Over the past five decades, obesity has become a major cause of increased healthcare costs and decreased quality of life in many countries. ¹⁾ Furthermore, obesity is strongly associated with hypertension, type 2 diabetes, and atherosclerotic cardiovascular disease.^1,2) A number of studies have reported that a high body mass index (BMI) is associated with different types of cancer.³⁾ The proposition that overweight or obesity is associated with a greater risk of mortality than normal weight is controversial.⁴⁾ This obesity paradox has been corroborated by studies involving a range of ethnic groups and patients with a chronic morbidity, such as hypertension, diabetes mellitus, chronic kidney disease, or heart failure.^5,6,7,8)

The World Health Organization (WHO) has reported that the worldwide prevalence of obesity has approximately doubled since 1980.⁹⁾ Studies have also found that the increased prevalence of obesity in South Korea is associated with an increased incidence of obesity-related diseases, with an estimated socioeconomic cost of 1,792.3 billion Korean won (KRW) in 2008.^10,11)

A study based on the Korea National Health and Nutrition Examination Survey data (2012) found that the prevalence of age-specific obesity, as defined by the WHO criteria, was not higher in South Korea than in other countries.¹¹⁾ However, for Asians, the WHO suggested using more lenient BMI categories compared with those used in Europe: normal, 18.5–22.9 kg/m²; overweight, 23–24.9 kg/m²; moderate obesity, 25–29.9 kg/m²; and severe obesity, ≥30 kg/m². This is because although South Korea has a lower obesity prevalence, a higher sensitivity in terms of obesity-related risk factors and cardiovascular-related mortality has been reported.^12,13,14,15)

The BMI criteria set out by the WHO have been generally accepted; however, many nations try to adjust the criteria for obesity according to the ethnic composition of their populations. A number of studies have reported the association between BMI and mortality in South Korea.^{4,16,17,18,19,20)} However, there are limited studies specifically investigating the association between BMI and mortality according to sex in the elderly.

The purpose of this study was to analyze the BMI in community-dwelling persons aged 65 and over in order to compare participants who survived a 3-year follow-up (survival group) and those who were registered as deceased at follow-up (deceased group). This study also evaluated the current criteria for the definition of obesity.

Table 1 shows the sociodemographic characteristics and functional status of participants. Of 10,613 subjects enrolled in the study, 6.2% (657) died and 93.8% (9,956) survived by 2011. The mean ages of participants in the death and survival groups were 76.99 and 72.63 years, respectively, and the proportion of male participants in each group was 57.2% and 40.7%, respectively. Subjects who were underweight (BMI <18.5 kg/m²) had a higher mortality rate (16.0% vs. 5.0%, P<0.001), whereas the moderate obesity group (25.0–29.9 kg/m²) showed lower mortality rates (17.2% vs. 27.7%, P<0.001). The participants who died during the follow-up period exercised less (≥150 minutes per week: 2.6% vs. 6.7%, P<0.001), were more dependent in terms of instrumental daily activities (K-IADL score ≥11: 45.1% vs. 23.2%, P<0.001), and had the highest score for symptoms of depression (GDS score ≥8, 37.6% vs. 24.6%, P<0.001). We also compared the two groups according to the following variables: smoking, alcohol intake, education level, employment status, residence, household income, cognitive functional status, and existence of hypertension, dyslipidemia, and diabetes mellitus. There were statistically significant differences between the two groups, except for smoking and education level.

The risk of mortality according to BMI in the groups of men, women, and both sexes, is shown in Table 2. The mortality risk according to the BMI in the unadjusted analysis (model 1) was significantly different from that of the diverse variant-adjusted analyses (models 2, 3, and 4). In the case of model 1, mortality in the underweight group (BMI <18.5 kg/m²) had a higher OR than that of the reference group across sexes (men: OR, 3.42; women: OR, 2.25; total: OR, 2.80; P<0.001). The moderate obesity (BMI 25.0–29.9 kg/m²) group showed a lower risk of mortality than the reference group across sexes (men: OR, 0.55; women: OR, 0.63; total: OR, 0.55; P<0.001). In model 2, adjusted for age, there was a significantly higher risk of mortality in the severe obesity group (BMI ≥30 kg/m²) than in the reference group in men and women. However, the decreased mortality risk in the moderate obesity group shown in the model 1 analysis was not observed. In the analysis adjusted for age, smoking, alcohol intake, and exercise level (model 3), the mortality was increased in the underweight group compared with the reference group across all groups stratified by gender, but decreased mortality in the moderate obesity group was found only in men. In model 4, adjusted for education level, employment, residence, household income, K-IADL, GDS, MMSE, and presence of hypertension, dyslipidemia, and diabetes mellitus in addition to those factors adjusted for model 3, showed the same trend as in model 3 (Table 2).

In this study, using the WHO criteria for the definition of obesity, adapted to the Asian context, the risk of mortality in the moderate obesity group (BMI 25.0–29.9 kg/m²) was lower than in the normal weight group (reference group, BMI 18.5–22.9 kg/m²), and the underweight group (BMI <18.5 kg/m²) had a 2.10-fold higher mortality rate than the reference group. These results are consistent with previous research and support the association between BMI and mortality in elderly people.^{16,21,22,23,24)} The lowest mortality rate was found in elderly male subjects in the moderate obesity group (BMI 25.0–29.9 kg/m²), although this was not observed for corresponding elderly female subjects.

In previous studies examining the association between BMI and mortality, an increased mortality rate was observed in the underweight and severe obesity group, which demonstrated a U-shaped association. ^{16,21,22,23,24)} The BMI category of 21.0–27.4 kg/m² has been identified as the group with the lowest mortality rates. Researchers have suggested that the lower and upper limit of the standard BMI range should be altered from the current value of 18.5 to 21.0 kg/m² and from 23.0 to almost 27.5 kg/m², respectively.¹⁷⁾ A number of large-scale cohort studies also recommend adjusting the standard BMI range.^16,18,21) Furthermore, another study proposed that the optimal BMI range should be age- and sex-specific as BMI has been shown to increase depending on age in both sexes.¹⁹⁾

Research in South Korea on the effects of frailty and BMI on mortality rates in community-dwelling elderly people showed that frailty and BMI significantly increased the mortality rate in men and women.²⁰⁾ In our study, the analysis was stratified according to sex, and the association between BMI and mortality rate was statistically significant in men, but not in women, after adjusting for confounding variables.

Dorn et al.²⁵⁾ reported that mortality secondary to cardiovascular disease, the most common cause of death in the elderly, is lower in women than in men in this age group. This may explain the lack of a statistically significant association between BMI and mortality rate in older women compared with men.

Furthermore, there are differences in body composition and body fat distribution between women and men. The BMI in men is largely determined by muscle mass rather than body fat mass, whereas the BMI in women is generally determined by the latter. Therefore, the BMI categories are not equally applicable to both sexes, and the waist to hip circumference ratio, which measures women's obesity by reflecting body fat distribution, appears to be a more accurate predictor of mortality than BMI.²⁵⁾ This is consistent with evidence suggesting that abdominal obesity is an independent risk factor for increased mortality and that both BMI and the waist to hip circumference ratio are important predictors of mortality.²⁶⁾ Therefore, in older women, it is important to examine alternative measurements of body fat distribution in addition to BMI.

There are several limitations to our study. First, the number of subjects was relatively small compared with earlier large-scale cohort studies. Furthermore, the limited 3-year follow-up period made it difficult to identify BMI or existing comorbidity as the primary cause of death.

Nonetheless, our study is distinct from other studies because we adjusted for older age-specific factors strongly associated with mortality in this age group, such as K-IADL, MMSE, and GDS scores. Moreover, this is the first study to investigate sex-specific mortality rates according to BMI. Further studies examining the association between BMI and increased mortality rates in men and women over a longer follow-up period are required.