The Impact of Using Different Reference Populations on Measurement of Breast Cancer-Related Cognitive Impairment Rates

Objective: To evaluate how use of different reference populations affects estimates of breast cancer-related cognitive impairment rates. Methods: Patients aged ≥60 years with stage 0-3 breast cancer (n = 371) and matched non-cancer controls (n = 370) completed 13 neuropsychological tests prior to systemic therapy or at enrollment (controls). The tests captured three domains: attention, processing speed and executive function; learning and memory; and visual-spatial function. Domain-specific impairment was defined as having one test score 2 SD below or two or more test scores 1.5 SD below the reference population means. Different reference populations were used to define impairment: published normative data, study-specific controls, age and education-stratified controls, and age and education-adjusted controls. The associations between the resultant impairment rates and breast cancer (vs. control) were evaluated using chi-square tests and logistic regression models. Cohen's kappa coefficients were used to evaluate agreement of impairment rates between study-specific control performance and the other reference population groups. Results: The patients and controls were aged 68.0 (SD 6.0) and 67.9 (SD 7.0) years, respectively. The association of breast cancer-control status with impairment did not differ based on reference group. Cognitive impairment based on published normative data yielded less agreement on impairment rates (κ = 0.22-0.89) than study-specific age and education-stratified control performance (κ = 0.62-1.00). Conclusion: The choice of reference populations did not affect conclusions about the association of cognition with breast cancer. However, while study-specific reference populations provided greater internal consistency in defining cognitive impairment, benchmarking against published normative data will enhance the ability to compare results across studies.