CONTEXT: The increasing prevalence of childhood asthma has been associated with low microbial exposure as described by the hygiene hypothesis.
OBJECTIVE: We sought to evaluate the evidence of association between antibiotic exposure during pregnancy or in the first year of life and risk of childhood asthma.
METHODS: PubMed was systematically searched for studies published between 1950 and July 1, 2010. Those that assessed associations between antibiotic exposure during pregnancy or in the first year of life and asthma at ages 0 to 18 years (for pregnancy exposures) or ages 3 to 18 years (for first-year-of-life exposures) were included. Validity was assessed according to study design, age at asthma diagnosis, adjustment for respiratory infections, and consultation rates.
RESULTS: For exposure in the first year of life, the pooled odds ratio (OR) for all studies (N = 20) was 1.52 (95% confidence interval [CI]: 1.30–1.77). Retrospective studies had the highest pooled risk estimate for asthma (OR: 2.04 [95% CI: 1.83–2.27]; n = 8) compared with database and prospective studies (OR: 1.25 [95% CI: 1.08–1.45]; n = 12). Risk estimates for studies that adjusted for respiratory infections (pooled OR: 1.16 [95% CI: 1.08–1.25]; n = 5) or later asthma onset (pooled OR for asthma at or after 2 years: OR: 1.16 [95% CI: 1.06–1.25]; n = 3) were weaker but remained significant. For exposure during pregnancy (n = 3 studies), the pooled OR was 1.24 (95% CI: 1.02–1.50).
CONCLUSIONS: Antibiotics seem to slightly increase the risk of childhood asthma. Reverse causality and protopathic bias seem to be possible confounders for this relationship.
Asthma, which has an estimated 10% worldwide prevalence among children aged 6 to 7 years, is one of the most common chronic childhood diseases.1 Observations that developed countries show higher prevalences of asthma than developing countries1 have led to the “hygiene hypothesis,” which postulates that childhood development in an overly hygienic environment reduces microbial exposure and promotes atopic immune responses and the risk of asthma.2 Results of experiments in mice have supported this hypothesis, demonstrating that bacterial gut colonization in early life plays a significant role in acquired immune system programming.3,–,6 In utero microbial exposures have also been shown to protect against allergic phenotypes in mice.7,–,9 Perturbation to normal microbial exposure during gestation or infancy, therefore, may elevate the risk of atopic disease including asthma. In consideration of this hypothesis, numerous studies to date have attempted to evaluate the association between antenatal and infant antibiotic usage and subsequent development of asthma.
This research is highly susceptible to several possible biases that confound the true association between antibiotics and asthma.10 Reverse causality, which occurs when the outcome precedes and increases the risk for the putative exposure, may be a strong source of bias for studies that fail to accurately determine the relative timing of antibiotic use and the onset of asthma. This is a particular concern because patients with asthma may receive antibiotics at elevated rates, either by misprescription or for treatment of comorbidities.11 Studies that assess outcomes of ever or current, but not incident (ie, onset), asthma may be susceptible to reverse causality even if measured after the time of exposure. A specific type of reverse causality is protopathic bias, which occurs when early symptoms of undiagnosed asthma are attributed mistakenly to respiratory infection and are treated with antibiotics. Diagnosis of asthma in children may occur up to several years after the onset of respiratory symptoms,12,13 which suggests that studies that use a short “wash-out” period between antibiotic exposure and outcome assessment may be susceptible to protopathic bias. Detection bias, which occurs when the likelihood of diagnosis is influenced by the patient's frequency of physician consultations, may also be an issue.12 Indication bias, which occurs when respiratory infections are treated with antibiotics while being independent risk factors for asthma, is likewise a strong possibility.14,–,16 Finally, recall bias may be a confounder for retrospective studies, because parents of children with asthma may be more likely to recall early life exposures.
The purpose of this review was to systematically evaluate evidence from studies that addressed 2 questions: “Does antibiotic exposure in the first year of life increase the risk of asthma in children aged 3 to 18 years?” and “Does antibiotic exposure during pregnancy increase the risk of asthma in offspring by the age of 18 years?” For the first question, the separation of at least 2 years between exposure and outcome was chosen to reduce the likelihood of reverse causality or protopathic bias in the included studies.
A previous systematic review in 2006 by Marra et al17 similarly examined the evidence for association between antibiotic exposure in the first year of life and childhood asthma, but the results of numerous studies that addressed this question have been published since then.
Evidence of a true association between antibiotic use and childhood asthma would further support the reduction of unnecessary antibiotic prescriptions during pregnancy and in early life and add support for the hygiene hypothesis.
A completed PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-analyses) 2009 checklist for this systematic review is available in Supplemental Table 3. A protocol for this review has not been previously published.
Search Strategy and Study-Selection Criteria
PubMed was searched between the date of index start (1950) and July 1, 2010. The search was not restricted by language. The following query string was used: (child or childhood or “early life” or “early child” or postnatal or pediatric or prenatal or antenatal or pregnancy or “in utero” or maternal or infant or mothers or mother or “maternal exposure” or “maternal exposures” or “prenatal exposure delayed effects” or “prenatal care”) and (asthma* or wheezing or atopy or COPD or “chronic obstructive pulmonary disease” or bronchitis or bronchiolitis) and (antibiotic* or antimicrobial or antibacterial or “anti-bacterial agents”).
A publication was selected for further evaluation if it met inclusion criterion of evaluating an association between exposure to antibiotics and an outcome of childhood asthma. Outcomes defined as “wheeze” or “wheezing” but not asthma were not included. A study was excluded from final analysis if it met 1 of the following criteria: (1) there was no reported age of exposure or outcome; (2) only mixed exposures or outcomes were assessed (ie, exposure to antibiotics and another factor or development of asthma and another condition); (3) antibiotic exposure was assessed after 1 year of age only; (4) asthma was assessed before 3 years only (for studies of childhood exposure); (5) outcomes were assessed after 18 years of age only; or (6) there was a lack of a reported statistical measure of association or enough data to allow for its calculation. If a study assessed asthma that could have been diagnosed before 3 years but also included any of years 3 to 18 in its follow-up, it was still included in this study. Each included study was scored on the basis of a modified Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) checklist,18 and studies were excluded if they had <50% of STROBE checklist criteria.
Studies were evaluated and abstracted by 2 independent reviewers (Mr Murk and Dr Risnes); discordant data abstraction was resolved by consensus and discussion with a third reviewer (Dr Bracken). For each included study, data were collected regarding study design, population size in each study group, criteria for eligibility, time of antibiotic use, type of antibiotic used, number of prescriptions for antibiotics, reasons for antibiotic use, method of determining antibiotic exposure, time of asthma diagnosis, method of determining asthma diagnosis, risk estimates with 95% confidence intervals (CIs), respiratory infections between ages 0 and 1 year, number of medical consultations (visits to a physician) between ages 0 and 1 year, family history of asthma, and other covariates used to adjust the risk estimates. If the authors of a study reported multiple comparisons of antibiotic and exposure, the comparison that was judged to be the least biased (eg, had an older first possible age of asthma diagnosis or was adjusted for respiratory infections or other factors mentioned above) was included for the final analysis. When the authors reported an adjusted estimate for the whole cohort, as well as for a subgroup with asthma first diagnosed at or after 2 years of life, the overall results were included in the appropriate main analysis, and the latter was included in comparisons of age of diagnosis-related subgroups.
Retrospective studies were defined as those studies that assessed exposures by self-report after they had already occurred. Database studies were assessed as a separate type of study design and were defined as those that assessed exposures via a medical practice or administrative database.
The effect measure used in the meta-analysis was the adjusted odds ratio (OR), relative risk (RR), or hazard ratio (HR) for asthma. If no adjusted effect measure was provided, then the crude measure was used. Meta-analysis was performed by using random-effects and fixed-effects models; the fixed-effects models were only used when low or moderate heterogeneity was detected. When not otherwise specified, pooled estimates mentioned in the text are from a random-effects model. Effect measures were weighted by log inverse variance. Heterogeneity was assessed by using Cochran's Q (χ2) statistic (P < .10 was considered significant evidence for heterogeneity), the I2 test (values of >40% and >75% were considered to indicate moderate and high heterogeneity, respectively), and visual inspection of Galbraith plots. Sources of heterogeneity were investigated by performing subgroup sensitivity analyses of studies on the basis of the following prespecified factors: study design (retrospective, database, or prospective study design); age of asthma onset (0–1 vs ≥2 years); asthma definition (doctor-diagnosed versus parental report of “having asthma”); use of an RR/HR or an OR effect estimate; adjustment for respiratory infections; adjustment for number of medical consultations; and adjustment for family history of asthma. Comparison of pooled effect estimates was performed by using a test of interaction for normally distributed differences in effect estimates.19 Publication bias was evaluated by visual inspection of funnel plots and by using Egger's asymmetric distribution test; P < .10 was considered significant.20
Statistical significance for pooled estimates was defined by using α = .05. Calculations were performed by using Review Manager (RevMan) 5.0 (Cochrane Collaboration, Oxford, United Kingdom [www.cc-ims.net/revman]) and SAS 9.0 (SAS Institute Inc, Cary, NC).
The literature search identified 2071 citations, of which 29 assessed the association between antibiotics and asthma (Fig 1). Of these 29 studies, 721,–,27 were excluded because they met exclusion criterion (see Supplemental Table 4): 3 studies only examined antibiotic exposures after 1 year of age, and 4 studies only examined mixed outcomes (asthma with another morbidity). The remaining 22 studies were included in the final analysis. Characteristics of the included studies that assessed exposure in infancy28,–,47 (n = 20) are listed in Table 1, and studies of in utero exposures40,48,49 (n = 3) are listed in Table 2. Two of the included studies met strict exclusion criteria by exceeding age of exposure (Wickens et al45; exposure age: 0–15 months) or outcome (Sobko et al43; asthma outcome age: 7–23 years [median age: 7]) limits, but they were retained because it was judged that these ages of exposure/outcome were not substantially different from those in the rest of the studies. Several included studies presented multiple estimates each for risk of asthma (ie, by conducting sensitivity or subgroup analyses); in such cases, the estimate that was judged to be the least susceptible to bias while retaining the most statistical power was included in the meta-analysis. A detailed list of all estimates eligible for inclusion, as well as rationale for selecting the estimate that was used, are provided in Supplemental Table 5 (early life exposures) and Supplemental Table 6 (in utero exposures).
Exposure to Antibiotics in the First Year of Life and Risk of Childhood Asthma
Study-specific and pooled risk estimates of asthma after any exposure to antibiotics in the first year of life are provided in Fig 2. The pooled OR for the adjusted estimates of all 20 applicable studies was 1.52 (95% CI: 1.30–1.77). However, this estimate showed high heterogeneity (I2 = 95%), which supports the need for subgroup analysis. For the main subgroup analysis, studies were analyzed according to design type (viz, retrospective, database, or prospective). Retrospective studies (n = 8) showed the highest pooled estimate of any subgroup (pooled OR: 2.04 [95% CI: 1.83–2.27]). These studies contributed significantly to the heterogeneity among all included studies while having moderate intragroup heterogeneity (I2 = 45%). Database studies had a significantly weaker pooled estimate (n = 6; OR: 1.35 [95% CI: 1.12–1.62]) (test for interaction, retrospective versus database subgroup: P = .0001). Prospective studies revealed the weakest pooled estimate, which revealed no significant effect of antibiotics on asthma (n = 6; OR: 1.07 [95% CI: 0.89–1.28]) (retrospective or database versus prospective subgroup: P < .000001 and P = .075, respectively). However, prospective studies were small (total cohort size: 5133 subjects) compared with the total database cohort size (400 167 subjects). Significant heterogeneity was detected among database studies (I2 = 96%) but not prospective studies (I2 = 0%).
To further investigate heterogeneity according to factors other than study design, sensitivity analyses were performed by using prospective and database studies combined (retrospective studies were excluded because of the clear effect on heterogeneity, as discussed in the previous section). The main findings for the sensitivity analyses are provided in Fig 3, and Supplemental Table 7 summarizes all sensitivity analyses in detail. The pooled OR for prospective and database studies combined was 1.25 (95% CI: 1.08–1.45; I2 = 92%; n = 12). Stratifying this group according to age of first possible asthma incidence revealed that studies with asthma at ages 0 to 1 year had a higher pooled estimate (n = 9; OR: 1.26 [95% CI: 1.08–1.47]; I2 = 51%) than those with asthma at ages 2 to 3 (n = 3; OR: 1.16 [95% CI: 1.06–1.25]; I2 = 43%), although this difference was not significant. This effect was driven by database studies, because there were no prospective studies that assessed asthma at the age of 2 years or older. Stratifying according to adjustment or no adjustment for respiratory infections revealed that studies that did not adjust had a higher pooled estimate (n = 7; OR: 1.38 [95% CI: 1.15–1.65]; I2 = 64%) than those that did adjust (n = 5; OR: 1.16 [95% CI: 1.08–1.25]; I2 = 30%) (test for interaction: P = .079). This result was also observed when restricted to database studies but not when restricted to prospective studies. Figure 3B provides a Galbraith plot that illustrates study heterogeneity when stratifying according to adjustment for respiratory infections. Note that the 2 largest studies, by Martel et al40 and Marra et al,39 strongly influenced study heterogeneity and the results of the stratification analyses. Although both of them were large database studies, Martel et al40 examined asthma at age 0 or older, did not adjust for respiratory infections, and reported a relatively strong estimate (RR: 1.59 [95% CI: 1.50–1.68]), whereas Marra et al39 examined asthma at the age of 2 years or older, adjusted for respiratory infections, and reported a relatively weak estimate (HR: 1.12 [95% CI: 1.08–1.16]). Note that Martel et al40 included a subgroup analysis of subjects with asthma at the age of 3 years or older and found a weaker effect estimate (OR: 1.29 [95% CI: 1.12–1.49]) than their main analysis of asthma (Supplemental Table 5); this subgroup was used as the estimate for the Martel et al40 study when sensitivity analyses for asthma diagnosis age were performed.
Subgrouping the studies by use of doctor-diagnosed asthma, adjustment for number of medical consultations in the first year of life, or adjustment for family history of asthma did not produce any clear differences in pooled effect estimates (Supplemental Table 7). Among database studies, those that used an RR or HR effect estimate type generally had higher pooled estimates than those that used OR estimates (Supplemental Table 7).
Two of the prospective studies, by Celedón et al30 and Kusel et al,37 recruited exclusively “high-risk” children with a family history of allergic disease. The pooled OR for these studies was 0.79 (95% CI: 0.48–1.28; I2 = 0%) (Supplemental Table 7). The pooled OR for other (non–high-risk) prospective studies was 1.12 (95% CI: 0.92–1.36; I2 = 0%), although this difference was not significant.
Funnel-plot inspection did not reveal evidence of publication bias when all studies were included (Supplemental Fig 6). Likewise, Egger's test for asymmetric distributions did not detect significant publication bias when considering all studies (P = .19), retrospective studies (P = .13), database studies (P = .52), and prospective studies (P = .77).
Analysis of Dose-Specific Exposure to Antibiotics in Early Life and Risk of Asthma
Ten studies that assessed the risk of asthma after early-life exposure to antibiotics included estimates by number of courses of antibiotics.28,–,30,36,–,39,41,44,46 Five of these studies28,29,36,39,41 allowed comparisons of 1 to 2, 3 to 4, and >4 courses versus none (Fig 4). Pooled ORs of these studies revealed that there was a stronger risk of asthma for increased numbers of doses. However, when the study by Ahn et al28 (the only retrospective study) was removed, the pooled estimates for >4 and 3 to 4 courses were nearly the same. No pairwise comparison with the 1- to 2-course subgroup showed a significant test for interaction for either analysis.
Five additional studies30,37,38,44,46 examined dose-specific effects but did not have directly comparable strata. When combined into strata with the previous 5 studies (n = 10), the pooled OR for assessment of >1 vs 0 or >2 courses vs 0 was 1.49 (95% CI: 1.24–1.80; I2 = 76%), whereas the pooled OR for assessment of 1 course vs 0 or 1 to 2 courses vs 0 was 1.24 (95% CI: 1.10–1.40; I2 = 72%) (test for interaction: P = .11). When the 2 retrospective studies were removed (Wickens et al46 and Ahn et al28) (n = 8), these pooled estimates became 1.37 (95% CI: 1.14–1.65; I2 = 74%) and 1.15 (95% CI: 1.08–1.23; I2 = 20%), respectively (test for interaction: P = .08).
Exposure According to Type of Antibiotics and Risk of Asthma
Four studies assessed the effect of type of antibiotic on the association between exposure in the first year of life and childhood asthma, but because of high heterogeneity they could not be meta-analyzed. McKeever et al41 reported that although amoxicillin, macrolides, cephalosporins, and augmentin were all associated with a significantly increased risk of asthma, no risk with penicillin was revealed. In contrast, Marra et al39 reported that penicillin exposure was associated with a relatively high risk for asthma and that amoxicillin, macrolides, cephalosporins, and sulfonamide were also significantly associated with asthma, whereas all other types of antibiotics were not. Kozyrskyj et al36 grouped antibiotics into categories of “narrow-spectrum” (penicillin, cloxacillin, cephalexin, cefadroxil, and erythromycin) and “broad-spectrum” (all other classes) antibiotics and found that the OR for narrow-spectrum antibiotics was 1.35 (95% CI: 0.29–6.23), whereas for broad-spectrum antibiotics the OR was 1.50 (95% CI: 1.16–1.93). Finally, Celedón et al29 reported that there was no association according to any class of antibiotic.
Exposure to Antibiotics During Pregnancy and Risk of Asthma
Three studies assessed whether exposure to antibiotics in pregnancy was associated with childhood asthma (Fig 5). The pooled estimate for these studies revealed an increased risk with borderline significance (pooled OR: 1.24 [95% CI: 1.02–1.50]; I2 = 80%). The only prospective study had the highest effect estimate (OR: 1.70 [95% CI: 1.11–2.60]) compared with the other 2 studies, which were of database design (pooled OR of the 2 database studies: 1.38 [95% CI: 1.07–177]; I2 = 73%). The 2 studies that did not adjust for medical consultation rates had a higher pooled estimate (pooled OR: 1.65 [95% CI: 1.32–2.07]; I2 = 0%) than the study that did (RR: 1.09 [95% CI: 1.06–1.12]). The estimate for this latter study, Martel et al,40 is the RR per each additional antibiotic prescription, which may explain the smaller effect estimate.
With this review we aimed to determine if exposure to antibiotics during pregnancy or in early life increased the risk for childhood asthma. On the basis of the heterogeneity analysis, nonretrospective studies that adjusted for respiratory infections likely had the greatest validity in assessing early-life exposures. The pooled estimate of effect for these studies (OR: 1.13 [95% CI: 1.10–1.17]; n = 5; fixed effects) is also in general agreement with stratification according to studies that only examined asthma at the age of 2 years or older (pooled OR: 1.13 [95% CI: 1.09–1.17]; n = 3; fixed effects), which suggests that there is a significant but weak association between antibiotic use in early life and childhood asthma.
Assuming a baseline childhood asthma incidence rate of 10%, an OR of 1.13 (95% CI 1.10–1.17) yields a number needed to harm of 87 (95% CI: 67–113); that is, for every 87 children exposed to antibiotics, 1 child will develop asthma because of his or her antibiotic exposure. However, this number is highly uncertain, because it is sensitive to the assumptions that were used to generate the effect estimate. Nevertheless, if there is a true causal relation between antibiotics and asthma, the total number of children who develop asthma as a result of this exposure is not insignificant even for small effect estimates, given the large proportion of children who are exposed to antibiotics.
We took 2 approaches in attempting to reduce the effect of protopathic bias: (1) stratifying studies according to their adjustment or lack of adjustment for respiratory infections; and (2) stratifying studies according to their age of first possible asthma diagnosis (age 0 or 1 vs ≥2 years). With both of these stratifications we were able to reduce the heterogeneity among the combined prospective and database studies, and studies that did not adjust for respiratory infections or used earlier ages of asthma incidence revealed a stronger relationship between antibiotic exposure and asthma, which supports the interpretation that protopathic bias was present among such studies. However, protopathic bias likely does not completely explain the observed association between antibiotics and asthma, because pooled estimates of studies with older ages of asthma incidence and those that adjusted for respiratory infections retained significant risk estimates. This insufficiency of protopathic bias to completely account for the association is emphasized by the Marra et al study,39 in which 2 sensitivity analyses were conducted: 1 that examined antibiotics not dispensed for respiratory infections or otitis media in the first year of life, which revealed that the association with asthma at ages 2 through 9 years was retained (HR: 1.13 [95% CI: 1.09–1.16]), and 1 that excluded children diagnosed with respiratory infections or otitis in the first year, which also revealed that the association was retained (HR: 1.17 [95% CI: 1.10–1.25]).
Retrospective studies (with self-reported exposure) revealed the highest pooled risk estimate of any subgroup, which strongly suggests recall bias. Conversely, all nonsignificant estimates of asthma risk were reported by prospective studies, which may be a result of the relatively small number of subjects in the prospective studies compared with the database studies. Moreover, it has been suggested that antibiotic use is not a risk factor in children with a genetic predisposition to asthma,17,36 which could explain the lack of an association in 2 of the prospective studies that exclusively studied children with a family history of allergic disease. To explore the effect of familial history of asthma, we conducted a subgroup analysis that revealed that, although the association was “protective” in the high-risk studies and harmful in the other prospective studies, the low numbers of participants yielded low precision, and we found no statistical difference between the 2 groups. The uncertainty of subgroup effects by parental asthma, therefore, challenges the interpretation of the lack of an association in the prospective study group.
Database studies included in this review mostly showed significant associations, but this study design may be vulnerable to recall biases in their original data collection and susceptible to detection bias, as well as other potential limitations.50 Because evidence of a significant association between antibiotics and asthma in the reviewed literature is entirely supported by retrospective and database studies, the conclusion that there is a causal relationship between antibiotics and asthma should be treated with caution until large, prospective studies can be conducted to examine this question.
For the 3 studies of in utero antibiotic exposure, the pooled estimate revealed a significantly increased risk of asthma, which supports an etiologic relation between antibiotic use and asthma, because reverse causality and protopathic bias are inherently avoided in this exposure. However, these studies may still be susceptible to sources of confounding such as prematurity, chorioamnionitis, and maternal smoking, all of which may be associated with antibiotic use during pregnancy and asthma or wheezing in children.51,52 Although Benn et al48 and Martel et al40 adjusted for low gestational age and small-for-gestational age, respectively, the McKeever et al49 study had no similar adjustment. In addition to chorioamnionitis, other types of maternal infections have also been associated with increased asthma risk in the child,49,53 which makes it difficult to disentangle the independent effects of antibiotics and infections, although the study by McKeever et al49 found that prenatal antibiotics significantly increased the risk for asthma even after adjusting for the number of maternal infections. Other sources of bias, including detection bias, may also remain. Intrapartum antibiotics, such as used for group B streptococcus prophylaxis, have been suggested to be another exposure source that could influence the risk of asthma,54 but investigation of this relationship remains underdeveloped in the literature and was not included in our review.
A primary strength of our systematic review is our focus on minimizing the most likely sources of bias that confound the relation between antibiotics and asthma, including subgrouping analyses according to study design, age of asthma diagnosis, and adjustment for respiratory infections. Limitations include the possibility that analyzing longer “wash-out” periods (ie, by using later ages of asthma diagnosis) between antibiotic exposure and asthma may overcompensate for protopathic bias, because it is possible that a true etiologic relation between antibiotic exposure and asthma may be more likely to occur on a shorter time scale than can be assessed by waiting until the age of 2 years or older to determine outcome. Moreover, fully adjusted effect estimates from most studies may have “overadjusted” the pooled estimate compared with what the true estimate should be.
A previous systematic review on childhood antibiotic exposure and asthma (by Marra et al17) included 8 studies in its meta-analysis. Our review included all of the studies from the previous review, except 1 that did not meet the criteria of exposure,22 together with an additional 13 studies for early-life exposure. The main conclusions of our review are in agreement with those of Marra et al17, who found that retrospective studies revealed a significant relation between antibiotic use and asthma but that prospective studies did not. An important addition to this topic are the large database studies, the results of which were published subsequent to the Marra et al17 review.
To date, evidence suggests that antibiotic exposure in early life or in utero slightly increases the risk of developing childhood asthma. However, this conclusion remains tentative, because evidence for protopathic bias was observed in the methodologically less secure studies. Future studies of large, prospective cohorts that adequately address concerns for protopathic bias are needed to definitively address this question.
Dr Risnes was financially supported by a grant from the Liaison Committee between the Central Norway Regional Health Authority and the Norwegian University of Science and Technology.
- Accepted February 3, 2011.
- Address correspondence to Michael B. Bracken, PhD, Center for Perinatal, Pediatric and Environmental Epidemiology, Yale University School of Public Health, 1 Church St, New Haven, CT 06510-3210. E-mail:
FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.
- CI =
- confidence interval •
- OR =
- odds ratio •
- RR =
- relative risk •
- HR =
- hazard ratio
- Lai CK,
- Beasley R,
- Crane J,
- Foliaki S,
- Shah J,
- Weiland S
- Yazdanbakhsh M,
- Kremsner PG,
- van Ree R
- Bashir ME,
- Louie S,
- Shi HN,
- Nagler-Anderson C
- Sudo N,
- Sawamura S,
- Tanaka K,
- Aiba Y,
- Kubo C,
- Koga Y
- Charlton I,
- Jones K,
- Bain J
- Bizal CL,
- Butler JP,
- Valberg PA
- Sigurs N,
- Bjarnason R,
- Sigurbergsson F,
- Kjellman B,
- Björkstén B
- Altman DG,
- Bland JM
- Egger M,
- Davey Smith G,
- Schneider M,
- Minder C
- Farooqi IS,
- Hopkin JM
- Illi S,
- von Mutius E,
- Lau S,
- et al
- von Mutius E,
- Illi S,
- Hirsch T,
- Leupold W,
- Keil U,
- Weiland SK
- Cohet C,
- Cheng S,
- MacDonald C,
- et al
- Foliaki S,
- Nielsen SK,
- Bjorksten B,
- Von Mutius E,
- Cheng S,
- Pearce N
- Marra F,
- Marra CA,
- Richardson K,
- et al
- Martel MJ,
- Rey E,
- Malo JL,
- et al
- Ponsonby AL,
- Couper D,
- Dwyer T,
- Carmichael A,
- Kemp A
- Murray CS,
- Pipis SD,
- McArdle EC,
- Lowe LA,
- Custovic A,
- Woodcock A
- Xu B,
- Pekkanen J,
- Jarvelin MR,
- Olsen P,
- Hartikainen AL
- Copyright © 2011 by the American Academy of Pediatrics