Identification and prediction of inter-individual differences in cognitive training trajectories : a growth mixture modelling approach

There is emerging evidence of inter-individual differences in cognitive training responsiveness. Conventional statistics do not adequately address heterogeneity and longitudinal performance trajectories. Generalised growth mixture modelling (GGMM; Muthén, 2004) was utilised to identify and predict heterogeneous longitudinal cognitive performance trajectories following training. Specific and generalised effects of training were examined. Baseline characteristics such as age, sex and proxies for cognitive reserve were also explored as predictors of trajectories. Data from 315 community-dwelling older adults (age 55–85 years) from the Active Cognitive Enhancement (ACE) Program training study were analysed. Short-term (VM) and long-term verbal memory (LTVM) and executive functioning (EF) were tested using the Rey Auditory Verbal Learning Test (RAVLT) and the CogState Ltd Groton Maze Learning Test (GMLT) at baseline and at 3-, 6- and 12-month follow-ups. Generalised growth mixture modelling demonstrated High, Moderate, and Low performance classes for memory performance. High and Low classes were identified for executive function. Also identified were demonstrable performance trajectory gains in the trained individuals of the Low class for executive function, those performing at a low normative level at baseline (Cohen’s d = 2.23). These results offer a novel contribution to the literature. Gains by those trained in the Low performing VM and LTVM classes’ performance trajectories were also shown (Cohen’s d = 4.48 and 1.38, respectively). However, the experimental participants were compared to a small number of controls (n = 2) thus no meaningful training effects on memory were identified. The GGMM models therefore demonstrated that the multidomain ACE cognitive training program produced some generalised cognitive improvement in healthy older adults, albeit to limited extent. Age and estimated premorbid IQ (a proxy for cognitive reserve) predicted Low EF performance trajectories compared to High class performances. Trained individuals were more likely to be older and have lower levels of estimated pre-morbid IQ. Individuals who demonstrated executive function performance gains were less likely to demonstrate verbal memory trajectory gains. These findings suggest distinct responses to training in different cognitive domains and/or distinctive inter-individual responses to elements of the multi-domain training program. Caution with interpretation of GGMM labels and predictive factors identified is necessary, given their relativity to the cohort. With this approach, current theories including compensation, magnification, ‘Use It or Lose It’, plasticity, flexibility, and cognitive reserve are supported. Application of GGMM can also further facilitate development of individually tailored and cost effective cognitive training programs