Introduction
Schistosomiasis is a common parasitic disease in the tropics and subtropics. An estimated 779 million individuals are at risk of acquiring schistosomiasis and more than 200 millions were infected in mid-2003 (Steinmann et al. Reference Steinmann, Keiser, Bos, Tanner and Utzinger2006). The World Health Organization (WHO) estimates that the global burden due to schistosomiasis may be as high as 4·5 million disability-adjusted life years (DALYs) (WHO, 2002). However, a meta-analysis suggests that disability weights might be 2–15 times higher than those used in the global burden of disease study (King, Dickman and Tisch, Reference King, Dickman and Tisch2005), and that the DALY underestimates the importance of chronic diseases like schistosomiasis (King and Bertino, Reference King and Bertino2008). This is further substantiated by the results of approaches using a quality of life questionnaire and decision-tree modelling (Jia et al. Reference Jia, Zhou, Wang, Utzinger, Steinmann and Wu2007; Finkelstein et al. Reference Finkelstein, Schleinitz, Carabin and McGarvey2008). From a public health perspective, the three most important schistosome species are Schistosoma haematobium (causing urinary schistosomiasis), and S. mansoni and S. japonicum (causing intestinal schistosomiasis).
Two drugs, metrifonate and praziquantel (PZQ), have been used extensively for urinary schistosomiasis (Cioli, Pica-Mattoccia and Archer, Reference Cioli, Pica-Mattoccia and Archer1995; Utzinger and Keiser, Reference Utzinger and Keiser2004). However, in the late 1990s, metrifonate was withdrawn from the WHO model list of essential medicines because it was considered clinically, economically and operationally inferior to PZQ as it is only active against S. haematobium, requires multiple administrations, and hence is less convenient in large-scale control programmes (Feldmeier and Chitsulo, Reference Feldmeier and Chitsulo1999). Thus, PZQ remains the only drug for clinical management and community-based control of schistosomiasis (Cioli, Reference Cioli2000; Fenwick et al. Reference Fenwick, Savioli, Engels, Bergquist and Todd2003; Utzinger and Keiser, Reference Utzinger and Keiser2004; Caffrey, Reference Caffrey2007; Doenhoff, Cioli and Utzinger, Reference Doenhoff, Cioli and Utzinger2008). Large-scale morbidity control programmes became feasible as the price of PZQ fell from approximately US$ 1.0 in the 1980s to less than US$ 0.1 per 600 mg tablet (Fenwick et al. Reference Fenwick, Savioli, Engels, Bergquist and Todd2003; Fenwick, Keiser and Utzinger, Reference Fenwick, Keiser and Utzinger2006; Doenhoff et al. Reference Doenhoff, Cioli and Utzinger2008). Paradoxically, this also stalled investments in the discovery and development of alternative control measures, such as other drugs, vaccines and diagnostics (Utzinger et al. Reference Utzinger, Xiao, Tanner and Keiser2007; Bergquist, Utzinger and McManus, Reference Bergquist, Utzinger and McManus2008). Research carried out over the past 15 years revealed the antischistosomal properties of artemisinin derivatives, which act especially on the young developing stages of the parasites (for a recent review see Utzinger et al. Reference Utzinger, Xiao, Tanner and Keiser2007) as opposed to PZQ, which acts against the adult worms and the very young schistosomula just after skin penetration (Sabah et al. Reference Sabah, Fletcher, Webbe and Doenhoff1986; Utzinger et al. Reference Utzinger, Xiao, Tanner and Keiser2007).
The use of PZQ has increased considerably after the 54th World Health Assembly set the target of at least 75% of school-aged children and other high-risk groups to be treated regularly with PZQ by 2010 in areas where the disease is highly endemic (WHO, 2002). At least 17 million doses of PZQ have been administered, primarily to school-aged children in selected African countries, since the launch of the ‘Schistosomiasis Control Initiative’ in 2003 (Fenwick et al. Reference Fenwick, Keiser and Utzinger2006). Relying on praziquantel alone for controlling a disease that affects millions of people is risky should resistance develop against this drug (Danso-Appiah and de Vlas, Reference Danso-Appiah and de Vlas2002; Doenhoff et al. Reference Doenhoff, Cioli and Utzinger2008).
In the light of this, we assessed the effects of PZQ and other antischistosomal treatments by conducting a Cochrane systematic review, including comparisons between PZQ and metrifonate and trials of combination treatments. During this process, we identified a number of methodological issues relevant to the interpretation of existing data that might help researchers to design more appropriate trials in the future. The full review is available on the Cochrane Library (Danso-Appiah et al. Reference Danso-Appiah, Utzinger, Liu and Olliaro2008). In this paper we highlight key findings of the review, discuss implications of various methodological limitations and consider future research needs.
SUMMARY OF COCHRANE SYSTEMATIC REVIEW
Inclusion criteria and search strategy
To qualify for inclusion, a study was (1) to be controlled, randomized or quasi-randomized, enrolling individuals infected with S. haematobium, as determined microscopically for S. haematobium eggs in a standard filtrate of 10 ml of urine or by haematuria (Feldmeier and Poggensee, Reference Feldmeier and Poggensee1993); and (2) to treat patients with either PZQ, metrifonate or artemisinin derivatives. An extensive, standard search was carried out, which included MEDLINE (1966 to August 2007), EMBASE (1974 to August 2007), LILACS (1982 to August 2007), conference proceedings and contacting specialists in the field (Danso-Appiah et al. Reference Danso-Appiah, Utzinger, Liu and Olliaro2008).
Data retrieval, quality assessment and analysis
Eligibility and methodological quality of the identified trials were assessed by the authors and the data analysed using Review Manager 4.2 (The Cochrane Collaboration, 2003). The main outcome measure was failure rate (the proportion of individuals remaining positive for eggs in their urine at follow-up). Comparisons between groups were expressed as relative risk (RR) with 95% confidence intervals (CIs) for both individual studies and on aggregate. The secondary parameter was egg reduction rate and was analysed using weighted mean difference with standard error. The proportion of patients with adverse events was compared between treatment and control arms.
Key findings of the Cochrane systematic review
The search identified 24 randomised controlled trials that together involved 6315 participants. Table 1 summarises key characteristics of these 24 trials. When used as monotherapy, both metrifonate and PZQ showed obvious benefit in terms of parasitological outcomes (Danso-Appiah et al. Reference Danso-Appiah, Utzinger, Liu and Olliaro2008). One trial (120 participants) of artesunate showed no obvious benefit over placebo.
Summary of the characteristics of randomised controlled trials included in our Cochrane systematic review (Danso-Appiah et al. Reference Danso-Appiah, Utzinger, Liu and Olliaro2008) evaluating antischistosomal drugs, used alone or in combination, for treating urinary schistosomiasis
* N – Number of communities involved in the trial.
** For urine filtration, 10 ml; 3 specimen vs. 10 ml; 1 specimen means pre- and post-treatment diagnosis varied; i.e. pre-treatment diagnosis involved 10 ml and 3 specimen but 10 ml and 1 specimen for post-treatment assessment.
For combination treatments, one trial studied the combination of PZQ with artesunate, but there was no obvious advantage over PZQ alone.
Metrifonate versus PZQ: comparisons and dose effects
Fig. 1 summarises the failure rate of metrifonate versus PZQ in the five trials meeting our inclusion criteria (McMahon, Reference McMahon1983; Pugh and Teesdale, Reference Pugh and Teesdale1983; Wilkins and Moore, Reference Wilkins and Moore1987; King et al. Reference King, Lombardi, Lombardi, Greenblatt, Hodder, Kinyanjui, Ouma, Odiambo, Bryan, Muruka, Magak, Weinert, Mackay, Ransohoff, Houser, Koech, Siongok and Mahmoud1988; Stephenson et al. Reference Stephenson, Kinoti, Latham, Kurz and Kyobe1989).
Risk ratio estimates (combined or total) of randomized controlled trial(s) of metrifonate (different regimens) versus PZQ (single dose, 30 mg/kg or 40 mg/kg) against S. haematobium. Rectangles represent risk ratios and sizes of the rectangles denote the weight given to each trial in the meta-analysis. Diamond and vertical broken line indicate combined (total) relative risk (RR). Horizontal lines indicate 95% confidence intervals. The dashed vertical line is the null value (RR=1; neither favouring metrifonate nor PZQ). Abbreviations: mo, month; we, week.
When metrifonate was introduced, some early studies investigated a single dose of 10 mg/kg (the standard dose is 7·5 to 10 mg/kg three times at 14-day intervals) versus the standard single dose of 40 mg/kg PZQ. Although the single metrifonate dose was inferior in three trials measuring failure at one to eight months, the 95% CIs were too wide for statistical significance (RR=2·31, 95% CI: 0·91–5·82; n=462 participants). This lack of significance is due to significant heterogeneity between trials (I 2=94%) likely to be associated with the duration of follow-up: the RR was 1·26 at one month of follow-up (Pugh and Teesdale, Reference Pugh and Teesdale1983), 2·23 at three months (Wilkins and Moore, Reference Wilkins and Moore1987) and 4·62 at eight months (Stephenson et al. Reference Stephenson, Kinoti, Latham, Kurz and Kyobe1989).
There was no significant difference in failure rates when metrifonate given as multiple doses (3×10 mg/kg fortnightly) was compared with PZQ (30 mg/kg) in a small trial involving 54 participants (McMahon, Reference McMahon1983). A trial comparing three doses of metrifonate at 10 mg/kg given once every four months with the standard 40 mg/kg PZQ (single dose) in school-aged children in Kenya detected no difference overall, but metrifonate was superior in the subgroup of children with a heavy infection (RR=0·88, 95% CI: 0·80–0·96; n=615 participants). However, as the subgroup was stratified after randomisation, this result should be interpreted with caution. Both metrifonate (two and three doses of 10 mg/kg) and PZQ (single dose 40 mg/kg) led to very high reductions in egg excretion (>98%) in two trials (McMahon, Reference McMahon1983; Doehring et al. Reference Doehring, Ehrich, Vester, Feldmeier, Poggensee and Brodehl1985).
One trial (n=54 participants) compared adverse events and reported similar minor events between metrifonate (3×10 mg/kg) and PZQ (30 mg/kg); no serious adverse events were noted (McMahon, Reference McMahon1983). Mild and transient abdominal pain was more common with triplicate metrifonate than single dose PZQ (75% versus 30%), but the dose of PZQ used (30 mg/kg) was lower than the one currently recommended (40 mg/kg) (WHO, 2002).
Metrifonate: dose comparisons
One trial with 201 participants compared 3×5 mg/kg metrifonate administered in a single day to 3×7·5 mg/kg given fortnightly (Aden Abdi and Gustafsson, Reference Aden Abdi and Gustafsson1989). There was no significant difference in parasitological failure and egg reduction rate; the geometric mean egg reduction rate was 96% for the one-day regimen and 97% for the fortnightly regimen. There was little difference in the percentage of patients with mild adverse events reported for the fortnightly regimen (7%) versus the one-day regimen (9%).
Three doses of metrifonate (10 mg/kg) were compared with one and two doses (Rey, Nouhou and Sellin, Reference Rey, Nouhou and Sellin1984). There was no significant difference in parasitological failure rates between two and three doses at one month and four months follow-up. Similarly, no significant differences in failure rate and egg reduction rates were detected between two and three doses of 10 mg/kg, given fortnightly in a trial of 81 participants (Rey et al. Reference Rey, Nouhou and Sellin1984). By contrast, there were fewer parasitological failures with the three-dose regimen over the one dose regimen at the one month follow-up (RR=2·75, 95% CI: 1·29–5·85; n=93 participants) and the four-month follow-up (RR=1·52, 95% CI: 1·03–2·25; n=111 participants).
Different PZQ doses versus standard regimen (1×40 mg/kg)
Ten trials compared the standard regimen of PZQ (single dose of 40 mg/kg) to various other doses (McMahon and Kolstrup, Reference McMahon and Kolstrup1979; Davis et al. Reference Davis, Biles, Ulrich and Dixon1981; Omer, Reference Omer1981; Oyediran et al. Reference Oyediran, Kofie, Bammeke and Bamgboye1981; Rey et al. Reference Rey, Sellin, Gazere, Ott, Reges and Garrouty1983; Kardaman et al. Reference Kardaman, Fenwick, el Igail, el Tayeb, Daffalla and Dixon1985; Wilkins and Moore, Reference Wilkins and Moore1987; Taylor, Murare and Manomano, Reference Taylor, Murare and Manomano1988; King et al. Reference King, Wiper, De Stigter, Peters, Koech, Ouma, Arap Siongok and Mahmoud1989, Reference King, Muchiri, Mungai, Ouma, Kadzo, Magak and Koech2002). In terms of parasitological failure, there was no significant difference between the standard regimen and 2×20 mg/kg (4 trials), a single dose of 30 mg/kg (6 trials), and a single dose of 20 mg/kg (2 trials). Similar results were found at one, three and six months follow-up. Losses to follow-up were generally high, but these did not differ across treatment and control groups within a single trial. There was no significant heterogeneity between the trials, and background endemicity did not seem to play a role. Examining for a differential effect between heavy and moderate or light infections with 30 mg/kg versus standard 40 mg/kg, a subgroup analysis of one small trial (King et al. Reference King, Wiper, De Stigter, Peters, Koech, Ouma, Arap Siongok and Mahmoud1989) did not show any difference (n=116 participants). Five trials showed no apparent differences in egg reduction rate (geometric mean); all had greater than 95% reduction in both arms.
Artesunate
Thus far, only one randomised controlled trial conducted in Gabon in schoolchildren compared the effects of artesunate combined with PZQ to each individual drug given as monotherapy (Borrmann et al. Reference Borrmann, Szlezák, Faucher, Matsiegui, Neubauer, Binder, Lell and Kremsner2001). Whilst the artesunate-PZQ combination resulted in a relatively higher egg reduction rate, it was not possible to identify an effect of artesunate, as no significant difference was observed in cure rates when compared to PZQ alone.
METHODOLOGICAL LIMITATIONS
Lack of standardization and quality data for the assessment of efficacy and safety of antischistosomal drugs was reported previously for S. mansoni (Danso-Appiah and de Vlas, Reference Danso-Appiah and de Vlas2002). In this Cochrane systematic review (Danso-Appiah et al. Reference Danso-Appiah, Utzinger, Liu and Olliaro2008), we identified a number of methodological limitations that raise issues with trial quality and the potential for bias, outlined below. Some of the shortcomings have implications for the interpretation of trials in schistosomiasis and other tropical diseases (responses to methodological limitations summarised in Box 1).
Box 1. Responses to methodological limitations of trials included in a Cochrane review of drugs for treating urinary schistosomiasis (Danso-Appiah et al. Reference Danso-Appiah, Utzinger, Liu and Olliaro2008)
Design issues1. There is the need for a unified study methodology in the design, collection and reporting of trials.
2. Trialists should be sensitized to the importance of proper sample size calculation to ensure that trials are sufficiently powered. High losses to follow-up in trials with small sample sizes further compromise the statistical power.
3. There is the need to describe the randomisation procedure clearly.
4. There is a need for standardized, quality-controlled diagnostic criteria within and between trials.
5. In high endemicity areas a follow-up time of 4 to 8 weeks is appropriate when investigating cure rates to avoid eggs released from dead worms and minimise the effect of re-infection.
Interpretation and reporting
6. Intensity of infection and egg reduction rate (ERR) should be reported in geometric mean, and intensity of infection should be based on egg count of only the positive cases and reported using the standard classification by the WHO (WHO, 2002).
7. Treatment outcomes need to be clearly defined and standardised across trials. Parasitological outcomes should be reported with; (a) level of edemicity, (b) diagnostic criteria, (c) dose used, (d) age of participants and (e) follow-up time.
1. There is the need for a unified study methodology in the design, collection and reporting of trials.
2. Trialists should be sensitized to the importance of proper sample size calculation to ensure that trials are sufficiently powered. High losses to follow-up in trials with small sample sizes further compromise the statistical power.
3. There is the need to describe the randomisation procedure clearly.
4. There is a need for standardized, quality-controlled diagnostic criteria within and between trials.
5. In high endemicity areas a follow-up time of 4 to 8 weeks is appropriate when investigating cure rates to avoid eggs released from dead worms and minimise the effect of re-infection.
Interpretation and reporting
6. Intensity of infection and egg reduction rate (ERR) should be reported in geometric mean, and intensity of infection should be based on egg count of only the positive cases and reported using the standard classification by the WHO (WHO, 2002).
7. Treatment outcomes need to be clearly defined and standardised across trials. Parasitological outcomes should be reported with; (a) level of edemicity, (b) diagnostic criteria, (c) dose used, (d) age of participants and (e) follow-up time.
(1) Some trials had no proper sample size calculation: this suggests the authors may not have considered whether their study was sufficiently powered to answer the question being posed. (2) Randomisation quality was not high: only four out of 24 trials (17%) met quality standards for adequate concealment of allocation and described the methods used (for quality standards see Jüni, Altman and Egger (Reference Jüni, Altman and Egger2001) and Higgins and Green (Reference Higgins and Green2008). Trials conducted in the early 1990s and earlier did not conceal allocation except one (Davis et al. Reference Davis, Biles, Ulrich and Dixon1981). Generation of allocation sequence was adequate in less than half of the trials included in our meta-analysis. For the others, the method used was unclear although all were reported as randomized controlled trials. (3) Losses to follow-up were often high in some trials, and increased proportionally with the duration of follow-up: 17 trials registered losses of <10% for short-term evaluations at one to three months, but losses reached up to 50% in some trials when follow-up time was longer than three months. (4) Diagnostic criteria were varied, vague and not standardised: among the trials included in our meta-analysis, the criteria for diagnosis varied greatly; some trials used three urine specimens on three consecutive days for microscopic examination, whilst others used a single specimen (Table 1). In some trials sampling criteria varied even between pre- and post-treatment using microscopy (e.g. three urine specimens for the pre-treatment diagnosis but only one for post-treatment follow-up assessment) while other trials lacked any criteria for diagnosis. (5) Classification of infection intensities lacked standardization: Table 2 shows considerable variation in the classification of infection intensity across trials. According to current WHO guidelines, infection intensity of S. haematobium is either light (1–49 eggs/10 ml of urine) or heavy (⩾50 eggs/10 ml of urine) (WHO, 2002). In the trials included in our meta-analysis, however, light infections were variably classified as 1–5, 1–29, 1–99, 60–249 or even 250–500 eggs/10 ml of urine. Accordingly, moderate and heavy infections varied from one trial to another. (6) Outcomes were reported in a variety of ways: in our review we defined primary outcomes as (i) parasitological failure and (ii) egg reduction rate. However, these two measures were variably reported as cure rate, failure rate, cumulative failure rate or prevalence for parasitological failure, and as a median, arithmetic mean or geometric mean for egg reduction rate. Even the calculation of geometric mean varied; some investigators considered only the egg-positive individuals, whilst others included the negatives and introduced a correction factor of plus 1. The latter becomes problematic after treatment when most of the remaining infections are light, as it may overestimate egg count values. (7) Timing of post-treatment assessments varied greatly: the majority of trials evaluated cure and egg reduction rate within one to three months; however, some trials did so at three weeks or earlier, or only six or even 12 months post-treatment. Results from studies on urinary schistosomiasis assessing outcomes earlier than three weeks or beyond three months post-treatment should be considered with caution. The reasons are that the development of S. haematobium worms takes approximately two months (Ghandour, Reference Ghandour1978); shorter follow-up is confounded by eggs of killed worms still being excreted, longer follow-up by re-infections, particularly in highly endemic settings (N'Goran et al. Reference N'Goran, Utzinger, N'Guessan, Müller, Zamble, Lohourignon, Traoré, Sosthene, Lengeler and Tanner2001; Tchuem Tchuenté et al. Reference Tchuem Tchuenté, Shaw, Polla, Cioli and Vercruysse2004; Satayathum et al. Reference Satayathum, Muchiri, Ouma, Whalen and King2006). Noteworthy, most of the trials in this review were conducted in high endemicity areas and there was no way in differentiating between re-infection and recrudescence.
Classification of different S. haematobium infection intensities in clinical trials included in a Cochrane systematic review (Danso-Appiah et al. Reference Danso-Appiah, Utzinger, Liu and Olliaro2008), n.c. not classified
Discussion
Public health implications
Despite the above-mentioned methodological limitations, the findings of our Cochrane systematic review have important public health implications. One of the most important findings is that both PZQ and metrifonate are efficacious and safe (Danso-Appiah et al. Reference Danso-Appiah, Utzinger, Liu and Olliaro2008). The failure rate with the recommended standard dose of PZQ (40 mg/kg) is 0–37%, whilst that of metrifonate (3×7·5–10 mg/kg given fortnightly) is 19–48% at one to three months follow-up. However, no trial included in our analysis directly compared the above-mentioned standard doses, therefore precluding any head-to-head assessment of the two treatments from currently available data.
Although the effects of PZQ against placebo are obvious, for some comparisons between regimens with both PZQ and metrifonate there was uncertainty around their effect estimates as shown by the wide 95% CIs. The small sample size in some of the trials may explain the levels of uncertainty. However, the magnitude of the effect is at times so dramatic that it is unlikely that methodological quality alone will have caused substantive biases to interfere with the marked effects and differences reported.
No difference was demonstrated with a single dose of 20 or 30 mg/kg of PZQ compared to the standard regimen (single oral dose of 40 mg/kg) in terms of all outcomes measured in this review. Given the current emphasis on controlling morbidity in high-burden areas (WHO, 2002), and morbidity, especially among school-aged children, being associated with the number of eggs in an individual, this finding suggests lower doses of PZQ may be effective in morbidity control. However, these results should be considered with caution as detection error can play a role, especially when the studies are few and sample sizes are small. While it is true that parasite load (with egg counts often used as a proxy measure) is an important factor in both morbidity for the individual patient and environmental contamination (WHO, 2002), a sub-curative dose may unduly put the drug under selective pressure and favour parasite resistance (Doenhoff, Reference Doenhoff1998; Doenhoff et al. Reference Doenhoff, Cioli and Utzinger2008). Pharmacokinetic data of different doses of PZQ are few and old, and have been obtained in healthy volunteers rather than in schistosome-infected patients (Leopold et al. Reference Leopold, Ungethum, Groll, Diekmann, Nowak and Wegner1978). An exponential increase was found in the area under the curve (AUC) with the PZQ dose in the range of 5 to 50 mg/kg, with a six-fold increase from 20 to 50 mg/kg (Leopold et al. Reference Leopold, Ungethum, Groll, Diekmann, Nowak and Wegner1978). However, these data do not come from infected patients, and hence cannot be extrapolated so easily. This calls for well designed trials incorporating also pharmacokinetic assessment, possibly with sparse sampling and population kinetic assessment. These trials should also control for food intake, as the bioavailability of PZQ depends upon taking it with food and the type of food matters (Mandour et al. Reference Mandour, el Turabi, Homeida, el Sadig, Ali, Bennett, Leahey and Harron1990; Castro et al. Reference Castro, Medina, Sotelo and Jung2000).
The rationale behind the widely spaced dosing intervals of metrifonate treatment derives from its long-lasting effect on red blood cells and plasma cholinesterases (Plestina, Davis and Bailey, Reference Plestina, Davis and Bailey1972). However, the clinical significance of this effect and why side effects disappear during the first 12–24 hours whereas the recovery of the enzymes takes more than 4–6 weeks is not known (Plestina et al. Reference Plestina, Davis and Bailey1972). Safety studies have shown no serious adverse events in patients treated with 5–10 mg/kg metrifonate daily for 6–12 days (Snellen, Reference Snellen1981), and various reviews of the toxicology and pharmacology of metrifonate during its extensive use for urinary schistosomiasis in the 1970s concluded that it had very few adverse events (Holmstedt et al. Reference Holmstedt, Nordgren, Sandoz and Sundwall1978; Feldmeier and Doehring, Reference Feldmeier and Doehring1987; Cioli et al. Reference Cioli, Pica-Mattoccia and Archer1995). Also, metrifonate is currently used in Alzheimer's disease in extended regimens, and a systematic review has concluded that overall tolerability is good with only mild to moderate adverse events (López-Arrieta and Schneider, Reference López-Arrieta and Schneider2006). In the current review, although drug safety was generally poorly reported and assessed in few trials, no trial recorded a serious adverse event, and no significant differences in the number and type of adverse events between metrifonate and PZQ were recorded, except for abdominal pain that was more frequent after metrifonate. Optimizing metrifonate treatment may provide a means of easing drug pressure exerted on schistosomes by the wide deployment of PZQ.
Immature schistosomes are less sensitive to PZQ than adult worms (Sabah et al. Reference Sabah, Fletcher, Webbe and Doenhoff1986), which has raised concern about controlling schistosomiasis effectively with this drug. Artemisinin derivatives proved to be effective against immature schistosomes in laboratory studies (Utzinger et al. Reference Utzinger, Keiser, Xiao, Tanner and Singer2003, Reference Utzinger, Xiao, Tanner and Keiser2007). However, this review shows that artesunate was not effective against S. haematobium infections (though evidence was derived from a single trial (Borrmann et al. Reference Borrmann, Szlezák, Faucher, Matsiegui, Neubauer, Binder, Lell and Kremsner2001)), and combining artesunate and 40 mg/kg PZQ did not improve efficacy over PZQ alone. In two non-randomised trials involving artesunate alone, results were relatively better (De Clercq et al. Reference De Clercq, Vercruysse, Kongs, Verlé, Dompnier and Faye2002; Inyang-Etoh et al. Reference Inyang-Etoh, Ejezie, Useh and Inyang-Etoh2004). The latter findings were confirmed in a recent trial; artesunate alone (4 mg/kg) resulted in a cure rate of 70·5%, whereas an artesunate-PZQ combination obtained a cure rate of 88·6% (Inyang-Etoh et al. Reference Inyang-Etoh, Ejezie, Useh and Inyang-Etoh2009). Finally, a recent trial in children under six years of age who were co-infected with Plasmodium falciparum and S. haematobium and who were treated with two different artemisinin-based combinations for malaria therapy showed good effects on S. haematobium. This trial, however, could not be included in the current analysis because there was no control group (Boulanger et al. Reference Boulanger, Dieng, Cisse, Remoue, Capuano, Dieme, Ndiaye, Sokhna, Trape, Greenwood and Simondon2007).
The need for good trial methods
The validity of randomized controlled trials rests in part on adequate allocation concealment and minimal losses to follow-up, and weaknesses in both these aspects were found in the trials included in the current meta-analysis (Table 1). Without adequate allocation concealment properly developed random allocation sequences can be subverted (Schulz and Grimes, Reference Schulz and Grimes2002). A likely explanation for only four trials (17%) included in our final analyses adequately concealing allocation is that this had not been identified as a particularly relevant issue at the time the trials were conducted (20–30 years ago). Even after the publication of the CONSORT statement (Begg et al. Reference Begg, Cho, Eastwood, Horton, Moher, Olkin, Pitkin, Rennie, Schulz, Simel and Stroup1996) and despite continued educational efforts, the quality of reporting of randomized controlled trials still needs improvement (Altman et al. Reference Altman, Schulz, Moher, Egger, Davidoff, Elbourne, Gotzsche and Lang2001; Moher, Schulz and Altman, Reference Moher, Schulz and Altman2001).
The effect of losing patients during follow-up on randomisation is crucial as this relates to the internal validity and the power of the trial. In our systematic review we could not do a sensitivity analysis to evaluate the effect of loss to follow-up because data were not sufficient. We encourage trialists to take particular note of this issue and ensure that losses are minimised and power is preserved in future trials. Also, we welcome debate on the most appropriate timing of follow-up in evaluating drug trials of both urinary and intestinal schistosomiasis. The biology of schistosomes suggests that treatment effects with antischistosomal drugs on parasitological parameters should be evaluated during a window of four to eight weeks post-treatment to avoid detecting the tail of eggs released from dead worms on one side, and re-infections on the other side. A first attempt has been made to evaluate this for intestinal schistosomiasis due to S. mansoni, and the authors concluded that three weeks after PZQ administration is an appropriate timing for drug efficacy evaluation (Scherrer et al. Reference Scherrer, Sjöberg, Allangba, Traoré, Lohourignon, Tschannen, N'Goran and Utzinger2009).
Safety is generally overlooked and when data are available they are poorly reported. It is important that trialists realise the importance of adequately documenting and reporting on tolerability.
Diagnostic concerns
The quality of diagnosis can influence the observed cure rates as clearly shown for both S. mansoni (de Vlas and Gryseels, Reference de Vlas and Gryseels1992; Booth et al. Reference Booth, Vounatsou, N'Goran, Tanner and Utzinger2003) and S. haematobium (N'Goran et al. Reference N'Goran, Utzinger, Gnaka, Yapi, N'Guessan, Kigbafori, Lengeler, Chollet, Xiao and Tanner2003). Sensitivity will affect in particular the detection of light infections during follow-up. We found considerable variation in diagnostic criteria not only between, but also within trials, also with regard to infection intensity. This may be explained by the fact that the WHO classification as light (1–49 eggs/10 ml urine) and heavy infection intensity (⩾50 eggs/10 ml urine) was endorsed only recently (WHO, 2002) and was not in use when the trials summarized here. Because of the different thresholds used for infection intensity, it was not possible to combine and analyse the data according to heavy infections, which is relevant to morbidity control.
Study population issues
The age of participants enrolled in randomized controlled trials may also influence results. Here, 22 trials out of 24 recruited school-aged children. Hence the overall effect estimates as reported in this review may be lower than studies including all-age subjects, as adults usually show lower infection intensities than school-aged children and, conversely, higher treatment efficacies. This issue has been documented for S. mansoni (Raso et al. Reference Raso, N'Goran, Toty, Luginbühl, Adjoua, Tian-Bi, Bogoch, Vounatsou, Tanner and Utzinger2004), and it is conceivable that the same holds for S. haematobium. It should also be noted that restricting treatment to school-aged children leaves untreated adults and pre-school children still excreting eggs to maintain transmission, if indeed transmission is a function of egg output. This brings us to two important sets of considerations. First, data should be reported separately for children and adults before, if necessary, pooling the data to assess overall effects. Second, the purpose of studies depends on the target population. Studies in children receiving antischistosomal treatment are more apt to assess the ‘true’ efficacy of the drug because drugs such as PZQ have an immune response-dependent component (Doenhoff et al. Reference Doenhoff, Sabah, Fletcher, Webbe and Bain1987), which is more active in adults, while whole-population studies are more suited to assess the programmatic effectiveness and effects of control interventions.
Implications for policy
Both PZQ and metrifonate are effective and safe for treating urinary schistosomiasis. Our systematic review and that of López-Arrieta and Schneider (Reference López-Arrieta and Schneider2006) indicate that metrifonate is well tolerated. Although in schistosomiasis control metrifonate has operational drawbacks, notably multiple administrations, which make it less convenient for large-scale morbidity control programmes than a single dose of PZQ, the two drugs have similar efficacy profiles. Furthermore, considering that schistosomes are under intense and growing drug pressure by PZQ and the inherent vulnerability of schistosomiasis control to parasite resistance, we suggest metrifonate should be reconsidered for the treatment of urinary schistosomiasis, to ease the drug pressure on PZQ. This implies continued availability (production and distribution) of the drug. It is also important to have an alternative drug for treating urinary schistosomiasis should PZQ resistance emerge.
Most of the trials contributing to this review were conducted more than a decade ago, and entail a series of methodological limitations. The new schistosomiasis trials must be conducted to contemporary standards of clinical research paying particular attention to quality issues we have raised, and adopt commonly agreed criteria.
Our findings point to new approaches worth being explored in well-designed trials such as: (1) reassessing appropriate dosing schedules for metrifonate, including compliance and feasibility in control programmes; (2) comparing standard metrifonate (3×7·5–10 mg/kg given fortnightly) and PZQ (1×40 mg/kg) doses; (3) evaluating artemisinin-based regimens and combination treatments where appropriate (areas where malaria and schistosomiasis are not co-endemic); and (4) obtaining pharmacokinetic/pharmacodynamic correlates for PZQ.
ACKNOWLEDGEMENTS
We thank Dr. Sake J. de Vlas for providing advice on technical issues. We also thank the Cochrane Infectious Diseases Group (CIDG) in Liverpool for providing administrative support and two anonymous referees for a couple of excellent points. This study received financial support from the Liverpool School of Tropical Medicine, North West Regional Health Authority (UK), Department for International Development (UK), European Commission (Directorate General XII; Belgium). JU is grateful to the Swiss National Science Foundation for sustained financial support through project no. PPOOB-102883 and PPOOB-119129.
DISCLAIMER
The opinions expressed in this paper are those of the authors and may not reflect those of their employing organizations. PLO is a staff member of the WHO; the authors alone are responsible for the views expressed in this publication and they do not necessarily represent the decisions, policy or views of the WHO.
