Four Blood Markers May Predict Who Becomes Frail in Old Age, Three Global Cohorts Suggest

A simple score built from four routine clinical measurements—systolic blood pressure, high-density lipoprotein cholesterol, glycated hemoglobin, and C-reactive protein—appears to flag older adults at heightened risk of developing frailty, according to a new longitudinal analysis published in BMC Geriatrics. The study, led by Zhaobei Cai of Peking University Aerospace Center Hospital and Zhenxiao Ren of the Hospital Nacional de Parapléjicos in Toledo, Spain, and RWTH Aachen University in Germany, set out to answer a deceptively narrow but consequential question: can a biomarker score that was originally calibrated for one aging-related outcome, mobility limitation, still predict a different outcome, incident frailty, without any re-tuning of its mathematical weights? The answer, drawn from more than 12,000 older adults across the United States, England, and China, is a cautious yes—with important caveats that the authors themselves are careful to spell out.

Frailty is one of the most consequential yet elusive syndromes in geriatric medicine. It describes a state of diminished physiological reserve in which relatively minor stressors—a urinary tract infection, a fall, a hospital admission—can trigger disproportionate declines in health, disability, and death. Researchers commonly quantify frailty using a frailty index, a composite measure built from deficits such as chronic diseases, functional limitations, and cognitive symptoms; in this study, a frailty index of 0.25 or higher defined the frail state. Because frailty emerges gradually over years, identifying modifiable physiological signals that precede its onset could open a window for prevention. Cardiometabolic dysfunction—elevated blood pressure, adverse cholesterol profiles, impaired glucose metabolism, and chronic low-grade inflammation—has long been suspected of contributing to this process, but the field has lacked a standardized, transportable way to capture that cumulative vulnerability.

That gap is precisely what the SHAC index was designed to fill. Originally developed for predicting incident mobility limitation, the index combines four biomarkers using fixed weights: systolic blood pressure, high-density lipoprotein cholesterol, glycated hemoglobin (HbA1c), and C-reactive protein (CRP). The crucial methodological twist in the new study is that the authors deliberately did not re-estimate those weights for frailty. Instead, they applied the original fixed-weight formula unchanged, testing whether a score developed for one functional aging outcome retains a meaningful association with a different one. This is a stricter test than most biomarker studies attempt, because re-fitting weights to a new outcome almost always improves apparent performance; a fixed-weight score that still works across outcomes and populations carries stronger evidence of capturing a genuine underlying biological construct.

To test transportability, the researchers turned to three of the world’s most extensively characterized aging cohorts: the Health and Retirement Study (HRS) in the United States, the English Longitudinal Study of Ageing (ELSA), and the China Health and Retirement Longitudinal Study (CHARLS). Each cohort harmonizes its core survey instruments across waves, allowing researchers to compare aging trajectories across very different economic, dietary, and healthcare contexts. The analysis included adults aged 65 and older who had complete SHAC biomarkers and a valid frailty index at a baseline assessment, and who were not already frail at that baseline. Incident frailty was then tracked longitudinally, defined as reaching a frailty index of at least 0.25 at a subsequent wave. The fully adjusted analytical samples comprised 6,965 HRS participants with 3,334 incident frailty events, 4,150 ELSA participants with 1,192 events, and 1,473 CHARLS participants with 476 events.

The statistical machinery behind the study was correspondingly rigorous. In each cohort, the researchers fitted Cox proportional hazards models estimating hazard ratios for incident frailty per one standard deviation higher SHAC score, using cohort-internal standardization so that the exposure scale was comparable across populations. The fully adjusted cohort-specific estimates were then pooled using random-effects meta-analysis, a technique that explicitly allows for between-cohort heterogeneity rather than assuming a single true effect. The headline result: hazard ratios of 1.12 (95 percent confidence interval 1.08–1.16) in HRS, 1.21 (1.14–1.29) in ELSA, and 1.19 (1.08–1.30) in CHARLS, pooling to a combined hazard ratio of 1.16 (1.10–1.23) with an I-squared heterogeneity statistic of 65.9 percent. In practical terms, each standard deviation increase in the SHAC score was associated with roughly a 12 to 21 percent higher hazard of becoming frail, an association that held in all three countries despite their markedly different baseline cardiometabolic profiles.

The authors did not stop at the primary model. A battery of sensitivity analyses probed the robustness of the finding from multiple angles: alternative ways of scaling the exposure, including a fixed ELSA development-reference scaling that retained positive associations; different definitions of the baseline frailty index; alternative timing of the outcome; various approaches to missing data and selection; competing mortality, addressed through subdistribution hazard models that account for the fact that some participants died before developing frailty; and longitudinally updated SHAC scores that incorporated repeated biomarker measurements over time. Across these analyses, the findings were broadly similar, lending weight to the conclusion that the association is not an artifact of any single modeling choice. Notably, when the models additionally adjusted for the baseline frailty index—a measure of how close participants already were to the frailty threshold—the associations attenuated but did not disappear, to hazard ratios of 1.05 (1.01–1.09) in HRS, 1.13 (1.06–1.20) in ELSA, and 1.12 (1.02–1.23) in CHARLS. This attenuation suggests that part, but not all, of the SHAC signal overlaps with pre-existing health deficits.

Among the four individual biomarkers, HbA1c emerged as the most consistently associated with incident frailty, a finding that aligns with a growing body of evidence linking glycemic dysregulation to accelerated functional decline in later life. Chronic hyperglycemia is implicated in microvascular damage, sarcopenia, and neuromuscular impairment, all of which feed into the frailty phenotype. The comparative performance of the composite score, however, tells a more sobering story. Adding SHAC to prediction models produced only small increases in apparent discrimination—the statistical ability to distinguish who will and will not experience the outcome—and, critically, SHAC did not clearly outperform HbA1c alone. For a score to justify clinical adoption, it must demonstrably add predictive value beyond its components; on that front, the study’s own data counsel restraint.

The authors are explicit about the limits of what their results support. Cohort-dependent scaling of the biomarkers, the attenuation observed when baseline frailty burden is accounted for, selection effects inherent to longitudinal cohort participation, the substantial between-cohort heterogeneity reflected in the I-squared statistic, and the modest discrimination gains all constrain stronger claims about universal transportability of the score or its use as a clinical prediction tool. In other words, the study establishes an association—consistent across three continents, robust to many sensitivity checks, and biologically plausible—but it does not establish that SHAC is ready for the clinic, nor that the same weight structure applies identically everywhere. This kind of calibrated interpretation is increasingly rare in biomarker research, where overclaimed transportability has repeatedly failed to survive external validation.

Nevertheless, the implications for aging research are significant. If a fixed-weight cardiometabolic vulnerability score, developed for mobility limitation, independently tracks incident frailty across American, English, and Chinese populations, it suggests that a shared physiological axis—spanning blood pressure regulation, lipid metabolism, glucose homeostasis, and systemic inflammation—underlies multiple manifestations of functional aging. That convergence matters for prevention: the same interventions that reduce cardiovascular risk, from glycemic control to anti-inflammatory strategies, might plausibly delay frailty as well. The study received no specific external funding, and it relied on de-identified secondary data from cohorts approved by institutional review boards in the United States, the United Kingdom, and China, with all participants having provided informed consent in the original surveys. Future work, the authors suggest, should focus on refining how such scores are scaled across populations, clarifying their relationship to baseline health status, and testing whether they can genuinely improve prediction beyond simple single-marker measures. For now, the message is measured but encouraging: the biology of frailty may be more unified, and more measurable, than previously thought.

Subject of Research: Longitudinal association between a fixed-weight cardiometabolic biomarker score and incident frailty in older adults across three international aging cohorts

Article Title: Cardiometabolic vulnerability and incident frailty across three aging cohorts: a multicohort longitudinal study of HRS, ELSA, and CHARLS

Article References: Cai, Z., & Ren, Z. (2026). Cardiometabolic vulnerability and incident frailty across three aging cohorts: a multicohort longitudinal study of HRS, ELSA, and CHARLS. BMC Geriatrics. https://doi.org/10.1186/s12877-026-08417-3

Image Credits: AI Generated

DOI: 10.1186/s12877-026-08417-3

Keywords: frailty, cardiometabolic vulnerability, SHAC index, HbA1c, C-reactive protein, systolic blood pressure, HDL cholesterol, HRS, ELSA, CHARLS, aging, Cox proportional hazards