Birth weight and ovulatory dysfunction

Authors


Abstract

Objective

To explore the association between birthweight and ovulatory dysfunction in adulthood.

Design

Case–control study.

Setting

Northeast of Scotland University Hospital, hosting the regional fertility centre and maternity unit.

Population

A total of 18 846 mother–daughter record pairs from the Aberdeen Fertility Centre Data Set and the Aberdeen Maternity and Neonatal Databank (AMND). Cases were the daughters with ovulatory dysfunction attending the Aberdeen Fertility Centre between 1992 and 2007, Control group 1 included the daughters attending the fertility centre with confirmed ovulation, and Control group 2 included all women naturally fertile who gave birth in Aberdeen during the same period.

Methods

The electronic maternity records of the mothers of women in the three groups were retrieved from AMND and compared.

Main outcome measures

Daughters' birthweight and standardised birthweight, characteristics of mothers and daughters at delivery and current daughters' characteristics.

Results

Cases, Control group 1 and Control group 2 included 466, 548 and 17 832 daughters, respectively. The mean birthweight (standard deviation) in grams was comparable between Cases 3203 (522), Control group 1, 3235 (482) = 0.30, and Control group 2, 3226 (495) P = 0.31. The proportions of daughters born small for gestational age, large for gestational age, or preterm were comparable between the Cases group and each Control group, as was the mode of delivery and Apgar scores at 1 and 5 minutes. The age at delivery, body mass index, social class or pregnancy complications were comparable in the mothers of the Cases and each Control group.

Conclusions

Ovulatory dysfunction does not appear to be related to birthweight or perinatal events.

Introduction

Several major diseases of adult life, including coronary heart disease, hypertension and type 2 diabetes, have their origins in impaired intrauterine growth and development. These conditions may be consequences of ‘programming’, whereby a stimulus or insult at a critical sensitive period of early life has permanent effects on structure, physiology and metabolism.[1-4] The so-called ‘fetal origins hypothesis’ or ‘Barker hypothesis’ proposes that alterations in fetal nutrition and endocrine status result in developmental adaptations that predispose individuals to cardiovascular, metabolic and endocrine disease in adult life.[1, 3] This ‘fetal reprogramming’ may be mediated by epigenetic modifications (defined as changes in the gene expression without changes in the DNA sequence) that are involved in the development and inheritance of many adult diseases with suggested fetal origins.

Ovulation is a complicated physiological process, involving many endocrine and paracrine regulatory and feedback mechanisms. Anovulation is a common disorder of female reproductive function, and is associated with heritable risk factors like obesity and insulin resistance. Ovulatory disorders are estimated to cause 21% of female infertility.[5] Group 2 ovulatory dysfunction according to the World Health Organization classification[6] accounts for 85% of ovulatory disorders and predominantly involves women with polycystic ovaries. It was demonstrated in a rat model of intrauterine growth restriction that ligation of the uterine arteries in female pups leads to a delay in the onset of puberty and anovulation.[7] Few studies have investigated fetal antecedents of female subfertility in humans and existing studies on the influence of birthweight on ovulation have produced contradictory results.

Li and Huang[8] have proposed the presence of ‘an epigenetic abnormality underlying the fetal origins of PCOS’, where disturbed reprogramming of the fetal reproductive tissue in utero results in postnatal polycystic ovarian syndrome (PCOS) phenotype in adulthood. It has been reported that lower birthweight in girls is associated with premature pubarche and subsequent ovarian hyperandrogenism.[9-13] Similarly, low birthweight[14] or being small for gestational age[15] were found to be associated with anovulation and PCOS. On the other hand, heavier newborns have also been reported to be at risk of developing anovulation associated with PCOS.[16, 17] Two further studies could not confirm a relationship between birthweight and anovulation and PCOS,[18] or self-reported symptoms suggestive of anovulation or PCOS.[19]

The results of these studies reveal the lack of consistency in the published literature. The outcome of interest in some of these studies was anovulation, whereas in others it was PCOS. The majority of these studies had relatively small sample sizes.

Considering the relatively high incidence of ovulatory dysfunction, and the impact on resources used to investigate and treat these women, it is important to understand the factors underlying anovulation and to distinguish between nonmodifiable and modifiable risk factors.

Aberdeen provides a unique opportunity to link intergenerational data on mothers' delivery outcomes, and infertility and subsequent ovulatory dysfunction in their daughters. The Aberdeen Maternity and Neonatal Databank (AMND) records and stores information on all pregnancy events occurring in Aberdeen city and district from 1950 onwards and currently holds records of approximately 200 000 pregnancies. As the only secondary, as well as tertiary, referral centre for fertility investigations and treatment, and as the only maternity unit for pregnancies complicated by growth restriction in the Grampian region, it allows population-based studies on a potential link between birthweight and ovulatory dysfunction. Since 1992, the Aberdeen Fertility Centre Database (AFCD) has recorded clinical data on all couples in Aberdeen city and district referred from primary care with a diagnosis of infertility. The juxtaposition of these two data sets provides a unique opportunity to perform data linkage studies on a defined population.

We designed this study to explore a possible link between birthweight and ovulatory dysfunction in adulthood.

Methods

Ethical approval

The study was approved by the local ethics committee, the Caldicott Guardians of the AMND and the AFCD. A case–control study was conducted.

Cases

Cases in the case–control study were women (daughters) with ovulatory dysfunction identified from the AFCD on the basis of clinical history (irregular periods) or a midluteal progesterone level below the ovulatory threshold of 30 nmol/l. Irregular periods were defined in the presence of cycle length <21 days or >35 days, or a cycle-to-cycle variation of ≥5 days. Women with ovarian failure or with hypogonadotrophic hypogonadism were identified from the database and were excluded from the Cases group. Women with PCOS were included in the Cases. The diagnostic criteria for PCOS were changed following the ESHRE/ASRM conference in Rotterdam in 2003, ‘the Rotterdam Criteria’.[20] For the sake of consistency, we chose to include a single ovulatory dysfunction population rather than divide them into two separate PCOS coded groups in the database based on alternative definitions before and after the introduction of the new criteria.

Control group 1 was selected from the AFCD and included women (daughters) who attended the fertility clinic with regular cycles and confirmed ovulation (serum midluteal progesterone level ≥30 nmol/l), but with tubal factor infertility or male factor infertility. We excluded women with unexplained infertility from this group.

Control group 2 included all women who delivered in Aberdeen between 1992 and 2007 and had no recorded contact with the Aberdeen Fertility Clinic. These women were identified from the AMND.

Two control groups were selected because there did not appear to be a single ideal comparison population. Control group 1 were known to have confirmed ovulation but still belonged to an ‘infertile’ population, whereas some members of Control group 2 could have had a period of infertility before spontaneous conception or conception following treatment outside Aberdeen.

Exclusion criteria

Women (daughters) with a twin sibling were excluded from all three groups. Women with unexplained infertility were excluded from Control group 1. Women who attended the Fertility Centre before achieving pregnancy were excluded from Control group 2. Women with a diagnosis of ovarian failure (hypergonadotrophic hypogonadism) and women with hypogonadotrophic hypogonadism were excluded from the Cases group.

We hypothesised that women with ovulatory dysfunction were born with a lower birthweight or standardised birthweight when compared with controls with normal ovulation or normal fertility.

Perinatal events for the three groups of women (daughters) now in their adulthood were compared by interrogating their mothers' maternity records. Cases group was compared with each one of the two Control groups in terms of birthweight adjusting for age, current body mass index (BMI), and number of cigarettes smoked; additionally the maternal and perinatal factors were compared.

Linkage of the three groups of ‘daughters' to their mothers' maternity records in the AMND was done by probabilistic methods (using the surname and the date of birth of the daughters to link to the surname and date of delivery of the mothers).

A list of the relevant obstetric and neonatal factors that could affect intrauterine and later development was drawn up and then extracted from the records of the ‘mothers' in the AMND. The variables included the mothers' age, social class, height and weight at booking, number of cigarettes smoked per day, mode of delivery, gestational age, pregnancy number, birthweight and standardised birthweight score, and Apgar score at 1 and 5 minutes. Data were also extracted on the social class of the fathers, and pregnancy complications like antepartum haemorrhage, diabetes mellitus or pre-eclampsia. Standardised birthweight score for the ‘daughters' was calculated using the following formula:

display math

Birthweight scores are standardised for sex, parity and gestational age and allow comparison between participants and the total population. A score of 0 implies no deviation from the population mean, whereas a score of +1 or −1 indicates a positive or negative deviation from the mean by 1 standard deviation (SD). We have an established system to check for systematic errors during data entry on databases, additionally, there are in-built consistency checks within the database that prevent data entry errors.

Statistical analysis and sample size calculation

Two separate comparisons were carried out. The first between the Cases group and Control group 1, and the second between the Cases group and Control group 2. No comparison between the two Control groups was carried out. The primary comparison measures were the birthweight (in grams) and the standardised birthweight score. The secondary comparisons included the current daughters' characteristics: age and BMI at presentation, and number of cigarettes smoked, the daughters' characteristics at birth including the mode of delivery, gestational age and Apgar score at 1 and 5 minutes and the mothers' characteristics including parity, BMI, age, partner's social class, smoking, antepartum haemorrhage, diabetes and pre-eclampsia.

The following sample size calculations assume that birthweights are normally distributed and use an estimated mean birthweight of 3349 g for the population controls, with 621 g as an estimate of the pooled standard deviation.[21] Assuming the difference in the mean birthweights between Cases and Controls is 150 g, to detect this difference at the 5% significance level, using a two-sided independent t test with 80% power, 271 women would be required in each group. For a two-sided independent t test to have 80% power, at the 5% level of significance, to detect a difference of 0.5 between the mean standardised birthweight scores, would require 126 women in each group, assuming a pooled standard deviation of 1.

For categorical variables, characteristics were compared using chi-square tests and absolute frequencies and percentages are reported. Continuous variables were compared using independent t tests for normally distributed and Mann–Whitney U tests for non-normally distributed variables. Means and standard deviations or medians and interquartile ranges (IQRs) were reported, as appropriate. We reported the numbers and percentages of all missing data. Missing data were excluded from the analysis and a P value <0.05 was regarded as statistically significant.

All analyses were performed using the statistical software package SPSS Statistics version 20.0 for Windows (IBM SPSS Inc., Chicago, IL, USA).

Results

Records of 7667 women from the Fertility Centre database were reviewed, of which 1632 (21.3%) were matched to their maternal records, indicating that the women were born in Aberdeen (see Figure 1). After excluding 22 women with a twin sibling, 487 (30.2%) women with ovulatory dysfunction, 789 (49.0%) with ovulatory levels of progesterone and 334 women (20.7%) with unclear ovulatory status were identified. Of the women with ovulatory dysfunction, 4 (0.8%) had a diagnosis of hypogonadotrophic hypogonadism and 17 (3.5%) had ovarian failure, leaving 466 (95.7%) women who were included as Cases. After excluding 241 women (30.5%) with unexplained infertility from the ovulatory participants, 548 (69.5%) women formed Control group 1.

Figure 1.

Women included in the study

During 1992 to 2007, 43 579 women gave birth at Aberdeen Maternity Hospital, of whom 17 832 (40.9%) could be matched with their mothers' maternity records, had no record of attendance at the Fertility Centre before delivery, and had no twin sibling. These 17 832 women formed Control group 2.

Missing data

Age at presentation of the daughters was not missing in any participant, the BMI at presentation was missing in 26.6% of the Cases, 32.7% of the Control group 1, but only in 4.4% of Control group 2 participants, indicating a stricter adherence to recording height and weight in the antenatal clinic compared with the fertility clinic. The current smoking status or number of cigarettes smoked per day was missing in 2.4% of Cases, 1.1% of Control group 1 and 1.5% of Control group 2. The length of the menstrual cycle was missing from the records of the Cases nine times more than in Control group 1 because of menstrual cycle irregularities and oligo/amenorrhoea in some women (21.5% versus 2.4%); the same trend was present for the midluteal progesterone level, for the same reasons (13.8% versus 1.8%).

Missing data from the maternity records of the mothers of the Cases group, Control group 1 and Control group 2, respectively, were as follows: mother's age at delivery was missing in 0.6%, 0.4% and 0.1%; mother's BMI in 58.7%, 70.3% and 56.8%; standardised birthweight in 15.6%, 15.7% and 27.5%; maternal smoking status in 42.4%, 46% and 40.9%; paternal social class in 23.1%, 20.1% and 22.9%; maternal diabetes, pre-eclampsia, antepartum haemorrhage, mode of delivery, pregnancy number and daughter's birthweight in <1% in all groups, Apgar scores at 1 and 5 minutes in 7.9%, 11.9% and 11.2%, and gestational age in 3.3%, 3.6%, and 3.5%.

Women in the Cases group were younger at presentation to the fertility clinic than women in Control group 1, mean (SD) age 28.4 (5.1) compared with 29.6 (5.2) years, but older than women in Control group 2 at the time of booking at the antenatal clinic, 27.3 (5.7) years. Additionally, women in the Cases group had a significantly higher BMI at presentation than women in control group 1 or control group 2; median BMI in kg/m2 (IQR) was 25.4 (22.1–30.1), 23.8 (21.5–28.0) P = 0.001 and 24.0 (21.7–27.3), P < 0.001, respectively. The proportions of nonsmokers, moderate smokers and heavy smokers were comparable between Cases and Control groups 1 and 2: P = 0.22 and P = 0.53, respectively, as shown in Table 1.

Table 1. Comparison of birthweight and current characteristics of women in the Cases and Controls groups
 Cases (anovulatory) (n = 465)Control group 1 (ovulatory infertile) (n = 547)P valueControl group 2 (fertile) (n = 17 698)P value
Birthweight in grams, mean (SD) 3203 (522)3235 (482)0.303226 (495)0.31
Standardised birthweight score, mean (SD) –0.10 (0.98)–0.14 (1.02)0.74–0.10 (1.00)0.97
Current body mass index in kg/m 2 , median (IQR) 25.4 (22.1–30.1)23.8 (21.5–28.0)0.00124.0 (21.7–27.3)<0.001
Mean age in years (SD) 28.4 (5.1)29.6 (5.2)<0.00127.3 (5.7)<0.001
Smoking status: n (%)
Nonsmoker309 (68)346 (64)0.2210 830 (67)0.53
Moderate (≤20/day)142 (31)186 (34)5129 (32)
Excessive (>20/day)4 (1)10 (2)243 (1)

The mean (SD) birthweight in grams for the whole study population was 3226 g (495). This was comparable between Cases (3203 [522] g) and Control group 1 (3235 [482] g, P = 0.30) and Control group 2 (3226 [495] g, P = 0.31). Similarly, the mean (SD) standardised birthweight score was comparable in the Cases (−0.10 [0.98]) and Control group 1 (−0.14 [1.02], P = 0.74), and control group 2 (−0.10 [1.00], P = 0.97), as shown in Table 1.

When categorised according to their birthweight, 9% of the Cases had low birthweight (defined as <2500 g) compared with 5% of Control group 1. The unadjusted odds ratio (95% confidence interval [95% CI]) of 1.73 (1.06–2.83) was statistically significant, but not the adjusted odds ratio when adjusted for confounders (current BMI in adulthood, age at presentation and number of cigarettes smoked) adjusted odds ratio (95% CI) 1.55 (0.80–2.98). When compared with Control group 2 with low birthweight rate of 7%, this was similar to the Cases with odds ratio (95% CI) of 1.36 (0.98–1.88) and an adjusted odds ratio of 1.15 (0.75–1.75).

The odds of birthweight below the 5th centile or over the 95th centile were comparable in Cases, Control 1 or Control 2 as shown in Table 2.

Table 2. Comparison of the categories of women in the Cases and Controls groups with a low or high birthweight
BirthweightCases (anovulatory) (n = 465)Control group 1 (ovulatory infertile) (n = 547)OR (95% CI)Adjusteda OR (95% CI)Control group 2 (fertile) (= 17) 698OR (95% CI)Adjusteda OR (95% CI)
  1. a

    Adjusted for daughters' age, body mass index and smoking category in adulthood at presentation.

Low birthweight <2500 g41 (9%)29 (5%)1.73 (1.06–2.83)1.55 (0.80–2.98)1178 (7%)1.36 (0.98–1.88)1.15 (0.75–1.75)
<5th percentile (2381 g)29 (6%)18 (3%)1.96 (1.07–3.57)1.91 (0.83–4.44)826 (5%)1.36 (0.93–1.99)1.21 (0.74–1.99)
>95th percentile (3970 g)26 (6%)28 (5%)1.10 (0.63–1.90)1.68 (0.83–3.39)880 (5%)1.13 (0.76–1.69)1.25 (0.80–1.96)

Perinatal outcomes in the three groups are shown in Table 3. The median (IQR) mother's age at delivery, the median (IQR) maternal BMI, social class, the parity number, incidence of antepartum haemorrhage, pre-eclampsia, or diabetes mellitus were comparable between the mothers of the Cases and the mothers of each one of the two Control groups. The number of mothers who smoked was significantly higher in the Cases group (55.6%) than the mothers of women in Control group 1 (42.7%, P = 0.002), but was comparable to that of the mothers of Control group 2 (58.5%, P = 0.93).

Table 3. Comparison of the mothers' characteristics, and the daughters' characteristics at birth
 Cases (anovulatory) (n = 465)Control group 1 (ovulatory infertile) (n = 548)P valueControl group 2 (fertile) (n = 17 832) P
  1. a

    For those women who smoked.

  2. b

    Mild, moderate and severe pre-eclampsia.

  3. c

    Before 37 weeks of gestation.

Mothers' age in years at delivery, median (IQR) 24 (21–28)24 (21–28)0.6724 (21–28)0.77
Mothers' BMI in kg/m 2 , median (IQR) 23.6 (20.9–26.0)22.8 (20.9–25.2)0.1823.1 (21.3–25.7)0.94
Parents’ social class, median (IQR) 4 (4–6)4 (4–6)0.794 (4–6)0.54
Mother smokes, n/N (%)148/266 (55.6)126/295 (42.7)0.0026162/10 541 (58.5)0.93
Cigarettes smoked by mother, median (IQR) a 9 (0–19)0 (0–19)0.027 (0–19)0.76
Mother's parity number of the index pregnancy, median (IQR) 1 (0–2)1 (0–1)0.901 (0–2)0.47
0, n (%)196 (41.2)224 (41.0) 7 100 (39.8) 
1149 (32)187 (34.2) 6 000 (33.6) 
270 (15.1)80 (14.6) 2 999 (16.8) 
≥350 (10.8)56 (10.2) 1733 (9.7) 
Mothers' diabetes mellitus, n (%)0 (0.0)0 (0.0)1.008 (0.0)1.00
Mothers' pre-eclampsiab, n (%)110 (23.7)113 (20.7)0.104 279 (24.0)0.87
Mothers antepartum haemorrhage, n (%)24 (5.2)23 (4.2)0.45858 (4.8)0.60
Mode of delivery, n (%)
Caesarean section31 (6.7)38 (6.9)0.871 275 (7.2)0.68
Instrumental58 (12.5)59 (10.8)2 132 (12.0)
Assisted breech4 (0.9)9 (1.6)323 (1.8)
Normal vaginal372 (80)441 (80.6)13 965 (78.3)
Apgar scores
Daughters' APGAR at 1 minute, median (IQR)9 (7–9)9 (7–9)0.289 (8–9)0.73
Daughters' APGAR at 5 minutes, median (IQR)9 (9–10)9 (9–10)0.789 (9–10)0.73
Daughters' gestational age in weeks, median (IQR) 40 (39–41)40 (40–41)0.7340 (39–41)0.20
Daughters born pretermc n/N (%)21/450 (4.7)12/528 (2.3)0.04689/17 214 (4.0)0.48

The proportion of Cases (daughters) who were born prematurely (<37 weeks of gestation) was slightly higher than in Control group 1 (4.7% versus 2.3%, P = 0.04), but was comparable to Control group 2 (4.0%, P = 0.48). The median (IQR) gestational age of women in the Cases group was 40 (39–41) weeks, comparable to those of Control group 1 of 40 (40, 41) weeks and Control group 2 of 40 (39, 41) weeks. The mode of delivery and Apgar scores at 1 and 5 minutes were comparable between Cases and each one of the two Control groups.

Discussion

Main findings

The results of this study fail to demonstrate a significant association between in utero factors represented primarily by birthweight and subsequently developing ovulatory dysfunction.

Strengths and weaknesses

Aberdeen city and district, in Scotland, provides a unique opportunity to conduct longitudinal studies because the population of the region is fairly stable.

We had the additional benefit of using the standardised birthweight score, which is corrected for gestational age, sex of the baby, and parity of the mother. It provides the most accurate way to study birthweight in epidemiological studies of this nature and allows a more robust measure of growth restriction than crude birthweight.

We used a second control group from the general fertile population, in addition to the first control group from the Fertility Clinic patients with confirmed ovulation. Although Control group 1 have confirmed ovulation, these women still come from the group of infertile women and may not fully represent the fertile population. Adding Control group 2 accounts for this possibility and increases the validity of our results.

We have an established system to check for systematic errors during data entry on databases. The completeness of data entry is checked against the NHS hospital returns for maternity and specified gynaecological admissions. There are in-built consistency checks within the database that prevent data entry errors and ad hoc validation exercises using case note review have found the data entry to be over 97% accurate.[22, 23]

Although the 21% match between daughters' and mothers' records may sound limited, this is consistent with results from previously published studies based on the intergenerational cohort within the AMND[24-26] and reflects the degree of outmigration of daughters. The inclusion of about 19 000 mother–daughter record pairs makes this one of the biggest reported studies of its kind to date to look at the effect of birthweight, antenatal events and subsequently ovulatory dysfunction in the female offspring.

Women with ovulatory dysfunction are a heterogeneous group, including women who have PCOS. We decided not to limit our study to this group of women, or to perform subgroup analysis with respect to PCOS, because the diagnostic criteria changed halfway through our study period with the introduction of the Rotterdam criteria.[20] Furthermore, in many instances the impression of a PCOS appearance of the ovaries on ultrasound scan was made subjectively by the person performing the scan, so it is difficult to be certain of the consistency in the criteria used to make the diagnosis. In addition to the above, we felt that ovulatory dysfunction from a clinical perspective is the outcome of interest, whether it was associated with PCOS or not. Our databases do not record data on the progress of the neonates during childhood and adolescence. We were therefore unable to comment on the possibility that a combination of a phase of catch-up growth during childhood, after growth restriction in intrauterine life, could be the precursor of ovulatory dysfunction in adulthood, rather than low birthweight alone.

Interpretation in light of other evidence

Ibanez has reported an association between low birthweight in girls who experience premature pubarche and ovarian hyperandrogenism.[9, 13, 27] In a study of 51 girls, a tendency to ovarian dysfunction in girls with low birthweight was demonstrated that was independent of BMI in adulthood.[10] In another study on 49 adolescent girls, the incidence of anovulation was shown to be higher in girls born small for gestational age compared with those born appropriate for gestational age.[12] Low birthweight was found to be associated with clinical and biochemical markers of PCOS in another study of 35 women from Italy, 19 of whom had been small for gestational age and the rest were premature, but with birthweight appropriate for gestational age.[14] Melo, in a recent study of 165 women from Brazil, has shown that women born small for gestational age are twice as likely to have PCOS than women born appropriate for gestational age.[15] All of these studies suggesting an association between low birthweight and PCOS later in life had limited sample sizes, and many of them included highly selected populations (adolescents with low birthweight, followed by premature menarche, and hyperandrogenism in the Ibanez series). Our study, which had a larger sample size, has examined women with ovulatory dysfunction (including, but not exclusively restricted to, women with PCOS) and did not confirm these findings, despite being adequately powered to show such an association if it existed.

Our findings therefore agree with those in the retrospective cohort study from the Netherlands by Sadrzadeh et al., who could not confirm a relationship between self-reported birthweight and anovulation or PCOS.[18] No data were available in this study on the gestational age at birth, and hence the appropriateness of the birthweight. Another longitudinal, population-based cohort study, from north Finland, showed that overweight and obesity in adult life, and not birthweight or intrauterine growth restriction, were associated with self-reported symptoms of PCOS.[19] Our results support this lack of association between ovulatory dysfunction and birthweight and once again confirm the well-established link between ovulatory dysfunction and BMI in adulthood.

Cresswell et al., in a study of 235 women from Sheffield (49 of them with a diagnosis of PCOS), found that obese PCOS women tend to have above average birthweight. This conclusion did not apply, in their study, to women with PCOS who were not obese.[16] Another way of interpreting the results of this study could simply be that having above average birthweight predisposes to obesity in adulthood, without implicating the diagnosis of PCOS in the conclusion. Polycystic appearance of the ovaries on ultrasound (not PCOS) was found to be associated with a higher ‘self-reported’ birthweight in mainly asymptomatic young women with normal androgen profile, some of whom were on the contraceptive pill.[28] These women do not necessarily have ovulatory dysfunction or PCOS. In a very recent national registry study from Denmark, Mumm et al.[17] reported an association between being born with gross macrosomia (birthweight ≥ 4500 g, which is above the populations' 98.5th centile) and increased incidence of PCOS in adulthood. This study did not report the characteristics of these individuals in adulthood. There is a well-established link between macrosomia at birth and obesity in adulthood[29, 30] (which is in turn a well-established risk factor for PCOS), and this study did not adjust the results for BMI in adulthood. Consequently it is difficult to ascertain if the association between neonatal macrosomia and PCOS in adulthood was a direct one, or mediated through obesity in adulthood. Furthermore, our study did not show an association between being born with a high birthweight (above the 95th centile) and ovulatory dysfunction later in life (with or without adjusting for BMI in adulthood).

Meanings of the results and implications for clinicians

Our results do not support a fetal origin theory of ovulatory dysfunction. We demonstrated an association between higher BMI in adulthood and ovulatory dysfunction when Cases were compared with women with normal ovulation or normal fertility. This emphasises that acquired (and modifiable) risk factors for anovulation, like obesity, remain the most important proven factors implicated in ovulatory dysfunction. Our large study population included women with ovulatory dysfunction, with or without PCOS. The findings of our study do not exclude the possibility of such a relationship in a subgroup of women who have ovulatory dysfunction associated with PCOS.

Future research

Large studies are required to assess early fetal and neonatal risk factors for PCOS using the standardised Rotterdam diagnostic criteria for PCOS and adjusting for the characteristics in adulthood like BMI, age and smoking. Further large studies following female neonates to childhood and adolescence, to test the theory that low birthweight followed by catch-up growth may lead to ovarian dysfunction, rather than the birthweight alone, are also recommended.

Conclusion

Ovulatory dysfunction in women does not appear to be related to birthweight.

Disclosure of interests

None.

Contribution to authorship

AGS designed the study protocol, applied for ethics approval, analysed the data, and wrote the paper. KH provided statistical advice on the power calculation, study design and analysis plan; had input into the interpretation of the results and edited the final draft. SB supervised the project and had input into the design, analysis and interpretation of the results; and edited the final draft. All authors approved the final draft of the paper.

Details of ethical approval

This study was approved by the North of Scotland Research Ethics Committee (REC reference number: 07/S0801/70), the Caldicott Guardians of the Aberdeen Maternity and Neonatal Database (AMND), and the Aberdeen Fertility Clinic Database (AFCD).

Funding

This study was internally funded by the University of Aberdeen, Aberdeen, UK.

Acknowledgements

The study team would like to thank Mrs Val Angus and Mrs Linda Murdoch for their assistance in data extraction from the AFCD and AMND.

Ancillary