Diabetes mellitus is a prevalent chronic metabolic disease worldwide and is frequently accompanied by serious complications (1). Its onset is typically preceded by a prodromal state—prediabetes—defined by elevated but nondiabetic blood glucose levels. Even at this stage, the risk of both micro- and macrovascular complications is markedly increased (2). Globally, more than 400 million adults currently have prediabetes, a figure projected to exceed 600 million by 2045 (3). In China, the age-standardized prevalence among adults is 35.7%, surpassing that of most other chronic disorders (4). Critically, intensive lifestyle or pharmacologic interventions can restore normoglycemia in a substantial proportion of individuals within a relatively short period, and the benefits accrue over time. In the Diabetes Prevention Program Outcomes Study, participants who reverted to normal glucose regulation—even once—experienced a ≥50% reduction in the subsequent risk of type 2 diabetes (T2D) (5, 6) and parallel declines in cardiovascular events, microvascular damage, nephropathy, and retinopathy (7). Given the substantial public health burden imposed by diabetes and its prequel state, a deeper understanding of the determinants of prediabetes is essential for the early identification of high-risk individuals and the timely deployment of effective preventive strategies.

Insulin resistance (IR), characterized by chronic compensatory hyperinsulinemia, is the fundamental pathophysiological driver of prediabetes and an independent risk factor for multisystem complications including cardiovascular disease, cellular senescence, and neurodegeneration, often preceding dysglycemia by years (8, 9). Direct quantification with the euglycemic-hyperinsulinemic clamp or intravenous glucose tolerance test is prohibitively expensive and technically demanding, precluding large-scale screening (10, 11). The triglyceride–glucose (TyG) index—calculated from fasting triglycerides (TG) and fasting plasma glucose (FPG)—has emerged as a practical and cost-effective surrogate marker of IR (12, 13)and has consistently demonstrated robust associations with incident T2D and cardiovascular events across diverse populations. More recently, obesity-adjusted derivatives such as the TyG-body mass index (TyG-BMI) have been developed to integrate IR with visceral adiposity, yielding superior metabolic risk prediction (14). Nevertheless, prior investigations have been limited by single-center designs and modest sample sizes, and large-scale studies systematically evaluating the predictive value of the TyG index and its obesity-adjusted variants for prediabetes in the Chinese population are lacking.

Although prior studies have linked the TyG index and its derivatives to incident diabetes mellitus (15, 16), their longitudinal value for identifying the earliest dysglycemic transition—prediabetes—remains undefined. Using a multicenter, retrospective Chinese cohort, we will quantify the temporal associations and predictive performance of TyG and TyG-BMI for future prediabetes onset, aiming to deliver an evidence-based tool for early identification of high-risk IR states.

In this large, population-based longitudinal cohort of Chinese adults, we report several novel findings. First, both the TyG index and TyG-BMI showed an independent, positive association with incident prediabetes. Second, these exposures demonstrated nonlinear, dose-response relations with prediabetes risk and exhibited clear threshold effects; Kaplan–Meier curves confirmed that higher TyG and TyG-BMI levels were associated with greater cumulative incidence of prediabetes. Third, age and sex each modified these associations synergistically, such that the combination of younger age (< 45 years) or female sex with elevated TyG or TyG-BMI conferred a supra-additive increase in prediabetes risk. Finally, sensitivity analyses indicated that these associations remained robust under multiple alternative assumptions.

Previous studies have established the triglyceride–glucose (TyG) index as a reliable surrogate marker of IR, showing strong correlations with both the hyperinsulinemic–euglycemic clamp and the homeostasis model assessment of IR (HOMA-IR) (21, 22). Consequently, elevated TyG levels have been linked to increased risks of T2D and cardiovascular disease (23, 24). However, investigations specifically targeting prediabetes remain scarce and are often limited by small sample sizes or cross-sectional designs (25). The TyG-body mass index (TyG-BMI), which integrates adiposity with dysregulated lipid–glucose metabolism, has also emerged as a robust predictor of T2D and metabolic syndrome (14, 26), yet its association with incident prediabetes in large, ethnically diverse Chinese cohorts has not been adequately characterized.

The robust associations of the TyG index and TyG-BMI with incident prediabetes directly reflect their role as sensitive indicators of underlying IR. Elevated fasting TG signify increased free fatty acid flux and augmented hepatic very-low-density lipoprotein secretion, both key drivers of hepatic IR (27). Concomitant hyperglycemia denotes impaired insulin-mediated glucose disposal in skeletal muscle and adipose tissue (28). This lipid–glucose dysregulation establishes a vicious cycle: IR exacerbates dyslipidemia and hyperglycemia, which in turn further impair insulin signaling and heighten β-cell stress (29). By incorporating adiposity, TyG-BMI captures obesity’s contribution to IR—through pro-inflammatory adipokines (e.g., TNF-α, IL-6) and reduced adiponectin—amplifying systemic inflammation and metabolic dysfunction (30). Critically, IR with compensatory chronic hyperinsulinemia is not merely a precursor to dysglycemia, but an independent pathological state that precedes prediabetes/T2D by years or decades (31, 32). It directly drives systemic damage, including cardiovascular remodeling (e.g., atherosclerosis, hypertension), accelerated cellular senescence, oncogenesis, and neurodegeneration (33). Thus, the TyG index and TyG-BMI parsimoniously encapsulate IR-driven derangements in lipid-glucose-adipose axis that initiate both metabolic dysregulation and extra-glycemic organ damage, long before overt dysglycemia develops.

Strengths of this study are fourfold. First, TyG and TyG-BMI can be derived rapidly from routine fasting lipid and anthropometric data, permitting low-cost prediabetes risk stratification without additional testing. Second, the cohort spans multiple regions of China, is large, and covers a wide age range, enhancing generalizability to the broader Chinese population. Third, we evaluated additive interaction within age- and sex-specific strata, aligning with contemporary public health decision-making frameworks (34). Finally, extensive sensitivity analyses confirmed the stability of our findings. Collectively, these results provide concise, efficient tools for early identification of IR—a fundamental driver of cardiometabolic and multiorgan pathology. Their use in risk stratification extends beyond predicting dysglycemia; it flags a high-risk metabolic state where interventions (e.g., lifestyle, metformin) can potentially prevent not only diabetes but also IR-related cardiovascular, neoplastic, and neurological complications.

Several caveats merit consideration. First, although the cohort was not designed as a nationally representative sample, its multicenter design and broad age span enhance reasonable generalizability to the Chinese population. Second, the applicability of our findings to populations with substantially higher obesity prevalence (e.g., European or North American cohorts) requires verification. In such settings, the strong correlation between adiposity and dysglycemia may diminish the incremental predictive value of TyG over BMI or waist circumference alone, as adiposity may dominate the dysglycemia risk signal. However, TyG-BMI—by integrating metabolic dysregulation and adiposity—likely retains discriminative superiority for prediabetes risk stratification even in high-adiposity populations. This is mechanistically plausible as the combined index captures both lipid-glucose dysregulation and adiposity burden, thereby mitigating signal saturation effects from either pathway alone. Therefore, future multi-ethnic studies with stratified obesity profiles (e.g., ≥30% obesity) are warranted to establish population-specific thresholds. Moreover, in this study, prediabetes was defined solely on the basis of FPG; consequently, some incident cases may have been missed. Nevertheless, we observed a significant association of both TyG and TyG-BMI with prediabetes in this restricted sample. In addition, we acknowledge that the absence of glycated hemoglobin and oral glucose tolerance test data limits the generalizability and completeness of our findings. Future work should incorporate these additional metrics to provide a more comprehensive and accurate assessment of diabetes risk. Furthermore, residual confounding—stemming from imprecisely measured or unmeasured factors such as physical activity and diet—may persist despite extensive adjustment. Prediabetes was defined exclusively by incident impaired fasting glucose, potentially overlooking cases identified through alternative glycemic criteria. The relatively brief follow-up yielded a modest absolute event count; nevertheless, TyG and TyG-BMI retained significant and consistent associations with prediabetes, suggesting that their predictive utility is likely to endure—or strengthen—over longer observation periods. Despite these limitations, our findings underscore that TyG and TyG-BMI are more than prediabetes predictors—they are accessible proxies for IR, a root cause of systemic disease burden. Prioritizing IR detection—years before dysglycemia emerges—could transform early prevention paradigms by targeting the root cause of multiorgan damage.

In this large, multicenter Chinese cohort, both the TyG index and TyG-BMI independently and non-linearly predicted incident prediabetes, with clear threshold effects at TyG of 9 and TyG-BMI of 200. The associations were amplified by older age and male sex, underscoring the value of sex- and age-specific risk stratification. Critically, as direct proxies for IR—an independent pathological state preceding dysglycemia by years—these indices empower clinicians to identify high-risk metabolic phenotypes for early interventions. Such interventions (ranging from intensive lifestyle optimization to targeted pharmacotherapy for eligible individuals) aim not only to prevent diabetes, but more importantly to mitigate IR-driven cardiovascular and neoplastic sequelae arising from chronic hyperinsulinemia.