Association between body mass index and prognosis of patients hospitalized with heart failure

The prognostic implications of very low body mass index (BMI) values remain unclear in patients with acute decompensated heart failure (ADHF). This study aimed to investigate the prognostic impact of BMI classification based on the World Health Organization criteria in patients with ADHF. Among 3509 patients with ADHF and available BMI data at discharge in 19 participating hospitals in Japan between October 2014 and March 2016, the study population was divided into five groups; (1) Severely underweight: BMI < 16 kg/m2, (2) Underweight: BMI ≥ 16 kg/m2 and < 18.5 kg/m2, (3) Normal weight: BMI ≥ 18.5 kg/m2 and < 25 kg/m2, (4) Overweight: BMI ≥ 25 kg/m2 and < 30 kg/m2 (5) Obese: BMI ≥ 30 kg/m2. The primary outcome measure was all-cause death. The median follow-up duration was 471 days, with 96.4% follow up at 1-year. The cumulative 1-year incidence of all-cause death was higher in underweight groups, and lower in overweight groups (Severely underweight: 36.3%, Underweight: 23.9%, Normal weight: 14.4%, Overweight: 7.9%, and Obese: 9.0%, P < 0.001). After adjusting confounders, the excess mortality risk remained significant in the severely underweight group (HR, 2.32; 95%CI, 1.83–2.94; P < 0.001), and in the underweight group (HR, 1.31; 95%CI, 1.08–1.59; P = 0.005) relative to the normal weight group, while the lower mortality risk was no longer significant in the overweight group (HR, 0.82; 95%CI, 0.62–1.10; P = 0.18) and in the obese group (HR, 1.09; 95%CI, 0.65–1.85; P = 0.74). Very low BMI was associated with a higher risk for one-year mortality after discharge in patients with ADHF.

Subgroup analysis. The BMI at discharge was significantly lower in the subgroups of women, and patients without edema at discharge than those without (Supplementary Table 2). There was no significant interaction between the subgroup factors and the effect of the each BMI group relative to normal weight group on all-cause death (Fig. 4). prognostic implications of BMi at admission. When we stratified the patients into 5 groups according to BMI at admission based on the WHO classification (Supplementary Table 5, Supplementary Fig. 6), the results were mostly consistent with the main analysis. The excess risk for all-cause death remained significant in the severely underweight group and in the underweight group, whereas overweight was associated with decreased risk of all-cause death and cardiovascular death in patients compared to normal weight ( Supplementary Figs. 7  and 8).

Discussion
The main findings of this study were as follows; (1) Lower BMI, especially severely underweight status, was associated with increased mortality in patients after discharge with HF; (2) Overweight and obesity based on WHO classifications were not associated with increased or decreased risk of death in patients compared to normal weight status; (3) The risk for HF hospitalization was not affected by BMI status. The association of BMI and prognosis in patients with HF has long been investigated. However, there is only one report on the prognostic significance of a severely underweight status in patients with HF. Matsushita et al. reported a severely low BMI was associated with mortality in the patients with ADHF, but the number of patients with a severely underweight BMI were limited 22 and the risk for death compared to that of normal weight status was unclear 22 . Using the large database in Japan, we showed that the severely underweight status was associated with all-cause, cardiovascular, and non-cardiovascular death. Our results are consistent with previous studies, which have shown a lower BMI is associated with a higher risk of death 4,5,7,8,23 . The classification of BMI did not influence the risk of hospitalization for HF in multivariable analyses, which is consistent with the results of the DIG and CHARM sub-studies 5,23 . The 1-year mortality after hospital discharge for ADHF is relatively low in the present study and ranged from 16.5% (cumulative 1-year mortality) to 22.2% in other Japanese studies 24 as compared with that in the United States 25 , despite that older patients were enrolled in the Japanese registries. This might be due to the differences in ethnicity, HF etiology, and enrollment timeframe. Despite these differences, the prognostic influence of low BMI was observed across studies worldwide.
The mechanistic link between underweight status and poor outcome in patients with HF has been proposed. A lower BMI reflects a decrease in skeletal muscle, implying the associated malnutrition and inflammation 17,26,27 . In fact, both the LVEF and NYHA status at presentation were not different among the BMI statuses. The albumin and hemoglobin levels was incrementally lower in the underweight groups. A reduction in food intake, gastrointestinal abnormalities, immunological and neurohormonal activation as well as an imbalance between anabolic and catabolic processes may be important mechanisms to understand these conditions [26][27][28][29] . After adjusting for confounders such as age, sex, and the presence of anemia, the association between being underweight and a poor prognosis remained significant.
In our study, the mortality was lowest in patients with an overweight status, followed by those with an obese status, although there was no significant difference from patients with a normal weight status based on the  30,31 . The patients in the overweight and obese groups were younger, had more metabolic diseases and decreased levels of BNP, and were more likely to be administered with an ACE-I/ARB or β-blocker. Low mortality rates in patients with higher BMI might be related to a greater metabolic reserve against stress 32 , a reduced cardiac sympathetic activity 33 , an attenuated neurohormonal response 34 , and a lower inflammatory cytokine levels, and lesser catabolic-anabolic imbalance 35 .
In the theory of obesity paradox, having a larger BMI is associated with better outcomes; however, many of previous studies stratified patients into two groups for comparisons [36][37][38] . In other studies, risk for all-cause death was lowest in obese patients (BMI ≥ 30 kg/m 2 ) 2,7 . In contrast, a sub-study of CHARM trial reported that patients with BMI ≥ 35 kg/m 2 tended to show a worse prognosis 5 . Nagarajan et al. from Cleveland Clinic HF program demonstrated a poor prognosis in very obese patients (BMI ≥ 40 kg/m 2 ) with advanced HF 39 . In the present study, the effect of higher BMI on mortality was inconclusive mainly due to small number of patients with higher BMI. The prevalence of overweight and obese HF patients was much lower than reports in previous studies based on randomized trials in Western countries 5 and previous studies from Japan conducted in 2004 40 and 2007 36 . The differences in patient backgrounds may be derived from the style of the study and the countries and periods of enrollment, focusing on the increase in aging patients with ADHF. The risk of all-cause death in obese patients was also inconclusive, but that in overweight patients became significant when we adopted the cutoffs for the Asian population. In Japan, the cutoff BMI is authorized by the guidelines of the Japan Society for the Study of Obesity and is basically identical to the WHO classification 30,31 . Defining the ideal BMI values in Japanese patient with ADHF is beyond the scope of the present study, and further studies are required to validate the cutoff BMI in Japan. Ideal body weight in patients with HF should be set individually and we should take ethnicity as well as comorbidities in consideration. Admission BMI was also associated with prognosis, even recognizing the setback of congestion. This result was consistent with the subgroup analysis stratified with or without edema at discharge. Considering the prognostic impact of BMI at admission and discharge, the evaluation of BMI is always critically important for the assessment of patients with ADHF. However, the BMI at admission can be more easily changed through the treatment for ADHF. In the present study, the mean difference between the BMIs at admission and discharge was 1.5 kg/m 2 (Table 1). Thus, the BMI at discharge would be a more reliable marker for patients with ADHF. Our study will be useful to understand the pathophysiology of ADHF and patients' conditions, and to evaluate the prognosis of patients with ADHF. When the BMI of a given patient is severely low, special attention should be paid to the worsening of HF and non-cardiovascular diseases. Future research would be warranted to identify and promote achieving the optimal BMI in individual patients.
Limitations. This study had several limitations. First, we could not determine the body weight at discharge was an optimal body weight without congestion in a given patient. Although the body weight at discharge was decreased compared to that at admission, a substantial proportion of patients had residual edema at discharge. Second, serum levels of cytokines, catecholamines, and renin and aldosterone were not collected. Thus, we can only speculate on the mechanistic link between the low BMI and poor outcome based on the available data in the present study. Third, residual unmeasured confounding factors could affect the results even after extensive adjustment. Due to lots of potential confounders, the conclusion should be treated with caution. Fourth, several subgroup analyses have a risk for multiple comparisons as well as a small sample size with low statistical power. Fifth, a selection bias might have been present. The patients with unavailable BMI data included older patients with anemia and hypoalbuminemia, and low ambulatory status. The lack of BMI data may be due to the patients' non-ambulatory status. The characteristics of the patients with unavailable BMI data were similar to those of underweight or severely underweight patients. The non-ambulatory patients showed worse outcomes 16 ; thus, excluding these patients may not change the result of this study. Sixth, owing to the short-term follow-up, the causal link between BMI and outcome is unclear. Further research studies are needed to clarify the causal link.
conclusion Very low BMI was associated with a higher risk for one-year mortality after discharge in patients with ADHF.