Prognostic value of elevated cardiac troponin I in patients with intracerebral hemorrhage

Abstract Background Although cardiac troponin has been well established as diagnostic and prognostic makers for acute coronary heart disease, the prognostic value of elevated cardiac troponin in patients with intracerebral hemorrhage (ICH) was inconsistent and not systematically evaluated. Hypothesis We proposed the hypothesis that the practical utility of cardiac troponin levels for prediction of mortality and poor outcome in ICH patients. Methods A total of 1004 patients with ICH were retrospectively reviewed and qualified for further analysis from June 2012 to December 2015. The patients were divided into different groups based on measurements of cardiac troponin I (cTnI) at the time of admission and the following day. Multivariate Cox proportional hazards analysis were performed to determine the independent prognostic value of the cTnI for patients in‐hospital mortality and poor outcomes, the receiver operator characteristic (ROC) analysis was performed to assess the predictive value of cTnI, ICH score, and combination of them. Results Serum cTnI level was an independent predictor in‐hospital mortality (positive vs negative, HR (hazard ratios) = 3.44, 95% CI (confidence interval) 1.66‐7.13, P < .001) and poor outcomes in patients with ICH (positive vs negative, HR = 6.69, 95% CI 4.25‐10.52, P < .001). Addition of cTnI to ICH score significantly improved the prognostic discrimination for both in‐hospital mortality and poor outcomes. Conclusion Serum cTnI levels may be valuable as predictor for in hospital mortality and poor outcomes and may be useful in the risk stratification of ICH during hospitalization.

22.8% to 27.5% for different race) and poor outcome of survivors. 4,5 Some inflammatory biological markers, including d-dimer level 6 and neutrophil-to-lymphocyte ratio, 7 have shown potential relevance with prognostic in patients with ICH, while they were not sufficiently validated and routinely used in clinical practice. In addition to those inflammatory markers, the ICH score has been established and commonly used for estimating the risk of short-term mortality or poor outcome of ICH patients from actual clinical data. [8][9][10][11] Given that biological markers involving in processes of ICH may add early prognostic information and be suitable targets for therapeutic research, other novel maker predisposing to ICH and its clinical sequel warrant further consideration.
Cardiac troponins, including cardiac troponin T and I, have been routinely used as sensitive and specific makers to diagnose acute coronary syndromes or myocardial damage such as acute myocarditis. 12 In the setting of these syndromes, elevation of cardiac troponins also served as prognostic markers to predict adverse consequences of heart failure even in the absence of coronary artery stenosis. 13 Moreover, cardiac troponin elevation has been reported in a number of nonmyocardial ischemic conditions including pulmonary embolism, sepsis syndrome, renal insufficiency and subarachnoid hemorrhage (SAH), and served as prognostic makers with different clinical implications. [13][14][15] Particularly, elevated cardiac troponin is associated with an increasingly delayed cerebral ischemia, poor outcomes and death in SAH patients, 14 playing an important role in predicting the prognostic value of the other type of hemorrhagic stroke.
Recently, elevation of cardiac troponin has been reported to occur in ICH patients along with only 1.2% of them died of cardiac causes. 16 Although previous studies have been tried to address hypothesis that elevated cardiac troponin might serve as prognostic markers for prediction of adverse clinical events, the results were largely inconsistent and inconclusive with regard to ICH. The previous study demonstrated that elevated troponin levels were associated with higher mortality following ICH. 16 Subsequently, the utility of elevated troponin levels for prediction of mortality was confirmed in surgical ICH patients, 17 but was not found to be consistently associated with in-hospital mortality in Chinese ICH patients. 18 Small sample sizes (less than 240 stroke patients) and ethnic variability probably contribute to the negative results and discrepancies.
This has prompted further efforts to determine whether cardiac troponin could provide valuable clinical prediction information for mortality as well as poor outcome in ICH patients with a relatively large cohort. We also confirmed the addition of cTnI to ICH score could significantly improve the prognostic discrimination for both mortality and poor outcome. and ICH due to a structural cause (arteriovenous malformation, aneurysm, cavernous angioma, venous angioma, dural fistula) were excluded from present study. In addition, hemorrhages associated with trauma, brain tumor, hemorrhagic cerebral infarction, or thrombolytic treatment of ischemic stroke during hospitalization did not meet included criteria. Initially, a total of 1319 potential cases of hemorrhagic stroke were abstracted, but 61 were removed from analysis because they had at least one exclusion criterion: trauma (n = 6), hemorrhagic transformation with infarction (n = 25), hemorrhage into tumor (n = 11), subdural hematoma (n = 7), age < 18 at the time of event (n = 12). Finally, a total of 1004 patients with a diagnosis of ICH in Tongji Hospital (Hubei province, China) were finally enrolled in the retrospective medical records review ( Figure S1). To determine whether these included patients were representative of all verified primary ICH patients and potential selection biases within the study, we compared the baseline characteristics and risk factors in original total sample (total group) and sub-sample after the exclusion (included group). As shown in Table S1, the proportion of hyperlipidemia accounted for lower proportion in total patients vs included subjects (31.40% vs 38.25%, P < .001). No significant difference of the other baseline characteristics and risk factors was found between the total and included group.
This study was approved by the institutional review board of Tongji Hospital (Wuhan, China) and the data protection authorities.
Owing to no breach of privacy and interference with clinical decisions related to patient care, informed consent was not required. All our present medical research was conducted according to the principles expressed in the Declaration of Helsinki.

| Data collection and definition of risk factors
We obtained clinical data and background information, including history of hypertension, diabetes, hyperlipidemia, cerebrovascular disease (ischemic stroke or heart disease), smoking, and drinks. Similarly, these risk factors were define as previously reported. 19

| Cardiac testing
Serum concentrations of cTnI were measured by using the Abbott-Architect Troponin I assay with the use of the Architect system (Abbott Diagnostics). According to the manufacturer and validation, 20 the limit of detection of this assay was 0.010 ng/mL, the 99thpercentile cutoff value was 0.028 ng/mL and the assay precision was 0.032 ng/mL, which achieved a coefficient of variation of less than 10%. An elevated cTnI was defined as plasma level greater than cutoff value of cTnI. Patients tested positive (cTnI ≥0.028 ng/mL) were divided into the dynamic group and stable group based on the dynamic changes of cTnI according to the previous studies. 21,22 Dynamic group was defined as a rise or fall of >30% with at least one cTnI value above the cutoff value, while the stable group was defined as cTnI with no significant dynamic changes.
When the patient's cTnI test is positive, a repeated measurement of cTnI was performed within 6 hours and a standard 12-lead ECGs was routinely recorded. All the obtained ECG results of patient were collected, during which the changes of rhythm, heart rate, PR interval, QRS duration, QTc interval, and ST segment/T-wave were analyzed.
All intervals were determined using commercial ECG analysis software (12-Lead Algorithm, GE Medical Systems, Connecticut). In addition, ECGs were employed as imaging evidence of new loss of viable myocardium or new wall motion abnormality by using a Vivid 7 ultrasound machine (GE Medical Systems).

| Hematoma volume measurement
In order to assess the volume of intracranial hematomas, ABC/2 technique was used based on the initial CT or MRI in ICH subjects as previously description, 19 where A referred to the maximum linear length in centimeter (cm), B related to the maximum width in cm, and C indicated the maximum depth in cm, respectively. The depth C was calculated by the number of slice where hematoma was visible multiplying by the slice thickness labeled in CT or MRI scan. 23

| Neurological scores assessment
Initially, brain injury of patients with ICH was assessed on admission using the Glasgow Coma Scale (GCS) score. 24 As previously description, the ICH score was calculated by comprising information, such as age, GCS score, ICH volume, intraventricular involvement, and supratentorial vs infratentorial origin. 25 While the clinical outcome was measured by modified Rankin Scale (mRS) scores at ICH day 14 (or discharge if earlier). 26

| In-hospital complications and poor outcome assessment
In present study, the in-hospital complications including abnormal ECG, left ventricle (LV) wall motion abnormality, hypotension, pulmonary edema, upper gastrointestinal hemorrhage, and acute renal insufficiency were recorded according to prespecified criteria based on the laboratory, imaging data as well as clinical manifestation. Abnormal ECG and LV wall motion abnormality were diagnosed by standard 12-lead ECGs and echocardiograms, respectively. Hypotension was defined as a MAP (mean arterial pressure) <90 mmHg. Pulmonary edema was defined as the development of at least of two clinical findings as follow: rales, hypoxemia (PO 2 /FiO 2 < 300) and pulmonary infiltrates on chest radiography. Upper gastrointestinal hemorrhage was defined as bleeding arising from the esophagus, stomach, or duodenum with observation of blood in vomit (hematemesis) or altered form in the stool (melena). As introduced by the kidney disease improving global outcomes (KDIGO) acute kidney injury work group, 27 acute renal insufficiency was diagnosed if any one of the following was present: (a) increase in serum creatinine by ≥0.3 mg/dL (≥26.5 μmoL/L) within 48 hours; (b) increase in serum creatinine to ≥1.5 times baseline, which has occurred within the prior 7 days; (c) urine volume < 0.5 mL/kg/h for 6 hours. Clinical poor outcome was defined as severe functional disability (dependency, mRS score 3-6) or mortality (death, mRS score 6).  Cox regression plots of the survival curves were generated according to cTnI levels and cumulative rates of survival between different groups using the log-rank test.

| Statistical analysis
Receiver operator characteristic (ROC) analysis was performed to determine the predictive value of cTnI, ICH score, and their combination value for mortality and poor outcomes. The optimal cutoff value was determined by maximal Youden index with sensitivity and specificity. 28

| Distribution of serum cTnI levels
Among 1004 patients with ICH, a total of 275 patients had serum cTnI level above the reference (≥0.028 ng/mL, 27.4%). As shown in Figure S2, the distributions of serum cTnI level deviated from a strict normal distribution, and median serum cTnI level in these patients was 0.010 ng/mL.

| Clinical characteristics of patients with ICH
Baseline characteristics and risk factors of patients with ICH according to serum cTnI concentration were presented in Table 1. Initially, we divided all the patients into two groups by cTnI reference value: a cTnI negative group (cTnI < 0.028 ng/mL, n = 729) and a positive group (≥0.028 ng/mL, n = 275). Positive group were further divided into two groups (constant, n = 187 and dynamic, n = 88) according to the elevation of cTnI. Compared with cTnI negative group patients, dynamic group has higher systolic blood pressure (P < .05), the percentage of history of heart disease and anti-coagulation (all the P < .01). However, there were no significant differences for other documented risk factors between the cTnI negative group and constant group as well as dynamic group. As expected, the admission characteristics such as ICH volume (>30 mL), GCS, ICH, and mRS scores were significantly different between the cTnI negative group with constant or dynamic groups (all the P value < .001, Table S2).

| In-hospital complications and clinical poor outcome
As shown in Figure 1, compared with cTnI negative group patients (constant, dynamic vs negative), patients with cTnI level above the reference (both constant and dynamic groups) had higher cumula-

| Correlation analysis between cTnI with GCS, ICH and mRSscores
Subsequently, we performed correlation analysis between serum peak cTnI levels and neurological scores. As shown in Table S3, cTnI levels had positive correlations with ICH score (r = .330, P < .001) and mRS score (r = .350, P < .001), in contrast, had negative correlations with GCS score (r = −.296, P < .001).

| Prognostic value of cTnI for prediction of inhospital short-term mortality of ICH patients
A total of 77 patients (7.7%) died during a median hospital stay of 5 days (IQR range from 2 to 12 days), among whom 61 patients (79.2%) died directly due to ICH. Baseline characteristics according to survival status were summarized in Table S4. As shown in Table 2 Figure 2A). When further elevated cTnI, the 30 days mortality significantly increased along with elevation of cTnI (Log rank P < .001, Figure 2B).

| Prognostic value of cTnI for predictors for prediction of poor outcomes of ICH patients
In total, clinical poor outcomes were observed in 217 patients (2.2%). 3.7 | Adding cTnI to ICH score improves the predicting value of in-hospital short-term mortality and poor outcome in ICH patients As shown in Table S6, the diagnostic accuracy of cTnI, ICH score, and combined of them in predicting of in-hospital short-term mortality and poor outcomes of ICH patients was moderate. In particular, the addition of cTnI to ICH score significantly improved the prognostic discrimination for both mortality and poor outcomes (both P < .001, Figure 3 and Table S7).

| DISCUSSION
In present study, we examined the utility of cTnI as prognostic makers in patients with ICH. The main findings of this study can be summarized as follows: (a) serum peak cTnI levels was correlated with GCS, ICH and mRS scores in patients with ICH; (b) serum cTnI elevated in ICH patients and along with high proportion of in-hospital complications and adverse events; (c) elevation of cTnI, especially dynamic change of cTnI, were independently associated with incidence of inhospital mortality and poor outcomes; (d) adding cTnI to ICH score further increased the prognostic discrimination for both in-hospital mortality and poor outcome of ICH patients. Thus, cTnI seems to be useful as an additive monitoring marker in patients with ICH.
The mechanisms of elevated cTnI in some of the patients with ICH was complicated and incompletely understood. Cardiac troponins are considered the most sensitive and specific biochemical markers for cardiac injury. 30 Thus, elevation and dynamic changes of serum cTnI levels are indicators of cardiac injury after ICH. The excess of catecholamines was found in patients with ICH, which peaked on the first week and then declined. 31 Given that catecholamine was capable of producing myocardial necrosis, even in the nonischemic heart, 32 the catecholamine surge theory has been reasonable linked to cardiac injury after ICH. 33 In addition, systemic inflammatory responses are activated after ICH and might contribute to myocyte injury and cell death. 34 Several other putative causes were also responsible for the observed elevated cTnI levels, including cardiopulmmonary disease, renal insufficiency, and ICH patients with neurological deterioration. 35 In present study, elevated cTnI was associated with high frequency of pulmonary edema and acute renal insufficiency.
F I G U R E 3 . Receiver operator characteristic (ROC) curves for in-hospital mortality and poor outcomes. Area under receiver operating characteristic (AUC) curve values for using cTnI, ICH score and combination of them in predicting in-hospital short-term (30-days) mortality (A) and poor outcomes (B) of patients with ICH Several important limitations of present study must be acknowledged. First, this was a single-center, retrospective study. Thus, the main findings of our study still need multi-centers and prospective studies to be further verified in order to assessment prognostic value of cTnI in patients with ICH. Second, whether adding troponin to the existing blood prognostic markers improves the predicting value of clinical outcome in ICH patients warrants further assessment. Third, the utility of cTnI for the identification of patients with ICH at highrisk of long-term mortality as well as adverse outcomes would greatly strengthen the importance but not determinate in present study. We consider that further long-term follow-up period, and many end-point studies are mandatory for confirming the correlation between cTnI and the prognosis of ICH. Fourth, because serial measurements of cTnI were not available, we could not address the important issue that whether any appropriate treatment to reduce cTnI might help us to improve outcome in patients with ICH. Finally, the weakness of present study is that current commercial version of high-sensitivity cardiac troponin I (hs-cTnI) assay was not routinely used in our central lab (before the year of 2016). Given that the diagnostic threshold to the 99th percentile is lower in hs-cTnI assay when compared with cTnI assay (0.014 ng/mL vs 0.028 ng/mL), the increased number of patients presenting with an elevated cTnI level that needs further assessment of the prognostic information.
In this study, we demonstrated that elevation of cTnI after ICH has important prognostic predictive value, particularly regarding dynamic changes in cTnI for prediction of in-hospital mortality and poor outcomes. In the future, we need larger prospective cohort study to confirm these findings and test the utility of risk stratification based on cTnI, which could guide clinicians to implement appropriate therapeutic strategies.

ACKNOWLEDGMENTS
This study was funded by the research grants from National "973" Project (No. 2014CB541601) and National Nature Science Foundation of China (Nos. 81470379, 81670215, and 81974047).