Simulation models predict that school-age children are responsible for most human-to-mosquito Plasmodium falciparum transmission in southern Malawi

Although malaria control interventions have reduced parasite prevalence in some regions, ambitious efforts have caused little reduction in many settings with high endemicity [1‐6]. Progress toward elimination requires understanding why current strategies fail. One potential explanation is that they do not adequately target sources of human-to-mosquito Plasmodium transmission. Defining the primary human reservoirs of transmission, or people who contribute most to infecting competent vectors, requires an understanding of infection patterns in human populations, variations in infectivity when bitten, and differential likelihoods of being bitten by female Anopheles.

Southern Malawi is one area where, despite high coverage of insecticide-treated nets (ITNs), the malaria burden remains high. Recent cross-sectional studies have found that 5–15-year old, school-age children (SAC) have the highest prevalence of Plasmodium falciparum infection [7], are more likely to carry gametocytes when parasitaemic [8], and are less likely to use bed nets [9] than children under five (U5s) and adults over 15 years. These findings suggest that SAC may be important human-to-mosquito transmission reservoirs in that area, but their contribution has not been quantified.

Several skin- or membrane-feeding studies in other sites have calculated age-specific contributions of human infectious reservoir populations [10‐18]. While they elucidated the difference between infected and infectious humans, most models did not consider whether mosquito feeding heterogeneity influences which people are key transmission reservoirs. Indeed, highly infectious people who are never bitten cannot contribute to transmission, while weakly infectious people who are frequently bitten might contribute considerably. Mosquito feeding frequency has been found to increase with the age of human hosts, possibly due in part to increasing body surface area [18‐22]. Recently, Gonçalves et al. [18] matched mosquito blood meals to their human sources and found evidence of extreme heterogeneity in biting frequency by age. In the high-transmission setting of Burkina Faso, adults were bitten 20 times as often as U5s, and SAC were bitten 7 times as often as U5s, and, accordingly, the adult contribution to transmission was underestimated when considering only the proportion of infectious people by age [18]. Assuming blood-feeding homogeneity limits the estimation of true transmission contributions.

Using recent molecular parasite data and reported ITN use in southern Malawi, this model estimated that SAC were responsible for > 60% of P. falciparum transmission to mosquitoes during both the dry and rainy seasons. The model estimated that young adults (16–30 years) were the second largest contributors to transmission; U5 children and adults over 30 years were each predicted to have caused < 10% of mosquito infections in most simulations. The proportion of mosquito infections attributable to each human age group did not differ considerably by season. Lacking human-to-mosquito infectivity data from skin or membrane feeding, sensitivity analyses were performed under a reasonable range of estimates of relative infectivity by age after accounting for gametocyte density distributions. If U5s are far more infectious to mosquitoes than all other age groups at a given gametocyte density, their estimated transmission contribution approached 20%, but SAC and young adults remained the largest and second largest contributors in these simulations. These findings may help explain P. falciparum’s persistence in the area despite considerable malaria control efforts in the past decade which targeted groups at high risk of disease—pregnant women and children under five.

Predictions about the transmission reservoir are sensitive to heterogeneity in mosquito biting, particularly in evaluating the contribution of U5s. Several prior lab studies of infectivity have made simple estimates of age-specific transmission contributions; however, those reports assumed homogeneous mosquito feeding [10‐17]. The model presented here, like those of Stone et al. [21] and Gonçalves et al. [18], demonstrated that such a homogeneity assumption considerably overestimates the contribution of U5s and underestimates that of older age groups. In their mosquito blood meal source matching study, Gonçalves et al. [18] found that the degree of heterogeneity in mosquito biting by human age was different for their high and moderate transmission sites. The biting frequency ratio in the high transmission site was even more extreme than that used by Stone et al. and the model presented here, at 1:7:20 for U5s compared to SAC and adults 16 years and older, but the increase in biting frequency with age was moderate (1:2.3:2.1) and did not achieve statistical significance in the low/moderate intensity transmission site in Kenya. Given the sensitivity of transmission dynamics to this parameter, future research should better quantify heterogeneity in human exposure to feeding mosquitoes to more accurately identify key P. falciparum transmission reservoirs, and assess the factors which drive variation in this age-specific biting heterogeneity in different geographic settings.

The results presented here using recent empirical data from southern Malawi predict high transmission contribution from SAC, and are consistent with those of Stone et al. [21] which were based on generic sub-Saharan African estimates of parasite prevalence and population distribution. This suggests that the findings may be generalizable to other sub-Saharan African settings with high P. falciparum endemicity. A worksheet of the model calculations that other researchers can utilize in their specific contexts with, at a minimum, age-specific data on (1) population distribution; (2) estimated PCR prevalence; and (3) frequency of ITN use is provided as Additional file 3. Results should be interpreted with caution in settings with different malaria epidemiology, particularly those near elimination or where non-falciparum malaria is common.

A potential limitation of the study was the use of four compartments for gametocyte density, and the subsequent need to average the density-dependent transmission probability within each compartment. The estimation of parameters for each compartment based on Churcher et al. [24] resulted in a lower relative likelihood of transmission for the highest density categories compared to that applied in some other models [26]; however, the very small proportion of high density gametocyte carriers in this community-based population makes the model relatively insensitive to changes in the value selected for that compartment. This limitation would be critical, however, if the few high density gametocyte carriers are disproportionately likely to be fed upon by mosquitoes, essentially driving transmission as ‘superspreaders’. As seen in this model, transmission patterns may be highly sensitive to heterogeneity in blood feeding. More research matching blood meals to specific human sources is needed to understand mosquito feeding patterns in real-world settings. If it is discovered that a select few superspreaders are responsible for the majority of Plasmodium transmission, this would have significant limitations for targeting interventions aimed at control and/or elimination.

Further, the model is static and depends on measured prevalence data, without regard for the causes of differential prevalence of infection by human age. While this should not limit inferences about transmission contributions for the time-period studied, design of effective interventions will require better understanding of whether differences in infection prevalence are a result of differences in risk of infection, natural infection clearance time, frequency of anti-malarial treatment, or some combination of these.

This model predicted that SAC and young adults were responsible for > 80% of all new mosquito infections in southern Malawi, with SAC contributing ~ 60% overall despite comprising only 30% of the population. A critical implication is that malaria elimination is unlikely in persistent settings like southern Malawi unless researchers and policy-makers develop interventions that can successfully reduce transmission from these groups, who are often asymptomatic and submicroscopically infected. More sensitive rapid detection methods and development of safe/effective drugs for transmission reduction are therefore fundamental to further progress toward malaria eradication.