Controlling the Risk of False Positive Clinical Trials
Statistical hypothesis testing is much like gambling. If, with gambling once, your chance of a prize is 5%, then, with gambling 20 times, this chance will be close to 40%. The same is true with statistical testing of clinical trials. If, with one statistical test, your chance of a significant result is 5%, then after 20 tests, it will increase to 40%. This result is, however, not based on a true treatment effect, but, rather, on the play of chance. In current clinical trials, instead of a single efficacy-variable of one treatment, multiple efficacy-variables of more than one treatment are increasingly assessed. For example, in 16 randomized controlled trials with positive results, published in the British Medical Journal (BMJ) in 2004 (Table 8.1), the numbers of primary efficacy-variables varied from 4 to 13. This phenomenon introduces the statistical problem of multiple comparisons and multiple testing, which increases the risk of false positive results, otherwise called type I errors. There is no consensus within the statistical community on how to cope with this problem. Also, the issue has not been studied thoroughly for every type of variable. Clinical trials rarely adjust their data for multiple comparisons. For example, none of the underneath BMJ papers did. The current chapter briefly summarizes the main methods for control in order to further emphasize the importance of this issue, and it gives examples.
KeywordsBritish Medical Journal Less Significant Difference Honestly Significant Difference Statistical Hypothesis Testing Less Significant Difference Test
- Pocock SJ (1988) Clinical trials. A practical approach. Wiley, New YorkGoogle Scholar