Use of Regional Brain Volume, the Montreal Cognitive Assessment, and Self-Reported Cognitive Complaint to Predict Future Cognition

Early and accurate prediction of future cognitive impairment can promote increased use of interventions, which in turn can lead to delayed decline and therefore higher quality and potentially longer length of life. There are a number of known risk factors/biomarkers that predict future cognition; however, prior literature is mixed and incomplete. Further, previous studies suggest that combining multiple risk factors may lead to increased utility above and beyond that of any single predictor, and a limited number of risk score algorithms have been derived accordingly, though more research is needed. The present study examined the ability of regional brain volumes, the Montreal Cognitive Assessment (MoCA), and self-reported cognitive complaint to predict future cognitive status (cognitively impaired or unimpaired), and a risk score algorithm was derived based on these findings. Exploratory analyses were also conducted to examine the ability of the aforementioned factors, excluding the MoCA, to predict future performance in various neuropsychological domains. Total brain volume, left hippocampal volume index, MoCA total score, and cognitive complaint significantly (all p's < .05) predicted diagnostic group (cognitively impaired or unimpaired), with a correct classification rate of 73.8%, controlling for sex (Nagelkerke R2 = .33, p < .001; Hosmer-Lemeshow p = .21). A risk score algorithm was created using these variables, which yielded an area under the curve (AUC) of .81 (p < .001). Regarding the exploratory analyses, whole brain volume was retained in models for all cognitive domain composites except verbal memory; and cognitive complaint was retained in models predicting global, language, and processing speed and attention composites. While a number of regional brain volume variables were retained in exploratory analyses, the majority of results were opposite the direction expected. The present study yielded a theoretically and statistically sound approach in producing a simple risk score algorithm to predict future cognitive decline that requires minimal resources and can be implemented across clinical settings, pending cross validation. Its use could lead to earlier intervention and ultimately improved preparation for, as well as reduced risk, delayed onset, and slower rate of, cognitive decline, thereby resulting in higher quality and potentially even length of life.