Associations of Cognitive Function and Education Level With All-Cause Mortality in Adults on Hemodialysis: Findings From the COGNITIVE-HD Study

Rationale & Objective: In the general population, cognitive impairment is associated with increased mortality, and higher levels of education are associated with lower risks for cognitive impairment and mortality. These associations are not well studied in patients receiving long-term hemodialysis and were the focus of the current investigation. Study Design: Prospective cohort study. Setting & Participants: Adult hemodialysis patients treated in 20 Italian dialysis clinics. Exposures: Patients' cognitive function across 5 domains (memory, attention, executive function, language, and perceptual-motor function), measured using a neuropsychological assessment comprising 10 tests; and patients' self-reported years of education. Outcome: All-cause mortality. Analytical Approach: Nested multivariable Cox regression models were used to examine associations of cognition (any domain impaired, number of domains impaired, and global function score from principal components analysis of unadjusted test scores) and education with mortality and whether there were interactions between them. Results: 676 (70.6%) patients participated, with a median age of 70.9 years and including 38.8% women. Cognitive impairment was present in 79.4% (527/664; 95% CI, 76.3%-82.5%). During a median follow-up of 3.3 years (1,874 personyears), 206 deaths occurred. Compared to no cognitive impairment, adjusted HRs for mortality were 1.77 (95% CI, 1.07-2.93) for any impairment, 1.48 (95% CI, 0.82-2.68) for 1 domain impaired, 1.88 (95% CI, 1.01-3.53) for 2 domains, and 2.01 (95% CI, 1.14-3.55) for 3 to 5 domains. The adjusted HR was 0.68 (95% CI, 0.51-0.92) per standard deviation increase in global cognitive function score. Compared with primary or lower education, adjusted HRs were 0.79 (95% CI, 0.53-1.20) for lower secondary and 1.13 (95% CI, 0.80-1.59) for upper secondary or higher. The cognition-by-education interaction was not significant (P = 0.7). Limitations: Potential selection bias from nonparticipation and missing data; no data for cognitive decline; associations with education were not adjusted for other socioeconomic factors. Conclusions: Cognitive impairment is associated with premature mortality in hemodialysis patients. Education does not appear to be associated with mortality.