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


Menopause. 2021;28(6):678-685. 

In This Article


Study Participants

A total of 277 women (59.7 ± 5.4 y) with complete data for anthropometric measurements and BF% measured by DXA and who participated in a weight-loss program from five studies were pooled.[20–23] All participants were White and living in the Montreal, Sherbrooke, or Ottawa areas in Quebec and Ontario (Canada) (Table 1).

Anthropometric Indexes

Relative Fat Mass Index (RFM). RFM was calculated using the equation developed by Woolcott & Bergman in 2018 and expressed in kg:

Height and WC are expressed in meters while sex is categorized as 0 for men and 1 for women.

Body Mass Index (BMI) and Waist Circumference (WC). Body weight (± 0.2 kg) was measured using an electronic scale (SECA 707, Hamburg, Germany) and height (± 0.1 cm) with a wall stadiometer (Takei, Tokyo, Japan). BMI was calculated using the following formula: BMI = weight (kg)/height (m)2. WC (cm) was measured according to the Canadian Society for Exercise Physiology criteria (midpoint between the costal inferior border and the iliac crest).[24]

Body Fat Percentage (BF%). Total BF% (including visceral and subcutaneous fat) was measured using Prodigy DXA (DXA; GE, Prodigy Lunar, Madison, WI) and iDXA (iDXA, GE Healthcare, Chicago, IL). It has been shown that there was no significant difference between Prodigy DXA and iDXA to estimate BF%.[25] All DXA data were entered by two separate persons, independently of investigators. BF% was then estimated based on linear regression models using RFM, BMI, or WC [lm (DXA-BF% ~ RFM-BF% or BMI-BF%, or WC-BF%)] at baseline and after weight-loss intervention. Changes in measured and estimated BF% after weight-loss intervention were calculated using the following equation: Delta BF% = [(postintervention BF% − baseline BF%)/baseline BF%)] × 100.

Identification of Postmenopausal Women Living With Obesity. The World Health Organization BF% cutoff (BF% > 35) was used to identify postmenopausal women living with obesity.[26]

Statistical Analyses

Characteristics of participants are presented as mean ± SD and N (%) for continuous and categorical variables, respectively. Shapiro-Wilk test was performed to assess the normality of continuous variables distribution. The relationship between BF% and anthropometric indices (RFM, BMI, and WC) was also tested, by fitting a linear model and nonlinear models (quadratic and cubic). Then, ANOVA test was performed to assess whether the nonlinear models explain a significantly larger amount of variance compared with linear models. Overall, there is no difference between linear and nonlinear models, suggesting that the relationship between BF% and anthropometric indices is linear. All analyses were performed using R v3.5.0 software, with statistical significance set at P ≤ 0.05.

Assessment of Correlation Between Anthropometric Indexes and DXA-BF%

The correlation between anthropometric indexes (RFM, BMI, and WC) and BF% measured by DXA was quantified using pairwise Pearson correlation tests.

Comparison Between Anthropometric Indexes to Estimate DXA-BF%

Linear Regression Models. Linear regression models were performed to predict BF% measured by DXA before a weight-loss intervention using the RFM, BMI, and WC as predictors. The coefficient of determination (r 2) and the root mean squared error (RMSE) were used as performance indicators to compare accuracy between the anthropometric indexes to predict BF%. To predict change of BF% measured by DXA after a weight-loss intervention, linear regression models were also performed, and results are presented as estimate coefficient with 95% CI.

The Lin's Concordance Correlation Coefficient (CCC). The Lin's CCC was used to quantify the degree of agreement between BF% measured by DXA and BF% predicted by each anthropometric index (RFM vs BMI vs WC). Briefly, the Lin's CCC is calculated based on a correction factor bias (Cb) and Pearson correlation coefficient. The Cb is considered an indicator of accuracy that quantifies how far the best-fit line deviates from the 45° line, whereas the Pearson correlation coefficient is considered an indicator of precision. A Lin's CCC value of 0 can be interpreted as no agreement between two methods of measurements, while the value of 1 represents a perfect agreement.[27]

Bland and Altman Analyses. Bland and Altman analyses (1986) were used to assess the level of agreement between BF% measured by DXA and BF% predicted by each anthropometric index (RFM vs BMI vs WC). Briefly, the Bland and Altman analyses quantify the difference between two quantitative methods, which is different from correlation and regression models that quantify the association between two continuous variables.

The Bland and Altman plot was used, first, to evaluate the bias (difference between the two methods of measurement) and, second, to estimate an agreement limit, within which 95% of the differences of the two methods of measurement varies. If the BF% values estimated by the anthropometric indexes are equal to the values of BF% provided by DXA, the bias equals zero, and we can conclude that the two methods of measurement are equivalent. A difference between two methods of measurement is considered significant when: 1) the bias is different from zero and 2) the line of zero is not included in the 95% confidence interval of the mean difference between the two methods. For more information about Bland and Altman analyses, refer to the article by Giavarina.[28]

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

Receiver Operating Characteristic (ROC). Curves were used to assess sensitivity and specificity of each anthropometric index (RFM vs BMI vs WC) to correctly identify obese and nonobese postmenopausal women. The sensitivity [N true positives/(N true positives + N false negatives)] is considered the probability to correctly identifying obese postmenopausal women (true positive), while the specificity [N true negatives/(N false positives + N true negatives)] is considered the probability to correctly identifying nonobese postmenopausal women at a given BF% cutoff (DXA-BF%: women: > 35%).[26] The anthropometric index with the highest sum between sensitivity and specificity was considered as the best tool to correctly identifying obese and nonobese postmenopausal women.