Background
Sub-Saharan Africa, where 94% of the global malaria burden occurs, has seen a substantial reduction in cases and deaths due to malaria over the past two decades [
1]. While this reduction is primarily due to an increase in investment and expansion of interventions such as insecticide-treated nets (ITNs), indoor residual spraying (IRS), and diagnosis and treatment, urbanisation also has played a part.
Africa has experienced rapid urbanisation in recent years, rising from 31.5% of the population living in urban areas in 1990 to 42.5% in 2018. By 2050, approximately 60% of the population is expected to live in urban areas [
2]. Planned urbanisation will likely have a positive effect on reducing the malaria burden in Africa, as urban areas typically experience substantially lower rates of malaria transmission than rural areas [
3]. Primarily, this is thought to be due to improved housing and the reduced availability of suitable larval habitats for African
Anopheles vector species [
4].
The protective effect conferred by urbanisation may be partially lost with the invasion and establishment of
Anopheles stephensi.
An. stephensi is found throughout South Asia, where it is capable of transmitting both
Plasmodium falciparum and
P. vivax parasites [
5] in a diverse set of habitats, from rural to highly urban settings [
6]. The success of this vector in urban locations is due to its ability to utilise water tanks, wells and other artificial containers as larval habitats [
7,
8]. Furthermore, it has shown substantial resistance to water pollution [
9]. In the last decade,
Anopheles stephensi has been discovered outside of its traditional endemic region in Asia and was first detected in Djibouti in 2012 [
10].
Prior to 2013, Djibouti had reported less than 3000 cases of malaria per year. Following the year of initial detection of
An. stephensi, cases have increased substantially, and in 2019, there were 49,402 confirmed cases of malaria [
11]. While causation has not been established between increasing
An. stephensi detection and malaria incidence in Djibouti, it has been heavily implicated [
12].
Anopheles stephensi has now been found in Sudan, Somalia and Ethiopia [
13‐
17].
Here, we quantify the potential impact of
An. stephensi invasion in Djibouti on malaria transmission in order to make projections about what could happen in Ethiopia, where the species has been found at numerous sites and is spreading [
18]. Translating the invasion of
An. stephensi to its public health impact is difficult due to uncertainty in its vectorial capacity and how public health entities and governments will respond. Different illustrative scenarios are investigated, exploring the public health impact of different vector control interventions.
Discussion
The invasion and establishment of Anopheles stephensi represents an imminent and substantial threat to the Horn of Africa and wider region which could jeopardise progress achieved in malaria control. The modelled increase in malaria is highly uncertain, but without additional interventions, the impact could be considerable, with an estimated ~50% increase from the reported ~740,000 P. falciparum malaria cases in Ethiopia, 2019, to an estimated 1,130,000 (95% CI 843,000–1,404,000) after establishment and disease equilibrium has been reached. This assumes that malaria cases only increase in areas previously predicted as suitable for A. stephensi establishment located under 2000m. Numbers could be substantially higher if invasion is more widespread.
Sub-nationally, significant heterogeneity in public health impact is expected. Analysis suggests that low altitude urban areas at pre-existing low levels of malaria transmission may experience the largest increases in disease burden. In these areas, the absence of existing vector control and low human population immunity indicates the possibility for substantial increases in transmission. These areas are also the most uncertain, with the model predicting the increase that clinical cases could be negligible to more than that observed in areas where malaria has historically been widespread. Our finding that areas with low pre-existing levels of transmission (much of Ethiopia) take substantially longer to see increases in malaria after
A. stephensi introduction is particularly worrying. In these locations, in the absence of widespread and routine surveillance for
A. stephensi, the first signal detected could be an increase in malaria, which would occur a considerable amount of time after mosquito establishment. While this would not be of significant concern if the vector was easily removed or only led to a relatively minor and easily combatable increase in malaria, our model findings and what has been observed in Djibouti [
11,
12] suggest that this is not the case. Once
A. stephensi becomes established, it could lead to significant increases in malaria transmission that are not easily reversed. This unnoticed proliferation and subsequent increase in transmission were previously seen in
Anopheles arabiensis in North-Eastern Brazil, where its ‘silent spread’ led to a large malaria epidemic [
30,
61]. Nevertheless, this work shows that the absence of an increase in reported malaria cases following the identification of the presence of
A. stephensi should not be overly interpreted.
To combat the possible increase of malaria transmission following
An. stephensi establishment, the deployment of a wide array of vector control interventions should be considered. High levels of pyrethroid resistance observed in
A. stephensi captured in Ethiopia [
17,
18] suggest that pyrethroid-only ITNs (already in use across the country) will have a limited efficacy for control of malaria transmitted by
A. stephensi. Adoption of PBO-ITNs is predicted to be both highly impactful and cost-effective. The impact of different control interventions is unclear given the receptiveness of
A. stephensi in this new environment is unquantified. For example, if the mosquito feeds at night and rests inside structures sprayed with IRS, then the widespread use of this intervention alone could be sufficient to mitigate the public health impact (Additional file
1 Section 4). However, if the mosquito was less amenable to indoor vector control, then layering of ITNs and IRS may be insufficient. Within its endemic range,
A. stephensi shows a propensity to both crepuscular biting and resting outside of houses compared to African
Anopheles species, suggesting a reduced efficacy if it behaves as it does in its endemic range [
30] (Additional file
1 Section 2). Compounding this, high-density urban locations, where large-scale vector-control campaigns have been historically absent, will present a challenge for establishing ITN access and use as well as achieving sufficient IRS coverage. While
A. stephensi is primarily known as an urban vector of malaria, it is found in both urban and rural settings across its endemic range [
30] and in Ethiopia [
18], and so there is the potential for its impact on malaria transmission to be found across the country. Differences in environment, housing, culture, human and vector behaviour in urban and rural settings are likely to result in very different public health outcomes, even before considering the logistics of intervention deployment. Though these factors are likely to have a substantial effect, it is presently unknown how these combined factors will translate into malaria transmission and control. In the absence of quantified and validated input data, we have not stratified models by urbanicity at this present time. Nevertheless, additional data on
A. stephensi bionomics and the impact of vector control interventions, stratified across urban/rural areas and housing types, should therefore be collected as a matter of urgency to enable model estimates to be refined and contribute to the decision-making process (Additional file
1 Section 4). These compound uncertainties combined with the crude method of economic evaluation adopted here make cost projections highly variable. Though these are substantial ($74.6 million USD ($42.6–$105.4)), the economic burden of an additional ~368,000 (95% CI 103,000–664,000) malaria cases should not be underestimated.
This work was intended to provide initial estimates of the possible scale and impact of
A. stephensi invasion rather than detailed predictions of what will happen. As such, there are many limitations to this approach (further detailed in the Additional file
1 Section 4) and results should not be overly interpreted. The public health impact is likely to be underestimated as we only consider
P. falciparum malaria, despite
A. stephensi having been shown to be capable of efficiently transmitting
P. vivax in Ethiopia [
54]. Due to the presence of additional
Anopheles species in Djibouti, it is unlikely the increases we have seen are purely due to
A. stephensi, which we have assumed, and while highly correlated [
11], and increasingly implicated [
12], the causative role in
A. stephensi in the increases seen in Djibouti has not been established. However, the trends observed along with evidence of underreporting in Djibouti are cause for significant concern [
12]. Malaria burden is related to the extent to which
A. stephensi may invade a region (i.e. the number of mosquitoes per person and not just its presence) and the seasonality of the vector. Malaria transmission is often highly seasonal and an important factor when evaluating local epidemiology and vector control. However, the case of
A. stephensi seasonality is a complicated and currently unquantified unknown. Across its endemic range,
A. stephensi shows a variety of patterns from unimodal to bimodal to peaking in the dry season to the wet [
62], and within Ethiopia, there is a dearth of this data. As we are not modelling seasonally timed interventions, and are concerned with long term rather than intra-annual dynamics, the omission of seasonality in our model does not change model predictions or recommendations, though the quantification of
A. stephensi seasonality in the Horn of Africa should be a matter of urgent study.
Conclusions
While we have made use of available published and unpublished sources on A. stephensi bionomics, there is either insufficient data or substantial intra-species variability to simply ascribe a set of characteristics to how A. stephensi will interact with humans and control interventions in Ethiopia. In order to improve predictions, we have identified priority aspects of A. stephensi bionomics (such as anthropophagy, endophily, indoor biting), local transmission (existing distribution of malaria cases between rural and urban areas) and intervention parameters (efficacy of ITNs, IRS and LSM give a certain effort) that should be explored as a priority in order to inform additional mathematical modelling of the impact of A. stephensi on malaria transmission in Africa (see Additional Section 4).
Though there is substantial uncertainty in what will happen if An. stephensi becomes established in Ethiopia and other locations across Africa, the estimates provided here, and the situation seen in Djibouti, should be a stark warning against complacency and highlight the need to rapidly improve surveillance and evaluate effective control interventions in response to this developing threat.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.