Persistent Mood Symptoms in a Multiethnic Community Cohort of Pre- and Perimenopausal Women

Abstract

To further our understanding of the relation between mood and menopause, the authors examined 1) the association between persistent mood symptoms and menopausal status and 2) factors that increase a woman’s vulnerability to an overall dysphoric mood during the early perimenopausal period. The sample consisted of an ethnically diverse community cohort of 3,302 pre- and early perimenopausal women aged 42–52 years who were participants in the Study of Women’s Health Across the Nation, an ongoing US multisite longitudinal study of menopause and aging. At study entry (1995–1997), women reported information on recent menstrual regularity and premenstrual symptoms, as well as on sociodemographic, symptom, health, sleep, psychosocial, and lifestyle variables. Rates of persistent mood symptoms were higher among early perimenopausal women (14.9%–18.4%) than among premenopausal women (8%–12%). In analyses adjusting for major covariates and confounders, early perimenopausal women had higher odds of irritability, nervousness, and frequent mood changes but not of feeling “blue.” The effect of being early perimenopausal on overall dysphoric mood was greatest among women with an educational level of less than high school graduation. These findings suggest that persistent mood symptoms and overall dysphoric mood are associated with the early perimenopause, particularly among women with lower educational attainment.

Received for publication July 24, 2002; accepted for publication February 6, 2003.

Numerous studies have examined the association between menopausal status and mood, yet these findings are inconsistent. Many community-based studies have not found an association between menopausal status and depressive mood (1–6), individual symptoms such as irritability, anxiety, and depression (7–12), or groups of mood symptoms (4, 13–16). On the other hand, some studies have found an association between menopausal status and mood, primarily in the perimenopausal (transitional) period (17–21). These data, along with clinical and anecdotal reports of increased negative mood during perimenopause, suggest the need for further clarification of the relation between menopause and mood.

Given the lack of consistent findings on the association between menopause and mood and the fact that when such an association is found, it tends to be observed during perimenopause, it is reasonable to postulate that some phase of the menopausal transition is a time of particular susceptibility to negative mood for a subset of women. Such women may have poor health (2, 5, 22), lower educational attainment (2), or prior depression (18, 20) or may have experienced very stressful events and/or circumstances (2, 23–25). All of these factors have been associated with dysphoric mood in women during midlife, just as they are at other times.

Controversy continues over whether psychological symptoms are related to menopausal status. Almost all of the research has been conducted among White women. The multiethnic Study of Women Across the Nation (SWAN) provides an opportunity to examine the association between menopausal status and mood symptoms among US White, African-American, Chinese, Hispanic, and Japanese women.

Previous data from the SWAN screening survey showed that the presence of psychological distress (irritability, depression, and tension) in the previous 2 weeks was greater among early perimenopausal women than among pre- or postmenopausal women (19). Odds of psychological distress, adjusted for covariates and confounders, were lower among all minority ethnic groups than among Whites. However, these data examined only the presence or absence of three symptoms and included only a limited set of predictors. In the present analysis, we extend these earlier findings by determining whether they are replicated with the inclusion of one additional mood symptom—frequent mood changes; whether duration of mood symptoms is also associated with menopausal status; and whether there are subsets of women who are particularly vulnerable to an overall dysphoric mood during the early stages of the transition, that is, early perimenopause. Duration adds another dimension to presence versus absence, representing the extent to which such symptoms are persistent and thus potentially more impairing and distressing.

Specifically, we hypothesized that 1) early perimenopausal women are more likely than premenopausal women to experience persistent mood symptoms, independent of ethnicity and potential confounders and covariates, and 2) the effect of being early perimenopausal on dysphoric mood will be greater among women with certain characteristics, specifically women with low educational attainment, high financial strain, poor perceived health, low social support, or high stress.

MATERIALS AND METHODS

Study sample

SWAN is an ongoing community-based longitudinal study that recruited 3,302 women between November 1995 and October 1997 to track changes in women’s physical and mental health as they age and experience menopause. During 1995–1997, at least 450 women at each of seven sites were enrolled in the longitudinal study. Each sample included White women and women of one predetermined minority group (African-American women in Pittsburgh, Pennsylvania; Boston, Massachusetts; Detroit, Michigan; and Chicago, Illinois; Japanese women in Los Angeles, California; Chinese women in the San Francisco Bay Area, California; and Hispanic women in Newark, New Jersey). The investigators used a variety of recruitment strategies (e.g., census data, lists, random-digit dialing, and/or “snowballing” from current participants) that varied depending on the specific situation at each site and the designated ethnic group to be enrolled. Thus, multiple sampling frames and approaches were used to recruit a community-based sample of local women. (For detailed information, see the chapter by Sowers et al. in Menopause: Biology and Pathobiology (26)). Eligibility requirements (in addition to community residence) were: age 42–52 years; either White or site-predetermined other ethnicity; having had menses within the previous 3 months; having a uterus and at least one ovary; not being pregnant; and not having used reproductive hormones or birth control pills within the previous 3 months.

Of the women who completed the screening interview (usually by telephone), approximately 40 percent were eligible for the longitudinal study. Of the eligible women, 3,302 (50.7 percent) were recruited and completed their in-person baseline interview. In this paper, we report on data collected from women who completed the screening and baseline interviews.

Overview of procedures

Investigators at each site adhered to the relevant institutional review board’s guidelines for human research. At the beginning of the screening survey, verbal consent was obtained by trained interviewers, and for telephone interviews, computer-assisted telephone interviewing was used at all sites.

Eligible women were seen within 3 months of the screening survey for their initial assessment, and written informed consent was obtained. Assessments took approximately 3–4 hours to complete and consisted of detailed questions about medical, reproductive, and menstrual history, lifestyle, psychosocial factors, depressive symptoms, and physical and psychological symptoms. Questions either were administered orally by trained staff or were part of a paper-and-pencil form. Additional physiologic and anthropometric data were collected but are not relevant to the current report. All study forms and materials were available in English, Cantonese, Japanese, and Spanish, and bilingual staff were used as appropriate.

Measures

Dysphoric mood symptoms

Women were asked how frequently they had experienced each of 15 somatic, vasomotor, and mood symptoms during the previous 2 weeks: not at all (coded 0), on 1–5 days (coded 1), on 6–8 days (coded 2), on 9–13 days (coded 3), and every day (coded 4). The symptoms were obtained from symptom checklists used in many studies of midlife women (6, 27, 28). In this paper, we report on four mood symptoms: feeling “blue” or depressed, irritability or grouchiness, feeling tense or nervous, and frequent mood changes. To test the first hypothesis, we considered each symptom separately. On the basis of criteria for clinical disorders (e.g., duration of depressive symptoms for most of a 2-week period), a persistent symptom was defined as one that occurred on 6 or more days, and this cutpoint was used as the threshold for a potentially more severe symptom. We expected that the duration of symptoms would be a potential indicator of more severe and impairing symptoms, and we validated this with the Medical Outcomes Study Short Form 36 (SF-36) (29) scale on low role functioning due to emotional problems (the Role-Emotional (RE) scale). There were significant relations between the number of days for which a symptom was reported and a woman’s reporting that she accomplished less because of emotional problems (χ²’s > 370, p’s < 0.001). For example, of the women who reported feeling tense or nervous for 5 or fewer days during the previous 2 weeks, 23.5 percent reported accomplishing less, as compared with 58.3 percent of those who reported that the symptom occurred on 6–8 days, 73.3 percent of those who reported that the symptom occurred on 9–13 days, and 66.7 percent of those who reported having the symptom daily.

To test the second hypothesis, we summed the scores of the four symptoms to form a scale of overall persistent dysphoric mood that ranged from 0 to 16. The scale was used to create a dichotomized measure of high and low dysphoric mood, as described below.

Independent variables

Sociodemographic variables

The variables of interest were age, marital status, educational attainment, and ability to pay for basic necessities. Ethnicity was self-defined as African-American, Chinese, Japanese, Hispanic, or White. Prospective participants were asked, “What is your primary racial or ethnic group?” If they indicated mixed ethnicity, they were asked whether they were comfortable choosing one racial or ethnic category. By design, only women who self-identified with a particular category were included in the cohort. The racial/ethnic categories were defined as follows: Black or African-American (African origin or descent), Chinese or Chinese-American, Japanese or Japanese-American, Hispanic, and Caucasian/non-Hispanic White (European descent). We initially had separate categories for Hispanic groups (Puerto Rican, Mexican or Mexican-American, Dominican, Central American, Cuban or Cuban-American, South American, Spanish, and other Hispanic), but because of the small numbers in many groups, we combined these women into one group.

Menopausal status

Menopausal status was categorized according to bleeding patterns: premenopausal (no decrease in the predictability of menses onset in the prior 12 months) or early perimenopausal (less predictable onset of menses in the last 12 months). These are similar to the recommendations from the Stages of Reproductive Aging Workshop (30) for distinguishing the late reproductive years (premenopause) from the early phase of the menopausal transition (early perimenopause). All women had had at least one menstrual period in the previous 3 months.

Health variables

Health variables included items and subscales from the SF-36 (29), a standard, frequently used measure of health-related quality of life. Information was gathered on 1) self-perceived health (excellent, very good, good, fair, or poor); 2) limitations in activities due to any impairments or health problems (yes/no); and 3) the amount of body pain experienced in the previous 4 weeks according to a six-point scale, ranging from none to very severe (29). Other health variables assessed included 1) self-reported past or current arthritis, high blood pressure, fibroids, migraines, anemia, or an underactive thyroid gland, as identified by a health care provider; 2) use of medications for nervous conditions in the past month (verified by examining the bottle label); and 3) frequency of hot flashes or night sweats in the previous 2 weeks. The latter variables were part of the symptom checklist described above.

Possible premenstrual syndrome

Women were asked whether, during the past year, they had had any of the following during at least half of their menstrual periods or in the week before them—abdominal cramps, breast pain, weight gain/bloating, increased appetite or food cravings, mood changes, anxiety, back, joint, or muscle pain, or headaches—or had experienced interference with a job or home activities and whether these symptoms usually (more than half the time) disappeared within 1–3 days after their period started. Because our data were not sufficient to evaluate premenstrual dysphoric disorder (31), we defined “possible premenstrual syndrome” as the presence of at least one mood and one physical symptom that interfered with activities and usually disappeared after menses started.

Sleep

Participants were asked how often in the past 2 weeks (ranging from none to five or more times per week) they had had trouble falling asleep, had awakened several times during the night, or had woken up earlier than planned and been unable to fall asleep again. These were items from the Women’s Health Initiative (32). We defined a sleep disturbance as the reported occurrence of at least one of these three symptoms at least three times per week.

Psychosocial variables

Psychosocial variables evaluated included level of social support, the quality of the participant’s marital/partner relationship, and the presence of stressful events. Level of social support was assessed as high or low on the basis of summed ratings for each of four items from the Medical Outcomes Study Social Support Survey (33), which asked how often each of four kinds of support was available if needed. The quality of the woman’s marital/partner relationship was assessed with an item asking about the degree of happiness in the relationship (34), ranging from extremely unhappy to perfect, or no partner. A previous study of middle-aged women found that the quality-of-relationship item from the Spanier dyadic adjustment scale was highly correlated with the overall scale score (K. A. Matthews, University of Pittsburgh, personal communication, 1995). The presence of stressful events was assessed by asking about the occurrence of 34 stressful life events (e.g., job-related problems, money problems, divorce, violence, having to move, legal problems, the death of a loved one) in the past 12 months; this checklist was a modification of the Psychiatric Epidemiology Research Interview (35, 36). We summed the number of such events to create a variable with three levels of stressful events: none, one, and two or more.

Lifestyle

Cigarette smoking was categorized in terms of current smoking (yes/no). Recreational/sports activity was assessed on the basis of the frequency, intensity, and duration of the two sports or exercise activities the participant had most frequently engaged in during the previous year (37, 38). This scale was an adaptation of the Baecke questionnaire, which has been used extensively in epidemiologic research and has demonstrated reliability and validity (37). Scores ranged from 1 to 5, with higher scores indicating higher activity.

Data analysis

Logistic regression models were fitted separately for each of the four mood symptoms, with outcomes categorized as having experienced the mood on at least 6 days in the previous 2 weeks versus 0–5 days. Unadjusted odds ratios for early perimenopausal women versus premenopausal women were fitted on the entire analytical data set. Model-building to obtain the adjusted odds ratios comprised several steps. First, stepwise logistic regression was used with a p value of 0.15 among women with no missing data on any of the candidate independent variables, to identify those variables with at least moderately strong relations to the outcome, independent of other variables in the model. Menopausal status was forced into all adjusted models as the variable of interest. Site and ethnic group were forced into all adjusted models, because the sampling plan for the study was based on these variables. Age was forced into all adjusted models to separate the effects of chronologic aging and reproductive aging. Second, the model identified by the stepwise process was then fitted to the larger data set of women who had complete data on the independent variables included in the model but possibly missing data on variables not selected for the model. If all variables (except site, ethnic group, menopausal status, and age) were significant at the 0.05 level, the model-building process stopped. Otherwise, the least significant independent variable was identified and dropped from the model. This iteration process was repeated until all variables were significant at the 0.05 level (except those forced into the model).

The four individual five-level mood frequency scales were highly correlated (α = 0.86) and formed a single factor. Therefore, it was reasonable and more parsimonious to evaluate the vulnerability hypothesis with an overall indicator of persistent dysphoric mood. We added the original scores to form a scale with a range of 0 (none of these moods had occurred during the last 2 weeks) to 16 (all four moods occurred every day) and dichotomized the overall mood score as 0–6 versus 7–16, which constitutes approximately the top 10 percent of the group and indicates persistent dysphoric mood. This outcome was significantly associated with the SF-36 subscale on low role functioning due to emotional problems (the RE scale) (29), which is consistent with the negative impact of overall dysphoric mood on functioning. For example, 43 percent of the women with a high dysphoric mood score had the lowest RE functioning score (0) and 26 percent had the highest score (100). In contrast, among those women with a low dysphoric mood score, 9 percent had the lowest RE functioning score, whereas 71 percent had the highest score. We fitted a logistic regression model with high dysphoric mood score as the dependent variable and site, ethnicity, age, menopausal status, self-rated health, and all covariates selected for at least one of the four final models for individual moods as independent variables. We then fitted five separate logistic regression models including these main effects plus the interaction between menopausal status and one of the covariates that was hypothesized to modify the effect of menopausal status. Stepwise logistic regression, forcing in all main effects and selecting among the significant interactions, was used to determine the final logistic regression model for the overall mood score if more than one significant interaction was identified.

The Hosmer-Lemeshow test and logistic regression diagnostic techniques were used to assess the overall fit of each model and to ensure that unusual data points did not overly influence results.

RESULTS

Description of the sample

Women who enrolled in the longitudinal study had more education and higher incomes and were less likely to be smokers than women who did not enroll. Response rates did not vary by ethnicity, age, marital status, parity, number of close friends, quality of life, perceived stress, or menopausal status.

Of the 3,302 women in the cohort, 141 did not have complete data on dysphoric mood and menopausal status. Therefore, the analytical sample consisted of 3,161 women. Forty-seven percent of the women were in the early perimenopausal period at the time they entered the SWAN study. The women represented a diverse group with respect to demographic, health, psychosocial, and lifestyle factors (see table 1).

Association between individual mood symptoms and menopausal status

The first hypothesis addressed the association between individual symptoms and menopausal status. Table 2 shows the numbers and percentages of premenopausal and early perimenopausal women and women of each ethnic group who had experienced each mood on at least 6 days during the preceding 2 weeks. Tables 3 and 4 show the unadjusted and adjusted odds ratios and associated 95 percent confidence intervals for the comparison of early perimenopausal women with premenopausal women and the comparison of ethnic minority women with White women. In unadjusted analyses, early perimenopausal status was significantly related to each persistent mood symptom, with odds ratios ranging between 1.45 and 1.83 and p values less than 0.001 for each outcome (table 3). In adjusted analyses, these odds ratios decreased in magnitude but remained statistically significant, except for the “feeling blue” outcome (table 4). The adjusted odds ratios were highest for feeling nervous (odds ratio = 1.54) and for frequent mood changes (odds ratio = 1.52). The differences between the unadjusted and adjusted odds ratios were due primarily to the adjustment for other key variables and were not an effect of deleting records that had incomplete data on variables in the final models.

In the multivariable models, moderate or severe pain, poor sleep, one or two vasomotor symptoms, taking medication for nervous conditions, low social support, two or more very upsetting life events, and an unhappy relationship were significantly positively associated with each of the four moods. Increased age was significantly associated with a lower risk of mood symptoms, except for feeling blue. Ethnic groups had significantly different odds of feeling blue, feeling irritable, and being nervous. There were no significant differences between ethnic groups in the odds of reporting persistence of frequent mood changes.

Other covariates were related to some of the mood symptoms but not all. Compared with women who did not find it hard at all to pay for basic necessities, those who found it very hard or somewhat hard were more likely to feel blue, feel nervous, or experience mood changes. Compared with women without it, those with “possible premenstrual syndrome” were more likely to feel blue and to feel irritable. Higher recreational/sports activity scores were associated with lower odds of feeling blue, feeling irritable, or being nervous. Women with less than a college education were less likely to feel nervous than were those with at least a college education. Marital status, perceived health, a set of specific health conditions, health limitations, and current smoking were not significantly related to any of the four mood outcomes after adjustment for other key predictors.

Association between persistent dysphoric mood and menopausal status in subgroups

The second hypothesis addressed the issue of whether certain subgroups of women were more vulnerable to the effects of menopausal status on mood. Table 5 shows the numbers and percentages of premenopausal and early perimenopausal women who had a dysphoric mood score of 7 or greater and the unadjusted and adjusted odds ratios for early perimenopausal women compared with premenopausal women. In the unadjusted analysis, early perimenopausal status was significantly related to a dysphoric mood score of 7 or greater. In adjusted analyses, the odds ratio decreased but remained statistically significant at p < 0.001.

We hypothesized that five subgroups of women would be particularly vulnerable to the effect of perimenopausal status on mood: women with 1) lower educational attainment, 2) more difficulty paying for basic necessities, 3) a lower level of self-perceived health, 4) a greater number of upsetting life events, or 5) low social support. In separate models including all main effects plus the interaction of menopausal status with one of these variables, only the interaction of status with educational level (p = 0.037) was significant. Status did not interact with the other four variables (all p’s > 0.29). Table 6 shows odds ratios for the comparison between early perimenopausal women and premenopausal women in this interaction model. After adjustment for the other covariates, perimenopausal status was significantly related to a high dysphoric mood score among women with less than a college education and women with some college.

DISCUSSION

This cross-sectional study of mood and menopause found that being in the early perimenopausal period as compared with being premenopausal was associated with higher odds of experiencing irritability, nervousness, and frequent mood changes (but not feeling blue/depressed) on at least 6 days during the previous 2 weeks after adjustment for major covariates and confounders as well as for site, age, and ethnicity. Our results are consistent with those from several studies reporting that prevalences of mood symptoms, including irritability (20, 39) and psychological symptoms generally (40),are higher in perimenopausal women. Additionally, the current data support our earlier findings from the SWAN survey data in which early perimenopausal women had a significantly greater probability (odds ratio = 1.20) of experiencing dysphoric mood (irritability, nervousness, and feeling blue/depressed) in the 2 weeks prior to the interview than did premenopausal women (19). As we noted above, although the results were adjusted for a number of covariates, those data were based on only a yes/no response to a query about the presence of each mood symptom.

In the current analyses we have shown that, with the exception of feeling blue/depressed, early perimenopausal women as compared with premenopausal women are more likely to experience persistent mood symptoms (adjusted odds ratios = 1.33–1.54). Moreover, the higher odds remained after adjustment for confounding factors such as vasomotor symptoms and sleep disturbances. The latter two factors have often been thought to mediate the relation between menopausal status and mood. These data are important, because persistent symptoms are indicative of potentially more serious mood disturbance and are more likely to interfere with functioning. Few studies of menopause have assessed either duration or severity of individual mood symptoms (4, 20, 41), and neither Dennerstein et al. (41) nor Porter et al. (4) found an effect for menopausal status.

The fact that these mood symptoms show significant adjusted associations with menopausal status suggests the importance of focusing on the early stage of the menopausal transition and distinguishing between the early and later stages. Previous studies that have not distinguished the early stages of the transition from the later ones may have obscured the effect of the early transition, at least in part. For example, prior studies (1, 2, 4, 20) have compared pre-, peri-, and postmenopause with perimenopause defined as no menstrual bleeding in the previous 3 months, or they have combined perimenopause and postmenopause (41).

Our results for feeling “blue” are consistent with those of other studies (4, 5, 18, 42). They are particularly important because reviews of the literature have concluded that depression may have received too much focus and that irritability, nervousness, and mood lability have not been sufficiently studied (43).

Consistent with other studies, we found that African-American, Chinese, and Japanese middle-aged women, in comparison with White women, had lower odds for several of the individual mood symptoms (17, 44). Whether and to what extent these lower rates reflect cultural differences in symptom reporting or in actual mood symptoms remains unclear (45). For middle-aged African-American women, two small studies have reported opposite results. Pham et al. (46) found that scores on the Center for Epidemiologic Studies Depression Scale, percentages with scores of 16 or greater, and daily psychological symptom ratings were similar for African-American and White women, whereas Bromberger (unpublished data) found higher levels of depressive symptoms and tension in African Americans than in Whites. The similarity of odds ratios for frequent mood changes among all ethnic groups suggests that changeable or labile mood may be a universal experience in midlife.

A number of researchers have suggested that subgroups of women may be particularly vulnerable to the mood effects of the menopausal transition (43, 47–49). Our results only partially supported this hypothesis. Among women who attained less than a college level of education, early perimenopausal women had significantly higher odds of a high dysphoric mood score than did premenopausal women. Lower education may be a marker of long-term socioeconomic or social stress, because it has been associated with a variety of negative health outcomes (50). Alternatively, perhaps higher education buffers the impact of the transition, since it has been associated with a more positive outlook as well as with higher levels of support, fewer symptoms, and the expectation that the menopause will be a positive experience (51).

This analysis had several limitations. The fact that only 50 percent of eligible women chose to enter the longitudinal study could have biased our results. For example, given the greater proportion of less well educated women among the enrolled women and the fact that less well educated women had more symptoms, it is possible that the association between menopausal stage and dysphoric mood symptoms was underestimated for women with a low level of education. A higher response rate among the less well educated might have produced an even stronger effect of education. As we discussed above, refusers were similar to the enrollees with regard to many important factors, including menopausal status, age, perceived stress, and quality of life, which could have influenced our results. Another potential limitation is the fact that the overall dysphoric mood measure was developed specifically for the study and thus was not a standard measure. However, as we noted above, this outcome was significantly associated with the SF-36 subscale on low role functioning due to emotional problems (the RE scale) (29) and is consistent with the negative impact of overall dysphoric mood on functioning.

We also were limited by the absence of prospective menstrual diary data for determination of premenopause versus perimenopause, and we were limited to studying pre- and early perimenopausal women by the requirements for study entry. Moreover, the cross-sectional nature of the data precludes interpretation of causality in the association between menopausal status and mood. Longitudinal data will become available for future SWAN research as the women transition to postmenopause.

Despite its limitations, the current analysis had many strengths, including a focus on the early stage of the menopausal transition, a diverse community sample, a measure of symptom duration, the inclusion of a wide range of covariates and confounders, and the ability to examine subgroups of potentially vulnerable women. The identification of the initial phase of the transition (i.e., a change in the predictability of menstrual cycles) as one of susceptibility to mood symptoms may ultimately help us clarify the etiology of these symptoms.

ACKNOWLEDGMENTS

The Study of Women’s Health Across the Nation (SWAN) was funded by the National Institute on Aging, the National Institute of Nursing Research, and the Office of Research on Women’s Health of the National Institutes of Health. Supplemental funding was obtained from the National Institute of Mental Health, the National Institute of Child Health and Human Development, the National Center for Complementary and Alternative Medicine (National Institutes of Health), and the Office of AIDS Research (National Institutes of Health).

The authors thank the study staff at each SWAN site.

The manuscript was reviewed by the Publications and Presentations Committee of SWAN and has its endorsement.

SWAN centers and investigators: Clinical Centers: University of Michigan, Ann Arbor, Michigan—Dr. MaryFran Sowers, Principal Investigator (National Institute of Nursing Research grant U01 NR04061); Massachusetts General Hospital, Boston, Massachusetts—Dr. Robert Neer, Principal Investigator, 1995–1999; Dr. Joel Finkelstein, Principal Investigator, 1999–present (National Institute on Aging (NIA) grant U01 AG12531); Rush-Presbyterian-St. Luke’s Medical Center, Rush University, Chicago, Illinois—Dr. Lynda Powell, Principal Investigator (NIA grant U01 AG12505); University of California, Davis, California/Kaiser Permanente, Oakland, California—Dr. Ellen Gold, Principal Investigator (NIA grant U01 AG12554); University of California, Los Angeles, California—Dr. Gail A. Greendale, Principal Investigator (NIA grant U01 A12539); Medical School, University of Medicine and Dentistry of New Jersey, Newark, New Jersey—Dr. Gerson Weiss, Principal Investigator (NIA grant U01 AG12535); University of Pittsburgh, Pittsburgh, Pennsylvania—Dr. Karen Matthews, Principal Investigator (NIA grant U01 AG12546); Central Laboratory: University of Michigan, Ann Arbor, Michigan—Dr. Rees Midgley, Principal Investigator, 1995–2000; Dr. Dan McConnell, Principal Investigator, 2000–present (NIA grant U01 AG12495, Central Ligand Assay Satellite Services); Coordinating Center: New England Research Institutes, Watertown, Massachusetts—Dr. Sonja McKinlay, Principal Investigator, 1995–1999; Dr. Kay Johannes, Principal Investigator, 1999–2001 (NIA grant U01 AG12553); Epidemiological Data Center, Pittsburgh, Pennsylvania—Dr. Kim Sutton-Tyrrell, Principal Investigator, 2001–present (NIA grant U01 AG12546); Steering Committee: Dr. Chris Gallagher, Chair, 1995–1996; Dr. Jenny Kelsey, Chair, 1996–2002; Dr. Susan Johnson, Chair, 2002–present; National Institutes of Health Project Offices: National Institute on Aging, Bethesda, Maryland—Drs. Sherry Sherman and Marcia Ory; National Institute of Nursing Research, Bethesda, Maryland—Dr. Carole Hudgings.

References