Latent Change Score Modeling as a Method for Analyzing the Antidepressant Effect of a Psychosocial Intervention in Alzheimer's Disease

Background: Developing and evaluating interventions for patients with age-associated disorders is a rising field in psychotherapy research. Its methodological challenges include the high between-subject variability and the wealth of influencing factors associated with longer lifetime. Latent change score modeling (LCSM), a technique based on structural equation modeling, may be well suited to analyzing longitudinal data sets obtained in clinical trials. Here, we used LCSM to evaluate the antidepressant effect of a combined cognitive behavioral/cognitive rehabilitation (CB/CR) intervention in Alzheimer's disease (AD). Methods: LCSM was applied to predict the change in depressive symptoms from baseline as an outcome of the CORDIAL study, a randomized controlled trial involving 201 patients with mild AD. The participants underwent either the CORDIAL CB/CR program or standard treatment. Using LCSM, the model best predicting changes in Geriatric Depression Scale scores was determined based on this data set. Results: The best fit was achieved by a model predicting a decline in depressive symptoms between before and after testing. Assignment to the intervention group as well as female gender revealed significant effects in model fit indices, which remained stable at 6- and 12-month follow-up examinations. The pre-post effect was pronounced for patients with clinically relevant depressive symptoms at baseline. Conclusions: LCSM confirmed the antidepressant effect of the CORDIAL therapy program, which was limited to women. The effect was pronounced in patients with clinically relevant depressive symptoms at baseline. Methodologically, LCSM appears well suited to analyzing longitudinal data from clinical trials in aged populations, by accounting for the high between-subject variability and providing information on the differential indication of the probed intervention.Peer Reviewe