Comparison of racial/ethnic-specific BMI cutoffs for categorizing obesity severity: a multicountry prospective cohort study

Abstract

Objective:

To compare racial/ethnic-specific BMI cutoffs for three classes of obesity based on equivalent risk of type 2 diabetes (T2D).

Methods:

Participants without T2D were included from the UK Biobank, China Health and Nutrition Survey, and Singapore Chinese Health Study. Poisson regressions with restricted cubic splines were applied to determine BMI cutoffs for each non-White race/ethnicity for equivalent incidence rates of T2D at BMI of 30.0, 35.0, and 40.0 kg/m2 in White, respectively.

Results:

During a median follow-up of 13.8 years among 507,763 individuals, 5.2% developed T2D. In women, BMI cutoffs for an equivalent incidence rate of T2D as observed at 40.0 kg/m2 in White adults were 31.6 kg/m2 in Black adults, 29.2 kg/m2 in British Chinese, 27.3 kg/m2 in South Asian adults, 26.9 kg/m2 in Native Chinese, and 25.1 kg/m2 in Singapore Chinese. In men, the corresponding BMI cutoffs were 31.9 kg/m2 in Black adults, 30.6 kg/m2 in British Chinese, 29.0 kg/m2 in South Asian adults, 29.6 kg/m2 in Native Chinese, and 27.6 kg/m2 in Singapore Chinese. The racial/ethnic order was consistent when equivalent BMI cutoffs were estimated for class I and II obesity.

Conclusion:

Establishing a race/ethnicity-tailored classification of three classes of obesity is urgently needed.

Introduction

Obesity is a global public health crisis. The worldwide prevalence of obesity nearly tripled between 1975 and 2020, and more than 813 million adults aged 20 years and older were obese in 2020 [1, 2]. The prevalence of severe obesity has been increasing at a faster rate than that of obesity [3]. Heterogeneous risks of chronic diseases by different severity of obesity have been widely reported, among which type 2 diabetes is an exemplar [4–6]. Burden of type 2 diabetes is higher with greater body mass index (BMI) within the obesity range [7]. For example, the prevalence of diabetes was 16% in people with class I obesity (BMI of 30.0–34.9 kg/m²) and 37% in people with class III obesity (BMI of ≥40.0 kg/m²) in US adults in 2017–2018 [8]. Targeted prevention of diabetes and other obesity-related diseases cannot ignore severity of obesity.

Thirty years ago, the World Health Organization (WHO) proposed three classes of obesity, i.e., class I (BMI of 30.0–34.9 kg/m²), class II (BMI of 35.0–39.9 kg/m²), and class III (BMI of ≥40.0 kg/m²), but these BMI cutoffs were based primarily on evidence from studies of White populations. In fact, prevalence of metabolic diseases, body fat percentage, and body fat distribution differ markedly by race/ethnicity at the same BMI [9, 10]. It is inappropriate to use universal cutoffs. The National Institute for Health and Care Excellence (NICE) recommends BMI cutoffs of 27.5 kg/m², 32.5 kg/m², and 37.5 kg/m² for defining class I, II, and III obesity in Asians; however, these cutoffs are based on expert consensus and not supported by the literature [11, 12]. Given the increasing BMI and growing number of people with severe obesity globally as well as racial differences in body composition and fat distribution, using universal BMI cutoffs to classify severity of obesity cannot guide precision prevention and risk management of obesity-related diseases.

Significant differences in racial/ethnic-specific BMI cutoffs for obesity equivalent to a BMI of 30 kg/m² in White adults in relation to diabetes risk have been explored in several non-White groups [12–16]; however, they compared racial/ethnic groups residing in the same nation. China has the largest population of people with diabetes and a fast-growing population of people with obesity [17, 18]. Whether the relationship between BMI and risk of diabetes differs between Native Chinese and migrants such as British Chinese and Singapore Chinese is unclear. Using the UK Biobank, the China Health and Nutrition Survey (CHNS), and the Singapore Chinese Health Study (SCHS), this study aimed to compare racial/ethnic-specific BMI cutoffs for categorizing three classes of obesity in Black adults, South Asian adults, British Chinese, Native Chinese, and Singapore Chinese at the same risk of type 2 diabetes compared with White adults at BMI of 30.0, 35.0, and 40.0 kg/m², respectively.

Methods

Data sources

This study utilized data from three large prospective cohort studies from the UK, China, and Singapore. The UK Biobank is an ongoing population-based cohort study that recruited over 500,000 participants aged 37 to 73 years between 2006 and 2010 from various locations in the UK. At baseline, participants provided self-reported demographic, socioeconomic, and lifestyle information, as well as a range of physical measurements. Participants provided written informed consent to linking their records to hospital admissions and death registrations. The CHNS was a population-based longitudinal survey of ten waves (1989–2015) to investigate how social and economic transformation of Chinese society had affected the health and nutritional status of its population. Details of the CHNS design have been described elsewhere [19]. Briefly, participants were selected from 15 provinces in China using a multistage, random cluster sampling strategy. Since diabetes information was not collected in 1989–1993, data of the remaining seven waves from 1997–2015 were used. The SCHS is a population-based cohort study established between 1993 and 1998 in which investigators recruited Chinese aged 45–74 years in Singapore. All participants were southern Chinese descent. The SCHS participants were interviewed face-to-face using a structured questionnaire at the time of recruitment [20]. For this analysis, participants who were aged 20 years and above, not pregnant, without prevalent diabetes, without missing BMI or implausible weight data (weight less than 20 kg or greater than 300 kg) at baseline, and with follow-up information were included.

The UK Biobank received ethical approval from the North West Multi-Center Research Ethics Committee, Manchester, UK. The CHNS was approved by the institutional review committees of the University of North Carolina at Chapel Hill and the National Institute for Nutrition and Health, Chinese Center for Disease Control and Prevention. The institutional review board at the National University of Singapore approved the SCHS. All participants provided written informed consent.

Data collection

Demographic, socioeconomic and lifestyle information were self-reported, including age, sex (men, women), race/ethnicity, education (less than high school, high school, some college, college graduate or higher, and none of above), and smoking status (never, former, and current). In the UK Biobank, included race/ethnicity for this analysis was White (White, British, Irish, and any other white background), Black (Caribbean and African), South Asian (Indian, Pakistani, and Bangladeshi), and British Chinese. People from the CHNS were Native Chinese and the SCHS were Singapore Chinese. Body measurements at baseline were obtained by trained staff who followed standard operating procedures and used calibrated equipment in the UK Biobank and CHNS. Height was measured to the nearest 0.1 cm using a tape measure, and weight to the nearest 0.1 kg using a calibrated platform scale. In the SCHS, weight and height information were self-reported at baseline interview. Baseline BMI was calculated as weight (in kilograms) divided by height squared (in meters) and used for the present analysis.

Type 2 diabetes

Type 2 diabetes was defined using a validated algorithm in the UK Biobank, including self-reported medical history, glucose-lowering medications, and ICD-10 code E11 [21]. In the CHNS and SCHS, type 2 diabetes was defined by self-report in each interview. Participants were asked by well-trained staff whether they were diagnosed with type 2 diabetes by physicians. Validation studies conducted in the SCHS and Native Chinese populations reported a high accuracy rate of over 95% for self-reported diabetes diagnosis [22–24]. The primary outcome of this analysis was incident type 2 diabetes. Participants were censored at the time of type 2 diabetes diagnosis, death, or last follow-up time, whichever occurred first.

Statistical analysis

Men and women were analyzed separately. Baseline characteristics for White adults, Black adults, South Asian adults, British Chinese, Native Chinese, and Singapore Chinese were described as mean (standard deviation) for continuous variables and count (percentage) for categorical ones. To determine racial/ethnic-specific BMI cutoffs, Poisson regressions were used in which the relationship between BMI and type 2 diabetes incidence rate was modeled using restricted cubic splines with four knots placed at 5^th, 35^th, 65^th, 95^th percentiles. The offset variable was person-years of follow-up. An assumption that the mean value of data equaled the variance value (or so-called equidispersion) was tested. In line with previous analyses deriving BMI cutoffs for obesity, only age, race/ethnicity, and BMI were included for modeling [12, 13, 16]. Baseline year was also adjusted for because a range of time-varying variables such as socioeconomic factors, diet, lifestyle, and healthcare resources and policies may alter the relationship between race/ethnicity, BMI, and type 2 diabetes. Age was centered at the mean of the study sample. Three sensitivity analyses were conducted. First, only age, race/ethnicity, and BMI were included for modeling. Second, only participants aged 45 to 73 years were analyzed from all three cohorts, because three cohorts recruited people with different age ranges at baseline. Third, people who developed type 2 diabetes within the first two years of follow-up were excluded from analysis. From Poisson regression models, we estimated predicted incidence rates and their 95% confidence intervals (95% CIs) for the White population with BMI of 30, 35, and 40 kg/m². Then we identified the corresponding BMI values and 95% CIs for Black adults, South Asian adults, British Chinese, Native Chinese, and Singapore Chinese, respectively. Statistical analyses were performed using R 4.2.2 and SAS 9.4.

Results

A total of 507,763 individuals were included (Figure 1); 440,668 (86.8%) were White adults, 6,756 (1.3%) Black adults, 6,384 (1.3%) South Asian adults, 1,436 (0.3%) British Chinese, 14,090 (2.8%) Native Chinese, and 38,429 (7.6%) Singapore Chinese; 280,865 (55.3%) were women (Table 1). The mean BMI was the lowest in Singapore Chinese and highest in Black adults. After a median follow-up of 13.9 years (IQR 13.0–14.6) in the UK Biobank, 19,923 White participants were diagnosed with incident type 2 diabetes, 723 in Black adults, 870 in South Asian adults, and 80 in British Chinese. After a median follow-up of 9.0 years (4.0–15.0) in the CHNS, 619 Native Chinese were diagnosed with incident type 2 diabetes. After a median follow-up of 12.0 years (7.0–14.0) in the SCHS, 4,387 Singapore Chinese were diagnosed with incident type 2 diabetes. In women the incidence rate of type 2 diabetes was 2.6, 7.8, 10.1, 3.2, 4.4, and 10.5 per 1000 person-year in White adults, Black adults, South Asian adults, British Chinese, Native Chinese, and Singapore Chinese, respectively; in men the incidence rate of type 2 diabetes was 4.4, 9.1, 11.4, 5.9, 4.3, and 10.4 per 1000 person-year in White adults, Black adults, South Asian adults, British Chinese, Native Chinese, and Singapore Chinese, respectively.

Figure 1.

Flow chart of participants in the UK biobank study (UKB), the China Health and Nutrition Survey (CHNS), and the Singapore Chinese Health Study (SCHS).

Table 1.

Baseline characteristics of participants according to race/ethnicity and sex in the UK Biobank, the China Health and Nutrition Survey, and the Singapore Chinese Health Study.

	White	Black	South Asian	British Chinese	Native Chinese	Singapore Chinese
Women
N	243820	3924	3092	909	7504	21616
Age, mean (SD), y	56.5 (7.9)	51.5 (7.7)	52.7 (8.1)	52.3 (7.4)	44.8 (14.6)	54.6 (7.4)
Education, n (%)
<High school	24179 (10.1)	788 (20.7)	306 (10.6)	112 (12.9)	5587 (74.5)	8957 (41.4)
High school	70501 (29.4)	1113 (29.3)	737 (25.5)	174 (20.0)	866 (11.5)	4776 (22.1)
Some college	29311 (12.2)	253 (6.7)	259 (9.0)	79 (9.1)	472 (6.3)	643 (3.0)
College graduate or higher	75657 (31.5)	1245 (32.7)	1084 (37.5)	406 (46.8)	447 (6.0)	160 (0.7)
None of the above	40306 (16.8)	405 (10.6)	504 (17.4)	97 (11.2)	132 (1.8)	7080 (32.8)
Smoking status, n (%)
Never	142795 (58.8)	3049 (78.2)	2837 (92.4)	806 (88.9)	7166 (96.1)	20151 (93.2)
Former	78407 (32.3)	505 (13.0)	138 (4.5)	66 (7.3)	9 (0.1)	391 (1.8)
Current	21797 (9.0)	344 (8.8)	96 (3.1)	35 (3.9)	279 (6.7)	1074 (5.0)
Body mass index, mean (SD), kg/m2	26.8 (5.0)	30.1 (5.8)	27.2 (4.7)	23.3 (3.4)	23.0 (3.5)	23.0 (3.6)
Incidence rate of type 2 diabetes (95% CI), per 1000 person-year	2.60 [2.54, 2.65]	7.79 [7.04, 8.60]	10.10 [9.17, 11.20]	3.17 [2.25, 4.33]	4.41 [3.95, 4.91]	10.50 [10.10, 10.90]
Men
N	196848	2832	3292	527	6586	16813
Age, mean (SD), y	56.7 (8.2)	51.1 (7.9)	52.5 (8.5)	52.0 (7.9)	45.2 (14.5)	55.1 (7.5)
Education, n (%)
<High school	26384 (13.6)	426 (15.6)	293 (9.3)	44 (8.8)	4480 (68.0)	8143 (48.4)
High school	47757 (24.7)	631 (23.2)	642 (20.4)	51 (10.2)	1036 (15.7)	5730 (34.1)
Some college	20504 (10.6)	219 (8.0)	275 (8.7)	47 (9.4)	452 (6.9)	1001 (5.9)
College graduate or higher	66685 (34.4)	1020 (37.4)	1471 (46.8)	295 (59.0)	553 (8.4)	633 (3.8)
None of the above	32327 (16.7)	428 (15.7)	463 (14.7)	63 (12.6)	65 (1.0)	1306 (7.8)
Smoking status, n (%)
Never	96913 (49.4)	1698 (60.4)	2232 (68.8)	330 (62.9)	2405 (37.0)	7672 (45.6)
Former	75244 (38.4)	621 (22.1)	564 (17.4)	120 (22.9)	226 (3.5)	3352 (19.9)
Current	24038 (12.3)	491 (17.5)	447 (13.8)	75 (14.3)	3876 (59.6)	5789 (34.4)
Body mass index, mean (SD), kg/m2	27.6 (4.1)	28.2 (4.1)	26.8 (3.7)	25.1 (3.1)	23.0 (3.3)	22.9 (3.4)
Incidence rate of type 2 diabetes (95% CI), per 1000 person-year	4.39 [4.31, 4.47]	9.13 [8.17, 10.20]	11.4 [10.4, 12.4]	5.86 [4.20, 7.94]	4.30 [3.82, 4.83]	10.40 [9.97, 10.90]

Abbreviations: CI, confidence interval; SD, standard deviation.

The BMI cutoffs in women at which type 2 diabetes incidence rate equaled that observed at a BMI of 30.0 kg/m2 in White adults were 25.9 (95% CI: 25.7, 26.0) kg/m2 in Black adults, 24.6 (95% CI: 24.5, 24.7) kg/m2 in British Chinese, 23.3 (95% CI: 23.1, 23.4) kg/m2 in South Asian adults, 22.7 (95% CI: 22.5, 22.9) kg/m2 in Native Chinese, and 18.4 (95% CI: 18.1, 18.7) in Singapore Chinese; in men, the equivalent BMI cutoffs were 26.2 (95% CI: 26.1, 26.3) kg/m2 in Black adults, 25.5 (95% CI: 25.4, 25.6) kg/m2 in British Chinese, 24.5 (95% CI: 24.4, 24.7) kg/m2 in South Asian adults, 24.7 (95% CI: 24.6, 24.8) kg/m2 in Native Chinese, and 22.6 (95% CI: 22.4, 22.8) kg/m2 in Singapore Chinese (Figure 2; Table 2). The BMI cutoffs in women at which type 2 diabetes incidence rate equaled that observed at a BMI of 40.0 kg/m2 in White adults were 31.6 (95% CI: 31.3, 31.9) kg/m2 in Black adults, 29.2 (95% CI: 29.0, 29.4) kg/m2 in British Chinese, 27.3 (95% CI: 27.1, 27.4) kg/m2 in South Asian adults, 26.9 (95% CI: 26.8, 27.1) kg/m2 in Native Chinese, and 25.1 (95% CI: 24.9, 25.3) kg/m2 in Singapore Chinese; in men, the equivalent BMI cutoffs were 31.9 (95% CI: 31.6, 32.2) kg/m2 in Black adults, 30.6 (95% CI: 30.4, 30.9) kg/m2 in British Chinese, 29.0 (95% CI: 28.8, 29.2) kg/m2 in South Asian adults, 29.6 (95% CI: 29.4, 29.8) kg/m2 in Native Chinese, and 27.6 (95% CI: 27.4, 27.7) kg/m2 in Singapore Chinese. The BMI cutoffs for Black adults, South Asian adults, British Chinese, Native Chinese, and Singapore Chinese when the incidence rate equaled that in White adults at a BMI of 35.0 kg/m2 in women and men can be found in Figure 2 and Table 2.

Figure 2.

Racial/ethnic-specific body mass index cutoffs for an equivalent incidence rate of type 2 diabetes compared with White adults at BMI of 30.0 kg/m2, 35.0 kg/m2, and 40.0 kg/m2 by sex.

Poisson regression models were adjusted for race/ethnicity, BMI, age, and baseline year.

Table 2.

Body mass index cutoffs in non-White adults equaling 30.0 kg/m2, 35.0 kg/m2, and 40.0 kg/m2 in White adults for an equivalent incidence rate of type 2 diabetes by sex.

	White	Black	British Chinese	South Asian	Native Chinese	Singapore Chinese
Women
Class I obesity cutoffs, (95% CI), kg/m2	30.0	25.9 (25.7, 26.0)	24.6 (24.5, 24.7)	23.3 (23.1, 23.4)	22.7 (22.5, 22.9)	18.4 (18.1, 18.7)
Class II obesity cutoffs, (95% CI), kg/m2	35.0	29.0 (28.8, 29.1)	27.1 (27.0, 27.3)	25.7 (25.6, 25.8)	25.2 (25.1, 25.3)	23.3 (23.1, 23.5)
Class III obesity cutoffs, (95% CI), kg/m2	40.0	31.6 (31.3, 31.9)	29.2 (29.0, 29.4)	27.3 (27.1, 27.4)	26.9 (26.8, 27.1)	25.1 (24.9, 25.3)
Men
Class I obesity cutoffs, (95% CI), kg/m2	30.0	26.2 (26.1, 26.3)	25.5 (25.4, 25.6)	24.5 (24.4, 24.7)	24.7 (24.6, 24.8)	22.6 (22.4, 22.8)
Class II obesity cutoffs, (95% CI), kg/m2	35.0	29.3 (29.2, 29.5)	28.3 (28.2, 28.4)	27.1 (27.0, 27.2)	27.6 (27.5, 27.7)	26.0 (25.9, 26.1)
Class III obesity cutoffs, (95% CI), kg/m2	40.0	31.9 (31.6, 32.2)	30.6 (30.4, 30.9)	29.0 (28.8, 29.2)	29.6 (29.4, 29.8)	27.6 (27.4, 27.7)

Abbreviation: CI, confidence interval.

Poisson regression models were adjusted for race/ethnicity, BMI, age, and baseline year.

When only age, race/ethnicity, and BMI were included in the models, BMI cutoffs altered primarily in Singapore Chinese, with 1.0–2.3 kg/m² decreases (Figure S1; Table S1). After imposing age inclusion criterion of 45–73 years for analysis, the BMI cutoffs were 0.5–1.3 kg/m² lower compared with those in the primary analysis in Native Chinese people (Figure S2; Table S2); alterations in the BMI cutoffs of other racial/ethnic groups were smaller compared with the Native Chinese group. The results remained similar after excluding people who developed type 2 diabetes within the first two years of follow-up (Figure S3; Table S3).

Discussion

Based on this analysis of multicountry data from three large prospective cohort studies, in both men and women, the BMI cutoffs were significantly lower in non-White adults than White adults for an equivalent risk of type 2 diabetes observed in White adults with BMI of 30 kg/m², 35 kg/m², or 40 kg/m². A substantial difference in the BMI cutoffs was observed even within Chinese people residing in different countries. The BMI cutoffs associated with the same risk of type 2 diabetes from the lowest to the highest were Singapore Chinese, Native Chinese, South Asian adults, British Chinese, Black adults, and White adults. The differences in the BMI cutoffs between White adults and other racial/ethnic groups as well as within each racial group were consistently greater with increasing severity of obesity.

For the first time, this study explored BMI cutoffs to classify three classes of obesity using a prospective study design. Our investigation has significant public health implications given the fast-growing population with morbid obesity who are at substantially increased risks of obesity-related diseases and premature mortality [4, 25]. Precisely classifying people at different risks of obesity-related diseases is foundational for devising targeted approaches for precision risk management. Currently, obesity guidelines either do not include racial/ethnic-specific BMI cutoffs for different severity of obesity or recommend cutoffs primarily based on empirical knowledge. For example, the NICE and WHO guidelines recommend using 37.5 kg/m2 for defining class III obesity and 27.5 kg/m2 for class I obesity in Asians. This classification scheme has been questioned and is also not supported by our study [12]. The existing literature only studied racial/ethnic-specific BMI cutoffs in comparison to a BMI of 30 kg/m² in White adults, but not at a higher BMI such as 35 or 40 kg/m² [16]. For example, a study from Canada found that for an equivalent incidence rate of diabetes at a BMI of 30 kg/m² in White adults, the BMI cutoffs were 24 kg/m² in South Asian adults, 25 kg/m² in Chinese, and 26 kg/m² in Black adults, which is in line with our study [13]. Our study additionally found that several non-White groups had an equivalent risk of type 2 diabetes even when their BMI levels were below 30 kg/m², compared with White adults at a BMI of 40 kg/m². The primary objective of this exploratory study was to reveal substantial racial/ethnic differences in the BMI values for predicting identical diabetes risk and that using 30, 35 and 40 kg/m² for classifying diabetes risk uniformly is inappropriate. This study did not aim to establish racial/ethnic-specific BMI cutoffs for three classes of obesity and the study findings are relevant for precision risk management of type 2 diabetes only, not other diseases. Therefore, precisely mapping the obtained BMI cutoffs to the obesity cutoffs specified in the existing guideline recommendations and performing a direct comparison is discouraged.

There is a consistent racial/ethnic order of BMI cutoffs. Studies have reported that White adults had higher BMI cutoffs for class I obesity than non-White adults, with Black adults, British Chinese, and South Asian adults in descending order [12, 15], which aligns with the findings of our study regardless of the severity levels of obesity. This racial/ethnic variability is likely related to racial/ethnic differences in body composition, biochemical characteristics, lifestyle factors (e.g., physical activity or diet), and genetic backgrounds [26]. Socioeconomic status, physical environment, access to healthcare, and other social determinants of health contribute to the disparities in diabetes risk across race/ethnicity [27]. Black adults have been reported to have lower visceral fat such as liver and pancreatic fat compared to White adults, but Black adults may be more sensitive to adverse consequences of visceral fat accumulation [28]. The main reason for the BMI cutoffs of Asians being lower than those of White adults and Black adults may in part be due to greater body adiposity, especially abdominal visceral fat, which contributes to an increase in fatty acids in the portal vein vasculature, enhancing insulin resistance and contributing to the development of metabolic complications [29, 30]. Additionally, it is possible that South Asian adults secrete less insulin, which means they have fewer reserves to make up for their unhealthy lifestyles compared with White adults and Black adults [29].

Our study found that there were considerable differences in the BMI cutoffs for three classes of obesity at equivalent incidence rates of type 2 diabetes between immigrant Chinese and Native Chinese, with the cutoffs of British Chinese higher than Native Chinese and Singapore Chinese lower than Native Chinese. There was a descending trend in the obesity prevalence from England to China, and to Singapore, while the prevalence of diabetes appeared to ascend [31, 32]. In Singapore Chinese, there was a linear association between BMI and incident diabetes without an inflection point or a threshold across the spectrum of BMI [33]. Genetic backgrounds are likely similar between immigrant Chinese and Native Chinese. Differences in environmental factors, such as dietary patterns, lifestyle behaviors, socioeconomic factors, and healthcare resources and utilizations, may have largely driven the differences in the diabetes risk at the same BMI levels [34]. Acculturation has been associated with type 2 diabetes but it is dependable on the aforementioned environmental influences [35, 36]. Formulating geographically tailored BMI cutoffs for different severity of obesity that reflect cultural, behavioral, and social adaptations associated with migration and ultimately disease risks is needed.

In our study, women had lower cutoffs than men for defining the same severity of obesity, which is consistent with the findings from a previous UK study analyzing linked electronic health record data [12]. At the same cutoffs such as 30, 35, and 40 kg/m² in White adults, women had a lower incidence of type 2 diabetes than men, which led to lower cutoffs for women of non-White race/ethnicity at the same risk of type 2 diabetes than men. Although body fat percentage of women is higher than that of men, women respond to estrogen and tend to store fat in the subcutaneous tissue and lower extremities, which is less metabolically risky than visceral fat [37]. Conversely, men tend to store more fat in abdominal visceral areas, which is associated with greater metabolic risks [38]. Personalized management of diabetes and obesity should take sex into account.

This study did not aim to assess the independent association between race/ethnicity, BMI, and diabetes, nor to estimate the quantitative contributions of possible factors to the observed racial/ethnic differences in the BMI cutoffs in relation to the equivalent diabetes risk. Instead, this study focused on comparing incidence rate of type 2 diabetes across racial/ethnic groups and geographical locations, using BMI levels of White people as the reference. Therefore, in line with the previous studies [12, 15, 16], only demographic variables were included for adjustment. The changes in diet, lifestyle, socioeconomic factors, and other environmental factors over time may alter the relationship between race/ethnicity, BMI, and type 2 diabetes, which is supported by comparing the results with and without adjustment for baseline year. This finding also emphasizes the importance of updating BMI cutoffs over time to ensure that they reflect contemporary risk profiles more accurately.

Our investigation highlights a pressing need to improving current obesity classification scheme that uniformly categorizes obesity and severe obesity across racial/ethnic groups. The current use of a single standard across diverse populations fails to address the varying metabolic risks by race/ethnicity and severity of obesity, which may lead to mismanagement of metabolic parameters like blood pressure, lipid levels, and glucose, especially in non-White populations [8, 39–41]. This not only impedes individual management of diabetes and other metabolic complications but also poses a challenge to the global endeavor to combat these diseases. However, adopting tailored BMI cutoffs that consider the unique demographic profiles of populations enables precision risk assessment and targeted interventions. By embracing precision medicine approaches, we can enhance the cost-effectiveness of diabetes prevention and treatment strategies, ultimately improving public health outcomes and reducing the burden of metabolic diseases globally.

Other than being the first study to identify racial/ethnic-specific BMI cutoffs for three classes of obesity, strengths also include a prospective design, long follow-up, large sample sizes, and multiple non-White groups, particularly Chinese from three different countries. However, this study has limitations. First, we identified type 2 diabetes by self-report in the CHNS and the SCHS, which may have underestimated the risk of developing diabetes. However, this limitation if existed would lead to an overestimation of BMI cutoffs in Native and Singapore Chinese. Second, height and weight at baseline were self-reported in the SCHS. Self-reported BMI tends to be lower than the objectively measured BMI, but the difference is mostly within 1.0 kg/m² [42], which is less than the smallest difference of at least 1.6 kg/m² between the BMI cutoffs in Singapore Chinese and those in other racial/ethnic groups. Third, our findings of lower BMI cutoffs in non-White adults compared with White adults apply only to the outcome of incident type 2 diabetes, not other diseases. Fourth, we did not have body fat data measured by dual-energy X-ray absorptiometry or magnetic resonance imaging to estimate risk of diabetes, but BMI is easy to obtain and more practical to apply at scale. Fifth, results may not be generalized to young adult population because only a small number of young adults were included. Sixth, the three included cohorts are large, but the included participants may not be a representative subset of each racial/ethnic group, especially for the small-sized groups. Seventh, due to the lack of well-accepted criteria to define different severity levels of central obesity commonly assessed by waist circumference, this study focused only on overall obesity assessed by BMI.

Conclusion

In conclusion, our study highlights the necessity for sex- and racial/ethnic-specific BMI cutoffs for classifying three classes of obesity. Furthermore, it underscores the importance of considering geographical location of residence when defining these cutoffs, even within the same race. We recommend that future guidelines incorporate sex, racial/ethnic and geographical considerations in establishing BMI cutoffs for three classes of obesity. It is imperative to incorporate these tailored standards into practice and subsequently evaluate their effectiveness in preventing, screening, and managing type 2 diabetes and other obesity-related metabolic disorders.

Study importance.

What is already known about this subject?

• Heterogeneous risks of type 2 diabetes by different severity of obesity and race/ethnicity are known.

• The risk of type 2 diabetes, body fat percentage, and body fat distribution differ markedly by race/ethnicity at the same BMI.

• Racial/ethnic-specific BMI cutoffs for defining class I, II, and III obesity in non-White people are either unclear or based on expert consensus which is not supported by the literature.

What are the new findings in your manuscript?

• The BMI cutoffs were significantly lower in non-White adults than White adults for an equivalent risk of type 2 diabetes observed in White adults with BMI of 30 kg/m², 35 kg/m², or 40 kg/m².

• A substantial difference in the BMI cutoffs was observed even within Chinese people residing in different countries.

How might your results change the direction of research or the focus of clinical practice?

• We recommend that future investigations and guideline recommendations incorporate sex, race/ethnicity and geographical considerations in establishing BMI cutoffs for three classes of obesity.