Efficacy and safety of praziquantel 40 mg/kg in preschool-aged and school-aged children: a meta-analysis

Background Children carry most of the schistosomiasis burden. While school-aged children are the principal target group of preventive chemotherapy with praziquantel, limited information on efficacy and safety exists for preschool-aged children. Methods Here, we conducted a meta-analysis of clinical trials of praziquantel for treating children with any form of schistosomiasis. Efficacy was reported as cure rate (CR) and egg reduction rates (ERR); statistical corrections were applied based on methodological disparities across trials to derive the predicted geometrical mean ERR (pERRgm). Safety was reported as frequencies of adverse events. Results Forty-seven comparative and non-comparative studies were identified, enrolling 15,549 children of whom 14,340 (92%) were assessed between 3 and 8 weeks post-treatment with praziquantel 40 mg/kg (the WHO-recommended treatment, n = 8,380, 56%) or comparators (n = 5,960, 44%). The median age was 10 years (range 1–19), 11% (n = 1,694) were preschool-aged. The CR and pERRgm with praziquantel 40 mg/kg were respectively: S. haematobium, 73.6% (95% CI: 63.5–81.40, 25 study arms) and 94.7% (95% CI: 92.7–96.4); S. mansoni, 76.4% (95% CI: 71.5–81.0, 34 arms) and 95.3% (95% CI: 94.2–96.2); S. mansoni/S. haematobium, 67.6% (95% CI: 54.1–80.7, 5 arms) and 93.4% (95% CI: 89.9–96.2); S. japonicum, 94.7% (95% CI: 92.2–98.0) and 98.7% (95% CI: 98.3–99.2). Mixed-effect multivariate analysis found no significant difference between preschool- and school-aged children for CR or pERRgm in S. haematobium (P = 0.309 and P = 0.490, respectively) or S. mansoni (P = 0.982 and P = 0.895) after controlling for time of assessment, formulation, intensity of infection and detection method. Praziquantel was reportedly safe at all ages, with only mild reported adverse events which cleared rapidly after treatment. Conclusions Praziquantel 40 mg/kg was effective at reducing infection intensity in all Schistosoma species without differences between preschool- and school-aged children. However, conclusions should be tempered because of the limited number of preschool-aged children enrolled, disparities in study procedures and limited information made available in publications, as well as the current imperfect test-of-cure. Also, although reportedly well-tolerated, safety was inconsistently assessed. Studies in target groups, individual-data meta-analysis and standardised methodologies are needed for more robust evidence-base. Electronic supplementary material The online version of this article (doi:10.1186/s13071-016-1958-7) contains supplementary material, which is available to authorized users.


Background
Children carry most of the schistosomiasis burden [1]. Schistosomiasis treatment and control relies largely upon preventive chemotherapy (PC) with praziquantel (PZQ) directed primarily at school-aged children living in schistosomiasis-endemic areas [2]. Preschool-aged children have traditionally been excluded from PC and treated on confirmation of infection [3], but are recognized as a vulnerable group [4][5][6]. Emphasis is now shifting from control to elimination of schistosomiasis [7], meaning that all the infected groups should be treated, including preschool-aged children (1-4 years-old), provided a suitable formulation is available. PZQ is registered for use in children from four years of age, and can be given routinely in PC programmes to children measuring 94 cm or more (recently extended to 60 cm [8]) but the absence of paediatric formulations is a practical limitation to having younger children treated routinely.
Praziquantel has been in use now for over three decades, and has been given to millions of people, mostly children -close to 40 million school-aged children in 2013 [9]. However, several unresolved questions still exist about this drug (reviewed in Stothard et al. [10]), including its pharmacokinetic and pharmacodynamic properties [11].
The World Health Organization (WHO) recommends using PZQ at 40 mg/kg, a recommendation generally supported by systematic reviews [12][13][14] and to express treatment outcome in terms of egg reduction rate (ERR) calculated as the difference of the arithmetic mean egg counts (preferred over geometric means) between preand post-treatment samples [3]. It is worth adding, as a note of caution, that we only have imprecise methods to detect Schistosoma infection that are based on the detection of eggs in excreta (faeces or urine) as opposed to adult worms, and that these methods lack sensitivity (for instance, the widely-used Kato-Katz method cannot detect less than 24 eggs per gram of stools). Therefore, the (temporary) reduction or cessation of egg-excretion should not be interpreted as being equivalent to killing of adult worms. This has profound implications both for the estimation of schistosomiasis cases and for the assessment of drug efficacy [15].
This meta-analysis was conducted to assess the efficacy and safety of PZQ in preschool-and school-aged children and whether these outcomes differ with subject's age. This question is particularly relevant in view of the inclusion of preschool-aged children as a target of schistosomiasis control through PC, and whether any dose adjustment should be required.

Methods
This is a meta-analysis of aggregated data; eligible studies were comparative and non-comparative clinical trials where PZQ had been given at any dose to preschool-and school-aged children and adolescents) for treating intestinal or urinary schistosomiasis, and where outcome was assessed within 2 months (3 to 8 weeks) post-treatment. The study was conducted according to the PRISMA guideline (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) [16]. The PRISMA checklist was used to ensure inclusion of relevant information in the analysis (Additional file 1: Table S1).
Published studies were identified by the Cochrane collaboration through electronic searches from January 1, 1990, up to November 2015 of MEDLINE, EMBASE, LILACS, the Cochrane Infectious Diseases Group's trials register and the Cochrane Central Register of Controlled Trials (CENTRAL) using the search term "praziquantel", regardless of language (English, French, and Portuguese) and might have failed to identify articles published in other languages.
To qualify for inclusion, trials were to be on PZQ mono-therapy at any dosage and dosing regimen, using any formulation and brand, and could be either noncomparative or comparative (randomized controlled trial, quasi-randomised trials). The search identified 828 studies of PZQ in schistosomiasis in children, of which 47 had an assessment within 8 weeks (subjects treated with praziquantel, control treatment or placebo). A previous paper reported the results of a broader metaanalysis conducted up to 2012 including subjects of all ages [14].
Efficacy outcomes are reported as cure rate (CR) and egg reduction rate (ERR). CR, defined as the conversion from a positive test pre-treatment to a negative test up to 8 weeks post-treatment, is reported as provided in the articles on a per-protocol population. ERR, defined as the proportional reduction in the mean eggs per gram of faeces (intestinal schistosomiasis) or per ml of urine (urinary schistosomiasis) post-treatment vs pre-treatment is presented based on the arithmetic (ERRam) or geometric (ERRgm) mean ERR provided in the articles, respectively, 5 and 31 of the studies included in this review (11 studies did not report ERR). Since ERR was inconsistently assessed on all patients in the perprotocol population or only those who had countable eggs in their excreta post-treatment ('uncured'), a logarithmic correlation between the CR and the ERRgm (log-transformed) was applied to correct the ERRgm results based on uncured patients and to help predict the ERRgm in studies that reported only the CR.
Tolerability is assessed by calculating the incidence of adverse events (AE) defined as any sign or symptom occurring after treatment, irrespective of whether that sign or symptom was present at baseline or not, of its severity and drug-event relationship. The mean incidence is presented for the PZQ 40 mg/kg treatment groups (all brands) excluding PZQ 40 mg/kg syrup, and Levo-PZQ 20 mg/kg. The 95% confidence intervals (CI) for the mean CR, ERR, and AE are calculated using a bootstrap resampling method based on statistical bias between studies with a maximum of 1,000 replicates [17]. For randomized controlled trials assessing the efficacy (CR) and tolerability (AE) of PZQ vs other drugs, placebo, or comparing different PZQ dosing regimens, risk ratios (RR) with 95% CI, meta-analysis regression with random effect on the study/site is used and the pooled RR is presented using the DerSimonian & Laird procedure for random effects models [18]. Heterogeneity was expressed as I 2 [19].
Children's age groups were differently categorized whether the study was school-based (between 5 and 17 years of age) or community based including (i) preschool-aged (between 1 and 5 or 7 years) and (ii) school-aged (between 4 and 19 years) according to study definition. It should be noted that studies used categories which only partly conform with the WHO age classification for PC: preschool-aged 1-4-year-old; school-aged children 5-14-year-old.
Age-groups were inconsistently defined and generally broader in community-based studies; for example, children in nursery were enrolled as school-aged in a community based study [20]. As the studies had disparate age ranges, a median age (in year) was calculated from the reported range. A multivariate mixed-effect model was used to assess the age-dependency of CR and ERR with random effect on the site in an attempt to account for any potential statistical heterogeneity across studies while controlling for potential confounders such as: (i) endpoint of assessment (analysed as a continuous variable, in week); (ii) differences in diagnostic approaches (binary, single vs multiple smears); (iii) infection intensities before treatment, categorized as light (<100 egg per gram of faeces), moderate (100-399), and heavy (≥400) for S. mansoni, and categorized as light (<50 eggs/10 ml urine) and heavy (≥50) for S. haematobium; (iv) year of study.
Graphical displays for CR and ERR by age group were illustrated using forest plots [21]. Data were analysed using Stata v13 (Stata Corp.).
We also provide more detailed narrative summaries of the studies treating specifically preschool-aged children and those reporting results for narrower age-groups, where more direct comparisons are possible between more homogenous populations.
The 1,694 pre-school-aged children were treated at five sites in Uganda, Niger and Ivory Coast with PZQ 40 mg/kg (tablet or syrup) or PZQ 2*40 mg/kg for S. mansoni and S. haematobium. The attrition at the onemonth post-treatment assessment was 11%. Of the 1,506 children assessed, 12% (n = 179) were infected with S. haematobium and were either treated with PZQ 40 mg/ kg tablet in a single-arm study (n = 18) [22] or with syrup only also in a single-arm study (n = 161) [23] and 88% (n = 1,327) were infected with S. mansoni: a twoarm study comparing crushed tablets and syrup, both given at 40 mg/kg (n = 203) [24]; a two-arm study [25] comparing PZQ tablets given as a single dose of 40 mg/kg (n = 357) to a double-dose of 2*40 mg/kg (n = 339); two single-arm studies [22,26] with crushed tablets only (n = 369 and 35, respectively); and a single-arm study (with syrup only; n = 88) [23].
Treatment outcome was reported as CR in 14,157 children and as ERR (all calculations methods) in 10,033 patients overall, of whom 58% (n = 8,199) and 62% (n = 5,958), respectively, were treated with PZQ 40 mg/kg (Fig. 1). The breakdown by species and reported outcome was: both CR and ERR, 62% of the subjects; CR only, 33%; and ERR only, 5%. For S. haematobium, the outcome was reported as CR for 4,612 patients and 3,963 for ERR; for S. mansoni this was 6,897 and 5,138, respectively; for S. mansoni/S. haematobium, 695 and 111; and for S. japonicum, 1,196 and 403.
Of the 36 studies reporting outcomes as ERR, 31 used the ERRgm and 5 the ERRam (2 studies reported both) ( Table 2). The ERRgm was calculated using a logtransformation applied to the entire per-protocol population for 24% of subjects, on uncured patients only in 16%, and not specified in 26%. No study reported ERRam on uncured subjects. Egg counts were reported using different diagnostic approaches (number of specimens and tests): from one to three days and one to three stools per day tested in simple or duplicate or triplicate slides. The most frequent approach for S. mansoni was one test on one stool sample collected on the same day and for S. haematobium was      (Table 3).
The ratio between predicted and reported logtransformed ERRs averaged 1.1 (0.93-1.4). The sensitivity analysis restricted to the studies reporting logtransformed ERRgm showed that the pERRgm was on average 0.8% higher than the reported ERRgm (using logarithmic transformation). Similarly, the pERRgm was on average 24% higher than the reported ERRgm based on uncured patients, 12% higher than the reported ERRgm with unspecified method and 3% higher than the reported ERRam.

Safety: adverse events (AEs)
Overall, 22 studies recorded AEs in 6,713 children. One additional tolerability study was included [68] to the 21 studies considered for the efficacy analysis. The proportion of children treated with PZQ 40 mg/kg was 67% (n = 4,480), 22% of whom were preschool-aged (n = 980). The syrup formulation was tested on half of the preschool aged children. Out of the three studies reporting AEs, one compared PZQ 40 mg/kg tablet and syrup formulation [24]. The two other studies were not comparative: one tested PZQ 40 mg/kg tablet [22] and the other one the PZQ 40 mg/kg liquid syrup formulation [23]. Five thousand, nine hundred and twenty-six AEs were reported within 48 h of treatment in PZQ 40 mg/kg recipients. Most of the signs and symptoms recorded were gastrointestinal, neurological and dermatological. On average 55.5% (95% CI: 39.4-78.2) of the 1,606 patients treated in 10 studies with PZQ 40 mg/kg experienced at least one of the AEs listed (Fig. 4). Signs and symptoms were not sought systematically in all studies. The incidence of AEs ranged from 6.1% for itching/rash (studied in 10 studies) to 35.5% for drowsiness (studied in 3 studies). In preschool-aged children in Uganda [18], no significant difference was detected between PZQ 40 mg/kg tablet (n = 254) and syrup (n = 249) in 15 different signs or symptoms screened.
One study analyzed age trends in schoolchildren aged between 6 and 19 years old [46]. Authors found more AEs in older subjects (P = 0.008) and differences between sites.
In Navaratnam et al. [24], 203 preschool-aged children aged 1.5 to 5 years were randomly assigned to PZQ Fig. 4 Adverse event incidence, preschool and school children, PZQ 40 mg/kg 40 mg/kg (crushed) tablet or syrup formulation. The pre-treatment arithmetic mean egg count was 317.6 in the tablet group and 292.1 in the syrup group. No difference was detected between the syrup and tablet formulations for CR (80.9 vs 81.7%) and ERRam (86.1 vs 89.0%). There was a significant correlation between pretreatment infection intensity and efficacy when assessed as CR: 88.6% in light, 74.5% in moderate and 67.4% in heavy infection. Non-compliance to PZQ syrup was 11.1% (95% CI: 8.6-14.0) and that of PZQ tablet was 14.7% (95% CI: 12.0-18.0) with no significant difference detected (OR = 1.33, 95% CI: 0.93-1.92, P =0.110). Infants 12 months of age were 20.3 times more likely (P = 0.001) to react negatively to the syrup formulation than 5 year olds. Amelioration of vomiting (OR = 1.65, P = 0.006) and peri-rectal bleeding (OR = 1.40, P = 0.007) was significantly higher with crushed PZQ tablet. The most common AEs in both treatment arms were dizziness, drowsiness and fatigue. This was the only study to report dizziness in preschool-aged children, and the incidence seemed to be higher than what reported in 13 studies in school-aged children (Table 4).
In Sousa-Figueiredo et al. [26], 305 patients aged from 5 months to 7 years were treated for S. mansoni. The overall pre-treatment egg count was 144.1 or 5.99 eggs using the arithmetic and geometric mean respectively. Children under 4 years-old had a lower CR (40.8%) and ERRam (75.4%), than 4 to 7 years-old (60.8 and 82.9%, respectively). Children with history of previous treatment had a lower CR (41.7%) and ERRam (72.8%) than treatment-naive children (77.6 and 92.1%, respectively). PZQ tablets proved to be safe, with only mild, transient reported adverse events. At the6-month and 12-month follow-up times, the number of cases of dizziness, sleepiness, fatigue, cramps, nausea, sweating and night fevers post-treatment decreased significantly compared to baseline. No symptom became more prevalent with time.
In Nalugwa et al. [25], Ugandan preschool-aged children aged from 12 to 60 months were randomized to receive PZQ 40 mg/kg or 2*40 mg/kg tablets (2 weeks apart). By week 4, no difference was detected between treatment for CR (83.2 and 85.5%, P = 0.390, respectively) or ERRgm (98.9 and 99.3%, respectively). Efficacy was lower in children with higher infection intensity and higher in younger children.

Discussion
This analysis is based on a systematic review of comparative and non-comparative studies of PZQ for treating intestinal and urinary schistosomiasis, and concentrates on the efficacy and safety of the WHO-recommended standard dose of 40 mg/kg in preschool-and school-aged children. The main question that we attempted to answer was whether efficacy varied with the subject's age, and more specifically if PZQ at 40 mg/kg is as effective in preschool-aged children as in older children, so that it can be used more broadly in this younger age category. These analyses do not indicate any age-effect on efficacy, whether expressed as CR or ERRgm, or a dose-effect However, one should bear in mind the limitations of this aggregate-data meta-analysis. First, only five studies compared efficacy explicitly between age-groups [25,26,38,52,54]. Secondly, age trends could not be calculated reliably because age categories as reported in the papers are often broad and inconsistent across studies, so that the calculated median age may not reflect the range of responses within each individual group. Following the WHO age classification for PC, the preschool-aged children age-group should include 1-4year-olds and the school-aged group 5-14-year-olds. Thirdly, preschool-aged children are under-represented: 11% of all subjects treated with PZQ 40 mg/kg. Lastly, and importantly, methodology was insufficiently standardized: studies differed in terms of diagnostic approach used (number of samples and smears), outcome measures (CR, ERR using geometric or arithmetic mean), duration of follow-up, and safety reporting. ERR was predominantly calculated using geometric means, which are less apt to identify outliers (patients responding less well) and tend to show higher efficacy, compared to arithmetic means or individual-patient response distributions [3,69]. It is therefore difficult to apply the WHO-recommended threshold of 90% ERRam [3] to the results of these studies and meta-analysis. Moreover, when the geometric mean egg count was used, some studies calculated the ERR using the geometric mean as the difference between egg counts on all the patients, pre-treatment, and only the positive subjects at the time of the endpoint assessment. This method is inappropriate and biases the result because the denominator changes from the baseline to the endpoint time, overestimates the egg counts posttreatment and underestimates efficacy (in our calculations here by~24%).
To mitigate these shortcomings, we recalculated the ERRgm and applied multivariate analyses. The predicted ERRgm (pERRgm) provided a good fit with the reported ERRgm using log-transformation; only three data points (the lowest reported ERRgm) out of 34 fell well below the trend line, which demonstrates a good regression fit. Of these, two are from the same study (S. mansoni infection, moderate intensity, Kenya [62]). There might be multiple reasons for this. One is mathematics (the geometric mean wipes out extreme values, so the contribution of individual outliers in the response distribution to the overall response is minimised). Another one is genetic heterogeneity between schistosome populations from different geographical regions, which means that otherwise generalizable trends might not apply to some specific local situations. A third one is the background prevalence of infection (which could not be estimated here, but appears to play an important role [10]). Predicting ERRgm from CR has therefore its limitations, but it allowed including a broader range of studies and a larger number of subjects that would have otherwise been possible, making it possible to expand the utility and comparability of older studies; it is expected that this will not be required with newer studies adopting the WHO recommendation to express results as ERRam [3].
With respect to the WHO-recommended threshold of 90% efficacy by ERRam, and with all the abovementioned provisos, using our statistical correction, we found that 85% of the study arms treated with PZQ 40 mg/kg for any species would meet this criterion.
The multivariate analyses with random effect on the study site to account for differences between studies using PZQ 40 mg/kg in S. mansoni or S. haematobium found no relationship between efficacy (whether measured as CR or ERR), and age (preschool vs school children aged), duration of follow-up (one vs two months), formulation (syrup vs tablet), diagnostic approach (single vs multiple smears), and infection intensity (light vs others). It is however worth noting that the effects of risk factors on efficacy varied across individual studies. Sousa-Figueiredo et al. (Uganda, S. mansoni) [26] found that the risk of not clearing the infection was significantly higher in children less than four years-old than four to seven year-old by a factor of 3.5, and in children with heavy infections by a factor of circa two. Thiong'o et al. [62] (Kenya, 5-9, 10-14, 15+ years) found a significant trend in four study sites for efficacy measured as CR to decrease with age: ranging 59-92, 72-79, 80-93, 52-78%, and ERR ranging 99-100, 99-100, 93-98, 75-91); of note, older children were more heavily infected and had higher ERRgm than younger children with lower infection intensity before treatment. Gryseels et al. [54] in Burundi found opposite trends in CRs after treatment with doses of 20, 30 and 40 mg/kg in subjects aged < 20 (from~52 to 78%) and 20+ years (from 80 tõ 94%). A recent study conducted on 508 Zimbabwean children treated with PZQ 40 mg/kg for S. haematobium not included in this review [70] found no difference in CR and ERR between children aged one to five (both 100%) and six to 10 years (94% and 97.9%, respectively). A previous systematic review and meta-analysis [10] which compared CR in similar numbers of participants (1,026 preschool-aged and 12,906 school-aged children), also did not detect significant differences in the combined CRs for S. mansoni and S. haematobium (73.6 vs 71.1%, respectively), but rather found a correlation between lower CR and higher background infection prevalence, especially for younger children. Here, we confirmed and extended these findings by including also ERRgm (see above), though both CR and ERRgm are probably inaccurate for the reasons listed above. These disparate findings likely reflect a range of situations in terms of infection prevalence and intensity in younger and older children which also depends on the deployment and coverage of preventive chemotherapy cycles in the population.
These findings must be considered in the light of the current imperfect test-of-cure; counting eggs in excreta (using the available low-sensitivity methods) cannot distinguish between definitive (adult worm killing) and temporary effects (suspended egg excretion). These methods, while useful in quantifying effects of PC programmes at the population level, are inaccurate when it comes to assess the real drug efficacy and detect early signals of reduced efficacy and potential resistance.
The other question was whether higher levels of efficacy could be achieved using 60 mg/kg instead of 40 mg/kg (and in which age-group). No evidence of difference emerged from this meta-analysis in children or previous meta-analyses in overall population [12][13][14]. However, a definitive conclusion cannot be reached on this question, both because the numbers are relatively small (S. mansoni: PZQ 40 mg/kg, n = 492 vs PZQ 60 mg/kg, n = 473; S. haematobium: PZQ 40 mg/kg, n = 74 vs PZQ 60 mg/kg, n = 77), and because ERRs do not capture the distribution of individual responses, and particularly outliers [69]. We also know that better efficacy can be achieved with longer treatment and higher doses -conditions that are however not feasible in routine PC.
Compared to our previous systematic review and meta-analysis, this was focused on the efficacy and safety of PZQ in children; here, the calculated median age was 10 years old, and the range across all studies was one to 19 years, for a total of 14,340 subjects with an efficacy outcome. The previous analysis (n = 19,499) had all treatments and age-groups (age range 1-75 years old). In addition, here we explored different approaches for ERR calculations methods and applied a statistical correction to standardize ERR outcomes.
Praziquantel was well tolerated in all age groups, though approximately half of the subjects experienced an adverse event. It must be however noted that safety was variably and inconsistently assessed; few AEs were assessed systematically across most studies. The frequency of certain AEs in each age-group may simply reflect the fact that these were recorded in certain studies, and not others. Also, it is very difficult to detect true signal from background noise: the absence of pre-exposure information on presence and severity of signs/symptom, prevents assessing whether the events reported were related or not to treatment. It is all-important that clinical researchers be reminded of the critical relevance of collecting safety information in the controlled environment of clinical trials, even for drug which are widely regarded as welltolerated: this is especially true for drugs, like praziquantel, that are used predominantly in children, who might not complain and cannot verbalize discomfort, with the risk of risks being underrated events being underreported [71]. A simple way to doing that is to collect and grade signs and symptoms at enrolment to allow comparison between pre-and post-drug exposure, to identify treatment-emergent signs and symptoms. Child-friendly questionnaires including drawings should be considered, especially in young children. and meta-analyses; PZQ: Praziquantel; RR: Risk ratio; sh: S. haematobium; sj: S. japonicum; sm: S. mansoni; SP: Sulfadoxine-pyrimethamine; WHO: World Health Organization