Table 2.

LME Results for pTau Analytes

	pTau181		pTau217		pTau231
Predictors	Estimates	CI	Estimates	CI	Estimates	CI
(Intercept)	2.54 ***	2.39 – 2.70	0.40 ***	0.35 – 0.44	11.61 ***	10.28 – 12.95
IR	0.09	−0.00 – 0.19	−0.01	−0.04 – 0.01	0.40	−0.45 – 1.25
WHR	−0.32 **	−0.53 – −0.12	−0.06 *	−0.12 – −0.00	−1.77 *	−3.51 – −0.03
Age	0.04 ***	0.02 – 0.05	0.01 ***	0.01 – 0.01	0.20 **	0.05 – 0.34
Gender	0.24	−0.02 – 0.49	0.04	−0.03 – 0.11	−0.11	−2.24 – 2.01
Amyloid status	0.66 ***	0.42 – 0.89	0.40 ***	0.33 – 0.46	4.39 **	1.43 – 7.36
eGFR					−1.01 *	−1.88 – −0.13
Random Effects
σ2	0.40		0.04		3.20
τ00	0.57 Subject		0.04 Subject		10.76 Subject
ICC	0.59		0.54		0.77
N	272 Subject		272 Subject		70 Subject
Observations	463		463		79
Marginal R2 / Conditional R2	0.164 / 0.654		0.321 / 0.689		0.320 / 0.844
AIC	1242.203		102.476		432.504
log-Likelihood	−613.101		−43.238		−207.252

^*p<0.05

^**p<0.01

^***p<0.001

Three linear mixed models were estimated for each pTau analyte. For pTau181 and pTau217, the models included random intercepts with fixed effects of insulin resistance (IR) z-scored, waist-to-hip-ratio (WHR) binarized, mean-centered age, sex, and amyloid status (positive vs. negative). For pTau231, the model was the same with estimated glomerular filtration rate (eGFR) added as a fixed effect. Model comparison (AIC/BIC) indicated that models without eGFR performed better for pTau181 and pTau217 whereas the model with eGFR performed better for pTau231. All models were fitted in R version 4.4.2 using the lme4 package. Results indicate that WHR, age, and amyloid status were all significant predictors of longitudinal pTau concentrations for all analytes. For pTau231, eGFR was a significant predictor, indicating that lower kidney function was associated with higher plasma pTau231 concentrations over time.