Carbohydrate antigen 125 and risk of heart failure readmissions in patients with heart failure and preserved ejection fraction

We aimed to assess the association between CA125 and the long-term risk of total acute heart failure (AHF) admissions in patients with an index hospitalization with AHF and preserved ejection fraction (HFpEF). We prospectively included 2369 patients between 2008 and 2019 in three centers. CA125 and NT-proBNP were measured during early hospitalization and evaluated as continuous and categorized in quartiles (Q). Negative binomial regressions were used to assess the association with the risk of recurrent AHF admission. The mean age of the sample patients was 76.7 ± 9.5 years and 1443 (60.9%) were women. Median values of CA125 and NT-proBNP were 38.3 (19.0–90.0) U/mL, and 2924 (1590–5447) pg/mL, respectively. During a median follow-up of 2.2 (0.8–4.6) years, 1200 (50.6%) patients died, and 2084 AHF admissions occurred in 1029 (43.4%) patients. After a multivariate adjustment, CA125, but not NT-proBNP, was positively and non-linearly associated with the risk of cumulative AHF-readmission (p < 0.001). Compared to Q1, patients belonging to Q2, Q3, and Q4 showed a stepwise risk increase (IRR = 1.29, 95% CI 1.08–1.55, p = 0.006; IRR = 1.35, 95% CI 1.12–1.63, p = 0.002; and IRR = 1.62, 95% CI 01.34–1.96, p < 0.001, respectively). In conclusion, CA125 predicted the risk of long-term AHF-readmission burden in patients with HFpEF and a recent admission for AHF.

www.nature.com/scientificreports/ excluded patients with left ventricle ejection fraction (LVEF) < 50% (n = 2204), and those who underwent valve surgical replacement or transcatheter valvular intervention (n = 128), or died during hospital stay (n = 111). The final study sample included 2369 patients ( Supplementary Fig. 1). Pre-established electronic questionnaires were used during admission to record information related to demography, medical history, vital signs, physical examination, 12-lead electrocardiogram, echocardiogram, and medical treatment on discharge. The study protocol conformed to the ethical guidelines of the 1975 Declaration of Helsinki (revised in 1983), as reflected by an a priori approval by the institution's human research committee. and the local ethics committee (Comité Ético de Investigación del Hospital Clínico Universitario de Valencia) approved the study. All participants provided written informed consent. Patients were not involved in the design and conduct of this research.
Biomarker assessment. Plasma CA125 and NT-proBNP were measured together within the first 24-48 h after admission and analyzed in the local laboratory at each center using commercially available immunoassays (Elecsys® NT-proBNP assay, Roche Diagnostics; Elecsys® CA125 II assay, Roche Diagnostics). For CA125, the intra-assay precision (coefficient of variation) is 1.4%-2.0%, and the inter-assay precision (coefficient of variation) is 0.0%-0.9%, with an analytical range of 0.6-5000 U/mL 12 . For NT-proBNP, the intra-assay precision (coefficient of variation) is 1.2%-1.5%, and the inter-assay precision (coefficient of variation) is 4.4%-5.0%, with an analytical range of 5-35,000 pg/mL 13 . Echocardiographic evaluation. After clinical stabilization, a comprehensive transthoracic echocardiographic examination was performed using commercially available systems (Agilent Sonos 5500 or IE33 Philips, MA, USA). Two-dimensional and Doppler measurements were performed and analyzed by trained cardiologists using standard views and techniques. Preserved LVEF was defined as ≥ 50% in the transthoracic echocardiographic examination performed during admission 10 . Follow-up and outcomes. Total HF readmissions occurring during the follow-up were selected as the endpoint of interest. Additionally, we evaluated all-cause mortality and cardiovascular mortality as endpoints. We identified HF-admission and fatal events from the electronic clinical records of the Regional Health Care System. The personnel in charge of endpoint adjudication were not aware of the patient's levels of both biomarkers.
Statistical analysis. Continuous variables are expressed as mean ± standard deviation (SD) or median [interquartile interval (IQI], as appropiate. Categorical variables are presented as percentages. Baseline characteristics among CA125 and NT-proBNP quartiles were compared by ANOVA, Kruskal-Wallis, or chi-squared tests, as appropriate. Rates of events were presented as per 100 person-years (P-Y). To account for the positive correlation between HF-hospitalization and mortality, we fitted the Famoye bivariate Poisson regression model. The number of admissions (as counts) and mortality (as the terminal event) were modeled simultaneously and linked by shared frailty. To account for differences in the time to each recurrent event, the log of follow-up time was included as an offset in each submodel. Crude and adjusted rates (number of events per 100 P-Y) are presented among the groups tested. We selected explanatory variables for the initial multivariate model based on subject-matter knowledge. Then, using a backward elimination procedure that included a polynomial transformation for continuous variables, we obtained a final model. In some instances, however, the automatic selection procedure was overridden by leaving well-known predictors in the setting of HF regardless the p-value, as priorly described in previous works 14 .
The final covariates included in the recurrent HF-readmission, all-cause mortality, and cardiovascular mortality model were: age, sex, hypertension, first admission for AHF, last New York Heart Association (NYHA) class III vs I-II under stable condition, ischemic and heart valve disease, Charlson comorbidity index, atrial fibrillation, heart rate, and their interaction (atrial fibrillation*heart rate), hemoglobin, estimated glomerular filtration rate (eGFR), left atrial diameter, tricuspid annular plane systolic excursion (TAPSE), furosemide equivalent dose at discharge, and treatment at discharge with angiotensin-converting enzyme inhibitors (ACEI) or angiotensin receptor blockers (ARB), aldosterone antagonists, and beta-blockers.
Risk estimates are presented as incidence rate ratios (IRRs). We set a two-sided p-value < 0.05 as the threshold for significance. All analyses were performed in Stata 15.1 (Stata Statistical Software, Release 15 [2017]; Stata-Corp LP, College Station, TX, USA). We used the "Bivcnto" Stata module for multivariate and bivariate Poisson analyses.

Results
Baseline characteristics. The mean age was 76.7 ± 9.5 years, 1443 (60.9%) patients were women, and 880 (37.1%) had a prior diagnosis of HF. Most of the patients had a prior history of hypertension (83.4%) and were admitted for the first time (68.4%). Median (IQR) values of CA125 and NT-proBNP were 38.3 (19.0-90.0) U/ mL, and 2924 (1590-5447) pg/mL, respectively. The baseline characteristics across CA125 and NT-proBNP quartiles are presented in Tables 1 and 2.
Patients in the upper quartiles of CA125 displayed more conditions of valvular etiology and more features of congestion. They also showed higher atrial fibrillation rates, lower sodium and TAPSE, and higher NT-proBNP (Table 1).
Contrary to CA125, after multivariate analyses, NT-proBNP was no longer associated with the burden of total HF readmissions when evaluated along the continuum (Fig. 1b) or categorized in quartiles (Fig. 2).

Risk of AHF-readmission: subgroup analyses.
With the exception of diabetes (p-value for interaction = 0.005), there was no evidence for a differential predictive value of CA125 among the most relevant subgroups. Thus, CA125 along its continuum was homogeneously associated with higher readmission risk across age, sex, renal function, comorbidity burden, atrial fibrillation, and ischemic heart disease. In patients with diabetes, CA125 portrends a greater risk ( Supplementary Figs. 2 and 3).

Mortality risk.
A total of 1200 (50.7%) patients died during the follow-up, 825 of them were cardiovascular deaths (34.8% of the patients). The crude incidence rates for all-cause mortality among CA125 and NT-proBNP quartiles were higher when moving from lower to higher quartiles (Tables 1 and 2). Kaplan Meier plots showed divergent trajectories among quartiles of both biomarkers throughout the follow-up (Fig. 3).
Under the same multivariate setting and accounting for HF-readmission burden, CA125 remained associated with a higher risk of all-cause mortality (Fig. 4). This relationship was positive and non-linear, as shown in Fig. 4. Compared to Q1, those in Q2, Q3, and Q4 exhibited an stepwise risk increase (Fig. 2). Likewise, NT-proBNP was independently and linearly associated with a higher risk of this endpoint (Fig. 4). Estimates of risk for quartiles of NTproBNP are presented in Fig. 2. Similar results were found for CA125 and NTproBNP when cardiovascular death was analysed (Supplementary Fig. 4).

Discussion
In this study, which included a large cohort of patients with HFpEF discharged after an episode of AHF, CA125, and not NT-proBNP predicted the long-term burden of total HF admissions. The predictive value of CA125 regarding HF-readmission was endorsed by consistent findings in the most representative clinical subgroups. Both biomarkers were associated with the risk of long-term all-cause mortality.  www.nature.com/scientificreports/ Hospitalizations due to HF decompensation are the leading cause of admission in patients older than 65 years in Western countries 15 . They account for the more significant part of HF-related morbidity and health care expenditure of the HF syndrome 2,4 . Prior studies assessing the factors associated with a higher risk of time to the first readmission have failed to find well-recognized prognostic factors. Additionally, most of the models have revealed a low discriminative ability for predicting readmissions 16 . Several reasons might explain these discouraging findings. Among them, the lack of homogeneous criteria for diagnosis and admission seems relevant. The variability in healthcare resources has also been postulated as crucial 16,17 .
Additionally, this "time-to-first" event methodology has been extensively criticized because it ignores all the subsequent outcomes occurring after the first event. Thus, given that a substantial number of patients experience recurrent admissions, this approach does not seem an accurate metric for measuring the morbidity burden of HF 9 . More recently, several authors have argued in favor of replacing analyses of time-to-first readmission by other statistical approaches that include the total number of admissions occurring during the course of the disease. Thus, recent studies and clinical trials have included total admissions during the study follow-up as objectives [18][19][20] . For instance, in the PARAGON trial, that evaluated the effect of sacubitril-valsartan vs. valsartan in 4822 patients with HFpEF, the primary endpoint was a composite of total HF-admission and death from cardiovascular causes 19 . This repeated events methodology may increase the ability to detect treatment effects to a greater extent than the classic "time-to-first" event methodology 9 . For instance, a post-hoc reanalysis of the CHARM-Preserved trial showed a superiority of candesartan when recurrent events were analyzed 21 .
HFpEF constitutes a syndrome with a wide variability of phenotypes and heterogeneous clinical course 2,22 . The risk of HF-admission is significantly higher in the first months following a decompensation 4,10 regardless of LVEF. In a recent study of our group, including 2013 patients with AHF admission, we found that almost 70% of the patients were readmitted in the follow-up at least once, with up to nearly 30% of them having three or more readmissions 4 . We found that the total readmission burden was similar in HFpEF and HFrEF, while readmissions for non-cardiovascular causes were more freqüent in patients with HFpEF 4 .
CA125 has emerged as a valuable biomarker of congestion in AHF 5 . Although the pathophysiological mechanisms responsible for the increase of the synthesis of CA125 in AHF are not fully understood, mesothelial cells activation in response to increased hydrostatic pressure, mechanical stress and/or cytokine activation have been suggested as the crucial mechanisms 5,23,24 . Recently, our group reported that the most important factors related to CA125 in patients with AHF were, in order of importance, the presence of pleural effusion and the severity of tricuspid regurgitation 25 . Thus, we envision CA125 levels as a proxy of fluid overload and right-sided HF 25 . Thus, given that presence of fluid overload is highly prevalent in decompensated HF patients, it seems feasible to speculate that higher CA125 identified more advanced patients with greater congestion and a higher risk of new HF decompensations.
Conversely, natriuretic peptides are the standard HF biomarker accurately reflecting the high filling ventricular pressures and myocardial stretch 26 . The role of natriuretic peptides in AHF has been extensively evaluated in patients with HFrEF, and the evidence is scarcer for HFpEF 27,28 . Most prior studies in HFpEF evaluated mortality or the composite of death and readmission, with a prognostic value not different from those with HFrEF 27,28 . The reasons behind the lack of predictive ability of NT-proBNP for predicting total AHF readmission in this sample remain elusive. However, some reasons have been postulated.
1. The higher prevalence of right-sided dysfunction and systemic congestion in HFpEF over HFrEF has been previously reported 29 . Thus, the relevance of CA125 can be expected to be a surrogate of systemic congestion over NT-proBNP as a proxy of left-sided filling pressure for predicting morbidity burden in this population.  Beyond the pathophysiology supporting the positive association between CA125 and burden of total HFadmission, we envision that the assessment of CA125 during decompensation might be a helpful complementary tool for predicting the risk of subsequent new HF decompensations. Thus, circulating levels of CA125 may play a role in planning the intensity of depletion therapy 32,33 , length of stay 34 , and frequency of postdischarge monitoring as reported in recent studies in which high CA125 identified patients that benefit from more intensive diuretic regimens, longer hospital stays and close postdischarge follow-up [32][33][34] . The current study expands the relevant role of CA125 as a circulating biomarker in patients with HF by confirming its value for predicting the morbidity burden in a frequent syndrome in elderly and comorbid patients in which most of the available therapeutic strategies remain empirical.
Some limitations need to be acknowledged. First, this study has the inherent limitations of being an observational study in which HF admission policies/criteria may differ from other healthcare systems. Additionally, despite a robust multivariate adjustment, other confounders may be involved. Furthermore, our conclusions cannot be extrapolated to patients with stable chronic HFpEF. Third, information regarding LVEF values prior to admission was not recorded, so a possible differential risk prediction in those with recovered ejection fraction was not evaluated. Fourth, we cannot evaluate the prognostic role of serial CA125 and NT-proBNP measurement with the current design. Lastly, both bioamarkers were measured at early hospitalzations. It may have not relevant consequences for CA125 because it long-half life (7-12 days) 5 . However, it might represent a limitation for NTproBNP in which predischarge assessment has shown a prognostic superiority over early hospital assessment 31 .
In conclusion, in patients with HFpEF discharged after an episode of AHF, CA125 predicted the risk of the total burden of AHF-readmission. Further studies should confirm these findings. www.nature.com/scientificreports/ Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.