Overweight and obesity are associated with a multitude of health conditions, and the anthropometric indicators used to define obesity can help to identify individuals or populations at specific risk of a diverse range of health problems.1 The many available anthropometric obesity indicators include body mass index (BMI), waist circumference (WC), waist-hip ratio (WHR), and waist-height ratio (WHtR). Associations have been sought between these indicators and important cardiovascular and metabolic risk factors, such as impaired glucose metabolism, hypertension (HT), dyslipidemia, insulin resistance, and metabolic syndrome.

The consensus view is that cardiometabolic risk is better predicted by abdominal obesity based on WC, WHR, or WHtR than by general obesity based on BMI. In the 1980s, the abdominal obesity indicator WHR attracted much attention, and associations were reported with cardiovascular disease, stroke, and diabetes mellitus.2,3 More recently, WHtR has emerged as a better predictor of metabolic risk than general obesity.4,5 Height alters little during adulthood, and therefore changes in WHtR are assumed to reflect changes in WC; in contrast, WHR is more sensitve to changes in body shape because it tracks proportional changes in WC and hip circumference.6 The results of 3 meta-analyses have indicated that abdominal obesity based on WHtR is more strongly associated than BMI with diabetes, metabolic syndrome, and other cardiovascular risk factors.7–9

A Spanish study from the Canary Islands found that WHtR was better than BMI, WC, and WHR at detecting diabetes mellitus and other cardiovascular risk factors such as HT, hyperlipidemia, and impaired fasting plasma glucose.10

In a subsequent study of diabetes risk, WHtR was found to be one of the main predictors of the disease, together with impaired fasting glucose, Canarian ancestry, and insulin resistance.11 Another Spanish study found that WHtR and WC had a greater capacity than BMI to predict diabetes mellitus, dyslipidemia, and metabolic syndrome, whereas BMI was better at predicting HT.12

These previous studies evaluated the relationship between anthropometric obesity indicators and impaired fasting glucose; however, they did not evaluate whether abdominal obesity, in particular WHtR, is more strongly associated than general obesity with prediabetes under a broader definition that includes altered glycosylated hemoglobin (HbA1c). Here, we present the findings of another Spanish study evaluating the magnitude of the association between different anthropometric obesity indicators and HT, dyslipidemia, and the presence of prediabetes, defined broadly as altered fasting glucose and/or altered HbA1c.

A total of 2022 patients participated in the study: 589 women and 595 men with prediabetes and 450 women and 388 men with normal carbohydrate metabolism. Stratification of the prediabetes and nonprediabetes groups according to a range of characteristics is shown in Table 1. For both women and men, stratification by age revealed statistically significant between-group differences, Moreover, women and men with prediabetes more frequently had a family history of diabetes, HT, dyslipidemia, or obesity than did their counterparts with normal carbohydrate metabolism.

The association of obesity indicators with HT and dyslipidemia is shown in Table 2. The discriminatory power of all logistic regression models was > 0.5, with the area under the curve varying from around 0.70 for HT to around 0.60 for hypertriglyceridemia and low HDL-C. Among women, HT showed the strongest association with general obesity (age-adjusted OR, 3.01; 95% confidence interval [95%CI], 2.24-4.04). In contrast, among men the association was strongest with abdominal obesity based on WHtR (age-adjusted OR, 3.65; 95%CI, 2.66-5.01). The risk factors most strongly associated with WHtR-estimated abdominal obesity among women were hypertriglyceridemia (age-adjusted OR, 2.49; 95%CI, 1.68-3.67) and low HDL-C (age-adjusted OR, 2.70; 95%CI, 1.89-3.86).

In contrast, among men, hypertriglyceridemia and low HDL-C were more strongly associated with general obesity, with age-adjusted OR values of 2.06 (95%CI, 1.56-2.73) and 1.68 (95%CI, 1.21-2.33), respectively.

The association between obesity indicators and the presence of prediabetes is presented in Table 3. The discriminatory power of all logistic regression models was > 0.5. The area under the curve in the models that included all adjustment variables was approximately 0.70. Prediabetes showed a stronger association with abdominal obesity (using either of the 2 criteria) than with general obesity. Prediabetes among women showed the strongest association with abdominal obesity based on WHtR (age-adjusted OR, 3.02; 95%CI, 2.29-3.98). After further adjusting for family history of diabetes, smoking status, alcohol consumption, lipid-lowering drug therapy, HT, and dyslipidemia, the OR decreased to 2.48 (95%CI, 1.85-3.33). Among men, prediabetes showed the strongest association with abdominal obesity based on WC (age-adjusted OR, 2.85; 95%CI, 2.18-3.73). After further adjusting for family history of diabetes, smoking status, alcohol consumption, lipid-lowering drug therapy, HT, and dyslipidemia, the OR decreased to 2.33 (95%CI, 1.75-3.08).

Finally, we assessed the association of obesity indicators with each of 3 types of prediabetes (Table 4). Among women, abdominal obesity based on WHtR showed the strongest association with all 3 types of prediabetes. The OR values adjusted for all variables (age, family history of diabetes, smoking status, alcohol consumption, lipid-lowering drug therapy, HT, and dyslipidemia) were as follows: 1.54 (95%CI, 0.95-2.50) for impaired fasting glucose, 1.98 (95%CI, 1.34-2.92) for impaired HbA1c, and 4.02 (95%CI, 2.66-6.08) for alterations to both fasting plasma glucose and HbA1c. Among men, abdominal obesity based on WC showed the strongest association with impaired glucose and with impaired HbA1c; however, general obesity based on BMI showed the strongest association with the simultaneous alteration of fasting plasma glucose and HbA1c (all-variable-adjusted OR, 2.90; 95%CI, 2.05-4.08).

Of the 3 anthropometric indicators studied, WHtR showed the strongest association with dyslipidemia in women and with HT in men, whereas BMI showed the strongest association with HT in women and with dyslipidemia in men. Compared with BMI, both abdominal obesity indicators showed a stronger association with prediabetes, except among men who met both prediabetes critieria: impaired fasting glucose and impaired HbA1c.

Previous studies indicated that abdominal obestity can induce a state of insulin resistance, characterized by a defective response to insulin in peripheral tissues (musculoskeletal tissues, liver, and adipose tissue) and resulting in altered glucose uptake and utilization.18,19 This situation leads to elevated plasma glucose and a compensatory increase in insulin, accompanied by other changes such as HT, dyslipidemia, and fatty liver. The results of a large and diverse body of research, including the present study, reveal that prediabetes is more strongly assocated with indicators of abdominal obesity than with general obesity. However, this pattern is not observed for HT or dyslipidemia.

Indeed, previous research into how anthropometric measurements relate to HT and dyslipidemia has produced inconsistent findings. For example, a Spanish study of elderly individuals at high cardiovascular risk found that WC and WHtR were superior to BMI at distinguishing dyslipidemia, whereas BMI was better at distinguishing HT.12 This lack of consistency is evident from the meta-analyses cited above, which sought to determine which anthropometric indicator best distinguishes dyslipidemia, HT, and diabetes mellitus. In 1 study, WHtR showed no clear superiority over the other indicators in distinguishing dyslipidemia, but was more strongly associated with HT in men.7 Another meta-analysis found that abdominal obesity indicators were not clearly superior at distinguishing dyslipidemia in men or HT in women.8 The results presented here are consistent with these meta-analyses in producing heterogeneous results, with differences between men and women.

Nonetheless, other studies have found altered fasting plasma glucose to be more strongly associated with abdominal obesity than with general obesity,20–22 and similar results have been obtained in studies carried out in Spain.11,12,23 For example, in 2 studies, altered fasting plasma glucose was more strongly associated with WC or WHtR than with BMI.11,12 Another study also revealed a stronger assocation of impaired fasting glucose with WC than with BMI; however, this study did not measure WHtR.19 In the present study, we obtained similar results from an analysis of the assocation in men and women between different anthropometric indicators and the presence of prediabetes defined as alterations to fasting plasma glucose and/or HbA1c. The only exception was observed in men who met both prediabetes criteria, among whom the strongest association with prediabetes was shown by general obesity based on BMI. This result has important implications for the management of male patients, since individuals who meet both prediabetes criteria have a 5-fold higher risk of progressing to diabetes than those who meet only one.24,25

The findings of this study do not demonstrate a clear superiority of WHtR over WC. The association with prediabetes was stronger with WC among men but with WHtR among women. Between 5% and 10% of prediabetes patients progress to diabetes per year,26 and early diagnosis and appropriate lifestyle interventions can reduce this progression by more than 50%.27 Abdominal obesity indicators may therefore be the most appropriate anthropometric indicators for identifying the possible presence of prediabetes. However, BMI remains an important measurement for men, among whom it predicts the type of prediabetes with the highest risk of progressing to diabetes.

Compared with the general obesity indicator, abdominal obesity indicators show a stronger association with prediabetes, except among men with alterations to both fasting plasma glucose and HbA1c. The results do not show a consistent pattern of association between anthropometric indicators and the presence of HT and dyslipidemia

This study received funding from Novartis and Sanofi for the development of the telematic data-collection platform, researcher meetings, and study monitoring. Sanofi and Novartis did not participate in the study design, data analysis and interpretation, manuscript writing, or the decision to submit the manuscript for publication. This study was possible thanks to the infrastructure provided by the Fundación redGDPS (Spanish acronym for Research Network into Diabetes in Primary Health Care).

None declared.

.