School children attending schools in rural areas of Pichincha Province in Ecuador were sampled within a cluster-randomized study that examined the effect of albendazole treatment on the prevalence of allergy. The study design is described in detail elsewhere 19. Briefly, mestizo children attending the second to seventh year of primary education in 68 rural schools were recruited. Schools were randomized to receive either albendazole (single doses of 400 mg provided every 2 months for a period of 12 months [total of 7 treatments] or no treatment. Albendazole treatments were directly observed and a placebo was not used in the no-treatment group. A total of 1,632 children (68.8% of 2,373 children that were recruited at the start of the study – follow-up rates were similar in treatment (67.4%) and no-treatment (70.1%) groups) – completed follow-up at 12 months in the 68 schools but 23 of these children with no baseline stool results were excluded from the analysis. Four samples of each of 100 individuals were selected randomly from the remaining 1,609 children to represent the four possible combinations of treatment and baseline infection as follows: 1) infection/no treatment (sampling fraction, 100/622) – children in the no treatment group that were infected at the beginning of the study with geohelminths; 2) no infection/no treatment (100/214) – children in the no treatment group that were not infected with geohelminths at the beginning of the study; 3) infection/treatment (100/543) – children in the treatment group that were infected with geohelminths at the beginning of the study; and 4) no infection/treatment (100/230) – children in the treatment group that were not infected with geohelminths at the beginning of the study. The study protocol was approved by the Ethics Committees of St George' University of London, London, UK, and the Hospital Pedro Vicente Maldonado, Pichincha Province, Ecuador. Informed written consent to participate in the study was obtained from a parent or guardian of all children.

Stool samples were collected from children at the beginning of the study (i.e. before receiving the 1^st dose of albendazole) and at 12 months (i.e. before receiving the 7^th dose of albendazole). Blood samples (3 mL) were drawn from the antecubital fossa into Vacutainers (Becton Dickinson) containing sodium heparin as anticoagulant at the beginning of the study and 7 days after receiving the 7^th dose of albendazole. The blood was centrifuged and the plasma aliquotted and stored at -20C until use. Total IgE was measured in IU/mL as described previously 1. Levels of A. lumbricoides – specific total IgG antibodies were measured by an antibody-capture ELISA as described 14. Briefly, plates were coated with a PBS-soluble extract of A. lumbricoides adult female worms (obtained by expulsion after anthelmintic treatment) at 2.5 μg/mL in carbonate buffer. After incubation with plasma samples diluted at 1/100, the plates were incubated with alkaline phosphatase conjugated anti-human IgG (Jackson ImmunoResearch), developed with p-nitrophenyl phosphate (Sigma) in sodium carbonate buffer, and read on a microtiter plate reader. Unknown values were interpolated from a standard curve derived from pools of positive serum samples, and antibody levels were expressed as arbitrary units. The cut-off for a positive serologic test for Ascaris lumbricoides-specific IgG was defined as mean plus 3 standard deviations of antibody levels obtained from 8 individuals resident in the town of Quinindé in Esmeraldas Province with no previous history of geohelminth infection. Stool samples were examined using the modified Kato-Katz and formol-ether acetate concentration methods as described 20.

The random samples that comprised the four study groups were generated using computer software (Stata 7, Stata Corporation, College Station, Texas, USA). Children were analysed according to their original treatment allocation. The primary outcomes were baseline levels of IgE and fold change in total IgE expressed as a percentage (100*posttreatment IgE/pretreatment IgE). Secondary outcomes were baseline levels of anti-A. lumbricoides IgG and changes in anti-A. lumbricoides IgG expressed as a percentage also. All analyses used logarithmically transformed data for the dependent continuous variables of baseline IgE and changes in IgE. The results of transformed data for IgE parameters were back-transformed to compare explanatory variables in terms of ratios (fold differences) of geometric means. Changes in IgE levels compared to baseline were assessed using the one-sample t test. Inter-group differences for IgE outcomes were assessed using one-way analysis of variance (for 4 group comparisons) and Student's t test (for 2 group comparisons). Multivariable analyses for IgE and anti-A. lumbricoides IgG outcomes used linear regression and controlled for the a priori confounding factors, age, sex, body mass index, socioeconomic level and household crowding. Clustering by school was adjusted for using the sandwich estimator ('robust cluster' in STATA). Four models were used to explore the associations between geohelminth infections and the primary IgE outcomes. The models were chosen to investigate the independent effects on these outcomes of different parameters of geohelminth infections including the presence of any infection and exposure to any infection, the presence of infections with different geohelminth parasites, and infection intensities with A. lumbricoides and T. trichiura (the predominant geohelminth parasites in the study area). The models were: Model 1 – association with any geohelminth infection; Model 2 – associations with individual geohelminth parasite infections; Model 3 – associations with infection intensities with A. lumbricoides and T. trichiura; and Model 4 – associations with geohelminth prevalence, and exposure to geohelminth parasites. Exposure to geohelminth infection was measured by anti-A. lumbricoides IgG antibodies – because there is extensive immunologic cross-reactivity between antigen preparations from different geohelminth parasites 21, the presence of IgG antibodies to A. lumbricoides is likely to indicate past and or present exposures to geohelminth infections including A. lumbricoides and antibody levels are likely to be associated with the intensities of current and recent infections. Linear regression models for changes in IgE included also parameters for geohelminth infection at 12 months. The secondary outcomes, baseline levels and changes in anti-A. lumbricoides IgG, were logarithmically transformed and analysed using Model 3 only. Bonferroni corrections were used to conservatively reduce the probability of Type I errors in those multivariable analyses involving multiple geohelminth-associated variables – the original significance level of 5% was divided by the number of such variables in each analysis to obtain an adjusted significance level for them. Therefore, the multiples used for Models 1, 2, 3, and 4 for total IgE and changes in total IgE were: 1, 4, 6, and 3 and 2, 8, 12, and 6, respectively, Bonferroni corrections for multivariable analyses of baseline and changes in anti-A. lumbricoides IgG were as for Model 3. Socioeconomic status was calculated as a score divided into quartiles and based on paternal and maternal educational level (categorized as illiterate [lowest] to completed higher education [highest]) and occupation (i.e. categorized as day-labourer [lowest] to professional [highest]), and the number of material goods in the household (radio, television, refrigerator, and sound system). Each variable was assigned a score and the range of scores was divided into quartiles numbered 1 to 4 representing the richest to poorest quartiles, respectively. Each individual was assigned a final quartile socioeconomic level score of 1 to 4. Geohelminth infection intensities were stratified into three groups (negative, and two positive groups [above and below the median]), and anti-A. lumbricoides IgG, when used as an independent variable, into tertiles. Tests for trend were included in the regression models by allocating ordinal scores (0, 1 and 2) to the categories in ascending order, and retaining the other model covariates. Analyses were done with Stata version 8.0.

The baseline characteristics of the study groups are shown in Table 1. The mean age, sex distribution, and mean levels of crowding (number of persons per sleeping room) were similar between the 4 groups. Socioeconomic level was significantly different between the 4 study groups (P < 0.0001) and tended to be lower (i.e. higher score) in infected children in both treatment groups compared to the respective non-infected groups. A. lumbricoides and T. trichiura infections were the main geohelminth infections present in this population, and the prevalence of all 4 geohelminths present in this sample were similar between the 2 infection groups (infection/no treatment and infection/treatment). A total of 185 of the 200 (93%) individuals in the two treatment groups (infection/treatment and no infection/treatment) received all 7 albendazole treatments, and only 4 individuals received 3 or fewer treatments. Treatment was incomplete in the treatment group as a consequence of failure of a few children to attend the school on treatment days or moving house from a treatment to a no treatment school over the course of the study (3 children). Five individuals in the no treatment groups had documented albendazole treatments during the study as a consequence of moving to a treatment school during the 12-month observation period.

Changes in the prevalence of geohelminth infections are shown in Figure 1. The prevalence of infection with any geohelminth parasite declined significantly from 100% to 67% in the infection/no treatment group (P < 0.001) and group from 100% to 24% (P < 0.001) in the infection/treatment group. The treatment effect on geohelminth prevalence at 12 months was statistically significant (OR 0.39, 95% CI 0.29–0.54, P < 0.001), indicating a 61% relative reduction in geohelminth prevalence in the treatment compared to the no treatment group. Anthelmintic treatment was more effective for A. lumbricoides than T. trichiura. Forty-two percent of individuals in the no infection/no treatment group acquired geohelminth infections over the 12-month period compared to 9% of individuals in the no infection/treatment group. No infections with hookworm and S. stercoralis were observed in either of the two treatment groups at 12 months. The decline in geohelminth prevalence in the no treatment group may have been caused by treatment contamination. Anthelmintic treatment is widely available without prescription in pharmacies and dispensaries throughout Ecuador, and mothers frequently administer anti-parasite drugs to their children without medical consultation. Five individuals in the non-treated communities actually moved from treated to non-treated communities over the 12 month study period and had documented treatments with albendazole.

Geometric mean levels of IgE in the study population were 1,004 IU/mL (range 12–22,608 IU/mL) and were similar between the infection (infection/no treatment, 1,315 IU/mL vs. infection/treatment, 1,566 IU/mL, P = 0.34) and the no infection groups (no infection/no treatment, 825 IU/mL vs. no infection/treatment, 599 IU/mL, P = 0.09). There were significant differences in the levels of total IgE between infected and non-infected children within the treatment groups (infection/no treatment vs. no infection/no treatment, P < 0.001; infection/treatment vs. no infection/treatment, P < 0.001). Relatively few children (16%) had IgE levels below the upper range of normal for children living in industrialized countries 22 (age 7 years [<248 IU/mL], 18.9%; age 8 years [<280 IU/mL], 16.5%; age 9 years [<304 IU/mL], 17.3%; age 10 years [<328 IU/mL], 24.7%; age ≥11 years [<114 IU/mL], 5.3%). None of the a priori confounding factors or treatment group was associated with levels of total IgE at baseline in any of the 4 models (Table 2). Parameters of geohelminth infection were strongly positively associated with fold-increased levels (OR>1) of total IgE in all models: a) any geohelminth parasite (Model 1, P < 0.001; Model 4, P = 0.001); b) infections with T. trichiura (Model 2, P < 0.001); c) increasing infection intensities with T. trichiura (Model 3, test for trend across tertiles, P < 0.001); and d) increasing levels of anti-A. lumbricoides IgG antibodies (Model 4, test for trend across tertiles, P < 0.001).

Geometric mean levels of total IgE at 12 months and geometric mean percent changes in IgE levels (brackets) in the 4 study groups were: infection/no treatment, 1,045 IU/mL (79.5% or a posttreatment fall (-) of 20.5% compared to pretreatment levels); no infection/no treatment, 833 IU/mL (101.0% or +1.0%); infection/treatment, 999 IU/mL (63.8% or -36.2%); and no infection/treatment, 395 IU/mL (66.0% or -34.0%). Overall in the treatment and no treatment groups, IgE levels fell significantly by (-)35.1% (P < 0.0001) and (-)10.4% (P = 0.0004), respectively. Significantly greater declines in IgE levels were observed in the treatment groups compared to the no treatment groups for both non-infected (P < 0.001) and infected (P = 0.0003) children (Figure 2A). Anthelmintic treatment had relatively little impact on the proportion of children with levels of IgE within the normal range for children in industrialized countries 22: (age 7 years [<248 IU/mL), 17.6%; age 8 years [<280 IU/mL], 21.5%; age 9 years [<304 IU/mL], 17.3%; age 10 years [<328 IU/mL], 33.8%; age ≥11 years [<114 IU/mL], 12.6%). The factors associated with changes in IgE levels are shown in Table 3. Receipt of albendazole treatment was strongly associated with falls (OR<1) in total IgE over the 12-month period in all models (Table 3). The following variables were also significantly associated with falls in IgE: a) infection with any geohelminth at baseline (Model 1, P = 0.01); b) increasing infection intensities with A. lumbricoides at baseline (Model 3, test for trend across tertiles, P = 0.05) and c) increasing levels of anti-A. lumbricoides IgG at baseline (Model 4, test for trend across tertiles, P = 0.02). Increases in IgE over the study period (OR>1) were associated with the presence of any geohelminth infection at 12 months (Models 1 and 4, P = 0.02), and increasing levels of anti-A. lumbricoides IgG at 12 months (Model 4, test for trend across tertiles, P = 0.03).

All children (100%) had positive serologic assays for anti-A. lumbricoides IgG before and after treatment. Per cent changes in levels of specific IgG in the 4 study groups (Figures 2B) were: infection/no treatment, 90.2% (-9.8%); no infection/no treatment 108.9% (+8.9%); infection/treatment, 74.4% (-25.6%); and no infection/treatment, 80.3% (-19.7%). Significantly greater reductions in levels of specific IgG antibodies were observed in treatment compared to no treatment groups (IgG, P≥0.001 for both infected and non-infected children).

Infections with A. lumbricoides and T. trichiura, but not hookworm or S. stercoralis, were positively associated with anti-A. lumbricoides IgG at baseline. The A. lumbricoides infection intensity stratum of 1–8,000 epg was associated with increased baseline antibody levels compared to the egg-negative group (1.5-fold increase, 95% CI 1.22–1.85, P < 0.001), and a similar but non-significant trend was observed for the A. lumbricoides infection stratum of >8,000 epg. Increased baseline antibodies were strongly associated with the heavy infection stratum (>500 epg) of T. trichiura infection, (1.55, 95% CI 1.23–1.95, P < 0.001). Tests for trend across strata of infection intensities for A. lumbricoides and T. trichiura were not statistically significant. Percent change in levels of anti-A. lumbricoides IgG antibodies was negatively associated with A. lumbricoides infection at baseline, but not at 12 months, and was not associated with the other geohelminth parasites. Antibody levels fell by a factor of 0.81 (95% CI 0.69–0.93, P = 0.004) in the >8,000 epg stratum of A. lumbricoides infection compared to egg-negative individuals.