br Although the global burden

Although the global burden of malaria has been greatly reduced through, for example, vector control, the disease is still a major concern and there is renewed interest in mass drug administration (MDA) as an additional control intervention. A Cochrane review of MDA for malaria, which included 32 studies, concluded that MDA substantially reduces the risk of infection leading to malaria. However, this study also raised the concern that these reductions are often not sustained. Another review from 2015 suggested that, although MDA is successful in controlling and eliminating disease, there remain substantial knowledge gaps and further studies are essential, particularly on optimal size of the target population, methods to improve coverage, and drug safety. Oliver Brady and colleagues address some of these research gaps on MDA for malaria in (July 2017 issue). Given the relatively small number of studies of the pkc inhibitor impact of MDA, computational models that can simulate the underlying biological processes of malaria in human and mosquito populations and also be fitted to the available data represent attractive “what-if?” tools. However, several approaches to modelling malaria epidemiology exist, and Brady and colleagues present a model comparison analysis done by the Bill & Melinda Gates Foundation-funded Malaria Modelling Consortium (MMC) on the effectiveness of MDA in different settings, identifying the most important determinants of MDA effectiveness. They compare a specific model outcome (the Parasite Rate following 2 years of MDA) generated by four established models, which represent different, yet reasonable, ways of simulating malaria transmission and MDA. The study provides much needed support to researchers responsible for quantitative analysis in the service of malaria control programmes, who are often faced with a plethora of mathematical models, run under non-standardised sets of interventions, and reporting different outcomes. Until recently, these analysts would typically depend on the work of a single modelling group. This approach was clearly less than optimal, since it often failed to take into account the complementary work of several other research groups around the world. Thanks to the leadership of organisations such as The Bill & Melinda Gates Foundation, there has been a shift during the past decade towards modelling consortia (eg, the MMC), in which representatives from several modelling groups meet frequently and evaluate scientific or operational questions collectively. Brady and colleagues, as members of the MMC, compare the quantitative predictions of a set of four malaria computational models. There is no simple recipe for performing such a comparison. Various modelling consortia have had to determine the best way to combine outcomes into consensus reports for diseases such as HIV, tuberculosis, and neglected tropical diseases. In the non-communicable disease domain, the US National Cancer Institute\'s Cancer Intervention and Surveillance Modeling Network (CISNET) is charged with investigating various questions in a comparative fashion, such as those relating to the potential medical and cost-benefit of cancer screening programmes. The points of agreement of the models in the study by Brady and colleagues make sense—eg, it is intuitive that the post-MDA prevalence of malaria is strongly related to the proportion of individuals in the community missed by an MDA campaign, as well as the pre-MDA endemic prevalence. Although it may be more difficult than discovering and documenting consensus, explaining discrepancy between models is also useful. We quickly move into philosophical territory when we ask questions such as: what does it mean to make a consensus prediction based on model structures that disagree? Or can Modelling Consortia devise experiments to reconcile (or refute) competing model structures? Forecasting and classification based on the averaged outcomes of an ensemble of models has become a standard method in machine learning; but ensemble approaches have not yet been taken up by the infectious disease modelling community, whose models are more mechanistic and therefore have a stronger link to underlying biological processes than machine learning or statistical approaches. However, when more data become available from, and become more consistent between, control programmes, we can expect to see statistical forecasts going head-to-head with mechanistic models in model comparison studies.
In , Jean T Coulibaly and colleagues report the efficacy and safety of different dosages of praziquantel in preschool children. In a randomised dose-ranging trial in southern Côte d\'Ivoire, they compared the cure rate obtained with three different doses of praziquantel or placebo, in preschool and school-aged children infected with . 161 (24%) of 660 eligible preschool-aged children had a detectable infection and 154 received treatment. 62% of preschool-aged children were cured with the 20 mg/kg dose, 72% with the 40 mg/kg dose, 71% with the 60 mg/kg dose, and 37% with placebo. 180 (80%) of 225 school-aged children were infected and 178 received treatment. In that group, the 20 mg/kg dose resulted in cure in 30% of the children, the 40 mg/kg in 69%, the 60 mg/kg in 83%, and placebo in 12%. The authors conclude that in the absence of other treatment options, a single dose of praziquantel of 40 mg/kg can be endorsed for preventive chemotherapy programs in children younger than 5 years of age.