Abstract
BACKGROUND: Dementia has become an important public health challenge as the aging population in China accelerates, highlighting the need to identify modifiable risk factors. The cumulative atherogenic index of plasma (CumAIP), a marker reflecting long-term dyslipidemia, may contribute to dementia via vascular and inflammatory pathways, but longitudinal evidence in Chinese adults around 60 years of age remains scarce. METHODS: Data were extracted from three waves of the China Health and Retirement Longitudinal Study (CHARLS): 2012 (Wave 1), 2015 (Wave 3), and 2018 (Wave 4). A total of 6473 participants with complete AIP measurements at 2012 and 2015 were included; CumAIP was computed as the time-weighted average of AIP values, normalized by the 2012-2015 observation duration. Dementia risk (primary outcome) was assessed via the Rotterdam Basic Dementia Risk Model (BDRM) using 2018 data, reflecting a 3-year follow-up with 2015 as the baseline. To validate the BDRM, we conducted a secondary analysis: Spearman rank correlation between 2018 BDRM scores and 2015 CHARLS cognitive function scores. Statistical analyses included: multivariable linear regression with three progressive adjustment tiers (sociodemographic, clinical, lifestyle factors); restricted cubic spline (RCS) curves (3 knots at 10th/50th/90th CumAIP percentiles) for linearity testing; quartile-based analyses (CumAIPQ1-Q4, Q1 as reference) for dose-response relationships; and subgroup (by gender, age, residence, and lifestyle/metabolic factors) and sensitivity analyses [multiple imputation, complete-case analysis, linear regression of CumAIP vs. 2015 cognitive function scores, linear regression of low-density lipoprotein cholesterol (LDL-C) vs. BDRM scores] to validate robustness. RESULTS: First, the 2018 BDRM score was significantly negatively correlated with the 2015 cognitive function score (Spearman's r = - 0.26, P < 0.001). Higher CumAIP was associated with elevated BDRM scores, remaining significant after full adjustment (β = 0.058, 95% CI 0.018-0.098, P = 0.004). RCS confirmed a linear relationship (nonlinear term P = 0.9671), and quartile analyses showed a dose-response trend (Q4 vs. Q1, β = 0.103, 95% CI 0.009-0.197, P = 0.031). Subgroup effects were more pronounced in females, those aged ≥ 60 years, rural residents, nondrinkers, smokers, and BMI ≥ 24 kg/m(2). Sensitivity analyses validated robustness: (1) CumAIP correlated positively with 2015 cognitive function (β = 0.123, 95% CI 0.008-0.237, P = 0.037); (2) LDL-C correlated positively with BDRM scores (β = 0.019, 95% CI 0.009-0.028, P = 0.0001), with a smaller effect than CumAIP. CONCLUSION: Cumulative elevated CumAIP is independently associated with higher BDRM-estimated dementia risk in Chinese adults (mean age 61.4 ± 8.5 years), with linear and dose-dependent relationships. The association is more prominent in high-risk subgroups: females, aged ≥ 60 years, rural residents, nondrinkers, smokers, and BMI ≥ 24 kg/m(2). The BDRM's validity supports its reliability. CumAIP's stronger association with dementia risk than LDL-C underscores its value as a comprehensive lipid marker for risk stratification. Dynamic lipid monitoring to maintain low CumAIP may aid dementia prevention in these high-risk groups.