Relatively preserved cognition despite elevated adjusted serum neurofilament light in U.S. adults aged 60–75 years: a survey-weighted NHANES analysis

Abstract Background Studies of circulating neurofilament light chain (NfL) and cognition have predominantly examined population-average associations, leaving individuals with relatively preserved cognition despite elevated biomarker burden insufficiently characterized. We investigated whether educational, socioeconomic, and behavioral resources were associated with this NfL–cognition discordance phenotype in older adults. Methods We conducted a cross-sectional complex-survey analysis of participants aged 60–75 years in the 2013–2014 National Health and Nutrition Examination Survey (NHANES). The primary analytic cohort included 502 participants with valid serum NfL measurements and complete cognitive assessments. Serum NfL was adjusted for age, sex, and estimated glomerular filtration rate, whereas global cognition was adjusted for age, sex, and test language. Participants were classified into four joint NfL–cognition phenotypes, with the primary analysis comparing participants with relatively preserved-cognition discordance against those with concordant impairment among individuals with elevated adjusted NfL. Survey-weighted regression models incorporated multiple imputation and prespecified robustness analyses. Results The weighted proportion meeting the study-defined criteria for relatively preserved-cognition discordance was 16.3% (95% CI, 12.0%–21.4%), whereas 8.9% (95% CI, 5.5%–13.6%) met the criteria for concordant impairment. These proportions are conditional on the internally defined NfL and cognition thresholds and should not be interpreted as clinical prevalence estimates. Among participants with elevated adjusted NfL, educational attainment was the only candidate correlate associated with relatively preserved cognition in the primary categorical analysis (overall p = 0.010). Compared with less than high school education, the odds of relatively preserved cognition increased across higher educational categories, whereas poverty-income ratio and recreational physical activity were not associated with the categorical phenotype. In complementary continuous analyses, education, poverty-income ratio, and physical activity were associated with better-than-expected cognition. Higher adjusted NfL showed a generally inverse pattern across cognitive outcomes, with the clearest evidence for global cognition (β=−0.117; 95% CI, − 0.206 to − 0.028) and Digit Symbol Substitution Test (DSST) performance (β=−0.086; 95% CI, − 0.172 to − 0.001). Findings were generally consistent across most stable prespecified sensitivity analyses, alternative cognitive definitions, inverse-probability weighting, and leave-one-primary-sampling-unit-out analyses, although evidence was attenuated under some alternative threshold definitions. Conclusions Among U.S. adults aged 60–75 years with elevated adjusted serum NfL in the NHANES serum-NfL subsample, cognitive performance was heterogeneous. Educational attainment was the most consistent correlate of relatively preserved-cognition discordance, although the data did not show that education modified the association between NfL and cognition. Joint consideration of biomarker burden and cognitive performance may provide complementary information beyond evaluating either measure alone. The phenotype framework should presently be regarded as an analytical research approach rather than a clinically deployable classification system, and longitudinal validation in independent cohorts is required.