This study provides essential clues on the bio-efficacy of public health insecticides currently being used for malaria vector control, either singly or in combination, particularly how these insecticides are likely to perform in an area where the primary malaria vectors are An. arabiensis, that are still susceptible to commonly used insecticides, albeit with clear signs of that this susceptibility is declining. This is an increasingly common scenario in East Africa where high pyrethroid treated bed net coverage has favoured An. arabiensis over An. gambiae s.s or An. funestus[8–10].
Several efforts are now being made to determine whether LLINs and IRS when used together can confer greater benefits than when the two are used alone
[1, 5, 28, 29]. Nearly all the studies concluded so far have assumed that LLINs and IRS treatments are not affected by time dependent loss of efficacy, and as such this subject has not been previously examined
[1, 5, 28, 29]. For example, in a recent commentary N’Guessan and Rowland
 noted that a study previously published by Corbel et al.
 showing no added value of combining LLINs with IRS or insecticidal wall linings, may have been limited by the failure to consider periods when the IRS treatment was actually active. However, Bradley et al. working in Bioko Island, Equatorial Guinea have reported that limited residual life of insecticides can lead to increased malaria risk, but that LLINs can reduce this effect when used alongside IRS
Based on percentage mortalities observed in the bioassays, where contact between mosquitoes and sprayed surfaces is ensured, this study shows that activity of the tested IRS compounds can decline significantly within the first few months after spraying, in many cases becoming ineffective earlier than the time when they would normally be due for re-spraying
. According to recommendations made by WHO, DDT should be re-sprayed after every six to 12 months, lambda cyhalothrin every three to six months and pirimiphos methyl, every two to three months
. As an example, this study shows that pirimiphos methyl EC caused merely 42.5% mortality on ceilings and only 55.0% on walls by the 3rd month after spraying, down from 100% mortality in the first month. Interestingly, in a concurrent study evaluating LLIN-IRS combinations (Okumu et al., unpublished), pirimiphos methyl was also shown to be the most lethal of all the tested chemicals against the mosquitoes, yet the mean An. arabiensis mortality over six months was less than 30%. These results may suggest that while standard WHO bioassays have been widely used to assess efficacies of insecticidal interventions over a given period of time, overreliance on those findings without corroborative experimental hut trials, may miss the subtleties associated with insecticide decay or mosquito behaviour in human dwellings, both of which can lead to significantly low mortality rates in human occupied houses.
If a practical situation where malaria control programs can conduct a maximum of two spray rounds per year is considered, it becomes apparent that all the tested IRS compounds in their existing formulations would be minimally appropriate for use in this study area or in areas with similar vector populations and where people use similar construction materials for walls and ceilings, though high toxicity of pirimiphos methyl as observed in our experimental hut study (Okumu et al., unpublished), may make it suitable for short term use, especially against epidemics. The low mortalities observed in the bioassays, would also suggest much poorer performance of the IRS treatments in actual human houses that are correctly using LLINs that prevent mosquitoes feeding, given that unfed mosquitoes may not rest long enough on treated surfaces to pick up lethal doses. Indeed studies have now shown that in areas dominated by An. arabiensis, as the main malaria vector, the effect of insecticidal interventions is limited
. A concurrent study evaluating LLINs and IRS, An. arabiensis mortalities were consistently low in huts with LLINs or IRS confirming the more limited value of intradomicilliary insecticidal interventions against this species that is more adaptable to feed at times and places where hosts are available than the strictly anthropophagic and endophagic vectors An. gambiae s.s. and An. funestus s.s. (Okumu et al., unpublished).
The findings of net bioassays were surprising, especially given that LLINs should retain their insecticidal activity for at least three years and 20 washes
. The tests described here depict a very rapid loss of the mosquitocidal activity of the candidate LLINs, especially Olyset® nets; even in the wire ball tests. Whereas LLINs are expected to last at least three years
, with some such as the Olyset® nets designed to have up to five years of effective life
, the tests described here show that insecticidal activity can decline significantly within the first few months. For example, Olyset® nets killed only 68.9% of An. arabiensis mosquitoes exposed in the wire ball assays and only 25.5% of those exposed in the cone assays by the second month of use, and by the sixth month, only 33.3% and 14.6% of the mosquitoes died when exposed to this net in either wire ball tests or cone tests. By the sixth month when this study ended, only the PermaNet 2.0® nets retained toxicity greater than 80%. Despite this rapid decline, it is equally important to note that in this study, we also observed that all candidate LLINs retained high knock-down rates (>90% in wire ball tests and >80% in cone tests) on the exposed mosquitoes, except Olyset® nets whose knock-down activity was slightly reduced to 72.7% on wire ball tests and 62% on cone tests by the sixth month. In some of the bioassays on LLINs and the IRS treated walls and ceilings, there seemed to be unexpectedly low mortalities in month two, such that when the data was plotted, month 2 was slightly out of the general monthly decay trend. Figures
3 show observed mortality was often higher in month three than month two or in month four than month two. While there is no obvious explanation for this observation, it could possibly be an artifact resulting from mosquito rearing conditions, and is unlikely to change the general inferences from these findings.
One important aspect to consider here is the fact that in this study the nets were not washed at any time during the course of the study, but were instead only dusted occasionally to remove dust. The lack of washing could explain the observation that LLINs such as Olyset® nets, which are known to possess regenerative properties (normally activated after lengthy periods of use, washing or exposure to heat
[33, 34]), exhibited a decline in activity during this study. Nevertheless, it is reasonable to be concerned about the quality of marketed LLINs, and all stakeholders including, net manufacturers, public health implementers and net users could benefit from improved quality control. Similar losses of LLIN activity have actually been reported also in community level trials, one example being a study in Kenya where 79.6% of Olyset® nets were found to have failed (i.e. having bioassay mortality rates of less than 50% in two consecutive monthly tests) after two years of use, compared to 17.8% of PermaNet nets
. The apparent loss of insecticidal activity also may suggest that long before the expected 3–5 years life of LLINs, the only effects of nets that would be left, is the physical barrier effect, where nets work simply by preventing mosquitoes from feeding upon the net occupants rather than killing the mosquitoes. Indeed, in a concurrent evaluation of LLINs and IRS it was determined that intact non-insecticidal nets equally prevent mosquitoes from blood feeding upon net users, just as intact insecticidal nets (Okumu et al., unpublished).
In addition to enabling the assessment of bio-efficacy and residual activity, the wall and ceiling bioassays also highlighted how differences in treatment surface substrates can affect insecticidal efficacy. That is to say, efficacy of active ingredients on mosquitoes is modulated by type of substrate onto which the compound is applied
. In this study, two of the IRS insecticides, pirimiphos methyl EC or lambda cyhalothrin CS, killed 100% of mosquitoes exposed to the Mikeka ceilings, while DDT WP sprayed on Mikeka ceilings killed a modest 85% in the first month. However, on the mud walls sprayed with the same chemicals, we observed 100%, 90.0% and 97.5% mortality respectively in the same period. It seems therefore that, whereas lambda cyhalothrin CS, performed better on ceilings than on mud surfaces, the DDT formulation was clearly better when used on mud walls than when used on Mikeka ceilings, from which the water-based wettable powder would more easily have flaked off over time. Similar arguments have been put forth by many previous authors
[36–40], and it is thought that such differences are associated with differences in adsorptive properties of the substrates. For instance, mud surfaces can be highly porous and adsorptive to insecticides, and substrates containing alkaline substances may degrade the candidate insecticide faster than substrates without alkaline contents
[36, 41] In one study where pyrethroids were tested on different substrates, it was found that porous surfaces such as mud can show variability in insecticidal activity, presumably due to absorption of the insecticides, while less porous surfaces such as wood would result in higher insecticidal activity for long periods due to lower rates of insecticide absorption
. More recently, Etang et al.
, also observed variations of insecticide residual bio-efficacy on different types of wall surfaces in Cameroon and, therefore, suggested that local construction materials should be considered when determining lengths of spray cycles as is also demonstrated from this data. Similarly, a recent study in an area in Ghana where malaria vectors were resistant to pyrethroids, organochlorides and carbamates, but not organophosphates showed that pirimiphos methyl formulation, which in this current study was nearly fully deteriorated by the third month, remained effective when sprayed onto cement walls, and can continue to kill An. gambiae s.l for up to fifteen weeks, matching existing WHO recommendations
With regard to bioassays on nets, it is also clear that the two methods used in this study, i.e. the plastic cone and wire ball method
, can give different outcomes, and therefore a more careful interpretation is required. The LLINs generally killed more mosquitoes in the wire ball assays than in the cone assays. According to the current LLIN testing guidelines
, there are two possible alternatives to the WHO cones, which can also be used to assess residual efficacy of insecticidal nets, namely: 1) the use of WHO test tubes (cylinders) lined on the inside with the test nets, and 2) the wire-ball test as used in this study. It is however also suggested that further calibration against the WHO cones is required before the alternative methods can be widely used in testing and evaluation of insecticide for treatment of mosquito nets, an explanation which also suggests an expectation that the two test methods would give different results.
It can be argued that since the wire ball offers no alternative resting sites (unlike in the cone assays, where mosquitoes can occasionally rest on the cotton plug used to seal the insertion hole on top of the cone and, therefore, fail to make adequate contact with the test surfaces), mosquitoes are more likely to be killed in the balls than in the cones. Furthermore, if the active ingredient has irritant properties, which prevent mosquitoes from resting on treated surfaces for extended periods of time, it is possible that exposed mosquitoes would tend to frequently move from point to point making multiple contacts with the treated surfaces and, therefore, leading to greater exposure and higher percentage mortality. In this study however, no mosquitoes were seen avoiding tarsal contact with the netting material during the cone bioassays; neither did we observe many mosquitoes landing on the cotton wool that was used to plug the plastic cones, which would have indicated a significant role of irritancy
[43, 44]. It is more likely therefore that the reason more mosquitoes died in wire ball assays than the cone assays was the greater total surface area of LLINs and consequently the greater overall quantities of insecticide that these insects were exposed to in the wire balls relative to the cones.
Insecticide susceptibility is usually classified based on proportions of mosquitoes that die when exposed to diagnostic concentrations of test chemicals as follows: 98-100% mortality indicates susceptibility, 80-97% mortality indicates signs of resistance that need to be confirmed and less than 80% mortality indicates that there is insecticide resistance
. In a previous nationwide study in Tanzania, where insecticide resistance was assessed in several districts, it was shown that susceptibility of mosquito populations to lambda cyhalothrin, deltamethrin and permethrin had started to diminish in most of the sentinel districts in the country, including Kilombero district, which neighbours Ulanga district where this current study was conducted
. In that study, standard WHO insecticide susceptibility tests on An. gambiae s.l from Kilombero district, showed 93.9% mortality after exposure to 0.05% lambda cyhalothrin, 96% mortality after exposure to 0.75% permethrin and 90.3% mortality after exposure to 0.05% deltamethrin
. Results from this current study (Table
1), depict a closely similar outcome two years later, i.e. full susceptibility to DDT, and reduced susceptibility to lambda cyhalothrin (mortality = 90.2%), permethrin (mortality = 95.2%) and deltamethrin (mortality = 95.8%). While the resistance limits in this area have not yet reached a state where vector control interventions such as pyrethroid based LLINs and IRS with DDT would be considered ineffective, the potentially declining susceptibility to common vector control insecticides clearly calls for continued constant monitoring and resistance management planning to be incorporated in any future vector control campaigns.
The good news, however, was that both the bioassays and the molecular analysis conducted to detect kdr alleles, confirmed absence of target site resistance to pyrethroids and DDT, which is one of the mechanisms linked to genetic mutations in the para-sodium channels in several insects
. Pyrethroid-DDT cross-resistance currently presents, what is perhaps the greatest challenge to insecticide based malaria interventions in Africa
[47, 48]. Susceptibility surveys have thus become standard pre-requisites, providing baseline data on insecticide susceptibility status, to support large scale LLINs and IRS campaigns in Africa
[48, 49]. Two different kdr mutations have been found in the African malaria vector An. gambiae s.s, including one in West Africa, which is caused by a leucine to phenylalanine substitution (L1014F)
[23, 26] in the genetic sequence coding for the sodium channels, and a different mutation in East Africa, caused by leucine to serine substitution at the same amino acid position (L1014S)
[24, 25]. Though the kdr-detection protocol used in this study could detect either of the two mutations
, they were not detected in any of the mosquitoes, and instead, there was a 100% kdr-negative rate in all the samples. It should be noted however that these tests may not be conclusive as there may be other modes of physiological resistance, which will also need to be investigated in the future.
If the results of this study are interpreted in the context of an increasingly important question of whether there are any added advantages of combining LLINs with IRS, relative to using each individual application separately
[4–6, 28], they provide evidence to support the need for adding LLINs where IRS is the only existing intervention. Since most of the IRS candidate insecticides decay so quickly, and since it can be difficult to regularly re-spray houses at the frequencies stipulated by WHO
, addition of LLINs in such houses would provide additional reduction in indoor mosquito biting rates or potentially kill additional mosquitoes, and would add the temporal overlap necessary to protect house occupants during the period after which the IRS is no longer efficacious. Similar suggestions have been made by Bradley et al. who recently attributed increased malaria risk to the time dependent insecticidal decay
. On the other hand, to benefit from adding IRS into houses where LLINs are already being used, especially where the predominant vector is An. arabiensis, it will be necessary to ensure that the IRS campaigns are properly timed, regular, quality controlled and use highly mosquitocidal chemicals, preferably organophosphates or carbamates rather than pyrethroids as in all current LLINs, a strategy which would help mitigate against spread of insecticide resistance
. Where IRS treatments significantly deter malaria mosquitoes from entering houses, it may also be important to ensure implementation at consistently high household coverage to ensure that even those mosquitoes that are deterred from one house to not find blood meals in other unprotected households