Jerome A. Yesavage

TL;DR: The study found an overall prevalence of tardive dyskinesia using strict diagnostic criteria in a large sample of chronic mental patients, and found the disorder was significantly less frequent in males than females.

Abstract: To the Editor: I read the article by Yesavage et al.1 with interest because I have a private pilot license. The efficacy of acetylcholinesterase inhibitors in healthy persons is of interest as is the performance of aircraft pilots. Correspondingly, the authors argue with the “face validity” of the task, which they believe exceeds that of so-called “mere laboratory measures of cognition.” The perceived high validity of the task or, more precisely, of this complex bundle of tasks does not imply similarly high qualities with regard to other test-theoretical properties. In many cases, there is an inverse relation between validity and face validity and repeat reliability of a measurement. It is necessary to isolate functions and to develop laboratory tests. Most of our current knowledge probably rests upon our ability to subdivide questions. Therefore, the introductory notes relating to reliability of measurements and face validity of tasks appear to be misleading. Both validity and reliability of the score in use have not been evaluated. There is clear indication of a low reliability in the data. Because measurements were not repeated under truly identical conditions, but only after a month of training, performance might well increase not only with donepezil but also with placebo or even without any further intervention. This is the main scientific reason for introducing a placebo group. Decreasing performance with placebo in the given setting, however, is difficult to understand except as the visible expression of a high measurement error and a low repeat reliability. Thus, the observed worsening with placebo probably represents nothing more than a random effect, and so does the improvement with donepezil, which was even substantially lower. Unfortunately, the authors preferred to ignore this important problem and to treat their results as if proven. On the contrary, in view of the questionable quality of the data, I would strongly recommend not drawing any conclusions except that more scientifically rigorous studies are needed.