Association between changes in predicted body composition and occurrence of heart failure: a nationwide population study

Background Large population-based studies on the association between changes in body composition and the occurrence of heart failure (HF) are rare. We aimed to determine the association between changes in body composition, including the predicted body fat mass index (pBFMI), predicted appendicular skeletal muscle mass index (pASMI), and predicted lean body mass index (pLBMI), and the occurrence of HF. Methods For present study, 2,036,940 people who consecutively underwent national health examinations from 2010~2011 (baseline period) to 2012~2013 (follow-up period) were included. The pBFMI, pASMI, and pLBMI were indirectly investigated using validated anthropometric prediction equations from the Korean National Health and Nutrition Examination Survey cohort. The outcome was defined as at least two or more claims of HF. Results During a median of 7.59 years of follow-up, 22,172 participants (event rate, 1.09%) with HF were observed. Decreased changes in the pASMI and pLBMI were associated with the occurrence of HF among males (hazard ratio [HR] 0.966, 95% confidence interval (CI) [0.944-0.988]; HR 0.939, 95%CI [0.923-0.955], respectively) and females (HR 0.924, 95%CI [0.900-0.947]; HR 0.951, 95%CI [0.939-0.963], respectively). An increased change in the pBFMI was associated with the occurrence of HF in males (HR 1.017, 95%CI [1.001-1.034]). However, paradoxically, a change in the pBFMI was associated with the occurrence of HF in females (HR 0.925, 95%CI [0.909-0.942]). Conclusion Decreased skeletal muscle mass was related to the occurrence of HF. However, the relationship between a change in fat mass and the occurrence of HF was different and even paradoxical depending on sex.


Introduction
Heart failure (HF) is a clinical syndrome that is related to decreased cardiac contractility accompanied by impairment of the ejection of blood or ventricular filling (1).HF is a common cardiovascular disease and a leading cause of hospitalization, especially in older adults.The economic burden and prevalence of HF have been increasing worldwide (2).Despite the development of treatment and prevention tools for HF, the morbidity and mortality rates associated with HF are still high (2).Therefore, it is important to identify the hidden risk factors associated with HF.To date, risk factors for HF, including hypertension, diabetes mellitus, coronary artery occlusive disease, aortic atheroma, poor oral hygiene, smoking, and cardiomyopathy, have been suggested.However, information regarding further modifiable associations or occurrence for HF is still lacking (3,4).
Obesity is defined as an abnormal accumulation of healthimpairing fat mass, commonly assessed as a determination of a body mass index (BMI) ≥30 kg/m 2 (5).The prevalence of obesity has increased over the past few decades throughout the world, and the global burden of obesity is currently increasing rapidly (6).It is well known that obesity, especially greater levels of adiposity, is a well-established risk factor for cardiovascular disease and is associated with cardiovascular risk factors, including hypertension, insulin resistance, lipoprotein metabolism, and inflammation (7,8).In a previous study, obesity and abdominal fat were found to be associated with an increased incidence of HF (9).However, when obesity was evaluated using BMI in patients with HF, the obesity paradox that obese patients have better longterm and short-term prognosis was observed (10).Furthermore, because BMI includes not only fat mass but also skeletal mass and lean body mass, cannot distinguish between lean body mass and fat mass, and provides no information on body fat distribution, it is necessary to check the correlation with HF using body composition rather than BMI (11).
Nevertheless, there have been few longitudinal studies targeting a general population of large sample sizes on the association of changes in body composition with the occurrence of HF.Therefore, we aimed to investigate the association between changes in body composition, including the predicted appendicular skeletal muscle mass index (pASMI), predicted body fat mass index (pBFMI), and predicted lean BMI (pLBMI), and the occurrence of HF in a longitudinal setting.

Participants
This study was performed using the National Health Insurance Service-Health Screening (NHIS-HEALS) dataset supplied by the Korean government.In South Korea, adults older than 40 years of age are advised to undergo free health examinations in alternate years.The Korean government combined the health screening dataset with age, sex, household income, and clinical information, including diagnostic codes, prescriptions, information on treatment or procedure, hospitalization, and mortality date (12, 13).For our study, 2,139,856 people who consecutively underwent national health examinations from 2010~2011 (baseline period) to 2012~2013 (follow-up period) were included (dataset number NHIS-2021-1-715) through an identification and validation process (14)(15)(16).Among 2,139,856 participants, those (n=85,947) with at least one missing data were excluded.Those with a previous history of HF (n=16,969) were excluded.Finally, 2,036,940 participants were included in the present study (Figure 1).The present study was approved by the Institutional Review Board of Ewha Womans University Seoul Hospital (Institutional Review Board approval number: SEUMC 2022-02-018).

Change in predicted body composition and covariates
The pBFMI, pASMI, and pLBMI were indirectly evaluated by validated anthropometric prediction equations from the Korean National Health and Nutrition Examination Survey cohort (11).A previous study suggested that body fat mass, appendicular skeletal muscle mass, and lean body mass were identified using dual-energy X-ray absorptiometry, and then a prediction equation was composed using different combinations of age, anthropometric measurements (weight, height, waist circumference), serum creatinine levels, physical activity, smoking status, and alcohol intake (11).These predictive equations were validated as having high predictive power, a moderate agreement rate, and low bias in the Korean general population (11).In the present study, these prediction equations were applied to evaluate the participants' body composition.Body fat mass, appendicular skeletal muscle mass, and lean body mass were presented as indices (weight [kg] divided by height squared [m 2 ]); thus, pBFMI, pASMI, and pLBMI, respectively (Supplementary Methods 1).The change in the pBFMI, termed DpBFMI, was calculated from the difference between the baseline and follow-up periods pBFMI, and the The flowchart of participant selection.changes in the pASMI and pLBMI (DpASMI and DpLBMI) were also calculated similarly.

Outcomes
The index date was defined as the date of the national health examination.The most recent health examination data were used for statistical analysis if more than one examination was performed during the period.Outcomes were defined as two or more claims for HF (21).Previously, the diagnostic accuracy of the ICD-10 code (I50) for HF using the NHIS dataset has been validated (3,21).Monitoring was conducted until December 31, 2020, or the first occurrence of death or HF.

Statistical analysis
Chi-square and paired t-tests were used to compare categorical and continuous variables according to sex.The Cox proportional hazards model presented with a hazard ratio (HR) with a 95% confidence interval (CI) was used to evaluate the effect of changes in body composition (1 kg/m 2 increments in the DpBFMI, DpASMI, and DpLBMI) between the baseline and follow-up periods on the occurrence of HF after adjusting for all potential confounding factors.In addition, we performed an additional multivariable analysis after adjusting for all potential confounding factors, except for variables used in the prediction equations to calculate the predicted body composition, including age, smoking status, alcohol intake, and physical activity.Furthermore, we provided restricted cubic splines of change in the predicted body composition to visually estimate the association between the change in the predicted body composition (DpBFMI, DpASMI, and DpLBMI) and the occurrence of HF.Four knots were placed at the 5th, 35th, 65th, and 95th percentiles of the change in the predicted body composition.Stratified analyses of the effects of changes in body composition on the occurrence of HF were conducted according to the subgroups of age, household income, alcohol intake, smoking status, physical activity, systolic blood pressure, fasting serum glucose, total cholesterol, CCI, and BMI at baseline and the follow-up categories: normal weight (18.5-24.9kg/m 2 ), overweight (25-29.9kg/m 2 ), and obese (≥ 30 kg/m 2 ).To evaluate the combined effects of changes in the BMI, pASMI, pBFMI, and pLBMI, stratified analysis was additionally performed to evaluate the effect of changes in BMI during the two health examinations on the occurrence of HF according to changes in body composition.Statistical analyses were performed using R software, version 3.3.3(R Foundation for Statistical Computing, Vienna, Austria) and SAS version 9.4 (SAS Inc., Cary, NC, USA).Two-sided P-values less than 0.05 were considered significant.

Results
Table 1 shows the results of comparing the baseline characteristics of the 2,036,940 participants according to sex (n=1,070,377 for males, n=966,563 for females).During a median of 7.59 years of follow-up, 22,172 participants (event rate, 1.09%) with HF were observed.The mean age of males was 49.60 ± 13.45 years, and of females was 52.32 ± 13.65 years.Significant differences in the BMI at baseline and follow-up, household income, smoking status, alcohol intake, physical activity, systolic blood pressure,   1).Multivariate analysis showed a relationship between the occurrence of HF and changes in body composition (Table 2; Supplementary Table 1).In males, with a 1 kg/m 2 increase in the DpBFMI, the HR for the occurrence of HF was 1.017 (95%CI: 1.001-1.034),and the HRs for a 1 kg/m 2 increase in the DpASMI and DpLBMI were 0.924 (95%CI: 0.900-0.947)and 0.951 (95%CI: 0.939-0.963),respectively.In females, with a 1 kg/m 2 increase in the DpBFMI, the HR for the occurrence of HF was 0.925 (95%CI: 0.909-0.942),for a 1 kg/m 2 increase in the DpASMI and DpLBMI were 0.966 (95%CI: 0.944-0.988),and 0.939 (95%CI: 0.923-0.955),respectively.In contrast, decreases in the DpASMI and DpLBMI were associated with an increased incidence of HF in both sexes.Furthermore, a decrease in the DpBFMI was associated with a decreased HF incidence in males, whereas a decrease in the DpBFMI was associated with an increased HF incidence in females (Table 2; Supplementary Table 1 and Figure 2).For the stratified analysis, these trends and results were similar regardless of the BMI status at baseline.The occurrence-reducing effects of increased DpLBMI and DpASMI on the occurrence of HF were more prominent when the participants were less obese.Furthermore, the occurrence-reducing effect of an increased DpBFMI on the occurrence of HF was more prominent when female participants were less obese.In contrast to females, the occurrence-increasing effect of increased DpBFMI on the occurrence of HF was more prominent when male participants were less obese (Table 3; Supplementary Table 2).
The occurrence-reducing effects of the increase in DpASMI and DpLBMI were observed in male participants whose BMI increased from normal to obese or overweight and from overweight to obese.The occurrence-reducing effect of the increase in the DpASMI and DpLBMI was also observed in female participants whose BMI increased from normal to overweight and obese.The decrease in the DpBFMI showed an occurrence-reducing effect in male participants whose BMI decreased from overweight or obese to normal and from obese to overweight.However, the increase in the DpBFMI showed an occurrence-reducing effect in female participants whose BMI increased from normal to obese or overweight and from overweight to obese (Table 4; Supplementary Table 3).In the subgroup analysis, there was no h e t e r o g e n e i t y a m o n g t h e g r o u p s , e x c e p t f o r a g e (Supplementary Table 4).

Discussion
In this large-scale cohort study, we found that decreased DpASMI and DpLBMI were negatively correlated with the occurrence of HF, regardless of sex.The occurrence-reducing effects of increased DpASMI and DpLBMI on the occurrence of HF were alleviated when the participants were more obese and when their BMI increased.In males, an increased DpBFMI showed a positive relationship with the occurrence of HF.The occurrenceincreasing effect of increased DpBFMI on the occurrence of HF was more prominent when male participants were less obese and when the BMI of participants decreased.In contrast, decreased DpBFMI showed a negative relationship with the occurrence of HF in females.The occurrence-reducing effect of increased DpBFMI on the occurrence of HF was alleviated when the female participants were obese and when their BMI increased.
Previous studies showed that sarcopenia, which is primarily characterized by a progressive and widespread decline in skeletal muscle mass and muscle dysfunction, is common in patients with HF (about 10%) and is strongly related to the prognosis of HF (22).Also, loss of lean body mass was related to the occurrence of HF, particularly among older males (23).Various mechanisms, including malnutrition and anorexia, physical inactivity, and insulin resistance, cause an increase in muscle protein catabolism, resulting in a loss of muscle or body weight, which is associated with HF (24).Although present study could not confirm the causal relationship, in other hand, some studies suggested that muscle wasting is a consequence of HF and HF lead to several complications including musculoskeletal abnormalities and sarcopenia (25,26).The present study showed that increased change in muscle mass had a negative relationship with the occurrence of HF, and the negative correlation was more attenuated when the participants were more obese and when the BMI of participants increased.
This study investigated sex-related differences in the association between changes in body fat mass and HF.Although the fat distribution varies according to sex, the role of obesity in the occurrence of HF in females may differ from that in males.Generally, males store sufficient fat in a visceral distribution, while females store sufficient fat in a peripheral subcutaneous distribution (27).These differences in location for stocking excessive fat have been shown to persist even after menopause.In males, visceral fat mass is associated with adiposopathy, which includes fat cell hypertrophy, increased circulating free fatty acids, and inflammatory and immune responses, leading to common metabolic diseases and HF (28,29).The present study showed that an increased change in the body fat mass had a positive relationship with the occurrence of HF, and the occurrenceincreasing effect of the increased change in body fat mass was more noticeable when the male participants were less obese and Association of the changes in the predicted body composition index with heart failure.Association of the changes in the predicted appendicular skeletal muscle mass index, body fat mass index, and lean body mass index (BMI) with heart failure.The solid lines indicate the hazard ratios, and the shaded regions show 95% confidence intervals from restricted cubic spline regression.Restricted cubic splines were constructed with four knots placed at the 5 th , 35 th , 65 th , and 95 th percentiles of the change in the predicted appendicular skeletal muscle mass index, body fat mass index, and lean BMI.The hazard ratios (95% confidence interval) were calculated by Cox proportional hazards regression analysis after adjusting for age, BMI at baseline and follow-up, household income, smoking status, alcohol intake, physical activity, systolic blood pressure, fasting serum glucose, total cholesterol, and Charlson Comorbidity Index.when the BMI of participants decreased.In females, peripherally distributed fat has a negative impact on exercise intolerance (30).Also, female patients with HF were found to be in a catabolic state and have more metabolic reserves.If patients experienced malnutrition, the burden of morbidity and mortality for HF increased, and loss of muscle mass occurred, leading to advanced HF (31).In contrast, obese patients showed low adiponectin levels and a decreased catecholamine response, which led to improved HF (32).Finally, obese females showed higher fatty acid uptake and less myocardial glucose utilization compared with obese males, which resulted in differences in myocardial metabolism (33,34).The present study showed that increased changes in the body fat mass had a negative relationship with the occurrence of HF, and the occurrence-reducing effect of the increased change in body fat mass was more alleviated when the female participants were obese and when the BMI of participants increased.The strength of our study is that it showed an association between changes in the body composition indices and the occurrence of HF in a large sample of the Korean general population in a longitudinal setting.However, our study has several limitations.First, present study has been conducted using anthropometric prediction equations to estimate body composition.Due to time constraint and cost, measurement of body composition using accurate modalities remains challenging issue in large epidemiological studies.Therefore, we decided to proceed with present study based on previous study which developed prediction equations to indirectly assess body composition using data obtained with dual-energy X-ray absorptiometry as reference (11).These anthropometric prediction equations from the Korean National Health and Nutrition Examination Survey cohort was validated, but it is major limitation of our study.Second, the equation for predicted body composition was estimated for the Korean population aged > 19 years; therefore, these findings may not be generalizable to other ethnicities and may not be directly applicable to different geographic and/or demographic settings.Third, although it was conducted with a longitudinal design, this is a retrospective study; therefore, we could not confirm the causal relationship or exclude confounders.Finally, because this is an epidemiologic study, our study cannot explain the basic mechanism of the association between changes in body composition and HF.
In conclusion, changes in body composition are associated with the occurrence of HF.Decreased DpASMI and DpLBMI are associated with the occurrence of HF.Furthermore, in males, increased DpBFMI was associated with the occurrence of HF.However, in females, decreased DpBFMI was associated with the occurrence of HF.

TABLE 1
Baseline characteristics of the study participants.

TABLE 2
Hazard ratios and 95%CI of heart failure per 1 kg/m 2 increase in the change in the predicted body composition index.
The multivariable model was used for the BMI at the baseline and follow-up period, household income, systolic blood pressure, fasting serum glucose, total cholesterol, and Charlson Comorbidity Index.BMI, body mass index; CI, confidence interval; HR, hazard ratio; DpASMI, change in predicted appendicular skeletal muscle mass index; DpBFMI, change in the predicted body fat mass index; DpLBMI, change in the predicted lean body mass index.

TABLE 3
Hazard ratios and 95%CI of heart failure per 1 kg/m 2 increase in the change in the predicted body composition index stratified by the BMI group.The multivariable model was used for the BMI at the baseline and follow-up periods, household income, systolic blood pressure, fasting serum glucose, total cholesterol, and Charlson Comorbidity Index.BMI, body mass index; CI, confidence interval; HR, hazard ratio; DpASMI, change in the predicted appendicular skeletal muscle mass index; DpBFMI, change in the predicted body fat mass index; DpLBMI, change in the predicted lean body mass index.

TABLE 4
Hazard ratios and 95%CI of heart failure per 1 kg/m 2 increase in the change in the predicted body composition index stratified by a change in the BMI group.

TABLE 4 Continued
Multivariable model was adjusted for age, BMI at baseline and follow-up, household income, smoking status, alcohol intake, physical activity, systolic blood pressure, fasting serum glucose, total cholesterol, and Charlson Comorbidity Index.BMI, body mass index; CI, confidence interval; HR, hazard ratio; DpASMI, change in the predicted appendicular skeletal muscle mass index; DpBFMI, change in the predicted body fat mass index; DpLBMI, change in the predicted lean body mass index.