Risk factors associated with rhinitis, allergic conjunctivitis and eczema among schoolchildren in Uganda

Background The prevalence of allergy-related diseases (ARDs), including rhinitis, allergic conjunctivitis and eczema, is on the increase globally. The causes of this increase are not well established. Objectives To investigate the risk factors associated with ARDs among schoolchildren in Uganda. Methods We conducted a secondary data analysis of a large asthma case–control study involving 1700 schoolchildren, 5–17 years, in urban Uganda. ARDs were defined according to the International Study of Asthma and Allergies in Childhood (ISAAC) questionnaire. Skin prick testing (SPT) was conducted using standard procedures and allergen-specific IgE (asIgE) using ImmunoCAP®. We employed inverse probability weighted analysis to generate estimated prevalence data and weighted odds ratios. Results The lifetime estimated weighted prevalence of reported rhinitis, allergic conjunctivitis and eczema was 43.3%, 39.5% and 13.5%; weighted prevalence in 12 months was 10.1%, 9.1% and 2.3%, respectively. There was overlap of ARDs, with 66.3% of 1193 schoolchildren who reported having ever an ARDs (including asthma) reporting two or more. Risk factors associated with reported rhinitis in the last 12 months were city residence at birth [adjusted odds ratio (95% confidence interval) 2.66 (1.42–4.99) compared to rural]; father’s [2.62 (1.79–3.83)] and mother’s history of allergic disease [2.12 (1.48–3.02)]; frequent de-worming in the last 12 months [2.01 (1.30–3.11), ≥2 versus none]; current high frequency of ‘trucks passing on the street near home’ [2.59 (1.48–4.52), ‘almost all the time’ versus rarely] and positive SPT [1.54 (1.09–2.18)] but not asIgE [1.38 (0.60–3.15)]. The same pattern of risk factors was observed for allergic conjunctivitis and eczema. Conclusion We found extensive multi-morbidity of, and overlap in the risk factors for, rhinitis, conjunctivitis and eczema—similar to asthma risk factors—among schoolchildren in urban Uganda. This suggests a similar underlying cause for all ARDs, associated with exposure to urban lifestyles and environment in Uganda.


| INTRODUC TI ON
Allergy-related diseases (ARDs) including rhinitis, allergic conjunctivitis and eczema are on the increase in most parts of the world, 1,2 but the causes of these diseases are generally not established. 3 These chronic recurrent conditions cause significant physical and psychological distress, sleep disturbance and reduced quality of life among people of all ages, but particularly among children. 3,4 The worldwide prevalence is estimated as 42% for rhinitis, 2 25% for allergic conjunctivitis 4 and 8% for eczema. 5 Studies from high-income countries have demonstrated that these ARDs often coexist, 6,7 but data from population studies in sub-Saharan Africa are scarce.
In Africa, and other low-and middle-income countries (LMICs), the prevalence of these conditions is higher in urban than rural areas. [8][9][10][11] Although data on risk factors for ARDs from Africa are scarce, there is evidence to suggest that there may be important differences in risk factors between high-income countries (HICs) and LMICs. For example, the International Study of Asthma and Allergies in Childhood (ISAAC) study reported a weaker association between ARDs and allergic sensitization in LMICs than in HICs. 12 These differences were also supported by results from our own work on risk factors for asthma, another important ARD, among Ugandan schoolchildren. 13 Children born in rural areas had the lowest asthma risk, but this was not associated with exposure to farm animals in early life 13 as has been found in Europe 14,15 and North America. 16,17 Childhood asthma was also associated with a higher parental education and social economic status, 13 in contrast to HICs where asthma is associated with low parental education and social economic status. 18,19 Understanding the risk factors for ARDs in Africa is key to identifying the causes of these conditions and will inform local intervention strategies for prevention and treatment. We therefore undertook a secondary data analysis of a large asthma case-control study involving schoolchildren in urban Uganda, in order to investigate the risk factors associated with rhinitis, allergic conjunctivitis and eczema.

| ME THODS
This was a secondary analysis of data from an asthma cross-sectional case-control study (ie disease status-case or control-is measured at one point in time). 13 For this paper, we used statistical tools that enabled us to analyse the data as a general cross-sectional study.
The statistical tool is inverse probability weighted analysis (detailed explanation in the statistics section below). The data that support the findings of this study are available in London School of Hygiene & Tropical Medicine Data Compass at http://datac ompass.lshtm.
ac.uk/1761/. Data access is restricted due to the presence of potential identifiers. To gain access, interested parties are asked to complete a data request, indicating the requested values (codebook for details freely available), and sign a data agreement. We report our findings according to the STROBE guidelines. 20

| Study population and enrolment procedures
Schoolchildren, 5-17 years, were enrolled from both primary and secondary schools in an urban area of Wakiso district in central Uganda between May 2015 and July 2017; further details of the study are described elsewhere. 13,21 For each child with asthma ('cases'), two children without a history of asthma symptoms ('controls') were randomly selected from the class register, using a random number generator programme in STATA (StataCorp). The parents or guardians of potential participants were contacted to attend a meeting, using invitation cards delivered by the children or telephone calls by the study team. During the meeting, parents/guardians who were interested in having their children participate provided written informed consent. Children 8 years or older provided written informed assent.

| Study procedures
Data on rhinitis, allergic conjunctivitis and eczema were collected using the widely used and validated ISAAC questionnaire, 22 which was administered by the study team in either English or Luganda (a language widely understood by the study population). Rhinitis was defined as 'a problem with sneezing, or a runny, or a blocked nose when you did not have a cold or the flu'; allergic conjunctivitis was defined as 'recurrent itchy-watery eyes'; eczema was defined as 'an itchy rash which was coming and going for at least 6 months, and in any of the following places: the folds of the elbows, behind the knees, in front of the ankles, under the buttocks or around the neck, ears or eyes'. We included an additional question for urticarial rash which was defined as 'itchy rash associated with wheals (ebilogologo in Luganda, a well-known terminology). Questions were answered Conclusion: We found extensive multi-morbidity of, and overlap in the risk factors for, rhinitis, conjunctivitis and eczema-similar to asthma risk factors-among schoolchildren in urban Uganda. This suggests a similar underlying cause for all ARDs, associated with exposure to urban lifestyles and environment in Uganda. by either the parents or the participant themselves (for adolescents).
We used the ISAAC environmental questionnaire 23 to collect data on risk factors for ARDs and added questions relevant to this setting, such as residence at birth and in the first 5 years of life [rural or urban (small town or the city Kampala)] and frequency of de-worming in the last 12 months.
We conducted assessments for allergic sensitization. Skin prick testing (SPT) was conducted using standard procedures 24   We generated an additional variable for any allergy-related disease which we defined as existence of any of rhinitis, allergic conjunctivitis, eczema and asthma. This was in recognition of the evidence from literature that these conditions frequently coexist, 6,7 suggesting the possibility of similar underlying risk factors that this study aimed to investigate.

| Statistical considerations
This was a secondary analysis of data from an asthma casecontrol study, and therefore, it was not a random population sample. We therefore used inverse probability weighting which takes into account the fact that asthma cases and non-asthma controls were sampled with different probabilities of selection into the case-control study. 29 This tool uses differential weighting, by giving each individual's data a weight inversely proportional to their probability of selection. Inverse probability weighting was implemented as follows. The population prevalence of asthma in urban Uganda is approximately 12%, 30 and the original asthma casecontrol study enrolled all identified asthma cases in the population under study [in Wakiso District] and randomly selected twice the number of controls. 13 One might therefore expect that for every 100 children in the source population, there would be 12 asthma cases and 24 non-asthmatic controls (out of 88 non-asthmatics) selected. We therefore assumed that all cases in the source population were sampled and that the sampling fraction in the non-asthmatics was 24/88 (0.273). We used inverse probability weighting to account for these sampling fractions in all our analyses, that is asthma cases received a weight of 1, whereas non-asthmatics received a weight of 3.67 (=1/0.273).
Using the weighted analysis, we estimated the weighted prevalence of rhinitis, allergic conjunctivitis and eczema. We used chisquared tests for comparison of the different risk factors among children with or without a given allergy-related condition. We built each multiple logistic regression model by adding one confounder (identified in literature and in preliminary analyses of the data) at a time and noted the change in effect side, we stopped adding when there was no change in effect size. 31 Variables that were strongly related (such as mother's and father's education or area of residence at birth and area of residence in the first 5 years) were not included in the same model in order to minimize problems of collinearity. 31 In order to investigate whether the results on risk factors associated with rhinitis, allergic conjunctivitis and eczema were being driven by asthma comorbidity, we conducted a subgroup analysis among schoolchildren with and without asthma separately and obtained similar findings. We present results from the weighed analysis.

| RE SULTS
We present results from the weighted analysis that included all 1700 schoolchildren enrolled in the original asthma case-control study 32 ; the detailed participant flow diagram has been published previously. 13

| Weighted prevalence of rhinitis, allergic conjunctivitis and eczema
The lifetime estimated weighted prevalence of reported rhinitis, allergic conjunctivitis and eczema was 43.3%, 39.5% and 13.5%, respectively, while the weighted prevalence in the last 12 months was 10.1%, 9.1% and 2.3%, respectively (Table 1). There was overlap of these ARDs (Table 1); of the 1193 schoolchildren who reported having ever had ARDs, 791 (66.3%) reported either two, three or four ARDs. The lifetime weighted prevalence of urticarial rash and weighted prevalence in the last 12 months was 31.9% and 1.8%, respectively ( Table 1). The overall weighted prevalence of positive skin prick test (to any of seven allergens) at enrolment was 34.7%.
For the rest of the analysis, the main outcomes are reported rhinitis in the last 12 months, the lifetime history of allergic conjunctivitis, eczema (because of larger numbers compared to the prevalence of these same conditions in the last 12 months) and the combined variable of any allergy-related diseases (that includes asthma).

| Risk factors associated with rhinitis
Schoolchildren with reported rhinitis in the last 12 months were more likely than their counterparts without rhinitis to report a father's [adjusted odds ratio (95% confidence interval), 2.  Table 2). Similar results were observed for lifetime history of rhinitis, albeit narrower confidence intervals were observed, due to bigger numbers (Table S1).

| Risk factors associated with allergic conjunctivitis
Schoolchildren with a lifetime history of allergic conjunctivitis were more likely than their counterparts without allergic conjunctivitis to report a father's [2.06 (1.56-2.71)] and mother's history of al-  (Table 3). Similar results were observed for allergic conjunctivitis in the last 12 months, albeit with wider confidence intervals due to much smaller numbers.

| Risk factors associated with eczema
Children with a lifetime history of eczema were more likely than their counterparts without eczema to report a father's [2. 19

| Risk factors associated with ARDs combined
Because of the high rate of multi-morbidity of ARDs (  (Table 5).
These results were similar for ARDs in the last 12 months, except this time the association with asIgE was statistically significant [1.85 (1.14-3.01)] (Table S2). Similar results were observed for risk factors associated with ARDs among non-asthma controls only (Table S3).

| Risk factors associated with positive skin prick test
Children with positive SPT were more likely than SPT-negative chil- For urticarial rash, the statistically significant associated risk factors were a father's and mother's history of allergic disease, reported increased frequency of de-worming and of 'trucks passing on the street near home' in the last 12 months, but not SPT, asIgE and total IgE (Table S4).
Other potential factors, including maternal smoking during pregnancy (3%), current smoking by anyone in the household (including the child, 11%) and breastfeeding history (74% of children breastfed to more than 1 year), were not associated with ARDs. Of the 1700 children enrolled in the study and tested for HIV, only 12 were positive and all were already receiving the recommended antiretroviral therapy.

| D ISCUSS I ON
We found extensive multi-morbidity and substantial overlap of the risk factors for rhinitis, allergic conjunctivitis, eczema, urticarial rash and atopic sensitization among schoolchildren in urban Uganda.
The most consistent risk factors included parental history of allergic disease, city residence at birth, current proximity to a busy road, frequent de-worming, positive SPT and elevated FENO, but not allergen-specific IgE or total IgE to crude allergen extracts.
We found that children born in the city (a proxy for mother's residence during pregnancy) were at a higher risk of ARDs than their counterparts born in the village. Of note, the schools were with studies from other developing countries. 35,36 We found a positive association between ARDs and positive SPT responses and with elevated FENO, which is consistent with underlying allergic inflammation that is common to these conditions. 3,4 However, there was a lack of association between asIgE (to three crude allergen extracts) and rhinitis, conjunctivitis and eczema, despite the standard asIgE assessments and cut-off points used, and a strong association between asIgE and SPT reactivity. This lack of association may be explained, at least in part, by high levels of cross-reactive IgE to other environmental allergens, particularly to cross-reactive carbohydrate determinants, 37 as reported by studies in Ghana. 38 Indeed, our recent work in Uganda showed that associations between asIgE or SPT sensitization and clinical allergy outcomes were weak among participants from rural, compared to urban, settings. 39 Rural residents had a higher prevalence of helminths-an important source of 'environmental' antigens. 39 Our observations are consistent with the ISAAC study, which reported a weaker association between ARDs and atopic sensitization in LMICs than in HICs. 12 This weak/lack of association between ARDs and asIgE has important implications for diagnosis and treatment of these conditions in this setting: what proportion of these conditions is allergic, and, would immunotherapy or biologicals treat as effectively as in HICs? To answer these question would require the use of component-resolved diagnosis (using purified single protein allergens) to determine IgE sensitization, with high cost implications that may be prohibitive for routine use in this setting.
The participant characteristics that were consistent across all ARDs were the reported high frequency of 'trucks on the street near home' and reported frequent use of de-worming medication.
We think 'trucks near home' is a proxy for proximity to a busy road, and this has been previously found to be positively associated with ARDs 40,41 and asthma morbidity, 42 and this is probably related to increased pollution. Air pollution as a possible risk factor is further supported by our observation that children with current rhinitis were more likely to report indoor cooking with firewood/charcoal stove as main fuel sources, similar to a South African study. 43 With regard to the reported increased frequency of de-worming among children with the different ARDs including asthma, 13 we have no ready explanation. In Uganda, schoolchildren routinely receive mass drug administration with albendazole once a year, but in this study, children with ARDs were more likely to report being de-wormed twice or more in the last 1 year. There was a trend towards low prevalence of helminths among children with ARDs but this was statistically significant for only 'rhinitis ever'. The inverse association between rhinitis and the tuberculin skin test has been reported previously in South Africa. 44 Among the study limitations, first, we performed standard tests for asIgE and total IgE on a randomly selected 400 schoolchildren, not the entire sample size of 1700, limiting our power to detect weaker associations with rhinitis, conjunctivitis and eczema (we had good power to detect strong associations). Secondly, the average age of the participants was 10 years, and as such, we could not obtain detailed information on some exposures in early life.
Nevertheless, we highlight the importance of environmental exposures in early life in increasing the risk of ARDs in the urban setting.
Thirdly, we acknowledge that estimates obtained by design weights tend to be less precise than the unweighted estimates from an unweighted study. Nevertheless, this statistical approach has enabled us to cost-effectively contribute valuable information on risk factors for ARDs from sub-Saharan Africa, using existing data from an asthma case-control study. Lastly, the variable 'father's (and

CO N FLI C T O F I NTE R E S T
None of the authors had any conflict of interest to declare.

AUTH O R CO NTR I B UTI O N S
Mpairwe H made substantial contributions to conception and design, acquisition of data, analysis and interpretation of data,

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available in