Transcript Statistics in Drug Development

Statistics in Drug Development
Mark Rothmann, Ph. D.*
Division of Biometrics I
Food and Drug Administration
* The views expressed here are those of the author and not
necessarily those of the FDA.
1
Various Indications for Drugs
For example: Headache/Pain medicine, psychiatric drugs,
AIDS treatments, Cancer drugs, etc.
Details about the design and goal of a study depends on the
indication of the drug (how the drug will be used).
2
Goal of New Drug Development
Develop a safe and effective drug.
3
Phases of Drug Development
Pre-clinical - Animal testing
Phase 1 - Dose ranging (toxicity)
Phase 2 - Use of the drug in a small number of studies
(efficacy screening)
Phase 3 - Comparative study with a placebo or an active drug
(usually the standard therapy)
Phase 4 - Post-marketing studies
4
Comparative Phase 3 Trials
Aspects of a quality comparative study
- Randomization (patients are randomly divided into groups)
- Stratification
- Double-Blind (eliminate a placebo-effect and diagnosis bias)
- Control
5
Endpoints (Variables of Interest)
Examples of Oncology Endpoints:
Survival Time
Tumor response (binary or ordinal variable)
Time to tumor progression
Quality of Life
6
Hypotheses
H0: experimental drug and the standard drug (or placebo)
have the same effectiveness
H1: experimental drug and the standard drug (or placebo)
have different effectiveness
Alternative hypotheses are two-sided. Hypotheses are formally
for those patients in the study.
One or more endpoints may tested.
7
Potential Errors
Type I error: Rejecting H0 when H0 is true
(false positive rate)
Type II error: Failing to reject H0 when H1 is true
( or for the drug company the type II error of
interest is failing to conclude the drug is
effective when it is effective)
An overall probability of a type I error is maintained at 0.05
for the primary efficacy analysis. If more than one endpoint is
involved in the primary efficacy analysis, individual type I errors
are adjusted for the total number of comparisons.
8
Sample Size Determination
Aspects considered for the sample size calculation:
- A primary method of analysis is selected.
- Desired Type II error probability at a clinically meaningful
effectiveness alternative.
- Accrual Period
- Follow-up time
- Fraction of dropouts
9
Analysis Populations
Intent-to-treat Population: All patients as randomized
Evaluable Population: All patients who received study
therapy that have measurements for the primary efficacy
endpoint and comply with the protocol.
10
Statistical Conclusions
Patients in the study are volunteers - not randomly
selected from some group. Formally, any conclusions
of statistical significance is good only for that set of
patients in the study.
11
Other Issues
- Formal Definitions of Endpoints
- Missing Data (common in quality of life measurements)
- Crossover to other therapies
- Censoring
- Robustness with respect to the method of analysis chosen
- Interim Efficacy Analyses
12