The authors have declared that no competing interests exist.
Analyzed the data: NS. Wrote the paper: NS PV JU. Led the design of the analysis and supervised the implementation: PV. Conceptualized the work and gave critical input: JU.
After many years of neglect, schistosomiasis control is going to scale. The strategy of choice is preventive chemotherapy, that is the repeated large-scale administration of praziquantel (a safe and highly efficacious drug) to at-risk populations. The frequency of praziquantel administration is based on endemicity, which usually is defined by prevalence data summarized at an arbitrarily chosen administrative level.
For an ensemble of 29 West and East African countries, we determined the annualized praziquantel treatment needs for the school-aged population, adhering to World Health Organization guidelines. Different administrative levels of prevalence aggregation were considered; country, province, district, and pixel level. Previously published results on spatially explicit schistosomiasis risk in the selected countries were employed to classify each area into distinct endemicity classes that govern the frequency of praziquantel administration.
Estimates of infection prevalence adjusted for the school-aged population in 2010 revealed that most countries are classified as moderately endemic for schistosomiasis (prevalence 10–50%), while four countries (i.e., Ghana, Liberia, Mozambique, and Sierra Leone) are highly endemic (>50%). Overall, 72.7 million annualized praziquantel treatments (50% confidence interval (CI): 68.8–100.7 million) are required for the school-aged population if country-level schistosomiasis prevalence estimates are considered, and 81.5 million treatments (50% CI: 67.3–107.5 million) if estimation is based on a more refined spatial scale at the provincial level.
Praziquantel treatment needs may be over- or underestimated depending on the level of spatial aggregation. The distribution of schistosomiasis in Ethiopia, Liberia, Mauritania, Uganda, and Zambia is rather uniform, and hence country-level risk estimates are sufficient to calculate treatment needs. On the other hand, countries like Burkina Faso, Mali, Mozambique, Sudan, and Tanzania show large spatial heterogeneity in schistosomiasis risk, which should be taken into account for calculating treatment requirements.
More than 200 million people are affected by the snailborne disease schistosomiasis. The main strategy to control schistosomiasis is to regularly treat school-aged children with the drug praziquantel. The frequency of praziquantel treatment depends on the average prevalence of schistosomiasis, which can be defined as low (prevalence <10%), moderate (10–50%), or high (>50%). However, it remains unclear at which geographical scale these prevalence levels should be considered to avoid unnecessary treatments but still comply with local needs. We investigated the effect of the geographical scale for an ensemble of 29 West and East African countries using previously published model-based schistosomiasis risk estimates obtained at high spatial resolution. These estimates allow spatial risk aggregation at different geographical scales (i.e., country, region, district, or pixel level). More than 70 million praziquantel treatments are required every year for school-aged children if countrylevel estimates are used. On a more refined geographical scale (i.e., province), annualized praziquantel treatments increase by 12%. Depending on the averaged schistosomiasis prevalence and the spatial risk variation across a country, the difference in the estimated amount of praziquantel between country-level aggregation and other geographical scales might be very important, as for example in Burkina Faso, Ghana, and Mali.
Schistosomiasis is a snail-borne parasitic disease, which affects more than 200 million people globally based on 2003 population estimates
The global strategy for the control of schistosomiasis, as recommended by the World Health Organization (WHO), is preventive chemotherapy, that is the repeated, large-scale administration of the antischistosomal drug praziquantel to at-risk populations
The aim of this paper was to assess the effect of the geographical scale of schistosomiasis risk estimates on the amount of praziquantel treatment needed in the school-aged population in selected parts of Africa. In contrast to existing treatment need calculations based on crude country prevalence estimates
In our previous work, we determined the spatial distribution of
The risk predictions of Schur and colleagues were originally summarized as country-specific
Population count data at 1×1 km spatial resolution in Africa for 2008 were obtained from the LandScan global population database (
National preventive chemotherapy programs are usually implemented at specific administrative levels, for example at country, province, and district level. Shape files containing geographical information on the administrative boundaries of these levels were downloaded from the Map Library (
Population-adjusted prevalence at a given administrative region was calculated by summing all infected school-aged children at pixel-level and dividing by the total school-aged population in that region. This calculation takes into account that pixels are not equally populated, and hence do contribute to the area-specific prevalence in the same way.
The aggregated and pixel-level data on the school-aged children and population-adjusted prevalence were converted to the required amount of drugs per year (“annualized treatment needs”) using WHO schistosomiasis control guidelines
For high endemicity areas, annual treatment of all school-aged children (and other high-risk groups, e.g., fishermen) is proposed. With regard to treatment needs in the school-aged population, we therefore estimated the amount of annualized praziquantel treatment needs in these areas to be equal to the number of school-aged children. In areas with moderate endemicity, it is recommended to treat school-aged children every other year. Hence, we consider half of the school-aged population for annual praziquantel treatment in our calculation of annualized needs. Areas of low endemicity warrant treatment of school-aged children twice during primary schooling (on entry and just before leaving school). Assuming an average duration of 6 years in primary school, we estimate annualized praziquantel needs by considering one third of the school-aged population for treatment every year.
Country-specific estimates on the prevalence of schistosomiasis among the school-aged population and the number of infected school-aged children are summarized in
The endemicity levels at country-level (A) and pixel-level (B) are based on previously published geostatistical model-based prevalence estimates
Country | Total number of children aged 5–14 years (×106) | Number of |
Prevalence adjusted to age group 5–14 years (%) | Treatment needs (10% and 50% cut-offs) | Treatment needs (10% and 25% cut-offs) | ||||||
Country-level (×106) | Province-level (×106) | District-level (×106) | Pixel-level (×106) | Country-level (×106) | Province-level (×106) | District-level (×106) | Pixel-level (×106) | ||||
Benin | 2.323 | 0.988 | 42.5 | 1.161 | 1.440 | 1.488 | 1.516 | 2.323 | 2.323 | 2.183 | 2.125 |
Burkina Faso | 4.099 | 1.969 | 48.0 | 2.049 | 2.916 | 2.978 | 3.006 | 4.099 | 4.032 | 4.032 | 4.024 |
Burundi | 1.719 | 0.655 | 38.1 | 0.859 | 0.949 | 1.020 | 1.072 | 1.719 | 1.719 | 1.655 | 1.599 |
Cameroon | 4.703 | 0.996 | 21.2 | 2.352 | 2.052 | 2.175 | 2.253 | 2.352 | 2.657 | 2.724 | 2.731 |
Côte d'Ivoire | 4.621 | 1.775 | 38.4 | 2.311 | 2.835 | 2.790 | 2.908 | 4.621 | 4.254 | 4.179 | 3.870 |
Djibouti | 0.041 | 0.009 | 21.3 | 0.020 | 0.020 | 0.020 | 0.020 | 0.020 | 0.029 | 0.028 | 0.025 |
Eritrea | 0.818 | 0.341 | 41.7 | 0.409 | 0.639 | 0.547 | 0.563 | 0.818 | 0.808 | 0.746 | 0.727 |
Ethiopia | 14.800 | 3.745 | 25.3 | 7.394 | 7.394 | 7.687 | 7.639 | 14.800 | 9.509 | 10.900 | 10.500 |
The Gambia | 0.424 | 0.061 | 14.4 | 0.212 | 0.182 | 0.184 | 0.189 | 0.212 | 0.182 | 0.199 | 0.215 |
Ghana | 5.395 | 2.945 | 54.6 | 5.395 | 4.676 | 4.618 | 4.470 | 5.395 | 5.395 | 5.395 | 5.333 |
Guinea | 2.456 | 1.027 | 41.8 | 1.228 | 1.664 | 1.588 | 1.571 | 2.456 | 2.456 | 2.375 | 2.285 |
Guinea-Bissau | 0.374 | 0.078 | 20.9 | 0.187 | 0.187 | 0.185 | 0.184 | 0.187 | 0.240 | 0.233 | 0.239 |
Kenya | 6.881 | 2.682 | 39.0 | 3.441 | 3.989 | 4.484 | 4.492 | 6.881 | 6.533 | 6.126 | 5.827 |
Liberia | 0.779 | 0.469 | 60.2 | 0.779 | 0.770 | 0.749 | 0.718 | 0.779 | 0.779 | 0.779 | 0.777 |
Malawi | 2.692 | 1.276 | 47.4 | 1.346 | 1.980 | 1.725 | 1.862 | 2.692 | 2.692 | 2.692 | 2.665 |
Mali | 3.577 | 1.758 | 49.1 | 1.789 | 2.814 | 2.680 | 2.654 | 3.577 | 3.566 | 3.488 | 3.415 |
Mauritania | 0.792 | 0.218 | 27.5 | 0.396 | 0.396 | 0.403 | 0.405 | 0.792 | 0.601 | 0.604 | 0.589 |
Mozambique | 4.264 | 2.495 | 58.5 | 4.264 | 3.748 | 3.686 | 3.660 | 4.264 | 4.264 | 4.260 | 4.236 |
Niger | 3.944 | 0.851 | 21.6 | 1.972 | 1.835 | 1.844 | 1.863 | 1.972 | 2.561 | 2.540 | 2.472 |
Nigeria | 35.300 | 15.000 | 42.5 | 17.700 | 22.800 | 23.100 | 23.000 | 35.300 | 31.500 | 32.200 | 32.000 |
Rwanda | 1.872 | 0.627 | 33.5 | 0.936 | 0.936 | 1.067 | 1.064 | 1.872 | 1.584 | 1.636 | 1.575 |
Senegal | 3.112 | 0.567 | 18.2 | 1.556 | 1.403 | 1.421 | 1.451 | 1.556 | 1.694 | 1.846 | 1.741 |
Sierra Leone | 1.553 | 0.921 | 59.3 | 1.553 | 1.411 | 1.427 | 1.357 | 1.553 | 1.553 | 1.553 | 1.553 |
Somalia | 1.770 | 0.685 | 38.7 | 0.885 | 1.006 | 1.030 | 1.019 | 1.770 | 1.657 | 1.690 | 1.649 |
Sudan | 7.652 | 2.657 | 34.7 | 3.826 | 4.117 | 4.358 | 4.621 | 7.652 | 7.179 | 6.625 | 6.279 |
Tanzania | 7.396 | 2.467 | 33.4 | 3.698 | 4.089 | 4.269 | 4.289 | 7.396 | 5.792 | 5.796 | 5.689 |
Togo | 1.327 | 0.535 | 40.3 | 0.663 | 0.943 | 0.865 | 0.862 | 1.327 | 1.123 | 1.208 | 1.161 |
Uganda | 6.315 | 1.309 | 20.7 | 3.158 | 3.158 | 3.113 | 3.160 | 3.158 | 3.866 | 3.942 | 3.700 |
Zambia | 2.225 | 0.576 | 25.9 | 1.112 | 1.112 | 1.153 | 1.159 | 2.225 | 1.651 | 1.650 | 1.599 |
|
|
|
|
|
|
|
|
|
|
|
|
The estimates are based on the median schistosomiasis risk of the posterior predictive distribution for 29 West and East African countries. Treatment needs were calculated based on previously published geostatistical model-based prevalence estimates
Country | Treatment needs (10% and 50% cut-offs) | Treatment needs (10% and 25% cut-offs) | ||||||
Country-level (×106) | Province-level (×106) | District-level (×106) | Pixel-level (×106) | Country-level (×106) | Province-level (×106) | District-level (×106) | Pixel-level (×106) | |
Benin | 1.161–2.323 | 1.161–2.172 | 1.184–2.215 | 1.166–2.170 | 2.323–2.323 | 1.877–2.323 | 1.612–2.323 | 1.404–2.323 |
Burkina Faso | 2.049–4.099 | 2.188–4.032 | 2.158–4.078 | 2.079–4.071 | 4.099–4.099 | 3.827–4.099 | 3.733–4.099 | 2.865–4.099 |
Burundi | 0.859–1.719 | 0.859–1.719 | 0.854–1.713 | 0.779–1.704 | 1.719–1.719 | 1.065–1.719 | 1.004–1.719 | 0.866–1.719 |
Cameroon | 2.352–2.352 | 2.052–3.017 | 2.009–2.968 | 1.888–3.021 | 2.352–4.703 | 2.657–3.126 | 2.611–3.259 | 2.186–3.408 |
Côte d'Ivoire | 2.311–2.311 | 2.311–3.982 | 2.302–4.127 | 2.146–4.326 | 4.621–4.621 | 3.267–4.621 | 3.205–4.621 | 2.411–4.569 |
Djibouti | 0.020–0.041 | 0.018–0.029 | 0.017–0.031 | 0.015–0.033 | 0.020–0.041 | 0.018–0.041 | 0.017–0.040 | 0.015–0.040 |
Eritrea | 0.409–0.818 | 0.409–0.818 | 0.409–0.818 | 0.409–0.818 | 0.409–0.818 | 0.409–0.818 | 0.409–0.818 | 0.409–0.818 |
Ethiopia | 7.394–7.394 | 7.318–8.864 | 7.260–9.605 | 6.171–13.600 | 14.800–14.800 | 9.509–14.800 | 8.578–14.800 | 6.313–14.700 |
Gambia, The | 0.212–0.212 | 0.161–0.212 | 0.160–0.212 | 0.157–0.216 | 0.212–0.212 | 0.161–0.324 | 0.160–0.310 | 0.157–0.338 |
Ghana | 2.697–5.395 | 2.697–5.395 | 2.677–5.362 | 2.713–5.347 | 5.395–5.395 | 5.395–5.395 | 4.664–5.395 | 4.041–5.395 |
Guinea | 1.228–2.456 | 1.228–2.269 | 1.252–2.375 | 1.170–2.379 | 2.456–2.456 | 1.964–2.456 | 1.715–2.456 | 1.397–2.454 |
Guinea-Bissau | 0.187–0.187 | 0.160–0.212 | 0.159–0.232 | 0.142–0.246 | 0.187–0.374 | 0.160–0.374 | 0.159–0.368 | 0.143–0.365 |
Kenya | 3.441–3.441 | 3.441–5.371 | 3.562–5.792 | 3.083–5.728 | 6.881–6.881 | 6.233–6.881 | 5.156–6.719 | 3.691–6.804 |
Liberia | 0.389–0.779 | 0.435–0.779 | 0.420–0.779 | 0.403–0.778 | 0.779–0.779 | 0.779–0.779 | 0.711–0.779 | 0.599–0.779 |
Malawi | 1.346–2.692 | 1.346–2.692 | 1.424–2.692 | 1.378–2.672 | 2.692–2.692 | 2.692–2.692 | 2.583–2.692 | 1.925–2.692 |
Mali | 1.789–3.577 | 2.077–3.478 | 1.987–3.488 | 1.755–3.432 | 3.577–3.577 | 3.478–3.577 | 3.223–3.577 | 2.468–3.577 |
Mauritania | 0.396–0.396 | 0.362–0.601 | 0.367–0.612 | 0.313–0.580 | 0.792–0.792 | 0.488–0.792 | 0.412–0.792 | 0.319–0.789 |
Mozambique | 4.264–4.264 | 2.966–4.264 | 2.575–4.252 | 2.514–4.261 | 4.264–4.264 | 4.088–4.264 | 3.825–4.264 | 3.361–4.264 |
Niger | 1.972–1.972 | 1.835–2.409 | 1.765–2.498 | 1.561–2.495 | 1.972–3.944 | 2.561–3.133 | 2.287–3.222 | 1.776–3.437 |
Nigeria | 17.700–35.300 | 17.700–32.600 | 17.000–33.400 | 16.600–34.300 | 35.300–35.300 | 29.300–35.300 | 23.500–35.300 | 19.500–35.300 |
Rwanda | 0.936–1.872 | 0.897–1.584 | 0.890–1.714 | 0.797–1.701 | 0.936–1.872 | 1.130–1.872 | 1.024–1.872 | 0.855–1.872 |
Senegal | 1.556–1.556 | 1.403–1.507 | 1.326–1.537 | 1.255–1.772 | 1.556–1.556 | 1.694–2.125 | 1.707–2.234 | 1.430–2.216 |
Sierra Leone | 0.776–1.553 | 0.776–1.553 | 0.776–1.553 | 0.783–1.553 | 1.553–1.553 | 1.411–1.553 | 1.427–1.553 | 1.170–1.553 |
Somalia | 0.885–0.885 | 0.885–1.621 | 0.872–1.654 | 0.802–1.756 | 1.770–1.770 | 1.467–1.770 | 1.178–1.770 | 0.872–1.767 |
Sudan | 3.826–3.826 | 3.826–5.217 | 3.997–6.065 | 3.222–7.468 | 7.652–7.652 | 7.179–7.652 | 5.881–7.652 | 3.296–7.646 |
Tanzania | 3.698–3.698 | 3.879–5.301 | 3.880–5.506 | 3.321–6.114 | 7.396–7.396 | 5.251–7.388 | 4.979–7.237 | 3.791–7.258 |
Togo | 0.663–1.327 | 0.663–1.123 | 0.663–1.110 | 0.626–1.137 | 1.327–1.327 | 1.123–1.327 | 0.943–1.327 | 0.731–1.298 |
Uganda | 3.158–3.158 | 3.158–3.158 | 3.026–3.181 | 2.798–4.390 | 3.158–6.315 | 3.866–5.492 | 3.633–5.443 | 2.960–5.971 |
Zambia | 1.112–1.112 | 1.112–1.474 | 1.055–1.519 | 0.906–1.778 | 2.225–2.225 | 1.651–2.225 | 1.248–2.190 | 0.926–2.196 |
|
|
|
|
|
|
|
|
|
The estimates are based on 50% confidence intervals of the posterior predictive distribution for 29 West and East African countries. Treatment needs were calculated based on previously published geostatistical model-based prevalence estimates
Over 5 million annual praziquantel treatments for the school-aged population are required in Ethiopia, Ghana, and Nigeria, while less than half a million treatments are needed for preventive chemotherapy in Djibouti, Eritrea, The Gambia, Guinea-Bissau, and Mauritania. These calculations are based on country-wide risk estimates and they vary when risk estimates are available at higher level of disaggregation.
Considering treatment in entire communities in high endemicity areas and 20% of the non-school-aged population in moderately endemic areas (as recently considered by WHO
Next, selected country examples are given regarding annualized praziquantel treatment needs in the school-aged population. These examples highlight that the required amount of praziquantel depends on the level of aggregation of the schistosomiasis risk estimates, and is particularly pronounced in countries where the distribution of schistosomiasis shows high focality.
Provincial (A), district (B), and pixel-level (C) endemicity and pixel-level prevalence (D). The country-specific prevalence in Ghana is 54.6% based on previously published geostatistical model-based estimates
Endemicity classes at province (called regions) and district level (called provinces) in Burkina Faso are shown in
Provincial (A), district (B), and pixel-level (C) endemicity and pixel-level prevalence (D). The country-specific prevalence in Burkina Faso is 48.0% based on previously published geostatistical model-based estimates
The estimated country-specific prevalence of schistosomiasis in school-aged children in Zambia is 25.9%, which classifies the country as moderately endemic. A switch from country to province level revealed no changes in endemicity class (
Provincial (A), district (B), and pixel-level (C) endemicity and pixel-level prevalence (D). The country-specific prevalence in Zambia is 25.9% based on previously published geostatistical model-based estimates
Among the 29 countries included in the current analysis, Senegal has the second lowest national schistosomiasis prevalence in school-aged children (18.2%; moderately endemic).
Provincial (A), district (B), and pixel-level (C) endemicity and pixel-level prevalence (D). The country-specific prevalence in Senegal is 18.2% based on previously published geostatistical model-based estimates
There are ongoing discussions within WHO and other organizations and consortia whether the threshold to separate between moderate and heavy schistosomiasis endemicity should be halved from 50% to 25%. Of note, the lower threshold is already adopted by the Schistosomiasis Consortium for Operational Research and Evaluation (SCORE) in the frame of large-scale programs that aim at gaining (prevalence ≥25%) and sustaining (prevalence 10–24%) schistosomiasis control (
Preventive chemotherapy is the current mainstay for morbidity control due to schistosomiasis and other helmintic diseases
Contrary to what we expected, the total amount of praziquantel treatments is higher at a smaller administrative unit of schistosomiasis prevalence aggregation (i.e., district level
Utzinger et al. (2009)
In early 2012, WHO released new estimates on the total number of individuals requiring preventive chemotherapy with praziquantel
Several issues warrant discussion regarding our geostatistical model-based analysis and its implication for preventive chemotherapy. First, human, financial, and technical resources for the control of neglected tropical diseases, including schistosomiasis, have increased considerably in recent years
Our praziquantel treatment needs were calculated based on schistosomiasis risk estimates summarized at different scales. Administrative boundaries were used because large-scale preventive chemotherapy programs are usually planned and conducted at specific administrative levels (e.g., district). However, given the strong focality of schistosomiasis, there is strong heterogeneity within administrative boundaries. For instance, villages in close proximity to freshwater bodies, where intermediate host snails proliferate, are often highly endemic for schistosomiasis, while villages farther away (but in the same administrative region) might have prevalence levels below 10%, even though the overall endemicity in the considered administrative unit is moderate. Hence, stratification of areas should preferably be based on the geography and suitable freshwater bodies (e.g., ecozones), which is the recommended strategy by WHO
Large parts of Africa are co-endemic for several neglected tropical diseases
Considering that, on average, three tablets of praziquantel are needed to treat a school-aged child for schistosomiasis, more than 210 million tablets would need to be administered every year to this population group in the 29 East and West African countries considered here. The distribution of such large amounts of praziquantel is a formidable challenge, especially in remote rural areas. It will require strong health systems and good supply strategies, which are currently lacking in large parts of Africa. Even though the price of a single tablet of praziquantel is now below US$ 0.10
With the transition from morbidity control to transmission control of schistosomiasis and other helminthic diseases, preventive chemotherapy alone is unlikely to achieve this goal
In conclusion, the geographical scale to estimate schistosomiasis risk should take into consideration the expected country-specific prevalence level and the diversity of schistosomiasis risk in the country in order to accurately calculate praziquantel treatment needs. In countries with rather homogeneous schistosomiasis distribution, it would be sufficient to implement control strategies based on schistosomiasis prevalence aggregated at country level. However, in countries with a highly focal distribution of schistosomiasis, province- or even district-level calculations of prevalence have to be considered in order to provide sufficient amounts of treatments to those at risk of morbidity and to reduce the costs in areas requiring less frequent rounds of preventive chemotherapy. The work reported here can be readily adapted to other neglected tropical diseases that have also endorsed preventive chemotherapy as the strategy of choice for morbidity control. In addition, it will be interesting to repeat our analysis in the frame of ongoing large-scale control programs, which are constantly changing prevalence levels and disease, and hence treatment needs.
Ideas to pursue this analysis stem from an informal consultation on schistosomiasis control that took place at WHO headquarters in Geneva between March 30 and April 1, 2011. We are grateful for continued discussions and most valuable input by Dr. Lester Chitsulo.