Zanzibar’s efforts in malaria vector control, largely based on IRS and LLINs, have contributed to a substantial decrease in malaria transmission over the years. These great achievements have raised hopes for the islands to enter malaria pre-elimination and elimination phases. However, results reported here raise serious concerns and demonstrate the substantial challenges that threaten the sustainability of the gains achieved.
The apparent near elimination of An. gambiae s.s. on the island has been observed in other parts of sub-Saharan Africa  and is usually attributed to the success of IRS and LLINs in controlling the endophilic and anthropophilic An. gambiae s.s. However, this leaves An. arabiensis, which is an equally efficient malaria vector, feeds indoors and outdoors, on human and non-human hosts and can rest outdoors , therefore is not optimally controlled by indoor interventions. The control of An. arabiensis for malaria elimination will require more tools in addition to LLINs and IRS.
In Zanzibar, IRS has been conducted for over six years now with more than 90% coverage of the island using lambda-cyhalothrin. LLINs impregnated with permethrin have been rolled out for more than three years to every household in Zanzibar. The decision to use the same class of insecticide (pyrethroid) for both IRS and LLINs at the same time has undoubtedly increased the selection pressure on the vectors to develop insecticide resistance .
Results from the standard WHO resistance assays in this study revealed a very pronounced and increasing pyrethroid resistance in An. arabiensis on Pemba Island over three collection periods within one year. Thirty years ago, An. gambiae s.s. was shown to be resistant to DDT on Pemba , while Lines and Nassor  reported the spread of DDT resistance to Unguja island. However, the current tests showed a 100% susceptibility to DDT for wild mosquitoes on Pemba, indicating that the lack of IRS using DDT, i.e. no DDT selection pressure, might have allowed the population to revert to the susceptible state. This suggests that the pyrethroid resistance now being documented is mediated by a metabolic and not a target site mechanism (cross-resistance between DDT and pyrethroids often suggests the presence of a kdr mutation [3, 19]). It is unclear why pyrethroid resistance in Zanzibar is highly prevalent on Pemba while the mosquitoes on Unguja remain largely susceptible. The ZMCP implements the same vector control interventions on both islands and the distribution of malaria vector species is more or less the same (predominantly An. arabiensis).
Although the question of keeping livestock was not investigated in our study, farmers in many areas use pyrethroids for the control of ticks and other veterinary pests. This may contribute to the selection pressure for cattle feeding mosquitoes such as An. arabiensis and might accelerate the emergence of resistance. However, it is believed that there is no significant difference in keeping livestock between the two islands of Zanzibar (personal communication from the Zanzibar Veterinary Officer). In addition, heavy use of similar classes of insecticides in agriculture may contribute to the selection for insecticide resistant mosquitoes [22, 23]. More investigations are required, therefore, to ascertain the origin and mechanism of resistance on Pemba and the difference between the two islands. Moreover, it is important to regularly assess the efficacy of lambda-cyhalothrin, which shows 5% survival in An. arabiensis on Unguja that may quickly become a serious resistance problem as it has on Pemba.
Physical inspection of LLINs distributed in early 2008 by ZMCP in the West district of Zanzibar has shown that 68% of the nets were in poor physical condition and below standard . This could be attributed to the structure of the beds used in the villages, the common type being that of wood connected by metal clamps with protruding bolts from all four corners of the frame. Another type of bed that is locally made uses wood branches connected with rope. These kinds of beds damage the nets very quickly, particularly on the lower portion where the nets are touching the bed frame or are tucked under the mattress. Similar results were found in Ghana and rural Uganda where the number of holes increased towards the lower part of the nets [14, 24]. Holes on the upper portion of the nets could be caused by tearing with sharp pointed sticks and other household materials. In some cases, burning was another cause of net damage as a result of exposure to naked flames from wood fires and kerosene lamps. Seam failure was responsible for damage found on the upper portion of a small number of the nets. The extent and size of holes and the frequency of seam failures demonstrates the need to raise the performance requirements for LLINs in terms of both fabric quality and seam strength .
In this study, two thirds of LLINs were damaged within three years of distribution. In the highlands of Ghana 84.8% of the nets were intact after 4–5 years of field use but only 56.6% of nets in lowland villages were intact 3–4 years after distribution . In another study conducted in Lao, about 40% of nets were in poor physical condition after 2–3 years of field use . These results suggest that net longevity is highly dependent on where they are being used and the type of bed structure.
While the present study showed (based on the cone bioassays) that new unwashed Olyset™ nets were completely effective against laboratory reared and wild-caught malaria vectors, the probability of LLINs being effective for vector control for any length of time is very low considering how rapidly the efficacy of the nets was reduced with washing (Figure 5). The mortalities achieved on the washed nets for the fully susceptible An. gambiae s.s. laboratory colony were far less than the minimum effectiveness recommended by WHO . The use of wild-caught anophelines (largely An. arabiensis) showed even lower mortalities after only six washes indicating a serious lack of efficacy in the presence of pyrethroid resistant vectors. While there is some doubt about the reality of 25 nets being washed 36 times over 3 years, the results are no different to the groups that were washed only 6 or 12 times. Other studies using the same WHO cone method have revealed that LLINs had a knockdown rate of 95% and mortality rate of 80% after 36 months of field use in Uganda . Similar results were observed after three years of net distribution in Buie and Fentalie districts of Ethiopia  and in a Tanzanian village after seven years of Olyset™ use . The current results also contradict previous findings in Tanzania reporting that old Olyset nets that had been in domestic use for four years were as good as new nets [28, 29]. However, WHO  reported in an overview of the variation of mortality and knockdown as a result of cone tests (on three year old nets collected from Angola, Ghana, Kenya, Madagascar, Togo and Zambia), that only 33% of 120 nets tested fulfilled the bioassays criteria. Furthermore, when the 80 nets that failed on the bioassay cone tests were subjected to a tunnel test, only 61% fulfilled the WHOPES criteria.
The insecticide retention analyses on Olyset™ nets carried out in the present study showed that they retained high permethrin concentrations within the fibers, but significantly less than new nets. The important aspect in mosquito mortality induced by LLINs is the surface concentration of insecticide, which is affected by a number of factors including frequent washing . The present study revealed that 40% of nets were dried in direct sunlight, which might also have affected the performance of the nets. There was no significant difference in mortality on LLINs washed between six and thirty-six times but the reduction in efficacy between unwashed and those washed was highly significant. Although reports on washing frequency are not always accurate, the most significant and worrying finding of this investigation is that regardless of the number of washes, LLINs failed every criteria for efficacy within only three years of use.
Based on the low malaria prevalence in Zanzibar, the question has been raised as to when the country-wide indoor residual spraying, which is expensive, should be stopped. The decision would be based on both entomological and parasitological findings. If IRS is stopped, LLINs would be the only remaining strong vector control tool to combat the disease at this critical stage when Zanzibar is approaching the pre-elimination phase. The very low mortality achieved by old used nets tested on susceptible An. gambiae s.s and compounded even further by resistance in wild mosquitoes in Pemba, showed that the LLINs distributed in 2008 were no longer effective against malaria vectors in Zanzibar after three years of use. The necessity for net replacement therefore, should be considered well before three years, instead of the five years suggested by the manufacturer. More importantly, the evolution of pyrethroid resistance in the Zanzibar vector population calls for urgent implementation of insecticide resistance management using different classes of insecticides and currently this can only be achieved through IRS. The results reported here have prompted the ZMCP to start using bendiocarb for the current IRS campaigns. However, this switch is likely to be more costly for the program. Most of the cost of implementing IRS is operational (http://www.pmi.gov/resources/reports/irs_iqc08.pdf) and since bendiocarb has a relatively short residual life (2–3 months), this would require more rounds of annual spraying. The insecticide resistance mitigation plan for Zanzibar has, therefore, adopted a targeted approach to IRS with bendiocarb in areas with higher malaria prevalence instead of blanket coverage of the whole island. In parallel, it was also decided to redistribute new LLINs to every household in 2012. However, in order to remain effective, these LLINs would have to be replaced again before 2015.