To the Editor: We reviewed with great interest the recent article by Kate L. Harkness, Ph.D., and Scott M. Monroe, Ph.D. (1), demonstrating a strong link between early childhood trauma and endogenous depression, in contrast to prior work linking early adversity with nonendogenous subtypes of depression (2). However, we are skeptical of these new findings in view of critical exclusions in their data collection, including a lack of consideration of atypical depression and overly narrow inclusion criteria.

Atypical depression, by definition, is characterized by mood reactivity, hypersomnia, and increased eating, in contrast to the lack of mood reactivity, insomnia, and loss of appetite seen in endogenous depression. Subjects with atypical depression thus provide a natural comparison group for linking particular vulnerability factors with endogenous depressive subtypes. Using this latter approach in a community-based sample of 653 individuals with major depression, we previously demonstrated a significant association between early childhood physical and/or sexual abuse and atypical but not endogenous symptoms of depression (3). Biological studies also point to a mechanistic link between atypical depression and severe trauma. More specifically, low baseline cortisol levels and hypersensitivity to low doses of dexamethasone have been found in both posttraumatic stress disorder (4) and atypical depression (5, 6), consistent with a hypothalamic-pituitary-adrenal (HPA) axis that is hyperregulated. In contrast, a large body of work has demonstrated hypercortisolemia and resistance to dexamethasone in endogenous depression, consistent with a stress response that is chronically overactivated and hyporegulated (7). This further suggests that severe trauma is more closely linked with atypical than endogenous depressive symptoms.

Given these links between atypicality and early trauma, the sampling procedure of Drs. Harkness and Monroe becomes problematic in that individuals with chronic depression and many with a "comorbid exclusionary diagnosis" were excluded from consideration in their study. Chronicity and comorbidity are hallmarks of atypical depression (8–10) and are themselves associated with early trauma; thus, elimination of subjects with these characteristics establishes a significant sampling bias.

Another issue relates to the cross-sectional nature of their study, in that many patients with recurrent depression have a fluctuation of neurovegetative symptoms over time; i.e., they may appear endogenously depressed at one point in time, while exhibiting atypical symptoms at other times (11). This is not a trivial point in that we have previously reported that depressed individuals with this fluctuating course have particularly high rates of severe childhood abuse (3). We speculate that several individuals labeled endogenous in the current study would have been designated as fluctuating had this been a consideration, further diluting the authors’ ability to find a putative link between early trauma and nonendogenous symptoms.

In sum, it may be that in a heterogeneous group of individuals with major depression, particularly high rates of trauma will be found in those with atypical symptoms, while in a more homogeneous group with nonchronic illness without certain comorbidities, a link with endogenous symptoms will emerge. This is an important consideration in designing and interpreting research linking early trauma with particular depressive subtypes.