Statistical Aspects of Clinical Trials

This seminar is designed specifically to enable nonstatisticians to understand statistical aspects of clinical trials. As well as learning how to communicate effectively with statisticians using statistical language, participants will obtain an understanding of the main concepts of statistical methodology sections in study protocols, reports and publications and will have the opportunity to explore key statistical aspects of the regulatory framework (ICH, EMEA, FDA) for clinical trials.

This seminar is aimed at non-statisticians working in the pharmaceutical industry who wish to acquire a better understanding of statistical aspects of clinical research. It is of particular interest to:

• Medical Advisors
• Clinical Research Associates
• Clinical Data Managers
• Medical Writers
• Quality Assurance personnel

On completion of this course you will appreciate the role of statistics and the statistician in clinical drug development.

You will be familiar with the notions of sampling, bias, randomization and blinding, standard errors, confidence intervals, correlation and regression, the concept of statistical power and its importance in sample size planning, and the distinction between clinical and statistical significance.

You will have a clear understanding of the application of statistical methodology in clinical trials, making it easier for you to understand protocols and discuss statistical aspects of study design with medical and clinical research professionals.

Statistical Aspects of Trial Design

• Basic ideas and issues in statistical design
• Confirmatory vs. exploratory analyses
• Concepts of superiority, equivalence and non-inferiority trials
• Distinction between ‘between patient’ designs and ‘within patient designs’
• Need to maximize the ‘signal’ whilst controlling the ‘noise’
• Statistical analysis process from the design stage to statistical analysis
• Planning, blind review and statistical analysis and reporting

Sampling and Standard Errors

• Interrelation between the sample and the target population
• Purpose of inferential statistics
• Measures of location and of variability
• Normal probability distribution
• Standard error as a measure of the ‘reliability’

Confidence Intervals and Hypothesis Tests

• Confidence interval and its appropriate interpretation
• Framework for hypothesis testing, p-value and its calculation
• Generalization to practical settings
• Definition of null distribution, testing of statistic and significance level
• One-sided and two-sided tests, type I and type II errors
• Introduction of two-sample t-test and paired t-test as examples

Intent-to-Treat and Analysis Sets

• Principle of Intention-to-treat (ITT)
• Relation to the choice of analysis set
• ‘Full Analysis Set’ and ‘Per Protocol Set’ according to ICH
• Practical issues, such as handling of missing values

Adjusted Analyses

• Two-way analysis of variance (ANOVA
• Investigation of treatment mean effects, center effects
• Treatment by center interactions
• Generalizability to more complex situations

Correlation, Regression, ANCOVA

• Correlation
• Linear regression as a technique for studying dependence
• Multiple linear regression
• Analysis of Covariance as a means of comparing groups
• Use and misuse of baseline testing

Analysis of Binary, Categorical and Ordinal Data

• Chi-squared test for binary data
• Generalization to categorical data
• Use of Mantel-Haenszel chi-squared test for ordinal data
• Odds ratio, relative risk, and number needed to treat
• Role of Fisher’s Exact Test

Power, Sample Size, Statistical and Clinical Significance

• Statistical power in relation to calculation of sample siz
• Role of the standard deviation in this calculation
• Re-evaluation of sample size as trial data accumulates
• Link between p-values and confidence intervals
• Limitations of p-values
• Distinction between clinical and statistical significance

Non-Parametric Methods

• Assumptions underlying ‘parametric tests’
• Transformations to recover normality
• Generalization of the t-test to deal with non-constant variance
• Non-parametric between and within-patient tests
• Wilcoxon Rank Sum and the Wilcoxon
• Signed Rank tests
• Advantages / disadvantages of the non-parametric framework

Practical Issues in Multiple Testing

• Problems associated with undertaking large numbers of statistical tests
• Inflation of the type I error rate
• Areas in which such multiple testing arises and offers solutions

Testing for Equivalence or Non-Inferiority

• Testing for superiority
• Testing for equivalence or non-inferiority
• Confidence interval approach to the equivalence / non-inferiority setting
• Choice of equivalence margins
• Problems associated with sensitivity to drug effects

Meta-Analysis

• Purpose and framework of meta-analysis
• General method for combining results in a meta-analysis
• Role of meta-analyses in the regulatory setting

Methods for the Analysis of Survival Type Data

• Difference of time-to-event data
• Kaplan-Meier estimates
• Log-rank test
• Gehan-Wilcoxon test
• Analyses adjusted for the influence of prognostic variables
• Assumption of independent censoring

Crossover Trials

• Problems associated with carryover effects in 2 x 2 crossover trials
• Advice on when to use and when not to use crossover trials
• Methods of analysis

Jamie Langley, RN MSN
Global Head of Parexel Academy

Phone: +1 339 223 0653

Email: Send email