Lower prevalence of non-cavity-distorting uterine fibroids in patients with polycystic ovary syndrome than in those with unexplained infertility.

We compared demographic characteristics for PCOS and UI patients ( Table 1 ). Patients with PCOS had a lower age (P<0.001), higher BMI (P<0.001), higher waist circumference (P<0.001), higher waist/hip ratio (P<0.001), and higher hirsutism score (P<0.001) when compared to those with UI. Patients with PCOS were more likely to be Hispanic (P<0.001), Black (P<0.001), to have a lower percentage with a professional occupation (P<0.001) and a lower annual household income (P<0.001), when compared to those with UI. Comparison of medical history and lifestyle factors between patients with PCOS and those with UI is shown in Table 2 . Patients with PCOS had a lower percentage of a prior conception (P=0.002) and pregnancy loss (P<0.001), a higher percentage of having a prior diagnosis of infertility (P<0.001), a higher percentage of having a history of high blood pressure (P<0.001), a higher percentage of current smokers (P<0.001) and a lower percentage of current alcohol users (P<0.001). Supplemental table 1 compares circulating hormones between patients with PCOS and those with UI. As expected, patients with PCOS had a higher value of total testosterone (P<0.001), free androgen index (P<0.001), fasting insulin (P<0.001), proinsulin (P<0.001), fasting glucose (P<0.001), and homeostatic model assessment-insulin resistance (HOMA-IR) than those with UI. Patients with PCOS also had lower sex hormone-binding globulin (SHBG) than those with UI (P<0.001). Table 3 shows the prevalence of women with non-cavity-distorting fibroids in patients with PCOS and those with UI. The prevalence of women with non-cavity-distorting fibroids was 6.7% (92 out of 1371) for patients with PCOS and 12.4% (109 out of 878) for patients with UI (p<0.001), with no difference existed between PPCOSI and PPCOSII patients [6.6% (41 out of 625) vs 6.8% (51 out of 746) respectively, p=0.84, data not shown]. In addition, for PPCOSII patients, no difference in the prevalence non-cavity-distorting fibroids was found between those with ovulatory dysfunction and hyperandrogenism (1990 NIH criteria PCOS) and those with ovulatory dysfunction with polycystic ovaries (non-hyperandrogenic PCOS) [6.9% (48/695, and 7.1% (3/42) respectively, p=1.00, data not shown]. When stratified by Black and non-Black race groups, the prevalence of women with non-cavity-distorting fibroids was significantly higher in patients with UI than those with PCOS in both race groups, with a prevalence of 10.7% versus 35.9% among Black women with PCOS and UI, respectively (p<0.001, Table 3 ) and 6.0% versus 10.1% in the non-Black women with PCOS and those with UI, respectively (p=0.001). Similarly, when stratified by 3 different BMI groups, the prevalence of women with non-cavity-distorting fibroids was significantly lower in patients with PCOS than those with UI in all of the groups, with a prevalence of 2.8% vs 10.6% (p<0.001), 5.7% vs 12.8% (p=0.02), and 7.8% vs 15.3% (p<0.001) for patients with PCOS vs those with UI in the BMI =30 groups, respectively ( Table 3 ). No significant difference, in the prevalence of non-cavity-distorting fibroids with any dimension >=4 cm nor the volume of the largest fibroid, existed between patients with PCOS and those with UI for the overall patients or the sub-groups with different races and various BMIs (all the p values >0.20, Table 3 ). The independent risk factors for, and the adjusted risk of PCOS on the prevalence of non-cavity-distorting fibroids were obtained by using the multivariable analysis. All the variables that survived in the final model are shown in Table 4 . After adjustment for age, waist circumference, ethnicity, race, HOMA-IR, prior therapy for infertility, and history of alcohol use, patients with PCOS showed a significantly lower prevalence of fibroids than those with UI (Odds Ratio=0.54, 95% confidence interval (CI), 0.36–0.79, p=0.002). The ROC curves, AUC, and the HL goodness fit test for the final models are shown in Supplemental figure 1 . The AUC of 0.76 and a p value of 0.87 for the HL goodness fit test indicate that the model has a good discrimination between patients with non-cavity-distorting fibroids and those without, and that the model has a good calibration. The association between PCOS and non-cavity-distorting fibroids was stronger among Black women than non-Black women ( Table 3 ), but there was no statistically significant evidence of heterogeneity across race groups (test for interaction=0.06). BMI did not remain in the final model; when forcing BMI and its interaction with patient group in the model, no evidence of heterogeneity across BMI groups was found (p=0.26 for the interaction).

This is a secondary analysis utilizing data from the PPCOS I ( http://clinicaltrials.gov : NCT00068861 ), PPCOS II ( NCT00719186 ), and AMIGOS ( NCT01044862 ) trials. PPCOS I was a multicenter, double-blind, randomized clinical trial, comparing live-birth rate in response to treatment with clomiphene citrate plus placebo, metformin plus placebo, or combination clomiphene citrate plus metformin for up to a total of five cycles ( 32 ). In brief, 626 oligoovulatory infertile women, aged 18–39 years, with PCOS diagnosed by Rotterdam criteria ( 24 ) were enrolled. Participants with other causes of infertility were excluded. Documentation of a normal uterine cavity and at least one patent fallopian tube were required prior to enrollment, as well as a semen analysis with at least 20 million sperm per milliliter for the male partners. All participants were in good health with no major medical disorders. PPCOS II was a multicenter, double-blind, randomized clinical trial comparing live-birth rate in response to treatment with clomiphene citrate or letrozole for up to a total of five cycles ( 33 ). In brief, 750 women with PCOS were enrolled. Females were between 18 and 40 years, with anovulation, combined with either hyperandrogenism or polycystic appearing ovaries by ultrasound. Documentation of tubal patency and a normal uterine cavity was required prior to enrollment, as well as a semen analysis with at least 14 million sperm per milliliter for the male partners. Key exclusion criteria included history of major medical diseases, including diabetes and cavity distorting uterine fibroids. AMIGOS was a prospective, multicenter randomized clinical trial that evaluated the outcomes of live birth and multiple gestations associated with ovarian stimulation and intrauterine insemination in couples with UI ( 34 ). In brief, 900 couples were enrolled. Participating women were between 18 and 40 years with regular ovulatory menses, had a normal uterine cavity with at least one patent fallopian tube, and had a male partner with a semen specimen with at least 5 million motile sperm in the ejaculate. Key exclusion criteria were history of major medical diseases, including diabetes, endometriosis, PCOS, and cavity distorting uterine fibroids. For all 3 studies, Institutional Review Board approval was obtained at each study site, and all participants gave written informed consents. The investigation was monitored by an independent Data and Safety Monitoring Board. Demographic characteristics, medical history, and lifestyle profiles were obtained using standardized forms. Patients also underwent Ferriman-Gallwey hirsutism scoring measurements. Fasting blood collected at baseline or screening was used for all hormonal assays. Samples were batched and analyzed at the Ligand Assay and Analysis Core Laboratory at the University of Virginia. All had acceptable quality control measures by Partnership for the Accurate Testing of Hormones standards ( 35 ). We have previously published the quality control standards of these assays and all had inter- and intra-assay coefficients of variation < 10% ( 27 , 28 ). Presence or absence of fibroids, and three dimensions of the largest fibroid (PPCOS II and AMIGOS patients only) were recorded during transvaginal ultrasound (using a 5–9 mHz vaginal probe, most centers used 5 mHz) ( 27 , 28 ). Per the original protocols, the location, total number and aggregate volume of all fibroids were not recorded. Analysis was conducted using the combined patients from PPCOS I (n=626), PPCOS II (n=750), and AMIGOS (n=900) studies. Of these 2276 patients, 27 were excluded from analysis because they had a history of surgery to remove fibroids (1 from PPCOS I study, 4 from PPCOS II study, and 22 from AMIGOS study). The main outcome of interest is the presence of non-cavity-distorting fibroids (fibroids with a normal uterine cavity), identified by transvaginal ultrasound performed at either the screening or baseline visit (before taking any study medication). Other outcomes include the presence of any fibroid with a dimension >=4 cm and the volume (cm 3 , length x width x height) of the largest fibroid. Baseline patient demographic characteristics, medical history, lifestyle factors, and circulating hormonal levels were selected as covariates. Continuous variables were expressed as median (interquartile range, IQR), with a Wilcoxon rank-sum test (also called the Mann–Whitney U test) used for testing differences between patients with PCOS and those with UI. Categorical data were reported as frequencies and percentages; differences in these measures between the two groups were assessed by a chi-square analysis, with Fisher’s exact test used for expected frequencies of less than 5. Logistic regression was used to establish a final model for the presence of fibroids. All variables were introduced into a multivariable logistic regression analysis in a stepwise fashion, using a P value of <.10 to enter and a P value of <.05 to remain. To assess the predictive/discriminative power of final models, we constructed receiver operating characteristic (ROC) curves and calculated the areas under the curve (AUC). A model with perfect discrimination has an AUC of 1.0, while a model with no discrimination has an AUC of 0.5. We also performed the Hosmer-Lemeshow (HL) test for goodness-of-fit of the final logistic regression models, with a P value of >.05 indicating a good fit of the data. In addition, for each covariate selected from the stepwise procedure, the interaction effects between the patient group and selected covariates were then assessed, and those interactions that were significant at the 0.05 level were then included in the final model. Analyses were performed with SAS software, version 9.4 (SAS Institute). Statistical significance was defined as a two-sided P value of less than 0.05.

We found that infertile patients with PCOS had a lower prevalence of non-cavity-distorting fibroids at baseline than those with UI. The association was similar between Black and non-Black women, and among women across different BMI categories. No difference existed for prevalence of non-cavity-distorting fibroid with any dimension >=4 cm nor the total volume of the largest fibroid between the two groups. This is the first study showing that patients with PCOS have a lower prevalence of non-cavity-distorting fibroids than those with UI, consistent with one previous finding that patients with PCOS had a lower percentage of fibroids when compared to those with normal ovaries ( 25 ). It differs from another previous finding that PCOS was associated with a 65% increased risk of fibroids among Black women ( 18 ). When limiting our analysis to include only the Black women, a diagnosis of PCOS was still associated with a significantly lower prevalence of non-cavity-distorting fibroids with [Odds ratio=0.26, 95% CI, 0.12–0.58, p<0.001, data not shown) or without ( Table 3 ) adjustment for other covariates. Thus, the difference in the two study findings is unlikely to be due to difference in race of study participants. It may be related to the fact that the previous study relied upon self-report to diagnose PCOS and did not systematically screen all patients for the presence of fibroids with the use of transvaginal ultrasound ( 18 ). Another possible explanation for the lack of agreement between our study and the previous report is that our study was limited to non-cavity-distorting fibroids, while the previous study considered submucosal as well as non-cavity-distorting fibroids. One advantage of our study is the systematic characterization and more accurate diagnosis of women with PCOS and the confirmation of the presence of fibroids by transvaginal ultrasound in all study participants prior to any treatment. In other words, the data analyzed herein were of higher quality. It has been shown that the size of non-cavity-distorting fibroids plays important roles in their effect on infertility; those with a size >4 cm had a lower pregnancy ( 12 ) and delivery rate ( 11 ). Our finding showed that no difference existed in the prevalence of non-cavity-distorting fibroids with any dimension >=4 cm nor the volume of the largest fibroids between patients with PCOS and those with UI, suggesting that the underlying mechanisms for UI may only influence the occurrence but not the growth/development of the non-cavity-distorting fibroids, and thus may have limited impact on pregnancy outcome when compared to those with PCOS. Since the patients in the original studies were highly selected, it would also be possible that most of the patients with large fibroid size were excluded from the studies and thus a small number of patients with fibroids >=4 cm would have limited the power to detect any difference ( Table 3 ). Since the information about the total number and the aggregate volume of all fibroids was not available, whether the overall fibroid burden was different between the two groups is unclear and warrants future investigation. The reason for a lower prevalence of non-cavity distorting fibroids in the women with PCOS compared to those with UI is unknown. Patients with PCOS have anovulation, with resultant limited exposure of uterine myometrium to progesterone ( 36 ). Progesterone has been shown to stimulate the growth of leiomyoma through a set of key genes that regulate both apoptosis and proliferation ( 37 , 38 ). Thus, the reduced exposure of women with PCOS to progesterone may explain our observed findings. Consistent with these observations we found that women with previous treatment for infertility, with probable increased ovulation rates and progestin exposure, had a significantly higher prevalence of fibroids than in those not having previous infertility treatment [Adjusted OR=1.53 (CI 1.10–2.13)] ( Table 4 ). Additionally hyperinsulinemia, insulin resistance, increased level of LH, diabetes, and hypertension ( 23 , 24 ) are associated with PCOS. Previous findings that hyperinsulinemia and diabetes were inversely associated with fibroids ( 16 – 18 ) were consistent with our findings. However, our results contradict other previous reports. For example, other investigators reported that an increase in LH levels was associated with increased fibroid development ( 15 ) which we did not find (data not shown); hypertension was reportedly associated with an increased risk of fibroids ( 21 , 22 ) but we did not observe this association. Our findings that higher waist circumference and HOMA-IR were associated with an increased risk of non-cavity-distorting fibroids ( Table 4 ) suggests that hyperinsulinemia is not the reason for the decreased prevalence of non-cavity-distorting fibroids observed in patients with PCOS. Although the prevalence and severity of fibroids were different between Black and non-Black women ( 13 , 29 ), and PCOS characteristics may differ between lean and obese women ( 30 , 31 ), our results that patients with PCOS was associated with a lower prevalence of non-cavity-distorting fibroids than those with UI was not significantly different between Black and non-Black women, nor among the women with different BMIs, suggests that no evidence of heterogeneity of the observed association existed across different race and BMI groups. Our findings that older age and Black race were associated with higher prevalence of non-cavity-distorting fibroids ( Table 4 ) are consistent with the work of others ( 4 , 13 , 16 , 29 , 39 ). In addition, we showed that previous or current use of alcohol was independently associated with a decreased prevalence of non-cavity-distorting fibroids, which differs from prior findings that alcohol use was associated with an increased prevalence of fibroids ( 40 – 42 ). The underlying mechanisms for this association are unknown. The present study has several strengths and weaknesses. Among the strengths is the large sample size, and multicenter design of our three studies, which represent the largest US-based cohort of infertile couples studied to date. The complete characterization of the women with demographics, medical history, lab tests, and ultrasound findings are additional strengths. One weakness of our study is that women with abnormal uterine cavity were excluded from our original studies and that we excluded any patients who had a prior surgery to remove fibroids. How the exclusion of these patients affects the association that we have identified is unknown; however this exclusion was applied across all women with either PCOS or UI. Another weakness is the fact that the three included trials were not conducted at the same time period; PPCOSI was conducted much earlier than PPCOSII and AMIGOS trials. Also difference may exist in ultrasound equipment and the data interpretation among sonographers. However, a standard protocol was used for obtaining sonographic measurements for PPCOSII and AMIGOS studies; more than half of the subjects were recruited at the same sites for all three studies and 88% of subjects were recruited at the same sites for both PPCOSII and AMIGOS studies. The very similar prevalence of non-cavity-distorting fibroids between PPCOSI and PPCOSII patients suggests it is unlikely that the lower prevalence of fibroids in PCOS patients was due to the different time periods when the trials were conducted and the difference in ultrasound equipment and the data interpretation among sonographers. In addition, participants for all the original studies were all infertile patients. The results may be different for reproductive-aged fertile women; thus firm conclusions about how representative our patients are to the general population should be made with caution. In summary, this study showed that infertile oligoovulatory women with PCOS had a reduced prevalence of non-cavity-distorting fibroids compared to regularly ovulating women with unexplained infertility. This association was not different with regard to Black and non-Black race groups, nor the patients with different BMIs. With a high prevalence of both PCOS and fibroids in reproductive-age women, future studies are needed to further evaluate the association between PCOS, fibroids and infertility in a more general population.

Uterine leiomyomata (fibroids) are benign monoclonal tumors that develop from the myometrium of the uterus. They are the most common tumors of the female reproductive tract, clinically apparent in 20% to 40% of women in their reproductive years ( 1 , 2 ). Fibroids are heterogeneous in pathophysiology, size, location, and clinical symptomatology ( 3 , 4 ). Intramural and/or subserosal fibroids that do not distort the uterine cavity can be asymptomatic or symptomatic ( 4 ); their effect on pregnancy outcomes remains poorly understood with conflicting results. Some studies showed that no difference existed in pregnancy outcome between patients with and without non-cavity-distorting fibroids ( 5 – 7 ). In contrast, many studies reported that uterine fibroids that do not distort the uterine cavity were associated with a decreased implantation rate and pregnancy rate ( 8 – 10 ), especially for the intramural fibroids with a size >4 cm ( 11 , 12 ). Many factors have been associated with the occurrence of fibroids (regardless of cavity-or non-cavity-distorting fibroids), including advancing reproductive age, Black race, obesity, and nulliparity ( 13 , 14 ). Recent studies showed that some hormonal and metabolic factors are also associated with the development of fibroids ( 4 ). While increasing luteinizing hormone (LH) levels were associated with increased fibroid development ( 15 ); hyperinsulinemia and diabetes were inversely associated with the prevalence fibroids ( 16 – 18 ) in some studies. However, other investigators have hypothesized that insulin stimulates fibroid growth ( 19 , 20 ). Hypertension, which often accompanies hyperinsulinemia in conditions such as the metabolic syndrome, has been associated with an increased risk of fibroids ( 21 , 22 ). All of these attributes: hyperinsulinemia, insulin resistance, increased LH, and hypertension are common in patients with polycystic ovary syndrome (PCOS) ( 23 , 24 ), which affects 5 to 10% of reproductive age women and is the most common cause of anovulatory infertility. Whether PCOS is associated with an increased or decreased risk of fibroids has received scant attention in the medical literature, with one study indicating a 65% increased risk of fibroids among Black women with PCOS when compared to those without PCOS ( 18 ) and another study showing a decreased risk for fibroids for patients with PCOS when compared to those with normal ovaries ( 25 ). The National Institute of Child Health and Human Development Cooperative Reproductive Medicine Network Pregnancy in Polycystic Ovary Syndrome I (PPCOS I) ( 26 ), Pregnancy in Polycystic Ovary Syndrome II (PPCOS II) ( 27 ), and Assessing Multiple Intrauterine Gestations from Ovarian Stimulation (AMIGOS) ( 28 ) studies provide a unique opportunity to investigate whether an association exists in women with PCOS and non-cavity-distorting fibroids. All patients enrolled in PPCOS I and PPCOS II were infertile oligoovulatory women with PCOS; women enrolled in AMIGOS were diagnosed with unexplained infertility (UI), and served as a control group. As an inclusion criterion, these three studies required every patient to have a normal uterine cavity at the time of enrollment. However, a small number of patients underwent uterine surgeries to remove fibroids before enrollment, and subsequently they met this inclusion criterion by the time of enrollment. To account for the possibility that this subgroup of patients might be different from the remaining patients, the current analysis excluded all females with any history prior to enrollment of fibroids which distorted the uterine cavity. Here, we focus on the comparison of the prevalence of uterine intramural and/or subserosal (non-cavity distorting) fibroids between infertile patients with PCOS and those with UI. Since a much higher prevalence and greater severity of fibroids have been consistently observed in African American women ( 13 , 29 ), and PCOS characteristics differ in lean and obese women ( 30 , 31 ), we stratified our analysis by race and various body mass index (BMI) groups. In addition, since the size of the fibroid may play a role in its effect on infertility ( 11 , 12 ), we compared the prevalence of the non-cavity-distorting fibroids with any dimension >=4 cm and the total volume of the largest fibroid between the two groups.

Supplemental figure 1. Receiver operating characteristic (ROC) curves and areas under the curve (AUC) of the final model for the presence of women with non-cavity-distorting fibroids. P=0.87 for the Hosmer and Lemeshow (HL) goodness-of-fit test.