Natural trajectory subclasses of cognitive impairment in breast cancer patients experiencing insomnia

Abstract

Background: Cancer-related cognitive impairment (CRCI) has traditionally been assessed in a dichotomous manner. Identifying subclasses of CRCI and novel biomarkers can improve the accuracy of identifying patients most at risk for CRCI.

Methods: A total of 139 breast cancer patients undergoing chemotherapy completed neurocognitive batteries over 12 months. Growth mixture modeling (GMM) was used to determine latent subgroups based on different trajectories of cognitive test performance across the four time points. Additionally, the authors collected peripheral blood to measure neuron-derived exosomes (NDE).

Results: Mean cognitive performance improved significantly over time (p < .001). However, GMM identified three distinct latent subgroups: patients with stable, high performance (class 1, N = 45), patients with variable low performance (class 2, N = 15), and patients with average performance who improved over time (class 3, N = 79). Cognitive subclass 2 was characterized by significantly lower education levels than the other two classes (p = .001). Cognitive subclass 1 had fewer racial/ethnic minority patients than the other two classes (p = .015). Cognitive subclasses did not differ significantly in any other demographic or clinical characteristic. There were no significant differences observed by NDE.

Conclusions: There are multiple distinct longitudinal trajectories of CRCI and these may be influenced by social determinants of health such as education and race/ethnicity. Future research can focus on ways to administer interventions earlier to those at most risk for CRCI and continue to explore novel biomarkers of CRCI.

Keywords: breast cancer; cognitive; gynecologic cancer; neuron‐derived exosomes; survivorship needs.

FIGURE 1

Longitudinal trajectories for cognitive test scores and NDE level. Growth mixture modeling indicated three latent subclasses for cognitive performance and NDE level across the four time points. Cognitive trajectories are derived from the multivariate growth mixture model, which jointly models the outcomes while accounting for collinearity. The same latent class structure underlies all outcomes, providing a unified representation of cognitive change across domains. Cognitive subclasses were characterized by a group of patients with relatively stable above average performance (class 1, “above average”), a group with low, variable performance (class 2, “below average”), and a group whose performance improved over time (class 3, “improvers”). Exosome subclasses indicated groups of patients with high NDE levels across time that spiked at 26 weeks (class 1, “high spikers”), average NDE levels (class 2, “steady average”) and low NDE levels (class 3, “steady low”). Shading around trend lines represents the 95% confidence interval. COWA indicates controlled oral word association; HVLT, Hopkins verbal learning test; NDE, neuron‐derived exosome.