1 1Department Of Obstetrics and Gynaecology, Diakonessen Hospital Utrecht, 2 Institute of Animal Sciences, Wageningen University, Wageningen, 3 Julius Center for Patient Oriented Research, Utrecht University, 4 Department of Human Genetics, 5 Department of Obstetrics and Gynaecology, University Medical Centre Utrecht, The Netherlands and 6 Department of Animal Science, University of Illinois, Urbana, USA
![]() |
Abstract |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Key words: fertility/genetics/heritability/menopause/reproductive failure
![]() |
Introduction |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
There is a wide variation in the age at which normal menopause commences, varying from between 40 to 60 years. The role of environmental and lifestyle factors in the timing of age at natural menopause has been thoroughly investigated. Although analysis shows significant influence of factors such as smoking, parity and possibly also oral contraceptive use on menopausal age, only a small part of the large variance in menopausal age can be explained by these factors (,van Noord et al.1997) and the high significance resides more in the large sample sizes employed than in the magnitude of the induced changes.
Cramer and co-workers suggested that menopause could be under the control of genetic factors, given their finding of a positive correlation between the menopausal ages of mothers and daughters (,Cramer et al.1995). Measurements of the genetic contribution to a phenotypical feature, in this case menopausal age, are referred to as heritability estimates, or the proportion of the phenotypical feature that is determined by genetic factors. Heritability is frequently expressed as a percentage. Recently, Snieder used twin data to estimate the effects of genetic and environmental factors on age at menopause, and reported a heritability of 63% (,Snieder et al.1998
). In a further twin study, based on Australian data, a heritability of between 31 and 53% was found (,Treloar et al.1998
).
In this article we estimate the heritability for age at natural menopause based on data collected from a population sample of pairs of sisters (singleton sisters) and twins previously randomly selected to participate in a breast cancer screening project [Doorlopend Onderzoek Morbiditeit, Mortaliteit (DOM) project] (,de Waard et al.1984). These subjects have been followed since 1974, and in contrast to previous studies on the heritability of menopause, their menopausal phenotype had not played a role in their inclusion.
![]() |
Materials and methods |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
We wished to include those women from the DOM project who had experienced natural menopause. Therefore, the only women included in our study were older than 57 years at the time of the latest survey (97.5 % of women at age 57, being the mean age at menopause +2 SD, can be expected to be post-menopausal). Records were excluded if menopause was due to surgical intervention, if menopause was obscured by hormone replacement therapy or if the menopausal age was out of the range of 4063 years. After these restrictions, data on age of natural menopause for 5877 women (menopause database) were available.
Full sisters were identified in the menopause database based on maiden name, current age of their mother or age of death of their mother, number of brothers, number of sisters, consistent birth order and consistent birth dates in relation to birth order. Application of these criteria identified 118 families containing at least two menopausal sisters: one family contained four sisters, five families contained three sisters and 112 families contained two sisters to give a total of 243 individuals. Twins, identified as sisters with the same birth date, were specifically excluded from the data set. This data set will subsequently be referred to as the singleton data set.
The singleton data were drawn from two DOM cohorts. Of the 243 individuals 209 women were part of DOM I. These women were born between 1911 and 1925 and all had completed five rounds of screening. The initial screening took place in 1975 and the last screening was in 1984 (,van Noord et al.1997). Women in this cohort were at least 59 and on average 64.5 years of age at the last screening. DOM II contributed 34 women to the singleton data set. DOM II started in 1981 and consisted of women born between 1926 and 1931. All women included in this study completed the second screening which took place in 1984. The women selected from this DOM-cohort were at least 57 years of age and on average 61.4 years of age at this last screening round. The mean age difference between all sisters in the singleton data set was 3.9 years.
A second data set consisting of twins was constructed. The twin data set was based on twins in which both twin sisters were included in the menopause database (12 twin pairs) and on women in the menopause database who reported having a twin sister who was not included in the initial data set. In 1997 retrospective data for menopausal age were obtained from these twin sisters directly by means of a questionnaire. Only twin pairs in which both members had experienced natural menopause were included, resulting in a data set of 22 monozygotic and 37 dizygotic twin pairs. Twelve pairs had been excluded because of: premenopausal state (one pair), surgical menopause (nine pairs) or hormone replacement therapy (two pairs). In 1997 the ages of the twin pairs ranged from 5779 years, with a mean of 66.8 years.
Methods
The phenotype is determined by the genotype and the environment. The values of a certain phenotypic feature for all individuals in a population have a mean and variance around this mean. The phenotypic variance (VP) is determined by the genotypic variance (VG) and the environmental variance (VE), or VP = VG + VE. Heritability (h2) is a statistical parameter that indicates the genetic component of the phenotypic feature under study and can be defined as the genotypic variance divided by the phenotypic variance, or h2 = VG /VP. The value of heritability ranges from 01 (1100%), a low value indicates a small contribution of genetic factors to the phenotype and a high value a strong contribution of genetic factors to the phenotype.
The phenotypic correlation coefficient for relatives (r) is defined as the phenotypic variance between relatives divided by the phenotypic variance in the population. There is a relation between heritability and the correlation coefficient, which depends on the degree of genetic similarity between relatives. Fisher found that on average the genetic similarity between first degree relatives is approximately 50% and the genetic similarity between monozygotic twins is 100% (Fisher, 1918). Consequently, the formulas for the relations between heritability and the correlation coefficients, relevant to the present study, are: h2 = 2r for singleton sisters and dizygotic twins and h2 = r for monozygotic twins.
Analysis of variance (ANOVA) is a relatively simple method for providing the terms of the heritability formula for both the singleton and twin data in the present study. The heritability estimate for age at natural menopause is based on the variance in natural menopausal ages occurring between sisters, relative to the variance occurring in the total menopause database. Importantly, the mean and variance estimates of the total singleton data set were almost identical to those observed for the complete menopause database of 5877 women. The estimate of the heritability of natural menopause for the singleton and dizygotic twin sisters was obtained by taking twice the sister variance for menopausal age and dividing this value by the total variance of menopausal age (h2 = 2r), for the monozygotic sisters by dividing the sister variance by the total variance of menopausal age (h2 = r).
In general, the presence of common environmental factors can lead to an overestimation of the heritability (Falconer, 1989). Twin studies, comparing the results of mono- and dizygotic twin data, have traditionally been used to distinguish between genetic and environmental contributions to the phenotype. The underlying assumption is that differences in the phenotype between monozygotic twins are solely caused by the environment. The Mx statistical method has been developed for heritability estimates in twin studies and is based on the principles of ANOVA (Neale, 1997
). The Mx procedure is a combined analysis of data from di- and monozygotic twins in combination with a correction for environmental factors. Consequently, the outcome is expressed as a combined twin heritability value. In our Mx procedure two different models were fitted. One model included only an additive genetic and an environmental component and a second model which additionally included a shared environmental component.
Heritability estimates are frequently used in animal breeding, in which it is essential to distinguish exactly between genetic and environmental contributions. For this purpose the Gibbs sampler analysis has been developed, which is a Markov Chain Monte Carlo technique (Gelfland and Smith, 1990; Casella and George, 1992
). In this analysis the parameters of the statistical model are continuously varied until the most likely set of conditions is reached which match the observed distribution of the phenotypical feature. In distinction to ANOVA and Mx analysis, this is a maximum likelihood estimate of heritability and has proven to be far more robust in establishing the relative contribution of genetic and environmental contributions. Outcomes are, in the case of age at menopause, the posterior (or post analysis) probability distributions of the heritability of menopause. These distributions can be used to determine the posterior mean and the 90% highest posterior density region. The latter parameter reflects the accuracy of the heritability estimate and can be used as a confidence interval. For a more technical description of our Gibbs sampler analysis, see Appendix A.
Essentially all three employed methods, ANOVA, Mx and Gibbs, deliver correlation coefficient values for sisters for menopausal age, which are corrected to heritability values by multiplying with their degree of genetic similarity.
![]() |
Results |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
|
|
The combined twin data set offered the possibility of estimating both additive genetic and common environmental effects. For the Mx analysis of the twin data, the results of the model including additive genetic and environmental variances gave the simplest explanation of observed variance of the data. The inclusion of common environmental effects in the second model did not provide a significant improvement of the explanation of the data. The estimated heritability for the combined twin data was 0.72.
From the analysis using the Gibbs sampler the estimated heritability for age of natural menopause based on the singleton sister data was 0.85. The 90% highest posterior density region of the heritability estimate ranged from 0.610.92. In the analysis of twins, shared environmental factors and genetic factors were both modelled as random factors. This analysis resulted in a heritability estimate of 0.71, with the 90% highest posterior density ranging from 0.550.83, which does not deviate significantly from the estimate obtained from the singleton data. The estimated variance due to shared environmental effects was essentially zero and corresponded with the result obtained from the Mx analysis.
The heritability estimates provide some measure of what can be expected for the same feature in other family members. If one family member exhibits a low age at natural menopause, other family members are likely to show the same trend. Similarly, if one family member has a high age at natural menopause, in general the same will be true for other family members also. For example, if we use our heritability estimate of 0.85 for singletons to assess the expected menopausal age of a woman whose sister became menopausal 9 years earlier than the population average of 51, i.e. at 42 years of age, her own expected menopausal age, on average, will be 51 [9x0.85x0.5] = 47 years of age. However, it should be noted that in individual cases it is not possible to predict the impact of heritability on other family members exactly, without determining the specific genetic variation in the genes contributing to this trait.
![]() |
Discussion |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
For this study we used unique data from a longitudinal study on breast cancer. Participating women were randomly selected, without their menopausal phenotype playing a role in their inclusion. As data on menopausal age were obtained shortly after reaching menopause, recall errors are expected to be small (den Tonkelaar, 1997). For this analysis we only used data from pairs of women who were full sisters of different ages or twin sisters and in which both sisters had experienced natural menopause. On the one hand this places a mild limitation on the size of the study group; on the other hand every case in the sample is informative and uncertainties introduced by modelling the most likely age of menopause for those sisters who are still premenopausal, or with an obscured menopause, are avoided. Only post-menopausal sister pairs older than 57 years at the time of the latest survey were included in our study. This might have a small effect on the total heritability estimate because this type of censoring would exclude pairs of sisters in which one of them has not yet experienced natural menopause by the age of 57. Therefore, excluding such individuals from the analysis reduces the number of sister pairs with possible high discordance in age of menopause because one of them is an outlier at the high end of the distribution. This would increase the average similarity between the remaining sister pairs in the data and increases the estimated heritability accordingly. However, effects on outcomes were expected to be small since
97.5 % of women at age 57 are already menopausal and 99.8% of women are menopausal by 60 years. In fact, 85% of women in the survey were over 60 years of age, from which only three experienced menopause after the age of 60 years.
Although no significance was reached, mean ages at natural menopause tended to be lower in twins, especially dizygotic twins. The same has been observed in earlier studies (,Martin et al.1997; ,Snieder et al.1998
). Dizygotic twinning is a familial trait, in which an explanation of possible earlier menopause could be higher concentrations of circulating follicle-stimulating hormone, causing polyovulation and earlier depletion of the follicle store (,Martin et al.1991
).
In general, heritability estimates based on singleton and dizygotic twin data might be expected to be positively biased due to shared environmental effects. Here, we used our combined twin sample to determine the effect of common environmental factors on menopausal age. Our twin sample was too small to discriminate exactly between additive genetic, non-additive genetic and common environmental effects. However, according to our results there was a non-significant effect of common environmental factors on the heritability estimate of menopausal age, which agrees with the conclusions from a larger twin sample (,Snieder et al.1998).
In addition to the relatively simple and well known methods of estimating heritability such as analysis of variance and Mx modelling for twin studies, analyses were performed using an advanced Markov Chain Monte Carlo model, as advocated by Treloar (,Treloar et al.1998). This model has two advantages above the other two. Firstly, it can combine all degrees of genetic relationship between individuals and so, using a single model, all analyses for both singleton and twin groups could be carried out. Secondly, obtained solutions enable the determination of confidence intervals. Importantly, all three analytic methods delivered almost identical overall heritability estimates, with the exception of the heritability estimate of 0.96 for dizygotic twins with ANOVA. This may be due to the limited number of observations. Interestingly, the same estimate using the more refined Gibbs sampling is 0.85, which seems more in line with the estimates for the singletons and monozygotic twins.
Age-related decline of fertility in females starts from 30 years of age; the average monthly pregnancy rate is halved by age 35, has dropped to one quarter by age 38 and is essentially zero by age 41, some 10 years preceding the average age of commencement of menopause (Wood, 1989; ,van Noord-Zaadstra et al.1991
). The decrease of fertility with age is further illustrated by IVF-treatments where implantation rates per embryo (,van Kooij et al.1996
) and probability of pregnancy (FIVNAT, 1993
) markedly diminish from 3738 years of age onwards. These events are chronologically paralleled by the rate of follicle atresia (,Faddy et al.1992
).
When data from cross-sectional studies on mean ages of reproductive events in females are combined, it appears that menopause is preceded by the onset of subfertility, infertility and irregularity of menses at distinct mean time intervals of 20, 10 and 6 years respectively (te ,Velde et al.1998). It is a plausible hypothesis that the ongoing depletion of the ovarian follicle store reflects a common underlying aetiology for all reproductive phases related to ovarian aging and that differences in the rate of depletion between women are related to differences in the chronological age at which individual women commence natural menopause. In contrast to the other events, age at menopause is an unambiguously and easily defined parameter, which has clear advantages for epidemiological purposes and can be used as a marker for the timing of the preceding reproductive events.
According to our heritability estimate, a woman with one or more first degree relatives with a history of early menopause is liable to experience earlier menopause herself. Further, this same woman is also expected to start becoming less fertile and to be completely infertile at an earlier age. Accordingly, she is at a greatly increased risk of remaining childless if she delays childbearing, as do many women in this day and age for socio-economic reasons. In the present study heritability estimates are based on sister pairs, theoretically the same heritability estimates and their implications are valid for motherdaughter pairs.
The underlying assumption in heritability analyses for continuously varying traits in which the variation is predominantly genetically determined, as in the case of age of menopause studied here, is that several to many genes independently make an additive contribution to the variation. The results of our own and related studies should encourage further research into defining the number, identity and genetic variation of the genes which determine variation in age of menopause by appropriate genome analyses. The imminent availability of accurate DNA sequence data for the whole human genome combined with information on the location and variation of all genes will assist enormously in this endeavour. There is a distinct possibility that the same genes which determine variation in the age of menopause also regulate the rate and age at which women become infertile prior to menopause. Such research will demonstrate which specific variation within the genes concerned is associated with early reproductive failure and enable establishment of global DNA profiles to identify women with an increased risk of premature infertility.
![]() |
Appendix A |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
![]() |
Acknowledgements |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
![]() |
Notes |
---|
![]() |
References |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Cramer, D.W., Xu, H. and Harlow B.L. (1995) Family history as a predictor of early menopause. Fertil. Steril., 64, 740745.[ISI][Medline]
Faddy, M.J., Gosden, R.G., Gougeon, A. et al. (1992) Accelerated disappearance of ovarian follicles in mid-life: implications for forecasting menopause. Hum. Reprod., 7, 13421346.[Abstract]
Falconer, D.S. (1989) Introduction to quantitative genetics. Longman Scientific and Technical, New York, p. 438.
FIVNAT (French In vitro National) (1993) French national IVF registry: analysis of 19861990 data. Fertil. Steril., 59, 587595.[ISI][Medline]
Fisher, R.A. (1918) The correlation between relatives on the supposition of Mendelian inheritance. Trans. R. Soc. Edin., 52, 399433.
Gelfland, A.E. and Smith, A.F.M. (1990) Sampling based approaches to calculating marginal densities. J. Am. Stat. Assoc., 85, 398409.[ISI]
Janss, L.L.G., Thompson, R. and van Arendonk, J.A.M. (1995) Application of Gibbs sampling for inference in a mixed major gene-polygenic inheritance model in animal populations. Theor. Appl. Genet., 91, 11371147.[ISI]
van Kooij, R.J., Looman, C.W.N., Habbema, J.D.F. et al. (1996) Age dependent decrease in embryo implantation rate after in vitro fertilization. Fertil. Steril., 66, 769775.[ISI][Medline]
Martin, N.G., Healey, S.C., Pangan, T.S. et al. (1997) Do mothers of dizygotic twins have earlier menopause? A role for fragile X? Am. J. Med. Genet., 69, 114116.[ISI][Medline]
Martin, N.G., Shanley, S., Butt, K. et al. (1991) Excessive follicular recruitment and growth in mothers of spontaneous dizygotic twins. Acta Genet. Gemellol., 40, 291301.
Neale, M.C. (1997) Mx: statistical modeling, 4th edn. Department of Psychiatry, Richmond, USA.
van Noord, P.A.H., Dubas, J.S., Dorland, M. et al. (1997) Age at natural menopause in a population-based screening cohort: the role of menarche, fecundity, and lifestyle factors. Fertil. Steril., 68, 95102.[ISI][Medline]
van Noord-Zaadstra, B.M., Looman, C.W.N., Alsbach, H. et al. (1991) Delaying childbearing: effect of age on fecundity and outcome of pregnancy. Br. Med. J., 302, 13611365.[ISI][Medline]
Snieder, H., MacGregor, A.J. and Spector, T.D. (1998) Genes control the cessation of a woman's reproductive life: a twin study of hysterectomy and age at menopause. J. Clin. Endocrinol. Metab., 83, 18751880.
Den Tonkelaar, I. (1997) Validity and reproducibility of self-reported age at menopause in women participating in the DOM project. Maturitas, 27, 117123.[ISI][Medline]
Treloar, S.A., Do, K-A. and Martin, N.G. (1998) Genetic influences on the age at menopause. Lancet, 352, 10841085.[ISI][Medline]
te Velde, E.R., Dorland, M. and Broekmans, F.J. (1998) Age at menopause as a marker of reproductive ageing. Maturitas, 30, 113121.[ISI][Medline]
de Waard, F., Collette, H.J.A., Rombach, J.J. et al. (1984) The DOM project for the early detection of breast cancer, Utrecht, The Netherlands. J. Chron. Dis., 37, 144.[ISI][Medline]
Wood, J.W. (1989) Fecundity and natural fertility in humans. In: Milligen, S.R. (ed.) Oxford reviews of reproductive biology, vol. 2. Oxford University Press, Oxford, pp. 61109.
Submitted on November 10, 2000; accepted on May 8, 2001.