Progression From Metabolically Benign to At-Risk Obesity in Perimenopausal Women: A Longitudinal Analysis of Study of Women Across the Nation (SWAN)

Abstract

Background:

Little is known about the natural history of progression from a metabolically benign overweight/obese (MBO) to at-risk overweight/obese (ARO) phenotype. Improved understanding would help clinicians focus on controlling risk factors that predispose an obese individual to progression.

Methods:

Using discrete-time proportional hazard modeling on data from the Study of Women's Health Across the Nation (SWAN), we examined the incident progression from MBO (less than two metabolic syndrome abnormalities) to ARO (two or more metabolic syndrome abnormalities) and factors associated with progression over a 7-year period.

Results:

Of 866 MBO women at baseline, 43% progressed to the ARO phenotype. Compared with those who remained MBO, those who progressed had higher baseline BMI and a higher prevalence of cardiometabolic abnormalities (elevated glucose, triglycerides, blood pressure and low high-density lipoprotein cholesterol). In multivariable analyses, an increase in body mass index was associated with a modest increase in the risk of progression. Although all cardiometabolic abnormalities were associated with an increased risk, the baseline impaired fasting glucose showed the strongest association with the risk of progression [hazard ratio 3.24; 95% confidence interval 2.10, 4.92; P < .001]. Physical activity played a protective role in decreasing the risk of progression [hazard ratio 0.86; 95% confidence interval 0.80, 0.92; P < .001].

Conclusions:

Increasing obesity and the presence of cardiometabolic abnormalities increase the risk of progression, whereas physical activity is the only lifestyle factor protective against progression from metabolically benign to the at-risk overweight/obese phenotype, a state that is unanimously associated with an elevated risk of cardiovascular morbidity and mortality.

Despite the heterogeniety in cardiometabolic abnormalities in obese individuals, the difference in cardiovascular disease (CVD)-related morbidity and mortality between metabolic phenotypes remains controversial (1–5). Many studies report a comparable CVD risk between normal weight and metabolically benign obese individuals over a 3- to 15-year period (1, 2, 5, 6), whereas others report an increased 16-year risk with obesity, even in the absence of comorbidities (7). In addition, we (8) and others (3, 9) have shown that compared with normal weight and at-risk obese adults, metabolically benign overweight/obese (MBO) adults have intermediate levels of subclinical atherosclerosis and adipokines (10). These findings suggest that MBO individuals do have an increased risk of incident CVD compared with normal-weight individuals, but they may experience a delay in the onset of clinical CVD compared with their at-risk counterparts.

Given the reported difficulties in weight loss maintenance over time, sustained status in the MBO phenotype may be a clinically useful strategy for delaying clinical CVD risk. However, little is known about the natural history of progression from the metabolically benign to at-risk phenotype and the factors that trigger this progression. Answering these questions will help tailor clinical practice to focus on the specific abnormalities that predispose an individual's progression to the at-risk phenotype. Thus, the aim of this study was to identify the incidence of transition from the MBO at baseline to the at-risk overweight/obese (ARO) phenotype over 7 years and to identify the factors associated with the transition from the MBO to the ARO phenotype. We used data from the Study of Women's Health Across the Nation (SWAN), a cohort of women in the menopausal transition, who were followed prospectively for CVD risk outcomes to address these questions.

Materials and Methods

Study population

The current analyses use women participating in the SWAN Study; a multicenter, multiethnic longitudinal study designed to characterize the biological and psychosocial changes that occur during the menopausal transition in a community-based sample. Details of the study design and recruitment have been previously published (11). Briefly, SWAN is being conducted at seven sites: Boston, Massachusetts; Chicago, Illinois; the area of Detroit, Michigan; Los Angeles, California; Newark, New Jersey; Pittsburgh, Pennsylvania; and Oakland, California. Data from baseline visit and follow-up visits at years 1, 3, 4, 5, 6, and 7 were available for analysis. Of the 3302, 42- to 52-year-old women enrolled from 1996 to 2002, 1308 were normal weight [body mass index (BMI) <25 kg/m²]; 42 were missing BMI information, and 211 were missing information needed to define the MBO/ARO phenotype. Of the 1741 women with BMI of 25 kg/m² or greater at baseline, 866 (49.7%) fulfilled the criteria of the MBO phenotype. Of these, 482 were overweight (BMI 25–29.99 kg/m²) and 384 were obese (BMI >30 kg/m²). The institutional review boards of the participating institutions approved this study, and informed consent was obtained prior to participation and annually from all participants.

Physical measures

Height and weight were measured in light clothing and without shoes using a stadiometer and balance-beam scale, respectively. BMI was calculated as weight in kilograms divided by height in square meters. Blood pressure (BP) was measured in the right arm with the participant seated, after at least 5 minutes of rest. Two sequential BP values were obtained and averaged. Waist circumference was measured with the participant in nonrestrictive undergarments, at the level of the natural waist, defined as the narrowest part of the torso as seen from the anterior aspect. For cases in which the waist narrowing was difficult to determine, the measure was taken at the smallest horizontal circumference in the area between the ribs and the iliac crest.

Other measures

Age was calculated as the date of the visit minus the date of birth. Race, current smoking status, alcohol consumption, and education status (high school or less/past high school/college or higher), and family history of cardiovascular disease were obtained from a self-reported questionnaire at baseline. Race/ethnicity was self-defined as black or African American, non-Hispanic Caucasian, Chinese or Chinese American, Japanese or Japanese American, or Hispanic (Central American, Cuban or Cuban American, Dominican, Mexican or Mexican American, Puerto Rican, South American or Spanish, or other Hispanic). In addition, respondents could specify a race/ethnicity other than the defined categories or indicate mixed or no primary affiliation.

By design, all women were pre- or early perimenopausal at baseline and menopausal transition status was assessed annually based on bleeding criteria. Categories were as follows: 1) premenopause (no decrease in regularity in menstrual bleeding during the last year); 2) early perimenopause (decreased menstrual regularity in the past year but at least one period in the past 3 months); 3) late perimenopause (no period in the past 3–11 months); 4) postmenopause (>12 consecutive months of amenorrhea); 5) surgical menopause (menopause induced by hysterectomy with/without oophorectomy); or 6) hormone replacement therapy (HRT) users (use of HRT before the documentation of a final menstrual period). Physical activity was measured using a modified version of the Kaiser Physical Activity Survey (KPAS) (12) at baseline and at annual visits 3, 5, and 6. Adapted from the Baecke physical activity questionnaire, (13), the KPAS assesses activity levels during the previous 12 months in various domains including active living, household/caregiving, and sports/exercise. The KPAS has been validated against activity logs, accelerometers, and maximal oxygen consumption, (12) with 1-month test-retest reliability of 0.81–0.84 (varying by domain) (12, 14). Scores were calculated based on ordinal Likert scale categorical responses, with higher scores indicating greater activity (range 3–15).

Blood assays

Standard cardiovascular risk factors were assayed annually at the Medical Research Laboratories (Lexington, KY), which is certified by the National Heart Lung and Blood Institute, Centers for Disease Control and Prevention Part II program, as previously described (15). Low-density lipoprotein was calculated using the Friedewald equation, (16) excluding women with values of triglycerides of 400 mg/dL or greater. The homeostasis model assessment insulin resistance index (HOMA-IR) was calculated from fasting insulin and glucose as (fasting insulin × fasting glucose)/22.5 (17). High-sensitivity C-reactive protein (CRP) levels were measured using an ultrasensitive rate immunonephelometric method (BN 100; Dade-Behring).

Definitions of metabolic phenotypes

Overweight/obese women (BMI ≥25 kg/m²) were characterized as either metabolically benign or metabolically at risk by assessing their cardiometabolic risk burden at baseline and annually using four criteria from the Revised Adult Treatment Panel III definition of metabolic syndrome (18) as follows: 1) systolic/diastolic BP of 130/85 mm Hg or greater or antihypertensive medication use; 2) fasting triglycerides of 150 mg/dL or greater; 3) fasting high-density lipoprotein (HDL) cholesterol levels of 50 mg/dL or less or lipid-lowering medication use; and 4) fasting glucose of 100 mg/dL or greater or self-reported use of antidiabetic medications. The waist circumference criterion was not used because we were analyzing only the overweight/obese group. Women who had fewer than two abnormalities were defined as metabolically benign and those with two or more abnormalities were defined as being at risk. The incident ARO phenotype was identified as the first visit in which a woman had two or more cardiometabolic abnormalities.

Statistical analyses

The incidence of progression to the at-risk overweight/obese phenotype was calculated for each study visit using the Kaplan-Meier plot.

Differences in baseline demographic characteristics, anthropometric measures, lifestyle factors, and metabolic profiles between the women who remained in the MBO category across the follow-up period and those who progressed to the ARO phenotype were examined using t tests or Wilcoxon tests for continuous variables and the χ² test for categorical variables.

We used discrete time proportional hazard modeling to obtain summary measures of relative risk of each variable of interest and the progression from MBO to ARO phenotype. Using women who remain in the MBO phenotype across the follow-up period as the reference category, we calculated the hazard ratio for each independent variable, which may be interpreted as an estimate of relative risk of progression from MBO to ARO phenotype. The time span being analyzed (1–7 y) was divided into intervals within which the proportional hazards assumption were deemed reasonable for all variables entered in the initial model. A separate model was fit for each time interval with only those surviving (ie, remaining metabolically benign to the beginning of the interval included in the computations). Once a woman transitioned to the ARO phenotype, she was censored thereafter.

To compare the independent effects of each risk factor on the progression to the ARO phenotype, we constructed separate multivariable discrete-time proportional hazard models using standardized estimates of each risk factor at baseline. The demographic, personal, and family history variables that are well known to affect CVD risk were selected a priori and added to the model.

Finally, we constructed a multivariable model with all known cardiometabolic abnormalities (elevated blood glucose, BP, triglycerides and CRP levels, and low HDL levels) and adjusted for both static and time-varying variables. Both baseline and time-varying BMI and menopausal status were included to not only determine how a woman's risk of progression to ARO depended on her baseline characteristics but also how the risk for women at a certain BMI and a certain menopausal status changes over time. We did not include time-varying physical activity scores because the means were very similar at each year showing that physical activity did not change much over the years of follow-up. In addition, we found significant multicollinearity in the model on adding both baseline and time-varying physical activity. Time varying cardiometabolic abnormalities were not included in the regression model because they are part of the outcome.

Sensitivity analyses included the following: 1) using a different definition of metabolic phenotypes in which the MBO phenotype was defined as having no metabolic abnormality and transition to the ARO phenotype defined as having one or more metabolic abnormalities; 2) censoring women at the time of initiation of menopausal hormonal therapy use; and 3) excluding women from the New Jersey site (due to the small number). To further understand the effect of obesity, we added an interaction term of obesity to each CVD risk factor in the final model. In addition, we constructed separate multivariable models for overweight and obese women to examine differences in associations.

Analyses were performed with SAS version 9.2 (SAS Institute). All tests were two sided at α = .05.

Results

Of the 866 MBO women at baseline, 43% progressed to the at-risk phenotype during follow-up and 493 (57%) remained metabolically benign at the end of the study period. The incidence of progression slowed with time, such that with each passing year, fewer women progressed to the at-risk phenotype (Figure 1).

Figure 1.

Kaplan-Meier curve showing time to progression from metabolically benign to at-risk overweight/obese phenotype.

Table 1 shows the differences in baseline demographic, anthropometric, and cardiometabolic risk factors (BP, triglycerides, glucose, insulin, HOMA, CRP, and HDL-cholesterol levels) between women who remained MBO until the end of follow-up and those who progressed to the ARO phenotype. Compared with women who remained metabolically benign, the group who progressed to the ARO phenotype had fewer white women; was less highly educated; reported less physical activity; had higher BMI and waist circumference; and had a poorer cardiometabolic risk profile (higher triglycerides, glucose, and CRP levels and lower HDL levels) at baseline.

Table 1.

Baseline Characteristics of Overweight/Obese Women Who Remained Metabolically Benign and Women Who Progressed to At-Risk Phenotype

Metabolically Benign Overweight/Obese Women at Baseline (n = 866)	Remained Metabolically Benign Until End (n = 492)	Progressed to At-Risk Phenotype (n = 374)	P Value
Age, ya	46.2 (0.1)	46.5 (0.1)	.113
Race/ethnicity, %, n			.020
Caucasian	53 (261)	46 (171)
Black	36 (179)	37 (139)
Chinese/Chinese American	2 (10)	3 (12)
Japanese/Japanese American	2 (9)	5 (19)
Hispanic	7 (33)	9 (33)
Education, %, n			.039
High school or less	20 (93)	26 (97)
High school or higher	34 (160)	34 (124)
College/postcollege	47 (221)	40 (146)
Current smokers, %, n	12 (59)	16 (59)	.102
Family history of CVD, %, n	50 (218)	55 (184)	.237
Current alcohol use, %, n			.258
Low, ≤1/mo	47 (229)	51 (191)
Moderate, >1/mo to ≤2/wk	34 (166)	29 (107)
High, 2+/wk	20 (97)	20 (76)
Menopausal status at baseline			.170
Premenopausal	54 (256)	49 (181)
Early perimenopausal	46 (221)	51 (189)
Physical activity scorea	7.7 (0.1)	7.4 (0.1)	.001
BMI, kg/m2b	28.7 (26.6–32.4)	30.1 (27.3–34.8)	<.001
Waist circumference, cma	88.9 (0.5)	93.8 (0.7)	<.001
Systolic BP, mm Hga	114.4 (0.6)	120.7 (0.8)	<.001
Diastolic BP, mm Hga	73.1 (0.4)	76.4 (0.5)	<.001
HDL-cholesterol, mg/dLa	60.4 (0.6)	54.3 (0.6)	<.001
Triglycerides, mg/dLb	77.0 (60–100.0)	95.0 (77–123.0)	<.001
Glucose, mg/dLb	89.0 (86.0–94.0)	91.0 (87.0–96.0)	<.001
HOMA-IRb	1.8 (1.4–2.5)	2.3 (1.7–3.2)	<.001
CRP, mg/dLb	1.9 (0.8–5.2)	2.6 (1.2–6.1)	<.001

Definition of menopausal status is as follows: 1) premenopausal is menstrual periods in the past 3 months, with no irregularity in the past 12 months; or 2) early perimenopause is monthly bleeding with a perceived change in cycle interval but at least one period within the past 3 months.

^aMean (SD).

^bMedian (interquartile range).

Using the women who remained metabolically benign until the end of follow-up as a reference, Table 2 shows the risk of progression to the ARO phenotype for baseline and time-varying variables of interest in the univariate analyses. Compared with Caucasians, both Hispanic and Japanese women had a higher risk of progression to the at-risk phenotype. Also, those with lower levels of education were at a higher risk of progression. Of the lifestyle factors, neither smoking nor alcohol consumption was associated with an increased risk of progression. Physical activity at baseline and over time showed a protective effect on the risk of progression. Obesity at baseline, and an increase over time (as seen by BMI), was associated with a significant risk of progression. The presence of any cardiometabolic abnormality at baseline was associated with a modest but significant risk of progression from the metabolically benign to the at-risk phenotype.

Table 2.

Univariate Associations Between Demographic and Cardiometabolic Risk Factors and Progression to At-Risk Phenotype in Overweight/Obese Women^a

	Hazard Ratio	95% CI	P Value
Age at baseline, y	1.04	1.00, 1.08	.044
Age (time varying), y	1.04	1.00, 1.08	.053
Race/ethnicity (Caucasian as reference)			<.001
Black	1.04	0.83, 1.29	.760
Chinese	1.35	0.71, 2.32	.319
Hispanic	1.85	1.24, 2.65	.002
Japanese	2.34	1.41, 3.66	<.001
Education (college/postcollege as reference)			.009
High school or less	1.48	1.14, 1.91	.003
Higher than high school	1.08	0.85, 1.37	.519
Menopausal status at baseline (premenopausal as reference)
Early perimenopause	1.09	0.89, 1.34	.393
Menopausal status (time varying) (premenopausal as reference)			.452
Early perimenopause	1.45	1.02, 2.11
Late perimenopause	1.36	0.82, 2.23
Postmenopausal	1.33	0.86, 2.09
Surgical menopause	1.49	0.84, 2.59
HRT use	1.20	0.72, 1.97
Smoking (nonsmokers as reference)	1.02	0.85–1.28	.886
Alcohol use (<1/mo as reference)			.676
Moderate	0.96	0.74–1.23
High, 2+/wk	0.89	0.66–1.17
Physical activity score at baseline	0.90	0.84, 0.95	<.001
Physical activity score (time varying)	0.84	0.77, 0.92	<.001
BMI at baseline, kg/m2	1.03	1.01, 1.05	<.001
BMI, kg/m2, time varying	1.05	1.03, 1.06	<.001
HTN at baseline (SBP ≥130 mm Hg and/or DBP ≥85 mm Hg or antihypertensive medication)	1.66	1.32, 2.07	<.001
Low HDL at baseline (HDL <50 mg/dL or lipid lowering medication)	1.74	1.40, 2.16	<.001
High TG at baseline (≥150 mg/dL)	1.89	1.19, 2.83	.004
High fasting glucose at baseline (≥100 mg/dL or antidiabetic medication use)	1.39	1.001, 1.89	.049
Elevated CRP at baseline (CRP ≥3.0 mg/dL)	1.33	1.08, 1.63	.006

Abbreviations: DBP, diastolic BP; HTN, hypertension; SBP, systolic BP; TG, triglycerides.

^aTime-varying associations were calculated using discrete-time proportional hazard modeling.

Separate multivariable models comparing standardized estimates of each CVD risk factor at baseline, and adjusted for known demographic and lifestyle risk factors, showed that hazard ratios were similar, with a 1-SD increase in each variable associated with a 14%–39% increase in the risk of progression from the MBO to ARO phenotype (Table 3). Both baseline HDL-cholesterol levels and physical activity scores conferred a modest but significant protection against progression.

Table 3.

Multivariable^a Models^b showing Risk of Progression to At-Risk Overweight/Obese Phenotype Comparing Standard Estimates of Each Cardiometabolic Risk Factor at Baseline

CVD Risk Factors	1 SD	HR	95% CI	P Value
BMI	5.73 kg/m2	1.18	1.06, 1.31	.003
Triglycerides	33.09 mg/dL	1.39	1.27, 1.51	<.001
HDL	12.29 mg/dL	0.61	0.53, 0.70	<.001
Glucose	13.03 mg/dL	1.14	1.05, 1.22	.001
Systolic BP	15.02 mm Hg	1.34	1.20, 1.50	<.001
Diastolic BP	9.86 mm Hg	1.27	1.13, 1.43	<.001
CRP	6.65 mg/L	1.02	0.92, 1.11	.665
Physical activity score	1.74	0.77	0.68, 0.87	<.001

^aEach model is additionally adjusted for age, site of recruitment, education status, race-ethnicity, smoking, alcohol use, family history of CVD, and menopausal status.

^bEach row is a separate multivariable model and includes the CVD risk factor of interest listed and adjusted for the variables given above.

When all CVD risk factors were added to the multivariable model, associations between educational status and the risk of progression from the MBO to ARO phenotype were no longer statistically significant (Table 4). Japanese women continued to show an increased risk of progression compared with their Caucasian counterparts. Interestingly, age acted as a protective factor, whereas menopausal status at baseline or over time showed no association with the risk of progression to the ARO phenotype. Although a lower baseline BMI was protective, an increase in BMI over time continued to show a modestly increased risk of progression. All of the cardiometabolic abnormalities were associated with a significant increase in risk of progression, with impaired fasting glucose showing the strongest association. Physical activity continued to be protective against progression after adjusting for all cardiometabolic risk factors.

Table 4.

Multivariable Model Showing Risk of Progression to At-Risk Phenotype in Overweight/Obese Women^a

n = 866	Hazard Ratio	(95% CI)	P Value
Age at enrollment, y	0.96	0.94, 0.98	<.001
Race/ethnicity (Caucasian reference)			.035
Black	0.79	0.58, 1.08
Chinese	1.33	0.59, 2.86
Hispanic	1.61	0.65, 4.58
Japanese	2.63	1.28, 5.47
Education (college/postcollege as reference)			.218
High school or less	1.16	0.84, 1.59
Higher than high school	0.86	0.64, 1.15
Baseline menopausal status (premenopausal as reference)
Early perimenopausal	0.99	0.77, 1.29	.959
Menopausal status; time varying (premenopausal as reference)			.191
Surgical	0.67	0.32, 1.31
Postmenopausal	0.81	0.48, 1.33
Late perimenopausal	0.73	0.40, 1.29
Early perimenopausal	1.08	0.74, 1.63
HRT	0.74	0.41, 1.32
Baseline physical activity	0.86	0.80, 0.92	<.001
Baseline BMI, kg/m2	0.87	0.82, 0.92	<.001
BMI (time varying)	1.16	1.10, 1.22	<.001
HTN (SBP ≥130 mm Hg and/or DBP ≥85 mm Hg) at baseline or antihypertensive medication use	3.00	2.11, 4.27	<.001
HDL <50 mg/dL at baseline	2.85	2.07, 3.95	<.001
TG ≥150 mg/dL at baseline or lipid-lowering medication use	2.91	1.62, 4.97	<.001
Fasting glucose ≥100 mg/dL at baseline or antidiabetic medication use	3.24	2.10, 4.92	<.001
CRP ≥3.0 mg/L at baseline	1.20	0.93, 1.55	.163

Abbreviations: DBP, diastolic BP; HTN, hypertension; SBP, systolic BP; TG, triglycerides.

^aModel was additionally adjusted for site of recruitment, smoking, alcohol use, and family history of CVD.

Sensitivity analyses

We did not find a significant interaction between obesity and any of the cardiometabolic abnormalities. In reconstructing the multivariable model separating overweight and obese women, we found that even though not always significant, the hazard ratios remained similar in magnitude (see Supplemental Table). Hazard ratios were similar when the metabolic phenotype definition was changed to one or more metabolic abnormalities defining the ARO phenotype. Removing New Jersey participants from the analysis and censoring women when they started using HRT did not alter the results significantly (data not shown).

Discussion

Although a significant number of women remained in the MBO category, 43% of metabolically healthy overweight/obese women progressed to the at-risk phenotype over a 7-year period. This high prevalence is concerning, especially given the relatively young age of these perimenopausal women. Thus, it is important to identify not only the triggers that predispose progression to the at-risk phenotype but also the factors that may impede or slow this progression. We find that not only an increase in obesity over time but also the presence of any cardiometabolic abnormality at baseline increases this risk significantly. At the same time, high physical activity scores at baseline protect against progression.

In our cohort, independent of obesity, glucose dysregulation at baseline was strongly associated with an increased risk of progression to the at-risk phenotype. Multiple data indicate the role of diabetes in augmenting future cardiovascular disease risk, independent of obesity (19–21), and other studies have suggested a potential role of impaired glucose metabolism (22) in the progression. Similar to these findings, in our cohort of overweight/obese metabolically benign perimenopausal women, we show that impaired fasting glucose acts as a significant predictor of progression, even before the clinical diagnosis of diabetes, and this risk remains significantly high, even after accounting for baseline and changes in obesity. Thus, we believe that the recent recommendations to monitor glucose levels in overweight and obese individuals at regular intervals appear prudent (23).

Vascular maladaptation acts as a major contributor to the development of cardiovascular disease. Both hypertension and atherogenic dyslipidemia have been associated with subclinical and clinical atherosclerosis (24–26). In our cohort, hypertension increased the risk of progression to the ARO phenotype 300-fold compared with those who did not have hypertension. Similarly, dyslipidemia (low HDL and high triglyceride levels) was a significant predictor to progression to the at-risk phenotype, even after adjustment for obesity. Multiple prospective studies show that a 1% decrease found in the plasma levels of HDL cholesterol is associated with a 2%–3% increase in the risk of CVD, independent of other risk factors, including age, BP, smoking, BMI, and low-density lipoprotein cholesterol (27–30). Our study reconfirms the role of lipid levels in CVD risk assessment (31, 32), and the argument that the Adult Treatment Panel III criteria may identify different subgroups at risk of CVD (33, 34) will need to be studied at greater length.

Underlying systemic inflammation has been considered an instigating factor for metabolic risk (35, 36), but in our cohort, the association between CRP and the risk of progression was lost after adjusting for known cardiometabolic risk factors. The benefit of adding CRP to CVD risk profile assessment remains controversial (37, 38). Although our results do not show any added value to measuring CRP levels, it is possible that other inflammatory markers may be more sensitive in assessing cardiovascular risk.

Equally of interest are variables that served as protective factors to progression to the at-risk phenotype. Women who remained metabolically benign were more physically active than their peers, and the protective effect of physical activity was also noted over time; and this protective effect persisted, even after adjusting for known cardiometabolic risk factors (Table 4). These results suggest that higher levels of physical activity could potentially allay the risk of progression from the MBO to the ARO phenotype. The role of physical activity on metabolic phenotypes remains controversial. Although some studies report comparable levels of the resting metabolic rate and peak energy expenditure between the two phenotypes (22), others report higher antioxidant activity after exercise in metabolically benign overweight/obese women (39). In addition, multiple studies show that even without weight loss, physical activity improves fat mobilization from the liver (40) and visceral adipose depot, (41) and is associated with a decrease in subclinical atherosclerosis (42). Therefore, physical activity may be able to protect against progression to the ARO phenotype, even in the presence of other cardiometabolic risk factors.

Interestingly, age appears to be protective for the whole group and in the obese category when stratified by BMI. In addition, the incidence of progression to the at-risk phenotype decreased over time. Although the duration of obesity has been linked to the development of metabolic syndrome (43), we found the highest risk of progression from metabolically benign to the at-risk phenotype occurred in the early years after recruitment. These results suggest that a longer duration in the MBO phenotype demonstrates an overall degree of metabolic health. In addition, baseline BMI, although increasing the risk of progression in univariate analysis, served as a protective factor when adjusted for all other cardiometabolic risk factors (as seen in Table 4). Although counterintuitive, it is possible that if an individual has remained metabolically benign to a certain BMI, a further increase in the BMI may act as the tipping point in increasing the risk of progression.

Our study focuses on women beginning the menopausal transition. Although we did not find an association between menopausal status at baseline or over time with progression to the at-risk phenotype, low SHBG levels and a high free androgen index have been associated with elevated CVD risk factors in postmenopausal women (44, 45). Similar associations have been observed in a cross-sectional analysis of the SWAN cohort (15). A closer look at our cohort shows that 52% women were premenopausal at the time of recruitment, and years 1 and 3 showed the highest transition from premenopause to early perimenopause (12% and 9%, respectively). Although it may be hypothesized that this transition was responsible for the progression to the at-risk phenotype in the early years of follow-up, a longitudinal analysis of carotid intima-media thickness in the SWAN cohort reports that the rate of carotid intima-media thickness progression is maximal during late perimenopause and then tapers off postmenopause, irrespective of age and ethnicity (46). Because 25% women in our study were still in early perimenopause at the end of follow-up, it is possible that a significant percentage of women have not experienced the hormonal changes and the vessel remodeling associated with late perimenopause that puts them at a higher risk of developing atherosclerosis. It would therefore not be prudent to generalize the natural history of progression to at-risk obesity of this cohort to a younger or older group of women without addressing the association between the changing menopausal hormonal milieu and CVD risk factors.

Our results should be viewed in light of certain limitations. Certain diets improve metabolic health, regardless of weight loss, and this information would add to the importance of our findings. Although BMI has been used as a surrogate for obesity, we recognize that a direct measure of abdominal visceral adipose tissue and subcutaneous adipose tissue would give more precise associations. Finally, it is also possible that women who remain in the MBO phenotype have other protective factors at play such as higher adiponectin levels conferring metabolic health.

Despite these limitations, by studying risk factors that can trigger progression to the at-risk phenotype in a large, multiethnic, multicenter cohort, our results are applicable to midlife women of our target ethnic groups. By prospectively following up participants over 7 years, we are able to report incidence of the progression to the at-risk phenotype in our cohort and examine putative and protective factors that are clinically measurable and treatable. Using standardized estimates improves our ability to compare the effect among known cardiometabolic risk markers on progression to the at-risk phenotype.

We show that an increase in obesity over time and the presence of any known metabolic abnormality increase the risk of progression, with glucose dysregulation being the best predictor of progression to the at-risk phenotype. On the other hand, higher levels of physical activity play a protective role in decreasing the risk of progression. Controlling these factors could potentially allay the elevated risk of cardiovascular morbidity and mortality unanimously associated with the at-risk phenotype. During the present obesity epidemic, public health resources need to have a multiprong approach to not only focus on obesity prevention and treatment but also include early identification and treatment of glucose dysregulation and other metabolic abnormalities. In addition, the promotion of physical activity should be emphasized in clinical encounters as a tool to improve cardiometabolic health.