Introduction

Frailty is a geriatric syndrome characterized by a declined physiological reserve and an increased vulnerability for adverse outcomes1. With the global population rapidly aging, frailty has become a significant health burden. Approximately 50% of older adults are in pre-frail status and 20% are in frail status2,3,4. Despite varying degrees of frailty, both pre-frail and frail statuses increase the risk of morbidity, disability, and medical expenditures5,6,7,8. Furthermore, the one-year mortality risk can be three times higher in frail than in pre-frail status9. Although frailty may be reversible under appropriate intervention, individuals in a frailer state are less likely to recover. Therefore, preventing frailty onset is critical for public health. Specifically, identifying pre-frail status is just as important as identifying frail status in frailty management.

Several frailty assessments have been proposed in recent decades, and one of the most used is the frailty index (FI)10. FI encompasses domains of physical function, mental health, and medical conditions. This comprehensive evaluation enables FI to identify more pre-frail and frail individuals than other methods in community settings, such as the frailty phenotype11 and the comprehensive geriatric assessment12. Moreover, FI has been shown to be superior in predicting adverse outcome13,14. Notably, some healthcare institutions have integrated FI into routine frailty evaluation15,16. Nevertheless, the dozens of items and diverse health domains in FI make its assessment time-consuming and require experienced evaluators10. Exploring simpler markers for frailty could cost-effectively facilitate frailty detection and management.

Recently, the triglyceride-glucose (TyG) index, a simple surrogate marker of insulin resistance (IR)17, has shown significant association with frailty18. This association is biologically plausible, as IR may impair skeletal muscle glucose uptake and promote chronic low-grade inflammation through cytokines released by adipose tissue19,20. These metabolic and inflammatory mechanisms jointly contribute to the onset and progression of frailty. However, current evidence between TyG index and frailty is mainly cross-sectional, failing to capture the dynamic and potentially reversible nature of frailty. For example, Yin et al. demonstrated a significant association between TyG and frailty, yet the study cannot determine whether TyG predicts transitions between frailty states18. Thus, longitudinal studies with repeated frailty assessment are needed to establish temporal relationships and to explore the potential value of TyG in frailty risk stratification and early identification.

To bridge these knowledge gaps, our study employed a 7-year longitudinal design using repeated FI assessments from the nationally representative China Health and Retirement Longitudinal Study (CHARLS). This design enables us to investigate the temporal association between TyG index and frailty progression and transition, and to identify potential TyG thresholds that may help in early frailty screening and intervention. Given the dynamic characteristic of frailty and its tendency to impact younger adults, we utilized data from the middle-aged and elderly in the CHARLS, which encompassed repeated FI measurements. Our objectives were: (1) to evaluate the association between TyG index and changes in FI, (2) to investigate the relationship between TyG index and various frailty transition patterns, and (3) to identify the TyG index thresholds for frailty transition.

Method

Study design and participants

This longitudinal study utilized a representative sample of community adults aged 45 years or older from the CHARLS cohort (2011–2018). CHARLS includes participants from 28 provinces in China using a multistage stratified probability proportional-to-size sampling strategy. It consists of five waves conducted every two to three years through face-to-face questionnaire interviews and blood sample analysis. Detailed information on sampling, data collection, and measurements is documented in other publications21,22. Participants underwent the evaluations of TyG index and FI at baseline were eligible for this study. In the first part, participants were excluded if no FI assessment was available during the 7-year follow-up. In the second part, only participants who were robust at baseline and completed consecutive assessments of FI were analyzed. The flowchart of this study is shown in Fig. S1.

Exposure and outcome

Fasting triglycerides and fasting plasma glucose were determined from blood samples obtained after an overnight fast. The TyG index was calculated using the formula: Ln [fasting triglyceride (mg/dL) x fasting plasma glucose (mg/dL)/2].

The study outcomes were frailty progression and frailty transition evaluated by FI. The FI was calculated using a standard procedure: the sum of existing deficits divided by the total number of 29, ranging from 0 to 1 score5,10. A higher FI value indicates worse frailty status. The items included in FI were identical in each wave and shown in Table S1. Three frailty status was determined by FI value: (1) robust: FI ≤ 0.10, (2) pre-frail: 0.10 < FI ≤ 0.21, and (3) frail: FI > 0.21. Frailty progression was defined by changes in FI during follow-up. Frailty transitions were defined by baseline-to-final follow-up status; participants with intermediate status changes (e.g., robust→pre-frail→robust) were classified by their final status, consistent with prior FI transition studies5.

We included covariates according to literature review and clinical experience. The sociodemographic variables included age (year), sex (male/female), education level (less than elementary school/middle school or higher), residence (urban/rural), marital status (currently married/other). Lifestyle behaviors included current smoking (yes/no) and alcohol consumption (< 3 times per week/≥3 times per week). Body measurements included systolic blood pressure (SBP, mmHg), diastolic blood pressure (DBP, mmHg), and body mass index (BMI, kg/m2). Clinical conditions and treatments were also included. Conditions were determined by whether participants suffered hypertension, diabetes mellitus, cardiovascular disease, stroke, dyslipidemia, lung disease, kidney disease, and cancer, as reported by participants or physician diagnosis. Medications used to lower blood pressure, blood glucose, and lipids were obtained.

Statistical analysis

Continuous variables were assessed for normality using Shapiro-Wilk tests and visual inspection of histograms and QQ plots. Normally distributed variables were presented as mean (standard deviation, SD), whereas skewed variables were presented as median (interquartile range, IQR). Categorical variables were presented as counts and percentages. T-test, Wilcoxon rank sum test, and chi-squared test were used to compare group differences. All participants were divided into 4 groups according to TyG index quartiles.

In the first part, a linear mixed model (LMM) was constructed to explore the longitudinal relationship between TyG index and the annual change rate of FI score over a 7-year follow-up23,24. Given that LMM could appropriately manage the randomly missing dependent variables, no further imputations were applied25. Five frailty transition patterns were identified: (1) stable: maintained robust/pre-frail/frail status, (2) worsened transition to pre-frail status: from robust to pre-frail, (3) worsened transition to frail status: from robust/pre-frail to frail, (4) improved transition from pre-frail status: from pre-frail to robust, and (5) improved transition from frail status: from frail to pre-frail/robust. Multinomial logistic regression (MLR) was used to evaluate the associations between TyG index and these transition groups. In the second part, Kaplan-Meier curves and log-rank tests were used to illustrate and compare the cumulative incidences of pre-frail and frail transitions from 3,957 robust participants across TyG index quartiles. The proportional hazards assumption was tested using Schoenfeld residuals for all covariates. A multivariable Cox proportional hazards regression model was then developed. Restricted cubic splines (RCS) using three knots at the 10th, 50th, and 90th percentiles of the TyG index were performed to assess the dose-response relationships26. If a non-linear association was observed between TyG index and frailty transition, such as a U-, J-, L-, reverse J-, or reverse L-shaped pattern, a two-line piecewise linear model with a single change point was applied to explore the TyG index cut-off value with the highest likelihood27,28. Furthermore, the prospective associations of the TyG index with the outcomes below and above the threshold were re-examined. Subgroup analyses of age (< 60 years versus ≥ 60 years), sex (male versus female), and BMI (< 24 kg/m2 versus ≥ 24 kg/m2) were examined. All the above models were adjusted by multiple covariates.

Several sensitivity analyses were conducted. First, a nonresponse analysis was performed by comparing the baseline characteristics of the included and the excluded participants. Second, we used a group-based trajectory model (GBTM) to identify FI change trajectories during the follow-up years and tested their relationships with TyG index. Third, the main analyses were performed again using FI > 0.25 for frail diagnosis. All the above analyses were performed using R (version 2023.09.1) and SAS (version 9.4). Two-sided p < 0.05 was considered statistically significant.

Results

Study population

A total of 8,510 participants were included in the final analysis, with a mean age of 58.6 (8.9) years and 51.9% male. The mean TyG index at baseline was 8.70. Participants with a higher TyG index tended to be male, have a higher BMI, not currently smoke, and consume alcohol more frequently. Regarding medical conditions, participants with higher TyG index were more likely to have CVD, dyslipidemia, and DM and to be taking related medications (Table 1).

During follow-up, 61.7% (5,249) participants experienced FI increase, with a median annual change rate of 0.004. In frailty transition, 56.9% (4, 845) remained stable, whereas 15.0% (1,276), 13.7% (1,166), 7.4% (627), and 7.0% (596) experienced robust to pre-frail, robust/pre-frail to frail, pre-frail to robust, and frail to pre-frail/robust transitions, respectively.

Associations between TyG index and frailty progression

In the LMM analysis, higher TyG index quartiles were associated with greater increase in FI after adjusting multiple covariates. Compared with the lowest TyG index quartile, the annual FI change rates were significantly higher in Quartile 2, 3 and 4 at 0.00113 (95%CI: 0.00002-0.00223, p = 0.046), 0.00113 (95%CI: 0.00002-0.00224, p = 0.046), and 0.00239 (95%CI: 0.00127–0.00351, p < 0.0001), respectively (Table 2). Each one-unit increment in TyG index was associated with a 0.00126 (95%CI: 0.00068–0.00184, p < 0.0001) increase in FI per year.

Table 1 Participants’ baseline characteristics according to TyG index quartiles.
Full size table
Table 2 Association between TyG index and frailty index during 7-year follow-up.
Full size table

Associations between TyG index and frailty transition

Compared to the group of maintained stable at robust/pre-frail/fail status, TyG index was positively associated with worsened frailty transitions (Fig. 1). The OR was 1.14 for both worsened to pre-frail (95%CI 1.02–1.26, p = 0.018) and frail (95%CI 1.03–1.27, p = 0.015) status, respectively. Considering the improved frailty transitions, the associations were not significant in the overall sample. However, an inverse association was observed among adults younger than 60 years (OR: 0.68, 95%CI: 0.53–0.88, p = 0.003) and women (OR: 0.76, 95%CI: 0.61–0.94, p = 0.010) in the improved transition from frail status (Fig. S2), but not the worsened transition patterns (Fig. S3). A significant interaction was found between TyG and sex for the transition from frail to pre-frail/robust (p = 0.021), whereas the TyG × age group interaction was not significant (p = 0.988) (Table S2).

Fig. 1
Fig. 1
Full size image

Associations between TyG index and various frailty transition patterns. All the models were adjusted for age, sex, BMI, residence, marriage, education, habits of smoking and drinking, physical activity, CVD, hypertension, dyslipidemia, diabetes mellitus, cancer, lung disease, stroke, kidney disease, medications, SBP, baseline FI, and follow-up time. TyG index, triglyceride-glucose index; BMI, body mass index; CVD, cardiovascular disease; SBP, systolic blood pressure; FI, frailty index.

TyG index thresholds for frailty transition

In the analysis of 3,957 robust participants at baseline, 41.1% (1,625) participants became pre-frail and 10.6% (421) became frail. Kaplan-Meier curves showed that participants in higher quartiles of TyG index had a significantly higher incidence of transition from robust to pre-frail (p = 0.022) and frail (p = 0.006) status, respectively. Schoenfeld residuals confirmed that the proportional hazards assumption was satisfied for the Cox models (GLOBAL p = 0.274 for pre-frail transition and p = 0.91 for frail transition). In multivariable Cox proportional hazards regression analysis, continuous TyG index was positively associated with pre-frail (HR: 1.11, 95%CI: 1.02–1.20, p = 0.014) and frail (HR: 1.22, 95%CI: 1.03–1.45, p = 0. 024) transition (Fig. 2).

Fig. 2
Fig. 2
Full size image

Associations of TyG index and incidence of pre-frail and frail statustransition. (A) Kaplan-Meier curve for TyG index and incidence of pre-frail status transition. (B) Adjusted association between TyG index and incidence of pre-frail status transition. (C) Kaplan-Meier curve for TyG index and incidence of frail status transition. (D) Adjusted associationbetween TyG index and incidence of frail status transition. All the adjusted models includedcovariates of age, sex, BMI, residence, marriage, education, habits of smoking and drinking, physical activity, CVD, hypertension, dyslipidemia, diabetes mellitus, cancer, lung disease, stroke, kidney disease, medications, and SBP. TyG index, triglyceride-glucose index; BMI, bodymass index; CVD, cardiovascular disease; SBP, systolic blood pressure.

A J-shaped dose-response association was depicted in Fig. 3, and the non-linear test was marginally significant for pre-frail transition (p = 0.092) and not significant for frail transition (p = 0.225). The estimated TyG index cut-off values were 9.10 and 9.14 for pre-frail transition and frail transition, respectively. Below the thresholds, the TyG index was not significantly associated with either frailty transition pattern. However, these associations became significant when above the thresholds. A 1-unit increase in TyG index was associated with a 35% and 67% higher risk of pre-frail (HR 1.35, 95%CI: 1.11–1.65, p = 0.003) and frail (HR 1.67, 95%CI: 1.04–2.68, p = 0.035) transition, respectively (Fig. 4). During the 7-year follow-up of robust participants, those with TyG ≥ 9.10 had a 6.4% higher absolute risk of transitioning to pre-frailty and those with TyG ≥ 9.14 had a 1.2% higher absolute risk of transitioning to frailty, compared to participants with lower TyG values, highlighting the clinical relevance of elevated TyG levels (Table S3).

Fig. 3
Fig. 3
Full size image

Dose-response associations of TyG index and pre-frail and frail status transition. (A) Restricted cubic splines for TyG index and incidence of pre-frail status transition. (B) Restricted cubic splines for TyG index and incidence of frail status transition. All models adjusted for covariates of age, sex, BMI, residence, marriage, education, habits of smoking and drinking, physical activity, CVD, hypertension, dyslipidemia, diabetes mellitus, cancer, lung disease, stroke, kidney disease, medications, and SBP. TyG index, triglyceride-glucose index; BMI, body mass index; CVD, cardiovascular disease; SBP, systolic blood pressure.

Fig. 4
Fig. 4
Full size image

Associations of continuous TyG index and incidence of pre-frail and frailstatus transition. All the adjusted models included covariates of age, sex, BMI, residence, marriage, education, habits of smoking and drinking, physical activity, CVD, hypertension, dyslipidemia, diabetes mellitus, cancer, lung disease, stroke, kidney disease, medications, and SBP. TyG index, triglyceride-glucose index; BMI, body mass index; CVD, cardiovascular disease; SBP, systolic blood pressure.

Varying cut-off values in different subgroups are shown in Table 3, ranging from 8.21 to 9.78 for pre-frail transition and from 8.41 to 9.14 for frail transition. The positive associations remained significant in men (Table S4).

Table 3 Estimated TyG index cut-off values.
Full size table

Sensitivity analysis

The comparison between excluded and included participants is shown in Table S5. Compared with included participants, excluded participants were generally older age (mean 60.98y (10.86)) and had similar TyG index levels. Over a 7-year follow-up, four distinct FI trajectories were identified using GBTM: Trajectory 1: robust and stable, Trajectory 2: frail and rapidly increased, Trajectory 3: pre-frail and mildly increased, Trajectory 4: severely frail and mildly increased (Fig. S4). The results showed that a high TyG index level was positively associated with higher FI values and rapid FI increase (Table S6). The positive association between TyG index and pre-frail transition remained consistent using 0.25 for frail diagnosis (Fig. S5-S6), and the cut-off value was a comparable 9.02 for pre-frail transition (Table S7-S8).

Discussion

In this nationwide cohort with a 7-year follow-up, we examined the associations of TyG index with FI, a well-accepted frailty assessment score, and different frailty transition patterns among the middle-aged and elderly dwellers. We found a positive association between TyG index and FI progression, with an annual FI increase rate of 0.00126 for every unit increment in TyG index. Furthermore, we observed a J-shaped dose-response association between TyG index and worsened frailty transition patterns. Specifically, we identified cut-off values of 9.10 and 9.14 for pre-frail and frail transition, respectively. Each increment in TyG index above the thresholds increased the risk of the transition from robust to pre-frail and frail status by 35% and 67%, respectively. To our knowledge, this study is the first to evaluate the prospective value of TyG index in long-term frailty progression and transition using multiple repeated frailty assessments. Our findings suggest that TyG index, a simple and feasible IR biomarker, has the potential to identify individuals at higher risk of frailty, especially in the early onset of pre-frail transition.

For the first time, our study investigated a potential and simple biomarker for frailty transition and demonstrated that TyG index is a valuable predictor for the early onset of frailty. Numerous studies have identified various biomarkers associated with frailty29. However, most of these findings stem from cross-sectional studies, and many biomarkers, such as α1-microglobulin30, TNF-alpha31, and leptin32, are limited to tertiary hospitals. Recently, two studies found that TyG index shows positive association with frail status18,33. One study under cross-sectional design and utilized 7,965 American residents aged 50 years and older in the National Health and Nutrition Examination Survey (NHANES). Another study utilized an urban cohort of 1,866 Chinese community adults aged 60 years and older. Moreover, the cohort observed that a high-stable TyG index trajectory over 10 years of follow-up was associated with pre-frail status. However, the baseline frailty status was not available in the cohort study, failing to explicitly reveal frailty transition patterns. By utilizing a larger and more representative national cohort from CHARLS, our study observed consistent findings in terms of frail status. More importantly, with multiple repeated frailty evaluations from baseline, our results highlighted the profound predictive value of TyG index in pre-frailty occurrence. With a one-unit elevated in TyG index, the risk of pre-frail and frail transition may increase by 11%. In addition, after adjusting baseline frailty status and follow-up years, one-unit increment in TyG index was associated with a 0.00126 increase in FI per year. The positive associations were also significant regarding TyG index and FI increased trajectories. Collectively, these findings suggested that TyG index could serve as a monitoring biomarker for FI progression and early frailty onset, which has important public health and clinical implications. Since lipid and glucose tests are widely available in primary institutions, TyG index is easy to obtain and monitor. Moreover, although substantial evidence demonstrates that FI-based frailty evaluation has significant prognostic value for risk stratification and health outcomes, its implementation in the practice has been challenging. Major obstacles include the complex structure and the time-consuming nature of the evaluation procedure. Thus, adding TyG index measurement to routine primary health checks could be a feasible alternative method to monitor FI progression and detect the early phase of frailty transition.

The prognostic value of TyG index for health outcome has been widely recognized and optimal thresholds have been proposed in recent years. Substantial evidence indicates that TyG index can predict the development of atherosclerosis34, cardiovascular disease35, and heart failure36. Moreover, a representative study with a sample of 141,243 individuals from five continents found that a higher TyG index level is significantly associated with increased risk of cardiovascular mortality37. Similar relationship was found in a Chinese cohort of 3.5 million community adults and it revealed a reverse L-shaped pattern between TyG index and all-cause mortality28. The study suggested that 9.75 is the optimal TyG index threshold for predicting mortality. Other TyG index thresholds for mortality in prior research range from 8.50 to 9.8338,39,40. Our study extends this research field to incident frailty and identifies similar thresholds of 9.10 for pre-frail transition and 9.14 for frail transition. A one-unit increase in TyG index above the thresholds was associated with a 1.35- and 1.67-fold higher risk of pre-frail and frail transition, respectively. Our findings suggest that a TyG index exceeding 9.10 might be a sight for early frailty onset.

Compared to worsened frailty transitions, evidence linking candidate biomarkers to frailty improvement is limited. Leveraging a large sample size, our study was the first to examine the relationship between TyG index and improved frailty transitions. Although interaction analysis confirmed a significant TyG × sex effect but not a TyG × age group effect, subgroup analyses revealed that a lower TyG index was significantly associated with frailty improvement in women and middle-aged adults (< 60 years), but not in men or older adults. This sex disparity may be explained by the impact of sex hormone on insulin sensitivity, as estrogen may enhance glucose utilization and mitigates IR-related inflammation among women41. This finding also aligns with the sex-frailty paradox, suggesting that women exhibit greater frailty yet lower mortality risk39,41,42. Regarding age group difference, the predictive value of TyG index appears to be stronger in adults under 60 years. Age-dependent changes in metabolic dysregulation throughout life may be the cause of this pattern. IR tends to peak in midlife, whereas after age 60, cellular senescence and other aging mechanisms become dominant drivers of frailty, reducing the influence of IR on frailty improvement43,44. Consequently, frailty recovery may be less detectable using IR markers in older adults, underscoring the importance of early detection and intervention2. Taken together, these findings underscore the need for sex- and age-specific frailty prevention strategies.

The mechanisms underlying the relationship between TyG index, a surrogate for IR, and frailty may be explained by the pathogenic roles associated with IR. IR is closely linked to muscle function through its impact on whole-body metabolism45. It may reduce glucose intake in muscle tissue, thereby potentially contributing to impaired muscle function. Additionally, IR may dysregulate lipid and adipose metabolism, which could result in ectopic fat accumulation in skeletal muscle. Increased adipose tissue is thought to induce chronic low-grade inflammation through various cytokines, which may be associated with declines in skeletal muscle quality19,20. Moreover, IR has been linked to oxidative stress, endothelial dysfunction, and altered autophagic process46,47. These metabolic imbalances could contribute to systemic inflammation and vascular remodeling, which are consistent with mechanisms underlying cardiometabolic diseases. Furthermore, emerging evidence from animal and human studies suggests that IR may be related to cognitive decline48,49. IR has been proposed to influence neurodegeneration by affecting cerebral bioenergetics, synaptic viability, and amyloid β peptide clearance, and tau phosphorylation. Collectively, these potential mechanisms are consistent with prior evidence linking IR-related metabolic disturbances to frailty, although causality cannot be established in the present study.

Strengths and limitations

This study has several strengths. First, the study sample consisted of a large, nationally representative cohort of middle-aged and older community-dwelling adults. Second, with a long-term follow-up of 7 years and more detailed frailty stratification, a comprehensive transition of frailty status was presented, providing greater insight into the relationship between TyG index and frailty transition. Third, extensive sensitivity analyses using different methods and cut-off values were performed to ensure the robustness of our main results. This study has several limitations. First, some of the deficits included in the FI calculation were self-reported, which may have introduced information bias. Second, the study population was exclusively from China, so further research in other countries may be needed to generalize the findings. Third, due to the nature of observational study, residual confounding cannot be eliminated. Fourth, the TyG index serves only as a surrogate marker of IR and does not directly measure IR or related biological pathways such as inflammation, oxidative stress, or endothelial dysfunction. Therefore, the proposed mechanistic interpretations should be regarded as hypothesis-generating rather than causal explanations. These potential mechanisms are speculative and based on previous experimental and epidemiological studies. Further research is needed to verify these biological pathways.

Conclusions

Based on a national cohort, this study demonstrated that TyG index could serve as a feasible method for screening and monitoring frailty progression and transition among middle-aged and older adults. Specifically, a TyG index above 9.10 and 9.14 significantly increased the risk of transition from robust to pre-frail and frail statuses by 35% and 67%, respectively. Further cautions and evaluations should be given to individuals with a TyG index exceeding 9.10 for early frailty management.