To the Editor In a recent issue of JAMA Dermatology, Anderson et al1 aimed to compare skin cancer diagnostic accuracy of physician assistants (PAs) with that of dermatologists (MDs) and used the number needed to biopsy as the main outcome measure.
Number needed to biopsy is a poor index to compare diagnostic accuracy between groups because it is contingent on disease prevalence and the examiner’s sensitivity threshold.2 Controlling for prevalence requires randomizing patients by skin cancer risk factors, which, as seen in Tables 1 and 2 of the report, was not performed even for the strongest risk factor—history of melanoma.1 Furthermore, biases that may have existed between clinicians and patients include (1) patient crossover, (2) factors influencing scheduling that would influence biopsy decisions, (3) differences in visit types (initial vs follow-up) and previous screening activities, (4) dermoscopy training, and (5) clinical experience. Controlling for biases would likely blurr the differences in number needed to biopsy and rate of melanoma in situ detection between the PAs and MDs.