Figure 1.
Schematic view of a networked landscape of Schistosoma transmission.
The diagram indicates the typical overlapping transmission links (gray lines) between human habitation sites (villages V1–V10, black dots) and their respective primary and secondary water contact sites (ponds P1–P5, gray dots) where people become exposed to cercariae from infected host snails. While each village may have its own principal transmission characteristics, this linked meta-population structure can foster more rapid reintroduction of infection in previously treated communities [22], [23].
Table 1.
2006 World Health Organization categorization of communities at risk for schistosomiasis [16].
Table 2.
WHO 2006 recommendations for population-based treatment of at-risk communities according to prevalence category [16].
Figure 2.
High correlation between pre- and post-treatment prevalence of Schistosoma infection at the local village-level.
Village-level S. haematobium prevalence values for school-age children are plotted for 10 neighboring communities included in the Msambweni Study in coastal Kenya [24], [25], [29]. Pretreatment prevalence values from Fall 1983 are indicated on the x-axis, while post-treatment values for 2000 are indicated on the y-axis. During the 1983–2000 interval, the participating communities received school-based targeted drug administration from 1984–1992, after which a funding lapse led to suspension of treatment. The 2000 values thus indicate a robust return of infection prevalence after an 8 year hiatus of control. Pearson correlation (R = 0.83, P<0.001) indicated a strong association between pre- and post-control village-level prevalence values. Corresponding linear regression (y = 1.08x−0.32) indicated that the post control prevalence was, on average, 32 percentage points lower than before control. Nevertheless, all villages relapsed to the moderate or high S. haematobium prevalence categories, with pre-control prevalence an excellent predictor of the level of post-treatment rebound. These and post-treatment data from additional yearly surveys were used to calibrate the SWB model used in this study.
Figure 3.
SWB model validation: prediction of 2009 village-level S. haematobium prevalence among school-age children.
Following initial calibration of the programmed SWB model against human and snail S. haematobium prevalence data from 1983–1987 and 2000–2006 [22], we checked the predictive accuracy of the model against new survey data obtained for two study villages in 2009 after a round of community-wide treatment in 2006. The observed S. haematobium prevalence among school-age children (y-axis) is indicated by black dots for villages 6 and 7. The corresponding gray dots indicate the predictions of the model for these two villages using the model's best-fit parameter values. The box-plots indicate the median, interquartile range, and 95% range of the SWB model predictions in sensitivity analysis, in which model input parameters were allowed to vary at random over a ±20% range. The concordance between observed and predicted values in this and other validation testing [22] provided support for the accuracy of model projections in the present study simulations.
Figure 4.
Modeling time-to-success for current WHO treatment strategies, as a function of program adherence.
In this SWB model simulation for the 10 interlinked villages in Figure 1, WHO treatment assignment was made based on current recommendations for high- and moderate-prevalence communities (see Table 2). The bars indicate the number of years required for the overall area school-age population prevalence to drop below 10% (y-axis). This working target is a prevalence level is felt to be associated with minimal risk of advanced schistosomiasis [16]. As shown in the different bars along the x-axis, the time to needed to achieve the <10% global prevalence target varied according to the population participation (adherence) with the implemented treatment program. At 80–90% adherence among children (achieved in many research control programs), it took a minimum of 4 years of treatment (as assigned in year 1, based on starting prevalence) to achieve the <10% target. Where adherence was less good (60–70%, typical of many non-research national programs), control was projected to take appreciably longer (6–8 years on the assigned treatment strategy).
Table 3.
Performance of WHO 2006 strategies with or without yearly pretreatment screening with possible treatment reassignment.
Figure 5.
Projected rebound of infection in villages V2 and V6 after community-based treatment for two years.
Graphs indicating the projected 20-year annual infection prevalence experience of two area villages, one of initially high prevalence (village 6, panel A) and one of moderate prevalence (village 2, panel B) during and after a 2-year annual community-wide drug administration campaign followed by a 2-year treatment hiatus (C-C-H-H). The timing of initial treatments is indicated by downward arrows. In the high-prevalence village (panel A), community coverage at 70% adherence among children and 50% adherence among adults reduces school age prevalence below the desired 10% target (gray zone) in two years. However, subsequent implementation of a treatment ‘holiday’ for 2 years results in rapid return of school age prevalence to moderate-level prevalence. At that point, re-implementation of treatment through an annual school age coverage results in continued suppression of prevalence to the <10% level. However, if school age treatment is again suspended after 4 years (dashed line) or 8 years (solid line) of retreatments, childhood prevalence is projected to slip above 10% within 3 years. By contrast, in panel B (lower prevalence Village 2), school age prevalence is adequately suppressed by 2 years of community-based treatment, after which local school age prevalence is expected to remain below 10% (gray zone) for at least 4 years. Here, dashed and solid lines indicate that the duration of school age treatment of village 6, whether for shorter (dashed line) or longer periods (solid line), has only marginal effect on the rebound of infection prevalence that is expected in Village 2.
Figure 6.
Likely impact of treatment adherence on long-term suppression of village prevalence using current WHO strategies.
In this simulation, village-level outcomes are projected in terms of the number of post-treatment ‘safe’ years (i.e., years with school age prevalence <10%) that are likely if assigned treatment is ended once an areas achieves a <10% prevalence in all villages. Here, control is achieved using standard WHO treatment assignments, with the differently shaded bars indicating different levels of treatment adherence. (As detailed in Figure 4, such control requires 8 years of treatments when adherence is 60%, 6 years at 70% and 4 years at 80–90% adherence.) Of note, the duration of control varies strongly depending on each village's pre-intervention level of school age infection prevalence. Where adherence is low, the initially high-prevalence villages very quickly rebound to >10% prevalence and associated risk for disease. Where overall coverage is much higher (90%), every village is projected to have at least 3 years of ‘safety’ below 10% school age prevalence. Independent of adherence levels, the 5 villages with the lowest initial prevalence levels were projected to have at least 2 ‘safe’ years after the suspension of program-based treatment.
Figure 7.
Projected impact of different SCORE treatment strategies on village level prevalence following 4 years intervention.
Large randomized trials of different schedules of anti-schistosomal drug delivery are now underway, under the auspices of the Schistosomiasis Consortium for Operational Research and Evaluation (SCORE) [21]. In this simulation, we examined the likely outcomes of the six different SCORE control strategies on village level prevalence at after 4 years of treatment. School age adherence was set to 70%, and in community-based treatments, adult adherence was assumed to be 50%. C indicates a year of community-based treatment, S indicates a year of school age treatment, and H indicates a ‘holiday’ or non-treatment year. Of note, in the low-prevalence villages, all strategies achieved the goal of reducing village prevalence <10%. Control was obtained in the high prevalence villages only when there were no gaps in treatment, i.e., C-C-C-C, C-C-S-S, or S-S-S-S, with no ‘holidays’ in the schedule.
Figure 8.
Likely impact of different SCORE strategies on long-term suppression of village prevalence of Schistosoma infection.
As in Figure 6, we simulated the duration of long term post-control suppression of Schistosoma infection to safer levels below 10%. In this case, overall school age adherence was estimated at 70%, and comparisons were made between the 3 most successful four-year SCORE strategies, and standard implementation of the WHO strategy (which takes 6 years to achieve area control). The more aggressive, every-year SCORE strategies resulted in more prolonged post-treatment suppression of infection prevalence than the WHO protocols (in which some communities get assigned to every-other-year treatment). This was apparent in the six villages with the lowest pre-treatment prevalences (16–44%). In the villages with highest pre-intervention levels of infection (57–69%), all of the strategies were very limited in their ability to effect post-treatment suppression (0 to 2 years only).