Lower respiratory tract infections in early life are associated with obstructive sleep apnea diagnosis during childhood in a large birth cohort

Abstract

Study Objectives

Several birth cohorts have defined the pivotal role of early lower respiratory tract infections (LRTI) in the inception of pediatric respiratory conditions. However, the association between early LRTI and the development of obstructive sleep apnea (OSA) in children has not been established.

Methods

To investigate whether early LRTIs increase the risk of pediatric OSA, we analyzed clinical data in children followed during the first 5 years in the Boston Birth Cohort (n = 3114). Kaplan–Meier survival estimates and Cox proportional hazards models adjusted by pertinent covariates were used to evaluate the risk of OSA by the age of 5 years between children with LRTI during the first 2 years of life in comparison to those without LRTI during this period.

Results

Early life LRTI increased the risk of pediatric OSA independently of other pertinent covariates and risk factors (hazard ratio, 1.53; 95% CI, 1.15 to 2.05). Importantly, the association between LRTI and pediatric OSA was limited to LRTIs occurring during the first 2 years of life. Complementarily to this finding, we observed that children who had severe respiratory syncytial virus bronchiolitis during infancy had two times higher odds of OSA at 5 years in comparison with children without this exposure (odds ratio, 2.09; 95% CI, 1.12 to 3.88).

Conclusions

Children with severe LRTIs in early life have significantly increased risk of developing OSA during the first 5 years of life. Our results offer a new paradigm for investigating novel mechanisms and interventions targeting the early pathogenesis of OSA in the pediatric population.

pediatric OSA, lower respiratory tract infection, OSA, children, bronchiolitis

Statement of Significance

For the first time in a large birth cohort (>3000 infants), we demonstrated that lower respiratory tract infections (LRTI) in early life increase the risk of pediatric OSA. Children with a history of severe respiratory syncytial virus (RSV) LRTI during early infancy had more than twofold increased odds of developing OSA during the first 5 years of life. The results suggest that RSV LRTI may contribute to the pathophysiology of OSA in children. This raises the possibility that primary prevention strategies can hinder the initial establishment of OSA following early viral LRTIs. Primary prevention of OSA in children would have a dramatic effect in reducing the increasing incidence of this condition and in preventing its detrimental effects on childhood health and beyond.

Introduction

Obstructive sleep apnea (OSA) affects 1%–5% of the pediatric population [1–4] and is disproportionately prevalent among children with disadvantaged backgrounds [3, 4]. Serious adverse health outcomes associated with untreated pediatric OSA include severe hypoxemia, hypercarbia, sleep fragmentation, cognitive deficits, behavioral problems, impaired school performance, and failure to thrive [1–4]. Moreover, there is increasing evidence that pediatric OSA can also lead to early cardiovascular dysfunction [5, 6], obesity, insulin resistance [5, 7], systemic inflammation [8], and worsening of the asthmatic condition [9–11]. Decreasing OSA rates in children, particularly among minority high-risk populations, would ostensibly reduce health-related costs and improve the well-being and sleep quality of hundreds of thousands of children and families affected by this serious condition.

The efforts to reduce OSA rates in children have primarily aimed at the treatment of nasopharyngeal obstruction caused by adenotonsillar enlargement [1–4]. Indeed, adenotonsillectomy (AT) is considered the primary treatment for childhood OSA and is effective in about 70%–80% of the cases [12]. Over 500 000 ATs are performed each year for OSA in the United States[13], resulting in substantial health care utilization. Primary prevention of pediatric OSA would be a more cost-effective strategy; however, the causative factors of nasopharyngeal obstruction leading to OSA during early childhood are unknown and understudied. Filling this gap is critically important to design novel interventions that prevent the initial development and progression of this common condition in children.

Several birth cohorts have demonstrated the pivotal role of early lower respiratory tract infections (LRTI) in the development of respiratory conditions in children [14–17]. These studies have established that LRTIs caused by rhinovirus or respiratory syncytial virus (RSV) during the first 2–3 years of life increase the risk of wheezing illnesses and asthma beyond 6 years of age [17, 18]. Previous studies have suggested a similar link between LRTIs and the risk of pediatric OSA [19, 20]. However, the association between early LRTIs and the incidence of OSA in children has not been investigated in a large prospective longitudinal cohort. Given that the nasopharynx is the entry point of respiratory viruses [21], it is possible that respiratory infections in infants not only contribute to the inception of wheezing and asthma, but also the development of nasopharyngeal obstruction and pediatric OSA.

Our goal was to test the hypothesis that LRTIs occurring during early childhood (first 2 years of life) confer significantly higher risk of OSA in children. Our study includes a large, low-income, minority birth cohort of children longitudinally followed during the first 5 years of life (n = 3114). This is an ideal cohort for our research’s goal as it represents the group with the highest risk for LRTIs and OSA in the pediatric population (1, 4).

Methods

Study population

The Boston Birth Cohort (BBC) is an ongoing prospective, longitudinal birth cohort of newborns recruited at the Boston Medical Center (BMC), encompassing a predominantly low-income, minority, inner-city population in Boston, MA. All children enrolled in the BBC between October of 1998 and March of 2019 (n = 8509) were screened for eligibility. We only included children with complete baseline and follow-up data available (n = 3152). Exclusion criteria included immunodeficiency, cystic fibrosis, trisomy 21, or cleft palate. We also excluded infants who had OSA during the first 3 months of life (Figure 1). Additional details on the study population are provided in Supplementary Material.

Figure 1.

Study population and outcomes.

Early life LRTI

As described by our team and others [22–24], LRTI were defined as those infections that affect airways below the epiglottis and include acute manifestations of bronchitis, bronchiolitis, lung infections, or any combination among them according to the International Classification of Diseases (ICD) (Supplementary Table E2). LRTI information (binary variable) was collected using two separate strategies. First, we used a structured review of the primary care electronic medical record (EMR) to extract ICD-9/ICD-10 codes directly from the patients’ medical record into the BBC database (Supplementary Table E2). Second, we conducted separate analyses using survey data collected independently of EMR during in-person follow-up visits in the BBC. For these analyses, we focused on severe RSV bronchiolitis episodes (binary variable) defined as hospitalization due to this virus during the first year of life documented by self-report (parental questionnaires).

Clinical outcomes and covariates

As previously described by our team [25], the outcome for this study was defined as the child has a diagnosis of OSA (ICD-9/ICD-10 codes; Supplementary Table E2) in the EMR at any age between 4 and 60 months of age. ICD-9/ICD-10 codes were selected using an EMR-based algorithm for OSA identification that was recently validated across six health systems in the United States [26]. Our covariates included prenatal and perinatal characteristics linked to respiratory morbidities such as maternal race, maternal history of asthma, maternal level of education, pre-pregnancy maternal body mass index (BMI) category (normal, underweight, overweight, and obese), prenatal smoking (continuous smoking during pregnancy) and infant’s sex, gestational age (GA), breastfeeding status [25, 27, 28]. We also examined other relevant factors such as maternal age and parity, pregnancy complications (maternal diabetes and hypertensive disorders of pregnancy), season of birth, and the presence of childhood obesity (Table 1) [25, 29–32].

Table 1.

Summary of baseline characteristics by study group (OSA vs. no OSA during the first 5 years of life) in 3114 children included in the study

Variable	Children without OSA (n = 2876)	Children with OSA (n = 238)	P
Child’s sex (n, %)			0.108
Female	1437 (50.0)	106 (44.5)
Male	1439 (50.0)	132 (55.5)
Maternal race (n, %)
White	213 (7.4)	13 (5.5)	0.264
Black	1695 (58.9)	129 (54.2)
Hispanic	628 (21.8)	65 (27.3)
Asian/Pacific Islander	47 (1.6)	4 (1.7)
Other/mixed race	293 (10.2)	27 (11.3)
Maternal age in years
Median (IQR)	28.1 (23.2–33.3)	28.7 (23.9–34.0)	0.106
Maternal education (n, %)
No school/elementary	123 (4.3)	10 (4.2)	0.505
Some secondary school	679 (23.6)	54 (22.7)
High school graduate	1040 (36.2)	86 (36.1)
Some college	610 (21.2)	60 (25.2)
College degree and above	402 (14.0)	28 (11.8)
Unknown*	22 (0.8)	0
Multiparous mothers (n, %)	1651 (57.4)	134 (56.3)	0.741
Body mass index (BMI)			<0.000
Median (IQR)	25.1 (21.9–29.3)	27.1 (23.2–32.3)
Unknown* (n, %)	154 (5.4)	16 (6.7)
Maternal BMI category (n, %)			0.001
Normal (BMI 18.5–24.9)	1224 (42.6)	79 (33.2)
Underweight (BMI < 18.5)	125 (4.4)	7 (2.9)
Overweight (BMI 25–29.9)	755 (26.3)	58 (24.4)
Obese (BMI ≥ 30)	618 (21.5)	78 (32.8)
Unknown*	154 (5.4)	16 (6.7)
Maternal asthma (n, %)			0.506
Yes	330 (11.5)	33 (13.9)
No	2028 (70.5)	161 (67.7)
Unknown*	518 (18.0)	44 (18.5)
Maternal diabetes (n,%)			0.907
Gestational diabetes	219 (7.6)	20 (8.4)
Type 1 or 2 diabetes mellitus	132 (4.6)	11 (4.6)
Hypertensive disorders of pregnancy (n, %)			0.353
Preeclampsia/eclampsia/HELLP	248 (8.6)	19 (7.9)
Chronic hypertension	186 (6.5)	20 (8.3)
Unknown status*	23 (0.8)	0
Gestational age (weeks) (n, %)			0.000
Median (IQR)	38.7 (36.6- 40.0)	38.1 (34.6–39.6)
Prematurity (<37 weeks) (n, %)			0.000
Mild (33–36 weeks)	532(18.5)	43 (18.1)
Severe (≤32 weeks)	268 (9.3)	42 (17.7)
Type of delivery			0.060
Vaginal delivery	1857 (64.6)	138 (58.0)
C-section	1006 (35.0)	100 (42.0)
Unknown*	13 (0.5)	0
Pregnancy smoking (n, %)			0.214
Smoking in pregnancy	313 (10.9)	20 (8.3)
Unknown*	19 (0.7)	0
Season of birth			0.406
Summer	768 (26.7)	71 (29.8)
Fall	748 (26.0)	68 (28.6)
Winter	699 (24.3)	52 (21.9)
Spring	661 (23.0)	47 (19.8)
Breastfeeding status (n, %)
Breastfed infant (exclusive/partial)	1895 (65.9)	159 (66.8)	0.196
Unknown*	320 (11.1)	18 (7.6)
Childhood obesity (n, %)	215 (7.5)	45 (18.9)	0.000

*Missing observations.

There were significant differences in the proportion of severely premature infants (born at 32 weeks of GA or earlier), in children born to obese mothers (pre-pregnancy BMI of 30 or greater) and in children with obesity between the two groups. No significant differences were observed in maternal baseline characteristics (age at delivery, race, level of education, parity, and history of asthma), pregnancy complications (maternal diabetes and hypertensive disorders of pregnancy), child’s baseline characteristics (sex, type of delivery), prenatal/perinatal exposures associated with respiratory morbidity (pregnancy smoking, season of birth), or in breastfeeding status. Statistically significant P-values are in bold.

Statistical analysis

Exploratory data analysis was performed to inspect all variables and define baseline characteristics. Covariates with missing observations were imputed using the multiple imputations (mi) functions in STATA 14 [33] and we performed sensitivity analysis comparing imputed and non-imputed data to verify regression results were comparable. The cumulative risk of incident OSA after LRTI was estimated by Kaplan–Meier survival analysis and Cox proportional hazards models. Children entered the study at birth and those who developed OSA left the study at the age of first OSA diagnosis on record. Time to incident OSA was calculated as the earliest age at which a diagnosis of OSA is recorded. Children who did not develop OSA during the observation period were censored at 60 months of age or at the age of their last visit if they were lost to follow-up. Logistic regression and adjusted odds ratios (OR) were used to examine the association between severe RSV bronchiolitis in the first year of life (self-reported) and the development of OSA by 5 years of age. Statistical analyses were conducted using the software STATA version 14 (StataCorp. Stata Statistical Software: Release 14. College Station, TX, 2015) and R studio (RStudio: Integrated Development for R. RStudio, PBC, Boston, MA, 2020).

Results

Study population and baseline characteristics

Of the 8509 mother–infant pairs enrolled in the BBC until March 2019, we found a total of 3152 children with complete baseline and follow-up data (Figure 1). We excluded children with immunodeficiency (n = 16), cystic fibrosis (n = 1), trisomy 21 (n = 11), cleft palate (n = 4), and infants who had OSA during the first 3 months of life (n = 6). Baseline characteristics of included and excluded individuals are presented in Supplementary Table E1. Of the 3114 children included, 98.8% had follow-up at 1 year (n = 3078), 94.3% at 2 years (n = 2935), 90% at 3 years (n = 2803), 86.3% at 4 years (n = 2686), and 80.6% at 5 years (n = 2508) (Supplementary Table E3). We identified that 18.8% of the children included (n = 585) had developed LRTI by 2 years of age. During the preschool years (2–5 years of age), an additional 6.3% of children (n = 196) had their first LRTI episode. We found that of the 2788 children with available self-reported data on infections during the first year of life, 3% (n = 84) had severe RSV bronchiolitis that required hospitalization.

A total of 7.6% of children developed OSA in our study (n = 238). Children who developed OSA did not differ with respect to maternal race, level of education, smoking during pregnancy, and breastfeeding status (Table 1). Notably, as previously reported by our team and others [25, 29–31], there were significant differences in the prevalence of severe prematurity (GA ≤ 32 weeks), maternal obesity (BMI ≥ 30), and in childhood obesity between children who developed OSA and those that did not (Table 1).

Early life LRTI increases the risk of OSA during the first 5 years of life

We next examined the longitudinal association between LRTIs occurring in the first 2 years of life and the risk of developing OSA. For this analysis, we excluded 4 children who developed OSA before the exposure (LRTI between 0 and 2 years of age). As shown in Figure 2, the Log-rank test for comparison of survival function of the 3110 children included demonstrated a significant difference in incident OSA between individuals with early-life LRTI and those without (Supplementary Table E4). The yearly cumulative risk estimates of OSA by the age of 5 years in each group are displayed in Figure 2. Severe prematurity, maternal obesity, male sex, and childhood obesity also showed significant associations with the development of OSA (Table 2 and Supplementary Figure E3). Remarkably, the adjusted hazard ratio (HR) for incident OSA during the first 5 years after early life LRTI was 1.53 (95% CI 1.15 to 2.05, p-value = 0.004) independently of maternal characteristics (race/ethnicity, pre-pregnancy BMI category, maternal education level, and pregnancy smoking), children’s baseline characteristics (sex, prematurity category), breastfeeding status, and preschool-age obesity (Table 2 and Supplementary Figure E3).

Table 2.

Adjusted Cox proportional hazard model assessing the relation between early life lower respiratory tract infections (LRTI) and obstructive sleep apnea (OSA) during the first 5 years of life

Outcome	Early life LRTI (n, %)(n = 581)	No early life LRTI (n, %)(n = 2529)	Adjusted hazard ratio (95% CI)	P
OSA at 5 years (n, %)			1.53 (1.15–2.05)	0.004
Yes	67 (11.5)	167 (6.6)
No	412 (70.9)	1884 (74.5)
Unknown†	102 (17.6)	478 (18.9)
Child’s sex			1.31 (1.01–1.69)	0.044
Female	251 (43.2)	1289 (51.0)
Male	330 (56.8)	1240 (49.0)
Maternal race/ethnicity
White	38 (6.5)	188 (7.4)	Ref
Black	332 (57.1)	1489 (58.9)	0.79 (0.44–1.44)
Hispanic	129 (22.2)	563 (22.3)	1.22 (0.65–2.29)
Asian/Pacific Islander	9 (1.6)	42 (1.7)	1.26 (0.40–3.97)
Mixed/another race	73 (12.6)	247 (9.8)	1.12 (0.56–2.22)
Prematurity
Full-term	367 (63.2)	1861 (73.6)	Ref
33–36 weeks	127 (21.9)	448 (17.7)	1.07 (0.76–1.51)
≤32 weeks	87 (15.0)	220 (8.7)	1.84 (1.28–2.63)	0.001
Maternal BMI
Normal	219 (37.7)	1082 (42.8)	Ref
Underweight	18 (3.1)	114 (4.5)	0.94 (0.44–2.06)
Overweight	166 (28.6)	645 (25.5)	1.07 (0.76–1.52)
Obesity	154 (26.5)	542 (21.4)	1.66 (1.19–2.30)	0.003
Unknown*	24 (4.1)	146 (5.8)
Pregnancy smoking			0.77 (0.47–1.25)
Yes	503 (86.6)	2255 (89.2)
No	75 (12.9)	258 (10.2)
Unknown*	3 (0.5)	16 (0.6)
Maternal education
No school/elementary	23 (4.0)	110 (4.4)	Ref
Some secondary school	149 (25.7)	584 (23.1)	1.17 (0.59–2.33)
High school graduate	208 (35.8)	916 (36.2)	1.24 (0.63–2.43)
Some college	122 (21.0)	548 (21.7)	1.41 (0.71–2.81)
College degree and above	75 (12.9)	353 (14.0)	1.15 (0.54–2.44)
Unknown*	4 (0.7)	18 (0.7)
Breastfeeding
Yes	370 (63.7)	1681 (66.5)	0.88 (0.65–1.20)
No	157 (27.0)	565 (22.3)
Unknown*	54 (9.3)	283 (11.2)
Childhood obesity	70 (12.1)	189 (7.5)	2.17 (1.54–3.04)	0.000

*Missing observations imputed for adjusted analysis. The combined inferences from 20 completed datasets are presented.

^†Children lost to follow-up during the first 5 years of life were assumed to have the same conditional probability of developing OSA during the observation period as those who remained under follow-up and were included in the analysis.

Of 3110 children included, 581 had LRTI in the first 2 years of life and had a 53% increased risk of developing OSA after adjusting by other relevant covariates included in the model. Severe prematurity (GA of 32 weeks or less), maternal pre-pregnancy obesity, male sex, and childhood obesity were also significantly associated with an increased OSA hazard. Statistically significant HR and P-values are in bold.

Figure 2.

Cumulative risks of OSA during the first 5 years of life in children with and without early life LRTI. Kaplan–Meier comparison of survival function of the 3110 children included demonstrates a significant difference in incident OSA between individuals with LRTI during the first 2 years of life and those without LRTI during this period (p-value < 0.000).

Preschool-age LRTI in relation to the development of OSA during the first 5 years of life

We also examined the time window of the longitudinal association between the first LRTI and the development of OSA during the first 5 years of life. For this analysis, we compared the risk of OSA by the age of 5 years between 2333 children with no history of LRTI and 196 children who had their first LRTI between 2 and 5 years of age (pre-school-age LRTI). Kaplan-Meier analyses demonstrated that children who had their first episode of LRTI between 2 and 5 years of age did not have increased OSA risk during the study period (Supplementary Figure E1 and Supplementary Table E5). Adjusted Cox regression confirmed that children with preschool-age LRTI (2–5 years) had no significantly different risk of incident OSA, compared with those without LRTI (HR 0.50, 95% CI 0.24 to 1.08, p-value = 0.080) (Table 3). These data indicate that the association between LRTI and pediatric OSA by the age of 5 years is limited to infections occurring during the first 2 years of life as there was no increased risk of OSA by the age of 5 years among children who had their first LRTI episode after their second birthday.

Table 3.

Unadjusted and adjusted hazard ratio (HR) estimates of OSA after exposure to LRTI during preschool age (2–5 years) obtained from proportional hazards regression

Variable	Crude HR (95% CI) (n = 2518)	Adjusted HR (95% CI) (n = 2518)	P
First LRTI 2–5 years	0.52 (0.25–1.12)	0.50 (0.24–1.09)
Child’s sex		1.29 (0.94–1.78)
Maternal race/ethnicity
White		Ref
Black		0.85 (0.41–1.73)
Hispanic		1.17 (0.55–2.51)
Asian/Pacific Islander		1.41 (0.37–5.37)
Mixed/another race		1.22 (0.54–2.77)
Prematurity
Full-term		Ref
33–36 weeks		0.94 (0.61–1.46)
≤ 32 weeks		1.68 (1.03–2.74)	0.039
Maternal BMI*
Normal		Ref
Underweight		0.55 (0.16–1.80)
Overweight		1.05 (0.69–1.60)
Obesity		1.48 (0.99–2.22)
Pregnancy smoking*		0.76 (0.42–1.39)
Maternal education*
No school/elementary		Ref
Some secondary school		1.73 (0.67–4.45)
High school graduate		1.53 (0.60–3.90)
Some college		1.89 (0.73–4.90)
College degree and above		1.65 (0.60–4.56)
Breastfeeding*		0.77 (0.53–1.13)
Childhood obesity		2.45 (1.60–3.75)	0.000

*Variables with imputed observations. Missing observations in these variables were imputed (Supplementary Table E7) and combined inferences of 20 completed datasets are presented.

Eleven children who were diagnosed with OSA before their first LRTI episode were excluded. Of 2518 children included in this analysis, 185 (7.4%) between were exposed. In comparison with children without a history of LRTI (n = 2333), the risk of OSA during the first 5 years was not increased among children who had their first LRTI during preschool age. Statistically significant HR and P-values are in bold.

Infants with severe RSV bronchiolitis have significantly increased OSA risk during the first 5 years of life

We conducted separate analyses to confirm our results using the self-report of a history of RSV bronchiolitis hospitalization. These data were obtained directly via parental surveys instead of ICD-9 or ICD-10 codes. We found that there were 84 children hospitalized with RSV bronchiolitis during the first year of life among 2788 children with available self-reported data (Supplementary Figure E2). Adjusted analysis of the association between severe RSV bronchiolitis and OSA at the age of 5 years revealed a twofold increase in the odds of OSA among children with a history of hospitalization due to RSV bronchiolitis (OR 2.09, 95% CI 1.12 to 3.88, p-value = 0.020). This association was independent of maternal factors (race, education, BMI category, and smoking) and children characteristics (sex, GA, breastfeeding, and childhood obesity) (Table 4).

Table 4.

Unadjusted and adjusted logistic regression models revealed an association between a history of severe RSV/bronchiolitis during infancy and OSA during the first 5 years of age

Variable	Crude OR (95% CI) (n = 2787)	Adjusted OR (95% CI) (n = 2787)	P
RSV hospitalization	2.34 (1.28–4.26)	2.09 (1.12–3.88)	0.020
Child’s sex		1.23 (0.93–1.64)
Maternal race/ethnicity
White		Ref
Black		0.88 (0.48–1.63)
Hispanic		1.32 (0.67–2.57)
Asian/Pacific Islander		1.05 (0.26–4.17)
Mixed/another race		1.43 (0.70–2.90)
Prematurity
Full-term		Ref
33–36 weeks		1.10 (0.76–1.59)
≤ 32 weeks		2.20 (1.44–3.38)	0.000
Maternal BMI*
Normal		Ref
Underweight		0.96 (0.42–2.22)
Overweight		1.09 (0.74–1.60)
Obesity		1.69 (1.21–2.38)	0.002
Pregnancy smoking*		0.88 (0.53–1.46)
Maternal education*
No school/Elementary		Ref
Some secondary school		0.97 (0.47–2.00)
High school graduate		1.07 (0.53–2.18)
Some college		1.33 (0.63–2.78)
College degree and above		1.07 (0.47–2.40)
Breastfeeding*		0.87 (0.62–1.20)
Childhood obesity		2.17 (1.48–3.18)	0.000

*Variables with imputed observations. Missing observations in the main outcome (OSA by 5 years of age) and maternal BMI category, pregnancy smoking, maternal education and breastfeeding status were imputed (Supplementary Table E8) and combined inferences of 20 completed datasets are presented.

We found that 84 children were hospitalized with RSV bronchiolitis during the first year of life among 2788 children with available self-reported questionnaires. One child diagnosed with OSA before their first LRTI was excluded. Statistically significant odds-ratios (OR) and P-values are in bold.

Discussion

In a large birth cohort, we demonstrated for the first time that LRTIs occurring in early childhood (0–2 years of age) significantly increase the risk of pediatric OSA by the age of 5 years of age. Complementarily to this novel finding, we demonstrated that children with a history of severe RSV bronchiolitis during early infancy had more than twofold increased odds of developing OSA during the first 5 years of life. Notably, the association between early life LRTI and OSA was independent of major risk factors for OSA and other pediatric respiratory diseases (e.g. prematurity or obesity).

Our study adds OSA to the pediatric respiratory disorders linked to LRTIs during early infancy [17, 34, 35]. The current findings are in line with a previous study by Snow et al., describing the association between severe RSV bronchiolitis and OSA in 21 children with OSA and 63 controls [19]. Another study reported a link between recurrent respiratory infections and OSA in preschool-aged children [20]. However, our study is the first to demonstrate that early-life LRTIs are strongly associated with the development of pediatric OSA in a large birth cohort. We cannot draw conclusions about causality because it is possible that infants who have RSV bronchiolitis and later present OSA were born with an intrinsic airway predisposition to develop both respiratory conditions. Nonetheless, the fact that the exposure to LRTI preceded the diagnosis of OSA suggests that respiratory viruses could mechanistically contribute to the development of pediatric OSA and motivates further research to establish whether early life LRTI could be a cause and/or a marker of an underlying susceptibility to OSA in the pediatric population.

Our finding that severe RSV infections are strongly associated with increased OSA risk generates important questions about how respiratory viruses may relate to the development of nasopharyngeal obstruction. RSV is a top cause of severe LRTI during early infancy and increases the risk of respiratory morbidity beyond childhood [36, 37]. RSV triggers airway hyperresponsiveness [38] and neuroimmunomodulatory pathways that collectively may alter lymphoproliferative and inflammatory responses in the nasopharynx [19, 39, 40]. In addition, since the nasal epithelium is the entry site of respiratory viruses [21], RSV and other pathogens may dysregulate infant’s nasal mucosal immunity and companion microbiome [41–44], further contributing to inflammation and obstruction. Several studies have reported that viruses are commonly detected in tonsils and adenoids of children [45–48]; thus, recurrent and/or chronic asymptomatic infections may also play a role in perpetuating lymphoid proliferation overtime. Respiratory viruses and airway inflammation may also affect the upper airway neuromotor control [49–51]. Altogether, these findings suggest an important role of respiratory infections in the development of adenotonsillar hypertrophy and nasopharyngeal airway obstruction. Nonetheless, much is unknown, and a more in-depth understanding of how viral infections disrupt airway homeostasis is required to elucidate the earliest origins of pediatric OSA and prevent its multiple consequences in pediatric health.

We also found that pediatric OSA was strongly associated with a history of severe prematurity as well as with maternal and childhood obesity. These factors have been previously identified as early determinants of childhood OSA by our group and others [20, 25, 29–31]. Interestingly, these are also major factors implicated in the development of recurrent wheezing and asthma [52], and the coexistence of pediatric OSA with these lower airway disorders has been demonstrated [9, 11]. In our current study, we found that the association of LRTIs and the subsequent development of OSA in children was independent of severe prematurity and maternal or childhood obesity. Therefore, the complex interactions between these independent pediatric OSA risk factors warrant further investigation. It is possible that there is a bidirectional link between OSA and the generation of other respiratory conditions in children (e.g. wheezing and asthma). This possibility deserves a more in-depth integrative research approach toward the study of common mechanisms and susceptibility factors of pediatric upper and lower airway disorders [53].

Our finding that the association between LRTI and pediatric OSA is specific to the first 2 years of life is also significant. This is in line with the prevailing notion that there is an early window of susceptibility to noxious exposures determining long-term trajectories of airway development and respiratory health [28]. The time-limited effect of LRTI on the risk of OSA is also in agreement with extensive evidence showing that viral respiratory infections during the first year of life increase respiratory morbidity beyond childhood [17, 35] and that the environment during this period may have protective or deleterious effects [54, 55].

Furthermore, our observation that maternal obesity increases OSA risk during the first 5 years of life suggests an important prenatal contribution. In-utero exposures and other nonheritable factors are associated with infant respiratory morbidity [56], and prenatal interventions may modulate this risk [57]. Thus establishing the precise timing when specific environmental perturbations in the prenatal and postnatal periods affect the respiratory system development would have a major impact on the early detection and prevention of pediatric OSA, and other highly prevalent chronic respiratory disorders.

Our study has some limitations. First, our definition of OSA was based on EMR, which made feasible the inclusion of a large number of children (>3000), but we did not have confirmation with polysomnography (PSG). However, EMR-based identification of OSA using an ICD-9/ICD-10 algorithm was recently validated against “true cases” (e.g. defined by PSG or home sleep tests) in a large multi-centric study [26]. This EMR algorithm showed excellent performance in identifying OSA cases with a positive predictive value of 97.1 and a negative predictive value of 95.5 and, although performed in adults, this method offers an alternative in OSA research studies, particularly when it is not feasible to perform PSG in all research subjects [26]. Second, although LRTI preceded OSA diagnosis in our study, we were unable to determine the precise onset of OSA symptoms, the temporal relationship between LRTI and OSA development, as well as the lag between OSA symptoms and diagnosis. Third, our LRTI definition did not include virus confirmation, thus the possibility that in addition to RSV other pathogens (e.g. rhinovirus) contribute to the development of pediatric OSA remains to be determined. Fourth, although we included most of the known determinants of childhood respiratory health [58], family history of atopy or OSA, relevant environmental exposures (e.g. pollution), nutrition and lifestyle factors, and relevant socioeconomical information (e.g. household income, housing) of participants were not accounted. Because the BBC excludes children with major anatomical anomalies and genetic syndromes, we did not actively look for craniofacial abnormalities other than cleft palate. Importantly, although the association between early-life LRTI and pediatric OSA was clearly time-specific, our study only evaluated OSA risk during the first year of life and we cannot exclude the possibility that LRTI occurring after the age of two, may increase OSA risk beyond 5 years of age. Finally, as our sample is an inner-city, minority population at high risk for LRTI and OSA, our findings may not be extrapolated to other populations. Nonetheless, the major strength of our study is that it represents the only large prospective birth cohort published to date with detailed information about LRTI and OSA incidence during the first 5 years of life. The availability of comprehensive clinical information in all study subjects allowed us to provide the first evidence that LRTIs occurring during early childhood increase the risk of developing pediatric OSA independently of other risk factors.

In conclusion, our results offer a new paradigm for investigating mechanisms implicated in the early pathogenesis of OSA in the pediatric population. Specifically, as most OSA cases in young children are caused by nasopharyngeal obstruction and adenotonsillar enlargement [1–4], our results suggest that during the first 2 years of life, respiratory infections can alter the nasopharyngeal tract and critically contribute to the pathogenesis of pediatric OSA. Furthering our understanding of the mechanisms mediating the earliest origins of OSA will be highly impactful and relevant to manage this condition in children. Importantly, our findings also raise the possibility that novel anticipatory strategies and interventions can be developed to identify and prevent the initial establishment of OSA following severe viral respiratory infections during early infancy. This novel primary prevention approach for pediatric OSA would have a dramatic effect in reducing the increasing incidence of this condition and its multiple detrimental effects on childhood health and beyond.

Funding

The Boston Birth Cohort received support from the National Institutes of Health (NIH) grants U01AI090727, R21AI079872, R01HD086013, 2R01HD041702, and R01HD098232 (to XW). MJG was supported by the Johns Hopkins University Clinician Scientist Award, the American Academy of Allergy, Asthma and Immunology (AAAAI) Foundation Faculty Development Award and NIH grant K23HD104933. GN was funded by NIH grants R01HL141237, and R41HL145669, and R21AI130502. The funding agencies had no involvement in the writing of this review or in the decision to submit the article for publication.

Disclosure Statements

Financial Disclosure: None.

Nonfinancial Disclosure: None.

Data Availability

The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request after review and approval of the Institutional Review Board.

Author notes

M.J.G. and G.N. equal first author contribution.