Progressive myocardial injury is associated with mortality in the acute respiratory distress syndrome

Abstract

Purpose:

Myocardial injury connotes worse prognosis in the Acute Respiratory Distress Syndrome (ARDS), however the prognostic connotation of changes in cardiac troponin (cTn) levels in ARDS patients is not known.

Methods:

We performed a study of 908 ARDS patients enrolled in two previously completed ARDS Network trials. We obtained plasma samples via the NIH BIOLINCC repository and measured cTn using the ARCHITECT STAT high sensitivity troponin-I assay (Abbott Laboratories) at trial day 0 and 3. We constructed Cox proportional hazard models to determine the association between 60-day mortality and quintiles of percentage change in high-sensitivity troponin (ΔhsTnI).

Results:

The median percent change in hsTnI (%ΔhsTnI) from day 0 to day 3 was −58.2% (IQR −79.0 to 0%). After multivariable adjustment, participants with a 32.1% or greater increase in hsTnI between day 0 and day 3 (highest quintile) had a 2.27 fold increased risk for mortality (95% CI 1.29 – 3.99, p=0.002) as well as fewer ventilator-free and ICU-free days compared to the lowest quintile.

Conclusion:

Progressive myocardial injury in ARDS patients is associated with worse outcome, independent of severity of critical illness. Investigation of the mechanisms underlying this relationship is warranted to guide possible strategies to mitigate myocardial injury in ARDS.

Introduction

The acute respiratory distress syndrome (ARDS) is a common cause of respiratory failure with substantial associated morbidity and mortality (1, 2). Cardiac involvement in ARDS, in the form of pulmonary vascular and right heart disease (3–5) and myocardial injury (6–11), is associated with adverse outcome in ARDS, and strategies to reduce myocardial injury in critical illness are attractive as potential therapies. Most studies have assessed the association between myocardial injury and ARDS outcomes using measures of myocardial injury at a single time point (6, 8, 9, 11), however serial measurements of myocardial injury could give clinicians an additional risk discriminator or opportunity to intervene. Serial measurements could also help distinguish ongoing myocardial injury from chronic myocardial injury. Peak troponin levels were associated with worsening echocardiographic indices of right heart function in a small study of 42 patients with ARDS (12), and increasing troponin levels on serial examinations was associated with worse outcome in sepsis and other non-cardiac critical illness (13, 14). Data assessing the association of progressive myocardial injury with ARDS outcomes are lacking.

Current highly-sensitivity assays for cardiac troponin (hsTnI) detect troponin in over 90% of patients with ARDS (8) and provide an opportunity to characterize ARDS-associated myocardial injury with sensitivity and precision. We performed a multi-center cohort study of ARDS patients to determine whether the change in hsTnI over 3 days is associated with mortality, hypothesizing that those patients with rising hsTnI levels over 3 days would have increased mortality. We also investigated patient and disease-specific factors associated with progressive myocardial injury.

Methods

Patient population

Our study population was comprised of patient cohorts from two NHLBI-sponsored ARDS Network clinical trials. The ALVEOLI trial enrolled patients in 23 ICUs, and randomized 549 subjects with ARDS already receiving low tidal volume ventilation to receive treatment with a lower or a higher positive end-expiratory pressure (PEEP) strategy (15). The Fluid and Catheter Treatment Trial (FACTT) included 1000 subjects with ARDS across 20 ICUs in a 2×2 design, randomizing subjects to conservative or liberal fluid management strategies and to treatment guided by a pulmonary artery catheter or central venous catheter (16, 17).

The inclusion criteria were similar in both trials. All patients had ratio of partial pressure of oxygen (PaO₂) to fraction of inspired oxygen (fiO₂) less than 300, bilateral pulmonary infiltrates, and no clinical evidence of left atrial hypertension. Patients with acute myocardial infarction and those patients with severe chronic pulmonary or neuromuscular disease were excluded. Only 27 of 1000 patients enrolled in FACTT had a history of prior heart failure, whereas the ALVEOLI trial did not report information on history of heart failure. Patients on hemodialysis were excluded from the FACTT trial, and hemodialysis status was not reported in the ALVEOLI trial.

We obtained the trial data sets and plasma samples from the NIH Biologic Specimen and Data Repository Information Coordinating Center (BioLINCC) (18, 19). For the present study, we included 908 ALVEOLI and FACTT subjects who were intubated within the 24 hours prior to study enrollment (day 0), were alive at study day 3, and who had an available plasma sample from study day 0 and 3. Study flow is displayed in the Supplemental Figure 1. Both initial trials were approved by participating Institutional Review Boards (IRB) (15–17). For this study, all data were publicly available from BioLINCC and no protected health information was included in study data sets. Thus, the Johns Hopkins Hospital IRB exempted our study from IRB review.

High sensitivity troponin I (hsTnI) measurements

We measured hsTnI in EDTA-anticoagulated plasma samples using the ARCHITECT STAT hsTnI assay (Abbott Laboratories). The limit of detection (LoD) was 2 ng/L. For subjects with values below the LoD, the value was set at LoD divided by $\sqrt{2}$ (1.3 ng/L). The upper reference limit (99^th percentile value of a healthy reference population) was 26 ng/L (20, 21).

Study endpoints

The primary outcome was in-hospital death within 60 days of enrollment. The secondary outcomes were ICU-free days (defined as the number of days not in the ICU within the first 30 days of trial enrollment) and ventilator free days (defined as the number of days not receiving mechanical ventilation within the first 30 days of trial enrollment).

Other variables

The Sequential Organ Failure Assessment (SOFA) score on day 0, an indicator of global illness severity, was used to adjust for the heterogeneity in injury severity within the study cohort (22). Other notable covariates included demographics, ventilator parameters, primary cause of ARDS, and vasopressor requirement.

Statistical analysis

The main exposure variable of interest was the percentage change in hsTnI (%ΔhsTnI), defined as [(hsTnI_{Day 3} − hsTnI_{Day 0})*100]/hsTnI_{Day 0}. For the primary analysis, we categorized %ΔhsTnI into quintiles. We report categorical and continuous demographic and clinical data across quintiles of percentage change in troponin. We performed a trend analysis to determine whether a linear trend exists in the association between demographic/clinical variable and %ΔhsTnI category. We utilized univariate logistic regression models for categorical variables and linear regression models for continuous variables. The P value reported in Table 1 is from an F test for the null hypothesis that model coefficients are zero. We used Cox proportional hazards models to estimate the hazard ratios (with 95% confidence intervals [CI]) for all-cause mortality comparing %ΔhsTnI quintiles to the lowest quintile. Patients were followed-up from day 3 until death, hospital discharge, or trial day 60, consistent with the parent trials. For secondary endpoints, we used multiple linear regression to estimate the average difference (with 95% CI) in ICU-free days and ventilator-free days comparing %ΔhsTnI quintiles to the lowest quintile.

Table 1.

Characteristics of study participants by quintile of percent change in high sensitivity troponin I levels between the day of intubation (day 0) and day 3.

		Quintile of percent change in HS-troponin (range)
	Overall	First (−98.5 to −81.8%)	Second (−81.8 to −67.6%)	Third (−67.6 to −40.7%)	Fourth (−40.7 to 32.1%)	Fifth (>32.1%)	P value
Number	908	181	182	181	182	182
Age, years	50.4	47.1	49.1	51.5	53.8	51.4	0.005
Ethnicity, %							0.09
White	68.4	62.4	68.7	76.8	71.4	62.6
Black	19.2	24.3	17.6	14.4	18.1	21.4
Other	12.4	13.3	13.7	8.8	10.4	15.9
Male sex, %	52.3	47.5	54.9	50.8	55.5	52.7	0.54
Primary cause of ARDS, %							0.04
Aspiration	15.5	14.4	17.6	14.9	15.4	15.4
Pneumonia	48.2	43.1	48.4	48.1	53.8	47.8
Sepsis	22.0	22.7	22.0	27.1	14.8	23.6
Transfusion	1.8	5.0	1.1	0.6	1.6	0.5
Trauma	4.4	7.2	3.3	5.5	2.7	3.3
Other	8.0	7.7	7.7	3.9	11.5	9.3
Day 0 variables
HS-troponin, ng/L	978.9	1,843.2	1,947.6	362.8	569.2	173.0	--
Temperature, °C	37.5	37.8	37.7	37.5	37.5	37.2	<0.001
Heart rate, beats per min	101.7	105.4	102.7	99.4	99.6	101.1	0.03
Systolic BP, mmHg	112.3	111.0	112.9	115.6	112.7	109.2	0.05
Diastolic BP, mmHg	58.8	59.9	60.3	59.3	58.6	55.9	0.004
Vasopressor use, %	31.1	35.9	29.7	24.3	29.1	36.2	0.07
Height, cm	169.7	169.4	169.9	169.5	170.4	169.0	0.78
Weight, kg	79.6	80.0	80.4	80.6	80.0	77.3	0.62
SOFA score	7.6	7.5	7.5	7.6	7.5	8.0	0.50
Tidal volume, mL	479.0	488.1	481.0	474.8	485.5	465.5	0.33
PEEP, cm H2O	9.4	9.9	9.1	9.2	9.3	9.4	0.44
Plateau pressure, cm H2O	26.0	27.0	26.3	25.4	25.7	25.8	0.28
FiO2	0.65	0.66	0.63	0.60	0.65	0.70	<0.001
PaO2/FiO2 ratio	149.5	152.9	148.7	160.6	146.0	139.5	0.04
Lung compliance, mL/cmH2O	32.8	32.0	31.6	33.2	34.1	33.2	0.73
Murray Lung Injury score	2.8	2.9	2.9	2.8	2.8	2.8	0.79
pH	7.37	7.38	7.37	7.37	7.37	7.34	0.004
pCO2, mmHg	39.6	38.1	39.8	40.1	40.4	39.9	0.23
pO2, mmHg	89.7	94.3	87.2	90.1	87.6	89.4	0.51
Creatinine, mg/dL	1.51	1.40	1.39	1.67	1.50	1.58	0.32
Day 3 variables
HS-troponin, ng/L	386.9	178.7	470.2	158.5	403.2	721.7	--
Temperature, °C	37.3	37.2	37.4	37.4	37.3	37.3	0.23
Heart rate, beats per min	93.7	92.4	93.0	95.6	92.4	94.8	0.38
Systolic BP, mmHg	122.4	121.1	123.9	122.0	123.2	121.8	0.80
Diastolic BP, mmHg	63.2	63.5	64.4	62.3	63.0	63.0	0.67
Vasopressoruse, %	15.2	10.5	11.0	13.3	17.6	23.6	0.003
Weight, kg	84.0	84.0	83.6	84.0	84.3	83.9	0.99
Tidal volume, mL	396.8	392.7	401.7	404.3	398.5	386.7	0.34
PEEP, cm H2O	8.5	8.3	7.7	8.4	8.7	9.5	0.003
Plateau pressure,cm H2O	23.0	22.6	22.6	22.8	22.8	24.4	0.10
FiO2,	0.48	0.47	0.46	0.48	0.49	0.5	0.14
PaO2/FiO2 ratio,	185.8	198.7	187.7	187.2	182.2	173.3	0.08
Lung compliance, mL/cmH2O	32.3	33.1	31.7	32.5	33.8	30.1	0.57
pH	7.40	7.42	7.41	7.39	7.39	7.37	<0.001
pCO2, mmHg	42.8	41.9	42.0	42.8	43.3	44.2	0.41
pO2, mmHg	81.1	84.2	81.0	80.2	80.8	79.0	0.52
Fluid balance, mL	4,554.2	4,364.8	3,109.6	3,839.2	4,668.9	6,783.6	<0.001

Values in the Table were calculated using multiple imputation to take missing data into account (see Supplemental Table 1 for description of data available for each variable).

For all analyses, we used 3 models with increasing adjustment for potential confounders defined a priori. Model 1 was adjusted for age (continuous), sex, randomized trial assignment, and hs-troponin levels at day 0 (quintiles). Model 2 was further adjusted by SOFA score at day 0 (continuous). Model 3 further adjusted for use of vasopressors at day 0 and day 3, heart rate at day 0 and day 3 (continuous), and total fluid balance between day 0 and day 3 (continuous).

To evaluate determinants of hsTnI progression between day 0 and day 3, we used multiple linear regression with log_e-transformed Tn_{Day 3} as the dependent variable and day 0 and 3 clinical covariates as independent variables. The final model included variables with p <0.1 in univariate analyses and was further adjusted via stepwise backwards selection removing variables with p >0.1 to arrive at the final multivariate model.

Data were complete for hsTnI_{Day 0} and hsTnI_{Day 3} and for all outcomes. Covariates missing data are shown in the Supplemental Table 1. Multiple imputation using chained equations with 50 imputed data sets was performed to account for missing data. The main results using multiple imputation and using complete case analysis were similar. All results reported herein were obtained using the multiply imputed dataset. A two-tailed p-value <0.05 was considered statistically significant. Statistical analyses were performed using Stata version 15.0 (StataCorp Inc, College Station, TX).

Results

Patient characteristics and demographics are shown in Table 1, stratified by quintile of %ΔhsTnI. Patients with greater positive change in hsTnI were older, had worse hypoxemia on day 0, had lower pH on days 0 and 3,were more likely to require vasopressors and higher PEEP on day 3, and had a more positive cumulative fluid balance between days 0 and 3.

On average, hsTnI values decreased between day 0 and day 3 (Fig 1). The median hsTnI_{Day 0} was 36.2 ng/L (interquartile range 9.1-175.3 ng/L), the median hsTnI_{Day 3} was 18.2 ng/L (5.3-97.7 ng/L), and the median %ΔhsTnI was −58.1% (inter-quartile range of percentage change −78.9 to 0%). The majority of subjects (71.9%) had a decrease in hsTnI levels. hsTnI levels at days 0 and 3 were correlated (Spearman correlation coefficient 0.77), but there was substantial between subject variability in ΔhsTnI (Supplemental Figure 2).

Fig 1.

Histogram of the distributions of day 0 and day 3 high-sensitivity troponin levels.

The white bar represents those subjects with undetectable high-sensitivity troponin levels. The dotted line represents a high-sensitivity troponin level of 26 ng/L, corresponding to the 99^th percentile reference value of a healthy population.

In multivariable models, significant predictors of day 3 hsTnI included day 0 hsTnI, age, day 0 temperature and day 0 heart rate, requirement for vasopressors on day 3, arterial pH on day 3 and peak creatinine level (Table 2). If only day 0 variables were considered in the final model, day 0 hsTnI, age, day 0 temperature, heart rate, creatinine, and FIO2 were strong predictors of day 3 troponin level (Supplemental Table 2). Significant predictors of the percentage change in troponin between day 0 and day 3 included only lower arterial pH on day 3 and lower temperature on trial day 0 (supplemental Table 3). Of note, neither PEEP nor total fluid balance was associated with day 3 troponin or trend in troponin in adjusted models. We defined “pulmonary ARDS” as ARDS due to aspiration or pneumonia and “extrapulmonary ARDS” as ARDS due to sepsis, transfusion, trauma or other causes. There were no significant differences in either day 0 troponin or trend in troponin between pulmonary or extarpulmonary ARDS.

Table 2.

Determinants of high sensitivity troponin I levels at day 3 after intubation.

	Ratio of geometric means (95% confidence interval)	P-value
High-sensitivity troponin, day 0†	46.44 (37.01 – 58.28)	<0.001
Age	1.46 (1.16 – 1.84)	0.001
Temperature, day 0	0.63 (0.50 – 0.78)	<0.001
Heart rate, day 0†	1.20 (0.95 – 1.52)	0.009
FiO2, day 0	1.26 (0.99 – 1.60)	0.06
Creatinine, day 0†	4.06 (1.72 – 9.59)	<0.001
Vasopressoruse, day 3 (yes vs. no)	1.33 (1.06 – 1.69)	0.016
pH, day 3	0.48 (0.39 – 0.60)	<0.01

Values in the Table are ratios of geometric means calculated from linear regression models with log-high sensitivity troponin I levels at day 3 as outcome. For categorical variables, the ratios are compared with respect to the reference category. For continuous variables, the ratios compare the 90^th to the 10^th percentiles of the predict or variable in the overall study sample. The 90^th and 10^th percentiles were, for log high-sensitivity troponin at day 0, 6.71 and 1.06 log(ng/L); for age, 73 and 30 years; for temperature at day 0, 38.8 and 36.3 °C; for heart rate at day 0, 127 and 74 beats per minute; for FiO₂ at day 0, 1.0 and 0.4; for creatinine at day 0, 2.7 and 0.5 mg/dL; and for pH at day 3, 7.48 and 7.28.

Ratios of geometric means were adjusted for other variables in the Table. Variables were included in the multivariable model if P <0.10 in the univariable model, and after a stepwise selection method with retention if P<0.10 in the multivariable model (see Methods for details). Values in the Table were calculated using multiple imputation to take missing data into account (see Supplemental Table 1 for description of data available for each variable).

^†Modeled as restricted cubic splines to accommodate non-linear associations (see Methods for details).

In-hospital mortality within 60 days of enrollment increased in a dose-dependent fashion with increasing %ΔhsTnI (Fig 2). After multivariable adjustment for trial assignment, age, sex, hsTnI at day 0, SOFA score at day 0, vasopressor use at day 0 and 3, heart rate and day 0 and 3, and total fluid balance between day 0 and 3, participants with a 32.1% or greater increase in hsTnI between day 0 and day 3 (highest quintile) had a 2.27 fold increased risk for mortality (95% CI 1.29 – 3.99, p=0.002; Table 3) as well as approximately 3.2 fewer ventilator-free and ICU-free days (Supplemental Table 4). Results were directionally and qualitatively similar when the absolute change in hsTnI between day 0 and day 3 was considered instead of %ΔhsTnI to assess hsTnI trajectories (Supplemental Table 5) with a statistical trend towards association of absolute change in hsTnI and outcome.

Fig 2.

Kaplan-Meier curves for cumulative mortality by quintiles of percent change in high-sensitivity troponin levels between day 0 and day 3.

Table 3.

Hazard ratios for 60-day mortality by quintile of % change in high sensitivity troponin I levels between the day of intubation (day 0) and day 3.

	Quintile of % change in hs-troponin between day 0 and day 3 (range)
	First (−98.5 to −81.8%)	Second (−81.8 to −67.6%)	Third (−67.6 to −40.7%)	Fourth (−40.7 to 32.1%)	Fifth (>32.1%)	P trend
Number of participants	181	182	181	182	182
Mean hs-troponin at day 0, ng/L	1,843.2	1,947.6	362.8	569.2	173.0
Mean hs-troponin at day 3, ng/L	178.7	470.2	158.5	403.2	721.7
Mean change in hs-troponin (%)	−89.1	−75.4	−56.7	−10.3	1,442.2
Number of deaths (%)	27 (14.9)	34 (18.7)	49 (27.1)	47 (25.8)	68 (37.4)
Model 1
Hazard ratio (95% confidence interval)	1.00 (reference)	1.40 (0.84 – 2.35)	2.14 (1.32 – 3.49)	2.35 (1.41 – 3.91)	3.88 (2.39 – 6.32)	<0.001
Model 2
Hazard ratio (95% confidence interval)	1.00 (reference)	1.42 (0.80 – 2.49)	1.99 (1.16 – 3.41)	2.11 (1.20 – 3.70)	3.22 (1.86 – 5.59)	<0.001
Model 3
Hazard ratio (95% confidence interval)	1.00 (reference)	1.28 (0.72 – 2.27)	1.67 (0.96 – 2.89)	1.63 (0.92 – 2.90)	2.27 (1.29 – 3.99)	0.002

Values in the Table were calculated using multiple imputation to take missing data into account (see Supplemental Table 1 for description of data available for each variable).

Model 1: Adjusted for age, sex, randomized trial assignment, and hs-troponin levels at day 0 (quintiles).

Model 2: Further adjusted for SOFA score at day 0.

Model 3: Further adjusted for use of vasopressors at day 0 and day 3, heart rate at day 0 and day 3, and total fluid balance between day 0 and day 3 (continuous).

Discussion

We present a cohort study assessing the prognostic association of trajectories in cardiac troponin levels in patients with ARDS. We report several major findings. First, most patients with ARDS have detectable troponin which declines over time, however approximately 28% of patients with ARDS will manifest increasing hsTnI levels reflecting progressive myocardial injury over the first 72 hours of treatment. Second, several factors including age, vasopressor requirement, peak creatinine, temperature, and arterial pH are associated with progressive myocardial injury. Third, the relative change in hsTnI levels over the first 72 hours of treatment was a graded strong predictor of survival, ICU-free days, and ventilator free days. Our study suggests that progressive myocardial injury could represent a major determinant of outcome in ARDS patients, and serial hsTnI levels in these patients may have prognostic value.

Our group previously reported that over 90% of patients with ARDS manifest detectable troponin when evaluated with a highly sensitive troponin assay (8). The present study extends these findings, and we report that the majority of these patients will subsequently have declining troponin levels over their first 3 days of support. Only one other study has reported on the longitudinal trend in troponin levels in ARDS. In a cohort of 42 ARDS patients who had troponin-I (conventional assay) checked on ICU admission and every 12 hours thereafter for 72 hours (12), 16 of 42 patients had detectable troponin-I on admission. Overall, 25 of the 42 subjects ultimately had detectable levels over the study period, including 26% of those who had an initially negative troponin (12). In contrast, we report that most patients with initial detectable troponin will have subsequent level that declines. Differences between this small study and our results are likely explained by study size, single center versus multicenter design, and different troponin assay utilized. The study by Lazzeri and colleagues used a troponin-I assay that has a 99^th percentile upper reference limit of 100ng/L (12). In contrast, we used a highly sensitive troponin assay with 99^th percentile reference range of 26 ng/L and limit of detection of 2 ng/L, which is much more sensitive to detect myocardial injury. Our findings are similar to those of Masson et al in the Albumin Italian Outcome Sepsis (ALBIOS) study of crystalloid versus albumin (13). They report a cohort of patients with severe sepsis, the majority of whom were receiving mechanical ventilation. Between trial day 1 and 2, 21% of patients (92 of 436) had troponin levels measured with a highly sensitive assay that increased by 20% or more (13). Thus, our results and those of Masson and colleagues corroborate that most patients with critical illness have detectable troponin and most, but not all, of the myocardial injury improves over time with ICU management.

Elevated troponin can be secondary to type I myocardial infarction related to coronary plaque rupture, type II myocardial infarction related to mismatch between myocyte oxygen demand and supply, or myocardial injury not meeting criteria for myocardial infarction (23–25). In our study, patients were excluded prior to trial enrollment if they had active cardiac ischemia. Thus we suspect that the myocardial injury in our cohort relates to the underlying critical illness. We previously reported that creatinine and SOFA score, arterial pH and pCO2, and heart rate and body temperature were associated with initial day 0 hsTnI levels. In the present analysis, we report that day 0 hsTnI, temperature, heart rate, vasopressor use, and arterial pH and the peak creatinine were associated with day 3 troponin. These factors all relate to the underlying illness more so than to modifiable clinical factors. Although a fluid conservative strategy reduced ventilator free days in FACTT (17), 72-hour fluid balance was not associated with trend in hsTnI in the present work. This observation is consistent with prior data that brain natriuretic peptide levels were not associated with outcome in ARDS (26). It does not appear that troponin trends are associated with fluid toxicity due to the fact that fluid balance was not associated with troponin levels in multi-variable models. Similarly, none of the ventilator variables assessed-including oxygenation, tidal volume, and plateau pressure- were associated with trend in hsTnI. Thus, progressive myocardial injury seems related more to the underlying critical illness than to modifiable treatment related factors, and future studies should assess mechanisms by which myocardial injury could be modified.

We report that progressive myocardial injury is independently associated with increased 60-day mortality and with fewer ventilator and ICU free days, even adjusting for underlying critical illness with SOFA score, vasopressor requirement, fluid balance, heart rate, and age. This independent association of hsTnI trajectories contrasts with our observation that initial hsTnI levels were associated with ARDS outcomes, but the association virtually disappeared after adjusting for underlying critical illness (8). It is likely that introducing two time points for hsTnI assessment integrates risk across day 0 to day 3 and adds an independent prognostic element beyond a single day risk score, such as SOFA score. Our findings are again consistent with those of Lazzeri and colleagues in ARDS and Masson and colleagues in severe sepsis, who both note that increasing troponin was associated with worse outcomes (12, 13). A rising troponin level could thus identify a subgroup of patients with ongoing organ injury in response to their underlying illness. This subgroup could be well suited for novel therapeutics, for enrollment in clinical trials in need of a high-risk patient pool, or for expanded diagnostics. For example, in Lazzeri and colleagues’ study, increasing troponin was associated with worse right heart function on echocardiography (12). Echocardiography to guide management of ARDS and titration of therapy to right ventricular function is attractive (5, 27–30) but its efficacy is unproven. Patients with progressive myocardial injury may be an attractive subgroup for initial right ventricular protective treatment strategies. Other attractive cardioprotective treatments in these patients could include beta blockade (31) and extracorporeal support (32, 33).

Limitations of our study include its observational design, which limits our ability to establish causal connections. Not all patients had available plasma on trial day 0 and trial day 3 and by design, our trial excluded patients who died before trial day 3 representing a form of survivor bias. Blood was only sampled on day 0 and 3, whereas clinically troponin levels can be checked more frequently. More granular information on troponin trajectories may provide additional prognostic information. The FACTT and ALVEOLI protocols did not include electrocardiography or echocardiography. Few patients in these trials received prone positioning, thus these data cannot investigate the association of prone positioning in ARDS with myocardial injury. Thus, we could not evaluate cardiac structural or electrical correlates of troponin trajectories. The mechanism of death was not ascertained in the parent trials, and troponin trajectories may have different prognostic implications in patients with cardiovascular versus non-cardiovascular deaths. The trials comprising our cohort were published in 2004 and 2006, and it is possible that patients in the present study do not reflect completely ARDS patients in the modern era. Finally, data on baseline medical therapy including beta-blockers, aspirin, and anticoagulants was not available in either parent trial and vasopressor choice and dose was not standardized in these trials but rather at clinical discretion which limits the ability to perform granular analyses of vasopressors and myocardial injury

In conclusion, progressive myocardial injury occurs in a significant number of patients with ARDS and is associated with higher 60-day mortality independent of underlying critical illness. ARDS patients with progressive myocardial injury represent a high-risk subgroup and future studies should assess the underlying pathophysiology of progressive myocardial injury and potential treatments.

Highlights.

• Most patients with ARDS suffer myocardial injury detectable with a highly sensitive troponin assay, and 72% of ARDS patients with myocardial injury have declining troponin levels over 3 days

• In-hospital mortality for ARDS patients increased in a dose-dependent fashion with increasing percentage change in troponin over 3 days

• After multivariable adjustment participants with a 32.1% or greater increase in troponin between day 0 and day 3 had a 2.27 fold increased risk for mortality (95% CI 1.29 – 3.99, p=0.002; Table 3) as well as approximately 3.2 fewer ventilator-free and ICU-free days