Introduction

The prevalence of in-hospital mortality among patients in intensive care units (ICUs) is as high as 20–30% [1], and ICUs often account for 20–50% of all hospital deaths [2]. However, in general hospitals, the ICU accounts for 10%–15% of hospital beds, including 22% of the total costs [3]. Patients in the ICU typically experience an extended mean length of stay and face an exceptionally high mortality rate, imposing a substantial burden on families and society [4]. Hence, predicting risks in advance is crucial for providing medical treatment guidance. The ICU encompasses a spectrum of various diseases, from sepsis to trauma, comma, and ischemic stroke [5]. However, most ICU patients exhibit increased insulin resistance due to inflammatory stress, which is a marker of both systemic inflammatory responses and metabolic disorders [6]. Previous studies demonstrated a 50–70% reduction in insulin sensitivity in critically ill patients and reported that this reduction is associated with illness severity rather than various admission diagnoses [4, 7]. Furthermore, insulin resistance has been related to significant morbidities in intensive care [8]. Thus, changes in insulin resistance are considered useful indicators of stress response-associated inflammation [9]. The triglyceride-glucose (TyG) index, which contains two biomarkers, fasting triglyceride (TG) and fasting blood glucose (FBG), is an easily accessible, cost-effective, and reliable surrogate indicator of insulin resistance [10] that can facilitate its application in clinical practice. Several studies have shown that TyG has a positive relationship with many diseases, such as cardiovascular diseases [11], diabetes [12], bladder cancer [13], and hypertension [14]. In addition, an observational study reported that the TyG index could be used as a predictor of hospital and ICU mortality in critically ill stroke patients [15]. A recent study revealed a strong association between the TyG index and increased all-cause mortality in critically ill patients [10]. Another scholar reported that the TyG index could be a significant predictor of severe impairment of consciousness and in-hospital death in patients with cerebrovascular disease in the ICU [16]. Furthermore, many investigations have verified the positive relationship between the TyG index and coronary atherosclerosis progression [17]. This index is a valuable approach for simultaneously analyzing lipid metabolism and glucose status, which is why it has been widely assessed in many different diseases [18]. Although many studies have investigated the relationship between the TyG index and different diseases, no studies have compiled this evidence and reached a clear conclusion. Therefore, we performed this systematic review and meta-analysis to determine whether the TyG index is associated with the risk of in-hospital and ICU mortality.

Methods

The current study followed the guidelines of the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) statement. The protocol of this study was registered in the International Prospective Register of Systematic Reviews (PROSPERO) with the registration ID CRD42024541439.

Search strategy

A comprehensive and systematic literature search, with no restriction on publication date, was performed through online databases, including PubMed, Scopus, and Web of Science, up to 21 Jan 2024 to find related studies. The search terms “TYG” and “triglyceride-glucose index” in combination with “in-hospital” and “ICU”-related mortality were included in the search strategy (Supplementary Table 1). Furthermore, the reference lists of eligible studies, and the first 4 pages of Google Scholar, were manually checked to find other potentially relevant studies.

Study selection

EndNote software version 20.4.1 was used to manage all the retrieved articles. After removing duplicate studies, two reviewers independently assessed the eligibility of the remaining studies according to the PECOS (population, exposure, comparator, outcome and setting) scheme, which considered hospitalized patients aged >18 years old as the population; the “TYG” index as the exposure; the highest compared to lowest “TYG” index as a comparator; “in-hospital” or “ICU” all-cause mortality as the outcome; and all prospective or retrospective cohort studies as the setting. In this review, letters, comments, reviews, randomized controlled trials, case reports, cross-sectional and case-control studies, studies on Coronavirus disease 2019 (COVID-19) patients, and those with insufficient data were excluded. Two reviewers cross-checked the excluded articles, and any discrepancies were resolved through discussion.

Data extraction

Two researchers screened the titles, abstracts, and full texts of the obtained articles to identify eligible studies. Then, all necessary data were extracted according to an arranged screening form that included the following items: the first author’s name, publication year, cohort-based population, study design, outcome, characteristics of participants, number of participants with mortality, mean or range of age, sex and any adjustment for confounding variables. If a study reported different risk estimates for mortality, the effect size in the fully adjusted model was selected and used in the meta-analysis. In cases where adequate data for outcomes were not reported, we contacted the corresponding author by email. Discrepancies between authors were resolved at any stage through consultation and discussion with the principal investigator.

Quality assessment of the studies

The Newcastle‒Ottawa Scale (NOS) was used to assess the quality of the included studies. The current checklist includes three domains: (1) selection (population representativeness, which can take a maximum of 4 stars), (2) comparability (controlling appropriate confounders in study design or during statistical analyses with a maximum of 2 stars), and (3) outcomes, with a maximum of 3 stars. A maximum of 9 stars for each study indicates the highest quality. Likewise, studies with more than 6 stars were considered high-quality, those with 4–6 stars were rated as medium-quality, and studies with 0–3 stars were considered low-quality. The quality assessment was independently checked by two authors, and any disagreements between the researchers were resolved through discussion with the principal investigator.

Statistical analysis

In all included studies, the relative risks (RRs) and their 95% confidence intervals (CIs) were considered the main effect sizes. A random effects model taking the between-study heterogeneity into account was used to estimate the pooled effect size. The assessment of interstudy heterogeneity was based on Cochrane’s Q test and the I-squared test (I2). Additionally, subgroup analysis based on participants’ disease (cardiovascular or cerebrovascular diseases) was performed to explore the possible source of heterogeneity among studies using the random effects model.

To calculate the linear dose-response relationship across categories of the TYG index, the methods proposed by Greenland and Longnecker [19] and Orsini et al. [20]. were applied where the number of mortality cases and the adjusted effect size (ORs, RRs, or HRs) for at least 3 categories of the TYG index were needed. Based on this method, the midpoint of exposure in each category of the TYG index was considered the relevant effect size for each study. In cases where the dose of the exposure was expressed in a range, the midpoint in each single category was calculated via the mean of the minimum and maximum values. To detect whether the overall pooled results might be affected by a specific study, a sensitivity analysis was performed by removing the studies from the analysis one by one. This also was the case for the study by Cheng et al., where we had to manually calculate the unadjusted effect size based on the reported crude data [21]. Likewise, publication bias tests were performed using visual examination of funnel plots and checked statistically by Egger’s and Begg’s regression tests. If the findings of publication bias were significant, a trim-and-fill analysis was performed to detect the effect of possible unpublished articles on the pooled results. All the statistical analyses were performed with Stata, version 14 (Stata Crop). Values < 0.05 were considered to indicate statistical significance.

Results

Literature search

The detailed processes of the study selection are provided in Fig. 1. In our primary search, 1048 studies were identified. After excluding duplicates and irrelevant studies, 30 potentially eligible studies remained for further assessment. Finally, by reviewing the full texts, an additional 11 studies were removed for the following reasons: nine studies reported outcomes not relevant to the purpose of the current study, one did not assess the TYG index, and one study by Şaylık et al. [22]. did not provide sufficient data. Thus, 19 unique eligible cohort studies [4, 10, 15, 16, 21, 23,24,25,26,27,28,29,30,31,32,33,34,35,36] were included in the final meta-analysis.

Fig. 1
figure 1

The process of study selection.

Full size image

Study characteristics

Table 1 shows the characteristics of the included studies evaluating the associations of TYG with in-hospital and ICU mortality. The outcomes were in-hospital mortality in 18 studies [4, 10, 15, 16, 21, 30,31,32,33,34,35,36] and ICU mortality in 8 studies [4, 10, 15, 23, 29, 30, 34, 36]. Among the 42,525 participants, 5233 in-hospital and 1754 ICU mortality cases were reported by the included studies. All studies included both males and females with an age range of 40–79 years. Five studies included patients with cerebrovascular diseases [10, 15, 16, 30, 31], eight were conducted on patients with cardiovascular diseases (CVD) [21, 23, 32,33,34], two were performed on critically ill patients [4, 24], and individual studies were conducted on adult patients who were in their first hospital and ICU admission [25], patients with sepsis [36], patients suffering from severe fever with thrombocytopenia syndrome [35] and liver transplantation recipients [26].

Table 1 Main characteristics of studies examining the association between TYG index and the risk of in-hospital and ICU all-cause mortality.
Full size table

Except for two studies [27, 31], all the included studies had a retrospective design. A study by Chen et al. [16] used two different populations (patients with nontraumatic cerebral hemorrhage and patients with cerebral infarction). Therefore, we considered this study to have two separate effect sizes (Chen et al. A and Chen et al. B). Of the included studies, 10 used data extracted from the Medical Information Mart for Intensive Care (MIMIC) [4, 10, 16, 21, 25, 29, 30, 32, 34, 36], and three used the eICU Collaborative Research Database [15, 23, 33]. Moreover, except for a study conducted by Cheng et al. [21], all included studies adjusted for potential confounders, including age, sex, smoking status, body mass index, and some conventional risk factors, including alcohol intake, chronic disease morbidity, diabetes, dyslipidemia, hypertension, and race. Generally, the included studies originated from China (n = 6) [24, 31, 35] and the United States of America (USA) (n = 13) [4, 10, 15, 16, 21, 23, 25, 29, 30, 36].

Meta-analysis

TYG index and in-hospital mortality

Nineteen effect sizes from 18 studies investigated the association between the circulating TYG index and in-hospital all-cause mortality. The pooled analysis revealed a significant positive association between the TYG index and in-hospital mortality; in this way, each one-unit increase in the TYG index was associated with a 33% increase in the risk of in-hospital mortality (RR = 1.33; 95% CI: 1.23, 1.43). However, high heterogeneity was detected among studies (Cochrane’s Q test, I squared = 90.3%; P < 0.001) (Fig. 2). Similarly, subgroup analysis revealed that the TYG index was more strongly associated with the risk of in-hospital mortality in patients with cardiovascular disease compared to those with cerebrovascular disease (p for heterogeneity between groups = 0.014) (Fig. 3).

Fig. 2
figure 2

TYG index and in-hospital all-cause mortality.

Full size image
Fig. 3
figure 3

Subgroup analysis for TYG index and in-hospital all-cause mortality.

Full size image

TYG index and ICU mortality

The association between the TYG index and all-cause mortality in the ICU was assessed in eight studies. Overall, each unit increase in the TYG index was significantly associated with a 45% increase in the risk of all-cause mortality in the ICU (RR: 1.45; 95% CI: 1.25, 1.67). The results of Cochrane’s Q test revealed a low level of heterogeneity among studies (Cochrane’s Q test, I squared = 44.8%; P = 0.08) (Fig. 4). Subgroup analysis showed no significant interstudy heterogeneity (p heterogeneity between groups = 0.069) (Fig. 5).

Fig. 4
figure 4

TYG index and ICU all-cause mortality.

Full size image
Fig. 5
figure 5

Subgroup analysis for TYG index and ICU all-cause mortality.

Full size image

Quality assessment

Table 2 shows the detailed quality assessment of the included studies. The quality scores were 9 in 8 studies, 7 in one study, and 8 in 10 studies. Hence, the overall quality of all the studies was rated as high.

Table 2 Quality assessment of included studies on the association between TYG and in-hospital and ICU mortality.
Full size table

Publication bias

Although visual inspection of the funnel plot for studies on the association between TYG and in-hospital mortality showed asymmetry, Begg’s regression test indicated no significant publication bias (P = 0.11). Regarding ICU mortality, funnel plot examination and Egger’s regression test showed some evidence of possible publication bias. Therefore, the trim-and-fill method was used, and the results showed that the overall effect size did not change with the addition of missing studies (RR: 1.31, 95% CI: 1.123, 1.527).

Sensitivity analysis

Sensitivity analysis revealed that the overall association of TYG with in-hospital (95% CI: 1.23, 1.43) and ICU (95% CI: 1.44, 1.67) mortality was not affected by a single study.

Discussion

To the best of our knowledge, the present study was the first to pool the results of previous studies to provide a quantitative measure of the association between the TYG index and the risk of in-hospital and ICU all-cause mortality. Our findings showed that each one-unit increase in the TYG index was positively associated with a 33% and 45% increase in the risk of in-hospital and ICU mortality, respectively. As the primary studies were heterogeneous in terms of their participants, besides applying the random model effects, a subgroup analysis based on participants' health status was performed to find the source of heterogeneity. Notably, the subgroup analysis indicated a stronger association between the TYG index and the risk of in-hospital mortality in patients with cardiovascular diseases compared to those with cerebrovascular diseases.

These results indicate that the TyG index could serve as a valuable marker for risk stratification and predicting outcomes in critically ill patients. In critical illness, the metabolic response is an integral part of the body’s adaptive response and encompasses various organ systems, wherein energy resources are shifted to the area of utmost necessity [21]. Multiple mechanisms are triggered to increase the provision of energy to vital tissues, including the stimulation of the sympathetic nervous system, the release of pituitary hormones, and an increase in peripheral resistance to the impacts of anabolic factors [37]. During the early phase of the metabolic response, the oxidation of carbohydrates is significantly more enhanced than the oxidation of fats and proteins [38]. Subsequently, there is a decrease in glucose oxidation, an increase in fat turnover, and a reduction in muscle and visceral protein mass, resulting in wasting [39]. The final common metabolic pathways activated in response to critical illness include uncontrolled catabolism and the development of resistance to anabolic signals such as insulin. This process aims to reset the hierarchy of energy supplies by prioritizing the allocation of energy substrates to vital tissues over insulin-dependent organs [40].

The TyG index has been well identified as an available and reliable substitute for insulin resistance (IR) since it does not necessitate insulin levels and can be utilized by all individuals, including both patients and healthy subjects [25]. Concurrently, the fasting levels of triglycerides and glucose, which are essential for calculating the TyG index, can be easily obtained clinically [16]. In a recently published meta-analysis, a significantly higher TyG index was observed in patients with obstructive sleep apnea, compared to the healthy subjects [41].

Moreover, in a study involving 6091 patients, the TyG index was reported to be a stronger predictor of metabolic syndrome than the homeostatic model assessment of insulin resistance (HOMAIR) [42].

The relationship between the TYG index and the risk of cardiovascular or cerebrovascular diseases has also been investigated in several studies [11, 43]. Findings from a meta-analysis indicated a potential linear dose-response association between TYG and the risk of cerebrovascular disease [43]. Likewise, another meta-analysis conducted by Khalaji et al. revealed a significant positive association between the TYG index and the risk of heart failure [44]. Furthermore, Tao et al. demonstrated [11] that the TyG index can serve as a reliable and available substitute for IR and can be improved for risk classification and prognostication of CVD outcomes. However, data regarding critically ill patients are limited. According to a study on critically ill stroke patients, the TyG index could predict in-hospital and ICU mortality [15]. Moreover, a recent study by Zhang et al. reported a significant linear correlation between the baseline TYG index and in-hospital and ICU mortality in critically ill patients suffering from coronary heart disease [34]. Similarly, our findings indicated linear dose-response associations between the TYG index and in-hospital and ICU mortality. Additionally, findings from the subgroup analysis showed that the TYG index had a stronger association with the risk of in-hospital mortality in patients with cardiovascular diseases compared to those with cerebrovascular diseases. Recent epidemiologic studies have indicated that the TyG index can serve as an independent predictor for the risk of cardiovascular diseases and consequent outcomes related to both cardiovascular and cerebrovascular diseases [23, 45, 46]. In this study, the greater association of TYG with cardiovascular diseases might be due to the larger effect sizes (7 vs 6 effect sizes) and a greater number of total participants in this subgroup than in the “cerebrovascular diseases” subgroup (20,347 vs 12,954 patients).

Notably, ICU patients are typically confronted with unstable and progressive disorders, and the progression of acute illnesses such as sepsis, shock, or trauma can result in stress-induced hyperglycemia. Therefore, this condition may bias the diagnostic value of the TyG index [11]. In addition, the TyG index is influenced by several factors, such as race and alcohol intake [16]. Since the majority of the included studies in this meta-analysis used the baseline TYG index as a biofactor to predict mortality and did not control for potential confounders such as race and alcohol intake, our findings should be interpreted with caution.

Strengths and limitations

The strengths of the current study include a meticulous and comprehensive systematic search in online databases, the use of a thorough and rigorous methodology to pool the results of the available studies, the use of subgroup analysis to determine the source of heterogeneity, the assessment of the impact of a separate study on the overall effect measured by sensitivity analysis, and the examination of the effect of unpublished studies on the overall result using publication bias tests. However, it should be noted that the present study has certain limitations that have the potential to undermine our results. As the majority of the included studies had a retrospective design, the dynamic measures of TYG, to control the effect of unstable stress-induced hyperglycemia throughout the studies, were not assessed. Although all the included studies controlled for confounding variables such as age, sex, smoking status, and body mass index, most of them did not adjust for other potential confounders such as race and alcohol intake.

Conclusion

The outcome of this meta-analysis demonstrated that the TYG index had a significant relationship with the risk of in-hospital and ICU mortality. Furthermore, subgroup analysis revealed a stronger association between the TYG index and the risk of in-hospital mortality. Based on these findings, the TyG index has the potential to predict the risk of in-hospital and ICU mortality in hospitalized patients.