Background

With the aging population expanding, it is expected that by 2050, there would be 2.1 billion people over 60 years old [1]. As age increases, pathophysiological changes such as rise in proinflammatory factors, oxidative stress, mitochondrial dysfunction, and DNA damage can contribute to the development of metabolic disorders [2, 3]. Patients over the age of 60 years make up the majority of diabetes mellitus (DM) cases [4], with the numbers expected to increase further [5]. Elderly diabetic patients are more susceptible to a range of health complications, such as frailty, sarcopenia, cognitive impairment, multimorbidity, coronary artery stenosis, heart failure, stroke and death [6,7,8,9]. Early identification of the high-risk patients is essential for improving prognosis.

Insulin resistance (IR) is a state of decreased responsiveness to insulin, and is one of the key pathophysiological features of DM. Increasing evidence suggests that IR contributes to the development of cardiovascular disease (CVD) in individuals with DM and is also an indicator of a poorer prognosis. Due to the high cost and inconvenience associated with traditional methods of evaluating IR, the triglyceride-glucose (TyG) index was developed [10]. TyG index is highly related to hyper-insulinemic euglycemic clamp and the homeostatic model assessment of insulin resistance index (HOMA-IR) [11, 12], and its elevation is positively correlated with poor prognosis [13,14,15,16]. Body mass index (BMI) is a simple anthropometric measure of nutritional status, and recent findings indicated that the combination of TyG and BMI (TyG-BMI index) increased effectiveness on assessing IR than TyG index alone [17, 18]. However, the relationship between TyG-BMI index and mortality in elderly patients with DM are still unclear. Thus, the study aimed to investigate the association between TyG-BMI with all-cause and cardiovascular mortality among elderly DM patients in the United States (US).

Methods

Study population

Data of this study were collected from the National Health and Nutrition Examination Survey (NHANES), a program managed by the Centers for Disease Control and Prevention (CDC) and the National Centers for Health Statistics (NCHS) in the United States (US). This crucial research program that mainly aims to evaluate the health and nutritional status of Americans and follows the STROBE guidelines for reporting observational studies. The protocols of NHANES have been approved by the Research Ethics Review Board of NCHS and the informed written consent were obtained from all of the participants involved in the study. We extracted the datasets from the NHANES website covering five survey cycles between 2007 and 2016 (https://www.cdc.gov/nchs/nhanes/index.htm). The anonymized NHANES database employs unique identifiers (SEQN) for matching participant data with NDI death records, adhering to CDC's established data linkage protocols. The data generated and analyzed in the study included demographic information, examination data, laboratory data and questionnaire data. Elderly population was defined as older than 60 years old. A total of 50,588 participants were screened in the NHANES cohort between 2007–2016. After excluding those with missing data on fasting plasma glucose (n = 34,915), triglycerides (n = 382) and BMI (n = 185) and those younger than 60 years old or without T2DM (n = 13,743), we included 1363 eligible elderly patients with DM in the final analysis (Fig. 1).

Fig. 1
figure 1

Flowchart of the sample selection from NHANES 2007–2016

Full size image

Diagnosis of diabetes

According to the ADA’s diabetes diagnostic criteria, DM was defined as having any of the following: (a) hemoglobin A1C concentration ≥ 6.5% or a fasting plasma glucose level ≥ 126 mg/dL [2]; (b) for those who responded ‘yes’ to the question: ‘Doctor told you have diabetes?’ or ‘Taking insulin now?’.

Assessment of covariates

Data of demographic, examination, laboratory and questionnaire were collected. BMI was calculated as weight in kilograms divided by height in meters squared. Race/ethnicity was categorized as Mexican American, non-Hispanic white, non-Hispanic black and other race. Alcohol intake was shown as non-drinker, 1 to < 5 drinks/month, 5 to < 10 drinks/month, or 10 + drinks/month. Smoking status was classified as never smoker, former smoker and current smoker. Hypertension was determined by a diagnosed hypertension history by physician, or non-same-day randomized records of 3 times systolic blood pressure ≥ 140 mmHg or diastolic blood pressure ≥ 90 mmHg from baseline interview to 1-year follow-up. Coronary heart disease (CHD) was diagnosed as self-reported CHD, angina pectoris or myocardial infarction. Heart failure was determined by a response of ‘yes’ to the question “Ever told you had heart failure?”. Cancer was defined as self-reported cancer or malignancy. Stroke was determined by a response of ‘yes’ to the question “Ever told you had a stroke?” and chronic obstructive pulmonary disease (COPD) was determined by a response of ‘yes’ to the question “Ever told you had COPD?”. Laboratory data such as fasting plasma glucose, glycosylated hemoglobin type A1c (HbA1c), triglycerides, total cholesterol, low-density lipoprotein cholesterol (LDL-C), and high-density lipoprotein cholesterol (HDL-C) were exported from the NHANES website.

Assessment of TyG-BMI index

TyG-BMI index = BMI × TyG, where BMI = weight/height2, and the TyG index = Ln [fasting triglycerides (mg/dl) × fasting plasma glucose (mg/dl)/2) [1]. The triglycerides and fasting plasma glucose were measured through enzymatic assays on Roche Modular P and Roche Cobas 6000 chemistry analyzers, respectively. All patients were classified into four groups (Q1, Q2, Q3, Q4) by the quartiles of TyG-BMI index, and the group of Q1 was used as the reference group.

Ascertainment of mortality

The mortality data of the current study were extracted from the National Death Index (NDI) death certificate records provided by NCHS, and the linked mortality files were updated to December 31, 2019. The study endpoints were all-cause mortality and cardiovascular mortality. The International Statistical Classification of Diseases, 10th Revision (ICD-10) was used to determine the reason of deaths. All-cause mortality was defined as death from any cause, including malignant neoplasms (019–043), chronic lower respiratory diseases (082–086), cardiovascular diseases (054–068), and other causes. During follow-up, death due to heart diseases was defined as cardiovascular mortality. The follow-up time was calculated from the baseline interview to the date of death or December 31, 2019.

Statistical analysis

R software (Version 4.3.2) was used to perform the statistical analyses and create the tables and figures. As required to analyzed the NHANES data, sample weights, clustering, and stratification were incorporated in all analyses [19]. The baseline characteristics were grouped according to quartiles (Q1-Q4) of the TyG-BMI index. Values are means ± standard deviation (SD) for continuous variables. Unweighted frequency counts and weighted percentages are shown for categorical variables. P Value was calculated by weighted linear regression model for continuous variables and weighted chi-square test was performed for categorical variables. The incidence rates of all-cause mortality for each TyG-BMI quartile group were computed during the follow-up. In addition, multivariate Cox proportion hazards regression model was applied to identify the independent predictive value of the TyG-BMI index. The Cox regression model included three models to confounding factors. Model 1 was not adjusted any covariates. Model 2 was adjusted for age, gender, race, and total cholesterol. Model 3 was adjusted for age, gender, race, hypertension, total cholesterol, HDL-C, LDL-C, CHD, heart failure, cancer, stroke, and COPD. For missing values, we used multiple imputation. To explore the association between TyG-BMI index and all-cause mortality, Cox proportional hazards regression models with restricted cubic spline (RCS) analyses with four knots. In the RCS model, we also adjusted for confounding factors: age, gender, race, smoking status, alcohol use, hypertension, HDL-C, LDL-C, CHD, heart failure, cancer, stroke, and COPD. Two-piecewise Cox proportional hazards model on both sides of the inflection point was used to investigate the association between SHR and the risk of mortality. Subgroup analyses were also performed to assess the influence of TyG-BMI index on all-cause mortality in different subgroups stratified by age (60–69 years old, and ≥ 70 years old), gender, and race (Mexican American, non-Hispanic white, non-Hispanic black, and other races). A P value < 0.05 was defined as the statistical threshold for all analyses.

Results

Baseline characteristics

A total of 1363 elderly patients with diabetes were included in the current study cohort (weighted population, 7,261,878). Table 1 showed the baseline characteristics of study patients stratified by quartiles of the TyG-BMI index. The average age of the including patients was 70.0 ± 6.8 years, and 48.6% patients were female. Among these participants, 544 (39.9%) were non-Hispanic White, 154 (11.3%) were Mexican American, 355 (26.0%) were non-Hispanic Black, and 310 (22.7%) were of other races. The mean TyG-BMI index was 286.3 ± 66.5. As for the comorbidities including CHD, heart failure, stroke, cancer and COPD, there was no significant differences between groups (all P values > 0.05). Patients with a higher TyG-BMI index were more likely to be younger compared to patients in the lower TyG-BMI index. Importantly, significant differences were observed between the different quartiles, with patients in the highest quartiles have significantly higher levels of total cholesterol and lower levels of HDL-C.

Table 1 Baseline characteristics of study participants
Full size table

Multivariate Cox regression analysis for all-cause and cardiovascular mortality

During a mean follow-up period of 77.3 months, 429 all-cause deaths occurred, with an all-cause mortality rate of 402/100000 person-years; 123 cardiovascular deaths occurred, with a cardiovascular mortality rate of 117/100000 person-years. As shown in Table 2, we tried three Cox regression models to investigate the relationship between TyG-BMI index and all-cause mortality. Before adjusted covariates, the hazard ratios (HRs) and confidence intervals (CIs) from the first quartile to the fourth quartile were 1.00 (reference), 0.75 (0.56, 0.99), 0.52 (0.38–0.71), and 0.56 (0.40–0.78), respectively. In model 2, we adjusted for age, gender, race and total cholesterol and the HRs and 95% CIs were 1.00 (reference), 0.85 (0.64–1.15), 0.64 (0.46–0.89), and 0.92 (0.62–1.35), respectively, for all-cause mortality. While in model 3, after adjusting age, gender, race, hypertension, total cholesterol, HDL-C, LDL-C, CHD, heart failure, cancer, stroke, and COPD, the HRs and 95% CIs were 1.00 (reference), 0.83 (0.61–1.13), 0.65 (0.43–0.96), and 0.87 (0.56–1.36), respectively. The multivariate Cox regression models for cardiovascular mortality was shown in Table 3. After adjusting age, gender, race and total cholesterol in model 2, the HRs and 95% CIs from the highest quartile to the lowest quartile were 1.83 (1.01–3.30), 0.73 (0.33–1.61), 1.09 (0.62–1.91), and 1.00 (reference), respectively. In addition, in model 3, we further adjusted hypertension, HDL-C, LDL-C, CHD, heart failure, cancer, stroke, and COPD beyond the model 2. The HRs and 95% CIs were 2.45 (1.07–5.57), 0.99 (0.42–2.32), 1.24 (0.64–2.43), and 1.00 (reference), respectively.

Table 2 Association between TYG/BMI index and all-cause mortality
Full size table
Table 3 Association between TYG/BMI index and cardiovascular mortality
Full size table

Relationships between TyG-BMI index and mortality

We used a Cox proportional hazards regression models with RCS to model the relationship between the TyG-BMI index and the risk of mortality in elderly patients with DM. We chose to use four knots, placed at the 25th, 50th, 75th, and 100th percentiles of the distribution of the TyG-BMI index in our study population, to allow for flexible modeling of the non-linear relationship between the TyG-BMI index and mortality. Figure 2A shows the estimated spline function for the relationship between the TyG-BMI index and the risk of all-cause mortality in elderly patients with DM, based on our RCS model. The results indicated a “U-shaped” association after adjusting for age, gender, race, smoking status, alcohol use, hypertension, HDL-C, LDL-C, CHD, heart failure, cancer, stroke, and COPD. P value for nonlinearity was < 0.05, when TyG-BMI index was < 264.18, the HR was decreased as the TyG-BMI index increased, while it increased when the TyG-BMI index was > 264.18. The value of TyG-BMI index corresponding to the lowest risk of all-cause mortality on multivariate-adjusted RCS analyses was 264.18 for the population (Fig. 2). Figure 2B shows the estimated spline function for the relationship between the TyG-BMI index and the risk of cardiovascular mortality in elderly patients with DM, based on our RCS model. This RCS model also adjusted the same variables as all-cause mortality. The curve showed an increasing trend in the risk of cardiovascular mortality as the TyG-BMI index increased, and there was a linear relationship with cardiovascular mortality (P for non-linear = 0.665). When TyG-BMI index was > 264.18, the HR was increased sharply and TyG-BMI index was served as a risk factor.

Fig. 2
figure 2

Association between TyG-BMI index and all-cause (A) and cardiovascular mortality (B) in elderly patients with DM

Full size image

Subgroup analysis

The survival advantages of a higher TyG-BMI index (≥ 264.18 for all-cause mortality) compared to a lower TyG-BMI index (< 264.18 for all-cause mortality) among elderly patients with DM was consistent across various subgroups based on age, gender and race as depicted in Table 4. There was no significant interaction between the TyG-BMI and stratified variables.

Table 4 Subgroup analyses of the association between TYG-BMI index and all-cause mortality
Full size table

Discussion

To the best of our knowledge, this is the first study to evaluate the association between TyG-BMI with all-cause and cardiovascular mortality among elderly diabetes patients. The major findings of the current study were as follows: (1) a significant correlation between the TyG-BMI index and all-cause and cardiovascular mortality in old diabetic patients, even after adjusting for potential confounding variables; (2) a U-shaped relationship was observed between baseline TyG-BMI index and all-cause mortality; (3) a linear relationship was observed between baseline TyG-BMI index and cardiovascular mortality.

IR is well acknowledged for its role in pathogenesis of type 2 DM, dyslipidemias, and obesity in general population [20, 21]. In addition, IR is linked to the development of CVD, and serves as a predictor for cardiovascular outcomes and all-cause mortality [22,23,24]. TyG-BMI index, which integrates TG, fasting plasma glucose and BMI, serves as a simple and feasible assessment tool for IR [25]. Many studies supported the close association between TyG-BMI index and visceral adiposity [26], type 2 DM [27], and metabolic syndrome [28]. The clinical value of TyG-BMI index on predicting occurrence of hypertension [29], nonalcoholic fatty liver disease [30], DM [31], and severe coronary artery disease [32] have been demonstrated in previous studies. Apart from the substantial correlation with various illness, TyG-BMI index was also useful for predicting clinical outcomes. In patients with heart failure, lower TyG-BMI index was associated with higher risk of one-year mortality [33]. In elderly population undergoing percutaneous coronary intervention, lower TyG-BMI index was correlated with higher risk for major adverse cardiac and cerebrovascular events [34]. These evidences implied that TyG-BMI index could become a valuable predictor for prognosis.

In our study, TyG-BMI index revealed similar predictive value for all-cause mortality in old diabetic patients; specifically, a U-shaped relationship between baseline TyG-BMI index and all-cause mortality was revealed. The U-shaped association between IR indicator and poor prognosis has been established in previous studies utilizing NHANES data. Zhang et al. observed a U-shaped association between TyG index with all-cause mortality in CVD patients with DM or pre-DM [35]. Sun et al. found a similar U-shaped association between TyG index and all-cause mortality in elderly US population [36]. The results were consistent in Li et al.’s study, where a U-shaped association was revealed with all-cause mortality in patients with CVD [37]. The mechanisms underlying this U-shaped association remain uncertain, but they may encompass the following aspects.

To begin with, a high TyG index is indicative of hyperglycemia and high TG level, were associated with adverse outcomes. Persistent hyperglycemia is related to a higher risk of death in nondiabetic patients with intracerebral hemorrhage [38]. For ischemic stroke patients, hyperglycemia is correlated with early neurological degeneration, 3-month poor outcome, and mortality after thrombolysis [39]. Uncontrolled hyperglycemia in the long run is associated with increased risk of micro- and macrovascular complications and mortality [40, 41]. High TG level, on the other hand, is an important contributor of residual cardiovascular risk. It may play a role on the development of CVD [42]. Marston et al. conducted meta-analysis of randomized controlled trials, and supported that lowering TG was associated lower risk of major vascular events [43]. Both hyperglycemia and high TG level can initiate oxidative stress, intensify inflammatory responses, promote the formation of foam cells, impair endothelial function, stimulate the proliferation of smooth muscle cells, escalate sympathetic nervous system activity, enhance renal sodium retention, raise blood pressure, and consequently increase cardiac load and lead to vascular damage [42, 44, 45].

Furthermore, a low TyG index reflects hypoglycemia and low TG level, and could also be harmful. Firstly, hypoglycemia induces platelet hyperactivity through the elevation of inflammatory and oxidative stress markers [46, 47]. It stimulates the production of adrenaline, a potent counter-regulatory hormone that further enhances platelet activation [48]. As a result, the platelet-related protein responses triggered by hypoglycemia increase the risk of cardiovascular and cerebrovascular events [49, 50]. Secondly, prior researches indicated that low TG level were associated with increased risk for hemorrhagic stroke [51,52,53,54], cardiac death in heart failure [55], in-hospital and late major adverse events in patients with ST-segment elevation myocardial infarction treated with primary percutaneous coronary intervention [56], increased incident DM in normoglycemic males with non-alcoholic fatty liver disease [57], and mortality in septic patients [58]. The situation is referred as “TG paradox”. The underlying mechanism is yet unclear, but might be explained by the inability to maintain a stable state of cell membrane and poor nutrition under the circumstance of low TG level [59].

Last but not least, BMI measures the amount of body fat and offers insights into nutritional status, and when compared to TyG index alone, TyG-BMI index is believed to possess an augmented risk assessment capacity, owing to its inclusion of obesity- and malnutrition-related information [27]. The relationship between both overweight or obese individuals and increased mortality risk has been demonstrated in large cohort studies recruiting general population [60]. Obesity is directly linked to the development of dyslipidemia, type 2 DM, and hypertension [61], and is associated with CVD and cardiac death. The primary mechanisms underlying these associations involve the production of inflammatory mediators, reactive oxidative stress, and IR [62]. More recent findings, however, found a U-shaped relationship between BMI and mortality in patients with DM, hypertension, heart failure, and CAD [63, 64]. Malnutrition is a risk factor for unfavorable outcomes in hospitalized patients [65], possibly via increasing circulatory inflammatory mediators, worsen IR and disturbing metabolic states [66]. Particularly, in old population, malnutrition is associated with frailty, cachexia, sarcopenia [67], and a strong predictor for worse clinical outcomes [68].

To sum it up, TG, fasting plasma glucose and BMI are integral components of the TyG-BMI index. Previous studies have demonstrated a U-shaped association of each of these components with clinical outcomes. These findings, in conjunction with our own, highlight the clinical significance of maintaining an optimal TyG-BMI index, as deviations to either extremely high or low levels can result in detrimental health consequences.

Our study had limited data access to diabetic-significant variables such as previous hypoglycemic events, which may significantly influence the outcomes in our study population. Future research should integrate a more comprehensive array of variables that have demonstrated significance in influencing outcomes among diabetic populations. Second of all, despite adjusting for potential confounding factors in the multivariate model and subgroup analyses, the inherent nature of observational studies and the restricted sample size for each subgroup necessitate careful interpretation of results due to the possibility of residual confounders. Finally, the lack of significance in the univariate cox regression for TyG-BMI and cardiovascular death and the absence of a similar U-shaped association could be attributed to the significantly younger age in the highest TyG-BMI quartile, which was not adjusted for in the univariate analysis but was in multivariate models 2 and 3. As our observational retrospective cohort study has a low incidence of cardiovascular death, the relationship between TyG-BMI and cardiovascular outcomes should be further investigated in randomized controlled trials with larger cohorts.

Conclusions

In the current study, the TyG-BMI index is a valuable index for predicting the risk of all-cause and cardiovascular mortality in elderly patients withDM, the association between the TyG-BMI index and all-cause mortality is U shaped. While the association between the TyG-BMI index and cardiovascular mortality is linear. The TyG-BMI index can serve as a convenient tool for identifying IR, may be beneficial in assessing the risk and predicting the prognosis of elderly diabetes patients.