Social Science and Medicine 52 (2001) 345–356

Is there a menopausal syndrome? Menopausal status andsymptoms across racial/ethnic groups

Nancy E. Avisa,*, Rebecca Stellatoa, Sybil Crawforda, Joyce Brombergerb,Patricia Ganzc, Virginia Caind, Marjorie Kagawa-Singere

aNew England Research Institutes, 9 Galen Street, Watertown, MA 02472, USAbDepartments of Epidemiology and Psychiatry, University of Pittsburgh, Pittsburgh, PA, USA

cUniversity of California at Los Angeles, School of Medicine, Los Angeles, CA, USAdOffice of Behavioral and Social Sciences Research, National Institutes of Health, Bethesda, MA, USA

eUniversity of California at Los Angeles, School of Public Health, Los Angeles, CA, USA

Abstract

In recent years, research on menopausal symptomalogy has focused on identifying symptom groupings experiencedby women as they progress from premenopausal to postmenopausal status. However, most of these studies have been

conducted among Caucasian women from western cultures. This leaves open the question of whether the findings fromthese studies can be extended to women of other racial/ethnic groups or cultures. Furthermore, many of the previousstudies have been conducted on relatively small samples. This paper addresses the diversity of the menopause experience

by comparing symptom reporting in a large cross-sectional survey of women aged 40–55 years among racial/ethnicgroups of women in the United States (Caucasian, African-American, Chinese, Japanese, and Hispanic). Evaluation ofthe extent to which symptoms group together and consistently relate to menopausal status across these five samples

provides evidence for or against a universal menopausal syndrome. The specific research questions addressed in thispaper are: (1) How does the factor structure of symptoms among mid-aged women compare across racial/ethnicgroups? (2) Is symptom reporting related to race/ethnicity or menopausal status? and (3) Does the relation between

menopausal status and symptoms vary across racial/ethnic groups?Analyses are based on 14,906 women who participated in the multi-ethnic, multi-race, multi-site study of mid-aged

women called the Study of Women’s Health Across the Nation (SWAN). Study participants completed a 15-mintelephone or in-person interview that contained questions on a variety of health-related topics. Items of interest for these

analyses include symptoms, menstrual history (to assess menopausal status), health status, and sociodemographics.Factor analysis results showed that across all five racial/ethnic groups, two consistent factors emerged; one consisting

of clearly menopausal symptoms } hot flashes and night sweats } and the other consisting of psychological and

psychosomatic symptoms. Results of regression analyses showed racial/ethnic differences in symptom reporting, as wellas differences by menopausal status. Controlling for age, education, health, and economic strain, Caucasian womenreported significantly more psychosomatic symptoms than other racial/ethnic groups. African-American women

reported significantly more vasomotor symptoms. Perimenopausal women, hormone users, and women who had asurgical menopause reported significantly more vasomotor symptoms. All of these groups, plus postmenopausalwomen, reported significantly more vasomotor symptoms than premenopausal women. The pattern of resultsargues against a universal menopausal syndrome consisting of a variety of vasomotor and psychological

symptoms. # 2001 Elsevier Science Ltd. All rights reserved.

Keywords: Menopause; Symptoms; Ethnicity; Culture; United States

*Corresponding author. Tel.: +1-617-923-7747 ext. 310; fax: +1-617-926-8246.

E-mail address: nancya@neri.org (N.E. Avis).

0277-9536/01/$ - see front matter # 2001 Elsevier Science Ltd. All rights reserved.

PII: S 0 2 7 7 - 9 5 3 6 ( 0 0 ) 0 0 1 4 7 - 7

Introduction

A variety of symptoms are frequently reported asbeing part of a menopausal syndrome. These include hotflushes or flashes, night sweats, menstrual irregularities,

vaginal dryness, as well as other symptoms such asdepression, nervous tension, palpitations, headaches,insomnia, lack of energy, difficulty concentrating, anddizzy spells (WHO Scientific Group, 1996). The question

of whether a universal ‘‘menopausal syndrome’’ existshas been debated for some time (Donovan, 1951;Neugarten & Kraines, 1965; McKinlay & Jefferys,

1974; Holte & Mikkelsen, 1991). Clinicians havesuggested a constellation of symptoms experienced bymost women due to declining levels of estrogen as they

pass through menopause. An opposing view is that nouniversal syndrome exists. Rather, women experience avariety of symptoms and the ones experienced and

reported by women are due to individual and culturalinfluences.The question of whether a single menopausal

‘‘syndrome’’ exists and whether this is a universal

phenomenon can be addressed in several ways: (1) useof data reduction techniques such as studies of clustersor groupings of symptoms, (2) the consistency of

associations between specific symptoms and menopausalstatus across studies, and (3) comparisons of symptomreporting according to menopausal status across cul-

tures.This paper directly addresses the diversity of the

menopause experience by comparing symptom reportingamong five racial/ethnic groups of women in the US

who are all part of a single study and all aged 40–55 atthe time of survey: (non-Hispanic) Caucasian, African-American, Chinese, Japanese, and Hispanic. While all

the women reside in the United States, the culturalexperiences of the various groups differ widely. Evalua-tion of the extent to which symptoms group together

and consistently relate to menopausal status across thesefive populations provides evidence for or against amenopausal syndrome and the universality of this

experience. The specific research questions addressed inthis paper are: (1) How does the factor structure ofsymptoms among mid-aged women compare acrossracial/ethnic groups? (2) Is symptom reporting related

to race/ethnicity or menopausal status? and (3) Does therelation between menopausal status and symptoms varyacross racial/ethnic groups?

Factor analyses of symptoms

A number of researchers have used factor analysis andrelated approaches to determine how menopausalsymptoms group together. One of the earliest such

studies was conducted by McKinlay and Jefferys (1974)in which they assessed relationships between symptoms,

as well as sociodemographic variables among a sampleof 638 women aged 45–54. Pairwise correlation coeffi-

cients were transformed to represent distances betweentwo-dimensional points such that these distances in-dicated strength of pairwise associations. Results of this

analysis showed that, except for hot flashes, none of thesymptoms was related to menopause status. A similaranalysis was conducted by Boulet, Oddens, Lehert,Vemer and Visser (1994) among 2992 women from seven

south-east Asian countries. This analysis yielded twofactors. One factor, consisting of vasomotor symptoms,dizziness, palpitations, and incontinence, was associated

with perimenopausal status. The other factor, whichincluded depression, insomnia, headache, anxiety, andirritability, was more closely associated with postmeno-

pausal status.The first reported factor analysis of symptoms was

conducted by Greene (1976) who factor analyzed a list

of 30 symptoms from a study of 50 British women aged40–55 who were attending a hormone replacement clinicin London. Greene found three factors, which he labeledvasomotor, somatic, and psychological. Since Greene’s

small, clinic-based study, numerous large, populationand clinic-based studies have conducted factor analysesof symptom lists.

Holte and Mikkelsen (1991) reported results of afactor analysis of 21 symptoms from 1566 Norwegianwomen aged 45–55 who were not taking HRT that

yielded five factors which they labeled vasomotorcomplaints, mood lability, nervousness, vague somaticcomplaints, and urogenital complaints. Dennerstein etal. (1993) conducted a factor analysis of 22 symptoms

among a population-based sample of 1897 women aged45–55 in Melbourne Australia and found seven firstorder factors: dysphoria, vasomotor, cardiopulmonary,

skeletal, digestive, respiratory, and general somatic. Asecond order analyses, yielded three factors: vasomotor,general respiratory, and psychosomatic. Collins and

Landgren (1995) report results of a study of 1324Swedish women all aged 48 years. From a 20-itemsymptom list, they found the following four factors:

negative mood, vasomotor symptoms, decreased sexualdesire, and positive well-being. This was the only studythat included aspects of positive mood. Other factorstudies have also been conducted in the US (Mitchell &

Woods, 1996) and Britain (Hunter, Battersby & White-head, 1986; Kuh, Wadsworth & Hardy, 1997).One study reported by Avis, Kaufert, Lock, McKin-

lay and Vass (1993) compared the factor structure ofsymptoms in three distinct populations (Canada, US,and Japan). All three studies were based on similar

sample characteristics (i.e., community-based; aged 45–55; and pre-, peri-, and naturally postmenopausalwomen). Avis et al. found somewhat different factor

structures across the three samples, which could havebeen due to the differences in number of symptoms

N.E. Avis et al. / Social Science and Medicine 52 (2001) 345–356346

assessed among the three studies or to true differences inthe types of symptoms experienced. For each popula-

tion, however, the vasomotor symptoms loaded on afactor separate from any of the psychological or somaticsymptoms.

These studies differ in terms of the specific symptomsstudied, number of symptoms included in the list(ranging from 20 to 36), time frame for symptomreporting (from the past two weeks up to one year), as

well as cut-point for factor loadings. Studies also differin sample characteristics such as age of sample (twostudies included only women aged 47 or 48, while others

included wider age ranges), composition of sample(some excluded women taking estrogen), and whetherthe sample was clinic- or community-based. Studies were

conducted in a variety of countries including the UnitedStates (Avis et al., 1993; Mitchell & Woods, 1996),Canada (Avis et al., 1993; Kaufert, Gilbert & Hassard,

1988), Australia (Dennerstein et al., 1993), Great Britain(Hunter et al., 1986; Kuh et al., 1997), Sweden (Collins& Landgren, 1995), Norway (Holte & Mikkelson, 1991),Japan (Avis et al., 1993), and south-east Asia (Boulet et

al., 1994). Despite these differences, the results areoverwhelmingly consistent in one respect; in every study,vasomotor symptoms load on a separate factor from

psychological symptoms.Except for the Avis et al. and Boulet et al. studies,

these analyses are all based on samples of Caucasian

women. We have little data on how symptoms grouptogether for non-Caucasian women.

Relation between menopausal status and symptoms

The majority of studies examining the relationbetween menopausal status and symptoms find vaso-

motor symptoms occurring more frequently amongpostmenopausal women (Collins & Landgren, 1995;Hunter et al., 1986; Kuh et al., 1997; Dennerstein et al.,

1993; Dennerstein, Smith & Morse, 1994; Kaufert et al.,1988). However, findings from these and other studiesare less consistent for psychological symptoms. The

majority of cross-sectional studies do not find a relationbetween menopausal status and various measures ofmood disturbance (Dennerstein et al., 1993; 1994; Holte

& Mikkelsen, 1991; Hunter et al., 1986; Kuh et al., 1997;Porter, Penney, Russell, Russell & Templeton, 1996;Ballinger, 1976; Cawood & Bancroft, 1996), althoughseveral studies have found greater psychological or

mood symptoms among perimenopausal women (Bal-linger, 1975; Avis et al., 1993; Hunter et al., 1986). Thesestudies, however, have generally not controlled for

vasomotor symptoms. When vasomotor symptoms areincluded in the model, this relationship has attenuated(Avis et al., 1994; Collins & Landgren, 1995). These

studies are based on Western, largely Caucasian,women.

Cross-cultural comparisons

Several studies of non-Western women suggestcultural differences in menopausal symptoms. Forexample, in a study of 483 Indian women of the Rajput

caste in India, Flint (1975) found that few women hadany problems with menopause other than cycle changes;they reported no depression, dizziness or incapacitation.Lock has extensively studied Japanese women and finds

that only a small proportion of Japanese women aged45–55 experience depressive symptoms or irritability andthese symptoms varied little with menopausal status.

Rates of hot flushes and night sweats were low incomparison with those reported in Western culture(Lock, 1986). In a cross-cultural comparison of the rates

of somatic and psychological symptoms, Avis andcolleagues (1993) reported that rates of almost everysymptom were lower in the Japanese than in groups of

US and Canadian women of similar ages. Furthermore,the percentage reporting five or more symptoms wassubstantially lower in the Japanese compared to the USand Canadian women.

Two recent studies found considerable variation insymptom reporting among women in different south-east Asian countries (Boulet et al, 1994; Punyahotra,

Dennerstein & Lehert, 1997). These samples, however,were not representative and varied considerably withregard to many sociodemographic variables that might

affect symptoms. The authors cite these differences, butdid not conduct analyses controlling for these variables.Other cross-cultural studies have significant metho-

dological limitations, such as small sample sizes, use of

patient samples, lack of a comparison group, use of non-standardized measures, and/or unclear definitions ofmenopausal status (Davis, 1982; Indiria & Murthy,

1980; Flint & Garcia, 1979).Cross-cultural/ethnic studies of menopause are very

limited in the United States. One study in the

Philadelphia area found similar levels of psychological,vasomotor, genitourinary, and general physical symp-toms in a convenience sample of 33 African-American

and 35 Caucasian premenopausal women, aged 44–49(Pham, Grisso & Freeman, 1997). In contrast, thePittsburgh Healthy Women Study (unpublished data)found that African-American women, compared to

Caucasian women, had significantly higher levels ofdepressive symptoms and tension. The size of theAfrican-American sample, however, was very small.

In summary, epidemiologic studies of menopausehave been carried out largely in Western cultures onCaucasian women. Those conducted in other cultures

have used differing measures, varying sampling metho-dology and composition, and yielded inconsistentfindings. Thus, we do not know to what extent these

findings are generalizable to non-Western and for non-white women.

N.E. Avis et al. / Social Science and Medicine 52 (2001) 345–356 347

Methods

Data in these analyses are from the Study of Women’sHealth Across the Nation (SWAN), a multi-race, multi-ethnic, multi-site study of middle-aged women from

across the United States. SWAN is comprised of twostages: a cross-sectional telephone or in-home surveyconducted between November 1995 and October 1997and a longitudinal investigation to track changes in

women’s physical and mental health as they age andtraverse the menopause transition. SWAN’s studydesign has been described in detail elsewhere (Sowers

et al., 2000). Briefly, community-based samples ofwomen were drawn from seven geographic locations:Boston, Massachusetts; Chicago, Illinois; Detroit, Mi-

chigan; Los Angeles, California; Newark, New Jersey;Oakland, California; and Pittsburgh, Pennsylvania.Women self-identifying primarily with one or more of

the following five racial/ethnic groups were interviewed:Caucasian, African-American, Chinese, Hispanic, andJapanese. Each site studied Caucasians and one otherethnic group; African-Americans were sampled at four

sites, and each of the other non-Caucasian ethnic groupswas sampled at a single site. A 15-min telephone or in-home interview was conducted in order to determine

eligibility for SWAN’s longitudinal cohort phase. Inaddition to cohort eligibility criteria, informationregarding other characteristics was collected, including

sociodemographic factors and symptoms. To be eligiblefor participation in this cross-sectional phase of SWAN,a woman had to reside in an appropriate geographicarea, to speak a designated study language (English,

Cantonese, Japanese, or Spanish), to be aged 40–55 atthe time of initial contact, and to be cognitively able toprovide verbal informed consent. A total of 16,065

women across the seven sites completed a cross-sectionalinterview.The cross-sectional instrument was translated into

Cantonese, Japanese, and Spanish. During the first fewmonths of recruitment, it was determined that severalitems for the Cantonese and Japanese instruments were

not accurately translated, and the translations wererevised accordingly. Relevant to the analyses presentedhere, data from 131 women receiving the originalJapanese translation were coded as missing for the two

symptoms with problematic translations: trouble sleep-ing and leaking urine. Because this constituted over 16%of the Japanese sample, these two symptoms were

omitted from analyses and are not discussed further.The Cantonese translation revision did not affect any ofthe variables considered here.

Measures

The SWAN screening questionnaire included ques-tions on health, reproductive history, lifestyle, stress,

social support, and sociodemographics. The items ofinterest for these analyses include symptoms, menopau-

sal status, health status, and sociodemographics (age,education, and financial strain).

Symptoms

The questionnaire inquired about the presence orabsence of twelve symptoms in the last two weeks. The

symptom list was selected from both endocrinologic andepidemiological literature (Avis et al., 1993; Matthews,Wing, Kuller, Meilahn & Plantinga, 1994; Neugarten &

Kraines, 1965) and from focus groups with women ofthe racial/ethnic groups being studied. Women wereasked whether or not they had experienced each of the

following symptoms within the last two weeks: nightsweats; stiffness or soreness in joints, neck or shoulders;headaches; hot flashes or flushes; forgetfulness; feeling

tense or nervous; feeling blue or depressed; vaginaldryness; irritability or grouchiness; and heart poundingor racing.

Menopausal status

Menopausal status was determined from answers to a

series of questions about menstrual patterns andgynecologic surgery. Women who had undergone ahysterectomy and/or a bilateral oophorectomy were

defined as surgically menopausal. Women who had nomenstrual bleeding in the previous twelve months (notdue to medication, pregnancy, or severe weight loss)

were defined as postmenopausal. Women with menses inthe previous twelve months but not in the previous threewere considered late perimenopausal. Early perimeno-pausal women were those who had menstrual bleeding

in the previous three months, but who had experiencedincreasing irregularity in cycle length over the past year.Because the late perimenopausal group was small, early

and late perimenopausal women were combined in allanalyses. Premenopausal women were those who re-ported menses in the previous three months, with no

increase in irregularity. All peri- and premenopausalwomen who reported using female hormones in the pastthree months were kept in a separate stratum becausethe predominant female hormone used in this age group

of women is HRT, and their observed menstrualbleeding could have been due to the HRT rather thanto true menopausal status.

Race/ethnicity

Race/ethnicity was self-determined by respondents,and was obtained by asking the following open-endedquestion: ‘‘How would you describe your primary racial

or ethnic group?’’. The responses then were categorizedas Caucasian, African-American, Chinese, Japanese,

N.E. Avis et al. / Social Science and Medicine 52 (2001) 345–356348

and Hispanic. The Hispanic category included women ofPuerto Rican, Dominican, Cuban, Mexican, Central

American, South American, Spanish, or other Hispanicdescent; these groups were combined in analyses in asingle Hispanic category due to small sample sizes. A

relatively small number of women self-reported aprimary ethnicity other than the above categories(N=357) or a mixed ethnicity (N=44). An additional49 women did not provide information on race or

ethnicity. Because these subgroups did not constitutehomogeneous groups with respect to race/ethnicity,interpretation of racial/ethnic comparison with the other

five ethnic categories would be difficult at best. Conse-quently, these 450 women were omitted from analysespresented here.

Covariates

Several covariates were included in the regressionmodel because previous research has shown they arerelated to both race/ethnicity and symptoms. Theseincluded age, education, self-assessed health, and

financial strain. Age was a continuous variable from40 to 55. Education was categorized as less than highschool, high school degree, some college or vocational

school, college graduate, or postgraduate degree. Wo-men were asked to rate their health as excellent, verygood, good, fair, or poor. The numbers of women

reporting fair or poor health were small, and the twocategories were combined for analyses. Because manysurvey respondents are reluctant to report income these

data were not collected in the interview. Ability to payfor basics such as housing, food and clothing was usedas a measure of economic strain. Income was notmeasured. Respondents were asked how hard it was to

pay for basics: very hard, somewhat hard, or not at allhard.

Analysis

Of the 16,065 women who participated in the cross-

sectional survey, 450 women were excluded from theseanalyses because they did not supply their racial/ethnicidentification, or because they self-identified as ‘‘other’’

(N=357) or ‘‘mixed’’ (N=44). For another 341 women,menopausal status could not be determined and 55women were pregnant or breastfeeding, all of whomwere excluded. An additional 313 women were excluded

due to incomplete data on symptoms or covariates.After all exclusions, 14,906 women remained in theanalytic sample.

Women with any missing data comprised a smallportion (2.1%) of the total sample available for analysis,and the rates of missingness for individual variables

ranged from 0.08% for headaches to 0.51% for hotflashes. Comparing women with any missing data to

women with no missing data, the former group wasmore likely to report forgetfulness (46.6% vs 39.9%,

p=0.02), hot flashes (34.3% vs 27.5%, p=0.02) andnight sweats (30.9% vs 23.5%, p=0.004), less likely toreport excellent or very good health (41.0% vs 51.3%,

p=0.001 for overall health comparison), and were morelikely to be African-American (37.1% vs 27.9%) and lesslikely to be Caucasian (40.3% vs 50.0%, p=0.003 foroverall comparison). Missingness was unrelated, how-

ever, to any other subject characteristics consideredhere. Moreover, given that only 2% of subjects had anymissing data, any nonresponse bias in estimates should

be very small.In preliminary analyses, the five racial/ethnic groups

were compared with respect to the 10 symptoms and the

predictors and covariates of interest, using chi-squaretests and analysis of variance. For the first study aim,factor analyses of the ten symptoms questions were

carried out to assess the grouping of symptoms. Theseanalyses were conducted for the full analytic sample andfor each racial/ethnic group separately. For all factoranalyses, varimax rotation was used and the number of

factors was left unspecified (Johnson & Wichern, 1982).The resulting racial/ethnic-specific factor structures werecompared across the five racial/ethnic groups.

The primary purpose of the factor analysis was to‘‘identify empirically derived clusters of items that tendto behave as groups’’ (Kim & Mueller, 1978). Perfor-

mance of factor analysis is best for normally distributedvariables; however, Kim and Mueller note that factoranalysis can be used in order to search for suchclustering patterns of binary variables. As has been

suggested previously for the case of binary items(Muthen, 1986; Nisenbaum, Reyes, Mawle & Reeves,1998; Schoenberg & Arminger, 1986), we estimated

exploratory factor analyses on a random half sample foreach ethnic group, and estimated a confirmatory factoranalysis (Floyd & Widaman, 1995) on the remaining

half sample, using the structure indicated by the analyseson the first half-sample. For all racial/ethnic groups, thefit of the confirmatory analysis was excellent } good-

ness-of-fit index (Floyd & Wadaman, 1995) was 0.90 orhigher } indicating high stability of the groupings ofitems.Addressing the second aim required identifying a set

of symptoms common to all racial/ethnic groups withsimilar covariance structure, i.e., similar factor loadings,then creating a separate scale outcome for each factor.

As presented in the following section, factor analysesidentified two factors. To construct scales from thesefactor, individual symptoms were required to have a

factor loading of 0.40 or greater, and not to load onmore than one factor. For each factor, a scale wascomputed as the sum of the symptom variables that

loaded on that factor. This resulted in two scales, oneranging from 0 to 5 and the other from 0 to 2. For these

N.E. Avis et al. / Social Science and Medicine 52 (2001) 345–356 349

two outcomes, separate multiple regression models wereestimated to investigate the relationships with ethnicity,

controlling for potential confounders.A linear regression model (Weisberg, 1980) was

estimated for the first scale. An ordinal logistic regres-

sion model (Agresti, 1984) was estimated for the secondscale, but the model assumption of proportional oddswas not satisfied; that is, the regression coefficients werenot similar for distinguishing 0 vs 1 symptom and 1 vs 2

symptoms. Consequently, polychotomous (multinomial)logistic regression was employed instead (Hosmer &Lemeshow, 1989). This approach was used rather than

modeling of hot flashes and night sweats separately,because it allowed an assessment of predictors of oneversus two symptoms; moreover, as implied by the

factor structure, predictors were similar for the twoindividual binary symptom variable. Models for bothoutcomes controlled for age, education, self-assessed

health, financial strain, and geographic site. Possibleinteractions between race/ethnicity and site were as-sessed by including interaction terms between African-American race and site in the multivariate models,

thereby testing whether the difference between African-Americans and Caucasians at the same site varied acrossfour of the seven sites. Note that estimation of site

interactions with other racial/ethnic groups (Chinese,Hispanic, and Japanese) was not possible, because eachof these three groups were studied at a single site. Based

on the multivariate models, adjusted means for thecontinuous outcome and adjusted probabilities for thecategorical outcome were calculated for the five racial/ethnic groups.

For the third study aim, interaction terms betweenrace/ethnicity and menopausal status were added toregression models including race/ethnicity and status,

and their statistical significance was assessed.

Results

Sample

Table 1 shows the characteristics of the analyticsample for the full sample and by race/ethnicity. All

variables, including menopausal status, were related toethnicity. Compared with all other groups surveyed,African-American women were much more likely tohave had a surgical menopause, while Chinese and

Japanese women were less likely to have had a surgicalmenopause than Caucasian or Hispanic women. TheChinese and Japanese women were more likely to be

premenopausal. Hormone use was highest among theCaucasian women and lowest among the African-American and Hispanic women. Because menopausal

status and hormone use are related to race/ethnicity andare thought to be related to symptom reporting, it is

critical to include menopausal status and hormone use inany analyses of race/ethnicity and symptoms. Analyses

comparing women without any symptoms with thosewho reported one or more symptoms indicated that allcovariates were related to symptoms as well as race/

ethnicity. Therefore, all potential covariates wereincluded in the regression models.

Factor analyses by ethnicity

Our first research question addressed the comparisonof symptom reporting across race/ethnicity. Table 2

shows the frequency distributions of the individualsymptoms across each of the racial/ethnic groups andfor the whole sample. Tensed, depressed, irritable,

headache, and stiffness were all frequently reportedsymptoms. In general, the Japanese and Chinese wereless likely than the other groups to report symptoms.

Table 3 presents the results of the factor analysis foreach racial/ethnic group. These results are most remark-able for their consistency across groups. While symp-toms vary in their loadings across groups, there are more

similarities than differences. Across all groups there aretwo factors. Factor 1 consists of a combination ofpsychological or mood symptoms and vague somatic

symptoms. Factor 2 primarily consists of hot flashes andnight sweats, the primary symptoms of menopause.While vaginal dryness also loads on this factor, its

loading is considerably lower than hot flashes and nightsweats. For all racial/ethnic groups, the vasomotorsymptoms associated with menopause load on a factor

distinct from the psychological and somatic symptoms.These results suggest a highly similar factor structure ofsymptoms across racial/ethnic groups.Our second research question addresses the associa-

tion between symptom reporting and race/ethnicity andmenopausal status. Addressing this question, requiredfirst determining a set of symptoms common to all

racial/ethnic groups that could form a scale. Based onour criteria that a symptom had to load at least 0.40 on afactor for all groups and not load on more than one

factor, the following symptoms formed the scales.Factor 1 consists of tense, blue or depressed, irritable,forgetfulness, and headaches. We labeled this factorpsychosomatic symptoms. Factor 2 consists of hot flashes

or flushes and night sweats (vaginal dryness did notreach the 0.40 cut-point for the Caucasian women). Welabeled this factor vasomotor symptoms.

Menopausal status, race/ethnicity and psychosomaticsymptoms

Table 4 presents the results of the multiple linearregression of the continuous outcome psychosomatic

symptoms on menopausal status and race/ethnicity,controlling for age, education, health, and economic

N.E. Avis et al. / Social Science and Medicine 52 (2001) 345–356350

Table 1

Description of the sample. Participants were 14,906 women aged 40–55 who completed the SWAN cross-sectional questionnaire, 1995–

1997

Racial/Ethnic Group

Whole sample African-American Hispanic Chinese Japanese Caucasian

N 14,906 4163 1859 625 811 7448

Agea Mean (SD) 47.1 (4.3) 47.3 (4.3) 46.8 (4.4) 46.2 (3.9) 47.0 (4.3) 47.1 (4.4)

Educationa

5high school 1461 (9.8%) 355 (8.5%) 658 (35.4%) 102 (16.3%) 5 (0.6%) 341 (4.6%)

HS graduate 3910 (26.2%) 1164 (28.0%) 510 (27.4%) 109 (17.4%) 156 (19.2%) 1971 (26.5%)

Some college 4720 (31.7%) 1652 (39.7%) 419 (22.5%) 141 (22.6%) 292 (36.0%) 2216 (29.7%)

College grad 2453 (16.5%) 480 (11.5%) 200 (10.8%) 161 (25.8%) 247 (30.5%) 1365 (18.3%)

Postgraduate 2362 (15.8%) 512 (12.3%) 72 (3.9%) 112 (17.9%) 111 (13.7%) 1555 (20.9%)

Menopausal statusa

Surgical 3093 (20.8%) 1332 (32.0%) 334 (18.0%) 36 (5.8%) 81 (10.0%) 1310 (17.6%)

Post 2182 (14.6%) 549 (13.2%) 356 (19.2%) 73 (11.7%) 90 (11.1%) 1114 (15.0%)

Peri 4060 (27.2%) 1139 (27.4%) 464 (25.0%) 175 (28.0%) 203 (25.0%) 2079 (27.9%)

Pre 4426 (29.7%) 1001 (24.0%) 629 (33.8%) 288 (46.1%) 365 (45.0%) 2143 (28.8%)

Hormone user 1145 (7.7%) 142 (3.4%) 76 (4.1%) 53 (8.5%) 72 (8.9%) 802 (10.8%)

Self-assessed healtha

Excellent 2862 (19.2%) 532 (12.8%) 180 (9.7%) 95 (15.2%) 152 (18.7%) 1903 (25.5%)

Very good 4789 (32.1%) 1139 (27.4%) 304 (16.4%) 178 (28.5%) 249 (30.7%) 2919 (39.2%)

Good 4573 (30.7%) 1525 (36.6%) 715 (38.5%) 205 (32.8%) 239 (29.5%) 1889 (25.4%)

Fair 2193 (14.7%) 793 (19.0%) 549 (29.5%) 131 (21.0%) 146 (18.0%) 574 (7.7%)

Poor 489 (3.3%) 174 (4.2%) 111 (6.0%) 16 (2.6%) 25 (3.1%) 163 (2.2%)

Difficult to pay for basicsa

Not at all 8330 (55.9%) 2152 (51.7%) 437 (23.5%) 424 (67.8%) 578 (71.3%) 4739 (63.6%)

Somewhat 4850 (32.5%) 1490 (35.8%) 866 (46.6%) 165 (26.4%) 208 (25.6%) 2121 (28.5%)

Very 1726 (11.6%) 521 (12.5%) 556 (29.9%) 36 (5.8%) 25 (3.1%) 588 (7.9%)

Sitea

Detroit 2416 (16.2%) 1437 (34.5%) 1 (0.1%) 0 (0.0%) 0 (0.0%) 978 (13.1%)

Boston 2012 (13.5%) 1082 (26.0%) 12 (0.6%) 0 (0.0%) 0 (0.0%) 918 (12.3%)

Chicago 1322 (8.9%) 805 (19.3%) 1 (0.1%) 0 (0.0%) 0 (0.0%) 516 (6.9%)

Oakland 1427 (9.6%) 19 (0.5%) 19 (1.0%) 624 (99.8%) 1 (0.1%) 764 (10.3%)

Los Angeles 2096 (14.0%) 13 (0.3%) 29 (1.6%) 1 (0.2%) 810 (99.9%) 1243 (16.7%)

Newark 3130 (21.0%) 14 (0.3%) 1796 (96.6%) 0 (0.0%) 0 (0.0%) 1320 (17.7%)

Pittsburgh 2503 (16.8%) 793 (19.1%) 1 (0.1%) 0 (0.0%) 0 (0.0%) 1709 (23.0%)

aAll differences by ethnicity significant (p50.0001).

Table 2

Racial/ethnic-specific prevalence of individual symptoms

Racial/ethnic group

Symptom Whole sample African-American Hispanic Chinese Japanese Caucasian

Tense 7747 (51.9%) 1923 (46.2%) 1117 (60.1%) 245 (39.2%) 279 (34.4%) 4183 (56.1%)

Depressed 6039 (40.5%) 1699 (40.8%) 962 (51.8%) 200 (32.0%) 180 (22.2%) 2998 (40.2%)

Irritable 7701 (51.6%) 2108 (50.6%) 816 (43.9%) 271 (43.4%) 324 (40.0%) 4182 (56.1%)

Headache 7115 (47.7%) 1850 (44.4%) 990 (53.2%) 243 (38.9%) 307 (37.9%) 3725 (50.0%)

Stiffness 8102 (54.3%) 2346 (56.3%) 884 (47.6%) 300 (48.0%) 403 (49.8%) 4169 (55.9%)

Hot flashes 4098 (27.5%) 1617 (38.8%) 484 (26.0%) 97 (15.5%) 96 (11.8%) 1804 (24.2%)

Night sweats 3512 (23.6%) 1352 (32.4%) 459 (24.7%) 68 (10.9%) 75 (9.3%) 1558 (20.9%)

N.E. Avis et al. / Social Science and Medicine 52 (2001) 345–356 351

Table 3

Factor loadings for all items on Factor I (psychosomatic) and Factor II (vasomotor), by racial/ethnic group

African-American Hispanic Chinese Japanese Caucasian

Item Factor I Factor II Factor I Factor II Factor I Factor II Factor I Factor II Factor I Factor II

Tense or nervous 0.75 0.11 0.72 0.11 0.70 0.08 0.74 0.04 0.73 0.05

Blue or depressed 0.73 0.07 0.71 0.11 0.70 0.12 0.64 0.11 0.67 0.10

Irritable 0.70 0.05 0.64 0.09 0.67 0.01 0.70 ÿ0.04 0.67 0.03

Forgetfulness 0.50 0.30 0.55 0.17 0.59 0.18 0.44 0.30 0.48 0.25

Headaches 0.48 0.09 0.52 0.10 0.40 0.02 0.59 ÿ0.01 0.51 0.06

Stiffness 0.45 0.28 0.42 0.33 0.51 0.13 0.40a 0.26 0.45 0.20

Heart pounding 0.44 0.26 0.48 0.34 0.38 0.36 0.36 0.27 0.40a 0.32

Hot flashes 0.08 0.83 0.15 0.80 0.10 0.76 0.02 0.74 0.04 0.83

Night sweats 0.11 0.81 0.14 0.77 0.02 0.73 ÿ0.04 0.74 0.06 0.82

Vaginal dryness 0.19 0.45 0.12 0.55 0.10 0.51 0.18 0.45 0.16 0.36

aThese factor loadings were actually below 0.40, but appear as 0.40 when rounded for this table.

Table 4

Results of multiple linear regression for Factor I: psychosomatic symptoms

Predictor Parameter estimate�SE p-value

Age ÿ0.03� 0.003 50.0001

Education

Less than high school ÿ0.07� 0.06 0.2158

High school degree/equivalent ÿ0.19� 0.04 50.0001

Some college or vocational ÿ0.08� 0.04 0.0397

College graduate ÿ0.12� 0.04 0.0059

Postgraduate degree (ref)

Self-assessed health

Excellent (ref)

Very good 0.47� 0.04 50.0001

Good 0.80� 0.04 50.0001

Fair/poor 1.36� 0.04 50.0001

Difficult to pay for basics

Very hard 0.81� 0.04 50.0001

Somewhat hard 0.34� 0.03 50.0001

Not at all hard (ref)

Racial/ethnic group

African-American ÿ0.43� 0.03 50.0001

Hispanic ÿ0.24� 0.05 50.0001

Chinese ÿ0.78� 0.08 50.0001

Japanese ÿ0.73� 0.07 50.0001

Caucasian (ref)

Menopausal status

Premenopausal (ref)

Perimenopausal 0.37� 0.03 50.0001

Postmenopausal 0.08� 0.05 0.0863

Surgically menopausal 0.26� 0.04 50.0001

Hormone users 0.44� 0.05 50.0001

Geographic site

Boston ÿ0.29� 0.05 50.0001

Chicago 0.04� 0.05 0.4322

Detroit (ref)

Los Angeles ÿ0.13� 0.06 0.0237

Newark ÿ0.25� 0.05 50.0001

Oakland 0.08� 0.06 0.1942

Pittsburgh ÿ0.01� 0.04 0.8209

N.E. Avis et al. / Social Science and Medicine 52 (2001) 345–356352

strain. The p-values for all sets of categorical predictorswere highly significant (p50.00005). Psychosomatic symp-toms differ by both race/ethnicity and menopausal status.Compared to Caucasian women, all other racial/ethnic

groups report significantly fewer symptoms. Chinese andJapanese women differ most from the Caucasian women.Fig. 1 graphically shows the fully adjusted mean scores of

psychosomatic symptoms by race/ethnicity.Psychosomatic symptoms also differ by menopausal

status with perimenopausal women, hormone users, and

women who have had a surgical menopause reportingsignificantly more symptoms than premenopausal wo-men. Postmenopausal women do not differ significantlyfrom premenopausal women. Relevant to the third aim,

we did not find a significant interaction between statusand race/ethnicity, or between site and race/ethnicity.Looking at other factors, age was negatively related to

symptoms, although the magnitude of the associationwas small. Symptom reporting differed somewhat byeducational level, but there was no monotone trend, nor

were the coefficients large in magnitude; thus, educa-tional levels appeared to play a minimal role inexplaining psychosomatic symptoms. Women reporting

better health had lower symptom levels, as did womenwith less difficulty in paying for basics. Residual analysesindicated a satisfactory fit of the linear regression to thedata, despite the limited range of the outcome variables;

thus no transformation of the outcome was indicated.

Menopausal status, ethnicity and vasomotor symptoms

For ease of interpretation of predictors of the secondoutcome, displayed in Table 5, are the results of two

dichotomous (binomial) logistic regression models, onesymptom versus none and two symptoms versus none,controlling for age, education, health, and paying for

basics. Again, the p-values for all sets of categoricalpredictors were highly significant (p50.00005). Looking

at one symptom versus none, we see a significantdifference by race/ethnicity, with African-American

women significantly more likely to report a vasomotorsymptom than Caucasian women; this difference variessomewhat across the four sites studying African-Amer-

icans, with a significantly smaller difference at Bostonthan the other three sites. Japanese and Chinese womenwere significantly less likely to report a vasomotorsymptom than Caucasian women measured at the same

site. Hispanic women did not differ significantly fromCaucasian women at the same site. This pattern isidentical in analyses comparing two symptoms to no

symptoms, with sharper ethnic differences.Not surprisingly, there was also a significant difference

in vasomotor symptoms by menopausal status. Peri- and

postmenopausal women, HRT users, and women whohad a surgical menopause were all significantly morelikely to report vasomotor symptoms as compared to

premenopausal women. These differences were greaterfor two symptoms versus none than for one symptomversus none. The postmenopausal women had a slightlyhigher odds ratio than the other groups. Regarding the

third aim, the interaction between menopausal status andrace/ethnicity was not significant.Symptom reporting was negatively related to educa-

tional level and self-assessed health, and positivelyassociated with difficulty in paying for basics. As forpsychosomatic symptoms, the magnitude of the associa-

tion with educational level was relatively small, eventhough it was statistically significant. The Hosmer–Lemeshow goodness-of-fit chi-square statistic (Hosmer,Taber & Lemeshow, 1991) was used to access model fit.

This statistic can be problematic in the case of smallexpected cell counts (Pigeon & Heyse, 1999a,b); however,this not an issue here given the very large sample sizes.

The Hosmer–Lemeshow statistic indicated satisfactorymodel fits for both binomial logistic regression models.

Discussion

This paper sought to determine the evidence for amenopausal syndrome across five racial/ethnic groups.The results of the five factor analyses reported in this

study do not support a single syndrome consisting ofboth menopausal and psychological symptoms. Acrossall five racial/ethnic groups, two factors emerged; oneconsisting of clearly vasomotor symptoms } hot flashes

and night sweats } and the other consisting ofpsychological and psychosomatic symptoms. Theseresults, along with other factor analytic studies, clearly

argue against a ‘‘menopausal syndrome’’ consisting of avariety of vasomotor and psychological symptomstogether.

The regression analyses of race/ethnicity on psycho-somatic and vasomotor symptoms support previous

Fig. 1. Mean number of psychosomatic symptoms by racial/

ethnic group, adjusted.� All racial/ethnic groups differ sig-

nificantly from Caucasian women } referent group.

N.E. Avis et al. / Social Science and Medicine 52 (2001) 345–356 353

findings of racial/ethnic differences in symptom report-

ing. Consistent with other findings (Avis et al., 1993;Boulet et al., 1994; Lock, 1986; Tang, 1994), the twoAsian groups of women reported the fewest symptoms.

We do not know of any large studies conducted amongHispanic or African-American women. These analysescontrolled for differences across racial/ethnic groups

(in education, health, and economic strain) which havebeen shown to be related to symptoms. Addi-tional analyses from the SWAN cross-sectional studydata have shown independent effects of race/ethnicity

and lifestyle factors (e.g. physical activity and smoking)

on reporting of physical symptoms (Gold et al., 2000).

Future data from the SWAN cohort will examine ingreater depth influences such as lifestyle, reproductivehistory, psychosocial factors, and acculturation on

symptom reporting to determine whether they furtherexplain the racial/ethnic differences demonstrated in thecurrent study. It has been hypothesized that the high soy

diet of Japanese may account for their low rates of hotflashes (Aldercreutz et al., 1992).We also found that perimenopausal women reported

more psychosomatic symptoms than pre- or postmeno-

pausal women. Postmenopausal women, were more

Table 5

Results of two multiple logistic regressions for Factor 2: vasomotor symptoms

Predictor 1 symptom vs 0 OR (95% CI) 2 symptoms vs 0 OR (95% CI)

Age (1 year) 1.02 (1.00–1.03) 1.05 (1.04–1.06)

Education

Less than high school 1.29 (1.07–1.57) 1.70 (1.37–2.12)

High school degree/equivalent 1.25 (1.08–1.45) 1.69 (1.42–2.01)

Some college or vocational 1.22 (1.06–1.40) 1.53 (1.29–1.82)

College graduate 0.95 (0.81–1.12) 1.03 (0.84–1.26)

Postgraduate degree (ref) (ref)

Self-assessed health

Excellent (ref) (ref)

Very good 1.38 (1.21–1.58) 1.69 (1.43–2.00)

Good 1.79 (1.56–2.05) 2.22 (1.88–2.62)

Fair/poor 2.58 (2.21–3.02) 3.72 (3.10–4.46)

Difficult to pay for basics

Very hard 1.30 (1.12–1.50) 1.55 (1.33–1.81)

Somewhat hard 1.15 (1.04–1.27) 1.18 (1.06–1.32)

Not at all hard (ref) (ref)

Racial/ethnic group

African-American: Boston 1.00 (0.79–1.28) 1.28 (1.08–1.58)

African-American: Chicago 1.41 (1.06–1.88) 1.95 (1.39–2.73)

African-American: Detroit 1.59 (1.29–1.95) 1.90 (1.54–2.35)

African-American: Pittsburgh 1.21 (0.97–1.51) 1.82 (1.43–2.35)

Hispanic 0.92 (0.76–1.12) 0.77 (0.62–0.96)

Chinese 0.60 (0.45–0.80) 0.29 (0.19–0.45)

Japanese 0.65 (0.50–0.84) 0.25 (0.17–0.38)

Caucasian (ref) (ref)

Menopausal status

Premenopausal (ref) (ref)

Perimenopausal 1.96 (1.74–2.20) 3.43 (2.93–4.02)

Postmenopausal 1.90 (1.62–2.24) 3.81 (3.14–4.62)

Surgically menopausal 1.78 (1.55–2.04) 3.35 (2.82–3.97)

Hormone users 1.69 (1.40–2.03) 3.68 (2.96–4.58)

Geographic sitea

Boston 1.00 (0.78–1.27) 0.78 (0.59–1.02)

Chicago 1.08 (0.89–1.57) 0.84 (0.60–1.17)

Detroit (ref) (ref)

Los Angeles 0.96 (0.76–1.20) 0.75 (0.59–0.97)

Oakland 1.26 (0.99–1.60) 0.97 (0.74–1.27)

Newark 0.88 (0.70–1.10) 0.79 (0.63–1.00)

Pittsburgh 1.04 (0.84–1.28) 0.74 (0.59–0.93)

aOdds ratio computes at ethnicity=Caucasian.

N.E. Avis et al. / Social Science and Medicine 52 (2001) 345–356354

likely to report one or more vasomotor symptom. Thisfinding further argues against a menopausal syndrome

consisting of a variety of symptoms.Notably, women who were hormone users in the

previous 3 months were as likely to report vasomotor

symptoms as peri- or postmenopausal women. This issomewhat surprising given HRT’s well-established effecton relief from vasomotor symptoms. However, not all ofthese women were current HRT users, since this

category included women who had stopped using HRTwithin the last 3 months and were using oral contra-ceptives. There are also clearly selection factors related

to taking HRT, with vasomotor symptoms being amajor factor. Therefore, women who had used hor-mones may have done so because they experienced more

vasomotor symptoms. Further, this could be due in partto the limitations of the symptom questions, which onlyasked whether or not a symptom had been experienced

in the past two weeks. This question does not ascertainfrequency or intensity or symptoms.There are several limitations of this study. As

mentioned, the symptom data are dichotomous variables

based on whether or not a symptom was experienced atall in the past 2 weeks. Baseline and annual question-naires from the SWAN cohort will collect more detailed

symptom data allowing for future confirmatory analysesof the factor structures reported in this paper. Anotherlimitation is the brevity of the symptom list. A longer

symptom list may have yielded more factors andprovided differences across ethnicity. The Avis et al.(1993) study which found different factor structuresamong women from three countries used a somewhat

longer list of symptoms. Differences between African-American and Caucasian women in vasomotor symptomreporting varied somewhat across site. For the three

groups studied at a single site (Chinese, Hispanic, andJapanese), it was not possible to estimate interactionsbetween race/ethnicity and site. Note, however, that

geographic locations for sampling were selected in largepart in order to be able to obtain sufficient numbers ofracial/ethnic minority subjects, rather than to provide a

basis for geographic comparison across field sites or toyield nationally representative } or even locally repre-sentative } data. Thus, estimated model coefficients forfield sites should not be interpreted as indicative of

regional differences in symptom reporting. A keyadvantage of the sampling design, however, was thatcomparisons of minorities were made with Caucasians at

the same site with similar geographic influences, therebyreducing confounding race/ethnicity and region. More-over, to a large extent, associations between symptoms

and predictors were of similar magnitude and in the samedirection across the seven sites (data not shown).Despite these limitations, the strengths of this study lie

in the racial/ethnic comparisons provided in a singlestudy with consistent questions and methodology across

all groups, the large size and the community-basedsample. Although the study sample all came from a single

country, their backgrounds and cultures varied widely.In summary, the data presented here go far toward

filling in gaps in knowledge about the nature of symptoms

experienced by a wide range of ethnically, racially, andculturally diverse women in the United States. The lack ofa single set of menopausal symptoms and the findings thatthe type and number of symptoms vary with race/ethnicity

suggest the need to continue to explore the complexrelationship between the physiological changes occurringduring menopause and the symptoms experienced by

women. Data from the prospectively studied SWANcohort will provide unique insight into the mechanismsthat might explain these racial/ethnic differences in

symptom reporting, as well as the increased symptomreporting among the perimenopausal women.

Acknowledgements

The Study of Women’s Health Across the Nation(SWAN) was funded by grants from the NationalInstitute on Aging and the National Institute of NursingResearch of the National Institutes of Health to the

following participating centers: University of Michigan,Ann Arbor (U01 NR04061; U01 AG12495); Massachu-setts General Hospital, Boston, MA (U01 AG12531);

Rush University, Rush–Presbyterian–St. Luke’s MedicalCenter, Chicago, IL (U01 AG12505); University ofCalifornia, Davis (U01 AG12554); University of Cali-

fornia, Los Angeles (U01 A12539); University ofMedicine and Dentistry of New Jersey–New JerseyMedical School, Newark (U01 AG12535); University

of Pittsburgh, Pittsburgh, PA (U01 AG12546); NewEngland Research Institutes, Watertown, MA (U01AG12553). The manuscript was reviewed by thePublications and Presentations Committee of SWAN

and has its endorsement. We thank the study staff at eachsite and all of the women who participated in the SWAN.

References

Agresti, A. (1984). Analysis of ordinal categorical data. New

York: Wiley.

Aldercruetz, H., Hamalaiven, O., & Gorbach, J., et al. (1992).

Dietary phytoestrogens and the menopause in Japan.

Lancet, 339, 1233.

Avis, N. E., Kaufert, P. A., Lock, M., McKinlay, S. M., Vass,

K. (1993). The evolution of menopausal symptoms.

Bailliere’s Clinical Endocrionology and Metabolism, 7, 17–31

Avis, N. E., Brambilla, D., McKinlay, S. M., & Vass, K. (1994).

A longitudinal analysis of the association between meno-

pause and depression: results from the Massachusetts

Women’s Health Study. Annals of Epidemiology, 4, 214–220.

Ballinger, C. B. (1975). Psychiatric morbidity and the meno-

pause: screening of general population sample. British

Medical Journal, 3, 344–346.

N.E. Avis et al. / Social Science and Medicine 52 (2001) 345–356 355

Ballinger, C. B. (1976). Psychiatric morbidity and the

menopause: clinical features. British Medical Journal, 1,

1183–1185.

Boulet, M. J., Oddens, B. J., Lehert, P., Vemer, H. M., &

Visser, A. (1994). Climacteric and menopause in seven

south-east Asian countries. Maturitas, 19, 157–176.

Cawood, E. H. H., & Bancroft, J. (1996). Steroid hormones, the

menopause, sexuality and well-being of women. Psycholo-

gical Medicine, 26, 925–936.

Collins, A., & Landgren, B. -M. (1995). Reproductive health,

use of estrogen and experience of symptoms in perimeno-

pausal women: a population-based study. Maturitas, 20,

101–111.

Davis, D. L. (1982). Women’s status and experience of the

menopause in a Newfoundland fishing village. Maturitas, 4,

207–216.

Dennerstein, L., Smith, A. M. A., Morse, C., Burger, H.,

Green, A., Hopper, J., & Ryan, M. (1993). Menopausal

symptoms in Australian women. The Medical Journal of

Australia, 159, 232–236.

Dennerstein, L., Smith, A. M. A., & Morse, C. (1994).

Psychological well-being, mid-life and the menopause.

Maturitas, 20, 1–11.

Donovan, J. C. (1951). The menopausal syndrome: a study of

case histories. Am. J. Obstet. Gynecol., 62, 1281–1291.

Flint, M. (1975). The menopause: reward or punishment?.

Psychosomatics, 16, 161–163.

Flint, M. P., & Garcia, M. (1979). Culture and the climacteric.

Journal of Biosocial Sciences, 6(Suppl.), 197–215.

Floyd, F. J., & Wadaman, K. F. (1995). Factor analyis in the

development and refinement of clinical assessment instru-

ments. Psychol. Assess., 7, 286–299.

Gold, E. B., Sternfeld, B., Kelsey, J. L., Brown, C., Mouton, C.,

Reame, N., Salamone, L., Stellato, R. (2000). The relation

of demographic and lifestyle factors to symptoms in a multi-

racial/ethnic population of women 40–55 years of age. Am.

J. Epidemiol., [in press]

Greene, J. G. (1976). A factor analytic study of climacteric

symptoms. Journal of Psychosomatic Research, 20, 425–430.

Holte, A., & Mikkelsen, A. (1991). The menopausal syndrome:

a factor analytic replication. Maturitas, 13, 193–203.

Hosmer, D. W., & Lemeshow, S. (1989). Applied logistic

regression. New York: Wiley.

Hosmer, D. W., Taber, S., & Lemeshow, S. (1991). The

importance of assessing the fit of logistic regression models:

A case study. Am. J. Public Health, 81, 1630–1635.

Hunter, M., Battersby, R., & Whitehead, M. (1986). Relation-

ships between psychological symptoms, somatic complaints

and menopausal status. Maturitas, 8, 217–228.

Indiria, S. N., & Murthy, V. N. (1980). Nature of a psychiatric

disturbances in menopausal women. Indian Journal of

Clinical Psychology, 7, 7–11.

Johnson, R. A., & Wichern, D. W. (1982). Applied multivariate

statistical analysis. Englewood Cliffs, NJ: Prentice-Hall.

Kaufert, P. A., Gilbert, P., & Hassard, T. (1988). Researching

the symptoms of menopause: an exercise in methodology.

Maturitas, 10, 117–131.

Kim, J., & Mueller, C. W. (1978). Factor analysis } statistical

methods and practical issues. Beverly Hills, CA: Sage.

Kuh, D. L., Wadsworth, M., & Hardy, R. (1997). Women’s

health in midlife: the influence of the menopause, social

factors and health in earlier life. British Journal of Obstetrics

and Gynaecology, 104, 923–933.

Lock, M. (1986). Ambiguities of aging: Japanese experience and

perceptions of menopause. Culture, Medicine & Psychiatry,

10, 23–46.

Matthews, K. A., Wing, R. R., Kuller, L. H., Meilahn, E. N.,

& Plantinga, P. (1994). Influence of the peri-menopause

on cardiovascular risk factors and symptoms of middle-

aged healthy women. Archives of Internal Medicine, 154,

2349–2355.

McKinlay, S. M., & Jefferys, M. (1974). The menopausal

syndrome. British Journal of Medicine, 28, 108–115.

Mitchell, E. S., & Woods, N. F. (1996). Symptom experiences

of midlife women: observations from the Seattle Midlife

Women’s Health Study. Maturitas, 25, 1–10.

Muthen, B. O. (1986). Dichotomous factor analysis of symptom

data. Soc. Methods Res., 18, 19–65.

Neugarten, B. L., & Kraines, R. J. (1965). Menopausal

symptoms in women of various ages. Psychosomatic

Medicine, 27, 266–273.

Nisenbaum, R., Reyes, M., Mawle, A. C., & Reeves, W. C.

(1998). Factor analysis of unexplained severe fatigue and

interrelated symptoms: overlap with criteria for chronic

fatigue syndrome. Am. J. Epidemiol., 148, 72–77.

Pham, K. -T. C., Grisso, J. A., & Freeman, E. W. (1997).

Ovarian aging and hormone replacement therapy: hormonal

levels, symptoms, and attitudes of African-American and

white women. Journal of General Internal Medicine, 12,

230–236.

Pigeon, J. G., & Heyse, J. F. (1999a). A cautionary note about

assessing the fit of logistic regression models. J. Appl. Stat.,

26, 847–853.

Pigeon, J. G., & Heyse, J. F. (1999b). An improved goodness of

fit statistic for probability prediction models. Biometrical J.,

1, 71–82.

Porter, M., Penney, G. C., Russell, D., Russell, E., &

Templeton, A. (1996). A population based survey of

women’s experience of the menopause. British Journal of

Obstetrics and Gynaecology, 103, 1025–1028.

Punyahotra, S., Dennerstein, L., & Lehert, P. (1997). Meno-

pausal experiences of Thai women. Part 1: symptoms and

their correlates. Maturitas, 26, 1–7.

Schoenberg, R., & Arminger, G. (1986). Latent variable models

in dichotomous data: the state of the method. Soc. Methods

Res., 18, 164–182.

Sowers, M., Crawford, S., & Sternfeld, B., et al. (2000). Design,

survey sampling and recruitment methods of SWAN: A

multi-center, multi-ethnic, community-based cohort study

of women and the menopausal transition. In R. A. Lobo, J.

Kelsay, & R. Marcus (Eds.), Menopause: Biology and

pathobiology. San Diego: Academic Press.

Tang, G. W. K. (1994). The climacteric of Chinese factory

workers. Maturitas, 19, 177–182.

Weisberg, S. (1980). Applied linear regression. New York: Wiley.

WHO Scientific Group (1996). Research on the menopause in

the 1990s. In WHO Technical Report Series, 866. Geneva:

World Health Organization.

N.E. Avis et al. / Social Science and Medicine 52 (2001) 345–356356