Conclusions Inconsistent With Results With Citalopram

To the Editor: In their article, Eric J. Lenze, M.D., et al. (1), after noting three limitations of their trial (its small size, problems with random assignment, and diagnostic heterogeneity in their study groups) reported positive summary statements in their conclusions and elsewhere. For example, they wrote, “Notwithstanding these limitations, this study suggests that, as in younger people, SSRIs [selective serotonin reuptake inhibitors] are efficacious and well tolerated in the treatment of anxiety disorders in elderly persons” (p. 149).

The authors’ statements regarding the efficacy/tolerability of citalopram in geriatric anxiety disorders may be true, but their published repetitive positive conclusions are not consistent with their evidence. For example, the Results section demonstrated that 11 members of a treatment group of 17 subjects experienced less anxiety than a placebo group. However, 88% of the treatment group were women, and 47% received various doses of lorazepam while they received various doses of citalopram (according to their clinical response). This treatment group was compared to an unmatched placebo group, 65% of whom were men and (only) 24% of whom were administered various doses of lorazepam.

The treatment group experienced more adverse side effects (intolerable sedation, nausea, and fatigue) and had a higher dropout rate than the members of the placebo group, who experienced fewer side effects and had a lower dropout rate. Furthermore, despite random assignment, the placebo group—when evaluated for mean scores for anxiety and depression—was more symptomatic than the treatment group before the initiation of any “treatment,” and conversely, the “treatment” group was less symptomatic at baseline on both measures, skewing the statistical endpoint contrasts of “treatment effect.”

Design difficulties and the questionable interpretation of results were distorted by graphical analysis. In Table 1, the mean baseline rating of anxiety for the placebo group is 23.1, whereas the corresponding number for the citalopram group is 21.4. However, despite the apparent use of data from Table 1 as the basis for Figure 1, the graphical analysis shows that the subjects taking citalopram began with a higher anxiety score than the placebo group, creating the impression that the citalopram “treatment” group had a more pronounced decrease in anxiety after treatment than it, in fact, did.

The evidence shows that the citalopram group did not tolerate its treatment as well as the placebo group, nor did the authors establish the efficacy of treatment since the groups were not comparable. Unmatched intergroup mean baseline scores for the symptoms of anxiety or depression skewed the statistical analysis, and inaccurate graphical representation of the results distorted the findings. The conclusions of this research effort, funded by Forest Pharmaceuticals and three grants from the National Institute of Mental Health, are misleading and inconsistent with the authors’ data.