Circulating biomarkers improve prediction of postoperative outcome after aortic valve surgery

Abstract

Background:

Predicting adverse outcomes following aortic valve surgery remains challenging. This study aimed to investigate the association between preoperative circulating biomarkers and postoperative mortality and morbidity, with the goal of improving established risk stratification tools.

Methods:

Between March 2018 and May 2022, 492 patients underwent surgical aortic valve replacement or repair and were included in a registry. Preoperative blood biomarkers, including hemoglobin, creatinine, high-sensitivity troponin I (hsTrop-I), GOT, GPT, INR, CRP, NT-proBNP and WBC, were sampled at baseline. Logistic regression analysis adjusted for EuroSCORE-II tested associations between biomarker levels and VARC-III adjudicated endpoints. Model fit was assessed using Akaike's Information Criterion, and likelihood ratio tests compared different prediction models.

Results:

Preoperative hemoglobin (OR 0.70; 95% CI: 0.57, 0.87; p < 0.001), creatinine (OR 4.09; 95% CI: 1.63, 10.26; p = 0.003), high-sensitivity troponin I (OR 1.48; 95% CI: 1.11, 1.95; p = 0.007), GOT (OR 2.49; 95% CI: 1.06, 5.86; p = 0.036), INR (OR 4.74; 95% CI: 1.32, 17.05; p = 0.017), CRP (OR 2.14; 95% CI: 1.53, 3.00; p < 0.001), NT-proBNP (OR 2.06; 95% CI: 1.41, 3.00; p < 0.001), and WBC (OR 5.82; 95% CI: 1.88, 17.97; p = 0.002) were independently associated with 30-day mortality after adjustment for EuroSCORE-II. Models combining biomarkers with EuroSCORE-II outperformed those predicting mortality by EuroSCORE-II or biomarkers alone, as indicated by the lowest Akaike's Information Criterion and likelihood ratio tests.

Conclusions:

Combining established risk stratification models with preoperative biomarkers was associated with improved predictive performance for adverse outcomes after aortic valve surgery and may support heart team decision-making when choosing between surgical and transcatheter aortic valve replacement.

Introduction

According to current guidelines (1), a crucial part of interdisciplinary heart team discussions involves the assessment of anticipated procedural risk and differential allocation between surgical aortic valve replacement (SAVR) and transcatheter aortic valve replacement (TAVR) in patients with relevant aortic valve disease. However, prediction of mortality or postoperative complications after cardiac surgery is challenging. In addition to established risk stratification tools (e.g., EuroSCORE-II (ES-II) or Society of Thoracic Surgeons Score (STS score)), preoperative determination of biomarkers may be helpful to predict postoperative adverse outcomes. To further elucidate this topic, we aimed to investigate the association between preoperative measured circulating biomarkers with postoperative complication rates after aortic valve surgery.

Material & methods

Study population

Between March 2018 and May 2022, 492 consecutive patients underwent aortic valve surgery at the University Heart & Vascular Center Hamburg due to severe aortic valve disease according to current guidelines (1) and were included for analysis. Prospectively, 265 patients were included. To increase the analyzed cohort, another 227 patients were added retrospectively. Patients undergoing concomitant procedures such as coronary artery bypass graft, mitral and tricuspid valve surgery were also included. Concomitant aortic root surgery (Bentall's, David's, Yacoub's or Ross' procedures) were excluded, since patients with thoracic aortic surgery have higher risk for peri- and postoperative stroke as shown by Demal et al., with an incidence of 7.0% (2), whereas SAVR patients have an occurrence of stroke of 1.1% in the German Aortic valve registry (3).

A larger and more heterogeneous cohort was chosen to increase the number of total events, thereby allowing the inclusion of more covariates. The ES-II was added as a covariate to balance out the influence of baseline parameters, endocarditis and combined procedures, so that multiple biomarkers could be analyzed. Furthermore, two subgroup analysis were conducted (1): excluding patients with endocarditis and (2): including only patients undergoing isolated SAVR.

This study was conducted in accordance with the Declaration of Helsinki and approved by the Hamburg Ethics Committee (protocol code: PV4304, date of approval: 31 January 2013). Patients either consented to data collection as part of the prospective cohort study (“HARbOR”, NCT 04227002) or ethical approval and informed consent to participate was waived in compliance with the local legislation and institutional requirements (§12 HmbKHG).

Biomarker analysis

Patients underwent routine preoperative sampling of the following biomarkers: hemoglobin, creatinine, hsTrop-I, glutamate oxaloacetate transaminase (GOT), glutamate pyruvate transaminase (GPT), INR (International normalized ratio), C-reactive protein (CRP), N-terminal Pro-B-Type Natriuretic Peptide (NT-proBNP), and white blood cells (WBC). Peripheral venous blood was drawn less than three days before surgery into pyrogen-free tubes (S-Monovette®, Citrat, EDTA, Lithium-Heparin or Serum-Gel Sarstedt, Nuembrecht, Germany). Detailed protocols of biomarker measurements are found in the methods section of the supplemental material.

Statistical analysis

Continuous variables are reported as median and Interquartile range (IQR). Binary variables are shown as counts and percentages.

Associations between the outcome parameters 30-day mortality, renal failure, stroke, life-threatening bleeding type 3-4 as well as the combined 30-day VARC-III endpoints early safety and clinical efficacy (dependent variables) with the tested biomarkers (independent variables) were assessed using logistic regression analysis. All biomarkers were log-transformed before analysis. Regression analysis was adjusted for ES-II and INR. Due to a low event rate a Firth correction was performed. Finally, accuracy of three different regression models for prediction of 30-day mortality was compared. In these models, independent variables were: Model 1, ES-II only; Model 2, Biomarkers only; Model 3, a combination of both ES-II and biomarkers. The model fit was compared using Akaike's Information Criterion (AIC). Additionally, likelihood ratio tests (LRT) for nested models were performed to compare the adjusted biomarker model (Model 3) to the univariable ES-II model (Model 1). For non-nested models, LRTs (Model 1 vs. Model 2) are impossible, since two logistic models with different variables cannot be compared using a p-value. All statistical analyses were performed using R version 4.2.1 (R Foundation for Statistical Computing). P-values <0.05 were considered statistically significant.

Results

Patient characteristics and biomarker levels are presented in Table 1 and revealed a predominantly male, generally low-risk patient, heterogeneous patient population with regard to the leading indication for surgery. Procedural data are listed in Table 2. Surgical aortic valve replacement was performed in 469 (95.4%) and aortic valve repair in 23 patients (4.6%). Concomitant procedures were performed in 174 patients (35.4%).

Procedural and 30-day outcome parameters are listed in Table 3. Overall 30-day mortality, disabling stroke and myocardial infarction were 3.5% (n = 17), 2.2% (n = 11) and 1.2% (n = 6) respectively. In 6.3% (n = 31) of the patients, bleeding type 3 or 4 was documented. Combined VARC-III endpoints early safety and clinical efficacy were reached in 73.9% (n = 362) and 93.6% (n = 457) of patients, respectively. Device success was present in all patients except cases with early mortality (95.7%).

Associations of the biomarkers with 30-day mortality and other postoperative outcome parameters are shown in Figure 1 and Table 4. Preoperative hemoglobin (OR 0.70; 95% CI: 0.57, 0.87; p < 0.001), creatinine (OR 4.09; 95% CI 1.63, 10.26; p = 0.003), hsTrop-I (OR 1.48; 95% CI 1.11, 1.95; p = 0.007), GOT (OR 2.49; 95% CI 1.06, 5.86; p = 0.036), INR (OR 4.74; 95% CI 1.32, 17.05; p = 0.020), CRP (OR 2.14; 95% CI 1.53, 3.00; p < 0.001), NT-proBNP (OR 2.06; 95% CI 1.41, 3.00; p < 0.001), and WBC (OR 5.82; 95% CI: 1.88, 17.97; p = 0.002) were all significant and independent predictors (adjustment for ES-II) for 30-day mortality. After adjustment for Coumarin intake, INR could not be confirmed as an independent predictor for 30-day mortality (OR 4.22; 95% CI 0.47, 37.81; p = 0.200). In Figure 2 the distribution of significant biomarker levels in 30-day-survivors vs. non-survivors are shown as violin plots.

Figure 1

Figure 2

Renal failure was predicted by hemoglobin (OR 0.69; 95% CI 0.62, 0.77; p < 0.001), creatinine (OR 11.84; 95% CI 5.74, 24.42; p < 0.001), hsTrop-I (OR 1.38; 95% CI 1.16, 1.63; p < 0.001), GOT (OR 2.42; 95% CI 1.36, 4.32; p = 0.003), INR (OR 4.21; 95% CI 1.68, 10.58; p = 0.002), INR adjusted for Coumarin intake (OR 4.45; 95% CI 1.15, 17.29; p = 0.031), CRP (OR 1.71; 95% CI 1.43, 2.06; p < 0.001), NT-proBNP (OR 1.72; 95% CI 1.45, 2.04; p < 0.001), and WBC (OR 3.21; 95% CI: 1.74, 5.91; p < 001).

Independent predictors for life-threatening bleeding (VARC bleeding type 3/4) were hemoglobin (OR 0.80; 95% CI 0.68, 0.94; p = 0.007), creatinine (OR 2.92; 95% CI 1.31, 6.52; p = 0.009), and CRP (OR 1.33; 95% CI 1.00, 1.76; p = 0.049).

None of the tested biomarkers independently correlated postoperative stroke occurrence.

The combined VARC-III endpoint early safety was predicted by hemoglobin (OR 1.38; 95% CI 1.24, 1.53; p < 0.001), creatinine (OR 0.22, 95% CI 0.12, 0.41; p < 0.001), hsTrop-I (OR 0.83, 95% CI 0.71, 0.97; p = 0.020), INR (OR 0.25; 95% CI 0.10, 0.61; p = 0.003), INR adjusted for Coumarin intake (OR 0.26; 95% CI 0.07, 0.98, p = 0.047), CRP (OR 0.59; 95% CI 0.49, 0.71; p < 0.001), NT-proBNP (OR 0.73; 95% CI 0.62, 0.84; p < 0.001), and WBC (OR 0.34; 95% CI: 0.19, 0.62; p < 0.001).

Independent predictors for the combined VARC-III endpoint clinical efficacy were hemoglobin (OR 1.27; 95% CI 1.08, 1.50; p = 0.004), creatinine (OR 0.35; 95% CI 0.16, 0.79; p = 0.011), hsTrop-I (OR 0.69; 95% CI 0.52, 0.90; p = 0.006), INR (OR 0.30; 95% CI 0.09, 0.97; p = 0.044), CRP (OR 0.63; 95% CI 0.48, 0.82; p < 0.001), NT-proBNP (OR 0.69; 95% CI 0.53, 0.90; p = 0.006), and WBC (OR 0.29; 95% CI: 0.12, 0.72; p = 0.008).

Comparisons of the different models predicting 30-day mortality are shown in Figure 3. Model 1 is characterized by the prediction using ES-II only, Model 2 by biomarkers only, and Model 3 by the combination of both ES-II and biomarkers.

Figure 3

Here, the combination of the ES-II and biomarkers predicts 30-day mortality best as indicated by lowest AIC value for hemoglobin (AIC 140.4), creatinine (AIC 144), hsTrop-I (AIC 78.5), INR (AIC 133.8), CRP (AIC 132), NT-proBNP (AIC 92) and WBC (AIC 145.6). Since every biomarker is its own model, the AIC cannot be compared between different biomarkers.

Accordingly, the LRT also suggests that using the combination of ES-II and biomarkers results in a better prediction of 30-day mortality than the use of ES-II alone if hemoglobin (p = 0.001), creatinine (p = 0.009), hsTrop-I (p = 0.016), CRP (p < 0.001), NT-proBNP (p < 0.001) or WBC (p = 0.020) are used as predicting biomarkers.

Figure 4 shows receiver operating characteristic (ROC) curves of significant biomarkers adjusted by ES-II predicting 30-day mortality. Predictive performance was good using hemoglobin, hsTrop-I, INR, NT-proBNP, and WBC with area under the curve (AUC) values between 0.7 and 0.8. CRP performs best with the highest AUC of 0.8.

Figure 4

As a sensitivity analysis, we examined the subcohort of patients undergoing isolated SAVR (see Supplementary Tables S1–S4). In this group, 30-day mortality was predicted by preoperative levels of hemoglobin (OR 0.55; 95% CI: 0.34, 0.88; p = 0.013) and CRP (OR 2.64; 95% CI: 1.43, 4.86; p = 0.002). INR (OR 5.23; 95% CI: 0.99, 27.62; p = 0.051) and hsTrop-I (OR 1.49; 95% CI: 0.99, 2.24; p = 0.053) also showed a trend towards predicting 30-day mortality.

Furthermore, we examined the cohort of patients without endocarditis (see Supplementary Tables S5–S8). Here, 30-day mortality was predicted by creatinine (OR 4.84; 95% CI: 1.20, 19.50; p = 0.027), NTproBNP (OR 2.23; 95% CI: 1.30, 3.85; p = 0.003). A trend predicting 30-day mortality was shown by CRP (OR 1.94; 95% CI: 0.99, 3.80; p = 0.055).

Discussion

Our analysis shows a strong association between preoperatively tested routine blood biomarkers and early outcome after SAVR. The analysis indicates that multiple biomarkers combined with ES-II may provide better prediction in SAVR patients than the ES-II alone. The ES-II is widely used in the preoperative risk assessment in cardiac surgery. It shows a high prediction accuracy as demonstrated by prior studies (4, 5). Nonetheless, the analysis of Chalmers et al. (6) showed that the ES-II did not improve on the original EuroSCORE in risk prediction in isolated SAVR procedures. Additionally, multiple studies (7, 8) indicate that the ES-II is not a reliable predictor in high-risk patients. Accordingly, Grant et al. (5) and Borde et al. (9) demonstrated an over-estimation of mortality in high-risk patients by ES-II.

The use of biomarkers in the prediction of the postoperative outcome after cardiac surgery has been proposed by several prior studies (10). Here, especially creatinine (11) and NT-proBNP (12) have shown reasonable accuracy. Additionaly, Salama et al. demonstrated that high-sensitivity troponin T has prognostic value for major adverse cardiovascular events in patients with severe aortic stenosis and preserved left ventricular ejection fraction undergoing SAVR (13). Prior to our analysis, Spampinato et al. investigated blood biomarkers predicting postoperative outcomes after SAVR for aortic stenosis (14). Here, combining the biomarkers NT-proBNP, hsTrop-T, and CRP with the EuroSCORE results in a better prediction profile than using EuroSCORE alone. However, their study used the original EuroSCORE which includes data of 19,030 patients, from which a high rate underwent coronary artery bypass grafting and only 5,671 patients received valve surgery (15). In our study, the more accurate ES-II was used, which is based on a larger cohort of 22,381 patients, of whom 10,353 of the patients underwent valve surgery (16).

Although strong correlations were found between several tested blood biomarkers and outcome parameters after SAVR, the causal connection between biomarker levels and occurring complications was not investigated. Whereas the correlations between bleeding events and preoperative INR as well as between creatinine and renal failure appear self-explanatory, other associations are more difficult to understand. However, high levels of NT-proBNP and hsTrop-I may indicate preoperative heart failure and/or cardiac ischemia impairing the periprocedural outcome. Elevated WBC and/or CRP levels may either indicate perioperative inflammation and/or infections or reflect an increased cardiovascular risk and are therefore associated with impaired postprocedural outcome. Noteworthy, both WBC and CRP demonstrate additional predictive performance beyond the information provided by the presence or absence of endocarditis. This is because the predictive models for CRP are adjusted for ES-II, within which endocarditis is included as a binary parameter. Finally, low hemoglobin levels may be caused by malnutrition or chronic diseases like malignant tumors or chronic inflammatory syndromes which may elevate the periprocedural risk.

The main limitation of this study is the single center design. In addition anemia-related parameters reflecting iron stores were not assessed. We therefore propose prospective multicenter analyses in future studies on risk predicting blood biomarkers in cardiac surgery.

In the context of preconditioning, the identified biomarkers reflect potentially modifiable factors, indicating that targeted optimization before surgery could be beneficial. Structured patient blood management and assessment of systemic inflammation may therefore represent important elements of prehabilitation and should be considered within Enhanced Recovery After Surgery (ERAS) programs.

Conclusion

This analysis identified the biomarkers hemoglobin, hsTrop-I, INR, NT-proBNP, WBC and CRP that predict postprocedural clinical endpoints after aortic valve surgery including 30-day mortality. We cautiously assume that these biomarkers may indicate baseline cardiac and/or end-organ injury. The prediction model combining biomarkers and ES-II showed a superior accuracy when compared to the models using ES-II or biomarkers alone. After validation of the proposed biomarkers in larger multicenter cohorts, their use may simplify heart team discussions and decision-making between SAVR and TAVR in patients with aortic valve diseases.

References

• 1.

  VahanianABeyersdorfFPrazFMilojevicMBaldusSBauersachsJet al. ESC National cardiac societies. 2021 ESC/EACTS guidelines for the management of valvular heart disease: developed by the task force for the management of valvular heart disease of the European Society of Cardiology (ESC) and the European association for cardio-thoracic surgery (EACTS). Eur. Heart J. (2022) 43(7):561–32. 10.1093/eurheartj/ehab395

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 2.

  DemalTJSitzmannFWBaxLvon KodolitschYBrickwedelJKonertzJet al. Risk factors for impaired neurological outcome after thoracic aortic surgery. J. Thorac. Dis. (2022) 14(6):1840–53. 10.21037/jtd-21-1591

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 3.

  BeyersdorfFBauerTFreemantleNWaltherTFrerkerCHerrmannEet al. Five-Year outcome in 18 010 patients from the German aortic valve registry. Eur J Cardiothorac Surg. (2021) 60(5):1139–46. 10.1093/ejcts/ezab216

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 4.

  VikholmPIvertTNilssonJHolmgrenAFreterWTernströmLet al. Validity of the Swedish cardiac surgery registry. Interact Cardiovasc Thorac Surg. (2018) 27(1):67–74. 10.1093/icvts/ivy030

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 5.

  GrantSWHickeyGLDimarakisITrivediUBryanATreasureTet al. How does EuroSCORE II perform in UK cardiac surgery; an analysis of 23 740 patients from the society for cardiothoracic surgery in Great Britain and Ireland national database. Heart. (2012) 98(21):1568–72. 10.1136/heartjnl-2012-302483

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 6.

  ChalmersJPullanMFabriBMcShaneJShawMMedirattaNet al. Validation of EuroSCORE II in a modern cohort of patients undergoing cardiac surgery. Eur J Cardio-Thorac Surg Off J Eur Assoc Cardio-Thorac Surg. (2013) 43(4):688–94. 10.1093/ejcts/ezs406

  • CrossRef
  • Google Scholar
  • View reference in article

• 7.

  HowellNJHeadSJFreemantleNvan der MeulenTASenanayakeEMenonAet al. The new EuroSCORE II does not improve prediction of mortality in high-risk patients undergoing cardiac surgery: a collaborative analysis of two European centres. Eur J Cardio-Thorac Surg Off J Eur Assoc Cardio-Thorac Surg. (2013) 44(6):1006–11discussion 1011. 10.1093/ejcts/ezt174

  • CrossRef
  • Google Scholar

• 8.

  MastroiacovoGBonomiALudergnaniMFranchiMMaragnaRPirolaSet al. Is EuroSCORE II still a reliable predictor for cardiac surgery mortality in 2022? A retrospective study study. Eur. J. Cardio-Thorac. Surg. Off. J. Eur. Assoc. Cardio-Thorac. Surg. (2022) 64(3):ezad294. 10.1093/ejcts/ezad294

  • CrossRef
  • Google Scholar
  • View reference in article

• 9.

  BordeDGandheUHargaveNPandeyKKhullarV. The application of European system for cardiac operative risk evaluation II (EuroSCORE II) and society of thoracic surgeons (STS) risk-score for risk stratification in Indian patients undergoing cardiac surgery. Ann. Card. Anaesth. (2013) 16(3):163–6. 10.4103/0971-9784.114234

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 10.

  VernooijLMKleiWAMoonsKGTakadaTWaesJDamenJA. The comparative and added prognostic value of biomarkers to the revised cardiac risk Index for preoperative prediction of Major adverse cardiac events and all-cause mortality in patients who undergo noncardiac surgery. Cochrane Database Syst. Rev. (2021) 2021(12):CD013139. 10.1002/14651858.CD013139.pub2

  • CrossRef
  • Google Scholar
  • View reference in article

• 11.

  GriffinBRBronsertMReeceTBPalJDClevelandJCFullertonDAet al. Creatinine elevations from baseline at the time of cardiac surgery are associated with postoperative complications. J Thorac Cardiovasc Surg. (2022) 163(4):1378–87. 10.1016/j.jtcvs.2020.03.174

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 12.

  FellahiJLDaccacheGRubesDMassettiMGérardJLHanouzJL. Does preoperative B-type natriuretic peptide better predict adverse outcome and prolonged length of stay than the standard European system for cardiac operative risk evaluation after cardiac surgery?J. Cardiothorac. Vasc. Anesth. (2011) 25(2):256–62. 10.1053/j.jvca.2010.05.009

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 13.

  SalamaAIbrahimGFikryMElsannanMHEltahlawiM. Prognostic value of high-sensitive troponin T in patients with severe aortic stenosis undergoing valve replacement surgery. Indian J. Thorac. Cardiovasc. Surg. (2024) 40(2):142–50. 10.1007/s12055-023-01594-5

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 14.

  SpampinatoRABochenRSiegFWeissSKornejJHaunschildJet al. Multi-Biomarker mortality prediction in patients with aortic stenosis undergoing valve replacement. J. Cardiol. (2020) 76(2):154–62. 10.1016/j.jjcc.2020.02.019

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 15.

  RoquesFNashefSAMMichelPGauducheauEde VincentiisCBaudetEet al. Risk factors and outcome in European cardiac surgery: analysis of the EuroSCORE multinational database of 19030 patients. Eur. J. Cardiothorac. Surg. (1999) 15(6):816–23. 10.1016/S1010-7940(99)00106-2

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 16.

  NashefSAMRoquesFSharplesLDNilssonJSmithCGoldstoneARet al. EuroSCORE II. Eur. J. Cardiothorac. Surg. (2012) 41(4):734–45. 10.1093/ejcts/ezs043.

  • Pubmed Abstract
  • CrossRef
  • Google Scholar