Cervical Length and Androgens in Pregnant Women With Polycystic Ovary Syndrome: Has Metformin Any Effect? Free

Abstract

Context:

Women with polycystic ovary syndrome (PCOS) have increased risk of preterm delivery. Shortening of the cervix is a sign of preterm delivery.

Objective:

This study aimed to investigate potential effect of metformin on cervical length and whether androgen levels correlate with cervical length in PCOS pregnancies.

Design and Setting:

This was a sub-study of a randomized, placebo-controlled, multicenter study (The PregMet study) performed at 11 secondary or tertiary centers from 2005 to 2009.

Participants:

Two-hundred sixty-one pregnancies of 245 women with PCOS, age 18–42 years participated.

Interventions:

Participants were randomly assigned to metformin or placebo from first trimester to delivery.

Outcome Measurements:

We compared cervical length and androgen levels in metformin and placebo groups at gestational weeks 19 and 32. We also explored whether cervical length correlated with androgen levels.

Results:

We found no difference in cervical length between the metformin and the placebo groups at gestational week 19 and 32. Dehydroepiandrosterone (DHEAS) tended to be higher in the metformin group. There were no correlations between androgens and cervical length at week 19. At gestational week 32, androstenedione (P = .02) and DHEAS (P = .03) showed a trend toward negative correlation to cervical length. High androstenedione level correlated with shortening of cervical length from week 19 to 32 when adjusted for confounders (P = .003). T (P = .03), DHEAS (P = .02), and free testosterone index (P = .03) showed a similar trend.

Conclusion:

Metformin in pregnancy did not affect cervical length in women with PCOS. High maternal androgen levels correlated with cervical shortening from the second to the third trimester of pregnancy, as a sign of cervical ripening.

Despite extensive use of tocolytic agents, antibiotics, and sick-leave, preterm delivery remains one of the major challenges in obstetrics with effects on short- and long-term health of the offspring. The prevalence of preterm deliveries varies between countries. Interestingly the United States has a significantly higher prevalence of preterm deliveries compared with other developed countries. An increase from 9% in 1982 to 12% in 2003 was reported (1). In 2012 the rate of preterm birth in the United States decreased to 9.7% (2, 3). In Norway, approximately 6% of all deliveries occur before gestational week 37 (4). The prevalence seems to be slowly increasing in most Western countries.

Polycystic ovary syndrome (PCOS) is a common condition, and recent studies suggest a prevalence of 15% in a general population (5, 6) and up to 30% among women who had experienced preterm delivery (7). They are also more prone to suffer pregnancy complications than women without PCOS. Increased risk for preeclampsia, gestational diabetes mellitus, and preterm delivery is reported (8–10). Women with PCOS have increased androgen levels throughout life compared with controls (11) and most of them are hyperandrogenic both in the nonpregnant and pregnant state (5). Androgen levels in relation to cervix length are not described to any extent. In pregnancy, androgen levels tend to be higher than in the nonpregnant state (12). Androgen excess is only one of several possible etiologic factors in preterm delivery and might be a causal factor in cervix remodelling in women with PCOS. Hyperandrogenicity is also associated with increased levels of inflammatory mediators (13) which may contribute to both shortening of the cervix and/or preterm labor (14, 15).

Metformin may have an effect on premature labor in PCOS pregnancies (16). As shortened cervix is a risk factor of preterm labor, a possible association between cervical length and metformin exposure is of importance to explore. Pooling results from two randomized controlled trials [The Pilot study and the PregMet study (17, 18)] on metformin or placebo treatment in pregnant women with PCOS, we found a decreased prevalence of preterm delivery in the metformin group compared with the placebo group (16). Cervical length measured by ultrasound is used to assess women with risk for preterm birth (19, 20). Shortening of the cervix is a sign of cervical ripening and is associated with increased risk of imminent preterm delivery.

Hormonal regulation is an important but not fully understood mechanism of cervical remodelling (21). There is increasing body of evidence that vaginal administration of progesterone decreases the prevalence of preterm delivery in singleton pregnancies (22, 23). Also, alteration in estrogen levels has been shown to affect late cervical fibroblast function in vitro (21). There is also some evidence that androgens may play a role in regulating cervical function, but this has been little studied (24).

The aims of this study were to investigate the potential effect of metformin 1) on cervical length measured by vaginal ultrasound examinations or 2) on androgens, and 3) to explore whether there was any correlation between androgen levels and cervical length in PCOS pregnancies.

Materials and Methods

The PregMet study

The present study is a substudy of the Metformin Treatment in Pregnant PCOS Women (PregMet) study, which was a prospective, randomized, double-blind, multicenter trial, comparing 2000 mg metformin daily with placebo from the first trimester to delivery (18).

In short, the inclusion criteria were: 1) PCOS diagnosed before pregnancy according to The Rotterdam Criteria (25), 2) age 18–45 years, 3) gestational age between 5 and 12 weeks, and 4) a singleton viable fetus shown on ultrasonography. The exclusion criteria were alanine aminotransferase greater than 90 IU/L, serum creatinine concentration greater than 130 μmol/L, known alcohol abuse, previously diagnosed diabetes mellitus or fasting serum glucose greater than 7.0 mmol/L at the time point of inclusion, treatment with oral glucocorticoids, or use of drugs known to interfere with metformin.

In all, 273 pregnancies in 257 women were randomly assigned to either metformin or placebo treatment (16 women participated twice). Randomization, blinding, and performed measurements are described in detail elsewhere (18). All participants received written and individual counseling on diet and lifestyle at inclusion. Thereafter, treatment with metformin 500 mg (metformin hydrochloride, Metformin, Weifa AS) or identically coated placebo tablets was initiated. The participants were instructed to take one tablet twice daily during the first week and two tablets twice daily for the rest of the study period (ie, until delivery). Study medication was stopped at delivery.

Standardized interviewer-administered questionnaires were used to obtain self-reported data on former medical and gynaecologic/obstetric history, ethnicity, employment, and smoking habits. Biometric variables, including height, weight, and blood pressure (BP) were recorded. BP and heart rate were measured while the patient was in the sitting position after at least 10 minutes of comfortable rest in a chair. The BP was measured three times, two minutes apart with digital devices. The mean of the second and third measurements was calculated. Body weight was recorded with light clothes on and without shoes.

Ultrasound measurements

The length of the cervix was measured by ultrasound, in a longitudinal midsection plane, where the cervical channel is visualized. This was performed with a vaginal probe, without pressure on the cervix. Funnelling was noted. The mean of three measurements was registered. All ultrasound examinations were performed at the different study sites, with up-to-date ultrasound equipment in clinical use present at each site. Gestational age was determined by midpregnancy ultrasound, measuring biparietal diameter, femur length, and mean abdominal diameter of the fetus.

Assays

Venous blood samples were drawn from an antecubital vein between 0800 and 1100 hours after an overnight fast in both week 19 and week 32 of pregnancy. The specimens were centrifuged within 1 hour and stored at 5°C. Within 24 hours, the specimens were frozen at −80°C. All blood samples were analyzed in 2009, after the last study participant had delivered. Dehydroepiandrosterone (DHEAS) and SHBG were analyzed by the ELISA technique with the reagents and calibrators supplied by the manufacturer (DRG Instruments, GmbH). We used organic solvent extraction (dichloromethane for T and ethyl ether for androstenedione) prior to quantification to analyze serum T and androstenedione. For quantification we used the ELISA technique for T (DRG Instruments, GmbH) and Coat-A-Count RIA kits (Diagnostic Products Corporation) for androstenedione. The intra- and interassay coefficients of variation were 6.6 and 4.0% for DHEAS, 5.3 and 2.8% for androstenedione, and 11.9 and 9.1% for T, respectively. Free testosterone index (FTI) was calculated; T (nmol/L) × 100/SHBG (nmol/L). Insulin was measured by the ELISA technique using kits and reagents supplied by the manufacturer (DRG Instruments, GmbH). The intra-and interassay coefficients of variation were 2.7 and 7.6%, respectively.

The study was approved both by the Regional Committees for Medical and Health Research Ethics and The Norwegian Medicines Agency. Written informed consent was obtained from each patient before inclusion and the declaration of Helsinki was followed throughout the study. The study was conducted according to principles of “Good Clinical Practice” and the trial is registered at www.clinicaltrials.gov as trial number NCT00159536.

The present substudy

In all, cervix measurement was performed in 261 of the pregnancies assessed in the PregMet study (pregnancies of 245 women), either in week 19 or week 32 of pregnancy, or both. These pregnancies comprise the study population in the present substudy. Measurements from both time points were available in 236 women, 18 had their cervix measured only in week 19, and seven only in week 32 of pregnancy. Of the 18 women who had their cervices measured only in week 19 of pregnancy, four dropped out of the study and five had preterm deliveries before week 32. The remaining nine women met at the week-32 followup but their cervixes were not measured. They were included in the analysis, as were those who did not have their cervix measured in week 19, but in week 32 of pregnancy. Androgen analysis demands also available blood samples, thus leading to minor variations in numbers included in each analysis (Tables 1 and 2).

Statistical analyses

All statistical procedures were performed using the SPSS version 22.0 (SPSS, Inc). Observed data are reported as mean ± SD. We used t test for independent samples to examine potential differences between the metformin and placebo groups. To evaluate associations between cervix length and hormone levels in gestational weeks 19 and 32 we used multivariate linear regression analysis, after testing the study material for normality and assuming equal variances. Comparison of gestational length, cervical length, androgens, and insulin levels in gestational weeks 19 and 32 were done both as intention-to-treat and per protocol analysis. We also explored a possible association between the change in cervical length between weeks 19 and 32 (cervical length week 19 − cervical length week 32 = Δ cervix) and hormonal levels in week 32. In all, in our analysis we performed adjustments for the following possible confounders: body mass index (BMI), age, randomization to metformin/placebo, previous vaginal delivery, and cervix conisation. To adjust for multiple testing we chose a pragmatic approach and considered a P ≤ .01 as significant. P < .05 and > .01 were considered as trends toward significance. Results are given as β-coefficients with 95% confidence intervals (CI), where CI not including zero suggests significance or a trend toward significance.

Results

Baseline characteristics of the study population are shown in Table 3. No differences between the metformin and placebo groups were observed.

Metformin effect on cervical length and hormone levels

We found no difference in cervix length between the metformin and placebo groups, either at gestational week 19 or 32 (Table 2). Per-protocol analyses did not change the results. Androgen and insulin levels at both gestational weeks 19 and 32, and gestational lengths were also similar in the metformin and placebo groups. When performing per protocol analyses (Table 2) we found higher DHEAS levels in the metformin group compared with the placebo group (−0.07 [−0.12; −0.02]; P = .01) at gestational week 32.

At gestational week 19, no correlations were found between cervical length and androgen or insulin levels (Table 3). At gestational week 32 cervix length tended to correlate negatively with androstenedione (−0.2 [0.3; −0.03]; P = .02) and DHEAS (−0.83 [−1.5; −0.1]; P = .03), ie, the higher androgen level, the shorter the cervix.

At gestational week 32 androstenedione correlated positively with Δ cervix from week 19 to week 32 when adjusted for cervical conisation, previous vaginal deliveries, randomization metformin/placebo, BMI, age, and for interval between examinations (0.2 [0.05; 0.35]; P = .01) (Table 3). Equally adjusted analyses showed similar trends for DHEAS (1.01 [0.16; 1.88]; P = .02), T (0.42 [0.04; 0.79]; P = .03) and FTI (1.31 [0.11; 2.50]; P = .03). We found no correlation between insulin of SHBG and Δ cervix.

Discussion

The most important findings of our study were 1) that metformin had no effect on cervical length, and 2) that shortening of the cervix (Δ cervix) from the second to the third trimester correlated with increased levels of androstenedione. Increased levels of DHEAS and FTI showed a trend toward correlation to Δ cervix. Metformin exposure did not affect cervical length or androgen levels at gestational week 19 or 32.

This study comprised women from The PregMet study. An epianalysis of the Pilot Study and the PregMet study showed a possible protective effect of metformin on the combined end-point; preterm delivery and second trimester miscarriages (16). Our study only evaluates cervical length in week 19 and 32 of pregnancy. Most of the preterm deliveries in the PregMet study were “late preterm” (ie, after gestational week 34), which may explain why we did not find any difference in absolute cervical length at gestational week 32. If metformin has a protective effect on preterm deliveries it might be exerted through pathways other than androgen synthesis or shortening of the cervix. Metformin could modulate inflammation, by reducing inflammatory cell numbers or inflammatory markers. Metformin has been shown to regulate the endometrial cell function directly (26). Another theoretical mechanism could be lowering of contractility threshold of the myometrium: one study reports that metformin does not exert effect on the spontaneous or oxytocin-induced myometrial contractions in an in vitro model (27).

In the third trimester of pregnancy, at week 32, high maternal androstenedione and DHEAS levels showed a tendency toward correlation to a short cervix. The cervix consists of fibroblasts, smooth muscle, epithelial, and immune cells that secrete extracellular matrix (28). Androgens, estradiol, relaxin, and prostaglandin can alter the extracellular matrix composition and affect the mechanical strength of the cervix (12, 29).

The idea of androgens affecting cervical ripening and by that causing preterm labor is supported by animal studies in which administration of supra-physiological doses of androstenedione was injected in pregnant rhesus monkeys causing preterm labor (30). This is in line with the present observation that androstenedione correlated negatively with cervical length and change, whereas the other androgens investigated showed a tendency toward correlation.

In other animal studies, administration of dihydrotestosterone resulted in decreased cervical resistance, whereas administration of the androgen receptor antagonist flutamide increased the resistance. Cervical resistance was determined by the “cervical creep method,” an experimental model to evaluate the effect of hormone administration on cervix fibers exposed to a certain load (24). Shortening of the cervix can be considered as a physiological sign of decreased cervical resistance.

Human studies on DHEAS administration and cervix ripening, increased collagenase, or IL-8 activity, which correlate with shorter time to delivery (31–34). These studies were performed later in pregnancy than ours (ie, in gestational week 38–41), they are mostly small, and rarely measure DHEAS levels, but investigate the effect of DHEAS injections late in normal pregnancies. One study also reported that high DHEAS levels correlated with higher Bishop score (more ripe cervix) in women with spontaneous cervix ripening compared with women in need of induction of labor (35). Our results (Table 3) are in line with these findings as we found that increased DHEAS level tended to correlate with shortening of the cervix, a marker of cervical ripening. DHEAS is a precursor for most other androgen hormones and might explain why other androgens are elevated too. Importantly, the circulating level of DHEAS is approximately 1000 times higher than those of androstenedione and T. The role of DHEAS in PCOS pregnancies has not been investigated to any extent. DHEAS, a precursor in the synthesis of androgens, correlates with increased shortening of the cervix. Why our participants showed higher DHEAS concentration in the metformin group than in the placebo group remains unclear. However, this observation is in line with at least one other study performed on women with PCOS and metformin treatment, where pre-in vitro fertilization (IVF) treatment of patients with PCOS taking metformin was shown to increase DHEAS levels while other androgens remained unchanged (36). DHEAS is converted to androgens and estrogens intracellularly in the peripheral tissues. Thus, they may exert their activity in the cells of origin (ie, intracrine effect) or on the surrounding cells (ie, paracrine effect) rather than appearing in the systemic circulation (37, 38). Most androgens tend to increase with gestational age in the normal pregnancy. The reported normal ranges are rather wide. Total T, free T, and SHBG increase from the second until the third trimester in normal pregnancy (12). It has been hypothesized that androgens may influence remodeling of the cervix, especially ripening of the cervix at term, and maybe also in preterm deliveries (31–34).

Strengths and limitations

Our study was performed in a setting with relatively low prevalence of preterm labor (3, 4). If performed in a setting with higher prevalence, such as in the United States (2), it might have altered the results. This study is a substudy of an randomized controlled trial, which might be a limitation as such. Also, the number of women included in each arm of the study represents a limitation. We have performed multiple analyses, which increases the possibility of finding statistical associations by chance. By using the pragmatic approach and setting our significance level at P = .01 we adjusted for this. The strengths of our study are that our findings are in line with former animal studies and smaller clinical studies, and together they represent a consistency. Few studies are done on PCOS in pregnancy and, although being a substudy, the original study (18) was a randomized, placebo-controlled trial of good quality, ie, representing a strength per se (39).

Conclusion

Metformin does not seem to affect cervical length in pregnant women with PCOS. If metformin plays a role to prevent preterm delivery, it is possibly by other mechanisms. In women with PCOS, shortening of the cervix in the latter half of the pregnancy, as a sign of cervical ripening, correlated with elevated androgen levels.

Acknowledgments

This study was registered in ClinicalTrials.gov as trial number NCT00159536.

This work was supported by the Norwegian University of Science and Technology, The Medical Faculty.

Disclosure Summary: The authors have nothing to disclose.

Abbreviations

 
• BMI

  body mass index

 
• BP

  blood pressure

 
• CI

  confidence interval

 
• DHEAS

  dehydroepiandrosterone

 
• FTI

  free T index

 
• IVF

  in vitro fertilization

 
• PCOS

  polycystic ovary syndrome

 
• PregMet

  Metformin Treatment in Pregnant PCOS Women Study.

References