Mood Improvement From Transcranial Magnetic Stimulation

To the Editor: I am writing regarding the study by Mark S. George, M.D., and colleagues (1) that reported mood improvement from transcranial magnetic stimulation. I believe that the statistical analyses performed were incorrect and that the proper analyses would show nonsignificant effects.

The data presented by the authors show that patients who received active treatment began with a Hamilton Depression Rating Scale score of 30; this fell to 23 after 2 weeks of active medication and increased to 26 after a subsequent 2 weeks with a placebo (sham treatment). The other patients (who were given the sham treatment first) began with a baseline Hamilton rating scale score of 26; this increased to 30 after 2 weeks on the sham treatment and fell to 27 after the subsequent 2 weeks on active treatment. Note that the group that received the sham treatment first was actually doing somewhat worse after active treatment than it had been at the beginning of the study.

The main problem with the analysis is that the authors used “change in Hamilton score” (treatment minus baseline) as their major outcome variable, but there was no washout period, so the value they used as the baseline for the second period of treatment was the rating at the end of the first phase of treatment. This, in a sense, doubles the importance of the rating at the end of the first period and, according to some authors (2), is clearly improper; Hills and Armitage (2) state that “there can be no second baseline without a washout period” (p. 16).

Also, the standard approach in analyzing a crossover study is to look first at period (not order) effects and period-by-treatment interactions. If there is a period effect or a period-by-treatment interaction, the second phase of the crossover study should not be analyzed. Dr. George and colleagues did not perform these initial analyses, and it seems quite possible that there was a significant period-by-treatment interaction because almost all of the evidence that magnetic stimulation was helpful occurred in the first period of treatment.

There may be other reasons for further evaluating trans­cranial magnetic stimulation. However, I believe that a more traditional analysis of this data would not yield a statistically significant evidence of benefit, so this should not be considered a positive study.