Racial and Ethnic Disparities in the Perinatal Health of Infants Conceived by ART

We carried out a population-based study including singleton births in the United States from 2016 and 2017. We used publicly available data from birth and death certificates and fetal death certificates (cohort-linked files) obtained from the National Center for Health Statistics. These files included information about maternal demographic and clinical characteristics, pregnancy complications, and birth outcomes for all live born infants and stillbirths in the United States. We excluded births occurring before 20 weeks’ and after 44 weeks’ gestation and those with missing data on gestational age, plurality, race and ethnicity, and mode of conception.

Birth and fetal death certificates included self-reported information regarding treatments used to facilitate conception based on a question with 2 options worded as follows: “Pregnancy resulted from infertility treatment, if yes, check all that apply (1) fertility-enhancing drugs, artificial insemination or intrauterine insemination; and (2) assisted reproductive technology (eg, in IVF, gamete intrafallopian transfer [GIFT])”.^20,21  We, therefore, examined 3 groups of women and their infants who were conceived by: (1) non-ART MAR, (2) ART, and (3) spontaneous conception (ie, no reported MAR), which served as the reference category. Infants of women who indicated both non-ART MAR and ART were included in the ART group only. The categories of self-reported maternal race and ethnicity were based on previous literature^2–4  and the terminology used in the US National Vital Statistics Reports²¹  as follows: (1) non-Hispanic White, (2) non-Hispanic Black; (3) American Indian and Alaska Native (AIAN); (4) Asian or Pacific Islander (Asian Indian, Chinese, Filipino, Japanese, Korean, Vietnamese, other Asian, Native Hawaiian, Guamanian or Chamorro, Samoan, and Other Pacific Islander); and (5) Hispanic (all women of Hispanic origin: Mexican, Mexican American, Chicana, Puerto Rican, Cuban, and other Hispanic origin, regardless of race).

The primary outcome was neonatal death, defined as death within 28 days after birth. The secondary outcomes included fetal death, preterm birth <34 weeks, preterm birth <37 weeks, small-for-gestational-age and large-for-gestational-age infant (SGA and LGA, defined as <10^th and >90^th centile of birth weight-for-gestational age, respectively),²³  macrosomic infant (≥4500g), NICU admission, and perinatal (fetal and neonatal) death. Secondary outcomes also included a composite outcome of severe neonatal morbidity (SNM), defined as any of the following: Apgar score at 5 minutes ≤ 3, neonatal seizures or serious neurologic dysfunction, and assisted ventilation for >6 hours. Gestational age at birth was based on the obstetric estimate of gestation, with last menstrual period dating used if the obstetric estimate was unknown.

Rates of neonatal death, SNM, and other adverse neonatal outcomes were expressed per 100 live births, whereas fetal death and perinatal death rates were expressed per 100 total births. The primary independent variable was race and ethnicity, whereas the mode of conception was examined as a factor that potentially modified the effect of race and ethnicity on adverse neonatal outcomes. For rates with a low number of adverse birth outcomes (<30), exact binomial 95% confidence interval (CI) was calculated; for all other rates, the normal approximation was used to calculate CIs. Non-Hispanic White race was the reference category for racial and ethnic comparisons. Rate ratio (RR) and 95% CI were used to assess racial and ethnic differences in adverse outcomes rates in spontaneous conception (reference), non-ART MAR, and ART groups. We also compared all groups with respect to maternal demographic and clinical characteristics, obstetric history, and labor and delivery characteristics (gestational diabetes, hypertension in pregnancy, chorioamnionitis, induction of labor, and mode of delivery).

We used Poisson regression models with robust error variance to assess racial and ethnic differences in adverse fetal and infant outcomes, with associations expressed using adjusted rate ratios (aRR). Further, pairwise differences in these adjusted rate ratios between mode-of-conception strata were assessed on the basis of the statistical significance of interaction terms (using ARRs in the spontaneous conception group as the reference). aRRs were adjusted for maternal demographic and pregnancy characteristics, including prepregnancy BMI, smoking during pregnancy, parity (nulliparous, parity 1–3, parity ≥4), age (<25, >34 years vs 25–34 years), education (high school or higher versus less than high school), the type of health care insurance (Medicaid, self-paid, other versus private), prenatal care (no prenatal care versus some prenatal care), previous fetal death or termination of pregnancy, previous infant death, and fetal sex (male versus female). BMI (kg/m²) was categorized as follows: normal BMI (18.5–24.9), underweight (<18.5), overweight (25.0–29.9), and obesity class 1 (30.0–34.9), class 2 (35.0–39.9), and class 3 (≥40). Regression models for all neonatal outcomes were adjusted for congenital anomalies (any of the following: anencephaly, meningomyelocele or spina bifida, cyanotic congenital heart disease, congenital diaphragmatic hernia, omplalocele, gastroschisis, limb reduction defects, cleft lip, cleft palate, Down syndrome, suspected or confirmed chromosomal disorders, and hypospadias).

The significance of interaction terms between race and ethnicity and the mode of conception was examined after adjustment for other covariates.²⁰  Regression models for fetal and perinatal death could not be adjusted for the type of health care insurance, because this information was not available on fetal death certificates. AIAN race was not included in the multivariable analyses because of a relatively low number of AIAN women in our study. Records with missing values for any covariates (<3% in total) were excluded from multivariable analyses.

All analyses were performed on publicly accessible deidentified data and thus did not require ethics approval. All analyses were carried out by using SAS version 9.4 (SAS Institute, Inc).

Overall, 7 576 417 singleton births occurred to women residing in the United States in 2016 and 2017; 30 612 (0.4%) births were excluded on the basis of exclusion criteria, and the remaining 7 545 805 singletons were included in the study (Supplemental Fig 2). Of these, 37 074 (0.5%) were conceived by non-ART MAR, whereas 56 395 (0.8%) were conceived by ART.

We observed large racial and ethnic differences in demographic and clinical risk factors in all mode-of-conception groups. In general, non-Hispanic Black, AIAN, and Hispanic mothers were younger and more likely to be obese, using Medicaid, unmarried and less educated as compared to non-Hispanic White mothers (Table 1). In particular, AIAN mothers had the highest prevalence of smoking during pregnancy, obesity, and prepregnancy diabetes, whereas Asians and Pacific Islanders had the highest proportion of underweight mothers in all mode-of-conception groups (Tables 1 and 2). The rates of gestational diabetes were high among Asian and Pacific Islanders (10.9% in the spontaneous conception group, 19.3% in the non-ART MAR group, and 18.8% in the ART group).

In the spontaneous conception group, 53.1% of mothers were non-Hispanic White, 15.4% were non-Hispanic Black, 0.9% were AIAN, 7.1% were Asian or Pacific Islanders, and 23.3% were Hispanic (Table 1). In the ART and non-ART MAR groups, the proportions of non-Hispanic White mothers was much larger, 72.3% and 77.2%, respectively; the proportions of non-Hispanic Black mothers were 5.4% and 4.8%, AIAN mothers were 0.2% and 0.3%, Asian or Pacific Islanders were 13.7% and 0.3%, and the proportions of Hispanic mothers were 8.3% and 8.9%, respectively.

In addition, women who used MAR (either non-ART or ART) were older, with a larger proportion of married women, and those with private health care insurance (Table 1). Some risk factors were less frequent among mothers who conceived by MAR versus spontaneous conception, eg, maternal age <25 years, less than high school education, smoking, Medicaid insurance, and absence of prenatal care. Obesity, nulliparity, prepregnancy diabetes and hypertension, previous fetal death or termination of pregnancy, and pregnancy complications were more prevalent among mothers who conceived by MAR (Table 1 and 2).

In women who conceived spontaneously, rates of all adverse birth outcomes, except for macrosomia and LGA infant, were elevated in non-Hispanic Black compared to non-Hispanic White women (Fig 1 and Supplemental Fig 3, Supplemental Table 4). This discrepancy was further pronounced in women who conceived by MAR. For example, among infants of non-Hispanic Black mothers, neonatal mortality was 0.5% in the spontaneous conception group, 1.3% in the non-ART MAR group, and 1.6% in the ART group; the corresponding rates among infants born to non-Hispanic White women were 0.3%, 0.4%, and 0.3%, respectively (Fig 1, Supplemental Table 4).

The largest racial and ethnic differences were observed in neonatal mortality among infants of women who used ART (Table 3). Infants of non-Hispanic Black versus non-Hispanic White mothers who conceived spontaneously had a twofold higher rate of neonatal death (aRR = 2.0, 95% CI: 1.9–2.0), whereas these rates were more than threefold higher among infants of non-Hispanic Black versus non-Hispanic White mothers who used non-ART MAR (aRR = 3.3, 95% CI: 2.0–5.3), and fourfold higher in infants of non-Hispanic Black versus non-Hispanic White mothers who used ART (aRR = 4.1, 95% CI: 2.9–5.9; Table 3).

Among women who conceived spontaneously, and adjusted for other covariates, the rates of perinatal death were approximately twofold higher in non-Hispanic Black women (aRR = 1.9, 95% CI: 1.9–2.0), 10% higher in Hispanic women (aRR = 1.1, 95% CI: 1.1–1.1), and 10% lower in Asian women (aRR = 0.9, 95% CI: 0.9–0.9) than in non-Hispanic White women (Table 3). These racial differences in perinatal death were magnified among women who used non-ART MAR (non-Hispanic Black women had a 3.1-fold higher rate, and Hispanic women had a 70% higher rate). Racial and ethnic differences in perinatal mortality were also significantly increased in all specified race and ethnic groups compared to non-Hispanic White women in the ART group. The large racial and ethnic disparity in fetal death rates among women who conceived using non-ART MAR (eg, aRR = 3.1, 95% CI: 2.1–4.7 in non-Hispanic Black versus non-Hispanic White women; Table 3) was particularly noteworthy.

Interactions between race and mode of conception, adjusted for other covariates, were observed for several other adverse birth outcomes as well (eg, preterm birth < 34 weeks, Table 3). Whereas both Asian and Hispanic women who conceived spontaneously had significantly lower rates of SNM as well composite SNM and perinatal death compared to non-Hispanic White women, Hispanic women had higher rates of perinatal death or SNM compared to non-Hispanic Whites among women who conceived using non-ART MAR. Increases in the disparity in preterm birth rates <34 weeks were observed in Hispanic women who used non-ART MAR and ART, Asian women who used ART, and non-Hispanic Black women who used non-ART MAR. Non-Hispanic Black, Asian or Pacific Islander, and Hispanic women had lower rates of LGA infant compared to non-Hispanic White women irrespective of mode of conception (Table 3). However, the ARR was closer to null among non-Hispanic Black versus non-Hispanic White women who used ART.

Our study presents new findings showing substantially larger racial and ethnic disparities in adverse birth outcomes among women who used MAR compared to racial and ethnic disparities among women with a spontaneous mode of conception. Among women who conceived spontaneously, neonatal death rates were twofold higher among non-Hispanic Black women and 1.1-fold higher among Hispanic women compared to non-Hispanic White women, whereas these differences were significantly larger among non-Hispanic Black women who used non-ART MAR (3.3-fold higher than among non-Hispanic Whites), non-Hispanic Black women who used ART (4.1-fold higher), and Asian and Pacific Islander and Hispanic women (both 1.9-fold higher). Racial and ethnic disparities were also significantly larger in relation to several other adverse birth outcomes.

Racial and ethnic disparities in adverse birth outcomes of singleton pregnancies in the United States, the United Kingdom, and Canada have been previously documented.^19,24–30  In the United States, non-Hispanic Black women have an approximately twofold higher risk of stillbirth, neonatal death, preterm birth, and SGA live birth,^3,27–30  and a threefold higher risk of infant death³¹  compared to non-Hispanic White women. Our results pertaining to women who conceived spontaneously are similar, which is not surprising because the vast majority of singleton pregnancies result from spontaneous conceptions. In addition, we observed increased risks of SNM and NICU admission in infants of non-Hispanic Black mothers. In contrast, Asian women who conceived spontaneously had lower rates of perinatal death, although their adjusted rates of preterm birth were higher compared to non-Hispanic White women. Previous studies have also reported relatively favorable fetal or infant outcomes among Asian women.³¹  Asian mothers in the United States are relatively older and highly educated, with a higher proportion of immigrant women and women with private health care insurance,³²  all of which may suggest higher SES. We observed an elevated risk of perinatal death, preterm birth, SGA infant, and NICU admission among Hispanic women and their infants conceived spontaneously, which is congruent with previous reports,³³  although their rates of severe neonatal morbidity were lower compared to rates among non-Hispanic Whites.

Assisted reproduction is associated with elevated rates of perinatal death and other adverse outcomes in singleton pregnancies, depending on the type of fertility treatment and the underlying causes of subfertility or infertility.³⁴  Previous studies in the United States suggest that non-Hispanic Black women are less likely to access fertility treatment,¹⁵  and if they do, they are less likely to achieve live birth.³⁵  Reports from the Society for Assisted Reproductive Technology Clinic Outcome Reporting System show that non-Hispanic Black women have a 30% lower rate of live birth and an almost 40% lower cumulative rate of live birth from the initial ART cycle.³⁵  A previous report from 3 US states in 2000 to 2010 also showed a twofold higher risk of low birth weight and a 40% higher risk of preterm birth in singletons born to non-Hispanic Black versus non-Hispanic White women who used ART. However, this previous study did not assess race-ART interactions and ART techniques included in this study do not reflect contemporary practices.³⁶  Our study thus adds to the literature by describing birth outcomes on a range of adverse birth outcomes in women who used non-ART MAR and ART.

The larger racial and ethnic disparity was unexpected given that, on average, women undergoing ART are more likely to be college-educated, married, privately insured, and to have initiated prenatal care early in pregnancy.³⁸  However, as our results show, large racial and ethnic differences in socio-demographic and clinical factors exist even among women who use MAR, for both ART and non-ART. Additional underlying medical risk factors may further amplify disparities in health outcomes among disadvantaged groups, especially in non-Hispanic Black women, as shown in our study. Nevertheless, racial and ethnic disparities in birth outcomes are not solely medical matters but rather symptoms of larger social, economic, and political factors. Socioeconomic disadvantages, poor neighborhood conditions, lack of access to health care, psychosocial stress, racial discrimination, and systemic racism also contribute significantly to racial disparities in reproductive health.^25,38–44 

Our study has several strengths. We used an extensive population database containing consistently collected and detailed information on the key demographic and clinical characteristics of women who gave birth in the United States. This information provided requisite control for confounding through adjustment for maternal prepregnancy BMI and previous fetal death. We had sufficient statistical power to study interactions between race and ethnicity and MAR despite addressing associations involving relatively rare adverse outcomes such as neonatal death.

Our study has a few limitations. First, the self-reported information on the mode of conception in birth certificates likely includes some misclassification errors. However, previous US studies also show that 0.8% of singletons were conceived by ART³⁶  (similar to the estimate in our study), and large differences by race and ethnicity groups in the degree of underreporting are unlikely. Second, race and ethnicity was self-reported, and women of mixed race and ethnicity were categorized according to the National Center for Health Statistics algorithm, which could have resulted in some approximations.³²  Third, we lacked details about specific treatment procedures, including intrauterine insemination and various ovulation stimulation regiments, as well as ART modalities such as in-vitro fertilization and intracytoplasmic sperm injection, the number of IVF cycles, days of oocyte hatching, donor eggs, etc. The use of gamete intrafallopian transfer declined significantly in the United States before 2000,⁴⁵  and the ART group most likely included women with other forms of ART. Similarly, we were unable to account for racial and ethnic differences in underlying morbidity leading to low fertility or infertility (male infertility, polycystic ovary syndrome, etc), duration of infertility, preferences for the mode of conception, treatment, etc. Compared to spontaneous conception, ART singletons are more frequently the result of spontaneous fetal demise (vanishing twins) or selective fetal reduction after infertility treatment. We did not have information on racial and ethnic differences in the proportion of MAR-initiated pregnancies that were reduced from multiple to singleton pregnancy whether spontaneously or after selective fetal reduction.⁴⁶ 

Finally, we could not adjust for SES, systemic racism, and other factors that are associated with race and ethnicity and affect birth outcomes. We did not have information on whether the birth occurred in states that mandate employment-related health insurance for infertility treatments.

Our large-scale contemporary population-based study describing racial and ethnic differences in adverse birth outcomes fills a gap in the scientific literature by reporting on racial and ethnic disparities among women who conceive by MAR. We show that these racial and ethnic disparities were larger in women who conceive by MAR compared to women who conceived spontaneously, especially for the most severe outcomes, including fetal and neonatal death. Future research should focus on identifying vulnerable women using risk factor and treatment information to improve management and outcomes for this high-risk subpopulation.

AIAN

American Indian and Alaskan Native

ART

assisted reproductive technology

aRR

adjusted rate ratio

CI

confidence interval

IVF

in-vitro fertilization

LGA

large for gestational age

MAR

medically assisted reproduction

PTB

preterm birth

RR

rate ratio

SES

socioeconomic status

SGA

small for gestational age

SNM

severe neonatal morbidity