To the Editor: Peter de Jonge, Ph.D., and colleagues (1) report that the measurement of somatic symptoms of depression following myocardial infarction is confounded by physical health and that somatic symptoms, but not cognitive symptoms, predict cardiac prognosis. Their results, however, do not appear to support these conclusions, primarily because the method used to delineate somatic and cognitive factors did not create sufficiently distinct constructs for comparison.
The authors utilized exploratory factor analysis to derive a factor structure in which five items (sadness, pessimism, dissatisfaction, social withdrawal, indecisiveness) load robustly on both the somatic/affective and cognitive/affective factors, artifactually inflating the correlation between the factors. The authors do not report factor correlations, but factor scores calculated using the authors’ paradigm correlated highly in samples of 285 myocardial infarction patients (r=0.79) (2) and 913 patients with unstable angina or myocardial infarction (r=0.78) (3). This high correlation and substantial item overlap, however, raises the question of what is behind the differential relationship of the factors with physical health indicators. Do only one or two items produce this difference? If so, which ones?
Further evidence that the report by Dr. de Jonge and colleagues on somatic/affective and cognitive/affective factors may not represent clearly differentiated constructs is found in the cardiac prognosis analyses. After controlling for physical health and other depression factors, only the relationship between the somatic/affective factor and cardiovascular death is significant. Compared to the univariate analysis, however, the inclusion of additional predictors (all of which correlate positively with the somatic/affective factor and cardiac prognosis) counterintuitively increases the hazard-rate of the somatic/affective factor dramatically (from 1.64 to 3.91), widens its confidence interval substantially, and transforms the cognitive/affective factor into a protective factor. The authors incorrectly interpret this as the cognitive/affective factor being “(over)corrected by the somatic/affective symptoms” (p. 141). Instead, this is a textbook example of multicollinearity, which occurs when highly correlated predictors, often measuring the same or similar constructs, are included in the same model. When this occurs, results are not interpretable (4). Thus, while their findings are intriguing, Dr. de Jonge and colleagues have not sufficiently untangled the somatic and cognitive aspects of depression in myocardial infarction patients adequately enough to make the case that their respective relationships to cardiac prognosis differ.
1.de Jonge P, Ormel J, van den Brink RH, van Melle JP, Spijkerman TA, Kuijper A, van Veldhuisen DJ, van den Berg MP, Honig A, Crijns HJ, Schene AH: Symptom dimensions of depression following myocardial infarction and their relationship with somatic health status and cardiovascular prognosis. Am J Psychiatry 2006; 163:138–1442.Bush DE, Ziegelstein RC, Tayback M, Richter D, Stevens S, Zahalsky H, Fauerbach JA: Even minimal symptoms of depression increase mortality risk after acute myocardial infarction. Am J Cardiol 2001; 88:337–3413.Grace SL, Abbey SE, Pinto R, Shnek ZM, Irvine J, Stewart DE: Longitudinal course of depressive symptomatology after a cardiac event: effects of gender and cardiac rehabilitation. Psychosom Med 2005; 67:52–584.Cohen J, Cohen P, West SG, Aiken LS: Applied Multiple Regression/Correlation Analysis for the Behavioral Sciences (3rd ed.). Hillsdale, NJ, Lawrence Erlbaum Associates, 2003