Maternal obstructive sleep apnea and neonatal birth outcomes in a population based sample

Abstract

Objective:

Evaluate the association of OSA with birth outcomes including the risk of congenital anomalies and the need for a higher level of clinical care at delivery.

Methods:

Population-based study that linked newborn records with maternal records. Data from 95 perinatal centers across all geographic census divisions of the U.S. of women with a delivery diagnosis from 2010 to 2014 whose records could be linked to the corresponding newborn record. An International Classification of Diseases, ninth Revision (ICD-9) code for sleep apnea was used to identify exposure and outcome variables. Univariate and multivariate logistic regression analyses were performed with a model that included substance use, obesity, diabetes, maternal co-morbidities, and pregnancy complications.

Results:

In this study, 1,423,099 maternal records were linked to live newborn records. OSA was associated with a higher risk for congenital anomalies in offspring (aOR 1.26, 1.11 to 1.43), with the highest risk being that of musculoskeletal anomalies (aOR 1.89, 1.16 to 3.07) after adjusting for comorbidities and potential teratogens. Neonates born to mothers with OSA were more likely to be admitted to the intensive care unit (25.3% vs. 8.1%, p < 0.001), require resuscitation (aOR 2.76, 1.35 to 5.64) and have a longer hospital stay (aOR 2.25, 1.85 to 2.65).

Conclusions:

Although our study does not establish causation, it is the first to demonstrate a higher risk of congenital anomalies and resuscitation at birth in neonates of mothers with OSA, emphasizing the importance of identifying OSA in pregnant women and women of reproductive age.

1. Introduction

Obstructive sleep apnea (OSA), a condition on the spectrum of sleep disordered breathing, is characterized by recurrent intermittent hypoxemia, airflow limitation and repetitive arousals from sleep. Dynamic pregnancy physiology predisposes women to the development of OSA. Upper airway narrowing and edema related to decreased oncotic pressure and hormonal influences, reduced functional residual capacity and chest wall compliance, as well as possible weight gain may all predispose pregnant women to the development or worsening of preexisting sleep disordered breathing. Furthermore, OSA is a component of sleep disordered breathing [1]. In addition, nearly half of women of childbearing age are entering pregnancy either overweight or obese in countries like the United States [2], and the prevalence of OSA now exceeds 10% in pre-menopausal women [3]. OSA likely affects a significant proportion of women entering pregnancy, with a higher proportion of women having OSA in late pregnancy [4].

Possibly due to repetitive hypoxemia and sympathetic activation, OSA and sleep disordered breathing have been linked to maternal outcomes such as hypertensive disorders of pregnancy and gestational diabetes [4–7], as well as neonatal outcomes including prematurity [8] and growth restriction [8,9], even after adjustment for multiple covariates. In one study, neonates born to mothers with OSA have also been shown to require an intensive care unit admission more often than those born to mothers without OSA [10]. However, there have been no reports to date examining an association of sleep disordered breathing with congenital anomalies.

The etiology of congenital malformations is usually multifactorial and includes an inflammatory or infectious milieu, genetic susceptibility, or environmental and epigenetic changes [11–14]. Given a link between OSA and inflammation, and epigenetic changes, we hypothesized that exposure to maternal OSA in-utero would be associated with an increased risk of congenital anomalies. The aim of this study was to examine whether infants born to women with OSA are at an increased risk of having congenital anomalies, and to confirm prior reports of a potential risk for preterm birth, growth restriction, and a need for intensive neonatal care. As congenital anomalies are a rare outcome of pregnancy, large datasets are needed to examine such associations. Furthermore, adverse neonatal outcomes and congenital anomalies have devastating effects on life-long health and are associated with high costs to individuals and society. As OSA is easy and safe to treat, identification of an association of OSA with these negative outcomes paves the way for future research; which could examine causality, assessing and optimizing the impact of therapy on these outcomes, and holds the prospect of offering potential preventive strategies to women at risk.

2. Material and methods

2.1. Study population, setting, and study period

The National Perinatal Information Center (NPIC) data is a membership organization consisting of 95 perinatal centers across all geographic census divisions of the U.S. defined by the American Hospital Association [15], that submit clinical and financial information quarterly. Data are validated and compiled into the Perinatal Center Data Base (PCDB), which consists of both maternal and neonatal hospital discharge data, the latter occurring from birth to 28 days after birth. Multiple levels of comparison are then provided for every metric reported. Submitted data are routinely compared to a peer subgroup, PCDB as a whole, a trend database, and other national benchmarks. This validation report is then communicated back with respective hospitals to address inconsistencies. Hospitals then examine their metrics, address potential issues or discrepancies with provider documentation, coding, or quality, and correct their data before those are included in the final dataset. The PCDB dataset was used to identify women with a delivery discharge from 2010 to 2014 and to link those records to the corresponding newborn record. The dataset consists of maternal characteristics, billing information, gestational age, birthweight, and diagnosis and procedure codes based on the International Classification of Diseases, ninth Revision (ICD-9). Patients were not involved in this research and an examination of de-identified hospital collected data was performed. The study was reviewed and approved by review boards of two separate institutions (no. 881483 and no. 894311).

2.2. Demographic and social characteristics

Demographic and social characteristics included substance use. Women with a history of tobacco, drug, or alcohol use were identified by appropriate ICD-9 codes. Maternal characteristics and co-morbid conditions including those known to be associated with adverse neonatal outcomes or OSA such as hypertensive disorders of pregnancy or gestational diabetes, obesity, chronic hypertension, and pre-gestational diabetes mellitus were investigated using diagnosis codes (see Supplement Table 1).

2.3. Maternal obstructive sleep apnea and covariates

Members of the research team, with both clinical and research expertise in pulmonary (GB), sleep (GB), obstetric medicine (GB, KRM), maternal fetal medicine (HSL), pregnancy (MHB), perinatal epidemiology (VAD), birth defect surveillance (HSL), and population-based datasets (JM) considered exposure, covariate and outcome diagnoses. Diagnosis of OSA was established if a diagnosis code for OSA was present on the delivery discharge record (Supplement Table 1). Women with a diagnosis code of OSA were considered to have OSA, and those without that diagnosis code were considered to not have OSA. This method of diagnosis identification cannot differentiate timing of the diagnosis of OSA in regards to pregnancy, disease severity, or therapy.

2.4. Neonatal outcomes

We examined higher-level admission status (intermediate or intensive care), total length of stay, gestational age at birth, birthweight, and Apgar score, all obtained from the medical record. Small for gestational age (<10%-tile) and large for gestational age (>90%-tile) were calculated based on Oken et al.’s proposed normative values [16]. These calculations accounted for newborns’ sex and gestational age. Preterm birth was defined as birth occurring up to 36/6 weeks of gestation. Gestational age was examined as <34 weeks, 34–36 completed weeks, 37–39 weeks and >39 weeks. Current procedural terminology was used to collect outcomes such as airway intubation and resuscitation at birth.

2.5. Congenital anomalies

When evaluating congenital anomalies, all infants and pregnancies impacted by chromosomal anomalies were excluded. We focused on birth defects that would either require medical attention or would significantly impact health or long-term function. Cases of birth defects that resulted in pregnancy termination or miscarriage are not included in this dataset. Selected birth defects were grouped by organ system (see Supplement Table 1).

2.6. Statistical analyses

Descriptive statistics of hospital and maternal characteristics were reported overall and by OSA status. Mean values and standard deviations were used to describe maternal age and hospital length of stay and compared using t-tests. Hospital characteristics, demographic data, and co-morbidities predating pregnancy were reported using numbers and percentages and compared with chi-square tests. Nearly 2.6% of mothers had twin or multiple gestation (see Fig. 1), which falls within expected multiples rate (~3%). Maternal records consisted of one maternal record per delivery, rather than one record per mother. The data were analyzed by controlling for newborns born to the same mother at the same time (twin/multiples) or at any time during the study period (non-multiples). Covariates included demographics, multiple births, pre-pregnancy hypertension and diabetes, obesity, substance use, geographic location. We then examined neonatal outcomes by OSA status. Univariate and multivariable logistic regression analyses were used to calculate adjusted odds ratios (aOR) and 95% confidence intervals (CI) for neonatal outcomes. Neonatal outcomes, including congenital anomalies and obstetric outcomes, hypothesized to be related to sleep apnea were identified prior to analysis. Given the a priori nature of these hypotheses, and interest in individual and not overall significance, we did not correct analytic interpretations for multiple comparisons, despite the availability of numerous methods for multiple comparisons [17–19].

Fig. 1.

Consort diagram of total and linked maternal and newborn records. Records included multiples and newborns born to the same mother (see Methods section).

The adjusted model for neonatal outcomes included age, race, multiple births, pre-pregnancy hypertension and diabetes, maternal obesity, gestational diabetes, hypertensive disorders of pregnancy, alcohol, tobacco, and drug use. In addition, sensitivity analyses were performed to examine how gestational age at birth could impact associations between OSA and neonatal outcomes.

3. Results

3.1. Maternal characteristics

From 1,606,190 newborn records, a total of 1,423,099 live newborns were linked to maternal records after excluding duplicates (Fig. 1). Hospitals contributing to the data have been detailed previously [5]. In brief, hospitals in the southern part of the U.S. contributed 43.3% of all patients to the dataset. Hospitals in metropolitan areas and teaching hospitals contributed the highest proportion of records to the dataset (97.2% and 73.6% respectively). Hospitals with more than 5000 deliveries per year contributed 45.1% of all patients.

Mothers with OSA appeared to be older, more likely to be Black and to have a diagnosis of obesity. In addition, mothers with OSA were more likely to have diagnosis codes of tobacco use and drug use, but not alcohol use (Table 1).

Table 1.

Maternal characteristics (N = 1,423,099).

	OSA	No OSA	p-value for t-test or Chi square
	N = 1739	N = 1,421,360
	N (%) or mean (SD)	N (%) or mean (SD)
Age (years)	32.5 (6.0)	29.7 (5.9)	<0.001
Race			<0.001
Asian	30 (1.7)	65,619 (4.6)
Black	569 (32.7)	254,911 (17.9)
Native Hawaiian	6 (0.4)	7660 (0.5)
Hispanic	72 (4.1)	95,073 (6.7)
American Indian	7 (0.4)	5170 (0.4)
Unknown	318 (18.3)	338,687 (23.8)
White	737 (42.4)	654,240 (46.0)
Obesity	1041 (59.9)	84,073 (5.9)	<0.001
Pre-gestational HTN	498 (28.6)	39,945 (2.8)	<0.001
Pre-gestational DM	226 (13.0)	16,914 (1.2)	<0.001
Gestational HTN	168 (9.7)	64,877 (4.6)	<0.001
Preeclampsia	239 (13.7)	66,834 (4.7)	<0.001
Gestational DM	337 (19.4)	101,421 (7.1)	<0.001
Multiple gestations	133 (7.7)	62,938 (4.4)	<0.001
Tobacco use	156 (9.0)	50,996 (3.6)	<0.001
Alcohol use	1 (0.1)	1517 (0.1)	0.530
Illicit drug use	40 (2.3)	19,149 (1.4)	<0.001

OSA: obstructive sleep apnea; HTN: hypertension; DM: diabetes mellitus; SD: standard deviation.

3.2. Neonatal outcomes

When compared to newborns born to women without OSA, newborns born to mothers with the diagnosis had a higher prevalence of preterm birth (31.3% vs. 13.0%, p < 0.001) (Supplement Table 2). In addition, infants born to women with OSA required airway intubation and resuscitation at birth more frequently (14.4% vs. 4.8%, p < 0.001 and 0.5% vs. 0.1%, p < 0.001, respectively), had a longer stay in the hospital (8.28 + 14.5 vs. 3.97 + 8.63 days, p < 0.001), and were admitted to intensive care or special care neonatal units more commonly. Newborns of mothers with OSA had lower birth weight and a younger gestational age at birth than those born to mothers without OSA. However, the risk for small babies for gestational age was not significantly elevated.

In a multivariate analysis that adjusted for the extensive list of covariates discussed above, the risk for preterm birth and the need for a longer hospital stay and a neonatal intensive care unit (NICU) admission remained significantly elevated (Table 2), however, the risk for small babies for gestational age remained non significant. A sensitivity analysis was performed by newborn race and did not show any significant associations (crude or adjusted) between a diagnosis of OSA and the risk of small babies for gestational age. Although crude, odds ratio was elevated for the risk of large babies for gestational age, this was no longer significant in the adjusted model.

Table 2.

Odds ratios of neonatal outcomes by OSA status, controlling for newborns born to the same mother during the study period including twin/multiple gestations.

Outcomes	Crude OR or β (95% confidence interval)	OR or β (95% CI), fully adjustedc
Admitted to NICU	3.87 (3.45–4.33)	1.92 (1.69–2.19)
Admitted to special care nursery (intermediate or intensive care)	3.41 (3.08–3.78)	1.73 (1.54–1.94)
Preterm birth	3.07 (2.73–3.45)	1.48 (1.29–1.69)
Resuscitation at birth	5.75 (2.86–11.5)	2.76 (1.33–5.72)
Intubation at birth	3.30 (2.88–3.79)	1.63 (1.39–1.90)
Birth trauma	0.76 (0.24–2.36)	0.53 (0.17–1.65)
Bronchopulmonary dysplasia	2.49 (1.44–4.31)	0.98 (0.56–1.73)
Total length of stay (days)a	4.31 (3.56–5.05)	2.11 (1.39–2.83)
Birthweight (grams)a	−136 (−90 to −183)	−48 (−7 to −90)
Gestational age at birth (completed weeks) N = 1,190,758a,b	−1.4 (−1.2 to −1.5)	−0.5 (−0.3 to −0.6)
5-min Apgar scorea	−0.3 (−0.3 to −0.4)	−0.2 (−0.1 to −0.2)
Small for gestational age (<10% tile at birth) N = 1,182,151b	1.16 (0.99–1.35)	0.97 (0.82–1.15)
Small for gestational age (<10% tile at birth) excluding multiples N = 1,125,497b	1.17 (0.99–1.38)	1.01 (0.85–1.21)
Large for gestational age (>90%tile at birth) N = 1,182,151b	1.69 (1.36–1.85)	1.10 (0.94–1.28)

OSA: Obstructive sleep apnea; DM: diabetes mellitus. NICU: Neonatal intensive care unit.

^aModels adjusted for maternal obesity, race, age, multiple gestation, pre-pregnancy DM, gestational DM, pre-pregnancy HTN, gestational HTN, preeclampsia, tobacco use, alcohol use, and drug use.

^bN corresponds to the number of newborns with available gestational age data.

^cModels adjusted for maternal obesity, race, age, multiple gestation, pre-pregnancy diabetes, gestational diabetes mellitus, pre-pregnancy hypertension, gestational hypertension, preeclampsia, tobacco use, alcohol use, and drug use.

Moreover, when comparing gestational age at birth between the two groups, infants born to women with OSA were, on average, born 0.5 weeks earlier (1.4 weeks before adjusting), compared to infants unexposed to maternal OSA. When birth weight for gestational age was calculated based on neonatal sex, OSA was not associated with an increased risk for small birth weight for gestational age.

Sensitivity analyses stratified by gestational age were performed. Number of neonates born at each gestational age are presented in a footnote in Table 3 and showed 3% of all neonates were born before 34 weeks of gestation while 61% were born at or after 39 weeks. Sensitivity analysis showed that in infants born to women with OSA, the effect on being admitted to the NICU was significant at all gestational ages but strongest at <34 weeks’ gestational age (Table 3). On the other hand, the effect of OSA on airway intubation appeared to be most noteworthy at 39 weeks.

Table 3.

Sensitivity analysis for odds ratios of select neonatal outcomes by OSA status, stratified by gestational age <34 weeks, 34–36 weeks, 37–38 weeks, and ≥39 weeks, and controlling for newborns born to the same mother during the study period including twin/multiple gestations. (N = 1,190,758; that is the number of newborns with available gestational age data).

Outcomes overall and stratified by gestational age	Crude OR or β (95% confidence interval) for OSA association with neonatal outcome	OR or β (95% CI), fully adjusteda for OSA association with neonatal outcome
Admitted to NICU
<34 weeks	3.01 (1.59–5.70)	2.51 (1.31–4.78)
34–<37 weeks	2.37 (1.87–3.01)	1.85 (1.45–2.37)
37–38 weeks	2.92 (2.20–3.87)	1.59 (1.18–2.16)
≥39 weeks	2.63 (1.94–3.57)	1.75 (1.27–2.42)
Intubation at birth
<34 weeks	1.10 (0.78–1.56)	0.93 (0.65–1.34)
34–<37 weeks	1.89 (1.46–2.44)	1.54 (1.18–2.00)
37–38 weeks	2.69 (1.81–3.99)	1.46 (0.97–2.21)
≥39 weeks	2.58 (1.75–3.79)	1.78 (1.20–2.64)
Total length of stay (days)
<34 weeks	4.39 (−0.29 to 9.06)	1.93 (−2.85 to 6.71)
34–<37 weeks	1.88 (0.97–2.80)	1.14 (0.21–2.07)
37–38 weeks	1.29 (0.90–1.68)	0.75 (0.36–1.14)
≥39 weeks	0.64 (0.46–0.83)	0.35 (0.17–0.54)

OSA: Obstructive sleep apnea. NICU: Neonatal intensive care unit.

Note: <34 weeks: 36,619 (3%).

34–36 weeks: 117,864 (10%).

37–38 weeks: 313,989: (26%).

≥39 weeks: 722,286: (61%).

^aModels adjusted for maternal obesity, race, age, multiple gestation, pre-pregnancy DM, gestational DM, pre-pregnancy HTN, gestational HTN, preeclampsia, tobacco use, alcohol use, and drug use.

After excluding newborns with chromosomal anomalies, newborns born to mothers with OSA had a higher prevalence of congenital anomalies (any type) than those not exposed to OSA, (17.3% vs. 10.6%, p < 0.001) (Supplement Table 3), in addition to a higher prevalence of digestive, ocular, circulatory, central nervous system, musculoskeletal, and integument anomalies. Following multivariate regression analyses we controlled for maternal age; race and obesity; pre-pregnancy diabetes mellitus; gestational diabetes mellitus; hypertensive disorders of pregnancy; alcohol, tobacco, and drug use. Associations with congenital anomalies of the circulatory system, the musculoskeletal system, and the central nervous system remained significant (Table 4).

Table 4.

Odds ratios of congenital anomalies by OSA status, excluding n = 3250 with chromosomal anomalies and controlling for newborns born to the same mother during the study period including twin/multiple gestations. (N = 1,419,849).

Outcomes	Crude OR or β (95% confidence interval)	OR or β (95% CI), fully adjusteda
Congenital anomaly (CA) – all	1.76 (1.56–2.00)	1.26 (1.11 – 1.44)
CA 740 – nervous system (anencephalus)	–	–
CA 741 – nervous system (spina bifida)	3.25 (0.81 – 13.1)	2.14 (0.52–8.71)
CA 742 – nervous system (other anomalies of nervous system)	3.14 (1.78–5.55)	2.05 (1.16–3.66)
CA 743 – ophthalmologic	2.41 (0.99–5.80)	1.84 (0.76–4.43)
CA 744 – otologic and face	1.02 (0.46–2.27)	0.94 (0.42–2.10)
CA 745 – cardiac (septal defects)	2.26 (1.76–2.89)	1.16 (0.90–1.50)
CA 746 – cardiac (other anomalies of heart)	2.86 (1.84–4.45)	1.23 (0.78–1.94)
CA 747 – circulatory	2.96 (2.38–3.69)	1.37 (1.09–1.72)
CA 748 – respiratory system	1.68 (0.75–3.74)	1.02 (0.46–2.30)
CA 749 – cleft palate and cleft lip	1.29 (0.41–4.01)	0.99 (0.32–3.11)
CA 750 – gastrointestinal (upper alimentary tract)	1.32 (0.89–1.96)	1.22 (0.82–1.82)
CA 751 – gastrointestinal (intestinal and biliary)	2.76 (1.43–5.31)	1.80 (0.92–3.51)
CA 752 – genital organs	1.36 (0.94–1.98)	1.10 (0.76–1.61)
CA 753 – urinary system	1.34 (0.80–2.23)	0.99 (0.59–1.66)
CA 754 – musculoskeletal (certain congenital musculoskeletal deformities)	1.76 (1.06–2.92)	1.43 (0.85–2.39)
CA 755 – musculoskeletal (other anomalies of limbs)	1.38 (0.86–2.22)	1.05 (0.65–1.70)
CA 756 – musculoskeletal (other musculoskeletal anomalies)	2.52 (1.54–4.13)	2.05 (1.24–3.38)
CA 757 – integument	1.35 (1.07–1.70)	1.18 (0.93–1.48)
CA 759 – other and unspecified anomalies	1.93 (0.92–4.06)	1.42 (0.67–3.04)

OSA: Obstructive sleep apnea.

^aModels adjusted for maternal age, race, and obesity, pre-pregnancy diabetes mellitus, gestational diabetes mellitus, pre-pregnancy hypertension, gestational hypertension, preeclampsia, tobacco use, alcohol use, and drug use.

4. Discussion

This study is the first to examine and show an association of in-utero exposure to OSA with the presence of congenital anomalies in the offspring. The study also demonstrates a higher risk of admission to the NICU, a higher risk of requiring resuscitation and airway intubation at birth, and for being born preterm. There was no significant association between OSA status and small baby size for gestational age, calculated based on birthweight, sex, and gestational age. After adjusting for covariates, the risk for large for gestational age was no longer significantly elevated.

Most novel of our findings is the association with a higher risk of congenital anomalies even after adjusting for obesity and multiple variables including teratogens such as diabetes, alcohol, tobacco, and drug use. The pathophysiology of congenital malformations is thought to be multifactorial with complex interactions between genetic, social, medical, and environmental factors. As OSA is linked to obesity and diabetes, these conditions may represent important confounders. Obesity has been suggested as a potential teratogen in previous studies [20]. A recent national cohort based on the Swedish birth register data of 1.2 million women examined the impact of body mass index on congenital anomalies [11]. The study showed a dose response relationship between body mass index and congenital anomalies including cardiac, genital, and central nervous systems. Understandably, the study did not examine OSA as a potential covariate. When our data is interpreted in the context of previous literature [11], they suggest that both obesity and OSA may both impact the risk of congenital anomalies, with OSA affecting the risk independently of obesity. Given the strong correlation between OSA and body habitus, however, it may prove difficult to isolate the effect of OSA from that of obesity, and BMI data were not available to examine in the current dataset. Furthermore, hyperglycemia and diabetes have been associated with an increase in the risk of birth defects [21]. Experimental data suggest that exposure to hyperglycemia induces oxidative stress with resultant production of free radical species, which have the potential of being teratogenic [22]. OSA has also been hypothesized to result in oxidative stress due to the exposure to intermittent hypoxia [23]. However, the role of oxidative stress in the pathogenesis of OSA in pregnancy is not clear. Preliminary data by the investigators suggest that pregnant women with OSA do not have an enhanced oxidative or carbonyl stress profile compared to controls [24], possibly due to the protective effect of estradiol [25]. Other potential mechanisms underlying the risk of birth defects may include inflammation associated with OSA [26,27], or vascular dysfunction and abnormal placental metabolism as seen in visceral obesity [28] and potentially with OSA [29].

On the other hand, recent literature suggests that exposure to intermittent hypoxia in animals results in epigenomic alterations and inflammation in the offspring with resultant metabolic dysfunction reflected by higher body weight and adiposity index [30]. Similar findings were reported by the same group of investigators following the exposure of mice in late gestation to sleep fragmentation [31]. Human studies evaluating this question are scarce. A recent study showed that infants born to mothers at high risk for sleep disordered breathing by questionnaire screening have shorter telomeres in cord blood [32]. These data suggest that exposure to either intermittent hypoxia or sleep fragmentation as seen in OSA may impact the in-utero environment, leading to epigenetic or genetic changes potentially influencing long-term outcomes in the offspring. Though there is biological plausibility for OSA to be associated with birth defects and findings of an association in our study even after adjusting for known teratogens, a causal relationship cannot be inferred from current data.

Our findings of an association of OSA with preterm birth are similar to prior studies. Our study was able to show an average reduction of 1.4 weeks in gestational age at delivery in pregnant women with OSA by examining the gestational age at birth, and a higher risk of delivering preterm based on ICD-9 diagnosis. Similarly to the study by Bin et al., which linked maternal and neonatal records, OSA was not significantly associated with growth restriction after controlling for covariates [10]. However, similar studies of national cohorts or population based samples which examined growth restriction by diagnosis code –rather than a calculation-and small cohort studies [33,34] did show a significant association [8,9] while another study has shown associations between slowing of fetal growth and SDB [35]. Alternately, despite a higher prevalence of large for gestational age in this cohort, the increased risk was no longer significant once adjusted for known confounders such as maternal obesity, metabolic disease, and demographics. This is in contrast to a recent cohort study that showed an association between mild SDB in non-obese pregnant women and accelerated growth [36]. Outcome definition and exposure, method used to define SDB, disease severity, gestational age at the time of development of SDB, and record (maternal versus neonatal) used to confirm the diagnosis, as well as the assessment of potential confounders including the significant association of SDB with diabetes and obesity (conditions associated with increased birth weight), and the lack of certain parental characteristics in large datasets may explain these discrepancies. Furthermore, it has been recently suggested that “customized” weight prediction data, which account for parental ethnicity, height, weight and parity, perform better than population based weight prediction [37]. Therefore, our study could have missed an association since it used prediction values from population based studies which, do not account for certain parental variables, but also predate the current study by a decade, potentially misrepresenting demographic changes in the population. Bin et al., also examined the risk of admission to the intensive care unit and showed a higher need for an admission to the NICU among infants exposed to OSA in-utero [10]. This was consistent with our findings, which added to the existing literature by showing a longer hospital stay in infants born to mothers with OSA and a higher risk of requiring intubation and resuscitation, the former being most pronounced after 39 weeks of gestation. Of note, our findings suggest that although preterm newborns exposed to OSA had a longer hospital stay than unexposed newborns, full-term babies exposed to OSA in utero had a shorter hospital stay. It is unlikely that the latter unexpected findings could be explained by higher adherence to treatment for OSA. In addition, although pregnancies that ended at term may have been associated with a milder degree of severity of OSA than those that delivered preterm, based on our current knowledge of OSA in pregnancy, it is unlikely that mild OSA has a protective effect on fetal health. Hence, our findings could have been due to residual confounding. Furthermore, as this positive outcome is the first to be described in association with OSA, it would need to be interpreted with caution given the multitude of negative outcomes known to be linked to perinatal SDB.

There are numerous strengths to our study. Sample size and study design are optimal for evaluating rare outcomes such as congenital anomalies and neonatal intensive care admissions. The population studied is geographically, racially, and ethnically diverse and hospitals included vary in terms of teaching status, delivery volume and population served. The dataset used is robust and collects many variables that are relevant to the study of OSA in pregnancy and neonatal outcomes beyond ICD-9 codes. Moreover, data collected undergo many levels of validation before they are included in the database. In addition, the composition of this sample is similar to those of previously published cohorts of pregnant women with OSA despite vastly different samples sizes (<300 participants, vs. approximately 1.5 million in this sample) and study designs [38–40], including a high prevalence of obesity and pregnancy complications. However, results need to be interpreted in light of some limitations. Our dataset did not include women with an early termination or miscarriage that may possibly be due to severe birth defects; this may have restricted the examination of an association with non-fatal. Additionally, a potential ascertainment bias is possible in our study. When SDB was defined by an apnea hypopnea index using the more inclusive 3% desaturation criterion for hypopnea, recent prospective data showed a prevalence of SDB of 3% in early pregnancy [4] in women who were screened for OSA irrespective of symptoms. Although it is likely that the 0.12% prevalence in our study suggests under-diagnosis and under-coding, the definition of SDB in the two studies is completely different. In our population women carry the diagnosis due to having had symptoms that warranted the investigation whereas in the prospective trial described, any woman with an apnea hypopnea index >5 events per hour was considered to have the condition. Under-diagnosing OSA in pregnancy is likely quite common as most obstetricians rarely ask questions to identify women at risk for OSA [41]. Diagnosis code based datasets may introduce an ascertainment bias to the data as patients labeled as having OSA in ICD-9 based observational studies will be a mixture of true and false OSA [42], thus underestimating associations with adverse outcomes [42–44]. Our OSA sample may also be biased towards having women with more significant disease as non-pregnancy conditions appearing on the delivery record are likely conditions of enough importance to the pregnancy to be coded. This may extend to other maternal characteristics and complications of pregnancy. As obesity is based on an ICD-9 code rather than actual body mass index (BMI) data, it is possible that residual confounding by BMI may also exist. Moreover, we were unable to differentiate treated from untreated women. Assuming that therapy would impact this association, it would likely bias our associations towards the null. Further, the inability to ascertain the timing of OSA in regards to pregnancy limits the interpretation of our results where OSA occurring during or prior to organogenesis would impact the neonate differently than OSA developing later in pregnancy. However, concurrent data in a large sample of pregnant women suggest that women are rarely screened for SDB and even more rarely investigated during prenatal care [41], potentially suggesting that the diagnosis of OSA likely precedes pregnancy. In addition, having OSA in pregnancy has been associated with severe maternal morbidity and a higher risk of maternal ICU admission [5]. Therefore, it is possible that being more clinically challenging, these women were more likely to be treated with pharmacotherapy or be exposed to radiation, which may also contribute to the risk of teratogenesis. The current data did not collect radiation exposure or pharmacotherapy, thus limiting our ability to determine their potential impact on birth outcomes. The role of obesity may not have been fully accounted for, as obesity is likely under-coded. Finally, social determinants of health, including poverty and access to health care, were not measured in the current study yet have been associated with infant morbidity and mortality in past studies [45,46].

5. Conclusions

In conclusion, infants born to mothers with OSA are at a higher risk of having congenital anomalies compared to unexposed infants. In addition, OSA diagnosis is associated with an elevated risk of neonatal intensive care admission, higher level of care and preterm birth. Though this study does not establish causation between OSA and congenital anomalies and the occurrence of the latter is related to complex interactions between the genetic, environmental and the biological milieu, the data provide new and valuable evidence highlighting the importance of identifying OSA in women of reproductive age. Future research should focus on understanding pathophysiologic mechanisms underlying OSA in pregnancy and the impact of therapy on perinatal outcomes.

Abbreviations:

aOR

adjusted odds ratios

CI

confidence intervals

DM

diabetes mellitus

HTN

hypertension

ICD-9

international classification of diseases, ninth revision

NPIC

National Perinatal Information Center

NICU

neonatal intensive care unit

OSA

obstructive sleep apnea

PCDB

Perinatal Center Data Base

SD

standard deviation