In an era where vector control strategies have been able to greatly reduce disease burden across the malaria endemic world, many countries are now evaluating how to transition programmes from control to elimination. Mass distribution of anti-malarials, with or without a diagnostic, are under consideration in many settings to markedly reduce the human parasite reservoir to allow a shift to case detection via surveillance as an intervention. Results of this study are consistent with both modelling and empirical evidence that has shown implementation of drug-based strategies to be effective within a transmission season in settings where sufficient coverage of the population can be achieved [
36]. Yet, recent studies on screen and treat for strategies targeting the whole community [
37] and specific risk groups [
38] have been unable to show sustained impact. This study examined how the OpenMalaria platform, parameterized according to the malaria transmissions dynamics in southern Zambia, can assist the Zambia NMCC to increase the effectiveness of these planned interventions through changes in implementation design.
Results suggest the optimal implementation strategy for deployment of community-based anti-malarial treatments requires a consistent approach addressing the variation in baseline malaria prevalence, rate of imported infections, and the achievement of high coverage in the population. Simulations conducted for this study suggest that an area such as Southern province with heterogeneous transmission and an uncertain infection importation rate, drug-based interventions based on detecting infections using current RDTs will not be sufficient to interrupt transmission. Indeed, as observed in the field study, there was no difference detected at follow-up in parasite prevalence between MTAT intervention and control in either high or low transmission areas [
3]. In this setting, a high proportion of low-density infections will be missed by RDTs that would be treated and cleared with MDA.
Simulation results suggest the most important determinant of success in reducing prevalence is the coverage of the population achieved in the campaign. The Southern province field study was designed to show how much impact would be achieved after 1 year of campaign implementation in order to move towards surveillance as an intervention by focal MDA (fMDA) and case investigation. However, even with high coverage with MDA in areas with a pre-intervention all-age parasite prevalence of less than 10 %, simulations suggest that elimination will require more than 1 year of campaign implementation. The inclusion of SLD primaquine (a gametocytocide), and ivermectin (an endectocide), to the drug regimen did not further reduce parasite prevalence within this 1 year post-campaign. It is important to note that mean parasite prevalence measured at earlier time points will yield different results, as the greatest effects will be seen several months post-campaign. Simulation of ACT regimens in a generic setting have been investigated elsewhere, both with and without the option of including SLD primaquine and/or ivermectin [
39,
40]. The simulation results presented here are less optimistic about the additional added benefit of PQ and ivermectin in interrupting transmission, although, importantly, the results are evaluated at different time points.
Success of anti-malarial drug administration campaigns depends on the sustained high coverage of vector control interventions. Zambia has been able to achieve high rates of household LLIN ownership and has steadily increased its within-household availability of LLINs through successive mass distribution campaigns. However, LLINs must be used or at least deployed in the household for their full impact to be attributable. Improving simulation estimates by including more direct measures of full coverage and usage of LLINs may yield greater overall reductions in predicted parasite prevalence in combination with the anti-malarial campaigns. Similarly with IRS, improved strategies to target the application of insecticides could be evaluated for specific benefits. Ensuring that transmitting mosquitoes are susceptible to IRS chemicals, that IRS is prioritized to transmission areas, and that full coverage of targeted areas is achieved will maximize the contribution of this intervention where drug strategies are attempted.
Planning for elimination scenarios: limitations and considerations
In the process of this simulation, we have identified key aspects of both the intervention tools and their delivery; improved data from the field may help in the future to improve the model and its ability to predict outcomes. As noted above, intervention coverage is a critical variable for these population-wide treatment efforts and intervention programming should improve on the quality of the coverage data. For example, most coverage estimates only examine the coverage of the household members in households that were reached; a more accurate estimation of the population denominator and coverage could perhaps be achieved by comparing satellite imaging of houses to actual field geo-position data of which houses were reached by a campaign.
Analysis was conducted with the end-point measurement of reduction in parasite prevalence. This is, indeed, an easily-measured quantity in the field. However, future applications of the OpenMalaria model in the study area should also include an end-point measurement that corresponds to potential vector control interventions, including ivermectin. The goal of ivermectin is not to reduce prevalence, but to reduce transmission. Further analysis measuring reduction in EIR or probability of interrupting transmission would be a more useful way of evaluating the impact of this intervention. In addition to the clear need for direct application of field-measurable quantities, it would be helpful to have a standard definition for defining interruption of transmission for model outputs that is operationally relevant to programme settings. Ideally this would include identifying the number of cases per week or per month in a given health facility catchment area that the health system would be able to handle.
The model has numerous limitations that are inherent in the introduction of many assumptions required for model parameterization. For example, the modelled analysis included a sensitivity analysis for different levels of adherence to AL with the assumption that the majority of non-adherers missed the final dose in the six-dose regimen, with results showing little impact on overall effectiveness. However, non-adherence would have a far greater impact on the cure rate of DHAP if a single dose of this three-dose regimen is missed [
41], something not included in this experiment.
The field implementation targeted health facility catchment areas for complete campaign coverage, while the simulation coverage parameterization and analysis is applied to the general population without explicit geographical limits. While model parameterization assumes an individual has the same independent likelihood of receiving the intervention, there are some groups that have a higher risk of being infected. These groups, including adults aged 25–49 that are more highly mobile than the rest of the population, were targeted in the simulation study design by varying the age groups targeted by the campaign. However, because these groups may be less likely to be reached during a general campaign, this higher risk is not captured in the baseline scenario of the model and may affect the impact of the simulated interventions even if a certain level of coverage is reached for the population as a whole.
The model parameterization assumes a conservative value for AL and DHAP prophylaxis based on modelling pharmacokinetics [
41]. However, a recent publication of post-treatment prophylaxis of anti-malarials suggests both a shorter duration of protection and a greater difference between the duration of protection between AL and DHAP in people over 5 years of age [
42]. This greater difference in prophylaxis benefit would affect the simulation results by increasing the additional transmission prevented by DHAP as compared to AL.
For the ivermectin analysis, the model assumes a common effect of ivermectin no matter the age of mosquito post-emergence, which has been shown to be different in trial settings [
14]. Several studies [
15] have found a higher mortality rate in mosquitoes exposed to ivermectin than the 85 % assumed in this experiment; this could indicate the experiment design underestimates the effect.
The model assumes a constant rate of yearly case importation (20 cases per 1000 individuals); this rate will undoubtedly vary by season and from location to location even within southern province, and data is not currently available to be able to parameterize this value correctly. Because the OpenMalaria simulations assumed an ongoing rate of imported infections, it is not straightforward from model outputs to distinguish between cases arising from imported infections and cases that were a result of local transmission. This parameter is likely a key driver of the differences in the simulated versus observed results of the model validation shown in Fig.
3. If possible, future iterations of field work could improve the ascertainment of mobility and likely imported versus locally transmitted cases. This includes methods such as parasite genotyping or using human movement via mobile phone data as a proxy of parasite flow [
43]. Understanding these dynamics will aid not only in developing more relevant baseline model parameterization, but understanding thresholds for importation in a population on the effect of intervention mixes. Combined with transmission estimates, these metrics can provide a more granular view of what intervention mix to deploy, and where.
Demonstrating elimination, and evaluating the ability of interventions to interrupt transmission, requires metrics that are practical in operational settings. The reproductive number under control settings,
R
c
, has been proposed as a metric to determine whether an area has achieved elimination [
44] by calculating the ratio of local cases to imported infections. However, the operational scale at which this should be applied is unclear, as is the geographical definition of infection importation in this context. For example, should importation from higher-transmission zones in the same province be considered as part of the definition, or only importation from a neighbouring country? The area in Southern province currently evaluating anti-malarial administration strategies is bound by the vast Lake Kariba to the east and south, and much lower malaria transmission areas to the north and west, making it an ideal setting to test elimination strategies. A country level metric is likely unsuitable for a country like Zambia which has highly heterogeneous endemicity in different parts of the country. On the other hand, applying such a metric at a more granular administrative level would not confer recognition of elimination by the WHO. Country-specific elimination strategic plans should address how progress will be tracked at both national and sub-national levels.
Other simulation studies and study data suggest the proportion of sub-patent infections missed by RDTs is higher in areas of lower transmission [
37,
45,
46], consistent with the simulation results from this study (Fig.
4). Transmission due to asymptomatic infections becomes increasingly important (in relative terms) as cases in a country become fewer and national control programmes approach elimination and prevention of reintroduction. As such, further validation and calibration of models to the prevalence of sub-patent transmissible infections in different epidemiological settings needs to be conducted as new study data becomes available.
Finally, an important limitation for interpreting results is the non-spatially explicit structure of the OpenMalaria model. Focal test and treat and focal MDA strategies that involve targeting specific households, geographic areas or risk groups, and reactive case detection are all implementation strategies in the study area but are not evaluated in the context of this experiment. In addition, there are options for anti-malarial drug administration intervention design that will be relevant in the Southern province that are not explored in this study. Future analyses could examine the potential impact of this larger package of intervention strategies as they are implemented in the same or adjacent geographic areas; such analyses could further contribute to the future programme design by the NMCC.