Article Text
PDF
Abstract
Background There is increasing interest in the role physical activity (PA) can play in the development and management of asthma. Understanding whether PA can have a positive effect is hindered by the potential influence of asthma on PA and a lack of relevant longitudinal data, leading to a debate on the existence and direction of these links. The aim of this study was to explore whether having asthma results in lower PA levels, and/or whether lower PA levels lead to more asthma in children and adolescents.
Methods In a population-based study of 4983 children, data on asthma and PA were collected via questionnaires and time use diaries biennially, between the ages of 6 and 14. Current asthma was defined as use of asthma medications or wheeze in the past year, and incident asthma was defined as doctor’s diagnosis since the previous wave. PA was time spent doing moderate-to-vigorous physical activities in a day. Bidirectionality of this relationship was investigated using cross-lagged structural equational models.
Results PA was not longitudinally associated with incident or current asthma. Similarly, there was no evidence that incident or current asthma predicted PA at any of the ages.
Conclusions Using a novel strategy to investigate bidirectionality between PA and asthma, our results suggest that asthma and PA participation are not longitudinally associated in either direction. Our findings suggest that PA does not play an important role in the development or persistence of asthma.
- adolescents Cg
- asthma
- child health
- physical activity
Statistics from Altmetric.com
Introduction
While the increase in the prevalence of asthma appears to have plateaued in recent years, it remains high globally,1 especially in children and adolescents.2 The reason for this high prevalence is multifactorial; however, several lifestyle factors, including increased screen time and decreased physical activity (PA), have been identified as possible contributors to the perpetuation of asthma as an important public health problem.3 It is estimated that globally, a large proportion of children and adolescents do not attain the recommended 60 min of moderate-to-vigorous PA (MVPA) per day,4 and there are concerns that an even smaller proportion of children and adolescents with asthma meet these recommendations.2
The relationship between PA and asthma is complex and may in fact be bidirectional. Some previous research has investigated the effect of PA on asthma in children, and this research was collated in a systematic review published in 2016 by Lochte et al.5 Their meta-analysis of three longitudinal studies concluded that lower levels of PA were associated with a 35% increased risk of new-onset asthma and wheeze in children (fixed-effect pooled OR 1.35; 95% CI 1.13 to 1.62; I2=60.6%).5 However, these studies varied in their assessment and definition of the exposure; while one looked at sedentary time,6 the second and third questioned the amount of time spent in sports.7 8 In addition to this, the scarcity of longitudinal studies in child and adolescent populations means that further studies are warranted.
On the other hand, an asthmatic state has been found to reduce levels of PA in affected individuals,9–11 probably through self-imposed or parentally imposed behavioural changes.2 12 Since strenuous exercise is a known trigger for asthma, studies have shown that individuals with the condition tend to refrain from participation in strenuous activity to avoid triggering an asthma attack.10 Despite this, several large international studies found no evidence that children with asthma participate in less PA than children without asthma.13–16 Similarly, our recent systematic review and meta-analysis concluded that children with and without asthma engaged in equivalent amounts of objectively measured PA.17 However, the majority of the studies included in the review were cross-sectional, with only one longitudinal study. Further longitudinal research would clarify the extent of any effect of childhood asthma on PA levels over time.
In summary, although the relationship between PA and asthma has been reviewed and documented in the literature,5 17 the temporality of the association has not yet been elucidated. There is both evidence for and against the effect of PA on asthma and the effect of asthma on PA. Inference on causality and direction of the relationships between asthma and PA are inhibited primarily by limited numbers of longitudinal studies in children and the potential for reverse causality. Importantly, to date, there have been no studies that have investigated these links between PA and asthma while taking into account the possibility of the bidirectionality.
Therefore, this study aimed to investigate the possibility of a reciprocal relationship between asthma and PA in children and adolescents in a longitudinal manner using a novel analytical approach. That is, this analysis aimed to investigate (i) the effect of asthma status at an earlier age on PA level at a later age and (ii) the effect of PA level at an earlier age on asthma status at a later age. We hypothesised that there would be evidence of associations in both directions.
Methodology
Longitudinal Study of Australian Children (LSAC)
This study used longitudinal data from the Growing Up in Australia: Longitudinal Study of Australian Children (LSAC) whose sampling and recruitment methodology has previously been described18 19 and is accessible on the LSAC website (www.growingupinaustralia.gov.au). The present analysis was conducted in one of the two nationally representative LSAC cohorts, namely the ‘K’ cohort which consisted of 5000 children aged 4–5 years at recruitment (wave 1). Participants were followed up biennially, and data have been collected via interviews, questionnaires, time use diaries (TUDs) and direct anthropometric measurements. This study used the data collected at waves 2 to 6 (ages 6 to 14) for the exposure and outcome variables; since asthma diagnosis in children is contentious before the age of 5 years,20 wave 1 (age 4) data were included for analytic adjustment purposes only.
Physical activity
PA data were extrapolated from the completion of TUDs. The use of TUDs in LSAC has previously been published in detail,21 but their use for this analysis is described below. The TUDs used at waves 1–3 differ from those used at waves 4–6. In earlier waves,22 at the face-to-face interview, parents were shown how to complete the TUD which divides the 24-hour day into 15 min intervals.22 Parents were asked to complete the child’s activity diary (by selecting from 26 precoded activities, presented in online appendix 1), indicating location and with whom the activity was performed for each 15 min interval of the day, for 1 weekday and 1 weekend day.22 23 From wave 4 onwards, children were asked to complete a single TUD themselves by recording the activities they performed throughout the day on any 1 day.22
Supplemental material
For waves 2 and 3 where two TUDs were completed, we randomly selected one TUD per participant for analysis to ensure consistency across all waves. Since the TUDs did not capture the intensity of PA, the following coded activities within the TUDs were considered MVPA (with a metabolic equivalent score of 3.0 or greater) based on the Compendium of Energy Expenditure for Youth categorisations24: structured or unstructured active play, organised sports/lessons/activities, riding bike/scooter/skateboard and walking/running skipping. For each wave, a continuous variable for total time spent in MVPA for each child was created by summing the time in minutes spent in all coded physical activities for the day the data were collected.
Asthma measurement
The present analysis investigated two asthma outcomes:
incident asthma since the last wave was defined as present if the response was ‘yes’ to the question ‘Has a doctor ever told you that child has asthma?’ at waves 1 to 4 (when the answer had been ‘no’ in previous waves) or ‘Since the last interview, has a doctor told you that study child has asthma?’ at waves 5 and 6;
current asthma was considered to be present if there was an affirmative response to either ‘In the last 12 months, has child taken any medication for asthma?’ or “In the last 12 months, has the child had an illness with wheezing in the chest which lasted for a week or more?’.25
Covariates
Three time-varying covariates, resident older siblings,26 27 socioeconomic position (SEP)28 29 and body mass index (BMI),30 and one non-time varying covariate, gender,30 were selected based on the literature and availability of data. Gender and the presence of one or more older siblings residing in the home with the study child were collected in the questionnaires and analysed as binary variables. SEP ranked each LSAC family based on parental income, education and occupation and was included in the models as a z-score. The development of this variable has been described extensively in a paper published by LSAC.31. Height was measured using an Invicta stadiometer in waves 2 and 3, and in waves 4, 5 and 6, a laser stadiometer was used.32 Weight was measured using HoMedics digital BMI bathroom scales in waves 2 and 3, and Tanita body fat scales were used in waves 4 to 6.32 BMI at each wave was categorised as ‘underweight’, ‘normal weight’ or ‘overweight’ based on the age appropriate cut-offs proposed by Cole et al.33
Statistical methodology
Stata/MP (StataCorp LLC) statistical software was used for all analyses. Initially, descriptive analyses of participant characteristics were undertaken by calculating frequency and percentages for categorical variables and means and SD for continuous variables. Time spent in MVPA was converted to hours for reporting purposes.
To examine the bidirectionality of the association between asthma and PA, we looked at three models using generalised structural equational modelling (GSEM) for both incident asthma and current asthma. GSEM is a statistical modelling technique used to analyse structural relationships between multiple variables. It works by first postulating what we believe to be the relationship between several variables, this can be based on expert knowledge input or literature review. The data are then used to confirm the postulated relationship. We used both linear and logistic regressions as appropriate. GSEM intrinsically accounts for missing and incomplete data using maximum likelihood estimation which assumes missingness at random. More details can be found in the reference.34
Our first model estimated the lagged effects of PA at each wave on asthma at the next wave using logistic regression and will be referred to as Model 1x (‘a’ for incident asthma, ‘b’ for current asthma). Through this model, we postulated that PA at a previous wave affects asthma at the current wave. For the second model, the lagged effects of asthma at each wave on PA at the next wave were investigated with linear regression in GSEM (Model 2x). In this model, we implied that incident asthma status at a previous wave influences the amount of time spent doing physical activities. The final model estimated the cross-lagged effects of asthma on PA and PA on asthma simultaneously (Model 3x). The model assumed that PA and asthma status have simultaneous causal effects on each other, that is, PA at a previous wave causes asthma at the current wave, and asthma status at a previous wave has a causal effect on PA (figure 1).
Schematic representation of the cross-lagged GSEM model (Model 3x). BMI, body mass index; GSEM, generalised structural equational modelling; SES, socioeconomic status.
Models 1 and 2 were compared with the cross-lagged model (Model 3x) using the likelihood ratio test (LRT) to determine if one of the parsimonious models was adequate. This test served two purposes. The first was to check if the two models are equally likely. The second was to test if the effect was evident in one direction only; that is, if either Model 1x or 2x was better than the cross-lagged model (Model 3x). Small p values from the LRT would favour the more complex model. Further estimation to compare models was performed using Akaike’s Information Criterion (AIC), where lower AIC scores indicated a better fit than higher scores.35
For GSEM Model 1b (current asthma) only, an interaction term was fitted to investigate whether the association between PA and odds of subsequent current asthma differentially varied depending on the presence or absence of current asthma at a younger age.
Results
Descriptive analyses
Of the 4983 children initially recruited, 4464, 4332, 4169, 3956 and 3537 children participated at waves 2 through 6, respectively. Documents detailing non-response at each wave are available online (www.growingupinaustralia.gov.au). Female participants constituted 49% of the study sample at each wave (table 1). The percentage of children who experienced incident asthma and current asthma at each wave decreased over time. For all children, the mean time spent in MVPA increased from ages 6 to 8 years, then decreased over time. The largest reduction in time spent in PA was observed between ages 8 and 10 (waves 3 and 4).
An overview of data collection and participant characteristics across waves
Investigation of bidirectionality
Results of LRT comparing the single direction GSEM Models 1 and 2 to the cross-lagged GSEM Model 3 indicated that neither model was clearly better than the others for both current and incident asthma (table 2). Similarly, the AIC confirmed this result, showing little difference between models. Hence, each of the single-directional models was the most parsimonious models and was used to estimate the individual effects. GSEM Model 3 results are presented in online supplementary table 1.
Comparison of three GSEM models using the AIC results comparing each model and LRT results comparing each single directional model to the cross-lagged model
Individual effect estimates
Effect of PA on subsequent asthma
The adjusted logistic regression GSEM Model 1a indicated a minimal effect from PA on risk of incident asthma for ages 8–12 years (table 3). There was moderate evidence (p=0.08) of association between PA and asthma at age 10 where a 1-hour increase in PA reduced the odds of incident asthma by 10% (95% CI 0.81 to 1.01). At age 12, increasing PA increased the odds of incident asthma at age 14. However, the associated 95% CIs for each age group traversed the null value, and p values larger than 0.05 suggested that there was little evidence of association. Similarly, although PA decreased the odds of current asthma at 8, 12 and 14 years, there was little evidence of association.
Logistic regression coefficients in ORs for the GSEM Model 1 (a—incident asthma; b—current asthma) investigating the lagged effects of physical activity at each wave on incident and current asthma at the next wave as the outcome
The p values associated with the fitted interaction term at each age (p=0.1, 0.5, 0.1 and 0.3 at ages 8 to 14, respectively) indicated that having previous asthma symptoms did not modify the association between PA and current asthma. When the mean time spent in PA at each age was stratified by the presence and absence of current asthma symptoms at the previous wave, it could be seen that that previous asthma symptoms had a minimal differential impact on future time spent in PA (data not shown).
Effect of asthma on subsequent PA
Results of the adjusted GSEM Model 2a suggested that incident asthma had little effect on the average time spent doing MVPA at ages 8, 10 and 14 (table 4). There was weak evidence (p=0.14) of an association at age 12, although again the associated 95% CI traversed the null. Similarly, there was moderate evidence (p=0.09 and 0.06) for the GSEM Model 2b at ages 8 and 10, respectively, but there was no evidence to suggest that current asthma at ages 10 and 12 affects subsequent PA at ages 12 and 14, respectively.
Linear regression coefficients for the GSEM Model 2 (a—incident asthma; b—current asthma) investigating the lagged effects of asthma at each wave on physical activity at the next wave as the outcome
Discussion
This present study is the first to investigate bidirectionality of the association between asthma and PA. In a large cohort of children and adolescents aged 6 to 14 years, this study found no evidence that current asthma influenced time spent in PA nor was any effect seen from incident asthma. In addition, there was no evidence to suggest that PA longitudinally affected either incident or current asthma at any age.
Influence of PA on asthma
Contrary to the findings by Lochte et al,5 our analysis found no evidence that PA influences either current or incident asthma. The inconsistency between our results and those of the systematic review5 may be due to fundamental differences in the exposure and outcome definitions and measurement techniques employed by our team as compared with those within the systematic review. For example, while our study looked at time spent in PA, the studies contained within the review used number of team sports played,8 sports participation frequency7 and duration of tv viewing6 as their PA exposure. Additionally, these studies differed in the definition of the outcome. Two looked at doctor-diagnosed new-onset asthma,6 8 and the third investigated new-onset wheeze.7 On the other hand, a recently published Norwegian cohort study produced similar results to our own. They too concluded that low PA level at ages 3–6 and 6–10 years was not longitudinally associated with current asthma at age 13.36 Despite emerging evidence that PA may have an anti-inflammatory effect on airways,37 more evidence regarding the long-term longitudinal effect of PA in epidemiological studies is required.
Furthermore, we also found no evidence of interaction between PA and current asthma symptoms on time spent in future PA. That is, PA does not appear to differentially modify the odds of asthma in children with and without current asthma at an earlier age.
Influence of asthma on PA
We found no evidence that children with current or incident asthma participate in less PA than their peers. Despite indications by previous studies, self-imposed or parentally imposed restraints against PA involvement as a means of avoiding exacerbations2 12 are not being applied in these children, suggesting that these children and their caregivers do not perceive asthma alone to be a barrier against PA. This could be attributed to a good understanding of an individual’s specific triggers and adequate management and control of asthma symptoms, although asthma control could not be measured in the present study due to a lack of relevant data. Appropriate control of asthma symptoms, through the strategic administration of asthma medication, enables children with asthma to attain similar levels of PA as their unaffected peers.38 This is evident in our results and in the results of other international studies,13–16 many of which measured PA through accelerometry.13–16 On the contrary, many studies which reported an association collected PA information via questionnaire.9–11
Strengths and limitations
This analysis had several important strengths and limitations. Use of the data collected by the LSAC was a major strength of this study, as the repeated, cross-sectional structure of the LSAC data is particularly conducive to the cross-lagged analytical method used to investigate bidirectionality; bidirectional investigations such as this are only possible in longitudinal studies with repeated measurements of both exposure and outcome variables, and hence previous studies have conducted similar statistical analyses using data from LSAC with different outcomes.39 40 A second strength was the large sample size of the LSAC which powers the associations observed, despite incomplete data and the inevitable attrition over time.
An important limitation of this study is the use of subjective data collection techniques, such as questionnaires and TUDs. Although commonly used in research, the imprecision of these methods of measurement exposes the analysis to a substantial amount of misclassification of both the exposure and outcome variables. For example, the LSAC instrument did not capture details about the formulation, dosage and frequency of asthma medications used, thus inhibiting an investigation of the effect of asthma severity or level of asthma control. Additionally, there are limitations in extrapolating the results to habitual PA as the measurements were collected on a single day, and the intensity of the PA was not formally recorded. Coded variables that directly correspond to physical activities and exercise, rather than work-related activities such as cleaning or tiding up, were included. While this means that the estimated amount of PA is more conservative than the actual amount of PA performed, this restriction reduced the ambiguity related to certain coded activities. Hence, to minimise the possibility of incorrect classification, we restricted the activities of interest to those listed above. Finally, we were unable to adjust for familial history of asthma and atopy.
Conclusions
Our novel analysis found that in childhood and adolescence, neither current nor incident asthma was longitudinally associated with reduced reported PA. Similarly, we found little evidence to suggest that PA has a longitudinal effect on asthma development or symptoms, although more longitudinal research with objective PA measurement is warranted.
What is already known on this subject
Previous research investigating the effect of either asthma on physical activity (PA) or PA on asthma has yielded inconsistent results. However, the possibility of a reciprocal relationship has not yet been explored.
What this study adds
This study uses a novel technique to investigate the possibility of bidirectionality between asthma and PA in children and adolescents. The results imply that a physically active lifestyle should be encouraged in all children, as it does not appear to have any effect on asthma. Similarly, asthma does not appear to be a hindrance to PA in children.
Acknowledgments
The authors wish to thank all the participants of the Longitudinal Study of Australian Children.
References
Footnotes
Contributors All authors contributed to the conception or design of the work; data acquisition, analysis and interpretation of data; drafting and critical revision of the paper; checking it for important intellectual content; approved the final version to be published and agreed to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.
Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.
Competing interests None declared.
Patient consent Not required.
Provenance and peer review Not commissioned; externally peer reviewed.