of inhaled corticosteroid prescriptions in asthma patients after a fixed date).

• Start of follow-up is defined by a clinical event, but exposure is defined by start of prescription following the clinicalevent (eg, cardiac drugs dispensed following hospital discharge for acute myocardial infarction).

• Start of follow-up in the treatment group is defined by the first prescription of a product, but the follow-up in thecomparison group is based on a date of diagnosis or start of another treatment (eg, COPD patients prescribedinhaled corticosteroids compared with bronchodilators, where many of the corticosteroids patients had a prior dateof diagnosis or had received bronchodilators prior to corticosteroids).

Immortal time bias: HRT and CHDTo clarify these concepts, the following is a more detailed example of immortal time bias associated with hormonereplacement therapy (HRT) and coronary heart disease (CHD) in women.

HRTs have been shown to be effective for menopause, mitigating symptoms such as hot flashes and joint pain, and toprevent bone loss. By 1992, HRTs became some of the most commonly prescribed products, not only because of theirbeneficial effects for menopause, but also because of their putatively beneficial impact on CHD. Multiple cohort studiessupported a positive effect on CHD, with a 1998 meta-analysis showing an average protective relative risk of 0.70 (95%CI; 0.65-0.75) associated with estrogen only HRTs and 0.66 (95% CI: 0.53-0.84) associated with estrogen-progestin HRTs.4

However, in 2002, the Women’s Health Initiative, a clinical trial with over 16,000 women, published 5-year follow-upresults that showed, in fact, an increased risk associated with HRTs, with a relative risk of 1.29 (95% CI: 1.02-1.63) forCHD, and 1.41 (95% CI: 1.07-185) for stroke.5

As Suissa6 described, many of the observational studies concluding that HRTs were beneficial to CHD were subject toimmortal time bias. In one example, a study evaluated the effect of HRT on mortality following coronary artery bypassgraft (CABg) surgery in 1,098 women between 1972 and 1989. Cohort members were classified as having received HRTif they were treated with HRT at the time of admission or if they reported HRT during follow-up. Follow-up beganfollowing CABg. A total of 92 women were reported to have received HRT, 42 at the time of CABg and 50 during follow-up. Cox regression analysis showed a 62% reduction in mortality associated with HRTs, with a relative risk of 0.38,p < .001. However, the immortal time bias was created when the 50 women who reported use of HRT during follow-upwere classified as HRT patients from the time of CABg (Figure 2). In other words, it was possible in cases where HRT usewas reported for the first time even several years after CABg, that women were considered to have been exposed toHRT all along, and the person-years of exposure prior to HRT was included in the HRT group.

FIGURE 2: IMMORTAL TIME BIAS IN THE HRT AND CHD STUDY

Source: Adapted from Suissa S. Randomized trials built on sand: Examples from COPD, hormone therapy, and cancer.Rambam Maimonides Medical J, 2012, Jul 31; 3(3)

HRT {Non-HRT

n = 50

n = 42

n = 1,006

Immortal time Time to event

Time to event

Time to event

Cohort entry (CABg) HRT



STATISTICAL REMEDIESStatistically, a number of methods have been employed to adjust for the effects of immortal time bias. These include:excluding the immortal time from the analysis or including it in the unexposed group; conditional landmark analysis;time-varying Cox regression modeling; and Inverse Probability Weighting (IPW).7 Each of these is discussed below.

Excluding immortal time or adding immortal time to unexposed groupPerhaps the simplest approach to try and correct for immortal time bias is excluding the immortal time from theexposed group or moving the immortal time from the exposed group to the unexposed group. However, simpleexclusion of the immortal time results in biased comparisons because this time represents unexposed time and mustbe included with the unexposed cohort. On the other hand, adding the immortal time to the unexposed group mayyield unbiased estimates if the risk of the event of interest is constant over time, but may result in biased estimates ifthe risk changes over time.

Conditional landmark analysisLandmark analysis defines a specific point in time during follow-up, called the landmark time. The patient groups aredefined at this point in time and then compared, based on the time to the event going forward from the landmark time.Patients who experience the event of interest prior to the landmark time are dropped from the comparison. Thus, theresults must be interpreted relative to that time point. This approach has some drawbacks, including the fact that thelandmark time may be clinically arbitrary, and there could be some loss of power from dropping cases. Researchers whouse this approach often select more than one landmark time-point to evaluate consistency of results.

Time-varying Cox regression modelingTime-varying Cox regression employs a time-varying covariate in the standard Cox regression model representing timedependent exposure status. In this approach, all patients and all data are used. All patients initially would be classifiedas unexposed, and patients in the treatment group would change classification to exposed later in the follow-up. Thisapproach produces unbiased group comparisons and allows the underlying risk of the event of interest to vary in time,although the proportional hazards assumption must still hold. It is often considered the gold standard approach toadjusting for immortal time bias.

Inverse probability weightingIPW involves a two-stage process. In the first stage, the follow-up time is divided into intervals. The probability ofpatient exposure during that interval – the interval dependent propensity score – is predicted, based on patientdemographic and clinical characteristics, either from baseline or time-varying characteristics up to the start of theinterval. In the second stage, the inverse of the predicted probabilities are then used as weights in a Cox regressionmodel to compare the groups. Like time-varying Cox regression, IPW makes use of all the patients and all the data. Itmay be more complex to employ but, like other propensity score methods, can be used to understand whethertreatment effectiveness varies by the likelihood of receiving the treatment.

CONCLUSIONSComputerized databases have facilitated the growth of epidemiological investigations of drug effectiveness in the real-world setting. In some studies, immortal time bias may arise by virtue of the definitions of cohort entry, productexposure and beginning of the follow-up. Immortal time bias can lead to exaggerated beneficial effects of thetreatment of interest. Fortunately, readily available statistical methods are available to correct for the bias, to yield validestimates of treatment effectiveness •

1 Sturmer T, Funk MJ, Poole CM, Brookhart MA. Nonexperimental comparative effectiveness research using linked healthcare databases. Epidemiology, 2011; 22:298-3012 Suissa, S. Immortal time bias in pharmacoepidemiology. American Journal of Epidemiology, 2008; 167:492-4993 gail MH. Does cardiac transplantation prolong life? A reassessment. Ann Intern Med, 1972; 76:815-8174 Barrett-Connor E, grady D. Hormone replacement therapy, heart disease and other considerations. Ann Rev Public Health, 1998; 19:55-72 5 Writing group for the Women's Health Initiative Investigators: Risks and benefits of estrogen plus progestin in healthy menopausal women. Principal results from the

Women's Health Initiative randomized controlled trial. JAMA, 2002; 288 (3):321-3336 Suissa S. Randomized trials built on sand: Examples from COPD, hormone therapy, and cancer. Rambam Maimonides Medical J, 2012, Jul 31; 3(3) 7 giobbe-Hurder A, gelber R, Regan M. Challenges of guarantee-time bias. Journal of Clinical Oncology, 2013; 31:2963-2970

INSIGHTS | FRENCH DIABETES COHORT

Xavier Ansolabehere, MSC is Senior Consultant, RWE Solutions & HEOR, IMS [email protected]

Philippe Le Jeunne, MD is Medical Director, IMS [email protected]

The authors


Recent clinical guidelines have redefined the benchmarks for diabetesmanagement in France. With questions around the alignment of currentpractice and the new treatment goals, a retrospective study reveals thepower of IMS Diabetes Cohort in enabling pivotal, real-world insights forimproved health approaches and outcomes.

Nathalie Grandfils, MSC is Director, RWE Solutions & HEOR, IMS [email protected]

Real-world insights from IMS LifeLinkTM Diabetes Cohort Diabetes is recognized by the WHO and International Diabetes Federation as asignificant and growing health problem. It already affects more than 371 millionadults worldwide, a figure which is expected to reach 552 million by 2030.1

Type 2 diabetes mellitus (T2DM) is the most common form of the disease, accountingfor around 90% of all diabetes cases. Patients with diabetes are at greater risk for developing numerous co-morbidities, resulting in increased healthcareexpenditures and higher mortality. Around one-third of patients with T2DM develop some form of clinical kidneydamage due to high levels of blood glucose, and patients have a progressively increasing risk of developing chronickidney disease.

Management strategies principally consist of controlling glycemia level which is crucial to preventing or delayingassociated complications2. Control is typically measured on the basis of glycated hemoglobin or HbA1c, which gives anaverage of the blood glucose over a 3-month period. On a day-to-day basis, self-monitoring of blood glucose (SMBg)contributes to improving glycemia control and reducing hypoglycemic risk by allowing self-adjustments.

NEW FRENCH TREATMENT GUIDELINESIn France, new clinical guidelines for T2DM, introduced by the National Authority for Health (Haute Autorité de Santé [HAS]) early in 2013, set out a number of therapeutic recommendations for its management.3 Specifically,these stipulate:

• Diet, exercise and education as the mainstay of any treatment program for T2DM

• Regular re-evaluation of management given the progressive nature of diabetes

• Individualized glycemia targets and glucose-lowering treatments according to specific patient profile

• Strategy based on target gap, efficacy, tolerance and cost of available treatments

• Use of metformin as preferred first-line monotherapy in the absence of contraindications, with priority then givento the association of metformin with a sulfonylurea (SU) when a bi-therapy is needed. Insulin to be prescribed whenoral anti-diabetics do not achieve glycemia target.

• SMBg recommended only for patients treated with anti-diabetics known to induce a hypoglycemia event and onlyif results of SMBg are likely to modify diabetes management.

The guidelines further distinguish five patient profiles, each separated into sub-categories for which the HbA1c targetis individualized. These constitute:

1. Elderly patients

2. Patients with previous cardiovascular (CV) events

3. Patients with renal impairment

4. Pregnant or planned-pregnancy patients

5. T2DM patients without any of the above profiles

FRENCH DIABETES COHORT | INSIGHTS

Promoting best practice treatmentfor Type 2 diabetes in France





GUIDELINES VERSUS CURRENT TREATMENT As the burden of diabetes continues to grow and more new oral anti-diabetic drugs become available, anunderstanding of real-world approaches to treatment is essential to promoting appropriate and cost-effective clinical practice in accordance with current recommendations.

In order to evaluate the consistency between T2DM management by French gPs in 2012 and the 2013 HAS guidelines,determine the adequacy of glycemic targets, and identify opportunities for improvement, a retrospectivepharmacoepidemiological study was conducted using IMS LifeLinkTM Diabetes Cohort linked to the EMR database, IMS Disease Analyzer™. Disease Analyzer is a longitudinal patient database of anonymous data on drug prescribing,systematically and continuously collected during patient visits from a nationally representative sample of about 1,000gPs. The process is strictly observational.

IMS LifeLink Diabetes Cohort is a comprehensive and clinically rich representative sample of patients with Type 1 andType 2 diabetes in France (Figure 1).4 Launched in 2011 and based on the continuous input of data by more than 250 gPs at every patient consultation, the cohort is prospective, longitudinal, open, observational and multi-centric.More than 8,000 diabetes patients are currently included and the number continues to grow. Linkage of DiabetesCohort to Disease Analyzer allows assessment of the way in which diabetes patients are managed in real life by FrenchgPs, with information on patient profiles, therapeutic regimen, treatment duration, adherence, and the patient journeythrough primary and secondary care.

Specifically, the goal was to classify patients with diabetes according to the profiles defined by the HAS guidelines,create a mutually exclusive groups of patients with a common target of HbA1c, evaluate for each of the groups theproportion of patients reaching their HbA1c target, and describe their treatment.

FIGURE 1: IMS LIFELINk DIABETES COHORT CONTAINS RICH CLINICAL DETAIL ON DIABETES PATIENTS IN FRANCE

Patient characteristicsType of diabetes (Type 1, Type 2)

History of diabetes (date of first diagnosis; date of first drug treatment; long-term complications, eg, eye, foot, kidney conditions)

Body Mass Index (BMI)

Risk factors (eg, family history, hypertension, smoking, alcoholabuse, abdominal perimeter, etc)

Diabetes complications (eg, heart disease, stroke, neuropathy,retinopathy, kidney disease, foot problems, etc)

Healthcare pathwaysDiabetologist/Endocrinologist/gP: Exams performed,prescriptions issued, follow-up

Ophthalmologist visit: Exams performed, prescriptions issued,diagnosis given

Podologist visit: Exams performed, prescriptions issued,diagnosis given

Cardiologist visit: Exams performed, prescriptions issued,diagnosis given

Hospitalizations

TreatmentsAnti-diabetes drugs prescribed, dosage, treatment changes and reasons for change

Co-prescriptions, prescriptions of blood glucose meters and test strips

Hypoglycemic events

Lab test results, diagnostic and imaging test findings

HbA1c, fasting blood glucose, albuminuria, proteinuria, urinarycreatinine, serum creatinine, total cholesterol (LDL, HDL),triglycerides concentration.

Eye examination

ECg

Vascular doppler ultrasound scan

FRENCH DIABETES COHORT | INSIGHTS


MethodologyThe transversal study thus investigated T2DM patient profiles (age, disease history and complications, renalimpairment, CV events, co-morbidities), HbA1c, hypoglycemic risk, SMBg and body mass index (BMI) according tocurrent treatment. Cohorts were defined for four of the five HAS patient categories, the exception being pregnant orplanned pregnancy women, for whom it was not possible to create a cohort from the database. Medical history anddrug markers were used as inclusion criteria and all inclusion decision algorithms were validated by a clinical expert.

Variables of Diabetes Cohort were used as measurements of: renal function based on the most recently availablecreatininemia value; duration of diabetes; risk factors; complications; hypoglycemic risk; chronic co-morbidities and CV diseases; hypoglycemia events; level of HbA1c; therapeutic scheme, classes and molecule.

Current treatment was defined as the last anti-diabetic treatment prescribed by the gP during the study period on thebasis of: therapeutic scheme (mono/bi/tri/insulin); therapeutic class or association of therapeutic classes (metformin,glinides, sulfonylurea, DDP-4 inhibitor (DDP-4i), gLP-1 receptor agonist, insulin); molecule(s) and daily dosages. The main part of the study was a cross-sectional analysis of the last gP consultation. Some analyses required the use oflongitudinal data to assess evolution of certain clinical parameters, such as weight change.

Key findingsA total of 6,680 T2DM patients were included in the 2012 study cohort (56% male; 44% female). Of these, 25% wereaged 75 years or over (the mean age was 66.6 years); 19% had moderate to severe renal impairment; and 56% had ahistory of CV disease. Among the key findings from the study were:

• HbA1c targets: The new HAS guidelines divide patients into several categories with associated HbA1c objectivesvarying from 6.5% to 9%, according to age, frailty, diabetes history and co-morbidities. The analysis showed thatcurrently in France, only 39% of patients have an HbA1c target below 7 while 43% and 18% respectively have atarget at 8% and 9%. This suggests that most patients, depending on their health status and presence of co-morbidities, have the least restrictive HbA1c objectives.

given the modulation of glycemic objectives as set out in the new guidelines, more patients now reach their HbA1ctargets (74%) when compared against previous guidelines (about 55%).5 This proportion increases with the HbA1c target:96% of patients with a 9% HbA1c target reach their objective versus 37% of patients with a 6.5% objective (Figure 2).

• Risk of hypoglycemia: It appears that 55% of the currently treated diabetic population in France is at risk ofhypoglycemia, with age and long diabetes history being the major risk factors.

• Appropriate prescribing: It is notable that, among patients who reach their target HbA1c objective, 11% areprescribed contraindicated treatments. Among patients who do not reach their HbA1c target (26%), 33% aretreated with a monotherapy, 34% with a bi-therapy and 33% with a tri-therapy or more, suggesting the need insome cases for drug escalation.

• Diabetes duration and complications such as renal impairment are predictive factors of using SMBg.Nevertheless, hypoglycemia risk as well as renal impairment do not seem to be appropriately assessed or managedby French gPs; T2DM patients who suffer from renal impairment or who are at risk of hypoglycemia for anotherreason are often treated with contraindicated treatments or without dosage adaptation. In this representativecohort, SUs were found to be prescribed with no dosage adjustment to patients and even more frequently topatients at risk of hypoglycemia (60.6%) while DDP4i were prescribed to “healthy patients” mostly not at risk ofhypoglycemia (51.5%).

The database analysis using Diabetes Cohort offers important real-worldinsights into the current management of T2DM by gPs in France in relationto recommended guidance.




FIGURE 2: kEY FINDINGS IN RELATION TO HBA1C TARGET

TOTAL Patients T2DM

Elderly ≥ 75 years

Healthy, vigorous

Frail

With multiple disease

Patients < 75 years with renal impairment

Moderate impairment

Severe impairment

Patients <75 years, without renal impairment, with history of CV events

Non-advanced vascular complications

Advanced vascular complications

Other patients

Recently diagnosed, uncomplicated

Longstanding diabetes (>10 years) complicated

Other cases

6,680

1,706

137

386

1,183

613

249

364

2,235

146

2,089

2,126

226

1,818

82

n % TargetHbA1c

% of patient whoreach target

100%

25.5%

2.1

5.8

17.7

9.2%

3.7

5.4

33.5%

2.2

31.3

31.8%

3.4

27.2

1.2

≤ 7%

≤ 8%

≤ 9%

≤ 7%

≤ 8%

≤ 7%

≤ 8%

≤ 6.5%

≤ 8%

≤ 7%

74%

91%

63%

85%

96%

61%

60%

90%

79%

57%

82%

58%

37%

93%

59%

CONCLUSIONNotwithstanding its inherent limitations, the database analysis using Diabetes Cohort offers important real-worldinsights into the current management of T2DM by gPs in France in relation to recommended guidance. Modification ofthe glycemic control objective depending on the patient’s profile seems to allow a higher percentage of patients to be“at goal”. However, while more patients attempt goals, some do not receive the appropriate treatment. Patients not atgoal for their individual objectives are not always treated with recommended drugs. Adaptation of treatment accordingto the evolution of both diabetes and patient profile appears not to be systematic. And while SMBg is required forpatients on insulin, SUs and glinides, in practice its use does not seem to be related in any way to recommendations.

These revelations, made possible through the cohort analysis, have important implications from a clinical andeconomic perspective, highlighting key areas of unmet need and confirming the role and value of real-world data insupporting best clinical practice. Moreover, this type of database enables a wide range of questions to be answered in avery short timeframe. The results of this large-scale study, for instance, were available less than four months after thenew guidelines were introduced •

1 International Diabetes Federation. Diabetes Atlas Update, 2012. Accessed 25 October, 2013 at: http://www.idf.org2 Clar C, Barnard K, Cummins E, et al. Self-monitoring of blood glucose in type 2 diabetes: Systematic review. Health Technol Assess, 2010;14(12):1–1403 HAS et ANSM, Stratégie médicamenteuse du contrôle glycémique du diabète de type 2 Méthode « Recommandations pour la pratique clinique »,

guidelines January, 20134 Le Jeunne P, IMS LifeLinkTM Cohorte Diabète : Validité et utilité, La revue du praticien médecine générale, tome 26, N° 889, November, 20125 Echantillon national témoin représentatif des personnes diabétiques (Entred) 2007-2010 Diaporama : Caractéristiques des personnes diabétiques, risque vasculaire,

complications et prise en charge médicale (mise à jour le 12 March, 2010).http://www.invs.sante.fr/surveillance/diabete/entred_2007_2010/resultats_metropole_principaux.htm

INTERPRETINg ICERs | INSIGHTS


Challenges in interpreting cost-effectiveness (CE) results in theSouth-West quadrant of CE planes

Anamaria Jugrin, MSC is Consultant RWE Solutions & HEOR, IMS [email protected]

The authors

Mark Lamotte, MD is Senior Principal RWE Solutions & HEOR, IMS [email protected]

Can ICERs falling in the South-West quadrant be considered cost-effective?It is established practice in developed healthcare systems to evaluate newinterventions under the framework of incremental cost-effectiveness ratios (ICERs),whereby costs and health outcomes of new interventions are compared with thoseof the existing standard of care. ICERs can be easily interpreted on a 2 by 2 matrix,illustrated by a cost-effectiveness (CE) plane (Figure 1).

• New interventions that are both cost-saving and elicit health gains over existing interventions should always beconsidered acceptable (ICERs fall in the South-East (SE) quadrant of the CE plane – the dominant position).

• New interventions that come at higher costs and do worse in terms of health outcomes should never be consideredacceptable (ICERs fall in the North-West (NW) quadrant – the dominated position).

In these two instances, the interpretation of CE results is straightforward. However, most new interventions will come athigher costs and elicit higher benefits over the existing interventions (ICER falls in the North-East (NE) quadrant). Insuch cases, ICERs are typically judged in the context of acceptability levels, or CE willingness-to-pay (WTP) thresholds,and the question to be answered is: do the additional benefits offset the additional costs? Less often is the case that a new intervention is cost-saving and projects a certain loss in health with respect to thecurrent standard of care (ICER falls in the South-West (SW) quadrant). Yet, with increasing use of uncertainty analyses,and the construct and presentation of CE scatter plots, it isoften the case that part of the scatter plots fall in the SWquadrant of the plane. In such cases, the interpretation ofICERs is less straightforward and can be hampered by twomain issues:

1. Positive incremental costs and positive incrementalbenefits in the NE quadrant, and negative incrementalcosts and negative incremental benefits in the SWquadrant, both yield positive ICERs. Should ICERsresulting from negative incremental costs andnegative incremental benefits be read in the same wayas ICERs resulting from positive incremental costs andpositive incremental benefits? Is the decision rule inthe SW quadrant the same as in the NE quadrant?

2. Having overcome this issue, the next question is: arewe clear at what thresholds we judge ICERs in the SWquadrant? Are those thresholds the same as forpositive costs and positive benefits (NE quadrant)?

Challenges in interpreting cost-effectiveness (CE) results in theSouth-West quadrant of CE planes

North-Eastquadrant

Incremental costs

IncrementalQALYs

South-Eastquadrant

North-Westquadrant

South-Westquadrant

FIGURE 1: ILLUSTRATION OF A COST-EFFECTIVENESS PLANE

INSIGHTS | INTERPRETINg ICERs





CHALLENGE 1: IS THE DECISION RULE IN THE SW QUADRANT THE SAME AS IN THE NE QUADRANT?Consider the following two scenarios:

• Scenario A: Intervention X results in an increase in costs with 15,000 Euro and an increase in quality-adjusted life-years (QALYs) with 0.80 QALY, compared with Intervention Y

• Scenario B: Intervention X results in cost savings of -50,000 Euro but a loss in QALY of -0.2 QALY, compared withIntervention Y.

The projected ICER is positive in both scenarios: 18,750 Euro/QALY and 250,000 Euro/QALY respectively.

At an assumed threshold of 30,000 Euro/QALY, a naive interpretation of these results would consider Intervention X as being cost-effective in Scenario A and not cost-effective in Scenario B. Is this the correct interpretation? No: thedecision rule in the SW quadrant is the exact opposite of that in the NE quadrant.

Net monetary benefitsTo illustrate this, results can be re-arranged in terms of net monetary benefits (NMB). The NMB can be defined as theincrease in effectiveness (ΔE), multiplied by the amount the decision maker is willing to pay per unit of increasedeffectiveness (WTP), less the increase in cost (ΔC): NMB = WTP * ΔE – ΔC. A program is deemed cost-effective if theNMB>0.In the current case, Intervention X yields a NMBA of: 30,000 Euro/QALYs * 0.80 QALY - 15,000 Euro = 9,000 Euro inScenario A; in Scenario B it yields a NMBB of: 30,000 Euro/QALYs * -0.20 QALY – (-50,000) Euro = 44,000 Euro. Thus,Intervention X seems more cost-effective in Scenario B than in Scenario A (NMBB>NMBA). Taking a lower acceptabilitylevel, say 22,000 Euro, Intervention X can no longer be considered cost-effective under Scenario A, while in Scenario B it remains acceptable.

A correct interpretation is given when all ICERs falling to the right side of the CE threshold line are considered as beingcost-effective (Figure 2).1 This means that in the SW quadrant, ICERs need to be higher than the threshold value inorder to be considered cost-effective.

A solution that solves the ambiguity of the SW quadrant isuse of the NMB approach. The NMB “unambiguously” sortsout the acceptability of CE results on the CE plane, assuggested by Briggs, et al.2 Indeed, if the decision rule isbased on the NMB results, it is enough to say that anintervention can be considered cost-effective when theproduct of WTP and ΔE is higher than ΔC (ie, NMB>0).

The same decision rule can be applied when interpretingprobabilistic results of CE analyses. More specifically, theNMB approach can and should be used whendetermining the level of CE acceptability for theconstruction and illustration of cost-effectivenessacceptability curves (CEACs), thus ruling out any potentialfor errors.

FIGURE 2: ALL ICERS TO THE RIGHT OF THE CE THRESHOLD ARECONSIDERED COST-EFFECTIVE

ICERs in this area considered cost-e�ective

A: ICER is 18,750 cost/QALY

-2.00 -1.50 -1.00 -0.50 0.50 1.00 1.50 2.00

NMB = 9,000

NMB = 44,000B: ICER is 250,000 cost/QALY

Incremental QALYs

-60,000

-40,000

-20,000

0

20,000

40,000

60,000

Incr

emen

tal c

osts

€

INSIGHTS | INTERPRETINg ICERs


CHALLENGE 2: ARE WE USING THE SAME THRESHOLD TO JUDGE ICERs IN THE SW QUADRANT?In the earlier example, it was assumed that theacceptability threshold is a straight line through the originof the CE plane. However, this may or may not be the case,bringing further challenges to the interpretation of resultsin the SW quadrant.

Based on the earlier work of Black in 1990,3 Drummond etal,1 and Briggs et al,2 illustrated the CE plane with the WTPthreshold drawn as a straight line through the origin. AllICERs falling to the right side of the threshold (whether inthe SW, NE or SE quadrants) are deemed cost-effective(Figure 2). CEACs constructed in TreeAge Pro® will, bydefault, apply the same acceptability level in the NE andSW quadrants.

In 2002, O’Brien et al4 suggested that the acceptability linemay not be straight in the origin, but rather that there is adisparity between WTP for an additional unit of healthgained (the value of the threshold line in the NE quadrant)and willingness to accept (WTA) health forgone (the valueof the threshold line in the SW quadrant). They showedthat individuals consistently attached a higher value tolosses than to gains (WTA>WTP).

Indeed, the acceptability threshold has, in fact, differentmeanings in the two quadrants:

• In the NE quadrant, the threshold reflects theadditional price we are willing to pay for a unit ofhealth gained = WTP

• In the SW quadrant, the threshold shows at what priceare we willing to give up a unit of health for it to beallocated elsewhere = WTA

Falling on the observed disparity between WTP and WTA,the threshold line becomes kinked in the origin of the CE plane (Figure 3).

In 2006, Annemans5 illustrated the CE plane with thethreshold line drawn in the NE quadrant only, and arguedthat given uncertainty as to whether policy makers wouldbe willing to forgo health losses, it is not possible tocontinue the threshold line into the SW quadrant. Thethreshold in the SW quadrant coincides with the 0y axisand WTA is infinite (Figure 4).

FIGURE 3: THRESHOLD IS kINkED IN THE ORIGIN

FIGURE 4: THRESHOLD COINCIDES WITH THE 0Y AXIS

ICERs in this area considered cost-e�ective

-60,000

-40,000

-20,000

0

20,000

40,000

60,000

-2.00 -1.50 -1.00 -0.50 0.50 1.00 1.50 2.00 Incremental QALYs

Incr

emen

tal c

osts

€

Reject

Accept

WTA>WTP

WTP=WTA

Source: Annemans L. L’économic de la santépour non économists, Academia Press, 2006

-60,000

-40,000

-20,000

0

20,000

40,000

60,000

-2.00 -1.50 -1.00 -0.50 0.50 1.00 1.50 2.00 Incremental QALYs

Incr

emen

tal c

osts

€

Reject Decision

Decision Accept

Accept

Threshold line

A correct and transparent interpretation of the results in the SW quadrant, and explicitillustration and description of cost-effectiveness acceptability curves, are warranted.



IMPLICATIONSThe assumption that there is little or no (political)willingness to “take away” existing health has ledresearchers and industry alike to sometimes “ignore” ICERsfalling in the SW quadrant or to label them as inferior. Thishas direct implications on the way CEACs are constructedand presented. If a proportion of the probabilistic ICERsfall in the SW quadrant – and all results in the SWquadrant are deemed as not being cost-effective – theCEAC shows a lower probability of the intervention to beconsidered cost-effective, independent of the valueattributed to the threshold in the NE quadrant.

On the other hand, if WTA is lower than infinite (and atminimum equal to the WTP), the proportion ofprobabilistic ICERs falling to the right side of the thresholdline in the SW quadrant are counted under the CEACcurve, and the likelihood of the intervention to beconsidered cost-effective increases. Unfortunately, whenresults are presented on CEACs, it is almost never clearwhether results in the SW quadrant were included and, ifthey were, how they were judged (under or above thethreshold). This can generate false discrepancies in thepresentation of probabilistic analyses results and canhamper the decision-making process.

Figure 5 illustrates how CEAC lines can change at various levels of WTA.

CONCLUSIONCE planes are a useful tool in visualizing the results of CE analyses. Nonetheless, issues arise when interpreting results inthe four quadrants. As shown, these are due to the nature of the CE plane (Challenge 1) and to a lack ofagreement/knowledge with regards to the level of acceptability in the SW quadrant (Challenge 2). Challenge 1 can beeasily overcome through a correct interpretation of the ICER points using the NMB approach.

Challenge 2 gives rise to more theoretical questions: are we willing to forgo some health units in order to recovermonetary resources which can be then allocated elsewhere? If so, does the monetary value we attach to losses equal themonetary value we attribute to (health) gains? Whose WTP/WTA should we measure: individual, societal or political?

In our view, the assumption of an infinite WTA does not hold under the scope of welfare economics, which maximizes asocial welfare function under limited resources. However, a certain value of a WTA threshold is not easy to acknowledge.Unfortunately, less transparent ways seem to prevail when the level of healthcare assistance is reduced (eg, waiting lists).With the recent economic crisis, such measures are likely to increase. With these, the need for appropriate, transparenttools to inform and support decision making will also increase. In this context, we strongly believe that ICERs falling inthe SW quadrant should, at least, not be completely ignored by health economists. A correct and transparentinterpretation of the results in the SW quadrant, and explicit illustration and description of CEACs, are warranted •

FIGURE 5: CEACS DEFINED BY DIFFERENT LEVELS OF WTA

40

50

60

70

0 10k 20k 30k 40k 50k 60k 70kPr

obab

ility

cost

-e�e

ctiv

e %

Willingness-to-pay threshold €

WTA =∞ WTA=30k WTA=50k WTA=100k

1 Drummond MF, Sculpher MJ, Torrance gW, O’Brien BJ, Stoddart gL. Methods for the economic evaluation of health care programmes, Oxford University Press, Fifthedition, 2005

2 Briggs A, Claxton K, Sculpher M. Decision modelling for health economic evaluation, Oxford University Press, 20063 Black WC. The CE plane: A graphic representation of cost-effectiveness. Med Decis Making. 1990, Jul-Sep;10(3):212-4 4 O’Brien BJ, gertsen K, Willan AR, Faulkner LA. Is there a kink in consumers’ threshold value for cost-effectiveness in health care? Health Econ, 2002, Mar; 11(2):175-80 5 Annemans L. L’économic de la santé pour non économists, Academia Press, 2006

Providing equitable, accessible and affordable healthcare is a critical priority forIndia today but one that remains largely aspirational, despite the progress beingmade. A new study from the IMS Institute for Healthcare Informatics revealspotential to reduce out-of-pocket expenditure on healthcare by more than 40%through improvements across four key, interrelated dimensions of access.

Murray Aitken, MBA is Executive Director, IMS Institute for Healthcare [email protected]

The author

PAGE 48 IMS HEOR & REAL-WORLD EVIDENCE SOLUTIONSPage 1 IMS HEALTH ECONOMICS AND OUTCOMES RESEARCH

INSIGHTS | HEALTHCARE ACCESS IN INDIA


A roadmap for future actionThe extent of change and improvement in India’s healthcare system over the pastdecade is remarkable, with multiple government and private-sector programsundertaken to improve healthcare access. Some of these have proved enormously successful: the maternal mortality rate has decreased by ~50%; the infantmortality rate fell by more than 25% from 2000-2009; immunization coverage has increased significantly, in the case ofHepatitis B from 68% in 2005 to 91% in 2010; and India is now polio-free. Yet, despite this progress and continuedinvestments, significant challenges persist in providing quality healthcare on an equitable, accessible and affordablebasis across all regions and communities of the country.

To do more, and at a faster rate, an understanding of the current state of healthcare is one important and foundationalelement for determining priorities, resource allocations and goals for the future. This study brings a fresh, objectiveperspective to healthcare access in India, and offers the most comprehensive view of this issue since 2004.

STUDY OBJECTIVES AND METHODOLOGYThe objectives of the study were to map the current status of healthcare in India, prioritize the challenges in terms oftheir relative impact on healthcare access, and provide a roadmap to guide future improvements. The findings aredrawn from an extensive nationwide survey covering 14,746 households representative of geography and incomesegments prevalent in India. The analysis was bolstered by interviews with more than 1,000 doctors and a panel ofhealthcare experts to provide rich and qualitative inputs. In addition to the primary survey, a detailed review of currenthealthcare policies, various healthcare schemes (both at the central and state level), and available data in the publicdomain was also taken into consideration.

Improving healthcare access in India

Healthcare Access Study. Findings from Primary and Secondary Research

Stages of healthcare access

1Physical

accessibility/location

4A�ordability

2

Availability/C

apacity

3

Quality/Functionality

Location:Rural vs Urban

IP vs OPAcute vs Chronic

Channels:Private vs PublicImpact on usage

Components:IP vs OP

Acute vs ChronicIncome levels

FIGURE 1: DIMENSIONS OF HEALTHCARE ACCESS

ACCESSPOINT • VOLUME 3 ISSUE 5 PAGE 49OUTCOMES - Issue 1 Page 1

HEALTHCARE ACCESS IN INDIA | INSIGHTS





DEFINING HEALTHCARE ACCESS Healthcare access has a varying meaning in different countries. As shown in Figure 1, for the purposes of this study, inorder to ensure a framework that was relevant to the population of India, it was defined across four dimensions:

1. Physical accessibility of required healthcare facilities for a patient2. Availability/capacity of the resources required for patient treatment3. Quality/functionality of the resources providing care4. Affordability of the complete treatment to the patient

For a person to have access to healthcare in India, a facility must be reachable within five kilometers and offer availabledoctors, drugs, and treatment options that satisfy both acceptable and cost and quality-of-care standards. Even if onlyone of the components is missing, a patient is unlikely to receive the right treatment in the most appropriate andefficient manner. It is therefore essential to consider all four dimensions in order to assess the state of healthcare access.

KEY FINDINGS The study revealed insights across each dimension of healthcare access:

• Physical reach of any healthcare facility is a challenge in rural areas, particularly for patients with chronicailments. Patients in rural areas must travel more than five kilometers to access an inpatient facility 63% of the time.Difficulty in accessing transportation options and the loss of earnings as a result of travel time lead to treatmentbeing deferred, or facilities selected that may be closer but are not cost-effective or best suited to patient needs.This is especially true for patients suffering from chronic ailments. In urban areas, accessibility is less of a challenge,due to the overall higher number of available facilities.

• The provision of healthcare services in India is skewed toward urban centers and the private sector. Urban residents, who make up 28% of India’s population, have access to 66% of the country’s available hospitalbeds, while the remaining 72% who live in rural areas have access to just one-third of the beds. Similarly, thedistribution of healthcare workers, including doctors, nurses and pharmacists, is highly concentrated in urban areasand the private sector.

• Private healthcare facilities are being used by an increasing proportion of patients due to gaps in quality andavailability of public facilities. Over the past 25 years, both rural and urban patients have increased their use ofprivate service providers over public options. In 2012, 61% of rural patients and 69% of urban patients chose privateinpatient service providers, up from 40% reported in a 1986/87 government survey. Long waiting times and theabsence of diagnostic equipment in public facilities were cited as key reasons by more than 40% of those surveyed.Better quality of treatment in private, inpatient centers was cited as an additional reason by 38% of survey respondents.

• Availability of doctors is a key reason for selecting private facility outpatient treatments. Across both urban andrural sectors, and among the poor and affording populations, at least 60% of those surveyed considered doctoravailability as a significant reason for selecting private facilities for outpatient treatment. However, patients wouldreadily switch to public healthcare centers if issues regarding inpatient and outpatient care were addressed.

• Patients using private facilities face greater affordability challenges. The cost of treatment at private healthcarefacilities is between two and nine times higher than at public facilities. For example, poor patients receivingoutpatient care for chronic conditions at a private facility spent, on average, 44% of their monthly householdexpenditure per treatment, compared to 23% for those using a public facility. However, due to lack of physicalreach, availability of quality treatment and other practices, patients are forced to use more expensive privatefacilities, thus exacerbating affordability challenges.

• Medicine costs as a proportion of out-of-pocket (OOP) healthcare expenses remain high but stable. On average,across public and private facilities, and for chronic and acute diseases, medicines account for more than 60% ofpatients’ total OOP expenses for outpatient treatments, and 43% for inpatient treatments. This share ofexpenditures for medicines has not increased since 2004 for inpatient treatments, and has decreased for outpatienttreatments. Low insurance penetration – and current insurance plans that do not cover drug costs – make the totalcost of medicines a continuing, significant burden for a majority of the population.



Overall, while there are pockets of improvements, significant healthcare access challenges continue to exist for theIndian population, especially in rural areas (Figure 2).

KEY LEVERS FOR IMPROVING ACCESS The survey and analysis show that the components that define healthcare access are not independent of each other:lower physical reach of public facilities reduces access and also increases costs by diverting patients to higher costalternatives; lack of availability of good doctors and resources in public facilities impacts affordability of healthcare in asimilar manner.

The levers of improvement in access can be broadly categorized into the following:

1. Improve physical reach of healthcare facilities, especially in rural areas of India

2. Improve availability and resourcing of public facilities (eg, by addressing concerns on availability of physicians andessential medicines, quality of care and prompt access at public healthcare facilities)

3. Make higher-cost channels more affordable or better financed (eg, by price regulations, subsidization of treatmentcosts, increasing insurance penetration and including drug reimbursement as part of insurance coverage)

From a patient cost-of-treatment perspective, modeling each of the levers can reveal their relative impact. The leverwith the maximum impact on OOP spend is improvement in availability and quality of public healthcare services, whichwould curb the diversion of patients to private channels enabling more patients to utilize lower-cost facilities. Thecumulative reduction in OOP expenditure possible is about 40% for outpatient treatments and 45% for inpatienttreatments.

FIGURE 2: SUMMARIZED ASSESSMENT OF HEALTHCARE ACCESS FOR INDIAN POPULATION

No concern Some concern Concern areas

No gaps in access

Physical reach Availability Quality A�ordability

Large gaps in access

UrbanA�ording No signi�cant gaps except

a�ordability of IP care

Quality and a�ordability of all HC servicesPoor

RuralA�ording Availability of HC services;

a�ordability of chronic care & IP care

Physical reach, availability, quality and a�ordability of all HC servicesPoor


RECOMMENDATIONS Some of the issues in healthcare access revealed in the survey and analysis are directly linked to deficiencies in supplyor quality; others are symptoms or consequences of gaps elsewhere in the healthcare system. As the government seeksto expand its expenditure on healthcare, it must select a strategy that provides the greatest healthcare access benefit tothe Indian population. Sustainable policy solutions to healthcare financing, infrastructure and human resourcechallenges are critically needed. These call for active collaboration between the public and private sectors in a system-level, coordinated approach, focused on creating and implementing initiatives in three priority areas:

1. Improve availability of healthcare services Availability is the front door to the healthcare system, determining whether the patient enters or not, whether he/shereceives care, and from whom. In India, this is currently characterized by lack of doctors, healthcare personnel, clinicsand hospitals, particularly in rural areas. Efforts will need to look at system availability and attack the bottlenecks.

Appropriately trained and adequately supported physicians and healthcare workers with relevant expertise is amedium to long-term investment. In the shorter term, some availability issues can be addressed by better matchingcertain needs with currently available capacity elsewhere. Notable successes include the National Rural Health Mission,which aims to improve basic healthcare delivery systems in rural areas by integrating organizational structures andoptimizing health manpower.

Private sector examples of bridging availability challenges include the use of telemedicine to connectphysicians/healthcare workers to specialists or supervising physicians who can assist in consultations and delivery ofclinical services. Additional ways to hasten the increase in capacity include more public-private partnerships which mayaddress any bureaucratic hurdles or cumbersome business processes, insufficient resourcing, and inadequacies in anylocal supervision. Ultimately, ensuring broad availability will not only improve overall access to healthcare, but alsoprovide multiple options for seeking affordable treatment and diagnosis.

Alongside these investments, measurable standards of performance will need to be established; technology andinformation leveraged to build appropriate metrics and monitoring systems; investments made to bring non-functioning facilities up to standard; and appropriate training and incentives provided for health workers. Effectiveenforcement will require tighter governance processes and efficient and transparent work and decision-makingprocesses. The best operational practices of successful healthcare centers should be replicated to others.

For the public channel, decentralization of healthcare delivery can lead to better governance and functioning. In India,local governance and involvement by the Panchayats, local communities, and NgOs will need to be strengthened tobring delivery at public facilities up to the desired quality and standards.

2. Raise performance of healthcare delivery organizations in terms of service quality As availability of healthcare services is improved, quality in both new and existing capacities will need to be ensured.The survey indicates that the perception of better quality of care is another reason why patients seek help in the privatesector, driven by such factors as perceived superior training of physicians, shorter wait times, and facility quality.Competitiveness and incentives in the private sector have created highly efficient and high-performance organizations.This knowledge and experience should be leveraged and applied to the operations of public healthcare facilities. Ifquality of basic healthcare was considered to be equal between private and public facilities, patients could be free toseek care at facilities that may be more affordable for them. The government should engage the private sector for helpto improve quality of care and healthcare services.

3. Expand and accelerate affordability of healthcare Effective financing mechanisms play a pivotal role in healthcare affordability for patients. Payments can come fromgovernment, health insurance companies, or the patients themselves. For the poor, affordability of healthcare is one ofthe most prominent issues in having good access. This segment will need the most help from the government toensure receipt of healthcare.



As already discussed, improving the availability of healthcare workers and facilities can indirectly address theaffordability issue. Additionally, by ensuring basic access to essential medicines, OOP expenses can be reduced,allowing for more income to address other needs, which may or may not be healthcare related. The government hasalready rolled out an ambitious and well-funded program to provide free essential medicine for all attending agovernment healthcare facility. Implementation of this program should be monitored and adjusted as necessary toensure its success.

government insurance schemes that pay for treatment in private facilities can also play an important role. Althoughprivate and public insurance programs are successfully covering more people, there are still many individuals who arenot aware of or fully understand them. To more rapidly increase insurance penetration and to avoid fraud, the poorand the lesser privileged population should be clearly informed about such government-run and public programs andtheir benefits.

Beyond these initial efforts, insurance penetration needs to be increased across all segments of the population, as wellas insurance coverage for more healthcare services, including outpatient care and prescription medicines. Moreexpansive efforts will be needed, such as increased investment in healthcare through sustainable financing, to have areal impact on healthcare affordability.

CONCLUSIONSRecent progress and commitments by the government and private sector suggest a willingness to invest andoperationalize the changes needed to broaden healthcare access in India. Recognizing that not everything can bechanged at once and that the timescale is long, a roadmap is essential to ensuring that gaps are prioritized,interconnections and dependencies recognized, resources directed to the right areas, targets defined, progressmeasured, and the community integrally involved along the way. The results of this survey provide a solid foundationfor the necessary discussion and debate that is required to align all stakeholders’ efforts in advancing this goal,improving availability, raising performance levels and expanding affordability in the long-term •

This article is summarized from “Understanding healthcare access in India: What is the current state?” published by theIMS Institute for Healthcare Informatics in June, 2013. The funding of the study by the Organisation of PharmaceuticalProducers of India (OPPI) and the Pharmaceutical Research and Manufacturers of America (PhRMA) is gratefullyacknowledged as is the support of the Indian Drug Manufacturer’s Association (IDMA). The full report is available todownload at www.theimsinstitute.org


The lever with the maximum impact on out-of-pocket spend isimprovement in availability and quality of public healthcare services.

Despite the high stakes and intensity of political and industry debate onpharmaceutical pricing, available comparative data are relatively scarce. A new study from IMS Health highlights the challenges of price comparisons,while revealing some intriguing paradoxes, in analyzing the impact of Frenchprice regulation on price gaps with other EU countries.

Nathalie Grandfils, MSC is Director RWE Solutions & HEOR, IMS [email protected]

PAGE 54 IMS HEOR & REAL-WORLD EVIDENCE SOLUTIONSPage 1 IMS HEALTH ECONOMICS AND OUTCOMES RESEARCH

INSIGHTS | PHARMACEUTICAL PRICE COMPARISONS

The authors

Claude Le Pen, PHD is a Member of the Strategic Committee, IMS Health and Professor of Health Economics, Paris-Dauphine [email protected]


A price comparison of recent pharmaceutical launches Healthcare databases are unique in their ability to capture relevant and timelyinformation on drug prescribing and effectiveness in conditions of real-worldpractice across large sample patient populations. In France, new products applying for reimbursement by the Public Health Insurance (PHI) must be first evaluated by thetransparency commission (TC)i to determine whether the '“service rendered” (SMR: Service Médical Rendu) is “enough”to warrant this coverage.1 The TC also evaluates the “improvement in medical benefit” (ASMR: Amelioration du ServiceMédical Rendu) in relation to existing alternative therapies. Drugs are classified on a 5-degree scale ranging from ASMR 1(major breakthrough) to ASMR 5 (no improvement over existing drugs).

Although in principle the TC advice is based exclusively on medical criteria, it is nevertheless used downstream byanother state committee, the CEPS (Comité Economique des Produits de Santé) which makes the final decision on price.According to the official CEPS doctrine, the “price of a medicine [should reflect] the health benefits it delivers”. This canbe seen as a variant of informal value-based pricing (VBP), where products considered innovative (ASMR 1-3) are in aposition to negotiate a premium price in relation to their comparators, while the price of those considered lessinnovative (ASMR 4 or 5) will be systematically discounted by a variable percentage which tends to rise.

In this context, several interesting questions arise:

• Does the same price hierarchy exist in the other four large EU markets, germany, Italy, Spain, UK (Top4 EU) and, if so,is the premium for innovative products of the same magnitude as in France?

• Is there evidence to suggest a convergence of prices across the different countries in recent years? Has theinternational price gap widened or narrowed?

• To what extent is any price convergence affected by the ranking on the ASMR scale? To explore these issues, a price comparison study was performed on a common set of new pharmaceutical productslaunched onto the retail market in France and the Top4 EU from 2008-2012.

METHODOLOGY Product selectionAll new chemical or biological entities applying for the first time for reimbursement by the PHI in France and assessedby the TC between January 1 2008 and June 30 2012 were included. From this list, the subset of products present onthe community pharmacy market in at least one other country in June 2012 was retained.

Product categorizationProducts were categorized by ASMR level, being:

• ASMR 1-2-3: Deemed innovative, these products can pass through a simplified price notification procedureprovided the ex-manufacturer price lies in the lower band of a European price corridor, including the Top4 EU.According to this element of external reference pricing (ERP),2 products in the ASMR 1-2-3 category would beexpected to exhibit the greatest price convergence.

Evaluating the impact of French price regulation in Europe

ACCESSPOINT • VOLUME 3 ISSUE 5 PAGE 55OUTCOMES - Issue 1 Page 1

PHARMACEUTICAL PRICE COMPARISONS | INSIGHTS

i The Transparency Commission is part of the “Haute Autorité de Santé” (HAS), the state agency for the evaluation of quality of care, the accreditation of hospitals, and theevaluation of medical goods and procedures.





• ASMR 4: This category raises the difficult and controversial question of evaluating “incremental innovation” whichmay be appreciated differently according to the parameter being evaluated (efficacy, tolerance, patient quality oflife, ease or comfort of use, etc). A common complaint among manufacturers is that “incremental innovation” inparticular is not appropriately recognized and rewarded.

• ASMR 5: For products considered not to offer any improvement, the law explicitly states that their price should belower than that of comparatorsii.

Data collectionASMR were obtained from a database of the LEEM, the association of the pharmaceutical manufacturers in France. Forproducts with several ASMR (due to multiple indications), the highest was retained as the determinant for pricing. All productsclassified as ASMR 1-3 were grouped into a single category since they were few in number and obey the same pricing logic.

Economic data (pharmacy prices, total volumes (pharmacies and hospitals)), were collected from the IMS Health MIDASdatabase. Volumes were measured in standard units (SU), ie, the smallest dispensation entity, to eliminate the effect ofpacket sizes which differ greatly among countries.

Ex-manufacturer prices as of June 2012 were retained. UK prices were converted into Euro using the current exchangerate (June 2012). In the remainder of this article, the term “price” refers to the price per standard unit (PSU).

Computation of price indexesPrices per dosages and forms were averaged using local volumes as weights. For each of the three ASMR categories andfive countries, a weighted price index was then computed by averaging the product prices obtained previously. Twoweights were used (pharmacy volumes, total volumes (pharmacies and hospitals)), and two weighting systems (Paasche,Laspeyres). The first weights each price of the five countries by French volumes, while the second uses domestic volumes.These two indexes allow the effect of consumption structures on the average price per country to be measured: whetheraverage prices in a country are lower because individual drug prices are lower or because, within a category, lower pricedproducts are more often prescribed and consumed in this country than in the others.

Uncertainty managementThe equality of weighted average prices between France and one of the other four countries was tested using a classicpaired 2-sample t-test. The influence of the methodological choices, especially regarding the weighting system, wasassessed through one-way sensitivity analyses.

RESULTS Final products setThe starting list of products comprised 472 presentations, corresponding to 237 molecules. Pediatric forms of the samemolecule were considered as different products. Some products and/or dosages were eliminated from the final list forone or several reasons (eg, restricted to hospital use only in France; not priced in France; no specified ASMR; not found

TABLE 1: COMPARATIVE CHARACTERISTIC OF PRODUCTS COMMON TO THE FIVE PHARMACY MARkETS AND PRODUCTS THAT ARE MISSING INAT LEAST ONE OF THE FOUR PHARMACY MARkETS

Prescription restricted to specialists only (in France)Orphan drugsAntiviralsAnticancer / ImmunomodulatorsMedian price per SU

NB: All differences statistically significant at 5%

Common set (n=51) Non-common set (n=56)

6%2%

10%0%

1.1€

28%13%20%18%9.9€

ii The choice of the company is then to market the product without public reimbursement or, most of the time, simply not to market it.



in any of the Top4 EU). Finally, a complete set of data was retrieved for 107 products and 245 dosages. Among these,only 51 (48%) were marketed simultaneously in the community pharmacies of all five countries.

In order to understand whether there was some rationale behind the loss of almost half the original set of products, theset of the 51 “common products” was compared to the remaining set of 56 products absent from at least one of theother markets (Table 1).

This comparison showed that the “non common set” comprised mainly more expensive products that have beensubjected more often to specific prescription conditions in their label, benefited more often from orphan drugdesignation, and are more often indicated in severe diseases. Logically, these products would be expected to be madeavailable under the same conditions to patients in the top five European countries. However, the results suggest that, tothe contrary, these products are those most likely to be banned from patient access in the retail market. This may be theresult of, for example, pharmacoeconomics (UK), restrictions to hospital use (Italy, Spain) or delays in products launches,and reflects the reality that market access in Europe is paradoxically more difficult for products classified as being themost innovative.

On a bi-lateral basis, the number of molecules common to France and to another country is obviously higher: 94 forgermany; 79 for UK; 71 for Spain; 69 for Italy. It is thus in germany where the supply of products in the pharmacymarket is most similar to France.

Comparative price weighted averagesThe results were then considered in the form of average prices per standard unit in euro, weighted by French pharmacysales, and ratios of these average pricesiii, for both the common set of products (multilateral analysis) and the setscommon to France and a single other country (bilateral analysis).

Although to be considered with caution because of the (very) small number of products in some categories (especiallyASMR 1-2-3), some key points merit attention:

• Only a few differences appear to be statistically significant, confirming the idea that a price convergence does existin Europe despite limited local variations.

• French prices are generally equal to or lower than prices in the Top4 EU. The only significant exceptions are UK pricesfor products ASMR 1-2-3, which are about 20% less expensive depending on the set of products.

• As expected, there is a clear price hierarchy across the different ASMR categories: around 20 €/SU for ASMR 1-2-3;1.60 €/SU for ASMR 4; and 0.50 €/SU for ASMR 5. Most innovative products, in a French TC sense, are clearlyassociated with higher prices in all the reviewed countries.

• germany clearly has a different price pattern from the other countries, with prices regularly and significantly highercompared to France and also to other countries.

• Prices of products ASMR 1-2-3 range from 48% above the average French prices to 18% below (bilateral analysis).Conversely, products with ASMR 4 and 5 have quite homogenous price levels across the five countries, none beingstatistically different from the French average, except products ASMR 5 in germany. This is paradoxical, since ASMR1-2-3 products are those that are explicitly submitted to an ERP procedure in France.

• Among products ASMR 1-2-3, the outlier appears to be the UK where prices are significantly below the level of othercountries. This potentially reflects health economic thresholds which ban a number of expensive products that arecurrently marketed in other countries, eventually with a restriction to hospital use only.

• Multilateral and bilateral results do not differ substantially, even if statistically significant differences are morefrequent in bilateral analysis, certainly because of the greater number of products.


iii A value greater than 1 means that the average price is lower in France compared to the considered country


Change over timeTo determine whether prices have converged under the influence of ERP, Table 2 compares the results of the presentanalysis with those of a previous study which used the same methodology and product classification to compare ex-manufacturer prices (2007) of products introduced in France from 2004-2007.3 To eliminate the effect of inflation andexchange rate fluctuations, results are expressed as ratios of weighted prices in one country and in France.

Under the influence of the ERP in France, ratios at 2012 would be expected to be closer to 1 than at 2007, especially forproducts ASMR 1-2-3. However, with the exception of the UK, this is not the case, although information is lacking tocompute statistical tests. To the contrary, the price gaps for these products paradoxically seem to have widened(unfavorably for French prices). During the same time, for products ASMR 4, the price gap has narrowed. Not only hasthere been little price convergence during the last five years, but also, ultimately, the convergence has been moreeffective for products that are not submitted to a formal ERP procedure in France.

Sensitivity analyses and effect of weighting optionsSeveral sensitivity analyses were performed to test the influence of the methodological choices, including use ofdifferent weighting variables, prices and weighting systems. The combination of these options leads to a greaternumber of variants than it is possible to reproduce here, but the results are sensitive to the weighting options.

Table 3 shows the effect of the weighting options on the price index of all the products, reporting prices and ratios

TABLE 2: RATIOS OF WEIGHTED AVERAGE PRICES IN A COUNTRY AND IN FRANCE BY CLASSES OF ASMR, 2007 AND 2012(MULTILATERAL ANALYSIS – RETAIL SALES – FRENCH VOLUMES)

ASMR 1-2ASMR 3ASMR 4ASMR 5All ASMR

0.691.061.051.021.00

1.14

1.011.131.11

0.781.151.060.980.99

1.37

0.951.051.07

0.811.071.341.451.25

0.77

0.961.071.02

0.861.081.411.231.22

1.21

1.021.411.32

GERMANY2007 2012

SPAIN2007 2012

ITALY2007 2012

UK2007 2012

TABLE 3: RATIOS OF AVERAGE PRICES IN A COUNTRY AND IN FRANCE UNDER DIFFERENT WEIGHTING OPTIONS(BILATERAL ANALYSIS – RETAIL AND TOTAL SALES – LOCAL VOLUMES)

France (€)Germany (€)Germany/France

France (€)Spain (€)Spain/France

France (€)Italy (€)Italy/France

France (€)UK (€)UK/France

3.164.921.56

2.501.470.59

1.681.440.86

2.553.991.56

3.104.851.56

1.982.091.06

1.671.911.14

2.514.351.73

0.561.192.13

0.440.611.39

0.460.340.74

0.50.621.24

0.591.282.17

0.100.323.20

0.470.450.96

0.530.781.47

0.891.852.08

0.600.691.15

0.580.380.66

0.790.770.97

0.991.961.98

0.610.781.28

0.590.570.97

0.881.111.26

14.9519.121.28

20.928.160.39

19.481.070.05

13.724.290.31

14.0518.791.34

18.210.060.55

16.4616.571.01

13.187.560.57

ASMR 1-2-3Retail Total

ASMR 4Retail Total

ASMR 5Retail Total

ALLRetail Total



obtained through domestic weights (Laspeyres) using retail sales and total sales. Results change dramatically fromthose reported in using French volumes (Paasche), especially for products ASMR 1-2-3 in Spain, Italy and UK. Theaverage price per SU for these products in Italy, for example, changes from 28.86 € when using French retail sales asweights, to 1.07€ when using Italian retail sales – the consequence of a policy tending to restrict the prescribing ofexpensive products to hospitals.

CONCLUSION In general, prices of new products in community pharmacies in France are equal to or lower than prices in othercountries, with the exception of the most advanced products in UK. This is a significant finding from a Frenchperspective since it contradicts a widely held view. Other key findings are:

• The ASMR classification can discriminate products according to their average price level in all Top4 EU. ASMR 1-2-3products are always more expensive than ASMR 4 products which in turn are always more expensive than those inASMR 5. While this validates the ASMR scale to some extent, a reverse interpretation is also possible: thedetermination of price obeys a logic which is independent of the rating systems for innovation that a country canset up for its own use. The results (prices) overcome the process (pricing systems).

• Price heterogeneity is paradoxically greater for the advanced products, which are theoretically submitted to EPR,than for products providing a small therapeutic benefit, which are not. In that case, price comparison is biased sincethe most advanced products are not made available under the same conditions in all countries: a number arerestricted to hospital use in some countries. This forbids a meaningful comparison given the substantial andconfidential nature of rebates which affect hospitals prices. Other products may simply not be recommended forhealth economic reasons.

• Price comparisons – and hence price regulation through EPR – bear on subsets of products that are allowed in thepharmacy market. This means a significant portion of the most advanced and expensive products escape this priceregulation, being previously submitted to a “non price” or “access” regulation which may be more important than theprice regulation itself. This obviously limits the full extension of ERP to products obeying the same prescription andavailability status.

This last point is important: pharmaceutical price comparisons are meaningless if they do not take into account themarket structures in each examined country, the restrictions to marketing conditions, the real availability of drugs topatients, and so on. According to Danzon and Furukawa,4 for example, higher US averages prices could be partiallyexplained by the tendency to use more recent products at higher dosages than in other countries. Price differencescannot be separated from the characteristics of drug use in different countries. For these reasons, international PPCslargely remain a topic for academic research and cannot be part of an administrative process without caution,explanations and limits •

Insights from this IMS Health study were highlighted in the 2012 Rapport d’activite of the Comite Economique de Produits deSante in France, published September, 2013. Available at: http://www.sante.gouv.fr/IMG/pdf/RA_2012_Final.pdf, page 127

1 HAS. La Commission de la Transparence, évaluation des médicaments en vue de leur remboursement. Available at: www.has-sante.fr (May 2013)2 Espin J, Rovira J, Olry de Labry A. External Reference Pricing, Working Paper 1. WHO Project on Medicine Prices and Availability. May 20113 geoffard PY, Sauri L. International comparison of prices for new drugs. Unpublished manuscript, May 20084 Danzon P, Furukawa MF. International Price and Availability of Pharmaceuticals in 2005. Health Affairs, 2008; 27-1: 221-233


Decision makers commonly rely on health economic modelspopulated with clinical trial data to inform their initialassessments about treatment selection, coverage andreimbursement. In addition to a reliance on trial-based efficacyand safety measures, peri-launch models may also requireassumptions (eg, to simplify pathways or address data gaps),utilize data inputs culled from a combination of disparatesources (eg, clinical trial, published literature, administrativedatasets), and often reflect how the new drug or intervention isanticipated to be (rather than actually) used in practice.

given these acknowledged but accepted uncertaintiesassociated with peri-launch model projections, model re-analyses leveraging RWE could help to transform thedecision-making landscape. However, to date, few suchpublished examples exist, perhaps in part due to accesschallenges to the original model or to real-world data sources,lack of centralized resources or expertise to conduct the dataanalyses or model re-analyses, or insufficient pressure ordemand for model re-analyses in the past. Yet, with the recentand tremendous growth of RWE and an ever-increasing need todemonstrate value for money, healthcare stakeholders are likelyto demand evidence of real-world health economic value tosupport ongoing access and/or pricing agreements.

PREVIOUS TRIAL-BASED HEALTH ECONOMIC MODEL INVENOUS THROMBOEMBOLISMIMS Health had previously developed a decision-analytic modelto compare the costs and outcomes of a new oral anticoagulantvs. a low molecular weight heparin (LMWH) as prophylaxis forvenous thromboembolism (VTE) in patients undergoing totalhip replacement (THR) or total knee replacement (TKR). At a highlevel, the model was divided into three modules representingdifferent periods following index hospitalization: prophylaxis,post-prophylaxis and long-term complications (Figure 1).

The model utilized short-term clinical outcomes data obtained fromPhase III trials, as well as data on long-term complications andsequelae as captured in observational studies and databases. Itconsidered the costs of prophylaxis drugs, administration,monitoring, diagnosis and treatment of VTE, as well as treatment ofpost-thrombotic syndrome and recurrent VTE. All costs wereestimated from the perspective of a US healthcare payer in 2010 USD,where resource use and unit cost estimates were drawn from acombination of published estimates, Medicare reimbursementrates, and modeling assumptions. Results of the cost-effectivenessanalysis were reported in terms of symptomatic VTE events avoided.

RWE has a key role to playin populating andvalidating peri-launchhealth economic modelsbased on actual treatmentexperience and outcomes

Validating a trial-based cost-effectiveness analysisleveraging RWE

The authorsHuimin Li, MS is Consultant RWE Solutions & HEOR, IMS [email protected]

Julie Munakata, MS is Principal RWE Solutions & HEOR, IMS [email protected]

Michael Nelson, PHARMDis Senior Principal, RWE Solutions & HEOR,IMS [email protected]

PROJECT FOCUS | VENOUS THROMBOEMBOLISM


NEW EVIDENCE TO EVALUATE REAL-WORLDPERFORMANCE OF ORAL ANTICOAGULANTIn 2013, nearly three years post preparation of the peri-launch cost-effectiveness analysis, sufficient treatmentexperience now exists with the oral anticoagulant in thereal world. Moreover, the primary outcome of interest(VTE events) from the original Phase III trials and in thetrial-based model has been captured, along with medicaland pharmacy costs, in the fully adjudicated UShealthcare claims data of IMS PharMetrics Plus™ database.

IMS Health therefore sought to assess the real-worldhealth and economic outcomes associated with the oralanticoagulant vs. LMWH in the prophylaxis indication andcompare the results of the health economic modelpopulated with clinical trial data vs. RWE.

RETROSPECTIVE DATABASE ANALYSIS OF REAL-WORLD TREATMENT OUTCOMESPatients ≥18 years of age who underwent THR or TKRbetween July 2011 and June 2012 and had at least oneclaim for the oral anticoagulant or LMWH of interest, wereidentified in PharMetrics Plus. Patients were required tobe continuously enrolled six months pre- and post-indexand were excluded if treated with multiple anticoagulantswithin 10 days post-index. A propensity score matchingtechnique was employed to reduce selection bias.

Patient characteristics, inpatient-related VTE events and healthcare costs were determined.

A total of 14,880 patients were identified (7,440 oralanticoagulant, 7,440 LMWH) in the database. In bothgroups, mean age was 59, and 53% were female.Compared with the original clinical trial, the followingdifferences were noted in the database sample:

1. Commercially insured population with fewer femalesand lower average age

2. VTE events were limited to inpatient encounters (did not include outpatient events)

3. Follow-up period up to 12 months (vs. 30-35 days)

MODEL RE-ANALYSIS USING RWE INPUTS FROMRETROSPECTIVE DATABASE ANALYSISTo perform the model re-analysis, the previouslydeveloped cost-effectiveness analysis model waspopulated with updated clinical inputs from the claimsdatabase analysis. For the purposes of these initialanalyses, no updates on cost inputs were made in themodel. New model results were generated, includingVTE-related costs and number of symptomatic VTE events.The focus was on validating the model in terms of short-term, one-year results.

VENOUS THROMBOEMBOLISM | PROJECT FOCUS

FIGURE 1: THREE-MODULE DECISION ANALYTIC MODEL FOR VENOUS THROMBOEMBOLISM

Prophylaxis (1-month) Post-prophylaxis(extrapolation)

Long-termcomplications

Asymptomatic DVTto symptomatic VTE

Fatal PE

Non fatal PE

Pulmonary embolism (PE)

Symptomatic VTE

VTEProphylaxis-related bleeding

{+} No prophylaxis-related bleeding

Extrapolate forsymptomatic VTE(up to 90 days)



Symptomatic DVT

Asymptomatic VTE

No VTE

PROJECT FOCUS | VENOUS THROMBOEMBOLISM


VALIDATION OF ORIGINAL FINDINGSCompared with LMWH, oral anticoagulant use wasassociated with fewer symptomatic VTE events over oneyear. When repopulated with clinical inputs from claimsdata, the model projected similar VTE event differences as the trial-based model (-0.023 vs. -0.015).

Costs (per patient/year) in the oral anticoagulant andLMWH groups were consistent across the trial-basedmodel ($385 vs. $1,011), claims-based model ($437 vs. $1,290), and direct reported results from claims analysis ($506 vs. $1,125) (Figure 2).

CONCLUSIONSUse of RWE is a practical and objective way to validate a trial-based health economic model. Future work should consider study design issues and practical use of the results •

FIGURE 2: RE-ANALYSIS VALIDATED ORIGINAL FINDINGS WITH CONSISTENT COSTS ACROSS GROUPS

Original model(Trial data)

$625.27

$853.12$619.00

Comparator

Comparator

Updated model(PMTX + Data)

Claims analysis(PMTX + Data)

$1,400.00

$1,200.00

$1,000.00

$800.00

$600.00

$400.00

$200.00

$ -

Comparator

Intervention InterventionIntervention

Medicare is the federal health insurance program in the UScreated for people age 65 or older, people under age 65 withcertain disabilities, and people of all ages with end-stage renaldisease.1 In 2012, Medicare covered over 50 million Americansand its spending accounted for 21% of overall national healthspending.2 Among the total 49.4 million Medicare beneficiariesin 2012, about 12.7 million (or 25.7% of the total) were coveredby Medicare Advantage through privately managed healthplans.3 US commercial payer databases (eg, IMS PharMetricsPlus™) do capture the share of Medicare patients in Advantageplans. However, for a complete capturing of Medicarebeneficiary types, researchers still need to work with the Centersfor Medicare and Medicaid Services (CMS) and its contractor toseek permission for data access.

With Medicare covering so many people, including Americans athigher risk for certain diseases given their age, Medicare datahas long been an important resource for researchers andpolicymakers who study the US healthcare system. As its releaseis governed by a series of federal laws to protect data privacyand security, the process to acquire and use the data (eitherresearch-identifiable files or limited dataset files) hastraditionally been complex and time consuming.

In order to achieve a better balance between privacy protectionand analytic utility, the CMS recently increased public access toits Medicare data with the release of a new data type - MedicareClaims Public Use Files (PUFs). These contain non-identifiable,claim-specific information based on the 5% random sample fee-for-service claims files, and are available for use within the publicdomain.4 To date, two years of de-identified data files (2008 and2010) have been released. Although these claim data files do notlend themselves to longitudinal analyses, they are similar instructure to some popular US national surveys for medicalservice use, eg, National Ambulatory Medical Care Survey(NAMCS) and National Hospital Ambulatory Medical Care Survey(NHAMCS), which include drug mentions that allow estimates ofprescription rates.

Medicare claims public usefiles, combined with privatedata, can enable moreefficient and credibleresearch insights withacknowledgement of theirinherent limitations

The authorsJuliette Chen, MPA is Director RWE Solutions & HEOR, IMS [email protected]

Vernon Schabert, PHD is Senior Principal RWE Solutions & HEOR,IMS [email protected]

ALZHEIMER’S DISEASE | PROJECT FOCUS


Assessing the value of an alternative data source for the Medicare-covered US population


PROJECT FOCUS | ALZHEIMER’S DISEASE


MEDICARE CLAIMS PUBLIC USE FILESThe PUF data source offers an option to use publiclyavailable data to look into patients covered by Medicareplans and implement some comparative effectivenessresearch without going through the paperwork hassles of the traditional application process for Medicare dataaccess. All personally identifying information has beenremoved from PUFs in compliance with the HealthInsurance Portability and Accountability Act (HIPAA)Privacy Rule. To guarantee the confidentiality of Medicarebeneficiaries’ Protected Health Information, extensivede‐identification procedures and rigorousre‐identification analyses have been employed.

The available files include service claim PUF of each caretype (ie, for hospital inpatient, hospice, skilled nursingfacility, home health agency, outpatient, physician/other,durable medical equipment supplies, and prescriptiondrug) and chronic conditions PUF (ie, an aggregated file inwhich each record is a profile defined by thecharacteristics of all combinations of age category,gender, various chronic conditions, dual-eligibility status,and services and payment by plan part of thebeneficiaries).

PRACTICAL APPLICATION: ALZHEIMER’S DISEASEAs shown in Figure 1, PUF data has both strengths andlimitations. To ensure efficient use and interpretation ofthe data, it is important to understand these in ascientifically rigorous manner. The following case studyexample compares characteristics and drug mentions ofpatients with Alzheimer’s disease in the ambulatorysetting, between PUF data versus NAMCS/NHAMCS.

Study designThe analysis sought to compare the findings amongpatients with a same type of condition (Alzheimer’sdisease), a same type of demographic characteristic (aged65+), covered by a same type of health insurance(Medicare), and with medications prescribed in a sametype of service setting (physician office and hospitaloutpatient department). Specifically, the comparison,based on two-year data (2008 and 2010), focused on:

• Demographics and clinical characteristics of patientswith Alzheimer’s disease

• Number of prescription mentions per visit associatedwith Alzheimer’s disease

• Top chronic conditions for patients with Alzheimer’s disease

Weighted national estimates of ambulatory visits made bythe identified patients with Alzheimer’s disease wereextracted and calculated. Their demographic profile (age,gender) and clinical characteristics (presence of chronicconditions such as cancer, heart failure, diabetes,rheumatoid arthritis) during visits were also analyzed.Numbers of prescriptions per visit in patients with andwithout Alzheimer’s disease diagnosis were estimatedthrough multivariate regressions, controlling fordemographic and clinical characteristics.

ResultsThe study found that the patients identified fromMedicare PUFs tend to be older, more often female, andmore likely to have co-morbidities than the patients withsimilar characteristics identified from the two surveys. Thismay help to explain why the Medicare elderly patientswith Alzheimer’s disease from Medicare PUFs wereassociated with more medication mentions comparedwith other patients who were identified fromNAMCS/NHAMCS. A few findings from this case study inrelation to age and sex distribution, co-morbidities andnumber of scripts per visit, are shown in Figure 2.

FIGURE 1: STRENGTHS AND LIMITATIONS OF MEDICARE PUF DATA

STRENGTHS

• Easy access - availablein public domain fordownloading

• Free of paperworkhassles

• No charge for use• Claim-level data of all

care types from 5%sample beneficiaries,including pharmacy datafrom Part D

LIMITATIONS

• De-identification in bothbeneficiaries andproviders

• Disjoint 5% samples -beneficiaries can appearin only one PUF

• Coarsening data (eg, agein intervals, 3-digit ICD-9diagnosis, 2-digit ICD-9procedure)

ALZHEIMER’S DISEASE | PROJECT FOCUS

DISTINGUISHING FEATURESThe foremost feature, and also limitation, of the MedicarePUF data source is that all directly identifiableinformation was removed and none of the potentialpatient identification variables were included in the datarelease. This means that users of Medicare claim PUFs canonly undertake analysis within a particular service settingor rely on the chronic condition data file that includes theaggregate-level data for any cross-setting analysis.Furthermore, a much shorter list of variables with lessspecific information is provided in the Medicare PUFs,which is another limitation when using the data.

On the plus side, compared to NAMCS/NHAMCS,Medicare PUFs collect data from a much larger and morerepresentative data source for the Medicare beneficiarypopulation, and offer a lot more claim-level details,through which a broader range of diseases, therapiesand medical services can be identified.

In addition to the analytical areas explored in the study,Medicare PUF data can also help to answer other types ofresearch questions, eg, using the claims files of each caretype to estimate the cross-sectional burden of a diseasewithin a care setting, or to determine the most frequentdiagnoses/procedures/medications and their associatedpayments; or using the chronic condition file to providevarious measures of utilization and amount as averagesfor different groups of Medicare beneficiaries or profiles.

CONCLUSIONThis new Medicare data source helps to meet growinginterest in understanding the utilization patterns amongMedicare beneficiaries. As the importance and frequencyof outcomes research using real-world data increase, so too will scrutiny and questions regarding thecompleteness of the data sources being used forverification. Medicare claims data is a vast pool ofinformation about the way in which healthcare is beingdelivered in the US which, if combined with private data,could facilitate more accurate measurement of a researchquestion. Accounting for its inherent biases/limitationswill not only make the study more credible with theintended audience, but also address the research andbusiness need more efficiently •

FIGURE 2: THE STUDY REVEALED DIFFERENCES IN AGE ANDGENDER DISTRIBUTION, CO-MORBIDITIES AND SCRIPTS

1 Centers for Medicare and Medicaid Services (CMS). Medicare Program - generalInformation. Accessed on September 30, 2013 at:http://www.cms.gov/Medicare/Medicare-general-Information/MedicaregenInfo/index.html

2 Congressional Budget Office (CBO) Medicare Baseline, February 20133 Kaiser Family Foundation. 2012 State Health Facts on Medicare. Accessed onSeptember 30, 2013 at: http://kff.org/state-category/medicare/

4 Centers for Medicare and Medicaid Services (CMS) Basic Stand Alone (BSA)Medicare Claims Public Use Files (PUFs). Accessed on September 30, 2013 at:http://www.cms.gov/BSAPUFS/

100%

90%

80%

70%

60%

50%

40%

30%

20%

10%

0%

Age ≥ 85Age 80-84Age 75-79Age 70-74Age 65-69

Claims: Age 65+, Medicare Covered (n,%)

Agedistribution %

Surveys: Age 65+, Medicare Covered (n,%)

14.2

17.5

20.9

21.2

2.6

45.5

25.4

16.2

8.64.3

100%

90%

80%

70%

60%

50%

40%

30%

20%

10%

0%

MaleFemale


Sexdistribution %


47.1

53.0

32.8

67.2

5.0

4.5

4.0

3.5

3.0

2.5

2.0

1.5

1.0

0.5

0.0

Raw #AD attributable #


# scripts pervisit


4.4

1.72.2

1.1

100%

90%

80%

70%

60%

50%

40%

30%

20%

10%

0%

With co-morbidities

Without co-morbidities


% With co-morbidities


67.8

32.2

88.2

11.8


Enabling your real-world successIMS Health offers a spectrum of world-class expertise in RWE and HEOR to deliver thelocal excellence you need.

In a future where healthcare efficiency and quality are measured through the lens of ‘real-world’ insights, externalvalidity demands a focus on the right data sources, scientifically credible research and actionable communication.IMS Health is committed to helping you succeed:

• Largest multi-disciplinary team of RWE and HEOR experts, as well as leading scientists inepidemiology, drug safety and risk management, based in 18 countries worldwide

• Credible scientific voice and deep therapy area knowledge, captured in over 2,500 publications• Market leadership in developing and adapting robust economic models• Most advanced capabilities in RWE management and analysis, leveraging relevant IMS proprietary

and other key external, third-party data assets• Proven expertise in generating and communicating RWE to advance stakeholder engagement at all levels

IMS Health is a leading independent provider of RWE, outcomes research, epidemiology, economic modeling andmarket access solutions, and value communication.

Our unique, data-agnostic market position can help you develop and support the evidence required to engage globaland local healthcare stakeholders, with deep insights into product safety, efficacy, cost, value for money and affordability.

IMS RWE SOLUTIONS & HEOR | OVERVIEW

We offer a wide spectrum of solutions

Outcomes Research• Evidence generation• Late-phase studies• Patient-Reported Outcomes (PROs)• Database studies• Mixed methods• Comparative effectiveness research• Quality of Life

Epidemiology, Drug Safety and Risk Management• Study design & protocol development• Signal detection & safety surveillance• Drug utilization studies• Comparative safety & outcomes studies• Natural history & burden of illness• Market projections• EU-RMP & REMS assessments

Market Access• Value development planning• Market access strategy• Core value dossiers & local adaptations• HTA readiness• Value communication• Reimbursement submissions

Technology• Platform engines• Data warehouse/data marts• Encryption systems & linking technology• Meta-data repository• User interface & sophisticated

analytics library• Electronic data capture

Real-World Evidence Solutions• Data sourcing & validation• Data integration & linking• Data management & curation • Platform development• Customized analytics & reporting• Evidence planning

Health Economic Modeling• Health economic evaluations• Core models & local adaptations• Budget impact• Meta-analyses• Indirect comparisons• IMS CORE Diabetes Model

• Health plan claims• PharMetrics PlusTM

• Longitudinal Rx

• Electronic medical records (EMR)• Hospital disease & treatment• Oncology

• Diabetes• Custom data sourcing

IMS LifeLinkTM Largest collection of scientifically-validated, anonymized patient-level data assets:


Global scope, local expertiseIMS RWE Solutions & HEOR experts are located in 18 countries worldwide and they havepublished on projects completed in more than 50 countries on all continents.

Dr. Michael NelsonSenior Principal IMS Health1725 Duke Street, Suite 510Alexandria, VA 22314USATel: +1 703 837 [email protected]

Dr. Jacco KejaSenior Principal IMS Health210 Pentonville RoadLondon N1 9JYUKTel: +31 (0) 631 693 [email protected]

Jon ResnickVice President and General ManagerIMS Health1725 Duke Street, Suite 510Alexandria, VA 22314USATel: +1 703 837 [email protected]

David GrantSenior Principal IMS Health210 Pentonville RoadLondon N1 9JYUKTel: +44 (0) 20 3075 [email protected]

yOUR PRIMARy CONTACTS

NORTH AMERICAREGIONAL HEADQUARTERS11 Waterview BoulevardParsippany, NJ 07054USATel: +1 973 316 4000

UNITED STATES1725 Duke StreetSuite 510Alexandria, VA 22314USATel: +1 703 837 5150

One IMS DrivePlymouth MeetingPA 19462USATel: +1 610 834 0800

CANADA16720 Route TranscanadienneKirkland, Québec H9H 5M3CanadaTel: +1 514 428 6000

LATIN AMERICAREGIONAL HEADQUARTERSInsurgentes Sur # 23755th Floor, Col. TizapanMexico City D.F. - C.P. 01090 MexicoTel: +52 (55) 5062 5239

EUROPEREGIONAL HEADQUARTERS210 Pentonville RoadLondon N1 9JYUnited KingdomTel: +44 (0) 20 3075 4800

BELGIUMMedialaan 381800 VilvoordeBelgiumTel: +32 2 627 3211

FRANCE29ème EtageTour Ariane5-7 Place de la Pyramide92088 La Défense CedexFranceTel: +33 1 41 35 1000

GERMANYErika-Mann-Str. 580636 MünchenGermanyTel: +49 89 457912 6400

ITALYViale Certosa 220155 MilanoItalyTel: +39 02 69 78 6721

SPAINDr Ferran, 25-2708034 BarcelonaSpainTel: +34 93 749 63 00

SWEDENSveavägen 155/Plan911346 StockholmSwedenTel: +46 8 508 842 00

SWITZERLANDTheaterstr. 44051 BasleSwitzerlandTel: +41 61 204 5071

UNITED KINGDOM210 Pentonville RoadLondon N1 9JYUnited KingdomTel: +44 (0) 20 3075 4800

ASIA PACIFICREGIONAL HEADQUARTERS8 Cross Street #21-01/02/03PWC BuildingSingapore 048424Tel: +65 6412 7365

AUSTRALIALevel 5, Charter Grove29-57 Christie StreetSt Leonards, NSW 2065AustraliaTelephone: +61 2 9805 6800

CHINA7/F Central TowerChina Overseas PlazaJianguomenwai Avenue, Chaoyang DistrictBeijing 100001ChinaTel: +86 10 8567 4255

SOUTH KOREA9F Handok Building735 Yeoksam1-dongKangnam-ku Seoul135-755S. KoreaTel: +82 2 3459 7307

TAIWAN8th FloorNo 2, Tun Hwa South RoadSection 1Taipei 10506TaiwanROCTel: +886 2 2721 5337

For further information, email [email protected] or visit www.imshealth.com/rwe

IMS RWE Solutions & HEOR office locations

LOCATIONS | IMS RWE SOLUTIONS & HEOR


Expertise in depthIMS Health has one of the largest global teams of experts in RWE, HEOR and market access ofany organization in the world.We apply unrivalled experience and specialist expertise to help our clients meet the demands of an increasingly complex global, regional and localpharmaceutical landscape. Our highly qualified, multi-disciplinary consultants, researchers and leading scientists in epidemiology, drug safety andrisk management, have proven skills and capabilities across all key therapy areas. Spanning industry, consulting, government and academia, theirexpertise reflects a global grasp, local experience and a unique, inside market perspective.

yumiko Asukai, MSC• Yumi Asukai is Principal RWE Solutions & HEOR at IMS Health, specializing in the development of economic models across

the product lifecycle and the interpretation of model outputs for strategic market access and value demonstration. Herexpertise in this field spans from early strategic modeling through to global core cost-utility models.

• Yumi’s background includes roles at Fourth Hurdle Consulting and in healthcare and business consulting in San Franciscoand Tokyo, where she focused on comparative studies of health policies between Japan and the US complemented byanalyses of primary data. Yumi has worked extensively in the cardiovascular, oncology and respiratory disease areas andshe is part of a global modeling taskforce for COPD composed of academic and industry members.

• Yumi holds a Master's degree in Health Policy, Planning and Financing from the London School of Hygiene & TropicalMedicine and the London School of Economics, and a Bachelor's degree in Political Science from Stanford University.

Our senior team

Nevzeta bosnic, bA• Nevzeta Bosnic is Principal at IMS Brogan, where she manages projects to meet the broad spectrum of client needs in the

Canadian pharmaceutical market.• Formerly Director of Economic Consulting at Brogan Inc, Nev has led many strategic consulting, policy and data analyses

for pharmaceutical clients, government bodies and academic institutions in Canada. She has extensive knowledge ofpublic and private drug plans across the country and in-depth expertise and experience on the drug reimbursement process.

• Nev holds a Bachelor’s degree in Business Economics from the School of Economics and Business at the University ofSarajevo, Bosnia-Herzegovina.

Karin berger, MbA• Karin Berger is Principal RWE Solutions & HEOR at IMS Health, with a focus on RWE, PROs and cost-effectiveness

evaluation analyses at a national and international level.• Formerly Managing Director of MERG (Medical Economics Research Group), an independent German organization

providing health economics services to the pharmaceutical industry, university hospitals and European Commission, Karin has more than 15 years experience in the health economics arena. She lectures at several universities, has publishedextensively in peer-reviewed journals, and regularly presents at economic and medical conferences around the world.

• Karin graduated as Diplom-Kaufmann (German MBA equivalent) from the Bayreuth University, Germany, with a specialfocus on health economics.

Joe Caputo, bSC• Joe Caputo is Regional Principal RWE Solutions & HEOR, Asia Pacific at IMS Health, leveraging more than 20 years

experience in the pharmaceutical sector to help clients address the challenges of global reimbursement and marketaccess throughout the drug development program. He has led numerous projects involving payer research, valuedossiers, local market access models and HTA submissions.

• Joe's background includes industry roles in drug development, sales and marketing, and UK and global health outcomes,as well as consulting in health economics. He has wide-ranging knowledge of the drug development process at bothlocal and international level and a unique understanding of evidence gaps in light of reimbursement and market access requirements.

• Joe holds a BSc in Applied Statistics and Operational Research from Sheffield Hallam University, UK.

IMS RWE SOLUTIONS & HEOR | EXPERTISE


Frank-Ulrich Fricke, PHD, MSC• Dr. Frank-Ulrich Fricke is Principal RWE Solutions & HEOR at IMS Health and Professor for Health Economics, Georg-

Simon-Ohm University of Applied Sciences, Nuremberg in Germany, with a focus on health economic evaluations,market access strategies and health policy.

• Formerly a Managing Director of Fricke & Pirk GmbH, and previously Head of Health Economics at NovartisPharmaceuticals, Frank-Ulrich has conducted health economic evaluations across a wide range of therapeutic areas,developing a wealth of experience in pricing, health affairs and health policy. As a co-founder of the NIG 21 association,he has forged strong relationships with health economists, physicians and related researchers working in the Germanhealthcare system.

• Frank-Ulrich holds a PhD in Economics from the Bayreuth University, and an MBA equivalent from the Christian-Albrechts-University, Kiel.

Mitch DeKoven, MHSA• Mitch DeKoven is Principal RWE Solutions & HEOR at IMS Health, leading teams in a variety of projects, including value

development plans, retrospective database studies and observational surveys. • Prior to joining IMS Health, Mitch was an Associate Director of Reimbursement and Market Access at ValueMedics

Research LLC. His previous roles include Manager of Reimbursement Services at United BioSource Corporation’s Centerfor Pricing & Reimbursement, Consultant with CHPS Consulting, and Program Manager of the Center for Cancer andBlood Disorders Children’s National Medical Center in Washington, DC, a position he held after completing anadministrative fellowship with the Johns Hopkins Health System.

• A past president of the board of directors of the Lupus Foundation of America Greater Washington Chapter, Mitch serveson six editorial advisory boards and is a peer reviewer for a number of international healthcare journals. He has alsoauthored several articles. Mitch holds an MHSA from the University of Michigan School of Public Health and a Bachelor’sdegree in Spanish from Washington University in St. Louis.

Carl de Moor, PHD• Dr. Carl de Moor is Senior Principal Epidemiology, RWE Solutions at IMS Health, with leadership responsibility in this field.

He has 25 years experience in epidemiology, biostatistics and health outcomes research in retrospective and prospectiveobservational studies, secondary database analyses, clinical and patient-reported outcomes studies, economic analysisand design of registries.

• Previously VP Epidemiology North America at MAPI, Carl has also held roles as Executive Director, Epidemiology & HealthOutcomes at PPD, and VP Health Outcomes and Pharmacoeconomics at Supportive Oncology Services, Inc. Prior tojoining the industry, he was Associate Professor, Department of Psychiatry at Harvard Medical School and AssociateProfessor, Department of Biostatistics at MD Anderson Cancer Center.

• Carl has authored more than 100 peer-reviewed publications, served as co-investigator or co-principal investigator onover 40 NIH funded grants and contracts, performed editorial reviews for 12 industry publications, and served as thesisand dissertation advisor to numerous graduate students. He holds a PhD in Biostatistics from the University ofWashington with an emphasis on epidemiologic methods, and a Bachelor’s degree in Biology from San Diego StateUniversity.

David Grant, MbA• David Grant is Senior Principal RWE Solutions & HEOR at IMS Health, specializing in reimbursement and market access,

environmental analysis, prospective and retrospective data collection and communications for product support.• A co-founder and former Director of Fourth Hurdle, David’s experience spans more than 10 years in health economics and

outcomes research consulting, and 15 years in the pharmaceutical industry, including roles in clinical research, newproduct marketing and health economics in the UK and Japan.

• David holds an MBA from the London Business School and a degree in Microbiology.

EXPERTISE | IMS RWE SOLUTIONS & HEOR


Joshua Hiller, MbA• Joshua Hiller is Senior Principal RWE Solutions at IMS Health, supporting the strategic planning and development of

IMS Health capabilities for data sourcing, integration, analytics and studies. He is also currently serving as Alliance Directorin the company’s collaboration with AstraZeneca for the advancement of RWE.

• During a career that includes roles in market analytics, government and healthcare consulting in both the US and UK,Joshua has led a wide range of projects for clients in the pharmaceutical and biotech sector as well as industryassociations. He has extensive experience in pharmaceutical pricing, contracting, market landscape development, supplymanagement, cross border trade, lifecycle management, competitive defense, generics market drivers and accountmanagement, with expertise across US and European markets.

• Joshua holds an MBA (Beta Gamma Sigma) from Columbia Business School, New York, and a BS in Mathematics fromJames Madison University, Virginia.


Mark Lamotte, MD• Dr. Mark Lamotte is Senior Principal RWE Solutions & HEOR at IMS Health, responsible for project management and

quality assurance within his team, and for leadership of health economic modeling. • A medical doctor specialized in cardiology, Mark spent six years in clinical practice before joining Rhône-Poulenc Rorer as

Cardiovascular Medical Advisor and later becoming Project Manager and Scientific Director at the Belgian researchorganization, HEDM. He has worked on over 300 cardiovascular, pulmonary, diabetes, urology and oncology projects,incorporating expert interviews, patient record review, modeling and report writing. Many of these projects have resultedin peer-reviewed publications.

• Mark holds an MD from the Free University of Brussels (Vrije Univeristeit Brussel, Belgium) and is fluent in Dutch, French,English and Spanish.

Won Chan Lee, PHD• Dr. Won Chan Lee is Principal RWE Solutions & HEOR at IMS Health, specializing in prospective and retrospective health

economics research.• Over the course of his career, Won has completed numerous international economic evaluations employing a variety of

analytical methods across a range of diseases and geographies. His expertise includes econometric database analysis,quality of life assessment and advanced economic modeling to establish the economic and humanistic value of new andexisting therapeutic interventions.

• Won holds a PhD in Economics from the Graduate Center of the City University of New York and a Master’s degree inEconomics from the University of Grenoble II.

Jacco Keja, PHD• Dr. Jacco Keja is Senior Principal RWE Solutions & HEOR at IMS Health, drawing on deep expertise in global market access,

operational and strategic pricing, and health economics and outcomes research. • Jacco’s background includes four years as global head of pricing, reimbursement, health outcomes and market access

consulting services at a large clinical research organization and more than 13 years experience in the pharmaceuticalindustry, including senior-level international and global roles in strategic marketing, pricing and reimbursement andhealth economics.

• Jacco holds a PhD in Biology (Neurophysiology) from Vrije Universiteit in Amsterdam, a Masters in Medical Biology, and an undergraduate degree in Biology, both from Utrecht. He is also visiting Professor at the Institute of Health Policy & Management at Erasmus University, Rotterdam.

Tim Kelly, MSC, bS• Tim Kelly is Vice President RWE Solutions at IMS Health, with responsibility for the company’s RWE data assets and data

architecture backbone, and for overseeing platform delivery infrastructure and engagements to ensure at-scale, high-quality data mart deployment. He also leads the client services team supporting data and technology applications.

• Tim’s background includes two decades of life-science experience managing large-scale data warehousing, technology,and analytic applications and engagements. He has worked with many clients in the pharmaceutical and biotech sectors,leveraging deep expertise in information management and modeling, commercial operations and analytics, advancedanalytics, business intelligence, data warehousing and longitudinal analytics.

• Tim holds a Master’s degree in Management Science from Temple University, Philadelphia and a Bachelor’s degree inQuantitative Business Analysis from Penn State University.



benjamin Hughes, PHD, MbA, MRES, MSC• Dr. Ben Hughes is Senior Principal RWE Solutions at IMS Health, leading the development of the RWE service offering. He

has helped many clients in the pharmaceutical industry to articulate and implement their RWE strategies, throughdefinition of RWE vision, business cases for RWE investments, capability roadmaps, partnerships, brand evidence reviews,HEOR function design, RWE training programs and related clinical IT strategies.

• Previously head of the European RWE service line at McKinsey & Co, Ben has extensive experience advising healthcarestakeholders on health informatics and RWE-related topics. This includes work on France’s electronic health recordstrategy, EMR adoption strategy for governments across Europe and Asia, data releases to support the UK’s transparencyagenda, and the development of payer health analytics and RWE capabilities across countries in Europe.

• A widely published author on health informatics, Ben holds a PhD in Medical Informatics from ESADE Barcelona, an MBAfrom HEC Paris, and Masters’ degrees in Research from ESADE Barcelona and in Physics from University College, London.

Claude Le Pen, PHD• Dr. Claude Le Pen is a member of the strategic committee of IMS Health and Professor of Health Economics at

Paris-Dauphine University, providing expert economic advisory services to the consulting practice.• A renowned economist, leading academic and respected public commentator, Claude has served as an appointed senior

member of several state commissions in the French Ministry of Health and is an expert for a number of parliamentarybodies, bringing a unique perspective and unparalleled insights into the economic evaluation of pharmaceuticaltechnologies at the highest level.

• Claude studied Business Administration in HEC Business School in Paris and holds a PhD in Economics from Panthéon-Sorbonne University.

Frédérique Maurel, MS, MPH• Frédérique Maurel is Principal RWE Solutions & HEOR at IMS Health, with a focus on observational research and health

economics studies.• A skilled consultant and project manager, Frédérique has extensive experience in the economic evaluation of medical

technologies gained in roles at ANDEM, Medicoeconomie, and AREMIS Consultants.• Frédérique holds a Master’s degree in Economics – equivalent to an MS – and completed a post-graduate degree

equivalent to an MPH with a specialization in Health Economics at the University of Paris-Dauphine (Paris IX) as well as adegree in Industrial Strategies at the Pantheon-Sorbonne University (Paris I).

Joan McCormick, MbA• Joan McCormick is Principal at IMS Brogan, leading a team providing strategic advice to companies with new products

coming to market and ongoing consultation on the rules for existing drugs post launch. • Formerly Head of Price Regulation Consulting at Brogan Inc, Joan has supported many major pharmaceutical companies

with the preparation of pricing submissions to the Patented Medicine Prices Review Board (PMPRB), gaining extensiveinsights into the operation of the Canadian pharmaceutical market.

• Joan holds an MBA from the University of Ottawa, Canada and a Bachelor’s degree in Life Sciences from Queen’s Universityin Kingston, Canada.

Charles Makin, MS, MbA, MM, bS• Charles Makin is Principal RWE Solutions & HEOR at IMS Health, leading value strategy development, economic

evaluations and health outcomes research studies and observational research. He has served as principal investigator on numerous US-based and global database analyses, adherence interventions, systematic literature reviews, and PRO research.

• Charles has deep insight into best practices in global research and market access. He was previously Global Head ofResearch Design and Proposal Development at UnitedHealth Group. He also worked as Research Operations Manager atWellPoint, leading project teams to execute HEOR projects.

• A widely published author, Charles holds a Master’s degree in Pharmacy Administration from Purdue University, Indiana,an MBA (summa cum laude) in Marketing Management and a Master’s degree in Management (summa cum laude), bothfrom Goldey Beacom College, Delaware, and a Bachelor’s degree in Pharmacy from the University of Pune, India.

Julie Munakata, MS• Julie Munakata is Principal RWE Solutions & HEOR at IMS Health, with a focus on global economic modeling, value

development planning, and survey data analysis.• An accomplished researcher and author of more than 25 original articles, Julie has extensive experience in managing

clinical trials, health economic studies and decision analytic modeling work, gained in senior roles at ValueMedicsResearch LLC, the VA Health Economics Resource Center and Stanford Center for Primary Care & Outcomes Research, andWyeth Pharmaceuticals.

• Julie holds an MS in Health Policy and Management from the Harvard School of Public Health and a BS in Psychobiologyfrom the University of California, Los Angeles.


Adam Lloyd, MPHIL, bA• Adam Lloyd is Senior Principal RWE Solutions & HEOR at IMS Health, with a focus on economic modeling and the global

application of economic tools to support the needs of local markets.• A former founder and Director of Fourth Hurdle, and previously Senior Manager of Global Health Outcomes at

GlaxoWellcome, Adam has extensive experience leading economic evaluations of pre-launched and marketed products,developing submissions to NICE and the SMC, decision-analytic and Markov modeling, and in the use of health economicsin reimbursement and marketing in continental Europe.

• Adam holds an MPhil in Economics and a BA (Hons) in Philosophy, Politics and Economics from the University of Oxford.

Michael Nelson, PHARMD • Dr. Michael Nelson is Senior Principal RWE Solutions & HEOR at IMS Health, with particular expertise in retrospective

database research, prospective observational research, health program evaluation, and cost-effectiveness analysis.• During a career that includes leadership roles in HEOR at PharmaNet, i3 Innovus, SmithKline Beecham, and

DPS/UnitedHealth Group, Mike has gained extensive experience in health information-based product development,formulary design, drug use evaluation, and disease management program design and implementation.

• A thought leader in health economics for more than 20 years, Mike holds a doctorate in Pharmacy and a Bachelor ofScience degree, both from the University of Minnesota College of Pharmacy. He also served as an adjunct clinical facultymember at the University of Minnesota whilst in clinical pharmacy practice.



IMS HEALTH ECONOMICS & OUTCOMES RESEARCHPage 72 IMS HEALTH ECONOMICS & OUTCOMES RESEARCH

Amy O’Sullivan, PHD• Dr. Amy O’Sullivan is Principal RWE Solutions & HEOR at IMS Health, with a focus on global economic modeling to support

product reimbursement in jurisdictions around the world. • Highly experienced in the economic evaluation of medical technologies, Amy’s background includes roles at Policy

Analysis Inc. (PAI) and most recently in a senior capacity at OptumInsight. She has led numerous pharmacoeconomic andoutcomes research studies including cost-effectiveness analyses, budgetary impact analyses, burden-of-illness studies andpiggyback economic evaluations. Her research spans a wide range of therapeutic areas, including autoimmune conditions,CV disease, CNS, metabolic disorders, musculoskeletal conditions, oncology, respiratory disease and women’s health.

• Amy holds a PhD in Health Economics from the Johns Hopkins University Bloomberg School of Public Health, Baltimore,and a BA in Economics and English from Boston College.

Carme Piñol, MD, MSC• Dr. Carme Piñol is Principal RWE Solutions & HEOR at IMS Health, with more than 20 years experience in the

pharmaceutical industry spanning clinical research, health economics and market access. • Previously Head of Market Access for Spain at Bayer, a role that included pricing, HEOR, advocacy and institutional

relations with the Regions, Carme is a Board member of the Spanish Association of Health Economics as well as the ISPORSpain Regional Chapter and coordinator of the Pharmacoeconomics Interest Group of the Spanish Association of Medicineof the Pharmaceutical Industry (AMIFE). She has authored more than 60 communications in international and nationalcongresses and more than 20 papers in peer-reviewed journals.

• Carme holds an MD from the Autonomous University of Barcelona, an MSc in Pharmacoeconomics and Health Economics fromPompeu Fabra University, an MSc in Health Research from Castilla-La Mancha University (UCLM), and an Executive ProgramDegree from ESADE Business School, Barcelona, Spain. She is currently completing a PhD in Health Research at UCLM.

Stefan Plantör, PHD, MbA, MSC• Dr. Stefan Plantör is Principal RWE Solutions & HEOR at IMS Health, with a focus on AMNOG-related projects, including

benefit dossiers, as well as reference price management, health economic evaluations and health policy analyses.• Stefan’s background includes roles as a researcher and five years experience in the pharmaceutical industry. He has also

served as a board member of ProGenerika, the German pharmaceutical association. Over the course of his career, Stefanhas broadened his expertise to include data analyses and decision analytic modeling, authored a number of publicationsin international journals and presented his research at major congresses.

• Stefan holds a PhD in Biology from the University of Tübingen, an MBA in International Marketing from the EuropeanBusiness School, Reutlingen and an MSc in Microbiology from the Eberhardt-Karls-University (Tübingen).

Shibani Pokras, MPH • Shibani Pokras is Principal Product Offerings, RWE Solutions & HEOR at IMS Health, applying her outcomes research skill

set in developing and marketing new data products for commercial, academic and government researchers.• An award-winning outcomes researcher, Shibani has worked with clients in the life science industries to design analyses

for multiple research databases and convert these into novel interactive payer-ready applications. She has extensiveexperience in developing product-specific health economic value strategy informed by close interactions with managedcare payers, clinicians and scientific research.

• Shibani’s background includes roles at ValueMedics Research LLC, and a fellowship award through NovartisPharmaceuticals and the Duke University Center for Clinical and Genetic Economics. Shibani holds an MPH from the YaleSchool of Public Health and Yale School of Management, and a BS with Honors in Life Sciences and Biochemistry from St.Xavier’s College in Mumbai, India.


Jon Resnick, MbA• Jon Resnick is Vice President and General Manager RWE Solutions at IMS Health, leading the company’s global RWE &

HEOR business, including the development of RWE strategy, offerings, collaborations and foundational technologies tomeet the RWE needs of healthcare stakeholders.

• A former Legislative Research Assistant in Washington DC and member of the Professional Health and Social Security stafffor the US Senate Committee on Finance, Jon has 10 years consulting experience at IMS. He was most recently responsiblefor leading the European management consulting team and global HEOR business teams of 300 colleagues, advisingclients on a wide range of strategic, pricing and market access issues.

• Jon holds an MBA from the Kellogg School of Management, Northwestern University, with majors in Management andStrategy, Finance, Health Industry Management, and Biotechnology.


ACCESSPOINT • VOLUME 4 ISSUE 7 PAgE 73AccessPoint - Issue 3 Page 73


Arnaud Troubat, PHARMD, MbA, MHEM• Dr. Arnaud Troubat is Principal RWE Solutions & HEOR at IMS Health. He has extensive consulting experience and special

expertise in the development of registration dossiers and market access strategies across a large number of therapeutic areas. • A pharmacist by training, Arnaud began his career at the French pharmaceutical industry association (LEEM). He then

spent a number of years in the pharmaceutical affairs department at ICI, leading regulatory work on registrationsubmissions and reimbursement strategies, before subsequently moving into consulting. Most recently he was Director atCarré-Castan Consultants, managing a research team.

• Arnaud holds a Doctor of Pharmacy degree and an MBA from IAE Paris and a Master’s degree in Health Economics andManagement from Paris-Dauphine University.

Núria Lara Surinach, MD, MSC• Dr. Núria Lara is Senior Principal RWE Solutions & HEOR at IMS Health, with a focus on the design and coordination of local

and international observational and patient-reported outcomes studies.• A former practicing GP and clinical researcher, Núria’s experience spans roles in outcomes research at the Institute of

Public Health in Barcelona and in Catalan Health Authorities, and consulting positions within the pharmaceutical andmedical device industries focusing on medical regulatory and pricing affairs, pharmacoeconomics and market access strategies.

• Núria holds an MD (specializing in Family and Community Medicine in Barcelona), and a Master’s degree in Public Healthfrom the London School of Hygiene and Tropical Medicine and London School of Economics.

Massoud Toussi, MD, MSC, PHD, MbA • Dr. Massoud Toussi is Principal and Medical Director RWE Solutions & HEOR at IMS Health, applying his expertise to assure

the quality of outcomes research. He is also the representative of IMS Health in ENCePP.• Previously head of Global Clinical Research Operations at Cegedim, Massoud has also worked with the French High

Authority for Health (HAS) and various CROs as Project Lead, Scientific Manager and Operations Director. His experienceincludes drug safety reporting, natural language processing, database linkage and drug utilization studies.

• Massoud holds an MD from Mashad University in Iran, an MSc in Medical Informatics and Communication Technologyfrom Paris VI, a PhD in Medical Informatics from Paris XIII University, and an executive MBA from a joint program ofUniversities of Paris-Dauphine and Quebec à Montreal.

Rolin Wade, RPH, MS• Ron Wade is Principal, RWE Solutions & HEOR at IMS Health, and a recognized expert in the applications and limitations of

using large retrospective datasets and late-phase datasets for health economics and outcomes research.• Prior to joining IMS Health, Ron served as a Healthcare Executive and Principal Investigator with Cerner Research and as a

Research Director at HealthCore. He also has experience generating evidence to support value messages to managedcare, government payers and public health associations, gained in leadership roles within the pharmaceutical industry.

• A widely published author with expertise in many therapy areas, Ron lectures at colleges of pharmacy and he has hadleadership roles with the American College of Clinical Pharmacy and the Academy of Managed Care Pharmacy. He is alicensed pharmacist and holds an MS in Pharmaceutical Sciences from the University of the Pacific, California and a BS in Pharmacy.

Jovan Willford, MbA• Jovan Willford is Principal RWE Solutions at IMS Health, supporting growth strategy, offering development and

commercialization of RWE solutions. • Jovan’s background includes more than 10 years of strategic advisory experience across payers, providers, life science

organizations and technology companies, including several cross-industry collaborations to advance quality and value ofcare delivery.

• Jovan holds an MBA from the Kellogg School of Management, Northwestern University, with majors in Management andStrategy, Managerial Economics and International Business, and an undergraduate degree from the University of NotreDame with majors in Marketing and Philosophy.

Vernon Schabert, PHD• Dr. Vernon Schabert is Senior Principal RWE Solutions & HEOR at IMS Health, with a focus on the assessment and

validation of PRO instruments, retrospective analyses of database, and primary data collection surveys.• A founder and former President of Integral Health Decisions, Inc, Vernon has extensive experience in conducting claims

analyses, creating custom administrative databases, developing business intelligence software, and leading nationalquality research projects, gained in roles with Thomson Reuters, Strategic Healthcare Programs LLC, and CIGNAHealthCare. His expertise spans numerous disease areas and diverse research topics.

• Vernon holds a PhD in Personality and Social Psychology from Stanford University and a BA in Psychology from Princeton University.


About IMS Health

IMS Health is a leading worldwide provider of information, technology, and services dedicated to making healthcareperform better. With a global technology infrastructure and unique combination of real-world evidence, advancedanalytics and proprietary software platforms, IMS Health connects knowledge across all aspects of healthcare to helpclients improve patient outcomes and operate more efficiently. The company’s expert resources draw on data fromnearly 100,000 suppliers, and on insights from 39 billion healthcare transactions processed annually, to serve more than5,000 healthcare clients globally. Customers include pharmaceutical, medical device and consumer health manufacturersand distributors, providers, payers, government agencies, policymakers, researchers and the financial community.Additional information is available at www.imshealth.com

©2013 IMS Health Incorporated and its affiliates. All rights reserved. Trademarks are registered in the United States and in various other countries.

ACCESSPOINT1113

EUROPE210 Pentonville RoadLondon N1 9JYUnited KingdomTel: +44 (0) 20 3075 4800

[email protected]/rwe

THE AMERICAS1725 Duke StreetSuite 510Alexandria, VA 22314USATel: +1 (703) 837 5150

ASIA PACIFIC8 Cross Street #21-01/02/03PWC BuildingSingapore 048424Tel: +65 6412 7365

LATIN AMERICAInsurgentes Sur # 2375 5th FloorCol. TizapanMéxico D.F. - C.P. 01090México Tel: +52 (55) 50 62 52 00

IMS REAL-WORLD EVIDENCE SOLUTIONS AND HEALTH ECONOMICS & OUTCOMES RESEARCHis based in 18 countries worldwide with regional headquarters in:

ims health rwe accesspoint 7 - november 2013

Documents

rwegenerated insights

rwe landscape

thepower of rwe

ims health economics

safety7 ims institute

realworld successsolutions

leading experts

new databasetechniques