[Skip to Content]
Access to paid content on this site is currently suspended due to excessive activity being detected from your IP address Please contact the publisher to request reinstatement.
[Skip to Content Landing]
Comment & Response
December 2015

Error in Calculating Main Outcome in Gamma Ventral Capsulotomy for Obsessive-Compulsive Disorder Randomized Clinical Trial

Author Affiliations
  • 1Department of Psychiatry and Psychotherapy, University of Cologne Medical School, Cologne, Germany
JAMA Psychiatry. 2015;72(12):1257-1258. doi:10.1001/jamapsychiatry.2015.0667

To the Editor In many ways, the Lopes and coauthors randomized clinical trial1 on gamma ventral capsulotomy (GVC) in patients with treatment-refractory obsessive-compulsive disorder stands out. Ethics approval and informed consent procedures are exemplary for a contemporary study in psychosurgery. Patients were followed up for an exceptionally long period. The article is among the few in psychiatry documenting blinding of raters and patients, although such data are not only important, as presented, at follow-up but particularly immediately after surgery (as apparently ascertained).2 Further, the authors make their study transparent in providing a wealth of outcome data.

It appears from these data that a small error occurred in calculating the main outcome: Table 1 and the P value for comparing the decrease in Yale-Brown Obsessive-Compulsive Scale (Y-BOCS) scores between treatment and control indicate the baseline Y-BOCS score of the third patient in the treatment group (ATa3) was not 36 but 30. With a follow-up score of 22, the decrease was 26.7%, rather than 38.9%, as stated in Table 2. Apparently, however, Lopes et al concluded from the latter figure that patient 3 was a responder because the decrease exceeded 35%. Without this patient, the number of responders under GVC stands at 2 of 8, with no responder among controls (Fisher exact test: P = .47). The absolute risk reduction, therefore, is reduced from statistically significant 37.5% (95% CI, 4% to 71%) to 25% (95% CI, −5% to 55%; nonsignificant), and number needed to treat increases to 4 (95% CI, 1.7 to ∞). Given group sizes of 8, Y-BOCS score change, when compared as a continuous variable, should be analyzed using the exact Mann-Whitney U test and not its asymptotic version.3 Thus, P slightly rises from .046 to .050 (or to .0499, to be exact). Similarly, when recalculated, the P value for change in Dimensional Y-BOCS score increased from 0.01 to 0.02. These differences are not entirely moot in a study with a host of outcomes and not adjusted for multiple tests. As a consequence, corrected figures for both dichotomous and continuous outcomes (calculated with Open Epi4 and SPSS version 22 [IBM]) suggest that chance cannot be ruled out as a factor in explaining the results.