Objective: Although obesity is typically associated with increased cardiovascular risk, a subset of obese individuals display a normal metabolic profile (“metabolically healthy obese,” MHO) and conversely, a subset of nonobese subjects present with obesity-associated cardiometabolic abnormalities (“metabolically obese nonobese,” MONO). The aim of this cross-sectional study was to identify the most important body composition determinants of metabolic phenotypes of obesity in nonobese and obese healthy postmenopausal women.
Design and Methods: We studied a total of 150 postmenopausal women (age 54 ± 7 years, mean ± 1 SD). Based on a cardiometabolic risk score, nonobese (body mass index [BMI] ≤ 27) and obese women (BMI > 27) were classified into “metabolically healthy” and “unhealthy” phenotypes. Total and regional body composition was assessed with dual-energy X-ray absorptiometry (DXA).
Results: In both obese and nonobese groups, the “unhealthy” phenotypes were characterized by frequent bodyweight fluctuations, higher biochemical markers of insulin resistance, hepatic steatosis and inflammation, and higher anthropometric and DXA-derived indices of central adiposity, compared with “healthy” phenotypes. Indices of total adiposity, peripheral fat distribution and lean body mass were not significantly different between “healthy” and “unhealthy” phenotypes. Despite having increased fat mass, MHO women exhibited comparable cardiometabolic parameters with healthy nonobese, and better glucose and lipid levels than MONO. Two DXA-derived indices, trunk-to-legs and abdominal-to-gluteofemoral fat ratio were the major independent determinants of the “unhealthy” phenotypes in our cohort.
Conclusions: The “metabolically obese phenotype” is associated with bodyweight variability, multiple cardiometabolic abnormalities and an excess of central relative to peripheral fat in postmenopausal women. DXA-derived centrality ratios can discriminate effectively between metabolic subtypes of obesity in menopause.
Although obesity is associated with an increased risk of diabetes and cardiovascular disease , not all obese individuals present with cardiometabolic abnormalities, and conversely, not all normal-weight subjects maintain a normal metabolic profile . The distinction between “metabolically healthy” and “unhealthy” phenotypes among nonobese and obese individuals introduces the entity of “metabolic obesity”, which is far more predictive of cardiometabolic risk than body mass index (BMI)-defined obesity .
The heterogeneous spectrum of “metabolic obesity” consists of two representative phenotypes, which have received considerable scientific attention during the last decade: the “metabolically obese nonobese” (MONO) and the “metabolically healthy obese” (MHO). The term “MONO” is used to describe a unique subset of nonobese subjects, who despite their apparently reassuring body weight, display a variety of obesity-related metabolic risk factors such as insulin resistance, atherogenic dyslipidaemia, and elevated blood pressure . On the other hand, the term “MHO” refers to a distinct subgroup of obese individuals, who despite their excess fat accumulation, appear to be protected against the development of obesity-associated cardiometabolic comorbidities, and are characterized by high insulin sensitivity, normal lipids and no signs of hypertension, inflammation, or fatty liver disease . Despite increasing clinical awareness of these phenotypes, there remains substantial residual uncertainty as to the exact constellation of their characteristics. It has been suggested that a complex interplay between genetic, metabolic, behavioral, and lifestyle factors plays an important role. However, the most prominent factor that can explain the variance in the metabolic profile of MONO and MHO individuals is body composition and regional fat distribution .
Identifying subtypes of obesity is of particular value in postmenopausal women, since menopause-related hormonal changes confer an increased cardiometabolic risk in all postmenopausal women, either they are obese or nonobese . Considering that body composition has been recognized as a major determinant of cardiometabolic risk after menopause , a strong correlation is expected to exist between body composition and metabolic subtypes of obesity in postmenopausal women. In this population, several studies have addressed the characteristics of MHO phenotype , while existing data are very scarce regarding MONO phenotype.
Dual-energy X-ray absorptiometry (DXA) constitutes a well-validated method of estimating total and regional fat and lean body mass (LBM), since it is far more accurate and reproducible than anthropometry, and far more feasible and cost-effective than computed tomography and magnetic resonance imaging . Besides, considering that the majority of postmenopausal women are advised to undergo DXA scans for measuring their bone mineral density, it seems convenient to combine their osteoporosis monitoring with an additional evaluation of body composition.
The aim of this cross-sectional study was to evaluate body composition and fat distribution in several metabolic phenotypes in a population of nonobese and obese Caucasian postmenopausal women and identify the most important body composition determinants of these phenotypes, in addition to anthropometric, clinical, and biochemical characteristics.
Methods and Procedures
We studied a total of 150 non-smoking postmenopausal women, who visited consecutively the Endocrine Unit of Attikon University Hospital for their annual osteoporosis screening between September 2009 and May 2011 and fulfilled the inclusion criteria of our study. The postmenopausal state was determined by reporting more than a year since last menstruation and follicle stimulating hormone levels were above 30 U/l. The exclusion criteria of the study were (i) current or prior use of hormone replacement treatment, (ii) clinical evidence of overt cardiovascular disease, (iii) current treatment with antidiabetic, hypolipidaemic, antihypertensive, or other medications that could interfere with body composition or cardiometabolic parameters (including corticosteroids and antidepressants), (iv) uncontrolled thyroid disease, and (v) changes of more than 5% of body weight in the previous 6 months.
The study protocol was approved by our institutional Biomedical Research Ethics Committee, and a written informed consent was obtained from all participants before the initiation of the study.
Demographic data and anthropometric measurements
A detailed medical history was obtained for all women, including data about duration of menopause, body weight changes, nutritional habits, and physical activity. The stability of body weight was assessed with the following question: “Have you experienced significant changes in your body weight during your adult life (except for the period around pregnancy) or has your weight been relatively stable?” A validated Mediterranean Diet Score, ranging from 0 to 55, was used to assess women's adherence to traditional Mediterranean dietary patterns . In addition, the short version of the International Physical Activity Questionnaire was used to assess the frequency, intensity, and duration of women's physical activity during an average week of their life .
All anthropometric measurements were performed in duplicate by the same trained physician, after women had removed their shoes and heavy clothing. Body weight was measured to the nearest 0.1 kg on a calibrated electronic scale, and height was measured using a standard wall-mounted stadiometer at the nearest 0.5 cm. BMI was calculated as body weight in kilograms divided by height in meters squared (kg/m2). Women with BMI >27 were defined as obese (OB) and those with BMI ≤27 were defined as being nonobese (NO), in accordance with a considerable number of studies using the same cut-off point to define obesity in sedentary postmenopausal women [18, 19]. Furthermore, BMI cut-off of 27 coincided in our study with the cut-off of 26.9, which has been suggested as an appropriate threshold for identifying obesity among white postmenopausal women . Waist, hip, mid-thigh, and mid-arm circumferences of non-dominant extremities were obtained using a flexible steel metric tape according to standard procedures. Waist-to-hip (WHR) and waist-to-height ratios (WHtR) were derived by dividing waist circumference (centimeters) by hip circumference and height (centimeters), respectively.
All women underwent a comprehensive assessment of cardiometabolic risk, using parameters such as blood pressure, fasting glucose homeostasis, lipid indices, liver function tests, and markers of subclinical inflammation. After an overnight fast of 12 h, venous blood samples were collected for the biochemical measurements. Fasting plasma glucose (FPG) was measured with spectrophotometry, using the method of glucose oxidase (BT T.A.R.G.A. 3000 Plus; Biotecnica Instruments S.p.A., Rome, Italy), while fasting plasma insulin (FPI) was measured with radioimmunoassay (Wallac Wizard automatic gamma counter). Based on FPG and FPI, Homeostasis Model Assessment (HOMA) index for Insulin Resistance was derived for an evaluation of insulin resistance . Lipids, uric acid, and liver transaminases were measured with standard enzymatic colorimetric techniques. Plasma fibrinogen levels were measured with the Clauss assay of clotting time, and serum levels of high-sensitivity C-reactive protein (hs-CRP) were assessed with nephelometry. In a subgroup of 83 women not taking calcium/vitamin D supplementation, serum 25-hydroxy-vitamin D levels were measured with radioimmunoassay. Systolic and diastolic blood pressure (SBP and DBP) were measured with an aneroid sphygmomanometer in women resting for at least 10 min in a sitting position. The average of three consecutive measurements at 5-min intervals was used as the final values. Mean blood pressure was estimated from the formula MBP = DBP + 1/3 × (SBP–DBP).
Based on previous studies [13, 22], a multicomponent cardiometabolic risk score was used to determine the metabolic profile of nonobese and obese women, consisting of the following abnormalities: SBP/DBP ≥130/85 mm Hg, FPG ≥100 mg/dl (5.6 mmol/l), HDL cholesterol ≤50 mg/dl (1.3 mmol/l), LDL cholesterol ≥130 mg/dl (3.4 mmol/l), triglycerides ≥150 mg/dl (1.7 mmol/l), HOMA index ≥highest tertile (2.4 for NO and 3.4 for OB), and hs-CRP ≥highest tertile (3.0 mg/l for NO and 4.7 mg/l for OB). This risk score was mainly based on the criteria proposed by Karelis et al.  and Wildman et al. , although it is important to acknowledge that there is a great heterogeneity in the criteria applied by different researchers. We opted for the approach of using a cardiometabolic risk score rather than focusing on insulin sensitivity, on the grounds that such a detailed risk score would more accurately reflect the global cardiometabolic burden associated with these phenotypes. For most parameters, we used the widely accepted cut-off values suggested by international scientific societies for defining components of metabolic syndrome . For HOMA index and hs-CRP which are greatly influenced by obesity state, different thresholds were used for NO and OB subgroups. In line with other investigators , abnormal values were defined by the highest tertile, since there are no established cut-off values for insulin resistance and inflammation in the general population.
Based on the above risk score, nonobese and obese women of our study were divided into “metabolically healthy” and “unhealthy” phenotypes as following:
- MHNO: BMI ≤27 kg/m2 and 0-1 abnormalities.
- MONO: BMI ≤27 kg/m2 and ≥2 abnormalities.
- MHO: BMI >27 kg/m2 and 0-1 abnormalities.
- Metabolically unhealthy obese (MUHO): BMI >27 kg/m2 and ≥2 abnormalities.
Whole-body and regional body composition was assessed using DXA (QDR bone densitometer, software version 12.3, Hologic Discovery-W, Bedford, MA, USA). Body composition scans were performed in a supine position, with an average time of measurement approximately 7 min, and minimal exposure to radiation. To ensure proper function of QDR system, daily quality control procedures were performed, and a step phantom calibration was performed on a weekly basis.
Total fat mass (FM), FM percent (%) and FM index (defined as FM in kilograms divided by height in meters squared) served as indicators of overall adiposity. Regional fat distribution was assessed with conventional indices provided by DXA analysis such as FM in the trunk, arms, and legs, which have been validated in postmenopausal women , as well as by additional regions of interest, which were manually determined by the same trained researcher based on prominent anatomic landmarks. In this study, the following regions of interest were evaluated: (1) the thoracic subregion, defined as a quadrilateral area extending from the sternal end of the clavicles to the lower edge of T12 thoracic vertebra, (2) the abdominal subregion, defined as a quadrilateral area extending from the upper edge of L2 lumbar vertebra to the horizontal pelvis line , and (3) the gluteofemoral subregion, defined as a quadrilateral area extending from the horizontal pelvis line to the level of the knee joints.
Central fat distribution was represented by the following indices: (i) thoracic FM (TFM), (ii) abdominal FM (AFM), (iii) trunk FM (TrFM), (iv) abdominal-to-gluteofemoral fat ratio (AFM/GFM), and (v) trunk-to-legs fat ratio (TrFM/LFM). Accordingly, peripheral fat distribution was assessed by the following indices: (i) arm FM (ArFM), (ii) leg FM (LFM), and (iii) gluteofemoral FM (GFM). Abdominal fat assessed by DXA and the centrality ratios TrFM/LFM and AFM/GFM have been widely used to assess regional fat distribution in several populations . Thoracic and gluteofemoral fat are newly proposed DXA-derived indices, which are defined based on clear-cut anatomic landmarks and it is therefore not arbitrary to assume that they reflect upper and lower body fat deposition, respectively.
The total amount and distribution of LBM, mainly composed of skeletal muscle mass, was also evaluated with DXA using the following indices: (i) total LBM, (ii) lean BMI, defined as LBM in kilograms divided by height in meters squared , (iii) trunk lean mass, indicating central distribution of lean tissue, and (iv) appendicular lean mass, indicating peripheral distribution of lean tissue.
Intra-observer reproducibility for the manually defined subregions of the study, expressed as coefficients of variation (CVs), was 2.1% for the thoracic, 3.5% for the abdominal, and 0.4% for the gluteofemoral region (n = 30). Inter-observer reproducibility for the same subregions was 6, 5.2, and 1.2%, respectively (n = 15). The in vivo reproducibility of DXA method was assessed by performing duplicate body composition scans at 1-min intervals in 10 subjects after repositioning. The estimated CVs for FM and LBM were 1.6 and 0.8%, respectively.
All numerical data were assessed for normality of their distribution with the non-parametric one-sample Kolmogorov–Smirnov test. All indices of body composition were normally distributed. The comparison of clinical and body composition parameters among the four metabolic phenotypes was performed with one-way analysis of variance (ANOVA), and Bonferroni correction was used to control for the inflation of type I error due to multiple comparisons. For data not following a normal distribution, the non-parametric variant of ANOVA (Kruskal–Wallis test) was applied, and the between-group differences were examined with Mann–Whitney U test. For categorical variables, such as history of body weight changes and level of physical activity, the significance of differences among the four phenotypes was evaluated with the χ2 test. Receiver operating characteristic (ROC) curve analysis was applied to evaluate the utility of several anthropometric and DXA-derived indices of body composition to determine metabolic phenotypes among nonobese and obese women. The independent association of clinical and body composition variables with metabolic phenotypes was assessed with stepwise multivariate binary logistic regression. P values (two-tailed) were considered statistically significant at the level of 0.05. All statistical analyses were performed with the SPSS Statistical Package (version 19.0, SPSS, Chicago, USA).
Tables 1 and 2 compare the major demographic, anthropometric, clinical, and body composition parameters of the four studied metabolic phenotypes (MHNO, MONO, MHO, and MUHO). Our total study population (n = 150) comprised 60 nonobese and 90 obese postmenopausal women.
Table 1. Demographic, anthropometric, and clinical parameters of the four studied metabolic phenotypes
|Age (years)||52 ± 6||54 ± 7||55 ± 5||55 ± 7|
|Duration of menopausea (years)||2 [0.9–4.5]||4 [1–11]||5.5 [1–9]||4 [1–11]|
|Age at menopause onseta (years)||50 [47–52]||49 [44–52]||50 [47–51]||50 [46–52]|
|Mediterranean diet score (0–55)||31 ± 4||32 ± 4||30 ± 4||30 ± 4|
|Physical activity level (low/moderate/high), %||18/47/35b||24/50/26||17/55/28||40/45/15b, c|
|History of significant body weight fluctuations, %||23.8||50d||54.5||77.1b, c|
|Body mass index (kg/m2)||23.7 ± 2||24.7 ± 2||31.6 ± 4e, f||33.5 ± 5b, g|
|Waist circumference (cm)||78 ± 6||85 ± 6d||96 ± 9e, f||101 ± 10b, c, g|
|Mid-thigh circumference (cm)||47.8 ± 3.2||47.6 ± 3.0||54.1 ± 5.9e, f||51.6 ± 5.7b, g|
|Mid-arm circumference (cm)||25.4 ± 1.7||26.2 ± 2.0||31.1 ± 3.1e, f||31.1 ± 3.3b, g|
|Waist-to-hip ratio (WHR)||0.82 ± 0.05||0.88 ± 0.06d||0.87 ± 0.05e||0.9 ± 0.06b|
|Waist-to-height ratio (WHtR)||0.5 ± 0.04||0.54 ± 0.05d||0.6 ± 0.05e, f||0.64 ± 0.07b, c, g|
|MBP (mm Hg)||85 ± 10||90 ± 8||92 ± 12||97 ± 11b, g|
|FPG (mg/dl)||86 ± 11||100 ± 10d||90 ± 11f||101 ± 12b, c|
|FPI (μIU/ml)||7.3 ± 2.2||9.9 ± 3.9||10 ± 4||13 ± 5b, g|
|HOMA index||1.6 ± 0.5||2.5 ± 1d||2.3 ± 0.8||3.1 ± 1.3b, c, g|
|Total cholesterol (mg/dl)||220 ± 24||243 ± 37||198 ± 43e||228 ± 37c|
|HDL cholesterol (mg/dl)||69 ± 15||68 ± 15||62 ± 11||58 ± 13b, g|
|LDL cholesterol (mg/dl)||136 ± 26||159 ± 31||123 ± 36e||152 ± 34c|
|Triglycerides (mg/dl)||84 ± 27||101 ± 57||84 ± 27||110 ± 38b, c|
|TC/HDL ratio||3.3 ± 0.7||3.7 ± 1.1||3.2 ± 0.6||4.1 ± 1.1b, c|
|LDL/HDL ratio||2.1 ± 0.6||2.5 ± 0.9||2 ± 0.5||2.8 ± 0.9b, c|
|Uric acid (mg/dl)||4 ± 0.8||3.8 ± 0.9||4.5 ± 1.2||4.8 ± 1b, g|
|SGOT (U/l)a||17.5 [14–18]||19 [16–23]||17 [16–20]||18 [16–21]|
|SGPT (U/l)a||14.5 [12–18]||17 [14–23]d||19 [16–23]||20 [17–29]e|
|γ-GT (U/l)a||14 [12–15]||14 [12–17]||14 [11–16]||17 [14–25]b, c|
|Plasma fibrinogen (mg/dl)||313 ± 47||366 ± 70d||364 ± 67||385 ± 75e|
|Serum hs-CRP (mg/l)a||0.7 [0.3–1.4] 26.4 ± 12||2.9 [1–3.3]d 30 ± 14||1.8 [0.7–2.8] 35 ± 9||3.4 [2.3–6.1]b, c, g|
|Serum 25-OH-Vit D (ng/ml)h||26.4 ± 12||30 ± 14||35 ± 9||28 ± 8|
Table 2. Body composition characteristics of the four studied metabolic phenotypes
|Total FM (kg)||20.7 ± 3.9||22.3 ± 3.7||32.8 ± 6.8a, b||36.1 ± 8.8c, d|
|FM (%)||35 ± 4||35.9 ± 4||41.5 ± 4a, b||43.2 ± 4.1c, d|
|FMI (kg/m2)||8.3 ± 1.6||8.8 ± 1.3||13 ± 2.5a, b||14.5 ± 3.4c, d|
|Central fat distribution|
|TrFM (kg)||8 ± 2.1||9.8 ± 2.2||15.1 ± 3.7a, b||17.4 ± 4.8c, d|
|AFM (kg)||1.5 ± 0.4||2 ± 0.6||3.2 ± 0.9a, b||4 ± 1.2c-e|
|TFM (kg)||3.1 ± 1||4.3 ± 1.3f||6.4 ± 1.8a, b||7.9 ± 2.4c-e|
|TrFM/LFM ratio||0.85 ± 0.2||1.14 ± 0.3f||1.19 ± 0.2a||1.37 ± 0.4c-e|
|AFM/GFM ratio||0.16 ± 0.03||0.2 ± 0.05f||0.22 ± 0.04a||0.26 ± 0.05c-e|
|Peripheral fat distribution|
|ArFM (kg)||2.6 ± 0.6||2.8 ± 0.6||3.8 ± 0.7a, b||4.4 ± 1.2c-e|
|LFM (kg)||9.4 ± 1.6||8.8 ± 1.7||13 ± 3.3a, b||13.2 ± 3.8c, d|
|GFM (kg)||9.9 ± 1.8||9.9 ± 1.7||15 ± 3.9a, b||15.5 ± 4.2c, d|
|Lean body mass|
|LBM (kg)||36.2 ± 3||37.7 ± 4||43.8 ± 5.9a, b||44.5 ± 5.2c, d|
|LBMI (kg/m2)||14.4 ± 0.9||14.9 ± 0.9||17.4 ± 1.9a, b||17.8 ± 1.7c, d|
|TrLM (kg)||17.9 ± 1.7||18.6 ± 1.9||21.7 ± 2.9a, b||21.9 ± 2.7c, d|
|AppLM (kg)||15 ± 1.3||15.7 ± 2.1||18.4 ± 3a, b||18.9 ± 2.6c, d|
(i) MONO versus MHNO
The majority of nonobese women were classified as MONO (n = 38, 63.3%), while only 22 women (36.7%) had the MHNO phenotype.
Women with the MONO phenotype reported more often body weight fluctuations (P = 0.04) compared with MHNO, while no significant differences were observed in age, duration, and age at menopause, level of physical activity, and degree of adherence to the Mediterranean diet (Table 1).
BMI values did not differ significantly between MONO and MHNO phenotypes. However, MONO women displayed significantly higher anthropometric indices of central adiposity such as waist circumference (P = 0.02), WHR (P = 0.001), and WHtR (P = 0.04), while there were no significant differences in mid-thigh and mid-arm circumferences (P = 0.9). In addition, MONO women exhibited a significantly worse cardiometabolic profile than MHNO, as expected by definition (Table 1).
Indices of total body fatness, LBM, and peripheral fat distribution were not significantly different between MONO and MHNO phenotypes, as shown in Table 2. However, for the same degree of overall and peripheral adiposity, women with the MONO phenotype displayed significantly higher DXA-derived indices of central fat distribution, such as TFM (P = 0.02), TrFM/LFM (P = 0.003), and AFM/GFM ratio (P = 0.006), compared with MHNO.
(ii) MUHO versus MHO
The majority of obese women were classified as being MUHO (n = 68, 75.6%), while only 22 women (24.4%) had the MHO phenotype.
Women with the MUHO phenotype reported a lower level of physical activity (P = 0.04) and experienced more commonly significant body weight changes (P = 0.046) compared with MHO. No significant differences were observed between the two groups in terms of age, age at menopause onset, duration of menopause, and adherence to the Mediterranean diet, as demonstrated in Table 1.
The two groups did not differ significantly in BMI values, mid-thigh, and mid-arm circumferences, but MUHO women displayed higher anthropometric indices of central fat distribution, such as waist circumference (P = 0.01) and WHtR (P = 0.04). Furthermore, MUHO women exhibited significantly higher cardiometabolic risk factors than MHO and higher serum 25-hydroxy-vitamin D levels, although this difference did not reach statistical significance.
Indices of total adiposity and LBM did not show significant differences between MUHO and MHO phenotypes. For the same degree of overall adiposity, MUHO women displayed significantly higher indices of upper body fat distribution, such as AFM (P = 0.02), TFM (P = 0.01), and AFM/GFM ratio (P = 0.01), compared with MHO (Table 2). Indices of lower body adiposity such as LFM and GFM were not significantly different between MUHO and MHO groups. The only index of peripheral adiposity which differed significantly between the two groups was ArFM, which was significantly higher in the MUHO group (P = 0.04).
(iii) Nonobese versus obese phenotypes
MHNO versus MHO/MUHO
Both obese phenotypes displayed higher anthropometric and DXA-derived indices of total and regional fat and LBM than MHNO women. No significant differences were found between MHNO and MHO phenotypes in cardiometabolic risk factors, age, menopause duration, diet, physical activity, and history of body weight changes. On the contrary, MUHO women reported lower physical activity and greater body weight variability than MHNO and displayed a significantly worse cardiometabolic profile, consisting of higher MBP, FPG, FPI, HOMA index, hepatic aminotransferases, atherogenic lipid ratios, uric acid, fibrinogen ,and hs-CRP (P < 0.05 for all).
MONO versus MHO/MUHO
Both obese phenotypes displayed higher anthropometric and DXA indices of fat and LBM than MONO, except for the ratios TrFM/LFM and AFM/GFM, which were not significantly different between MONO and MHO groups. MHO women exhibited significantly lower levels of FPG (P = 0.01), total and LDL cholesterol (P < 0.001) than MONO, while all the other cardiometabolic parameters did not show significant differences between the two groups. On the other hand, MUHO women displayed significantly higher levels of MBP (P = 0.008), FPI (P = 0.03), HOMA index (P = 0.02), uric acid (P < 0.001), and hs-CRP (P = 0.002), as well as lower HDL cholesterol levels (P = 0.001), compared with MONO.
(iv) ROC curve analysis and multivariate regression
Based on ROC curve analysis, two DXA-derived ratios of central-to-peripheral fat distribution displayed the highest area under the curve (AUC) for detecting the “metabolically unhealthy” phenotypes among nonobese and obese women of our study. As shown in Table 3, TrFM/LFM ratio had the greatest AUC for predicting the MONO phenotype among nonobese women (AUC: 0.817 ± 0.06, P < 0.001), and AFM/GFM ratio had the greatest AUC for predicting the MUHO phenotype among obese women (AUC: 0.706 ± 0.06, P = 0.004). The results of stepwise multivariate regression were in total agreement with ROC curve analysis, regarding the significant contribution of TrFM/LFM and AFM/GFM ratios (Table 4). Among indices of fat distribution which were strongly correlated with each other, we included those indices which provided the greatest AUC in the ROC curve analysis, in order to minimize collinearity between independent variables. As shown in Table 4, TrFM/LFM ratio was the only independent determinant of the MONO phenotype (P = 0.001), classifying correctly 76.3% of nonobese women into MONO/MHNO. As far as MUHO phenotype is concerned, the only independent predictor was AFM/GFM ratio (P = 0.01), classifying correctly 76.9% of obese women into MHO/MUHO.
Table 3. ROC curve analysis for detecting the “metabolically unhealthy” phenotypes among non-obese and obese women of our study
|BMI Waist circumference||0.667 ± 0.07 0.785 ± 0.06||0.53–0.81 0.66–0.91||0.04 <0.001||0.632 ± 0.07 0.669 ± 0.07||0.49–0.77 0.54–0.8||0.06 0.02|
|WHR||0.788 ± 0.06||0.67–0.91||<0.001||0.647 ± 0.07||0.51–0.78||0.04|
|WHtR||0.758 ± 0.07||0.63–0.89||0.001||0.699 ± 0.06||0.58–0.82||0.005|
|FM||0.615 ± 0.08||0.47–0.76||0.1||0.603 ± 0.07||0.46–0.74||0.1|
|TrFM||0.717 ± 0.07||0.58–0.85||0.006||0.638 ± 0.07||0.5–0.77||0.05|
|AFM||0.752 ± 0.07||0.62–0.88||0.001||0.67 ± 0.06||0.55–0.8||0.02|
|TFM||0.759 ± 0.06||0.63–0.89||0.001||0.692 ± 0.07||0.56–0.83||0.007|
|TrFM/LFM ratio||0.817 ± 0.06||0.71–0.93||<0.001||0.649 ± 0.06||0.53–0.77||0.04|
|AFM/GFM ratio||0.777 ± 0.06||0.65–0.9||<0.001||0.706 ± 0.06||0.58–0.83||0.004|
|LBM||0.63 ± 0.07||0.49–0.78||0.1||0.545 ± 0.08||0.39–0.7||0.5|
Table 4. Stepwise multivariate logistic regression analysis for the “metabolically unhealthy” phenotypes among non-obese and obese women
Our study compared a variety of clinical, biochemical, and body composition parameters in “metabolically healthy” and “unhealthy” phenotypes of nonobese and obese postmenopausal women. Compared with “healthy” phenotypes, the “unhealthy” phenotypes were associated with body weight fluctuations, central fat distribution and elevated markers of insulin resistance, hepatic steatosis, and subclinical inflammation, in both nonobese and obese women. Focusing on body composition, our data have shown that total adiposity, peripheral fat distribution, and LBM were not able to discriminate effectively between metabolic phenotypes. The most important body composition determinant of cardiometabolic risk was the relative excess of central-to-peripheral fat, assessed by two easy-to-obtain DXA-derived indices, TrFM/LFM, and AFM/GFM.
The exact prevalence of metabolic phenotypes of obesity in postmenopausal women remains unclear, due to the lack of a standardized definition and the heterogeneity of existing criteria. In the general population, the prevalence of MONO phenotype ranges between 5 and 45% depending on the criteria used, age, BMI, and ethnicity , while the prevalence of MHO phenotype is approximately 30% . In our study, there was an increased prevalence of “metabolically unhealthy” phenotypes, both in nonobese and obese women (63.3 and 75.6%, respectively), which is consistent with the postmenopausal state of our study population, since menopause is associated with increased cardiometabolic risk, regardless of body weight .
A recurrent characteristic of “unhealthy” phenotypes in both nonobese and obese women was the history of body weight fluctuations during adult lifetime. Although the question posed may be considered simplistic since it does not capture the exact magnitude of body weight changes and relies on self-reported data, it is actually a simple convenient question, reflecting the well-established metabolic consequences of weight cycling or else yo-yo phenomenon . The level of physical activity was also found to differ significantly between “healthy” and “unhealthy” phenotypes in the obese women of our study, lending support to existing evidence .
Concerning body composition, this study has shown that upper body fat accumulation constitutes a potent determinant of metabolic phenotypes of obesity, in both obese and nonobese postmenopausal women. Our findings add to the existing strong evidence that central adiposity is a prominent body composition factor, which is associated with “metabolically obese” phenotypes, irrespective of BMI [4, 12, 34]. We showed that conventional indices such as trunk fat were not significantly different between “healthy” and “unhealthy” phenotypes. On the contrary, more specific indices based on manually defined subregions, such as thoracic fat and AFM/GFM, were more informative, supporting previous observations of our group . Among all examined indices of body composition, two DXA-derived ratios reflecting the excess amount of central in relation to peripheral fat emerged as the most informative parameters for detecting the “metabolically unhealthy” phenotypes in our cohort. TrFM/LFM and AFM/GFM were the most significant independent determinants of MONO and MUHO phenotypes in multivariate regression, classifying correctly nearly 80% of studied nonobese and obese women into “healthy” and “unhealthy” phenotypes. To the best of our knowledge, our study is the first to suggest that central-to-peripheral fat ratios obtained by DXA analysis are the major body composition determinants of “metabolically unhealthy” phenotypes in postmenopausal women. Our findings substantiate the existing body of evidence, that centrality ratios are more accurate predictors of cardiometabolic risk than central fat indices alone, unadjusted for peripheral adiposity [31, 32].
Measures of overall adiposity did not differ significantly between “healthy” and “unhealthy” phenotypes, confirming a large number of studies showing comparable total body fatness in MHO and MUHO postmenopausal women [12, 14]. In contrast, studies in younger women have shown an increased amount of FM in normal-weight women with the MONO phenotype [5, 6]. The discordant findings could be explained by the marked differences in body composition between pre- and postmenopausal women. Since menopause transition increases fat accumulation in all postmenopausal women, either they are normal weight or obese, total FM may lack sufficient discriminant power to differentiate between phenotypes in nonobese women.
Regarding peripheral adiposity, we could not find any significant differences in peripheral fat indices between metabolic phenotypes, indicating a minor contribution of peripheral fat to “metabolic obesity” after menopause. Although many studies have shown protective effects of lower body adiposity on cardiometabolic parameters in postmenopausal women , very few data actually suggest peripheral fat as a potent body composition determinant of metabolic phenotypes of obesity, as defined in this study. The only peripheral fat depot, which differed significantly between metabolic phenotypes of obese women, was FM in the upper extremities. This finding is consistent with previous data, suggesting unfavorable metabolic effects of ArFM in postmenopausal women, due to its upper body localization and strong positive correlation with central fat .
Although some studies have shown that LBM is significantly increased in the MUHO phenotype of postmenopausal women [12, 36], and significantly decreased in the MONO phenotype of young women , we found in this study no significant differences in LBM indices between metabolic phenotypes of either obese or nonobese postmenopausal women. This lack of significance indicates an overriding effect of central fat distribution on metabolic risk in our study population, outweighing the potential contribution of other body composition parameters, such as LBM. Beyond that, the observed discrepancy might be partly explained by the different criteria used by several investigators to define “healthy” and “unhealthy” phenotypes, making thus their results difficult to compare.
Our study has also performed comparisons between nonobese and obese phenotypes. It was found that MHO women, despite their excess adiposity, displayed comparable cardiometabolic parameters with MHNO women, and even better glucose and lipids, compared with MONO. Furthermore, MUHO women displayed higher blood pressure, insulin resistance, uric acid, and hs-CRP compared with MONO, indicating that concurrent existence of “metabolic” and BMI-defined obesity, confers a higher degree of cardiometabolic dysregulation, compared with “metabolic obesity” alone.
The major limitations of this study include the inability of DXA technique to differentiate between subcutaneous and visceral fat, the potential recall bias related to self-reported information about body weight changes, and the lack of more sophisticated methods of estimating insulin sensitivity such as the clamp technique due to feasibility reasons. On the other hand, our study is strengthened by the following: we performed a comprehensive evaluation of total and regional body composition through a variety of upper and lower body depots, and tried to identify the most important determinants of metabolic phenotypes in a considerable number of healthy postmenopausal women. Furthermore, we provided for the first time a detailed characterization of clinical and body composition characteristics of MONO postmenopausal women.
In conclusion, our data indicate that the “metabolically obese” phenotype is associated with fasting hyperglycaemia, insulin resistance, low-grade inflammation, hepatic steatosis, central fat distribution, and temporal body weight variability. This study has also shown that two DXA-derived ratios of central-to-peripheral fat distribution, TrFM/LFM, and AFM/GFM are the major determinants of the “metabolically obese” phenotype among both obese and nonobese postmenopausal women. These ratios can help discriminate between “high risk” and “low risk” phenotypes among postmenopausal women and may facilitate informed therapeutic decision making, so that appropriate interventions are preferentially targeted to those postmenopausal women expected to benefit the most.