Integrating Frailty and Cumulative Lipid Burden for Stroke Risk Stratification: A Machine Learning–Guided Athero -Frailty Score From the CHARLS Cohort

Integrating Frailty and Cumulative Lipid Burden for Stroke Risk Stratification: A Machine Learning–Guided Athero -Frailty Score From the CHARLS Cohort | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Integrating Frailty and Cumulative Lipid Burden for Stroke Risk Stratification: A Machine Learning–Guided Athero -Frailty Score From the CHARLS Cohort jingxian sun, Daikang Xu, Guangyu Du, Tongyu Jia, Xing Han, Yi Yu, and 5 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8788634/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 31 Mar, 2026 Read the published version in Lipids in Health and Disease → Version 1 posted 10 You are reading this latest preprint version Abstract Background Traditional lipid indices, such as the cumulative Atherogenic Index of Plasma (cumAIP), demonstrate inconsistent predictive performance for stroke in older adults, likely due to the modifying effects of complex medication regimens. We hypothesized that incorporating the Frailty Index (FI)—a measure of cumulative physiological deficits—could capture the residual risk missed by lipid markers alone. We developed and evaluated a novel Athero-Frailty Score (AFS) to address these limitations. Methods We analyzed 3,690 participants from the China Health and Retirement Longitudinal Study (CHARLS) using a landmark analysis design (baseline: 2015). An exploratory XGBoost model with SHAP analysis was used for variable screening. Based on the orthogonality of lipid and frailty metrics, AFS was constructed as an additive composite score of cumAIP and FI. Associations with incident stroke were assessed using Cox proportional hazards models and restricted cubic splines (RCS). Clinical utility was evaluated via C-statistics, Net Reclassification Improvement (NRI), and Integrated Discrimination Improvement (IDI). Results SHAP analysis identified FI as having a larger average SHAP contribution than cumAIP and medication indicators. Survival analyses suggested that the association between cumAIP and stroke was attenuated after adjustment for antihypertensive and lipid-lowering therapies (HR 1.20, P = 0.042), and RCS indicated a non-linear pattern with an apparent plateau at higher cumAIP levels. In contrast, AFS was associated with an approximately linear dose–response relationship independent of medication use. In the fully adjusted model, each 1-SD increase in AFS was associated with a higher risk of stroke (HR = 2.75, P < 0.001). Crucially, the addition of AFS yielded significant improvement in reclassification (NRI = 15.4%, P < 0.001), while the improvement in integrated discrimination was modest (IDI = 0.007, P = 0.088). Conclusion The AFS effectively addresses the predictive limitations of traditional metabolic markers in medicated older adults. By integrating metabolic burden with systemic vulnerability, this novel composite score offers a linear and robust predictive approach for stroke, supporting a multidimensional approach to vascular risk stratification in aging populations. Stroke Athero-Frailty Score cumAIP Frailty Index Machine Learning Figures Figure 1 Figure 2 Figure 3 Introduction Stroke imposes substantial mortality and long-term disability worldwide, and improving the identification of individuals at elevated risk remains a central goal of primary prevention [ 1 , 2 ]. Dyslipidemia is a well-established contributor to atherosclerotic disease, yet lipid measurements obtained at a single time point may not fully capture an individual’s long-term exposure to atherogenic lipids[ 3 ]. To better approximate cumulative metabolic burden, repeated-measure constructs such as the atherogenic index of plasma (AIP; log[TG/HDL-C]) have been proposed[ 4 ]. By incorporating information across multiple assessments, cumulative or time-averaged lipid metrics have been suggested to improve risk stratification compared with baseline measurements alone[ 5 ]. Risk assessment in older adults, however, presents additional challenges. Medication use is highly prevalent in aging populations[ 6 ], and antihypertensive or lipid-lowering therapies can substantially modify circulating biomarker levels. As a result, measured metabolic indices may no longer fully reflect the underlying burden of vascular injury accumulated over time, a phenomenon increasingly recognized in geriatric cardiology[ 7 , 8 ]. Meanwhile, data from major outcomes trials—such as REDUCE-IT and PROMINENT—indicate that vascular events continue to occur at high rates despite achieving guideline-recommended treatment targets or biomarker reduction[ 9 , 10 ]. Furthermore, large-scale meta-analyses suggest that the relative risk reduction conferred by statin therapy may be attenuated in older adults compared with younger populations [ 11 ]. Together, these observations support the "residual risk" hypothesis[ 12 ] and raise the possibility that reliance on metabolic markers alone is insufficient for stroke risk stratification in older, treated individuals[ 13 ]. Frailty represents a critical, yet underutilized, dimension of residual risk. Conceptualized as multisystem vulnerability and reduced resilience to physiological stressors[ 14 ], frailty is driven by "inflammaging"—chronic, low-grade inflammation that promotes both atherogenesis and plaque instability[ 15 ]. The frailty index (FI), based on the deficit accumulation framework, summarizes impairments in chronic conditions, functional status, cognition, and psychological health into a single continuous measure of global deficit burden[ 16 ]. A growing body of evidence has linked frailty to adverse cardiovascular outcomes, including incident stroke, independent of chronological age and traditional risk factors[ 17 , 18 ]. Importantly, frailty burden correlates poorly with traditional lipid markers, suggesting that it captures complementary information related to systemic vulnerability that is not reflected by the lipidome alone[ 19 ]. In this study, we leveraged data from the large, community-based China Health and Retirement Longitudinal Study (CHARLS)[ 20 ] to develop and validate a novel integrated risk metric. Using an exploratory machine-learning approach to guide variable selection, we constructed the Athero-Frailty Score (AFS) as a parsimonious composite of cumulative lipid burden and frailty index. We evaluated the dose–response relationship between AFS and incident stroke and quantified its incremental predictive utility over standard clinical risk factors. Our primary objective was to determine whether this multidimensional construct can overcome the limitations of traditional metabolic markers and enhance risk stratification in older adults. Methods Study Population This study analyzed data from the China Health and Retirement Longitudinal Study (CHARLS), a nationally representative longitudinal cohort of residents aged 45 years and older. The study design and sampling procedures have been described in detail previously. The baseline survey was conducted in 2011 (Wave 1), with biennial follow-up assessments in 2013 (Wave 2), 2015 (Wave 3), and 2018 (Wave 4).For the current analysis, participants were initially screened if they had complete biomarker data in both 2011 and 2015 to allow for the calculation of cumulative indices. The exclusion criteria were as follows: (1) a self-reported doctor diagnosis of stroke or cardiovascular disease (CVD) at or before the 2015 baseline; (2) missing data for variables required for the Frailty Index (FI) construction; or (3) loss to follow-up during the outcome ascertainment period (2015–2018). The final analytic cohort consisted of 3,690 participants. The study protocol was approved by the Ethical Review Committee of Peking University, and written informed consent was obtained from all participants. Assessment of Cumulative AIP (cumAIP) Blood samples were collected after an overnight fast. Triglycerides (TG) and High-density lipoprotein cholesterol (HDL-C) were measured using enzymatic colorimetric tests. The Atherogenic Index of Plasma (AIP) was calculated as log(TG / HDL-C).To capture the long-term burden, cumAIP was calculated as the weighted sum of the average AIP between consecutive waves (2011–2015), multiplied by the time interval (years). The formula was defined as: cumAIP = Σ [(AIP_2011 + AIP_2015) / 2 × (2015 − 2011)]. Assessment of Frailty Index (FI) The Frailty Index (FI) was constructed using the standard deficit-accumulation approach. A total of 32 health deficits were included, encompassing multiple domains: chronic comorbidities (e.g., hypertension, diabetes, dyslipidemia), functional limitations (Activities of Daily Living [ADL] and Instrumental ADL), cognitive function, and depressive symptoms. Each deficit was dichotomized or graded, mapping to a score between 0 (absence of deficit) and 1 (presence of deficit). The FI score was calculated as the ratio of the number of deficits present to the total number of deficits considered, yielding a continuous variable ranging from 0 to 1. Construction of the Athero-Frailty Score (AFS) To jointly capture cumulative metabolic burden and global deficit burden, we constructed the Athero-Frailty Score (AFS). Cumulative atherogenic lipid burden was quantified using the cumulative atherogenic index of plasma (cumAIP), derived from repeated AIP measurements between 2011 and 2015, while frailty burden was assessed at baseline (2015) using a 32-item frailty index (FI) based on the deficit accumulation approach. To place the two components on a comparable scale, both cumAIP and FI were standardized to z-scores. AFS was then defined as a weighted linear combination of these standardized variables, with weights (β coefficients) obtained from multivariable Cox proportional hazards models: AFS = β₁ × Z(cumAIP) + β₂ × Z(FI). AFS was analyzed as a continuous variable and, for descriptive and stratified analyses, further categorized into tertiles based on its distribution in the study population. An additive formulation was prespecified for conceptual and methodological reasons. Conceptually, cumAIP and FI represent complementary but distinct dimensions of stroke risk—metabolic lipid burden and multisystem vulnerability, respectively—and an additive model provides a parsimonious and interpretable approach to integrating these domains in a prediction-focused framework. Methodologically, multiplicative or higher-order combinations (e.g., cumAIP × FI) can generate highly skewed distributions and disproportionately amplify extreme observations, potentially leading to model instability and overfitting in finite samples. In addition, selecting a functional form after exploring multiple nonlinear combinations may introduce model-selection optimism and inflate apparent predictive performance. To minimize these risks and improve interpretability and transportability, we specified an additive linear composite as the primary model a priori. Potential nonlinearity and interaction between cumAIP and FI were examined in sensitivity analyses. Outcome Assessment The primary outcome was incident stroke, defined as the first occurrence of a stroke event during the follow-up period (2015–2018). Stroke events were ascertained based on self-reported doctor diagnoses ("Have you been diagnosed with stroke by a doctor?") during the 2018 wave. Follow-up time was calculated from the date of the 2015 interview to the date of stroke diagnosis or the date of the 2018 interview (censored), whichever occurred first. Covariates Covariates were collected at baseline (2015) and included:Demographic: Age, sex, marital status, education level, and residence (urban/rural).Lifestyle: Smoking status (current/former/never) and alcohol consumption.Clinical: Systolic blood pressure (SBP), fasting plasma glucose (FPG), C-reactive protein (CRP), estimated glomerular filtration rate (eGFR), blood urea nitrogen (BUN), and body mass index (BMI).Medication History: Self-reported use of antihypertensive, antidiabetic, and lipid-lowering agents. Statistical Analysis Baseline characteristics were compared across AFS tertiles using ANOVA, Kruskal-Wallis tests, or Chi-square tests.Machine Learning Screening: To contextualize predictor importance, we employed the XGBoost algorithm. The SHAP (SHapley Additive exPlanations) method was used to interpret the model and rank variables based on their global contribution to stroke prediction.Survival Analysis: Cox proportional hazards models were used to estimate Hazard Ratios (HRs) and 95% Confidence Intervals (CIs). Model 1 was unadjusted; Model 2 adjusted for age and sex; Model 3 further adjusted for clinical biomarkers; and Model 4 (fully adjusted) included medication history.Linearity and Incremental Value: Restricted Cubic Splines (RCS) were used to visualize the dose-response relationship between AFS and stroke risk. The incremental predictive value of AFS over cumAIP was assessed using the C-index, Net Reclassification Improvement (NRI), and Integrated Discrimination Improvement (IDI).All statistical analyses were performed using R software (version 4.5.2). A two-sided P-value < 0.05 was considered statistically significant. Results Predictive Limitations of cumAIP and the Association With Medication Use The final analytic cohort consisted of 3,690 participants, with baseline demographic and clinical characteristics stratified by AFS categories presented in Table 1. Regarding lipid burden, initial stratification by cumAIP tertiles confirmed a positive association with stroke risk, with participants in the highest tertile (T3) exhibiting a higher incidence compared with those in the lowest tertile (Table S1). In unadjusted and minimally adjusted models, cumAIP was a significant predictor of incident stroke (Table 2). However, with further adjustment for clinical covariates and specifically medication use (including antihypertensive and lipid-lowering therapies), the association between cumAIP and stroke risk was progressively attenuated. In the fully adjusted model (Model 4), the independent association of cumAIP with stroke risk was modest and of borderline statistical significance (HR=1.20, 95% CI =1.01–1.43, P = 0.042; Table 2). Restricted cubic spline (RCS) analyses further demonstrated a non-linear association, characterized by an apparent plateau at higher cumAIP levels (Figure 1A). Together, these findings suggest that cumulative lipid burden alone may have limited discriminatory capacity for stroke risk stratification in the context of widespread medication use. Exploratory Machine-Learning Analysis To contextualize the relative contributions of candidate predictors, we applied an exploratory XGBoost model with SHAP value interpretation. In this model, the Frailty Index (FI) demonstrated a larger average SHAP contribution to stroke prediction than cumAIP, eGFR, and medication indicators (Figure 2 and S2). Correlation analyses showed minimal correlation between FI and cumAIP ( r = 0.04; Figure S2), indicating that the two measures capture distinct, orthogonal dimensions of risk. As prespecified, this machine-learning analysis was utilized strictly for hypothesis generation to inform the selection of variables for the composite score, rather than for final model specification. Association Between the Athero-Frailty Score and Incident Stroke Baseline characteristics stratified by AFS tertiles are summarized in Table 2. Participants in the highest AFS tertile were older, more likely to be female, and exhibited both a less favorable metabolic profile and substantially higher frailty burden compared with those in the lowest tertile. Notably, AFS stratification revealed a pronounced gradient in frailty burden that was not apparent when stratifying by cumAIP alone. In multivariable Cox models, AFS was strongly and independently associated with incident stroke (Table 1). In the fully adjusted model, a 1-SD increase in AFS was associated with a 2.75-fold increased risk of stroke (P < 0.001). Unlike cumAIP, RCS analyses for AFS demonstrated an approximately linear dose–response relationship across its distribution, with no evidence of a plateau (Figure 1B). Kaplan–Meier analyses showed clear and early separation of cumulative stroke incidence across AFS tertiles (log-rank P < 0.001), whereas separation by cumAIP tertiles was less pronounced (Figure S3). Incremental Predictive Performance and Sensitivity Analyses The addition of AFS to the baseline risk model was associated with improved model discrimination, increasing the C-statistic from 0.681 to 0.706 (P < 0.001; Figure S4). Notably, AFS yielded a substantial and statistically significant improvement in Net Reclassification Improvement (NRI = 15.4%, P < 0.001), indicating that the score correctly reclassified a significant proportion of individuals into appropriate risk categories. The Integrated Discrimination Improvement (IDI) showed a positive trend but was modest (0.007, P = 0.088). In contrast, the addition of cumAIP alone did not result in a statistically significant improvement in reclassification (NRI = 11.1%, P = 0.076). The association between AFS and stroke risk was consistent across subgroups stratified by age, sex, and hypertension status, with no significant interactions observed (P_interaction > 0.05; Figure 3). Crucially, sensitivity analyses were conducted to compare the additive AFS specification against alternative strategies, including a multiplicative product score (cumAIP×FI) and an explicit multiplicative interaction term. Although the product term was associated with stroke when modeled alone, it did not improve discrimination or reclassification compared with the additive AFS, and adding the interaction term to the model provided no incremental predictive value (Table S3). Consistently, the multiplicative interaction analysis did not show strong evidence of synergy beyond main effects, supporting the robustness and parsimony of the additive AFS specification (Table S4). Discussion In this prcospective cohort of community-dwelling middle-aged and older adults, we evaluated whether integrating cumulative atherogenic lipid burden with frailty burden improves short-term stroke risk stratification. Three main findings emerged. First, although cumulative AIP was positively associated with incident stroke, this association was attenuated after adjustment for medication use and exhibited a distinct non-linear pattern with a plateau at higher levels, corroborating recent observations regarding lipid variability and risk plateauing[ 21 , 22 ]. Second, frailty burden, as assessed by the Frailty Index, provided complementary risk information that was largely orthogonal to cumulative lipid burden. Third, a parsimonious additive composite integrating cumAIP and FI—the Athero-Frailty Score (AFS)—demonstrated a robust linear association with stroke risk and significantly improved discrimination and integrated discrimination, with more modest improvement in reclassification beyond standard risk factors. The observed attenuation of the association between cumulative lipid burden and stroke risk after accounting for medication use is consistent with the "medication paradox" or "reverse epidemiology" often seen in aging populations[ 23 , 24 ]. In older adults, biomarker levels measured during treatment may not fully reflect lifetime exposure or the extent of underlying vascular injury[ 25 ]. Moreover, individuals receiving antihypertensive or lipid-lowering therapies often represent a higher baseline risk group, confounding interpretation in observational settings—a phenomenon known as "confounding by indication"[ 26 ]. Our restricted cubic spline analyses suggest that cumulative AIP alone has limited incremental discriminatory capacity at higher exposure levels once these clinical factors are taken into account. This aligns with data from the PROSPER trial and other geriatric cohorts suggesting that the predictive strength of traditional lipid markers diminishes with age[ 27 , 28 ], potentially due to the competing risks of non-cardiovascular mortality and the catabolic effects of chronic disease[ 29 ]. Importantly, these findings reflect the challenges of risk stratification in medicated populations rather than a lack of biological relevance of lipids[ 30 ]. Frailty captures multisystem vulnerability and reduced physiological reserve, integrating information across comorbidities, functional limitations, and psychosocial domains. Consistent with previous studies linking frailty to adverse cardiovascular outcomes[ 31 , 32 ], FI showed a strong association with incident stroke in our analyses. Mechanistically, this link may be mediated by shared pathways such as oxidative stress, mitochondrial dysfunction, and dysregulated coagulation, which are hallmarks of both frailty and cerebrovascular disease [ 33 , 34 ]. The weak correlation between FI and cumAIP observed in our study suggests that frailty burden and lipid burden represent distinct pathophysiological axes[ 35 ]. From a prediction perspective, this orthogonality provides the biological rationale for their joint consideration. Rather than replacing metabolic markers, frailty assessment captures the "additional vulnerability"—or residual risk—related to biological aging that lipid indices miss [ 36 ]. A critical methodological insight from our study concerns the structural relationship between metabolic and functional risk. While biological systems often exhibit complex interactions, our rigorous sensitivity analyses comparing additive versus multiplicative specifications provided strong support for an additive accumulation model. The inclusion of multiplicative interaction terms (cumAIP × FI) failed to provide incremental improvement in discrimination or reclassification compared with the simpler additive AFS. This finding has important conceptual implications: it suggests that metabolic atherogenicity and systemic frailty may operate as parallel, largely independent risk vectors rather than synergistic amplifiers[ 37 ]. By summing these independent deficits, AFS captures the "total physiological load" in accordance with the principle of parsimony[ 38 ], minimizing the risk of overfitting inherent in complex interaction models while maximizing interpretability and transportability across populations[ 39 – 41 ]. This additive approach parallels the construction of polygenic risk scores, where multiple small-effect variants are summed to estimate aggregate liability[ 42 ]. From a clinical perspective, our findings suggest that stroke risk assessment in older adults may benefit from incorporating information on global health status alongside traditional metabolic markers. Frailty assessment is increasingly feasible in both research and clinical settings, including through electronic health record–based indices[ 43 , 44 ]. An integrated score such as AFS may assist in identifying individuals at elevated short-term risk who might not be fully captured by lipid measures alone, aligning with recent AHA/ACC recommendations to incorporate geriatric assessment into cardiovascular care[ 45 , 46 ]. However, the observed improvements in predictive performance should be interpreted cautiously and do not, by themselves, imply changes to treatment thresholds or management strategies without further interventional evidence[ 47 ]. From a research standpoint, these results underscore the importance of considering multidimensional vulnerability in cardiovascular risk prediction. Future studies should evaluate the performance of AFS in other populations and over longer follow-up periods[ 48 ], and assess whether incorporating frailty into risk assessment frameworks improves clinical decision-making or patient outcomes via "frailty-guided" management strategies[ 49 – 51 ]. Strengths and Limitations This study has several strengths, including the use of a large, nationally representative cohort with repeated biomarker measurements and detailed assessment of frailty. The integration of exploratory machine-learning analyses with hypothesis-driven survival modeling provided complementary perspectives on predictor contributions. Several limitations merit consideration. Stroke events were based on self-reported physician diagnoses, which may introduce misclassification. The follow-up period was relatively short, and residual confounding cannot be excluded. In addition, although we examined multiple sensitivity analyses, external validation in independent cohorts is necessary before broader application. Conclusions In this cohort of middle-aged and older adults, integrating cumulative atherogenic lipid burden with frailty burden via a parsimonious additive score was associated with improved short-term stroke risk stratification. Our findings indicate that metabolic and functional deficits contribute independently to vascular risk, supporting a multidimensional "Athero-Frailty" framework for prediction in aging populations. Declarations Competing interests The authors declare that they have no competing interests. Funding Jingxian Sun received financial support from the Youth Research Fund of the Affiliated Hospital of Qingdao University (Grant. No. QDFYQN2023225); Chao Wang received financial support from the Youth Research Fund of the Affiliated Hospital of Qingdao University (Grant. No. QDFYQN2023114). Author Contribution JS and CW conceived and designed the study. JS, GD and DX performed the statistical analysis and machine learning modeling using R software. XH, YY, and JW contributed to data curation and data cleaning from the CHARLS database. ZY and SL assisted in the interpretation of the results and validation. JS and TJ drafted the original manuscript. CW and SC obtained funding, supervised the study, and critically revised the manuscript for important intellectual content. All authors read and approved the final manuscript. Acknowledgement We thank the China Health and Retirement Longitudinal Study (CHARLS) team for providing data and all participants for their contribution. The National School of Development at Peking University provided the data. We sincerely appreciate the YIWANDOU team for their invaluable contribution to data cleaning and processing. Data Availability The datasets generated and/or analyzed during the current study are available in the CHARLS repository, http://charls.pku.edu.cn/en. References Collaborators GBDS. Global, regional, and national burden of stroke and its risk factors, 1990–2019: a systematic analysis for the Global Burden of Disease Study 2019. Lancet Neurol. 2021;20(10):795–820. Kleindorfer DO, Towfighi A, Chaturvedi S, Cockroft KM, Gutierrez J, Lombardi-Hill D, et al. 2021 Guideline for the Prevention of Stroke in Patients With Stroke and Transient Ischemic Attack: A Guideline From the American Heart Association/American Stroke Association. Stroke. 2021;52(7):e364–467. Ference BA, Ginsberg HN, Graham I, Ray KK, Packard CJ, Bruckert E, et al. Low-density lipoproteins cause atherosclerotic cardiovascular disease. 1. Evidence from genetic, epidemiologic, and clinical studies. A consensus statement from the European Atherosclerosis Society Consensus Panel. Eur Heart J. 2017;38(32):2459–72. Dobiasova M, Frohlich J. The plasma parameter log (TG/HDL-C) as an atherogenic index: correlation with lipoprotein particle size and esterification rate in apoB-lipoprotein-depleted plasma (FER(HDL)). Clin Biochem. 2001;34(7):583–8. Domanski MJ, Tian X, Wu CO, Reis JP, Dey AK, Gu Y, et al. Time Course of LDL Cholesterol Exposure and Cardiovascular Disease Event Risk. J Am Coll Cardiol. 2020;76(13):1507–16. Qato DM, Wilder J, Schumm LP, Gillet V, Alexander GC. Changes in Prescription and Over-the-Counter Medication and Dietary Supplement Use Among Older Adults in the United States, 2005 vs 2011. JAMA Intern Med. 2016;176(4):473–82. Yi SW, Yi JJ, Ohrr H. Total cholesterol and all-cause mortality by sex and age: a prospective cohort study among 12.8 million adults. Sci Rep. 2019;9(1):1596. Ofori-Asenso R, Jakhu A, Zomer E, Curtis AJ, Korhonen MJ, Nelson M, et al. Adherence and Persistence Among Statin Users Aged 65 Years and Over: A Systematic Review and Meta-analysis. J Gerontol Biol Sci Med Sci. 2018;73(6):813–9. Bhatt DL, Steg PG, Miller M, Brinton EA, Jacobson TA, Ketchum SB, et al. Cardiovascular Risk Reduction with Icosapent Ethyl for Hypertriglyceridemia. N Engl J Med. 2019;380(1):11–22. Das Pradhan A, Glynn RJ, Fruchart JC, MacFadyen JG, Zaharris ES, Everett BM, et al. Triglyceride Lowering with Pemafibrate to Reduce Cardiovascular Risk. N Engl J Med. 2022;387(21):1923–34. Cholesterol Treatment Trialists C. Efficacy and safety of statin therapy in older people: a meta-analysis of individual participant data from 28 randomised controlled trials. Lancet. 2019;393(10170):407–15. Ridker PM. Residual inflammatory risk: addressing the obverse side of the atherosclerosis prevention coin. Eur Heart J. 2016;37(22):1720–2. Libby P. The changing landscape of atherosclerosis. Nature. 2021;592(7855):524–33. Clegg A, Young J, Iliffe S, Rikkert MO, Rockwood K. Frailty in elderly people. Lancet. 2013;381(9868):752–62. Ferrucci L, Fabbri E. Inflammageing: chronic inflammation in ageing, cardiovascular disease, and frailty. Nat Rev Cardiol. 2018;15(9):505–22. Searle SD, Mitnitski A, Gahbauer EA, Gill TM, Rockwood K. A standard procedure for creating a frailty index. BMC Geriatr. 2008;8:24. Veronese N, Cereda E, Stubbs B, Solmi M, Luchini C, Manzato E, et al. Risk of cardiovascular disease morbidity and mortality in frail and pre-frail older adults: Results from a meta-analysis and exploratory meta-regression analysis. Ageing Res Rev. 2017;35:63–73. Evans NR, Todd OM, Minhas JS, Fearon P, Harston GW, Mant J, et al. Frailty and cerebrovascular disease: Concepts and clinical implications for stroke medicine. Int J Stroke. 2022;17(3):251–9. Boehm FJ, Fritzenschaft L, Braig S, Denkinger M, Rothenbacher D, Dallmeier D. Sex- and system-specific analysis of blood-based biomarkers and frailty in older adults-the Activity and Function of the Elderly study. Age Ageing. 2025;54(9). Zhao Y, Hu Y, Smith JP, Strauss J, Yang G. Cohort profile: the China Health and Retirement Longitudinal Study (CHARLS). Int J Epidemiol. 2014;43(1):61–8. Varbo A, Nordestgaard BG. Remnant cholesterol and risk of ischemic stroke in 112,512 individuals from the general population. Ann Neurol. 2019;85(4):550–9. Johannesen CDL, Mortensen MB, Langsted A, Nordestgaard BG. Apolipoprotein B and Non-HDL Cholesterol Better Reflect Residual Risk Than LDL Cholesterol in Statin-Treated Patients. J Am Coll Cardiol. 2021;77(11):1439–50. Ravnskov U, Diamond DM, Hama R, Hamazaki T, Hammarskjold B, Hynes N, et al. Lack of an association or an inverse association between low-density-lipoprotein cholesterol and mortality in the elderly: a systematic review. BMJ Open. 2016;6(6):e010401. Nissen SE, Menon V, Nicholls SJ, Brennan D, Laffin L, Ridker P, et al. Bempedoic Acid for Primary Prevention of Cardiovascular Events in Statin-Intolerant Patients. JAMA. 2023;330(2):131–40. Navar AM, Kolkailah AA, Gupta A, Gillard KK, Israel MK, Wang Y, Peterson ED. Gaps in Guideline-Based Lipid-Lowering Therapy for Secondary Prevention in the United States: A Retrospective Cohort Study of 322 153 Patients. Circ Cardiovasc Qual Outcomes. 2023;16(8):533–43. Kyriacou DN, Lewis RJ. Confounding by Indication in Clinical Research. JAMA. 2016;316(17):1818–9. Shepherd J, Blauw GJ, Murphy MB, Bollen EL, Buckley BM, Cobbe SM, et al. Pravastatin in elderly individuals at risk of vascular disease (PROSPER): a randomised controlled trial. Lancet. 2002;360(9346):1623–30. Postmus I, Trompet S, Deshmukh HA, Barnes MR, Li X, Warren HR, et al. Pharmacogenetic meta-analysis of genome-wide association studies of LDL cholesterol response to statins. Nat Commun. 2014;5:5068. Eitmann S, Matrai P, Hegyi P, Balasko M, Eross B, Dorogi K, Petervari E. Obesity paradox in older sarcopenic adults - a delay in aging: A systematic review and meta-analysis. Ageing Res Rev. 2024;93:102164. Mortensen MB, Nordestgaard BG. Elevated LDL cholesterol and increased risk of myocardial infarction and atherosclerotic cardiovascular disease in individuals aged 70–100 years: a contemporary primary prevention cohort. Lancet. 2020;396(10263):1644–52. Pasieczna AH, Szczepanowski R, Sobecki J, Katarzyniak R, Uchmanowicz I, Gobbens RJJ, et al. Importance analysis of psychosociological variables in frailty syndrome in heart failure patients using machine learning approach. Sci Rep. 2023;13(1):7782. Fried LP, Tangen CM, Walston J, Newman AB, Hirsch C, Gottdiener J, et al. Frailty in older adults: evidence for a phenotype. J Gerontol Biol Sci Med Sci. 2001;56(3):M146–56. Walston J, McBurnie MA, Newman A, Tracy RP, Kop WJ, Hirsch CH, et al. Frailty and activation of the inflammation and coagulation systems with and without clinical comorbidities: results from the Cardiovascular Health Study. Arch Intern Med. 2002;162(20):2333–41. Iadecola C, Anrather J. The immunology of stroke: from mechanisms to translation. Nat Med. 2011;17(7):796–808. Ramsay SE, Arianayagam DS, Whincup PH, Lennon LT, Cryer J, Papacosta AO, et al. Cardiovascular risk profile and frailty in a population-based study of older British men. Heart. 2015;101(8):616–22. Afilalo J. Frailty in Patients with Cardiovascular Disease: Why, When, and How to Measure. Curr Cardiovasc Risk Rep. 2011;5(5):467–72. Knol MJ, VanderWeele TJ. Recommendations for presenting analyses of effect modification and interaction. Int J Epidemiol. 2012;41(2):514–20. Steyerberg EW, Harrell FE Jr., Borsboom GJ, Eijkemans MJ, Vergouwe Y, Habbema JD. Internal validation of predictive models: efficiency of some procedures for logistic regression analysis. J Clin Epidemiol. 2001;54(8):774–81. Harrell FE Jr., Lee KL, Mark DB. Multivariable prognostic models: issues in developing models, evaluating assumptions and adequacy, and measuring and reducing errors. Stat Med. 1996;15(4):361–87. Ren X, Mi Z, Georgopoulos PG. Comparison of Machine Learning and Land Use Regression for fine scale spatiotemporal estimation of ambient air pollution: Modeling ozone concentrations across the contiguous United States. Environ Int. 2020;142:105827. Kundi H, Wadhera RK, Strom JB, Valsdottir LR, Shen C, Kazi DS, Yeh RW. Association of Frailty With 30-Day Outcomes for Acute Myocardial Infarction, Heart Failure, and Pneumonia Among Elderly Adults. JAMA Cardiol. 2019;4(11):1084–91. Khera AV, Emdin CA, Drake I, Natarajan P, Bick AG, Cook NR, et al. Genetic Risk, Adherence to a Healthy Lifestyle, and Coronary Disease. N Engl J Med. 2016;375(24):2349–58. Clegg A, Bates C, Young J, Ryan R, Nichols L, Teale EA, et al. Development and validation of an electronic frailty index using routine primary care electronic health record data. Age Ageing. 2018;47(2):319. Kim DH, Glynn RJ, Avorn J, Lipsitz LA, Rockwood K, Pawar A, Schneeweiss S. Validation of a Claims-Based Frailty Index Against Physical Performance and Adverse Health Outcomes in the Health and Retirement Study. J Gerontol Biol Sci Med Sci. 2019;74(8):1271–6. Rich MW, Chyun DA, Skolnick AH, Alexander KP, Forman DE, Kitzman DW, et al. Knowledge Gaps in Cardiovascular Care of the Older Adult Population: A Scientific Statement From the American Heart Association, American College of Cardiology, and American Geriatrics Society. Circulation. 2016;133(21):2103–22. Forman DE, Maurer MS, Boyd C, Brindis R, Salive ME, Horne FM, et al. Multimorbidity in Older Adults With Cardiovascular Disease. J Am Coll Cardiol. 2018;71(19):2149–61. Temtem M, Mendonca MI, Gomes Serrao M, Santos M, Sa D, Sousa F, et al. Predictive improvement of adding coronary calcium score and a genetic risk score to a traditional risk model for cardiovascular event prediction. Eur J Prev Cardiol. 2024;31(6):709–15. Hoogendijk EO, Afilalo J, Ensrud KE, Kowal P, Onder G, Fried LP. Frailty: implications for clinical practice and public health. Lancet. 2019;394(10206):1365–75. Dent E, Martin FC, Bergman H, Woo J, Romero-Ortuno R, Walston JD. Management of frailty: opportunities, challenges, and future directions. Lancet. 2019;394(10206):1376–86. Ng TP, Feng L, Nyunt MS, Feng L, Niti M, Tan BY, et al. Nutritional, Physical, Cognitive, and Combination Interventions and Frailty Reversal Among Older Adults: A Randomized Controlled Trial. Am J Med. 2015;128(11):1225–e361. Walston JD, Bandeen-Roche K. Frailty: a tale of two concepts. BMC Med. 2015;13:185. Tables Table 1 to 3 are available in the Supplementary Files section. Additional Declarations No competing interests reported. Supplementary Files Table1.docx Table 1. Baseline characteristics of the study population stratified by tertiles of the Athero-Frailty Score (AFS). Table2.docx Table 2. Hazard ratios (95% CIs) for incident stroke associated with cumAIP and the novel Athero-Frailty Score (AFS). Table3.docx Table 3. Incremental predictive value of adding AFS to the baseline risk model for stroke. TableS4.docx TableS1.docx TableS2.docx TableS3.docx FigureS2.tif FigureS4.tif FigureS3.tif FigureS1.tif SupplementaryInfo.docx Cite Share Download PDF Status: Published Journal Publication published 31 Mar, 2026 Read the published version in Lipids in Health and Disease → Version 1 posted Editorial decision: Revision requested 23 Feb, 2026 Reviews received at journal 16 Feb, 2026 Reviewers agreed at journal 16 Feb, 2026 Reviews received at journal 16 Feb, 2026 Reviewers agreed at journal 16 Feb, 2026 Reviewers agreed at journal 09 Feb, 2026 Reviewers invited by journal 09 Feb, 2026 Editor assigned by journal 04 Feb, 2026 Submission checks completed at journal 04 Feb, 2026 First submitted to journal 04 Feb, 2026 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-8788634","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":588698986,"identity":"5c3d03a2-9ab4-4e6b-971e-e587b593666d","order_by":0,"name":"jingxian sun","email":"","orcid":"","institution":"Affiliated Hospital of Qingdao University","correspondingAuthor":false,"prefix":"","firstName":"jingxian","middleName":"","lastName":"sun","suffix":""},{"id":588698987,"identity":"af3625c1-e1de-4a2d-a7c9-e6bee48f6617","order_by":1,"name":"Daikang Xu","email":"","orcid":"","institution":"Affiliated Hospital of Qingdao University","correspondingAuthor":false,"prefix":"","firstName":"Daikang","middleName":"","lastName":"Xu","suffix":""},{"id":588698989,"identity":"3547c229-362d-44d2-9fab-79115817f9b8","order_by":2,"name":"Guangyu Du","email":"","orcid":"","institution":"Affiliated Hospital of Qingdao University","correspondingAuthor":false,"prefix":"","firstName":"Guangyu","middleName":"","lastName":"Du","suffix":""},{"id":588698990,"identity":"bb09d21e-ee60-4928-8aa2-f3eef2617c52","order_by":3,"name":"Tongyu Jia","email":"","orcid":"","institution":"Affiliated Hospital of Qingdao University","correspondingAuthor":false,"prefix":"","firstName":"Tongyu","middleName":"","lastName":"Jia","suffix":""},{"id":588698991,"identity":"65a15b88-77cc-4e7c-a672-795efe69fd69","order_by":4,"name":"Xing Han","email":"","orcid":"","institution":"Affiliated Hospital of Qingdao University","correspondingAuthor":false,"prefix":"","firstName":"Xing","middleName":"","lastName":"Han","suffix":""},{"id":588698992,"identity":"89adab07-e5a4-4b8b-bb39-9603d00661d0","order_by":5,"name":"Yi Yu","email":"","orcid":"","institution":"Affiliated Hospital of Qingdao University","correspondingAuthor":false,"prefix":"","firstName":"Yi","middleName":"","lastName":"Yu","suffix":""},{"id":588698993,"identity":"2bd594b7-9d59-4149-b0d8-0d1a01b6dc21","order_by":6,"name":"Jianpeng Wang","email":"","orcid":"","institution":"Affiliated Hospital of Qingdao University","correspondingAuthor":false,"prefix":"","firstName":"Jianpeng","middleName":"","lastName":"Wang","suffix":""},{"id":588698994,"identity":"52ed2243-e164-4996-b62b-7663da500ef5","order_by":7,"name":"Zhiyong Yan","email":"","orcid":"","institution":"Affiliated Hospital of Qingdao University","correspondingAuthor":false,"prefix":"","firstName":"Zhiyong","middleName":"","lastName":"Yan","suffix":""},{"id":588698995,"identity":"8b95aad5-4ecf-4508-af68-2812808838f9","order_by":8,"name":"Shifang Li","email":"","orcid":"","institution":"Affiliated Hospital of Qingdao University","correspondingAuthor":false,"prefix":"","firstName":"Shifang","middleName":"","lastName":"Li","suffix":""},{"id":588698996,"identity":"befb56fc-b08c-45f7-8f6a-b8c337ea9436","order_by":9,"name":"Chao Wang","email":"","orcid":"","institution":"Affiliated Hospital of Qingdao University","correspondingAuthor":false,"prefix":"","firstName":"Chao","middleName":"","lastName":"Wang","suffix":""},{"id":588698997,"identity":"2a8355fd-9da4-45a3-8ba9-b16e757a452c","order_by":10,"name":"Shusheng Che","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA+0lEQVRIiWNgGAWjYDACCQYGZhDNz8zYcOADhG1AnBbJduaDD2eQpMWgny3ZmIcYLfyze4w/F1TcsdvAzGMmbVNmndjA3rxNgqHmDm5L7pwxk55x5lnydpCWnHPpiQ08x8okGI49w6nFQCLHjJm37XCyZTNQS27b4cQGoIgEY8NhfFqMP/P+O5xscBioxRKkRf4NQS0G0rwNh+0MDgO9zwi2hQe/FokbaWXSPMcOJ0g2AwO551y6cRtPWrFFwjHcWvhnJG/+zFNz2J6f/2DDgR9l1rL97Ic33vhQg1sLDCQ2gCk2MGJgSCCogYHBngGmZRSMglEwCkYBOgAAbw1R5JerhyAAAAAASUVORK5CYII=","orcid":"","institution":"Affiliated Hospital of Qingdao University","correspondingAuthor":true,"prefix":"","firstName":"Shusheng","middleName":"","lastName":"Che","suffix":""}],"badges":[],"createdAt":"2026-02-04 16:10:24","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8788634/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8788634/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1186/s12944-026-02931-4","type":"published","date":"2026-03-31T15:58:37+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":102442203,"identity":"bd7ba12e-721f-4222-881d-99a90b82c54f","added_by":"auto","created_at":"2026-02-11 17:03:28","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":204289,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eDose-response relationships of cumulative lipid burden and the Athero-Frailty Score with incident stroke. \u003c/strong\u003e(A) Restricted cubic spline (RCS) analysis of the association between cumAIP and stroke risk, illustrating a non-linear pattern with an apparent plateau at higher levels. (B) RCS analysis of the association between AFS and stroke risk, demonstrating an approximately linear dose–response relationship. Solid lines represent estimated hazard ratios, and shaded areas indicate 95% confidence intervals. Histograms along the x-axis display the distribution of participants. Abbreviations: cumAIP, cumulative Atherogenic Index of Plasma; AFS, Athero-Frailty Score; CI, confidence interval.\u003c/p\u003e","description":"","filename":"OnlineFigure1.png","url":"https://assets-eu.researchsquare.com/files/rs-8788634/v1/33d286aeaf9d20d8863d336d.png"},{"id":102442207,"identity":"39a1e6d8-06e7-4e45-bf18-c203bd5a064b","added_by":"auto","created_at":"2026-02-11 17:03:28","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":171064,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eMachine learning-based identification of top predictors for stroke using the XGBoost model. \u003c/strong\u003e(A) Feature importance plot ranking clinical variables by their mean absolute SHAP (SHapley Additive exPlanations) values. The Frailty Index (FI) ranks higher than traditional lipid markers. (B) SHAP summary plot (beeswarm plot) illustrating the impact of feature values on stroke risk. Each dot represents a participant; red indicates high feature values, and blue indicates low values. Positive SHAP values indicate a higher predicted risk of stroke. Abbreviations: FI, Frailty Index; cumAIP, cumulative Atherogenic Index of Plasma; SHAP, SHapley Additive exPlanations.\u003c/p\u003e","description":"","filename":"OnlineFiguer2.png","url":"https://assets-eu.researchsquare.com/files/rs-8788634/v1/d0a3531873890235fd459f36.png"},{"id":102442212,"identity":"6be26900-dd66-4c7c-9a46-cdf4c09d2586","added_by":"auto","created_at":"2026-02-11 17:03:28","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":112568,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSubgroup analysis of the association between AFS and incident stroke.\u003c/strong\u003e Forest plot displaying the hazard ratios (per 1-SD increase in AFS) for stroke risk across prespecified clinical subgroups. The vertical line indicates the reference HR of 1.0. The association between AFS and stroke remained consistent across all strata, with no significant interactions observed (P for interaction \u0026gt; 0.05). Abbreviations: AFS, Athero-Frailty Score; HR, hazard ratio; CI, confidence interval; BMI, body mass index; SBP, systolic blood pressure.\u003c/p\u003e","description":"","filename":"OnlineFigure3.png","url":"https://assets-eu.researchsquare.com/files/rs-8788634/v1/7d20ca57d8019352343ac768.png"},{"id":106344977,"identity":"3c68bb9b-966f-4249-812c-6e0082e86863","added_by":"auto","created_at":"2026-04-07 16:17:31","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1514348,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8788634/v1/e77c0be9-ac43-4807-a9a3-38efb23cadd6.pdf"},{"id":102442210,"identity":"553ceaa6-06ac-4ed1-8cab-2d4bb7e31879","added_by":"auto","created_at":"2026-02-11 17:03:28","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":24751,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eTable 1. Baseline characteristics of the study population stratified by tertiles of the Athero-Frailty Score (AFS).\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"Table1.docx","url":"https://assets-eu.researchsquare.com/files/rs-8788634/v1/cdbb4510e7c2f3219c39813d.docx"},{"id":105032478,"identity":"b043fdb1-0bb2-4f03-9bd6-481e066f626e","added_by":"auto","created_at":"2026-03-20 06:57:37","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":54791,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eTable 2. Hazard ratios (95% CIs) for incident stroke associated with cumAIP and the novel Athero-Frailty Score (AFS).\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"Table2.docx","url":"https://assets-eu.researchsquare.com/files/rs-8788634/v1/0c2aaa8269cdbd13b86511ef.docx"},{"id":102442215,"identity":"ff2a908b-e537-4c4b-8753-c57cc308c4fe","added_by":"auto","created_at":"2026-02-11 17:03:28","extension":"docx","order_by":3,"title":"","display":"","copyAsset":false,"role":"supplement","size":14829,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eTable 3. Incremental predictive value of adding AFS to the baseline risk model for stroke.\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"Table3.docx","url":"https://assets-eu.researchsquare.com/files/rs-8788634/v1/b39209ce44c9772c378e29c5.docx"},{"id":102746182,"identity":"90499c7a-6eff-4d53-a203-b50b3b6d8666","added_by":"auto","created_at":"2026-02-16 08:56:01","extension":"docx","order_by":4,"title":"","display":"","copyAsset":false,"role":"supplement","size":16032,"visible":true,"origin":"","legend":"","description":"","filename":"TableS4.docx","url":"https://assets-eu.researchsquare.com/files/rs-8788634/v1/59f124bd1b61ee79c489e47b.docx"},{"id":102442224,"identity":"14d0ad41-ba6e-4e9c-a49b-2cb9c22ad03f","added_by":"auto","created_at":"2026-02-11 17:03:32","extension":"docx","order_by":5,"title":"","display":"","copyAsset":false,"role":"supplement","size":24635,"visible":true,"origin":"","legend":"","description":"","filename":"TableS1.docx","url":"https://assets-eu.researchsquare.com/files/rs-8788634/v1/9f315f83c270cbe2d8a278e9.docx"},{"id":102745830,"identity":"ce97de6a-83a1-421e-87ac-657e9aeb1cf6","added_by":"auto","created_at":"2026-02-16 08:54:14","extension":"docx","order_by":6,"title":"","display":"","copyAsset":false,"role":"supplement","size":54664,"visible":true,"origin":"","legend":"","description":"","filename":"TableS2.docx","url":"https://assets-eu.researchsquare.com/files/rs-8788634/v1/088e78d747d316e99798c101.docx"},{"id":102745576,"identity":"3e412a25-c5f1-440d-9aed-9c0f20493dfd","added_by":"auto","created_at":"2026-02-16 08:52:00","extension":"docx","order_by":7,"title":"","display":"","copyAsset":false,"role":"supplement","size":51734,"visible":true,"origin":"","legend":"","description":"","filename":"TableS3.docx","url":"https://assets-eu.researchsquare.com/files/rs-8788634/v1/2e27bed317b0809116cb1b7c.docx"},{"id":102745690,"identity":"c0df9ba9-a410-4899-9607-34d2a19c3938","added_by":"auto","created_at":"2026-02-16 08:53:22","extension":"tif","order_by":8,"title":"","display":"","copyAsset":false,"role":"supplement","size":1413036,"visible":true,"origin":"","legend":"","description":"","filename":"FigureS2.tif","url":"https://assets-eu.researchsquare.com/files/rs-8788634/v1/d75a11355811bfcc1825d66c.tif"},{"id":102442225,"identity":"4267a378-531a-4271-9f3d-3e048ce2bd77","added_by":"auto","created_at":"2026-02-11 17:03:32","extension":"tif","order_by":9,"title":"","display":"","copyAsset":false,"role":"supplement","size":1539504,"visible":true,"origin":"","legend":"","description":"","filename":"FigureS4.tif","url":"https://assets-eu.researchsquare.com/files/rs-8788634/v1/b12a877920abb580f70340e1.tif"},{"id":102442218,"identity":"5a671062-04f5-4a8b-a35f-d9055e46b06a","added_by":"auto","created_at":"2026-02-11 17:03:28","extension":"tif","order_by":10,"title":"","display":"","copyAsset":false,"role":"supplement","size":3288160,"visible":true,"origin":"","legend":"","description":"","filename":"FigureS3.tif","url":"https://assets-eu.researchsquare.com/files/rs-8788634/v1/1989d509be00e2da813f5a62.tif"},{"id":102442223,"identity":"e86105e0-9f84-4160-a2d7-9c125bcca66a","added_by":"auto","created_at":"2026-02-11 17:03:32","extension":"tif","order_by":11,"title":"","display":"","copyAsset":false,"role":"supplement","size":18015276,"visible":true,"origin":"","legend":"","description":"","filename":"FigureS1.tif","url":"https://assets-eu.researchsquare.com/files/rs-8788634/v1/fba77e14f2897002b6fa9928.tif"},{"id":102746184,"identity":"9bbf272d-2f7e-47d0-97fa-217ea60e9de7","added_by":"auto","created_at":"2026-02-16 08:56:02","extension":"docx","order_by":12,"title":"","display":"","copyAsset":false,"role":"supplement","size":14705,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryInfo.docx","url":"https://assets-eu.researchsquare.com/files/rs-8788634/v1/422e7519b3280da5cc8d219a.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Integrating Frailty and Cumulative Lipid Burden for Stroke Risk Stratification: A Machine Learning–Guided Athero -Frailty Score From the CHARLS Cohort","fulltext":[{"header":"Introduction","content":"\u003cp\u003eStroke imposes substantial mortality and long-term disability worldwide, and improving the identification of individuals at elevated risk remains a central goal of primary prevention [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e, \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e]. Dyslipidemia is a well-established contributor to atherosclerotic disease, yet lipid measurements obtained at a single time point may not fully capture an individual\u0026rsquo;s long-term exposure to atherogenic lipids[\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]. To better approximate cumulative metabolic burden, repeated-measure constructs such as the atherogenic index of plasma (AIP; log[TG/HDL-C]) have been proposed[\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e]. By incorporating information across multiple assessments, cumulative or time-averaged lipid metrics have been suggested to improve risk stratification compared with baseline measurements alone[\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eRisk assessment in older adults, however, presents additional challenges. Medication use is highly prevalent in aging populations[\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e], and antihypertensive or lipid-lowering therapies can substantially modify circulating biomarker levels. As a result, measured metabolic indices may no longer fully reflect the underlying burden of vascular injury accumulated over time, a phenomenon increasingly recognized in geriatric cardiology[\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e, \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e]. Meanwhile, data from major outcomes trials\u0026mdash;such as REDUCE-IT and PROMINENT\u0026mdash;indicate that vascular events continue to occur at high rates despite achieving guideline-recommended treatment targets or biomarker reduction[\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e, \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e]. Furthermore, large-scale meta-analyses suggest that the relative risk reduction conferred by statin therapy may be attenuated in older adults compared with younger populations [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]. Together, these observations support the \"residual risk\" hypothesis[\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e] and raise the possibility that reliance on metabolic markers alone is insufficient for stroke risk stratification in older, treated individuals[\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eFrailty represents a critical, yet underutilized, dimension of residual risk. Conceptualized as multisystem vulnerability and reduced resilience to physiological stressors[\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e], frailty is driven by \"inflammaging\"\u0026mdash;chronic, low-grade inflammation that promotes both atherogenesis and plaque instability[\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e]. The frailty index (FI), based on the deficit accumulation framework, summarizes impairments in chronic conditions, functional status, cognition, and psychological health into a single continuous measure of global deficit burden[\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e]. A growing body of evidence has linked frailty to adverse cardiovascular outcomes, including incident stroke, independent of chronological age and traditional risk factors[\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e, \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e]. Importantly, frailty burden correlates poorly with traditional lipid markers, suggesting that it captures complementary information related to systemic vulnerability that is not reflected by the lipidome alone[\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eIn this study, we leveraged data from the large, community-based China Health and Retirement Longitudinal Study (CHARLS)[\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e] to develop and validate a novel integrated risk metric. Using an exploratory machine-learning approach to guide variable selection, we constructed the Athero-Frailty Score (AFS) as a parsimonious composite of cumulative lipid burden and frailty index. We evaluated the dose\u0026ndash;response relationship between AFS and incident stroke and quantified its incremental predictive utility over standard clinical risk factors. Our primary objective was to determine whether this multidimensional construct can overcome the limitations of traditional metabolic markers and enhance risk stratification in older adults.\u003c/p\u003e"},{"header":"Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eStudy Population\u003c/h2\u003e \u003cp\u003eThis study analyzed data from the China Health and Retirement Longitudinal Study (CHARLS), a nationally representative longitudinal cohort of residents aged 45 years and older. The study design and sampling procedures have been described in detail previously. The baseline survey was conducted in 2011 (Wave 1), with biennial follow-up assessments in 2013 (Wave 2), 2015 (Wave 3), and 2018 (Wave 4).For the current analysis, participants were initially screened if they had complete biomarker data in both 2011 and 2015 to allow for the calculation of cumulative indices. The exclusion criteria were as follows: (1) a self-reported doctor diagnosis of stroke or cardiovascular disease (CVD) at or before the 2015 baseline; (2) missing data for variables required for the Frailty Index (FI) construction; or (3) loss to follow-up during the outcome ascertainment period (2015\u0026ndash;2018). The final analytic cohort consisted of 3,690 participants. The study protocol was approved by the Ethical Review Committee of Peking University, and written informed consent was obtained from all participants.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eAssessment of Cumulative AIP (cumAIP)\u003c/h3\u003e\n\u003cp\u003eBlood samples were collected after an overnight fast. Triglycerides (TG) and High-density lipoprotein cholesterol (HDL-C) were measured using enzymatic colorimetric tests. The Atherogenic Index of Plasma (AIP) was calculated as log(TG / HDL-C).To capture the long-term burden, cumAIP was calculated as the weighted sum of the average AIP between consecutive waves (2011\u0026ndash;2015), multiplied by the time interval (years). The formula was defined as:\u003c/p\u003e \u003cp\u003ecumAIP\u0026thinsp;=\u0026thinsp;Σ [(AIP_2011\u0026thinsp;+\u0026thinsp;AIP_2015) / 2 \u0026times; (2015\u0026thinsp;\u0026minus;\u0026thinsp;2011)].\u003c/p\u003e\n\u003ch3\u003eAssessment of Frailty Index (FI)\u003c/h3\u003e\n\u003cp\u003eThe Frailty Index (FI) was constructed using the standard deficit-accumulation approach. A total of 32 health deficits were included, encompassing multiple domains: chronic comorbidities (e.g., hypertension, diabetes, dyslipidemia), functional limitations (Activities of Daily Living [ADL] and Instrumental ADL), cognitive function, and depressive symptoms. Each deficit was dichotomized or graded, mapping to a score between 0 (absence of deficit) and 1 (presence of deficit). The FI score was calculated as the ratio of the number of deficits present to the total number of deficits considered, yielding a continuous variable ranging from 0 to 1.\u003c/p\u003e\n\u003ch3\u003eConstruction of the Athero-Frailty Score (AFS)\u003c/h3\u003e\n\u003cp\u003eTo jointly capture cumulative metabolic burden and global deficit burden, we constructed the Athero-Frailty Score (AFS). Cumulative atherogenic lipid burden was quantified using the cumulative atherogenic index of plasma (cumAIP), derived from repeated AIP measurements between 2011 and 2015, while frailty burden was assessed at baseline (2015) using a 32-item frailty index (FI) based on the deficit accumulation approach. To place the two components on a comparable scale, both cumAIP and FI were standardized to z-scores. AFS was then defined as a weighted linear combination of these standardized variables, with weights (β coefficients) obtained from multivariable Cox proportional hazards models: AFS\u0026thinsp;=\u0026thinsp;β₁ \u0026times; Z(cumAIP) + β₂ \u0026times; Z(FI). AFS was analyzed as a continuous variable and, for descriptive and stratified analyses, further categorized into tertiles based on its distribution in the study population.\u003c/p\u003e \u003cp\u003eAn additive formulation was prespecified for conceptual and methodological reasons. Conceptually, cumAIP and FI represent complementary but distinct dimensions of stroke risk\u0026mdash;metabolic lipid burden and multisystem vulnerability, respectively\u0026mdash;and an additive model provides a parsimonious and interpretable approach to integrating these domains in a prediction-focused framework. Methodologically, multiplicative or higher-order combinations (e.g., cumAIP \u0026times; FI) can generate highly skewed distributions and disproportionately amplify extreme observations, potentially leading to model instability and overfitting in finite samples. In addition, selecting a functional form after exploring multiple nonlinear combinations may introduce model-selection optimism and inflate apparent predictive performance. To minimize these risks and improve interpretability and transportability, we specified an additive linear composite as the primary model a priori. Potential nonlinearity and interaction between cumAIP and FI were examined in sensitivity analyses.\u003c/p\u003e\n\u003ch3\u003eOutcome Assessment\u003c/h3\u003e\n\u003cp\u003eThe primary outcome was incident stroke, defined as the first occurrence of a stroke event during the follow-up period (2015\u0026ndash;2018). Stroke events were ascertained based on self-reported doctor diagnoses (\"Have you been diagnosed with stroke by a doctor?\") during the 2018 wave. Follow-up time was calculated from the date of the 2015 interview to the date of stroke diagnosis or the date of the 2018 interview (censored), whichever occurred first.\u003c/p\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003eCovariates\u003c/h2\u003e \u003cp\u003eCovariates were collected at baseline (2015) and included:Demographic: Age, sex, marital status, education level, and residence (urban/rural).Lifestyle: Smoking status (current/former/never) and alcohol consumption.Clinical: Systolic blood pressure (SBP), fasting plasma glucose (FPG), C-reactive protein (CRP), estimated glomerular filtration rate (eGFR), blood urea nitrogen (BUN), and body mass index (BMI).Medication History: Self-reported use of antihypertensive, antidiabetic, and lipid-lowering agents.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec9\" class=\"Section2\"\u003e \u003ch2\u003eStatistical Analysis\u003c/h2\u003e \u003cp\u003eBaseline characteristics were compared across AFS tertiles using ANOVA, Kruskal-Wallis tests, or Chi-square tests.Machine Learning Screening: To contextualize predictor importance, we employed the XGBoost algorithm. The SHAP (SHapley Additive exPlanations) method was used to interpret the model and rank variables based on their global contribution to stroke prediction.Survival Analysis: Cox proportional hazards models were used to estimate Hazard Ratios (HRs) and 95% Confidence Intervals (CIs). Model 1 was unadjusted; Model 2 adjusted for age and sex; Model 3 further adjusted for clinical biomarkers; and Model 4 (fully adjusted) included medication history.Linearity and Incremental Value: Restricted Cubic Splines (RCS) were used to visualize the dose-response relationship between AFS and stroke risk. The incremental predictive value of AFS over cumAIP was assessed using the C-index, Net Reclassification Improvement (NRI), and Integrated Discrimination Improvement (IDI).All statistical analyses were performed using R software (version 4.5.2). A two-sided P-value\u0026thinsp;\u0026lt;\u0026thinsp;0.05 was considered statistically significant.\u003c/p\u003e \u003c/div\u003e"},{"header":"Results","content":"\u003cp\u003e\u003cstrong\u003ePredictive Limitations of cumAIP and the Association With Medication Use\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe final analytic cohort consisted of 3,690 participants, with baseline demographic and clinical characteristics stratified by AFS categories presented in Table 1. Regarding lipid burden, initial stratification by cumAIP tertiles confirmed a positive association with stroke risk, with participants in the highest tertile (T3) exhibiting a higher incidence compared with those in the lowest tertile (Table S1). In unadjusted and minimally adjusted models, cumAIP was a significant predictor of incident stroke (Table 2). However, with further adjustment for clinical covariates and specifically medication use (including antihypertensive and lipid-lowering therapies), the association between cumAIP and stroke risk was progressively attenuated. In the fully adjusted model (Model 4), the independent association of cumAIP with stroke risk was modest and of borderline statistical significance (HR=1.20, 95% CI =1.01\u0026ndash;1.43, P = 0.042; Table 2). Restricted cubic spline (RCS) analyses further demonstrated a non-linear association, characterized by an apparent plateau at higher cumAIP levels (Figure 1A). Together, these findings suggest that cumulative lipid burden alone may have limited discriminatory capacity for stroke risk stratification in the context of widespread medication use.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eExploratory Machine-Learning Analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo contextualize the relative contributions of candidate predictors, we applied an exploratory XGBoost model with SHAP value interpretation. In this model, the Frailty Index (FI) demonstrated a larger average SHAP contribution to stroke prediction than cumAIP, eGFR, and medication indicators (Figure 2 and S2). Correlation analyses showed minimal correlation between FI and cumAIP (\u003cem\u003er\u0026nbsp;\u003c/em\u003e= 0.04; Figure S2), indicating that the two measures capture distinct, orthogonal dimensions of risk. As prespecified, this machine-learning analysis was utilized strictly for hypothesis generation to inform the selection of variables for the composite score, rather than for final model specification.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAssociation Between the Athero-Frailty Score and Incident Stroke\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eBaseline characteristics stratified by AFS tertiles are summarized in Table 2. Participants in the highest AFS tertile were older, more likely to be female, and exhibited both a less favorable metabolic profile and substantially higher frailty burden compared with those in the lowest tertile. Notably, AFS stratification revealed a pronounced gradient in frailty burden that was not apparent when stratifying by cumAIP alone. In multivariable Cox models, AFS was strongly and independently associated with incident stroke (Table 1). In the fully adjusted model, a 1-SD increase in AFS was associated with a 2.75-fold increased risk of stroke (P \u0026lt; 0.001). Unlike cumAIP, RCS analyses for AFS demonstrated an approximately linear dose\u0026ndash;response relationship across its distribution, with no evidence of a plateau (Figure 1B). Kaplan\u0026ndash;Meier analyses showed clear and early separation of cumulative stroke incidence across AFS tertiles (log-rank P \u0026lt; 0.001), whereas separation by cumAIP tertiles was less pronounced (Figure S3).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eIncremental Predictive Performance and Sensitivity Analyses\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe addition of AFS to the baseline risk model was associated with improved model discrimination, increasing the C-statistic from 0.681 to 0.706 (P \u0026lt; 0.001; Figure S4). Notably, AFS yielded a substantial and statistically significant improvement in Net Reclassification Improvement (NRI = 15.4%, P \u0026lt; 0.001), indicating that the score correctly reclassified a significant proportion of individuals into appropriate risk categories. The Integrated Discrimination Improvement (IDI) showed a positive trend but was modest (0.007, P = 0.088). In contrast, the addition of cumAIP alone did not result in a statistically significant improvement in reclassification (NRI = 11.1%, P = 0.076). The association between AFS and stroke risk was consistent across subgroups stratified by age, sex, and hypertension status, with no significant interactions observed (P_interaction \u0026gt; 0.05; Figure 3). Crucially, sensitivity analyses were conducted to compare the additive AFS specification against alternative strategies, including a multiplicative product score (cumAIP\u0026times;FI) and an explicit multiplicative interaction term. Although the product term was associated with stroke when modeled alone, it did not improve discrimination or reclassification compared with the additive AFS, and adding the interaction term to the model provided no incremental predictive value (Table S3). Consistently, the multiplicative interaction analysis did not show strong evidence of synergy beyond main effects, supporting the robustness and parsimony of the additive AFS specification (Table S4).\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eIn this prcospective cohort of community-dwelling middle-aged and older adults, we evaluated whether integrating cumulative atherogenic lipid burden with frailty burden improves short-term stroke risk stratification. Three main findings emerged. First, although cumulative AIP was positively associated with incident stroke, this association was attenuated after adjustment for medication use and exhibited a distinct non-linear pattern with a plateau at higher levels, corroborating recent observations regarding lipid variability and risk plateauing[\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e, \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e]. Second, frailty burden, as assessed by the Frailty Index, provided complementary risk information that was largely orthogonal to cumulative lipid burden. Third, a parsimonious additive composite integrating cumAIP and FI\u0026mdash;the Athero-Frailty Score (AFS)\u0026mdash;demonstrated a robust linear association with stroke risk and significantly improved discrimination and integrated discrimination, with more modest improvement in reclassification beyond standard risk factors.\u003c/p\u003e \u003cp\u003eThe observed attenuation of the association between cumulative lipid burden and stroke risk after accounting for medication use is consistent with the \"medication paradox\" or \"reverse epidemiology\" often seen in aging populations[\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e, \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e]. In older adults, biomarker levels measured during treatment may not fully reflect lifetime exposure or the extent of underlying vascular injury[\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e]. Moreover, individuals receiving antihypertensive or lipid-lowering therapies often represent a higher baseline risk group, confounding interpretation in observational settings\u0026mdash;a phenomenon known as \"confounding by indication\"[\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e]. Our restricted cubic spline analyses suggest that cumulative AIP alone has limited incremental discriminatory capacity at higher exposure levels once these clinical factors are taken into account. This aligns with data from the PROSPER trial and other geriatric cohorts suggesting that the predictive strength of traditional lipid markers diminishes with age[\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e, \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e], potentially due to the competing risks of non-cardiovascular mortality and the catabolic effects of chronic disease[\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e]. Importantly, these findings reflect the challenges of risk stratification in medicated populations rather than a lack of biological relevance of lipids[\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eFrailty captures multisystem vulnerability and reduced physiological reserve, integrating information across comorbidities, functional limitations, and psychosocial domains. Consistent with previous studies linking frailty to adverse cardiovascular outcomes[\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e, \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e], FI showed a strong association with incident stroke in our analyses. Mechanistically, this link may be mediated by shared pathways such as oxidative stress, mitochondrial dysfunction, and dysregulated coagulation, which are hallmarks of both frailty and cerebrovascular disease [\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e, \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e]. The weak correlation between FI and cumAIP observed in our study suggests that frailty burden and lipid burden represent distinct pathophysiological axes[\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e]. From a prediction perspective, this orthogonality provides the biological rationale for their joint consideration. Rather than replacing metabolic markers, frailty assessment captures the \"additional vulnerability\"\u0026mdash;or residual risk\u0026mdash;related to biological aging that lipid indices miss [\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eA critical methodological insight from our study concerns the structural relationship between metabolic and functional risk. While biological systems often exhibit complex interactions, our rigorous sensitivity analyses comparing additive versus multiplicative specifications provided strong support for an additive accumulation model. The inclusion of multiplicative interaction terms (cumAIP \u0026times; FI) failed to provide incremental improvement in discrimination or reclassification compared with the simpler additive AFS. This finding has important conceptual implications: it suggests that metabolic atherogenicity and systemic frailty may operate as parallel, largely independent risk vectors rather than synergistic amplifiers[\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e]. By summing these independent deficits, AFS captures the \"total physiological load\" in accordance with the principle of parsimony[\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e], minimizing the risk of overfitting inherent in complex interaction models while maximizing interpretability and transportability across populations[\u003cspan additionalcitationids=\"CR40\" citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e]. This additive approach parallels the construction of polygenic risk scores, where multiple small-effect variants are summed to estimate aggregate liability[\u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e42\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eFrom a clinical perspective, our findings suggest that stroke risk assessment in older adults may benefit from incorporating information on global health status alongside traditional metabolic markers. Frailty assessment is increasingly feasible in both research and clinical settings, including through electronic health record\u0026ndash;based indices[\u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e43\u003c/span\u003e, \u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e44\u003c/span\u003e]. An integrated score such as AFS may assist in identifying individuals at elevated short-term risk who might not be fully captured by lipid measures alone, aligning with recent AHA/ACC recommendations to incorporate geriatric assessment into cardiovascular care[\u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e45\u003c/span\u003e, \u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e46\u003c/span\u003e]. However, the observed improvements in predictive performance should be interpreted cautiously and do not, by themselves, imply changes to treatment thresholds or management strategies without further interventional evidence[\u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e47\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eFrom a research standpoint, these results underscore the importance of considering multidimensional vulnerability in cardiovascular risk prediction. Future studies should evaluate the performance of AFS in other populations and over longer follow-up periods[\u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e48\u003c/span\u003e], and assess whether incorporating frailty into risk assessment frameworks improves clinical decision-making or patient outcomes via \"frailty-guided\" management strategies[\u003cspan additionalcitationids=\"CR50\" citationid=\"CR49\" class=\"CitationRef\"\u003e49\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e51\u003c/span\u003e].\u003c/p\u003e \u003cdiv id=\"Sec16\" class=\"Section2\"\u003e \u003ch2\u003eStrengths and Limitations\u003c/h2\u003e \u003cp\u003eThis study has several strengths, including the use of a large, nationally representative cohort with repeated biomarker measurements and detailed assessment of frailty. The integration of exploratory machine-learning analyses with hypothesis-driven survival modeling provided complementary perspectives on predictor contributions. Several limitations merit consideration. Stroke events were based on self-reported physician diagnoses, which may introduce misclassification. The follow-up period was relatively short, and residual confounding cannot be excluded. In addition, although we examined multiple sensitivity analyses, external validation in independent cohorts is necessary before broader application.\u003c/p\u003e \u003c/div\u003e"},{"header":"Conclusions","content":"\u003cp\u003eIn this cohort of middle-aged and older adults, integrating cumulative atherogenic lipid burden with frailty burden via a parsimonious additive score was associated with improved short-term stroke risk stratification. Our findings indicate that metabolic and functional deficits contribute independently to vascular risk, supporting a multidimensional \"Athero-Frailty\" framework for prediction in aging populations.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e \u003ch2\u003eCompeting interests\u003c/h2\u003e \u003cp\u003eThe authors declare that they have no competing interests.\u003c/p\u003e \u003c/p\u003e\u003ch2\u003eFunding\u003c/h2\u003e \u003cp\u003eJingxian Sun received financial support from the Youth Research Fund of the Affiliated Hospital of Qingdao University (Grant. No. QDFYQN2023225); Chao Wang received financial support from the Youth Research Fund of the Affiliated Hospital of Qingdao University (Grant. No. QDFYQN2023114).\u003c/p\u003e\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eJS and CW conceived and designed the study. JS, GD and DX performed the statistical analysis and machine learning modeling using R software. XH, YY, and JW contributed to data curation and data cleaning from the CHARLS database. ZY and SL assisted in the interpretation of the results and validation. JS and TJ drafted the original manuscript. CW and SC obtained funding, supervised the study, and critically revised the manuscript for important intellectual content. All authors read and approved the final manuscript.\u003c/p\u003e\u003ch2\u003eAcknowledgement\u003c/h2\u003e\u003cp\u003eWe thank the China Health and Retirement Longitudinal Study (CHARLS) team for providing data and all participants for their contribution. The National School of Development at Peking University provided the data. We sincerely appreciate the YIWANDOU team for their invaluable contribution to data cleaning and processing.\u003c/p\u003e\u003ch2\u003eData Availability\u003c/h2\u003e\u003cp\u003eThe datasets generated and/or analyzed during the current study are available in the CHARLS repository, http://charls.pku.edu.cn/en.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eCollaborators GBDS. Global, regional, and national burden of stroke and its risk factors, 1990\u0026ndash;2019: a systematic analysis for the Global Burden of Disease Study 2019. Lancet Neurol. 2021;20(10):795\u0026ndash;820.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKleindorfer DO, Towfighi A, Chaturvedi S, Cockroft KM, Gutierrez J, Lombardi-Hill D, et al. 2021 Guideline for the Prevention of Stroke in Patients With Stroke and Transient Ischemic Attack: A Guideline From the American Heart Association/American Stroke Association. Stroke. 2021;52(7):e364\u0026ndash;467.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFerence BA, Ginsberg HN, Graham I, Ray KK, Packard CJ, Bruckert E, et al. Low-density lipoproteins cause atherosclerotic cardiovascular disease. 1. Evidence from genetic, epidemiologic, and clinical studies. A consensus statement from the European Atherosclerosis Society Consensus Panel. Eur Heart J. 2017;38(32):2459\u0026ndash;72.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDobiasova M, Frohlich J. The plasma parameter log (TG/HDL-C) as an atherogenic index: correlation with lipoprotein particle size and esterification rate in apoB-lipoprotein-depleted plasma (FER(HDL)). Clin Biochem. 2001;34(7):583\u0026ndash;8.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDomanski MJ, Tian X, Wu CO, Reis JP, Dey AK, Gu Y, et al. Time Course of LDL Cholesterol Exposure and Cardiovascular Disease Event Risk. J Am Coll Cardiol. 2020;76(13):1507\u0026ndash;16.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eQato DM, Wilder J, Schumm LP, Gillet V, Alexander GC. Changes in Prescription and Over-the-Counter Medication and Dietary Supplement Use Among Older Adults in the United States, 2005 vs 2011. JAMA Intern Med. 2016;176(4):473\u0026ndash;82.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYi SW, Yi JJ, Ohrr H. Total cholesterol and all-cause mortality by sex and age: a prospective cohort study among 12.8 million adults. Sci Rep. 2019;9(1):1596.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOfori-Asenso R, Jakhu A, Zomer E, Curtis AJ, Korhonen MJ, Nelson M, et al. Adherence and Persistence Among Statin Users Aged 65 Years and Over: A Systematic Review and Meta-analysis. J Gerontol Biol Sci Med Sci. 2018;73(6):813\u0026ndash;9.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBhatt DL, Steg PG, Miller M, Brinton EA, Jacobson TA, Ketchum SB, et al. Cardiovascular Risk Reduction with Icosapent Ethyl for Hypertriglyceridemia. N Engl J Med. 2019;380(1):11\u0026ndash;22.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDas Pradhan A, Glynn RJ, Fruchart JC, MacFadyen JG, Zaharris ES, Everett BM, et al. Triglyceride Lowering with Pemafibrate to Reduce Cardiovascular Risk. N Engl J Med. 2022;387(21):1923\u0026ndash;34.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCholesterol Treatment Trialists C. Efficacy and safety of statin therapy in older people: a meta-analysis of individual participant data from 28 randomised controlled trials. Lancet. 2019;393(10170):407\u0026ndash;15.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRidker PM. Residual inflammatory risk: addressing the obverse side of the atherosclerosis prevention coin. Eur Heart J. 2016;37(22):1720\u0026ndash;2.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLibby P. The changing landscape of atherosclerosis. Nature. 2021;592(7855):524\u0026ndash;33.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eClegg A, Young J, Iliffe S, Rikkert MO, Rockwood K. Frailty in elderly people. Lancet. 2013;381(9868):752\u0026ndash;62.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFerrucci L, Fabbri E. Inflammageing: chronic inflammation in ageing, cardiovascular disease, and frailty. Nat Rev Cardiol. 2018;15(9):505\u0026ndash;22.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSearle SD, Mitnitski A, Gahbauer EA, Gill TM, Rockwood K. A standard procedure for creating a frailty index. BMC Geriatr. 2008;8:24.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVeronese N, Cereda E, Stubbs B, Solmi M, Luchini C, Manzato E, et al. Risk of cardiovascular disease morbidity and mortality in frail and pre-frail older adults: Results from a meta-analysis and exploratory meta-regression analysis. Ageing Res Rev. 2017;35:63\u0026ndash;73.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEvans NR, Todd OM, Minhas JS, Fearon P, Harston GW, Mant J, et al. Frailty and cerebrovascular disease: Concepts and clinical implications for stroke medicine. Int J Stroke. 2022;17(3):251\u0026ndash;9.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBoehm FJ, Fritzenschaft L, Braig S, Denkinger M, Rothenbacher D, Dallmeier D. Sex- and system-specific analysis of blood-based biomarkers and frailty in older adults-the Activity and Function of the Elderly study. Age Ageing. 2025;54(9).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhao Y, Hu Y, Smith JP, Strauss J, Yang G. Cohort profile: the China Health and Retirement Longitudinal Study (CHARLS). Int J Epidemiol. 2014;43(1):61\u0026ndash;8.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVarbo A, Nordestgaard BG. Remnant cholesterol and risk of ischemic stroke in 112,512 individuals from the general population. Ann Neurol. 2019;85(4):550\u0026ndash;9.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJohannesen CDL, Mortensen MB, Langsted A, Nordestgaard BG. Apolipoprotein B and Non-HDL Cholesterol Better Reflect Residual Risk Than LDL Cholesterol in Statin-Treated Patients. J Am Coll Cardiol. 2021;77(11):1439\u0026ndash;50.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRavnskov U, Diamond DM, Hama R, Hamazaki T, Hammarskjold B, Hynes N, et al. Lack of an association or an inverse association between low-density-lipoprotein cholesterol and mortality in the elderly: a systematic review. BMJ Open. 2016;6(6):e010401.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNissen SE, Menon V, Nicholls SJ, Brennan D, Laffin L, Ridker P, et al. Bempedoic Acid for Primary Prevention of Cardiovascular Events in Statin-Intolerant Patients. JAMA. 2023;330(2):131\u0026ndash;40.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNavar AM, Kolkailah AA, Gupta A, Gillard KK, Israel MK, Wang Y, Peterson ED. Gaps in Guideline-Based Lipid-Lowering Therapy for Secondary Prevention in the United States: A Retrospective Cohort Study of 322 153 Patients. Circ Cardiovasc Qual Outcomes. 2023;16(8):533\u0026ndash;43.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKyriacou DN, Lewis RJ. Confounding by Indication in Clinical Research. JAMA. 2016;316(17):1818\u0026ndash;9.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eShepherd J, Blauw GJ, Murphy MB, Bollen EL, Buckley BM, Cobbe SM, et al. Pravastatin in elderly individuals at risk of vascular disease (PROSPER): a randomised controlled trial. Lancet. 2002;360(9346):1623\u0026ndash;30.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePostmus I, Trompet S, Deshmukh HA, Barnes MR, Li X, Warren HR, et al. Pharmacogenetic meta-analysis of genome-wide association studies of LDL cholesterol response to statins. Nat Commun. 2014;5:5068.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEitmann S, Matrai P, Hegyi P, Balasko M, Eross B, Dorogi K, Petervari E. Obesity paradox in older sarcopenic adults - a delay in aging: A systematic review and meta-analysis. Ageing Res Rev. 2024;93:102164.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMortensen MB, Nordestgaard BG. Elevated LDL cholesterol and increased risk of myocardial infarction and atherosclerotic cardiovascular disease in individuals aged 70\u0026ndash;100 years: a contemporary primary prevention cohort. Lancet. 2020;396(10263):1644\u0026ndash;52.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePasieczna AH, Szczepanowski R, Sobecki J, Katarzyniak R, Uchmanowicz I, Gobbens RJJ, et al. Importance analysis of psychosociological variables in frailty syndrome in heart failure patients using machine learning approach. Sci Rep. 2023;13(1):7782.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFried LP, Tangen CM, Walston J, Newman AB, Hirsch C, Gottdiener J, et al. Frailty in older adults: evidence for a phenotype. J Gerontol Biol Sci Med Sci. 2001;56(3):M146\u0026ndash;56.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWalston J, McBurnie MA, Newman A, Tracy RP, Kop WJ, Hirsch CH, et al. Frailty and activation of the inflammation and coagulation systems with and without clinical comorbidities: results from the Cardiovascular Health Study. Arch Intern Med. 2002;162(20):2333\u0026ndash;41.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eIadecola C, Anrather J. The immunology of stroke: from mechanisms to translation. Nat Med. 2011;17(7):796\u0026ndash;808.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRamsay SE, Arianayagam DS, Whincup PH, Lennon LT, Cryer J, Papacosta AO, et al. Cardiovascular risk profile and frailty in a population-based study of older British men. Heart. 2015;101(8):616\u0026ndash;22.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAfilalo J. Frailty in Patients with Cardiovascular Disease: Why, When, and How to Measure. Curr Cardiovasc Risk Rep. 2011;5(5):467\u0026ndash;72.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKnol MJ, VanderWeele TJ. Recommendations for presenting analyses of effect modification and interaction. Int J Epidemiol. 2012;41(2):514\u0026ndash;20.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSteyerberg EW, Harrell FE Jr., Borsboom GJ, Eijkemans MJ, Vergouwe Y, Habbema JD. Internal validation of predictive models: efficiency of some procedures for logistic regression analysis. J Clin Epidemiol. 2001;54(8):774\u0026ndash;81.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHarrell FE Jr., Lee KL, Mark DB. Multivariable prognostic models: issues in developing models, evaluating assumptions and adequacy, and measuring and reducing errors. Stat Med. 1996;15(4):361\u0026ndash;87.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRen X, Mi Z, Georgopoulos PG. Comparison of Machine Learning and Land Use Regression for fine scale spatiotemporal estimation of ambient air pollution: Modeling ozone concentrations across the contiguous United States. Environ Int. 2020;142:105827.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKundi H, Wadhera RK, Strom JB, Valsdottir LR, Shen C, Kazi DS, Yeh RW. Association of Frailty With 30-Day Outcomes for Acute Myocardial Infarction, Heart Failure, and Pneumonia Among Elderly Adults. JAMA Cardiol. 2019;4(11):1084\u0026ndash;91.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKhera AV, Emdin CA, Drake I, Natarajan P, Bick AG, Cook NR, et al. Genetic Risk, Adherence to a Healthy Lifestyle, and Coronary Disease. N Engl J Med. 2016;375(24):2349\u0026ndash;58.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eClegg A, Bates C, Young J, Ryan R, Nichols L, Teale EA, et al. Development and validation of an electronic frailty index using routine primary care electronic health record data. Age Ageing. 2018;47(2):319.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKim DH, Glynn RJ, Avorn J, Lipsitz LA, Rockwood K, Pawar A, Schneeweiss S. Validation of a Claims-Based Frailty Index Against Physical Performance and Adverse Health Outcomes in the Health and Retirement Study. J Gerontol Biol Sci Med Sci. 2019;74(8):1271\u0026ndash;6.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRich MW, Chyun DA, Skolnick AH, Alexander KP, Forman DE, Kitzman DW, et al. Knowledge Gaps in Cardiovascular Care of the Older Adult Population: A Scientific Statement From the American Heart Association, American College of Cardiology, and American Geriatrics Society. Circulation. 2016;133(21):2103\u0026ndash;22.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eForman DE, Maurer MS, Boyd C, Brindis R, Salive ME, Horne FM, et al. Multimorbidity in Older Adults With Cardiovascular Disease. J Am Coll Cardiol. 2018;71(19):2149\u0026ndash;61.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTemtem M, Mendonca MI, Gomes Serrao M, Santos M, Sa D, Sousa F, et al. Predictive improvement of adding coronary calcium score and a genetic risk score to a traditional risk model for cardiovascular event prediction. Eur J Prev Cardiol. 2024;31(6):709\u0026ndash;15.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHoogendijk EO, Afilalo J, Ensrud KE, Kowal P, Onder G, Fried LP. Frailty: implications for clinical practice and public health. Lancet. 2019;394(10206):1365\u0026ndash;75.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDent E, Martin FC, Bergman H, Woo J, Romero-Ortuno R, Walston JD. Management of frailty: opportunities, challenges, and future directions. Lancet. 2019;394(10206):1376\u0026ndash;86.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNg TP, Feng L, Nyunt MS, Feng L, Niti M, Tan BY, et al. Nutritional, Physical, Cognitive, and Combination Interventions and Frailty Reversal Among Older Adults: A Randomized Controlled Trial. Am J Med. 2015;128(11):1225\u0026ndash;e361.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWalston JD, Bandeen-Roche K. Frailty: a tale of two concepts. BMC Med. 2015;13:185.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"},{"header":"Tables","content":"\u003cp\u003eTable 1 to 3 are available in the Supplementary Files section.\u003c/p\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"lipids-in-health-and-disease","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"lhad","sideBox":"Learn more about [Lipids in Health and Disease](http://lipidworld.biomedcentral.com/)","snPcode":"12944","submissionUrl":"https://submission.nature.com/new-submission/12944/3","title":"Lipids in Health and Disease","twitterHandle":"@BioMedCentral","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"BMC/SO AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Stroke, Athero-Frailty Score, cumAIP, Frailty Index, Machine Learning","lastPublishedDoi":"10.21203/rs.3.rs-8788634/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8788634/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBackground\u003c/h2\u003e \u003cp\u003eTraditional lipid indices, such as the cumulative Atherogenic Index of Plasma (cumAIP), demonstrate inconsistent predictive performance for stroke in older adults, likely due to the modifying effects of complex medication regimens. We hypothesized that incorporating the Frailty Index (FI)\u0026mdash;a measure of cumulative physiological deficits\u0026mdash;could capture the residual risk missed by lipid markers alone. We developed and evaluated a novel Athero-Frailty Score (AFS) to address these limitations.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e \u003cp\u003eWe analyzed 3,690 participants from the China Health and Retirement Longitudinal Study (CHARLS) using a landmark analysis design (baseline: 2015). An exploratory XGBoost model with SHAP analysis was used for variable screening. Based on the orthogonality of lipid and frailty metrics, AFS was constructed as an additive composite score of cumAIP and FI. Associations with incident stroke were assessed using Cox proportional hazards models and restricted cubic splines (RCS). Clinical utility was evaluated via C-statistics, Net Reclassification Improvement (NRI), and Integrated Discrimination Improvement (IDI).\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e \u003cp\u003eSHAP analysis identified FI as having a larger average SHAP contribution than cumAIP and medication indicators. Survival analyses suggested that the association between cumAIP and stroke was attenuated after adjustment for antihypertensive and lipid-lowering therapies (HR 1.20, P\u0026thinsp;=\u0026thinsp;0.042), and RCS indicated a non-linear pattern with an apparent plateau at higher cumAIP levels. In contrast, AFS was associated with an approximately linear dose\u0026ndash;response relationship independent of medication use. In the fully adjusted model, each 1-SD increase in AFS was associated with a higher risk of stroke (HR\u0026thinsp;=\u0026thinsp;2.75, P\u0026thinsp;\u0026lt;\u0026thinsp;0.001). Crucially, the addition of AFS yielded significant improvement in reclassification (NRI\u0026thinsp;=\u0026thinsp;15.4%, P\u0026thinsp;\u0026lt;\u0026thinsp;0.001), while the improvement in integrated discrimination was modest (IDI\u0026thinsp;=\u0026thinsp;0.007, P\u0026thinsp;=\u0026thinsp;0.088).\u003c/p\u003e\u003ch2\u003eConclusion\u003c/h2\u003e \u003cp\u003eThe AFS effectively addresses the predictive limitations of traditional metabolic markers in medicated older adults. By integrating metabolic burden with systemic vulnerability, this novel composite score offers a linear and robust predictive approach for stroke, supporting a multidimensional approach to vascular risk stratification in aging populations.\u003c/p\u003e","manuscriptTitle":"Integrating Frailty and Cumulative Lipid Burden for Stroke Risk Stratification: A Machine Learning–Guided Athero -Frailty Score From the CHARLS Cohort","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-02-11 17:03:12","doi":"10.21203/rs.3.rs-8788634/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2026-02-23T08:12:01+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-02-17T04:06:40+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"279185380553791227034772265615826037644","date":"2026-02-16T23:31:49+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-02-16T11:58:54+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"217348577457564473012046133603008481923","date":"2026-02-16T06:33:35+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"59251637114525108456667827703322568941","date":"2026-02-09T13:28:03+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2026-02-09T13:10:41+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-02-04T20:40:07+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-02-04T20:38:59+00:00","index":"","fulltext":""},{"type":"submitted","content":"Lipids in Health and Disease","date":"2026-02-04T15:52:37+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"lipids-in-health-and-disease","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"lhad","sideBox":"Learn more about [Lipids in Health and Disease](http://lipidworld.biomedcentral.com/)","snPcode":"12944","submissionUrl":"https://submission.nature.com/new-submission/12944/3","title":"Lipids in Health and Disease","twitterHandle":"@BioMedCentral","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"BMC/SO AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"6a08eecd-0619-40fc-a4a2-94af20f18807","owner":[],"postedDate":"February 11th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[],"tags":[],"updatedAt":"2026-04-07T16:16:05+00:00","versionOfRecord":{"articleIdentity":"rs-8788634","link":"https://doi.org/10.1186/s12944-026-02931-4","journal":{"identity":"lipids-in-health-and-disease","isVorOnly":false,"title":"Lipids in Health and Disease"},"publishedOn":"2026-03-31 15:58:37","publishedOnDateReadable":"March 31st, 2026"},"versionCreatedAt":"2026-02-11 17:03:12","video":"","vorDoi":"10.1186/s12944-026-02931-4","vorDoiUrl":"https://doi.org/10.1186/s12944-026-02931-4","workflowStages":[]},"version":"v1","identity":"rs-8788634","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8788634","identity":"rs-8788634","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}