The Risk of Birth Defects With Conception by ART

Barbara Luke; Morton B. Brown; Ethan Wantman; Nina E. Forestieri; Marilyn L. Browne; Sarah C. Fisher; Mahsa M. Yazdy; Mary K. Ethen; Mark A. Canfield; Stephanie Watkins; Hazel B. Nichols; Leslie V. Farland; Sergio Oehninger; Kevin J. Doody; Michael L. Eisenberg; Valerie L. Baker


Hum Reprod. 2021;36(1):116-129. 

In This Article

Materials and Methods

This study linked data from birth certificates to data from birth defects registries and the national ART database, the SART CORS, in four States (New York, Texas, Massachusetts and North Carolina). Data from birth certificates (2004–2013) were collected in a study of the risk of childhood cancer and ART (Spector et al., 2019). The remaining data were obtained in the current study of the risk of birth defects in ART. New York, Texas, Massachusetts and North Carolina were chosen for the current study because they are large and ethnically diverse, with birth defect registries utilizing the same case definitions and data collected. These four States ranked #2 #3, #6 and #12 in highest number of annual ART births in the USA, respectively, in 2016, accounting for 3.0%, 1.5%, 4.7% and 1.4% of all births in each State (Martin et al., 2018; Sunderam et al., 2019).


The SART CORS contains comprehensive information on ART procedures from more than 83% of all clinics providing ART and more than 92% of all ART cycles in the USA. Data are collected and verified by SART and reported 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) (Centers for Disease Control and Prevention, 2019). The Society makes data available for research purposes to entities that have agreed to comply with SART research guidelines. Patients undergoing ART at SART member clinics sign clinical consent forms that include permission to use their data for research with appropriate provisions for safeguarding confidentiality. Data are submitted by individual clinics and verified by the medical director of each clinic. Approximately 10% of clinics are audited each year to validate the accuracy of reported data (Centers for Disease Control and Prevention, 2019). During each audit visit, data reported by the clinic are compared with information recorded in the medical record; most data fields have discrepancy rates less than 2%.

Fertility Treatment Data

ART represents only a small portion of all infertility treatments used in the USA. The National Survey of Family Growth reported that infertility services included medical advice (29%), infertility testing (27%), ovulation drugs (20%), artificial insemination (7.4%), surgery or treatment for blocked tubes (3.2%) and ART (3.1%) (Chandra et al., 2014). Identifying non-ART treatments is challenging, as there is no national registry for these therapies. In the 2003 revision of the US Birth Certificate, a checkbox was added to indicate that the pregnancy resulted from infertility treatment (worded as: if yes, check all that apply): (i) Fertility-enhancing drugs, artificial insemination or intrauterine insemination; (ii) Assisted reproductive technology (e.g. ART (in vitro fertilization), GIFT (gamete intrafallopian transfer)). Of the four States in this study, Massachusetts has collected data on infertility treatment on its birth certificate since 1996 and adopted the other items in the 2003 revision in 2012; Texas adopted the revision in 2005; New York State in 2004, New York City in 2008 (New York City maintains a separate birth registry); and North Carolina in 2010. Births which linked to the SART CORS cycles were categorized as ART; births with an indication that they resulted from infertility treatment (via any infertility checkbox on the birth certificate) but that did not link to an ART cycle in the SART CORS were categorized as ovulation induction (OI)/IUI; the remaining births were categorized as naturally conceived. Since <1% of births were checked as OI/IUI, all births prior to implementation of the checkbox on each State's birth certificate were labeled as naturally conceived. We estimate that 7.8% of the naturally conceived births did not have the infertility checkboxes on their birth certificate during the study period. This nonresponse rate would have increased the number of OI/IUI births by 8.5%. However, only 41.8% of the ART-treated women had an infertility checkbox checked 'Yes,' indicating an under-response of 58.2% for the ART births. Assuming this would be true for the OI/IUI births as well, this would more than double the number of OI/IUI births.

Linkage Procedure

This study linked ART cycles reported to the SART CORS from 1 January 2004 to 31 December 2015 that resulted in live births from 1 September 2004 to 31 December 2016 in Massachusetts and North Carolina and to 31 December 2015 in New York and Texas, to the birth certificates and birth defects registries in these four States. Initially, study States linked the SART CORS data to birth certificates. Each State received a SART CORS file with identifiers for women with ART cycles resulting in a live birth who were residents of that State during the study time period. The States linked the SART CORS data to birth certificate data to identify the ART-conceived births; >90% of the ART-conceived births were linked to their respective birth certificates. States then matched ART mothers to all study years to identify naturally conceived siblings of the ART birth reported to the SART CORS (ART siblings group); we did not include any ART siblings with an indication of infertility treatment on their birth certificates (OI/IUI). Any ART sibling who was conceived with ART was included in the ART group.

There were 97 582 ART-treated mothers of 158 698 children: 78 362 ART singletons and 22 301 ART singleton siblings, and 56 689 ART twins and 1346 ART twin siblings. Among the 97 582 ART-treated mothers, 61 327 had one ART singleton, 8247 had two ART singletons and 360 had three or more ART singletons; 27 675 had one set of ART twins and 173 had two sets of ART twins; 17 675 had one singleton sibling, 2014 had two singleton siblings, 193 had three or more singleton siblings and 698 had twin siblings.

For each delivery identified as having been conceived by ART, we requested that the subsequent 10 deliveries (all liveborn infants from a pregnancy) be selected as the non-ART comparison group, although not all States implemented this request, providing the next 10 births (individual children) instead. Each child was then linked to their respective State's birth defects registry. The vital records/birth defects linked files were de-identified before being sent to the investigators. We then linked the de-identified files to ART treatment parameters from the SART CORS by the use of unique research identifiers to create the final analytic file. This study was approved by the Institutional Review Boards at Michigan State University, the University of Michigan, and each of the four study State Departments of Health. The Michigan State University IRB determined that this research did not involve human subjects, as defined in 45 CFR 46.102 (f), in review dated 13 November 2015.

Birth Defects

The four States participating in this project are current or former Centers for Disease Control (CDC) Centers for Birth Defects Research and Prevention. As such, they conduct enhanced birth defects surveillance in terms of scope and quality of data. Each State conducts active or a combination of active and passive population-based surveillance that includes the major birth defects. These States employ standard case definitions, as defined by the National Birth Defects Prevention Study and National Birth Defects Prevention Network (NBDPN), and code birth defects using the CDC coding system adapted from British Pediatric Association codes, which is more specific for birth defects than ICD-9 or ICD-10 coding (Supplementary Table SI) (National Birth Defects Prevention Network (NBDPN), 2004). They employ multiple quality assurance procedures including validity checks, double-checking of assigned codes, clinical review of at least a subset of cases and comparison/verification between multiple data sources. They collect key demographic and clinical variables as defined by the NBDPN guidelines for conducting birth defects surveillance ( For this study, we analyzed birth defects diagnosed within the first year of life, as defined in Supplementary Table SI. We then classified individuals with major birth defects as either 'chromosomal' (i.e. presence of a chromosomal defect with or without any other major defect) or 'nonchromosomal' (i.e. presence of a major defect but having no chromosomal defect). 'Any birth defect' is any ICD-9 code with the first 3 digits 740–759, and any ICD-10 code inclusive of Q00.0–07.9, 10–18.9, 20–28.9, 30–45.9, 50–56.4, 60–87.89 and 89–99.9.

Blastogenesis Defects

We chose also to include birth defects classified as a group by Halliday et al. (2010) as blastogenesis defects, defined on the basis of pathologic development rather than by organ system. This allowed us to define defects which were expected to originate within the first 4 weeks of gestation, excluding cardiac defects. Disorders of blastogenesis in the current study were defined as the presence of one or more of the following: abdominal wall defects, vertebral segmentation defects, tracheoesophageal fistula, diaphragmatic defects, neural tube defects, anal atresia, renal agenesis, caudal regression sequence, laterality defects, sirenomelia, sacrococcyeal teratoma, holoprosencephaly, acro-renal field defect and ammelia, based on Halliday's grouping. Among children with a blastogenesis defect, 4% also had a chromosomal defect.


As described above, births were defined based on the presence or absence of subfertility/infertility and the method of conception. Births were categorized as natural-conceived, OI/IUI, ART, and natural-conceived ART siblings. The ART births were further divided into four subgroups depending on the combination of oocyte source (autologous or donor) and embryo state (fresh or thawed), based on our prior analyses indicating associations of these combinations with adverse perinatal outcomes (Luke et al., 2019, 2020). From these subgroups, children born to ART-treated women from cycles using autologous oocytes and fresh embryos (AF) without the use of ICSI were physiologically most similar to fertile births. The reference group, natural-conceived births, were compared to OI/IUI births, ART siblings and children born to ART-treated women from AF cycles without ICSI, and children born to ART-treated women from AF cycles with ICSI with or without the diagnosis of male factor infertility. When comparing within the ART subgroups, the reference group was children born to ART-treated women from AF cycles, stratifying by the use of ICSI for fresh cycles (data on ICSI was not available for thawed embryos). When modeling ICSI (the injection of a single spermatozoon into an oocyte) and assisted hatching (perforating the zona pellucida to facilitate hatching of the embryo and subsequent implantation), we restricted the analysis to ART cycles that had the responses 'All' or 'None' for these two variables to avoid cycles in which some, but not all embryos were treated with these procedures.

Independent Variables

Independent variables were selected a priori for inclusion in the models based on established associations with birth defects and/or ART. These included maternal age at delivery (grouped as 18–29, 30–34, 35–37, 38–40, 41–43 and ≥44 years), race (White, Black, American Indian/Alaskan Native, Asian/Pacific Islander, other or missing), Hispanic ethnicity, education (less than high school graduate, high school graduate or general educational development, some college or associate degree, bachelor's degree, post-graduate education or missing), parity (nulliparous, primiparous or multiparous prior to the index pregnancy), BMI (weight/height2) (≤24, underweight or normal weight; 25–29, overweight, and ≥30, obese, or missing) calculated from height and pre-pregnancy weight reported on the birth certificate, diabetes (pregestational and/or gestational), hypertension (chronic/pregestational and/or gestational and/or eclampsia) and infant sex, as well as State and year of birth.

Birthweight z-score was calculated as ((actual weight − reference weight)/standard deviation for the reference population), as recommended by Land (2006), using sex-specific national standards (Talge et al., 2014). Infants with z-scores of ≤−1.28 were categorized as small-for-gestation (SGA) and infants with z-scores of ≥1.28 were categorized as large-for-gestation (LGA). ART factors and treatment parameters included infertility diagnoses (male factor, endometriosis, ovulation disorders, diminished ovarian reserve, tubal ligation, other tubal factors, uterine factor, unexplained, other (immunologic, chromosomal or other serious disease) and other-non-infertile (single woman or same sex partners)); sperm source (partner, donor, or mixed); use of assisted hatching and ICSI. Twin births were analyzed separately. Triplets and higher-order multiples were excluded, as well as all births of women with the infertility diagnosis of PGD. Singleton data are shown in the Table I, Table II, Table III and Table IV and twin data are presented in the Supplementary Table SII, Supplementary Table SIII and Supplementary Table SIV.

Statistical Analysis

Data from each State were processed to generate a common dataset. The only exclusions were a mother or father who was younger than 18 years of age or implausible values (gestational age <22 weeks or birthweight <300 g even if indicated as a live birth). Because most independent variables were categorized, missing values were included as a separate category. Based on expected birth defects rates per 10 000 live births averaged across the four study States, we expected our naturally conceived and ART study populations to provide 90% power to detect an effect size of 6–8% with a two-sided α of 0.05 for major defects and cardiovascular defects, and an effect size of 15–30% for blastogenesis defects, genitourinary defects, orofacial defects and gastrointestinal defects. We used logistic regression to model the risk of any birth defect, a major nonchromosomal birth defect (i.e. major defect not accompanied by a chromosomal defect), blastogenesis defects, cardiovascular defects, orofacial defects, gastrointestinal defects, genitourinary defects in male children, musculoskeletal defects, chromosomal defects and any defects by group (as defined previously), with naturally conceived children as the reference.

Within the ART group, risks were modeled by oocyte source-embryo state combinations, and among infants born from cycles using fresh embryos, additionally by the use of ICSI. All analyses were performed using SAS Version 9.4 software (SAS, Cary, NC, USA). We could not properly account for correlation within twin pairs because data on twinship were inconsistently available (data were not consistently provided for both twins in a pair in the natural and OI/IUI conceived births). The number of fetal heartbeats greater than plurality at birth was added to the models and changed the point estimates by at most 0.02; most were unchanged, so this factor was not retained in the models.

Given that the time period of this study was 2004–2016, there were women with more than one delivery resulting in a live birth. We were able to identify women who had more than one live birth among those who were ART-treated, but not among the natural and OI/IUI conceived women. Of the ART women who had a singleton live birth, 85.3% had only one singleton live birth; 13.4% had two singleton live births and 1.2% had 3–5 singleton live births. If all the children delivered by women with more than one live birth had the same birth defect, this would increase the estimate of the standard error by ~7%; however, only 12% of the children from a mother who had a child with one defect had more than one child with a birth defect. Therefore, the effect on the standard error and the resulting CI is ~1%. It is likely that the repeat live birth rate in fertile women is higher than that for ART women, but the rate of repeat defects is not likely to exceed that of the ART births. Even if the rate of repeat pregnancies was as high as 50%, the effect on the estimate of the standard error would be <5%.