Worldwide time trends in prevalence of symptoms of rhinoconjunctivitis in children: Global Asthma Network Phase I

Abstract Background The Global Asthma Network (GAN), by using the International Study of Asthma and Allergies in Childhood (ISAAC) methodology, has updated trends in prevalence of symptoms of childhood allergic diseases, including non‐infective rhinitis and conjunctivitis (‘rhinoconjunctivitis’), which is reported here. Methods Prevalence and severity of rhinoconjunctivitis were assessed by questionnaire among schoolchildren in GAN Phase I and ISAAC Phase I and III surveys 15–23 years apart. Absolute rates of change in prevalence were estimated for each centre and modelled by multi‐level linear regression to compare trends by age group, time period and per capita national income. Results Twenty‐seven GAN centres in 14 countries surveyed 74,361 13‐ to 14‐year‐olds (‘adolescents’) and 45,434 6‐ to 7‐year‐olds (‘children’), with average response proportions of 90% and 79%, respectively. Many centres showed highly significant (p < .001) changes in prevalence of rhinoconjunctivitis in the past year (‘current rhinoconjunctivitis’) compared with ISAAC. The direction and magnitude of centre‐level trends varied significantly (p < .001) both within and between countries. Overall, current rhinoconjunctivitis prevalence decreased slightly from ISAAC Phase III to GAN: −1.32% per 10 years, 95% CI [−2.93%, +0.30%] among adolescents; and −0.44% [−1.29%, +0.42%] among children. Together, these differed significantly (p < .001) from the upward trend within ISAAC. Among adolescents, centre‐level trends in current rhinoconjunctivitis were highly correlated with those for eczema symptoms (rho = 0.72, p < .0001) but not with centre‐level trends in asthma symptoms (rho = 0.15, p = .48). Among children, these correlations were positive but not significant. Conclusion Symptoms of non‐infective rhinoconjunctivitis among schoolchildren may no longer be on the increase globally, although trends vary substantially within and between countries.


| INTRODUC TI ON
Non-infective rhinitis and conjunctivitis ('rhinoconjunctivitis') are common manifestations of allergic disease among children, and their prevalence varied substantially around the world during the 1990s, as documented by the International Study of Asthma and Allergies in Childhood (ISAAC) Phase I. 1 Approximately seven years later, a comparison of ISAAC Phase III with ISAAC Phase I assessed time trends in annual period prevalence of rhinoconjunctivitis symptoms among almost half a million children from 106 centres in 56 countries. 2 Although no consistent global pattern emerged, the average prevalence of rhinoconjunctivitis symptoms increased among both 6-to 7-year-olds and 13-to 14-year-olds. Greater increases were evident in centres from low-and middle-income countries, but prevalence decreased in many centres with the highest rates in ISAAC Phase I, suggesting that rhinoconjunctivitis symptoms may have peaked in those generally more affluent countries. 2 In this paper, we extend those earlier ISAAC time trend comparisons to include more recent surveys using identical methodology, which were conducted by the Global Asthma Network 3 in 27 centres that had previously participated in ISAAC. This offers the opportunity to assess time trends over a longer period in both higher and lower income countries. We sought to evaluate whether the prevalence of symptoms of rhinoconjunctivitis among children has continued to rise, or has plateaued, or indeed started to decline, during the first two decades of the 21 st century. We also compared this trend with that for symptoms of asthma (wheeze) and eczema (flexural itchy rash). Charlotte Rutter, UKMRC grant number MR/N013638/1.
Conclusion: Symptoms of non-infective rhinoconjunctivitis among schoolchildren may no longer be on the increase globally, although trends vary substantially within and between countries. | 3 of 12 STRACHAN eT Al.

| ME THODS
The Global Asthma Network (GAN) was established in 2012 as a successor to ISAAC, in collaboration with the International Union Against Tuberculosis and Lung Disease. GAN Phase I, adapting the ISAAC approach and methods, not only focuses upon global surveillance of prevalence and severity of asthma symptoms, but has also included ISAAC questionnaires on symptoms of rhinoconjunctivitis and eczema.
Elsewhere, we have published the rationale and study design for GAN Phase I, 3,4 the scope of completed fieldwork and its geographical overlap with ISAAC 5 and the results for time trends in prevalence of asthma symptoms, among GAN Phase I centres that previously participated in ISAAC. 6 GAN Phase I surveys followed the standardized and validated ISAAC methodology, 7-11 and a specified protocol. 3 Cluster sampling was employed, selecting from a geographically defined sampling frame (the 'study centre') at least 10 schools at random (or all schools if <10), from which all children of the relevant age (or class or grade) were surveyed. All centres studied 13-to 14-year-olds ('adolescents'), who self-completed written questionnaires at school.
Additional inclusion of 6-to 7-year-olds ('children') was optional, and their questionnaires were completed at home by their parents.
Sample sizes of at least 1000 and preferably 3000 were sought for each age group.
The symptom definitions used for comparisons in this paper were identical to those used in previous ISAAC rhinitis-related publications 1,2 : Statistical analysis used Stata version 15. 13 We derived estimates of the absolute ten-yearly rate of change in prevalence of rhinitis ever, current rhinitis, current rhinoconjunctivitis, severe rhinoconjunctivitis and hay fever ever for each centre. The standard error (SE) of this change was calculated, allowing for school-level clustering. Random-effects meta-analysis investigated heterogeneity of centre-level trends within and between countries and age groups.
Additional meta-analyses compared trend estimates from the 'earlier period' (ISAAC Phase I to ISAAC Phase III) and the 'later period' (ISAAC Phase III to GAN Phase I) for the subgroup of centres that had participated in all three surveys.
Mixed-effects linear regression models were used to compare prevalence trends from ISAAC Phase III to GAN Phase I with those from ISAAC Phase I to Phase III (including non-GAN centres) as previously published. 2 These models were fitted for each of the five symptom definitions separately. We included country-and centrelevel random intercepts to model within-centre absolute changes in percentage point prevalence per 10-year interval. Data from both age groups were combined to improve model efficiency but we included age group, region and country income group as confounders and tested for these as effect modifiers.
The relationships between observed centre-level time trends in rhinoconjunctivitis, asthma and eczema symptoms were assessed by rank correlation. For comparison between trends in the three aller- Many centre-specific changes in rhinoconjunctivitis prevalence differed from zero at conventional levels of statistical significance. Substantial and statistically significant diversity was also seen for other common outcomes (rhinitis ever, current rhinitis and hay fever). Even severe rhinoconjunctivitis, with much lower prevalence, changed significantly in several centres in both age groups (Supplementary Tables S1 and S2).

| Comparison of within-centre trends across symptoms, age groups and diseases
Among adolescents, centre-specific trends in current rhinoconjunctivitis from ISAAC Phase III to GAN correlated very closely with those for rhinitis ever and current rhinitis (both rho = 0.90, p < .0001, N = 26 centres) and to a moderate but significant degree with trends in severe rhinoconjunctivitis (rho = 0.64, p = .0005) and lifetime hay fever (rho = 0.54, p = .005). Among  When current rhinoconjunctivitis trends were compared between the two age groups, the correlation was weak and non-significant There was only a weak rank correlation between trends in asthma symptoms and current rhinoconjunctivitis among adolescents (rho = 0.15, p = .48), and none of the cross-disease correlations in the younger age group were significant. The correlation between rhinoconjunctivitis trends and eczema trends among adolescents was evident within each of four groups of countries defined by GNI.

| Comparison of time trends by period in centres with data at three time points
When the analysis was restricted to centres participating in all three surveys (13 contributing results for adolescents and 9 contributing results for children), the rate of change in prevalence of current rhinoconjunctivitis (pooled across age groups) was significantly (p < .001) lower after ISAAC Phase III than before. The inversion in slope (from positive to negative) was similar in both age groups (Table 3). This is consistent with the pattern shown for current rhinoconjunctivitis in Table 4 below.

| Modelling of time trends combining GAN and ISAAC data
Multi-level modelling compared trends in 26 GAN and ISAAC centres (the 'later period') with results from 110 ISAAC centres participating in both Phases I and III (the 'earlier period'). Within each of these two periods, a single centre could contribute data for one or both age groups surveyed at two time points.
Modelling of the combined results for current rhinoconjunctivitis found no significant difference between the age groups (interaction p = .28), nor was there effect modification by grouped WHO region (p = .31). However, there was significant heterogeneity across countrylevel income group (interaction, p < .001) and evidence of non-linearity of the trend across the time period (p = .02 for quadratic term).
When earlier and later periods were considered separately (

| DISCUSS ION
This is the most comprehensive analysis hitherto of time trends in symptoms related to allergic rhinitis among schoolchildren, across diverse study centres around the world using a standardized methodology. We followed ISAAC conventions by focusing on noninfective rhinitis symptoms accompanied by itchy-watery eyes, a symptom combination closely related to allergic sensitization, particularly to seasonal allergens, among adults 16,17 and children 18,19 in Europe. Even in high-income countries, atopy appears less relevant to rhinitis without conjunctivitis, and in less affluent settings, the symptom associations with allergic sensitization are much weaker. 19 Therefore, a global perspective on trends in these symptoms requires cautious interpretation.

Studies in Nordic countries suggest a marked increase in
prevalence of allergic rhinitis among children 20 Interval_2   3,067  3,036  1,538  2,404  2,516  2,099  3,509  2,707  1,936  3,206  3,474  1,885  3,030  3,024  2,091  3,000  3,437  3,379  1,338  2,750  3,  Note: Results from mixed models with random intercepts for country and centre, by age group and country-level income group, separately for two time periods.
Estimates for the sentinel symptom "current rhinoconjunctivitis" are shown in bold.
a Adjusted for income group. b Adjusted for age group.
Despite the smaller number of GAN centres compared with ISAAC and the incomplete overlap between these two lists, sufficient GAN centres had participated in both ISAAC Phases to allow a 3-point within-centre analysis. This clearly demonstrates a slowing or reversal of the rate of increase in prevalence of rhinoconjunctivitis previously seen within ISAAC. 2 This conclusion is robust to inclusion or exclusion of Ibadan, which was a notable outlier in the ISAAC Phase I prevalence data. 1 Furthermore, it is consistent with the broader comparison of trends in the earlier and later periods, using all available centres irrespective of overlap (Table 4).
Our analysis focused on current rhinoconjunctivitis, but the conclusions generally apply to other rhinitis-related symptoms, whereas the patterns for trends in lifetime prevalence of hay fever were somewhat different. Hay fever is a label for seasonal allergic rhinitis and/or conjunctivitis in temperate climates but is a less familiar concept in subtropical and tropical regions, where many of our centres are located.
A potential limitation is our reliance upon symptoms reported

| CON CLUS ION
The trends we observed varied substantially and significantly both within and between countries, limiting the internal and external generalizability of conclusions. Local investigation is therefore important for understanding local trends and their implications for healthcare decision-making. Nevertheless, our wide international coverage, including many centres in low-or middle-income countries, provides a global perspective, which suggests that the prevalence of symptoms of non-infective rhinoconjunctivitis may no longer be increasing among children, as it was previously.

ACK N OWLED G EM ENTS
We are grateful to the children, parents, adults who willingly partici-

CO N FLI C T O F I NTE R E S T
The authors declare that they have no conflict of interest.

PEER R E V I E W
The peer review history for this article is available at https://publo ns.com/publo n/10.1111/pai.13656.

DATA AVA I L A B I L I T Y S TAT E M E N T
ISAAC data are already deposited for wider use at the UK Data Archive: http://disco ver.ukdat aserv ice.ac.uk/catal ogue?sn=8131 (https://doi.org/10.5255/UKDA-SN-8131-1). The GAN Phase I data, including de-identified individual participant data, will be made available on the Global Asthma Network website http://www.globa lasth manet work.org/ within 12 months of all GAN Phase I analyses being published. Access will require a formal request, a written proposal and a signed data access agreement.