Association of Insulin Resistance With Cardiovascular Disease and All-Cause Mortality in Type 1 Diabetes: Systematic Review and Meta-analysis

Cardiovascular disease (CVD) is the main driver of morbidity and mortality in individuals with type 1 diabetes (T1D) (1). Despite current guidelines emphasizing the importance of managing conventional cardiovascular risk factors alongside glycemic control (2–4), the absolute risk of cardiovascular events remains high in this population, suggesting that our knowledge of CVD risk factors and mechanisms in T1D is incomplete (5) and that other mechanisms or risk factors may be involved.

Mechanistic studies have shown that insulin resistance (IR) elevates cardiovascular risk by mediating the development of metabolic disorders (hyperglycemia, hypertension, and dyslipidemia), endothelial dysfunction, and a state of low-grade inflammation (6,7). The presence of IR, a pathogenic basis of type 2 diabetes (T2D), has been shown to also be common in T1D, both in youth and adults (8,9). For example, findings from the U.S. population in the Diabetes Control and Complications Trial/Epidemiology of Diabetes Interventions and Complications (DCCT/EDIC) study revealed that 36% of patients with T1D showed evidence of IR (10). More importantly, IR is clearly present in people of normal weight with T1D and is exacerbated by obesity, which is increasing among individuals with T1D (11,12). Recently, it was proposed that IR in T1D may account for excess cardiovascular risk (1,13).

The gold standard test for measuring insulin sensitivity is the euglycemic-hyperinsulinemic clamp. However, this technique is relatively invasive, time consuming, and expensive, and may therefore not be suitable for large-scale clinical use (13). Subsequently, a surrogate measurement of the whole-body, namely, the estimated glucose disposal rate (eGDR), has been proposed (14). This technique was originally developed and validated with the euglycemic-hyperinsulinemic clamp in a subset of 24 adults with T1D by the landmark Pittsburgh Epidemiology of Diabetes Complications (EDC) study. The eGDR (in mg·kg⁻¹·min⁻¹) can be calculated using the following routine clinical measures: glycated hemoglobin (HbA_1c), presence of hypertension, and waist circumference or waist-to-hip ratio (14). The eGDR has been used as a diagnostic criterion for IR, with lower eGDR values indicating higher degrees of IR (15). eGDR values have been reported to exhibit good correlation with IR, as measured by the euglycemic-hyperinsulinemic clamp and validated for the estimation of insulin sensitivity in individuals with T1D (16).

In recent years, studies have reported that the eGDR is closely associated with the risk of CVD and all-cause mortality in individuals with T1D. One study found the eGDR is an independent predictor for major cardiovascular events and all-cause mortality after adjustment for multiple confounders, including IR-related parameters (17). Despite these findings, the potential application of the eGDR index in managing cardiovascular and all-cause mortality risk in T1D remains to be validated. Addressing this gap, we performed a systematic review and meta-analysis of available evidence from observational studies to evaluate the associations between IR, as assessed by the eGDR, and the risk of incident CVD and all-cause mortality in individuals with T1D.

We performed a literature search in accordance with the recommendations of the Meta-analysis of Observational Studies in Epidemiology (MOOSE) group (18). The protocol was registered prospectively with PROSPERO (CRD42023464717 [https://www.crd.york.ac.uk/prospero/display_record.php?ID=464717]). Electronic databases (PubMed, Embase, and the Cochrane Library) were searched, from inception to 15 March 2023, by using a text search strategy combining Medical Subject Headings and multiple terms associated with “insulin resistance,” “cardiovascular disease,” “all-cause mortality,” and “type 1 diabetes mellitus.” There were no restrictions on language. The search was updated on 31 October 2023. Supplementary Table 1 provides full details of the search strategy. In addition, we manually searched the reference lists of included articles.

Studies were included for analysis if they met the following criteria: 1) observational studies (prospective or retrospective cohort studies or post hoc analysis of clinical trials) on patients with T1D; 2) studies in which the eGDR was calculated; 3) studies reporting adjusted hazard ratios (HRs), with 95% CIs, for composite CVD, all-cause mortality, or type-specific CVD associated with the eGDR in individuals with T1D; and 4) studies that did not impose restrictions on age and included both young people and adults.

Studies were excluded if they were cross-sectional studies, case-control studies, reviews, editorials, and comments, and if the length of follow-up was <1 year. If multiple articles were derived from the same cohort and reported the same associated outcome, only the latest published report was included for analysis. Potential studies in languages other than English were translated with the aid of translation software or translators, if necessary. Study selection was performed in two phases: primary screening of titles and abstracts, followed by a full-text review of potentially eligible articles. Two review authors (R.S. and J.W.) independently evaluated eligibility, with discrepancies resolved by a third investigator (J.L.).

Two authors (B.L. and M.L.) independently extracted data from eligible studies using piloted data extraction sheets. Extracted data included the first author, publication year, country, duration of follow-up, study design, data source (national or local database), enrollment period, mean age, sex, sample size, calculation of the eGDR, outcome variables of interest, definition of cardiovascular diseases, and adjustment variables.

The primary outcome was the association of IR with composite CVD and all-cause mortality. The secondary outcome was the association between IR and type-specific CVD. CVD was defined as the composite of CVDs (including angina pectoris, myocardial infarction, coronary artery disease [CAD], cardiovascular procedures, heart failure [HF], cardiovascular death, peripheral arterial disease, and stroke).

The eGDR (in mg·kg⁻¹·min⁻¹) was calculated using three equations: Eq. 1 = 24.31 − (12.22 × WH) − (3.29 × HT) − (0.57 × HbA_1c); Eq. 2 = 24.4 − (12.97 × WH) − (3.39 × HT) − (0.60 × HbA_1c); Eq. 3 = 21.158 − (0.09 × WC) − (3.407 × HT) − (0.551 × HbA_1c), where WC is waist circumference, WH is waist-to-hip ratio, and HT represents history of hypertension (0 = no, 1 = yes). Hypertension was defined based on a reported history or current diagnosis of physician-diagnosed hypertension (blood pressure ≥140/90 mmHg) or treatment with antihypertensive medication.

Two reviewers (R.S. and J.W.) independently assessed the quality of the selected studies using the Newcastle-Ottawa quality assessment scale (19): studies were assessed based on selection (four items, 1 point each), comparability (one item, up to 2 points), and exposure/outcome (three items, 1 point each), with a maximum score of 9 points. The quality of the studies was graded as poor (<4 points), fair (4 to 6 points), and good (≥7 points).

Two investigators (J.W. and Y.P.) independently assessed the certainty of evidence for each outcome, with discrepancies resolved by a third reviewer (R.S.). Certainty of evidence was evaluated using the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) framework, which divides evidence into the categories of very low, low, moderate, and high certainty levels (20). The quality of evidence from observational studies was initially categorized as low and then upgraded or downgraded based on predefined criteria. Quality can be upgraded for large effect sizes (risk estimates of >2 or <0.5 in the absence of plausible confounders), dose-response gradients, or attenuation of the pooled risk estimates by plausible confounders. Conversely, quality can be downgraded for risk of bias (≥25% of the contributing studies were assessed as having serious risk of bias), inconsistency (substantial between study heterogeneity, I² ≥50%), indirectness (presence of factors limiting the generalizability of the results), imprecision (if 95% CIs for risk estimates are wide or cross a minimally important difference of 10% for outcomes [HR, 0.9–1.1]), and publication bias (evidence of small-study effects).

All statistical analyses were performed using Stata 16.0 statistical software. The P values were two-sided, with the values of 0.05 considered to indicate statistical significance. Random-effects models (DerSimonian and Laird method) were used to calculate pooled HRs with 95% CIs for the association of IR with risk of CVD and all-cause mortality in T1D. We selected risk estimates from the multivariate models that were fully adjusted for confounders. Studies reported HRs in different ways. These including HRs describing those as categories variable, and per-unit increase or per-SD increase in the eGDR. To enable direct comparison of study-specific HRs, the studies that reported HRs were recalculated as the HR for per-unit increase of eGDR for studies not reporting HRs for per-unit increase of the eGDR (21). Using dose-response relationships, we transformed HRs of categories into HRs of per-unit eGDR increase (22). The method was based on data in each study of categories of eGDR levels, number of events and participants in each group, and the adjusted HR and its corresponding 95% CI in each group (22). This transformation was performed under the assumption that the exposure variable is normally distributed and there is a log-linear association between exposure and outcome, as previously described (23–24). The heterogeneity across studies was quantified using the I² statistic (0–25%, low heterogeneity; 25–50%, moderate heterogeneity; 50–75%, substantial heterogeneity; and 75–100%, high heterogeneity).

To identify the subgroup differences and potential sources of the observed heterogeneity, subgroup analyses were performed after stratifying for study design (prospective vs. retrospective vs. post hoc analysis of clinical trials), CVD at baseline (yes or no), median duration of follow-up (>10 years vs. ≤10 years), calculation of the eGDR (Eq. 1 vs. Eq. 2 vs. Eq. 3), median sample size (>1,000 vs. ≤1,000), and level of adjustment (++ vs. +++). A P value of <0.10 for differences in estimates between these subgroups was considered to indicate significance. Given the nature of secondary data capture and analysis, the patients and public were not involved in the design or interpretation of this study and its findings. The results of this review will be disseminated to appropriate audiences with proper caution.

In our initial and updated search, we identified 1,664 records after the removal of duplicates. After titles and abstracts were screened, the full text of 41 articles was subsequently reviewed. Eight studies were eligible for data extraction and quantitative analysis, reporting data on 21,930 participants (10,17,25–30). Supplementary Fig. 1 shows the flow of records through the review, and Supplementary Table 2 summarizes the characteristics of the included articles. The included studies were based on six data sets and published between 2002 and 2021. Of these studies, three were from the U.S. (25,26,28), one each from Canada and the U.S. (10), Italy (17), Sweden (27), Finland (30), and the Netherlands (29). The studies comprised prospective (17,25–27,29,30) or retrospective cohort designs (27) or post hoc analysis of randomized clinical trials (10). The data for three studies were sourced nationwide (10,27,30), whereas the remaining four reported local data (17,25,26,28,29). Five studies reported the risk of composite CVD (10,17,27–29), five studies reported the risk of all-cause mortality (17,26,27,29,30), and two reported the risk of lower-extremity arterial disease (LEAD) (25) and CAD (17). The eGDR was calculated using three equations. Supplementary Table 2 describes the details of the eGDR reported in the included studies. Supplementary Table 3 shows the details of the levels of covariate adjustment used in the included studies. Overall, according to the Newcastle-Ottawa quality assessment scale, seven studies were graded as good quality. Supplementary Table 4 presents a detailed assessment of quality for each domain.

Five studies (10,17,27–29) involving 19,960 individuals with T1D from five data sets reported data for the association between the eGDR and risk of composite CVD outcome in T1D. Table 1 summarizes the characteristics of the included articles for CVD outcome. The median follow-up time was 10 years (range, 7.1–25 years). Among 19,960 patients with T1D, 9,237 participants (46.3%) were women, and 2,149 developed the composite CVD outcome. Pooled data showed that for every 1-unit increase in eGDR, the risk of composite CVD decreased by 17% (HR 0.83, 95% CI 0.78–0.90). Heterogeneity was observed for HRs of composite CVD (I² = 58.9%) (Fig. 1A). Supplementary Table 5 summarizes the quality of evidence based on the GRADE framework. For the outcome of composite CVD, these findings were considered as very low-quality evidence.

Five studies (17,26,27,29,30) involving 19,403 individuals with T1D from five data sets reported data for the association between eGDR and risk of all-cause mortality in T1D. Table 2 summarizes the characteristics of the included articles for all-cause mortality outcome. The median follow-up time was 10 years (ranging from 7.1 to 16.6 years). Among 19,403 patients with T1D, 8,571 (44.2%) participants were women, and 1,279 experienced all-cause mortality events. When pooling data from the studies, for every 1-unit increase in eGDR, the risk of all-cause mortality decreased by 16% (HR 0.84, 95% CI 0.81–0.87, I² = 0%) (Fig. 1B). Supplementary Table 5 summarizes the quality of evidence based on the GRADE framework. For the outcomes of all-cause mortality, these findings were considered as very low-quality evidence.

Two studies reported the association between eGDR and the risk of type-specific CVD in T1D, respectively (17,25). Garofolo et al. (17) examined the association between the eGDR and risk of CAD in 774 individuals with T1D (35 CAD events, 52.6% women), and found that patients had an HR of 0.82 per 1-unit increment of the eGDR index (95% CI 0.68–0.98). Olson et al. (25) investigated the association between the eGDR and risk of LEAD in T1D through a study involving 586 individuals with TID (70 LEAD events, 51.4% women), and found that female patients had an HR of 0.66 per 1-unit increment of the eGDR index (95% CI 0.55–0.79).

To investigate the presence of subgroup differences for the outcome of composite CVD, we conducted subgroup analyses based on the characteristics of eligible studies, including study design, CVD at baseline, follow-up period, measurement for the eGDR, sample size, and level of adjustment. Significant differences between subgroups were detected by measurement for the eGDR (P = 0.019) (Fig. 2).

The findings of this systematic review and meta-analysis of eight studies indicate that a lower eGDR is associated with incident CVD events and all-cause mortality in T1D. We found that for every 1-unit increase in eGDR, the risk of composite CVD outcome decreased by 17% and that the risk of all-cause mortality decreased by 16%. Hence, IR in T1D could, at least partially, account for excess cardiovascular risk and all-cause mortality in individuals with T1D.

To our knowledge, this systematic review and meta-analysis represents the first comprehensive examination of the associations of IR with the risk of incident CVD and all-cause mortality in patients with T1D.

In recent years, a growing body of observational studies has highlighted that IR is associated with an elevated risk of adverse cardiovascular outcomes and all-cause mortality in T1D. A previous meta-analysis concentrated solely on exploring the relationship between IR and T1D (31). For instance, a meta-analysis examining IR in adults with T1D, as assessed by euglycemic-hyperinsulinemic clamp, found that IR is a prominent characteristic in this population (31). In our systematic review and meta-analysis, we found that the pooled maximally adjusted HRs for CVD per 1-unit change in the eGDR was 0.83 (95% CI 0.78–0.90). Additionally, two studies in the included literature investigated the association between the eGDR and risk of in type-specific CVD, reporting a significant association of the eGDR with CAD (HR per 1-unit increment of the eGDR index: 0.82, 95% CI 0.68–0.98) and LEAD (HR per 1-unit increment of the eGDR index: 0.66, 95% CI 0.55–0.79). Furthermore, our data indicated that the pooled maximally adjusted HRs for all-cause mortality per 1-unit change in eGDR was 0.84 (95% CI 0.81–0.87).

In the subgroup analyses, associations between IR and CVD outcomes in patients with T1D varied depending on the eGDR calculation. The eGDR was originally developed and validated with the euglycemic-hyperinsulinemic clamp in a subset of 24 adults with T1D by the landmark Pittsburgh EDC study, which found that the eGDR values had a high correlation with those obtained with the clamp (r = 0.79) (13). Subsequent researchers have found that it has also been used in children and adolescents with T1D (5,25), with a recent attempted use in T2D (32). The equation for the calculation of eGDR was originally as follows: 24.31 – (12.22 × WH) − (3.29 × HT) − (0.57 × HbA1). Subsequently, this equation was modified for the use of HbA_1c in place of HbA1, directly replacing the latter with the former, as in Eq. 1: 24.31 – (12.22 × WH) − (3.29 × HT) − (0.57 × HbA_1c) (29) or as shown in Eq. 2: 24.4 − (12.97 × WH) − (3.39 × HT) − (0.60 × HbA_1c) (30). Additionally, some studies substituted WC for waist-to-hip ratio (WHR) in the eGDR formula owing to the fact that the research does not include WHRs, which is described as Eq. 3: 21.158 − (0.09 × WC) − (3.407 × HT) − (0.551 × HbA_1c) (27). A recent study has also used BMI as a substitute for WC, which is described as: 19.02 − (0.22 × BMI) − (3.26 × HT) − (0.61 × HbA_1c) (33).

In our study, we observed that three equations were used to calculate the eGDR in the included studies. Furthermore, we found that IR calculated using Eq. 1 [24.31 − (12.22 × WH) − (3.29 × HT) − (0.57 × HbA_1c)] demonstrated a stronger association with CVD. However, given the limited number of reports included in each subgroup, we were unable to rule out spurious results in multiple subgroup analyses. Furthermore, in the original study by Williams et al. (13). WC and BMI were also tested, with an association comparable to that for WHR.

IR is characterized by reduced sensitivity and responsiveness to the action of insulin (34). In individuals with T1D, factors such as obesity, a sedentary lifestyle, a family history of T2D, and weight gain associated with intensive insulin therapy may contribute to IR. Exogenous insulin therapy might induce IR in patients with T1D (35). Individuals with T1D exhibit heightened peripheral hyperinsulinemia compared with those without diabetes owing to the subcutaneous injection of exogenous insulin, which may potentially exacerbate peripheral IR, partly through the downregulation of insulin receptors and GLUT-4 (10). Additionally, reduced hepatic insulin exposure leads to a decrease in circulating IGF-1, which, coupled with a parallel increase in growth hormone and IGF-binding proteins, may also contribute to heightened peripheral IR (36). Furthermore, obesity-induced IR has been proposed to be an environmental contributor to the increased incidence of T1D (12). Although the traditional perception remains that individuals with T1D are lean with a normal body weight, the prevalence of overweight and obesity in those with T1D has shown an alarming rise in parallel with the global population trends in weight gain (37). This increase can be partly attributed to general population trends and, additionally, to iatrogenic weight gain caused by intensive insulin treatment (12). Finally, it is plausible that the genetic and lifestyle factors associated with T2D may occur at a similar frequency in individuals with T1D, aligning with the substantial data on a family history of T2D (35).

Numerous molecular mechanisms underlie the association between IR and CVD, including metabolic flexibility, endothelial dysfunction, coagulation disorders, and smooth muscle cell dysfunction (7). Individuals with IR are prone to developing various metabolic disorders, such as hyperglycemia, dyslipidemia, and hypertension, all of which are strongly linked to poor CVD outcomes (10). IR induces endothelial cell dysfunction by reducing the production of nitric oxide from endothelial cells and increasing the release of procoagulant factors, leading to platelet aggregation (34). Moreover, IR has been found to be associated with reduced cardiac autonomic function, low-grade inflammation, and macrophage activity (34,38,39). These multiple mechanisms collectively contribute to the formation of atherosclerotic plaques and CAD.

Additionally, previous studies have shown that IR promotes excessive glycosylation, which can stimulate smooth muscle cell proliferation, collagen cross-linking, and collagen deposition. These pathological events contribute to increased diastolic left ventricular stiffness, cardiac fibrosis, and, ultimately, HF (7,10).

Our findings have implications for clinical practice in the T1D population: specifically, the results indicate that clinicians should focus more on the identification of patients with elevated IR, alongside managing glycemic control and traditional cardiovascular risk factors. This could be achieved through simple measurement of WC or WHR in addition to regular blood pressure and HbA_1c checks. Further studies, including those examining correlations with lifestyle modifications (diet and exercise) and pharmacological interventions, are necessary to validate these findings.

This study had several advantages over previous meta-analyses. First, we examined the association between IR and both CVD as well as all-cause mortality in individuals with T1D. Second, we assessed IR using the surrogate measurement of the eGDR, which is simple and noninvasive. Third, we excluded cross-sectional studies, thereby avoiding potential recall bias associated with study designs of a cross-sectional nature.

However, we acknowledge that the current study has some limitations. First, the lack of access to individual participants’ data prevented us from completely ruling out potential confounders. Indeed, some studies did not consistently collect data for relevant cardiovascular risk factors in T1D such as smoking, diabetes duration, family history of T2D, and race and ethnicity. Second, evidence on the association between IR (as estimated by the eGDR) and CVD in T1D patients is, to a large extent, limited.

The small number of available studies restricted our ability to perform meta-regression analyses or more detailed subgroups analyses (e.g., by sex, race and ethnicity, and duration of diabetes). Therefore, the results need to be further validated. Furthermore, the assessment for publication bias could not be conducted owing to the relatively small number (n < 10) of studies. Finally, as with all meta-analyses, the current study was limited by the quality of the included studies. Interpretation of evidence from observational studies requires caution, given these study types are prone to selection bias, recall bias, and exaggeration of associations. Using the GRADE framework, we identified very low-quality evidence for study outcomes in this review. We expect the quality of evidence to improve with future up-to-date research and more high-quality studies.

The evidence from this systematic review and meta-analysis demonstrates that IR, as calculated by the eGDR, may be an additional risk factor for CVD and all-cause mortality in individuals with T1D. Modification of IR through lifestyle interventions or pharmacological treatments may be beneficial in managing these risks and improving CVD outcomes for individuals with T1D.

Funding. This work was supported by the National Key R&D Program of China (Grant No. 2022YFC3500101).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. R.S., J.W., and L.Z. contributed to the study concept and design and revised the draft. R.S., J.W., M.L., Y.P., B.L., and G.Y.H.L. were involved in acquisition of data and statistical analyses. R.S., J.W., and J.L. performed the search strategy and contributed to database research. All authors participated in data analysis, reviewed the manuscript, and read and approved the final manuscript. L.Z. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Handling Editors. The journal editors responsible for overseeing the review of the manuscript were Elizabeth Selvin and Kristen J. Nadeau.