Jumping the crevasse between assertions of drug interactions and 

clinical relevanceDaniel C. Malone, RPh, PhD

Professor The University of Arizona

You wouldn’t believe how many BIG BAD drug‐drug interactionsthere are.  Just ask your doctor about all the DDI alerts she gets!

The Problem!

Prescriber’s KnowledgeComputer Screening 

Pharmacist’s Knowledge

Latent Failures

Patient Risk Factors

Patient Education

Monitoring

ADR

A + B“When the Holes Line Up”

Defenses

Hansten PD, Horn JR. Modified from: James Reason, Human Error, 1990

Drug Administration

Market Removals Due to Drug‐Drug Interactions

• Terfenadine (Seldane®) – 1998• Mibefradil (Posicor®)‐ 1998• Astemizole (Hismanal®) – 1999• Cisapride (Propulsid®) – 2000• Cerivastatin (Baycol®) – 2001

5

VA practitioner knowledge of drug-drug interactions

• 168 responses to a postal survey• 135 physicians• 22 nurse practitioners• 11 physician assistants

• Clinicians correctly categorized 53% of drug-drug interactions

• But, only• 64% correctly answered sildenafil-isosorbide (28% not sure)• 58% correctly answered cisapride-erythromycin (27% not

sure)• 43% correctly answered phenelzine-sertraline (46% not sure)

Source: Glassman PA et al. Medical Care 2002; 40:1161-1171

6

National Survey of Prescribers’Knowledge of DDIs - Methods

• Postal survey of prescribers (12,500)• Sample: Identified via pharmacy claims to a pharmacy benefit

manager• Cases – history of 1 or more DDI’s• Controls – match on prescribing either objective or precipitant

medication• Practice characteristics• Respondents asked to classify 14 drug pairs

• Contraindicated• May be used together but with monitoring• No interaction• Not sure

• Usual source of drug-drug interaction information

Ko et al. Drug Saf. 2008;31(6):525‐536. 

7

Summary of Prescriber DDI Knowledge

• Correct classification of drug-pairs• Mean (SD) = 6.0 (3.1)• Overall – 42.7% of drug pairs correctly

identified• 30% or more of respondents answered “unsure”

for 8 of the 16 the drug pairs• 2 combinations are contraindicated

• Drug‐drug interaction alerts can be useful• Prescriber knowledge of DDIs is lacking1,2

• 42.7% of drug pairs correctly identified1

• VA practitioners generally agree that DDI alerts are useful3

1) Ko et al. Drug Saf. 2008;31(6):525‐536. 2) Glassman. Med Care. 2002;40(12):1161‐1171. 3) Ko et al. JAMIA 2007;14:56‐64

Background

Sensitivity of Computer Software to Detect Drug Interactions in Arizona Pharmacies (N=64)

89%

86%

88%

45%

81%

90%

75%

84%

87%

83%

80%

70%

75%

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%

Carbamazepine + clarithromycin

Digoxin + amiodarone

Digoxin + clarithromycin

Digoxin + itraconazole

Nitroglycerin + sildenafil

Simvastain + itraconazole

Simvastatin + amiodarone

Simvastatin + gemfibrozil

Warfarin + amiodarone

Warfarin + fluconazole

Warfarin + gemfibrozil

Warfarin + naproxen

Warfarin + sulfamethoxazole/trimethoprim

Saverno et al. JAMIA; 2011:18:32-37

DDI Prevalence in Elderly

• Elderly veterans with new DDI at ED discharge:1 13%

• Older adults exposed to a “major” DDI:2 4%

• Medicare Part D enrollees exposed to certain DDIs: 7.3%

1) J Am Geriatr Soc. 2008;56:875‐80.  2) JAMA. 2008;300:2867‐78. 

Who Prescribes Drug‐Drug Interactions?

010203040506070

Potential Drug‐Drug Interactions by the Same Prescriber

Concordance Among DDI Compendia

Methods• Review of print versions of compendia for “major” interactions• Inclusion Criteria:

• DDI listed in at least 3 compendia

• Available in US for human use

• Medications likely to be dispensed in community pharmacy

• Medications likely captured in electronic database

• Interacting medications not used for therapeutic benefit

Abarca J et al. J Am Pharmacist Assn 2003: 44:136-141.

Rating Systems & Selection Criteria

• Evaluation of Drug Interactions• Uses 4-item summary measure based on:

• Potential harm to the patient• Frequency and predictability of occurrence• Degree and quality of documentation

• Code 1: highly clinically significant• Code 2: moderately clinically significant• Code 3: minimally clinically significant• Code 4: not clinically significant

Abarca J et al. J Am Pharmacist Assn 2003: 44:136-141.

Rating Systems & Selection Criteria

• Drug Interaction Facts• Uses 5-item summary measure based on:

• Severity (i.e., major, moderate, minor)• Documentation (i.e., established, probable,

suspected, possible, unlikely)• 1: major/established, probable, suspected• 2: moderate/established, probable, suspected• 3: minor/established, probable, suspected• 4: major,moderate/possible• 5: minor/possible or any/unlikely

Abarca J et al. J Am Pharmacist Assn 2003: 44:136-141.

Rating Systems & Selection Criteria

• Drug Interactions: Analysis and Management• Used 5-item summary measure based on:

• Severity• Corresponding documentation• Availability of alternatives are considered

• 1: Avoid combination• 2: Usually avoid combination• 3: Minimize risk• 4: No action needed• 5: No interaction

Abarca J et al. J Am Pharmacist Assn 2003: 44:136-141.

Rating Systems• Drug-REAX (MicroMedex)

• Used 5-item severity scale• Major• Moderate• Minor• None• Not specified

• No summary measure

Abarca J et al. J Am Pharmacist Assn 2003: 44:136-141.

Results• 62 ‘major’ DDIs identified• Additional criteria:

• 18 ‘major’ DDIs excluded :• 8 DDIs - not available in U.S. (e.g., terfenadine, mibefradil)• 4 DDIs – not dispensed from a community pharmacy• 4 DDIs – not likely to be captured in electronic database (e.g,

ethanol, tyramine-containing foods)• 1 DDI – occurs upon discontinuation (clonidine-β blockers)• 1 DDI – used for therapeutic benefit (phenothiazine-SSRI)

Abarca J et al. J Am Pharmacist Assn 2003: 44:136-141.

DDIs in 4 of 4: 2.2% (9/406) DDIs in 3 of 4: 8.6% (35/406) DDIs in 2 of 4: 17.4% (71/406) DDIs in 1 of 4: 71.7% (291/406)

Intra-class Correlation Coefficient: -0.092

Agreement Among Four Drug Interaction Compendia

Abarca J et al. J Am Pharmacist Assn 2003: 44:136-141.

Clinical Pharmacology and Therapeutics 2011: 87:48‐51 

“Only 13.6% of “critical interactions” listed in both compendia

Quantity and Quality of DDI Evidence – Interactions with Macrolides

Harper, Jackson, and Malone – unpublished data

Reasons for Differences in Drug‐Drug Interaction Compendia

• Rating systems are different• Lack of high‐quality studies

• Case reports• Small pharmacokinetic studies

• Rating is subjective• Different editors/contributors

• Few “tools” for evaluating “poor quality” evidence

Hierarchy of Evidence

Systematic Reviews

RCTs

Controlled Clinical Trials and Observational Studies

Uncontrolled Observational Studies

Case reports and case series

Expert Opinions

Lowest risk of bias

Hypothesis Testing

Hypothesis Generating

Theophylline – Allopurinol:A Major Drug Interaction?

Theophylline

3-Methylxanthine

1-methylurinc acid

1,3 dimethyluric acid

1-methylxanthine35%1A2 3A, 2E140%

16%1A2

Xanthine Oxidase

Theophylline – Allopurinol:A Major Drug Interaction?

• Listed as major in several CDS databases• Two studies found no change in

theophylline PK with allopurinol 300 mg daily x 7 days.1

• One study found ~25% increase in AUC and half-life after 2 weeks of concurrent allopurinol 300 mg BID.2

1. Vozeh S et al. CPT. 1980;27:194‐7.  2. Manfredi RL et al. CPT 1981;29:224‐9

Theophylline Label

Drug Type of Interaction Effect

Alcohol A single large dose of alcohol (3 mL/kg of whiskey) decreases theophylline clearance for up to 24 hours.

30% increase

Allopurinol Decreases theophylline clearance at allopurinol doses ≥600 mg/day.

25% increase

Table II. Clinically significant drug interactions with theophylline*. 

Source: http://dailymed.nlm.nih.gov/dailymed

Azithromycin Product Label

• “The following drug interactions have not been reported in clinical trials with azithromycin; however, no specific drug interaction studies have been performed to evaluate potential drug‐drug interaction. Nonetheless, they have been observed with macrolide products. Until further data are developed regarding drug interactions when azithromycin and these drugs are used concomitantly, careful monitoring of patients is advised:• Digoxin–elevated digoxin levels.• Ergotamine or dihydroergotamine–acute ergot toxicity 

characterized by severe peripheral vasospasm and dysesthesia.• Triazolam–decrease the clearance of triazolam and thus may 

increase the pharmacologic effect of triazolam.• Drugs metabolized by the cytochrome P450 system–elevations 

of serum carbamazepine, cyclosporine, hexobarbital, and phenytoin levels.”

Source: http://dailymed.nlm.nih.gov

Health Systems Approach to DDIs• Evidence for DDIs is lacking

• Very few well-controlled studies• Lack of concordance among DDI compendia1

• Differing severity rating systems, terminology, methodologies

• Limitations of DDI clinical decision support2-4

• “Alert fatigue”• High rates of alert override

1) Abarca et al. J Am Pharm Assoc (2003). 2004;44(2):136-141. 2) Grizzle et al. Am J Manag Care. 2007;13(10):573-578. 3) Murphy et al. Am J Health Syst Pharm. 2004;61(14):1484-1487. 4) Abarca et al. J Manag Care Pharm. 2006;12(5):383-389.

Problems With DDIs in CDS Systems

• Classification systems that are based on rules of questionable relevance

• Reliance on label without informed review or evaluation

• Assumption of ‘class–based’ interactions

• Conclusion: “the current DDI alert system is broken”1

1Hatton RC, et al. Ann Pharmacother. Mar 2011;45(3):297‐308. 

Model for DDI alerting in CPOE

Integrationsoftware

Knowledge‐base (rules)

Patient database (meds, etc.)

Knowledge engineering

Literature, domain expertise

Potential DDI report

MDPharmacistNurse

Too many alerts!

• Organizational alert override rates are high• 96% in Netherlands1

• US medical center2

• 461 different physicians, 18,354 medication orders

• 2,455 alerts• DDI override rate: 95.1%

1van der Sijs H, Aarts J, Vulto A, Berg M. Overriding of drug safety alerts in computerized physician order entry. J Am Med Inform Assoc. 2006 2006 Mar‐Apr;13(2):138‐47. 2Bryant AD, Fletcher GS, Payne TH. Drug interaction override rates in the Meaningful use era: no evidence of progress. Applied Clinical Informatics 2014; 5:802‐813

Case Study ‐ Sentara Healthcare• Amiodarone + warfarin DDI fires 1000 times

• 960 – alert overridden/bypassed• 40 – warfarin dose adjusted • 10 – INR or CBCs ordered• 30 – INR or CBC frequency adjusted• 50 – pt. education or anticoagulation clinic scheduled• 5 – alert overridden, INR > 5

Practitioners’ Views on DDI Alerts

• Methods• Random sample of 100 to 125 practitioners

in outpatient clinics at 6 VAMCs

• Results (N = 258) (Response rate: 36%)• Internal medicine – 31%• Primary care – 29%

• Years in practice (mean) - 14.8 (10.6)• Rx’s written per week – 97.7 (155.5)

Ko et al. JAMIA 2007; 14: 56-6434

Practitioners’ Views of Alerts

Prescribers’ Views on DDI Alerts Mean Response

I am satisfied with the accuracy of DDI alerting system

3.1

DDI alerts provide me with information I already know

3.4

The DDI system provides alerts that seems to be just about exactly what I need

2.7

DDI alerts change my initial prescribing decisions

1.9

1= strongly disagree; 2 = disagree; 3 = neither disagree or agree;4 = agree; 5 = strongly agree

35

Attitudes Toward Improving DDI Alerts

Attribute Mean Response

DDI alerts should be accompanied by management alternatives

3.8

DDI alerts should be accompanied by more detailed information about the interaction

3.7

DDI alerts should only appear once in the order entry process 3.6

DDI alerts are presented in a useful format 3.2

1= strongly disagree; 2 = disagree; 3 = neither disagree or agree;4 = agree; 5 = strongly agree

36

Top Ranked Changes to VA’s Alerting System

Change Weighted Preference Score

Provide management options for DDIs 1.53

Show DDI alerts one at a time 1.24

Make it more difficult to override lethal interactions

0.95

Eliminate all DDI alerts 0.05

Rank 1 = 3 points, Rank 2 = 2 points, Rank 3 = 1 point37

“Asthma sufferer wins $28.6 million award” –Seattle Time (9/3/94)

• 24 year old man on theophylline went to ER with an infection, prescribed ciprofloxicin by ER physician

• Theophylline levels doubled, patient had permanent brain damage

• Patient awarded $22.5 million in pain and suffering

Tranylcypromine (Parnate®) +Phenylpropanolamine (Tavist‐D®)

• Patient on Parnate went to ER for URI: received Rx for Tavist-D®

• Rx filled at his regular pharmacy (technician ignored computer warning)

• Patient developed an acute hypertensive episode resulting in a stroke

• Patient committed suicide, leaving a note that the disabilities induced by the stroke were intolerable

Williams KG. Am J Health-Syst Pharm 1996;53:1709.

Current “Lists” of Important DDIsSource Purpose Systematic 

ReviewExpert Panel

Consensus Process

Classen 2011 / Leapfrog List

Verify whether an inpatient COPE system has the potential to intercept critical DDIs 

Unknown Unknown Unknown

CMS FTag 329‐Unecessary Medications

Medications requiring increased involvement from consultant pharmacists. 

Unknown Yes  Unknown

Malone 2004To develop a list of clinically important DDIs in outpatient setting

Yes; primary literature and 

tertiary references

YesModified Delphi 

process

Phansalkar 2012 / ONC List

To reduce alert fatigue and describe the most clinically significant high‐priority DDIs

Unclear Yes Mixed approach

Pharmacy Quality Alliance

Evaluate prescription drug plans and ambulatory/community pharmacists

No No Yes

van Roon 2005 / G‐standard 

Dutch database for CDS in NetherlandsYes Yes Yes

Rationale for Changing the Approach

• Drug interaction clinical decision support shouldimprove patient safety

• Instead….• Evidence lacking to support effectiveness• Excessive irrelevant alerts• Wide variation across health systems for DDI CDS• Pandemic clinician annoyance 

• What’s needed• Guiding principles for evidence‐based, clinically relevant, 

consistent alerts with improved usability• Evidence of effectiveness

Rationale for Changing the Approach

• Concerns about process used to generate current “lists”

• No well‐defined, broadly accepted, uniform standard for rating:• Strength (quality) of DDI evidence• Strength of recommendations for patient risk management

• Concerns about liability

• DDI CDS conference series to:• Develop guidelines for systematic appraisal of DDI evidence (Evidence Workgroup)

• Recommend principles for including DDIs in CDS (Content Workgroup)

• Establish preferred strategies for presenting DDI CDS (Usability Workgroup)

• Consensus recommendations by international experts 

AHRQ Conference Series

https://sites.google.com/site/ddiconferenceseriessite

Project At A Glance

DDI CDS Conference Website

https://sites.google.com/site/ddiconferenceseriessite/home

Category EvidenceSufficientSufficient evidence to evaluate a clinically relevant drug interaction

One or more of the following:• Well‐designed and executed prospective controlled studies• Well‐designed and executed retrospective controlled studies• Case reports or series demonstrating probable or highly 

probable causality of an interaction (DIPS score of 5‐10)*• Reasonable extrapolation on the basis of:

• Studies of drugs with similar pharmacologic properties • Studies with in vitro substrate data• Human genetic polymorphism studies 

DRug Interaction eVidence Evaluation(DRIVE) Instrument

* Drug Interaction Probability Scale; Horn et al. Ann Pharmacother 2007;41:674‐80. 

DRug Interaction eVidence Evaluation(DRIVE) Instrument

Category EvidenceInsufficient Insufficient evidence to evaluate a clinically relevant drug interaction 

One or more of the following, without supporting evidence from the “sufficient” category:• Extrapolation on the basis of studies with in vitro 

inhibitor or inducer data• Case reports or series demonstrating only possible or 

doubtful causality of an interaction (DIPS score of <5)• Studies of poor design or execution • Hypotheses‐generating research methods• Unsupported data on file or unsupported 

recommendations from product manufacturer • Animal data 

* Drug Interaction Probability Scale; Horn et al. Ann Pharmacother 2007;41:674‐80. 

AHRQ Conference SeriesWorkgroup Recommendations

• Establish national expert consensus panel• Oversight by national organization • Develop and maintain standard set of DDIs for CDS• Employ transparent, systematic, evidence‐driven process 

• Grade quality of evidence• Incorporate expert and clinical advice• Provide graded recommendations for risk management• Collect and incorporate user feedback• Ongoing reevaluation and updates

1. Evidence Synthesis & Grading

2. Expert Advice

3. Consensus Graded 

Recommendations4. CDS 

Implementation

5. Stakeholder Feedback

6. Reevaluation& Updates

National DDI Expert Panel 

Oversight by Central Organization 

Transparent, Systematic Process for aStandard Set of DDIs

Support

Agency for Healthcare Research and Quality • Grant #1R13HS021826‐01 (PI – Malone)• Grant #1R13HS018307‐01 (PI – Malone)National Library of MedicineGrant #R01LM011838‐01 (PI – Boyce)

Additional Support• Cerner• Epocrates• First Databank• Truven Health Analytics• Wolters Kluwer Health• Elsevier

Acknowledgements• Lisa Hines, PharmD, University of Arizona• Richard T. Scheife, PharmD, FCCP, Tufts University• Darrell R. Abernethy, MD, PhD, Food and Drug Administration• Richard Boyce, PhD, University of Pittsburgh• Clarissa Borst, PharmD, Elsevier• Sophie Chung, PharmD, Epocrates, athenahealth, Inc.• Susan Comes, PharmD, Epocrates, athenahealth, Inc.• John Horn, PharmD, University of Washington• Gretchen Jones, PharmD, (formerly) Epocrates, athenahealth, Inc.• Jeremiah Momper, PharmD, PhD, University of California, San Diego• Alissa Rich, PharmD, MBA, (formerly) Cancer Treatment Centers of America• Stephen J Sklar, PharmD, Wolters Kluwer Health• Christine D Sommer, PharmD, FDB (First Databank, Inc.)• Tricia Lee Wilkins, PharmD, MS, Office of the National Coordinator for Health Information Technology• Michael A Wittie, MPH, Office of the Chief Medical Officer, Office of the National Coordinator for Health 

Information Technology• Samantha K Wong, BSPharm, RPh, Cerner

Acknowledgements• Lisa Hines, PharmD, University of Arizona• Hugh Tilson, MD, DrPH, University of North Carolina • David W Bates, MD, Harvard Medical School• Joseph T Hanlon, PharmD, MS, University of Pittsburgh• Philip Hansten, PharmD, University of Washington• Amy L Helwig, MD, MS, Office of the National Coordinator for HIT • Stefanie Higby‐Baker, RPh, MHA, CPHIT, Cerner Multum• Shiew‐Mei Huang, PhD, Food and Drug Administration• David R Hunt, MD, FACS, Office of the National Coordinator for HIT• Marianne le Comte, PharmD, Royal Dutch Association for the Advancement of Pharmacy• Karl Matuszewski, MS, PharmD, FDB (First Databank, Inc.)• Gerald McEvoy, PharmD, ASHP• Anthony Perre, MD, Cancer Treatment Centers of America• Lynn Pezzullo, RPh, CPEHR, Pharmacy Quality Alliance• John Poikonen, PharmD, MedVentive• Kathy Vieson, PharmD, Elsevier Clinical Solutions• David M Weinstein, RPh, PhD, Lexicomp, Wolters Kluwer Health• Michael A Wittie, MPH, Office of the National Coordinator for HIT

Acknowledgements• Thomas H Payne, MD, FACP, University of Washington • Bruce W Chaffee, PharmD, University of Michigan Health System• Raymond C Chan, PharmD, Sentara• Peter A Glassman, MBBS, VA, Greater Los Angeles Healthcare System • Brian Galbreth, PharmD, PeaceHealth Southwest Medical Center• Christian Hartman, PharmD, MBA, FSMSO, Pharmacy OneSource, Wolters Kluwer Health• Seth Hartman, PharmD, Oregon Health & Science University• Joan Kapusnik‐Uner, PharmD, FDB (First Databank, Inc.) • Gilad J Kuperman, MD, PhD, New York‐Presbyterian Hospital• Gordon Mann, RPh, Clinical Informatics, Epic• Shobha Phansalkar, RPh, PhD, Wolters Kluwer Health (formerly with Partners Healthcare System)• Alissa Russ, PhD, Roudebush VA Medical Center • Hugh Ryan, MD, Cerner Corporation• Howard Strasberg, MD, MS, Wolters Kluwer Health• Amanda Sullins, PharmD, Cerner• Vicki Tamis, PharmD, BCPS, PeaceHealth Southwest Medical Center• Heleen van der Sijs, PharmD, PhD, Erasmus Medical Center

Thanks!  Questions and Comments?

Answers (Even Better)?!