A novel nutritional index as a predictor of mortality in dilated cardiomyopathy: a retrospective study

Methods

Study population

This retrospective cohort study enrolled 445 patients with discharge diagnosis of DCM at first admission to the First Affiliated Hospital of Guangxi Medical University between January 2015 and June 2020. DCM was defined by (a) the presence of left ventricle dilation with left ventricular end-diastolic diameter (LVEDD) > 5.0 cm (female) or > 5.5 cm (male); (b) left ventricular ejection fraction (LVEF) < 45% and left ventricular fractional shortening (LVFS) < 25%; and (c) the exclusion of significant CAD, hypertension, primary valve disease, and congenital heart disease. In addition, we excluded patients with cancer, known inflammatory or infectious diseases, and renal failure. The detailed study flow chart is shown in Fig. 1. The primary endpoint of the study was all-cause mortality. Follow-up outcomes were obtained from hospital medical records or telephone interviews, and follow-up time ended on June 2021. The study protocol was approved by the Human Research Ethics Committee of the First Affiliated Hospital of Guangxi Medical University, China (NO.:2021-KY-E-158) and fulfilled all principles of the Declaration of Helsinki. Given the retrospective study design, no informed consent was required.

Relevant definition

Body weight, height and the laboratory data were all reviewed from the first admission record with DCM discharge diagnosis. Body mass index (BMI) was calculated as weight in kilograms divided by height in meters squared. Diabetes mellitus was defined as fasting glucose ≥7.0 mmol/l, glycated hemoglobin (A1C) ≥6.5%, or self-reported history of diabetes mellitus.

GNRI = [1. 489 ×albumin (g/L)] + 41.7× (measured body weight (kg)/ideal body weight (kg)] (Bouillanne et al., 2005) Ideal body weight = height (cm) −100 − [(height (cm) − 150)/4]c for men, or height (cm) –100 − [height (cm) − 150)/2.5] for women (Bouillanne et al., 2005).

PNI = serum albumin (g/L)+ (0.005 × total lymphocyte count in mm³).

TCBI = triglycerides (mg/dL) × total cholesterol (mg/dL) × body weight (kg)/1000.

CONUT was calculated based on serum albumin concentration, cholesterol level, and lymphocyte count with different point values assigned to various ranges of the laboratory measures (Ignacio de Ulíbarri et al., 2005).

Statistical analysis

Continuous variables were described as mean ± standard deviation (SD) (for normal distributions) or median (interquartile range) (for skewed distributions), and categorical variables were presented as counts (percentages). The TCBI was stratified in tertiles: <701 (Q1), 701–1212 (Q2) and >1212 (Q3), and differences in baseline characteristics of the three groups were assessed by One-way ANOVA, Kruskal–Wallis test, chi-squared test, or Fisher’s exact test as appropriate. Spearman rank correlation was performed to evaluate the association between TCBI and other nutritional indexes. The prognostic value of TCBI was determined using Kaplan–Meier survival analysis with Log-rank test and Cox proportional hazards regression analysis. To assess the prognostic independence of TCBI, we created multivariable Cox regression models. Confounders were selected based on the clinical judgement and P ≤ 0.05 in the univariate analysis. Model 1 was adjusted for age and sex. Model 2 was adjusted for age, sex, and the variables in MAGGIC (Meta-Analysis Global Group in Chronic Heart Failure) risk score (Pocock et al., 2013), including BMI, current smoking, diabetes, symptom duration, systolic blood pressure, New York Heart Association (NYHA) class, LVEF, creatinine, use of beta–blocker, and use of angiotensin–converting enzyme inhibitor/angiotensin receptor blockers (ACEI/ARBs). Model 3 was adjusted for age, sex, and the other potential confounders in this study, including prior stroke, hemoglobin, neutrophils, albumin, uric acid, Na+, high-density lipoprotein cholesterol, low-density lipoprotein cholesterol, lipid–lowering drugs, LVEDD, and N-terminal pro-brain natriuretic peptide (NT-proBNP). To evaluate the predictive efficiency of different nutritional indexes, we compare the time-dependent area under the receiver operating characteristic (ROC) curve (AUC) using the R package ‘timeROC’. Also, multivariable Cox regression with the forward stepwise selection was built to investigate which nutritional indexes had independent prognostic value. To investigate the added prognostic value of TCBI beyond GNRI or NT-proBNP, C-index, continuous net reclassification improvement (NRI), and integrated discrimination improvement (IDI) at 3 years were calculated using the R packages‘survC1’ and‘survIDINRI’. All analysis was conducted with the use of R statistical software version 4.0.3 (R Foundation for Statistical Computing, Vienna, Austria). A two-tailed P value <0.05 was regarded statistically significant.

Baseline characteristics

Baseline clinical characteristics according to the tertiles of TCBI are shown in Table 1. Of note, lower tertile groups of TCBI were likely to be older, female, and had lower BMI, blood pressure, lymphocyte, hemoglobin, albumin, Na+, left ventricular end-diastolic diameter, and a lower frequency of ACEI/ARBs treatment. In addition, lower tertile groups of TCBI had higher levels of NT-proBNP, NYHA class, and a higher frequency of hypolipemia.

Association between TCBI and other nutritional indexes

Spearman correlation analysis showed that TCBI were positively correlated with GNRI (r = 0.461, P < 0.001), PNI (r = 0.322, P < 0.001), but negatively correlated with CONUT (r = −0.516, P < 0.001).

Kaplan–Meier survival and Cox regression analysis

During a median follow-up duration of 2.8 years, the all-cause mortality was observed in 135 out of 445 patients (30.3%). In Kaplan–Meier analysis, the third TCBI tertile had a lower mortality risk (T3 vs. T2 vs. T1: 16.9% vs. 35.1% vs. 38.9%; log-rank P < 0.001; Fig. 2).

All the variables with P ≤ 0.05 in univariable Cox regression analysis were shown in Table 2. Of note, each 1-SD increase in TCBI was associated with a lower risk of death (hazard ratio [HR]: 0.54, 95% confidence interval [CI] [0.39–0.75]; P < 0.001). The risk remained significant after adjustment of age and sex in model 1 (HR per 1-SD increase: 0.54, 95% CI [0.38–0.75]; P < 0.001), after adjustment of the predictors in MAGGIC risk score in model 2 (HR per 1-SD increase: 0.65, 95% CI [0.46–0.92]; P = 0.015), and after adjustment of the other potential confounders in model 3 (HR per 1-SD increase: 0.59, 95% CI [0.40–0.88]; P = 0.010) (Table 3). Moreover, when TCBI was categorized into tertiles, patients in the third tertile were associated with a decreased mortality, whereas there was no significant difference between the T2 and T1 groups (Table 3).

Comparison between TCBI and other nutritional indexes for predicting mortality

The time-dependent AUC of TCBI for predicting mortality was shown in Fig. 3A. At 1 year, the AUC of TCBI, GNRI, PNI and CONUT were 0.630, 0.623, 0.623, and 0.639 (all P < 0.01) respectively (Fig. 3B). At 3 years, the AUC of TCBI, GNRI, PNI and CONUT were 0.631, 0.612, 0.634, and 0.622 (all P < 0.01) respectively (Fig. 3C). There were no statistically significant differences between TCBI and other nutritional indexes for prognostic prediction at 1 or 3 years (AUC comparison, all P > 0.05). In multivariable Cox regression including the four nutritional indexes, only TCBI (HR per 1-SD increase: 0.64, 95% CI [0.46–0.90]; P = 0.009) and GNRI (HR per 1-SD increase: 0.78, 95% CI [0.65–0.95]; P = 0.013) were independent predictor of mortality after the forward stepwise selection, suggesting that TCBI and GNRI might be the better predictors of mortality. We next investigated whether TCBI could significantly improve the prediction over GNRI. The addition of TCBI to GNRI showed no improvements in C-index (0.639 vs. 0.615; P = 0.134). However, TCBI significantly improved continuous NRI (NRI: 0.161, 95% CI [0.023–0.290]; P = 0.027) and IDI (IDI: 0.020, 95% CI [0.002–0.048]; P = 0.020) over GNRI.

Incremental prognostic value of TCBI over NT-proBNP

Considering NT-proBNP as a classic prognostic biomarker of HF, we investigated whether TCBI could significantly improve the prediction of all-cause death over NT-proBNP. The addition of TCBI to NT-proBNP was not associated with a significant improvement in the C-index for prognostic prediction (0.691 vs. 0.679; P = 0.362). However, TCBI significantly improved continuous NRI (NRI: 0.189, 95% CI [0.050–0.305]; P = 0.006) and IDI (IDI: 0.024, 95% CI [0.005–0.053]; P = 0.008) over NT-proBNP.