Combined effectiveness of anthelmintic chemotherapy and WASH among HIV-infected adults

Introduction Current global helminth control guidelines focus on regular deworming of targeted populations for morbidity control. However, water, sanitation, and hygiene (WASH) interventions may also be important for reducing helminth transmission. We evaluated the impact of different potential helminth protective packages on infection prevalence, including repeated treatment with albendazole and praziquantel with and without WASH access. Methodology/Principal findings We conducted a cohort study nested within a randomized trial of empiric deworming of HIV-infected adults in Kenya. Helminth infections and infection intensity were diagnosed using semi-quantitative real-time PCR. We conducted a manual forward stepwise model building approach to identify if there are packages of interventions that may be protective against an STH infection of any species (combined outcome) and each helminth species individually. We conducted secondary analyses using the same approach only amongst individuals with no anthelmintis exposure. We used interaction terms to test for potential intervention synergy. Approximately 22% of the 701 stool samples provided were helminth-infected, most of which were of low to moderate intensity. The odds of infection with any STH species were lower for individuals who were treated with albendazole (aOR:0.11, 95%CI: 0.05, 0.20, p<0.001), adjusting for age and sex. Although most WASH conditions demonstrated minimal additional benefit in reducing the probability of infection with any STH species, access to safe flooring did appear to offer some additional protection (aOR:0.34, 95%CI: 0.20, 0.56, p<0.001). For schistosomiasis, only treatment with praziquantel was protective (aOR:0.30 95%CI: 0.14, 0.60, p = 0.001). Amongst individuals who were not treated with albendazole or praziquantel, the most protective intervention package to reduce probability of STH infections included safe flooring (aOR:0.34, 95%CI: 0.20, 0.59, p<0.001) and latrine access (aOR:0.59, 95%CI: 0.35, 0.99, p = 0.05). Across all species, there was no evidence of synergy or antagonism between anthelmintic chemotherapy with albendazole or praziquantel and WASH resources. Conclusions/Significance Deworming is effective in reducing the probability of helminth infections amongst HIV-infected adults. With the exception of safe flooring, WASH offers minimal additional benefit. However, WASH does appear to significantly reduce infection prevalence in adults who are not treated with chemotherapy. Trial registration ClinicalTrials.gov, NCT00507221.


Methodology/Principal findings
We conducted a cohort study nested within a randomized trial of empiric deworming of HIVinfected adults in Kenya. Helminth infections and infection intensity were diagnosed using semi-quantitative real-time PCR. We conducted a manual forward stepwise model building approach to identify if there are packages of interventions that may be protective against an STH infection of any species (combined outcome) and each helminth species individually. We conducted secondary analyses using the same approach only amongst individuals with no anthelmintis exposure. We used interaction terms to test for potential intervention synergy. Approximately 22% of the 701 stool samples provided were helminth-infected, most of which were of low to moderate intensity. The odds of infection with any STH species were lower for individuals who were treated with albendazole (aOR:0.11, 95%CI: 0.05, 0.20, p<0.001), adjusting for age and sex. Although most WASH conditions demonstrated minimal additional benefit in reducing the probability of infection with any STH species, access to safe flooring did appear to offer some additional protection (aOR:0.34, 95%CI: 0.20, 0.56, p<0.001). For schistosomiasis, only treatment with praziquantel was protective (aOR:0.30 95%CI: 0.14, 0.60, p = 0.001). Amongst individuals who were not treated with albendazole or praziquantel, the most protective intervention package to reduce probability of STH infections included safe flooring (aOR:0.34, 95%CI: 0.20, 0.59, p<0.001) and latrine access (aOR:0.59, 95%CI: 0. 35

Introduction
The neglected tropical diseases (NTDs) are a group of parasitic and bacterial infections that are associated with substantial global disease and disability, particularly in low income communities. Among the NTDs, helminth infections including schistosomiasis and the soil transmitted helminthiasis (STH) are highly prevalent, infecting over two billion people [1]. Mass drug administration (MDA), or the presumptive treatment of all at risk individuals with preventive chemotherapy, is used to control helminth associated morbidity in endemic areas. In 2001 the World Health Organization (WHO) endorsed MDA as the recommended strategy for controlling STH and schistosomiasis, with a goal of reaching at least 75% of at risk populations and up to 100% of school-aged children [2,3]. The effectiveness of MDA as a control strategy is influenced by the treatment coverage of affected populations, the prevalence and intensity of different helminth species, drug susceptibility and socio-behavioral patterns that influence infection and re-infection rates [4]. In addition, while MDA reduces morbidity by decreasing parasite burden, treatment offers only temporary benefit, as individuals remain at risk of reinfection following treatment. Other strategies, including water, sanitation, and hygiene (WASH) interventions may be needed to reduce the risk of reinfection and ultimately to break transmission of schistosomiasis and STH [5][6][7]. The presence or absence of WASH, in addition to other social and economic factors, may influence the success of helminth control programs as the intensity of transmission (as estimated by the reproductive number, Ro) is influenced by the survival of free living helminth stages in the surrounding environment [8].
Recent modeling efforts highlight the potential importance of WASH in controlling helminth infections, particularly in high prevalence areas [9,10]. Data from a meta-analysis of water and hygiene protective measures suggest that drinking treated water, soap use, and wearing shoes were associated with approximately 45-60% lower odds of infection [11]. Access to and utilization of sanitation facilities is also associated with significant reductions in the prevalence of all three major STH infections; A. lumbricoides (OR: 0.54, 95% CI: 0.43-0.69), T. trichiura (OR: 0.58, 95% CI: 0.45-0.75), and hookworm species (OR: 0.60, 95% CI: 0.48-0.75) [12].
Despite these associations, data from randomized trials evaluating the impact of WASH interventions have demonstrated minimal or no benefit in reducing STH prevalence or intensity in children [13,14]. However, evaluations of the impact of school-based deworming programs indicate that treatment programs have a larger impact on reducing helminth prevalence in areas with improved sanitation facilities, as compared to areas with unimproved or no sanitation [15]. Integrating WASH and MDA strategies may be a promising approach to eliminating disease transmission, however data are limited [16,17]. One cluster-randomized trial in Kenya found a 44% reduction in the odds of A. lumbricoides infection (but not in other STH species) among children receiving both WASH and MDA, as compared to MDA alone [18]. However, another randomized participatory hygiene and sanitation behavior change intervention trial found no significant difference in helminth infection between the group receiving both WASH and deworming, as compared those receiving deworming alone [19].
Using data from a randomized controlled trial of deworming among HIV infected adults in Kenya, we evaluated associations between WASH access and utilization and helminth prevalence following repeated rounds of treatment with albendazole and praziquantel.

Study design and population
We conducted a retrospective cohort study nested within the Helminth Eradication to delay ART Trial (HEAT) [20]. HEAT was a randomized trial of single-dose albendazole (400 mg) provided every 3 months plus single dose praziquantel (25 mg/kg) provided annually for 2 years to HIV-infected adults. Participants were recruited from HIV clinics at three sites in Kenya (Kisii Provincial Hospital, Kisumu District Hospital, and Kilifi District Hospital) between February 6, 2008, and June 28, 2010. Individuals were eligible for inclusion if they were 18 years of age or older, were HIV seropositive, were not pregnant at the time of enrollment, and did not meet current criteria for ART initiation (on the basis of documented WHO disease stage and CD4+ cell count within the previous 3 months and a clinical assessment at enrollment). Participants were followed every 3 months for 24 months after enrollment.
A single stool sample was obtained from participants after 24 months of follow up. All stool samples were initially examined with a combination of direct microscopy techniques, Kato-Katz and formol-ether concentration by trained laboratory technologists at each of the study sites. Stool aliquots of approximately 1 gram were placed in cryotubes and frozen at −80˚C for subsequent DNA extraction and examination at the Leiden University Medical Center (the Netherlands). For this analysis, intensity of helminth infections was assessed by PCR only.
Following a bead-beating step, DNA was isolated using DNeasy 96 Blood & Tissue Kit spin columns in accordance with the manufacturer's instructions (Qiagen, Hilden, Germany) [21,22]. Phocin herpes virus-1 (PhHV-1) was added to the lysis buffer in each sample as an internal control and virus-specific primers and detecting probe were included in each reaction mixture. Two different multiplex real-time PCR detection panels were used for parasite specific DNA detection of STH and Schistosoma: (i) Ancylostoma spp (i.e. detecting both A. duodenale and A. ceylanicum), Necator americanus, Ascaris lumbricoides, and Strongyloides stercoralis, and (ii) Schistosoma spp (i.e. detecting both S. mansoni and S. haematobium DNA without differentiation) and Trichuris trichiura [21]. For the DNA isolation and setup of the PCR reactions a custom-made Hamilton robot platform was used, while amplification, detection and analysis were performed using the CFX real-time detection system (Bio-Rad Laboratories, USA). Negative and positive control samples for each parasite species were included in each PCR run. Fifty PCR amplification cycles were run per sample, with the output expressed as a cycle threshold (Ct). Ct reflects the amplification cycle in which the level of fluorescent signal exceeds background fluorescence and thus species-specific DNA loads in stool samples. For each parasite-specific target, DNA loads were arbitrarily categorized into the following intensity groups, according to previous publications: low (35 Ct<50), moderate (30 Ct<35) and high (Ct<30) [23].
In addition to deworming history, four WASH conditions were assessed; access to safe water via access to piped water, purchasing of purified water or treating water independently (filtration, chlorination, boiling, etc), access to a flush latrine or pit latrine in a house or on compound premises, consistent hand washing after using the latrine, and residence in a household in which the floors of the house are made of cement, iron, stone, or timber (i.e. not earthen). Latrines were not identified by improved or unimproved status. Access to these protective WASH conditions were ascertained during longitudinal HEAT surveys. If any of the responses regarding access to WASH resources varied over follow-up, we utilized a participant's response from the end of the study (when stool samples were collected), or from the preceding three months if not available at study conclusion. Study teams were able to validate reported access to latrines and safe flooring during household visits. Hand washing and safe water access were self-reported by study participants.
The primary outcome of interest was the detection of any STH species infection after 24 months follow-up. As numerous studies have found differential effects of both WASH and deworming by specific helminth species, secondary outcomes of interest included the detection of specific species of helminths [18]. Intensity of infection was examined but could not be assessed due to a lack of statistical power and precision.
We conducted a manual forward stepwise model building approach to identify the potential package of interventions most protective against a helminth infection of any STH species (combined outcome). Individual sex and age were included as covariates in the forward stepwise model building process due to their hypothesized role as potential confounders in the relationship between WASH access and helminth infection. We did not further adjust for SES as the specific WASH resources of interest lie on the causal pathway between SES and helminth infections and access to these WASH resources is a modifiable characteristic that may influence susceptibility to helminth infections.
Multiple imputation using the chained equations (MICE) method was used to address missing values for four of the exposure variables; water treatment status (22% missing), hand washing status (22% missing), safe latrine status (2% missing), and safe flooring status (2% missing). Among individuals not asked about water treatment and hand washing status during surveys, missingness was assumed to be random. Logistic regression was used to impute missing values with ten imputations. Derived estimates from each imputation were combined using Rubin's methods [24]. We also performed a sensitivity analysis including only complete cases. Estimates from the complete case analysis did not differ qualitatively from those produced with multiple imputation. Therefore, only the effect estimates, 95% confidence intervals (CIs), and p-values derived through imputation are presented.
We used logistic regression and a cutoff threshold of p 0.1 on the Wald Test as criteria for graduating a variable into the next stepwise model (i.e. branching), or as criteria for removing a variable from a model (i.e. pruning). We started with univariate models that included each independent variable and the outcome. We branched variables that met the cutoff threshold in univariate models into a single multivariate model. After pruning back any variables that did not meet the cutoff threshold, we introduced interaction terms for all independent variable combinations. The purpose of the interaction terms was to identify any potential synergy associated with access to more than one intervention (i.e. a larger protective effect than would be possible with access to one intervention alone). We pruned back any higher level interaction terms that did not meet the cutoff threshold. We then inspected the lower order interaction terms, and pruned back terms that did not meet the cutoff threshold. The branching and pruning stopped once all remaining terms met the cutoff threshold (i.e. equilibrium). Odds ratios (ORs) and their 95% CIs are reported, with significance determined at the alpha = 0.05 level.
We also utilized the manual forward stepwise model building method to identify the optimal package of protective resources relevant to each helminth species. This process was undertaken separately for A. lumbricoides, hookworm species (all identified infections were N. americanus), T. trichiura, S. stercoralis and Schistosoma spp. Finally, we evaluated the package of WASH resources most protective against helminth infections in HIV-infected adults without access to anthelmintic treatment (the HEAT placebo group) using the manual forward stepwise model building method described above.

Ethics statement
The University of Washington (UW) Human Subjects Review Committee and the Kenya Medical Research Institute (KEMRI) Ethical Review Committee approved the HEAT study protocol and the trial was registered (ClinicalTrials.gov, number NCT00507221). Study participants were adults who provided written consent in their preferred language (Kiswahili, Kisii, Luo, Giriama, or English) or, if unable to give written consent, provided oral consent in the presence of a witness and confirmed by thumbprint on HEAT study consent forms. Witnesses also signed the study consent forms.
Exactly half (50% (n = 348)) of the participants received albendazole and praziquantel in the 6 months preceding stool collection and half (n = 353) received placebo. Nearly 60% (n = 418) of the participants reported access to safe or treated drinking water and 56% (n = 393) reported consistent handwashing after using the latrine. Latrines were accessible in the homes or on the compounds of 66% (n = 480) of study participants and 46% (n = 325) lived in homes with safe, non-earthen flooring. Some of the participants (n = 59, 8%) had access to deworming treatment and all of the protective WASH resources. Only 6 individuals (0.9%) of the participants reported no protective WASH access of any kind.
In total, 152 helminth infections were detected amongst 137 infected individuals using realtime PCR, including: 21 A. lumbricoides, 60 N. americanus, 43 Schistosoma spp., 8 S. stercoralis, and 20 T. trichiura infections. There were 96 STH infections detected, none of which were Ancylostoma spp. (Table 2). When evaluating intensity of infection by qPCR, the majority of hookworm, A. lumbricoides, and T. trichiura infections were of moderate infection intensity while the majority of S. stercoralis and Schistosoma infections were of low intensity. S. stercoralis had the highest percentage of high intensity infections of any of the helminth species (25%), while T. trichiura had the lowest percentage of high intensity infections (5%) ( Table 3). During the stepwise model building process pruning and branching steps were undertaken to derive the package of potential interventions most protective against any STH infection (combined outcome). Only two of the protective interventions remained in the final model: deworming treatment and safe flooring. The odds of infection with any STH species was lower for individuals who were treated relative to those who were not (aOR: 0.11, 95% CI: 0.05, 0.20, p<0.001). The odds of infections with any STH species was also lower for individuals with access to safe flooring relative to those without (aOR: 0.34, 95% CI: 0.20, 0.56, p<0.001) ( Table 4). All interaction terms were pruned out of the model, indicating no statistical synergy (or antagonism) across interventions.
In the absence of anthelminthic treatment, the most protective package for any STH infection included safe flooring (aOR: 0.34, 95% CI: 0.20, 0.59, p<0.001) and latrine access (aOR: 0.59, 95% CI: 0.35, 0.99, p = 0.05) ( Table 5). For the specific outcome of N. americanus, safe flooring (aOR: 0.25, 95% CI: 0.13, 0.50, p<0.001) and latrine access (aOR: 0.47, 95% CI: 0.26, 0.86, p = 0.01) appeared most protective. For A. lumbricoides, the most protective package not We assessed the pairwise correlation between the independent variables in order to understand if multicollinearity was affecting the variance in the models, and thus causing some independents variables to appear statistically insignificant and drop out of the stepwise model building process. Correlation appeared to be mild, ranging from -0.05 (treatment and safe water) to 0.17 (safe flooring and toilet access). We also performed a sensitivity analysis including only complete cases. Estimates from the complete case analysis did not differ qualitatively from those produced with multiple imputation. Therefore, only the effect estimates, 95% CIs, and p-values derived through imputation are presented.

Discussion
Given recent calls for the elimination of several NTDs by 2020, evaluation of multilateral approaches to morbidity control and interruption of disease transmission through combination interventions including MDA and WASH programs is critically important. In this analysis, both treatment with albendazole and praziquantel and access to some WASH resources demonstrated benefit in reducing the probability of STH and schistosomiasis infections. Despite this, in this study there was no evidence of additional benefit (synergy) associated with access to more than one WASH resource when combined with chemotherapy. However, access to several WASH resources did appear to demonstrate benefit in reducing STH prevalence and intensity in adults who did not receive albendazole or praziquantel. Treatment with albendazole and praziquantel was effective in preventing all helminth infections with the exception of S. stercoralis. This is consistent with previous studies documenting high treatment efficacy of albendazole for A. lumbricoides (cure rate of 88%) and hookworm species (cure rate of 72%), and of praziquantel for schistosomiasis (cure rate of 76%). It is also consistent with the documented poor treatment efficacy of albendazole for S. stercoralis. [25]. The finding of consistent benefit of albendazole for T. Trichiura was somewhat surprising given previous data suggesting relatively poor treatment efficacy for this organism (28%), and may be due to the fact that individuals in this study were treated with albendazole repeatedly over 24 months [4,26].
In this study, 60% and 69% of participants had access to safe drinking water and latrine facilities, respectively. However, among individuals who received deworming medications, access to safe water did not appear to provide additional benefit in reducing helminth infections and access to latrines only appeared to reduce infection with hookworm species. The apparent contradiction between these findings and understood mechanisms of helminth exposure are echoed in previous studies [14,27], including a three year cluster randomized trial with over 50,000 participants in India that found no association between increased latrine coverage and reductions in helminth infection or intensity [13]. These findings suggest that the additional benefit of reducing exposure in individuals receiving repeated rounds of chemotherapy may be limited. In addition, access to safe water and latrines is not equivalent to consistent or effective uptake of the resources, and cultural norms regarding use of these resources greatly influence helminth exposure.
Despite the inconsistent benefit of WASH demonstrated in this study, reducing exposure through primary transmission routes may be particularly important in populations not targeted by global guideliens for routine deworming, including adults. Amongst adults who did not receive albendazole or praziquantel, the impact of WASH in reducing STH infection was more pronounced. In untreated participants, access to latrines reduced the probability of infection of both hookworm and S. stercoralis and access to safe water reduced the probability of A. lumbricoides infection. This finding is consistent with previous evidence suggesting that untreated individuals without access to improved latrines have an increased odds of infection with skin penetrating helminths such as hookworm and S. stercoralis (combined aOR 3.9, 95% CI: 2.6-5.9) and that unimproved drinking water is associated with increased odds of infection with orally-ingested helminths such as A. lumbricoides and T. trichiura (combined aOR = 2.2; 95% CI: 1.3-3.7) [28].
In many settings, handwashing has also been associated with prevention of intestinal parasitic infections [29][30][31][32][33]. However, predictive models from the data presented here suggest that consistent handwashing may not provide benefit and, in fact, may be associated with an increased probability of Schistosoma infection. This may be due to self-reporting bias that misrepresents the true number of people practicing consistent hand washing. Alternatively, this observation may be due to a practice of handwashing with water from Schistosoma contaminated water sources. Individuals who live in close proximity to water sources may be more likely to practice handwashing and also more likely to be exposed to infective cercariae. Handwashing with soap is recommended when washing hands with fresh water in endemic areas in order to reduce the infectivity of cercariae, and data regarding individual soap use was not collected in this study [34].
In this analysis, the WASH resource that appeared to have the largest and most consistently protective effect was residence in a household with non-earthen floors. Earthen floor have been shown to be a significant risk factor for hookworm infection, responsible for as much as 86% of all hookworm infections in West Africa [35]. Given that 44% of helminth infections in this cohort were hookworm species (as expected in an adult population), it is likely that safe flooring was particularly important for preventing infection in this group. Although we did not measure shoe wearing practices, consistent use of footwear has also been shown to lower the odds of hookworm infection [18,36].
While biologic rationale would seem to indicate that integration of deworming and WASH interventions would have a synergistic effect on helminth prevalence through dual removal of extant infections and sources of exposure, there was no indication of such synergy in our study. These observations are in contrast to findings from a randomized trial that identified a significant reduction in A. lumbricoides infections amongst school children who received a school-based WASH program plus albendazole as compared to children who received albendazole alone [18]. Other studies have also found that health education, latrine construction, and deworming significantly reduce STH prevalence [37], and a systematic review concluded that combined sanitation, education, and deworming reduced STH prevalence by more than 60% relative to sanitation and education alone [38]. New efforts are underway to understand the influence of integrated WASH activities and mass deworming of entire communities using a cluster randomized trial design [39].
There are several notable strengths of this analysis. These data were rigorously collected as part of a randomized trial of deworming interventions and highly sensitive molecular diagnostic tools were used to detect helminth infections. Given the poor sensitivity of microscopic techniques in low prevalence areas, including this study site [40], and the fact that 39% of infections in this cohort were low intensity, it is likely that we detected infections that would have gone undetected using standard microscopic methods. However, there were also several limitations associated with this retrospective cohort study. First, we were unable to document true WASH usage, adherence, and behavior patterns and using WASH access as a surrogate for WASH usage may be problematic [41]. In addition, our data do not provide insight into food preparation or shoe wearing practices, which could also influence helminth exposure and infection estimates. Although individuals did report whether they had access to latrines, we did not assess whether or not the latrines were improved or unimproved or if they were well maintained as original HEAT surveys did not capture UNICEF Joint Monitoring Programme indicators. Additionally, it is possible that repeat surveys regarding WASH access affected participant behavior, such that the final survey was not necessarily a representative index of risk throughout follow-up. Also, although the Schistosoma PCR targeted both S.mansoni and S. haematobium, our ability to detect S. haematobium DNA in stool may be limited. In the current study we think we have not missed many S. haematobium infections because urine microscopy revealed less than 1% prevalence. Finally, because the study was conducted in HIV-infected adults, the generalizability of our findings may be limited. However, there is some evidence that HIV infected adults may benefit from deworming and the potential additive role of WASH in this population warrants further investigation [42]. Lastly, if helminth acquired immunities are age-dependent, one might expect these protective resources to have more pronounced influences in younger populations.

Conclusion
Access to WASH is unequivocally essential, it is a human rights necessity and a hallmark of the Sustainable Development Goals. However evidence supporting the additive benefit of integrated deworming and WASH strategies has been mixed, in part due to the logistical and ethical challenges surrounding the design and implementation of large randomized WASH studies [43]. In this study, anthelmintic treatment was shown to be highly effective in reducing the probability of helminth infections. WASH resources, with the exception of safe flooring and latrine access, did not appear to provide substantial additive benefit to chemotherapy. However, among individuals not receiving chemotherapy, there was clear benefit of WASH in reducing the probability of helminth infections. Given the complexity of integrating multisectorial approaches to disease control [44], researchers and NTD programs should carefully consider if and how pursuing integrated chemotherapeutic and WASH strategies may influence capacity to deliver high quality interventions and achieve targeted outcomes.