Reporting statistics in clinical trials published in Indian journals: A survey

Objectives : Clinical trials are having very important place in the hierarchy of evidence based medicine. It has been observed that current methods of use and reporting of statistics of clinical trials are responsible for errors in the interpretation of results. So we decided to evaluate clinical trials published in three Indian journals of 2007 and 2008 to analyse statistical issues which may affect the interpretation of results.
Settings and Design: Retrospective observational study
Materials and Methods : We analyzed all the clinical trials (46) published in Indian pediatrics, Indian journal of pharmacology, journal of postgraduate medicine and of 2007-2008.
Results : We found that median number of end points reported in clinical trials was 4 and median number of end points which were used for testing of significance was also 4. (21) 45% of trials reported repeated measurement. 18 (39%) trials had three or more than three treatment groups. median number of test of significance was 15. post hoc subgroup analysis was done in 19% (9) of trials. P value was the sole criteria for interpretation of results in most of the trials, confidence interval was calculated in 11 (23%) trials. Baseline comparison between the study groups was done in 41 (89%) trials. in all cases comparison was done by statistical tests. Exact sample size was calculated in 18 (39%) trials.
Conclusion: There are great chances of having error during the interpretation of results of these trials was because of multiple treatment groups, repeated measurements of endpoints, multiple tests of significance, over reliability on P value and less use of confidence interval. Statistical methods available for reducing chances of errors should be used and result should be interpreted accordingly.