Table 2.
Plasma NfLv1 and δNfL and their relation to all-cause mortality by sex: Cox PH hazards models
| Overall (N = 694), β ± SEa | Women (N = 401), β ± SE | Men (N = 293), β ± SE | Psexb | |
|---|---|---|---|---|
| NfLv1 | ||||
| Logetransformed,z-score | ||||
| Model 1d | + 0.275 ± 0.157, p = 0.080 | + 0.696 ± 0.247, p = 0.005 | + 0.060 ± 0.236, p = 0.80 | 0.018 |
| HR = 1.317 | HR = 2.006 | HR = 1.062 | ||
| Model 2d | + 0.241 ± 0.170, p = 0.16 | + 0.557 ± 0.273, p = 0.041 | + 0.072 ± 0.257, p = 0.78 | 0.07 |
| HR = 1.273 | HR = 1.745 | HR = 1.075 | ||
| Above vs. below mediane | ||||
| Model 1d | + 0.177 ± 0.352, p = 0.61 | + 0.477 ± 0.527, p = 0.36 | + 0.104 ± 0.486, p = 0.83 | 0.34 |
| HR = 1.194 | HR = 1.611 | HR = 1.110 | ||
| Model 2d | 0.095 ± 0.359, p = 0.79 | + 0.274 ± 0.566, p = 0.63 | − 0.041 ± 0.509, p = 0.94 | 0.43 |
| HR = 1.010 | HR = 1.315 | HR = 0.956 | ||
| δNfLc | ||||
| Logetransformed,z-score | ||||
| Model 1d | + 0.223 ± 0.141, p = 0.11 | + 0.436 ± 0.223, p = 0.051 | + 0.060 ± 0.197, p = 0.76 | 0.11 |
| HR = 1.250 | HR = 1.547 | HR = 1.062 | ||
| Model 2d | + 0.201 ± 0.144, p = 0.16 | + 0.500 ± 0.235, p = 0.034 | + 0.047 ± 0.200, p = 0.81 | 0.092 |
| HR = 1.223 | HR = 1.649 | HR = 1.048 | ||
| Above vs. below mediane | ||||
| Model 1d | + 0.448 ± 0.346, p = 0.20 | + 1.363 ± 0.649, p = 0.036 | − 0.194 ± 0.451, p = 0.67 | 0.015 |
| HR = 1.565 | HR = 3.908 | HR = 1.214 | ||
| Model 2d | + 0.440 ± 0.351, p = 0.21 | + 1.595 ± 0.695, p = 0.022 | − 0.266 ± 0.464, p = 0.57 | 0.011 |
| HR = 1.553 | HR = 4.928 | HR = 1.305 | ||
Abbreviations: Bayes empirical Bayes estimator, CES-D Center for Epidemiologic Studies-Depression, δ annualized rate of change, HANDLS Healthy Aging in Neighborhoods of Diversity Across the Life Span, HEI Healthy Eating Index, HR hazard ratio, NfL plasma neurofilament light, v1 visit 1, v2 visit 2, v3 visit 3
aValues are Loge(HR) ± SE, p from Cox PH hazards models associated with each NfL exposure of interest. Hazard ratio (HR) point estimates are also presented. 95% CI for HR can be calculated as follows: lower confidence limit (LCL): exp[LogeHR − 1.96×SE(LogeHR)]; upper confidence limit (UCL): exp[LogeHR + 1.96×SE(LogeHR)]
bP-value for the 2-way interaction term between plasma NfL exposure and sex in a separate unstratified model
cAnnual rate of change in NfL between v1 and v3 using the empirical Bayes estimator predicted from a mixed-effects linear regression model with NfL as the outcome and TIME as the only predictor, validated against the observed annualized change between v1 and v3 (Pearson’s r > 0.80)
dModel 1 adjusted for age at v1, sex, race, poverty status, and the inverse mills ratio; model 2 additionally adjusted for education, HEI-2010 total score, mean energy intake (kcal/day), current tobacco use, current illicit drug use, and the CES-D total score
eSD values of continuous exposures can be derived from Table 1. Binary NfLv1 and δNfL (above vs. below or = median) based on distributions in the final selected sample with a cutoff of 1.966279 for NfLv1 and 0.0466214 for δNfL