Determining Body Composition‐Based BMI Cut‐Offs for Ethiopian Adults in Northwest Ethiopia: A Population‐Based Cross‐Sectional Study

ABSTRACT

Background

Body mass index (BMI) is widely used as a population‐level measure of obesity, given its robustness and inclusion in health surveys. However, for Ethiopians, the applicability of universally recognized anthropometric cutoffs may be inadequate. This study aimed to determine body composition‐based BMI cutoffs for the healthy adult population of Ethiopia.

Method

A population‐based cross‐sectional study was conducted in the Amhara region of Ethiopia from June to August 2023, collecting anthropometric data and body fat percentage from 838 adult participants. The body fat percentage was measured using the Omron Body Composition Monitor and Scale, which utilizes the bioelectrical impedance analysis (BIA) method to estimate various body measurements. Data were analyzed using Stata and MedCalc software. BMI cutoff values were determined using the receiver operating characteristics (ROC) analysis, and performance was assessed using area under the curve (AUC), Youden index, sensitivity, and specificity.

Results

The optimal BMI cutoff values for defining obesity were 24.8 kg/m² for young adult men and 23.2 kg/m² for young adult women. For mature adult men and women, the values were 25.4 kg/m² and 26.3 kg/m², respectively. These cutoff values showed the highest effectiveness in defining obesity. The optimal BMI cutoff values for underweight, normal weight, and overweight categories were: ≤ 18.0 kg/m², 18.1–20.5 kg/m², and 20.6–24.8 kg/m² for young adult men; ≤ 18.3 kg/m², 18.4–21.1 kg/m², and 21.2–23.2 kg/m² for young adult women; ≤ 16.5 kg/m², 16.6–20.8 kg/m², and 20.9–25.4 kg/m² for mature adult men; and ≤ 17.1 kg/m², 17.2–21.1 kg/m², and 21.2–26.3 kg/m² for mature adult women. The proposed BMI cutoffs performed well in identifying underweight and overweight individuals.

Conclusion

The identified BMI cutoff values for Ethiopians were lower than international standards. Adopting these country‐specific cutoffs would be more appropriate for clinical practice and research in Ethiopia.

Abbreviations

BF%

body fat percentage

BIA

bioelectrical impedance analysis

BMI

body mass index

GTIFRDC

Guna Tana Integrated Field Research and Development Center

NIH

National Institutes of Health

ROC

Receiver Operating Characteristics Curve

WHO

World Health Organization

1. Introduction

The primary premise underlying BMI guidelines is that body mass, when adjusted for squared stature, is strongly linked to body fatness as well as the resulting morbidity and mortality [1]. Since, it is challenging to assess body fat at the population level, body mass index (BMI) is a measurement alternative for body fat [2]. Body mass index (BMI) is a common metric for comparing adiposity across different heights and weights. However, it may misclassify individuals due to the varying impacts of bone density, muscle mass, and fluid on body weight [3]. To reduce the risk of overdiagnosis and underdiagnosis of obesity, excess adiposity should be verified using at least one other measurement, such as waist circumference or direct fat measurement [4].

BMI cut‐off values in Africa vary by country or region [5, 6, 7], but the WHO's classifications [2] are commonly used as a reference including in Ethiopia. The optimal BMI may vary among different populations [8]. Caucasians generally have higher BMIs than Chinese, Ethiopians, and Polynesians at the same age, sex, and body fat level [8]. Specifically, BMIs in Ethiopians are 4.6 kg/m² lower than in Caucasians, suggesting an underestimation of obesity levels among Ethiopians [9].

Moreover, BMI fails to consider the extensive diversity in body fat distribution, and may not accurately reflect the level of adiposity or the related health risks across various individuals and populations [10]. By the virtue of this, BMI's reliability as an individual‐level body fat measurement tool is questionable. Direct body fat percentage (BF%) measurements are suggested as a more reliable tool for diagnosing obesity [11]. Various techniques are used to measure body composition in clinical and research settings [12, 13, 14, 15]. However, many in vivo methods are impractical or error‐prone for epidemiological research [16, 17]. Despite their objective data, they often have drawbacks such as high costs and limited portability, making them unsuitable for large‐scale obesity assessments [18]. Each method has its limitations and assumptions [19]. Therefore, Bioelectrical Impedance Analysis (BIA) was selected for its simplicity, cost‐effectiveness, and user‐friendliness in large studies [20], proving effective in healthy individuals and those with stable hydration [21].

Despite the definitional challenges associated with BMI, it continues to be the primary measurement used in clinical settings to categorize obesity. However, there is a lack of agreement between the suggested thresholds for anthropometric adiposity indicators, such as the ideal BMI, and the most effective thresholds based on total body fat percentage [22, 23, 24]. The WHO has established cut‐off points for obesity based on BMI values [2], but these cut‐off points may not be applicable to other populations because they were determined using research on the association between BMI and morbidity and mortality in Western populations [25, 26]. Hence, it is important to note that the commonly utilized cutoff points of body mass index (BMI) for diagnosing overweight and obesity may not be universally applicable across both genders, ethnicity and different age groups [7, 27, 28, 29]. However, obesity (both preclinical and clinical) is primarily defined by BMI, but this broad classification risks overdiagnosis, leading to potential negative effects on clinical, economic, and political levels [4].

Different relationships between BMI and BF% have been found in recent studies among various ethnic groups [9, 30]. Epidemiological studies have shown that the ideal body mass index (BMI) may differ for different populations [8]. There is an increasing body of evidence that the relationship between BMI and body fat percent distribution differs across populations [31]. Differences in cut‐off points affect the estimates of obesity prevalence [10].

Different threshold levels have been employed in prior studies to define obesity, leading to a diverse range of cutoff values being utilized. Some studies relied on the World Health Organization's (WHO) guidelines (BMI: ≥ 30 kg/m²) [6, 7, 28], while others utilized locally recommended or region‐specific cutoff values specific to the Asian Pacific region, as suggested by the WHO [22, 27, 29, 32, 33, 34, 35, 36, 37, 38, 39, 40].

In spite of inherent measurement errors, BMI will persist in their use for evaluating patients in low‐resource settings [22]. As anthropometric parameters are influenced by factors such as sex, age, race/ethnicity, and geography, it is essential to establish country‐specific cutoff values for anthropometric adiposity measures [9, 22]. Evidence suggests that the universally recognized anthropometric cutoffs utilized to identify obesity may not be suitable for Ethiopians [41]. Besides, the utilization of WHO thresholds for defining obesity has been subject to limitations and criticisms [3, 42]. As a result, the use of a universal BMI cutoff point for obesity is not considered appropriate [43], emphasizing the need for population‐specific cutoff values that can vary among different populations or ethnic origins.

Each population should identify appropriate anthropometric measurement values for disease risk screening [44], with cut‐off points tailored to different ethnic groups based on reliable evidence [43]. Establishing specific reference values for the Ethiopian population is essential for accurately defining obesity. However, there is insufficient community‐level data in Ethiopia [44]. Thus, this study aims to determine body composition‐based BMI cut‐offs for healthy adults in the North‐West region of Ethiopia.

2. Methods

2.1. Study Area and Period

The study area was near the highest point of Guna Mountain (4231 m altitude) and within a 60–70 km radius of lowland areas. This is unique as few places worldwide undergo such significant environmental changes in a short distance, especially in densely populated areas [45]. The study took place at the Guna Tana Integrated Field Research and Development Center (GTIFRDC) established by Debre Tabor University in Northwest Ethiopia. The center is surrounded by Guna Mountain to the east, Lake Tana to the West, Ribb River to the North, and Gumara River to the South. The catchment area includes three ecological zones: “woorch” (above 3500 m), “Dega” (2500–3500 m), and “Woyna‐Dega” (1500–2500 m above sea level) [46]. The study was conducted from June to August 2023 at the Guna Tana Integrated Field Research and Development Center in the South Gondar zone of the Amhara region in Northwestern Ethiopia.

2.2. Study Design, Population and Sample Size

A population‐based cross‐sectional study was conducted. The study included apparently healthy adults of both sexes, aged between 18 and 64 years, consisting of young adults from 18 to 45 years and mature adults over 45 years old. Participants who had been residing in the study area for at least 1 year were included in this study. However, individuals with physical disabilities such as kyphosis or scoliosis, limb deformities preventing them from standing upright, and pregnant women were excluded.

2.3. Sample Size Determination

To determine the appropriate sample size for establishing cut‐off values for body mass index among apparently healthy adults, several factors were considered: the sensitivity estimation formula [47], a 33.3% prevalence rate of central obesity in adults in Southwest Ethiopia [48], a 5% margin of error, a 95% confidence level, and an anticipated sensitivity (SN) of 90%.

Hence, the minimum estimated sample size was determined to be 415 using the sample size formula for sensitivity analysis. After accounting for a 10% nonresponse rate, the adjusted required sample size was 457. However, to improve precision and address another objective of a larger study to establish laboratory reference intervals for Ethiopian adults, anthropometric data were collected from a total of 838 adults.

2.4. Sampling Procedures

Once the total household size of each kebele was determined or accessed from the local woreda administration, we allocated the sample size for each kebele proportionally to their sizes. The study participants were then drawn in proportion to the population size in each of the 10 representative kebeles (the smallest administrative unit) of GTIFRDC. A systematic random sampling technique was employed to select individual households. We then sampled the eligible individuals in those selected households. However, if more than one eligible subject was available within a selected household, a lottery method was used to select one individual per household.

2.5. Study Variables

The dependent variable is body mass index (BMI), while socio‐demographic characteristics, behavioral conditions (alcohol consumption, exercises or physical activities and smoking), and participants' diet and nutrition history serve as explanatory variables. Alcohol Consumption: Refers to any local or standard drink consumed by participants in the 48 h prior to the study. The amount of alcohol was measured in bottles for beer, glasses for wine, and glasses for local drinks Tela, Araki and Teji. All participants were non‐smokers.

2.6. Data Collection Tools and Procedures

Data from eligible participants were collected through a structured questionnaire utilizing the face‐to‐face interviews and anthropometric measurements. Height was measured using a portable and adjustable wooden stadiometer with an adjustable headpiece. Weight was measured using a digital and portable SECA scale from Germany. The measurements were conducted in accordance with the guidelines outlined in the WHO STEPS protocol (Part 3: data collection manual) [49] and the National Health and Nutrition Examination Survey (NHANES): Anthropometry Procedure Manual; CDC, 2009 [50]. Body mass index (BMI) was calculated using the formula BMI = body weight (kg)/height^2 (m^2). Body fat percentage measurements were performed using the Omron Body Composition Monitor and Scale, which actually utilizes bioelectrical impedance analysis (BIA) techniques. Before participants used the body composition measurement devices, we had to input their age, sex, and height. Once this information was entered, participants could step onto the devices, which would then produce outputs such as weight, BMI, body fat percentage, visceral fat level, skeletal muscle percentage, resting metabolism, and body age. Finally, we recorded all this information in the questionnaire.

2.7. Standardization of the Anthropometric Measurements' Procedures

Procedures were standardized in both English and the local language Amharic for accurately measuring weight and height. Detailed step‐by‐step instructions and procedures were prepared for each anthropometric measurement. Accordingly, before starting to measure the height, ask the participant to remove footwear and headgear. Instruct the participant to stand on the board, feet together, heels against the back board, and knees straight. Ensure the participant's back of head, shoulders, buttocks, and heels make contact with the backboard. Instruct the participant to look straight ahead, not tilting their head. Gently move the measuring arm down onto the participant's head and ask them to stand as tall as possible, take a deep breath, and hold the position. Finally, record the height in centimeters. Likewise, in order to measure the weight, prepare a portable digital weighing scale and a stiff wooden board to place under the scales. Ensure the scales are on a firm, flat surface. Instruct the participant to remove heavy or thick dressings (light indoor clothing was allowed), footwear and socks. Ask the participant to step onto the scale with one foot on each side, and stand still, facing forward, with arms at the sides. Finally, record the weight in kilograms.

Data collectors and supervisors were responsible for ensuring strict adherence to these measurement instructions to maintain accuracy and consistency. The weight and height measurements were rounded to the nearest 0.1 kg and 0.1 cm, respectively.

2.8. Data Quality Control and Calibration of Anthropometry Measuring Devices

To ensure accurate anthropometric measurements, trained personnel such as nurses, health officers, and health extension workers conducted the data collection. A 2‐day training session was conducted for both data collectors and supervisors, including individuals with expertise in public health and biostatistics. The measurement tools were pretested before the actual data collection to ensure their effectiveness. Regular checks of data completeness and quality were carried out by supervisors and principal investigators with necessary adjustments made as needed.

To maintain accuracy, the weight scale was calibrated daily at the start and end of each session using two sealed 2 kg bottled water, which provided precise weight measurements ranging from 3.85 to 4.15 kg. Similarly, the stadiometer was calibrated weekly using an 80 cm rod.

2.9. Body Composition Measurement

In the current study, body fat percentage was evaluated using the Omron body composition monitor (Omron model HBF‐214, Omron Healthcare, Kyoto, Japan) in accordance with the manufacturer's instructions [51]. The Omron Body Composition Monitor and Scale utilize the bioelectrical impedance analysis (BIA) method to estimate various body measurements, including body weight, body mass index (BMI), body fat percentage, skeletal muscle percentage, visceral fat level, body age, and resting metabolism [51]. BIA is a safe, rapid, noninvasive, and cost‐effective method for evaluating body composition [12]. It estimates total body water (TBW) by passing an electrical current through body segments and predicts body fat (BF) and fat‐free mass (FFM). Participants followed specific guidelines before the body composition assessment. They abstained from alcohol, pharmacologically active substances, and intense exercise for set time periods. Measurements were taken in the morning after fasting overnight, emptying the bladder, and resting for 15 min. The device was cleaned and operated according to the manufacturer's instructions.

2.10. Data Processing and Analysis

Data were entered into Epi info 7.2.5.0 software and exported to Stata 17.0 and MedCalc v 20.215 software for processing and analysis. Descriptive statistics, such as frequency, percentage, mean, and corresponding confidence intervals, were reported. An independent sample t‐test was used to compare anthropometric measurements between men and women.

The study utilized the World Health Organization (WHO) criteria as the gold standard for obesity classification, with body fat percentages exceeding 25% for males and 35% for females (18–45 years old) [3, 52]. For individuals over 45 years old, obesity was defined as a body fat percentage of 40% for women and 28% for men [1]. These thresholds were used in Receiver Operating Characteristic (ROC) curve analyses to determine anthropometric cut‐off values.

BMI cut‐off values for defining the normal range of body fat percentage were determined based on the body fat percentage ranges proposed by Gallagher et al. (2000) [1]. For males, the range of 8%–20% was used, while for females, the range of 21%–33% was considered. Additionally, overweight cutoff values extended from the upper value of the normal body fat percentage (20% for males and 33% for females) up to the lower body fat percentage cutoffs defined by theWHO [3] (25% for young men and 35% for young women). Normal weight ranges for mature adults were defined based on body fat percentages of 11%–22% for men and 23%–34% for women. Overweight mature adults were defined as having body fat percentages of 22%–28% for men and 34%–40% for women. Body fat percentages below 11% for men and 23% for women were used to define underweight mature adults [1]. To assess the performance of the proposed BMI cut‐off values, the Area Under the Curve (AUC), sensitivity, specificity, and Youden's index values were calculated for both the normal ranges and obesity classification [53].

2.11. Operational Definitions [54, 55]

• Intense Exercise: Refers to vigorous‐intensity activities that significantly elevate heart rate and breathing, requiring a high level of effort, such as athletics and other competitive sports.

• Physical Training: A structured regimen designed to improve physical fitness and performance over time. This includes various types of training, such as strength and aerobic training.

3. Results

3.1. Sociodemographic Characteristics of the Respondents

Data were successfully collected from 838 samples. The respondents' mean (±) age were 39.74 ± 0.75 and 39.46 ± 0.53 for men and women, respectively, with an age range of 18–64 years. Among the respondents, 522 (62.29%) were females. The majority of respondents, identified as Orthodox Christians 795 (94.87%) and Amhara 838 (99.64%) in terms of their ethnic origin. Around one‐third (36.52%) had not received formal education, while 31.26% had attended college or university. The majority of participants (68.62%) were married, and 18.62% were single (Table 1).

TABLE 1.

Sociodemographic characteristics of the study participants, GTIFRDC, 2023.

Variables	Frequency (n)	Percentage (%)
Age (years)	M: 39.74 ± 0.75
(Mean ± sd)	W: 39.46 ± 0.52
Sex
Male Female	316 522	37.71 62.29
Religion
Orthodox Muslim Protestant Others	795 34 1 8	94.87 4.06 0.12 0.95
Ethnicity
Amhara Tigre Oromo	835 2 1	99.64 0.24 0.12
Marital status
Single Married Divorced/separated Widowed Cohabit	156 575 59 47 1	18.62 68.62 7.04 5.61 0.12
Educational status
No formal education Primary school Secondary & preparatory school College/university and above Refuse	306 138 130 262 2	36.52 16.47 15.51 31.26 0.24
Occupation
Gov't employee NGO employee Private company employee Private skill workers Farmer Merchant Students Housewife Others	156 7 5 61 188 66 82 222 51	18.62 0.84 0.60 7.28 22.43 7.88 9.79 26.49 6.09
Annual income
≤ 12,000 Birr 12,000–18,000 Birr 18,000–23,000 Birr 23,000–30,000 Birr > 30,000 Birr Don't know Refuse	264 132 98 100 172 60 12	31.50 15.75 11.69 11.93 20.53 7.16 1.43

Abbreviations: GTIFRDC, Guna Tana integrated field research and development center; M, men; W, women; Birr (1 Birr = 0.0177$ at time of data collection).

3.2. Behavioral Related Characteristics of the Respondents

A significant majority of respondents 568 (67.78%) reported not consuming alcohol within the 48 h prior to data collection. However, nearly a third of them (32.2% out of 838) had consumed alcohol within 24–48 h prior to data collection. Additionally, a vast majority (94.99%) did not engage in regular physical exercise. Over half of the respondents (51.1%) did not participate in moderate physical exercises lasting at least 10 min, and a similar percentage (64.7%) did not have work involving vigorous activity. However, a notable portion of participants (35.3%) were engaged in work requiring vigorous physical activity. Conversely, a considerable number of respondents (48.9%) reported engaging in moderate exercises for at least 10 min (Table 2).

TABLE 2.

Behavioral related characteristics of the study participants, GTIFRDC, 2023.

Variables	Frequency (n)	Percentage (%)
Drunk alcohol in past 48 h
No Yes a	568 270	67.78 32.22
Regular exercise
No Yes	796 42	94.99 5.01
Do you do moderate physical exercise for at least 10 min
No Yes	428 410	51.1 48.9
Does your work involve vigorous activity
No Yes	542 296	64.7 35.3

Abbreviation: GTIFRDC, Guna Tana integrated field research and development center.

^a Alcohol drinks include: local alcohols like Tella, Araki, Gin, wine & beer.

3.3. Food and Nutrition History of the Respondents

Regarding the food intake and nutrition history of the respondents, the majority relied on teff‐based Injera (44.39%) and Injera made of mixed grains (39.86%) as their staple diets. A significant number of participants (88.78%) reported eating three meals a day, and almost all of them (98.57%) followed a non‐vegetarian diet. Beef (61.46%) and goat/sheep (34.49%) were the preferred sources of meat consumption. Sunflower oil (68.38%) and palm oil (27.68%) were commonly used for cooking among the respondents (Table 3).

TABLE 3.

Food and nutrition history of the study participants, GTIFRDC, 2023.

Variables	Frequency (n)	Percentage (%)
Stable diet
Injera made of teff a  Injera made of wheat and barely  Injera made of mixed grains b Rice	372 116 334 16	44.39 13.84 39.86 1.91
Dishes per day/frequency of eating
Twice per day Three times per day Four times per day	73 744 21	8.71 88.78 2.51
Are you vegetarian
No Yes	826 12	98.57 1.43
If non vegetarian, types of meat commonly eat/consume
Beef Goat/sheep Chicken Others	515 289 25 9	61.46 34.49 2.98 1.07
Types of cooking oil
Vegetable oil Sunflower oil Palm oil Butter Margarine/coconut Non in particular No oil uses Others	2 573 232 1 4 2 3 21	0.24 68.38 27.68 0.12 0.48 0.24 0.36 2.51

Abbreviation: GTIFRDC, Guna Tana integrated field research and development center.

^a Injera is a local pancake; Teff is a tiny grain scientifically named “Eragrostis tef”.

^b Injera made of mixed grains contains (teff, wheat, sorghum/maize, rice, millet).

3.4. The Prevalence of Overweight and Obesity by Various Criteria

The proposed BMI cut‐off indicates that the prevalence of overweight and obesity among young adult men is 41 (21.2%) and 14 (7.3%), respectively. For young adult women, these figures are 69 (19.8%) for overweight and 84 (24.1%) for obesity. Using the WHO cut‐off of ≥30 kg/m², obesity prevalence drops to 1 (0.5%) among young men and 13 (3.7%) among young women. The newly proposed BMI cut‐off for obesity in mature adults revealed a prevalence of obesity of 12 (10.3%) in men and 18 (10.4%) in women. However, the WHO cut‐off shows obesity prevalence at 0 (0%) for men and 7 (4.04%) for women. Based on body fat percentage, 18 (9.3%) young men and 59 (16.9%) young women are categorized as obese, while for mature adults, 10 (8.5%) men and 11 (6.4%) women are identified as obese (Table 4).

TABLE 4.

Prevalence of overweight and obesity status by various criteria among study participants, GTIFRDC, 2023.

Men (18–45 years) (n = 193)			Women (18–45 years) (n = 348)
Criteria	Weight category	Prevalence n (%)	Criteria	Weight category	Prevalence n (%)
BMI cut off by WHO a			BMI cut off by WHO a :
25.0–29.9 ≥ 30	Overweight Obese	12 (6.2) 1 (0.5%)	25.0–29.9 ≥ 30	Overweight Obese	38 (10.9%) 13 (3.7%)
Local BMI cut off b			Local BMI cut off b :
20.6–24.8 > 24.8	Overweight Obese	41 (21.2%) 14 (7.3%)	21.2–23.2 > 23.2	Overweight Obese	69 (19.8%) 84 (24.1%)
Body fat% c			Body fat% c
20%–25% > 25%	Overweight Obese	27 (13.9%) 18 (9.3%)	33%–35% > 35%	Overweight Obese	36 (10.3%) 59 (16.9%)

Men (> 45 years) (n = 117)			Women (> 45 years) (n = 173)
BMI cut off by WHO a			BMI cut off by WHO a :
25.0–29.9 ≥ 30	Overweight Obese	12 (10.3) 0 (0%)	25.0–29.9 ≥ 30	Overweight Obese	19 (10.9) 7 (4.04%)
Local BMI cut off b			Local BMI cut off b :
20.9–25.4 > 25.4	Overweight Obese	38 (32.5%) 12 (10.3%)	21.2–26.3 > 26.3	Overweight Obese	57 (32.9%) 18 (10.4%)
Body fat% c			Body fat% c
22%–28% > 28%	Overweight Obese	44 (37.6%) 10 (8.5%)	34%–40% > 40%	Overweight Obese	51 (29.5%) 11 (6.4%)

Abbreviations: %, percentage; GTIFRDC, Guna Tana integrated field research and development center; n, frequency.

^a WHO classification, 1995 & 2008.

^b The BMI cuts off proposed in the current study.

^c Body fat percentage category according to Gallagher et al. (2000).

3.5. BMI Cut‐Offs for Obesity Based on Body Fat Percentage in Adults

For young men, the BMI cut‐off was > 24.8 kg/m², and for young women, it was > 23.2 kg/m². The overall group had a cut‐off of > 22.9 kg/m². Sensitivity and specificity values were calculated: men had a sensitivity of 66.6% and specificity of 98.8%, while women had a sensitivity of 96.6% and specificity of 90.6% at the proposed cut off values. The Youden index yielded values of 0.66 for men and 0.88 for women, indicating moderate and high performance, respectively (Table 5). Additionally, AUC‐ROC analysis evaluated BMI's ability to define obesity using body fat percentage thresholds of 25% for men and 35% for women. The AUC values were 0.88 (p < 0.001) for men and 0.96 (p < 0.001) for women, indicating significant performance in distinguishing obesity status. These findings are visually presented in (Table 5 and Figure 1).

TABLE 5.

Derivation of BMI cut‐offs for obesity defined by body fat percentage among study participants (young and matured adults), GTIFRDC, 2023.

Sex/bf%	BMI cut off	Youden index (se)	AUC (se)	Sensitivity (95% CI)	Specificity (95% CI)	Lr+ (95% CI)	Lr− (95% CI)
Young adults (18–45 years)
Men a > 25%	> 24.8	0.66 (0.11)	0.88 (0.04) p < 0.0001*	66.67 (41.0–86.7)	99.43 (96.9–100.0)	116.67 (16.08–846.26)	0.34 (0.17–0.64)
Women a > 35%	> 23.2	0.89 (0.02)	0.96 (0.01) p < 0.0001*	96.61 (88.3–99.6)	92.04 (88.3–94.9)	12.4 (8.18–18.02)	0.03 (0.01–0.14)
Both M & W	> 22.9	0.83 (0.03)	0.94 (0.02) p < 0.0001*	92.21 (83.8–97.1)	90.52 (87.5–93.0)	9.72 (7.29–12.98)	0.08 (0.04–0.19)
Matured adults (> 45 years)
Men b > 28%	> 25.4	0.80	0.89 (0.07) p < 0.0001*	81.82 (48.2–97.7)	98.11 (93.4–99.8)	43.36 (10.69–175.9)	0.19 (0.05–0.65)
Women b > 40%	> 26.3	0.88	0.96 (0.03) p < 0.0001*	91.67 (61.5–99.8)	96.27 (92.1–98.6)	24.60 (11.01–54.93)	0.08 (0.01–0.57)
Both M & W	> 25.6	0.82	0.93 (0.03) p < 0.0001*	86.96 (66.4–97.2)	94.76 (91.4–97.1)	16.58 (9.72–28.29)	0.14 (0.05–0.40)

Note: The bold value indicates a statistically significant difference in the area under the curve (AUC) values. An AUC of 0.5 suggests no differences.

Abbreviations: M, men; W, women.

^a Obesity cuts‐off were generated from the ROC analyses using WHO cuts‐off for obesity based on body fat percent > 25% for men and > 35% for women as binary classifier [3, 52].

^b Obesity cuts‐off were generated from the ROC analyses using Gallagher etal. (2000) [1] cuts‐off for obesity based on body fat percent > 28% for men and > 40% for women as binary classifier.

*p‐statistically significance level (AUC = 0.5, where AUC = area under the ROC curve).

FIGURE 1.

AUC for obesity definition based on body fat percentage thresholds (25% for men, 35% for women, young adults).

Furthermore, the optimal BMI cut‐off values for defining obesity in mature adults were also determined. For men, the cut‐off was > 25.4 kg/m², and for women, it was > 26.3 kg/m². The overall group had a cut‐off of > 25.6 kg/m². Sensitivity and specificity values were calculated: men had a sensitivity of 81.82% and specificity of 98.11%, while women had a sensitivity of 91.7% and specificity of 96.3%. The Youden index was used to assess BMI's performance in defining obesity based on the cut‐off points, resulting in an index value of 0.88 for both men and women.

Likewise, AUC‐ROC analysis evaluated BMI's ability to define obesity using body fat percentage thresholds of 28% for men and 40% for women. The AUC values were 0.89 (p < 0.001) for men and 0.96 (p < 0.001) for women, indicating a significant performance in distinguishing obesity status. Likewise, the likelihood ratio positive and negative values were also greater than 10 and 0.1, respectively, indicating good performance of the proposed cut off values. These findings are visually presented in (Table 5 & Figure 2).

FIGURE 2.

AUC for obesity definition based on body fat percentage thresholds (28% for men, 40% for women, mature adults).

3.6. BMI Cut off Values for Underweight, Normal, and Overweight Categories

The BMI cutoff values for underweight, normal weight, and overweight young adult men were ≤ 18.0 kg/m², 18.1–20.5 kg/m², and 20.6–24.8 kg/m², respectively. For women in the same age group, these values were ≤ 18.3 kg/m², 18.4–21.1 kg/m², and 21.2–23.2 kg/m², respectively. Similarly, for mature men and women, the BMI cutoff values for underweight, normal weight, and overweight categories were ≤ 16.5 kg/m², 16.6–20.8 kg/m², and 20.9–25.4 kg/m² for men, and ≤ 17.1 kg/m², 17.2–21.1 kg/m², and 21.2–26.3 kg/m² for women, respectively (Table 6).

TABLE 6.

Performance of BMI cut off values for underweight, normal, and overweight based on body fat percentage in study participants, disaggregated by sex and age, GTIFRDC, 2023.

Sex & body weight category based on bf% a , b	18–45 years
	n	BMI cut off	Youden index	AUC (se)	SN (95% CI)	SP (95% CI)	Lr+ (95% CI)	Lr− (95% CI)
Men
Low (< 8%)	13	≤ 18	0.79	0.94 (0.02)	100.0 (75.3–100)	78.9 (72.2–84.6)	4.74 (3.57–6.28)	0.00
Normal (8%–20%)	135	≤ 20.5	0.56	0.74 (0.05)	88.2 (81.5–93.1)	67.2 (53.7–79.0)	2.69 (1.85–3.91)	0.18 (0.11–0.3)
Overweight (20%–25%)	27	> 20.5	0.75	0.88 (0.03)	92.6 (75.7–99.1)	81.9 (75.2–87.5)	5.12 (3.6–7.2)	0.09 (0.02–0.3)
Obese (> 25%)	18	> 24.8	0.66	0.89 (0.05)	66.7 (41.0–86.7)	99.4 (96.9–100)	116.7 (16.1–846)	0.3 (0.2–0.6)
Women
Low (< 21%)	22	≤ 18.3	0.67	0.89 (0.04)	81.8 (59.7–94.8)	84.9 (80.6–88.7)	5.4 (3.93–7.53)	0.21 (0.08–0.5)
Normal (21%–33%)	231	≤ 21.1	0.56	0.81 (0.03)	74.9 (68.8–80.3)	81.2 (72.9–87.8)	3.98 (2.7–5.85)	0.3 (0.24–0.4)
Overweight (33%–35%)	36	> 21.1	0.63	0.77 (0.02)	100.0 (90.3–100)	62.5 (56.9–67.9)	2.7 (2.3–3.1)	0.00
Obese (> 35%)	59	> 23.2	0.89	0.97 (0.01)	96.6 (88.1–99.6)	92.0 (88.3–94.9)	12.14 (8.18–18.0)	0.04 (0.01–0.1)

		> 45 years
Men
Low (< 11%)	5	≤ 16.5	0.98	0.98 (0.01)	100 (47.8–100)	98.2 (93.7–99.8)	56.0 (14.2–221.2)	0.00
Normal (11%–22%)	58	≤ 20.8	0.71	0.83 (0.05)	93.1 (83.3–98.1)	77.9 (65.3–87.7)	4.2 (2.6–6.9)	0.1 (0.03–0.2)
Overweight (22%–28%)	44	> 20.8	0.68	0.80 (0.05)	86.0 (72.1–94.7)	82.4 (71.8–90.3)	4.9 (2.9–8.1)	0.2 (0.1–0.4)
Obese (> 28%)	10	> 25.4	0.80	0.89 (0.07)	81.8 (48.2–97.7)	98.1 (93.4–99.8)	43.4 (10.7–175.9)	0.2 (0.05–0.7)
Women
Low (< 23%)	15	≤ 17.1	0.71	0.89 (0.06)	80.0 (51.9–95.7)	91.1 (85.6–95.1)	9.03 (5.2–15.8)	0.2 (0.1–0.6)
Normal (23%–34%)	96	≤ 21.4	0.67	0.81 (0.04)	91.7 (84.2–96.3)	75.3 (64.2–84.4)	3.7 (2.5–5.5)	0.1 (0.05–0.2)
Overweight (34%–40%)	51	> 21.1	0.77	0.87 (0.03)	98.0 (89.4–99.9)	78.9 (70.6–85.7)	4.6 (3.3–6.5)	0.02 (0.01–0.2)
Obese (> 40%)	11	> 26.3	0.88	0.96 (0.03)	91.7 (61.5–99.8)	96.3 (92.1–98.6)	24.6 (11.0–54.9)	0.08 (0.01–0.6)

Abbreviations: AUC (se), area under the ROC curve with its standard error; Lr−, negative likelihood ratio; Lr+, positive likelihood ratio; n, frequency; SN, sensitivity; SP, specificity.

^a Body fat percentage (bf%) category according to Gallagher et al. (2000) [1].

^b The normal and overweight BMI cut‐off was determined as a bf % range between low/underweight & overweight, and between overweight and obese.

The performance of the newly proposed BMI cutoffs in identifying underweight, normal weight, and overweight individuals was evaluated. The results showed that the Youden index values and area under the curve (AUC) values ranged from 0.75 to 0.79 in young men and 0.63 to 0.67 in young women for Youden index, and 0.88 to 0.94 in young men and 0.77 to 0.89 in young women for AUC. Correspondingly, the Youden index and area under the curve values for underweight, normal weight, and overweight mature men ranged from 0.68 to 0.98 for the Youden index and from 0.80 to 0.98 for the area under the curve. For mature women, these values ranged from 0.71 to 0.77 for the Youden index and from 0.87 to 0.89 for the area under the curve (Table 6).

Additionally, the sensitivity and specificity values varied from 92.6% to 100% and 78.9%–81.9% in young men, and from 81.8% to 100% for sensitivity and 62.5%–84.9% for specificity in young women. Among mature adults, the sensitivity and specificity of the BMI cutoffs for underweight, normal weight, and overweight categories ranged from 86% to 100% and 82.4%–98.2% in men, while in women, the sensitivity ranged from 80% to 98% and specificity ranged from 78.9% to 91.1% (Table 6).

4. Discussion

In the current study, the optimal BMI cut‐off values for defining obesity in young adult men and women were found to be 24.8 kg/m² and 23.2 kg/m², respectively. For mature adult men and women, these values were 25.4 kg/m² and 26.3 kg/m², respectively. The lower optimal BMI cut‐off points for obesity in young men and women align with previous studies conducted in South West Ethiopia [32], Indonesia [29], China [56], India [35], and Sri Lanka [33]. In these studies, various BMI cut‐off values were reported, including 22.2 kg/m² for males and 24.5 kg/m² for females [32], 21.9 kg/m² for males and 23.6 kg/m² for females [29], 22.5 kg/m² for males and 23.5 kg/m² for females [56], 22.0 kg/m² for both sexes [35], and 22.9 kg/m² for men and 23.3 kg/m² for women [33].

However, the BMI cut‐offs of 25.4 kg/m² and 26.3 kg/m² used to define obesity for mature adult men and women in the current study are relatively higher than those reported in other previous studies [29, 32, 35, 56]. Interestingly, these values align with the BMI cut‐offs reported for Asians [57] and in Sri Lanka [33], where obesity was defined based on body fat percentage thresholds of 30% for men and 40% for women.

The higher BMI cut‐off of 24.5 kg/m² for women in a previous study in South West Ethiopia [32] and the 24.8 kg/m² cut‐off for young men in the current study may result from differences in study populations. The earlier study focused on university staff, who likely have distinct nutritional habits and resource access compared to rural participants, many of whom are Orthodox Christians observing fasting that affects body fat levels. Additionally, the current study analyzed BMI cut‐off values across age categories, acknowledging that body fat percentage typically increases with age and is generally higher in women than in men. However, these differences may not be fully captured by BMI alone [3].

The discrepancy in BMI cut‐off values across studies may also be influenced by differences in body composition assessment methods [29, 32, 35], differences in the study populations [32], variations in the body fat thresholds used to define obesity [29, 35], as well as differences in body composition and ethnic diversity [7, 29, 35].

The BMI cut‐off values for obesity in the current study were 5 kg/m² and 7 kg/m² lower than those set by theWHO [3] and NIH [58]. This difference may be due to the distinct body shape of Ethiopians, who typically have shorter stature and leaner body structures. Additionally, while the WHO classification is age and sex‐independent, this study used age and sex‐specific body fat percentage thresholds proposed by Gallagher et al. (2000) for both young and mature adults [1].

The BMI cut‐off for obesity in this study suggests that the current WHO [3] threshold (≥ 30 kg/m²) underestimates obesity prevalence and associated health risks. Furthermore, Ethiopians of the same age, gender, and body fat levels have BMIs that are 4.6 kg/m² lower than those of Caucasians, indicating a potential underestimation of obesity levels among Ethiopians [9]. The proposed cut‐off values for overweight (≥ 23.0 kg/m²) and obesity (≥ 25.0 kg/m²) for Asians, as defined by WHO experts [16, 57], align with the current study's findings. The BMI cut‐off values for overweight in this study ranged from 20.6 to 24.8 kg/m² for young men, 21.2–23.2 kg/m² for young women, 20.9–25.4 kg/m² for mature men, and 21.2–26.3 kg/m² for mature women.

The BMI cut‐off values for overweight in Ethiopians vary by gender, being 4–5 units lower for young men and 4–7 units lower for young women compared to WHO classifications [3]. A nearly 4‐unit difference for mature adults also exists. These disparities may stem from the WHO's cut‐off points [3], which are based primarily on studies from Europe and the USA, focusing on the relationship between BMI and health outcomes. This study established cut‐off values using body fat thresholds recommended by WHO [3, 52] and Gallagher et al. (2000) [1]. It aligns with findings that Caucasians have higher BMIs than Chinese, Ethiopians, and Polynesians of the same age, sex, and body fat [8], suggesting that BMI may not accurately reflect body fat across different ethnic groups [31]. Considering that, the WHO [10] has proposed lower cut‐off points for overweight/obesity specifically for Asian and Pacific populations (BMI ≥ 25 kg/m²) to promote healthy lifestyles and weight control.

In this study, the BMI cut‐off for overweight in most groups closely aligns with the WHO's range of 25–29.9 kg/m² [3], indicating that these cut‐off ranges accommodate similar weight fluctuations. However, such wide ranges can lead to weight cycling, which is linked to increased morbidity and mortality. The WHO's broad normal weight range (BMI: 18.5–25 kg/m²) [3] should not imply that individuals can fluctuate freely within it without risk. This study found narrower normal weight BMI ranges, with differences of 2.4–4.2 units for young and mature men and 2.7–4 units for young and mature women. These findings highlight the importance of minimizing weight fluctuations to reduce health risks.

The BMI cut‐off values for underweight in the current study were found to be ≤ 18.0 kg/m² for young men and ≤ 18.3 kg/m² for young women, aligning with WHO standards [3]. However, the cut‐off values for mature adults were lower, at ≤ 16.5 kg/m² for men and ≤ 17.1 kg/m² for women, which differ from other studies [1, 6]. These findings correspond with WHO recommendations for severe underweight (≤ 16.0 kg/m²) and moderate underweight (17.0–18.49 kg/m²) [3, 16], as well as the BMI range for mild to moderate malnutrition reported in Southwest Ethiopia [32]. The lower BMI values observed in Ethiopians might be attributed to their slender body build, characterized by a predominantly small to medium size and slim physique with less prominent abdominal region [32]. Indeed, all the proposed BMI cuts off values were lowered than the WHO recommended cut off values [3]. Nonetheless, the body fat percentage of Ethiopians was found to be 10% higher than that of Caucasians when considering individuals of the same age, sex, and BMI level [9]. At the same body mass index (BMI), a person with a slender build will have a higher percentage of body fat compared to someone with a more muscular or stockier build [9].

The newly proposed BMI cut‐off values for obesity were more effective in women than in men. For mature women, a cut‐off of > 26.3 kg/m² yielded a Youden index of 0.88, AUC of 0.96, sensitivity of 91.7%, and specificity of 96.3%. In mature men, the corresponding values were 0.80, 0.89, 81.8%, and 98.1%. In young adults, men had a Youden index of 0.66, AUC of 0.89, sensitivity of 66.7%, and specificity of 99.4%, while women had 0.89, 0.97, 96.6%, and 92.0%. These findings align with previous studies from South West Ethiopia [32], Indonesia [29], India [35], and Sri Lanka [22, 33], which reported varying BMI cut‐off values. For males, cut‐offs ranged from 21.9 kg/m² to 22.2 kg/m², with AUC values of 0.88–0.92 and sensitivities of 83.9%–88%. For females, reported cut‐offs were between 23.6 kg/m² and 24.5 kg/m², with AUC values of 0.87–0.95 and sensitivities from 67.6% to 80%. In India, a cut‐off of 22.0 kg/m² showed 80% sensitivity and 70% specificity for both genders. A Sri Lankan study indicated that a body fat percentage of 33% corresponded to a BMI of 24.5 kg/m², while 35% body fat aligned with a cut‐off of 25.0 kg/m², demonstrating similar performance metrics.

The current findings have important implications for understanding weight‐related health conditions. The Lancet Diabetes & Endocrinology Commission [4] has highlighted that the existing definition of obesity lacks sufficient sensitivity and specificity for clinical use, raising concerns about treating obesity as a standalone disease. Disparities between our study and WHO [3] classifications suggest a need for revising the criteria for overweight and obesity. As a consequence, further research is essential to ensure accurate categorization and validate the newly proposed BMI cut‐off points, emphasizing the importance of ethnic‐specific values for assessing overweight and obesity among Ethiopians.

Our study focused on physically active individuals, primarily farmers, which may have led to lower fat mass and higher lean mass per BMI. Notably, most Ethiopian adults live in rural areas, as indicated by the national demographic and health survey [59]. Therefore, the BMI‐fat relationships observed in our study are based on a population sample that accurately reflects the physically active and mostly rural residents of Ethiopians.

This study has several strengths, including a large sample size that enhances the applicability of results. As a community‐based study, it offers unique insights into establishing anthropometric cut‐off values in Ethiopia, with diverse urban and rural representation. The use of bioelectrical impedance for body composition assessment adds practicality, while standard ROC techniques strengthen the determination of cut‐off values. However, the assessment method may be less reliable and population‐specific than advanced four‐compartment methods [16, 17], which can be costly and impractical for widespread obesity assessments [18]. Cross‐sectional studies capture data at a single time point, potentially missing changes in BMI trends or health status, as BMI can fluctuate seasonally. Selection bias may also affect the reliability of the cut‐off values, despite rigorous precautions like training for data collectors and standardized procedures. Although we do not expect significant differences among Ethiopia's ethnic groups, not all categories were included.

5. Conclusion

The identified BMI cutoff values were found to be lower than the internationally recommended threshold standards. The existing WHO‐recommended BMI cutoff value of 30 kg/m² has resulted in underestimation and possibly misclassification of individuals. Furthermore, to prevent the misclassification of adiposity status, it is important to consider age, sex and context when utilizing and deriving the BMI cut‐off points for categorizing overweight and obesity. Therefore, the application of country‐specific BMI cutoff values, as proposed in this study, would be more suitable for future clinical practice. Furthermore, it is recommended to conduct a comprehensive nationwide study to establish national‐level BMI cut‐off values.

5.1. Units of Measurement

Weight was measured in kilograms (kg), height was measured in centimeters (cm), and body mass index (BMI) was calculated as weight in kilograms divided by height in meters squared (kg/m²).

Author Contributions

M.A., A.T., A.A. and A.W. conceptualized the study. M.A., A.T., A.A. and A.W. analyzed and wrote the first draft of the manuscript. All authors commented on the design of the study and reviewed the final draft of the manuscript. All authors have read and approved the final manuscript.

Ethics Statement

The studies involving human participants adhered to the ethical standards set by the institutional review board of the College of Health Science at Addis Ababa University (Approval number: 090/22/SPH) and followed the principles outlined in the 1964 Helsinki declaration and its subsequent amendments or similar ethical standards.

Consent

Prior to their participation, all individual participants provided informed consent. Participants were informed that the information collected during the study as well as the study results may be published. We have assured them that their privacy and confidentiality will be protected, and no information that could be traced back to them will be disclosed. Participants' names and any other identifying details will be kept separate and will not be shared with any other individuals or entities.

Conflicts of Interest

The authors declare no conflicts of interest.

Data Availability Statement

The datasets generated or analyzed during the current study can be obtained from the corresponding author upon a reasonable request.