Sample Size Planning of Clinical Trials, Introduction

1
Sample Size Planning of Clinical Trials,
Introduction

• Dr. Gu Ming Gao
• Department of Statistics, CUHK

2
Clinical trials are important in

• Evidence-based medicine
• Compare different treatments, quality of care or
  medicine, study into Chinese medicine
• Development, improvement of clinical practice
  guidelines
• Finding new factors and relationships which
  effect our health, drug development

3
Two types of clinical trials

• Trials involve immediate responses
• Trials involve time-to-event responses
• The statistical procedures for handling those two
  types of data are very different
• Survival analysis
• Cox regression model

4
Trials with immediate responses, comparing two
samples

• Control vs. treatment
• Assumptions on the distribution
• Statistic T-test statistic
• Type-I error and power
• Formula for sample size calculations under normal
  distribution assumptions

5
Trials with time-to-event responses, comparing
two samples

• Normal assumption on the distribution of the
  life-time is always violated
• We usually only have vague ideas about the
  underlying distribution functions
• No exact simple size calculation formula
• Uses of Monte-Carlo simulations

6
Censoring

• Censoring caused by patient withdraw
• Censoring by other illness, competing risks
• Censoring by design of the trial
• Censoring caused by pre-scheduled date of
  termination of the trial

7
Noncompliance

• Noncompliance is also called crossovers
• Drop-in rates
• Drop-out rates

8
Log-rank statistics

• An intuitive view of Log-rank statistics boys
  computing with girls
• General payment functions, Wilcoxon statistic
• Rank statistics are more robust than the t-test,
  less assumptions on the distribution functions
• The Log-rank statistics are very powerful

9
Sequential monitoring of clinical trials

• Trials are monitored for safety reasons
• Group sequential trials, planning and design
• Complications caused what is my type-I error and
  what is my power?

10
Different stopping rules in clinical trials

• Pocock boundary
• OBrein-Fleming boundary
• Haybittle boundary
• Use function approach by Lan DeMetes
• Complications caused different stopping rules

11
The DSSP program

• Use DSSP program to find the power of the test
  for a given sample size
• Use the DSSP program for sample size.
• The program is window based and easy to use
• Intro to the program will be given next week
• Available at http//www.sta.cuhk.edu.hk/minggao/ct
  rials.asp