Down with odds ratios!

Abstract

As described in an earlier EBM note (1) and in our glossary, this journal reports the results of individual randomised trials in terms of relative risk reductions (RRRs), calculated by dividing the absolute difference in event rates between the control (control event rate [CER]) and experimental (experimental event rate [EER]) patients by the event rate for the controls: (CER – EER)/CER = RRR. From these same values, we also report the number of patients that would need to be treated (NNT) to prevent 1 additional event—1/(CER–EER)—or by its alternative calculation–1/(RRR × CER). Thus, in the example shown in Table 1, the RRR is 89% and the NNT is 4 (2).