To the Editor: In his letter, Dr. Grucza underlines the importance of appreciating that the relationship of anxiety symptoms and panic disorder to suicide attempts may differ between community samples and inpatient groups. We agree, but with an important caveat. Such comparisons should take into account other clinical factors thought to convey risk for suicidal behavior. One of the major limitations of the studies reviewed in the article by Johnson et al. (1990) was the fact that axis II comorbidity and the aggression/impulsivity dimension were not considered in the evaluation of suicidal status. Both cluster B personality disorder and aggression/impulsivity have been found to be highly associated with suicidal acts +(1). A high rate of cluster B personality disorders or aggressive-impulsive traits among the subjects with comorbid depression and panic could bias the studies. We stressed this limitation in the work of Johnson et al. +(2). Thus, the assessment of axis I and II disorders and their comorbidity, as well as aggression/impulsivity traits, should be considered when evaluating suicidal risk.

Dr. Grucza inquires about the details of the multivariate and logistic regressions that were performed in order to establish the direction of the association between the anxiety variables and the status of being a suicide attempter. In the interest of conciseness, we did not publish the quantitative details of the multivariate and logistic regressions. However, all of the anxiety variables reported in Table 2 and Table 3 were included in the analyses.

Dr. Grucza inquires about the exclusion of the Beck Depression Inventory score from the multivariate analysis. We did not include any self-report measures in the multivariate analysis in order to maintain uniformity of the assessment perspective. Also, since all of the subjects were depressed, we had already controlled for the relationship of depression to the status of attempter, and the hypothesis tested addressed the relationship of anxiety to suicide attempts, not to subjective depression.

Suicidal behavior is determined by multiple factors, and it is critical to assess all of the major factors in individual studies to avoid gathering misleading results for individual putative correlates or risk factors. Finally, we agree that longitudinal, prospective studies are needed to test the potential predictors of suicidal behavior, including anxiety disorders and symptoms.