1 Julius Center for Health Sciences and Primary Care, University Medical Center, PO Box 85500, 3508 GA Utrecht and 2 International Health Foundation, Europalaan 506, 3526 KS, Utrecht, The Netherlands
3 To whom correspondence should be addressed at: Julius Center for Health Sciences and Primary Care, Str. 6.119, University Medical Center Utrecht, PO Box 85060, 3508 AB Utrecht, The Netherlands. Email: s.elias{at}jc.azu.nl
![]() |
Abstract |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Key words: caloric restriction/famine/human/reproduction/subfertility
![]() |
Introduction |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Since this famine occurred in an otherwise well nourished population for a clearly demarcated period, it offers the opportunity to study its long-term effects and to pinpoint those effects to specific stages of human development. After birth, the hypothalamopituitarygonadal axis continues to mature until menarche and the first years of menstruation, which tend to be irregular and anovulatory (Harlow and Ephross, 1995). This development is characterized by changes in amplitude and pulsatility of hypothalamic GnRH (Yen et al., 1993
; Apter and Hermanson, 2002
). In order to establish regular ovulatory cycles, adequate hypothalamic and pituitary function is a prerequisite. Hypothalamic pulsatile secretion of GnRH results in a surge of LH being released from the pituitary. This surge of LH is necessary for ovulation (Filicori et al., 1993
). At reproductive ages, underfeeding can change the pattern of GnRH release towards a pre-pubertal state, resulting in menstrual disturbances or amenorrhoea (Warren, 1983
). We therefore hypothesize that since GnRH secretion is sensitive to caloric restriction, the latter may also interfere with the maturation of the hypothalamopituitarygonadal axis, leading to menstrual irregularities throughout life. Consequently, since irregular menses are less likely to be ovulatory (Harlow and Ephross, 1995
), the monthly chance of conception would be decreased, in severe cases to an extent of clinical subfertility. Moreover, menstrual disorders can be a reason for hysterectomy (van Hall, 1984
), resulting in the premature end of reproductive life. We used data on women exposed at young ages to the 19441945 Dutch famine, a cohort in which we previously found the famine to be associated with a decrease in age at natural menopause (Elias et al., 2003
), to explore its effect on these parameters.
![]() |
Subjects and methods |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Women who lived in the occupied parts of The Netherlands during the famine (n=7941) and could be classified according to their degree of famine exposure were eligible for analyses (n=6037). Individual famine exposure status was based on three separate questions on experiences of cold, hunger and weight loss during the famine period (ranked absent, moderate or severe). We considered women to be unexposed or severely exposed if they responded to at least two of these questions with absent or severe exposure, respectively. All remaining women were classified to moderate exposure (Elias et al., 2002).
In this cohort, we determined the effects of famine exposure in childhood on several indicators of subsequent reproductive capability. Ideally, one would measure the probability of conceiving by asking women about the time when they tried to become pregnant. However, this information was unavailable as the data were not obtained specifically for this study, and we therefore chose to use both the time between marriage and first childbirth and the time between first and second childbirth as an alternative. We also assessed the risk of a medical reason for having no or fewer children than wanted. Because a large proportion of the women were premenopausal at recruitment (72%), we used additional questionnaires to assess the risk of surgical menopause. These questionnaires were sent biennially until 1995, at which time 82% of the total cohort had complete follow-up (de Vries et al., 2001). Women reported the cause of menopause, but information on reason for surgery was not available.
Potential confounders
Several variables differed between famine exposure categories and could be related to fertility: socio-economic status was based on type of health insurance: public health insurance (lower status), civil servants insurance (intermediate status) or private insurance (higher status). Information on highest educational attainment (primary school, vocational training, high school or MSc/BSc) and cigarette smoking habits (ever/never smoker) was available from questionnaires.
Data analysis
From the study cohort (n=6037), we excluded seven women who had missing data on potential confounding variables. Thereafter, we selected appropriate subcohorts for the various analyses: (i) for time between marriage and first childbirth we selected all married women (n=5762) and excluded those who reported giving birth to their first child (n=780; 18%, 13% and 13% of the severely, moderately and unexposed women, respectively) or recalled entering menopause (n=6) before marriage or within 32 weeks afterwards (the former were excluded because date of marriage was unlikely to coincide with the start of attempting to get pregnant, and including these women could thus lead to a dilution of any effects due to increased misclassification); (ii) for time between first and second childbirth we selected all parous women (n=5395); (iii) the question regarding a medical reason for having no or fewer children than wanted was asked conditional on wanting to have more children or being nulliparous, which was indicated by 2051 women, from whom we excluded 675 women who answered I don't know; and (iv) the total study cohort could be used to assess the risk for surgical menopause.
To examine the effect of famine on interval between marriage and first childbirth, data were structured for time to event analyses with time coded in weeks from marriage onwards. Each woman contributed to the analysis the time until she either gave birth (event) or was not able to conceive a child due to entering the menopause (censoring event). Follow-up time ended at study recruitment for women who were at that time still premenopausal and nulliparous (censoring event). Data on interval between first and second childbirth were structured similarly. Hazard ratios (HRs) for first and second childbirth between famine exposure categories were estimated with Cox's proportional hazard analyses with unexposed as the reference category. Adjustments were made for age during the famine, socio-economic status at recruitment, highest educational attainment, cigarette smoking habits and age at baseline (age at marriage or age at first childbirth, respectively).
Data on surgical menopause were also analysed with Cox's proportional hazard models as well as with analysis of covariance in order to report adjusted means of menopausal age. For Cox's proportional hazard models, follow-up time in years was coded from famine exposure onwards. Surgical menopause was considered to be an event and all other women were censored at the last reported menstrual cycle. Famine exposure estimates were adjusted for age during the famine and socio-economic status at recruitment. We evaluated the proportionality of the hazards assumption for all variables by log minus log plots, which we found to be justified.
Whether famine exposure is related to the risk of a medical reason for having no or fewer children than wanted was analysed by means of logistic regression. First we assessed the crude odds ratios (ORs) of reporting such a medical reason between famine exposure categories with unexposed as the reference category. Adjustments were made for age during the famine and socio-economic status at recruitment. All models fitted well to the data according to the HosmerLemeshow goodness-of-fit test.
To explore the possibility of modification of the famine exposure effects by age during the famine, we included in the fully adjusted models interaction terms for famine exposure status and subgroups according to age at start of the famine (1 October 1944). These subgroups are based on stages in female development: 26 years (early childhood), 79 years (middle childhood) and 1012 years (later childhood) (Bogin, 1999).
Statistical analyses were performed with SPSS 11. Point estimates are given with corresponding 95% confidence intervals (CIs).
![]() |
Results |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
|
|
|
Previously, we have shown that the famine was associated with a decrease in subsequent age at natural menopause (Elias et al., 2003). Here we find that famine exposure was related to a higher occurrence of surgical menopause. From the total study cohort, 1369 (23%) women reported a surgical menopause. Mean age at surgical menopause, adjusted for age during the famine and socio-economic status, decreased with severity of famine exposure: 43.8 years (95% CI 43.444.1) in the unexposed group, 43.1 years (95% CI 42.643.5) in the moderately exposed group and 42.6 years (95% CI 41.743.6) in the severely exposed group (test of between-subjects effects P=0.022). Compared with the unexposed women, severely exposed women were at increased risk for surgical menopause (HR 1.53; 95% CI 1.271.84). Moderate famine exposure was also related to the risk of surgical menopause, but to a lesser extent (HR 1.11; 95% CI 0.991.24). These estimates did not change after adjustment for socio-economic status and age during the famine.
We did not find specific time-windows in female prepubertal development in which caloric restriction has a larger impact on the risk for surgical menopause or for having a medical reason for having no or fewer children than wanted (interaction terms non-significant).
![]() |
Discussion |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
As a result of the Dutch famine at the end of World War II, which struck an otherwise well-nourished population, we can now study the impact of short but severe starvation. We collected individual data on famine exposure for a large group of women. Although this famine exposure classification is based on recollection and thus prone to misclassification, we believe that it is more accurate than ecological comparisons. We found a good correlation between place of residence during the war and recalled famine exposure severity (Elias et al., 2002), reflecting differences in exposure between people living in rural versus urbanized areas at time of the famine (de Jong, 1981
). Also, the finding that the young were less exposed than the old is historically accurate (Burger et al., 1948
). From these circumstantial data, the individual famine exposure score seems valid. Misclassification owing to recall would most likely have underestimated the observed effects, since we believe that recall of famine exposure is not related to fertility experiences.
With regard to the analyses on time to childbirth, some issues on the validity of our findings need to be discussed. First of all, the participation rate in our study was 44%, which may have caused selection bias. Yet, in order to explain the observed results by selection bias, non-participation of women with a short waiting time should be higher among women who were severely exposed to the famine as compared with the unexposed. Or, vice versa, non-participation of women with a long waiting time should be higher among women who were unexposed to the famine as compared with the severely exposed. It is unlikely that such a bias due to selection has occurred.
Secondly, time to pregnancy is widely used as a measure for subfertility (Baird et al., 1986; Greenhall and Vessey, 1990
). As we were not able to directly determine time to pregnancy, we used the surrogate measures of time to childbirth after marriage or after birth of a first child. This should yield adequate estimation of fertility in cultures with a close link between marriage and procreation, as the Dutch culture was before the widespread use of oral contraceptives. In The Netherlands, oral contraceptives were introduced in 1962 (lynestrenol), mainly to treat abnormalities of the menstrual cycle. Within our analyses on time to first and second childbirth we did not control for oral contraceptive use, although we adjusted for socio-economic status, which did not change the results. Since we hypothesize that childhood caloric restriction leads to disturbances of the menstrual cycle and therefore to decreased chances of childbirth, it would be inappropriate to control for oral contraceptive use. However, analyses restricted to women who gave birth before 1962, 36% and 21% of the population on time to first and second childbirth, respectively, yielded similar results.
Thirdly, time to childbirth was taken as a proxy of female fertility. Of all cases of subfertility, 20% can be accounted for by a male factor only, and
27% by both male and female factors (Wong et al., 2000
). Male infertility may be related to lifestyle habits such as smoking and alcohol use (Wong et al., 2000
), and therefore to low socio-economic status. Since people tend to marry within their peer-group and women with a low socio-economic status were more often severely famine-exposed, a male cause of subfertility could have confounded the effect of famine exposure. Consequently, we adjusted for socio-economic status, which did not change our results. Still, this misclassification by male factors may have diluted the observed effects. Additionally, the measures of subfertility we used in our analyses are rather crude compared with a proper time-to-pregnancy analysis, where one would ideally ask women about the time they tried to become pregnant. Unfortunately, this information was unavailable in our cohort. It is very likely that the crudeness of our subfertility measures have led to considerable underestimation of contrasts and we therefore expect the true association between famine and subfertility to be of a larger magnitude. In order to minimize misclassification as much as possible, we excluded all women whose date of marriage could not possibly coincide with the date of trying to get pregnant. By making use of data on a comparable cohort of women we decided to exclude all women who gave birth within 32 weeks after marriage, representing the 5th percentile of gestation, and thereby limiting loss of valuable information and minimizing misclassification (Kloosterman, 1970
). Reanalysis of the data with all women included indeed yielded similar, albeit diluted, results (adjusted HR of first childbirth after marriage 0.93, 95% CI 0.880.99 for moderate exposure; and HR 0.90, 95% CI 0.811.01 for severe exposure).
Adjustment for potential confounders had remarkably little impact on the results (Table III), potentially partly because we adjusted rather crudely for cigarette smoking habits, which could have led to residual confounding. However, reanalysis of our data in only women who reported to have never smoked (57%) did not change the results. We could also not control for body mass index at conception, since these data were not available. Adjustment for body mass index at study recruitment did not change the results.
Data on age at menopause were obtained mostly in a prospective fashion until 82% complete follow-up (de Vries et al., 2001). The reproducibility and validity of recalled type and age of menopause is fairly high (Colditz et al., 1987
; den Tonkelaar, 1997
). Unfortunately, we did not have data on reason for surgery and therefore can only speculate on the underlying mechanisms that may have resulted in the higher risk of surgical menopause in the severely famine-exposed women. In line with our other findings, an effect on disorders of menstruation seems plausible, which was a medical indication in 23% of all, including postmenopausal, hysterectomies performed in The Netherlands around 1980 (van Hall, 1984
). Involvement of other indications, such as leiomyomata, can, however, not be ruled out. Similarly, we did not have information on the type of diagnosis when evaluating the risk of a medical reason for having no or fewer children than wanted. In a Western population, female subfertility is reported to be caused in 24% of cases by ovulatory disorders. Other causes include a male factor, as discussed above, and endometriosis (Evers, 2002
).
Literature on decreased female reproduction during periods of starvation is abundant (Bongaarts, 1980; Warren, 1983
; Harlow and Ephross, 1995
), in contrast to literature on childhood caloric restriction and subsequent fertility. Women exposed in utero to the 19441945 Dutch famine showed no differences in proportion childlessness, age at first delivery or family size (Lumey et al., 1995
). We also could not detect such effects. Data on Gainj women from highland Papua New Guinea show that the mean time to pregnancy is 1 month longer compared with a Western population. These women are frequently exposed to low-level undernutrition, which may play a role in these differences and is in line with our findings (Johnson et al., 1990
). One study specifically examined childhood nutrition and subsequent fertility milestones in Guatemala. Researchers randomly allocated two pairs of villages to low-energy or high-energy nutritional supplements. Women were exposed to these supplements already in utero and up to 3 years of age. High-energy supplementation led to increased chance of childbirth after first intercourse (HR 1.12; 95% CI 1.011.23) (Ramakrishnan et al., 1999
).
In conclusion, our findings support the view that childhood famine exposure impairs subsequent female reproductive function. These effects appear not to be related to specific time-windows in female prepubertal development, although the relationship between the famine and time to first childbirth after marriage seemed to be strongest in those aged 1012 years during the famine.
![]() |
Acknowledgements |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
![]() |
References |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Baird DD, Wilcox AJ and Weinberg CR (1986) Use of time to pregnancy to study environmental exposures. Am J Epidemiol 124, 470480.[Abstract]
Bogin B (1999) Patterns of Human Growth, 2nd edn. Cambridge University Press, Cambridge, UK.
Bongaarts J (1980) Does malnutrition affect fecundity? A summary of evidence. Science 208, 564569.[ISI][Medline]
Burger GCE, Sandstead HR and Drummond JC (1948) Malnutrition and Starvation in Western Netherlands, September 1944 to July 1945. Part I and II. General State Printing Office, The Hague, The Netherlands.
Colditz GA, Stampfer MJ, Willett WC, Stason WB, Rosner B, Hennekens CH and Speizer FE (1987) Reproducibility and validity of self-reported menopausal status in a prospective cohort study. Am J Epidemiol 126, 319325.[Abstract]
de Jong L (1981) Het Koninkrijk der Nederlanden in de Tweede Wereldoorlog (The Kingdom of The Netherlands in the Second World War). In Dutch. General State Printing Office, The Hague, The Netherlands. Part 106, 153266.
de Vries E, den Tonkelaar I, van Noord PA, van der Schouw YT, te Velde ER and Peeters PH (2001) Oral contraceptive use in relation to age at menopause in the DOM cohort. Hum Reprod 16, 16571662.
de Waard F, Collette HJ, Rombach JJ, Baanders-van Halewijn EA and Honing C (1984) The DOM project for the early detection of breast cancer, Utrecht, The Netherlands. J Chronic Dis 37, 144.[CrossRef][ISI][Medline]
den Tonkelaar I (1997) Validity and reproducibility of self-reported age at menopause in women participating in the DOM-project. Maturitas 27, 117123.[CrossRef][ISI][Medline]
Elias SG, van Noord PA, Peeters PH, den Tonkelaar I and Grobbee DE (2002) The 19441945 Dutch famine and age at natural menopausethe value and validity of individual exposure assessment. IARC Sci Publ 156, 311313.[Medline]
Elias SG, van Noord PA, Peeters PH, den Tonkelaar I and Grobbee DE (2003) Caloric restriction reduces age at menopause: the effect of the 19441945 Dutch famine. Menopause 10, 399405.[CrossRef][ISI][Medline]
Evers JL (2002) Female subfertility. Lancet 360, 151159.[CrossRef][ISI][Medline]
Filicori M, Cognigni G, Dellai P, Arnone R, Sambataro M, Falbo A, Pecorari R, Carbone F and Meriggiola MC (1993) Role of gonadotrophin releasing hormone secretory dynamics in the control of the human menstrual cycle. Hum Reprod 8 (Suppl 2), 6265.[ISI][Medline]
Greenhall E and Vessey M (1990) The prevalence of subfertility: a review of the current confusion and a report of two new studies. Fertil Steril 54, 978983.[ISI][Medline]
Harlow SD and Ephross SA (1995) Epidemiology of menstruation and its relevance to women's health. Epidemiol Rev 17, 265286.[ISI][Medline]
Johnson PL, Wood JW and Weinstein M (1990) Female fecundity in highland Papua New Guinea. Soc Biol 37, 2643.[ISI][Medline]
Kloosterman GJ (1970) On intrauterine growth: the significance of prenatal care. Int J Gynaecol Obstet 8, 895912.
Lumey LH, Stein AD and Ravelli AC (1995) Timing of prenatal starvation in women and birth weight in their first and second born offspring: the Dutch Famine Birth Cohort study. Eur J Obstet Gynecol Reprod Biol 61, 2330.[CrossRef][ISI][Medline]
Ramakrishnan U, Barnhart H, Schroeder DG, Stein AD and Martorell R (1999) Early childhood nutrition, education and fertility milestones in Guatemala. J Nutr 129, 21962202.
Stein Z and Susser M (1975) Fertility, fecundity, famine: food rations in the dutch famine 1944/5 have a causal relation to fertility, and probably to fecundity. Hum Biol 47, 131154.[ISI][Medline]
van Hall EV (1984) Hysterectomy in the Netherlands. In van Herendael H, van Herendael B, Riphagen FE, Goessens L, and van der Pas H (eds) The ClimactericAn Update. MTP Press Limited, Lancaster, UK, pp. 265270.
Warren MP (1983) Effects of undernutrition on reproductive function in the human. Endocr Rev 4, 363377.[ISI][Medline]
Wong WY, Thomas CM, Merkus JM, Zielhuis GA and Steegers-Theunissen RP (2000) Male factor subfertility: possible causes and the impact of nutritional factors. Fertil Steril 73, 435442.[CrossRef][ISI][Medline]
Yen SS, Apter D, Butzow T and Laughlin GA (1993) Gonadotrophin releasing hormone pulse generator activity before and during sexual maturation in girls: new insights. Hum Reprod 8 (Suppl 2), 6671.[ISI][Medline]
Submitted on August 4, 2003; resubmitted on November 15, 2004; accepted on April 25, 2005.