Limitations of safety data derived from trials: statistical reminders when interpreting regulatory trials for reference and biosimilar products
| Test | Uses and interpretation | Example | |
|---|---|---|---|
| Bonferroni correction41 | When you perform a hypothesis test in statistics, a p value helps you determine the significance of your results. The more often you repeat a test in a study, the more likely it is to show a positive result by chance (a false positive). The simplest way to compensate for this is to limit comparisons to only the most clinically relevant and critical outcome. If multiple comparisons are needed, the measure of statistical confidence (traditionally taken to be p≤0.05) needs to be adjusted downwards. The Bonferroni correction is the simplest method for this. | Number of tests | Bonferroni's adjusted p value |
| 2 | 0.025 | ||
| 3 | 0.0167 | ||
| 4 | 0.0125 | ||
| 10 | 0.005 | ||
| The ‘rule of 3’42 | In statistical analysis, the ‘rule of 3’ states that if a certain event did not occur in a sample with n subjects, then you can approximate that the upper limit of 95% CI=maximum risk=3/n. When n is greater than 30, this is a good approximation to results from more sensitive tests. | For example, if a drug was studied in 900 patients, you could be 95% certain that any unreported toxicity should occur at a rate of <1 in 300 (<3/900). To exclude less frequent events requires larger trial sizes and emphasises the importance of pharmacovigilance reporting. | |