adding patient heterogeneity into phase iii trials_getreal_helenekarcher

The research leading to these results has received support from the Innovative Medicines Initiative Joint Undertaking under grant agreement no [115303], resources of which are composed of financial contribution from the European Union’s Seventh Framework Programme (FP7/2007-2013) and EFPIA companies’ in kind contribution. www.imi.europa.eu

A practical guide to adding patient heterogeneity into Phase III trials

Case study in schizophrenia

GetReal meeting – April 18th, 2016

Hélène Karcher


Context

• Learning about a new drug’s effectiveness is essential to appraising its benefit/risk in each real-life care setting/country

• Double challenge

– Effectiveness is rarely measured pre-authorization • Clinical development focuses on uncovering efficacy

• Pragmatic trials are usually not implemented early

– Effectiveness is not universal • Effectiveness is specific to a care setting / country!


Two ways to improve learning about effectiveness early in clinical development

1. More pragmatic design, i.e., any aspect of study design : population, type of randomization, blinding, monitoring, etc.

Analyses

2. Better “analyses tools” , i.e., any aspect of data analyses : statistical or model-based analyses, predictive models, etc.

Design

Clinical development trial


Factors that drive

effectiveness

EFFICACY

EFFECTIVENESS

? Drug use factors

Patient population factors

Interaction

- Patterns of use, dose, treatment duration. Can be defined with: 1. Adherence of prescribers to label recommendations,

2. Adherence of patients to prescriptions

- Past history of exposure - Variability of diagnosis

- Patient physical and behavioral characteristics: age, gender, weight, ethnicity, smoking/eating/exercise habits, etc.

- Co-morbidities - Disease stage/severity - Co-prescriptions - Other baseline risk factors and genetics relevant to the disease/drug

Health system care delivery

factors

- Type of setting for care delivery (e.g, hospital, home) - Type of prescriber : general physician, specialist, oncologist, nurse practitioner... - Socio-economic situation of health system, prescribers and patients


Systematic review of methods to incorporate pragmatism pre-authorization: results*

* Karcher, Nordon, Neumann, Nikodem, Zyla, Chevrou-Severac, Jimenez, Bala, Abenhaim. Methods to Evaluate Real-World Effectiveness in Pre-Authorization Trials SLR. HTAi 2015.

1. Many (39) methodological papers were identified that recommend how to relax trial features to make them more pragmatic, and to adapt analyses

2. However, this does not translate into many actual Phase 2-3 trials with pragmatic elements – due to scientific and operational hurdles – Systematic review only identified 18 pre-authorization trials with

pragmatic elements – Typically only 1-2 selected features are pragmatic

• Features required to conduct the trial for authorization • Features that could demonstrate a benefit not present in an RCT setting


Hurdles to incorporating effectiveness before authorization* (review of 39 articles)

1. Known and unknown confounders in real-world trials may lead to inconclusive effect sizes18,25

2. Extensive cost of running such trials due to larger sample size required14

3. Operational difficulties in recruiting certain populations, and in minimising measurements/study visits30,31

4. Uncertainty in reactions from regulatory bodies30,32

* Karcher, Nordon, Neumann, Nikodem, Zyla, Chevrou-Séverac, Jimenez, Bala, Abenhaim. Methods to Evaluate Real-World Effectiveness in Pre-Authorization Trials SLR. HTAi 2015.


Case study : Test broadening of eligibility criteria in pre-authorization trials in

schizophrenia


Real-world study to test eligibility criteria in schizophrenia

• Objective – Explore how to mitigate strict eligibility criteria in

Phase 2/3/3b trials (=RCT, randomized controlled trials) with real-life population heterogeneity


* “Reverse” of the method used in Schneeweiss et al. Increasing Levels of Restriction in Pharmacoepidemiologic Database Studies of Elderly and Comparison With Randomized Trial Results. Med Care. 2007

Real-world study to test eligibility criteria in schizophrenia

• Method*: use real-world data to optimize clinical trials 1. Study patient characteristics and interplay between factors and outcome

in a real-life schizophrenia population (SOHO cohort)

2. Identify the subpopulation eligible for a typical pre-authorization trial : “reference RCT population”

3. Re-include in this “reference RCT population” a minimal subset of patients who would usually be excluded (=broaden the eligibility criteria)

• Method of quotas (stratification) for patient inclusion in trials

• Combined with predictive modeling to evaluate the outcome in the RW population

4. Measure how “efficient” each re-inclusion is


Data source : SOHO study

• A prospective, observational study of 10,218 schizophrenia patients – from 10 European countries

– followed over 3 years

– who initiated an antipsychotic treatment

Outcome: CGI-S score • Clinical Global Impression-Severity • Assesses severity of the patient’s mental illness at time of rating

with one question • 7-point scale: from 1 (not at all ill) to 7 (extremely ill) • We used mean CGI-S at 3 months (change from baseline) as

outcome.


* Leucht et al. Comparative efficacy and tolerability of 15 antipsychotic drugs in schizophrenia: a multiple-treatments meta-analysis. Lancet. 2013

Identify a reference RCT population within SOHO

2,132 patients were selected to define a “reference RCT population” with the following eligibility criteria (taken from a meta-analysis of 212 clinical trials*):

The patients were then distributed in groups according to the drug they initiated.

Eligibility criteria applied to define the reference RCT population • Age between 18 and 65 years old • Duration of illness superior to 3 years • BMI between 17 and 40 • No history of alcohol or drug abuse • Patient with compliance to prescribed antipsychotic therapy • Patient without any past suicide attempt • Patient included in public or combined practices


Identify a reference RCT population within SOHO (observational data source)

12

All typical RCT eligibility criteria applied


Drug effects in reference RCT vs SOHO population

• Real-life effect is slightly better than effect in RCT under all 4 drugs – Excluded patients respond better to drug (trend)

• Cannot (yet) compare between drugs since patients under different drugs may intersect and may have uncomparable characteristics

Patients

taking

drug:

Mean CGI-S at 3 months (change from baseline)

Reference RCT population SOHO (real-life population)

Drug R n=1127 -0.78 (SD = 0.95) n=5591 -0.88 (SD = 0.82)

Drug AE n=498 -0.66 (SD = 0.89) n=2188 -0.71 (SD = 0.94)

Drug AD n=199 -0.54 (SD = 0.99) n=847 -0.60 (SD = 0.99)

Drug H n=118 -0.77 (SD = 0.84) n=456 -0.87 (SD = 0.97)

Note : “R” , “AE”, “AD”, “H” are the most popular drugs in the SOHO study.

13


Predicting RW drug effects using data only from the reference RCT population

14


Enriching RCTs to improve predictions

15


Enriching RCTs to improve predictions

16


Patients eligible to enroll into an “enriched” RCT with 1 broadened criterion (as % of reference RCT population, patients under drug R)

Broadened eligibility criterion

Disease duration 1-3 years

1 past suicide attempt

Private practice

Age > 65 years

Alcohol abuse

Drug abuse

BMI < 17 or > 40

% %

%

% % %

%

Size of real-world subpopulations that can be re-included in trials differs with each criterion

Opening the trial to patients with a shorter disease duration enable to tap into a wider patient pool than opening it to patients with alcohol or drug

abuse. Assuming all patients are equally difficult to recruit, opening the trial to patients with e.g., a shorter disease duration, enables faster recruitment.




Private practice

Age > 65 years

Alcohol abuse

Drug abuse

CGI-S at 3 months in re-included patients under drug R

BMI < 17 or > 40

Mean

Median

Mean CGI-S in reference RCT population

Outcomes for re-included real-world subpopulations slightly differ from the reference RCT population.

As expected, certain subpopulations have a trend to benefit more from treatment. Their re-inclusion is a priori beneficial to trial success.




Private practice

Age > 65 years

Alcohol abuse

Drug abuse

BMI < 17 or > 40

% %

%

% % %

%

0

20

15

10

5

Pat

ien

ts e

ligib

le t

o e

nro

ll in

to a

R

CT

wit

h 1

bro

ade

ne

d c

rite

rio

n

(as

% o

f re

fere

nce

RC

T)

C

GI-

S at

3 m

on

ths

in

re

-in

clu

de

d p

atie

nts

The mean CGI-S at 3 months in reference RCT population



Private practice

Age > 65 years

Alcohol abuse

Drug abuse

BMI < 17 or > 40

Mean Median

-4

2

0

-2

Both size and outcome of real-world subpopulations need to be considered before re-inclusion!


The following ordered probit regression model was used:

• The above covariates were chosen among all available ones

through a Chi-square test for independence. • The model was fitted with either reference RCT or enriched

RCTs data, then used to predict the real-life drug effect in SOHO.

CGI-S at 3 months ~ (age + chronicity + gender + BMI + hospitalization + number of admissions in hospital + depression score + QOL score + patient compliance + country + work status + housing condition + social activities + relationship + negative symptom at baseline + positive symptom at baseline + cognitive symptom at baseline + dosage DDDeq) I {if initiated the drug at baseline}

Prediction of real-life effects using only data from reference RCT or enriched RCT populations

20


Evaluation of prediction accuracy

• As each enrichment requires random replacement of patients, 100 independent repetitions were performed.

• Several enrichment factors were studied: suicide attempts, duration of illness (chronicity), practice type, alcohol abuse, drug abuse and age.

21


Prediction accuracy from different RCTs enriched in patients with shorter duration of illness

reference RCT popualtion -> less robust prediction

Enriched RCTs -> better prediction

Enrichment of reference RCT in patients with chronicity between 1 and 3 years (# patients)

22

Prediction accuracy

Only reference RCT population -> less robust prediction



Prediction accuracy from different RCTs enriched in patients with past suicide attempts

Only reference RCT population -> less robust prediction


enrichment of reference RCT in patients with 1 suicide attempt (# patients) 23

Prediction accuracy


Prediction accuracy from 4 different RCTs enriched in patients with different characteristics

24

enrichment of reference RCT with 4 subpopulations (# patients under drug R)

Target RCT size = 1127

Prediction accuracy


Prediction accuracy from different RCTs enriched in patients with either or both characteristics

25

RCT enrichment with either subpopulations (# patients under drug R)

Pre

dic

tio

n a

ccu

racy

Target RCT size = 1127


The probability of Success of a parallel RCT increases with enrichment in RW subpopulations

26

Method: - Simulate a parallel

RCT with 250 patients in each arm (drug R vs AE)

- Random sampling within (enriched) RCT populations under drug R or AE

- Propensity score matching

- Simulate 1000 trials Sample size unchanged 250 patients/arm


Conclusion – methods

• We used RW data to guide addition of patient heterogeneity to standard Phase 3 trials in schizophrenia.

• The impact of the following trial design changes was assessed: – Relax a few, selected exclusion criteria in a controlled way – Quantify the gain in accuracy of effectiveness predictions – Measure probability of success of the new trial design

while keeping sample size constant

27


Conclusion – results

• The best choice of enrichment factor to predict real-life effects was found to be driven by:

– Size of the excluded real-life population. Re-including

“chronicity 1-3 years” and “number of past suicide attempts > 1” increased the most the pool of schizophrenia patients eligible for Phase 3 trials.

– Change in outcome in patients with this factor. Patients with

a practice type “private” and disease chronicity between 1-3 years had the most different outcome from typical Phase 3 patients.

• Enriching typical Phase 3 with selected factors improved the representability of real-life and as a result, it improved predictions of the real-life effects of the investigated drug.

28


Conclusion – take-home message

29

Including a more heterogeneous population in a Phase 3/3b trial: • Enables to gather appropriate data for more

accurate predictions of real-world effects, • Can increase both Probability of trial success &

recruitment speed, – if done in rationally relaxing a few, selected eligibility criteria


Acknowledgments

• LASER

– Shuai Fu, Clementine Nordon, Lucien Abenhaim

• Eli Lilly

– Mark Belger, Iris Goetz

• University of Ioannina

– Orestis Efthimiou

• Saint John of God Health Park in Barcelona, Spain

– Josep Maria Haro

30


Questions?

Hélène Karcher, PhD

LASER Analytica

[email protected]

mailto:[email protected]



adding patient heterogeneity into phase iii trials_getreal_helenekarcher

Documents