Assisted reproductive technology treatment increases obstetric and neonatal risks over that of the underlying infertility diagnosis

The study sample included singleton deliveries for women whose deliveries between October 1, 2013 and December 31, 2017 were linked to APCD but who at no time in that 4 year period had Medicaid (MassHealth). A comparison of singletons only allowed us to avoid the complications introduced by multiple pregnancy. Exclusion of MassHealth patients helped to ensure a more homogeneous population and minimize differences between the fertile and other groups. We included deliveries from October 1, 2013 rather than January of 2013 to allow us to have 9 months of APCD data in which to find an infertility code in our dataset prior to delivery if one existed. This retrospective study used information from the 1) SART CORS, 2) PELL, and 3) APCD data systems. The study was approved by the Massachusetts Department of Public Health (MDPH) and the Dartmouth-Hitchcock Health Institutional Review Board. A Memorandum of Understanding was executed among SART, MDPH, and the project principal investigators. The SART CORS database collects cycle-based data from ART clinics for reporting to the Centers for Disease Control and Prevention in compliance with the Fertility Clinic Success Rate and Certification Act of 1992 (Public Law 102–493). Cycle specific demographics, infertility diagnoses, ART treatment, and outcome data are collected by Redshift Technologies, Inc. under contract to SART. Approximately 90% of ART clinics in the USA and all Massachusetts clinics are included in the database. SART CORS data are validated annually by assessment of a random selection of clinics that have on-site visits for chart review. During these visits, data reported by the clinic are compared with information recorded in patients’ charts. In 2017, most data fields selected for validation were found to have discrepancy rates of ≤5% ( 16 ). PELL links Massachusetts birth certificates and fetal death records to corresponding hospital utilization data for the delivery and for longitudinal non-delivery hospital admissions, observational stays, and emergency room visits for both the mother and child. Since 1998, data have been linked for 98% of the births and fetal deaths for individual women and their children. The PELL linkage is performed using names, dates of birth, zip codes, infant sex, and medical record numbers. Linked records are identified through randomly generated unique IDs for mothers and infants. PELL data are housed at MDPH under the custodianship of MDPH and the Center for Health Information and Analysis (CHIA). The APCD is a repository for insurance claims from the majority of public and private insurance payers that provide insurance to Massachusetts residents and employees. APCD claims include those for medical, pharmacy, dental, vision, behavioral health, and specialty services. APCD claims data are available for women who delivered after January 1, 2013. The Massachusetts Outcome Study of Assisted Reproductive Technology (MOSART) database was formed by linking the SART CORS and PELL data systems for all Massachusetts resident women delivering in Massachusetts hospitals for deliveries from 2004–2017. Linkage was performed using a deterministic five phase linkage algorithm based on mother’s date of birth, her first name and last name, father/partner’s last name, baby’s date of birth, plurality, and infant sex ( 17 ). The linkage rate for 2004–2017 data was 91.5% overall and 94.9% for deliveries in which both mother’s zip code and clinic were located in Massachusetts. We linked MOSART data from 2013 through 2017 to the APCD under an MOU among CHIA, MDPH, and the project PIs as previously described ( 18 ). PELL variables for women from the MOSART database were submitted to CHIA for linkage. Variables included mother’s date of birth, last and first names, and zip codes. CHIA matched and then extracted the APCD non-MassHealth medical claim records for the linked mothers. Overall, 98.7% of the MOSART mothers were linked to the APCD of which, 81% had at least one APCD entry. We did not seek approval from MassHealth (the Massachusetts Medicaid program) to obtain APCD insurance claims data and thus MassHealth claims were not included in the data. The 10.3% of companies that did not enter data into APCD due to a lawsuit (SCOTUS Gobeille v. Liberty Mutual Insurance Company Decision, 2016) could also not be included. Deliveries were classified as ART-treated if they were linked to an ART cycle in the SART CORS database. Deliveries were defined by International Classification of Disease (ICD) 9 or ICD 10 code as having tubal disease, polycystic ovarian syndrome, other ovulatory disorders (this group did not include diminished ovarian reserve), or endometriosis ( Supplemental Table 1 ), if there had been an inpatient or outpatient claim for the diagnosis in APCD prior to the delivery date or, for ART treated women, if they had been designated as having one of these diagnoses in SART CORS as the reason for their infertility (rfa fields). Women were not excluded from a group for having more than one diagnosis and therefore some were included in more than one group. The fertile group contained deliveries that were in none of the above groups and that also did not contain subfertile or infertile women as defined previously ( 18 – 19 ). Briefly, subfertile deliveries were those that had one or more of the following 1) a marked checkbox for infertility treatment on the birth or death certificate, 2) an ICD9 or 10 code for infertility (ICD 9 codes 628 and V230; ICD 10 O09.00-O09.03 and N97.0-N97.9) during a prior hospitalization, 3) prior delivery with either a checkbox for infertility treatment or linkage to SART CORS. Infertile deliveries were defined by having had a provider confirmed diagnosis of infertility (see codes above) in an APCD outpatient or inpatient claim between LMP and prior to that delivery. Outcome measures included pregnancy hypertension, eclampsia/preeclampsia, gestational diabetes, and placental problems as well as delivery method and delivery complications of low birthweight, and prematurity. Pregnancy hypertension, preeclampsia/eclampsia, gestational diabetes, and placental problems (abruptio placenta, placenta previa, vasa previa, and placenta accreta) were obtained from birth certificates, death certificates and hospital discharge records. Birthweight and gestational age were obtained from birth certificates. Clinically determined gestational age was modified, when needed, by reported dates of last menstrual period with gestational ages outside of the range of 17–44 weeks set as missing. Maternal and paternal age, race/ethnicity, and education and maternal BMI, prior gravidity, insurance at delivery, parity and infant sex were obtained from birth certificate and fetal death records. Chronic hypertension and diabetes were obtained from a combination of birth certificates and hospital discharge data as were pregnancy bleeding, fetal distress, and PROM. Diagnosis of infertility was determined by having a defined infertility diagnosis in APCD. We used generalizing estimating equations (GEE) with Poisson distribution and exchangeable correlation structure to account for multiple deliveries by the same women and to estimate relative risk ratios (RRs) and 95% confidence intervals (CIs). Models were adjusted for mother’s age (< 30, 31–34, ≥ 35), race/ethnicity (NHW, Others, unknown), education (HS or =2), and BMI (<18.5, 18.5–24.9, 25–29.9, ≥ 30, missing). For delivery outcomes (low birth weight, prematurity) records with missing data were excluded from analysis. Pregnancy complications and neonatal deaths were coded as ‘No’ if no event was reported. Where necessary, categories were combined to enable convergence of the models. Analyses were performed in SAS software 14.3 (SAS Institute, Cary NC). Guidelines from CHIA required suppression of counts that were less than 11.

There were 80,333 singleton deliveries in the study sample of which 4,061 were ART treated. There were 2,454 deliveries to women with tubal disease, 3,916 to those with PCOS, 19,184 to those with other ovulatory conditions, and 1,585 to those with endometriosis with some women appearing in more than one group. Table 1 demonstrates that both maternal and paternal age were older in couples with ART treatment and women in all diagnostic groups had more obesity than those in the fertile group. There was more chronic hypertension among women in all diagnostic groups and more in the ART than in their respective non-ART groups. There were only minor differences in other characteristics with the exception of parity which was lower in all ART groups and in some groups with infertility diagnoses than in the fertile group. Over 95% of all women in the ART treated groups had a defined diagnosis of infertility in APCD whereas between 32% and 57% of those in the non-ART groups for each diagnosis had a defined diagnosis ( Table 1 ). Table 2 shows the pregnancy and delivery complications for each diagnostic group with and without ART treatment. All diagnostic groups had higher incidence of cesarean section than the fertile group and all ART-treated groups had higher incidence than their respective non-ART group. Pregnancy hypertension and eclampsia/preeclamsia showed similar trends for all but the tubal group. The percentage of patients with gestational diabetes was higher in the ART than non-ART groups in all but the PCOS group. Incidence of bleeding problems during pregnancy were elevated in the ART groups. The infant and delivery characteristics showing lower rates of preterm delivery and low birthweight in the fertile group, are shown in Table 3 . Adjusted relative risk for ART vs fertile, non-ART vs fertile and ART vs non-ART are shown in Table 4 for each diagnostic group. The risk of placental problems was higher in ART-treated women in all diagnostic groups when compared with either the fertile or the non-ART deliveries with the same diagnosis. The only group in which the non-ART deliveries had a greater risk of placental problems than the fertile group was that of tubal disease. Both non-ART and ART-treated PCOS patients had a greater risk of pregnancy hypertension, eclampsia/preeclampsia, and pregnancy diabetes than the fertile group but ART in PCOS patients did not increase the risk of eclampsia/preeclampsia or gestational diabetes as compared with the non-ART group with PCOS. Patients with other ovulatory problems had elevations in risk for all of these conditions with the exception of pregnancy hypertension and endometriosis patients who had elevations only among ART-treated women for all but pregnancy diabetes. There was a greater risk of low birthweight and prematurity in the ART-treated group compared to the fertile group for all diagnoses except for endometriosis which did not differ with regard to birthweight.

This study demonstrated an increase in pregnancy and delivery associated risks in singleton pregnancies conceived to women with a history of one of four infertility-related diagnoses: tubal disease, PCOS, other ovulatory conditions, or endometriosis as well as a further increase in risk as the result of ART treatment. The increased risk for the diagnostic groups alone without ART included those for hypertension, preeclampsia/eclampsia, placental problems, preterm delivery, and low birthweight. We found that for many of these outcomes, particularly that of placental problems, receiving ART treatment was associated with greater risk than that to women with the same diagnosis who conceived without ART treatment. Although multiple studies have shown that ART-conceived singleton pregnancies and deliveries are at higher risk than those to untreated fertile women ( 1 – 8 ), there has long been a debate as to whether the risks seen in singleton deliveries following ART are a function of the ART treatment or the underlying infertility ( 13 ). Studies of this issue have been complicated by the large number of differing diagnoses and underlying conditions that constitute the infertility which leads to ART treatment. Despite this, previous studies have been consistent in showing that both ART and other measures of infertility, subfertility, or non-ART medically assisted reproduction such as intrauterine insemination are all associated with an increase in risk in singleton pregnancies and deliveries over that of fertile women. ( 4 – 5 , 8 – 12 ). One intriguing paper questioning this paradigm is that of Libby et. al ( 20 ) that used the SART CORS database to compare the delivery risks of women undergoing ART for infertility to those undergoing it for prior tubal sterilization only. The authors found no difference in gestational age or birthweight in deliveries to these groups suggesting that ART alone contributes to the adverse outcomes. Unfortunately, that study had no fertile comparison group, but the suggestion that ART alone is a major contributor to adverse outcome remains. In a previous study we attempted to address the debate about whether underlying infertility or ART treatment affects outcomes by comparing ART and non-ART deliveries among diagnoses including endometriosis and ovulation disorders ( 14 ). Although risks were shown to increase in both the ART and non-ART groups, the study was limited by the fact that non-ART deliveries were identified by having had a previous hospital stay rather than a diagnosis in an outpatient office setting. Further, ART and non-ART were compared to the fertile group only and not directly to each other. The current study was designed to improve upon the prior analysis by using both inpatient and outpatient APCD claims to identify and then compare ART and non-ART deliveries in each of 4 diagnostic groups. We found, as expected, that among patients with tubal disease, PCOS, other ovulatory problems and endometriosis both ART-treated and non-ART groups showed increases in risk when compared to fertile deliveries. Further the comparison of ART to non-ART within these diagnostic groups demonstrated that ART increased the risk for many outcomes. It is important to note that, consistent with our hypothesis, even in the groups not treated with ART (the non-ART groups), there was increased risk for adverse outcomes over that seen in deliveries in the fertile group ( Table 4 ). For example, PCOS patients had increased risk of pregnancy hypertension, preeclampsia/eclampsia and pregnancy diabetes. This is consistent with previous studies that suggest a possible link between pregnancy diabetes and PCOS ( 21 – 23 ). PCOS was also associated with more preterm delivery and low birthweight. The diagnosis of other ovulatory conditions had elevated risk of preeclampsia/eclampsia and pregnancy diabetes while risk of eclampsia/preeclampsia for endometriosis patients did not reach statistical significance (ARR 1.28, 95% CI 0.99–1.64). Whether these risks are increased for these two diagnoses has been inconsistent in prior literature ( 24 – 28 ). Patients with tubal disease had an increased risk of preeclampsia/eclampsia, pregnancy diabetes, and placental problems. Why this group had an increased risk of these conditions remains to be explored. Taken together, the elevated risk in these groups suggest that they contribute to the elevated risk seen previously in subfertile/infertile populations. The infertility-related diagnoses in the non-ART groups in this paper were identified though the APCD which collects information on both inpatient and outpatient insurance claims. These claims do not differentiate between levels of disease severity or duration of the condition. We therefore had no information on the severity of disease in the ART-treated as compared to non-ART treated women and we cannot conclusively determine whether the increased risk in ART-treated women was a function of disease severity or the ART treatment itself. Of note is the fact that we found a diagnosis of infertility in over 95% of women with each of the underlying diagnoses in the ART-treated groups, but the in the groups without ART treatment this percentage was significantly lower ( Table 1 ). The higher rate of infertility in the ART group suggests that many women with these diagnoses in the non-ART group conceived without assistance and were thus never diagnosed with infertility ( 29 ). However, taken together with the consistent increase in risk in ART pregnancies and deliveries seen repeatedly in virtually all studies of ART ( 5 , 9 – 12 ), the current study suggests that ART treatment itself does indeed add additional risk. Of the risks that increased in ART-treated as compared with non-ART pregnancies, placental problems were consistently elevated in the ART-treated group. Risk ratios for placental problems in ART-treated vs. non-ART pregnancies were 1.72 for tubal disease, 1.49 for PCOS, 2.11 for other ovulatory conditions, and 2.86 for endometriosis patients. By contrast, PCOS, other ovulatory conditions, and endometriosis without ART treatment demonstrated no increase in placental problems over that in the fertile group ( Table 4 ). Previous studies have demonstrated an increase in placental abnormalities following ART treatment. Romunstad et. al ( 6 ) demonstrated an increased risk of placenta previa, Schachter et. al (Schachter, 2002) demonstrated an increase in vasa previa, and Esh-Broder et. al ( 30 ) demonstrated an increase in placenta accreta in ART deliveries. Further, we have demonstrated mediation of prematurity by placental problems in ART deliveries ( 31 – 32 ). These prior studies, along with the information in this study, lead us to hypothesize that abnormal placentation is a major mediating factor in the increased risk of adverse outcomes found in ART pregnancies and deliveries. Other adverse outcomes seen following ART may be more closely related to aspects resulting in underlying infertility. This study has several limitations. This is a retrospective study that relies on databases in which data are not collected for research purposes. There may thus be clinically relevant data entry errors and inaccuracies in the dataset. Although identification of ART deliveries based as it is on SART CORS, is likely to be highly accurate, identification of diagnostic groups may be less accurate since these rely on claims data and require that a patient have access to professional medical care to receive the diagnosis. Severity of the various disease conditions cannot be obtained from the claims data and thus we cannot directly compare patients who have similar patterns of each of the diseases evaluated. The fertile group may contain some unidentified infertile or subfertile deliveries and this could have decreased the magnitude of the observed differences. In addition, the MOSART database does not contain data on early pregnancy or pregnancy loss and thus we could not evaluate the effect of diagnosis on these stages of pregnancy. In addition, we note that we did not differentiate different ART treatment types such as day of transfer (day 2–3 or 5–7), fresh vs. frozen, ICSI vs. non-ICSI, or AH any of which could have had a differential impact on overall outcomes. Finally, the study results may not be generalizable to all states and countries as it was performed with data from a single state, Massachusetts. The study also has a number of strengths including that MOSART is a large population-based database and that it contains both maternal and infant hospital discharge records linked to birth certificate data. We also had access to linked outpatient claims data in APCD. Identification of ART deliveries was made through linkage to the gold standard of ART treatment, the SART CORS database, which contains all ART deliveries in Massachusetts. In conclusion, our study demonstrated that pregnancies and deliveries to women with several underlying infertility-related conditions were associated with increased risk of adverse pregnancy outcomes over those to fertile women and that in many cases ART treatment was associated with a further increase in that risk. Among the risks, placental abnormalities were particularly evident in the ART as compared with the non-ART deliveries. Future prospective studies will be needed to further explore these associations.

It is now well established that there is an increase in adverse obstetric and neonatal outcomes of singleton pregnancies following assisted reproductive technology (ART) ( 1 – 8 ), defined here as procedures involving manipulation of eggs and/or embryos in vitro. Adverse obstetric and perinatal outcomes studied have included gestational diabetes, pregnancy hypertension, placental abnormalities, and bleeding disorders, low birthweight, and prematurity ( 5 , 9 – 12 ). It is also well established that infertility itself is associated with an increase in adverse outcomes ( 4 – 5 , 8 , 13 ). What is less well established is whether underlying infertility-related diagnoses contribute to the adverse outcomes in the infertile populations in these studies and the extent to which ART adds additional risk over and above that associated with these individual diagnoses. In a previous attempt to address this question, we looked at underlying diagnoses of male factor infertility, endometriosis, ovulation disorders, tubal disease, and inflammatory disease and found that for many of these diagnoses both alone and in combination with ART there was increased risk of gestational diabetes, prematurity, and low birthweight ( 14 ). This prior study used the Massachusetts Outcome Study of ART (MOSART) database, that combined data from the Society for Assisted Reproductive Technology Clinic Online Reporting System (SART CORS) clinical ART database with the Pregnancy to Early Life Longitudinal (PELL) data system comprised of vital records and hospital stays for mothers and infants. The study did not fully address the question however, due to the necessity for a diagnosis in women without ART to be identified by prior hospital stays. This resulted in the population of non-ART treated women being only those whose infertility-related diagnosis reached the severity of having to have had either an inpatient hospital discharge, observational stay, or emergency room visit for the diagnosis in question. The diagnosis in these women was thus likely more severe than that expected in a general population of women with these conditions and was not an optimal comparison group. Most infertility is diagnosed and treated in the outpatient setting ( 15 ). Therefore, a more accurate sample of women with various infertility-related diagnosis would be obtained from outpatient visits rather than hospital stays alone. The goal of the current study was to identify women with infertility-related diagnoses using both inpatient and outpatient visits found in the Massachusetts All Payors Claims Database (APCD) and to use this population to compare obstetric and neonatal outcomes of women and their children for several important diagnostic categories both with and without treatment with ART.