To the Editor: The report by Tetsuya Sato, M.D., Ph.D., and colleagues (1) is a welcome addition to our understanding of symptom factors and clinical subtypes in mania. As they noted, their findings have substantial overlap with our own (2).

The two analyses are consistent in key respects. Neither found a factor denoting general severity of illness. Both found a depression factor that was preeminent in the purely manic subjects as well as in those with diagnoses of mixed manic episodes. Both found that sleep disturbance did not load with any of the classical or nonclassical symptom factors in mania. It appears that insomnia in mania is not simply a correlate of psychomotor activation. Both found that the DSM-III-R/DSM-IV criteria substantially undercount the number of manic patients with depressive syndromes who were identified by the multivariate analyses, emphasizing the need for new, data-derived criteria for mixed bipolar episodes.

The individual factors are in broad agreement for depressed mood, psychosis, and irritable aggression. The studies diverge on whether the depressive symptom factor is bimodally distributed, which we found it to be. The factor termed "mania" in the study by Dr. Sato et al. was represented in our study as two factors—hedonic activation and psychomotor acceleration—reflecting our inclusion of additional hedonic symptoms such as sexuality and humor. Likewise, Dr. Sato et al. found a new symptom factor, depressive inhibition, based on symptoms that we did not evaluate.

In the cluster analysis by Dr. Sato et al. (1), the largest subgroup, pure mania, appeared as a residual group, with no positive loadings for any identified factor—not even factor 5, mania (Table 2). The factor scores have extremely large standard deviations, denoting a wide overlap of scores among the identified clusters and calling into question the interpretation of Dr. Sato et al. that "depressive mood, irritable aggression, and psychosis…are unlikely to coexist" (p. 972). Moreover, although they found that depressed mood and depressive inhibition are independent symptom factors, their cluster analysis did not distinguish patients with these two characteristics. Thus, their cluster analysis does not positively support Kraepelin’s subclassification of mixed states.

In a report published after the article by Dr. Sato et al. was submitted, we described five subtypes of mania identified by grade-of-membership analysis (3). Type 1 is a nonpsychotic, relatively mild form of mania that corresponds to Kraepelin’s "hypomania" and to subgroup 1 in the current report. Type 2 is a severe form of classical mania, with high levels of psychomotor activity, irritability, and psychosis, which corresponds to Kraepelin’s "acute mania." Type 3 is a very delusional form of mania with relatively less severe classical manic symptoms that corresponds to Kraepelin’s delusional mania and perhaps to subgroup 3 in the report by Dr. Sato et al. Type 4 is a severe form of mania with high levels of dysphoric symptoms and the complete absence of grandiosity or euphoria that corresponds to Kraepelin’s anxious or depressive mania. Type 5 is an overall less severe form of dysphoric mania than type 4, with moderate degrees of depressive mood symptoms alternating or coexisting with grandiosity, humor, sexuality, and psychomotor acceleration. In the study by Dr. Sato et al., subgroup 4 seems to comprise patients similar to types 4 and 5 in our grade-of-membership analysis. Again, the similarities between the analysis by Dr. Sato et al. and our analysis are more impressive than any differences. Both groups agree that clinical care and research studies of manic patients may benefit from serious attention to the "rediscovered" heterogeneity of clinical subtypes.