Increased risk of congenital anomalies after fertility treatment is partially mediated by multiple birth: a population-based cohort study.

Congenital anomalies are a major contributor to infant morbidity and mortality and impose long-term burdens on families and health systems ( World Health Organization, 2023 ). Fertility treatment has been associated with a higher risk of congenital anomalies ( Hwang et al. , 2018 ; Luke et al. , 2021 , 2022 ). However, most studies have compared births after fertility treatment with births among women without infertility, which may result in confounding by underlying infertility ( Velez and Ray, 2026 ). Emerging evidence suggests that infertility itself, independent of treatment, may be associated with adverse pregnancy outcomes ( Richmond et al. , 2022 ; Velez et al. , 2022 ; Wang et al. , 2022 ). Fertility treatment also substantially increases the likelihood of multiple birth ( Velez et al. , 2014 , 2024c ), and multiple births are at an increased risk of congenital anomalies compared to singleton births ( Glinianaia et al. , 2008 ; Bateni et al. , 2016 ; Esposito et al. , 2024 ; Shin et al. , 2024 ). Therefore, the observed associations between fertility treatment and congenital anomalies may reflect both a direct association with treatment and an indirect association mediated through multiple birth. Previous studies have often excluded multiple births or adjusted for plurality as a covariate, approaches that may attenuate the estimated association between fertility treatment and congenital anomalies ( Terho et al. , 2022 ; Henningsen et al. , 2023 ). The extent to which multiple birth mediates the association between fertility treatment and congenital anomalies remains unclear. Therefore, we used a large population-based cohort within a universal healthcare system to quantify the mediating effect of multiple birth on the association between fertility treatment and congenital anomalies, using births among women with infertility who did not receive fertility treatment as the reference group.

There were 274 893 births included in the cohort ( Fig. 1 ). Of these, 229 932 (83.6%) were among women with infertility who did not use fertility treatment, 20 525 (7.5%) after OI/IUI, and 24 536 (8.9%) after IVF/ICSI ( Table 1 ). Compared with women with infertility who did not use fertility treatment, those who underwent fertility treatment were more likely to be older, nulliparous, and reside in higher income and urban areas. Women with infertility who did not use fertility treatment had higher rates of obesity and pre-existing diabetes than those who underwent IVF/ICSI, although the highest prevalence of both conditions was observed among those who underwent OI/IUI. Rates of pre-existing hypertension were similar among groups. Cohort creation flowchart . OHIP, Ontario Health Insurance Plan; ID, identifier. Characteristics of all infants born in Ontario, 2006–2021, according to exposure group. Small cells <5. P -values were obtained using ANOVA and chi-square tests. The rate of congenital anomalies was 5.9% among births to women with infertility who did not use fertility treatment, 6.7% following OI/IUI, and 7.6% following IVF/ICSI. Multiple birth rates were 2.3%, 7.2%, and 13.3%, respectively. Among births following OI/IUI, the risk of any congenital anomaly was higher than among births to women with infertility who did not use fertility treatment (adjusted relative risk [aRR] 1.11; 95% CI 1.05–1.17) ( Table 2 ). Increased risks were also observed for CNS (aRR 1.39; 95% CI 1.07–1.80) and respiratory system anomalies (aRR 1.53, 95% CI 1.07–2.17). At the level of specific anomalies, increased risks were observed for atrial septal defects (aRR 1.24; 95% CI 1.06–1.45), pulmonary artery stenosis (aRR 1.56, 95% CI 1.06–2.29), and unspecified hypospadias (aRR 1.51; 95% CI 1.11–2.05) ( Table 3 ). Relative risk of any congenital anomaly and organ system anomalies for births following fertility treatment compared to infertility without treatment. Relative risks were adjusted for maternal age at delivery, income quintile, pre-existing diabetes, chronic hypertension, smoking, alcohol use, substance use disorder, rurality, obesity, and year of birth. Relative risk of specific anomalies for births following fertility treatment compared to infertility without treatment. Relative risks were adjusted for maternal age at delivery, income quintile, pre-existing diabetes, chronic hypertension, smoking, alcohol use, substance use disorder, rurality, obesity, and year of birth. Small cells. Among births following IVF/ICSI, the risk of any congenital anomaly was higher than among births to women with infertility who did not use fertility treatment (aRR 1.28; 95% CI 1.22–1.34) ( Table 2 ). Increased risks were observed across multiple organ systems, including cardiovascular (aRR 1.31, 95 CI 1.17–1.45), CNS (aRR 1.67, 95 CI 1.35–2.08); gastrointestinal (aRR 1.14, 95% CI 1.03–1.26), genital (aRR 1.46, 95% CI 1.26–1.69), musculoskeletal (aRR 1.44, 95% CI 1.27–1.63), neoplasms and tumors (aRR 1.71, 95% CI 1.35–2.16); and respiratory anomalies (aRR 1.72; 95% CI 1.21–2.46). At the level of specific anomalies, increased risks were observed for atrial septal defects (aRR 1.32, 95% CI 1.14––1.51), pulmonary artery stenosis (aRR 1.57; 95% CI 1.11–2.23), Tetralogy of Fallot (aRR 1.67; 95% CI 1.07–2.60), unspecified hypospadias (aRR 1.99, 95% CI 1.54–2.58), hemangioma (aRR 2.22, 95% CI 1.55–3.19), plagiocephaly (aRR 1.81; 95% CI 1.13–2.92), and pilonidal cysts (aRR 1.35; 95% CI 1.03–1.77) ( Table 3 ). The proportion of the association between fertility treatment and congenital anomalies mediated by multiple birth varied by treatment group and anomaly type ( Tables 4 and 5 ). Among births following OI/IUI, the proportion mediated was greatest for pulmonary artery stenosis (33.9%), followed by CNS anomalies (30.2%) and atrial septal defect (27.1%). Among births following IVF/ICSI, the proportion mediated was greatest for pulmonary artery stenosis (44.4%), followed by CNS (32.2%), and cardiovascular system anomalies (28.6%). Mediation analysis of the effect of multifetal birth on the association between fertility treatment and risk of organ system anomalies compared to infertility without treatment. Relative risks were adjusted for maternal age at delivery, income quintile, pre-existing diabetes, chronic hypertension, smoking, alcohol use, substance use disorder, rurality, obesity, and year of birth. Statistically significant natural indirect effect at a P -value of less than 0.05. NA: Results are due to the natural direct effect and the natural indirect effect operating in opposite directions. The proportion mediated is rendered meaningless when the natural direct and indirect effects are approximately the same magnitude, but opposite signs. Mediation analysis of the effect of multifetal birth on the association between fertility treatment and risk of specific anomalies compared to infertility without treatment. Relative risks were adjusted for maternal age at delivery, income quintile, pre-existing diabetes, chronic hypertension, smoking, alcohol use, substance use disorder, rurality, obesity, and year of birth. Statistically significant natural indirect effect at a P -value of less than 0.05. NA: Results are due to the natural direct effect and the natural indirect effect operating in opposite directions. The proportion mediated is rendered meaningless when the natural direct and indirect effects are approximately the same magnitude, but opposite signs. E-value analyses were conducted for all statistically significant associations ( Supplementary Tables S3 and S4 ). The smallest E-values were observed for any congenital anomaly following OI/IUI (E-value: 1.46; lower confidence limit 1.28) and IVF/ICSI (E-value 1.74; lower confidence limit 1.28), indicating that relatively modest unmeasured confounding could explain these associations. In contrast, the largest E-values approached 4.0, suggesting that a substantially stronger unmeasured confounder would be required to fully explain the observed associations, which is less plausible.

This population-based cohort study used administrative health data and registries from the province of Ontario, Canada held at ICES. ICES is an independent, non-profit research institute with designation under the Ontario Personal Health Information Protection Act to collect and analyze healthcare data for Ontario residents, without consent, for health system research ( ICES, 2021 ). ICES maintains validated datasets that have a unique maternal-infant matching number to link mother-infant pairs and capture information on pregnancy and pregnancy outcomes. These databases house mandatory submissions from hospital and healthcare visits using billing codes for physician and hospital reimbursement. The cohort included all live births and stillbirths at ≥22 weeks’ gestation from 2006 to 2021 among women aged 15–50 years with a history of infertility. Pregnancy and infant characteristics were obtained from the Better Outcomes Registry & Network (BORN) Ontario, which captures births at or greater than 20 weeks’ gestation; however, only births ≥22 weeks were included in our cohort. A birth refers to the complete expulsion or extraction of one or more fetuses ≥ 22 weeks gestation ( Zegers-Hochschild et al. , 2026 ). Maternal demographics, and pre-existing health conditions were obtained through linkage of ambulatory and hospital databases; the data sources and variables used are described in Supplementary Table S1 . Mother–infant pairs with less than 5 years of eligibility in Ontario Health Insurance Program (OHIP) eligibility before pregnancy were excluded to minimize under-ascertainment of infertility diagnoses and healthcare utilization. Births resulting from gestational carriage were excluded, as they represent a distinct clinical context involving embryos genetically related to the intended parent(s) ( Velez et al. , 2024b , 2025 ). The study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting based on EQUATOR guidelines ( Vandenbroucke et al. , 2007 ). Exposure was categorized as (1) infertility without fertility treatment, defined as physician consultation(s) for infertility within 2 years before pregnancy (OHIP ICD-9 628), and no record of fertility treatment, (2) ovulation induction (OI) or IUI, and (3) IVF or ICSI, consistent with previous population-based studies ( Dayan et al. , 2022 ; Fine et al. , 2022 ; Richmond et al. , 2022 ; Wang et al. , 2022 ; Velez et al. , 2023 , 2024a ). Information on treatment characteristics, including fresh versus frozen embryo transfer, cleavage versus blastocyst-stage transfer, assisted hatching, and preimplantation genetic testing was not available in the study datasets. Congenital anomalies were identified from the Canadian Institute of Health Information’s infant Discharge Abstract Database (DAD) using ICD-10 codes within the first year of life: Q000-Q999 and were classified using the Metropolitan Atlanta Congenital Defects Program algorithm, a surveillance system of the US Center for Disease Control and Prevention established in 1967 ( Correa-Villaseñor et al. , 2003 ). The DAD collects information on obstetrics, newborns and stillbirths from all acute inpatient hospitals in Ontario. Newborn and childhood anomaly diagnoses in acute hospitals are recorded and classified according to the 10th International Statistical Classification of Diseases and Related Health Problems, Canadian Adaptation ( Canadian Institute for Health Information, 2016 , 2018 ). The following outcome categories were modeled: (1) any anomaly, (2) organ system anomalies, and (3) specific anomalies as in prior ICES studies ( Nelson et al. , 2016 ; Brogly et al. , 2017 ; Bowie et al. , 2022 ). We adjusted for possible confounders of the association between fertility treatment and risk of congenital anomalies, including maternal age at delivery, income quintile, pre-existing diabetes, chronic hypertension, smoking, alcohol use, substance use disorder, rurality, obesity, and year of birth (codes provided in Supplementary Table S1 and expanded justification in Supplementary Table S2 ) ( American Society for Reproductive Medicine, 2006 ; Luo et al. , 2006 ; Hackshaw et al. , 2011 ; Correa and Marcinkevage, 2013 ; Mavrogenis et al. , 2014 ; Fan et al. , 2017 ; Harris et al. , 2017 ; de Angelis et al. , 2020 ; Lorzadeh et al. , 2020 ; Mattsson et al. , 2021 ; Benavides et al. , 2022 ; Bowie et al. , 2022 ; Evans et al. , 2022 ; Gautam et al. , 2023 ; Imrie et al. , 2023 ; Pethő et al. , 2023 ; Scott and Sher, 2023 ; Dhage et al. , 2024 ; Hedermann et al. , 2024 ). Patient characteristics were compared by exposure group using χ 2 and ANOVA. Relative risks (RR) were derived using modified Poisson regression models with robust error variance to account for more than one birth by the same mother during the study period ( Zou, 2004 ). A causal mediation analysis was performed to estimate the mediating role of multiple birth on the association between fertility treatments and congenital anomalies. Only associations with a statistically significant increased risk between exposure and outcomes were modeled in mediation analysis. Causal mediation is based on the counterfactual framework and delineates the association between fertility treatment and risk of congenital anomalies (total effect), into the direct effect of fertility treatment on risk of anomalies and the indirect effect due to multiple birth ( VanderWeele, 2015 , 2016 ). The estimated proportion of the total effect mediated by multiple birth was reported. We conducted a quantitative bias analysis using the E-value to assess potential unmeasured confounding. The E-value represents the minimum strength of association that an unmeasured confounder would need to have with both the exposure and the outcome, conditional on the measured covariates, to fully explain away the observed association ( VanderWeele and Ding, 2017 ). The study was approved by the Queen’s Health Sciences and Affiliated Teaching Hospitals Research Ethics Board.

In this population-based cohort study, births following fertility treatment had a slightly higher risk of congenital anomalies compared with births to women with infertility who did not use fertility treatment. This association was only partially mediated by multiple birth. Multiple birth has been proposed as an important mechanism underlying the increased risk of congenital anomalies following fertility treatment ( Tang et al. , 2006 ; Glinianaia et al. , 2008 ). Multiple gestations are associated with intrauterine crowding and reduced nutrient availability, which have been linked to an increased risk of structural anomalies ( Zhu et al. , 2006 ; Grantz et al. , 2016 ). Twins, for example, have higher risks of specific defects compared with singletons, including Tetralogy of Fallot, pulmonary valve stenosis, coarctation of the aorta, and cleft lip with or without cleft palate ( Li et al. , 2003 ; Tang et al. , 2006 ). Consistent with this, a large European cohort study reported modestly increased risk of non-chromosomal congenital anomalies among multiple birth compared with singletons (RRs ranging from 1.19 [95% CI 1.07–1.33] to 1.31 [95% CI 1.24–1.38]) ( Marleen et al. , 2024 ). Despite these associations, multiple birth explained only a portion of the increased risk of congenital anomalies observed following fertility treatment. Few studies have evaluated the mediating role of multiple birth. A cohort study comparing fertility treatments with births to women without infertility reported that 47.5% of any congenital anomaly, 63.2% of cardiac anomalies, 36.2% of non-cardiac anomalies, 42.7% of Tetralogy of Fallot, and 50.2% of hypospadias anomalies were mediated through multiple birth ( Liberman et al. , 2017 ). These estimates were generally higher than those observed in our study. One possible explanation is the choice of reference group. Unlike that study, we compared births to women with infertility who did not use fertility treatment, thereby reducing confounding by underlying infertility ( Velez and Ray, 2026 ). Consequently, our findings may provide a more specific estimate of the proportion of risk attributable to multiple birth rather than to infertility itself. The observed multiple birth rates in this study likely reflect the inclusion of births from an era when double embryo transfer was more common in Canada. Previous population-based research from Ontario demonstrated a marked decline in multifetal pregnancy rates following the introduction of publicly funded IVF and an elective single embryo transfer policy in 2015 ( Velez et al. , 2024c ). We observed an increased risk of genital anomalies, including unspecified hypospadias, following fertility treatment that was not mediated by multiple birth. This finding is consistent with prior cohort studies reporting an increased risk of hypospadias after ICSI ( Wennerholm et al. , 2000 ; Fedder et al. , 2013 ). Hypospadias has been hypothesized to reflect underlying parental factors, particularly male factor infertility, through heritable alterations in androgen production or Y-chromosome microdeletions ( Feng et al. , 2008 ; Simpson, 2014 ). Supporting this hypothesis, a large Nordic cohort study reported an increased risk of hypospadias among births following ICSI for male factor infertility ( Henningsen et al. , 2023 ). However, parental factors associated with infertility may extend beyond male factor infertility. Recent population-based work from our group has demonstrated an increased risk of congenital anomalies among births to women with endometriosis, including hypospadias (aRR 1.47; 95% CI 1.04–2.09), among other anomalies ( Milne et al. , 2026 ). These findings suggest that the increased risk of congenital anomalies observed following fertility treatment may, at least in part, reflect underlying infertility-related factors rather than treatment itself. This study benefits from a large, population-based cohort from a diverse population in Ontario with publicly funded fertility treatment ( Bougie et al. , 2021 ; Velez et al. , 2022 ), and datasets capturing 99% of births ( Better Outcomes Registry & Network ). Ontario had a universal healthcare system and publicly funded fertility treatment during part of the study period. Differences in healthcare coverage, access to fertility treatment, eligibility criteria, and funding models may limit the direct generalizability of these findings to other jurisdictions. Although our models were adjusted for birth year to account for temporal changes in fertility treatment access and practice patterns, caution is warranted when extrapolating these findings to healthcare systems with substantially different models of fertility care. Unlike most prior studies, we used births to women with infertility without treatment as the reference group to reduce confounding by underlying infertility. However, given that this group was defined using administrative data, misclassification of infertility status is possible when billing codes capture consultations in which difficulty achieving pregnancy was discussed without a confirmed diagnosis, potentially introducing residual confounding. Furthermore, pharmaceutical dispensing data were not available in the study cohort, and some patients may have received oral OI medication outside of a fertility clinic or without a fertility treatment procedure code, which could result in exposure misclassification. However, the proportion of multiple birth in the infertility without treatment group was low, at 2.3%, compared with 7.2% after OI/IUI and 13.3% after IVF/ICSI, suggesting that substantial misclassification of OI exposure in the reference group is unlikely. Data regarding IVF cycle characteristics and laboratory procedures were also not captured within the available databases; therefore, the contribution of treatment-related factors such as fresh versus frozen embryo transfer, embryo stage at transfer, assisted hatching, and preimplantation genetic testing could not be assessed. As with all studies using administrative health data, unmeasured confounding remains possible. E-value analyses suggested that modest unmeasured confounding could explain some associations, whereas others would require stronger confounding. Potential unmeasured factors include race/ethnicity. Lastly, there was limited statistical power to assess outcomes, particularly specific congenital anomalies, among births following ICSI. Therefore, ICSI was combined with IVF into a single exposure category. In conclusion, this population-based study suggests that multiple birth partially mediates the slightly increased risk of congenital anomalies among births following fertility treatment, while a substantial proportion of the observed association appears independent of multiple gestation. These findings suggest that while continued efforts to reduce multiple pregnancy following fertility treatment may reduce part of the excess risk ( Velez et al. , 2013 , 2014 , 2024c ) further research is needed to disentangle the contribution of specific infertility diagnoses (i.e. endometriosis, polycystic ovary syndrome, male factor) and treatment-related factors (e.g., fresh versus frozen embryo transfer, cleavage versus blastocyst-stage transfer, assisted hatching) to the risk of congenital anomalies.