Ethical aspects

Since this is a systematic review based on published literature, the ethical requirements have been met previously for each individual study. Accordingly, the relevant approvals are stated in each original publication (article) included in our review. Written informed consent was obtained from the patient's guardian/parent/next of kin for the publication of each original article included in this report and any accompanying images.

Inclusion criteria for studies on PA and asthma diagnoses

We included longitudinal and cross-sectional studies that investigated asthma and PA in children and adolescents aged 0–18 years. PA was documented by either interviews or self-administered questionnaires. Childhood asthma was defined using parental reports of either physician diagnosis of asthma, “current” (within last 12 months) asthma, “ever” (lifetime) asthma, wheezing, exercise-induced asthma (EIA), or medical treatment of asthma symptoms. We defined new-onset asthma (incident asthma) as a physician diagnosis of asthma/or wheezing. Hence, for incident asthma, there was no sampling based on disease status [33]. We used asthma/or wheezing (a representative asthma symptom) [34] to capture the heterogeneous symptomatology of asthma in children [35].

We defined PA as a behavioral concept that varied according to “leisure time” or “sports and exercise” [36]. We recognized that PA can be further characterized by its dimensions as follows: (1) frequency, (2) intensity, (3) duration, and (4) type [37]. Intensity has been identified as the key dimension for possible dose-response relationships with either reduced or increased health risks for exercise-induced medical conditions [38]. This review did not distinguish between PA and exercise. The concept “PA” referred to general leisure-time PA, exercise, or sports during or outside of school hours [39]. High amounts of TV viewing (duration in hours) represented sedentary behavior [40, 41] and were used as a proxy for low PA. This approach was based on the previous use of TV viewing [24, 42] which validated that TV viewing could be used to represent PA in population surveys. It was beyond the scope of this review to discuss the scientific distinctions between sedentary activity and physical inactivity in children and adolescents.

Inclusion criteria for the meta-analysis

We adhered to appropriate standards [29] in defining our criteria for the meta-analysis, which were as follows: (1) broadly similar research questions, (2) comparable participant populations (children and adolescents), and (3) broadly similar research mechanisms.

Exclusion criteria

We excluded studies involving adults >18 years of age and non-English-language studies [43]. We also excluded single outcomes of intermediate phenotypes for childhood asthma (i.e., bronchial hyperresponsiveness [BHR], allergic rhinoconjunctivitis, atopic dermatitis, airway inflammation, eczema) and cumulative incidence along with studies that had fitness or body composition as their only outcomes. Studies that reported on only PA or asthma were excluded, as were clinical investigations (e.g., randomized controlled trial [RCT] designs) of training and/or medical treatment in children with asthma. If pediatric asthma or PA was explored using noncomparable (rare) methodologies or the studies excluded relevant participants, the studies were excluded. We excluded other reviews, methodology reports, validation studies, and studies that collected data for other purposes or had other non-applicable outcomes. The two stages of exclusion are illustrated in Fig. 1, and the articles excluded at each stage are grouped by exclusion rationale in Additional file 3A and B.

Open in a separate window

Fig. 1

Inclusion and Exclusion Criteria for Systematic Reviews. Numbers of search results from Medline, the Cochrane Library, and Embase

Search strategy

Data extraction and study quality

LL extracted information from the included studies. Table 2 shows the information points that were extracted from each study for the descriptive data synthesis. The extracted items represented adopted standards for methods, participants, outcomes, and results as defined in the checklist of The Cochrane Handbook for Systematic Review [44]. For the quantitative data synthesis (the meta-analysis), we extracted individual summary data [29] from each study that met the criteria for meta-analysis. We excluded BHR as an asthma phenotype and consequently were only able to obtain asthma severity data from a few of the reviewed studies [22, 24, 45].

Using GATE [28] entailed documenting the study population, representativeness, measurement(s), and timing. All data that were extracted to electronic GATE forms [46] are illustrated in Additional file 4.

Cohort studies

Measurements of new-onset (incident) asthma/or wheezing

Two studies [45, 51] described cases of new-onset asthma using a physician’s diagnosis of asthma (Table 3), and one study described new-onset wheezing [52]. We synthesized three cohort studies [45, 51, 52] that met the criteria for inclusion in our meta-analysis. The follow-up times (in years) were 6–7 [52], 10 [51], and 11.5 [45] (Table 3). In these studies [45, 51, 52], a total of 549 children had new-onset asthma/or wheezing, and the total number of cohort children studied was 6,037 (Table 3). The reported asthma prevalence was 6.0 % [45], the new-onset wheezing prevalence was 11.3 % [52], and the asthma incidence rate was 16.6 % per 1,000 person-years [51]. Overall, 57.7 % (317) of the cohort children with new-onset asthma/or wheezing had low PA [45, 51, 52] (Table 5).

Table 5

Distribution (N, %) of Children with New-Onset Asthma/or Wheezing and All Children According to PA

First author, publication year	PA-exposure levels	New-onset asthma outcome	New-onset asthma, N (%)	All children N (%)
Vogelberg et al., 2007 [52]	Low PA if sport freq≤once/wk High PA (rfgr) if sport freq≥two times/wk	Wheezing	Low PA: 199 (60.5) High PA: 130 (39.5)	Low PA: 1,470 (51.4) High PA: 1,388 (48.6)
Sherriff et al., 2009 [45]	Low PA if TV viewing ≥1 hr/day High PA (rfgr) if TV viewing=none or <1 hr/day	Asthma	Low PA: 61 (78.2) High PA: 17 (21.8)	Low PA: 1,100 (68.8) High PA: 499 (31.2)
Islam et al., 2009 [51]	Low PA if number of team sports=none High PA (rfgr) if number of team sports ≥1	Asthma	Low PA: 57 (40.1) High PA: 85 (59.9)	Low PA: 648 (41.0) High PA: 932 (59.0)

Open in a separate window

Details regarding data from the meta-analyzed studies

Freq Frequency, Hr/hrs Hour/hours, N Number, PA Physical activity, Rfgr Reference group, Wk Week

Results from meta-analysis

We conducted a meta-analysis using data on asthma and PA provided by three articles [45, 51, 52]. To combine the study results, we reclassified the exposure variables. The original PA variables were number of team sports played (none, 1–2, >2) [51], sports participation frequency (≤once per month, ≤once per week, 2–3 times per week, >3 times per week) [52], and duration of TV viewing (not at all, <1 hour per day, 1–2 hours per day, >2 hours per day) [45]; for the meta-analysis, we dichotomized the results into no team sports played (low PA) and ≥1 team sport played (high PA) [51], sports participation≤once per week (low PA) and ≥2 times per week (high PA) [52], and TV viewing ≥1 hour per day (low PA) and <1 hour per day (high PA) [45]. The reference category was high PA in both the random- and fixed-effects models. The overall meta-analysis results showed positive risks for new-onset asthma (OR [95 % CI] 1.32 [0.95; 1.84] [random effects] and 1.35 [1.13; 1.62] [fixed effects]) in children with low PA compared with high PA (reference). These results are illustrated in Fig. 2 (random effects) and Fig. 3 (fixed effects). I² was 60.6 % (χ²=5.08, P=0.079) for both random and fixed effects.

Open in a separate window

Fig. 2

Random-Effects Model: Study-Specific and Overall Odds Ratios (ORs) with 95 % Confidence Intervals (CIs). Data are derived from the meta-analysis of low physical activity (PA) and new-onset asthma during childhood. High PA: Reference category

Open in a separate window

Fig. 3

Fixed-Effects Model: Study-Specific and Overall Odds Ratios (ORs) with 95 % Confidence Intervals (CIs). Data are from the meta-analysis of low physical activity (PA) and new-onset asthma during childhood. High PA: Reference category

Consistency of meta-analysis results: risk of bias across studies

In Fig. 4, the studies that included larger numbers of asthmatic participants [51, 52] were positioned toward the top, i.e., the upper two-thirds of the funnel, representing large sample sizes and small standard errors. Figure 4 also shows that the studies in the meta-analysis [45, 51, 52] were within the 95 % confidence limits (diagonal, dashed lines) around the summary estimate.

Open in a separate window

Fig. 4

Funnel Plot with 95 % Pseudo Confidence Intervals (CIs). Data are from the meta-analysis depicting the log-scale odds ratios (logORs) (horizontal axis) for new-onset childhood asthma by low physical activity (PA) using individual study effect size data plotted against the standard errors (SEs) (vertical axis) of the logORs

Validity and quality: risk of bias within studies

Our review showed that these three studies [45, 51, 52] explored the role of the temporal sequence following quantified PA exposure and its effect on new-onset asthma/or wheezing in children and adolescents.

Cross-sectional studies

Validity and quality: risk of bias within studies

We applied the GATE approach developed for the critical appraisal of quantitative studies (electronic forms) [28, 46]. When data were available, we first extracted study numbers regarding exposure, comparison, and outcomes for the association between PA and childhood asthma (Additional file 4). We first used the GATE calculator (one-page Microsoft Excel format) and then transferred the calculated results to the GATE-lite form (one-page Microsoft Word format) [46]. We used GATE to illustrate individual study designs and study details as recommended for gauging bias risks [56].

We illustrated the study design using the acronym PECOT, i.e., extracted data on participants, exposure, comparisons, outcomes, and time. To assess study validity, we used the acronym RAMBOMAN, i.e., extracted data on recruitment, allocation, maintenance, blind or objective measurements, and analyses.

We applied the GATE approach to a total of eight non-meta-analyzed cross-sectional studies [9, 22–25, 53–55] that investigated asthma prevalence or asthma symptoms. The studies included a total of 4,155 children with current asthma/or wheezing, and the total number of participants was 41,770 children (Table ​(Table44 and Additional file 4). Unfortunately, in one study [23], the absolute number of participants was not given. The prevalence of asthma/or wheezing in six of these studies [9, 24, 25, 53–55] ranged from 3.8 % [55] to 23.7 % [24] (Table 4). Five cross-sectional studies originated from Europe [9, 24, 25, 54, 55], and two were from North America [22, 53]. One study was cross-national [23] and included data for 6–7-year-olds from 17 countries and data for 13–14-year-olds from 35 countries (Additional file 4). A majority of the eligible populations were derived from respective national surveys [9, 24, 25, 53–55] of children and adolescents (Additional file 4).

The GATE assessment showed that all eight studies [9, 22–25, 53–55] included measures of exposure and outcome and included a comparison group, and all authors reported the results of adjusted analyses; however, for six studies [22–25, 53, 54], we were unable to obtain data on either the exposure or the comparison groups (Additional file 4). The response rates were >50 % in seven [9, 23–25, 53–55] of the eight studies, although Mitchell et al. [23] observed a response rate <50 % for younger children (6–7 years of age) (Additional file 4). Two cross-sectional studies [22, 53] analyzed PA as an outcome.

The definitions of PA varied. Nystad [55] and Nystad et al. [9] measured PA outside of school hours (sports or exercise) that caused a child to become sweaty or out of breath. Lang et al. [22] registered the total minutes spent engaging in PA in one day, Jones at al. [53] assessed sufficient moderate PA (e.g., fast walking, slow bicycling), and Priftis et al. [24] examined sports-related PA (e.g., brisk walking, running, swimming). Corbo et al. [25] registered PA as regular sports, i.e., formal games or forms of aerobic exercise. Kosti et al. [54] observed leisure-time PA, i.e., unstructured outdoor PA involving play, walking, or cycling. Mitchell et al. [23] described PA as weekly vigorous activity that was sufficient to cause heavy breathing in the child.

Additionally, the definition of low PA varied, with some studies defining low PA as ≤1 hour per week [9], <30 min per day [22], once or twice per week [23], no participation in any PA [24], sports 1–2 times per week [25], sufficient moderate PA [53], leisure-time PA [54], and 1–3 h per week [55]. In four [9, 24, 25, 55], one [23], and one [54] of the eight [9, 22–25, 53–55] cross-sectional studies, the reference groups in the adjusted analyses were “low to no PA”, “no vigorous PA”, or “no leisure time”, respectively (Additional file 4).

In four of the eight studies [9, 22, 53, 55], we were able to extract data for the GATE calculator to estimate occurrences in exposure groups and/or exposure effects. In the studies that investigated distinct low PA (≤1 h per week [9], 1–3 h per week [55], <30 min per day [22]), the occurrences per 100 persons in the exposure groups (EGO) were 6.2 [9], 4.1 [55], and 14.6 [22]. Two [9, 55] of these studies provided sufficient data to estimate exposure effects in terms of relative risk (RR) (Additional file 4). In the remaining four studies [23–25, 54], we could not derive appropriate data for the calculations.

Although we found each of the eight cross-sectional studies [9, 22–25, 53–55] applicable to practice (Additional file 4), the GATE analysis illustrated variations across the studies. We concluded that the quality of these studies was high for cross-sectional designs, but the variation among the studies confirmed that individual study analysis (e.g., GATE assessment) as opposed to common estimation across studies (e.g., meta-analysis) was a sound approach that agreed with recommendations [29, 49, 57].

All studies - measurements PA

In all but one study, PA was assessed using a questionnaire that asked about sports participation (Tables 3 and ​and4).4). One cohort study [45] reported TV viewing (Table 3).

Main effect size and adjustment covariates

Tables 3 and ​and44 show that six studies reported positive associations between asthma/or wheezing and low PA [9, 22, 24, 25, 45, 53], and one study showed that asthma/or wheezing was positively associated with high PA [51]. Of the eight cross-sectional studies [9, 22–25, 53–55], three [22–24] indicated significant positive associations between childhood asthma or asthma symptoms and low PA (of which one [23] reported this association for 13- to 14-year-olds only and one [24] reported this association for boys only).

The adjustment covariates applied in the multivariate analyses varied across the reviewed studies. All authors adjusted for age, gender, weight, and/or smoking measures (Tables 3 and ​and4), three4), three studies included adjustments for asthma history [9, 25, 45], and eight studies adjusted for socioeconomic measures [23, 25, 45, 51–55].

The review commenced during LL’s initial research program at the University of Bristol (UoB), Bristol, United Kingdom. Therefore, the authors wish to express thanks to Medical Subject Librarian CF Borwick (CFB), UoB, for assistance in refining the database search strategies. We also thank Prof RT Jackson (RTJ) of the University of Auckland (Auckland, New Zealand) for reviewing our data, providing helpful guidance (which included providing savable files for GATE), and offering suggestions regarding our application of the GATE tool.