Biostatistical expertise is required when the clinical study is designed, and must continue to be available as data are collected, analysed and prepared for the final report. The misuse of statistical evaluation and abuse of statistical tests is common in all clinical research, particularly in relation to the “t test”. Statistical analysis should be made applicable to the nature of the data acquired and the clinical situation in the study. It should always be kept in mind that statistical significance is different from clinical significance. It should not always involve a simple “t test”. Every attempt should be made to avoid type II statistical errors and achieve at least an 80% statistical power although 90% would be ideal. Confidence limits should always be stated with the significance values. The value of small group studies can be enhanced by meta-analysis. Failure to complete protocol treatment should be recorded and analysed.
Statistical considerations will govern the number of patients needed to obtain a significant result from the study. The number of patients needed depends on the anticipated difference in the result between the treatment groups in the study. The plan for the statistical analysis to be used at the conclusion of the study must be determined in advance and specified within the protocol. When results are finally analysed, they should be presented in a form that facilitates clinical interpretation.