Body Size and Inflammation Track Closely With TyG-BMI in Pre-Frail Older Adults
A cross-sectional analysis of 2,249 adults aged 60 years and older found that TyG-BMI clustered most strongly with body size and a marker of systemic inflammation. Weight carried the largest positive standardized regression coefficient in the multiple linear regression model, followed by waist circumference, gender, and C-reactive protein. In practical terms, the surrogate index linked to insulin resistance rose as participants became heavier and more centrally adipose, while inflammatory burden also moved in the same direction.
TyG-BMI combines triglycerides, fasting glucose, and body mass index, so it sits at the intersection of adiposity, glycemic dysregulation, and lipid handling. In older adults already classified as pre-frail, that mix is biologically plausible: reduced muscle reserve, visceral fat accumulation, and chronic low-grade inflammation often travel together. A higher TyG-BMI in this setting may therefore reflect more than a laboratory abstraction; it may be a compact signal of metabolic stress in aging bodies that are already vulnerable to functional decline.
The study drew on baseline data from the 2015 China Health and Retirement Longitudinal Study and focused on pre-frail participants rather than the full older population. That choice matters because pre-frailty is a transitional state, not a fixed endpoint. It captures older adults who have already begun to accumulate physiologic vulnerability, making them a useful group for examining how metabolic patterns vary before overt frailty becomes entrenched.
Weight and Waist Circumference Dominated the Regression Signal
In the regression model, weight emerged as the strongest positive contributor to TyG-BMI, with waist circumference also showing a significant association. The pattern is consistent with the central role of adiposity in insulin resistance. Excess adipose tissue, particularly when concentrated around the abdomen, increases free fatty acid flux, worsens hepatic insulin signaling, and tends to push triglyceride levels upward while glucose control deteriorates. Because TyG-BMI incorporates both lipid and glucose measures alongside BMI, it is unsurprising that anthropometric measures remained prominent after adjustment.
Gender also entered the final model, suggesting that metabolic differences between men and women persisted even within a pre-frail cohort. CRP added an inflammatory dimension to the picture. Low-grade inflammation is common in older adults and can amplify insulin resistance through cytokine signaling and impaired metabolic flexibility. In this dataset, metabolic status was not separable from inflammatory burden; the two appeared to advance together.
The analysis did not stop at a standard regression framework. Investigators also used classification and regression tree methods to map subgroup structure. That second approach is useful when risk patterns are not purely linear and may depend on combinations of demographic and social variables rather than single predictors in isolation.
Seven Terminal Subgroups Revealed a Social Gradient in Metabolic Risk
The decision tree produced seven terminal subgroups, with gender as the root node. Residence, number of living children, and social interaction appeared in later splits, indicating that metabolic patterning extended beyond body composition alone. Social and family structure shaped where participants landed in the tree, suggesting that TyG-BMI variation in pre-frail older adults may reflect the lived environment as much as the metabolic phenotype.
One subgroup, labeled PCOA-6, had the highest observed mean TyG-BMI at 243.77. Another, PCOA-2, had the lowest at 194.92. Those values are descriptive rather than causal, but they provide a concrete signal that the cohort was not metabolically uniform. The spread across terminal nodes hints at distinct combinations of sex, residence, family structure, and social engagement that may align with different metabolic profiles.
This is where the biology and the social context meet. Older adults with fewer social contacts or different household structures may experience changes in physical activity, diet regularity, emotional stress, and health monitoring. Each of those can influence triglycerides, glucose regulation, and adiposity. In a pre-frail population, such shifts may be especially consequential because physiologic reserve is already reduced.
Apparent ROC Performance Favored the Regression Model, Not a Future Prediction Claim
For an exploratory discrimination analysis, participants were divided into high- and low-TyG-BMI groups using the sample median. Under that framework, the multiple linear regression model produced an apparent area under the curve of 0.873, with a 95% confidence interval of 0.859 to 0.888. The decision tree reached an AUC of 0.605, with a 95% confidence interval of 0.582 to 0.629. The gap between the two models suggests that the regression-based approach separated the median-defined groups more efficiently than the tree.
These AUC values are descriptive and should not be treated as evidence of future outcome prediction. They were calculated against groups defined by the sample median, not against incident frailty, cardiovascular events, diabetes, or mortality. In other words, the models describe separation within the dataset; they do not establish clinical forecasting power.
That caution is not academic hair-splitting. A marker that sorts participants well in a cross-sectional sample can still fail to predict downstream disease trajectories. Without longitudinal follow-up, the study cannot tell whether higher TyG-BMI in a given subgroup leads to worse functional decline, more hospitalizations, or increased mortality. The current analysis remains anchored in association, not prognosis.
What the Findings Can and Cannot Support in Geriatric Practice
The chief value of this work lies in pattern recognition. Among pre-frail older adults, TyG-BMI aligned most closely with weight, waist circumference, sex, and CRP, while subgroup structure was shaped by residence, living children, and social interaction. This combination points toward a metabolic phenotype influenced by both adiposity and social circumstance. For geriatric care, that suggests TyG-BMI may be worth examining alongside anthropometric and inflammatory measures when clinicians are thinking about metabolic burden in vulnerable older adults.
The limits are equally direct. The design was cross-sectional, so directionality cannot be inferred. No causal pathway can be assigned to weight, social interaction, or CRP from these data alone. The tree analysis is exploratory, the AUC values are apparent, and the subgroup labels require independent replication before they can be treated as clinically durable. Validation in other cohorts and longitudinal studies will be needed before TyG-BMI can be linked confidently to meaningful geriatric outcomes.