An observational analysis of the impact of indoor residual spraying with non-pyrethroid insecticides on the incidence of malaria in Ségou Region, Mali: 2012–2015

Over the past 15 years global efforts to control and prevent malaria have produced some astounding results. Estimates indicate that between 2000 and 2015 worldwide malaria incidence has declined almost 40% in areas at risk (from 146 cases per 1000 people per annum to 91 cases per 1000 people per annum) and malaria mortality rates have decreased by almost 60% in areas at risk (from 47 deaths per 100,000 people per annum to 19 deaths per 100,000 people per annum) [1]. This is reflective of worldwide progress in malaria that has come about, in part, through implementation of a package of standard evidence-based interventions of which malaria vector control has been a central component [1, 2]. Indeed, Bhat and others recently estimated that 81% of all malaria cases averted from 2000 to 2015 could be attributed to successful vector control interventions, namely the large-scale distribution and use of insecticide-treated nets (including long-lasting insecticidal nets, or LLINs) and the indoor residual spraying (IRS) of insecticides inside homes [3]. Even in Mali, where malaria control efforts are challenged by complex and intense transmission cycles as well as political instability, IRS and LLINs are thought to have contributed (along with improved malaria case management) to decreasing trends in all cause child mortality observed over the last 10 years [2, 4, 5].

While vector control is recognized as an integral part of the present and future global malaria control and elimination effort [6‐8], the continued success of currently available tools is threatened by the rapid emergence and spread of insecticide resistance in key mosquito populations [9, 10]. Particularly worrisome is resistance to pyrethroids, the most commonly used class of insecticide for malaria vector control (and currently the only class available on LLINs), which is now widespread and found throughout most of sub-Saharan Africa [1]. Accordingly, WHO and others have recognized an urgent need to develop next-generation, non-pyrethroid insecticide products in order to preserve our ability to utilize IRS and LLINs for malaria prevention [2, 10].

Despite this need, widespread adoption of new products for IRS is hampered by several complex factors among which are concerns about increased costs as well as gaps that exist in the evidence base needed to evaluate the impact of newer products in various malaria transmission settings and in combination with other malaria control interventions. Nonetheless, there are a few currently available pyrethroid alternatives recommended by WHO for IRS targeting malaria vectors. Two of the most widely available products include a wettable powder (WP) formulation of bendiocarb, a carbamate insecticide, and a micro-encapsulated formulation of pirimiphos-methyl (PM CS), an organophosphate insecticide [11], both of which have been used by the US President’s Malaria Initiative (PMI) Africa Indoor Residual Spraying (AIRS) Project in recent years [12, 13]. While the insecticidal efficacy of bendiocarb has been well established experimentally [14, 15], and there are indications that it can have a positive impact in reducing malaria transmission in areas with pyrethroid-resistant mosquitoes [16‐19], significant concerns about poor residual efficacy [20‐22] and emerging resistance [23, 24] have recently led to a shift in PMI/AIRS operations away from spraying bendiocarb and towards the use of PM, both throughout Africa in general [12] and in Mali specifically [25]. However, wider-spread adoption and scale-up of PM has been hampered by its higher cost. To help meet this challenge, the Next Generation IRS (NgenIRS) project has introduced a co-payment mechanism for PM as part of a broad market-shaping effort for third generation IRS (3GIRS) products that includes wide-ranging impact analyses. By definition, a 3GIRS product is one that is designed to be effective against pyrethroid-resistant mosquitoes and have an indoor residual efficacy of at least 6 months.

Here, an observational, retrospective (2012–2015) time-series analysis of the seasonal epidemiological impact of IRS with bendiocarb WP and PM CS in the Ségou Region (Ségou) of central Mali is presented to: (1) help further expand the evidence base needed to evaluate the public health impact and cost effectiveness of IRS campaigns utilizing non-pyrethroid, including 3GIRS products; and (2) help foster the use of quality-assured and validated routine malaria surveillance data for evidence-based decision making.

The success that has been achieved in reducing the global burden of malaria over the past 15 years is impressive, and should engender enthusiasm about the prospects for continued progress. Clearly though, the malaria control landscape is becoming increasingly complex and the need for new tools and further investment is becoming increasingly urgent. Presented here are results of a set of observational studies that show that IRS campaigns in the Ségou Region of central Mali likely prevented at least 280,000 cases of all-ages malaria from 2012 to 2014 at a cost of around 7.00 USD per person protected, with additional evidence that discontinuing IRS in Bla District in 2015 likely contributed to the sharp increase in the incidence of malaria observed that year. Collectively, this work supports both the public health utility and the cost-effectiveness of IRS with two different non-pyrethroid insecticides in communities with high rates of LLIN access and use as well as documented pyrethroid resistance.

Important limitations of this study include its observational nature and a lack of supportive datasets needed to control for potential confounding or co-variable factors or fundamental differences between intervention and non-intervention areas. Even though the current analysis is strengthened by the fact that the IRS and non-IRS districts of Ségou are similar ecologically and demographically and matched in time, specific limitations include the inability to adequately quantify the uncertainty around the proportion of reduced case numbers observed in IRS districts is attributable to the intervention, and not to chance or other factors. Additional limitations include reliance on malaria RDT results from patients who passively reported to the health system, which do not account for issues of patient access to health services that are likely to disproportionately affect more rural facilities, as well as the different geographical and temporal resolutions at which various data sets were available in Ségou. Although seasonal spray status and monthly malaria incidence rates were available at the health facility level, entomological, rainfall and SMC data were limited to the district and/or regional levels, preventing some more detailed analyses (e.g., multivariate regression, ANOVA, etc.) that could prove valuable in the future as various surveillance capacities grow and become more aligned. The basic impact model built to estimate the effect of changing the IRS campaign start date assumes that the magnitude of the monthly effect of IRS relative to the beginning of spray activities is independent of timing against the beginning of the rains (and therefore also independent of overall mosquito abundance), an assumption that is largely untested and likely to differ by active ingredient and target vector species. It is possible that IRS could be more or less effective as mosquito populations and environmental conditions fluctuate seasonally, though how this would impact overall disease transmission is uncertain. While this uncertainty is not likely to change the interpretation of the crude model presented here, future work could address these questions and help refine more detailed models of seasonal vector control impact.

Despite these limitations, the work presented here helps paint a picture of how IRS with non-pyrethroid insecticides has had a positive public health impact in central Mali by reducing malaria incidence rates, supporting similar positive results reported from transmission settings in Uganda [19, 44], Sudan [45], Tanzania [46], Zimbabwe [47], and Zambia [48], but contrasting less encouraging results from Benin [49] and The Gambia [50] that showed no benefit to IRS on top of existing interventions. It is clear (Fig. 2) that the IRS campaigns of 2012, 2013 and 2014 each had a sizable public health impact in Ségou by substantially reducing malaria transmission. While the incidence reductions observed after spraying bendiocarb WP in 2012 (320 fewer cases per 10,000 person-months) and 2013 (290 fewer cases per 10,000 person-months) were more similar than the reduction observed in 2014 after spraying PM CS (which resulted in 490 fewer cases per 10,000 person-months, around a 60% greater cumulative incidence reduction), it should be noted that malaria incidence was particularly high throughout Mali in 2014 and there is no indication of how well bendiocarb may have performed in this context. Any impact of having switched active ingredients is difficult to interpret given the inherent differences between the malaria transmission seasons of 2013 and 2014. Indeed, the crude impact modelling that was done here seems to indicate that the impact of the two chemicals was relatively similar in terms of their monthly per cent reductions in incidence compared to baseline (i.e., comparable unsprayed districts) by month: coefficients of variation were similar across years whether or not there was a shift in active ingredient (Additional file 2).

The overall effectiveness of IRS in Ségou is further evidenced by the observation that suspension of IRS operations in Bla District in 2015 was associated with a significant increase (70%) in malaria incidence that year, while no increase was observed in neighbouring Barouéli District, where IRS operations with Actellic were consistent across both years (Fig. 4). Interestingly, this increase in malaria transmission associated with the suspension of IRS from Bla is evident even though the district-wide SMC campaigns achieved high coverage there in both years as well as in Barouéli in 2015. In this context, and given recent indications that SMC campaigns can significantly reduce malaria incidence in western Africa [51‐54], these results from Bla could suggest a compelling argument for the simultaneous implementation of IRS and SMC strategies in high burden areas with good LLIN coverage.

Also apparent in the epidemiological curves from Ségou is that the timing of the IRS campaigns, which began in late July and early August, did not effectively target the first few weeks of increased transmission following the very beginning of the rainy seasons, which began in May (Fig. 2). While the planning of IRS operations is an extraordinary complex decision (that in Mali was informed by concerns about the length of residual efficacy of bendiocarb [33, 41]), it is nonetheless intriguing to speculate whether moving the IRS campaign start dates to an earlier month could have helped reduce these early season cases and, therefore, prevented more malaria overall. However, the model described here does not support this, with the increased coverage during the beginning of the rainy season achieved by spraying earlier counter-balanced by the fact that the period of maximum IRS impact (40–47% reduction) lasted for only 2 months after the end of the spray campaigns; interestingly, this was consistent across all years, regardless of which IRS product was used (Additional files 1, 2). When spray campaigns ended in August, these 2 months of maximum impact (September and October) aligned with the 2 months of otherwise peak transmission, maximizing the effect of the intervention in terms of total cases of malaria prevented. Although the 2012–2014 campaigns in Ségou appear to have been timed well to achieve maximum impact, it is worth noting that the model described here suggests that even with less-than-perfect timing, IRS can still have a substantial, cost-effective impact on reducing malaria burden, at least when erring on the side of spraying too early as opposed to spraying too late.

The estimates of cost-effectiveness calculated here are basic, and meant only to give a rough idea about the financial implications of IRS campaigns that may require the use of newer, more expensive insecticides in the face of resistance concerns. It is nonetheless encouraging that these estimates indicate that spray campaigns with non-pyrethroid, including 3GIRS, products in central Mali were cost-effective investments, ranging from 6.08 to 7.40 USD per person protected and 23.33 to 42.59 USD per case averted, regardless of which active ingredient was used. It is also encouraging that these estimates did not fluctuate much, but it should be stressed that the target vectors in Mali were fully susceptible to both chemicals throughout the study period and changes in resistance to a product or chemical class will likely reduce its efficacy, and therefore its cost-effectiveness.

It is also worthwhile to consider how these epidemiological trends fit with the entomological trends previously reported by PMI/AIRS [28, 33, 41, 42] (Fig. 5). Although the entomological surveillance datasets are of different spatial and temporal resolutions, and parallels with the incidence data should therefore be interpreted with caution, the PMI/AIRS reports have shown that IRS successfully controlled or reduced mosquito densities at IRS sentinel sites in Ségou, compared to the densities observed at the non-IRS sentinel site, results that support the epidemiological observations made here. The entomological trends held true for indoor resting mosquito densities collected via pyrethrum spray catch, which were 85% lower in IRS areas than control areas 2 months after the spray campaign in 2013 and 79% lower in 2014, and indoor biting rates calculated from human landing collections (HLCs), which were 45% lower in IRS areas than control areas 2 months after the spray campaign in 2013 and 73% lower in 2014 (Fig. 5). Interestingly, outdoor biting rates were also much lower at IRS sentinel sites than the control site, 80% lower after IRS in 2013 and 91% in 2014 (Fig. 5). This trend of higher malaria incidence rates corresponding to higher mosquito densities is also evident in Bla in 2015. After the cessation of IRS following the 2014 campaign, the 70% increase in malaria incidence observed in 2015 corresponds to a more than fivefold increase in peak (September) indoor HLC An. gambiae sensu lato (s.l.) densities (from 3 bites/person-night in 2014 to 2016 bites/person-night in 2015) (Additional file 3).

It is noteworthy that the waning residual efficacy of IRS, as measured using standard wall bioassays, corresponded to reduced entomological effect but not to reduced malaria prevention (Fig. 5). In 2013, the bendiocarb WP residual efficacy fell below the 80% wall test mortality threshold after just 1 month post-spray and mosquito densities returned to control levels by October, but the intervention appeared to contribute to a reduced malaria incidence for around 3 more months, until at least January (Fig. 5). After spraying PM in 2014, even though wall bioassay mortality stayed above 80% for a full 3 months and again maintained entomological control at the sentinel sites until the final collection, a similar duration of reduced malaria incidence was observed. There are many factors likely to influence these observations, among which are the natural lag time between the observed entomological and epidemiological effects of a malaria control intervention because of the external incubation period required, as well as the aforementioned fact that the disease surveillance data represent true district aggregates (cases reported from all health facilities located throughout an entire district) while the entomological surveillance, though robust, represents a much smaller geographical scale of just a few sentinel sites. Regardless, these observations highlight the need to develop better entomological indicators of potential epidemiological impact for various vector control interventions. Additionally, the data summarized here indicate that the residual efficacy of PM in central Mali was shorter than has been reported elsewhere [55‐59], and additional assessments are underway to try to assess if regional climatic and/or housing material factors may influence product performance.

Even when interpreted with appropriate caution, the observational analyses presented here indicate that the annual IRS campaigns from 2012 to 2015 in the Ségou Region of Mali were good public health investments. Around half a million people were protected for 3 years at a rough cost of around 7.00 USD or less per person per year regardless of which product was sprayed—similar to recent estimates of the cost of distributing LLINs to school children (around 5.50 USD per child [60]) or delivering community-based SMC (around 4.20 USD per child [54]) in Mali. These same IRS campaigns were also broadly associated in time and space with rapid decreases in malaria incidence and overall An. gambiae indoor resting and biting densities. It is encouraging that, despite the obvious need for an expanded IRS toolkit, campaigns with currently available non-pyrethroid products can still have a sizable, cost-effective impact in areas of high resistance and widespread use of LLINs. Nonetheless, the context of these analyses is relatively specific to central Mali, and further impact and cost-effectiveness studies in other regions and in other transmission settings would help enhance an understanding of how and when to maximize the impact of various vector control interventions.

Taken together, this work supports the growing evidence that shows that IRS with non-pyrethroid insecticides, including 3GIRS products, is a wise public health investment, especially in areas with documented pyrethroid resistance and high rates of LLIN coverage and where house structures and population densities are appropriate. Additionally, this work highlights the critical importance of quality-assured validated routine surveillance and well defined observational analyses to assess the impact of malaria control interventions in operational settings, helping to empower evidence-based decision making and to grow the evidence base needed to better understand when and where to utilize new vector control tools as they become available.