This is a cross-sectional descriptive survey. The study was conducted at Sagamu and Remo North local government area of Ogun State, Nigeria. The locations were chosen by convenience sampling. Approval for the study was obtained from the ethic and research committee of the Olabisi Onabanjo University Teaching Hospital, Sagamu, Ogun State. The consent of the king and community leaders was also sought. The consent of each participant was also obtained. The participants were apparently healthy individuals aged 20 - 70 years. The participants included market men and women, artisans, farmers, drivers, and civil servants. Initial sensitization visits were paid to the communities, and the people that met the inclusion criteria were told to meet at designated points such as motor parks, town hall or the king’s palace for screening. All the people who gave their consent were included in the study. Individuals who are wheelchair bound, pregnant women and lactating mothers were excluded.

Trained assistants helped with data collection and measurements.

Data were analysed using the statistical package for social sciences (SPSS) version 20.0 (Chicago, Illinois, USA). Continuous variables were expressed as means and standard deviation. Pearson correlation between BMI and the three central obesity indices was determined. MedCalc Statistical Software version 14.8.1 (MedCalc Software, Ostend, Belgium, 2014) was used to generate Receiver Operative Characteristics (ROC) Curve, in order to determine and compare the ability of the measures of central obesity to identify persons with generalized obesity and systemic hypertension. The area under curve (AUC) on the ROC was used to determine the best measure of central obesity which identifies individuals with generalized obesity and systemic hypertension.

Among the male participants, WC and WHtR correlated better with BMI compared with WHR , (r = 0.860, p < 0.001 vs r = 0.490, p < 0.001). Among the female participants, the correlation between indices of central and general obesity was best with WC, (r = 0.895, p < 0.001). There was a strong correlation among the three indices of central obesity (Table 2).

Data are means (SD), BMI, body mass index; WC, waist circumference; WHtR, waist-to-height ratio; WHR, waist-to-hip ratio; DBP, diastolic blood pressure; SBP, systolic blood pressure.

Figure 1 shows the Area under the Curve (AUC) of WC, WHR, and WHtR on the Receiver Operating Characteristics (ROC) curve (of male participants) for predicting general obesity and systemic hypertension. Figure 2 shows the AUC of WC, WHR, and WHtR on the ROC curve (of female participants) for predicting general obesity and systemic hypertension. In both genders, the areas were significantly greater than 0.5 (half of the total area) for the three obesity indices (p < 0.001). The AUC of WC and WHtR were however greater than that of WHR.

The values of the Areas Under the Curve (AUC) for predicting the presence of general obesity, hypertension, or general obesity/hypertension are shown in Table 3. In most cases the absolute value of AUC of WHtR was greater than those of WC and WHR. Among the three indices of obesity waist-to-hip ratio has the smallest AUC, suggesting that it predicted cardiovascular risk factors less than WC or WHtR.

Table 4 compares the AUC for detecting both general obesity and systemic hypertension. It showed that WHtR and WC were better than WHR at detecting the presence of both general obesity and hypertension in both males (WHtR and WHR {difference in areas = 0.131} p < 0.0001; WC and WHR {difference in areas = 0.132} p < 0.0001), and females (WHtR and WHR {difference in areas = 0.214} p < 0.0001; WC and WHR {difference in areas = 0.205} p < 0.0001). The AUC of WHtR and WC on the ROC for identifying the presence of both general obesity and hypertension were not significantly different in both genders (p > 0.05).

BMI, body mass index; WC, waist circumference; WHtR, waist-to-height ratio; WHR, waist-to-hip ratio. p < 0.001 for all the correlations.

p < 0.001 for all the AUCs, AUC: area under the curve, CI: confidence interval.

The AUC of WHtR and WC on the ROC for identifying the presence of both general obesity and hypertension were not significantly different in both genders (p > 0.05).

Consistent with previous reports [25] , there was a good correlation between body mass index and the indices of central obesity. Among the female participants, WC and WHtR correlated equally with BMI but both correlated better than WHR. This observation is similar to the report by previous authors [26] . This is not unexpected since both indices are measures of adiposity.

The ROC revealed that whereas, WHtR has the largest AUC in most instances, WHR has the smallest AUC. This suggests that WHtR identified individuals with cardiovascular risk factors better than WC and WHR. However subjecting the differences in the AUC to further analysis revealed that there was no significant difference between the AUC of WHtR and WC, whereas the AUC of WHR was significantly smaller than that of the other two indices of central obesity. This indicates that WHtR was as good as WC, and both were better than WHR in identifying individuals with cardiovascular risk factors. This is true in both men and women for identification of both general obesity and hypertension.

Previous studies that compared the ability of indices of central obesity to identify CVD risk factors yielded inconsistent results. Similar to our findings, a Nigerian study which compared the ability of WC and WHR to predict CVD risk factors found that WC was better than WHR [19] . However, the above study did not test the predictive ability of WHtR.

Bener et al., [17] in a study conducted in Qatar found that WC performed better than WHR and WHtR in identifying people with metabolic syndrome in both men and women. However, some researchers [16] [27] [28] [29] found that WHR predicted CVD risk factors better that WC in Turkey and Australia. Other studies conducted in Iran, Jordan, China, Chinese Taipei and Europe found that WHtR either identified [18] [30] , or predicted [15] [31] [32] cardiovascular risk factors than other central obesity indices.

Contrary to the above reports none of the three indices of obesity were found to be superior in predicting CVD by other workers in America and New Zealand [7] [33] . These inconsistencies may be due to different geographical locations and or methodologies. Indeed, it has been observed that ethnicity and race influence the cut-off points and/or use of different anthropometric indices to diagnose obesity and metabolic syndrome [34] . Ethnicity significantly affects the associations between anthropometric indices used to assess adiposity such as BMI and WC. This underscores the need for ethnic specific cut-off for WC in the definition of metabolic syndrome proposed by the IDF [9] . Lower cut-off for BMI was also proposed for the Asians who have smaller body frames, are more insulin resistant, and have greater body fat for the same BMI than the Caucasians [35] -[37] . Furthermore, ethnicity influences the association between anthropometric obesity measures and cardiovascular risk factors [38] . Molarius et al., therefore, suggested that the predictive power of anthropometric indices is ethnic and population-dependent [14] . In view of the above, it has also been suggested that ethnic-specific assessment, prevention and treatment of CVD risk be developed [8] [39] .

Apart from geographical locations, methodological differences in the above studies may also contribute to different conclusions. While some studies were prospective, others were cross-sectional in design.