Cutoff points of BMI for classification of nutritional status using bioelectrical impedance analysis

The prevalence of overweight and obesity is increasing at an alarming rate in developed and developing countries the world over [1,2]. Epidemiological studies indicate that overweight and obesity are important risk factors for diabetes, cardiovascular disease, cancer and premature death [2]. Thus, this increased prevalence, in combination with the concomitant health risks, becomes a particularly relevant worldwide public health challenge [3].

This increased prevalence of excess weight/obesity has been reported in various world regions and, in view of its growing importance for public health, precise estimates of the global load of this condition are crucial for the elaboration of public health strategies directed at its primary prevention and treatment [3].

According to the World Health Organization (WHO), obesity is defined as excess adipose tissue [4] and, in view of its increasing incidence, it is considered a worldwide epidemic affecting both developed and developing countries [5].

The WHO projected that by 2005 the world would have 1.6 billion people older than 15 years with excess weight (BMI ≥ 25.0 kg/m²) and 400 million people with obesity (BMI ≥ 30.0 kg/m²). The projection for 2015 was even more pessimistic, i.e., 2.3 billion people with excess weight and 700 million obese people, indicating a 75% increase in cases of obesity within 10 years, with Brazil occupying 77^th position in the WHO ranking [6].

The index universally accepted for the diagnosis and classification of obesity is the BMI, which is expressed by subject weight in kg divided by squared height in meters (weight/height²) [7]. In 1997, the WHO adopted this index as reference for the measurement of obesity, with overweight being defined as a BMI range of 25.0-29.9 kg/m² and obesity as a BMI above 30.0 kg/m² [8]. These values were obtained based on the associations between BMI and mortality in European populations, which formed a J curve between the values of 18.5 and 25.0 kg/m² [9]. Since then, these cutoff points have been used as standard for different populations and different ethnic groups based on the assumption that these different ethnic groups have a similar mortality/morbitity risk.

Recently, a new entity called normal weight obesity (NWO) was created, representing a condition in which an individual has an adequate BMI but an increased percentage of body fat and a higher risk of developing chronic diseases [10]. Although the investigations are recent, it has been estimated that about 30 million Americans are affected by NWO [11,12]. However, this prevalence has not been well established and there is wide variation among the studies conducted thus far, attributed to aspects such as ethnic differences, diverse methods used to assess body composition and different cutoff points established for a diagnosis of NWO.

In this respect, recent studies [13,14,15] have demonstrated that when the traditional BMI is considered, there still is controversy about the best BMI for the classification of obesity in different populations. In addition, recent evidence suggests that the relation between increased BMI and percent body fat also differs among various ethnic groups.

Thus, in view of this tendency, the objective of the present study was to propose a refinement of the reference points of the traditional BMI for the classification of nutritional status, especially overweight/obesity, in a sample of the Brazilian population, and to discuss the new proposed cutoff points for the BMI described in the literature in order to favor an earlier clinical intervention with the consequence of reduction of healthcare costs.

The study was conducted on 1301 volunteers, 386 (29.7%) of them men and 915 (70.3%) women. Mean age was 3.5±17.6 years for men and 35.4±17.6 years for women (p=0.3033) and mean weight was 75.8±15.9 kg for men and 64.2±14.8 kg for women (p<0.0001). Mean height was 173.9±8.2 cm for men and 161.1±7.1 cm for women (p<0.0001). Mean BMI was 24.9±5.5 kg/m² for all subjects as a whole and 25.0±4.6 kg/m² for men and 24.8±5.8 kg/m² for women (p=0.5471).

The following bioelectrical impedance results were obtained: mean total fat mass (%) 27.6±10.0%, fat-free mass (kg) 45.8±10.2, and total body water (%) 52.8±7.5. According to these data, the statistical differences between genders were p<0.0001 for total fat mass, p<0.0001 for fat-free mass, and p<0.0001 for total body water.

Figure 1 illustrates subject distribution separated by gender according to the classification ranges of the BMI suggested by the WHO [4]. A small percentage of students was found to be in the undernutrition range (BMI <18.5 kg/m²), which was slightly more elevated in men (7.7%) than in women (7.4%). Most of the subjects of both genders were of normal weight (BMI of 18.5 to 24.9 kg/m²), with values of 45.0% and 53.5% for men and women, respectively. On the basis of the BMI values that characterize excess weight, 32.6% of the men and 22.3% of the women were overweight (≥25.0 and <29.9 kg/m²), 12.2% of the men and 9.9% of the women were in the range of grade I obesity (BMI of 30.0 to 34.9 kg/m²), and a minority had grade II obesity (BMI of 35.0 and 39.9 kg/m²) and grade III obesity (≥40.0 kg/m²), with values of 1.8% and 0. 5% for men and 4.3% and 2.5% for women, respectively.

Figure 1

Table 2 lists the distribution of the subjects within each BMI category and also according to the body fat ranges (%) suggested by Gallagher et al. (2000) for the 20-39 year age range for a joint African American and Caucasian sample that was closest to the predominant sample of the present study considering the mean age of both genders [16]. It was possible to observe that only 52.3% (3.1%. 29.3%. 10.1% and 9.8%) of men and 68.4% (4.9%. 39.9%. 10.5% and 13.1%) of women were properly categorized by the BMI. Also according to BMI classification, 47.1% of men and 39.0% of women had excess weight (BMI of 25 to 29.9 kg/m² and above 30.0 kg/m².

Table 1

Anthropometric and body composition characterization of the study sample.

Table 2

Correspondence between the cutoff points of the BMI (kg/m²) and body fat ranges (%) according to Gallagher et al (2000).

Table 3 is a comparative presentation of the main studies cited above and their respective proposals of specific cutoff points according to ethnicity and/or gender. However, previous studies by our group conducted on smaller samples have already suggested the adoption of lower BMI cutoff points for the classification of obesity, i.e., 26.11 kg/m² for men and 25.3 kg/m² for women in a sample of 100 subjects [13]. In a later study conducted on 501 participants, the cut-off points detected for the BMI were: 22.67 kg/m² (for body fat = 30%) and 25.24 kg/m² (for body fat = 35%), both for women; 25.33 kg/ m² (for body fat = 20%) and 28.38 kg/m² (for body fat = 25%), both for men, for the classification of overweight and obesity, respectively, since there is a tendency to also reduce these body fat values to 20% for men and 30% for women in order to include a larger number of individuals and to permit an earlier intervention [17].

Table 3

Comparative summary of studies proposing new cutoff points for the BMI for the classification of overweight/obesity in men and women.

In the present study, which involved a larger sample size, the expectation was to maintain this tendency to a reduction of BMI values for the classification of obesity, also observed as a global tendency according to studies by other research groups. In addition, this larger sample (n=1301) is certainly more representative of the profile of the Brazilian population. On this basis, as shown in Figure 2, the final cutoff poins for the BMI suggested for the classification of excess weight in the present sample of the Brazilian population were 24.27 kg/m² (for body fat = 30%) and 27.02 kg/m² (for body fat = 35%), both for women; 25.24 kg/ m² (for body fat = 20%) and 27.15 kg/ m² (for body fat = 25%), both fr men, for the classification of overweight and obesity, respectively (Figure 2).

Figure 2

Figure 3 illustrates the different percentages of subjects to be classified as having excess weight (obesity) according to different criteria. Thus, if we adopt the cutoff point traditionally proposed for the BMI (more than 30 kg/m²), 14.5% and 16.7% of men and women, respcetively, are classified as obese. In contrast, using fat mass as reference (25% for men and 35% for women), the percentage of subjects to be included in this classification increases to 24.8% and 33.6% of men and women, respectively. Finally, when the new ranges for the classification of excess weight according to the BMI are adopted (27.15 kg/m² for men and 27.02 kg/m² for women), the number of subjects included becomes 28.7% for both genders in the sample studied here.

Figure 3

Obesity has been routinely described as an epidemic and a significant threat to global health and the use of the BMI takes on a central role in the definition of this excess weight. The BMI has a double function, i.e., it is used both to measure a body attribute (weight in relation to height) and to define a condition (obesity or escess weight).

This index, which is calculated using the formula weight/height², was suggested as a new index of excess body weight in 1972 and by the mid-nineties, it became the standard method for both the measurement and definition of obesity.

The subsequent steps involved the definition of its cutoff points, with the proposed values, especially for overweight and obesity, being derived from graphs of mortality rate against BMI. These graphs were often described as having a “J curve” shape, in which a BMI of 30.0 kg/m² or 31.0 kg/m² marked a point of significant increase of risk when the curve becomes much more accentuated. Even during the 1980′s and 1990′s, when the BMI was being extensively used, the specific cutoff points used varied among sudies. For example, in 1985 the US National Institutes of Health Conference decided to use a cutoff point for obesity as a BMI of 27.8 kg/m² for men and of 27.3 kg/m² for women (National Institutes of Health [NIH] Panel on Obesity, 1985), whereas most British investigations used a BMI of 25.0 kg/m² for overweight and a BMI of 30.0 kg/m² for obesity for both genders.

However, at the end of the 1990 decade, the cutoff points used by the US government were changed. Thus, when the existing cutoff points of the NIH were used, the prevalence of overweight in the adult population was 33.3% for men and 36.4% for women, whereas when the cutoff points of Bray and Garrow were applied the prevalence became 59.4% for men and 50.7 % for women. This led to the statement that “By simply changing the overweight cutoffs, the estimated number of overweight adults increases from 61.7 million (BMI ≥ 27.8 and 27.3) to 97.1 million (BMI ≥ 25.0), representing a difference of 35.4 million overweight adults. This example calls attention to the actual effect that a shift in BMI criteria can have on determining the population at risk” [18]. At the end of the 1990′s the BMI was sufficiently well established and was defined as a consensus by the WHO in a technical report on anthropometry (WHO, 1995) in which BMI values of 25.0 to 29.9 kg/m² were used for overweight and values above 30.0 kg/m² were used for obesity. And these have been the cutoff points traditionally used since then in studies using the BMI.

However, the precision of the BMI for the assessment of body fat will continue to be debated since the BMI, although extensively used as a substitute measurement of adiposity, is more a measurement of excess weight in relation to height than a measurement of excess body weight. The BMI does not distinguish between fat mass and fat-free mass nor does it consider the distribution of fat throughout the body. Thus, other measures of adiposity such as waist circumference and waist-hip ratio and information about skinfold thickness and body fat should also be adopted in addition to the BMI.

In this respect, several research groups have discussed the adoption of cutoff points mainly by considering ethnic and body composition differences among world populations and the resuls of these studies are summarized in Table 3. We emphasize the study by Deurenberg-Yap et al. (2000) who investigated the relation between percent body fat and BMI in three different ethnic groups in Singapore (Chinese, Malay and Indians) for a total of 291 subjects and observed that lowering the cutoff point of the BMI for the classification of obesity from 30.0 kg/m² to 27.0 kg/m² (for Chinese and Malay subjects) and to 26.0 kg/m² (for Indian subjects) could increase the prevalence of obesity from 6% to 16% in this population, having an important impact on public policies as well as massive economic implications [19].

Frankenfield et al. (2001), in a study of 141 North American individuals (53 men/88 women) using bioelectrical impedance proposed BMI values of 22.6 kg/m² for men and 20.1 kg/m² for women to determine obesity in that population [20]. In the same year, Ko et al. (2001) analyzed the relation between BMI and percent body fat obtained by electrical bioimpedance in 5153 Chinese subjects (3734 women and 1419 men) residing in Hong Kong and suggested that BMI values of 23.0 kg/m² and 26.0 kg/m² would be the most appropriate for the definition of cutoff points for overweight and obesity, respectively, in that population [21]. Dudeja et al. (2001) established appropriate BMI cutoffs for the definition of overweight considering the percentage of fat mass in a healthy Asian Idian population of Northern India [22]. In a study of 123 subjects, on the basis of a ROC curve, the authors proposed that lower BMI values of 21.5 kg/m² for men and 19.0 kg/m² for women showed optimum sensitivity and specificity and fewer errors in the identification of individuals with a high percentage of body fat.

Oh et al. (2004) studied a cohort of 773,915 Korean men and women aged 30 to 59 years followed up for about 10 years and observed that populations with BMI ≥ 25.0 kg/m² were increasing rapidly and had a substantial risk for disease. Thus, they suggested that Korea should consider the use of a BMI of 25.0 kg/m² as a point for the prevention of obesity and for the control of interventions [23]

Another study conducted on 139 young Japanese women proposed that a BMI value of 23.0 kg/m² reflects the current level of fat obtained by DXA better than a BMI of 25.0 kg/m² and therefore maximizes the effectiveness of the BMI as a screening tool in this specific population [24].

Bozkirli et al. (2007), based on a sample of 909 adult Turks (249 men and 660 women), proposed BMI values of 28.24 kg/m² for men and 28.02 kg/mb² for women, thus obtaining an increase from 361 to 457 (a 26.6% increase) in the number of obese women and an increasee from 79 to 132 (a 67.1% increase) in the number of obese men when these new cutoff points were used [25].

Romero-Corral et al. (2008), studying a large sample of 13,601 North American individuals of both genders, proposed cutoff BMI values of 25.8 kg/m² for men and 25.5 kg/m² for women for the classification of obesity [26]. Laughton et al. (2009), in a study of 77 subjects (63 men and 14 women) with spinal cord injury, demonstrated that a cutoff BMI value of 30.0 kg/m² failed to identify 73.9% of the obese participants and that subjects with a spinal cord injury and a BMI ≥ 22.1 kg/m² are considered to be at high risk to develop chronic diseases related to obesity [27].

Previous studies by our group have also confirmed this tendency. Mialich et al. (2011), in a study of 100 Brazilian subjects of both genders, detected BMI ranges of 21.84 – 26.11 kg/m² for men and 22.0 – 25.3 kg/m² for women for the classification of overweight and obesity in this population ⁽¹³⁾. And in 2014, Mialich et al. (2014) proposed cutoff points of 28.38 kg/m² and 25.24 kg/m² for a slightly larger sample of the same population [17].

More recent studies continue to raise questions. Gupta and Kapoor (2012) assessed 578 adult subjects (271 men and 307 women ) from Northern India and observed that the best cutoff points for the BMI for overweight and obesity capable of predicting hypertension would be 22.9 kg/m² and 28.8 kg/m² for men and women, respectively [14]. Gómez-Ambrozi et al. (2012) studied 6123 Caucasians originating from Spain (4208 women and 1915 men) and concuded that the most accurate current cutoff points for the diagnosis of obesity in this population were 29.0 kg/m² and 27.0 kg/m² for men and women, respectively, and that the inclusion of measurements of body composition in routine clinical practice could improve the diagnosis and the decision for the most appropriate treatment of obesity [15].

In the present study, we expected to obtain lower proposed cutoff points for the classical BMI than those traditionally adopted according to the classification ranges proposed by the WHO, i.e., 25.0 kg/m² for overweight and above 30.0 kg/m² for obesity. On this basis, the results obtained here were 27.15 kg/m² for men and 27.02 kg/m² for women, in agreement with this worldwide tendency to propose new cutoff poins for the BMI, most of them lower than those currently used and also specific for gender and ethnicity.

Thus, we may conclude that a diagnosis of overweight and obesity based on BMI classification can provide valuable information about the increase in body fat and permit a significant inter- and intrapopulation comparison of body weight. In addition, it can identify persons or groups at risk for morbidity and mortality, thus opening the way for the identification of priorities at the individual or community level and for the assessment of the efficacy of such interventions, especially when its reference points are adjusted to the ethnic and body composition characteristics of the population under study.

Thus, the development of the BMI represented an important factor of standardization, which solved concrete problems of investigation and led to the creation of new types of knowledge and new ways of working, consequently facilitating the development of science related to obesity. The adoption of cutoff points for the BMI was crucial for the definition of obesity as an epidemic, permitting the creation of large data banks for the screening of alterations at the population level both regarding mean body weight and a schematic presentation of these changes. These two new entities made it possible for researchers in this area and for others to argue that obesity should be defined as an epidemic.

Finally, we may conclude that, despite the many merits of the BMI, it is also essential to consider that the nutritional profile of the world population has changed considerably from the time when it was developed to the present day, when we are witnessing a significant increase in excess weight/obesity in the population. This justifies the need to propose a refinement of this index, i.e., a reevaluation of its cutoff points so that new classification ranges may be later proposed in order to contemplate a larger number of individuals and to take into account ethnic differences between populations, thus permitting an earlier clinical intervention with a consequent reduction of health care costs. Thus, further studies using appropriate methods are needed in order to fully explore the gamut of questions related to the new classification ranges and/or more appropriate and specific cutoff points for populations that use the BMI as at tool for the assessment of nutritional status.