The use of insecticides in malaria control programmes in Africa is expanding with the extensive and rapid roll out of long lasting insecticide-treated bed nets (LLINs) and indoor residual spraying (IRS) 1 . Twelve insecticides are approved by the World Health Organization (WHO) for IRS, but these belong to just four chemical classes (organochlorines, organophosphates, carbamates and pyrethroids) 2 . All four of these classes are nerve poisons and either target acetylcholinesterase in the synapses or the voltage-gated sodium channel on the insect neurones. For insecticide-impregnated material, such as LLINs, the chemical arsenal is even more limited with only six insecticides, all from the pyrethroid class, available 2 . These same insecticide classes are also widely used to control agricultural pests in Africa and this can pose additional selection pressure on mosquitoes when insecticide contaminated ground water permeates their larval habitats. This intensive exposure to insecticides has inevitably resulted in the evolution of insecticide resistance in the Anopheles mosquitoes that vector malaria. Resistance to the organochlorines DDT and the now obsolete dieldrin was first reported in African malaria vectors in the 1950s and 1960s 3 4 . Pyrethroid resistance was detected in African malaria vectors in 1993 5 . Since then there have been published reports of pyrethroid resistant populations of Anopheles gambiae s.l. in countries from west, central, east and southern Africa (see, for example 6 7 8 9 ) and Anopheles funestus in Ghana, Mozambique and South Africa 10 11 . Recently, carbamate and organophosphate resistant populations of An. gambiae have been reported in west Africa 12 .

The WHO has produced a series of guidelines for measuring insecticide resistance in disease vectors and determined a set of 'diagnostic doses' for many of the commonly used insecticides 13 . A WHO reference centre in Malaysia supplies insecticide bioassay kits and papers coated with these diagnostic doses of insecticides. This has facilitated standardization of bioassays, but even when these guidelines are followed, variations in methodologies can make inter study comparisons difficult. For example, although WHO protocols specify using 3-5 day old unfed females, the difficulties in measuring mosquito age in the field and the lack of even basic insectary facilities needed to rear adult mosquitoes from field collected larvae, sometimes leads to the use of wild caught mosquitoes, which are of mixed age and physiological status 14 . This may confound results, as both age and blood feeding status are known to influence the mosquito's response to insecticides 15 16 . Difficulties in procuring the insecticide-treated papers from the WHO reference centre can also disrupt planned insecticide resistance monitoring and can sometimes necessitate the adoption of alternative methodologies, or the use of insecticide papers procured from alternative sources. In some cases molecular assays to detect the resistance alleles have been used as a substitute for bioassays 17 . These molecular assays have an important role to play in insecticide resistance management. As they work at the individual rather than population level and are able to identify heterozygotes that may have no discernable phenotype, they are more sensitive than bioassays. However, current understanding of the molecular basis of insecticide resistance in malaria vectors is largely limited to target site resistance. Given that this is only one of several mechanisms implicated in conferring resistance, the molecular tools cannot currently be used as a substitute for bioassays.

The WHO/TDR network on insecticide resistance in African malaria vectors was established in 2008. One of the major objectives of this network is to improve the monitoring of insecticide resistance in malaria vectors, prioritizing areas where large scale insecticide based control programmes are being implemented or planned in the absence of data on vector susceptibility. Here steps taken to standardize resistance monitoring in each of the participant countries are described and results from year one for three countries, Burkina Faso, Chad and Sudan, are reported. The bioassay data indicate an alarmingly high distribution of resistance in members of the An. gambiae complex. The implications for the sustainability of malaria vector control in Africa are discussed.

The results of the insecticide bioassays are shown in Figures 2, 3 and 4. In Burkina Faso the highest percentage of mosquitoes surviving the WHO diagnostic doses were seen in the two agricultural districts in the southwest of the country, Soumousso and Kuinima. This agrees with earlier monitoring of resistance in this country completed in 2006 which found higher frequencies of resistance in the western part of the country than in the central and eastern parts 22 . However, the current study clearly demonstrates a general increase in the numbers of resistant An. gambiae in Burkina Faso during the past three years. Populations in the Ouagadougou district that were reported as fully susceptible to pyrethroids and DDT in 2006 22 now show permethrin and DDT resistance, with mortality levels less than 80 per cent recorded in October 2008. The Soumousso population also shows a much lower mortality to DDT and pyrethroids than was reported previously. Resistance to bendiocarb and fenitrothion was also detected in Kuinima and Soumousso. This is the first published bioassay data on these insecticides in Burkina Faso although resistance has been inferred by the presence of the ace-1 ^R mutation, in this region 23 .

Anopheles gambiae M and S forms and Anopheles arabiensis were found in all four study sites in Burkina Faso and the relative frequencies of each form varied between the two collection seasons (Figure 2). The species distribution within the DDT and permethrin survivors and dead progeny was compared to that of the general population (weighted according to the bioassay results) for the second collection season (where resistance was generally higher) for all sites in Burkina Faso (Table 1). With two exceptions (permethrin survivors in Kuinima and DDT survivors in Soumousso), the species distribution in the surviving and dead subset differed significantly from the distribution in the total population used for the bioassays. In general, the majority of the DDT and permethrin survivors are comprised of the S form while the biggest proportion of the dead progeny was identified as An. arabiensis. It is important to note that, although lower mortality with the diagnostic dose was generally observed for An. gambiae compared to An. arabiensis, some individuals from the latter species survived the diagnostic doses of DDT and pyrethroids in almost all assays indicating that resistance is widely distributed, albeit currently at lower frequencies, in An. arabiensis in Burkina Faso.

All of the bendiocarb and fenitrothion survivors in Soumousso belonged to the S form of An. gambiae. A similar pattern was seen in Kuinima with all fenitrothion survivors and 46/47 bendiocarb survivors belonging to the S form (one M form An. gambiae survived bendiocarb exposure in Kuinima).

In Chad, permethrin and/or deltamethrin resistance was detected in all four sites but only Donia, the cotton growing area, showed resistance to DDT (Figure 3). All populations were fully susceptible to bendiocarb and fenitrothion. Previous insecticide bioassay data from collections made in August and September 2006, reported resistance to permethrin and deltamethrin in Bongor, the vector control site on the Cameroon border used in the current study 24 . Comparing the results from 2006 with those of round two in 2008 indicates an increase in the frequency of permethrin resistant individuals with mortality with the diagnostic dose decreasing from 59.8% mortality to 43% and, for deltamethrin, from 72.5% to 31% mortality over the two-year period.

Anopheles arabiensis was the only member of the An. gambiae complex identified in N'djamena and Mandelia and the predominant species in Bongor (Figures 3). In Donia, in the south of the country, An. gambiae M and S forms, and An. arabiensis were present in large numbers. All species/forms showed resistance to pyrethroids at this site.

In Sudan, as in the other countries, large heterogeneities were observed in mortality between the different sentinel sites. Anopheles arabiensis is the only member of the An. gambiae complex present in Gezira state (Figure 4 and 18 25 ). Here, the highest numbers of bioassay survivors were observed in the urban agricultural site of Wad Medani, which is found within the Gezira irrigation scheme (Figure 4). Permethrin and DDT resistance was also observed in Roofa, which was selected as the low insecticide exposure site. In contrast, the cotton growing region of El Managil showed only 'suspected resistance' according to WHO definitions 13 . Permethrin and DDT resistance had been monitored in three of the four sites in Sudan in 2005/2006 25 . There has been little decrease in mortality over the past three years in El Managil and only a small decrease in Roofa. In Wad Medani, mortality after DDT exposure was 89.9% in 2005/2006 and 95.1% after permethrin exposure. It is difficult to directly compare these results with the current study as a different age group of mosquitoes were used (1-3 days in 25 versus 3-5 days in the current study) and because the results from the current study show very strong seasonal variations, discussed below, that were not accounted for in the earlier study. Nevertheless, it appears as if the frequency of DDT and permethrin resistant individuals has increased in the past three years in Wad Medani. Mosquitoes from all four sites were fully susceptible to bendiocarb and fenitrothion.

Using WHO definitions of resistance 13 , eight of the 12 populations in this study were resistant to DDT in at least one round of monitoring and a further three had suspected resistance. For the pyrethroids, all sentinel sites showed suspected or confirmed resistance to both permethrin and deltamethrin in at least one round of monitoring. Resistance to the acetylcholinesterase inhibitors, bendiocarb and fenitrothion, was more restricted with the four populations analyzed from Chad and the four from Sudan all remaining susceptible to these insecticide classes, although resistance is present in two sites in the southwest of Burkina Faso.

In the study design, sentinel sites with different levels of insecticide exposure were selected. The resources available did not permit a robust analysis of the effects of different selection pressures on the evolution of resistance but did enable a wider range of resistance phenotypes to be captured in each country. Surprisingly high levels of heterogeneities in resistance were observed even across relatively small distances. For example Wad Medani is less than 50 km southeast of El Magagil and yet, while the An. arabiensis population at the latter site showed only low or 'suspected resistance' to DDT and pyrethroids, the mortality rate to this insecticide was drastically reduced in Wad Medani.

Agricultural use of insecticides has been suggested to be one of the major drivers of insecticide resistance in malaria vectors 26 27 . Certainly the expansion of urban agriculture, which is often associated with intensive and under regulated insecticide use, is having a pronounced effect on the ecology and resistance levels in malaria vectors in several countries 28 . There are also many well-documented examples of IRS and LLINs directly selecting for insecticide resistance 8 29 30 . No IRS programmes are currently active in Burkina Faso or Chad and in these study sites the primary malaria vector control intervention is the use of LLINs. The distribution of these is very patchy within our study sites. Only Koupela in Burkina Faso and Bongor in Chad have been targeted for free net distribution to at risk populations but bed nets are used by some households in the other sites and the frequency of insecticide-treated nets in each site has not been documented or compared. In Sudan, larviciding with temephos is being used throughout Gezira State but only the Wad ElHadad site is targeted for IRS (with bendiocarb). LLINs can be found in all sites but, during the time of the study, coverage rates were highest in Wad ElHadad.

Pronounced differences were frequently observed in the mortality levels during the malaria transmission season. In general, higher numbers of bioassay survivors were observed towards the end of the malaria transmission season in Chad and Burkina Faso but this trend was reversed for DDT in two of the sites in these countries, Donia in Chad and Kuinima in Burkina Faso. Furthermore in Wad Medani, located in the Gezira irrigation scheme, lower survivorship was observed during the second collection season, which occurred in January 2009. The heterogeneity in resistance observed within short periods of times (typically four months elapsed between the beginning of the first monitoring round and the end of the second) may simply reflect stochastic variation and results from subsequent years will be needed to determine whether the trends observed in year one of the study are reproduced. In this study, bioassays were performed on adults raised from field caught larvae, in line with WHO recommendations 13 . Collecting immatures rather than raising progeny from blood fed females collected in houses increases the genetic heterogeneity of the samples and avoids selection biases that may be caused by collection of adults inside insecticide treated houses. A limitation of this approach is the lack of control over rearing conditions that would be achieved by using F1 progeny raised under standard conditions in the insectary. Hence the choice of using larval collections may have introduced additional 'noise' to the bioassay data. Nevertheless, there are several biological explanations that may account for this apparent seasonal heterogeneity.

Firstly, the results could reflect a change in species composition. Anopheles gambiae is a species complex composed of at least seven morphologically identical species, of which An gambiae s.s and An arabiensis are the most important in terms of malaria transmission. Anopheles gambiae s.s can be further divided into at least two different molecular 'forms'. Molecular analysis has found that the frequency of insecticide resistance alleles can vary dramatically within sympatric populations of An. gambiae s.l 31 . Therefore, the species and molecular form distribution was determined in both collection rounds in all sites. This was performed using a randomly selected subset of 40 mosquitoes from the control bioassay (Figure 2). However, it is important to note that, when the species distribution in this control subset was compared to the subset used for the insecticide bioassays in Burkina Faso, significant differences were observed in 3 of the 8 tests (data not shown). Further work in identifying the most appropriate sampling strategy to capture the species variation in a given study site is clearly needed.

Nevertheless, even when taking the above concerns into consideration, no clear trends in species distribution were observed that could explain the seasonal variation in insecticide susceptibility. For example, in Kuinima, Burkina Faso, the proportion of An. gambiae relative to An. arabiensis increased between collection rounds one and two. This was accompanied by an increase in the number of mosquitoes surviving pyrethroid exposure. However, in Soumousso, where the proportion of An. gambiae decreased between collection rounds, the frequency of pyrethroid resistant individuals in the population increased.

Many of the populations showed a significant difference in the species distribution of the survivors relative to the general population with the An. gambiae M or S form generally showing higher numbers of bioassay survivors than An. arabiensis. Care should be taken in interpreting these results due to sample size limitations imposed by the small number of survivors in some sites. However this analysis does demonstrate that resistance is not restricted to just one member of the An. gambiae complex and, while proportionally higher numbers of An. gambiae s.s. are generally observed in survivors, An. arabiensis is also resistant to pyrethroids.

In some sites, the change in observed susceptibility to insecticides between the two seasons may be related to exposure to insecticides in agriculture. The first round of mosquito collections in Wad Medani, Sudan, occurred in October/November 2008. During this time, crops in the Gezira irrigation scheme are regularly sprayed with insecticides. The second collection round occurred during the winter period, when crop production is lower and intensive insecticide spraying is reduced. The increased numbers of pyrethroid and DDT susceptible An. arabiensis individuals collected from breeding sites in this region in the winter months may possibly reflect this reduction in exposure to insecticides used in agriculture. A previous study in the cotton growing region of Pitoa, Northern Cameroon, observed a decrease in mortality between insecticide bioassays performed in July, at the start of the cotton spraying season, and those performed in late August/early September, mid way through the cotton crop and a period of intense insecticide application 26 . This observation was not repeated in the neighbouring cotton growing region of Donia, Chad, in this study. However, the increase in numbers of bioassay survivors in Soumousso seen in October versus August mirrors the decrease in mortality to pyrethroids reported during the rains in the cotton growing region of Burkina Faso in 1999/2000 27 .

A third explanation for the seasonal variation in mortality to pyrethroids in particular, is the cause for gravest concern. The results from Burkina Faso may be indicative of a very rapid selection for pyrethroid resistance in An. gambiae s.s. in West Africa. Although this network has only been in operation for one year, additional data from the region paint a worrying picture. The frequency of the kdr allele, which encodes an altered version of the voltage gated sodium channel and confers resistance to pyrethroids, is increasing at a rapid rate in the region 22 . Metabolic resistance mechanisms have also been described and, although tools are not yet available to track the frequency of these alleles as they spread through the population, laboratory analysis has identified several cytochrome P450 genes that are conferring pyrethroid resistance in multiple populations of An. gambiae s.s 32 33 . Similarly comparison of the results from the first year of this study with previously published reports from Chad and Sudan 24 25 also indicate rapid increases in the frequency of pyrethroid resistant mosquitoes. Clearly resistance monitoring, using both bioassays and molecular tools, needs to be incorporated as part of the monitoring and evaluation component of any existing or planned insecticide based malaria control activity. Selection of insecticides for IRS must consider the existing levels of insecticide resistance in the local vector populations and the alternative sources of selection pressure imposed on the local vector population.

With some notable exceptions 10 34 studies on the impact of insecticide resistance on malaria control activities are severely lacking, and indeed some studies have failed to observe any detrimental effect of resistance on mosquito control in experimental huts 35 but it would be naïve not to anticipate an eventual reduction in the efficacy of vector control interventions as a result of resistance. Indeed, a recent study in Benin, comparing the efficacy of LLINs and IRS in areas with largely susceptible or pyrethroid resistant populations of An. gambiae s.l., provided a stark reminder of how dependent malaria control in Africa is on a pyrethroid susceptible malaria vector population 36 . The current study shows that this trait cannot be taken for granted.

The authors declare that they have no competing interests.

HR is the PI of the WHO/TDR network. She conceived and designed the study and drafted the manuscript. HA, participated in the study design, led the field work in Sudan and carried out laboratory analysis. AB and WMG contributed to the field and laboratory work in Burkina Faso and the analysis of the data. CKH participated in the study design, led the field work in Chad and contributed towards the analysis. EYK contributed to the field work in Chad and carried out laboratory analysis. NFS participated in the study design, led the field work in Burkina Faso and contributed towards the analysis. FS participated in the study design, provided a critical review of the manuscript and participated in the data analysis for Chad. MC participated in the study design, provided a critical review of the manuscript and participated in the data analysis for Sudan. All authors read and approved the final manuscript.