Impact of endometriosis on in vitro fertilization outcomes: an evaluation of the Society for Assisted Reproductive Technologies Database

Endometriosis is a chronic benign gynecologic disease that affects 10% of women, and is a major cause of chronic pelvic pain and infertility ( 1 , 2 ). Anatomic distortion leading to tubal occlusion, poor oocyte quality, impaired implantation, and progesterone resistance have all been implicated; however, the mechanisms of endometriosis-associated infertility remain incompletely understood. A number of observational studies have sought to determine the effects of endometriosis on pregnancy rates, with some reporting negative associations, and others noting no association. A meta-analysis of the available observational data in 2002 suggested that patients with endometriosis-associated in fertility undergoing IVF had an absolute pregnancy rate (detection of serum human chorionic gonadotropin) almost half that of other diagnostic groups, with similar trends in other surrogate markers of IVF success including oocyte yield, fertilization rate, and implantation rate ( 3 ). However, more recent analyses have suggested that a diagnosis of endometriosis may be associated with comparable pregnancy outcomes compared to other infertility diagnostic groups. In a retrospective analysis of linked cycles from the SART database over a 7 year period, live birth rates were similar to other IVF diagnostic groups in both fresh and frozen cycles ( 4 ). However, this particular analysis reported on endometriosis as a single diagnosis. As it is more typical for endometriosis to present in conjunction with other diagnoses than in isolation, this can complicate counseling patients regarding IVF outcomes. To reconcile this controversy, the proposed study sought to assess the relationship between a diagnosis of endometriosis, either in isolation or in combination with other infertility diagnoses, and in-vitro fertilization outcomes using population level data from the Society for Assisted Reproductive Technologies Database with the hypothesis that endometriosis would be associated with lower live birth rates compared to other diagnostic groups particularly in endometriosis with concomitant infertility diagnoses.

This is a population-based retrospective study of subjects from the Society for Assisted Reproductive Technology’s national database from 2008-2010 representing the IVF cycles from >85% of infertility clinics in the United States. This study proposal was reviewed by the Institutional Review Board at the University of Pennsylvania and was deemed appropriate for full IRB review exemption due to use of de-identified data. Cycles were analyzed according to reported infertility diagnosis with endometriosis as the exposure of interest. Cycles were categorized as those having an isolated diagnosis of endometriosis (“Endometriosis Only”), endometriosis plus at least one other concomitant diagnosis (“Endometriosis Plus”), an isolated diagnosis of tubal factor infertility (“Tubal Factor”), or an isolated diagnosis unexplained infertility (“Unexplained”). Patients for whom the reason for infertility was a diagnosis other than endometriosis, tubal factor, or unexplained infertility (including those listed as “Other,” or “Other Noninfertile” with additional explanatory comments that excluded endometriosis or tubal factor) were classified as “All Other Diagnoses”. All donor, gestational carrier, and banking cycles were excluded. The primary outcome of interest was live birth rate, defined as delivering a liveborn infant after 22 weeks gestation. Secondary outcomes included oocyte yield, fertilization rate (number of embryos/oocyte yield), proportion of cycles resulting in blastocyst transfer, implantation rate (number of fetal hearts with detectable activity/number of embryos transferred), and early pregnancy loss rate (biochemical pregnancy, ectopic pregnancy, or miscarriage – [clinical intrauterine gestation resulting in pregnancy loss or abortion]). Analyses of oocyte yield were restricted to fresh cycles, and analyses of fertilization rate were restricted to fresh cycles with an oocyte yield >=1. All other analyses were restricted to those cycles in which an embryo transfer was performed to reduce bias from cancelled cycles due to inadequate response. Baseline and demographic characteristics were analyzed with ANOVA and Pearson chi-square testing as appropriate. Generalized linear regression models were used for multivariable modeling while analysis of count data and implantation rate was performed using poisson regression( 5 ). Comparisons of oocyte yield and fertilization rate were performed using negative binomial regression modeling to account for excess variability (over dispersion) in the rates. All other outcomes were analyzed with logistic regression modeling using backwards elimination. Models of proportion of blastocyst transfer, implantation rate, and pregnancy outcomes evaluated the potential for effect modification by cycle type (fresh versus frozen) with adjustment for significant confounders. Mixed cycles (those with both fresh and frozen embryos transferred) were excluded. Maternal age, body mass index (BMI), race, smoking history, number of prior treatment cycles, maximum follicle stimulating hormone level, prior parity, use of intracytoplasmic sperm injection (ICSI), assisted hatching, and year of treatment were considered as potential confounders in the relationships between infertility diagnosis and the outcomes of interest as appropriate. Missing data were given separate categorical indicators within each covariate for analysis to account for the effects of missing data. An a priori subanalysis of first IVF cycles was considered to address the influence of multiple or prior treatment cycles. To account for the influence of multiple comparisons and the impact of a large number of observations in this dataset, a p-value<0.001 was considered statistically significant. All data were analyzed using STATA v 12.0 (StataCorp, College Station, TX, USA).

Of the 400,059 cycles reported from 2008-2010, 347,185 were included in the analyses after excluding all donor, gestational-carrier, and banking cycles. There were 39,356 initiated cycles of patients with endometriosis, which comprised 11% of the study sample. Of these, 14,053 cycles (4%) were in women who had an isolated diagnosis of endometriosis (Endometriosis Only), while 25,303 cycles (7%) were in women who had a diagnosis of endometriosis and at least one additional diagnosis (Endometriosis Plus). Isolated tubal factor infertility (Tubal Factor) was representative of 25,906 cycles (7%), and 44,200 cycles (12.7%) were classified as unexplained infertility (Unexplained). Table 1 summarizes the characteristics of each diagnostic group. Notable differences include that women with isolated endometriosis were younger than other diagnostic groups. Those with tubal factor infertility had a higher BMI and were more likely to report African American race. Women with endometriosis in combination with another diagnosis (Endometriosis Plus) were more likely to have undergone a flare protocol for ovarian stimulation, ICSI, assisted hatching, and more likely to have had at least 1 prior IVF cycle. Of women with Endometriosis Plus, 65% had a single additional infertility diagnosis, 28% had 2 additional diagnoses, 6% had 3 additional diagnoses and the remaining 1% had 4 or more diagnoses reported. The distribution of concomitant diagnoses associated with endometriosis is presented in the Appendix ( Supplement Table 1 ). A total of 291,244 embryo transfer cycles were analyzed of which 11.5% had endometriosis: 4.2% (12,335) had an isolated endometriosis diagnosis (Endometriosis Only) and 7.3% (21,223) had endometriosis in combination with other diagnoses (Endometriosis Plus). Women with endometriosis and concomitant diagnoses were significantly more likely to have a cancelled cycle (11.3%) in comparison to women with isolated endometriosis, tubal factor, or unexplained infertility (8.5%, 8.3%, 8.1% respectively, p<0.0001). Ovarian stimulation and pregnancy outcomes for women based on diagnosis are reported in Table 2 . Overall, women with a diagnosis of endometriosis (either isolated endometriosis or concomitant with another diagnosis) had a reduction in oocyte yield (RR 0.91 [0.91-0.92]), implantation rate (RR 0.94 [0.93-0.96]), proportion of blastocyst transfer (RR 0.96 [0.93-0.99]) and a 6% reduction in in live birth rate compared to women without endometriosis in adjusted analyses of fresh cycles (RR 0.94 [0.91-0.97]). Fertilization rate was similar (RR 1.00 [1.00-1.01]). Implantation rate and live birth rate followed similar trends of poorer outcomes in endometriosis compared to those without endometriosis in frozen/thawed transfer cycles. We further examined the outcomes of women with isolated endometriosis (Endometriosis Only) and those with a concomitant infertility diagnosis (Endometriosis Plus) separately to test our a priori hypothesis that outcomes may differ in each subgroup and thus explain the differences in the findings of prior studies. These data are presented in Tables 3 and 4 . Table 3 presents the IVF outcomes of women with isolated endometriosis to women with other diagnoses. While not found in all comparisons to all subgroups, an isolated diagnosis of endometriosis (Endometriosis Only) was generally associated with a decrease in oocyte yield, a slightly lower or similar fertilization rate, blastocyst transfer rate, and pregnancy loss rate. However, women with isolated endometriosis were found to have a similar or higher liver birth rate compared to those with other infertility diagnoses. These findings were similar in fresh and frozen embryo transfer cycles. Table 4 presents the IVF outcomes of women with endometriosis and at least one other concomitant diagnosis (Endometriosis Plus) in comparison to other diagnostic groups. This subgroup was noted to have significantly poorer IVF outcomes compared to women with other infertility diagnoses. Oocyte yield was consistently 7-9% lower compared to unexplained, tubal factor, and all other diagnostic groups. Despite similar fertilization rates and blastocyst transfer rates, there was 11 – 17% reduction in implantation rates in Endometriosis Plus compared to unexplained infertility, tubal factor, and all other diagnostic groups combined. Live birth rates were reduced by 19-26% in fresh cycles. Trends were similar in frozen cycles, noting a 12-18% reduction in live birth rates. A restricted analysis of first cycles demonstrated no significant difference in reported trends in oocyte yield, implantation rate, or live birth rate. Subanalyses of mechanisms of early pregnancy loss demonstrated no differences in incidence of ectopic pregnancy, biochemical pregnancy, or miscarriage among women with Endometriosis Only compared to tubal factor and unexplained infertility in fresh and frozen cycles (p>0.05 for all comparisons). Interestingly, amongst those with a positive pregnancy test, those with Endometriosis Plus were significantly more likely to have a biochemical pregnancy or miscarriage compared to those with Unexplained Infertility and Tubal Factor Infertility in fresh cycles (RR 1.26 [1.16-1.37] p<0.0001 and RR 1.19 [1.08-1.31] p<0.0001, respectively).

The impact of Endometriosis on fecundity and IVF outcomes continues to be debated. This study confirms that endometriosis is associated with lower live birth rates than other infertility diagnoses, and specifically compared to tubal factor and unexplained infertility. As previously noted, this difference is more apparent after adjusted, rather than crude, analysis of the data ( 3 ). This may explain the discrepancy between these findings and the Clinic Summary Report of the Society of Assisted Reproductive Technology. However, the association of endometriosis and IVF outcomes is confounded by other infertility diagnoses. The majority of couples with a diagnosis of endometriosis presenting for IVF will have at least one other infertility diagnosis which contributes to clinical challenge of assessing the impact of a sole diagnosis on IVF outcomes. In fact, our analysis demonstrates that when endometriosis was seen in isolation, it was associated with a similar or even higher live birth rate compared to all other diagnoses in fresh autologous cycles, despite lower oocyte yield. These patients may represent a milder phenotype of endometriosis and thus may have a more favorable response to the specific benefits that IVF avails, including optimizing oocyte-sperm interaction outside the inflammatory peritoneal environment, and progesterone supplementation to overcome relative progesterone resistance. Importantly, this only applies to a minority of all patients presenting with endometriosis. Endometriosis more commonly presents in conjunction with at least one other fertility diagnosis, and as this larger subgroup of patients have poorer prognoses overall, as evidenced by the higher likelihood of prior IVF cycles and more aggressive (flare) stimulation protocols. The mechanism of endometriosis-related infertility, or its impact on IVF, has not been fully established ( 6 ). As endometriosis is a chronic and often progressive disease, it possible that as the disease advances it will result in alterations that will be categorized as other infertility related diagnoses. Thus, it is possible that women with isolated endometriosis represent a subgroup of women with “mild” disease. If so, these results are similar to previous findings ( 3 ) and those of a recent meta-analysis ( 7 ). In the latter, a 21% reduction in both implantation and clinical pregnancy rates in those with Stage III-IV endometriosis was noted (RR=0.79, 95% CI [0.67-0.93], p=0.0006 and RR 0.79 [0.65-0.91], p=0.0008, respectively), but no difference in live birth rates in Stage I-II or Stage III-IV endometriosis was observed (RR 0.92 [0.83-1.02] p=0.10 and RR 0.86 [0.68-1.08] p=0.19, respectively) ( 7 ). Similarly, a large single center cohort study spanning a 20 year period of autologous GnRH agonist cycles in Norway noted that cycles with endometriosis were associated with similar cumulative live birth rates compared to tubal factor infertility (Stage I-II 73%, 95% CI [58-75%]; Stage III-IV 58%, [22-94%], tubal factor 66%, [58-75%]) demonstrating clear heterogeneity within the population of those with endometriosis by stage ( 8 ). A possible mechanism of lower live birth rates amongst those with endometriosis appears to be linked to oocyte quality, as reflected by lower oocyte yield, as well as impaired implantation. Diminished ovarian reserve was a highly prevalent concomitant diagnosis in the Endometriosis Plus group. These women had a higher rate of cancellation compared to those with tubal factor and unexplained infertility, confirming poorer prognosis for this subgroup overall. The link between diminished ovarian reserve and endometriosis has been suggested by studies noting lower serum markers of ovarian reserve in patients with endometriosis compared to tubal factor infertility ( 9 , 10 ). In a retrospective cohort study of autologous and donor oocyte cycles in patients with endometriosis compared to other diagnoses, endometriosis was similarly associated with a lower pregnancy rate per transfer (p<0.0004) and implantation rate (p<0.0003) compared to tubal factor infertility ( 11 ). Furthermore, when analyzing the impact of endometriosis on uterine environment in oocyte donation cycles, there was a lower implantation rate after transfer of embryos from endometriotic ovaries into women without endometriosis, whereas there was no difference in pregnancy rates between women with endometriosis and tubal factor receiving donor oocytes suggesting that oocyte quality and not the uterine environment is the main contributor to lower pregnancy rates ( 11 ). These findings were later corroborated by a case-control analysis from the same group ( 12 ); however, these studies did not differentiate the contribution of the oocyte from the embryo. Our data does not suggest that endometriosis has a large impact on embryo progression to blastocyst as a surrogate of embryo quality. We noted that the rate of blastocyst transfer were similar if not higher in the Endometriosis Plus group compared to other diagnostic groups despite lower oocyte yield. However, the embryo-endometrial interaction, and subsequent impact on implantation rate may be associated with a reduction in live birth rate noted. Possible mechanisms for this finding may include altered HOXA10 gene expression ( 13 , 14 ), altered endometrial receptivity ( 15 - 17 ), and/or progesterone resistance ( 18 - 21 ). We note that the frozen embryo transfer cycles resulted in a lower pregnancy rates across all diagnostic subgroups compared to fresh cycles and did not differ in those with endometriosis compared to those without endometriosis overall. However, the majority of the cycles included in this analysis were fresh cycles; as such, caution should be taken in extrapolating these results to suggest inherent differences in fresh and frozen cycles, as the retrospective nature of this analyses is most certainly subject to selection bias with respect to cycle type (fresh versus frozen). These observations are likely due to the routine practice of selecting the best quality embryos for fresh transfer and cryopreserving supernumerary embryos with high frozen blastocyst transfer rates reflecting selective blastocyst cryopreservation. As such, the impact of peri-implantation environment would be better ascertained in a prospective controlled study. This study utilized the SART Database to capture infertility diagnosis and outcome data at a population level given that the over 345,000 cycles included represent the majority of IVF practices in the US over a 3 year period. While the size of this study strengthens the conclusions drawn and using a national database lends generalizability, we acknowledge that the findings may still be affected by confounding and bias. By accounting for relevant confounders including age, prior parity, FSH, prior cycles, micromanipulation, and year of treatment, this study design and analysis allowed for a conservative method of analyzing this population-based data to reduce the risk of overstating conclusions. We acknowledge that this study is limited by information bias and the abstracted data did not have identifiers for linking cycles within an individual. As patients in the Endometriosis Plus group were more likely to have had prior cycles represented within the dataset, this may have resulted in bias of the reported results away from the null. However, Kalra et. al. ( 22 ) were able to link data from multiple cycles per woman and estimated the within woman correlation of multiple cycles to be nearly zero (Sarah Ratcliffe personal communication) given that the majority of women contributed 1 cycle (59% in their study, 47% reported no prior cycles in the present study) we believe the impact of linking multiple cycles would have a negligible effect on our conclusions. As stage of endometriosis, the presence or absence of endometriomas, and prior interventions for endometriosis is not universally reported, the impact of disease severity and endometriosis treatment on IVF outcomes cannot be completely ascertained from this analysis. There is theoretical risk of diagnostic misclassification with respect to the endometriosis only and tubal factor only groups when using administrative data ( 21 ). However, IVF centers are able to report multiple SART diagnoses, (as seen by 29% of the Endometriosis Plus group reporting concomitant tubal factor). Thus any misclassification is likely non-differential, resulting in a bias toward the null. Of note, there is the possibility of diagnostic misclassification such that some of those with unexplained infertility may have undiagnosed endometriosis given the shift in clinical care away from routine diagnostic laparoscopy for all infertility patients. As such, misclassification could be differential, or unidirectional. It is unknown if correct diagnostic classification would result in bias towards the null or perhaps, an even more dramatic reduction in the live birth rates observed.

Endometriosis is a heterogeneous disease with respect to presentation and outcomes in those with infertility. IVF undeniably remains one of the most effective treatments for women with endometriosis-associated infertility - yet, there are nuances of this complex disease process that are important for counseling patients with respect to expected IVF outcomes. In general, endometriosis is associated with lower oocyte yield, lower implantation rates, and lower pregnancy rates. Endometriosis, when associated with other alterations in the reproductive tract (either as a result of progression, or by chance) has the lowest chance of live birth. In contrast, for the minority of women who have endometriosis in isolation, the live birth rate is similar or slightly higher compared to other diagnostic groups. Further studies are needed to assess the role of peri-implantation environment and endometrial receptivity, to understand the mechanism(s) of endometriosis-associated infertility, and how it may be overcome.