Assessment of cardio-renal-hepatic function in patients with valvular heart disease: a multi-biomarker approach—the cardio-renal-hepatic score

The China Valvular Heart Disease (China-VHD; NCT03484806) study was a nationwide, multicenter, prospective, observational study for adult patients (≥ 18 years) with significant VHD. Consecutive patients with at least moderate VHD, as identified by echocardiography, were enrolled between April and June 2018 from inpatient wards and outpatient clinics at 46 medical centers in China. Data collection and quality control of the China-VHD study have been described previously [17]. The study was performed in accordance with the Declaration of Helsinki and approved by the Institutional Review Board at Fuwai Hospital, National Center for Cardiovascular Diseases of China. Written informed consent was given by all eligible patients before registration. A total of 13,917 patients with various VHD were included in the China-VHD study. To conduct the present analysis, we excluded patients with moderate or greater tricuspid stenosis, pulmonary valve diseases, or mixed VHD. Patients with infective endocarditis, previous valvular interventions, the history of dialysis, missing value on heart, kidney, or liver biomarkers, as well as those without any follow-up information were also excluded. Finally, 6004 patients with aortic stenosis (AS), aortic regurgitation (AR), mitral stenosis (MS), mitral regurgitation (MR), tricuspid regurgitation (TR), and multiple valvular heart disease (MVHD) were included in the current study (Fig. 1; Additional file 1: Figure S1).

Details of the China Elderly Valve Disease (China-DVD; NCT02865798) study has been described previously [2, 24]. In brief, the China-DVD study was a nationwide, multicenter, prospective, observational study for elderly inpatients (≥ 60 years) with VHD. Inpatients with at least moderate VHD, as defined by echocardiography, were enrolled consecutively between September and December 2016 at 69 sites in China. The study protocol was approved by the central and site Institutional Review Board or Ethics Committees. Written informed consent was given by eligible patients before registration. Of 8929 patients enrolled in the China-DVD study, 3156 were included in the validation analysis (Fig. 1; Additional file 1: Figure S2).

In China-VHD study, comprehensive transthoracic two-dimensional and Doppler echocardiography was performed on all patients using standard ultrasound systems. The chamber quantification was performed based on the recommendations of American Society of Echocardiography and the European Association of Cardiovascular Imaging [30]. Left ventricular ejection fraction (LVEF) was calculated using the biplane modified Simpson method. Echocardiographic criteria of VHD were summarized in Additional file 2: Page S1. Quality control of echocardiography in the China-VHD study has been described previously [17]. The echocardiographic measurements, quality control, as well as echocardiographic criteria of significant VHD in the China-DVD study have been also described and published elsewhere [2].

Baseline venous blood samples were drawn after admission. Biomarker concentrations were determined during the same period of the baseline echocardiography. If there were multiple laboratory tests, the result of the first test after admission was collected in the databases. The plasma N-terminal pro-B-type natriuretic peptide (NT-proBNP) concentrations were measured using four assays, including Roche NT-proBNP Elecsys (Roche Diagnostics, Basel, Switzerland), Ortho Clinical Diagnostics Vitros ECi (Ortho Clinical Diagnostics, Raritan, New Jersey), BioMérieux NT-proBNP Vidas (Bio-Mérieux, Marcy, France), and Radiometer AQT90 Flex (Radiometer Medical Aps, Copenhagen, Denmark). The same antibodies and calibrator from the same vendor (Roche Diagnostics) were used by four assays.

The primary outcome of the present study was all-cause mortality. When evaluating outcome of patients under conservative treatment, follow up was censored at the time of valvular intervention if performed.

Data were presented as mean ± standard deviation (SD) or medians (interquartile range [IQR]) for continuous variables, and as counts (percentages) for categorical variables. Differences among groups were compared using Kruskal–Wallis test or Mann–Whitney U-test according to number of groups for continuous variables, and using Chi-square or the Fisher’s exact test for categorical variables as appropriate. The associations between the score and other variables were analyzed using Spearman correlation test and multiple linear regression models.

A total of 6004 patients with as least moderate VHD from the China-VHD study were included in the derivation cohort, and 3156 patients from China-DVD study were included in the validation cohort. The mean ages of two cohorts were 62.07 ± 13.80 and 71.24 ± 7.62 years, respectively. Baseline characteristics were summarized in Tables 1 and 2. The concentrations of NT-proBNP, creatinine, and albumin were 1447.22 pg/ml (468.00–3783.28 pg/ml), 82.44 μmol/L (69.00–100.00 μmol/L), and 3.94 ± 0.51 g/dL in the derivation cohort, and 1829.00 pg/ml (630.26–4510.07 pg/ml), 83.33 μmol/L (69.00–104.00 μmol/L), and 3.90 ± 0.62 g/dL in the validation cohort.

During a median follow up of 731 (704–748) days, 594 (9.9%) patients died in the China-VHD cohort. The cumulative survival at one and two years was 92.7% and 89.7%, respectively. The increasing levels of NT-proBNP, creatinine, and albumin were independently and monotonically associated with two-year mortality (Fig. 2; Additional file 1: Table S4; NT-proBNP: adjusted HR [95%CI], 1.749 [1.616–1.892], P < 0.001; creatinine: adjusted HR [95%CI], 1.744 [1.428–2.131], P < 0.001; albumin: adjusted HR [95%CI], 0.679 [0.570–0.809], P < 0.001). The multi-biomarker index, known as the CRH score, was developed based on the regression coefficients of these biomarkers (Additional file 1: Table S5), as below:where NT-proBNP was in pg/ml, creatinine in μmol/L, and albumin in g/dL. The median values of the CRH score were 3.86 (2.96–4.61), 3.47 (2.59–4.44), 3.55 (3.00–4.07), 4.22 (3.41–4.96), 4.20 (3.44–4.86), 4.64 (3.98–5.29), and 4.22 (3.41–4.96) in AS, AR, MS, MR, TR, MVHD, and total derivation cohort respectively, with detailed score distribution presented in Additional file 1: Figure S3.

The C index of the CRH score in total cohort was 0.78 (95%CI: 0.76–0.80), indicating high discrimination. The score also achieved satisfactory discrimination across all types of VHD (Additional file 1: Table S6; AS: 0.70 [0.61–0.80]; AR: 0.87 [0.84–0.91]; MS: 0.93 [0.88–0.97]; MR: 0.75 [0.71–0.79]; TR: 0.79 [0.75–0.83]; MVHD: 0.74 [0.71–0.77]). Regarding the calibration property of the CRH score, the calibration curves demonstrated excellent agreement between observed and predicted survival probabilities in both total cohort and various VHD (Additional file 1: Figure S4 and S5).

Correlations of left cardiac dimensions, function as well as other variables with the CRH score were presented in Fig. 3 and Additional file 1: Figure S6-S8. Significant positive correlations between left ventricular dimension and the score could be observed in AS, AR, MR, TR, and MVHD, while a negative correlation of LVEF with cardio-renal-hepatic function was found across all types of VHD (Fig. 3; Additional file 1: Figure S6). In multivariable analyses, LVEF was independently and negatively associated with the CRH score in all types of VHD (Additional file 1: Table S7), and was identified as the most important associated factor of cardio-renal-hepatic co-dysfunction in AS, AR, MR, TR, MVHD, as well as in total cohort (Additional file 1: Figure S9 and S10).

As a continuous variable, the CRH score was independently and strongly associated with mortality in overall population, with one-point increase carrying over two times of mortality risk (Table 3; overall adjusted HR [95% CI]: 2.095 [1.891–2.320], P < 0.001). The score was also a powerful predictor of mortality across all types of VHD in multivariable analyses (Table 3; adjusted HR [95%CI]: AS, 1.791 [1.018–3.148], P = 0.043; AR, 3.290 [2.245–4.821], P < 0.001; MS, 4.986 [2.069–12.016], P < 0.001; MR, 1.938 [1.573–2.389], P < 0.001; TR, 2.253 [1.808–2.808], P < 0.001; MVHD, 1.914 [1.603–2.284], P < 0.001). When analyzed categorically by quartile values, the score was significantly associated with mortality (Table 3; Fig. 4; Additional file 1: Figure S11; P_log-rank < 0.05 for all types of VHD).

During a median follow up of 730 (520.75–748) days, death occurred in 539 (13.2%) patients under medical treatment. The cumulative survival at one and two years were 90.1% and 85.6%, respectively. As in overall population, one-point increase of the CRH score was independently associated with more than two-fold risk of mortality in patients under conservative care (Table 4; adjusted HR [95%CI]: 2.168 [1.946–2.416], P < 0.001). The score was also an independent and powerful predictor of mortality in patients with AR (adjusted HR [95%CI]: 4.435 [2.781–7.072], P < 0.001), MS (adjusted HR [95%CI]: 4.401 [1.155–16.778], P = 0.030), MR (adjusted HR [95%CI]: 1.994 [1.601–2.483], P < 0.001), TR (adjusted HR [95%CI]: 2.200 [1.758–2.751], P < 0.001), and MVHD (adjusted HR [95%CI]: 1.977 [1.642–2.380], P < 0.001) under medical treatment. A borderline statistical significance was found in AS (adjusted HR [95%CI]: 2.499 [0.922–6.778], P = 0.072). The prognostic value of the CRH score was well retained in patients with primary VHD, as well as in most patients with LVEF ≥ 50% (AR, MS, MR, TR, and MVHD) (Table 4). In 1922 patients under valvular intervention, increasing level of the CRH score was also independently associated with higher risk of mortality (adjusted HR [95%CI]: 1.734 [1.220–2.464], P = 0.002). The survival benefit of valvular intervention over conservative therapy appeared to be constant across the range of the score (Additional file 1: Figure S12).

The addition of the CRH score to the base predictive model substantially improved the prognostic capability of the model in the overall population (Additional file 1: Table S8; CRH score + base model vs base model, C index, 0.81 [0.80–0.83] vs 0.78 [0.76–0.80]; NRI [95%CI], 0.255 [0.204–0.299], P < 0.001; IDI [95%CI], 0.055 [0.038–0.073], P < 0.001; likelihood ratio test P < 0.001). The score also provided significantly incremental prognostic value over conventional clinical and echocardiographic variables, including left cardiac dimensions and function, in AS, AR, MR, TR, and MVHD (Additional file 1: Table S8; likelihood ratio test P < 0.05). A similar result was obtained in MS when added the score into the minimally adjusted model incorporating age and sex (Additional file 1: Table S8; likelihood ratio test P < 0.001). The decision curve analysis further demonstrated better clinical utility after the inclusion of the score to base models (Additional file 1: Figure S13 and S14). Notably, the cardio-renal-hepatic function, evaluating by the CRH score, showed significantly better predictive performance than the hepatorenal function, represented by the Model for End-Stage Liver Disease excluding international normalized ratio (MELD-XI) score (Additional file 1: Table S9). Additional analyses also confirmed that the novel score performed better than its components alone, especially for creatinine and albumin (Additional file 1: Table S10).

Relative importance of variables was evaluated by the proportion of explainable log-likelihood ratio χ² statistics, best subset analysis, as well as the random survival forest. The cardio-renal-hepatic function, represented by the CRH score, was identified as the most important predictor of mortality in both total cohort and patients with AR, MS, MR, TR, and MVHD (Additional file 1: Figure S15-S19). In patients with AS, the importance of the score was ranked the second or third by different approaches, while previous myocardial infarction was consistently identified as the most contributive feature (Additional file 1: Figure S15-S19).

During a median follow up of 364 (215–381) days, 272 (8.6%) deaths occurred in the China-DVD cohort. The cumulative survival at one year was 90.8%. The C index of the CRH score in the China-DVD cohort was 0.72 (95%CI: 0.69–0.75), indicating satisfactory discrimination. The score also exhibited adequate predictive performance in all types of VHD except MS (Additional file 1: Table S6; AS: 0.82 [0.73–0.91]; AR: 0.66 [0.52–0.81]; MS: 0.62 [0.37–0.87]; MR: 0.73 [0.67–0.78]; TR: 0.70 [0.61–0.79]; MVHD: 0.72 [0.67–0.76]). Calibration curves demonstrated excellent calibration of the score in both total cohort and various VHD (Additional file 1: Figure S4 and S20).

Per one-point increase of the CRH score, the relative risk of mortality increased by 85.0% in the China-DVD cohort (adjusted HR [95%CI]: 1.850 [1.592–2.151], P < 0.001). The score was also strongly associated with mortality in patients with AS (adjusted HR [95%CI], 2.633 [1.151–6.026], P = 0.022), AR (adjusted HR [95%CI], 2.004 [1.080–3.717], P = 0.028), MR (adjusted HR [95%CI], 1.656 [1.275–2.150], P < 0.001), TR (adjusted HR [95%CI], 2.169 [1.358–3.464], P = 0.001), and MVHD (adjusted HR [95%CI], 2.243 [1.721–2.924], P < 0.001), but not in those with MS (adjusted HR [95%CI], 0.625 [0.273–1.427], P = 0.264).

A growing body of evidence implies the existence and development of multi-organ cross-talk in patents with VHD, most prominently the heart-liver and heart-kidney interactions [9, 11, 12, 14‐20]. Results from early studies showed a significant association of elevated kidney or liver function biomarkers with the severity of VHD [11, 14, 31, 32], which might be attributed to increased systemic venous congestion or impaired cardiac output in patients with more severe valvular lesions and cardiac dysfunction. In recent years, studies focusing on patients under valvular intervention further demonstrated the direct contribution of VHD-induced damage to extra-cardiac organ, with data showing that the adverse remodeling and dysfunction of extra-cardiac organs could be reversible after valvular corrections [9, 12, 16, 18, 20, 33]. Although the mechanisms of “cardiorenal syndrome” and “cardiohepatic interaction” remain to be further elucidated in VHD, the kidney and liver function indexes have emerged as prognostic indicators [10, 12, 15‐18, 28, 29]. However, given the relatively weak correlations of hepatorenal function indexes with echocardiographic findings in both prior analyses and the present study [17, 24], elevations of these parameters are unlikely to be mainly explained by the systemic consequences of cardiac dysfunction, or to reflect intrinsic changes of cardiac structure in patients with VHD. Therefore, measuring these indexes is insufficient to monitor progression of VHD, cardiac function, and systemic hemodynamic burden in reality. Expanding on previous findings, we proposed the concept of cardio-renal-hepatic co-dysfunction in patients with VHD, and hypothesized that it could promote better understanding of systemic hemodynamic impairments and improve risk stratification.

The VHD-related cardio-renal-hepatic co-dysfunction can be defined as a clinical syndrome. In this context, cardiac remodeling and dysfunction are induced or exacerbated by VHD, resulting in systemic venous congestion and decreased cardiac output, which further lead to the functional or structural impairments of liver and kidney. The most important feature of VHD-related cardio-renal-hepatic co-dysfunction is that if a successful valvular intervention is performed at an early stage, the structural and functional damages of heart, kidney, and liver will be reversible to some extent.

The multi-biomarker approach has been used to estimate event risk as well as identifying high-risk patients who tended to benefit from more intensive therapy in coronary artery disease [34, 35]. The merits of the multi-biomarker approach include its stable predictive value, user-friendly feature, as well as allowing an integrative consideration of multiple pathophysiological pathways of disease. To our best knowledge, the present study for the first time developed a novel multi-biomarker score with the integration of heart, kidney, and liver function biomarkers to enable assessment of multi-organ function, as well as the systemic condition in patients with VHD. The CRH score is a consequence-oriented multi-biomarker index which includes NT-proBNP, creatinine, and albumin as components quantifying cardio-renal-hepatic function. A prior study showed that NT-proBNP correlated well with echocardiographic parameters, and were independent prognostic factors in patients with significant AS, AR, MR, TR, and MVHD [24]. There was also evidence suggesting that renal and hepatic dysfunction, represented by the elevation of specific biomarkers such as creatinine, albumin, and bilirubin, played roles in prognostic evaluation in various VHD [8, 10, 15‐18, 28, 29]. The hepatorenal function indexes, which combined renal and liver function biomarkers, were independently associated with outcomes in patients with AS, MR, TR, and MVHD [15‐17, 28, 36, 37]. Currently, there is no approach measuring the cardio-renal-hepatic co-dysfunction, and no study evaluating its prognostic implications. Our study represents a novel step towards an ideal biomarker-based strategy for risk assessment in patients with VHD.

The present study demonstrated that the CRH score was independently and strongly associated with all-cause mortality in patients with AS, AR, MR, TR, and MVHD, and provided substantially incremental prognostic information over traditional risk factors. In patients with MS, the predictive performance of CRH score was inconsistent between the derivation and validation cohorts, which could be attributed to the relatively small sample size and number of events. Given the significant prognostic value of the CRH score in the China-VHD cohort, we believe that the score is still a valuable index in MS, and should be further validated in larger cohorts.

One interesting finding of the present study was that the CRH score could predict mortality risk not only in the entire VHD population, but also in the subset with LVEF ≥ 50%. It is well established that left ventricular systolic dysfunction indicates poor outcome in patients with VHD [23]. However, patients with VHD and preserved LVEF can also be in different risk profiles, and there is evidence supporting the prognostic utility of blood parameters in these patients [17, 24‐26]. Compared with LVEF, biomarker-based assessment may enable a more sensitive detection of early disease deterioration. In addition, although our results revealed the intimate relationship of LVEF with cardio-renal-hepatic function, as a cardiac index only assessing left ventricular systolic function, it is unlikely for LVEF to reflect systemic hemodynamic condition or prognosis better than the multi-biomarker index which allowed comprehensive evaluation of heart, kidney, and liver function.

In this study, we adopted both predictive modeling techniques and machine-learning approach to evaluate the relative importance of cardio-renal-hepatic function compared with other predictors. The CRH score was identified as the most powerful predictor in all types of VHD except AS, in which the score was also highly ranked as the second or third most important prognostic factor among clinical characteristics and echocardiographic findings. So far, there exists numerous studies investigating outcome determinants in patients with cardiovascular diseases. From the perspective of methodology, it is not difficult to identify new prognostic factors with independent prognostic value through traditional regression-based statistical analyses. However, the properties of an ideal marker are far beyond its independent prognostic effect among covariates, as the clinically useful marker should also be significantly more important and powerful than existing predictors, especially the determinants of current management decisions. There is scarce literature evaluating relative predictor importance of biomarkers in patients with VHD [17, 24]. Our previous analyses demonstrated that NT-proBNP was the most contributive prognostic factor among clinical characteristics and echocardiographic parameters in elderly patients with AS, AR, MR, and MVHD [24], and the hepatorenal function, measured by the modified MELD scores, was the most important predictor in patients with isolated TR [17]. The present study, using multiple methods including the machine-learning technique to evaluate variable importance, confirmed the crucial role of biomarker-based integrative cardio-renal-hepatic assessment in risk stratification in patients with VHD.

The present study was far beyond proposing a novel concept, as it also provided important information to clinical management of VHD. Progressive multi-organ dysfunction is a crucial landmark of systemic hemodynamic deterioration in patients with VHD. It must be taken seriously because the adverse cardiac remodeling, symptoms of heart failure, as well as extra-cardiac organ impairments are not always reversible. In fact, once severe cardiac or hepatorenal failure occurs, patients are less likely to respond well to valvular corrections, regardless of operative approach [21, 22, 38]. This is particularly notable in patients with MR or TR [21‐23, 36, 39, 40], and also merits attention in those with aortic valve disease and MVHD [8, 37, 41, 42]. The present study suggested that the CRH score could serve as a pragmatic tool to assess cardio-renal-hepatic function, and therefore could help identify high-risk patients as early as possible. Based on three readily accessible biomarkers, this multi-biomarker algorithm is easy to implement in routine clinical practice across all levels of medical institutions, which is of critical importance for monitoring disease progression closely [1].

The current study had several limitations. The addition of novel indexes to the present multi-biomarker score may provide new insights into cardio-renal-hepatic interactions and further improve the predictive performance of the score. Nevertheless, the CRH score has already included readily available biomarkers with ample evidence demonstrating their robust prognostic value in VHD. Serial measurements of cardio-renal-hepatic function may also further improve risk prediction in patients with VHD, which was not investigated in this study. However, the main purpose of this study was to develop a multi-biomarker index for assessing the cardio-renal-hepatic function as well as investigating its prognostic role. The association of longitudinal change in multi-organ function with outcomes will be an interesting topic for future investigations. In the multicenter cohort study, the NT-proBNP measurement was based on four assays, and the variation among laboratories was not evaluated. However, the four assays used same antibodies and calibrator from the same vendor, and previous data showed that the between-method variability of NT-proBNP was not the predominant component of total variability [43], and the imprecision performance of measurement of NT-proBNP in China had improved with the significant decrease of current coefficient of variations [44]. Finally, although the validation cohort of the present study allowed the CRH score to be externally tested in a population with a distinct age distribution, it had relatively shorter duration of follow up compared with the derivation cohort. More studies are needed to further validate this index.