Incidence and prognosis of myocardial injury in patients with severe trauma

Severely injured patients [defined as injury severity score (ISS) ≥ 16] and patients who were  ≥ 18 years of age admitted to the resuscitation room of the Emergency Department of the University Hospital Duesseldorf (Level I trauma center) between January 2016 and December 2019 were included in the study [11, 12]. Patients were excluded when no measurement of troponin at arrival in the resuscitation room was available. The treatment of all patients was standardized according to Advanced Trauma Life Support (ATLS) guidelines.

The main exposure was myocardial injury as defined by the fourth universal definition of myocardial infarction, detected using high-sensitive troponin T (hsTnT; Roche Diagnostics, Elecsys®)[6]. HsTnT was measured immediately after arrival in the resuscitation room (main exposure) as well as on day 1 (= 24 h after trauma) and day 2 (= 48 h after trauma) (secondary exposures). The main endpoint of this study was in-hospital mortality.

Due to the retrospective nature of the study, we did not conduct a formal sample size calculation. Between 2016 and 2019, on average approximately 100 patients/year met the inclusion criteria and were defined as having adult severe trauma (ISS ≥ 16). Based on the current literature, we expected an all-cause in-hospital mortality of approximately 25% (= 100 expected events in total) [13]. Assuming that hsTnT values would be available in > 90% of patients, a sample size of 360 patients was deemed appropriate to answer the research question including robust multivariable adjustment with nine co-variables according to Peduzzi et al. [14].

A complete case analysis was conducted. Categorical data were presented as absolute numbers (percent). Continuous data were presented as mean ± standard deviation or median (interquartile range) as applicable. The discrimination of hsTnT for in-hospital mortality was examined by receiver operating characteristic curve (ROC) and the area under the curve (AUC). To quantify the association between myocardial injury and the main endpoint, multivariate binary logistic regression with forced entry of predefined co-variables was conducted. For continuous variables, the established cutoff (hsTnT) or data-driven cutoff (age, base excess) was used. The coding and definitions for each covariable can be found in supplementary Table S3. Hosmer–Lemeshow test was performed to evaluate goodness of fit. In addition, we evaluated the additional value of myocardial injury when added to the ISS score in a separate logistic regression model. We also calculated the AUCs for all logistic regression models with and without myocardial injury to determine potential improvement and performed De Long-Test for comparison of AUCs. During review process, a “post-factum” Kaplan–Meier analysis was performed.

The following co-variables for multivariable adjustment were predefined: age, sex, ISS, pre-existing chronic kidney injury [as defined by Kidney Disease Improving Global Outcome (KDIGO) Criteria], pre-existing coronary artery disease, base excess[15], American Society of Anesthesiologists (ASA) classification and thorax trauma (defined as any form of physical injury to the chest including the ribs, muscles, heart and lungs). The choice of co-variables was driven by its potential impact on both troponin and outcome and based on literature research by two independent members of the study team.

The AUC of initial hsTnT for in-hospital mortality was 0.76 [95% confidence interval (CI) 0.71–0.82] (see Fig. 2). The AUC for hsTnT and in-hospital mortality on day 1 and 2 was 0.84 [95% CI 0.75–0.92] and 0.87 [95%CI 0.79–0.95], respectively (see Fig. 3).

Multivariate binary logistic regression revealed an adjusted odds ratio (OR) of 2.27 [95% CI 1.16–4.45] for myocardial injury and in-hospital mortality (see Table 2). Hosmer–Lemeshow test revealed a p value of 0.3. The full results of multivariate analysis are reported in Table 2. The results of univariate analysis for each covariable are presented in supplementary Table S1.

The AUC for the above described logistic regression model was 0.85 [95% CI 0.80–0.90] without myocardial injury and 0.86 [95% CI 0.81–0.90] including myocardial injury. De Long test showed that the difference between the AUCs was not significant.

Varying pathomechanisms leading to myocardial injury in severe trauma patients are discussed [9, 10, 16]. To date, only few studies investigated this aspect in the past [17‐19]. Keskpaik et al. retrospectively investigated 147 severely injured patients with chest trauma [defined as Chest Abbreviated Injury Scale (AIS) > 3][17]. The primary endpoints of this study were in-hospital mortality and 1-year mortality. The authors found that elevated troponin is associated with poor outcome compared to patients with normal troponin levels. As only patients with chest trauma were included, a selection bias might be present in this study not considering alternative pathomechanisms such as a mismatch between oxygen supply and demand. Edouard et al. postulated in another analysis including 728 patients that elevated troponin I is not associated with higher mortality in trauma patients [20]. However, the authors concluded that troponin I assessment may be used for initial screening in high-risk trauma patients to detect anatomical cardiac injuries. While interpreting the findings listed above, it has to be taken into account that definition of trauma and measurement of troponin (troponin I versus troponin T) differ between the examined cohorts. Moreover, Edouard et al. did not give clear definitions of “late mortality” and trauma severity and did only distinguish between troponin release yes/no rather than considering troponin as continuous variable. Mahmood et al. also identified troponin as a risk marker for mortality in severe trauma patients, but did predominantly concentrate on patients with chest trauma, in whom traumatic damage of the heart plays a crucial role [17, 21]. The authors performed a retrospective analysis of 993 blunt traumatic chest injury patients divided into two groups (positive vs. negative troponin T) and concluded that patients with elevated troponin tend to have the worst outcome even in the absence of clinical evidence of acute cardiac involvement.

The findings of our study add to the limited literature in this field. We investigated a broad and representative cohort of patients with severe trauma. Especially, we included all types of injuries taking into account other mechanisms of troponin release than chest trauma, and according to our multivariate analysis, the association between myocardial injury and mortality was independent of chest trauma. Although we performed a retrospective analysis, our data are based on a prospectively constituted database which ensures high data quality. In addition, we did not only investigate initial troponin, but also included measurements on day 1 and 2 after trauma which enables the interpretation of troponin values over time. Finally, we evaluated the additional value of myocardial injury when added to the established ISS and our results show clearly that the combination of myocardial injury and ISS is significantly better in terms of prognosis than ISS alone.

Our study has several strengths and limitations. Strengths: Because of the standardized collection of data in patients enrolled in the TraumaRegistry®, data quality of this study can be regarded as high. Moreover, we conducted robust multivariate regression including nine relevant covariates and the choice of these covariates was based on literature research. Limitations: One limitation of this study consists of the retrospective single-center design. However, our results for the main endpoint in-hospital mortality are in line with the current literature and thus can be considered as representative for our study cohort. A second limitation refers to ROC analysis for troponin and mortality on day 1 and 2. It is important to clarify that mainly patients with elevated initial troponin at the resuscitation room might have received further troponin sampling which surely led to relevant selection bias. Third, this study only investigated the association between myocardial injury and in-hospital mortality. Future studies should investigate the association with further relevant endpoints (e.g. in-hospital complications, hospital length of stay or long-term outcome).