Limited utility of number needed to treat and the polarity index for bipolar disorder to characterize treatment response.

The medical community increasingly supports the use of simplifying constructs or ratios to facilitate incorporation of evidence-based medicine into clinical practice such as number needed to treat (NNT) and polarity index (PI). Clinicians and teachers find them to be an appealing, easy-to remember integer that can be readily translated into clinical practice. However, serious questions have been raised with respect to the validity, reliability and value of these descriptors of response. This commentary identifies some of the specific limitations of the NNT and PI constructs when applied to treatments of bipolar disorder.