Prognostic value of high-sensitivity cardiac troponin for major adverse cardiovascular events in patients with diabetes: a systematic review and meta-analysis

Review protocol

This systematic review and meta-analysis was conducted according to the reporting guidelines of Meta-analysis of Observational Studies in Epidemiology (MOOSE) (Stroup et al., 2000) (checklist is shown in Appendix A). The protocol was published on the International prospective register of systematic reviews (PROSPERO) (CRD42021287365).

Data sources and searches

As for data sources, two authors (T.S. and Y.L.) searched MEDLINE, Embase, and the Cochrane Database for potential eligible original articles. The following keywords were used: “troponin T”, “troponin I”, “cardiac troponin”, “high-sensitivity”, “high-sensitive”, “diabetes mellitus”, “diabetes”. No language restriction was applied. We also manually screened the reference lists of included studies for potential articles meeting inclusion criteria. The electronic literature search was last updated on May 25th, 2023. The detailed search strategies are displayed in Appendix B.

Study selection

Two authors (T.S. and Y.L.) independently selected eligible studies based on standard inclusion and exclusion criteria. Disagreements were resolved by discussion with a third author (LJ). In this systematic review, observational cohort studies were included if they investigated the association between hs-cTn and risk of cardiovascular outcomes in patients with diabetes or prediabetes. Studies were excluded if they met the following standards: (1) the study did not report data regarding cardiovascular outcomes or mortality, (2) urine hs-cTn were measured instead of plasma hs-cTn.

Data extraction

Data collection was accomplished by two authors using a unified extraction form independently, and any discrepancy was handled by discussion from a third author (LJ). If one article contained data for different populations, for instance, prediabetes and diabetes, it was then extracted as separate studies.

The primary outcome was MACE, which was defined as the composition of incident cardiovascular events, and hospitalization or mortality with a primary or secondary diagnosis of cardiovascular diseases. Data of outcomes which involved definitions of incident cardiovascular diseases, or ischemic heart disease, or cardiovascular mortality were also extracted as MACE. The secondary outcomes included all-cause mortality, and incident heart failure.

Quality assessment

Two investigators independently assessed the risk of bias of included studies with Newcastle-Ottawa Scale (NOS) (Wells, O’connell & Peterson, 2000). The eligible cohort studies were scored on the following domains: representativeness of the study population, selection of non-exposed cohort population, assays of measuring hs-cTn, demonstration that outcomes of interest was not present at the beginning of study, comparability of cohort with appropriate adjustment for confounding factors, assessment of outcomes, adequate length of follow-up and loss to follow-up. Each included study was allocated stars based on the above domains (from zero to nine), and studies with a higher score than seven were determined to be high-quality.

Data synthesis and analysis

As our meta-analysis focused on long-term outcomes involving time-to-event data, hazard ratio (HR) and related 95% confidence interval (95% CI) were used to summarize the effect measures according to the Cochrane Handbook for Systematic Reviews of Interventions (version 6.2) (Higgins, Li & Deeks, 2021). The values of HR and 95% CI were obtained from Cox proportional hazards regression analyses on a continuous scale (per standard deviation (SD) change after log transformation) or a categorical scale (high versus low levels based on included studies). Generic inverse-variance methods were used to pool the results of the natural log transformed HR (lnHR) and related standard errors (SE (lnHR)) (Deeks, Higgins & Altman, 2021). As innate and unmeasurable heterogeneity may exist in study population and hs-cTn measurement assays across included studies, we adopted random-effect model to pool the results of each study (Deeks, Higgins & Altman, 2021).

The chi-squared (χ²) test and I² statistic were used to assess the inter-study heterogeneity (Higgins et al., 2003). The rough interpretation of the I² statistic was as follows: 0% to 40%, might not be important; 30% to 60%: moderate heterogeneity; 50% to 90%: substantial heterogeneity; 75% to 100%: considerable heterogeneity. Subgroup analyses were performed to address potential innate heterogeneity, stratified by types of diabetes and subunit of hs-cTn. In addition, several sensitivity analyses were performed to evaluate the robustness of main analyses: (1) only including perspective studies, as prospective studies usually have fewer potential sources of bias and confounding than retrospective studies; (2) only including high-quality studies derived from quality assessment; (3) only including studies adequately adjusted for demographics, cardiovascular risk factors, biomarker levels or medication use, etc.; (4) only including the population without major cardiovascular comorbidities at the start of study. In addition, Funnel plots and the Egger test were used to assess the potential for publication bias. Review Manager V.5.1 (The Nordic Cochrane Centre, København, Denmark) was used for statistical analysis.

Study characteristics

After screening 601 records retrieved from the systematic literature search, we finally identified 26 articles eligible for our study (Bidadkosh et al., 2017; Bluro et al., 2021; Busch et al., 2021; Cimaglia et al., 2021; Colombo et al., 2018; Costacou, Saenger, & Orchard, 2020; Galsgaard et al., 2017; Gori et al., 2016; Hendriks et al., 2016; Hillis et al., 2014; Junttila, 2016; Keller et al., 2018; Looker et al., 2015; Nguyen et al., 2020; Ohkuma et al., 2017; Pandey et al., 2021; Price et al., 2017; Resl et al., 2016; Saeed et al., 2021; Scirica et al., 2016; Sharma et al., 2020; Tang et al., 2020; Witkowski et al., 2021; Wong et al., 2019; Yiu et al., 2014; Zellweger et al., 2015). Four articles involved different cohorts, thus they were treated as separate studies (Galsgaard et al., 2017; Nguyen et al., 2020; Pandey et al., 2021; Tang et al., 2020). Therefore, a total of 30 studies were included in the quantitative analysis of our study. The detailed study selection flow diagram is shown in Fig. 1. Among the included studies, 29 were prospective cohort studies, one was a retrospective cohort study. Most studies measured the subunit of high-sensitivity cardiac troponin T (hs-cTnT), with only three examining high-sensitivity cardiac troponin I (hs-cTnI). Sixteen studies investigated population with type 2 diabetes, among which two studies included populations with prior stable coronary artery diseases or recent acute coronary syndrome, two studies with chronic kidney diseases. Four studies focused on patients with type 1 diabetes, and three on patients with prediabetes. In total, 62,419 participants were included. The median follow-up time of involved population was 5 (4.1–9.5) years. The detailed characteristics of involved studies are shown in Table 1.

Quality assessment

The median NOS score for included studies was 8, with 17 studies scored of 8, eight studies of 7, and five studies of 6. As for selection of study population, all included studies had optimal representatives of exposed cohort and selection of non-exposed cohort, as well as valid ascertainment of exposure, but ten studies partially contained patients with prior cardiovascular diseases. Most included studies (24/30) had adequate adjustment for potential confounding factors, involving demographics, comorbidities, biomarker levels, medications, etc. Detailed adjusted variables were shown in Table S1. Nearly all studies had a median follow-up period of over two years, but few of them mentioned the percentage of patients lost during follow-up. The summary of quality assessment was displayed in Fig. 2.

MACE

A total of 24 studies including 47,612 patients investigated the association between hs-cTn and risk of MACE. Eighteen studies evaluated the prognostic effect of hs-cTn from a continuous scale, and the pooled results suggested that elevation of hs-cTn was associated with a significantly increased risk of MACE in patients with diabetes (adjusted HR per standard deviation (SD) change 1.15, 95% CI [1.06–1.25], Fig. 3). No heterogeneity was observed among included studies (I² = 0%). When dividing the hs-cTn levels into high and low categories, we also found significant association between high hs-cTn and increased risk of MACE without heterogeneity (adjusted HR 1.73, 95% CI [1.11–2.68], I² = 0%). Most studies adopted 14 ng/L as the cutoff value of hs-cTn. No significant publication bias was detected in the Egger test (t = 0.52, p = 0.62) and funnel plot (Fig. S1).

When stratified by types of diabetes, the pooled results of 12 studies suggested that elevation of hs-cTn was associated with a significantly increased risk of MACE in type 2 diabetes (adjusted HR per SD change 1.14, 95% CI [1.04–1.25], I² = 0%), while no significant association was found between elevation of hs-cTn and risk of MACE in type 1 diabetes (adjusted HR per SD change 1.17, 95% CI [0.93–1.47], I² = 0%). As for the subunit of hs-cTn, we observed significant association of increased risk of MACE with hs-cTnT (adjusted HR per SD change 1.19, 95% CI [1.08–1.31], I² = 0%), but not with hs-cTnI (adjusted HR per SD change 1.06, 95% CI [0.91–1.24], I² = 0%). The results of sensitivity analyses were similar in perspective studies, high-quality studies, adequately adjusted studies, or population without major cardiovascular comorbidities at baseline (Table 2).

Heart failure

The pooled results of ten studies involving 28,159 diabetic patients suggested elevation of hs-cTn was associated with a significantly increased risk of heart failure (adjusted HR per SD change 1.33, 95% CI [1.08–1.63], I² = 0%, Fig. 4). Similar results were found in the multiple sensitivity analyses (Table 2). But after dividing the concentrations of hs-cTn with a cutoff value of about 6 ng/L, no significant association was observed between high hs-cTn and risk of heart failure (adjusted HR 1.59, 95% CI [0.91–2.78], I² = 0%).

In the subgroup analyses of prediabetes and diabetes population, the pooled results indicated elevation of hs-cTn was associated with a significantly increased risk of heart failure in patients with diabetes (adjusted HR per SD change 1.39, 95% CI [1.09–1.77], I² = 0%), but not in patients with prediabetes (adjusted HR per SD change 1.19, 95% CI [0.82–1.73], I² = 0%).

All-cause mortality

Eleven studies with 12,874 patients evaluated the association between hs-cTn and risk of all-cause mortality in diabetic population. The pooled results from four studies showed no significant association between elevation of hs-cTn and risk of all-cause mortality (adjusted HR per SD change 1.24, 95% CI [0.98–1.57], I² = 0%). The results remained similar after dividing the level of hs-cTn into high and low categories (adjusted HR 1.69, 95% CI [0.93–3.10], I² = 0%, Fig. S2). Four studies adopted a cutoff value of hs-cTn ranging from 8.2 to 14 ng/L, and the rest of the studies used 25 or 35 ng/L.