Implementation of highly effective methods of decreasing malaria transmission, including ITNs, IRS, and ACT, has led to renewed discussion about global eradication of malaria (4,11). A major goal in moving toward eradicating malaria is interruption or elimination of local malaria transmission. The World Health Organization has stated that “elimination has been achieved when the ‘prevention of reintroduction’, without local transmission by mosquitoes, has been successful for three or more consecutive years” (12). Interruption of local transmission is the step before elimination, in which it is documented that local transmission of malaria is absent in a previously malaria-endemic area for a specific period. The present study provides microscopy evidence of interruption of local malaria transmission in 2 adjacent highland areas of unstable transmission. Malaria could recur in these areas, and it is unclear precisely how much specific factors (e.g., IRS, ACT, and changes in rainfall and temperature) affected malaria transmission and incidence. Overall, however, data support the idea that in unstable transmission settings, combining regular, widespread IRS campaigns and use of ACT as first-line antimalarial treatment has the potential to interrupt local malaria transmission.

IRS probably played a major role in reducing malaria transmission in these areas for several reasons. First, sustained decreases in indoor resting Anopheles spp. mosquito density were seen after the IRS campaigns. Second, in both sites, malaria incidence decreased only after IRS was widely applied. In Kipsamoite, no large reduction in malaria incidence was seen until 2007, when spraying covered >70% of households. In Kapsisiywa, ≈50% of households were sprayed in 2005 and 2006, and a large decrease in malaria incidence was observed in both years. In 2007, after >90% of households were sprayed, malaria transmission was interrupted for the subsequent year. Third, in contrast to IRS, 2 factors that could affect vector density and therefore malaria incidence, rainfall and temperature, showed no clear relationship with either vector density or malaria incidence.

A reduction in vector density was seen in both sites in 2003 before the IRS campaigns. Potential reasons for this reduction include an unusual decrease in temperature during July–September 2003 (Figure 2). This decrease in temperature coincided with the first decrease in vector density; another possible reason for the decrease was pyrethrum spray catch testing of anopheline vectors conducted by our team, which started in April 2003. Although spraying with short-term insecticide does not usually affect vector density in areas of high transmission, spraying of approximately one sixth of all households every 2 weeks may have had an effect on the adult vector population, which led to a smaller breeding pool and lower overall vector density in this area of low transmission. The decrease in malaria incidence in Kipsamoite in 2005, after only 15% of households were sprayed, may in part reflect the effects of greater spraying in neighboring Kapsisiywa. The decrease may also reflect the combined effect of partial coverage with ITNs and additional coverage by IRS. Reduction of incidence in Kipsamoite in 2005 was not caused by concentrated spraying in areas of malaria clustering (5) because spraying was nearly absent in these areas. Reductions in malaria incidence were seen in Kapsisiywa in 2005–2006 after spraying of 40%–50% of households, but the small peak in incidence seen subsequently in these 2 years, but not 2007 (Figure 1), suggests that for interruption, a higher percentage of households (>70%) must be sprayed.

Treatment of malaria patients with co-artemether reduces gametocyte carriage and density in children and makes them less infectious to mosquitoes than treatment with sulfadoxine-pyrimethamine plus chloroquine (13). The effect of co-artemether on gametocytes may have been synergistic with the effect of widespread IRS on the Anopheles spp. vector in reducing malaria transmission. In symptomatic persons, gametocyte prevalence was always low, and it decreased before introduction of co-artemether, but it did not decrease to undetectable levels in Kipsamoite until after introduction of co-artemether. Among asymptomatic persons, studies during 1999–2002 generally demonstrated higher gametocyte prevalence (0%–5.7%) (7). Lower prevalence among asymptomatic persons in the current study (0%–0.2% during May 2007–April 2008) could reflect effects of co-artemether on gametocyte prevalence after introduction of co-artemether in late 2006–early 2007, but without interim data from 2002–2006, an association cannot be clearly demonstrated. As with IRS, however, absence of microscopy-positive malaria cases occurred only after introduction of co-artemether. Because co-artemether was first used during a time of low transmission of malaria, the contribution of ACT to the absence of malaria incidence could not be quantified in the present study. However, a much larger study in South Africa in which IRS treatment and ACT treatment of persons with clinical malaria were introduced sequentially demonstrated an additional reduction of malaria incidence after introduction of ACT (14) and this supports the idea of synergy between these 2 interventions.

Although ITNs are the preferred intervention for preventing malaria-related illness and death in areas of high transmission (1,15,16), ITNs probably did not play a major role in interrupting malaria transmission in the highland areas we studied. ITN coverage never exceeded 30% in either area, and use actually decreased over the study period. In areas of unstable transmission, IRS treatment once a year is likely to be easier, more effective, and more accepted than ITNs. The Roll Back Malaria program currently recommends IRS as the preferred method of reducing malaria in areas of low transmission (17); our study supports this recommendation.

Insecticide treatment using IRS is not without problems; chief among them is potential development of resistance to the insecticide. Lambda-cyhalothrin, the insecticide used for IRS in these areas by the Kenyan Ministry of Health, was used in Mozambique for IRS starting in 1993, but resistance developed to such an extent that lambda cyhalothrin was replaced by bendiocarb in 2000. By 2006, however, lambda cyhalothrin resistance had decreased in many areas (18). In a recent study in nearby areas of western Kenya, no phenotypic resistance to pyrethroid insecticides was seen, but 27% of anophelines carried the knockdown resistance (kdr) mutation associated with increased resistance to pyrethroids (19). Assessment for insecticide resistance in the highland areas of the present study will enable better policy decisions to be made about continued use of lambda cyhalothrin, use of alternatives such as bendiocarb or DDT, or cycling of insecticides when certain resistance thresholds are reached for a particular insecticide. P. falciparum resistance to sulfadoxine-pyrimethamine was present in 27% of infections in western Kenya as early as 1999 (20). Resistance to co-artemether has not yet been documented in Kenya, but development of resistance to components of artemether/lumefantrine in nearby populations with higher levels of transmission would also pose a threat to this highland population. Monitoring of drug resistance to ACT in all areas in which malaria is endemic will be critical for sustaining reduction of malaria incidence in sub-Saharan Africa.

Limitations of our study include missing monthly malaria incidence data among symptomatic persons during 2005–2006, the observational nature of the study, and the possibility that lack of parasitemia by microscopy and PCR in symptomatic and asymptomatic persons was caused by seasonal variation common in highland areas (21) and not by interruption of local transmission. However, 3 pieces of evidence from studies of asymptomatic and symptomatic persons support interruption of local transmission rather than seasonal variation. First, no parasitemia was seen in 2 successive microscopy and PCR surveys of asymptomatic persons, whereas in 5 earlier surveys of asymptomatic persons in Kipsamoite, the area of lower transmission, the frequency of infected persons was never zero (range 5.9%–14.5% by PCR) (7). Second, over 7 years of clinic surveillance of symptomatic persons in Kipsamoite and 5 years in Kapsisiywa, there was never a >4-month period in which there were no microscopy-positive cases of P. falciparum before March 2007. Thus, even with seasonal variation, a year with no microscopy-positive cases in these areas is unprecedented. Third, absence of gametocytemia by microscopy was documented among asymptomatic persons in 2 of 4 assessments and among symptomatic persons for a year (April 2007–March 2008), which suggested that the potential for local transmission was low or absent.

Because reverse transcription–PCR methods for detection of gametocytes have documented higher rates of gametocyte infection than microscopy (22,23), we are developing this testing method in our laboratory to confirm the absence of gametocytemia in the most recent samples from study participants. The presence of asexual P. falciparum infection by PCR in 15 symptomatic persons during April–June 2007 could reflect prolonged detection by this more sensitive method. The presence of only 2 PCR-positive cases in a 9-month period (July 2007–March 2008) suggests that malaria transmission was either interrupted, if these cases were caused by patients’ travel, or reduced to almost undetectable levels.

Sustained elimination of local malaria transmission in these areas will require ongoing surveillance of malaria incidence, anopheline vector density, and anopheline insecticide resistance, and correctly timed IRS campaigns with broad coverage of the area (24). The longer populations at these sites are unexposed to malaria, the more susceptible they are to malaria epidemics, which could occur if an increase in vector density occurs in conjunction with the arrival of infected persons or mosquitoes from an area of higher transmission of malaria. Because travel is increasingly frequent, true elimination of malaria in this and other highland areas will require reduction and eventual elimination of malaria in surrounding areas. Co-artemether must be consistently available to treat any infected and symptomatic travelers or immigrants to the area. Finally, as malaria cases decrease, microscopists will need to receive training to remain proficient in detection of malaria in blood smears.

In summary, this study demonstrates pronounced reduction and possible interruption of malaria transmission in 2 highland areas of Kenya for a 1-year period and provides evidence that interruption of transmission was related to widespread annual IRS insecticide treatment and use of ACT as first-line treatment for uncomplicated malaria. Although both areas remain at risk for recurrence of malaria epidemics, our study provides evidence that interruption and eventual elimination of malaria in areas of unstable transmission may be achievable.