Khanh Phu (12°14′N; 108°56′E) is a commune with ≈3,000 inhabitants, mainly of the Raglai ethnic minority, who live between the forested foothills on the east side of the Truong Son mountain range in south-central Vietnam (Khanh Hoa Province), an area where malaria was previously hyper- to holo-endemic (3). Since 1993, the Medical Committee Netherlands–Vietnam (a Dutch nongovernmental organization) has cooperated with the National Institute of Malariology, Parasitology and Entomology in Hanoi; the Institute of Malariology, Parasitology and Entomology in Qui Nhon; and the Malaria Control Centre of Khanh Hoa Province to set up and operate the Khanh Phu Malaria Field Research Unit to provide Vietnamese malaria researchers with the opportunity to study local malaria epidemiology and develop and test improved control methods. An. dirus species A is currently the only malaria-transmitting species of mosquitoes in Khanh Phu. Anopheles minimus mosquitoes, previously the major malaria vector in the region, disappeared from Khanh Phu after 1998, following the wide-scale use of insecticide-treated mosquito nets (3,12). While the average malaria prevalence in the human population has been greatly reduced (from >50% before 1998 to 2%–3% during 2003–2009), persons who sleep overnight in the forest still run a high risk for infection. Analysis of samples from infected persons by thin-smear microscopy showed that approximately two thirds were caused by P. falciparum, one third by P. vivax, and a very small number by P. malariae during the 2003–2009 study period. This malaria prevalence typically affects the poorest members of the local Raglai ethnic community, whose livelihood partly depends on excursions into the forest to collect products such as bamboo and rattan, or to cultivate their plots on the mountain slopes. All residences within the commune were mapped, and all persons were registered at the research station and assigned unique code numbers based on residential location and family relations. (All methods in this study that involved human participants in the field were certified as permitted standard procedures by the National Institute of Malariology, Parasitology and Entomology in Hanoi.)

Mosquitoes were collected by human-baited landing catches at 4 collection sites in the forest and forest fringe areas near Nga Hai village in the southern part of Khanh Phu commune, from January 2008 through February 2010. Mosquito collectors were adult men of the Raglai ethnic group. They were intensively screened for malaria and promptly treated with artemisinin combination therapy if infected. The collectors worked in teams of 2 over the whole night, 1 person collected from 6:00 pm to 12:00 am and the other from 12:01 am to 6:00 pm. The monthly collection effort ranged between 40 and 60 person-nights per month, a total 1,285 person-nights over the 26-month period.

Anopheles species were determined on the basis of morphologic features (13). All An. dirus group mosquitoes were assumed to be An. dirus species A on the basis of previous accurate identifications and the known distribution of this species (14,15). Female anopheline mosquitoes were dissected for salivary glands, midgets, and ovaries, and these were examined by microscopy for the presence of sporozoites, oocysts, and parity, respectively. Sporozoite-infected glands were applied to filter paper and dried in an ambient atmosphere before storage in closed vials at 4°C–6°C.

Blood samples were collected by 2 methods. First, to detect as many parasite-infected persons as possible, from March 8, 2009, through February 28, 2010, blood was collected by targeted active case detection from persons who had frequently worked in the forest or had reported fever. A total of 549 blood samples from 305 persons from 156 families were collected; 183 persons were sampled once, and no one was sampled >8 times; the sex ratio had a male bias (64% male). Sixty-nine of the 121 blood samples that were positive by microscopy, and an additional randomly selected 105 of the negative samples were processed by PCR to determine Plasmodium species. Second, from March 11, 2010, through March 30, 2010, a cross-sectional survey was undertaken in the 2 villages nearest to the mosquito collection sites. A single blood sample was taken from each of 624 residents, irrespective of symptoms or work place. One hundred and thirty-five persons in this sample had previously given samples as part of the targeted active case detection (ACD) survey; 489 had not given samples previously. From the 49 blood samples found positive by microscopy, 37 were randomly chosen for PCR processing.

Sixty-eight percent of all blood slides in both samples were from persons in the 2 southernmost villages of Khanh Phu (Nga Hai and Da Trai). This represents an area where ≈1,100 persons are living, spread over an area of 1.7 km² at 12°12.5′N and 108°55.5′E. All their houses are located <1 km from the nearest forest and 1–3 km away from the mosquito collection sites. An. dirus mosquitoes were rarely caught in these villages.

Blood was collected by finger-prick; thick and thin blood films were made for diagnosis, and blood was applied to filter paper for downstream molecular analyses. All adult volunteers provided informed consent and for children, consent was obtained from close relatives. All persons found to be infected with parasites (as diagnosed by microscopy) received treatment, according to the policy of the Vietnam Ministry of Health.

DNA was extracted from dried blood samples on filter paper and from sporozoite-infected glands, and subsequent malaria parasite species identification by PCR that targeted the 18S rRNA gene was carried out as previously described (11,16). Briefly, DNA was extracted by using the QIAamp DNA micro kit (QIAGEN, Tokyo, Japan). Extracted DNA samples were stored at –20°C until use. Plasmodium species–specific nested PCR assays to detect and identify human malaria parasites were performed as described (17). For detection of the P. knowlesi 18S rRNA gene, the primers Pmk8 and Pmk9 were used (5). Because this primer set can occasionally cross-react with P. vivax DNA and produce false-positive results (18), samples were also subject to a further PCR targeting of the P. knowlesi circumsporozoite protein (CSP) gene as described (8). PCR products were separated by electrophoresis on 1.5% agarose gels and stained with ethidium bromide. Primer sequences for the gene of human Plasmodium species (17), 18S rRNA of P. knowlesi (5), and the CSP gene of P. knowlesi (8) were as previously described. P. knowlesi H strain (American Type Culture Collection no. 30158) (kindly donated by Satoru Kawai, Dokkyou University, Japan) was used as a positive control. We verified that no cross-reaction occurred between the primer sets used to amplify P. vivax CSP and P. knowlesi CSP by using DNA extracted from single infections of both species (data not shown).

Student 1- or 2-tailed t tests were used for comparing means. Chi-square tests for association and confidence limits were calculated with Microsoft Excel (Microsoft Corp., Redmond, WA, USA).

From January 1, 2008, through February 22, 2010, a total of 6,834 female anopheline mosquitoes were captured, of which 83.2% belonged to An. dirus s.s., and 8.6% to An. maculatus, 5.4% to An. peditaeniatus; the remaining 2.8% were divided between a further 11 anopheline species in very small numbers. Of the 8,331 An. maculatus mosquitoes dissected in Khanh Phu over >15 years, only 1 mosquito was ever found to be infected with sporozoites. None of the other Anopheles species, with the exception of An. dirus, were ever found to be infected with sporozoites. Thus An. dirus mosquitoes are regarded as the only malaria vectors in the Khanh Phu forest.

An. dirus mosquitoes were found during every month of sampling with an average human-biting density of 4.4 bites/person-night (range 0.3–17.4; Table A1). As in previous years in Khanh Phu (3), these mosquitoes usually reached peak densities in the dry season (February–April). Numbers were lowest during the hottest part of the year (May–August), climbed to a second peak with the onset of the rains during September–November, and usually decline again following heavy rains during October–December. The monthly average parous rate fluctuated between 65% and 90% (average 77%).

Sporozoites were detected by microscopic screening in 89 (1.57%) of 5,663 dissected An. dirus mosquitoes, and oocysts were observed on 0.94% of the midguts. The sporozoite rate showed some seasonal fluctuations, but these did not clearly correlate with the fluctuations in biting density. Multiplying the sporozoite rate by the human-biting rate gave an average annual entomological inoculation rate of 25.4 infective bites per person per year. This is comparable with the average rate estimated during the previous 8 years in the forest of Khanh Phu (22 infective bites per person per year, unpub. data). However, both the sporozoite rate (4.6%) and human-biting density (12 bites/person-night) reached a maximum in the last 2 months of the survey (January and February 2010). This finding implies that persons who slept unprotected in the forest during this period would have had a >50% chance of being bitten by an infected An. dirus mosquito during any 1 night.

Of the 86 sporozoite-positive mosquitoes, 73 underwent PCR analysis for malaria parasite detection, of which 72 were successfully assayed. Thirty-one (43%) of these 72 salivary glands were PCR positive for P. knowlesi csp and for P. knowlesi 18S rRNA. Five additional specimens were positive for P. knowlesi 18S rRNA but negative for P. knowlesi CSP and were therefore regarded as P. knowlesi negative (Figure 1). One specimen was only positive for P. knowlesi CSP and not by any other test, and therefore the species could not be confidently determined. The frequencies of P. falciparum, P. vivax, and P. malariae were: 50%, 50%, and 6%, respectively. In 22 (71%) of 31 glands, P. knowlesi was found as a co-infection with one of these other species. The combination P. knowlesi + P. vivax (in 14 glands) was far more common than P. knowlesi + P. falciparum (1 case). The combination of P. knowlesi + P. vivax + P. falciparum was, however, also quite common (7 glands).

The frequency of P. knowlesi–positive mosquitoes did not significantly differ between the 4 collections sites, and P. knowlesi was detected during 8 months of the past year, except in August, September, November, and December when sporozoite-infected mosquitoes were few (Figure 2). However, the highest number of sporozoite-infected mosquitoes and 13 of the 31 P. knowlesi infections were found in January 2010.

Of the 549 blood samples collected by targeted ACD from March 2009 through February 2010, 121 (22.0%) were parasite positive by microscopy, including 94 (17.1%) that were positive for P. falciparum. The cross-sectional survey conducted in March 2010 showed a parasite prevalence of 7.9% (10.5% in male residents and 5.2% in female residents; χ² p = 0.013). The P. falciparum prevalence was 5.1% (7.3% in male residents and 2.9% in female residents; χ² p = 0.012).

The Table shows the results of parasite species identification by microscopy and PCR for both the targeted ACD and cross-sectional surveys. Both samples had a male bias. In the targeted ACD sample, this result reflected the skewed sex ratio in the total 549 blood samples. Among the 624 persons sampled during the cross-sectional survey the sex ratio was 50/50, but because infections were more common in men, the random subsample of 37 for PCR analysis of the 49 positive results by microscopy was also male biased. The mean age of women and men was similar, and the difference in mean age between findings of the targeted ACD and those of the cross-sectional survey was not significant (t test p>0.05). In the targeted ACD sample, 19 (18%) of 105 samples that were negative by microscopic examination were parasite positive by PCR. Five patients with positive results in the ACD were shown by PCR to have P. malariae as a co-infection, and 1 patient in the cross-sectional survey had a P. malariae single infection, all of which were not seen by microscopy. We were unable to determine the parasite densities of P. knowlesi in our samples because it was not possible to discriminate between parasite species in co-infections at low parasite densities.

A total of 32 blood samples were found to be infected with P. knowlesi on the basis of simultaneous positivity in the P. knowlesi 18S rRNA and P. knowlesi csp PCR tests: 19 (51%) of 37 malaria-infected persons in the cross-sectional survey, and 13 (15%) of 88 PCR-positive cases collected by targeted ACD. Four additional specimens in the ACD sample were positive for P. knowlesi 18S rRNA but negative for P. knowlesi csp, all in co-infections with other parasite species, most commonly P. vivax. Human blood samples positive for P. knowlesi csp and negative for P. knowlesi 18S rRNA were not found.

P. knowlesi infection was detected by targeted ACD in 8 of 12 months. On the basis of the PCR results of the samples from the ACD survey, the annual P. knowlesi incidence can be estimated at 10/1,000 person-years (95% confidence limit 4%–17%). The results from the cross-sectional survey imply a peak P. knowlesi prevalence in March 2010 among the population of the southern villages of 12.6% (95% confidence limit 6.3%–21.1%). The large increase in P. knowlesi co-infections in humans in March 2010 correlates with the increased frequency of this parasite in the mosquitoes in the preceding months.

Twenty-eight of the 32 P. knowlesi infections were found in persons of Raglai ethnicity, the dominant group in the study area. However, 2 of the 5 persons of Trinh ethnicity and 2 of the 7 persons of Kinh ethnicity sampled also had a P. knowlesi co-infection; all had asymptomatic cases detected during the cross-sectional survey.

The average age of persons with P. knowlesi co-infections was 15.0 years in the targeted ACD survey and 16.9 years in the cross-sectional survey; both ages were significantly lower than the ages of P. knowlesi–infected persons in both samples (24.8 and 26.2 years, respectively; 2-tailed t test p<0.001; Figure 3, panel A) and those of all other positive case-patients (24.8 and 24.1 years, respectively; 2-tailed t test p< 0.01; Figure 3, panel B). P. falciparum–infected persons had a mean age of 23.4 years, which did not differ significantly from the mean age of 25.6 years for P. falciparum–negative persons (2-tailed t test p>0.3; Figure 3, panel C).

Six (19%) of the 32 P. knowlesi co-infected persons had fever. This finding was not significantly different from the 10% with fever among 86 uninfected persons in the PCR-analyzed sample (χ² p = 0.23). Only persons with single P. falciparum infections were significantly more often febrile (30%) than uninfected persons (χ² p = 0.003).

In the human blood samples, P. knowlesi was only ever found in a co-infection, almost always with P. vivax and often with P. falciparum in addition. The lack of the P. knowlesi and P. falciparum without P. vivax combination (only 1 case each in the mosquito and human samples; Figure 4, panels A and B) contributed most strongly to the significance of the nonrandom distribution (χ² p<0.001). P. knowlesi infection was equally frequent among PCR-positive women and men, whereas P. falciparum was significantly more frequent among the PCR-positive men (Table).

These findings may fundamentally change the perspectives for the control of forest malaria. Previously, forest malaria may have been considered manageable because the parasite reservoir in the forest may be reduced through intensive case detection and treatment of human communities living in or near the forests. The likely presence of a nonhuman reservoir of P. knowlesi (and possibly other parasites) in monkeys and of a mosquito vector that intensively inoculates parasites among monkeys and humans reinforces the need to find methods of vector control or biting prevention that can be applied to An. dirus mosquitoes.