Assessment of reproductive ageing patterns by hormonal and ultrasonographic ovarian reserve tests

E. Tufan1, K. Elter1,2 and F. Durmusoglu1

1 Division of Reproductive Endocrinology and Infertility, Department of Obstetrics and Gynecology, Marmara University School of Medicine, Istanbul, Turkey

2 To whom correspondence should be addressed at: Kuyubasi S.Fenik Apt. No: 20/17, 34724 Feneryolu, Istanbul, Turkey. Email: korayelter{at}marmara.edu.tr


    Abstract
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 References
 
BACKGROUND: Tests to assess the change in ovarian reserve (OR) with age have been analysed only in monophasic or biphasic linear patterns. Our aim was to analyse an optimum curve that might define the relationship between different OR tests and age. METHODS: A total of 81 regularly menstruating women without a history of infertility were included in this prospective study. On cycle day 3, antral follicle (AF) counts, ovarian volume (OV), and serum FSH and estradiol levels were determined. Curve estimation was performed to determine the optimal relationship between age and OR tests. Optimum curves were also compared with monophasic and biphasic linear relationships. RESULTS: A quadratic model (y=axx2+bxx+c) had the highest coefficient of determination for the AF count, OV and serum FSH level. The predictive power of this quadratic model was comparable with biphasic linear models for the OV and serum FSH level, but was better than that of the AF count. CONCLUSIONS: The pattern of reproductive ageing as assessed by hormonal and ultrasonographic OR tests does not appear to show an abrupt change at a certain age, but follows a continuously increasing rate of decline in the third decade of life. The changes in serum FSH levels and ultrasonographic OR tests follow a quadratic model in regularly menstruating women.

Key words: age/antral follicle/FSH/ovarian reserve/ovarian volume


    Introduction
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 References
 
It is known that the decline in fecundity with female age occurs long before menopause (O'Connor et al., 1998Go). Fertility is remarkably reduced with increasing age of women in both spontaneous conceptions (Tietze, 1957Go) and assisted reproductive methods (Templeton et al., 1996Go). The decrease in fertility with ageing is due to decreases in both oocyte quality and quantity. Direct data on oocyte quality or number of follicles present in the ovaries in in vivo situations are limited for obvious reasons. Parameters such as early follicular serum FSH and estradiol (E2) levels and various stimulation tests, as well as ultrasound-based measurements, i.e. antral follicle (AF) count and ovarian volume (OV), have been used to predict advanced ageing of the ovary and the probability of treatment success in infertility populations. Therefore, the so-called ovarian reserve (OR) tests are frequently used to evaluate the probability of pregnancy in infertile women (Bukman and Heineman, 2001Go).

Many women in developed countries delay childbearing. Since age is not a good predictor of reproductive potential (Bukman and Heineman, 2001Go), use of OR tests for an accurate assessment of the reproductive potential in these older women may help clinicians during counselling. The determination of OR may help clinicians in making the decision regarding prophylactic oophorectomy in selected women (Barnes-Kedar and Plon, 2002Go). Younger women can be screened for early ovarian ageing, especially those in high-risk groups, as has been proposed recently for the prevention of subfertility (Nikolaou and Templeton, 2003Go). Detection of early ovarian ageing is not only important for infertility, but is also of interest in relation to general somatic ageing (Kirkwood, 1998Go). Therefore, the establishment of normal curves for OR tests is important. A normal curve for OV may also help to define abnormal values at different ages in screening for early ovarian cancer (DePriest et al., 1997Go).

It has been suggested that the normal rate of oocyte depletion follows a biphasic pattern, accelerating below a number of ~25 000 at a mean age of 37–38 years based on histological analyses of ovaries (Faddy et al., 1992Go). This suggested midlife acceleration in oocyte loss is of great interest. Although the OR tests have been studied in infertile women, there are few studies addressing the effects of age on OR tests in women without infertility (Ruess et al., 1996Go; Scheffer et al., 1999Go; Ng et al., 2003Go). Furthermore, these studies have not analysed an optimum curve for the relationship between OR and age. Some authors have suggested a monophasic linear relationship, and others have suggested a biphasic pattern without analysing a comparison between different models (Ruess et al., 1996Go; Ng et al., 2003Go).

Therefore, in this prospective study, we aimed to find an optimum curve that might define the relationship between different OR tests and age, by comparing the predictive power of monophasic and biphasic regression lines with different cut-off values of age.


    Materials and methods
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 References
 
Subjects
Eighty-one women presenting to our out-patient clinic for acute minor gynaecological problems such as vaginitis were included in this prospective study. To qualify for participation, subjects were required to (i) have regular 21–35 day cycles; (ii) be on no medications or exogenous hormones for 6 months before participation; (iii) have no evidence of endocrinological disease; and (iv) have no history of infertility. Seventy-three women had proven fertility, and the remaining eight (9.9%), all aged ≤26 years, were nulliparous with no desire for fertility. Each woman had both of her ovaries, and women with previous ovarian surgery, endometriomas or follicles measuring at least 10 mm on baseline ultrasound were excluded from the study. The study was approved by the institutional review board at Marmara University, and written informed consent was obtained from each subject.

On cycle day 3 of spontaneous bleeding, baseline vaginal ultrasonography was performed for bilateral AF count and OV determinations, and venous blood samples were withdrawn for serum FSH and E2 determinations.

Assays and ultrasonographic measurements
All blood samples were centrifuged within 2 h after withdrawal and stored at –20°C until assayed. Serum FSH and E2 concentrations were determined using the Immulite immunoassay system (Diagnostic Products Corporation, Los Angeles, CA). This assay is standardized to the World Health Organization Second International Reference Preparation 78/549. The inter- and intra-assay coefficients of variation were 6.6 and 5.4% for FSH, and 5.4 and 4.4% for E2, respectively.

Transvaginal ultrasound was performed by the same physician (E.T.), using a GE Logiq 200 Pro (GE Medical Systems, Milwaukee, WI) with a 6.5 MHz vaginal transducer. Round or oval echo-free structures were regarded as follicles, and all ovarian follicles measuring 2–10 mm on both ovaries were counted on cycle day 3. Ovarian volume subsequently was computed using the ellipsoid formula: OV = D1 x D2 x D3x{pi}/6, where D1, D2 and D3 are the maximal perpendicular diameters of each ovary. The volumes and AF counts of both ovaries were added, and the total number of follicles and total OV per patient were used for calculations.

Statistical analysis
Correlations between OR tests and age were analysed by using the Pearson's correlation test. For the significantly correlated parameters, curve estimation was performed to determine the optimal relationship between age and OR tests. For the curve estimation procedure, curve estimation regression statistics were performed for 11 different regression models, including linear, logarithmic, inverse, quadratic, cubic, power, compound, S-curve, logistic, growth, and exponential models. The model with the highest coefficient of determination (r2) was accepted as the optimal model for the relationship. The coefficient of determination is the amount of the scatter in one variable that can be explained by another. In the present study, it is the rate (e.g. 35% when r2=0.35) of variation in the values of the OR test which can be accounted for by knowing the age through using the relevant model.

Standard deviations for the OR test in different age groups, i.e. 21–25, 26–30, 31–35, 36–40 and 41–45 years, were determined. The predictive power of the model with the highest r2 value was analysed by determining the number of observed values in ‘the mean as determined by the regression equation±2 SDs’ range. Any observed value below or above the±2 SDs of the predicted value was accepted as the ‘outlying value’. The number of outlying values was determined and the ratio of outlying values to the total number of values (n=81) was used for the statistical analysis.

Linear regression equations were also determined. The regression coefficient {beta} and the y-intercept were determined for the regression formula: OR test = {beta} x age + y-intercept. The number of observed values below and above the±2 SDs of the predicted values was determined. Following determination of the monophasic line for the significantly correlated OR tests, the predictive power of biphasic lines was analysed. Each of the cut-offs between 30 and 38 for the age was analysed separately, and the number and ratio of outlying values to the total number of values (n=81) were determined for each biphasic line. The ratio of outlying values was compared by using the {chi}2 or Fisher's exact tests, where appropriate. SPSS, Release 11.5 (SPSS, Inc, Chicago, IL) was used for the statistical analysis, and a P-value of <0.05 was considered significant.


    Results
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 References
 
Means (±SD) for the age and body mass index (BMI) among women in the present study were 34.4 (±6.4) years (range 21–43) and 24.3 (±4.5) kg/m2. When the correlations between the OR tests and age were evaluated, OV, AF count and serum FSH level were found to be significantly correlated with age (Table I).


View this table:
[in this window]
[in a new window]
 
Table I. Correlations between age and ovarian reserve tests

 
The changes in OV and AF count with age are shown in Figure 1. The quadratic model (y = axx2 +bxx+c) had the highest coefficient of determination for both of these ultrasonographic OR tests (Table II). The regression equations for this quadratic model were as follows: AF count=–0.0165 x age2+0.7643 x age + 0.2793, and OV=–0.0196 x age2+1.0193 x age–1.9988 (Figure 1). Standard deviations in different age groups are shown in Table III. ‘The mean as determined by the regression equation±2 SD’ range for the AF count included, and therefore predicted, 97.5% (79 out of 81) of the observed values. ‘The mean as determined by the regression equation±2 SD’ range for the OV included, and therefore predicted, 96.3% (78 out of 81) of the observed values.



View larger version (24K):
[in this window]
[in a new window]
 
Figure 1. The change in OV and AF count with age. Dashed lines indicate the regression equations for the ovarian volume and AF count [AF count=–0.0165 x age2+0.7643 x age + 0.2793; OV (ml)=–0.0196 x age2+1.0193 x age–1.9988].

 

View this table:
[in this window]
[in a new window]
 
Table II. Coefficients of determination (r2) of different models for the relationship between age and OR tests

 

View this table:
[in this window]
[in a new window]
 
Table III. Standard deviations of OR tests in different age groups

 
The change in serum FSH level with age is shown in Figure 2. An exponential model had the highest coefficient of determination (Table II). Assessment of the 11 models after this logarithmic transformation revealed that the quadratic model improved the r2 value from 33 to 37%. The regression equation for this combined model was as follows: serum FSH level (mIU/ml) = e (0.0006xage2–0.0075xage+1.6793) (Figure 2). Standard deviations for the serum FSH level in different age groups are shown in Table III. ‘The mean as predicted by the regression equation±2 SD’ range for the serum FSH level included, and therefore predicted, 96.3% (78 out of 81) of the observed values.



View larger version (15K):
[in this window]
[in a new window]
 
Figure 2. The change in serum FSH level with age. The dashed line indicates the regression equation for the serum FSH level [serum FSH level (mIU/ml) = e(0.0006 x age2–0.0075 x age + 1.6793)].

 
The amount of decrease in the AF count and OV per year increased linearly with age, but it was <1 (n for the AF count and ml for the OV) per year for all ages between 21 and 43 years for both of these ultrasonographic variables (Figure 3). The amount of increase in the serum FSH level per year also increased with age (Figure 4). The percentage change in the serum FSH level per year increased linearly with age and it was between 2 and 5%/year for ages between 21 and 43 years (Figure 4).



View larger version (11K):
[in this window]
[in a new window]
 
Figure 3. Decline rate in the OV (ml per year) and AF count (n per year) with age. The dashed line indicates the decline rate for the AF count.

 


View larger version (11K):
[in this window]
[in a new window]
 
Figure 4. Increase in the serum FSH level (% per year).

 
The predictive power of monophasic trend lines and biphasic trend lines created by placing the cut-off point at various ages from 30 to 38 years is shown in Tables IV, V, and VI. The lowest number of outlying values was observed with biphasic lines with the cut-off values of 30 and 31 in both of the ultrasonographic OR tests. The rate of outlying values was comparable between the quadratic and biphasic linear models for OV and serum FSH level (three out of 81 versus three out of 81 in both of these variables). The rate of outlying values was higher in the biphasic linear model when compared with the quadratic model for the AF count (three out of 81 versus two out of 81; P=0.001).


View this table:
[in this window]
[in a new window]
 
Table IV. Regression equations for monophasic and different biphasic linear relationships between the AF count and age

 

View this table:
[in this window]
[in a new window]
 
Table V. Regression equations for monophasic and different biphasic linear relationships between the ovarian volume and age

 

View this table:
[in this window]
[in a new window]
 
Table VI. Regression equations for monophasic and different biphasic linear relationships between the serum FSH level and age

 

    Discussion
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 References
 
The present study showed that markers of OR change in a quadratic model with increasing age. This model is optimal after logarithmic transformation for the serum FSH level. The present study also showed that the rate of change increases with age. To our knowledge, a quadratic model has not been used previously as a means of testing the relationship between age and OR.

It has been suggested that oocyte depletion accelerates toward the onset of menopause, and this acceleration is characterized by a bend in the scatter of points on a log-linear plot of follicle number by age (Richardson et al., 1987Go; Faddy et al., 1992Go; Gougeon et al., 1994Go). This abrupt change in the rate of follicular loss has been suggested to occur at the age of 37.5 years (Faddy et al., 1992Go). Thus, follicular atresia has been interpreted as biphasic. A linear relationship after logarithmic transformation indicates that the rate of percentage change is constant. However, what happens at some critical age to make the rate of follicular depletion increase abruptly? Recently, this biphasic pattern has been questioned by Leidy et al. (1998)Go. They have tested four models with the use of data drawn from published studies and histological analyses of ovaries (Leidy et al., 1998Go). These models did not include a quadratic model. Leidy et al. (1998)Go concluded that follicular atresia was not linear on either original measurement or log-linear scales. They also suggested that the data on follicular atresia did not support the notion that an abrupt change in the exponential rate of decline exists, and concluded that a biphasic pattern was most likely to be an artefact (Leidy et al., 1998Go). It is more acceptable to hypothesize the existence of an increase in the decline rate of follicles through the years instead of an abrupt change at a certain age.

Faddy and Gosden also revised their original model, and suggested a gradual change in follicle depletion (Faddy and Gosden, 1996Go, 2000Go; Faddy et al., 1992Go). In their model, the follicles were classified into three stages of growth: stage I, primordial follicles; stage II, early growing forms; and stage III, more advanced stages. Following re-analysis of ovaries in their initial study (Faddy et al., 1992Go), they suggested a model that shows a gradual increase in follicle depletion based on either growth of follicles between these three compartments or atresia. Their model suggested that a proportional number of primordial follicles enter the growth phase. This model suggests that the AF count can reliably reflect the primordial follicle pool, which is the actual OR. This is consistent with the results of clinical studies, which showed that AF count is a reliable OR test. Since their model was based on histological findings and a three compartment model, we avoided detailed comparison between their model and that in the present study, which was based on OR tests. However, the similarity of the graph in the present study to that in their model supports the proportional relationship between the primordial follicle cohort and growing follicles (Faddy and Gosden, 2000Go).

Clinical experience and demographic data obtained from natural populations suggest that the period of optimal fertility only lasts until age 30–31 and decreases thereafter (van Noord-Zaadstra et al., 1991Go; te Velde et al., 1998Go). The suggested model for the AF count and OV in the present study indicates almost a constant OR between ages 21 and 30. The OV decreases by 1.7% and the AF count decreases by 7.7% during that time. Thereafter, although the amount of decrease per year is constant according to the model, the percentage change increases with age. Therefore, the apparent acceleration in decline is merely due to a constant depletion and a decreasing denominator; percentage change = amount of difference (which is constant)/previous amount (which decreases with age). The percentage change in the serum FSH level increases with age through all ages between 21 and 43. However, this increase is slight for ages between 21 and 30, i.e. 23% in 9 years.

These models may also help to analyse the mechanism for ovarian ageing. It has been suggested previously that the accelerated increase of oocyte depletion might be due to a rising serum FSH level (Gosden and Faddy, 1994Go; Gougeon, 1996Go). In the present study, we observed that the serum FSH level increases slightly at the third decade of age, but AF count and OV appear to decline slightly in the same period. The increase in serum FSH level was slightly faster than the decline in AF count. In the model, serum FSH level increased 60% between ages 25 and 40. The AF count decreased 51% during the same period.. These findings suggest that increasing serum FSH level may initiate the accelerated loss. However, it should be mentioned that the evidence that the FSH receptor is not expressed in follicles until they have reached a certain stage argues against this FSH-dependent hypothesis (Oktay et al., 1997Go). The underlying mechanism for the accelerated loss of oocytes remains unclear. However, the results of egg donation programmes suggest that the oocyte itself is responsible for the declining fertility in humans. It appears that ageing has some negative effects on the quiescent oocytes. This may be a result of either damage to DNA, decreasing density of oocytes or changes in growth factor tone (Faddy and Gosden, 2000Go).

In the present study, the quadratic model for the AF count was superior to any biphasic linear relationship for predicting the observed values. This model indicates a slower increase in the decline rate instead of an abrupt change. This relationship between age and AF count is similar to fertility change by age in different populations (Menken et al., 1986Go). Compared with women aged 20–24, fecundity is reduced on average by 6% for women aged 25–29, 14% for those aged 30–34, and 31% for women aged 35–39 (Menken et al., 1986Go). The corresponding values were 3, 14 and 34%, respectively, when the model in the present study for the AF count was used.

The model for the AF count in the present study means that the existence of an FSH-sensitive cohort will have ended at the age of 46.7 years. This seems biologically unlikely in view of the fact that the vast majority of women will still have cycles at this age, albeit irregular. Although this value is consistent with mean age at menopause in Turkey (47 years) (Neslihan Carda et al., 1998Go), it should also be mentioned that models in the present study are based on the data of women between the ages of 20 and 43 years, and therefore extrapolating these results to older ages may not be correct.

The AF count had the highest correlation with age among the OR tests in the present study. Scheffer et al. (2003)Go also showed that the number of AFs correlated much better with the age of the women evaluated in their study than other presumed basal markers for reproductive age, including FSH, inhibin B, E2 and OV. Therefore, results in the present study also support the hypothesis that the AF count is superior to other static measures of reproductive ageing.

Although nomograms for OV in post-menopausal women have been studied (Goswamy et al., 1983Go, 1988Go), data in premenopausal women are limited (van Nagell et al., 1990Go). Surprisingly, there are reports that used a constant cut-off value for the OV while studying the effectiveness of transvaginal sonography in screening for ovarian cancer in premenopausal women (DePriest et al., 1997Go). The present study showed that OV decreases with age, especially after the age of 30 years. Therefore, using a constant cut-off value for the OV in premenopausal women may not be appropriate for screening purposes.

In conclusion, the pattern of reproductive ageing as evidenced by hormonal and ultrasonographic OR tests does not show an abrupt change at a certain age, but follows a continuously increasing rate of decline in the third decade of life. The change in serum FSH level and ultrasonographic OR tests with age follows a quadratic model in regularly menstruating premenopausal women.


    References
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 References
 
Barnes-Kedar IM and Plon SE (2002) Counseling the at risk patient in the BRCA1 and BRCA2 era. Obstet Gynecol Clin North Am 29, 341–366, vii.[ISI][Medline]

Bukman A and Heineman MJ (2001) Ovarian reserve testing and the use of prognostic models in patients with subfertility. Hum Reprod Update 7, 581–590.[Abstract/Free Full Text]

DePriest PD, Gallion HH, Pavlik EJ, Kryscio RJ and van Nagell JR, Jr (1997) Transvaginal sonography as a screening method for the detection of early ovarian cancer. Gynecol Oncol 65, 408–414.[CrossRef][ISI][Medline]

Faddy MJ and Gosden RG (1996) A model conforming the decline in follicle numbers to the age of menopause in women. Hum Reprod 11, 1484–1486.[Abstract/Free Full Text]

Faddy MJ and Gosden RG (2000) Mathematical models for follicle development and depletion. In te Velde ER, Pearson PL, and Broekmans FJ (eds), Female Reproductive Ageing. Parthenon Publishing Group, Lancs, UK, pp. 71–78.

Faddy MJ, Gosden RG, Gougeon A, Richardson SJ and Nelson JF (1992) Accelerated disappearance of ovarian follicles in mid-life: implications for forecasting menopause. Hum Reprod 7, 1342–1346.[Abstract]

Gosden RG and Faddy MJ (1994) Ovarian aging, follicular depletion, and steroidogenesis. Exp Gerontol 29, 265–274.[CrossRef][ISI][Medline]

Goswamy RK, Campbell S and Whitehead MI (1983) Establishment of normal ranges for ovarian volumes and identification of enlarged ovaries by real-time mechanical sector sonar in postmenopausal women. Ultrasound Med Biol Suppl 2, 615–619.

Goswamy RK, Campbell S, Royston JP, Bhan V, Battersby RH, Hall VJ, Whitehead MI and Collins WP (1988) Ovarian size in postmenopausal women. Br J Obstet Gynaecol 95, 795–801.[ISI][Medline]

Gougeon A (1996) Regulation of ovarian follicular development in primates: facts and hypotheses. Endocr Rev 17, 121–155.[ISI][Medline]

Gougeon A, Ecochard R and Thalabard JC (1994) Age-related changes of the population of human ovarian follicles: increase in the disappearance rate of non-growing and early-growing follicles in aging women. Biol Reprod 50, 653–663.[Abstract]

Kirkwood TB (1998) Ovarian ageing and the general biology of senescence. Maturitas 30, 105–111.[CrossRef][ISI][Medline]

Leidy LE, Godfrey LR and Sutherland MR (1998) Is follicular atresia biphasic? Fertil Steril 70, 851–859.[CrossRef][ISI][Medline]

Menken J, Trussell J and Larsen U (1986) Age and infertility. Science 233, 1389–1394.[ISI][Medline]

Neslihan Carda S, Bilge SA, Ozturk TN, Oya G, Ece O and Hamiyet B (1998) The menopausal age, related factors and climacteric symptoms in Turkish women. Maturitas 30, 37–40.[CrossRef][ISI][Medline]

Ng EH, Yeung WS, Fong DY and Ho PC (2003) Effects of age on hormonal and ultrasound markers of ovarian reserve in Chinese women with proven fertility. Hum Reprod 18, 2169–2174.[Abstract/Free Full Text]

Nikolaou D and Templeton A (2003) Early ovarian ageing: a hypothesis. Detection and clinical relevance. Hum Reprod 18, 1137–1139.[Abstract/Free Full Text]

O'Connor KA, Holman DJ and Wood JW (1998) Declining fecundity and ovarian ageing in natural fertility populations. Maturitas 30, 127–136.[CrossRef][ISI][Medline]

Oktay K, Briggs D and Gosden RG (1997) Ontogeny of follicle-stimulating hormone receptor gene expression in isolated human ovarian follicles. J Clin Endocrinol Metab 82, 3748–3751.[Abstract/Free Full Text]

Richardson SJ, Senikas V and Nelson JF (1987) Follicular depletion during the menopausal transition: evidence for accelerated loss and ultimate exhaustion. J Clin Endocrinol Metab 65, 1231–1237.[Abstract]

Ruess ML, Kline J, Santos R, Levin B and Timor-Tritsch I (1996) Age and the ovarian follicle pool assessed with transvaginal ultrasonography. Am J Obstet Gynecol 174, 624–627.[ISI][Medline]

Scheffer GJ, Broekmans FJ, Dorland M, Habbema JD, Looman CW and te Velde ER (1999) Antral follicle counts by transvaginal ultrasonography are related to age in women with proven natural fertility. Fertil Steril 72, 845–851.[CrossRef][ISI][Medline]

Scheffer GJ, Broekmans FJ, Looman CW, Blankenstein M, Fauser BC, teJong FH and teVelde ER (2003) The number of antral follicles in normal women with proven fertility is the best reflection of reproductive age. Hum Reprod 18, 700–706.[Abstract/Free Full Text]

Templeton A, Morris JK and Parslow W (1996) Factors that affect outcome of in-vitro fertilisation treatment. Lancet 348, 1402–1406.[CrossRef][ISI][Medline]

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

Tietze C (1957) Reproductive span and rate of reproduction among Hutterite women. Fertil Steril 8, 89–97.[ISI]

van Nagell JR, Jr, Higgins RV, Donaldson ES, Gallion HH, Powell DE, Pavlik EJ, Woods CH and Thompson EA (1990) Transvaginal sonography as a screening method for ovarian cancer. A report of the first 1000 cases screened. Cancer 65, 573–577.[ISI][Medline]

van Noord-Zaadstra BM, Looman CW, Alsbach H, Habbema JD, te Velde ER and Karbaat J (1991) Delaying childbearing: effect of age on fecundity and outcome of pregnancy. Br Med J 302, 1361–1365.[ISI][Medline]

Submitted on March 26, 2004; accepted on July 13, 2004.