(41/55) = 0.75, while the proportion for all drugs is 754=591 958 = 0.0013: PRR = 586, chi-square = 22 736, p < 0.00001.
It should be realised that all of this process should be used for the purpose of signal detection, and even more importantly, for prioritisation of the detected signals to help decide which ones require most urgent further investigation. The basic data are still subject to biases; they are at very best observational data and to use a high value of a PRR or POR as the sole convincing evidence of causation is unwarranted. They raise a serious question that merits further study. At the same time, it should be remembered that where the reports are of suspected ADRs, then the reporter suspected a causal relationship and the fear that raised PRRs or PORs will generate too many false positive signals is probably also unjustified.
Was this article helpful?