Figures
Abstract
Background
Acute otitis media (OM) is a common disease which often develops through complex interactions between the host, the pathogen and environmental factors. We studied single nucleotide polymorphisms (SNPs) of genes involved in innate and adaptive immunity, and other host and environmental factors for their role in OM.
Methods
Using Sequenom Massarray platform, 21 SNPs were studied in 653 children from prospective (n = 202) and retrospective (n = 451) cohorts. Data were analyzed for the relationship between SNPs and upper respiratory infection (URI) frequency, risk of acute OM during URI episodes, and proneness to recurrent OM.
Results
Increased risk for OM proneness was associated with CX3CR1 (Thr280Met) SNP and with a jointly interactive group of IL-10 (−1082) SNP, IL-1β (−511) wild type genotype and white race. Family history of OM proneness independently increased the risk for frequent URIs, OM occurrence during URI, and OM proneness. Additionally, IL-1β (−31) SNP was associated with increased risk for frequent URIs, but IL-10 (−592), IL-1β (−511), IL-5 (−746) and IL-8 (−251) SNPs were associated with decreased risk of URI.
Citation: Nokso-Koivisto J, Chonmaitree T, Jennings K, Matalon R, Block S, Patel JA (2014) Polymorphisms of Immunity Genes and Susceptibility to Otitis Media in Children. PLoS ONE 9(4): e93930. https://doi.org/10.1371/journal.pone.0093930
Editor: Yungling Leo Lee, National Taiwan University, Taiwan
Received: December 9, 2013; Accepted: March 11, 2014; Published: April 9, 2014
Copyright: © 2014 Nokso-Koivisto et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: This work was supported by the National Institute of Deafness and Other Communication Disorders [grant number R01 DC005841], and from the National Center for Advancing Translational Sciences [grant number UL1TR000071], National Institutes of Health. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have the following interests: Dr. Stan Block is an employee of Kentucky Pediatric Research, Inc, Bardstown, Kentucky, which is a commercial company. This does not alter the authors′ adherence to PLOS One policies on sharing data and materials.
Introduction
Viral upper respiratory infection (URI) is the commonest infectious disease worldwide. Acute otitis media (OM) is a frequent complication of viral URI in children [1]. Some children are more susceptible to recurrent OM (OM-prone); the reasons for this are likely to be multifactorial, including the host genetic factors, the pathogen and environmental factors. High risk for OM proneness occurs in family clusters and specific ethnic populations, and twin and triplet studies have strongly suggested the heritability of recurrent OM [2]–[4].
Cytokines participate in the innate and adaptive immunity against infectious diseases. Many cytokines are actively induced in nasal secretions of children during viral URI, suggesting that these cytokines participate in regulation of virus-induced inflammation and recovery from infection. High levels of certain cytokines in respiratory secretions have been associated with the severity of respiratory disease [5]–[8]. We have also shown that high IL-1β levels in the nasopharynx are associated with the risk for acute OM during URI episode [9]. As single nucleotide gene polymorphisms (SNPs) of cytokine genes can modulate the production of respective cytokines, it is likely that these SNPs affect the risk for viral URI and OM. For example, we have shown an increased risk for URI as well as OM proneness with IL-6 (−174) and TNFα (−308) SNPs [10], [11]. OM proneness has been shown to be increased with IL10 (−1082), TLR4 (−299), CD14 (−159), and MBL (−54) SNPs [10], [12]–[15].
The present study aimed to investigate the role of several additional genes of innate and adaptive immunity in susceptibility to OM proneness and URI in association with host and environmental risk factors. Specifically, 21 SNPs of genes were selected based on their previously published roles in respiratory infections and OM [16], [17]. Since the study cohorts in this report were previously examined to evaluate the association of IL-1β (+3953), TNFα (−308), and IL-6 (−174) SNPs with URI and OM, data related to these three SNPs are not presented herein [10], [11].
Methods
Study Population
Children included in the present analysis are from two different retrospective and prospective study cohorts. The cohorts were enrolled from January 2003 through March 2007.
The retrospective study cohort.
This cohort was enrolled to assess the genetic risk factors for OM proneness in children, age 3 yrs or older, who had pre-determined OM prone or non-prone (control) status [10]. The children were enrolled at the outpatient general pediatrics clinics and the pediatric otolaryngology clinic at University of Texas Medical Branch, Galveston, TX and at the general pediatrics clinic at Kentucky Pediatric Research Office, Bardstown, KY. Clinical data were collected by interviewing the parents and by reviewing medical charts.
The prospective study cohort.
This cohort was followed longitudinally to investigate the incidence of URI and occurrence of AOM following URI [1]. In brief, healthy children, age 6 mos to 3 yrs were followed for one year to study the occurrences of URI and AOM. Parents informed the study personnel when the child developed URI symptoms (nasal congestion, rhinorrhea, cough and/or sore throat, with or without fever). Children were then seen by a study physician as soon as possible, and were followed for the occurrence of acute OM. At each visit, otoscopic and physical examinations were performed, and tympanometric data were recorded. Acute OM complicating URI was considered when it occurred within 28 days of the onset of URI. Acute OM was defined as 1) acute onset of symptoms, 2) signs of tympanic membrane inflammation, and 3) the presence of middle ear fluid as documented by pneumatic otoscopy and/or tympanometry.
In both cohorts, a blood sample or buccal mucosa swab was collected for DNA extraction at enrollment. Children in both cohorts were classified as OM-prone by one of the following criteria: 1) 3 or more episodes of OM within 6 mos.; 2) 4 or more episodes of OM within 12 mos.; 3) 6 or more episodes of OM by age 6 yrs., 4) first OM episode before age of 6 mos.; 5) history of tympanostomy tube placement for recurrent or persistent OM. Children were classified as non-OM-prone if they had only 0–1 episode of OM by age 2 yrs (with reliable documentation available in their medical records). Children with an anatomic or a physiologic defect of the ear or nasopharynx, known immunologic abnormality, or major medical conditions or treatment for chronic conditions were excluded from both cohorts.
Ethics Statement
The study was approved by the Institutional Review Board at University of Texas Medical Branch, Galveston, TX, USA. Informed written consent was obtained from the parents of all participating children.
Data Sharing Plan
We intend to make our original data available to interested nonaffiliated scientists. We will review each written request and honor each legitimate requests that will serve to validate our data and will advance the field provided that this request does not compromise our right to first publication of results that directly address the aims of the grant.
SNP Assays
The whole genome DNA was extracted from the peripheral blood mononuclear cells or buccal epithelial cells and stored at −70C until further use. The specific 21 SNPs studied are shown in Table 1; C–C chemokine receptor type 5 (CCR5) −2554, CX3C chemokine receptor 1 (CX3CR1) 280, Inter-Cellular Adhesion Molecule 1 (ICAM1) K469E and I20788, Interleukin 1β (IL-1β) −31, −511, Interleukin 2 (IL-2) −330, Interleukin 5 (IL-5) −746, Interleukin 8 (IL-8) −251, Interleukin 10 (IL-10) −1082 and −592, Interleukin 12 (IL-12) −1188, Interleukin 13 (IL-13) −1055, Interleukin 18 (IL-18) 133, Mannose-binding lectin (MBL) gly54asp, RANTES −403, Transforming growth factor β (TGF-β1) −509 Toll-like receptor (TLR4) Asp299Gly and Thr399Ile, Tumor necrosis factor α (TNFα) −238, −376.
The SNPs were analyzed at the Center for Genotyping and Analysis of the Broad Institute of Massachusetts Institute of Technology, Cambridge, MA, using Sequenom MassARRAY platform as previously described [18].
Statistical Analyses
An elastic net Poisson model [19] was used to model the number of URI episodes and OM occurrence during URI episodes. To model proneness, logistic regression with an elastic net penalty was used. All models included demographic factors such as breast feeding (any vs. none), day care (any vs. home care), exposure to cigarette smoke (any vs. none), family OM history (OM susceptibility in immediate family members; yes vs. no), as well as measured SNP genotypes which each had three categories ‘wild type’ (predominant genotype), ‘heterozygous polymorphism’ and ‘homozygous polymorphism’. Tuning parameters were set using out-of-sample error likelihood estimation based on 10-fold cross validation. All statistical procedures were run using libraries in the R programming environment (http://cran.r-project.org/). All of the models included IL-1β (+3953), TNFα (−308), and IL-6 (−174) SNPs, but the data related to these SNPs are not shown as they have been published previously. Additionally, since each study enrolled cohorts of differing ages, the age effect was not reported due to possible confounding.
Results
SNP Assay Evaluation
Altogether we had results of 21 SNPs from DNA of 747 children. However, 94 children were excluded from the analysis because of missing or inaccurate data related to OM proneness classification or environmental risk factors. In the final analysis, we used data from 653 children whose information was complete.
The distribution of the allele and genotype frequencies of 21 SNPs is shown in Table 1. The hetero- or homozygous SNP frequencies were less than 10% for TGF-β1 (−509), TLR4 (Asp299Gly), TLR4 (Thr399Ile), TNFα (−238) and TNFα (−376). The SNP frequencies exceeded 50% for CCR5 (−2554), ICAM1 (K469E), ICAM1 (20788), IL-1β (−31), IL-1β (−511), IL-5 (−746), IL-8 (−251), IL-10 (−1082), IL-10 (−592) and TGF-β1 (−509).
Population Characteristics
The demographic and clinical characteristics of 653 study children are given in Table 2. The median age of enrollment in the retrospective cohort (4 years) was higher than in the prospective cohort (1 year) due to the enrollment criteria for each cohort. The racial/ethnicity distribution of the study subjects in the retrospective cohort reflected the general population distribution at the two study sites (Texas and Kentucky). The prospective cohort represented the distribution of general population at the Texas study site only. The retrospective cohort also had a higher number of children with OM proneness because the study using this cohort was designed to compare equivalent number of children who were OM prone or non-prone. Statistical analysis revealed no site specific effect on the results reported below.
SNPs vs. Risk for OM Proneness
The risk of OM proneness was analyzed in 653 children from both the prospective and retrospective cohorts (Table 3). Increased risk for OM proneness was independently associated with CX3CR1 (Thr280Met) polymorphic genotype, family history of OM, attendance at day care, and lack of breastfeeding. Furthermore, IL-10 (−1082) SNP, IL-1B (−511) wild type genotype and white race predicted OM proneness only when analyzed together in a joint hypothesis (ie. these factors are not individually independent predictors of OM proneness). The risk of tympanostomy tube placement was positively associated with IL-2 (−330) hetero- or homozygous SNP, TGFβ1 (−509) wild type genotype, and male gender (data not shown).
SNPs vs. Risk for Frequent URI and OM Occurrence during URI Episode
In the prospective study cohort of 202 children, the number of URI episodes and occurrence of OM during URI episode during the one-year follow-up period were analyzed. IL-1β (−31) homozygous SNP was associated with increased risk for frequent URIs while IL-10 (−592), IL-1β (−511), IL-5 (−746) and IL-8 (−251) homozygous SNPs were associated with decreased risk (Table 4). Furthermore, decreased risk of OM occurrence during URI episodes was associated with IL-10 (−592) homozygous SNP (Table 5). Family history of OM proneness increased the risk for both frequent URIs and OM occurrence during URI episodes, while day care attendance was associated only with OM occurrence during URI episodes (Tables 4 and 5).
Discussion
We have previously shown that IL6 (−176) and TNFα (−308) SNPs are associated with frequent URI and OM proneness [10], [11]. In the present study, we further examined the role of additional 21 SNPs of immunoregulatory genes in URI and OM in children in the same study cohorts. Our data show the significant role of specific SNPs that promote or protect against URI and OM.
IL-1β is a major mediator of inflammation that plays an important role in tissue injury repair as well as in the defense against microbial pathogens. We found that IL-1β (−31) SNP was associated with increased risk for frequent URI. Chen et al have also shown increased susceptibility to chronic rhinosinusitis in association with this polymorphism [20]. On the other hand, our study showed that another functional IL-1β SNP, (−511), was associated with decreased risk for OM after URI. The reasons for the divergent results of two separate SNPs of the same cytokine gene are unknown. Watanabe et showed that IL-1β (−511) was associated with worsened systemic inflammation in sepsis [21]. In our previous study, we showed that an additional IL-1β SNP (+3954) was associated with more severe acute OM [22].
In this study, IL-10 (−592) was found to protect against both URI and occurrence of OM during URI episodes. This result is consistent with the known role of IL-10 as an anti-inflammatory cytokine; this property may result in reduced inflammation in the nasopharynx during viral infection. However, Alpert et al reported an increased risk for OM in children with IL-10 (−592) SNP during URI due to rhinovirus and respiratory syncytial virus [13]. The reasons for this discrepancy are not clear, but the subjects in different studies may have different counter-regulatory genes or differing local environmental influences that alter the susceptibility profile to disease. Furthermore, the size of our study population is much larger, and we studied the influence of genes on OM proneness to recurrent disease while Alpert et al studied the risk of OM after two specific viral upper respiratory infections. We also found that a different IL-10 (−1082) SNP was associated with OM proneness, although not with the increased frequency of URI or OM during URI episode.
IL-5 (−746) and IL-8 (−251) SNPs were found to be protective against frequent URIs. However, this effect differs from other reported effects in which they promote lower airway inflammation. For example, IL-5 (−746) SNP has been associated with worse lung functions in asthmatic children [23] and IL-8 (−251) SNP with an increased risk for bronchiolitis with respiratory syncytial virus infection [24]. It is likely that these polymorphic genes have a different influence on disease susceptibility at different body sites because different elements of host immune defenses may be required at the specific site of infection.
CX3CR1 (Thr280Met) SNP was associated with an increased risk for OM proneness. While this polymorphism is relatively rare, 9 out of 10 children with this polymorphism were OM prone which was the highest rate for any SNP. Previously, this genotype has been associated with an increased risk for lower airway complication due to infection with respiratory syncytial virus [25]. CX3CR1 is a cellular receptor on leukocytes which binds to CX3C chemoattractants. The enhanced chemotactic activity associated with the CX3CR1 (Thr280Met) may result in increased inflammation, thereby predisposing to OM.
In a subset of our subjects with tympanostomy tubes, the genotypes associated with the risk for tubes were different from the OM prone group as a whole. IL-2 (−330) hetero- or homozygous SNP, TGFβ1 (−509) wild type genotype increased the risk for tube placement. IL-2 (−330) SNP is known to increase the risk for respiratory infections [26]. TGFβ1 protein is involved in tissue healing leading to scar formation. Its role in chronic OM leading to granulation tissue in the middle ear has been demonstrated in experimental models [27]; however, the role of TGFβ1 (−509) SNP in URI and OM has not been explored.
As with many past studies, our study found that family history of OM proneness was associated with OM proneness in the child; however, we additionally showed that it also increases the risk for frequent URIs and occurrence of OM during URI. This observation strongly suggests that the URI and OM risks have strong familial pattern of aggregation which could be both due to genetic as well as environmental factors.
As with other published studies, we found that the history of breastfeeding was associated with a lower risk for OM proneness. However, we also showed that breastfeeding does not influence the risk for frequent URIs; this observation is similar to that of Chantry et al who studied U.S. children [28]. Duijts et al showed that Dutch children who were exclusively breastfed for at least 4 months had lower incidence of URIs, but protection lasted only during the first 6 mos of age [29]. Because our population was older than 6 months (median age of 1 yr), lack of effect on URI is consistent with previously published observation.
Overall, our study highlights the complex, interactive, positive and negative influences of the host immunoregulatory genes and environmental factors on susceptibility to URI and AOM. For example, IL-10 (−1082) SNP, IL-1B (−511) wild type genotype and white race predicted OM proneness only when analyzed together in a joint hypothesis (ie. these factors are not individually independent predictors). Additional statistical tools are needed to further dissect the gene, pathogen and environmental interactions. Some of the SNPs produced counterintuitive effects, suggesting that the immune regulatory genes may influence the disease manifestation through multiple intricate pathways with counter current loops of interaction.
The strength of our study is the large number of children who were studied and the wider selection of genetic SNPs that were analyzed. However, the study is limited by the relatively low frequencies of some of the SNPs in our population. Furthermore, since the predictive measures are relatively low, these models are best considered prognostic, characterizing the effective genetic component rather than predictive of any particular outcome. Currently, as none the SNP tests are available for routine clinical use, they cannot be used to guide patients for their individual risk. On the other hand, patients with OM proneness can be counselled that they may have inherited certain high risk genes as an explanation for their proneness. In addition, our study suggests that change in behavior such as increasing breastfeeding may be beneficial for disease reduction regardless of inheritance of disease susceptible genes.
Conclusion
In addition to previously reported TNFa (−308) and IL-6 (−174) SNPs, we found additional SNPs of immunoregulatory genes which were associated with frequent URIs or OM proneness; these are CX3CR1 (Thr280Met), IL-10 (−1082) and IL-1β (−31). Additionally, IL-10 (−592), IL-1β (−511), IL-5 (−746) and IL-8 (−251) SNPs were found to have a protective role in URI. Further studies are needed to better understand the host genetic, pathogen and environmental factors in order to predict the child at risk for frequent URI and OM and to develop novel interventions for prevention and treatment.
Acknowledgments
We thank Meagan Harden of The Center for Genotyping and Analysis at the Broad Institute of MIT and Harvard for DNA genotyping. We also thank Krystal Revai, Ron L Veselenak, Sangeeta Nair, M. Lizette Rangel, Kyralessa B. Ramirez, Syed Ahmad, Michelle Tran, Liliana Najera, Rafael Serna and Carolina Pillion for assistance with study subjects, and specimen collection and processing.
Previous Presentation
The data has been in part presented in Pediatric Academic Societies Annual Meeting, Boston, MA, May 2012 and in the 7th Extraordinary International Symposium on Recent Advances in Otitis Media, Stockholm, Sweden, June 2013.
Author Contributions
Conceived and designed the experiments: JN TC KJ RM SB JAP. Performed the experiments: JN TC. Analyzed the data: JN TC KJ RM JAP. Contributed reagents/materials/analysis tools: KJ SB. Wrote the paper: JN TC KJ RM SB JAP.
References
- 1. Chonmaitree T, Revai K, Grady JJ, Clos A, Patel JA, et al. (2008) Viral upper respiratory tract infection and otitis media complication in young children. Clin Infect Dis 46: 815–823.
- 2. Casselbrant ML, Mandel EM, Fall PA, Rockette HE, Kurs-Lasky M, et al. (1999) The heritability of otitis media: a twin and triplet study. JAMA 282(22): 2125–2130.
- 3. Rovers M, Haggard M, Gannon M, Koeppen-Schomerus G, Plomin R (2002) Heritability of symptom domains in otitis media: a longitudinal study of 1,373 twin pairs. Am J Epidemiol 155(10): 958–964.
- 4. Rye MS, Blackwell JM, Jamieson SE (2012) Genetic susceptibility to otitis media in childhood. Laryngoscope 122(3): 665–75.
- 5. Fritz RS, Hayden FG, Calfee DP, Cass LM, Peng AW, et al. (1999) Nasal cytokine and chemokine responses in experimental influenza A virus infection: Results of a placebo-controlled trial of intravenous zanamivir treatment. J Infect Dis 180: 586–93.
- 6. Hornsleth A, Larsen LL (2001) Cytokines and chemokines in respiratory secretion and severity of disease in infants with respiratory syncytial virus infection. J Clin Virol 21: 163–70.
- 7.
Skoner DP, Gentile DA, Patel A, Doyle WJ (1999) Evidence for cytokine mediation of disease expression in adults experimentally infected with influenza A virus. J Infect Dis 180-10-4.
- 8. Turner RB, Weingand KW, Chyon-Hwa Y, Leedy DW (1998) Association between interleukin-8 concentration in nasal secretions and severity of symptoms of experimental rhinovirus colds. Clin Infect Dis 26: 840–6.
- 9. Patel JA, Nair S, Revai K, Grady J, Chonmaitree T (2009) Nasopharyngeal acute phase cytokines in viral upper respiratory infection: Impact on acute otitis media in children. Pediatr Infect Dis J 28: 1002–7.
- 10. Patel J A, Nair S, Revai K, Grady J, Saeed K, et al. (2006) Association of proinflammatory cytokine gene polymorphisms with susceptibility to otitis media. Pediatrics 118(6): 2273–2279.
- 11. Revai K, Patel JA, Grady JJ, Nair S, Matalon R, et al. (2009) Association between cytokine gene polymorphisms and risk for upper respiratory tract infection and acute otitis media. Clin Infect Dis 49(2): 257–261.
- 12. Emonts M, Veenhoven RH, Wiertsema SP, Houwing-Duistermaat JJ, Walraven V, et al. (2007) Genetic polymorphisms in immunoresponse genes TNFA, IL6, IL10, and TLR4 are associated with recurrent acute otitis media. Pediatrics 120(4): 814–23.
- 13. Alper CM, Winther B, Hendley JO, Doyle WJ (2009) Cytokine polymorphisms predict the frequency of otitis media as a complication of rhinovirus and RSV infections in children. Eur Arch Otorhinolaryngol 266(2): 199–205.
- 14. Wiertsema SP, Herpers BL, Veenhoven RH, Salimans MM, Ruven HJ, et al. (2006) Functional polymorphisms in the mannan-binding lectin 2 gene: effect on MBL levels and otitis media. J Allergy Clin Immunol 117(6): 1344–50.
- 15. Wiertsema SP, Khoo SK, Baynam G, Veenhoven RH, Laing IA, et al. (2006) Association of CD14 promoter polymorphism with otitis media and pneumococcal vaccine responses. Clin Vaccine Immunol 13(8): 892–7. Erratum in: Clin Vaccine Immunol. 2006 13(12): 1375.
- 16. Miyairi I, DeVincenzo JP (2008) Human genetic factors and respiratory syncytial virus disease severity. Clin Microbiol Rev 21(4): 686–703.
- 17. Tregoning JS, Schwarze J (2010) Respiratory viral infections in infants: Causes, clinical symptoms, virology and immunology. Clin Micribiol Rev 23: 74–98.
- 18.
Gabriel S, Ziaugra L, Tabbaa D (2009) SNP genotyping using the Sequenom MassARRAY iPLEX platform. Curr Protoc Hum Genet. Chapter 2: Unit 2.12.
- 19. Friedman J, Hastie T, Tibshirani R (2010) Regularization Paths for Generalized Linear Models via Coordinate Descent. J Stat Softw 33(1): 1–22.
- 20. Chen N, Ni P, Cai C, Wang S (2011) Association of IL-1beta gene -31T/C polymorphism with chronic rhinosinusitis. Lin Chung Er Bi Yan Hou Tou Jing Wai Ke Za Zhi 25: 197–200.
- 21. Watanabe E, Hirasawa H, Oda S, Shiga H, Matsuda K, et al. (2005) Cytokine-related genotypic differences in peak interleukin-6 blood levels of patients with SIRS and septic complications. J Trauma 59: 1181–90.
- 22. McCormick DP, Grady JJ, Diego A, Matalon R, Revai K, et al. (2011) Acute otitis media severity: Association with cytokine gene polymorphisms and other risk factors. Int J Pediatr Otorhinolaryngol 75: 708–712.
- 23. Hong SJ, Lee SY, Kim HB, Kim JH, Kim BS, et al. (2005) IL-5 and thromboxane A2 receptor gene polymorphisms are associated with decreased pulmonary function in Korean children with atopic asthma. J Allergy Clin Immunol. 115: 758–63.
- 24. Hull J, Thomson A, Kwiatkowski D (2000) Association of respiratory syncytial virus bronchiolitis with the interleukin 8 gene region in UK families. Thorax. 55(12): 1023–7.
- 25. Amanatidou V, Sourvinos G, Apostolakis S, Tsilimigaki A, Spandidos DA (2006) T280M variation of the CX3C receptor gene is associated with increased risk for severe respiratory syncytial virus bronchiolitis. Pediatr Infect Dis J 25(5): 410–4.
- 26. Nieters A, Brems S, Becker N (2001) Cross-sectional study on cytokine polymorphisms, cytokine production after T-cell stimulation and clinical parameters in a random sample of a German population. Hum Genet 108(3): 241–8.
- 27. Lee YW, Chung Y, Juhn SK, Kim Y, Lin J (2011) Activation of the transforming growth factor beta pathway in bacterial otitis media. Ann Otol Rhinol Laryngol. 120(3): 204–13.
- 28. Chantry CJ, Howard CR, Auinger P (2006) Full breastfeeding duration and associated decrease in respiratory tract infection in US children. Pediatrics 117: 425–32.
- 29. Duijts L, Jaddoe VW, Hofman A, Moll HA (2010) Prolonged and exclusive breastfeeding reduces the risk of infectious diseases in infancy. Pediatrics 126: e18–25.