Among the patients with diagnosis of ADHF following current guideline [20] hospitalized in Kyungpook National University Hospital between January 2005 and July 2013, 482 patients who had tests for NT-proBNP or CA-125 at admission were retrospectively included. Diagnosis of ADHF was defined as the rapid onset or worsening of symptoms and/or signs of heart failure such as pulmonary crackles, peripheral edema and cardiomegaly. The patients were excluded before inclusion of the study if they had prior acute myocardial infarction in the previous 3 months, end stage renal or liver disease, pregnancy, malignancy, and age < 18 years (Supplementary Fig. 1). The diagnosis of heart failure was established on the basis of the current guidelines, symptoms, radiographic and echocardiographic findings, and biomarkers.

Data on the demographic characteristics, laboratory tests, medications, and prognosis were collected from the medical charts. Initial echocardiographic data were also collected from echocardiography reports and images from the database. Venous blood sampling for biomarkers was obtained on admission. Patients followed up for a period of 24 months after admission. The primary endpoint was all-cause mortality. Survival or death status was confirmed by reviewing the death certificates, telephone interviews, and data from the national health insurance system. The study protocol was approved by the Institutional Review Board of Kyungpook National University Hospital (KNUH 2016-09-013). Informed consent was waived because of the retrospective nature of the study.

As baseline characteristics, categorical variables were presented as number with percentage, and the differences were compared using the chi-square test. The distribution of continuous variables was presented as mean ± SD or median (interquartile range) and comparison between them was conducted using the independent t test. Normality of the continuous variables was tested using the Shapiro-Wilk test. For the normality of both CA-125 and NT-proBNP, log transformed values were used in all analyses. The multivariate Cox regression analysis was used to determine the significant factors related to the 2-year mortality. A receiver operating characteristic (ROC) curve was used to identify the optimal cutoff points of the CA-125 levels and NT-proBNP levels to determine the potential relationship between CA-125 and 2-year mortality. Sequentially, patients were divided into four groups according to the cutoff levels of NT-proBNP and CA-125. The Kaplan-Meier survival analysis was performed on the basis of those four category groups. Four Cox-proportional hazard models were built to test the incremental prognostic value of CA-125 in addition to established risk factors and NT-proBNP level. Harrell’s C-statistics and likelihood ratio test were performed to assess the discriminative ability of each model. The models served to estimate the integrated discrimination improvement (IDI), net reclassification improvement (NRI) with a category-free option, and global-chi-square test among the models following the methodology of Pencina et al. [21,22] for the incremental predictive value. All analyses were 2-tailed, p values < 0.05 were considered statistically significant.

Among the patients hospitalized in Kyungpook National University Hospital between January 2005 and July 2013, a total of 413 patients (64.1 ± 15.6 year-old, 214 men) have both NT-proBNP and CA-125 tests on admission. The study population included 103 patients (24.9%) with left ventricular ejection fraction > 40%. During follow-up (mean follow-up, 591 ± 233 days), a total of 109 deaths (26.4%) were recorded. Patients with mortality were more frequently men (60.6% vs. 48.6%, p = 0.044), older (age, 72.1 ± 13.6 years vs. 61.3 ± 15.3 years, p < 0.001), and had a lower body mass index (BMI) (21.8 ± 2.9 kg/m2 vs. 23.4 ± 4.2 kg/m2, p < 0.001), more frequent ischemic etiology (53.2% vs. 28.0%, p < 0.001), hypertension (56.0% vs. 43.1%, p < 0.028), diabetes (45.0% vs. 24.7%, p < 0.001), and New York Heart Association (NYHA) class III–IV dyspnea (93.6% vs. 68.1%, p < 0.001). Non-ischemic etiology was listed in Supplementary Table 1. β-Blocker and angiotensin-converting enzyme inhibitor (ACEI)/angiotensin receptor blocker (ARB) prescribed during admission were less frequently used in patient who died during the 2-year follow-up (β-blocker: 73.6% vs. 83.9%, p < 0.028; ACEI/ARB: 66.0% vs. 85.5%, p < 0.001). The serum sodium level was lower (137.7 ± 4.5 mmol/L vs. 139.0 ± 7.1 mmol/L, p < 0.031), and serum creatinine level was higher (1.5 ± 0.9 mg/dL vs. 1.1 ± 0.8 mg/dL, p < 0.001) in patients with mortality. Mean CA-125 and NT-proBNP levels were 83.5 ± 111.0 U/mL and 8,269.2 ± 10,544.1 pg/mL, respectively (Supplementary Fig. 2). Serum CA-125 and NT-proBNP levels on admission were higher in the patients with mortality (log CA-125: 4.2 ± 1.1 U/mL vs. 3.6 ± 1.2 U/mL, p < 0.001; log NT-proBNP: 8.9 ± 1.3 pg/mL vs. 8.0 ± 1.4 pg/mL, p < 0.001). The baseline characteristics of the patients are presented in Table 1.

The ROC curve analysis of CA-125 and NT-proBNP predicting the 2-year mortality is shown in Fig. 1. Areas under the curves were 0.635 and 0.666 for CA-125 and NT-proBNP, respectively. A baseline level of CA-125 > 54.5 U/mL could predict 2-year mortality with 58.6% sensitivity and 61.8% specificity; a baseline level of NT-proBNP > 5,269 pg/mL could the predict 2-year mortality with 58.6% sensitivity and 68.5% specificity (Fig. 2).

In the multivariate analyses using the Cox proportional hazard model for the 2-year mortality, both high CA-125 and NT-proBNP levels were independently associated with an increased 2-year mortality (CA-125: hazard ratio [HR], 1.230; 95% confidence interval [CI], 1.021 to 1.482; p = 0.030) (NT-proBNP: HR, 1.220; 95% CI, 1.006 to 1.480; p = 0.044) together with age, sex, NYHA class, and β-blocker use (Table 2).

We divided the study cohort into four groups according to the cutoff values of CA-125 and NT-proBNP described above: those with low NT-proBNP and low CA-125 (n = 173), low NT-proBNP and high CA-125 (n = 67), high NT-proBNP and low CA-125 (n = 68), and high NT-proBNP and high CA-125 (n = 105). The Kaplan-Meier survival analysis demonstrated that the group with both low marker levels showed best the 2-year survival (87.9%) followed by the group with low NT-proBNP and high CA-125 (76.1%), high NT-proBNP and low CA-125 (64.7%), and both high NT-proBNP and CA-125 levels (54.3%) (p < 0.001) (Fig. 2).

To investigate the incremental predictive value of CA-125 in addition to the established risk factors, we built four models: (1) baseline model including the established risk factors (age, sex, BMI, ischemic etiology, NYHA class, hypertension, diabetes, β-blocker use, ACEI/ARB use, and serum sodium and creatinine levels); (2) baseline model + NT-proBNP; (3) baseline model + CA-125; and (4) baseline model + NT-proBNP + CA-125. The global chi-square of model 1 was 130.2. The addition of serum NT-proBNP and/or CA-125 to the established risk factors increased the global chi-square and provided incremental information for predicting mortality; after adding the NT-proBNP, the global chi-square increased to 138.2 (p < 0.005); combination of CA-125 to the model comprising the established risk factors and NT-proBNP showed additional improvement of the prognostic power for the 2-year mortality (global chi-square from 1,382 to 142.9, p = 0.030) (Fig. 3).

Discrimination analysis of the four models described above is presented in Table 3. Addition of CA-125 alone to the baseline model yielded a category-free NRI (0.456, p < 0.001) and IDI (0.025, p = 0.004), improving the discriminative power of the baseline model; adding NT-proBNP alone to baseline model did not yield significant IDI (category-free NRI 0.262, p = 0.017; IDI 0.012, p = 0.061).