Obesity Among Postmenopausal Women

What Is the Best Anthropometric Index to Assess Adiposity and Success of Weight-loss Intervention?

Ahmed Ghachem, PhD; Alexis Marcotte-Chénard, MSc; Dominic Tremblay, MSc; Denis Prud'homme, MD, MSc; Rémi Rabasa-Lhoret, MD, PhD; Eléonor Riesco, PhD; Martin Brochu, PhD; Isabelle J. Dionne, PhD

Disclosures

Menopause. 2021;28(6):678-685. 

In This Article

Results

Characteristics of Participants at Baseline

Characteristics of participants are presented in Table 2. Overall, 277 women were included in these analyses. Even if participants were obese (BMI = 31.2 ± 4.9 kg/m2; DXA-BF% = 43.8 ± 5.4%), the majority had a generally healthy cardiometabolic profile, with only 27% having metabolic syndrome (based on the harmonized definition[29]).

Comparison Between Anthropometric Indexes to Estimate DXA-BF% Before Weight-loss Intervention

Coefficient of Correlation. Figure 1 shows the coefficient of correlation between each anthropometric index and DXA-BF%. BMI (r = 0.73; P < 0.001) displayed the highest correlation index with DXA-BF%, followed by RFM (r = 0.51; P < 0.001) and WC (r = 0.49; P < 0.001) although they were all significant.

Figure 1.

Correlation between anthropometric indexes and DXA-BF% before a weight-loss intervention. BMI, body mass index (kg/m2); DXA-BF%, percentage of body fat mass measured by DXA; RFM, relative fat mass (kg); WC, waist circumference (cm). Darker background color indicates stronger correlation. Pairwise Pearson correlation tests were performed, and all coefficients of correlation were significant at P < 0.001.

Linear Regression Models. Results of linear regression models are presented in Figure 2. BMI (r 2 = 0.52; RMSE = 3.7%; P < 0.001) was the best predictor of DXA-BF% compared with RFM (r 2 = 0.27; RMSE = 4.4%; P < 0.001) and WC (r 2 = 0.25; RMSE = 4.8%; P < 0.001).

Figure 2.

DXA-BF% estimation using anthropometric indexes before a weight-loss intervention. BMI, body mass index (kg/m2); DXA-BF%, percentage of body fat mass measured by DXA; R2, coefficient of determination; RFM, relative fat mass (kg); RMSE, root mean squared error; WC, waist circumference (cm). Linear regression models were performed.

Bland and Altman Analyses/Lin's CCC. The mean bias (quantitative difference between the two methods of measurement) between DXA-BF% and RFM-predicted BF% (0.8 [0.3, 1.4]) was higher compared with BMI-predicted BF% (−0.07 [−0.5; 0.3]) and WC-predicted BF% (−0.07 [−0.6; 0.4]), suggesting that the level of agreement between BF% measured by DXA and BF% predicted by BMI and WC is higher than with RFM. In addition, the difference between DXA-BF% and BF% estimated by RFM was statistically significant (the line of zero is not included in the 95% confidence interval) compared to BMI- and WC-BF% (the line of zero is included in the 95% confidence interval) (Figure 3).

Figure 3.

Bland-Altman plot to quantify agreement between DXA-BF% and BF% estimated by anthropometric indexes before a weight-loss intervention. The bold gray continuous line of zero corresponds to a perfect agreement. Dashed lines indicate bias (represented by the gap between the "zero" line and the mean of the difference) and 95% limits of agreement. Lin's concordance correlation coefficient (CCC) and accuracy (C b) are shown. BMI, body mass index (kg/m2); DXA-BF%, percentage of body fat mass measured by DXA; RFM, relative fat mass (kg); WC, waist circumference (cm).

Results from Lin's CCC are in line with those from Bland and Altman analyses, showing higher coefficients for BMI (Lin's CCC = 0.71) compared with RFM (Lin's CCC = 0.46) and WC (Lin's CCC = 0.41), suggesting a better agreement between DXA-BF% and BF% predicted using BMI compared to RFM and WC (Figure 3).

Comparison Between Anthropometric Indexes to Estimates Change in DXA-BF% After a Weight-loss Intervention

Linear regression models. Change in all anthropometric indexes was significantly associated with change in DXA-BF% after a weight-loss intervention. However, change in BMI (β=1.03 [0.87; 1.20]) had a better predictive power compared with RFM (β=0.65 [0.46; 0.84]) and WC (β=0.20 [0.13; 0.27) (Figure 4).

Figure 4.

Relationship between change in DXA-BF% and change in anthropometric indexes after a weight-loss intervention. Estimations (points) together with 95% CIs (segments) for relationships between body fat change (post- pre-intervention) and changes in anthropometrics index. Linear regression models were performed. When 95% CIs (segments) include the null value ("0", dash line) association is not significant (P > 0.05).

Bland and Altman Analyses/Lin's CCC. Based on results from Bland and Altman analyses as well as Lin's CCC, the level of agreement between the change in DXA-BF% after a weight-loss intervention and the change in BF% predicted by BMI (mean bias = −0.00004 [−3.87; 3.87]; Lin's CCC = 0.60) was better than RFM (mean bias = −0.24 [−4.98; 4.49]; Lin's CCC = 0.39) and WC (mean bias = 2.12 [−6.65; 10.90]; Lin's CCC = 0.26) (Figure 5).

Figure 5.

Bland-Altman plot showing limits of agreement between postintervention change of body fat percentage measured by DXA (Delta DXA-BF%) versus those estimated by the postintervention changes in RFM, BMI, and WC. The bold gray continuous line of zero corresponds to a perfect agreement. Dashed lines indicate bias (represented by the gap between the "zero" line and the mean) and 95% limits of agreement. Lin's concordance correlation coefficient (CCC) and accuracy (C b) are shown. BMI, body mass index; RFM, relative fat mass; WC, waist circumference.

Comparison Between Anthropometric Indexes to Correctly Identify Postmenopausal Women Living With Obesity

ROC curves analyses showed that BMI (sensitivity + specificity = 193) was the best anthropometric index to correctly identify postmenopausal women living with obesity (cut-off BF% > 35) compared with RFM (sensitivity + specificity = 152) and WC (sensitivity + specificity = 158) (Figure 6). BMI also had lower misclassification error (6%) compared with the other anthropometric indexes (data not shown).

Figure 6.

Anthropometric indexes to correctly identify postmenopausal women living with obesity (BF% > 35). AUC, area under curve (95% CI); BMI, body mass index (kg/m2); DXA-BF%, percentage of body fat mass measured by DXA; RFM, relative fat mass (kg); Se, sensitivity: probability of correctly identifying obese postmenopausal women having DXA-BF% > 35%; Sp, specificity: probability of correctly identifying postmenopausal women having DXA-BF% < 35%; WC, waist circumference (cm).

Comments

3090D553-9492-4563-8681-AD288FA52ACE

processing....