Cardiac Biomarkers Predict Long-Term Survival After Cardiac Surgery

Abstract

Background:

Cardiac biomarkers soluble ST-2 (sST-2) and N-terminal prohormone B-type natriuretic peptide (NT-proBNP) may be associated with long-term survival after cardiac surgery. We explored the relationship between long-term survival after cardiac surgery and serum biomarker levels.

Methods:

Patients undergoing cardiac surgery from 2004-2007 were enrolled in a prospective biomarker cohort in the Northern New England Cardiovascular Disease Study Group Registry. Preoperative, postoperative, and the change in serum biomarker levels were categorized by quartile. We used Kaplan-Meier survival analysis and Cox regression models adjusted for variables in the STS’s ASCERT long-term survival calculator to study the association of biomarker levels with long-term survival. Following Kaplan-Meier analysis, quartiles 2 and 3 were found to have similar survival and were therefore combined into one category.

Results:

In our cohort (n=1,648), median follow-up time was 8.5 years (IQR: 7.6-9.7), during which there were 227 deaths. The 10-year survival rate was 86%. Kaplan-Meier survival analysis demonstrated a significant (p<0.001) difference across quartiles of each biomarker levels measurement. After adjustment, pre-operative, post-operative, and the change in biomarker levels in quartile 4 (highest serum levels/change) were significantly predictive of worse survival (hazard ratio range 1.77- 2.89, all p<0.05) compared to quartile 1; however, levels of sST-2 and NT-proBNP in quartiles 2-3 demonstrated a non-statistically significant trend with long-term survival.

Conclusions:

Elevated pre- and post-operative levels of sST-2 or NT-proBNP and large changes in these biomarkers’ levels are associated with increased risk of worse survival after cardiac surgery. These biomarkers can be used for risk stratification or assessing postsurgical prognosis.

Introduction

Biomarkers have demonstrated their value in disease detection, surveillance of clinical conditions, and prediction of response to an intervention.^1-6 Yet, we have not fully explored their utility in cardiovascular care and outcomes. Despite declining mortality rates, heart disease remains the leading cause of death in the United States.⁷ Biomarker measurement presents a potential avenue through which we can track patient health and predict postsurgical outcomes to ultimately reduce heart disease-related mortality.

Soluble ST-2 (sST2) and N-terminal prohormone of B-type natriuretic peptide (NT-proBNP) are two promising contenders in this field. A member of the interleukin 1 receptor family, sST-2 is released from cardiomyocytes in response to mechanical strain and is involved in cardiac remodeling and fibrosis.^8-12 NT-proBNP is a natriuretic peptide secreted in response to myocardial wall stress and increased cardiac workload.^13-15 As such, increasing levels of serum sST-2 or NT-proBNP are associated with worse outcomes and disease severity in patients with acute myocardial infarction and chronic heart failure.^{8, 10-12, 14-16} The prognostic value of sST-2 and NT-proBNP in other cardiovascular conditions and their complementary properties make them suitable for study in patients undergoing cardiac surgery.

In this study, our objective was to evaluate the association of sST-2 and NT-proBNP serum levels with long-term mortality after cardiac surgery. Using data from a prospectively collected biomarker cohort, we studied preoperative, postoperative, and change in serum biomarker levels to explore their relationship to patient survival after discharge from a coronary artery bypass grafting (CABG) and/or valve replacement. Understanding the predictive ability of biomarkers measured at different stages in the surgical intervention process makes room for ways to identify and intervene for high-risk patients, thus reducing cardiovascular disease-related mortality.

Patients and Methods

Settings and Participants

The NNECDSG is a regional voluntary consortium founded in 1987 with a mission to improve outcomes for patients undergoing a cardiac procedure in Maine, New Hampshire and Vermont. Members include voluntary multi-disciplinary groups of clinicians, hospital administrators and health care research personnel from Eastern Maine Medical Center, Bangor, ME; Maine Medical Center, Portland, ME; Catholic Medical Center, Manchester, NH (affiliate Parkland Medical Center, Derry, NH), Dartmouth-Hitchcock Medical Center, Lebanon, NH; Concord Hospital, Concord, NH; and University of Vermont Medical Center, Burlington, VT.^{17, 18} An additional Maine and New Hampshire site are also included as they were members at the time of the study: Central Maine Medical Center, Lewiston, ME and Portsmouth Regional Hospital, Portsmouth, NH. The NNECDSG maintains a prospective registry of patient demographics, clinical characteristics, and in-hospital outcomes for cardiac surgical procedures including CABG and valve replacement.

We prospectively enrolled patients undergoing a CABG and/or valve replacement procedure from 2004-2007 in the New England Cardiovascular Disease Study Group (NNECDSG) for our biomarker study, as described in previous studies.^19-22 Patients (n=1,690) had 10 mL blood samples collected preoperatively (before skin incision) and postoperatively (approximately 24 hours post-procedure). After collection, blood samples were allowed to clot and separate at room temperature for 20 minutes. Samples were then centrifuged at 3500 rpm for a further 20 mins before storage at −80°C. After freezing, sera samples were transported to the Laboratory for Clinical and Biomedical Research in Colchester, Vermont and stored at −80°C until biomarker serum level measurement by Meso Scale Discovery multiplex assay (Rockville, MD). Patients enrolled in the Biomarker study were linked to the National Death Index to determine all-cause mortality through 2015. All patients provided signed informed consent for blood collection and biomarker measurement

Forming the Analytic Cohort

For our study, we excluded patients who died prior to discharge (n=42), were missing biomarker measurements (n=122 missing preoperative, n=374 missing postoperative) or underwent a valve replacement procedure (n=5). Our final cohort includes 1,643 patients with follow-up data, for which we were able to study the effect of biomarkers levels during three time frames: preoperative (n=1,526), postoperative (n=1,256) and the change in serum levels between the preoperative and postoperative period (n=1,152). The institutional review board at each hospital approve the collection of NNECDSG data. The Committee for the Protection of Human Subjects at Dartmouth College approved this study.

Measures

The primary outcome was long-term survival after cardiac surgery, as measured by time from index discharge to death or last date known alive. Our primary exposures were preoperative, postoperative, and change in serum sST-2 and NT-proBNP biomarker levels. Change in biomarker levels was defined as the difference between preoperative and postoperative measurements. Each exposure was categorized by quartile. Following Kaplan-Meier analysis, quartiles 2 and 3 were found to have similar survival and were therefore combined into one category.

Statistical Methods

All baseline characteristics were compared using chi-squared test for categorical variables or Student’s t-test for continuous variables. The threshold for significance was a two-tailed p-value < 0.05. We also present absolute standardized differences (d) between patients that died and survived, which is a measure of the effect size (difference in means or proportions divided by the standard deviation). Based on the literature, we considered d > 0.1 as an threshold for imbalance, or significant difference, between the groups²³. We used Kaplan-Meier survival analysis and Cox regression models to study the association of biomarker levels with long-term survival. As there were no deaths beyond seven years, we truncated our Kaplan-Meier survival graphs at this point. We adjusted for variables in the American College of Cardiology Foundation-Society of Thoracic Surgeons (STS) Collaboration on the Comparative Effectiveness of Revascularization Strategy (ASCERT) Long-Term Survival Probability Calculator^{24, 25} including age, weight, height, creatinine, ejection fraction, mean aortic gradient, sex, ethnicity, diabetes, cerebrovascular disease, cigarette smoking, congestive heart failure class, prior cardiac operation, cardiac status, number of diseased coronary vessels, myocardial infarction, and valve insufficiency. All statistical analyses were performed using Stata 15.1 (College Station, TX).

Results

Study Population

In our cohort (n=1,648), median follow-up time was 8 years (IQR: 7.5-9.7), during which there were 227 deaths. The majority of the cohort were men (78%) age 65 ± 10.1 years. Serum sST-2 and NT-proBNP levels showed substantial variability in the preoperative, postoperative, and change in levels. Mean values and cut-offs used for defining biomarker quartiles are listed in Table 1.

Table 1.

Cut-offs used for preoperative, postoperative, and change in sST-2 and NT-proBNP levels quartile categorization.

Time Measured	Statistic	sST-2 Serum Level (in ng/mL)	NT-proBNP Serum Level (in ng/mL)
Preoperative	Mean (SD)	5.9 (8.5)	7.5 (21.5)
	Quartile 1	< 3.2	< 1.0
	Quartile 2-3	3.2–5.7	1.0–5.6
	Quartile 4	> 5.7	> 5.6
Postoperative	Mean (SD)	63.8 (56.2)	21.9 (29.4)
	Quartile 1	< 28.0	< 7.9
	Quartile 2-3	28.0–82.0	7.9–24
	Quartile 4	> 82.0	> 24.0
Change	Mean (SD)	59.4 (56.8)	14.3 (18.1)
	Quartile 1	< 22.0	< 6.0
	Quartile 2-3	22.0–76.0	6.0–18.0
	Quartile 4	> 76.0	> 18.0

^asST-2= soluble ST-2; NT-proBNP= N-terminal prohormone of BNP

Compared to those who survived (Table 2), patients who died were older (69.6 vs 64.5, p<0.001) and more likely to have comorbidities such as atrial fibrillation (16% vs 5.6%, p<0.001) congestive heart failure (30% vs. 7.8%, p<0.001), diabetes (50% vs. 35%, p<0.001), vascular disease (42% vs. 25%, p<0.001), and any acute kidney injury (53% vs. 32%, p<0.001). During the procedure, patients who died commonly received packed red blood cells (61% vs. 34%, p<0.001), and when they did, received more units than those who survived (2.41 vs 0.95 mean units, p<0.001). The standard difference (d) for all these values were far greater than 0.10, highlighting the magnitude of imbalance between the two groups. Patients who died more commonly experienced in-hospital complications than those who survived (Table 3).

Table 2.

Characteristics of patients by survival status.

Characteristics		Died (n=226)	Survived (n=1,417)	Std. Diff. (d)	p-value‡
Demographics	Age, mean years (SD)	69.5 (10.1)	64.5 (9.9)	0.50	<0.001
	Women	28%	22%	0.14	0.046
	Body Mass Index, kg/m2				0.022
	< 18.5	2.2%	0.4%	0.16
	18.5-24.9	19%	17%	0.04
	25.0-29.9	41%	40%	0.02
	30.0-34.9	23%	28%	0.11
	≥ 35.0	15%	15%	0.01
Comorbidities	COPD	21%	12%	0.25	<0.001
	Smoker	27%	23%	0.10	0.148
	Diabetes	53%	36%	0.36	<0.001
	Vascular Disease	42%	25%	0.39	<0.001
	Preoperative serum creatinine, mg/dL (SD)	1.29 (0.80)	1.13 (1.05)	0.17	0.030
	AKI Network Stage				<0.001
	No AKI	47%	68%	0.44
	Stage 1	41%	29%	0.25
	Stage 2	6.0%	2.0%	0.21
	Stage 3	6.0%	1.0%	0.28
Cardiac History, Function, and Anatomy	Preoperative MI				0.008
	No	47%	57%	0.22
	< 24 hours	0.9%	1.9%	0.09
	> 24 hours, < 7 days	22%	18%	0.10
	> 7 days, < 365 days	15%	9.1%	0.17
	> 365 days	16%	14%	0.07
	Prior CABG	2.7%	1.9%	0.05	0.451
	Prior PCI	3.5%	3.7%	0.01	0.923
	Hypertension	85%	80%	0.12	0.109
	Atrial Fibrillation	16%	5.5%	0.33	<0.001
	Congestive heart failure	30%	7.8%	0.58	<0.001
	Unstable angina	54%	54%	0.004	0.955
	Ejection Fraction (SD)	49.1 (13.5)	54.6 (11.5)	0.44
	Left main, ≥ 50% stenosis	39%	33%	0.14	0.050
Procedural Practice	Priority				0.060
	Non-urgent	25%	32%	0.14
	Urgent	74%	66%	0.17
	Emergent	0.9%	2.1%	0.10
	Preoperative IABP	5.8%	4.2%	0.07	<0.001
	Received PRBC	60%	34%	0.55	<0.001
	PRBC, units (SD)	2.40 (3.68)	0.95 (1.73)	0.50	<0.001

^aSD= standard deviation; COPD=chronic obstructive pulmonary disease; AKI= acute kidney disease; MI= myocardial infarction; CABG= coronary artery bypass grafting; PCI= percutaneous coronary intervention; IABP= intraaortic balloon pump; PRBC= packed red blood cells

^bp-values calculated with chi-squared for categorical variables, Student’s t-test for continuous variables

Table 3.

In-hospital outcomes for patients by survival status.

Outcome	Died (n=226)	Survived (n=1,417)	Std. Diff. (d)	p-value‡
Stroke	8 (3.5%)	13 (0.9%)	0.18	0.001
Transient Ischemic Attack	1 (0.4%)	4 (0.3%)	0.03	0.685
Mediastinis Infection	5 (2.2%)	6 (0.4%)	0.15	0.006
Pneumonia	14 (6.2%)	15 (1.1%)	0.28	<0.001
Low Cardiac Output	21 (9.4%)	61 (4.4%)	0.20	0.001
Leg Infection	2 (0.9%)	13 (0.9%)	0.003	0.962
Return to Bypass Pump	10 (4.5%)	36 (2.6%)	0.10	0.113
Renal Failure	6 (2.7%)	2 (0.1%)	0.21	<0.001

Long-term survival

The 10-year survival rate in this cohort was 86%. The 30-day mortality was 0.9% (n=14). Kaplan-Meier survival analysis showed a dose-response relationship between sST-2 and NT-proBNP serum biomarker levels and patient survival, where increasing biomarker levels or change was associated with decreased survival rate. As such, patients in quartile 4, who are those with the highest preoperative/postoperative serum biomarker level or largest change, had the lowest survival rate (log-rank p<0.001 for all comparisons) (Figure 1).

Figure 1.

Kaplan Meier survival curves for (a) preoperative (b) postoperative and (c) change in sST-2 and NR-proBNP serum biomarker levels. This figure shows the relationship between long-term survival and quartile of biomarker serum level measured at different time points (log-rank p<0.001 for all comparisons).

We discovered this same trend in the unadjusted hazard ratios (HR) for all comparisons. Using quartile 1 as the reference group, patients in quartile 4 for preoperative, postoperative or change in biomarker levels were anywhere from 2-8 times more likely to die (p<0.05 for all comparisons) (Table 4). Though the size of the effect varied, this pattern was true for sST-2 and NT-proBNP.

Table 4.

Simple and adjusted effects of preoperative, postoperative, and chance in serum biomarker levels on long-term survival.

Biomarker	% Dead	Simple			Adjustedb			Adjusted + Other Biomarkerc
		HR	95% CI	p-value	HR	95% CI	p-value	HR	95% CI	p-value
Pre-operative (n= 1,526)
sST-2
Quartile 1	10%	1.0 (ref)	-	-	1.0 (ref)	-	-	1.0 (ref)	-	-
Quartile 2-3	15%	1.50	1.02-2.20	0.041	1.36	0.91-2.03	0.129	1.29	0.86–1.93	0.221
Quartile 4	22%	2.26	1.52-3.36	<0.001	1.45	0.94-2.24	0.092	1.33	0.85-2.07	0.206
NT-proBNP
Quartile 1	3.7%	1.0 (ref)	-	-	1.0 (ref)	-	-	1.0 (ref)	-	-
Quartile 2-3	15%	3.83	2.19-6.70	<0.001	2.43	1.36-4.32	0.003	2.33	1.31-4.16	0.004
Quartile 4	28%	7.66	4.37-13.43	<0.001	2.89	1.54-5.43	0.001	2.63	1.38–4.99	0.003
Post-operative (n= 1,256)
sST-2
Quartile 1	7.3%	1.0 (ref)	-		1.0 (ref)	-	-	1.0 (ref)	-	-
Quartile 2-3	14%	2.17	2.35-3.51	0.001	1.49	0.90-2.45	0.117	1.38	0.83-2.27	0.213
Quartile 4	27%	3.95	2.42-6.44	<0.001	2.57	1.52-4.35	<0.001	2.27	1.33–3.86	0.003
NT-proBNP
Quartile 1	6.0%	1.0 (ref)	-	-	1.0 (ref)	-	-	1.0 (ref)	-	-
Quartile 2-3	12%	1.81	1.08-3.01	0.023	1.26	0.74-2.14	0.391	1.17	0.69–1.99	0.564
Quartile 4	32%	5.28	3.21-8.66	<0.001	2.50	1.40-4.47	0.002	2.13	1.18–1.99	0.012
Change in Serum Level (n=1,152)
sST-2
Quartile 1	8.8%	1.0 (ref)	-	-	1.0 (ref)	-	-	1.0 (ref)	-	-
Quartile 2-3	13%	1.52	0.96-2.40	0.071	1.09	0.78-1.76	0.710	1.03	0.63-1.67	0.913
Quartile 4	25%	2.81	1.76-4.47	<0.001	1.77	1.07-2.95	0.027	1.50	0.87-2.53	0.131
NT-proBNP
Quartile 1	9.3%	1.0 (ref)	-	-	1.0 (ref)	-	-	1.0 (ref)	-	-
Quartile 2-3	12%	1.22	0.78-1.89	0.381	1.12	0.70-1.76	0.632	1.10	0.69-1.74	0.693
Quartile 4	28%	2.96	1.93-4.54	<0.001	1.85	1.13-3.02	0.014	1.69	1.02–2.79	0.040

^aHR= hazard ratio; CI= confidence interval; sST-2= soluble ST-2; NT-proBNP= N-terminal prohormone of BNP

^bAdjusted for ASCERT variables: age, weight, height, creatinine, ejection fraction, mean aortic gradient, sex, ethnicity, diabetes, cerebrovascular disease, cigarette smoking, congestive heart failure class, prior cardiac operation, cardiac status, number of diseased coronary vessels, myocardial infarction, and valve insufficiency.^24,25

^cAdjusted for all ASCERT variables and other biomarker measure (e.g sST-2 model adjusted for NT-proBNP level at the same time point, and vice versa).

After adjustment, patients in quartile 4 of postoperative and change in sST-2 levels remained 2.6 times (HR 2.57, 9% CI: 1.52-4.35) and 1.8 times (HR 1.77, 95% CI: 1.07-2.95) more likely to die after surgery, respectively. Risk adjustment mitigated the effect of preoperative sST-2 levels and they were no longer statistically significantly predictive of death. Similarly, we found that adjusted HRs for quartile 4 NT-proBNP levels were associated with an increased likelihood of death across preoperative (HR 2.89, 95% CI: 1.54-5.43), postoperative (HR 2.50, 95% CI: 1.40-4.47), and change in levels (HR 1.85, 95% CI: 1.13-3.02). We found similar trends even after adjusting the other biomarker level at the same time point (e.g adjust for NT-proBNP level in the sST-2 model).

Comment

The use of cardiac biomarkers such as sST-2 and NT-proBNP is limited in cardiac surgery, though measuring biomarkers serum levels can potentially facilitate improved patient outcomes. Our study demonstrated that elevated preoperative and postoperative measure of sST-2 and NT-proBNP, and large changes in serum levels of these biomarkers are predictive of poorer postsurgical survival, even after adjustment for patient comorbidities and procedural characteristics. These findings can be used to identify patients at high-risk for death after CABG and/or valve replacement. Providers can then offer targeted care such as more extensive evaluation, post-discharge home visits, closer surveillance by primary care physician, or earlier post-operative follow-up appointments for these patients, actions that might mitigate future adverse outcomes.

Our results are corroborated by the cardiovascular literature. Increasing levels of sST-2 and NT-proBNP have been successfully used to identify and predict adverse outcomes for heart failure and acute myocardial infarction.^{8, 10-12, 14-16, 26, 27} In fact, NT-proBNP serum levels have become an established clinical practice for heart failure patients.^{28, 29} The most recent update of the American College of Cardiology/American Heart Association Task Force on Clinical Practice Guidelines of the Management of Heart Failure also mentions the role of sST-2 in additive risk stratification for heart failure patients, however, calls for more validation studies in large cohorts.²⁹ In terms of cardiac surgery, a previous study of the NNECDSG registry evaluated the added value of preoperative serum biomarker levels, including NT-proBNP, to risk prediction models for in-hospital mortality after CABG and found that there was no significant improvement in prediction ability over using patient and disease characteristics alone.¹⁹ However since then, several other studies of the NNECDSG Biomarker Study have found prognostic value for sST-2 and NT-proBNP to predict short term outcomes such as 1-year mortality and readmission, though long-term outcomes remain unstudied.^{20, 21, 30} There is less research available about how NT-proBNP and ST2 biomarker levels change due to intervention (e.g. the change from preoperative to postoperative level) and how this affects long-term outcomes, for which our study is a novel contribution to the published literature.

Despite the promise of our findings, there were limitations to our study. We used data from the NNECDSG Biomarker Study, a large, multicenter prospective study that enrolled patients in Northern New England. This patient population might not be generalizable across the United States, hence further studies would be required to see how the effect size changes for a potentially younger, more racially diverse population. Furthermore, the primary outcome in this study was all-cause mortality, and thus, our focus on cardiovascular comorbidities might not capture the residual confounding for predicting this outcome.

In conclusion, elevated pre- and post-operative levels of sST-2 or NT-proBNP are associated with increased risk of worse survival after cardiac surgery. Future research should expand on the generalizability of these findings to establish clinically significant thresholds for risk stratification or estimating postsurgical prognosis in order to identify and target interventions to patients at high-risk for poor survival. Early intervention in high-risk cases can ultimately improve long-term outcomes for patients undergoing CABG or valve replacement procedures.