Bulletin of the World Health Organization

Prevention and control of neglected tropical diseases: overview of randomized trials, systematic reviews and meta-analyses

Shanthi Kappagoda a & John PA Ioannidis b

a. Division of Infectious Diseases and Geographic Medicine, Stanford University School of Medicine, Stanford, United States of America (USA).
b. Stanford Prevention Research Center, Stanford University School of Medicine, 1265 Welch Road, MSOB X306, Stanford, California, 94305-5411, USA.

Correspondence to John PA Ioannidis (e-mail: jioannid@stanford.edu).

(Submitted: 29 August 2013 – Revised version received: 18 December 2013 – Accepted: 02 January 2014 – Published online: 13 March 2014.)

Bulletin of the World Health Organization 2014;92:356-366C. doi: http://dx.doi.org/10.2471/BLT.13.129601

Introduction

More than one billion of the world’s poorest people are affected by neglected tropical diseases (NTDs), which are a group of parasitic, viral and bacterial infections that each year cause an estimated 534 000 deaths and a disease burden of 57 million disability-adjusted life–years (DALYs).1 The World Health Organization (WHO) advocates five strategies for preventing and controlling NTDs: preventive chemotherapy, intensified case management, control of disease vectors, provision of clean water and sanitation and veterinary public health measures.2 Historically, the development of drugs for these diseases has been limited by a lack of market incentives.3 More recently, the formation of public–private partnerships for drug development has increased investment in research and development but the results have been uneven, with some diseases benefiting more than others.4 For some NTDs, such as geohelminth infection, affordable and effective treatments do exist but their availability for people living in highly endemic areas is often limited.4 For many others, treatment is inconvenient, poorly tolerated and expensive. A rational and comprehensive approach to disease control may, therefore, involve: (i) prevention strategies, including combined preventive chemotherapy (i.e. the treatment of more than one disease by the mass administration of more than one drug concurrently); (ii) improved access to clean water and sanitation; and (iii) the reduction of disease transmission by insect vectors. In addition, integrating efforts to control several NTDs into a single programme may reduce costs and streamline implementation.1,58

Evidence from randomized controlled trials (RCTs) can provide valuable information about the relative merits of different preventive interventions. However, the evidence may be scattered across several different trials. Moreover, although several systematic reviews and meta-analyses have been carried out, typically each has considered only one or a few interventions for a single disease, thereby creating a fragmented picture of the evidence available from RCTs. To prioritize research into the control of NTDs and to make evidence-based decisions about prevention, we must know: (i) the extent to which the RCTs available and associated systematic reviews and meta-analyses address the most important questions about control and prevention; (ii) whether these studies leave modest or large gaps in evidence and (iii) whether any interventions have been found to be consistently effective.

To address these issues, we systematically collected evidence from RCTs available in the peer-reviewed literature on the prevention or control of the principal NTDs and from corresponding systematic reviews and meta-analyses. Our aims were to evaluate the evidence from RCTs, to identify interventions that were found to be effective in systematic reviews and meta-analyses, to determine whether different meta-analyses on the same topic yielded similar or conflicting conclusions and to identify gaps in the evidence available.

Methods

Randomized controlled trials

We searched PubMed and the Cochrane Central Register of Controlled Trials for RCTs published on or before 31 December 2012 that addressed the prevention or control of 16 NTDs: American trypanosomiasis (Chagas disease), Buruli ulcer, cysticercosis, dengue, dracunculiasis (guinea-worm disease), echinococcosis (hydatid cyst disease), foodborne trematode infection, geohelminth infection, human African trypanosomiasis, leishmaniasis, leprosy, lymphatic filariasis, onchocerciasis, rabies, schistosomiasis and trachoma. We sought additional trials by reviewing our own literature collections, English-language systematic reviews, meta-analyses and Cochrane reviews, and the references of eligible publications we identified. The search strategy is given in detail in Table 1 (available at: http://www.who.int/bulletin/volumes/92/5/13-129601).

For several diseases, prevention, control and treatment overlap. For example, preventive chemotherapy is a disease control strategy that encompasses treatment and prevention: in highly endemic areas, periodic mass drug administration both provides treatment for infected individuals and decreases the burden of disease in the community by reducing transmission and preventing new cases.9 Currently, preventive chemotherapy is used for schistosomiasis, lymphatic filariasis, geohelminth infection, onchocerciasis and trachoma.9 Our study included preventive chemotherapy trials, which were defined as trials in which chemotherapy was given to a group of participants regardless of their infection status (i.e. without testing or screening for disease), either by mass drug administration to the whole population or by targeting treatment at a known high-risk group (e.g. schoolchildren). We excluded trials of individual treatment in which only infected participants were randomized. We also excluded trials of diagnostic tests, pharmacokinetic studies in healthy volunteers, trials with nonhuman subjects, non- and pseudo-randomized trials and trials that addressed the prevention of disease complications (e.g. trials of footwear for preventing foot ulcers in leprosy patients). Furthermore, we excluded trials published only as abstracts, descriptions of planned studies, subgroup or secondary analyses of previously published RCTs and trials reported in languages other than Dutch, English, French, German, Portuguese or Spanish. When we found a preliminary report of clinical trial data that were later included in a more complete publication, we included only the final publication. Trials that addressed more than one disease were included in the data set only once, although, for completeness, they are listed in each relevant disease section in Table 2.

Systematic reviews and meta-analyses

We carried out a separate search of PubMed and the Cochrane Database of Systematic Reviews to identify English-language meta-analyses and systematic reviews on the control or prevention of NTDs published on or before 31 December 2012. Eligible reviews had to contain at least one RCT that had been reported in a peer-reviewed publication and had to address the efficacy of any interventions used to prevent or control one of the 16 diseases of interest. We excluded articles on treatment, diagnosis, epidemiology, disease burden or the molecular biology, evolution or ecology of the etiological agent. We also excluded reviews that exclusively addressed animals (for example, vaccines in livestock) and protocols for planned reviews. For Cochrane reviews, we included only the most recent update. Systematic reviews had to include a methods section that described a comprehensive search strategy, with inclusion and exclusion criteria. A subset of systematic reviews included a meta-analysis that provided a formal quantitative synthesis of study results. We excluded three publications that were primarily reviews of reviews rather than of primary research data, though we read these publications to identify any additional suitable RCTs or systematic reviews. The search strategy is described in detail in Table 3 (available at: http://www.who.int/bulletin/volumes/92/5/13-129601).

Data analysis

We extracted the following information from each published RCT: first author, publication year, journal title, study country or site, study design (i.e. cluster, crossover or neither), interventions, sample size and follow-up period. We used the “unit of randomization” for calculating sample sizes: for trials in which individual participants were randomized, the sample size was the total number of individuals randomized and, for cluster randomized trials, the sample size was the total number of clusters (for example, neighbourhoods or villages). For both individually and cluster randomized trials, the sample size was taken to be the total number of individuals or clusters initially randomized rather than the number remaining after accounting for those lost to follow-up. We noted whether the trial report included a funding statement and ascribed funding to one or more of four sources: (i) a government or other public agency, including WHO and national public organizations such as the National Institutes of Health in the United States of America; (ii) industry; (iii) a charity or foundation; and (iv) a university or hospital. For each of the 10 diseases for which data were available, we calculated the Pearson correlation coefficient for the correlation between the number of RCTs performed and the annual global burden of disease, expressed in DALYs, and between the total sample size and the annual global disease burden.10 We obtained data on the disease burden of American trypanosomiasis,11 dengue,11 leishmaniasis,11 leprosy,11 lymphatic filariasis,11 onchocerciasis,11 rabies,12 geohelminth infections,11 schistosomiasis11 and trachoma.11 No reliable estimates were available for Buruli ulcer.

For each systematic review, we extracted details of the first author, the publication year and the interventions addressed. Most meta-analyses and systematic reviews included both RCTs and nonrandomized studies. For each RCT, we extracted data on the sample size and the primary outcome. For reviews that included a meta-analysis, we recorded the effect size with 95% confidence intervals and, for all systematic reviews, we recorded the authors’ conclusions, including their views on whether the intervention could be classified as: (i) likely to be effective; (ii) likely to be ineffective or (iii) of unknown efficacy due to a lack of sufficient evidence. When two or more reviews were available on the same intervention, we recorded whether they had similar or conflicting conclusions. Finally, we determined which RCTs in our RCT data set had not been included in a systematic review.

Results

Randomized controlled trials

The initial literature search for RCTs identified 2855 publications (Fig. 1, available at: http://www.who.int/bulletin/volumes/92/5/13-129601), of which 223, containing 236 eligible RCTs, were retained for analysis. Five publications were added from our literature collection or from the bibliographies of the publications retained; 18 were added from the bibliographies of systematic reviews or meta-analyses. The final analysis included 246 publications containing details of 258 RCTs (Appendix A, available at: https://stanford.box.com/s/nm7ri7xnxq58m744gcl5).

Fig. 1. Literature search for randomized controlled trials on the prevention and control of neglected tropical diseases, to 2012
Fig. 1. Literature search for randomized controlled trials on the prevention and control of neglected tropical diseases, to 2012
RCT, randomized controlled trial.

Table 2 shows, for 11 NTDs, the number of RCTs on prevention or control performed, the total sample size and the annual global disease burden. Geohelminth infection was studied most (51 RCTs with a total sample size of 22 848), followed by leishmaniasis (46 RCTs with a total sample size of 22 758) and rabies (34 RCTs with a total sample size of 5176). No RCT had been performed on the prevention or control of five diseases: cysticercosis, dracunculiasis, echinococcosis, foodborne trematode infection and human African trypanosomiasis. There was no significant correlation between disease burden and either the number of RCTs (ρ = 0.12, P = 0.73) or the total sample size (ρ = −0.38; P = 0.28). Although the disease burden was greatest for lymphatic filariasis, only 10 RCTs had been performed.

Table 4 shows that most RCTs were conducted in the WHO Region of the Americas, the South-East Asia Region or the African Region. Only 10 RCTs (3.9%) were multicentre trials. There were 71 (27.5%) cluster randomized trials. Most trials were publicly funded (55.8%) or had no reported funding source (25.6%) and only 10.1% were funded by industry. The median sample size was 151 (interquartile range: 36 to 553).

Overall, 113 of the 258 RCTs (43.8%) involved vaccines, 99 (38.4%) involved topical or oral chemoprophylaxis and 39 (15.1%) involved interventions that targeted insect vectors, such as insecticide-treated bednets or indoor residual spraying. In addition, 80 (31.0%) had one or more preventive chemotherapy arms – either mass drug administration or targeted treatment. Few trials addressed the delivery of preventive chemotherapy: only 10 considered dosing intervals, 4 considered the choice of target population, 2 considered the population coverage needed for mass drug administration and 1 considered how best to deliver preventive chemotherapy. The other preventive chemotherapy trials either compared a drug with placebo or compared two or more drugs. Only 12 trials addressed co-treatment of more than one disease by mass drug administration. Preventive chemotherapy is the main intervention for four diseases – lymphatic filariasis, onchocerciasis, schistosomiasis and geohelminth infections – and is a key component in the control of trachoma.9,13 Although 80 of 112 RCTs on these five diseases had a targeted treatment or mass drug administration arm, again very few addressed how best to deliver preventive chemotherapy: only 10 investigated different time intervals for drug distribution, 2 considered population coverage, 4 considered the target population and 1 considered the method of drug distribution. Moreover, of the 80 RCTs, 27 compared mass treatment and placebo, whereas 24 compared different drugs.

Systematic reviews and meta-analyses

The literature search identified 31 publications that reported one or more systematic reviews, with or without a formal meta-analysis (Fig. 2, available at: http://www.who.int/bulletin/volumes/92/5/13-129601). They addressed 36 different interventions for nine different diseases: American trypanosomiasis, dengue, geohelminths, leishmaniasis, leprosy, lymphatic filariasis, onchocerciasis, schistosomiasis and trachoma. Of the 16 systematic reviews that included a meta-analysis, there were 2 on dengue,14,15 1 on leishmaniasis,16 5 on leprosy (3 on the bacillus Calmette–Guérin [BCG] vaccine1719 and two on chemoprophylaxis20,21), 2 on schistosomiasis,22,23 1 on onchocerciasis,24 4 on lymphatic filariasis2528 and 1 on trachoma.29 Of the 16 systematic reviews that did not include a meta-analysis (reported in 15 publications), there were 4 on trachoma,3033 3 on dengue,3436 1 on leprosy,37 2 on leishmaniasis,38,39 1 on onchocerciasis,40 3 on schistosomiasis,4143 1 on geohelminths43 and 1 on American trypanosomiasis.44 Details of the systematic reviews are given in Appendix B (available at: https://stanford.box.com/s/af4co496bynqxlfwnlgc).

Fig. 2. Literature search for systematic reviews and meta-analyses on the prevention and control of neglected tropical diseases, to 2012
Fig. 2. Literature search for systematic reviews and meta-analyses on the prevention and control of neglected tropical diseases, to 2012

Appendix C (available at: https://stanford.box.com/s/ftoxgslppc4d9qzno3j5) lists the RCTs included in each review. Overall, only 79 of the 258 RCTs (30.6%) were included in a systematic review: 31 (12.0%) were included only in a systematic review without a meta-analysis, 40 (15.5%) were included only in a systematic review with a meta-analysis and 8 (3.1%) were included in both a systematic review without a meta-analysis and one with a meta-analysis.

Nineteen interventions had been assessed by a single systematic review (Table 5). Of the 19, 14 were found to be effective, 3 were found to be ineffective and, for 2, there was insufficient evidence to judge efficacy. Another 17 interventions had been assessed by two or more systematic reviews (Table 6). Of the 17, 8 were consistently found to be effective (all had been assessed in a meta-analysis), 1 was consistently found to be ineffective and, for 8, different reviews produced conflicting conclusions. For 4 of the 8 interventions on which conclusions conflicted, different systematic reviews concluded either that the intervention was likely to be effective or that there was insufficient evidence; for 1 other intervention, different systematic reviews concluded either that the intervention was likely to be ineffective or that there was insufficient evidence; and for the remaining 3 interventions, different reviews reported all possible conclusions (i.e. likely to be effective, likely to be ineffective and insufficient evidence). Only interventions for trachoma, leprosy and schistosomiasis were consistently judged to be effective by more than one systematic review. However, interventions for American trypanosomiasis, geohelminth infection, onchocerciasis and lymphatic filariasis were judged effective by the one systematic review in which each had been assessed.

For Buruli ulcer and rabies, no systematic review of a prevention or control intervention containing a peer-reviewed RCT had been performed, though RCTs had been carried out. For cysticercosis, dracunculiasis, echinococcosis, foodborne trematode infection and human African trypanosomiasis, neither a systematic review nor an RCT had been performed.

Discussion

Our analysis of 258 RCTs and 32 systematic reviews provides a summary of the evidence available on the prevention and control of NTDs and identifies gaps in that evidence. Although prevention is likely to be more cost-effective than treatment for these diseases,45,46 far fewer trials of prevention than treatment have been carried out.47 We found that RCTs on prevention or control had been performed for only 11 of the 16 principal NTDs and that systematic reviews had been performed for only 9. Most RCTs had not been included in a systematic review. We identified 8 interventions that were consistently found to be effective in two or more systematic reviews: topical tetracycline and oral azithromycin chemotherapy for the prevention of trachoma; BCG vaccination, dapsone and acedapsone chemoprophylaxis for the prevention of leprosy; and artesunate, artemether and praziquantel chemoprophylaxis for schistosomiasis (Table 6). For another 14 interventions, a single review concluded that they were effective (Table 5). However, for 8 interventions for which two or more reviews were available, conclusions were conflicting (Table 6).

A range of nonpharmacological interventions for disease control was reported. Future meta-analyses would be made easier by standardizing the design of trials on vector control strategies, including habitat modification, the use of insecticide-impregnated materials and spraying for dengue and leishmaniasis. Furthermore, most vector control trials for leishmaniasis and dengue did not consider human disease as an outcome and there is a need for more research on the relationship between vector control and human disease to justify such interventions. In particular, research on leishmaniasis and dengue is very different from that on trachoma: the recent literature on trachoma reports that collaborative, large-scale studies have been carried out, study designs and protocols have been published and efforts have been made to standardize definitions and outcome measures.13,48,49

There was either limited evidence that trachoma could be prevented by environmental improvements, such as increased access to sanitation, or conflicting conclusions in meta-analyses and systematic reviews. However, since such interventions may have broader benefits for other conditions, such as childhood diarrhoea50,51 and geohelminth infection,52 and may reduce overall childhood mortality,53 it may not be a good use of resources to carry out further studies into their effect on this one disease.

Publications on the five NTDs for which no systematic review or RCT had been performed – cysticercosis, dracunculiasis, echinococcosis, foodborne trematode infections and human African trypanosomiasis – generally focused on treatment.47 These diseases all have a focal form of transmission, which means that controlled trials would require the collaboration of veterinary public health experts, clinical researchers and behavioural health experts. For two diseases – rabies and Buruli ulcer – RCTs had been carried out but no systematic review containing an RCT was found in the literature search. For example, the systematic reviews found on rabies prevention did not include RCTs from peer-reviewed publications.54,55

Limitations

Our study has several limitations. First, we did not consider unpublished findings or the results of trials that were published only as abstracts because such material is difficult to identify systematically and formal peer-reviews have not been carried out. Consequently, we do not know the extent to which publication or selective reporting bias may have influenced the reliability of the evidence available. However, it is unlikely that such biases would completely invalidate the large preventive effects observed. Second, we did not conduct our own systematic review or meta-analysis because this would have been difficult to achieve for the dozens of different interventions used for 16 diseases. However, the reviews and meta-analyses we identified provide a summary of the evidence available and our analysis highlights those interventions that remain controversial and indicates those that need to be evaluated in systematic reviews and meta-analyses and those that need to be tested in RCTs. Third, we avoided adopting a specific conclusion when different systematic reviews and meta-analyses reached different conclusions. Instead, we registered the discrepancy, which may have reflected differences in eligibility criteria, data analysis or interpretation.5658 For example, different eligibility criteria were used in the reviews of chemical vector control for dengue,14,34,35 different eligibility criteria and different methods for calculating summary effect measures were used in the meta-analyses of albendazole for lymphatic filariasis,27,28 and mainly the same evidence was interpreted in different ways in the reviews of eye or face washing and environmental improvements for trachoma.30,32,33

Although trials of individual treatments can help in selecting drugs for use in mass drug administration programmes, such trials are not usually designed to address rare but serious side-effects or drug resistance, both of which are concerns in large-scale programmes. In addition, these trials rarely include nursing or pregnant women, who could also benefit from preventive chemotherapy.5961 Furthermore, while the evidence supporting the use of individual drugs included in mass drug administration programmes may be adequate, there are few reports of prevention or treatment trials involving combinations of drugs for several diseases, which are essential for evaluating integrated preventive chemotherapy.47

In conclusion, our study provides an overview of what is and is not known about the effectiveness of prevention and control measures for the principal NTDs. Where strong evidence is available, it can be used to guide the introduction and scale-up of prevention and control programmes; where gaps in evidence have been identified, the result should be new RCTs on prevention and control measures, although carrying out such trials in endemic areas is challenging. Moreover, these trials may have to be large to detect significant effects in the population and the costs may be prohibitive. However, a considerable amount of money has already been invested in, for example, mass drug administration programmes. Such programmes could only become more acceptable and financially viable if the most effective and cost-effective ways of delivering preventive chemotherapy were identified.


Funding:

Shanthi Kappagoda was supported in part by grant HS000028 from the Agency for Healthcare Research and Quality. The funding body did not influence the design or implementation of the study or the decision to publish and the authors alone are responsible for the content of this paper.

Competing interests:

None declared.

References

Share