Helminths are positively associated with atopy and wheeze in Ugandan fishing communities: results from a cross-sectional survey

Background: Parasitic helminths are potent immunomodulators and chronic infections may protect against allergy-related disease and atopy. We conducted a cross-sectional survey to test the hypothesis that in heavily helminth-exposed fishing villages on Lake Victoria, Uganda, helminth infections would be inversely associated with allergy-related conditions. Methods: A household survey was conducted as baseline to an anthelminthic intervention trial. Outcomes were reported wheeze in last year, atopy assessed both by skin prick test (SPT) and by the measurement of allergen-specific IgE to dust mites and cockroach in plasma. Helminth infections were ascertained by stool, urine and haemoparasitology. Associations were examined using multivariable regression. Results: Two thousand three hundred and sixteen individuals were surveyed. Prevalence of reported wheeze was 2% in under-fives and 5% in participants ≥5 years; 19% had a positive SPT; median Dermatophagoides-specific IgE and cockroach-specific IgE were 1440 and 220 ng/ml, respectively. S. mansoni, N. americanus, S. stercoralis, T. trichiura, M. perstans and A. lumbricoides prevalence was estimated as 51%, 22%, 12%, 10%, 2% and 1%, respectively. S. mansoni was positively associated with Dermatophagoides-specific IgE [adjusted geometric mean ratio (aGMR) (95% confidence interval) 1.64 (1.23, 2.18)]; T. trichiura with SPT [adjusted odds ratio (aOR) 2.08 (1.38, 3.15)]; M. perstans with cockroach-specific IgE [aGMR 2.37 (1.39, 4.06)], A. lumbricoides with wheeze in participants ≥5 years [aOR 6.36 (1.10, 36.63)] and with Dermatophagoides-specific IgE [aGMR 2.34 (1.11, 4.95)]. No inverse associations were observed. Conclusions: Contrary to our hypothesis, we found little evidence of an inverse relationship between helminths and allergy-related outcomes, but strong evidence that individuals with certain helminths were more prone to atopy in this setting. LaVIISWA is being conducted in 26 fishing villages on the Lake Victoria islands of Koome subcounty, Mukono district, Uganda, a remote setting accessible in 2–3 h from Entebbe by powered canoe. Full details of the trial design are described elsewhere 29. The baseline household survey was conducted between October 2012 and July 2013, across all trial villages, immediately preceding intervention roll‐out. All households in participating villages were eligible for inclusion in the survey. Available household listings were checked and updated by the research team, and simple random samples of 45 households were selected from each village. In selected households, all members were eligible for inclusion in the survey. Questionnaires were completed regarding household features and individual social‐demographic characteristics. Information regarding asthma, eczema and allergy symptoms was obtained using questions from the International Study on Allergy and Asthma in Children (ISAAC) questionnaire, with supplementary questions from the UK diagnostic criteria for atopic eczema 30, 31. A general history and examination, including height, weight and hepatosplenomegaly, was performed. All individuals were examined for visible flexural dermatitis: for this, all team members were trained in the standardized approach described in 32. SPTs were performed on participants aged ≥1 year, using standard methods, with three allergens [Dermatophagoides mix, Blomia tropicalis and German cockroach (Blatella germanica)] and positive and negative controls (ALK‐Abelló; supplied by Laboratory Specialities (Pty) Ltd., Randburg, South Africa). Each participant was asked for one stool sample; mid‐stream urine samples were requested from all participants in the 15 villages surveyed from February 2013 onwards. Blood samples of 14 ml were obtained from individuals ≥13 years, 10 ml from children 5–12 years and 6 ml from children 1–4 years. Individuals were offered HIV counselling and testing in collaboration with local health service providers. Ethical approval was granted by the Research and Ethics Committee of the Uganda Virus Research Institute, the Uganda National Council for Science and Technology, and the London School of Hygiene and Tropical Medicine. Individual written informed consent (for adults ≥18 years and emancipated minors, and for children by a parent or guardian) and assent (for children 8–17 years) was sought for survey participation. Two slides from each stool sample were examined (by different technicians) using the Kato‐Katz method 33. The remaining sample was suspended in ethanol and stored at −80°C to allow further investigation for Necator americanus and Strongyloides stercoralis, and, among a subset of 200 participants, for Ancylostoma duodenale, using real‐time polymerase chain reaction (RT‐PCR) 34. Quality control for PCR assays was conducted at St Elisabeth’s Hospital, Tilburg, NL. The Uganda results were comparable for N. americanus and A. duodenale, but had a lower detection rate for S. stercoralis. The presence of circulating cathodic antigen (CCA) of S. mansoni in urine was assessed (Rapid Medical Diagnostics, Pretoria, South Africa). Infection intensity based on Kato‐Katz results was classified using WHO‐recommended cut‐offs 35. For PCR results, there are no standard cut‐offs for categorizing infection intensity; however, based on results from Verweij et al. 34, individuals with C t > 30 would have parasite loads difficult to detect by microscope. Mansonella perstans infection was determined by a modified Knott’s method 36; malaria was determined by thick blood film. IgE specific to Dermatophagoides and cockroach allergens was measured by ELISA 29. The lower detection limit for our in‐house ELISA was 15.6 ng/ml. We used 20‐fold diluted plasma samples in our assay; hence, the lower detection limit in undiluted plasma was calculated as 312 ng/ml. This was used as a cut‐off to create binary variables for detectable vs undetectable responses for each allergen. This was a cross‐sectional analysis of survey data. Outcomes were reported wheeze in the last 12 months for children <5 years and for participants ≥5 years; visible flexural dermatitis; atopy defined as positive SPT response to any allergen for participants ≥1 year; atopy assessed as concentration of asIgE and analysed both as a continuous outcome and as detectable/nondetectable using the cut‐off of 312 ng/ml. Exposures for the analysis were helminth infections. The following variables were considered as potential confounders: individual socio‐demographic characteristics (age, sex, birth order, number of siblings, area of birth, area resided in for first 5 years, preschool attendance; occupation, maternal tribe, paternal tribe); behavioural and clinical characteristics (hand‐washing behaviour, BCG scar, maternal or paternal allergy/asthma/eczema, immunization history, breastfeeding, exposure to anthelminthic treatment in utero, anthelminthic treatment in last 12 months, artemisinin combination treatment for malaria in last 12 months, malaria infection, HIV infection); and household characteristics (crowding, animal ownership, asset score, indoor cooking, toilet access, drinking water source, washing water source, malaria control measures). Assuming an average of 2.8 people per household, we expected that sampling 45 households per village would yield at least 3250 participants. For common exposures (prevalence ≥20%), assuming a design effect of 1.5 and outcome prevalence of 10%, the study would have over 80% power to detect risk ratios ≥1.5. All analyses employed the ‘svy’ survey commands in Stata to allow for clustering of respondents within villages using linearized standard errors 37 and for variable village sizes using weights. Village‐level weights were calculated based on the numbers of included and total households in each village. For binary outcomes, univariable and multivariable logistic regressions were used to obtain crude and adjusted odds ratios (OR) and 95% confidence intervals (CI). P‐values were calculated using Wald tests. Adjustment was made for any potential confounder for which there was evidence of crude association with the outcome or which was considered to have a possible role, a priori. Raw asIgE responses were skewed. Therefore, we used simple and multiple linear regressions to examine the association between helminth infections and log10 levels of asIgE, and back transformed results to obtain geometric mean ratios (GMRs) and 95% CIs. For all outcomes, the role of S. mansoni infection intensity was assessed using the test for trend. The population attributable fraction (PAF) for reported wheeze due to atopy was estimated as p’(OR−1)/OR with p’ the prevalence of a positive SPT response among individuals with reported wheeze and OR the odds ratio for the wheeze–SPT association. We prespecified that we would examine whether helminth infections modified the associations between the atopy and the wheeze outcomes, by fitting interaction terms in multivariable logistic regression models. We also undertook a series of additional exploratory interaction analyses between helminth infections for each outcome, in an attempt to understand the primary association findings. Finally, as most previous studies have been performed in children, we conducted an exploratory investigation into whether associations between allergy‐related outcomes and between helminths and allergy‐related outcomes differed by age group (<16 vs ≥16 years).