Does early growth influence timing of the menopause? Evidence from a British birth cohort

R. Hardy1 and D. Kuh

Medical Research Council National Survey of Health and Development, Department of Epidemiology and Public Health, Royal Free and University College Medical School, London WC1E 6BT, UK


    Abstract
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 Acknowledgements
 References
 
BACKGROUND: Few adult environmental or behavioural factors have been consistently associated with age at menopause. The peak number of follicles attained in utero or lost before ovulation begins may be more important. This study investigates whether birthweight, childhood body size, having been breastfed and early socioeconomic circumstances are associated with age at menopause. METHODS: Menopausal status and risk factor information have been collected prospectively from 1572 British women followed up since their birth in 1946, so far until 53 years. Cox's regression models were used to investigate the relationships between early life factors and rate of menopause. RESULTS: Age at menopause varied by duration of breastfeeding, weight at age 2 years, childhood socioeconomic status, but not birthweight. In a multiple regression model, women of low weight at 2 years had an earlier menopause [hazard ratio (HR) = 0.75 for highest versus lowest quarter: 95% confidence interval (CI) 0.54–1.02] and those who had been breastfed had a later menopause (HR = 0.69 for 7 months versus never breastfed: 95% CI 0.51–0.92) than others. The influence of socioeconomic status was attenuated. CONCLUSIONS: Early life influences may influence ovarian ageing, highlighting the importance of investigating factors from across the life course.

Key words: birth cohort/birthweight/breastfeeding/childhood growth/menopause


    Introduction
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 Acknowledgements
 References
 
Adult environmental factors have been shown to have relatively little impact on determination of the age at menopause, with only cigarette smoking (Jick et al., 1977Go; Kaufman et al., 1980Go; Willett et al., 1983Go; McKinlay et al., 1985Go; Luoto et al., 1994Go; Torgerson et al., 1994Go; Cramer et al., 1995aGo; Bromberger et al., 1997Go; van Noord et al., 1997Go; Do et al., 1998Go; Shinberg, 1998Go; Hardy et al., 2000Go) and parity (Stanford et al., 1987Go; Whelan et al., 1990Go; Torgerson et al., 1994Go; Cramer et al., 1995bGo; van Noord et al., 1997Go; Hardy and Kuh, 1999Go) showing consistent associations. Cigarette smokers and nulliparous women reach the menopause at younger ages.

Since a woman is born with an exhaustible and non-renewable supply of ovarian follicles, the peak number of follicles attained in utero or the number retained at birth have been proposed as the most important determinants of age at menopause (Ginsberg, 1991Go; Finch and Kirkwood, 2000Go). Very little is known about factors that determine the size of the initial pool of primordial follicles or about the subsequent rate of depletion in early postnatal life before ovulation begins (Finch and Kirkwood, 2000Go).

There is now considerable evidence that a number of chronic diseases are influenced by growth during the intrauterine period (Barker, 1994Go). Given that the pool of primordial follicles is formed during fetal development, it is plausible that retarded growth in utero could have a detrimental effect on the number of these follicles (te Velde et al., 1998Go) and therefore potentially lead to an earlier menopause (Cresswell et al., 1997Go). Few epidemiological studies have investigated the hypothesis. A cohort study of women in Hertfordshire, England (Cresswell et al., 1997Go) and a twin study in Australia (Treloar et al., 2000Go) have shown that lower birthweight was associated with later, rather than earlier, age at menopause, although in neither case was the association statistically significant. Low weight at 1 year was found to be associated with an earlier menopause in the Hertfordshire cohort (Cresswell et al., 1997Go). It is plausible that postnatal growth is associated with the extent of loss of follicles during childhood. Early life nutrition and socioeconomic conditions may be important in this respect since childhood growth trajectories may be influenced by such factors.

The MRC National Survey of Health and Development (NSHD), a birth cohort study where women have been followed up annually through the menopause, provides an ideal opportunity to investigate the hypothesis that early growth, either pre- or postnatal, influences age at menopause. The NSHD has detailed data regarding timing of the perimenopause and menopause, hysterectomy operations and hormone replacement therapy (HRT) use. It also has prospective measures of birthweight and height and weight throughout childhood, information on social conditions in early life and breastfeeding in infancy. In addition, a wealth of information on potential confounding factors throughout the life course has been collected.


    Materials and methods
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 Acknowledgements
 References
 
The MRC National Survey of Health and Development is a socially stratified birth cohort of 2547 women and 2815 men (Figure 1Go). There have been 20 follow-ups of the whole cohort between their birth in 1946 and age 53 years (Wadsworth, 1991Go, Wadsworth and Kuh, 1997Go). Those interviewed at age 43 years were, in most respects, representative of the native population of that age (Wadsworth et al., 1992Go). Since 1993 when cohort members were 47 years old, a postal questionnaire about health during the middle years of life has been sent annually to 1778 women study members with whom there was still contact (Kuh et al., 1997Go). Of the original cohort of women, 6% had died (n = 154), 9% were living abroad and were not in contact with the study (n = 232), 12% had refused to participate (n = 296) and 3% could not be traced (n = 87). During the 7 years (47–53 years) of the postal questionnaire survey, 1572 (88%) women have returned at least one postal questionnaire and the response rates in each year have varied from 84 to 90%.



View larger version (19K):
[in this window]
[in a new window]
 
Figure 1. Flow diagram indicating the losses from the original birth cohort.

 
Definition of menopausal status
Definition of menopausal status (premenopausal, perimenopausal and postmenopausal) was based on the criteria used in the Massachusetts Women's Health Study (Brambilla and McKinlay, 1989Go; Brambilla et al., 1994Go). Date of menopause was defined retrospectively following 12 months of amenorrhoea. The date of inception of the perimenopause was defined either as the date of last period if periods had stopped for between 3 and 12 months, or as the date when a woman reported that her menstrual cycle length had become more irregular in the preceding 12 months.

Women experiencing cessation of periods other than by a natural menopause were identified from the health questionnaires where they were asked to provide dates of hysterectomy or bilateral oophorectomy operations. A complete record of HRT use was also collected. The use of the majority of such preparations causes bleeding among post-menopausal women so it was not possible to define age at menopause, according to the definition of menopause based on cessation of menstruation used here, if HRT started before cessation of menstrual cycles was reported.

Explanatory variables
Birthweight of cohort members to the nearest quarter of a pound was extracted from medical records within a few weeks of delivery and converted into kilograms. Birthweight was categorized into a low birthweight group (<=2500 g) and three further groups (2501–3000, 3001–3500 and >=3501 g). Height and weight measured at 2 years was chosen for the analysis as they were the earliest postnatal measures of growth in the NSHD. Height and weight at age 7 years were also considered as these are the last premenarcheal measures for all the cohort.

Information on duration of breastfeeding was obtained from mother's reports to health visitors when the survey members were 2 years of age, and was here categorized into never breastfed, breastfed for <=3 months, 4–6 months and >=7 months. Father's social class and crowding in the household were used as markers of socioeconomic status in childhood. Social class was assigned from father's occupation when the survey member was 4 years of age, or if this was unknown, from the father's occupation when the survey member was 11 years of age or at the time of birth of the survey member. Those from non-manual and manual origins were distinguished for the purposes of the analysis. A measure of `crowding' at age 2 years was defined using the number of persons per room, with `crowded' being defined as two or more persons per room.

Potential confounding variables
Potential confounders, identified because of their possible association with age at menopause, were parity, smoking, body mass index (BMI), adult socioeconomic status and experience of a unilateral oophorectomy operation. Records of all live births have been collected throughout the adult life of the cohort. Smoking and BMI were defined using information given when the cohort were 36 years of age since this represents a premenopausal measure for all the cohort. Occurrences of unilateral oophorectomy (n = 26) were reported in the women's health questionnaires. Socioeconomic status was represented by spouse's occupational social class, based on the current or most recent occupation at the 43 year follow-up, and the women's own educational qualifications by the age of 26 years.

Statistical methods
Cox's proportional hazard models (Cox, 1972Go) were used to obtain estimates of the hazard ratio for the risk factors. Follow-up time was in months since age 25 years and was until menopause (n = 412), or, if no date of menopause was available, until the date of one of the following events: hysterectomy (or bilateral oophorectomy) (n = 290), start of HRT use (n = 414) or last returned questionnaire if pre-menopausal or peri-menopausal at that time (n = 398). Follow-up was treated as censored if the event was not menopause. Where the date of menopause was not specified to the nearest month, the mid point of the year of menopause was taken. The small number of women (n = 32) whose periods stopped for other surgical reasons (usually endometrial ablation) or medical treatment (for example, chemotherapy) before the menopause, and those who had or were taking oral contraceptives (n = 23) or for whom data were not complete (n = 3), were omitted.

The assumption of proportional hazards was checked both by inspection of plots and by the use of time-dependent covariates. Initially, unadjusted estimates of the hazard ratio for birthweight and each height and weight measure compared with a baseline category were obtained, as well as estimates for duration of breastfeeding and the socioeconomic indicators. Childhood heights and weights were split into quarters of the distribution. Tests for linear trend across the ordered categorical variables were carried out. Models including both birthweight and childhood heights and weights were then investigated in order to assess the relative importance of prenatal and postnatal growth (Lucas et al., 1999Go). Tests for interactions between birthweight and childhood heights and weights were carried out by testing whether any linear effect of later measures varied systematically across categories of birthweight. The early life socioeconomic factors and breastfeeding were added to the model judged to represent the effects on menopausal age of early growth. Finally, the influence of all the potential adult confounders was assessed.


    Results
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 Acknowledgements
 References
 
There was no evidence of an association between birthweight and age at menopause (test for linear trend P = 0.4) (Table IGo). There was a suggestion of a decreasing hazard of menopause with increasing weight at age 2 years (test for linear trend P = 0.08), indicating a later age at menopause with increasing weight. By the end of follow-up when the cohort was aged 53 years and 7 months, 51% of those in the lowest quarter of the weight at 2 years distribution had reached menopause compared with only 44% in the top quarter (Figure 2Go). At 50 years, the relative difference was greater, as 23% of the lightest group had reached menopause compared with 16% in the heaviest group. The estimates indicated that those in the top half of the weight distribution appeared to have a later menopause than those of lower weight. There was no evidence of any such trend with weight at age 7 years or either of the height measures. The pattern for height at age 2 years appeared to be curvilinear with those in the middle categories having a later age at menopause than those in the bottom and top groups. Addition of a quadratic term for height to the linear trend improves the fit of the model (P = 0.08).


View this table:
[in this window]
[in a new window]
 
Table I. Unadjusted hazard ratios for age at menopause by potential early life risk factors obtained from Cox's proportional hazard models
 


View larger version (21K):
[in this window]
[in a new window]
 
Figure 2. Kaplan–Meier estimates of the proportion of women who are still pre-menopausal or peri-menopausal by age for the top and bottom quarter of the distribution of weight at age 2 years.

 
Women who had been breastfed for at least 7 months had a 26% lower hazard rate than those not breastfed [HR (95% CI) = 0.74 (0.57, 0.97)]. There appeared to be an increased rate of menopause in those women from a manual social class in childhood compared with those from a non-manual social class and those who experienced overcrowding in early life compared to those who did not (Table IGo). The proportional hazards assumptions were judged to be reasonable in all cases, although there was some suggestion that the effect of weight at 2 years weakened with increasing age.

In the model including both birthweight and weight at age 2 years, the effect of weight at 2 years is strengthened (test for linear trend P = 0.06). The influence of weight at age 2 years is confounded with height at 2 years. No evidence of any interaction between the growth variables were found. Hence, birthweight and weight at 2 years are considered in further models. For parsimony, social class was omitted from further models as investigation (results not shown) suggested that crowding and social class were confounded with each other.

Breastfeeding and crowding were added to the growth model. In the subgroup of women restricted to those who have full early life information (n = 1238), the unadjusted effect of crowding was weaker (Table IIGo, unadjusted column) than that observed in the unrestricted sample (Table IGo). This effect was further reduced after adjustment. Little change was observed in the estimates for breastfeeding in the full model, while the linear effect of weight at 2 years became stronger (P = 0.04) with a 25% lower hazard rate in the top compared with the bottom group (Table IIGo).


View this table:
[in this window]
[in a new window]
 
Table II. Unadjusted and adjusted hazard ratios for age at menopause obtained from Cox's proportional hazard models
 
Further adjustment for the later life risk factors showed that the linear effects of weight at 2 years and breastfeeding remained associated (P <= 0.05 in each case) with menopausal age after addition of each potential adult confounder (parity, cigarette smoking, BMI, socioeconomic status and unilateral oophorectomy), added separately.

Of these adult variables, only parity had a strong association (P < 0.001) with age at menopause, with women with no children having an earlier menopause than women with children.


    Discussion
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 Acknowledgements
 References
 
The NSHD provides some evidence of a link between early life development and age at menopause. Women who were heavier at age 2 years were found to reach the menopause later than other women. This is consistent with the findings of a study of Hertfordshire women (Cresswell et al., 1997Go), which suggested that low weight at 1 year was associated with an earlier menopause. The same workers also observed that, in a smaller British cohort of women aged 40–42 years for whom detailed birth measurements were available, the 12 who had undergone an early menopause were shorter at birth and of a higher ponderal index (birthweight/length3) than the other women (Cresswell et al., 1997Go). They proposed that the evidence was consistent with the hypothesis that both poor growth in infancy and early menopause were associated through a common origin of low rates of linear growth in utero since short body length at birth has been associated with low rates of growth in infancy. The lack of any association between birthweight and age at menopause was consistent with previous studies (Cresswell et al., 1997Go; Treloar et al., 2000Go). Neither gestational age nor other measures of size at birth were recorded in the NSHD. Direct evidence of an association between poor intrauterine growth and number of primordial follicles is also scarce (De Bruin et al., 1998Go, 2001Go).

The possibility that the effect of weight at 2 years represents a postnatal influence on ovarian function remains. The effect was strengthened after adjustment for birthweight indicating that it was growth between birth and 2 years that is important. The fact that weight was attenuated in a model including height at 2 years suggests that the important aspect of size is difficult to distinguish. Additional analyses found no trend in menopausal age across categories of BMI at age 2 years, suggesting that adiposity is not the important factor. There was no association with body size measures at 7 years, suggesting that any effect is occurring early in postnatal development. Although 2 years is the first age after birth at which body size was measured, the effect of breastfeeding on age at menopause may be a marker of the effect of earlier postnatal growth. In this cohort, breastfeeding has previously been found to influence adult leg length but not trunk length, a finding likely to indicate an effect on early growth (Wadsworth et al., 2002Go). A study in New Guinea found that women who were currently undernourished and smaller in adult height and weight reached menopause 4 years earlier than a better nourished sample (Scragg, 1973Go). This could reflect the effects of lifelong poor nutrition and poor growth, as opposed to adult nutrition, on reproductive ageing. The association with breastfeeding may indicate the importance of early nutrition in relation to ovarian development.

Few studies have considered early socioeconomic conditions in relation to age at menopause (Shinberg, 1998Go; Hardy et al., 2000Go). In the unadjusted analyses those of a lower socioeconomic status in childhood had an earlier menopause, but the effect was much weaker in the adjusted model. This was due to the reduction in the size of the unadjusted effect in the restricted sample as well as to confounding with the growth and nutrition variables. A study of Wisconsin women (Shinberg, 1998Go) observed an effect of parental occupation. Women whose fathers were farmers had a later menopause compared with women whose fathers were not and it was suggested that this may be a result of a different lifestyle including nutrition. Alternatively it might be that stress (Bromberger et al., 1997Go), either early in life or across the life course, may be responsible for any social class differences in age at menopause.

All the early life influences identified are relatively small. Given the observed relationships between childhood growth and breastfeeding (Wadsworth et al., 2002Go), it is interesting that the effect of weight at age 2 years gets stronger rather than weaker after adjustment. This, together with the previously observed relationship between low cognitive function and earlier menopause in this cohort (Richards et al., 1999Go), suggests that there may be a number of independent pathways from childhood risk to age at menopause. All variables considered are proxy measures for the underlying biological mechanism or mechanisms and hence further research needs to consider what such processes might be and how best they can be measured. The lack of many strong adult environmental risk factors for age at menopause supports further exploration of early life influences on ovarian ageing.

The estimated median age at menopause in this cohort cannot yet be obtained, as not enough women have reached menopause. The estimate will thus be greater than 52 years and 7 months (1 year prior to current maximum follow-up), later than the most reliable estimates of median age at menopause from Western industrialized countries of between 50 and 52 years (McKinlay, 1996Go). There has been considerable debate as to whether there has been a secular increase in the age at menopause. Data have shown an increase between the mid-nineteenth and the mid-twentieth century (Flint, 1978Go), but it has been argued that this is a result of improving methodology in the study of menopausal age (McKinlay et al., 1992Go). An increasing trend would be consistent with the improvement in early nutrition and better childhood growth of recent cohorts of women, although any increase may have been masked by the effects of an increased prevalence during the twentieth century of the main adult risk factors for early menopause of nulliparity and cigarette smoking (Swerdlow et al., 2001Go). The increased use of HRT and the increased number of hysterectomy operations may have influenced the more recent estimates of menopausal age. A greater proportion of women were using HRT in the NSHD compared with most previously studied populations. This results in the follow-up of more women being censored, and, since women are likely to be approaching the menopause at the time that they start HRT use, the estimated median age may be biased upwards. In a competing risks framework (Farewell, 1979Go; Hardy et al., 2000Go), the median age at first event (either menopause or start of HRT use) in the NSHD is 50 years and 5 months, supporting this possibility.

The increased popularity in HRT use will be a constant problem in epidemiological investigation of the natural menopause if menstrual change is the indicator of menopausal status. The results presented in this paper can only be interpreted under the conditions existing in the study and may not be applicable to populations with different patterns of HRT use. Replication of the findings would be valuable in samples with lower HRT use. Since the NSHD is a prospective study, birthweight was obtained from medical records soon after birth and childhood heights and weights were measured. Confounding variable information was also obtained at follow-ups prior to menopause and hence was less subject to recall problems than if collected concurrently with menopausal status.

The independent associations of weight at age 2 years and breastfeeding suggest that early life factors acting in utero or in early childhood may play some part in determining a woman's age at menopause. The potential importance of a life course approach in understanding the menopause is thus highlighted (Hardy and Kuh, 2002Go). The previous inconsistency in the findings on adult behavioural and social factors and the age at menopause may be due, at least in part, to a failure to account for earlier life factors.


    Acknowledgements
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 Acknowledgements
 References
 
The authors wish to thank Professor Michael Wadsworth and Dr Marcus Richards for helpful comments on a previous draft of the paper. Funding for this study was provided by the Medical Research Council.


    Notes
 
1 To whom correspondence should be addressed. E-mail: rebecca.hardy{at}ucl.ac.uk Back


    References
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 Acknowledgements
 References
 
Barker, D.J.P. (1994) Mothers, Babies, and Disease in Later Life. BMJ Publishing Group, London.

Brambilla, D.J. and McKinlay, S.M. (1989) A prospective study of factors affecting age at menopause. J. Clin. Epidemiol., 42, 1031–1039.[ISI][Medline]

Brambilla, D.J., McKinlay, S.M. and Johannes, C.B. (1994) Defining the perimenopause for application in epidemiologic investigations. Am. J. Epidemiol., 140, 1091–1095.[Abstract]

Bromberger, J.T., Matthews, K.A., Kuller, L.H., Wing, R.R., Meilahn, E.N. and Plantinga, P. (1997) Prospective study of the determinants of age at menopause. Am. J. Epidemiol., 145, 124–133.[Abstract]

Cox, D.R. (1972) Regression models and life tables (with discussion). J. R. Statist. Soc., B, 74, 187–220.

Cramer, D.W., Harlow, B.L., Xu, H., Fraer, C. and Barbieri, R. (1995a) Cross-sectional and case-controlled analyses of the association between smoking and early menopause. Maturitas, 22, 79–87.[ISI][Medline]

Cramer, D.W., Xu, H and Harlow, B.L. (1995b) Does `incessant' ovulation increase risk for early menopause? Am. J. Obstet. Gynecol., 172, 568–573.[ISI][Medline]

Cresswell, J.L., Egger, P., Fall, C.H.D., Osmond, C., Fraser, R.B. and Barker, D.J.P. (1997) Is the age of menopause determined in-utero? Early Hum. Dev., 49, 143–148.[ISI][Medline]

De Bruin, J.P., Dorland, M., Bruinse, H.W., Spliet, W., Nikkels, P.G.J. and te Velde, E.R. (1998) Fetal growth retardation as a cause of impaired ovarian development. Early Hum. Dev., 51, 39–46.[ISI][Medline]

De Bruin, J.P., Nikkels, P.G.J., Bruinse, H.W., van Haaften, M., Looman, C.W.N. and te Velde, E.R. (2001) Morphometry of human ovaries in normal and growth-restricted fetuses. Early Hum. Dev., 60, 179–192.[ISI][Medline]

Do, K-A., Treloar, S.A., Pandeya, N., Purdie, D., Green, A.C., Heath, A.C. and Martin, N.G. (1998) Predictive factors of age at menopause in a large Australian twin study. Hum. Biol., 70, 1073–1091.[ISI][Medline]

Farewell, V.T. (1979) An application of Cox's proportional hazards model to multiple infection data. J. R. Statist. Soc., C, 28, 136–143.

Finch, C.E. and Kirkwood, T.B.L. (2000) Chance, Development and Aging. Oxford University Press, Oxford.

Flint, M. (1978) Is there a secular trend in age at menopause? Maturitas, 1, 133–139.[ISI][Medline]

Ginsberg, J. (1991) What determines the age at menopause? The number of ovarian follicles seems the most important factor. Br. Med. J., 302, 1288–1289.[ISI][Medline]

Hardy, R. and Kuh, D. (1999) Reproductive characteristics and age at inception of the perimenopause in a British national cohort. Am. J. Epidemiol., 149, 612–620.[Abstract]

Hardy, R. and Kuh, D. (2002) Menopause and gynaecological disorders: a life course perspective. In Kuh, D. and Hardy, R. (eds), A Life Course Approach to Women's Health. Oxford University Press, Oxford. pp 64–85.

Hardy, R., Kuh, D. and Wadsworth, M. (2000) Smoking, body mass index, socioeconomic status and the age at menopause transition in a British national cohort. Int. J. Epidemiol., 29, 845–851.[Abstract/Free Full Text]

Jick, H., Porter, J. and Morrison, A.S. (1977) Relation between smoking and age of natural menopause. Lancet, 1, 1354–1355.[Medline]

Kaufman, D.W., Slone, D., Rosenberg, L., Miettinen, O.S. and Shapiro, S. (1980) Cigarette smoking and age at natural menopause. Am. J. Public Health, 70, 420–422.[Abstract]

Kuh, D.L., Wadsworth, M. and Hardy, R. (1997) Women's health in midlife: the influence of the menopause, social factors and health in earlier life. Br. J. Obstet. Gynaecol., 104, 923–933.[ISI][Medline]

Lucas, A., Fewtrell, M.S. and Cole, T.J. (1999) Fetal origins of adult disease—the hypothesis revisited. Br. Med. J., 319, 245–249.[Free Full Text]

Luoto, R., Kaprio, J. and Uutela, A. (1994) Age at menopause and sociodemographic status in Finland. Am. J. Epidemiol., 139, 64–76.[Abstract]

McKinlay, S.M. (1996) The normal menopause transition: an overview. Maturitas, 23, 37–45.

McKinlay, S.M., Bifano, N.L. and McKinlay, J.B. (1985) Smoking and age at menopause in women. Ann. Intern. Med., 103, 350–356.[ISI][Medline]

McKinlay, S., Brambilla, D. and Posner, J. (1992) The normal menopause transition. Am. J. Hum. Biol., 4, 37–46.[ISI]

Richards, M., Kuh, D., Hardy, R. and Wadsworth, M.E.J. (1999) Lifetime cognitive function and timing of the natural menopause. Neurology, 53, 308–314.[Abstract/Free Full Text]

Scragg. R.F.R. (1973) Menopause and reproductive span in rural Niugini. Proc. Papua New Guinea Med. Soc., Port Moresby, Papua New Guinea, pp. 126–144. Papua New Guinea Medical Society, Port Moresby.

Shinberg, D.S. (1998) An event history analysis of age at last menstrual period: correlates of natural and surgical menopause among midlife Wisconsin women. Soc. Sci. Med., 46, 1381–1396.[ISI][Medline]

Stanford, J.L., Hartge, P., Brinton, L.A., Hoover, R.N. and Brookmeyer, R. (1987) Factors influencing the age at natural menopause. J. Chron. Dis., 40, 995–1002.[ISI][Medline]

Swerdlow, A., dos Santos Silva, I. and Doll, R. (2001) Cancer Incidence and Mortality in England and Wales: Trends and Risk Factors. Oxford University Press, Oxford.

te Velde, E.R., Dorland, M. and Broekmans, F.J. (1998) Age at menopause as a marker of reproductive ageing. Maturitas, 30, 119–125.[ISI][Medline]

Torgerson, D.J., Avenell, A., Russell, I.T. and Reid, D.M. (1994) Factors associated with onset of menopause in women aged 45–49. Maturitas, 19, 83–92.[ISI][Medline]

Treloar, S.A., Sadrzadeh, S., Do, K-A, Martin, N.G. and Lambalk, C.B. (2000) Birth weight and age at menopause in Australian female twin pairs: exploration of the fetal origin hypothesis. Hum. Reprod., 15, 55–59.[Abstract/Free Full Text]

van Noord, P.A., Dubas, J.S., Dorland, M., Boersma, H. and te Velde, E. (1997) Age at natural menopause in a population-based screening cohort: the role of menarche, fecundity, and lifestyle factors. Fertil. Steril., 68, 95–102.[ISI][Medline]

Wadsworth, M.E.J. (1991) The Imprint of Time: Childhood, History and Adult Life. Oxford University Press, Oxford.

Wadsworth, M.E.J. and Kuh, D.J.L. (1997) Childhood influences on adult health: a review of recent work in the British 1946 national cohort study, the MRC National Survey of Health and Development. Paediat. Perinat. Epidemiol., 11, 2–20.

Wadsworth, M.E.J., Mann, S.L., Rodgers, B., Kuh, D.J.L., Hilder, W.S. and Yusuf, E.J. (1992) Loss and representativeness in a 43 year follow-up of a national birth cohort. J. Epidemiol. Community Health, 46, 300–304.[Abstract]

Wadsworth, M.E.J., Hardy, R.J., Paul, A.A., Marshall, S.F. and Cole, T.J. (2002). Leg and trunk length at 43 years in relation to childhood health, diet and family circumstances; evidence from the 1946 national birth cohort. Int. J. Epidemiol., 31, 383–390.[Abstract/Free Full Text]

Whelan, E.A., Sandler, D.P., McConnaughey, D.R. and Weinberg, C.R. (1990) Menstrual and reproductive characteristics and age at natural menopause. Am. J. Epidemiol., 131, 625–632.[Abstract]

Willett, W., Stampfer, M.J., Bain, C., Lipnick, R., Speizer, F.E., Rosner, B., Cramer, D. and Hennekens, C.H. (1983) Cigarette smoking, relative weight and menopause. Am. J. Epidemiol., 117, 651–658.[Abstract]

Submitted on December 11, 2001; resubmitted on March 18, 2002; accepted on May 9, 2002.