Clinical and biochemical signs of polycystic ovary syndrome in young women born preterm

Objective It has been suggested that adverse early life exposures increase the risk of developing polycystic ovary syndrome (PCOS) in later life. We hypothesized that women born preterm would have more biochemical and clinical signs of PCOS than women born at term. Design The ESTER Preterm Birth Study participants were born in Northern Finland and identified from the Northern Finland Birth Cohort and the Finnish Medical Birth Register. Altogether, 74 women born very or moderately preterm (<34 gestational weeks, VMPT), 127 born late preterm (at 34–36 weeks, LPT), and 184 born full term (≥37 weeks, controls) were included in the analysis (mean age: 23.2 years). Methods We measured serum total testosterone and sex hormone-binding globulin (SHBG) and calculated the free androgen index (FAI). PCOS according to the clinical and biochemical signs was defined either as hirsutism and oligoamenorrhea (via questionnaire) or as oligoamenorrhea and elevated testosterone levels (>2.4 nmol/L). Results Women born VMPT/LPT exhibited 33.0% (8.7, 62.8)/16.4% (−2.0, 38.1) higher testosterone, 28.5% (5.3, 45.9)/24.1% (5.6, 38.9) lower SHBG levels, and 64.6% (19.4, 127.1)/42.5% (11.1, 82.9) higher FAI than controls after adjusting for age and recruitment cohort, maternal BMI, smoking, and pregnancy disorders, parental education, history of hypertension, diabetes, myocardial infarction or stroke, and subject’s birth weight s.d. Odds ratios for having PCOS were 1.67 (0.44, 6.23)/3.11 (1.26, 7.70). Conclusions Women born preterm have a more hyperandrogenic hormonal profile, and those born LPT are approximately three times more likely at risk to have PCOS compared to women born at term.

was categorized in four levels (dummy coded). Self-reported physical activity levels were converted to total metabolic equivalent hours per week (METh/wk) (21).

Definition of PCOS according to the clinical and biochemical signs
Several questions were asked to assess hirsutism, a clinical manifestation of androgen excess: "Do you have excessive growth of hair? Do you remove facial hair regularly (excluding the eyebrows)? How many times per month?" Hirsutism was considered present if a woman had excessive growth of hair or removed facial hair regularly at least four times per month. To assess menstrual cycle irregularity, one question inquiring about the minimum and maximum length and another inquiring about the regularity of the menstrual cycle without hormonal contraception was asked: "Is your menstrual cycle often (more than twice per year) longer than 35 days?" (22). Oligomenorrhea was defined if the participants answered "yes" to the aforementioned question, and irregular menses were defined if the length of the menstrual cycle regularly varied over seven days or if the participant had been treated for irregular menstrual cycle.
According to the Rotterdam definition and the International guidelines, the diagnostic criteria of PCOS require women to have two out of three of the following manifestations: 1) oligo-or anovulation, 2) clinical and/or biochemical signs of hyperandrogenism, and 3) polycystic ovaries at ultrasound (23,24). As the ultrasound of the ovaries had not been performed in the study population, the group referred in the present study as " women with PCOS" included the participants having 1) oligo-or anovulation and hirsutism or 2) oligo-or anovulation and elevated testosterone levels (>2.4 nmol/l).

Ethics
The research protocol was approved by the Coordinating Ethics Committee at Helsinki and Uusimaa Hospital District, and all participants provided written informed consent. increased regarding testosterone levels; they were attenuated regarding SHBG levels and FAI. However, with the same adjustments, the differences in SHBG levels and FAI between the LPT group and the controls remained similar (Model 3, Table 3).
When gestational age at birth was used as a continuous variable, a one-week higher gestational age was associated with a 2.8% (95% CI 0. 1, 5.4)
We conducted an additional analysis by excluding women born SGA. The mean difference in testosterone levels between the VMPT group and the controls increased after adjusting for age and recruitment cohort (Model 1) and attenuated when adjusted for parental and prenatal factors (Model 2); they were not statistically significant after adjusting all covariates (Model 3) compared to an analysis that included those born SGA. However, the differences in SHBG levels and FAI increased in both groups compared to controls after adjusting for Models 1-3; the ORs for PCOS were similar in all of the models.
When hormonal contraception was not a covariate in Model 3, the difference between the preterm and control groups decreased regarding testosterone levels, but it increased regarding SHBG and FAI levels.
The ORs for PCOS remained similar. We also performed a sensitivity analysis excluding the 154 women (40.0%) who used hormonal contraception. Differences in testosterone and FAI attenuated, but differences in SHBG and OR's for PCOS remained similar, although confidence intervals were wider (Supplementary Table 1). In addition, we excluded those subjects who had given birth within six months before participating to this study (2 subjects born VMPT, 3 born LPT and 6 controls). The results remained similar (Supplementary Table 2).

Discussion
We found that women born VMPT had higher serum total testosterone, while women born VMPT and LPT had lower SHBG levels and higher FAI. The differences were larger between the VMPT group and the controls than between the LPT group and the controls, which is consistent with a dose-response relationship between gestational age at birth and hormonal levels. In addition, women born LPT had elevated risk of having PCOS according to clinical and biochemical signs when adjusted for parental and perinatal factors and/or current characteristics.
In a previous study women born with VLBW had similar levels of reproductive hormone levels compared with their controls at 26-28 years of age (14). However, the number of subjects was small (24 VLBW women, 25 controls). Also elevated adrenal androgen levels has been reported in women born before 32 GW particularly in those born SGA, but testosterone levels was not reported (15). Women born SGA (but not preterm) had a higher prevalence of hyperandrogenism and/or PCOS (11,12). In the present study, the overall prevalence of PCOS according to the clinical and biochemical signs was 8.8% (5.4% of women in the early preterm group, 12.6% in the late preterm group, and 7.6% in the control group); this is in line with a prevalence of a 6-18% of PCOS as reported in other populations (6-8). Our results are also in line with the data of the NFBC1966 study in the same area, in which isolated oligoamenorrhea at age 31 years was reported by 11.2%, isolated hirsutism by 10.9%, and both symptoms by 4.2% of the women who returned the questionnaire (26,27).
As expected, use of hormonal contraception was associated with a less hyperandrogenic hormonal profile (lower total testosterone levels and FAI and higher SHBG levels) in the linear regression models.
However, differences in the use of hormonal contraception between groups were small and not statistically significant, and adjusting for this parameter did not alter the results. We also performed a sensitivity analysis excluding women who used hormonal contraception, in which differences in testosterone concentration and FAI attenuated. However, as 40% of women used hormonal contraception, the power of this analysis was limited and the results should be treated with caution.

Mechanisms
We have previously shown in this cohort that young adults born VMPT and LPT had higher rates of obesity and higher waist circumferences and body fat percentages than their peers born at term, although this was not statistically significant in this sample (5). Adjustment for body fat percentage attenuated our main findings, and larger body fat percentage among women born preterm could be among the underlying mechanisms.
Young adults born preterm have lower insulin sensitivity than their peers born at term (1,3,4). In the whole population of the ESTER Preterm Birth Study (including men), especially LPT birth was associated with decreased insulin sensitivity (5); this was mainly explained by increased body fat (5). Similarly, in the present study including only women, those born LPT were more hyperinsulinemic (higher mean fasting and two-hour insulin levels in the oral glucose tolerance test) and insulin resistant (higher HOMA-IR) than women born at term. Women with PCOS exhibit peripheral insulin resistance (28) and compensatory hyperinsulinemia, both of which are central pathogenetic features of the syndrome (29). Insulin resistance might contribute to hyperandrogenism by several mechanisms: Insulin acts synergistically with luteinizing hormone to improve androgen production, and high levels of insulin reduce circulating SHBG levels, thereby increasing the bioavailability of testosterone. In contrast, increasing evidence suggests that an excess of androgen might also contribute to insulin resistance by favoring a predominantly abdominal distribution of body fat and visceral adipose tissue dysfunction (30). Furthermore, abdominal adiposity and adipose tissue dysfunction may induce insulin resistance, and compensatory hyperinsulinemia favors further androgen excess, closing a vicious circle that contributes to the development of the metabolic and cardiovascular risks typically observed in women with PCOS. In line with this hypothesis, in the present study, LPT women exhibited significantly higher waist circumferences compared with controls despite having similar BMIs.
In the present study, neither low birth weight SD score nor being SGA explained the difference in clinical and/or biochemical signs of PCOS between women born preterm and women born at term. The exposure to maternal GDM in the present sample predicted lower total testosterone and lower SHBG levels, and maternal type 1 or 2 diabetes predicted PCOS according to clinical and biochemical signs in women, but this did not explain the difference between women born preterm and their peers born at term. Lower SHBG levels were also found in a parallel study which focused on maternal GDM as an exposure and included predominanty adults born at term (31). That study showed no difference in total testosterone and did not assess clinical signs of of PCOS. It is possible that the association between maternal GDM and hyperandogenemia in the present study is specific to our sampling frame focusing on adult outcomes of preterm birth.
The development of the hypothalamus-pituitary-gonadal-axis in preterm infants is not fully understood. Girls born preterm have highly elevated FSH and LH levels and immature ovarian hormone synthesis during early infancy (32) suggesting disturbance of development of HPG-axis. The origin of PCOS is multifactorial, but there is increasing evidence of a strong genetic background (33). It has been suggested that some genetic determinants are associated with preterm birth as well as with cardiovascular diseases: Women with a history of preterm delivery have an elevated risk of cardiovascular disease (34). Also women with PCOS are at greater risk for preterm delivery and have more pregnancy disorders such as pregnancyinduced hypertension, pre-eclampsia, and GDM (35). In total, our findings support the theory that early life factors may be associated with the development of PCOS in later life.

Limitations
The number of participants in the VMPT group resulted in limited power for the analysis of dichotomous variables (PCOS according to clinical and biochemical signs). Therefore, our findings of no difference in the presence of PCOS between the VMPT group and the controls should be interpreted with caution. Other limitations include the definition of PCOS according to clinical and biochemical signs. Hirsutism should ideally be quantified according to the modified Ferriman-Gallwey score as per common guidelines. However, this scoring was not performed during the extensive clinical examination of the cohort, which included a wide range of reproductive, cardiometabolic, mental health and functional outcomes. Hirsutism is subjective and may be easily over-reported. Nevertheless, we have previously shown that self-reported isolated hirsutism does correlate with increased androgen secretion; further, self-reported oligoamenorrhea and hirsutism can identify women with the typical endocrine and metabolic profiles of PCOS (22,36). We did not perform ultrasonography to assess the presence of polycystic ovaries (PCO). Women with PCO who might be symptom-free would exhibit milder hormonal and metabolic disorders than symptomatic women and they have been reported to show similar metabolic status than control healthy women (37). We would therefore expect that the differences between the clinical and biochemical signs of PCOS and control groups would have been similar or even greater if we would have been able to exclude women with PCO only.

Significance of the findings
The evidence of the association of preterm birth and CVD risk factors in adulthood is strong and consistent.
Although metabolic and cardiovascular disturbances (including abdominal adiposity, disorders of glucose regulation, dyslipidemia, metabolic syndrome, hypertension and cerebrovascular disease) are frequent in adolescent and adult women with PCOS (28,38-40), it is still uncertain whether PCOS confers a risk of CVD through mechanisms other than overweight, insulin resistance, diabetes, and metabolic syndrome. However, recent studies indicate that impaired glucose regulation, enhanced ovarian androgen secretion, and chronic low-grade inflammation observed in premenopausal women with PCOS persist after menopausal transition, emphasizing lifelong health risks related to this syndrome (41).