Antibiotics are frequently used to prevent and treat infections during pregnancy [1-3]. Antepartum, respiratory, and urinary tract infections as well as pelvic inflammatory disorder are common reasons for antibiotic use [3]. During the intrapartum period, the use of antibiotics is frequent in high-risk women to prevent vertical transmission and early-onset group B streptococcal (GBS) diseases in neonates with an estimated prevalence of GBS colonization in 10% to 30% of pregnant women [4-6]. Moreover, surgical antibiotic prophylaxis is administered to about 19% of women before cesarean incision [7-10]. In Canada, the United States, and Europe, about 40% of pregnant women received antibiotics according to studies using data between 1998 and 2010 [2,3,11]. Although antibiotics are indicated for obstetrical and nonobstetrical conditions in pregnant women, little is known about medium- and long-term health effects in the infant as the consequence of in-utero exposure to antibiotics.

The widespread use of antibiotics during antepartum and intrapartum periods has raised concerns about alteration of the microbiota of the maternal birth canal before birth and interference with the early microbial transfer from the mother to the fetus during pregnancy, delivery, and lactation. This may lead to disruption of the child’s gut microbiota [12,13]. The bacterial microbiome in infants undergoes changes until it develops into a more stable adult-like microbiome [14]. Early life disruptions in microbial colonization and maturation may have downstream consequences on the metabolism and health of a child [15,16]. However, it is unclear as to how the exposure to antibiotics in-utero affects the maturation from the infant to the adult microbiome and, in consequence, how this may affect the health of a child.

The aims of the Fetal Antibiotic EXposure (FAX) study are to examine the association between in-utero exposure to antibiotics and adverse childhood outcomes including body weight, atopic diseases, and autism spectrum disorders (Figure 1) and to investigate the role of other potential factors mitigating or moderating the risk for adverse outcomes. Here, we describe the methods, cohort characteristics, and retention of infants included in the study cohort.

The study cohort was comprised 114,464 male and 108,967 female infants born in KPSC hospitals between 2007 and 2015 to 177,666 mothers. About 67,050 out of 223,431 infants (30.00%) were born via cesarean section, whereas 17,191 out of 223,431 (7.69%) were born preterm. Approximately half of the infants were non-Hispanic whites (Multimedia Appendix 1). Almost 1 out of 10 infants (n=20,071, 8.98%) had insurance coverage through government health care assistance programs. Most mothers (n=199,099, 89.11%) initiated their prenatal care at KPSC in their first trimester of pregnancy; about 22.52% of mothers (n=50,312) were obese with 9.59% of mothers (n=21,436) having a BMI ≥35 kg/m².

The in-utero exposure to antibiotics was frequent with 42,511 out of 223,431 (19.03%) of infants exposed during the antepartum period, 66,896 out of 223,431 (29.94%) during the intrapartum period, and 3731 out of 223,431 (16.69%) exposed during both the antepartum and intrapartum periods. About one-third of infants (76711 out of 223431, 34.33%) were not exposed to antibiotics in-utero (Multimedia Appendix 1). Compared with non-Hispanic white infants, black infants were more likely to be exposed to antibiotics in-utero overall; Hispanic infants were more likely to be exposed to antibiotics during pregnancy, and API infants were more likely to be exposed to antibiotics during the intrapartum period. Infants born preterm, with insurance coverage through government health care assistance programs, born to older mothers, born to mothers with a higher BMI, and born to mothers who smoked during pregnancy were more likely to be exposed to antibiotics in-utero.

The 5-year retention of infants in the health plan was over 80% (Table 1) and it has been increasing steadily since 2007. Retention was the lowest for infants with unknown race or ethnicity and highest for Hispanic infants. The retention of infants did not differ between infants with and without insurance coverage through government health care assistance programs.

During the neonatal period, infants had a median of 3 BMI measurements recorded and during their first year of life, a median of 5 BMI measurements (Figure 3). The number of recorded BMI measurements per child and year then slowly declined to 2 in the fifth year of life. The availability of BMI measures was consistent across all birth cohorts from 2007 to 2015 (data not shown).

In total, 2 out of 3 infants included in the FAX study were exposed to antibiotics in-utero; the exposure primarily occurred during the intrapartum period. The use of antibiotics during pregnancy is a risk-versus-benefit decision because untreated infections are associated with significant risk for the unborn child [1,35,36]. Owing to the widespread use of antibiotics, understanding the short- and long-term risks of in-utero exposure to antibiotics is an important public health issue. Considering the high proportion of infants exposed, even small individual-level health risks from antibiotic exposure could result in significant population-level effects. Possible adverse childhood outcomes include microbiome disruption [37], obesity [38,39], and infections [40,41]. Future interventions can be targeted to attenuate the risk [42-47].

To address current gaps in the knowledge about the health consequences of in-utero antibiotic exposure, we created a large cohort study of infants with racial, ethnic and socioeconomic diversity. The large sample size is particularly useful to support the study of rare outcomes and smaller population segments defined by race and socioeconomic status. Available clinical information is robust and reflects real-world information that clinicians and health plans use to document health care rather than research-quality data collected at prespecified study intervals [48].

Weight and height were frequently measured and recorded in the electronic medical records for calculation of BMI [19]. The frequency of BMI measurements per child together with the large sample size will allow us to model nonlinear trajectories of BMI or identify distinct BMI trajectories. We will be able to model these trajectories for the overall cohort and for the strata of children defined by the in-utero exposure to antibiotics ante- and intrapartum.

Several challenges for studies using data from the FAX cohort are noted. The use of routinely measured clinical weight and height may increase variation, which may bias potential risks toward the null. However, this bias will be compensated by the large sample size. Another source of variation is the variation in medical practice. In an ideal world, acute medical conditions such as infections are treated according to published recommendations. This may not always be the case in clinical practice. On the contrary, practice variation and the varying probability to be treated can be addressed in analyses by using appropriate statistical methods. In fact, practice variation can add knowledge about real-world settings, which can be readily translated into clinical practice.

Missing values may be frequent in routinely collected electronic medical record data and may not always be missing at random but they reflect real-world settings. The data compiled for the proposed study were collected for routine clinical care and not specifically for research purposes. However, most information such as race, ethnicity, and breastfeeding were collected for almost all patients [33,49]. To address any possible bias because of missing baseline or follow-up outcomes or covariates, we will examine baseline characteristics to determine if there are any systematic differences between the groups with and without missing data. If differences are found, appropriate estimation methods can be used to account for biases because of missing data, such as multiple imputations. Although addressing these challenges poses difficulty in the proposed research, the use of real-world clinical data also represents a key strength of the study.

Another potential challenge is the possibility of attrition because of patients leaving the health plan. KPSC is an open health care plan and members can enroll and disenroll. However, retention of members in the health plan, especially in the targeted age group, is high. To better understand the effects of attrition on the cohort characteristics, we will compare the characteristics (eg, demographics and BMI) of those who remain enrolled in KPSC and those who discontinued.

The current cohort of children will provide a unique opportunity to address key questions regarding the long-term sequelae of in-utero exposure to antibiotics in real-world data. The cohort shows a high retention rate and provides data from multiple medical visits over time, allowing us to model the trajectories of BMI over time.