Objective: Several studies showed that neuropsychological performance might service as predictor of antidepressant response, but other studies did not. A main contributing factor for the inconsistency could be the heterogeneous samples in terms of different clinical depressive subtypes that have preferential response to antidepressants; in this study, we attempted to assess the predictive values of domains of cognitive function on antidepressant response in three depressive subtypes including melancholic, atypical and undifferentiated depression (defined as have neither melancholic nor atypical features). Additionally, we generated multiple regression models in order to evaluate the impacts of illness course factors (e.g. age of onset, number of episodes, and depressive symptoms) on neuropsychological performance.
Method: The study was a 6-week prospective, longitudinal, semi-naturalistic design. The sample comprised of 142 melancholic, 76 atypical, and 91 undifferentiated depression according to the Diagnostic and Statistical Manual for Mental Disorders, Fourth Edition (DSM-IV_TR). A comprehensive battery of neuropsychological tests was administered to all participants, assessing seven cognitive domains, including processing speed, attention, shifting, planning, verbal şuency, and verbal and visual spatial memory. Given the fact that the potential predictors of cognitive measures and clinical variables are highly correlated, in the regression model we calculated the relative weighs (RW) instead of relative importance (RI); the former could be interpreted as estimates of RI of the original set of predictors but are not correlated. We considered the neuropsychological measures at the baseline as potential predictors while improvement in HAM-D scores after six weeks of treatments was the dependent variable in our regression models; the HAM-D scores at the baseline was controlled for due to its well-known impacts on treatment outcome.
Results: The three regression models (each for a subtype) explained a significant proportion of variance of the criterion (for atypical depression: R2=0.564, F=4.00, p=0.001; for melancholic depression: R2=0.376, F=4.66, p<0.001; for undifferentiated depression: R2=0.597, F=7.96, p<0.001). We then computed the RW, and the corresponding significance tests. However, only the depressive symptoms at baseline (i.e., HAM-D scores) had a significant contribution in the prediction of antidepressants response (HAM-D scores improvement). We once recomputed the three regression analyses removing the depression at baseline (i.e., HAM-D scores). However, none of the three models managed to explain a significant proportion of variance of the criterion (i.e., significant tests suggested that R2 was zero in the population). Among the tested clinical variable—age at onset, duration, number of episodes, HAM-A, HAM-D, YMRS, psychotic symptoms and years of education—, only age at onset demonstrated a predictive power in the domain of processing speed (measured by symbol coding of WAIS-RC) in atypical depression (R2=0.447, p<0.001; RW=26.0, p=0.01).
Conclusion: our data do not lead support, at least at the practical level, to the notion that neuropsychological performance be able to predict antidepressant treatment outcome. Apart from age at onset, we did not find significant impacts of the illness course variables on neuropsychological performance in the three subtypes of major depression.