Obesity is estimated to be the third largest contributor to the overall burden of disease in Australia [1]. It is most commonly assessed using the body mass index (BMI) threshold of 30 kg/m² or greater [2]. The limitations of BMI have been discussed in the literature, including in a review which suggested that it is time we move beyond the use of BMI to assess obesity to actual measurements of body fat mass [3]. However, methods such as underwater weighing, deuterium dilution, dual-energy X-ray absorptiometry, or bioelectrical impedance analysis reliably estimate total body fat, but are too expensive and time-consuming for population-level use.

Consequently, which anthropometric measure best identifies high-risk adiposity has been much debated. Various alternative measures to BMI have been advocated to improve estimates of adiposity and/or to provide better prediction of ill-health, including waist circumference (WC), waist-hip ratio, and waist-height ratio [4]. WC has generally been shown to be a better predictor of metabolic risk and chronic diseases than BMI. This is likely due to a combination of WC better identifying individuals at increased health risk through greater total body fat, and by also including those with central adiposity who are not necessarily identified by BMI [5, 6]. However, the degree of improvement in prediction of outcomes has not generally been seen as sufficient to drive a shift from BMI to WC.

New data has recently emerged suggesting that 40–70 % of the population classified as having high-risk WC have a BMI that is below the obese range [7–9]. However, relatively little is known about their associated health risks. In order to determine if they are a group that should be routinely identified as obese, it is important to know whether their health risks are greater than individuals with a non-obese BMI and non-obese WC. The aim of our study was to compare the metabolic health risks across different combinations of BMI and WC categories in a population of Australian adults, and to explore whether including those with a non-obese BMI but a high risk WC with those classified as ‘obese’ by BMI alone improves prediction of adiposity-related metabolic outcomes.

The characteristics of the study population are presented in Table 1. The mean age of the participants was 48.2 years (SD 15.4 years) and 50 % were men. The majority of the population had their highest level of completed education as secondary school, and were Europid, sedentary and non-smokers. The prevalence of hypertension, diabetes, dyslipidaemia and CVD were 29, 7, 26 and 7 %, respectively. The mean BMI and WC were 26.6 kg/m² and 90 cm (96 cm for men and 84 cm for women). Around two-thirds of the population were non-obese by both BMI and WC (BMI^N/WC^N), while 12, 2 and 18 % were categorised as BMI^N/WC^O, BMI^O/WC^N and BMI^O/WC^O, respectively. Figure 1 presents the prevalence by sex.

	Total	BMIN/WCN	BMIN/WCO	BMIO/WCN	BMIO/WCO
	n = 10 659	n = 6 688	n = 1 594	n = 171	n = 2 206
Age (years)	48.2 (15.4)	46.1 (14.6)	56.2 (17.4)	42.1 (10.9)	51.6 (15.2)
Male sex, %	50.0	52.0	43.5	68.0	44.3
Education, %
Secondary school	37.0	33.7	44.2	35.1	44.7
Trade, technician’s certificate	32.2	32.1	31.9	33.5	32.8
Associate, undergraduate diploma,	11.7	12.3	10.6	11.0	10.5
nursing or teaching qualification
Bachelor degree, post-graduate	19.1	21.9	13.4	20.4	11.9
Europid, %	87.5	86.3	92.1	84.5	89.4
BMI (kg/m2)	26.6 (4.9)	24.2 (2.7)	27.8 (1.9)	31.0 (0.8)	34.3 (4.2)
Waist circumference (cm)	90.0 (13.6)	83.6 (9.5)	98.4 (8.0)	94.4 (5.8)	107.5 (11.2)
TV viewing time, %
<2 h/day	6.7	7.6	4.0	4.5	5.4
2- < 4 h/day	8.1	9.4	5.0	5.2	5.8
≥4 h/day	85.2	83.0	91.0	90.3	88.8
Alcohol consumption, ≤10 g/day (%)	62.1	59.4	64.5	63.6	70.5
Smoking status, %
Current smoker	16.4	17.0	16.5	21.2	13.4
Ex-smoker	25.9	23.2	33.1	24.1	31.3
Non-smoker	57.8	59.8	50.3	54.7	55.3
Systolic blood pressure (mmHg)	128.1 (18.3)	124.8 (16.6)	133.8 (21.4)	126.8 (12.1)	136.6 (19.5)
Diastolic blood pressure (mmHg)	70.8 (11.7)	69.4 (11.0)	71.9 (13.8)	72.9 (7.5)	75.2 (12.2)
Fasting plasma glucose (mmol/l)	5.5 (1.1)	5.4 (0.8)	5.7 (1.2)	5.6 (0.8)	6.0 (1.7)
Total cholesterol (mmol/l)	5.5 (1.1)	5.4 (1.0)	5.8 (1.2)	5.7 (0.8)	5.8 (1.2)
Hypertension, %	28.9	20.8	45.0	19.0	49.7
Diabetes, %	7.0	3.7	10.7	4.0	17.2
Dyslipidaemia, %	25.9	18.4	36.2	28.0	46.6
Cardiovascular disease, %	7.1	5.4	13.4	2.3	9.9

Data presented as mean (standard deviation) unless otherwise specified

Adiposity categories: BMI^N/WC^N: non-obese BMI and WC; BMI^N/WC^O: non-obese BMI, obese WC; BMI^O/WC^N: obese BMI, non-obese WC; BMI^O/WC^O: obese BMI and WC

Hypertension defined as blood pressure ≥140/90 mmHg or on antihypertensive medication; diabetes defined as fasting plasma glucose ≥7.0 mmol/l or 2-hour plasma glucose ≥11.1 mmol/l, or on oral hypoglycaemic medication or insulin; dyslipidaemia defined as triglycerides ≥2.0 mmol/l or HDL cholesterol <1.0 mmol/l; CVD defined as previous angina, stroke and/or coronary artery disease

Discrimination was significantly higher for obesity defined using a combination of BMI or WC compared to obesity defined using BMI alone (Table 2). Similar results were generally found when stratified by age and by sex, except for CVD in those aged <55 years and diabetes in those aged ≥55 years, where no significant difference in AUC between the two obesity definitions was observed (Additional file 1: Table S1).

Compared to those with BMI^N/WC^N, the odds for hypertension, diabetes and dyslipidaemia were increased in those with BMI^N/WC^O and those with BMI^O/WC^O (Fig. 2 and Additional file 2: Figure S1), with the greatest odds in the BMI^O/WC^O group. In contrast, the odds for CVD were similar for both the BMI^N/WC^O and BMI^O/WC^O groups compared to those with BMI^N/WC^N. In all analyses, greater odds ratios were observed for the BMI^O/WC^N group compared to the BMI^N/WC^N group, though the results were not statistically significant. The magnitude of the odds ratios were generally higher in women and in those aged <55 years compared to men and those aged ≥55 years respectively. Systolic blood pressure, fasting plasma glucose and total cholesterol were elevated in both those with BMI^N/WC^O and those with BMI^O/WC^O compared to those with BMI^N/WC^N (Table 3 and Additional file 3: Table S2). Total cholesterol, but not systolic blood pressure and fasting glucose, was also elevated in those with BMI^O/WC^N compared to those with BMI^N/WC^N.

	Adiposity categories
	BMIN/WCN Coefficient (Reference)	BMIN/WCO Coefficient (95 % CI)	BMIO/WCN Coefficient (95 % CI)	BMIO/WCO Coefficient (95 % CI)
Systolic blood pressure (mmHg)	0.0	3.0 (1.5, 4.6)	3.3 (−0.5, 7.0)	8.7 (7.4, 10.0)
Fasting glucose (mmol/l)	0.0	0.2 (0.1, 0.3)	0.2 (−0.0, 0.4)	0.6 (0.5, 0.7)
Total cholesterol (mmol/l)	0.0	0.2 (0.1, 0.3)	0.4 (0.1, 0.7)	0.3 (0.3, 0.4)

Analyses were adjusted for age, sex, education, country of birth, TV viewing time, alcohol consumption and smoking status

Adiposity categories: BMI^N/WC^N: non-obese BMI and WC; BMI^N/WC^O: non-obese BMI, obese WC; BMI^O/WC^N: obese BMI, non-obese WC; BMI^O/WC^O: obese BMI and WC

When we used the alternative BMI cut-point for obesity (≥28.5 kg/m²), the proportion of the population with BMI^N/WC^O decreased from 12 to 7 %, while the proportion with BMI^O/WC^N and with BMI^O/WC^O increased from 2 to 6 % and from 18 to 23 %, respectively. The proportion of hypertension, diabetes, dyslipidaemia and CVD attributable to obesity defined using the alternative BMI cut-point were higher compared to when obesity was defined using the conventional BMI cut-point of ≥30 kg/m², though still lower than the proportions attributable to obesity defined using a combination of BMI or WC (Table 2 and Additional file 1: Table S1). Similarly, while the AUCs for obesity defined using the alternative BMI cut-point were higher compared to obesity defined using the conventional BMI cut-point, they were still significantly lower than the AUCs for obesity defined using a combination of BMI or WC (Table 2 and Additional file 1: Table S1).

In the sensitivity analysis, including only Europid participants did not significantly alter our results.

In a population of Australian adults aged 25 years and over, around one in three were classified as obese by their BMI or their WC. The PAF and AUC for each metabolic health outcome were higher when obesity was defined as having either an obese BMI or obese WC, indicating an improvement in the prediction of adiposity-related metabolic outcomes, compared to when defined as obese BMI alone. When compared to those who were non-obese (BMI^N/WC^N), we found increased odds for hypertension, diabetes, dyslipidaemia and CVD in those with BMI^N/WC^O and those with BMI^O/WC^O. These findings highlight the importance of looking beyond BMI when assessing obesity in relation to health outcomes across the population, as even those with a BMI in the non-obese range may have elevated health risks that may be as high as those with an obese BMI.

Population attributable fraction estimates the proportion of disease that is attributable to a given risk factor, and thus represents the proportion of disease that theoretically would not have occurred had the risk factor not been present in the population [18]. In this study, we found that the proportion of metabolic outcomes attributable to an obese BMI was significantly lower than the proportion attributable to an obese BMI or WC combined. This addition of WC to BMI appeared to be beneficial as obesity defined using BMI alone had lower discriminative ability for outcomes compared to obesity defined using BMI and WC combined. Importantly, these differences persisted even after we lowered the BMI cut-point to increase the prevalence of the population classified as obese by BMI alone to the same as that identified by waist.

Our findings relating to obesity defined using BMI or WC combined and obesity defined using WC alone were very similar. However, we were unable to ascertain whether this reflects no added advantage in combining WC and BMI to define obesity rather than using WC alone, or whether this is because in this study, there was a modest number of participants with BMI^O/WC^N (2 %), and therefore the individuals captured by defining obesity as either an obese BMI or WC is virtually the same as those captured by obese WC alone. Nevertheless, these findings further highlight the importance of looking beyond BMI when setting targets for the population. By incorporating WC to identify those with high risk adiposity and who should be targeted for intervention, the potential exists to prevent more metabolic outcomes in the population.

There has been little systematic analysis of the health risks associated with different combinations of BMI and WC. One study found that within each BMI category, those with a high-risk WC were more likely to have hypertension, diabetes, dyslipidaemia and the metabolic syndrome compared to those with a normal-risk WC [19]. This finding, while in line with ours, reported odds ratios but not absolute risks, and without absolute risk, it was not possible to determine whether the risk of disease in a person with high risk WC and obese BMI differed from a person with a high risk WC and non-obese BMI. To our best knowledge, only one previous study has found that, among individuals with a non-obese BMI, those with a high-risk WC (BMI^N/WC^O) had an increased risk for hypertension compared to those with a non-high risk WC (BMI^N/WC^N) [20]. Thus our study is among the first to demonstrate that by not regularly identifying those with obesity exclusively according to their WC, we are missing a group who appear to be at an increased risk of metabolic outcomes.

Previous studies that have compared different anthropometric measures in relation to health outcomes have generally concluded that while WC predicts metabolic outcomes more strongly that BMI, the improvements tend to be relatively modest and insufficient to warrant a move away from BMI to WC [21, 22]. More recently, there is evidence that BMI and WC have been tracking differently over the past few decades. In a number of populations, greater increases in WC than BMI have been observed, and WC has continued to increase despite an apparent plateau in BMI trends [8, 23–27]. The potential implication of these trends is that even if WC is only equivalent to or modestly better at predicting metabolic outcomes than BMI, WC may be increasingly capturing a different group of the population who have increased health risks related to excess adiposity than would be captured by BMI alone.

Taken together, our findings reinforce the importance of WC measurement in obesity assessment to accompany measures of BMI. Nevertheless, current population monitoring for obesity and identification of high risk individuals in clinical practice still tend to rely heavily on BMI [28, 29]. While some national weight management guidelines suggest that health professionals consider both BMI and WC, how to do so remains unclear within the guidelines [29]. Further, there are no similar recommendations for population health surveys. There needs to be a systematic move away from the sole use of BMI in obesity assessment towards identification of high risk adiposity based on WC or a combination of adiposity markers. Cut-points to identify high risk adiposity using any anthropometric measure are arbitrary since associations between risk factors and diseases are continuous, without discrete thresholds separating disease and no-disease. Nevertheless, cut-points are valuable for population health monitoring, to classify and quantify likely disease burdens, and to target health promotion.

A limitation to the current study is the small number of participants in the BMI^O/WC^N category, which limited our ability to ascertain their odds for disease compared to those in other adiposity categories. Future studies in populations with a greater number of participants with BMI^O/WC^N may help elucidate whether combining BMI and WC for obesity assessment would be of benefit, or whether simply moving to another single anthropometric measure, such as WC, is sufficient.

In conclusion, one pitfall of current obesity monitoring is that it may be missing a large proportion of the population who have a high risk WC but a BMI below the obese range, who are at increased health risk through excess adiposity. Our findings support the need for a systematic approach to obesity assessment including identification of high risk adiposity based on other measures in addition to BMI. This will improve identification of those at increased risk of ill-health, which is important for better prioritisation of policy and resources.