Metabolic Syndrome and Perimenopause

Perimenopause & Metabolic Syndrome: Is there a link and what should be done about it?

Iris Thiele Isip Tan MD, FPCP, FPSEMMS Health Informatics (cand.)Clinical Associate Professor, UP College of MedicineSection of Endocrinology, Diabetes & MetabolismDepartment of Medicine, UP-Philippine General Hospital

Hypertension

High triglycerides

ObesityLow HDL-C

Glucose intolerance

Metabolic Syndrome

Cross-sectional data MetSyn increases

from premenopause to postmenopause,

independent of age.

Menopause

Hypertension

High triglycerides

Obesity

Low HDL-C

Glucose intolerance

(Peri)

sociated with weight gain, most studies do not reveal in-creases in BMI independent of normal aging (23, 24). Al-though it is estimated that middle-aged women gainapproximately 0.55 kg (!1 lb)/yr, there does not appear tobe an independent effect of menopause on body weight (25,26). However, even in the absence of weight gain, body fatdistribution changes across the menopause. Cross-sectional(27) and longitudinal studies (22, 28) have shown that themenopausal transition is associated with a preferential in-crease in abdominal adiposity, independent of the effect ofage and total body adiposity. Poehlman et al. (22) prospec-tively compared women who became postmenopausal toage-matched controls who remained premenopausal andfound that the transition to menopause was associated withan increase in the waist to hip ratio and total body fat.Abdominal fat, measured by CT scan, has also been shownto increase with menopause in both cross-sectional (29) andprospective studies (30). Visceral fat accumulation is thoughtby many to be the major determinant of the metabolicsyndrome.

Women with high amounts of visceral fat have an excessof cardiovascular mortality and associated metabolic abnor-malities (31). When Pascot et al. (32) matched women forabdominal fat (by CT scan) and menopausal status, the dif-ferences initially found in very low density lipoprotein(VLDL), LDL, HDL, large buoyant HDL2 particles, LDL par-ticle size, fasting glucose, C peptide, and blood pressure were

eliminated, implying that the differences in visceral fat andmenopausal status accounted for the metabolic differences.Regional differences in adipose tissue lipoprotein lipase ac-tivity in postmenopause may account for the menopausalchanges in fat accumulation, but results to date are conflict-ing (33, 34). Adiponectin, a novel adipocyte-derived peptide,may play a role in the metabolic syndrome, as concentrationsare inversely related to obesity and insulin resistance. How-ever, the only study evaluating adiponectin in menopauserevealed no difference in pre- and postmenopausal women(35).

Menopause is also associated with reduced lean bodymass (muscle) and this appears to be related to decreasedphysical activity (36). Lynch et al. (37) recently reported lowermaximal oxygen consumption (VO2 max) in sedentary post-menopausal (VO2 max) women compared with sedentaryage-matched premenopausal women and found an inverserelationship between visceral adiposity and maximal oxygenconsumption. The reductions in exercise capacity and activ-ity may contribute to the reduced lean body mass and in-creased central adiposity with menopause.

Effects of menopause on lipid metabolism

Although the association between abdominal adiposityand the constellation of lipid abnormalities is well known,the underlying pathophysiology is not clear. High amountsof abdominal fat are associated with increased insulin resis-tance, free fatty acid (FFA) levels, and decreased adiponectin(Fig. 2). These factors contribute to increased secretion ofapolipoprotein B (apo B)-containing particles, leading to hy-pertriglyceridemia and increased hepatic lipase (HL) activityresulting in a predominance of small dense LDL particles anda reduction in large antiatherogenic HDL2 particles. A similarpattern of lipid abnormalities emerges with menopause(Table 3).

Changes in LDL with menopause

Postmenopausal women have higher total cholesterol,LDL cholesterol, triglycerides (TG), and lipoprotein(a)[Lp(a)] levels and lower HDL cholesterol levels than pre-menopausal women (38–40). Although elevated LDL is nota component of the metabolic syndrome, LDL levels increaseby 10–20% (23, 41) with menopause, and the greatest change

FIG. 1. Patterns of body fat distribution.

FIG. 2. The interaction of genetic and environmental factors influences the manifestation of the metabolic syndrome. High amounts ofintraabdominal fat (IAF) are associated with increased insulin resistance (IR) and FFA levels, and decreased adiponectin. These factorscontribute to increased secretion of apo B-containing particles, leading to hypertriglyceridemia and increased HL activity, which lead to apredominance of small dense LDL particles and a reduction in the large antiatherogenic HDL2 particles.

2406 J Clin Endocrinol Metab, June 2003, 88(6):2404–2411 Carr • Menopause and the Metabolic Syndrome

by on August 4, 2009 jcem.endojournals.orgDownloaded from

⬆riskdiabetes hypertriglyceridemiasmall dense LDLhypertensionCVD

Estrogen promotes gluteo-femoral fat distribution

Carr M et al JCEM 2003;88:2404-11

Menopausal fat redistribution ➨ MetSyn













Carr M et al JCEM 2003;88:2404-11

Sex- and menopause-associated changes in body-fat distribution

Ley et al Am J Clin Nutr 1992;55:950-4

-

?:,

. i ): . .

1.

!

. .

-

Men21-79 yn = 103

Premenopausal women Regular menses

No menopausal sx 19-51 yn = 61

PostmenopausalAmenorrhea

Elevated LH & FSH 43-63 yn = 70

Good healthNot obese (BMI 18.8-28)

No meds affecting lipid/bone metabolism

DEXA measurement

Increased android fat in menopause


60

50

40

30













Pre-menopausal

Post-menopausal Males

%n = 61***

n = 70***

n = 103***

*** p <0.001

Decreased gynoid fat in menopause














60

50

40

30Pre-

menopausalPost-

menopausal Males

%

n = 61***

n = 70***

n = 103***

*** p <0.001

Hallmark of menopausal transition is

dramatic reduction in estradiol

Is androgen dominance in menopause associated with

development of MetSyn?

SWAN Cohort & Metabolic Syndrome

Study of the Women’s Health Across the Nationn = 949; 9-year follow-up

Premenopause or early perimenopause

Reached menopause during the studyNever took hormone therapyNo diabetes or MetSyn at baseline

Primary outcome: MetSyn (NCEP-ATP III)

Secondary outcome:MetSyn components

Serum T, SHBG and estradiol

Janssen I et al Arch Intern Med 2008;168(14):1568-75

Hormone Levels in SWAN Cohort With and Without MetSyn

Hormone levels, median (IQR) Total Cohort Cohort without

MetSynCohort

with MetSyn

Testosterone, ng/dL 38.0 (26.5-51.0)

37.6 (25.8-50.4)

⬆42.5*(30.4-53.9)

Bioavailable T, ng/dL 3.13 (1.78-5.32)

2.89(1.58-4.86)

⬆4.52**(2.48-7.08)

SHBG, ug/mL 4.88(3.37-6.80)

5.06(3.54-7.21)

⬇3.83*(2.52-5.46)

Estradiol, pg/mL 25.5(14.8-65.0)

26.6(14.7-69.5)

⬇21.8**(14.4-34.8)


* p<0.05 ** p<0.001


SWAN Cohort & Metabolic Syndrome

New-onset MetSyn by final menstrual period: 13.7%

Odds of MetSyn/year in perimenopause:1.45 (95%CI 1.35-1.56)

Odds of MetSyn increased by

10% for every 1-SD increase in bioavailable T levels*

13% for every 1-SD decrease in SHBG level*

*Adjusted for age, age at final menstrual period, ethnicity, study site, baseline BMI, change in BMI from baseline,

baseline education level, marital status and smoking

Body fat distribution andMetabolic Syndrome

Obesity clinic (Istanbul)

Osbey et al Endoc J 2002;49(4):503-9

Premenopausal n= 405Overweight/obese (BMI>27)mean BMI 37.8 + 6.9regular cycles FSH <30 IU/LWeight stable x 6 mosNormal resting ECG

Matched for BMI

Postmenopausal n= 405Overweight/obese (BMI>27)mean BMI 37.7 + 6.8abnormal menses x 12 mosFSH >30 IU/LExcluded women with premature menopause

Exclusion criteriaDiabetes, IHD, hypertension or chronic diseaseuse of OCP or HRT Medications affecting body composition/metabolism

Retro

spec

tive

Postmenopausal women had more intra-abdominal fat


Premenopausaln = 405

Postmenopausal n = 405

p

Age 33.28 + 7.62 52.36 + 7.50 <0.001

Weight (kg) 92.0 + 17.5 91.2 + 16.0 NS

Height (cm) 155.9 + 5.3 155.6 + 5.8 NS

BMI (kg/m2) 37.83 + 6.91 37.77 + 6.84 NS

Waist circ (cm) 99.19 + 13.45 103.34 + 13.20 <0.001

Hip circ (cm) 123.10 + 13.10 123.83 + 13.09 NS

WHR 0.80 + 0.07 0.84 + 0.08 <0.001

IAF (L) 3.19 + 1.42 3.97 + 2.17 <0.001

Matched for BMI

MetSyn components higher in postmenopausal women


Premenopausaln = 405

Postmenopausal n = 405

p

Age 33.28 + 7.62 52.36 + 7.50 <0.001

Systolic BP (mm Hg) 135.12 + 26.79 148.24 + 29.77 <0.001

Diastolic BP (mm Hg) 87.70 + 15.08 91.79 + 15.71 <0.001

Glucose (mg/dL) 99.41 + 19.19 109.68 + 33.62 <0.001

Uric acid (mg/dL) 4.34 + 1.12 4.78 + 1.47 <0.001

Cholesterol (mg/dL) 202.33 + 37.09 232.22 + 43.22 <0.001

HDL-C (mg/dL) 45.32 + 10.89 47.24 + 10.38 0.016

Triglyceride (mg/dL) 152.28 + 74.93 172.68 + 79.97 <0.001

Matched for BMI

Increase in abdominal fat and unfavorable risk factors despite unchanged total body weight and BMI during menopause transition

Menopause


Unclear whether menopause is a CV risk factor for all women or only those with a genetic predisposition to central obesity













Carr M et al JCEM 2003;88:2404-11

Epidemiologic data Central fat distribution

related to adverse psychological states

(i.e. depression and anxiety) and to social difficulties

(i.e. unemployment and divorce)

Epel E et al Psychosomatic Med 2000;62:623-32

Depressive Symptoms and Stressful Life Events Predict MetSyn

Raikkonen et al Diabetes Care 2007;30:872-7

Enrolled at 3-y follow-up 254 (58.8%) premenopause63 (14.6%) perimenopause115 (26.6%) menopause

Had MetSyn components at 3-yr follow-up and at least one later examination

Psychosocial measuresBeck Depression InventoryFramingham Tension ScaleSpielberger Trait Anxiety QSpielberger Anger QPerceived Stress Scale

Healthy Women Study cohortn = 432; 15-year follow-up

Psychosocial Factors and MetSyn

Depressive symptomsOR 1.39 (1.11-1.74)

Trait angerOR 1.40 (1.13-1.74)

Trait anxietyOR 1.03

(0.83-1.28)

Framingham tensionOR 1.27 (1.03-1.57)

Perceived stress

OR 1.19 (0.96-1.47)

None/mild vs at least one very severe

stressful life event OR 1.84 (1.20-2.81)

Prevalence of MetSyn (ATP III) over 15-y follow-up

Raikkonen et al Diabetes Care 2007;30:872-7

Adjusted for age, physical activity, alcohol consumption, current smoking status, use of HRT and level of education

Stress and Body ShapeHypothesis: Greater vulnerability to stress increases exposure to stress-induced cortisol ➜ central fat deposition

Do women with greater central fat (high WHR) adapt less effectively to repeated stress over time than those with low WHR?

Psychological/cognitive measures:• Coping• Mood• Rosenberg Self-

Esteem Scale

Salivary cortisol


induced disease in people other than those with arecurrent depressive disorder or pathologic disorder ofthe HPA axis, such as Cushing’s syndrome.

The primary goal of this study was to assess whetherwomen with greater central fat (high WHR) adapt lesseffectively to repeated stress over time, both physio-logically and psychologically, than women with rela-tively greater peripheral fat (low WHR). Two pilotstudies of obese women found that those with greatercentral fat secreted more cortisol in response to a novellaboratory challenge (15, 16). However, these studieshad methodological limitations that preclude conclu-sive interpretations and generalization of results. Forexample, in one of these studies, 32% of the samplehad a history of smoking, which itself can increaseboth cortisol and abdominal obesity (17). Furthermore,they did not assess concurrent psychological re-sponses to the stressors to determine whether therewere differences in stress appraisal or performancethat may have stimulated the greater cortisol secretion.It is important to determine whether greater physio-logical reactivity among women with central fat distri-bution is caused centrally (psychologically driven re-activity). Otherwise, we leave open the possibility thattheir physiological reactivity could be a mere periph-eral epiphenomenon or a consequence of their in-creased central fat rather than a result of psychologicaldifferences.

Another limitation of the prior pilot studies is thatthey assessed exposure to a single episode of stress.Cortisol responses to a single laboratory challenge maynot represent stable response patterns over time be-cause the HPA axis is sensitive to novelty. Cortisolsecretion to a novel challenge, as observed in thesestudies, may even represent a healthy response,whereas consistently high cortisol reactivity to re-peated familiar challenges (or “nonhabituation”) is anatypical response that may reflect chronic physiologi-cal stress (18). Kirschbaum et al. (18) found that thecortisol response does not habituate to repeated stressin one-third of healthy men. For cortisol reactivity tostress to affect the target tissue (visceral fat) enough toactually shape one’s body, the reactivity would have tobe consistent and exaggerated. Thus, we predicted thatpeople with greater central fat would be “nonhabitua-tors” in that they would reliably secrete cortisol inresponse to familiar stressful situations. We thereforemeasured cortisol responses to identical challengesover several days.

Lastly, both prior studies included only overweightwomen, so the question of whether fat distribution hassimilar correlates in lean women could not be ad-dressed. Obesity itself affects fat distribution, contrib-uting to greater central fat deposits. Although obesity

and central fat tend to covary, that is, women tend tobe lean with a low WHR (Figure 1, A) or overweightwith high WHR (Figure 1, D), overall fatness and fatdistribution can uncouple, creating the less frequentlyobserved combinations of being lean but carryinggreater central fat relative to peripheral fat (Figure 1, B)and, conversely, of being overweight but having lesscentral fat relative to peripheral fat (Figure 1, C). Oursecondary aim was to compare stress reactivity in leanwomen with central fat (Figure 1, B) with stress reac-tivity in lean women with peripheral fat (Figure 1, A),which has not been studied. This poses a more directtest of the association between stress and fat distribu-tion because it is an assessment of fat distribution inthe absence of general accumulation of fat, which canserve as a confounding factor. Whereas overweightwomen will inevitably have greater central fat as aresult of their excess fat, lean women are less likely tohave central fat. One possible contributor to central fat

Fig. 1. Fat distribution phenotypes by level of obesity. A, Leanwith peripheral fat. B, Lean with central fat. C, Overweightwith peripheral fat. D, Overweight with central fat.

E. S. EPEL et al.

624 Psychosomatic Medicine 62:623–632 (2000)


WHR <0.75n = 29

WHR >0.79n = 30

Stress session 1

Stress session 2

Stress session 3

Control session

45 mins/session

Day 1

Day 2

Day 3

Day 4

ArithmeticSerially subtract a

prime number from a large

number

Visuospatial puzzle

Replicate picture designs with

blocks

SpeechConvince

committee that she is best job

applicant

Trier Social Stress Test

weight women, had we been able to use a more accu-rate measure of visceral fat.

It is also important to note that genetics may play arole in the stress–central fat relationship, although wedid not examine this in the current study. Genetics canaccount for up to 50% of the variance in fat distribu-tion (36). That leaves another 50% of the variance to beshaped by environmental influences. There are alsogenetic influences on psychological coping with stress(37, 38) and cortisol reactivity (39), so it is possiblethat stress reactivity and central fat are geneticallylinked.

Nevertheless, experimental manipulations of stressshow a causal relationship between stress and fat dis-tribution in animal models, regardless of genetics. Invivo, visceral fat deposits increased in a dose-depen-dent manner in rats and primates randomly assignedto a chronic stress condition (40, 41). In vitro, cortisolincreases lipoprotein lipase (a fat-storing enzyme) infat tissue but has an especially exaggerated effect onvisceral fat tissue (11).

CONCLUSIONS AND IMPLICATIONS

Although our findings are strictly correlational, thepsychological and cortisol data are consistent with thehypothesis that greater life stress and stress reactivitycontribute to central fat among lean women. The con-sistency of findings is striking: Vulnerability to stresswas noted across both psychological and physiologicalmeasures among women with a high WHR. There isgrowing recognition that overexposure to cortisol canhave pathophysiological consequences on many organsystems (42), stress-induced damage that has been la-beled “allostatic load” (43). Central fat among leanwomen may serve as an indicator of one type of allo-static load, physical damage resulting from lack ofadaptation to stress, that can eventually result in dis-ease (43). Thus, lean women with a high WHR may beat higher risk of disease for two known and likelyinterrelated factors, greater exposure to cortisol andpossible metabolic aberrations associated with centralfat distribution, such as greater insulin resistance (2).

Only longitudinal and genetic studies will deter-mine conclusively whether stress and central fat, withits related metabolic profile, are causally related orparallel phenomena. Future research needs to betterdefine levels of risk and appropriate treatments basednot only on one’s girth but also on the multiple causesof central fat, such as genetics, behavior, general obe-sity, and chronic stress.

Funding for this study was provided by the John D.and Catherine T. MacArthur Foundation ResearchNetwork on Socioeconomic Status and Health. Addi-

Fig. 2. Average cortisol response to stress sessions among leanwomen with a high or low WHR. Top, Mean cortisol re-sponse to first exposure to stressors (novel stress). Middle,Mean cortisol response to second exposure to stressors(familiar stress). Bottom, Mean cortisol response to thirdexposure to stressors (familiar stress).

E. S. EPEL et al.










E. S. EPEL et al.










E. S. EPEL et al.



High-WHR vs low-WHR womengreater threat appraisal (p=0.030)

exerted increasingly less effort over time (p=0.05)

made more mistakes (p=0.002)

Cross-sectional data Stress-induced cortisol secretion

may contribute to central fat: link between psychological stress

and risk for disease


Androgen predominance

(Menopausal transition)

Centralobesity

Metabolic Syndrome

Genetic predisposition

Environmental influence

Metabolic Syndrome in postmenopausal women and CV mortality

Tanko et al Circulation 2005;111:1883-90

Community-based sampleCopenhagenPostmenopausal womenn = 557 (48-76 years)8.5 + 0.3 y follow-up

Enlarged waist with elevated triglycerides (EWET) and CV mortality

Tanko et al Circulation 2005;111:1883-90

EWET confers 4.7-fold* (95%CI 2.2-9.8; p<0.001) increased risk of fatal CV events * adjusted for age, smoking and LDL-C

What should be done about it?Identify women at risk for MetSyn.

Management of Cardiovascular Risk in the Perimenopausal Woman: A Consensus Statement of European cardiologists and gnyecologists.Collins et al Eur Heart J 2007;28:2028-40

Screening for CV risk in Perimenopause

Assess all perimenopausal women seeking medical help for menopausal symptoms for risk of

‣ developing CVD‣ complications in the presence of existing disease

Menopausal Symptoms and CV Risk Factors

Eindhoven Perimenopausal Osteoporosis Study cohortn = 5523 women aged 46 to 57 years

Do women with vasomotor symptoms differ from those without with respect to CV risk factors?

Self-reported dataNight sweats 38%Flushing 39%

Measurementstotal cholesterolblood pressureBMI

Gerrie-Cor et al Hypertension 2008;51:1492-8

Menopausal Symptoms and CV Risk FactorsEindhoven Perimenopausal Osteoporosis Study cohortn = 5523 women aged 46 to 57 years

Gerrie-Cor et al Hypertension 2008;51:1492-8

Hypercholesterolemia OR 1.52 (95%CI 1.25-1.84)

Hypertension OR 1.20 (95%CI 1.07-1.34)

Women with flushing ↑ cholesterol (0.27 mmol/L [95%CI 0.15-0.39])

↑ BMI (0.60 kg/m2 [95%CI 0.35-0.84])

↑ SBP (1.59 mm Hg [95%CI 0.52-2.67])

↑ DBP (1.09 mm Hg [95%CI 0.48-1.69])

What should be done about it?Address the obesity.

Abdominal ObesityNCEP ATP III (Waist circ.)Men >102 cm (40 in)

Women >88 cm (>35 in)

IDF (Waist circ.)Men >90 cm

Women >80 cm

Nutritional Risk and MetSyn in Women

Framingham Offspring-Spouse Study 300 healthy women (aged 30-69 y) free of MetSyn risk factors at baseline

12-y follow-up

Tertiles of nutritional risk based on intake of 19 nutrients

Outcomes: Abdominal obesityMetSyn

Selected macronutrients• Energy• Protein• MUFA• PUFA

Millen et al Am J Clin Nutr 2000;62:623-32

Risk nutrients• Total fat• Saturated fat• Alcohol• Cholesterol• Sodium

Protective nutrients• CHO• Dietary fiber• Calcium• Selenium• Vit C, B6, B12, E• Folate• β-carotene

Age-adjusted proportions of women who complied with NCEP-ATP III dietary guidelines


0

25

50

75

100

Saturated fat Total fat CHO Fiber CHON Cholesterol

123

Nutritional risk tertile

%

<7% 25-35% 50-60% 20-30 g/d 15% <200 mg/d

Nutritional Risk and MetSyn in Women


Outcome

Nutritional Risk Score Tertile

1(n = 91)

2(n = 109)

3(n = 100)

Abdominal obesity (WC >88 cm)

1.0 (ref)RR 1.1

(95%CI 0.6-1.9)RR 2.3

(95%CI 1.2-4.3)

Metabolic Syndrome

1.0 (ref)RR 0.8

(95%CI 0.3-2.5)RR 3.0

(95%CI 1.2-7.6)

Multivariate logistic regression model adjusted for age, smoking, physical activity and menopausal status

Physical Activity and Changes in Weight and Waist Circumference in Midlife Women

Study of Women’s Health Across the Nation cohortn = 3064 women aged 42 to 52 years; 3-y follow-up

Change in weight and waist circumferenceabsolute difference: baseline and 3 yearsrelative difference: % baseline value

Exposure variablesAgeMenopausal status Physical activity

Sternfeld et al Am J Epidemiol 2004;160:912-22

Physical Activity and Weight and Waist Change 917

Am J Epidemiol 2004;160:912–922

ficient, a one-unit difference in the mean level of sports/exer-cise was inversely associated with a 2.76-kg (standard error,0.36) difference between women in weight. Similarly, a one-unit within-woman increase in that domain over time (betacoefficient for change in activity) was associated with a0.32-kg (standard error, 0.08) decrease in weight. The influ-ence of the between-women variability in daily routineactivity was of even greater magnitude (! = –3.31 (standarderror, 0.46); p < 0.0001), but the magnitude of the within-woman variability was slightly lower (! = –0.21 (standarderror, 0.10); p = 0.03). In contrast, mean household/care-giving activity had no relation with weight, although anincrease in activity in this domain was marginally associatedwith a decrease in weight (p = 0.10). Racial/ethnic differ-ences in weight, which were apparent at baseline (the meanranged from a high of 84.2 kg (SD, 20.2) in African Ameri-cans to 58.2 kg (SD, 10.8) and 56.1 kg (SD, 8.8) in theChinese and Japanese, respectively), persisted over time,with the African-American women being heavier than Whitewomen and the Chinese, Hispanic, and Japanese womenbeing lighter than White women. Weight was also directlyrelated to having a chronic medical condition and poorerperceived health and inversely related to current cigarettesmoking. Exclusion of current smokers did not alter theserelations (data not shown).

Waist circumference, which increased as weightincreased, also increased independently with older baselineage and with time. Menopausal status was not significantlyrelated to waist circumference, but waist circumferencetended to be higher in women who became postmenopausalor whose status was undetermined. Sports/exercise activityand daily routine activity were inversely related; the vari-ability in waist circumference attributable to these twodomains of activity was due primarily to differences in meanlevel (between-women differences) and less to change inactivity (within-woman variability) (p = 0.18 and p = 0.14for change in sports/exercise and daily routine, respectively).Unlike the relations between race/ethnicity and weight, waistcircumference, after adjustment for weight, was slightlylower in the African Americans than in Whites and signifi-cantly higher in Chinese, Hispanics, and Japanese. In addi-tion, current cigarette smoking, after adjustment for weight,was associated with greater waist circumference, as werepoorer perceived health and having a chronic medical condi-tion.

Change in physical activity and within-woman change in weight and waist circumference

Figures 1, 2, 3, and 4 illustrate the mean within-womanchange in weight and waist circumference for different cate-gories of change in sports/exercise and daily routine activity,adjusted for covariates, including baseline activity. In thesedomains of activity, the groups that decreased their level ofactivity gained the most weight (adjusted least-squaresmean = 2.7 kg (95 percent confidence interval (CI): 2.2, 3.3)for sports/exercise and 2.4 kg (95 percent CI: 1.9, 2.8) fordaily routine activity), and those that increased their activitygained the least (least-squares mean = 1.0 kg (95 percent CI:0.5, 1.5) for sports/exercise and 1.4 kg (95 percent CI: 0.9,

1.8) for daily routine). Similar group differences in meanwaist gain were also observed. Neither weight change norchange in waist circumference varied by change in house-hold/caregiving activity (data not shown).

Factors associated with risk of substantial gain in weight and waist circumference

The results of the multivariable logistic regression anal-yses are summarized in table 3. In general, these results areconsistent with those of the longitudinal analysis. Risk of

FIGURE 1. Mean within-woman weight change between baseline(1996–1997) and year 3 (1999–2000) according to change in thelevel of sports/exercise (on a scale of 1–5), Study of Women’s HealthAcross the Nation. Results were adjusted for baseline age, baselinelevel of sports/exercise, race/ethnicity, the presence of chronic condi-tions, and study site. For the F statistic, p < 0.0001; all pairwise com-parisons showed a significant difference at p < 0.01. Bars, 95%confidence interval.

FIGURE 2. Mean within-woman change in waist circumferencebetween baseline (1996–1997) and year 3 (1999–2000) according tochange in the level of sports/exercise (on a scale of 1–5), Study ofWomen’s Health Across the Nation. Results were adjusted for base-line age, baseline level of sports/exercise, race/ethnicity, the pres-ence of chronic conditions, and study site. For the F statistic, p <0.0001; all pairwise comparisons showed a significant difference atp < 0.05. Bars, 95% confidence interval.

Mean within-woman weight change between baseline and year 3 accdg to change in level of sports/exerciseStudy of Women’s Health Across the Nation cohort n = 3064 women

Scale of 1-5* Adjusted for baseline age, baseline level of sports/exercise, race/ethnicity, the presence of chronic conditions and study site


p <0.01

Physical Activity and Weight and Waist Change 917

Am J Epidemiol 2004;160:912–922

ficient, a one-unit difference in the mean level of sports/exer-cise was inversely associated with a 2.76-kg (standard error,0.36) difference between women in weight. Similarly, a one-unit within-woman increase in that domain over time (betacoefficient for change in activity) was associated with a0.32-kg (standard error, 0.08) decrease in weight. The influ-ence of the between-women variability in daily routineactivity was of even greater magnitude (! = –3.31 (standarderror, 0.46); p < 0.0001), but the magnitude of the within-woman variability was slightly lower (! = –0.21 (standarderror, 0.10); p = 0.03). In contrast, mean household/care-giving activity had no relation with weight, although anincrease in activity in this domain was marginally associatedwith a decrease in weight (p = 0.10). Racial/ethnic differ-ences in weight, which were apparent at baseline (the meanranged from a high of 84.2 kg (SD, 20.2) in African Ameri-cans to 58.2 kg (SD, 10.8) and 56.1 kg (SD, 8.8) in theChinese and Japanese, respectively), persisted over time,with the African-American women being heavier than Whitewomen and the Chinese, Hispanic, and Japanese womenbeing lighter than White women. Weight was also directlyrelated to having a chronic medical condition and poorerperceived health and inversely related to current cigarettesmoking. Exclusion of current smokers did not alter theserelations (data not shown).

Waist circumference, which increased as weightincreased, also increased independently with older baselineage and with time. Menopausal status was not significantlyrelated to waist circumference, but waist circumferencetended to be higher in women who became postmenopausalor whose status was undetermined. Sports/exercise activityand daily routine activity were inversely related; the vari-ability in waist circumference attributable to these twodomains of activity was due primarily to differences in meanlevel (between-women differences) and less to change inactivity (within-woman variability) (p = 0.18 and p = 0.14for change in sports/exercise and daily routine, respectively).Unlike the relations between race/ethnicity and weight, waistcircumference, after adjustment for weight, was slightlylower in the African Americans than in Whites and signifi-cantly higher in Chinese, Hispanics, and Japanese. In addi-tion, current cigarette smoking, after adjustment for weight,was associated with greater waist circumference, as werepoorer perceived health and having a chronic medical condi-tion.

Change in physical activity and within-woman change in weight and waist circumference

Figures 1, 2, 3, and 4 illustrate the mean within-womanchange in weight and waist circumference for different cate-gories of change in sports/exercise and daily routine activity,adjusted for covariates, including baseline activity. In thesedomains of activity, the groups that decreased their level ofactivity gained the most weight (adjusted least-squaresmean = 2.7 kg (95 percent confidence interval (CI): 2.2, 3.3)for sports/exercise and 2.4 kg (95 percent CI: 1.9, 2.8) fordaily routine activity), and those that increased their activitygained the least (least-squares mean = 1.0 kg (95 percent CI:0.5, 1.5) for sports/exercise and 1.4 kg (95 percent CI: 0.9,

1.8) for daily routine). Similar group differences in meanwaist gain were also observed. Neither weight change norchange in waist circumference varied by change in house-hold/caregiving activity (data not shown).

Factors associated with risk of substantial gain in weight and waist circumference

The results of the multivariable logistic regression anal-yses are summarized in table 3. In general, these results areconsistent with those of the longitudinal analysis. Risk of

FIGURE 1. Mean within-woman weight change between baseline(1996–1997) and year 3 (1999–2000) according to change in thelevel of sports/exercise (on a scale of 1–5), Study of Women’s HealthAcross the Nation. Results were adjusted for baseline age, baselinelevel of sports/exercise, race/ethnicity, the presence of chronic condi-tions, and study site. For the F statistic, p < 0.0001; all pairwise com-parisons showed a significant difference at p < 0.01. Bars, 95%confidence interval.

FIGURE 2. Mean within-woman change in waist circumferencebetween baseline (1996–1997) and year 3 (1999–2000) according tochange in the level of sports/exercise (on a scale of 1–5), Study ofWomen’s Health Across the Nation. Results were adjusted for base-line age, baseline level of sports/exercise, race/ethnicity, the pres-ence of chronic conditions, and study site. For the F statistic, p <0.0001; all pairwise comparisons showed a significant difference atp < 0.05. Bars, 95% confidence interval.

Mean within-woman waist change between baseline and year 3 accdg to change in level of sports/exerciseStudy of Women’s Health Across the Nation cohort n = 3064 women

Scale of 1-5

* Adjusted for baseline age, baseline level of sports/exercise, race/ethnicity, the presence of chronic conditions and study site


p <0.05

918 Sternfeld et al.

Am J Epidemiol 2004;160:912–922

substantial weight gain was not associated with baseline age,baseline menopausal status, or change in menopausal status.Although baseline sports/exercise was not significantly asso-ciated with risk, change in that domain of activity wasinversely associated (odds ratio = 0.77, 95 percent CI: 0.66,0.89), as were both baseline level of daily routine activityand change in daily routine activity. Of the factors associatedwith substantial gain in waist circumference, the mostnotable was the 32 percent increase in risk associated with a1-kg increase in weight. Increases over time in the sports/

exercise and daily routine indices tended to be associatedwith lower risk, but only the relation with sports/exercisewas statistically significant. Baseline smoking status was notsignificantly associated with either substantial weight gainor substantial gain in waist circumference and was notincluded in the model. Although smoking cessation wassignificantly associated with risk of substantial weight gain,only 54 women gave up smoking during follow-up, andinclusion of this variable did not alter the other relations(data not shown).

DISCUSSION

In this study of a multiethnic cohort of midlife women,significant increases in mean weight and waist circumfer-ence occurred over the 3-year follow-up period. Time was arisk factor for weight gain, but change in menopausal statuswas not, whether the outcome was considered continuously,as a repeated measure over time, or categorically, in terms ofa substantial gain. Time was also a risk factor for increases inwaist circumference, as was baseline age. Although changein menopausal status was not significantly associated withchange in waist circumference, women who became post-menopausal or who began hormone therapy tended to havelarger waists than women who remained premenopausal orearly perimenopausal. In the longitudinal analysis, higherlevels of sports/exercise and daily routine activity were inde-pendently associated with lower weight and waist circumfer-ence. Greater levels of participation in both of these domainsof activity and increases in participation over time alsotended to be associated with lower risk of substantial weightgain. Furthermore, when within-woman changes in weightand waist circumference were examined in relation to cate-gorical change in activity, women who increased theiractivity had the least amount of gain, while those whodecreased their activity had the most.

The finding that weight increased steadily over a 3-yearperiod is consistent with the well-established observationthat weight tends to increase with age in young and middle-aged adults (31, 32). It is notable that the mean within-woman weight gain of 2.1 kg in the SWAN cohort is onlyslightly lower than the 2.4-kg weight gain seen in theHealthy Women’s Study, a cohort study of primarily Whitewomen, over a similar period of time (2), although it isgreater than the increase of 2.9–4.5 kg observed over a 6-year follow-up period in Scottish women (3). These findingsimply that women in their forties and early fifties may, onaverage, expect to gain approximately 1.5 pounds (0.7 kg)per year during their middle years, regardless of initial age,initial body size, or race/ethnicity. On the other hand, therewas a large degree of individual variability in the tendency togain weight over time, and almost one fourth of the SWANwomen lost at least 2 percent of their initial body weight overthe 3-year follow-up period.

In this study, differences in physical activity in thedomains of both sports/exercise and daily routine contrib-uted to the variability in weight over time. The findings fromthe longitudinal analysis demonstrated that variability inweight over time was attributable not only to variability inactivity level between women but also to within-woman

FIGURE 3. Mean within-woman weight change between baseline(1996–1997) and year 3 (1999–2000) according to change in thelevel of daily routine physical activity (on a scale of 1–5), Study ofWomen’s Health Across the Nation. Results were adjusted for base-line age, baseline level of daily routine activity, race/ethnicity, thepresence of chronic conditions, and study site. For the F statistic, p <0.001; the group with a decrease in activity (change <0) was signifi-cantly different from the other two groups at p < 0.01. Bars, 95% con-fidence interval.

FIGURE 4. Mean within-woman change in waist circumferencebetween baseline (1996–1997) and year 3 (1999–2000) according tochange in the level of daily routine physical activity (on a scale of 1–5), Study of Women’s Health Across the Nation. Results wereadjusted for baseline age, baseline level of daily routine activity, race/ethnicity, the presence of chronic conditions, and study site. For the Fstatistic, p < 0.002; the group with a decrease in activity (change <0)was significantly different from the other two groups at p < 0.05. Bars,95% confidence interval.

Mean within-woman weight change between baseline and year 3 by change in level of daily routine activityStudy of Women’s Health Across the Nation cohort n = 3064 women

Scale of 1-5

* Adjusted for baseline age, baseline level of daily routine activity, race/ethnicity, the presence of chronic conditions and study site


p <0.01Walking or biking for transportation

Hours of TV viewing

Mean within-woman waist change between baseline and year 3 by change in level of daily routine activityStudy of Women’s Health Across the Nation cohort n = 3064 women

Scale of 1-5

* Adjusted for baseline age, baseline level of daily routine activity, race/ethnicity, the presence of chronic conditions and study site


Walking or biking for transportation

Hours of TV viewing

918 Sternfeld et al.

Am J Epidemiol 2004;160:912–922

substantial weight gain was not associated with baseline age,baseline menopausal status, or change in menopausal status.Although baseline sports/exercise was not significantly asso-ciated with risk, change in that domain of activity wasinversely associated (odds ratio = 0.77, 95 percent CI: 0.66,0.89), as were both baseline level of daily routine activityand change in daily routine activity. Of the factors associatedwith substantial gain in waist circumference, the mostnotable was the 32 percent increase in risk associated with a1-kg increase in weight. Increases over time in the sports/

exercise and daily routine indices tended to be associatedwith lower risk, but only the relation with sports/exercisewas statistically significant. Baseline smoking status was notsignificantly associated with either substantial weight gainor substantial gain in waist circumference and was notincluded in the model. Although smoking cessation wassignificantly associated with risk of substantial weight gain,only 54 women gave up smoking during follow-up, andinclusion of this variable did not alter the other relations(data not shown).

DISCUSSION

In this study of a multiethnic cohort of midlife women,significant increases in mean weight and waist circumfer-ence occurred over the 3-year follow-up period. Time was arisk factor for weight gain, but change in menopausal statuswas not, whether the outcome was considered continuously,as a repeated measure over time, or categorically, in terms ofa substantial gain. Time was also a risk factor for increases inwaist circumference, as was baseline age. Although changein menopausal status was not significantly associated withchange in waist circumference, women who became post-menopausal or who began hormone therapy tended to havelarger waists than women who remained premenopausal orearly perimenopausal. In the longitudinal analysis, higherlevels of sports/exercise and daily routine activity were inde-pendently associated with lower weight and waist circumfer-ence. Greater levels of participation in both of these domainsof activity and increases in participation over time alsotended to be associated with lower risk of substantial weightgain. Furthermore, when within-woman changes in weightand waist circumference were examined in relation to cate-gorical change in activity, women who increased theiractivity had the least amount of gain, while those whodecreased their activity had the most.

The finding that weight increased steadily over a 3-yearperiod is consistent with the well-established observationthat weight tends to increase with age in young and middle-aged adults (31, 32). It is notable that the mean within-woman weight gain of 2.1 kg in the SWAN cohort is onlyslightly lower than the 2.4-kg weight gain seen in theHealthy Women’s Study, a cohort study of primarily Whitewomen, over a similar period of time (2), although it isgreater than the increase of 2.9–4.5 kg observed over a 6-year follow-up period in Scottish women (3). These findingsimply that women in their forties and early fifties may, onaverage, expect to gain approximately 1.5 pounds (0.7 kg)per year during their middle years, regardless of initial age,initial body size, or race/ethnicity. On the other hand, therewas a large degree of individual variability in the tendency togain weight over time, and almost one fourth of the SWANwomen lost at least 2 percent of their initial body weight overthe 3-year follow-up period.

In this study, differences in physical activity in thedomains of both sports/exercise and daily routine contrib-uted to the variability in weight over time. The findings fromthe longitudinal analysis demonstrated that variability inweight over time was attributable not only to variability inactivity level between women but also to within-woman

FIGURE 3. Mean within-woman weight change between baseline(1996–1997) and year 3 (1999–2000) according to change in thelevel of daily routine physical activity (on a scale of 1–5), Study ofWomen’s Health Across the Nation. Results were adjusted for base-line age, baseline level of daily routine activity, race/ethnicity, thepresence of chronic conditions, and study site. For the F statistic, p <0.001; the group with a decrease in activity (change <0) was signifi-cantly different from the other two groups at p < 0.01. Bars, 95% con-fidence interval.

FIGURE 4. Mean within-woman change in waist circumferencebetween baseline (1996–1997) and year 3 (1999–2000) according tochange in the level of daily routine physical activity (on a scale of 1–5), Study of Women’s Health Across the Nation. Results wereadjusted for baseline age, baseline level of daily routine activity, race/ethnicity, the presence of chronic conditions, and study site. For the Fstatistic, p < 0.002; the group with a decrease in activity (change <0)was significantly different from the other two groups at p < 0.05. Bars,95% confidence interval.

p <0.05

Maintain or increase physical activity in midlife

to prevent or attenuate increase in weight and

waist circumference

Androgen predominance

(Menopausal transition)

Centralobesity

Metabolic Syndrome

Genetic predisposition

Environmental influence

Thank Youhttp://www.endocrine-witch.info

http://www.endocrine-witch.info

http://www.endocrine-witch.info

Metabolic Syndrome and Perimenopause

Health & Medicine