During the dozen years or so that I have served as a biostatistical consultant to the editors of JAMA, the most frequent comment I have made in my reviews is the simple, straightforward question, "Can the authors provide any rationale or statistical power considerations underlying their choice of sample size?" From this experience, I infer that lack of adequate justification for sample size is a major statistical limitation that pertains to many of the manuscripts submitted to JAMA.
I greet with great enthusiasm any attempt to disseminate more widely the message of proper statistical planning of a research endeavor, particularly in regard to choice of study size. Hence, I am delighted to see the article by Arkin and Wachtel entitled "How Many Patients Are Necessary to Assess Test Performance?" appear in this issue of The Journal.1
The article breaks no new statistical ground. Essentially, the authors apply standard textbook
Colton T. The 'Power' of Sound Statistics. JAMA. 1990;263(2):281. doi:10.1001/jama.1990.03440020115045