Key result
The authors argue that the General Health Questionnaire is a non-specific indicator of psychiatric need, not a valid measure of depression, and criticize the methodology of a previous review.
Answers are grounded in this paper and cite their sources.
This letter highlights methodological flaws in a prior study assessing depression's impact on mortality, specifically criticizing the use of the GHQ and inappropriate covariate adjustment.
The title of the article by Stewart et al.1, in the November 2003 issue, implies that depression was assessed. However, the authors used the 30-item General Health Questionnaire with a cut-point ⩾5 to identify those with “depressive symptoms,” who are elsewhere in the manuscript referred to as the “depressed”. The GHQ is intended to screen primary care patients for probable need for psychiatric treatment.2 It is non-specific indicator of a variety of anxiety and mood disorders. It is not a measure of depressive symptoms or depression. Second, previous publications by the authors used a cut-point of 10 on this same measure to identify “GHQ cases of depression”.3 This mixing of terminology and shifting of cut-points is troubling. There are similar problems with the literature review. Table 5 represents “cohort studies published before March 2003 that reported at least 30 fatal events.” While the number of events is important for covariate control, the measures used and the timing of their administration are as important. Four of the studies listed did not use recognized measures of depression (Stewart et al1 and Jenkinson et al., 1993, Carinci et al., 1997 and Denollet et al., 1996). One assessed patients up to three years before an MI (Berkman et al., 1992), one assessed cardiac rehabilitation patients (Denollet et al., 1996), and one assessed patients taking amiodarone (Irvine et al., 1991). Three studies adjusted for covariates that are highly related to depression (Type-D personality (Denollet et al., 1996), social support and fatigue (Irvine et al., 1999), and perceived global health (Stewart et al.1)) and thus not true confounders.4 By doing so these studies explained away depression's impact on outcomes. In this context, the statement that “in eight of the 10 studies listed, there was no statistically significant association between depression and mortality after adjustment for potential confounders” is misleading.
No takes yet. Share an insight, caveat, or question.
Nancy Frasure‐Smith (2004) conducted a letter in Depression. The authors argue that the General Health Questionnaire is a non-specific indicator of psychiatric need, not a valid measure of depression, and criticize the methodology of a previous review.
Synapse has enriched 10 related papers. Consider them for comparative context: