The health of in vitro fertilization-conceived children: The Blind Men and the Elephant

There is a well-known poem by John Godfrey Saxe titled “The Blind Men and the Elephant,” based on a famous Indian legend:

It was six men of Indostan

To learning much inclined, Who went to see the Elephant (Though all of them were blind), That each by observation Might satisfy his mind […] And so these men of Indostan Disputed loud and long, Each in his own opinion Exceeding stiff and strong, Though each was partly in the right, And all were in the wrong!

As investigators, each of us is at least partly blind, partly in the right, and partly in the wrong. In an ideal world, we would know everything about the individuals we are studying. Factors would remain constant over time, and reporting would always be complete and accurate. However, we do not live in an ideal world, and in studying infertility, its treatment, and child health outcomes, several factors need to be considered. Each new analysis adds to our knowledge and should be put in context to prior published findings and hopefully identify areas in need of further investigation. In this issue of Fertility and Sterility, Farley et al. (1) presented an analysis evaluating the risk of neonatal and infant mortality among singletons conceived with infertility treatments using deidentified public-use national data on all US live births in 2015–2018. They concluded that infertility treatment was associated with a 51% increased risk of neonatal mortality and, through mediation analysis, concluded that 72% of the effect was through preterm birth. Of note, in their study, labor was induced for approximately one-fourth of the natural conception group compared with one-third of the infertility groups, and delivery was by cesarean section for approximately 30% in the natural conception group vs. 40%–48% in the infertility groups, indicating that some portion of the prematurity was likely to have been iatrogenic.

Conception using infertility treatments is a significant worldwide issue, for both maternal and child health. The International Committee for Monitoring Assisted Reproductive Technology estimates that 4.75 million in vitro fertilization (IVF) cycles are performed worldwide each year, resulting in the birth of 1.37 million children. In the United States, in 2019, nearly 84,000 infants were born from this technology, representing 2.2% of all US live births, a proportion that has tripled over the past 20 years. Although there is a federal mandate to report IVF cycle data in the United States, there is no such requirement for non-IVF infertility treatments, including ovulation induction and intrauterine insemination, which have been estimated by the Centers for Disease Control and Prevention scientists to represent 4 times as many US births as those conceived with IVF, bringing the actual number of children born annually from all infertility treatments closer to 420,000 or more than 11% of all annual US live births (2).

In vitro fertilization conception indicated on the birth certificate has been shown to only identify approximately 36.5%–50% of children, as acknowledged in this analysis (1). This means that 50%–63.5% of IVF-conceived children in this study are misclassified as being naturally conceived. Many women consider conception to be a very private issue and choose not to disclose how they became pregnant or the specifics of their infertility treatment to their obstetricians or midwives, making this a challenging subject to analyze solely on the basis of deidentified public-use vital records data.

In 2017, 60%–67% of all IVF transfer cycles used a single embryo, with rates declining with older maternal age. As a result, 73.6% of IVF-conceived infants were singletons, and 26.4% were multiples (twins accounted for 25.5%, and triplets and higher-order multiples accounted for 0.9%). The preterm birth rate among IVF-conceived infants was 27.8% overall and 14.0% among singletons, compared with 9.9% and 8.1%, respectively, among all US births in 2017 (https://www.cdc.gov/mmwr/volumes/69/ss/ss6909a1.htm). Because 33%–40% of IVF cycles that resulted in a live birth in 2017 involved the transfer of more than 1 embryo, it is likely that plurality at conception was greater than at birth, with a substantial proportion of singleton births the result of the vanishing twin syndrome or embryonic or fetal loss—an important and acknowledged factor increasing the risk of prematurity.

In 2019, more than 36% of all IVF cycles in the United States were freeze-all cycles, up from 10% in 2012. In cycles using thawed embryos or embryos conceived with donor oocytes, the corpus luteum is absent, which is known to have profound physiologic effects on the pregnancy, including increased risks of placental complications, fetal growth restriction, gestational hypertension and preeclampsia, and prematurity. Placental issues have been identified as critical factors in pregnancies among subfertile women and those conceived with assisted reproductive technology (ART) and IVF, including greater risks of abnormal cord insertions, placenta previa, vasa previa, and abruptio placenta; these risks are greater with the use of donor oocytes and thawed embryos (1, 3).

Birth defects are the leading cause of neonatal and infant mortality and have been shown to be more prevalent among ART- and IVF-conceived infants, with greater risks associated with male factor infertility and the use of intracytoplasmic sperm injection (4). The presence of birth defects increases the risk of childhood cancer, which is further magnified with IVF conception (5). The identification of birth defects from the birth certificate is known to have very low sensitivity, less than 25%. The investigators excluded 37,599 singleton births with major malformations identified on the birth certificate, for a rate of 0.25% (1). Population-based studies using registry-confirmed major birth defects indicate that the actual rate is between 2.0% and 2.5% or 10-fold higher than that indicated on the birth certificate (4). Farley et al. (1) did not present the causes of neonatal and infant mortality, which would have provided some indication of the underreporting of this factor on the birth certificate. This is another source of misclassification in their analysis. By limiting the dataset to live-born singletons, excluding birth defects, not accounting for pregestational or gestational diabetes, gestational hypertension, or preeclampsia, placental complications, or any IVF treatment parameters and not evaluating the causes of infant death (which is possible using the deidentified public-use linked birth-infant death files from the National Center for Health Statistics), the investigators presented a very limited view of a subset of pregnancies and births from ART—like The Blind Men and the Elephant (1). Moreover, no amount of advanced statistical modeling can provide unbiased estimates when there is omission of such critical variables (Fig. 1).

Figure 1.

Luke B, Brown MB, Wantman E, Seifer DB, Sparks AT, Lin PC, et al. Risk of prematurity and infant morbidity and mortality by maternal fertility status and plurality. J Assist Reprod Genet 2019;36:121–38.

Population-based studies are a starting point—known limitations should be acknowledged and overcome with linkages to other validated databases when possible—such as national IVF cycle databases and state birth defect and cancer registries. The true value of population-based analyses is finding patterns over time and identifying areas in need of more intense investigation, often using other types of data—such as histologic analyses of placentas and genomic findings from children or parental-child triads within fertility-treated families. The obstacles to developing such a complete and comprehensive dataset are both financial and logistical. Linking birth and infant death records on the state level is challenging but has been achieved among 14 states with the highest number of IVF-conceived births in a prior study on assisted reproduction and the risk of childhood cancer (R01 CA151973), with continued analyses in 4 of these states with comparable birth defect registries (R01 HD084377). The Massachusetts Outcomes Study of Assisted Reproductive Technologies has successfully linked IVF cycle data to vital records, hospital discharge, and early intervention data through their Department of Public Health’s Pregnancy to Early Life Longitudinal Data System (the PELL project) (R01 HD064595, R01 HD067270). Relevant to the analysis by Farley et al. (1), a recent publication from the Massachusetts Outcomes Study of Assisted Reproductive Technologies reported that among singleton births, the strongest effectors of prematurity were placental issues (adjusted odds ratio of 4.02 for births of 34–36 weeks and adjusted odds ratio of 10.28 for births of <34 weeks), pregnancy hypertension, and chronic hypertension; mediation analysis demonstrated a statistically significant indirect effect of placental issues for IVF and subfertile births (3).

The study by Farley et al. (1) is in line with prior published findings that prematurity is a major cause of neonatal and infant mortality, as shown in Figure 1 from our analysis of linked data from 14 states, but several important factors in the causal pathway were not identified in their analysis, including gestational hypertension and diabetes, growth restriction, and placental complications, as well as specific IVF treatment parameters. Like The Blind Men and the Elephant, we need to keep observing and debating to understand the relationships between infertility, its treatments, and subsequent maternal and child health.