Uncomplicated depression: new evidence for the validity of extending the bereavement exclusion to other stressors


Gordon Parker lucidly explains why the DSM-5's decision to eliminate the bereavement exclusion makes no clinical sense [1]. He suggests that, based on what we know about grief and depression, the exclusion should have been extended to other stressors instead. Newly emerging epidemiologic evidence shows that he is right.

It has been recognized since antiquity that normal grief and clinical depression share many symptoms and can be easily confused, yet that grief sometimes leads to genuine clinical depression [2]. Paula Clayton's early studies of normal grief clarified the situation by distinguishing those depressive symptoms common in normal grief (e.g., sadness, insomnia, decreased appetite, loss of interest in usual activities, difficulty concentrating) from more pathosuggestive symptoms characteristic of depressive disorder even during bereavement (e.g., suicidal ideation, psychotic ideation, psychomotor retardation, sense of worthlessness, severe impairment, lengthy duration) [3]. These findings were formalized in the DSM's bereavement exclusion (BE) to major depressive disorder (MDD). The BE classifies a depressive episode during bereavement as normal (or ‘uncomplicated’) if it involves only common general-distress symptoms. It conservatively classifies any other (‘complicated’) episode as MDD if it has even one of the six pathosuggestive features. Note that the BE was not a return to the discredited reactive/endogenous distinction, but rather a distinction among reactive depressions.

Clayton et al. [3] suggested extending the BE's rules to other stressors, but this was never seriously considered. The BE's rules for bereavement have recently been demonstrated to yield a valid distinction, as judged by concurrent and predictive pathology validators [4, 5], except that the BE's 2-month duration limitation for uncomplicated episodes has been shown to be too short for optimal validity [6].

Nonetheless, the DSM-5 is eliminating the BE as a formal MDD criterion. This means that any depressive episode during bereavement, whatever the symptoms, will now fall under major depression criteria if it lasts just 2 weeks after a loss. As Parker explains, an advisory note will be included, but one that lacks any diagnostic criteria, making it impossible to scientifically study this group and leaving the judgment of normal versus disordered reactions to each clinician.

Parker identifies a central argument that he sees as having swayed the DSM-5 debate, and his discussion reveals some remarkable semantic chicanery. Parker first observes that ‘DSM-IV criteria for major depressive episode are commonly viewed as excluding grief states; however, this is not exactly the case’. In fact, as we saw, the BE's rules are very narrowly drawn and exclude from MDD not grief states in general but only a minority of bereavement-related depressive episodes. Yet, those favoring the BE's elimination repeatedly mischaracterized it as excluding all bereavement-related episodes. Parker then explains that, once the evidence showed that uncomplicated bereavement-related and uncomplicated other stressor-related episodes are quite similar [7, 8], the DSM-5 Mood Disorders Work Group was influenced by Kenneth Kendler's ‘consistency’ argument: As bereavement-related depression is not special, to maintain consistency, either the BE must be eliminated from DSM-5 or extended to other stressors. However, the idea of extending the BE was rejected out of hand, partly because, parallel to the misstatement of the bereavement exclusion, the idea of extending the BE was mischaracterized as excluding all stress-related depression from MDD, an absurd idea. Yet, this is precisely how the ‘extend’ alternative was described by Kendler in his official statement on the DSM-5 website explaining the elimination of the BE: ‘Either the grief exclusion criterion needs to be eliminated or extended so that no depression that arises in the setting of adversity would be diagnosable’ [9]. In fact, of course, the BE's extension would exclude only the minority of adversity-triggered depressive episodes that satisfy the BE's demanding rules for being an uncomplicated, benign reaction [10].

Semantic confusions aside, we now know that the right empirical answer to the ‘eliminate or extend’ dilemma was to extend the exclusion. This is the answer provided by two recent studies evaluating the validity of extending the BE's exclusion rules to other stressors.

The first study used National Comorbidity Survey data to examine whether the uncomplicated depression construct continues to have high concurrent validity when applied to non-bereavement episodes [11]. After eliminating all bereavement-related episodes, MDD was divided into three categories: uncomplicated stress-triggered MDD, all other (‘complicated’) stress-triggered MDD, and endogenous (i.e., untriggered)/psychotic MDD. Validity was measured by eight pathology validators used to compare the pathology levels of these three groups' depressive episodes: number of symptoms, suicide attempt, melancholic depression, duration, interference with life a lot, hospitalized for depression, saw a mental health professional or general practitioner for depression, and took medication for depression.

Results were similar to previous validity results for bereavement-related episodes. For all eight validators, uncomplicated non-bereavement cases were significantly and substantially less pathological than either complicated triggered or endogenous/psychotic cases, which were generally similar to each other. For example, uncomplicated, complicated triggered, and endogenous/psychotic validator percentages were as follows: suicide attempt 0%, 8.6%, 17.2%; depression interfered with life a lot 7.6%, 43.5%, 49.3%; and melancholic symptoms 9.6%, 28.4%, and 33.2%, respectively [11]. Uncomplicated depression thus has strong concurrent validity.

Additional analyses systematically purged the analysis of semantic biases (e.g., the validator ‘suicide attempt’ was reanalyzed after removing the biasing criterion ‘no suicidal ideation’ from the uncomplicated criteria), and demonstrated that the uncomplicated/complicated differences were not due to such biases but reflected real syndromal differences that were stable under adjustments to the criteria. Furthermore, it was demonstrated that uncomplicated depression is not simply the same as mild depression (standardly defined by number of symptoms). Regression analysis revealed that uncomplicated status offers incremental validity over severity alone in predicting validators. These additional analyses support the construct validity of uncomplicated depression, which constituted about 15% of MDD in this sample.

A second study, using two-wave longitudinal Epidemiologic Catchment Area Study data, examined the crucial predictive validity of an extended exclusion [12]. Risk of depression recurrence, generally considered characteristic of depressive pathology and important to clinical decision-making, was used as the predictive validator. Wave 1 respondents were divided into three groups according to lifetime MDD history: no MDD history, uncomplicated MDD, all other MDD. Each group's rate of depression during the one-year follow-up period between the first and second interviews was calculated.

Results were that the recurrence rate for uncomplicated cases was not significantly higher than the occurrence rate in the no-MDD-history group (3.4% vs. 1.7%), but both were much lower than the rate for other MDD (14.6%). In contrast, the recurrence rates for mild depression (9.6%) and non-melancholic depression (10.6%) were significantly higher than no-MDD-history group (1.7%) [12].

In sum, the extension of the BE rules to other stressors yields a distinction that has strong concurrent and predictive validity for identifying a benign subtype of depressive reactions. Treatment protocols, research sample selection, and estimates of unmet need should be adjusted accordingly. The BE's rules work and the uncomplicated/complicated depression distinction should have been extended in the DSM-5 rather than eliminated. By ignoring the emerging evidence, the DSM-5 Task Force made the wrong judgment and opened a Pandora's box of new problems that will have to be addressed in the future.