Intensive vector control efforts have led to a dramatic decline in the global malaria burden in the past decade [1]. According to the World Health Organization’s (WHO) World Malaria Report 2015, malaria morbidity has decreased from an estimated 262 million in 2000 to 214 million in 2015, a reduction of 18% over the 15 years [2]. Four major malaria intervention methods are credited for the significant malaria control gains: long-lasting insecticidal nets, indoor residual spraying (IRS), diagnosis and management of malaria cases, and intermittent preventive treatment in pregnancy [2]. However, the emergence and spread of resistance in malaria vectors and parasite populations to insecticides and artemisinin, respectively, is threatening to slow and even reverse the gains made in malaria control. Insecticide resistance in the major malaria vectors, Anopheles gambiae (s.s.), An. arabiensis, An. coluzzii and An. funestus, has been reported from east [3–8], west [3, 9–16], central [3, 17–20] and southern Africa [3, 21–23]. Artemisinin resistance in Plasmodium falciparum has been reported from five countries, all in southeast Asia (Cambodia, the Lao People’s Democratic Republic, Myanmar, Thailand and Vietnam) [2].

If hard-won malaria control gains are to be sustained and further progress made toward the long-term goal of a malaria-free world, malaria control communities need to preserve the efficacy of existing tools even as they pursue innovative methodologies. In 2012, WHO developed and launched the Global Plan for Insecticide Resistance Management in coordination with the Roll Back Malaria partnership, to help guide and improve the planning and implementation of resistance management strategies in malaria-endemic countries [24].

Nearly all of Madagascar’s population is at risk of contracting malaria [25]. Anopheles gambiae (s.s.) is the primary vector on the east and west coasts, where malaria prevalence is stable and transmission lasts longer than six months per year [26]. Anopheles funestus and Anopheles arabiensis are secondary vectors in these coastal areas [25]. In the south sub-desert and central highlands (CHL), where An. funestus is the primary vector and An. arabiensis and An. mascarensis are secondary vectors [27], malaria is seasonal and prone to epidemics. Anopheles mascarensis is implicated as the primary and secondary vector in the South-East sub-desert and Sainte Marie District, respectively [28]. A recent study also implicated An. coustani involvement in the transmission of malaria in the CHL [29].

Malaria vector control with IRS has a long history in Madagascar. IRS with DDT started in 1949 [29]. Impressive results were achieved in the 1970s when this intervention was coupled with malaria case management using chloroquine [30]; malaria transmission was interrupted and An. funestus was eliminated from most parts of the CHL, and malaria morbidity significantly declined even in the perennial transmission areas. After IRS was stopped in 1979, An. funestus gradually re-emerged in the CHL as did an increased occurrence of malaria epidemics in 1986 [30]. In response to the resurgence of malaria transmission in the CHL, in 1993, IRS with DDT was re-introduced [30]. Currently, the two major malaria vector control methods used in Madagascar are IRS and insecticide-treated nets (ITNs) [31]. Madagascar adopted the policy of universal coverage of ITNs in 2008. A malaria indicator survey conducted in 2013 showed that 79% of the target population had access to at least one ITN [31]. A malaria case-control survey conducted in 2013 estimated that IRS and ITNs each provided a protective efficacy of 51% when implemented individually [32]. When the two interventions were implemented in combination, a protective efficacy of 72% was recorded [32]. Any change in the response of malaria vectors to the insecticides used in the two interventions could negatively affect malaria control efforts.

Anopheles gambiae (s.l.) resistance to DDT was first observed in Madagascar in the late 1990s, and appeared to be widespread in the CHL in the early 2000s [33, 34]. Initial evidence of resistance to permethrin was reported, but local populations of An. gambiae (s.l.) remained fully susceptible to lambda-cyhalothrin, deltamethrin, cyfluthrin and alpha-cypermethrin [33, 34], while An. funestus was fully susceptible to DDT and pyrethroids [33, 34]. However, these studies, conducted between 1996 and 2003, were limited in geographical scope to the CHL, and no updated resistance data have been published since 2003, despite the scale-up of insecticide-based vector control in Madagascar in the past decade. More recent studies from other countries have reported rapid changes in resistance patterns following increased use of insecticides for public health and/or agriculture [10–13]. For all these reasons, regular resistance monitoring from nationally representative sites is warranted to inform the planning and implementation of successful vector control and resistance management strategies.

The objective of the current study was to assess and report on the insecticide resistance profiles of three malaria vectors known to transmit the disease in Madagascar: An. gambiae (s.l.), An. funestus and An. mascarensis. The assessment was based on phenotypic resistance and resistance mechanisms. There has been no data published on the efficacy of public health insecticides since 2003; the previous studies covered only one out of five eco-epidemiological zones (the central highlands) and two classes of insecticides (pyrethroids and DDT). The present study will help to fill some of the information gap and for the first time report on the susceptibility of An. mascarensis to pyrethroids and pirimiphos-methyl.

This study’s main objective was to conduct and report on the distribution and frequency of insecticide resistance in the three known malaria vectors in Madagascar, in order to better guide insecticide-based vector control. The study covered three of the country’s five eco-epidemiological zones for An. gambiae (s.l.) and two for An. funestus and An. mascarensis.

Consistent with earlier studies [33, 34], susceptibility test results from this study indicate a high prevalence of An. gambiae (s.l.) resistance to DDT and permethrin relative to the other insecticides tested. Resistance to alpha-cypermethrin and lambda-cyhalothrin was observed in a few sites at low frequencies; no confirmed resistance to deltamethrin was recorded. Rakotondrainble et al. [33] reported no phenotypic resistance in An. gambiae (s.l.) in the CHL to deltamethrin and lambda-cyhalothrin, but they detected resistance to permethrin in one site, Alasora. Ratovonjato et al. [34] detected possible resistance to alpha-cypermethrin in two sites in the CHL. This study indicated suspected resistance to alpha-cypermethrin in two of seven tests, with mortality of 97% in Soavina (Betafo) and 95% in Imerina Imady, and full susceptibility in five of the sites. Hence, there is no evidence to support an increase in frequency of resistance to pyrethroids over the past decade (1996–2015); the frequency of resistance to DDT and pyrethroids appears stable.

As noted above, IRS was done consistently for malaria vector control in Madagascar from 1949 [29] until 1979, and was re-introduced in 1993 [30]; in 2008, the Ministry of Health began scaling up ITN distribution, with coverage quickly reaching 79% of households [31]. Despite the consistent use of these two vector control methods with insecticides that target the sodium channels of insect vectors that can confer resistance to both pyrethroids and DDT, molecular analyses by this study and an earlier one [34] observed no detectable level of kdr mutation. One explanation could be that Madagascar is geographically isolated, making the introduction of vector populations carrying kdr alleles associated with resistance unlikely. It also is possible that local selection pressure from the insecticides might not be enough to maintain naturally occurring, low-frequency mutations in the mosquito population, particularly if carrying the resistant genes comes with some fitness cost. The absence of kdr mutations in turn might totally, or partially, explain why the emergence and spread of DDT and pyrethroid phenotypic resistance in Madagascar has been slower than in most African countries, where there is easier flow of resistant genes between the mosquito populations of neighboring territories. Insecticide selection pressure from agriculture might also be different from mainland Africa.

With the exception of two sites, Imerina Imady and Bekily, the 2015/2016 resistance tests were conducted in sites different from sites used in 2013/2014; while this provides more representative data, it makes it hard to make comparisons between years. For example, in 2013/2014, An. gambiae (s.l.) from Imerina Imady was fully susceptible (mortality ≥ 98%) to alpha-cypermethrin, bendiocarb, DDT, deltamethrin, lambda-cyhalothrin, permethrin and pirimiphos-methyl, whereas in 2015/2016, possible resistance to the pyrethroids alpha-cypermethrin and lambda-cyhalothrin was recorded. The different results may have been due to differences in larval collection locations, timing of collections/testing, species composition or emergence over time of resistant individuals.

With regard to the distribution of members of the An. gambiae complex, only An. arabiensis was found in CHL. Anopheles gambiae (s.s.) was predominant in all east coast study sites except Vavatenina, where An. arabiensis was dominant. Anopheles gambiae (s.s.) was more prevalent in the south sub-desert. These members of the An. gambiae complex may respond differently to the various insecticides. Differences in resistance status between An. gambiae complex members have been seen elsewhere. Ochomo et al. [44] reported high kdr frequency in An. arabiensis (75.9%) compared with An. gambiae (s.s.) (11.1%) in the Bungoma district of Kenya, where the two species live in sympatry, as well as differences in phenotypic resistance. Conversely, in Tanzania, high phenotypic resistance to pyrethroids and DDT and high kdr frequencies were seen in An. gambiae (s.s.), while An. arabiensis populations were fully susceptible to all pyrethroids tested except permethrin with no kdr mutations [4].

In this study, no significant difference in resistance status was noted between An. arabiensis and An. gambiae in Mahambo district (east coast) when tested against alpha-cypermethrin. However, in Bekily (southern Madagascar), where An. arabiensis and An. gambiae larvae were collected from the same breeding sites, a significantly higher proportion of An. arabiensis survivors was observed as compared to An. gambiae when tested against deltamethrin and permethrin (Fisher’s exact test, P < 0.001). Olyset nets (impregnated with permethrin) and PermaNet (impregnated with deltamethrin) were deployed to Bekily in 2010 and 2015, respectively. A previous study in Kenya reported a correlation between ITNs scale-up and increased prevalence of insecticide resistance [45]. It is not clear whether selection pressure from ITNs has played a role in the differential response in the two species observed in this study. However, the bioassay results showed that three rounds of IRS application with a carbamate (bendiocarb WP 80) between 2010 and 2013 and one round of an organophosphate (pirimiphos-methyl 300 CS) in 2014 has not selected a detectable level of resistance to these two classes of insecticides; the vector was fully susceptible to both insecticides. Assessing the extent of use and the contribution of agricultural insecticides to the observed malaria vector resistance in Madagascar was not part of the scope of this study, but it should be included in future studies to better understand the evolution of resistance and to design evidence-based mitigation strategies.

Pre-exposing mosquitoes to synergists PBO, an inhibitor of oxidases, and DEF, an inhibitor of non-specific esterases (NSEs), prior to the bioassay tests restored at least partial susceptibility of An. gambiae (s.l.) to pyrethroids in CDC bottle bioassays, suggesting the two categories of detoxifying enzymes (NSE, and cytochrome P450 monoxygenases) have played an important role in the evolution and shaping of insecticide resistance in Madagascar. With the exception of one site, where only partial restoration of susceptibility to permethrin was recorded, complete pyrethroid resistance suppression was achieved for insecticides tested with pre-exposure to PBO. With pre-exposure to DEF, only partial suppression of resistance to permethrin was achieved in two of three vector populations. Nevertheless, pre-exposure to PBO or DEF fully synergized resistance to deltamethrin and alpha-cypermethrin. Inhibiting one of these two enzymes appeared sufficient to restore the effectiveness of these two insecticides to which resistance had been suspected (not confirmed). However, inhibition of NSE alone was not always sufficient to fully restore susceptibility to permethrin in areas where resistance to this insecticide had been confirmed, indicating the dominant role of mixed function oxidases (MFOs) in An. gambiae (s.l.) resistance to permethrin. Further investigation is needed to learn if MFOs also play crucial role in vector resistance to deltamethrin and alpha-cypermethrin where confirmed resistance to these insecticides exists.

Absence of kdr mutations coupled with partial or complete restoration of susceptibility to pyrethroids with vector pre-exposure to synergists indicates the critical role that metabolic resistance mechanisms play in shaping phenotypic resistance in Madagascar. Further biochemical analysis is needed to determine the specific gene or genes responsible for vector survival. The absence of cross-resistance between DDT and pyrethroids is further evidence of the absence of the role of kdr in determining vector resistance to these insecticides. Kdr-free pyrethroid resistance in An. arabiensis attributed to metabolic resistance was detected in Chad [46]. Similarly, Verhaeghen et al. [47] reported the absence of kdr in several vectors of malaria in the Mekong region resistant to DDT and pyrethroids with elevated detoxifying enzymes. Unlike kdr, all insects are reported to have enzyme-based adaptive mechanisms that would help them metabolize xenobiotics including insecticides [48]. This might be the most likely reason why resistance due to metabolic mechanism was observed in Madagascar though the country is geographically isolated and the gene flow from other malaria endemic countries is restricted.

Results from both WHO tube and CDC bottle bioassays demonstrated the full susceptibility of all three malaria vectors in the country to bendiocarb (carbamate) and pirimiphos-methyl (organophosphate) insecticides. The G119S mutation associated with resistance to carbamates and organophosphates was also not detected. This is encouraging because it means Madagascar has several effective insecticides options to choose from when planning insecticide-based vector control, unlike most other sub-Saharan Africa countries. The other secondary vectors, An. funestus and An. mascarensis, were fully susceptible to pyrethroids and organophosphates.

As noted above, the WHO tube bioassay was the only standard resistance testing procedure used in Madagascar until the CDC bottle bioassay method was deployed in 2012. There are some differences in the diagnostic dosages and what the two techniques measure. The WHO assay diagnostic dosage is twice the lowest concentration that produces 100% mortality after a 60 min exposure and a holding period of 24 h on a susceptible strain/population. The CDC bottle bioassay uses an insecticide that kills all (100%) susceptible mosquitoes within a given diagnostic time (45 min for DDT and 30 min for the other public health insecticides).

Of 95 tests (n = 9500) conducted using both methods, 90 concordant results (94.73%) and five discordant ones (5.26%) were observed in resistance classification. When the data were broken down by insecticide, all the test results from the two test methods converged and no discrepancy was observed in bendiocarb (κ = 1), pirimiphos-methyl (κ = 1) and lambda-cyhalothrin (κ = 0.70). The five discordant results all came from two insecticides, four from permethrin and one from deltamethrin. Further investigation is needed to learn why test results for permethrin differ between the two test methods (κ = 0.45). From our limited observation, the results from the two test methods seemed to vary as the frequency of resistance increased. The discordant results observed in deltamethrin, with mortality of 98% and 96% in one site, was not beyond variation that could be expected due to chance alone (κ = 0.77). Aïzoun et al. [49] reported comparable results from both assays, which is consistent with our results. The monitoring, separately or in parallel, for bendiocarb, pirimiphos-methyl, deltamethrin and lambda-cyhalothrin and give similar results in Madagascar where there is no or very infrequent resistance to these insecticides.

The resistance status of vectors in Madagascar shows the potential for malaria vector control to be effective there. The absence of widespread vector resistance to insecticides other than DDT and permethrin will enable the vector control program to develop a preemptive and evidence-based insecticide use strategy aimed at preventing the development of widespread resistance and preserving the ongoing impact of control efforts.

Though it yielded important findings, this study had limitations. Wild-caught female adult An. funestus and An. mascarensis (a mix of old and young age) mosquitoes were directly used for testing. The results indicated full susceptibility of the vectors to the insecticides tested. Previous studies have reported a correlation between mosquito age and mortality (increased mortality with increasing age) in An. gambiae (s.l.) [50] and An. funestus [51]. The absence of resistance in these two vector species, therefore, may be explained by either full susceptibility of the vector to the insecticides tested or by resistance possibly masked by the use of mixed mosquito ages. Among the mosquitoes exposed to insecticides, none survived. Hence, the first explanation appears more plausible as the use of mixed-age groups may not totally abolish resistance if it exists in the population.

Another limitation is that not all of the susceptibility tests were accompanied by a test of a susceptible mosquito colony as a reference (positive control) in the field due to the logistical challenges of safely transporting susceptible mosquitoes to the remote test sites and lack of an adequate number of mosquitoes in the insectary. However, the quality and efficacy of the WHO impregnated papers and insecticide stock solution used for CDC bottle bioassay were checked and confirmed for the test results reported in this paper.

In one of the study areas, Bekily, PBO-only was not included as a control during the synergist-insecticide testing due to an inadequate number of mosquitoes. Some of the mosquito mortality observed in Bekily could therefore be due to the synergist alone. However, there was no mosquito mortality attributable to PBO from the other sites tested, and therefore it is unlikely that PBO alone was the cause of any mortality in Bekily.

Finally, biochemical analysis was not performed to support phenotypic resistance data obtained with and/or without pre-exposing mosquitoes to the synergists. This was due to lack of a facility in country. Further study is needed to determine the specific genes responsible for metabolic resistance detected in Madagascar.

Despite these limitations, the strengths of this study are the inclusion of three known malaria vectors, the number and geographical distribution of sentinel sites (18 in three of the five eco-epidemiological zones) and the number and types of insecticides tested. The information gathered is substantial and will contribute to informing country malaria control policies.

The limited insecticide resistance observed in Madagascar’s malaria vector populations presents an opportunity for the country’s vector control program to maintain a range of viable insecticides for use in the ongoing mission to eliminate malaria. This situation provides a unique opportunity to implement preemptive insecticide rotation to preserve existing vector control tools and creates space for continued cost-effective vector control programming. If implemented successfully, a strong resistance management plan could provide a working model for other regions that are currently limited by widespread resistance to available insecticides, with the expectation that novel products will be available in the near future.