Statistics

6.1. Role of a Biostatistician

Involvement of a appropriately qualified and experienced statistician is necessary in the planning stage as well as throughout the Study.  The Bio-statistician’s should make a statistical model to help the Sponsor, CRO and / or the Investigator in writing the Protocol.  The number of Subjects to be included in the study is determined in relation to the statistical model on which the Protocol is based.

6.2. Study Design:

The scientific integrity of a Clinical Study and the credibility of its report depends on the design of the Study.  In comparative studies the Protocol should describe:

1.      an “a priori” rationale for the target difference between treatments that the Study is being designed to detect, and the power to detect that difference, taking into account clinical and scientific information and professional judgment on the clinical significance of statistical differences.

2.      measures taken to avoid bias, particularly methods of Randomisation.

6.2.1.  Randomisation and blinding:

The key idea of a clinical trial is to compare groups of patients who differ only with respect to their treatment.  If the groups differ in some other way then the comparison of treatment gets biased.  Randomisation, as one of the fundamental principles of experimental design, it deals with the possible bias at the treatment allocation. It ensures that the allocation of treatment to human subjects is independent of their characteristics. Another important benefit of Randomisation is that statistical methods of analysis are based on what we expect to happen in random samples from populations with specified characteristics. The Protocol must state the method used for Randomisation.

The Study should use the maximum degree of blindness that is possible. Study subjects, investigator or any other party concerned with the study may observe and respond by knowledge of which treatment was given.  To avoid such bias it is often desired that the patient or any other person involved with the study does not know which treatment was given. Where a sealed code for each individual treatment has been assigned in a blinded randomized study it should be kept both at the site of the investigation and with the sponsor. 

The Protocol must state the conditions under which the code is allowed to be broken and by whom.  The system of breaking the code should be such that it allows access to only one Subject’s treatment at a time.  The coding system for the Investigational Product(s) should include a mechanism that permits rapid identification of the products in case of a medical emergency, but does not permit undetectable breaks of the blinding.

6.3.  Statistical Analysis

The type(s) of Statistical Analyses to be used must be clearly identified and should form basis of the statistical model for the Study. Any subsequent deviation(s) should be described and justified in the Final Report.  The need and extent of an interim analysis must be specified in the Protocol.  The results of the statistical analyses should be presented in a manner that is likely to facilitate the interpretation of their clinical importance, e.g. by estimates of the magnitude of the treatment effect / difference and confidence intervals rather than sole reliance on significance testing.

Missing, unused and spurious data should be accounted for during the statistical analyses.  All such omissions must be documented to enable review.

back

Definitions, Pre-requiste for the study, Responsibilitiles, Record Keeping and Data Handline, Quality Assurance, Statistics, Special Concerns, Appendices