Do helminth infections underpin urban‐rural differences in risk factors for allergy‐related outcomes?

Summary Background It is proposed that helminth exposure protects against allergy‐related disease, by mechanisms that include disconnecting risk factors (such as atopy) from effector responses. Objective We aimed to assess how helminth exposure influences rural‐urban differences in risk factors for allergy‐related outcomes in tropical low‐ and middle‐income countries. Methods In cross‐sectional surveys in Ugandan rural Schistosoma mansoni (Sm)‐endemic islands, and in nearby mainland urban communities with lower helminth exposure, we assessed risk factors for atopy (allergen‐specific skin prick test [SPT] reactivity and IgE [asIgE] sensitization) and clinical allergy‐related outcomes (wheeze, urticaria, rhinitis and visible flexural dermatitis), and effect modification by Sm exposure. Results Dermatitis and SPT reactivity were more prevalent among urban participants, urticaria and asIgE sensitization among rural participants. Pairwise associations between clinical outcomes, and between atopy and clinical outcomes, were stronger in the urban survey. In the rural survey, SPT positivity was inversely associated with bathing in lakewater, Schistosoma‐specific IgG4 and Sm infection. In the urban survey, SPT positivity was positively associated with age, non‐Ugandan maternal tribe, being born in a city/town, BCG scar and light Sm infection. Setting (rural vs urban) was an effect modifier for risk factors including Sm‐ and Schistosoma‐specific IgG4. In both surveys, the dominant risk factors for asIgE sensitization were Schistosoma‐specific antibody levels and helminth infections. Handwashing and recent malaria treatment reduced odds of asIgE sensitization among rural but not urban participants. Risk factors for clinical outcomes also differed by setting. Despite suggestive trends, we did not find sufficient evidence to conclude that helminth (Sm) exposure explained rural‐urban differences in risk factors. Conclusions and clinical relevance Risk factors for allergy‐related outcomes differ between rural and urban communities in Uganda but helminth exposure is unlikely to be the sole mechanism of the observed effect modification between the two settings. Other environmental exposures may contribute significantly.


| INTRODUCTION
Advances in health and hygiene practices have transformed highincome countries into "cleaner" environments, with reduced infection exposure. Consequently, homeostatic immunomodulatory effects of exposure to microbes and parasites that co-evolved with mammalian species (the "old friends hypothesis") have been lost. 1 The surge in allergy-related diseases alongside other chronic inflammatory diseases in high-income countries over recent decades has been partly attributed to this phenomenon. 2 Although other environmental exposures 3 may contribute, substantial support for the "old friends hypothesis" comes from studies in high-income countries, 4-9 which show that traditional farming and related microbial exposures 10 are associated with protection against allergy-related diseases. Additional evidence suggests a parallel relationship between ongoing urbanization and increasing allergy-related disease prevalence in tropical lowand middle-income countries (LMICs). 11,12 Akin to farming environments in high-income countries, rural LMIC settings are relatively protected against allergy-related diseases. [13][14][15][16][17] Animal models and in vitro experiments in human samples have identified helminths as potent inhibitors of allergic reactions, [18][19][20] leading to the hypothesis that they are partly responsible for the low overall prevalence of allergy-related diseases in tropical LMICs and the observed rural-urban disparities in allergyrelated disease prevalence in the same settings. 16,21 Helminths may dissociate risk factors, such as atopy, from allergy-related disease: work in Ugandan children showing that hookworm infection dissociates allergen-specific IgE from the effector phase of the allergic response 22 is strongly suggestive. However, little comparative analysis of risk factors for allergy in rural vs urban LMIC settings has been conducted. Exploration of these factors in LMICs, where an epidemiological transition is ongoing, provides an unprecedented opportunity to better understand interactions between the environment and the allergic pathway and allergy-related disease outcomes.
Using data generated from two surveys in Uganda, one in rural helminth-endemic Lake Victoria island fishing villages and another in nearby mainland urban communities with lower helminth exposure, we investigated socio-demographic, behavioural, clinical and immunological characteristics as risk factors for allergy-related outcomes and assessed whether helminth infections contribute to ruralurban differences in these risk factors. The "rural survey" was part of the Lake Victoria Island Intervention Study on Worms and Allergy-related diseases (LaVIISWA; ISRCTN47196031), a cluster-randomized trial of standard vs intensive anthelminthic intervention, described elsewhere. 24,25 A baseline household survey preceded the trial intervention; helminth-allergy associations at baseline have been reported. 24 A household-based allergy outcomes survey (the "rural survey") was conducted between September 2015 and August 2016, following 3 years of anthelminthic intervention: there was no difference in the prevalence of allergy outcomes between the two trial arms. 26 Sampling for the survey involved random selection of 70 households from each village using a Stata program. All household members (1 year and older) of selected households were then invited to participate. Permission for household participation was granted by the household head.
The urban survey of allergy-related outcomes (September 2016-September 2017) was designed intentionally to collect data from Entebbe municipality for comparison with the helminth-endemic rural survey. Before the start of the survey, each sub-ward was mapped onto satellite imagery of the municipality. A random point generation function of ArcGIS software (version 10.4.1, Environmental Systems Research Institute, Redlands, CA) was then used to generate random starting points within each sub-ward. The number of starting points selected was proportional to the population size of the sub-ward. Coordinates of the random starting points generated were loaded onto geographic information system (GIS) devices (eTrex®, Garmin ™ Ltd, Olathe, KS). These devices were then used in the field to identify the selected random points, from which the nearest four houses were surveyed.
There was no randomization to intensive or standard anthelminthic treatment in the urban survey; however, all other procedures were designed to be identical in both the urban and the rural survey.
Following written informed consent and assent, questionnaires were completed for each participant, capturing socio-demographic, clinical and behavioural characteristics as well as asthma, eczema and allergy symptoms. The latter employed questions based on the International Study on Allergy and Asthma in Children (ISAAC) questionnaire. Blood, stool and mid-stream urine were collected.

| Allergy-related outcomes
Outcomes were skin prick test (SPT) reactivity to allergens common in our setting, 30  | 665 assessed using standard procedures. 31 SPT reactivity was defined primarily as a positive response to any of the three allergens. SPT reactivity was also analysed as a positive vs negative response to individual allergens.
Whole allergen (Dermatophagoides pteronyssinus, peanut [A hypogaea] and B germanica) extract-specific plasma IgE (asIgE) was measured by ImmunoCAP® (ThermoFisher Scientific, Uppsala, Sweden) in a sample of 780 and 345 rural and urban survey participants, respectively, randomly selected from those with sufficient volume of stored plasma. Allergen-specific IgE sensitization was defined as a positive ImmunoCAP response (IgE concentration ≥0.35kU/L) to any of the three allergens and as a positive vs negative ImmunoCAP response for individual allergens. ImmunoCAP IgE outcomes were also analysed as continuous variables.
Wheeze is considered a good proxy for asthma in epidemiological studies 32 and was assessed separately in two age groups (≥5 years and <5 years) using an interviewer-administered ISAAC questionnaire.
The principal age group of interest was ≥5 years because wheeze cannot be assumed to represent asthma in children below 5 years. 33 Data on recent rhinitis (runny/blocked nose or sneezing accompanied by watery and itchy eyes, in the absence of cold or "flu") and urticarial rash (pruritic rash with weals, known as "ebilogologo" in the local language [Luganda]) were obtained by questionnaire. Visible Stata "svy" commands were used to allow for clustering of participants within villages and for the non-self-weighting design of the rural survey 24 and for clustering by sub-ward in the urban survey.
Logistic regression was used to compare the prevalence of outcomes and other characteristics between the rural and urban survey and to assess associations between each pair of allergy-related outcomes in both surveys. Population attributable fractions (PAFs) for pairs of allergy-related outcomes were calculated. Interaction tests were done to assess whether these associations differed by setting.
Unadjusted and adjusted odds ratios (OR) for associations between exposures and allergy-related outcomes were estimated using univariable and multivariable logistic regression. Additionally, linear regression was used in secondary analyses of ImmunoCAP IgE outcomes as continuous variables. Age, sex (a priori) and factors showing evidence of crude association with an outcome (P < 0.05) were considered in multivariable analyses for that outcome. We hypothesized that helminth infections might be key mediating factors on the causal pathway between urban/rural residence and allergy-related outcomes; hence, helminths (and Sm-specific antibody responses and other "helminth-related" factors such as frequency of lake contact and occupation) were not included in multivariable analyses for other risk factors. The potential mediating role of helminths was then investigated separately by assessing whether associations between non-helminth-related risk factors and allergy-related outcomes changed substantially when adjusted for Sm infections and Schistosomaspecific antibody levels. These analyses were initially conducted separately for each survey. Subsequently, we merged data from the two surveys and tested for interaction between the rural and urban survey, to assess whether risk factors for allergy outcomes differed by setting. Here, we also assessed the potential role of helminths in urban-rural interactions by comparing interaction P values before and after adjusting for Sm infection. A 5% significance level was used for all analyses. Participant characteristics differed between the two study settings (Table 1). Significantly, rural, compared to urban participants, were more likely to be infected with helminths (including Sm), malaria and HIV, had higher median levels of Schistosoma-specific antibodies and were more likely to report anthelminthic or malaria treatment in the previous 12 months. Dermatitis and SPT reactivity were more prevalent among urban participants, while asIgE sensitization and urticaria were more common among rural participants (Table 1 and Figure 2A). The prevalence of wheeze and rhinitis was similar between the two communities.

| Associations between allergy-related outcomes
Crude associations between allergy-related outcomes are shown in Table 2. Individuals who were ImmunoCAP asIgE sensitized were more likely to have a positive SPT response in both surveys; the PAF for SPT reactivity associated with asIgE sensitization was 86.1% and 80.9% for the urban and rural survey, respectively. Atopy measures (asIgE, SPT) were generally more strongly associated with other allergy-related conditions in the urban compared to rural survey; asIgE-rhinitis (interaction P = 0.081), asIgE-urticaria (interaction P = 0.056), SPT-rhinitis (interaction P = 0.019) and SPT-urticaria (interaction P = 0.005) associations approached statistical significance. Another major difference was that urticaria was associated with wheeze, rhinitis and SPT reactivity in the urban survey, but not with any allergy-related outcome in the rural survey.
We hypothesized that helminth infection, particularly Sm infection, might mediate this effect modification between the urban and rural setting ( Figure 2B). However, the comparison of crude associations (reported above) with associations adjusted for current Sm infection (generally, or categorized by infection intensity) and Schistosoma-specific antibody concentrations did not show clear differences in the test statistics (Table S1); hence, any mediating role of current Sm infection, including effects on interactions between the rural and urban survey, was not evident. Table 3 and Table S2 show  Associations with SPT reactivity to individual allergens are summarized in Table S3, and paint a similar picture.

| Factors associated with skin prick test reactivity
Comparison of models with and without additional adjustment for current Sm infection (generally, or categorized by infection intensity) and Schistosoma-specific antibodies did not suggest any mediating role of Sm infection in associations between non-helminth-related risk factors and SPT reactivity, or in interactions between the rural and urban survey (Table S4A). Table 4 and Table S5 show factors associated with ImmunoCAP IgE sensitization to any of D pteronyssinus, A hypogaea or B germanica extracts. In the urban survey, the presence of younger siblings and SWA-specific IgG were associated with asIgE sensitization. Rural participants who washed hands after toilet use, slept under a mosquito net and/or had recently been treated for malaria were less likely to be asIgE sensitized. Engaging in agricultural/fishing/lake-related activities or being unemployed, Sm infection (KK) and intensity, and elevated SWA-specific IgE increased the odds of asIgE sensitization.

| Factors associated with allergen-specific IgE sensitization
The presence of younger siblings (interaction P = 0.008) and hand washing (interaction P = 0.003) were associated with reduced odds of asIgE sensitization in the rural but not the urban survey (Table 4). Adjusting for Sm infection in multivariable analysis models did not suggest a mediating role for Sm in these rural-urban differences (Table S4B). Table S6 summarizes factors associated with ImmunoCAP asIgE sensitization to individual allergens: Schistosoma-specific antibody levels and helminth infections were the predominant risk factors in both surveys. Hygiene practices (washing and bathing) reduced the odds of sensitization in the rural but not urban survey.

| Factors associated with clinical allergy-related outcomes
Factors associated with self-reported recent wheeze, urticarial rash and rhinitis are shown in Table S7. Risk factors for visible flexural dermatitis could not be assessed because it was rare in both settings. In the urban survey, the presence of older siblings, handwashing before eating, SWA-specific IgG and SEA-specific IgG were inversely associated with wheezing. In the rural survey, female sex and presence of any nematode infection were inversely associated with wheezing, while increasing age, SWA-specific IgG, SEA-specific IgG and paternal history of allergies increased the odds of wheezing. Non-Ugandan paternal tribe (interaction P < 0.001) increased the odds of wheezing in the urban but not rural survey, while SWA-specific IgG (P < 0.001) and SEA-specific IgG (P = 0.001) were positively associated with wheezing in the rural but not the urban survey.
Urban individuals who received any anthelminthic treatment in the previous 12 months were more likely to report urticarial rash. In  Percentages adjusted for survey design. Percentages that are significantly higher in one setting compared to the other (P ≤ 0.05) are highlighted in bold. Adjusting for age and sex differences had no significant impact on these differences. *P values obtained from survey design-based logistic regression. **P values obtained from survey design-based linear regression.
F I G U R E 2 Urban-rural differences in risk factors for allergy-related outcomes in Uganda: a role for helminths? A, summary of principal findings regarding prevalence of allergy-related outcomes in urban Uganda and in rural Ugandan fishing communities. B, Risk factors for allergy-related outcomes differed between urban and rural settings. Our data suggest that helminth exposure is unlikely to be the only factor involved in this effect modification. Additional hypothesized effect modifiers are indicated NKURUNUNGI ET AL. We did not find any evidence to suggest that current Sm infection influenced associations between non-helminth-related risk factors and clinical allergy-related outcomes, and interactions between the rural and urban survey (Table S4, C-E).

| DISCUSSION
We show risk factors for allergy-related outcomes in proximate Ugandan rural and urban settings. The rural setting was characterized by a significantly higher prevalence of Sm and nematode infections compared to the urban setting. The prevalence of SPT reactivity and visible flexural dermatitis was lower, and that of asIgE sensitization and urticaria higher, in the rural compared to urban setting. Risk factors for these outcomes differed by setting.
We investigated the hypothesis that rural-urban differences in risk factors for allergy were attributable to differences in current Sm Our rural and urban settings were atypical. Observations in the rural survey are against a backdrop of three years of well-organized community-level anthelminthic intervention 25 that led to a decline in helminth intensity in both standard and intensive treatment arms, but had no effect on overall Sm prevalence. 26 Before analysis of risk factors, we confirmed a lack of effect of the intensive (compared with standard) anthelminthic treatment on allergy-related outcomes.
The urban survey was done in the unusual context of a setting with considerable exposure to light Sm infection (inferred from 44% urine CCA positivity). However, this enabled us to adjust for Sm infection in both settings and hence explore the role of Sm in interactions between the settings. Recruitment of participants in the urban survey was done after conclusion of the rural survey; however, this is unlikely to account for observed urban-rural differences in allergy risk factors, as both surveys were conducted by the same research team, and covered approximately 1 year (so any seasonal effects were approximately matched). Another potential limitation was the large number of statistical tests, increasing likelihood of chance findings. However, we were cautious to look for patterns of association rather than interpreting individual results equally.
In keeping with the "old friends" hypothesis 1 and observations from several studies, [37][38][39] SPT reactivity was less prevalent in the helminth-endemic rural setting and was inversely associated with helminth infections in the same setting. The only exception was Trichuris trichiura infection, which was weakly positively associated with Dermatophagoides SPT (Table S3). This lone observation was also manifest in the same communities in a baseline household survey 3 years earlier, 24 although no other helminth species were associated with SPT then. The current observations beg further investigation into the impact of anthelminthic treatment on SPT-helminth associations in a helminth-endemic setting. In mice, allergic airway inflammation is increased during acute Sm infection but reduces drastically with progression to chronic infection. 40 In our urban setting, light Sm infection was positively associated with SPT reactivity while moderate and heavy infections were inversely associated with the same outcome (Table 3). "Helminth-related" behavioural characteristics were also inversely associated with SPT reactivity in the rural survey.
It is plausible that in these fishing communities, frequent lake contact, bathing in lakewater and handwashing, for example, increase the risk for Sm infection through contact with infected snails. Indeed, these characteristics were strongly associated with Sm infection (P < 0.001). However, the same characteristics were also inversely associated with asIgE sensitization in the rural survey but not in the urban survey.
As discussed earlier, Pinot de Moira and colleagues' study in a Ugandan village found that hookworm infection abrogated the predicted association between Dermatophagoides-specific IgE and basophil histamine release. 22 We postulated that the rural setting might interfere with the link between atopic sensitization (asIgE, SPT) and clinical outcomes (reported wheeze and rhinitis) through high helminth exposure. Indeed, we found that associations between asIgE or SPT sensitization and clinical outcomes were weak among participants from the rural compared to the urban setting. However, statistical analyses did not suggest that this Sm or Schistosoma-specific antibodies had no effect on this association.
Allergen-specific IgE sensitization, particularly to cockroach, was more prevalent in the rural compared to the urban setting, possibly due to the higher helminth prevalence in the former. Additionally, helminth infections and Schistosoma-specific antibody levels were positively associated with asIgE in both surveys. Our immunoassays measured IgE sensitization to crude allergen extracts; these may contain cross-reactive components that are conserved in several helminth antigens, 41-46 explaining the above associations. These crossreactive components may be less effective at mediating the effector phase of the allergic response, explaining the lower prevalence of SPT reactivity in the helminth-endemic rural survey.
Associations with wheeze and rhinitis should be interpreted with caution, because these outcomes were relatively rare. Furthermore, reported wheeze can easily be misclassified in these populations, because there is no direct translation of the word "wheeze" in the local languages. 24,47 Nonetheless, rural-urban differences in the risk factors for these outcomes were visible. For example, while Schistosoma-specific antibody levels were inversely associated with wheezing in the urban survey, the reverse was true in the rural setting.
Urticarial rash was a more common outcome, particularly in the helminth-endemic rural survey, where it may be indicative of parasiteinduced skin allergy 48 and reaction to parasite antigens following anthelminthic treatment. 49 Support for these deductions comes from our observations that recent anthelminthic treatment (urban survey) and SEA-specific IgE (rural survey) were associated with urticaria.
In conclusion, we show that risk factors for allergy-related outcomes differ between rural and urban communities in this tropical setting. However, our analyses did not confirm a role for current helminth (Sm) infection as the primary mechanism of the observed effect modification between the two settings, despite indicative trends. Differences in other environmental exposures may contribute significantly.

ACKNOWLEDG EMENTS
We thank Entebbe municipality and Koome sub-county community members for participating in the urban survey and the rural (LaVIISWA) study, respectively. These findings are presented on behalf of the following members of the LaVIISWA and urban survey research teams: project leaders, physicians, postdoctoral scientists:

CONF LICT OF I NTEREST
The authors declare no conflict of interest.