To the Editor: In their study, Giovanni P.A. Placidi, M.D., et al. (1) made an impressive effort to deconvolute the complex relationships among suicide attempts, panic disorder, and anxiety-related variables in the context of major depression. They found no greater risk of suicide attempts in depressed patients with comorbid panic disorder and suggested that anxiety may actually play a role in preventing suicide attempts. They effectively articulated the serious implications this finding might have for the treatment of comorbid anxiety and depression.

Several issues need to be considered before conclusions can be drawn regarding the exact relationship between comorbid anxiety and suicide attempts. Community-based studies have shown a greater level of past suicide attempts in subjects with comorbid disorders (2, 3). The apparent discordance among studies might easily be accounted for by sample differences, but the question remains: which sample is more representative of the patients in the typical psychiatric care setting? In addition, numerous investigations, including longitudinal studies (see reference 4 for a review), have suggested that panic disorder is associated with a greater risk of suicide in depressed patients.

The authors stated that, in a logistic regression model, "panic disorder and anxiety measures together, with adjustment for aggression, were significant predictors of attempter status" (p. 1615), with greater anxiety associated with a lower likelihood of suicide attempts. The details of the multivariate analysis used to reach those conclusions were not made clear. Not all of the unadjusted Hamilton Anxiety Rating Scale and Brief Psychiatric Rating Scale anxiety scores were significantly associated with attempter status, and it is unclear which anxiety scores were included in the model. This is particularly important given the likely distinct etiologies of cognitive and somatic anxiety. Moreover, it would be advantageous to know the quantitative relationships between the anxiety scores and attempter status and how those relationships might translate into clinically observable phenomena. It is particularly surprising that the Beck Depression Inventory score was not considered in the multivariate analyses given that the differences between the attempters and nonattempters in the Beck Depression Inventory scores were more significant than differences in any of the Hamilton anxiety scale scores.

It is intrinsically difficult to draw conclusions about suicidality from cross-sectional studies; however, this study demonstrated the value of measuring quantitative anxiety-related traits in assessing suicidal behavior. It is hoped that the investigators will consider such variables in the design of future longitudinal studies.