Using real-world data to characterize distinct experiences of and risk factors for worse cancer-related cognitive impairment.

Abstract

458 Background: Cancer-related cognitive impairment (CRCI) is a common and distressing symptom affecting decision-making, treatment adherence, and function. Limited research has used real-world data to identify patients at risk for worse CRCI. The purpose of this study was to use routinely-collected electronic patient-reported outcomes (ePRO) data to identify subgroups of patients with distinct CRCI experiences and factors associated with worse CRCI. Methods: Retrospective analysis of ePRO data submitted between 11/1/22 and 10/31/25. Patients rated “difficulty concentrating” on a scale of 0 (“none”) to 3 (“severe”) before each chemotherapy visit (approximately every 1-4 weeks). Eligible patients were adults treated for malignant solid non-CNS tumors who reported difficulty concentrating at least twice during the study period; data were retrieved using OncDRS. Latent class growth analysis (LCGA) identified subgroups of patients with distinct CRCI trajectories over 12 months; missing data were handled using Full Information Maximum Likelihood. Chi-square and ANOVA compared demographic and clinical characteristics across subgroups. Ordinal regression identified factors associated with worse difficulty concentrating, adjusting for age, race, ethnicity, marital status, education, employment, and disease site. This study received an exemption from the Dana-Farber Cancer Institute IRB. Results: Among 781 patients (Table 1), three subgroups were identified with persistent mild (n=193, 25%), moderate (n=480, 61%), and severe (n=108, 14%) difficulty concentrating over 12 months. Compared to the mild subgroup, the severe subgroup was younger and less likely to have graduated high school, be retired, or be married/partnered. Compared to the mild subgroup, the moderate subgroup was more likely to be disabled. In the adjusted model, predictors of worse difficulty concentrating included not completing high school, being diagnosed with breast cancer, and not being married/partnered (all p <0.05). Conclusions: Real-world data provides a powerful opportunity to characterize and predict interindividual variability in CRCI. Our findings suggest patients with less formal education or social support are at risk for worse CRCI and may benefit from proactive interventions. Patient characteristics across difficulty concentrating subgroups ( n = 781). Mild Moderate Severe Total n (%) n (%) n (%) n (%) All patients 193 (24.7) 480 (61.5) 108 (13.8) 781 (100) Age ( m, SD) 62.7 (13.0) 60.6 (13.0) 58.6 (14.2) 60.8 (13.2) Female 109 (24.0) 281 (61.8) 65 (14.3) 455 (100) Married or partnered 133 (27.7) 296 (61.7) 51 (10.6) 480 (100) Did not graduate high school 1 (5.0) 13 (65.0) 6 (30.0) 20 (100) Retired 79 (29.8) 160 (60.4) 26 (9.8) 265 (100) Disabled 11 (11.2) 73 (74.5) 14 (14.3) 98 (100) Diagnosed with breast cancer 17 (15.5) 70 (63.6) 23 (20.9) 110 (100)