The results obtained from the cohort model using FoI from two different individual-based stochastic models of STH transmission suggest that annual deworming treatment of adolescent girls reduces the prevalence of M&HI infections in this age group by up to 60% in moderate transmission settings and by 12-27% in high-transmission settings. This is comparable to the impact of SB treatment of pSAC/SAC. The results also suggest that treatment of WRA during pregnancy and lactation reduces the prevalence of M&HI infections by a small but significant fraction of less than 20%. In moderate prevalence settings where infection levels are uniformly distributed over age (ICL model), reductions of the FoI over time achieved by ongoing SB treatment can prevent morbidity in WRA. However, for highly endemic settings (both models) and settings where adults host the majority of the hookworm population (EMC model), morbidity in WRA will persist at low levels during SB treatment, which allows for (some) additional benefit of targeted deworming of WRA.
The lower impact of treatment during pregnancy and lactation on the prevalence of M&HI infections compared to annual treatment during adolescence or SB treatment is expected, as it corresponds to low-coverage treatment in adult women. Assuming on average four pregnancies per woman, each woman receives treatment on average every seven to eight years. This frequency is unlikely to completely prevent re-infection and the re-establishment of M&HI infections between treatments.
The differences in the predictions between the two models can be primarily explained by the different age-intensity profiles of infection in the two models and different assumptions about the form of density dependent egg production by adult female worms. The ICL model assumes that exposure to infection is constant over all age groups, such that the worm burden rises to a plateau as people age, as observed in the recent Tumikia study data from Kenya [14]. In contrast, the EMC model assumes that exposure and contribution are 0 at age 0 months, increase linearly up to age ten and stay constant thereafter. This produces a pattern of infection intensity that increases with age up to the age of about 15, matching some previously published age-intensity profiles from various countries [15]. Accordingly, SB treatment has a greater impact on the FoI in the ICL model compared to the EMC model in moderate-transmission settings.
A second explanation for the differences between predictions of the two models relates to density dependence in female worm fecundity, which describes how the egg production per female worm declines as the number of worms in a host increases [17]. The EMC model assumes a hyperbolic saturation of the density-dependent fecundity of female worms inside the host, while the ICL model assumes the density-dependent fecundity of female worms decays in an exponential manner with negative exponent to almost zero as worm burdens rise to high levels [17]. The two functions are similar for low and moderate worm burdens, but different for high worm burdens (eggs produced per female worm decline more rapidly with exponential saturation). As a result, in individuals with high worm burden, female hookworm produce more eggs in the EMC model than in the ICL model. Consequently, killing the same number of worms by treatment leads to a greater reduction in epg in the EMC model compared to the ICL model. Moreover, because in the ICL model density-dependent processes lead to a reduction in the total number of eggs produced inside a host when worm burdens are very high, a reduction in the worm burden following treatment (and re-infection) can result in higher egg production inside a host. This explains why the impact on the prevalence of M&HI infections in high-transmission settings is less in the ICL model than in the EMC model.
A third difference between the two models is the variation in worm lifespan which is assumed to be Weibull distributed in the EMC model and follows an exponential distribution in the ICL model. The modes of the Weibull and exponential distributions are different (2.12 vs. 0) which contributes to differences in the distributions of worms within hosts between the two models.
In the simulations, we assumed effective treatment coverage of 75% and random-access to treatment. Not all countries may achieve this, and non-access to treatment is likely non-random in reality. However, here our focus is on comparing two different treatment strategies. Provided that all assumptions are the same for simulations of the different scenarios that we compare, the results will be informative with respect to the treatment strategies.
When girls/women stop receiving regular SB MDA at effective coverage each year (at age 14 years according to the old guidelines, at age 18 according to the new guidelines), the prevalence of any infection and the prevalence of M&HI infection are expected to be low in the cohort directly after treatment stops. How long resurgence of prevalences to endemic equilibrium levels will take depends on the worm life expectancy (two or three years in the case of hookworm) and whether or not treatment has lowered the prevalence of any infection so much that it is close to the transmission breakpoint. The transmission breakpoint is the prevalence below which transmission of the parasite cannot be sustained and the parasite population becomes extinct. If the latter has been achieved, models predict that resurgence of infection can take many years and to depend on people movements reintroducing infective stages into the environments where transmission interruption has been achieved [18, 19]. If the prevalence of any infection is not close to the transmission breakpoint after treatment, the time to resurgence of infection in the population will be about equivalent to the worm life expectancy, i.e. 2-3 years. In the cohort model, we only look at the prevalence of infection in a cohort of girls/women and not at the background population that they are embedded in. Consequently, we cannot conclude from the results of the cohort model on its own if the transmission breakpoint has been reached or not. In the absence of any treatment (SB treatment or treatment of WRA), the prevalence and the prevalence of M&HI infections in the cohort at any age is always higher than if treatment at effective coverage is implemented (Supplementary Figure 1).
If treatment has not eliminated hookworm transmission, there will be a FoI that causes a resurgence of infection between treatments, but not back to the endemic equilibrium levels that would be observed in the absence of treatment (compare spikes between treatments and equilibrium prevalence of M&HI infection in Figures 1-3). Resurgence between treatments does not exclude reaching the target of an MDA programme (e.g. <2% M&HI infections). However, the treatment strategy must be intensive enough for a given intrinsic transmission intensity in a defined setting and must be applied for long enough to reach the target. This has been shown in previous simulation studies [8, 20].
Resurgence of infection may still occur quickly in locations where transmission intensity is high and the infectious reservoir has not been sustainably depleted. For example, Ortu et al. 2016 report that a single year of missed treatment led STH resurgence back to baseline levels in an MDA programme for STH in Burundi where the dominant species was A. lumbricoides [21]. In a systematic review and meta-analysis of re-infection with STH after MDA, Jia et al. 2012 found that re-infection between treatments occurs quickly. Twelve months after the last treatment prevalence resurged to 94% of baseline levels in A. lumbricoides, 82% of baseline levels in T. trichiura and 56% of baseline levels in hookworm. Therefore, Jia et al. recommended frequent antihelminthic treatment [22]. Similarly, Gunawardena et al. 2011 found a prevalence of M&HI infections of 11.6% four years after MDA ceased [23]. Appleton et al. 2009 reported rapid resurgence to baseline prevalence levels four to twelve months after treatment in slum-dwelling children in South Africa [24].
Another study that shows that M&HI infections can be acquired rapidly is Menzies et al. 2014. The study, set in Ecuador, investigated how quickly children are infected with STH during the first three years of life. The authors found that prevalence of any STH infection was about 25% at three years of age and that 10-15% of infected three-year-old children had M&HI infections [25]. This is accordance with the prevalence of M&HI infections a few years after treatment in our simulations.
However, rapid resurgence at the local level does not necessarily contradict sustained low prevalences and reaching the WHO 2030 morbidity target at the country level. At the country level many locations, especially those with low-moderate baseline prevalences, could successfully reach morbidity control, and prevalences could be brought down to levels near the transmission breakpoint from where resurgence, if it happens, takes longer [19]. If the baseline survey happened many years in the past, it is also likely that economic development has occurred in a country and reduced the transmission intensity in many locations, which is another explanation why resurgence to baseline levels is not observed at the country level.