Physical Activity and Asthma: A Systematic Review and Meta-Analysis

Introduction This review aims to give an overview of available published evidence concerning the association between physical activity and asthma in children, adolescents and adults. Methods We included all original articles in which both physical activity and asthma were assessed in case-control, cross-sectional or longitudinal (cohort) studies. Excluded were studies concerning physical fitness, studies in athletes, therapeutic or rehabilitation intervention studies such as physical training or exercise in asthma patients. Methodological quality of the included articles was assessed according to the Newcastle-Ottawa Scale (NOS). Results A literature search was performed until June 2011 and resulted in 6,951 publications derived from PubMed and 1,978 publications from EMBASE. In total, 39 studies met the inclusion criteria: 5 longitudinal studies (total number of subjects n = 85,117) with physical activity at baseline as exposure, and asthma incidence as outcome. Thirty-four cross-sectional studies (n = 661,222) were included. Pooling of the longitudinal studies showed that subjects with higher physical activity levels had lower incidence of asthma (odds ratio 0.88 (95% CI: 0.77–1.01)). When restricting pooling to the 4 prospective studies with moderate to good study quality (defined as NOS≥5) the pooled odds ratio only changed slightly (0.87 (95% CI: 0.77–0.99)). In the cross-sectional studies, due to large clinical variability and heterogeneity, further statistical analysis was not possible. Conclusions The available evidence indicates that physical activity is a possible protective factor against asthma development. The heterogeneity suggests that possible relevant effects remain hidden in critical age periods, sex differences, or extremes of levels of physical activity (e.g. sedentary). Future longitudinal studies should address these issues.


Introduction
The prevalence of asthma has increased significantly during the past decades [1]. Concurrently, the prevalence of overweight has increased, while physical activity levels have decreased substantially [2,3]. In 2005, less than half (49.1%) of US adults met the CDC/ACSM (Centers for Disease Control and Prevention/ American College of Sports Medicine) physical activity recommendation (at least 30 minutes of moderately intense activity on five days per week or vigorously intense activity for a minimum of 20 minutes on three days each week) [4]. Physical inactivity is an important risk factor, because it is potentially modifiable and therefore an opportunity for prevention. Several studies have shown that training improves cardiopulmonary fitness, asthma symptoms and quality of life in asthmatic subjects [5]. This evidence suggests that training and high levels of physical activity play a role in the course and severity of asthma. Besides this, an etiological relation between physical activity levels and development of incident asthma might also be possible. Different hypotheses have been suggested to explain the possible protective character of physical activity against asthma development such as reducing airway inflammation, a central feature of asthma [6]. Another explanation is that physical activity could positively influence the patency of bronchioles: poor mucociliary clearance from decreased epithelial stimulation secondary to decreased activity can cause excess mucus and airway edema. Decreased deep inspiration and sigh rate during physical inactivity could lead to smooth muscle latching and subsequent increased risk of asthmatic symptoms [7].
We performed a systematic literature review to evaluate the potential causal relation between physical (in)activity and asthma development, and a pooled analysis to estimate the effect size.

Search strategy
We conducted an electronic search in PubMed (US National Library of Medicine) and EMBASE to obtain all publications on studies that reported on physical activity and asthma published until June 2011. The PubMed search used the Medical Subject Headings (MeSH) terms ''motor activity'' or text word terms ''activity'', ''physical activity'', ''physical exercise'' or ''sedentary'', as well as the MeSH term ''asthma'' or text word terms ''asthma'', ''asthmatic'', ''wheeze'' or ''wheezing''. The EMBASE search used the MeSH terms ''motor activity'' or ''physical activity'' or text word terms ''physical activity'', ''physical exercise'', ''sedentary'', as well as the MeSH terms ''asthma'' or ''wheezing'' or the text word terms ''asthma'', ''asthmatic'', ''wheeze'' or ''wheezing''. These terms were searched using limits that included all articles published in the English language. There were no age restrictions.
We conformed to the MOOSE (Meta-analysis Of Observational Studies in Epidemiology) guidelines for reporting [8] and PRISMA (Preferred Reporting Items for Systematic reviews and Meta-Analyses) statement [9].

Inclusion
Our primary research question concerned the role of habitual physical activity in the development of incident asthma. Therefore, we searched for longitudinal studies in which the exposure (physical activity) precedes the outcome (onset of asthma). In addition, we included studies that looked into asthma prevalence in different physical activity levels. For this goal, we searched crosssectional studies that investigated physical activity levels in subjects with asthma compared to controls. For maximal sensitivity, a broad inclusion strategy was used. Inclusion criteria were: original articles in which physical activity as well as asthma was studied, and a control group consisting of healthy subjects or general population. Excluded were studies that did not concern habitual physical activity such as studies in athletes, physical fitness, therapeutic or rehabilitation intervention studies such as physical training in asthma patients. Two investigators (ME, MM) independently assessed whether articles met the inclusion criteria. In case of disagreement, consensus was reached through discussion.

Quality assessment and data extraction
Methodological quality of included articles was assessed according to the Newcastle-Ottawa Scale (NOS). This instrument was developed to assess the quality of nonrandomized studies. Its content validity and inter-rater reliability has been established [10]. The NOS gives predefined criteria, some of which have to be further specified for the specific topic. We specified these criteria in a consensus meeting with all authors (criteria are presented in figure S1 and S2) before assessing the studies. In short, longitudinal studies were assessed for quality of selection (representativeness, selection of controls, ascertainment of exposure, no asthma at start of study); comparability (confounding); and outcome (assessment of outcome, length and adequacy of followup). Gender, weight, and smoking were identified as important confounders. Studies could be awarded a maximum score of 9 points. Studies with scores of 5 points or more were considered to be of moderate to good study quality. However, all studies were used for analysis, irrespective of NOS score. Quality assessment was done by all five authors using the NOS. Each single article was assessed by at least three authors independently. In case of disagreement the other two authors were consulted. Quality assessment was completed before data extraction was started. Data were extracted from the full text article. Quantitative results were extracted from text and tables, choosing preferably those adjusted for important confounders (gender, weight, and smoking). Dataextraction in the longitudinal studies was performed independently by two authors (ME, CT). If essential data were lacking in the original studies, their authors were contacted.

Statistical analysis
Analyses were performed using the statistical software Review Manager version 5 [11]. Heterogeneity among studies was assessed using the chi-square test (significant at p,0.05) and the Higgins I 2 test [12]. A random effects model with the Mantel-Haenszel method was used for pooling the results of different studies. Pooled odds ratios (OR) with 95% confidence intervals (CI) were calculated for the longitudinal studies and a subgroup of the cross-sectional studies, namely those studies that used a motion sensor for measuring physical activity levels. We decided to refrain from statistical pooling of the other cross-sectional studies because of substantial clinical and methodological heterogeneity.

Literature search
The search resulted in 6,951 publications derived from PubMed, and 1,978 studies from EMBASE. Based on titles and abstracts, 8,790 articles were excluded at first screening because they did not meet the eligibility criteria, such as experimental studies with intermediate outcomes (such as inflammatory markers) but no asthma as clinical outcome, case reports and studies with case series without control group, and studies of exercise induced asthma in athletes. Full-text copies of the remaining 139 potentially relevant studies were obtained. Ninety-five studies were excluded because they did not meet the inclusion criteria. Four were excluded because they were duplicate publications of the same studies. One study was not available in full-text. The remaining 39 studies were included for this systematic review. Five studies were longitudinal studies and 34 were cross-sectional in design (figure 1).

Study quality
All 39 studies were assessed using the adjusted NOS scale (see figure S1 and S2 for the adjusted NOS scales), of which the majority (79%) scored 5 points or more, indicating a moderate to good study quality (table S1 and S2). When grouped by study quality, we could not detect a clear pattern in study results or study characteristics (tables 1 and 2). The authors of 4 articles [13,14,15,16] were contacted to obtain essential data that were lacking in the original studies, of which 3 replied [14,15,16].

Longitudinal studies
Study characteristics and odds ratios of the longitudinal studies [14,17,18,19,20] are summarized in table 1. All 5 studies looked into physical activity levels of subjects at baseline and incident asthma during follow up. Follow up duration ranged between 5 and 10 years. Physical activity was assessed by questionnaires. Different reference categories, subgroups and confounders were used (table 1). Asthma diagnosis was defined as doctor's diagnosis, either through self report or linkage to an insurance registry.

Statistical analysis and pooling
Data of the 5 longitudinal studies were pooled using a random effects model ( figure 2). Data on studies with more than two groups of different physical activity levels were converted into two groups, namely low physical activity and high physical activity, of which low physical activity was used as reference category. In case of an uneven number of groups, the reference category consisted of the lowest physical activity levels including the middle group.
Pooled odds ratio was 0.88 (95% CI: 0.77-1.01). Chi-square test for heterogeneity was borderline significant (p = 0.07). Higgins I 2 index was 45%, indicating moderate inconsistency. These results are not adjusted for potential confounders as the majority of studies did not provide adjusted results. When we restricted analysis to the studies with moderate to good study quality, identified by NOS scores of 5 or higher, 4 studies remained [14,17,18,19]. Sensitivity analysis showed a consistent result: the pooled odds ratio did not change much (0.87 (95% CI: 0.77-0.99)) but did reach statistical significance.
Physical activity was assessed by motion sensors in 7 studies [13,16,22,28,42,48,49], all examining children (n = 2916). In these studies, asthma was defined as self reported doctor's diagnosis or asthma symptoms. Two studies [48,49] combined self reported asthma diagnosis with spirometry. Not enough data were available for pooling in one study (i.e. standard deviations were missing despite contacting the authors). [13] Data of the other 6 studies using motion sensors were pooled: standard mean difference 20.19 (95% CI 20.69; +0.32) (figure 3). Testing for heterogeneity showed a highly significant result (p,0.00001) and high inconsistency (I 2 = 94%).

Walders-Abramson 2009
U.S. We found similar rates of objectively measured physical activity among youth with well controlled asthma and controls.
Overview of study characteristics, study quality based on the Newcastle-Ottawa Scale (NOS), odds ratios and author's conclusions of cross-sectional studies on physical activity and asthma prevalence. Odds ratios are noted here only if odds ratios or equivalents with 95% confidence intervals are specified in the article. Author's conclusions are noted only if the author mentions a conclusion on the relation between physical activity and asthma prevalence. If not, a conclusion was drawn based on the data in the article.

Discussion
This systematic review gives an overview of the published evidence concerning the association between physical activity and asthma. Our primary research question was aimed at the etiological association between different physical activity levels and subsequent asthma incidence. In an extensive search, we only found 5 longitudinal studies that met the inclusion criteria and could be of use in answering this question. Although the number of longitudinal studies was small, the total accrued number of subjects was considerable (n = 85,117). Pooling showed that subjects with higher physical activity levels might have lower risk of developing asthma.
Thirty-four studies were cross-sectional in design. Due to large clinical variability and heterogeneity we had to refrain from further statistical analysis, except for a small group of studies using a motion sensor to measure physical activity. Despite this limitation, however, we can draw some conclusions: a substantial number of included cross-sectional studies, with the largest total study population, did find an association between high physical activity levels and low asthma prevalence. This seems consistent with physical activity being protective against asthma. However, we can not rule out publication bias. Moreover, cross-sectional studies are not suited to give insight into the causal relation between physical activity and subsequent asthma incidence. Besides the hypothesis that subjects with higher physical activity levels have a lower risk of developing asthma (protective), reverse causality is also possible. There are several hypotheses why asthma patients (with asthma as exposure) could have lower physical activity levels (outcome), such as fear for symptoms of shortness of breath, wrongful education, or by asthma that is not well regulated.
In contrast to the studies that were cross-sectional in design, this reverse causality does not play a role in interpreting the results of the 5 longitudinal studies. In all 5 studies physical activity levels were measured before asthma was diagnosed. However, the results could be influenced by protopathic bias (e.g. physical activity restricted by respiratory complaints that precede an asthma diagnosis) or earlier diagnosis of asthma through exercise-induced symptoms. The first would lead to overestimation of the true association between low physical activity levels and subsequent asthma development; whereas the second would lead to an underestimation. Unfortunately, none of the longitudinal studies addressed these biases.

Limitations
It is important to realize that there are several limitations to this review. First of all, due to the fact that this research is based on published material, publication bias is an important factor. Furthermore, studies showed substantial heterogeneity in different areas such as population (number, age, gender, race, duration of follow-up), exposure variables (physical activity measured by questionnaires, whether or not validated, or measured by motion sensors) and outcome variables (asthma diagnosis as self reported doctor's diagnosis, asthma symptoms or spirometry). Analysis showed borderline statistical heterogeneity. The small number of longitudinal studies prevented us from performing meta-regression or subgroup analysis. Confounding is an important issue, because other risk factors (such as smoking and obesity) could be associated with both low habitual physical activity as well as asthma development. First and second-hand cigarette smoke exposure is already established as an independent risk factor for developing asthma. [53] It is suggested that obesity is a risk factor for asthma development. [54] In our meta-analysis of longitudinal studies, pooling of results adjusted for confounders was not sensible because only three studies presented such results. However, adjusted odds ratios were never lower than unadjusted odds ratios (see table S3), so that a pooled effect for the adjusted results would be higher than the odds ratio of 0.88 (95% CI: 0.77-1.01) found for the unadjusted result.
Another limitation might be the fact that the validity of the NOS score recently has been questioned by Stang who believes that the NOS provides a quality score that has unknown validity at best. [55] We noted that some methodological pitfalls were not well represented in the NOS scale: reverse causation and protopathic bias or confounding by indication (e.g. advice to remain physically active for children with respiratory complaints).

Conclusion
In conclusion, the results of available published evidence indicate that high physical activity levels are a possible protective factor against asthma development. The heterogeneity suggests that possible relevant effects remain hidden in critical age periods, sex differences, or extremes of levels of physical activity (e.g. sedentary). Future longitudinal studies should address these issues. Table S1 NOS scores of longitudinal studies. NOS: Newcastle-Ottawa Scale. Result of quality assessment of longitudinal studies on physical activity and asthma using NOS scores. We refer to figure S1 for the adjusted NOS for longitudinal studies, which was used as a scoring list. (DOC)

Supporting Information
Table S2 NOS scores of cross-sectional studies. NOS: Newcastle-Ottawa Scale. Result of quality assessment of crosssectional studies on physical activity and asthma using NOS scores. We refer to figure S2 for the adjusted NOS for crosssectional studies, which was used as a scoring list.

(DOC)
Table S3 Data extraction of longitudinal studies. CI; confidence interval, PA; physical activity, aHR; adjusted hazard ratio, OR; odds ratio; aOR; adjusted odds ratio, BMI; body mass index. Data extraction of longitudinal studies concerning baseline physical activity and asthma incidence. * p,0.05, #a used as reference category for pooling in this review, #1 adjusted for age, race, sex, center, and maximal education, #2 adjusted for BMI, smoking status, menopausal status, education level, working status, co-morbidities, #3 adjusted for age, atopy, and respiratory symptoms. (DOC)