Waist circumference thresholds predicting incident dysglycaemia and type 2 diabetes in Black African men and women

Abstract Aims To determine the waist circumference (WC) thresholds for the prediction of incident dysglycaemia and type 2 diabetes (T2D) in Black South African (SA) men and women and to compare these to the advocated International Diabetes Federation (IDF) Europid thresholds. Materials and Methods In this prospective study, Black SA men (n = 502) and women (n = 527) from the Middle‐aged Sowetan Cohort study who had normal or impaired fasting glucose at baseline (2011‐2015) were followed up until 2017 to 2018. Baseline measurements included anthropometry, blood pressure and fasting glucose, HDL cholesterol and triglyceride concentrations. At follow‐up, glucose tolerance was assessed using an oral glucose tolerance test. The Youden index was used to determine the optimal threshold of WC to predict incident dysglycaemia and T2D. Results In men, the optimal WC threshold was 96.8 cm for both dysglycaemia and T2D (sensitivity: 56% and 70%; specificity: 74% and 70%, respectively), and had higher specificity (P < 0.001) than the IDF threshold of 94 cm. In women, the optimal WC threshold for incident dysglycaemia was 91.8 cm (sensitivity 86%, specificity 37%) and for T2D it was 95.8 cm (sensitivity 85%, specificity 45%), which had lower sensitivity, but higher specificity to predict incident dysglycaemia and T2D than the IDF threshold of 80 cm (sensitivity: 97% and 100%; specificity: 12% and 11%, respectively)). Conclusions We show for the first time using prospective cohort data from Africa that the IDF Europid WC thresholds are not appropriate for an African population, and show that African‐specific WC thresholds perform better than the IDF Europid thresholds to predict incident dysglycaemia and T2D.


Conclusions:
We show for the first time using prospective cohort data from Africa that the IDF Europid WC thresholds are not appropriate for an African population, and show that African-specific WC thresholds perform better than the IDF Europid thresholds to predict incident dysglycaemia and T2D.
K E Y W O R D S metabolic syndrome, obesity, risk stratification, sub-Saharan African cohort

| INTRODUCTION
The global prevalence of type 2 diabetes (T2D) is increasing, with sub-Saharan Africa (SSA) having the highest projected relative rates of increase. 1 The burden of T2D in SSA is reflected by the high estimated T2D-associated deaths (312 000 deaths in 2017), with 73% of these being in people under the age of 60 years, a higher proportion than any other region in the world. 1 Within SSA, South Africa (SA) has the highest number of people with T2D, 1 and T2D was the second leading cause of death in SA in 2016 (5.5% of deaths), and the highest amongst women (7.2% of deaths). 2 Notably, SSA has the highest proportion (59.7%) of people with undiagnosed T2D. 1 Accordingly, risk stratification that is accessible and cost-effective is essential for the early detection of T2D to prevent or delay the progression of the disease.
Obesity, in particular central obesity, is an important risk factor for T2D. 3,4 Although imaging techniques are more accurate measures of total and central adiposity, they are not practical or affordable for routine practice and population-based risk stratification. Accordingly, anthropometric measures are used as surrogate markers for risk stratification for T2D. Body mass index (BMI) is the most commonly used proxy of total adiposity, while waist circumference (WC) is most often used as a proxy for central adiposity. [4][5][6] Waist circumference represents the sum of abdominal visceral (VAT) and subcutaneous adipose tissue (SAT), with VAT being the most significant determinant of T2D. 7,8 However, we and others have shown that for the same level of WC, Black Africans have less VAT than their white European counterparts. [9][10][11] Accordingly, the WC threshold used for defining risk for T2D may differ in Black Africans.
Indeed, both the World Health Organization (WHO) and the International Diabetes Federation (IDF) acknowledge that optimal thresholds for abdominal obesity vary across different ethnicities and population groups. 6,12 Although several studies in SSA have been undertaken to identify WC thresholds for risk, these have all been cross-sectional and relied on metabolic syndrome (MetS; excluding WC) as the outcome. [13][14][15][16] As there is no consensus on an appropriate WC threshold for Black Africans, the IDF has recommended the use of Europid thresholds (≥80 cm in women and ≥94 cm in men) for SSA. 6 Prospective studies are therefore required to identify the optimal WC thresholds that identify incident T2D in Black African men and women.
While WC is regarded as a useful primary screening tool for T2D, it is also a key feature of MetS, which is also typically used in risk prediction for T2D and cardiovascular diseases. 6 MetS represents a cluster of risk factors that occur together more often than by chance alone, and in addition to WC, include elevated blood pressure, fasting glucose and triglycerides, and low fasting HDL cholesterol concentrations. 6 However, it is not clear whether including these additional MetS risk factors improves the discriminatory ability to predict T2D in African men and women when compared to WC alone.
The aim of this study, therefore, was to determine the WC thresholds for the prediction of incident dysglycaemia (prediabetes and T2D) and T2D in Black SA men and women, and to compare these to the advocated Europid thresholds, as defined by the IDF. A secondary aim was to determine if the derived WC thresholds for the prediction of incident dysglycaemia and T2D performed similarly to the presence of MetS in Black SA men and women.

| Design, study population and setting
Baseline data collection was part of the African WITS-INDEPTH Partnerships for Genomic Research (AWI-Gen) Collaborative Centre, which is a Human Heredity and Health in Africa (H3A) Consortium study, 17,18 and included 1027 men and 1004 women aged 40 to 60 years living in Soweto, from which the participants in the Middle-aged Soweto Cohort (MASC) study were randomly selected (n = 1112). The MASC study is a longitudinal study of Black SA men and women residing in Soweto, South Africa, on whom baseline data were collected between 2011 and 2015, and again between January 2017 and August 2018 ( Figure 1). Data in this study were collected from a sample of 1029 participants (502 men and 527 women) who were representative of the AWI-Gen sample and did not differ in terms of age, sex, sociodemographic or lifestyle factors from the main cohort (Supplementary Table S1). Only participants with normal fasting glucose (NFG) or impaired fasting glucose (IFG), and anthropometric measures at baseline, as well as measures of glycaemia from an oral glucose tolerance test at follow-up, were included in this analysis ( Figure 1). Complete data were available for 890 participants (452 men and 438 women).
The study was conducted in accordance with the Declaration of

| Sociodemographic and health questionnaires
The same interviewer-administered questionnaire was completed at both time points and included age, current employment (employed/ not employed), and highest educational level attained (no formal schooling/elementary school, secondary school, tertiary education).
In addition, participants were asked to bring all other current medications, including diabetes medications, to the testing facility for recording and verification. Participants were classified as current smokers/non-smokers and current alcohol consumers/nonconsumers.

| Anthropometry and blood pressure
Weight was measured to the nearest 0.1 kg using a digital scale (model: TBF-410; Tanita Corporation, Arlington Heights, Illinois).
Height was measured to the nearest 0.1 cm using a wall-mounted stadiometer (Holtain, Crymych, UK). WC and hip circumference were measured to the nearest 0.1 cm with a nonstretchable tape. WC was measured in the mid-axillary line at the midpoint between the lower margin of the last palpable rib and the top of the iliac crest at the end of normal expiration, and hip circumference was measured at the greatest protrusion of the buttocks. 12 Systolic and diastolic blood pressure were measured on the left arm using a digital blood pressure monitor (Omron M6, Kyoto, Japan) and appropriate cuffs. After the participant had been seated for at least 5 minutes, three blood pressure readings were taken at 2-minute intervals. For each participant, the average of the second and third readings was used in the analyses.

| Glucose tolerance and MetS classification
Glucose tolerance was defined based on the WHO criteria. 20 At baseline, only fasting glucose samples were available, hence the participants were classified as having NFG if fasting glucose was <6.1 mmol/L, IFG if fasting glucose was 6.1 to 6.9 mmol/L, or T2D if fasting glucose was ≥7 mmol/L and/or if taking T2D medications.
Only those with NFG or IFG at baseline were included in this study.
At follow-up, glucose tolerance was defined based on both fasting were classified with T2D. At follow-up, dysglycaemia, which encompasses both prediabetes and diabetes, was defined as IFG and/or IGT and/or T2D.
Presence of MetS was based on the 2009 harmonized criteria. 6 Participants with three or more of the following components were classified as having MetS: (i) elevated WC (≥94 cm in men or ≥ 80 cm in women); (ii) elevated fasting triglycerides (≥1.7 mmol/L); (iii) reduced fasting HDL cholesterol (<1.0 mmol/L in men or < 1.3 mmol/L in women); (iv) elevated blood pressure (≥130 mmHg for systolic and/or ≥ 85 mmHg for diastolic blood pressure and/or using blood pressure medication); (v) elevated fasting glucose (≥5.6 mmol/L and/or using diabetes medication).

| Statistics
Data analysis was conducted in STATA SE Version 15 (StataCorp, College Station, Texas). Normality of the data was assessed using Shapiro-Wilks test. As all the descriptive variables were skewed, continuous variables are presented as median (25-75th percentiles) and categorical variables are presented as frequencies and percentages.
Wilcoxon-Mann-Whitney tests and chi-squared tests were used to compare continuous and categorical variables between men and women, respectively. Receiver-operating characteristic (ROC) area under the curves (AUCs) were used to assess and compare the ability of baseline WC to predict incident dysglycaemia (prediabetes or T2D) and T2D at follow-up. The standard AUC analysis was used as 46% of the participants were diagnosed with T2D at the follow-up visit,  6 and defined as meeting any three of the following criteria: elevated waist circumference (≥80 cm in women; ≥94 cm in men); elevated blood glucose (≥5.6 mmol/L and/or using diabetes medication); elevated fasting triglycerides (≥1.7 mmol/L); reduced fasting HDL cholesterol (<1.0 mmol/L in men, <1.3 mmol/L in women); elevated blood pressure (≥135/85 mmHg and/or using antihypertension medication).
which precluded time-to-event analysis. For the prediction of incident dysglycaemia, only those with NFG at baseline were included in the analysis, whereas for the prediction of incident T2D, those with NFG and IFG at baseline were included in the analysis. Optimal WC thresholds to predict incident dysglycaemia and T2D were determined using Youden's index in men and women separately. 21  3 | RESULTS

| Participant characteristics
At baseline, the sample were middle-aged (50 years), with men being slightly older than women (

| Performance of different WC thresholds to predict incident dysglycaemia and T2D
The ROC analyses for WC to predict incident dysglycaemia and T2D in men and women are presented in Figure 2. WC showed acceptable discrimination to predict dysglycaemia and T2D in men and women, with the AUCs being higher in men than women ( Figure 2). Based on F I G U R E 2 Receiver-operating characteristic area under the curve (AUC) of waist circumference to predict A, incident dysglycaemia and B, type 2 diabetes in Black African men and women the Youden's index, the optimal WC thresholds to predict incident dysglycaemia in men and women were 96.8 cm and 91.8 cm, respectively ( Table 2). In men, this threshold was similar to the IDF threshold of 94 cm, and accordingly had similar sensitivity (P = 0.250), but significantly higher specificity (P < 0.001). However, the threshold of 96.8 cm was higher than those derived from cross-sectional studies of other African populations to detect MetS (84-90 cm), with a resultant lower sensitivity, but higher specificity.
In women, the threshold of 91.8 cm to predict incident dysglycaemia was higher than the IDF-recommended threshold of 80 cm but was similar to most thresholds from other African studies to detect MetS (Table 2). Although the sensitivity was lower than the IDF threshold (0.86 vs. 0.97; P = 0.001), the specificity (0.37 vs. 0.12; P < 0.001) was higher using the derived threshold of 91.8 cm.
The optimal WC threshold to predict incident T2D in men (

| Comparative ability of derived WC thresholds versus MetS to predict incident dysglycaemia and T2D
We then determined whether including additional MetS risk factors together with the derived WC thresholds improved the prediction of incident dysglycaemia and T2D compared to the derived WC thresholds alone. In Tables 2 and 3, we showed that the threshold to predict incident dysglycaemia and T2D in men, and T2D in women were simi-  (Tables 2 and 3). In contrast, the sensitivity of MetS using a WC threshold of 96 cm to predict both incident dysglycaemia and T2D in women was lower than for the derived thresholds alone, while the specificity was higher (Tables 2 and 3).  . NPV = sensitivity/1-specificity. Youden's index = (sensitivity + specificity)-1. MetS was defined using a waist circumference threshold of 96 cm for men and women derived from this study. *P = 0.016 for the difference in the sensitivity between the derived and IDF thresholds. # P < 0.001 for the difference in the specificity between the derived and IDF thresholds.

| DISCUSSION
This is the first prospective study to examine WC thresholds to predict incident dysglycaemia and T2D in an African population and showed that the optimal thresholds differed from those in European populations. The optimal thresholds to predict incident dysglycaemia and T2D in Black SA men were 96.8 cm for both outcomes and in women they were 91.8 and 95.8 cm, respectively. Importantly, these African-specific WC thresholds had significantly higher specificity than the IDF Europid thresholds.
The WC thresholds to predict incident dysglycaemia and T2D in women were higher than the IDF threshold of 80 cm. This is consistent with the findings of several cross-sectional studies in SA that detected MetS, defined as at least two components of MetS excluding WC, and reported optimal thresholds of 90 to 94 cm. 13,14,16,24 We showed that in women the thresholds of 91. 8  The major strengths of this study are the prospective design and the diagnosis of incident dysglycaemia and T2D using an OGTT. To date, all studies that have explored thresholds for WC have been cross-sectional and were designed for identifying the optimal WC cutoff point for detecting MetS. 13,14,16,24 We used an OGTT to diagnose T2D at follow-up, which is considered the "gold standard", particularly in African populations where fasting glucose and glycated haemoglobin (HbA1c) may perform sub-optimally. 38 In conclusion, we show for the first time using prospective cohort data from South Africa that the IDF Europid WC thresholds are not appropriate for an African population and that African-specific WC thresholds perform better than the IDF Europid thresholds in predicting incident dysglycaemia and T2D in Black SA men and women. These findings verify existing evidence from cross-sectional studies showing suboptimal performance of currently recommended WC thresholds in African populations, 13,15,16 but require validation in an independent longitudinal cohort from Africa.