Edinburgh Research Explorer Population level changes in schistosome specific antibody levels following chemotherapy

Summary Aims Previous studies have reported that chemotherapy of schistosomiasis by praziquantel in humans boosts protective antibody responses against S mansoni and S haematobium. A number of studies have reported schistosome‐specific antibody levels before and after chemotherapy. Using these reports, a meta‐analysis was conducted to identify predictors of population level change in schistosome‐specific antibody levels after chemotherapy. Methods and results Following a systematic review, 92 observations from 26 articles published between 1988 and 2013 were included in this study. Observations were grouped by antigen type and antibody isotypes for the classification and regression tree (CART) analysis. The study showed that the change in antibody levels was variable: (a) between different human populations and (b) according to the parasite antigen and antibody isotypes. Thus, while anti‐worm responses predominantly increased after chemotherapy, anti‐egg responses decreased or did not show a significant trend. The change in antibody levels depended on a combination of age and infection intensity for anti‐egg IgA, IgM, IgG1, IgG2 and anti‐worm IgM and IgG. Conclusion The study results are consistent with praziquantel treatment boosting anti‐worm antibody responses. However, there is considerable heterogeneity in post‐treatment changes in specific antibody levels that is related to host age and pre‐treatment infection intensity.


| INTRODUC TI ON
Schistosomiasis is a water-borne parasitic disease of great public health importance mainly in sub-Saharan African countries. 1 Currently, there is a major effort to have schistosome control programmes in all schistosome endemic countries in Africa. [2][3][4] There is a need to understand all of the potential effects and implications of schistosome antihelminthic treatment in affected populations. One such potential impact of treatment is on the host immune response against schistosome parasites. 5,6 It is well documented that naturally acquired immunity against schistosome infections reduces both prevalence and infection intensity in the older age groups in endemic areas. 7,8 However, this protective immunity takes years of chronic infection to develop naturally. There are two reasons why anti-schistosome immunity takes long to develop, first; the parasite is capable of evading host immunity and second; the host requires a threshold of antigens to mount a protective immune response. Schistosome worms have an average life span of several years. [9][10][11] The antigens from adult worms only become accessible to the host immune system following praziquantel treatment. 5,[12][13][14] Praziquantel (PZQ) treatment has been reported to enhance host protective immunity by exposing the parasite's hidden antigens in large amounts, 5,6,15 and by removing the immunomodulatory effects of adult worms. 16,17 PZQ treatment allows hosts' immune systems to recognize schistosome parasite adult worm antigens that is required to develop protective immunity. 15 In 2001, a review of seven field studies reported high levels of heterogeneity in the type and magnitude of change in antibody levels after chemotherapy between different human populations. 18 To date, many potential factors have been suggested to explain this variation, such as pretreatment infection intensity, 19 level of schistosome endemicity, 19 age, 20,21 sex 21 and co-infection with human immunodeficiency virus (HIV). 22 This is important for understanding the consequences of the epidemiological transition occurring in populations subject to national schistosome control programs through mass drug administration both for the host and the parasite. 23 However, there has not been a systematic review of published studies investigating the potential determinants of the heterogeneity in post-treatment immune responses. Therefore, the objectives of this study are (a) to investigate the pattern of antibody levels change across different human populations and (b) to identify host and parasite factors that affect the antibody levels change after PZQ treatment in the population levels.

| Systematic review
An electronic literature search was conducted using six databases. Web of Science Core Collection, BIOSIS Citation Index and MEDLINE all of which were searched through Web of Science (www.webofknowledge.com). In addition, EMBASE, Global Health and Ovid Medicine were searched through Ovid (ovidsp.tx.ovid. com). The search terms were as follows: "schistosom*" AND ("antibod*" OR "IgA" OR "IgE" OR "IgM" OR "IgG*") AND ["albendazole" OR "metrifonate" OR "artesunate" OR "antihelmint*" OR "chemotherap*" OR "praziquantel" OR "oxamniquine" OR ("drug" AND "treatment")]. This electronic literature search was completed in January 2014. After removing duplicates, a total of 1366 unique articles were identified for consideration in the present study.
Titles and abstracts of articles were screened to exclude those that were clearly not relevant. Full texts of potentially relevant articles were then reviewed for further selection. Full texts of the relevant articles were sourced through the Web of Science, Ovid, Google Scholar (scholar.google.com), the University of Edinburgh library and the Inter Library Loan of the University of Edinburgh.
Non-English articles were included in this study, and several Chinese articles were identified and translated into English by a native Chinese speaker for the systematic literature review.

| Inclusion criteria
An article was included in this study if it met all of the following criteria: (a) human study (either sex), (b) infection with S mansoni and/ or S haematobium diagnosed by parasitological egg count, (c) participants treated with PZQ, (d) number of participants reported, (e) schistosome-specific antibody levels reported before and after PZQ treatment, (f) follow-up study conducted within 14-180 days after PZQ treatment, (g) schistosome worm antigen (WWA: whole worm homogenate and/or soluble worm antigen), and/or soluble egg antigen (SEA) used to measure antibody isotype levels, (h) participants potentially exposed to schistosome infection for their life-time and/ or longer than 1 year before the study, (i) participants' ages could be categorized as child (0-10 years old), adolescent (11-21 years old) or adult (≥21 years old).

| Exclusion criteria
Articles were excluded based on the following exclusion criteria: (a) participants had a previous history of antihelminthic treatment prior to the study participation, (b) participants were treated with any antihelminthic drug other than PZQ (eg, oxamniquine), (c) participants were specifically selected because of co-infection with HIV and/or soil transmitted helminths and/or plasmodium, (d) purified and/or recombinant schistosome antigens were used to measure antibody isotype levels, (e) participants were temporary visitors to endemic areas (ie, travellers), (f) participants were originally from endemic areas but had moved to non-endemic areas prior to the study (eg, immigrants), (g) participants were diagnosed with acute schistosomiasis, (h) clinical case reports from a single patient.
In schistosomiasis endemic areas, co-infection with soil transmitted helminths is frequently reported. 24 These studies were kept in the analysis only if they did not specifically select co-infected participants. Schistosome-specific antibody isotype levels before (baseline) and after (follow-up) treatment with PZQ were extracted from the selected articles. For those articles that reported results only in graphical format, the software DataThief III (2006) was used to extract the raw data, whenever the graph format allowed it.
In addition to antibody levels, the following information was also extracted from each article: the year of publication, article title, names of authors, study area (country, region and village), schistosome parasite species, co-infection status, co-infecting pathogen species, number of participants, age or age range, sex, height, weight, days between the treatment and follow-up, pre-  Appendix S1 and Table S1).
Potential predictors were selected according to their biological importance, as suggested by earlier studies 19,21 and if they were reported by the majority of articles included in this study. The following predictors were considered: age groups (0-10 years old, 11-21 years old, ≥21 years old), pre-treatment infection intensity [low (S mansoni: 1-99 eggs/1 g faeces, S haematobium: 1-49 eggs/10 ml urine) or high (S mansoni: ≥100 eggs/1 g faeces, S haematobium: ≥50 eggs/10 mL urine)], schistosome species (S mansoni, S haematobium, or co-infection of S mansoni and S haematobium), disease prevalence (low/moderate: <50% or high: ≥50%), and days between chemotherapy and follow-up. The "low" infection intensity in the current study is equivalent to "light" in the WHO's infection intensity categories both for S mansoni and S haematobium. However, "high" infection intensity is WHO's "heavy" for S haematobium but "moderate" or "heavy" for S mansoni. This was due to the small sample size of S mansoni studies: there were only eight observations from four articles that reported >400 eggs/1 g faeces that is "heavy" infection by WHO guideline.

| Software
Articles identified by the systematic review were recorded using

| Statistical analysis
The majority of studies included in this study used the enzyme-linked immunosorbent assay (ELISA) method to quantify antibody isotype levels and reported optical density (OD). However, OD values cannot be directly compared between studies conducted by different research groups. 25 Therefore, the outcome variable was categorized according to the direction of change in antibody levels from pre-treatment baseline to levels at follow-up. That is, pre-treatment and post-treatment antibody isotype levels were compared within the same population and the outcome was categorized as "increase" if the post-treatment level was higher than the pre-treatment level, and "decrease" if it was lower. There were seven observations that reported the exact same value of antibody isotype levels both preand post-treatment. 10,26,27 The number of those observations were too small to form their own category "no change"; therefore, they were categorized into "decrease" group in this study for analyses purposes. All observations were grouped according to the type of schistosome parasite antigens (WWA or SEA) that were used to measure antibodies and analysed separately. Prior to the data stratification, we conducted preliminary analyses. The analyses results supported the partitioning by antigen type but not parasite species; therefore, WWA and SEA were analysed separately but not parasite species.
There were 29 observations from four articles that failed to report pre-treatment infection intensity of study participants. In these cases, pre-treatment infection intensity was obtained from scientific publications describing the larger populations that contained the study populations (articles listed in Table S2). Similarly, there were three observations from two articles that did not report the schistosome infection prevalence in the study area. In these cases, prevalence was obtained from scientific publications or governmental reports from the same area or larger area that contained the study populations (Table S2).
In schistosomiasis endemic areas, infection intensity peaks in the young age group, giving a convex curve typically observed in schistosomiasis. 28 Furthermore, our preliminary analyses using a mixed effects logistic regression model suggested that there is an association between age and infection intensity (data not shown). To take this nonlinear association into account, a combination predictor for age and infection intensity was generated, with format age/infection   intensity as shown in

| Classification and Regression Tree models
Classification and Regression Tree (CART) models were used to identify influential predictors of the direction of change of schistosome specific antibodies after PZQ treatment. 29,30 Briefly, CART models are non-parametric and nonlinear analyses which allow us to investigate the pattern of data without making data distribution assumptions of both outcome and predictors. 31 CART models do not make assumption about the type of association between dependent variable and influential predictors. 32 In addition, CART models do not have mandatory minimal sample size that allows us to apply this model to study with a small sample sizes. 32 Confidence intervals or standard errors are the most common weighting methods for meta-regression 31 and for meta-CART. 33 However, in this analysis the measures of antibody levels were ELISA OD values, which cannot be compared directly when coming from different research groups. 25 Therefore, in this analysis, the sample size (the number of participants) was used for weighting. The number of participants across studies varied from 7 to 148. The potential predictors used for the analysis were as follows: schistosome species (S mansoni, S haematobium, or co-infection), days between treatment and follow-up, and disease prevalence (low/moderate or high), and age/infection intensity (child/low, adolescent/low, adult/low, child/high, adolescent/high, adult/high) ( Table 2).
The CART analysis was conducted to build a tree using the standard three steps: (a) growing a maximum-sized tree with largest number of subgroups, (b) pruning the tree to generate subtrees and (c) identifying the optimal sized tree with the minimal risk estimate following the methodology of Breiman. 30 Initially, the maximumsized complex trees were grown with data from all study variables for each antibody isotype. All potential predictors were compared using the Gini index to identify the optimum split of the dependent variable (increase or decrease in antibody level). Based on the Gini index, the strongest predictor variable and its splitting value, that is sub-groupings for categorical variables and cut-off values for continuous variables, were used to split the original data (ie, root node) into two subgroups (ie, daughter nodes). The subgroups were then divided repeatedly into smaller subgroups following the same procedure until they represented the most homogeneous subgroups achievable (ie, terminal node). In this study, terminal nodes of these maximum-sized trees were set to be pure (ie, the node contents only "increase" or "decrease" observations) or with only a single observation. Then a series of subtrees was generated by pruning the initial maximum-sized trees. To estimate the optimal subtree among the different sized subtrees, 10-fold cross-validation analysis was conducted for each subtree followed by the selection based on the one standard error (SE) rule. Briefly, the cross-validation analysis is used to estimate the risk of misclassification using a randomly selected subset (ie, test samples) of the original dataset (ie, learning sample). The optimal tree is the one that yields the minimal risk estimate. However, the noisy nature of the data and the instability of the cross-validation procedure can lead to the selection of unstable and large trees. 34 Therefore, following the one SE rule, the smallest tree that has a cross-validation risk estimate of less or equal to the minimal risk plus one SE of the minimal error was selected as the optimal tree 30,35 ( Figures S1 and S2).    (Table 3). On the other hand, none of the predictors included in the analyses had a significant overall effect on anti-SEA (IgG3, IgG4, IgG) and anti-WWA (IgA, IgE, IgG1, IgG2, IgG3, IgG4) ( Table 3).  The graph shows the fraction of observations that reported a decrease/no change (filled bar) or an increase (unfilled bar) in each antibody isotype. Chi-square tests were conducted for each pair of anti-SEA or anti-WWA antibody isotype. NS non-significant, *significant at p < 0.05, **significant at P < 0.01, ***significant at P < 0.001 a systematic review and meta-analysis to identify predictors that influence the direction of change in schistosome-specific antibody isotype levels after PZQ treatment in humans.

Anti-WWA antibodies
Our analyses showed that the antibodies' direction of change after chemotherapy is highly variable among different populations.
There was only single antibody isotype: anti-WWA IgA that have been reported to increase after PZQ treatment in all the human populations studied. Our analyses also showed that the antibodies to whole worm antigens (WWA) being reported to increase in the majority of studies. In contrast, studies of anti-SEA antibody isotypes showed a weak tendency towards decreased levels after PZQ treatment. PZQ treatment has been reported to damage adult worm tegument, therefore allowing the host immune system to detect schistosome worm antigens that would otherwise not be accessible until those worms die naturally. 5,15 This theory is in line with a previous report that PZQ treatment enhances the host immunological recognition of S haematobium specific proteins. 15 The results of the present study are consistent with this hypothesis and that the elevation of some anti-WWA antibodies (IgA, IgE, IgG1, IgG2, IgG4) is less likely to be affected by characteristics of the populations.
There are multiple schistosome-specific antibodies whose association with protection for schistosomiasis has been reported.
In particular, schistosome specific IgE and IgA are commonly associated with protection against re-infection after PZQ treatment in humans. [40][41][42] Ideally, vaccination programs would be the most to re-infection after treatment in schistosomiasis endemic areas. 45,46 In particular, IgG4 has been suggested as a possible blocking antibody that inhibits the action of protective IgE in both S haematobium and S mansoni infections. 39,47,48 Field studies have reported that the ratio of IgE to IgG4 has a positive influence on resistance to future S mansoni re-infection. 49,50 There were multiple observations that  This meta-analysis revealed that literature studies, on the whole, tended to report an increase of anti-WWA antibodies isotypes after PZQ treatment. Our analyses also show a considerable variability in change among different antigens, antibody isotypes and populations following treatment with PZQ, confirming the work of Mutapi (2001). 18 Although the combination of age and infection intensity, alongside the number of days after treatment, were identified as influential predictors for some antibody isotypes, there is no single predictor that consistently affects all antibody isotypes in the same way. Our results also demonstrated that antibody isotypes that have been reported to have a protective effect against future re-infection (anti-WWA IgA, IgE) can be stimulated by PZQ treatment in the majority of cases for a minimum of 6 months after treatment. These results, therefore, reinforce previously reported protection enhancement following PZQ treatment.

ACK N OWLED G EM ENTS
We would like to thank our colleagues from Epigroup and Parasite The funders had no role in study design, data collection and analysis, decision to publish or preparation of the manuscript.

D I SCLOS U R E S
None.