Arboviral Etiologies of Acute Febrile Illnesses in Western South America, 2000–2007

Background Arthropod-borne viruses (arboviruses) are among the most common agents of human febrile illness worldwide and the most important emerging pathogens, causing multiple notable epidemics of human disease over recent decades. Despite the public health relevance, little is know about the geographic distribution, relative impact, and risk factors for arbovirus infection in many regions of the world. Our objectives were to describe the arboviruses associated with acute undifferentiated febrile illness in participating clinics in four countries in South America and to provide detailed epidemiological analysis of arbovirus infection in Iquitos, Peru, where more extensive monitoring was conducted. Methodology/Findings A clinic-based syndromic surveillance system was implemented in 13 locations in Ecuador, Peru, Bolivia, and Paraguay. Serum samples and demographic information were collected from febrile participants reporting to local health clinics or hospitals. Acute-phase sera were tested for viral infection by immunofluorescence assay or RT-PCR, while acute- and convalescent-phase sera were tested for pathogen-specific IgM by ELISA. Between May 2000 and December 2007, 20,880 participants were included in the study, with evidence for recent arbovirus infection detected for 6,793 (32.5%). Dengue viruses (Flavivirus) were the most common arbovirus infections, totaling 26.0% of febrile episodes, with DENV-3 as the most common serotype. Alphavirus (Venezuelan equine encephalitis virus [VEEV] and Mayaro virus [MAYV]) and Orthobunyavirus (Oropouche virus [OROV], Group C viruses, and Guaroa virus) infections were both observed in approximately 3% of febrile episodes. In Iquitos, risk factors for VEEV and MAYV infection included being male and reporting to a rural (vs urban) clinic. In contrast, OROV infection was similar between sexes and type of clinic. Conclusions/Significance Our data provide a better understanding of the geographic range of arboviruses in South America and highlight the diversity of pathogens in circulation. These arboviruses are currently significant causes of human illness in endemic regions but also have potential for further expansion. Our data provide a basis for analyzing changes in their ecology and epidemiology.


Introduction
Over the past few decades there has been a global resurgence of arthropod-borne viral pathogens (arboviruses) worldwide [1,2], particularly those transmitted by mosquitoes. Despite the public health relevance, the geographic range, relative impact, and epidemiologic characteristics associated with arbovirus infection are poorly described in many regions of the world. Arboviruses are a heterogeneous group, but those of medical relevance largely belong to a few virus genera, including Flavivirus, Alphavirus, and Orthobunyavirus. Prominent examples of emergent arboviruses include West Nile virus (WNV; Flavivirus) in North America, Japanese encephalitis virus (JEV; Flavivirus) in Asia, chikungunya virus (CHIKV; Alphavirus) in the Indian Ocean region and dengue viruses (DENV; Flavivirus) worldwide. One common feature shared by many emergent arboviruses is the capacity to expand host and geographical range, owing in part to the plasticity of the RNA genome [3]. Some arboviruses have evolved to exploit humans as their primary reservoir, while others rely on birds or peridomestic animals, with human infection resulting from spill-over from zoonotic replication cycles. Human exposure to sylvatic arbovirus cycles is likely to increase as activities continue to encroach on forested areas worldwide. Tropical areas in particular, with yearround hot and humid conditions, are well-suited for maintaining arboviruses with potential to emerge as significant human pathogens [4]. In the neotropics alone, greater than 145 distinct arbovirus species have been identified [4], many of which have already been associated with human illness.
One limitation of conducting surveillance for arboviral diseases is the generic nature of disease presentation. While severe disease can result, such as hemorrhagic manifestations (DENV and yellow fever virus [YFV]) or neurological disease (WNV, JEV, and Venezuelan equine encephalitis virus [VEEV]), arbovirus infection typically results in mild to moderate febrile illness [2,5,6]. Particularly early in disease progression, patients commonly present with undifferentiated febrile illness [5,7] rendering clinical diagnosis unreliable [8]. In DENV-endemic areas, for example, diseases caused by co-circulating pathogens have been found to be often misclassified [8,9]. In light of the lack of distinct clinical presentation and the diversity of the etiologic agents, laboratory support has become a critical component of effective surveillance programs.
The impact on human health in endemic regions and the potential for broader spread underscore the importance of improving understanding of arbovirus transmission patterns. Currently the epidemiological characteristics and geographic range for many endemic arboviruses in South America are poorly understood. To begin to address this gap, we established a laboratory-supported clinic-based study to identify the etiologic agents associated with undifferentiated febrile illness in sites in Peru, Ecuador, Bolivia, and Paraguay. Herein we describe the geographic distribution of distinct arboviruses and their relative contribution to human febrile illness in these study sites. In addition, we present the temporal trends and epidemiological characteristics associated with arbovirus infection in Iquitos, Peru, a site where more extensive monitoring was conducted.

Study Locations
In 1990 the U.S. Naval Medical Research Center Detachment (NMRCD) initiated a clinic-based surveillance program to determine the etiologies of febrile illness in Iquitos, Peru [10][11][12]. In 2000 NMRCD collaborated with local Ministries of Health to expand the surveillance program into other regions of Peru and South America, including sites in Ecuador, Bolivia, and Paraguay. In addition to Iquitos, in 2000 the study was implemented at regional sites in or near Piura, Cusco, Tumbes, and Yurimaguas, Peru, as well as Santa Cruz, Bolivia ( Figure 1; Table 1). Additional sites were later added in Concepción, Magdalena, and Cochabamba (Villa Tunari and Eterazama), Bolivia; Guayaquil, Ecuador; Asunción, Paraguay; and La Merced and Puerto Maldonado, Perú. Participants were recruited when reporting with acute febrile illness to public, private, or military health facilities in and around these regional centers. Details of the study sites are described in Table 1. Study sites were selected based largely on locations in hot and humid climates conducive for arbovirus transmission, typically situated near or in tropical rainforest regions. Notable exceptions include Piura, Tumbes, and Cusco, which are located in coastal desert (Piura and Tumbes) or highlands (Cusco; Figure 1 and Table 1) regions. It should be noted that the study staff in Cusco (Hospital Regional) on occasion attended to participants arriving from surrounding highlands rainforest regions.

Study subjects
Study subjects included patients 5 years of age or older who presented in outpatient clinics or hospitals with acute, undifferentiated, febrile illness (greater than or equal to 38uC for 7 days duration or less) along with one or more of the following symptoms: headache, muscle, ocular and/or joint pain, generalized fatigue, cough, nausea, vomiting, sore throat, rhinorrhea, difficulty breathing, diarrhea, jaundice, dizziness, disorientation, stiff neck, or bleeding manifestations. Children younger than five years of age were included if they presented with hemorrhagic manifestations indicative of dengue hemorrhagic fever (DHF), including epistaxis, pleural effusion, platelets less than 100,000/ml, petechiae, or bloody stool or vomit. Exclusion criteria included fever in excess of seven days or an identifiable focus of infection, such as sinusitis, pneumonia, acute otitis media, or acute urinary tract infection. Demographic data, medical history, and clinical features for each patient were obtained using a standard questionnaire. In malaria-endemic regions if malaria was suspected, capillary blood from febrile patients was screened for Plasmodium spp. by clinic or hospital personnel according to routine diagnostic procedures at each site. Peripheral blood samples were screened by microscopic analysis of stained thick

Author Summary
Over recent decades, the variety and quantity of diseases caused by viruses transmitted to humans by mosquitoes and other arthropods (also known as arboviruses) have increased around the world. One difficulty in studying these diseases is the fact that the symptoms are often non-descript, with patients reporting such symptoms as low-grade fever and headache. Our goal in this study was to use laboratory tests to determine the causes of such non-descript illnesses in sites in four countries in South America, focusing on arboviruses. We established a surveillance network in 13 locations in Ecuador, Peru, Bolivia, and Paraguay, where patient samples were collected and then sent to a central laboratory for testing. Between May 2000 and December 2007, blood serum samples were collected from more than 20,000 participants with fever, and recent arbovirus infection was detected for nearly one third of them. The most common viruses were dengue viruses (genera Flavivirus). We also detected infection by viruses from other genera, including Alphavirus and Orthobunyavirus. This data is important for understanding how such viruses might emerge as significant human pathogens. smear slides. In some sites, owing to the possibility of arbovirus coinfection, malaria-positive patients were subsequently invited to participate in the NMRCD study, with malaria results recorded along with symptoms and demographic information.
During the acute phase of illness blood samples were obtained from each patient, and when possible, convalescent samples were obtained 10 days to 4 weeks later for serological studies. For patients older than 10 years of age, up to 15 mL of blood was collected, and for patients younger than 10 years of age, up to 7 mL of blood was collected. Trained phlebotomists collected blood samples via arm venipuncture using standard methods and universal precautions.

Laboratory Procedures
Virus isolation. Acute-phase serum samples were transported on dry ice to the U.S. NMRCD laboratory in Lima and stored at 280uC. Serum samples were thawed and diluted 1:10 in Eagle's minimum essential medium (EMEM) containing 2% heat-inactivated fetal bovine serum and antibiotics (200 U/ml penicillin and 200 mg streptomycin). African Green Monkey Vero (37uC) and Aedes albopictus C6/36 (28uC) cell cultures were each inoculated with 200 ml of the diluted serum in 25 ml flasks. Upon observation of cytopathic effect (CPE) with a light microscope, or ten days post-inoculation if no CPE was observed, cells were removed from the flasks, collected by centrifugation for 10 minutes, and placed on 12-well glass spot-slides for microscopic examination by standard indirect immunofluoresence assay (IFA). Viral antigens were detected with hyperimmune mouse ascitic fluid (HMAF) produced in the NMRCD-Lima laboratory against the arbovirus isolates listed below, followed by the addition of fluorescein-conjugated goat anti-mouse IgG, similar to previously described [11,13,14]. DENV serotypes were identified using serotype-specific monoclonal antibodies (DENV-1: 15F3, DENV-2: 3H5; DENV-3: 5D4; DENV-4: 1H10).
Serology. IgM titers were determined by using IgM-capture enzyme-linked immunosorbant assays (ELISA), as previously described [15,16]. Briefly, microtiter plates (96-well format) were coated with anti-human IgM (Tago, Inc., Burlingame, CA) diluted 1:800 in phosphate-buffered saline (PBS) and incubated overnight at 4uC. Participant serum was diluted 1:100 and incubated in coated wells for 1 h at 37uC, followed by the addition of viral antigen and incubation at 37uC for 1 h. Viral antigens were detected with HMAF (produced by inoculation of mice with the respective viral strains), followed by horseradish peroxidaseconjugated goat anti-mouse IgM + IgG (Pierce, Rockford, IL). Following the addition of ABTS (2,29-azino-bis-[3ethylbenzthiazoline-6-sulfonic acid]) colorimetric substrate, plates were read at 410 nm with a Dynex ELISA MRX Revelation absorbance reader (Dynex Technologies, Chantilly, VA). All acute-and convalescent-phase samples were initially screened at 1:100. Previously identified reactive patient sera were used as positive controls, and normal human serum was used as a negative control. Samples exceeding the reference cut-off value, calculated as the mean of seven antibody-negative samples (normal human serum) plus three standard deviations, were considered IgMpositive. Positive samples were subsequently re-tested at four-fold serial dilutions (1:100, 1:400, 1:1600, and 1:6400). The highest dilution at which the OD still exceeded the cut-off value was considered the final titer.
Case Definitions. Diagnoses were considered confirmed if they met any of the following criteria: isolation of virus from the specimen, virus identification by RT-PCR, or 4-fold or greater increase in IgM antibody titers between acute and convalescent samples. Diagnoses were considered presumptive if elevated IgM titers (positive at dilutions greater than or equal to 1:100) were detected in the acute or in both acute-and convalescent-phase samples without a four-fold rise between samples. Participants with evidence of recent infection by more than one arbovirus are included in each arbovirus category for which such evidence was observed, as the etiologic agent could not be definitively identified. The exception to this was within the Flavivirus genus. Since we observed significant IgM cross-reactivity between DENV and YFV antigen (see Results), participants were only considered YFVpositive by serology in the absence of reactivity to DENV antigen in the IgM ELISA. In the absence of laboratory evidence of any of the above pathogens, cases were classified as arbovirus-negative.

Statistical Analyses
Statistical analyses (Chi-square, Fisher's exact test, and logistic regression) were performed in R version 2.8 (The R Foundation for Statistical Computing, Vienna, Austria) [20]. The significance level was set at a = 0.05.
Of the 20,880 cases of febrile illness included, paired acute and convalescent-phase samples were collected from 13,259 participants (63.5%), while acute-phase only (without convalescent samples) were available from 7,621 participants (36.5%; Table 2). Most participants for whom data was available reported to a health center within four days following disease onset (15,911/ 19,632; 81.0%). A subset of participants (n = 9,800; 46.9%) were screened for malaria by thick smear prior to enrollment. Of these, 584 (6.0%) were found to be positive. Patient samples were screened for recent infection by members of the Flaviviridae,

Malaria co-infections
DENV was the most common malaria co-infection, observed for 11.3% (66/584) of participants reported as malaria-positive, including 17 DENV-3 isolates and one DENV-1 isolate. The rate of DENV infection was significantly lower for malaria-positive participants than for malaria-negative participants (11.3% vs 32.9%; p,0.0001). In contrast, VEEV infection was more common among malaria-positive participants (7.4%) as compared to malaria-negative participants (2.7%; p,0.0001). There were no significant differences between malaria thick smear-positive participants and thick smear-negative participants for the other arboviruses studied (data not shown).

Flavivirus
DENV serotypes were the predominant arboviruses detected, accounting for 26.0% of febrile episodes analyzed (Table 5), based on virological (2,662 virus isolates or RT-PCR positives, with or without supporting serology) and serological (1,058 IgM seroconversions and 1,700 participants with elevated DENV IgM, without accompanying positive results by virus isolation or RT-PCR) evidence. Considerable YFV cross-reactivity was observed for DENV-positive samples. Based on the 2,662 cases with definitive DENV diagnosis (IFA or RT-PCR confirmation in the acute sample), 847 (32.0%) also had IgM reactive to YFV antigen in the acute or convalescent sample.
DENV-3 was most commonly isolated serotype, accounting for 81.1% (2,159/2,662) of DENV isolates over the course of the study. In our study, DENV-3 was first detected in sites along the northern coast of Peru (Piura and Tumbes) in 2000 ( Figure 2) during a large outbreak of dengue fever in the region [21,22], although DENV-1 and DENV-2 were the most commonly isolated serotypes during this outbreak. DENV-3 quickly became the dominant serotype in the northeastern rainforest (Iquitos and Yurimaguas), with limited DENV-1 co-circulation in the region in subsequent years (  Figure 2). DENV-4 was rarely detected during the study period, with only four isolates from study participants. However, more recently this situation has changed dramatically with the 2008 emergence of DENV-4 in northern Peru [23].
Overall, DENV infection was more common among female participants than male participants (28.1% vs 25.5%; p,0.0001), with a statistically significant bias towards older participants (28.0% of participants 30 or older were DENV-positive as compared with 25.9% of participants younger than 30; p = 0.001). However, the epidemiology of DENV infection varied by study site, particularly within Peru. The prevalence of DENV infection was higher among older participants in Puerto Maldonado (p = 0.01), La Merced (p,0.0001), Piura (p,0.0001), and Tumbes (p = 0.015), while no elevated DENV prevalence was observed for older participants in Yurimaguas and Iquitos. In addition, DENV infection was more common among female participants in Puerto Maldonado (p,0.0001), La Merced (p,0.0001), and Tumbes (p = 0.028), but no statistically significant difference was observed in Iquitos, Yurimaguas, or Piura.
Other than DENV, the only flavivirus isolated during the course of the study was YFV, which was isolated from four participants ( Table 4). In addition to the four isolates, serological evidence of recent YFV infection (without evidence of DENV infection) was detected in an additional 494 participants, including 143 who seroconverted between acute and convalescent samples. Overall, data on prior vaccination was available for 17,816 participants, 10,667 (59.9%) of whom reported having received YF vaccination. YF vaccine coverage varied widely by study site, ranging from less than 10% in non-endemic sites along the northern coast of Peru (Piura and Tumbes) to 59% in Iquitos, 70% in Yurimaguas, and 77% or greater in Cochabamba (Villa Tunari and Eterazama), Concepción, Junin, Magdalena, and Puerto Maldonado. Study participants with evidence of recent YFV infection based on IgM were significantly more likely than the rest of the overall study population to have reported receiving YF vaccination within the previous 6 months (30.0% vs. 6.7%; p,0.0001).

Alphavirus
Recent alphavirus infection was detected for 3.1% (n = 645) of febrile patients (Table 5), including 102 VEEV isolates and 40      Presumptive cases (elevated IgM without 4-fold rise between acute and convalescent). 3 Percent of total febrile cases from this location. 4 Many of the participants with YFV-reactive serum are likely due to recent vaccination or Flavivirus cross-reactivity, and thus are not reflective of natural YFV infection (see text). 5 Only limited serologic testing for GROV IgM was performed prior to 2007. Virus isolates (  Unlike the flaviviruses, little serologic cross-reaction (or, alternatively, concurrent infection) was observed among alphaviruses. For the 72 participants with the most well-defined VEEV infections (IFA-positive, plus a convalescent sample available for testing), 7 (9.7%) had IgM reactive to MAYV antigen in either the acute or convalescent sample. For the 24 cases where MAYV was isolated and a convalescent sample was available for testing, no crossreactivity was observed in the VEEV IgM ELISA.

Orthobunyavirus
Arboviruses belonging to the Orthobunyavirus genus of the Bunyaviridae family accounted for approximately 2.5% of all febrile cases (Table 5). In total there were 54 orthobunyavirus isolates, including 30 Group C viruses, 18 OROV isolates, and six GROV ( Table 4). The Group C virus isolates were not definitively identified; however, based on serological techniques (ELISA and PRNT), ten were antigenically related to CARV and six were antigenically related to MURV, while 14 could not be antigenically distinguished. Nearly all Simbu Group (OROV) and Group C virus isolates were collected from patients reporting to clinics in Iquitos, Madre de Dios, and Yurimaguas, Peru, while three out of six GROV isolates were obtained from patients in La Merced, Peru, in January and February of 2007 (Table 4). As with the alphaviruses, little serologic cross-reaction was observed within the Orthobunyavirus genus. For the 24 participants IFA-positive for a Group C virus and a convalescent sample available, only 2 had IgM reactive to OROV antigen in either the acute or convalescent sample (8.3%); for the 12 OROV IFA-positive participants with a paired convalescent sample, no reactivity with CARV or MURV antigen was observed.

Incidence rates and epidemiology of arbovirus infection in Iquitos, Peru
The Iquitos health centers included in this study cover a geographically stratified area of the city and in 2007 represented nearly 20% (10 of 55) of civilian public health centers in the greater urban health network. Based on the populations assigned to each health center by the local ministry of health (Dirección Regional de Salud -Loreto), in 2007 clinics included in this study were designated to serve approximately 43% of the population of the Iquitos region. Using the population numbers assigned to each health center by DIRESA-Loreto, we estimated incidence rates for the most common arboviral infections in Iquitos beginning with the first full year of the study (Table 6). Over the course of the study, there were 855.9 acute undifferentiated febrile episodes per 100,000 people per year, peaking during periods of highest dengue activity (2002 and 2004; Table 6). DENV incidence rates varied greatly, peaking in 2002 with 554.0/100,000 following the introduction of DENV-3 and averaging 274.7/100,000 over the 7-year period. The average symptomatic incidence rates for other predominant arboviruses were 28.1/100,000 for VEEV, 8.5/ 100,000 for MAYV, 14.3/100,000 for OROV, and 14.2/100,000 for Group C viruses. Peak transmission rates were observed for these four viruses between 2004 and 2006, including a previouslydescribed outbreak of VEEV in 2006 [24].
It should be noted that the incidence rates above only reflect the participants enrolled in the study. Starting in 2006, demographic data was collected for those who reported to Iquitos health centers and fulfilled the inclusion criteria (acute undifferentiated febrile illness of fewer than 7 days in duration) but declined participation in the febrile surveillance study. In 2006 and 2007, 3,385 and 3,283 febrile patients, respectively, fitting the case definition were examined by study personnel, with 43.3% (n = 1,433) and 36.5% (n = 1,197) of patients agreeing to provide venous blood samples for the surveillance program. During these two years, 94.3% of febrile patients were first screened for malaria by thick smear, with 32.4% of those screened classified as positive. Malaria-negative patients were significantly more likely to accept participation in the surveillance study (51.3%; 2,185/4,256) than malaria-positive patients (9.5%; 193/2,036; p,0.001). Children were significantly less likely to participate than adults (25.3% of eligible children vs. 44.0% of eligible adults chose to participate; p,0.001).
To begin to describe the epidemiology associated with these arboviruses in the Iquitos, demographic characteristics of participants with recent infection by the most common pathogens -DENV serotypes, VEEV, MAYV, OROV, and Group C viruses -were compared with the rest of the participating febrile population in Iquitos. YFV infection, as determined by positive IgM ELISA, was significantly associated with self-reported recent YF vaccination (OR 2.30, 95% CI 1.44-3.57); no similar association was observed for other arboviruses. Thus no further analyses were conducted for YFV IgM-positive participants. Overall, male participants were more common than female (51.4% vs. 48.6%), consistent with the population of Loreto Department as a whole (51.2% male; p = 0.77) [25]. The median age of study participants was 23 (average 26.1), with the highest percentage of participants between the ages of 15 and 29. Both MAYV (p = 0.003) and VEEV (p = 0.009) infection were significantly more common among males, and this effect was only observed among the older age groups (15 years or older), suggesting an occupational exposure. A similar trend for higher prevalence of Alphavirus infection among males was observed in Yurimaguas and Puerto Maldonado, although these analyses were limited by small sample size. Group C virus infection was more common in males in these three sites, although the differences were only statistically significant in Puerto Maldonado (p,0.01). In Iquitos no significant differences were observed between sexes for DENV or OROV (Table 7). DENV infection was significantly more common among participants younger than 15 in Iquitos (p = 0.005); however, this effect was only observed during the earlier years of the study (2002 and 2003 in particular). OROV infection in Iquitos was significantly more common among age groups 15 or older (p = 0.007). For the 30-44 year old age group, MAYV infection was significantly more common than for participants younger than 15 ( Table 7).
The health centers in the Iquitos area included in this study were predominantly public clinics and hospitals located within the urban area of the city (n = 8), although the study was also conducted in three military clinics located within the urban area Table 6. Yearly incidence rates of arboviral diseases in Iquitos, Peru, per 100,000 residents. and two public clinics located in rural zones between approximately five and ten kilometers outside the city limits. The majority of participants in the Iquitos area were recruited in the urban clinics (79.6%), while 10.6% and 9.8% were recruited at the two rural clinics and three military clinics, respectively. Using the different categories of clinics as a proxy for potential differences in arbovirus exposure, we compared the relative prevalence of arboviruses among those reporting to the urban, rural, or military clinics. DENV infection was far more common in participants reporting to the urban clinics than rural clinics, whereas VEEV (p = 0.018), MAYV (p,0.001), and Group C viruses (p,0.001) were more common among those reporting to the rural clinics. For OROV infection there was no statistically significant differences among the types of health centers (Table 7). Participants were recruited year-round, with a peak in December that was largely due to DENV transmission ( Figure 3A). Transmission of the arboviruses peaked during different months of the year. Over the course of the study, DENV transmission was most common between October and December with lowest levels between June and August ( Figure 3A). Alphavirus transmission was highest between February and July ( Figure 3B), with reduced transmission during the second half of the year, while the highest percentage of Group C virus cases was observed between December and February ( Figure 3C).

Discussion
Tropical areas, with year-round hot and humid conditions, are particularly well-suited for maintaining arboviruses with both current public health importance as well as the potential to emerge as significant human pathogens [4]. Therefore in this study we focused on arbovirus transmission in tropical regions of four countries in South America. Our data demonstrate that arboviruses are a common cause of human febrile illness in these sites in South America, accounting for greater than 30% of the febrile episodes analyzed. Importantly, arbovirus-associated disease was not restricted to DENV in most of the locations studied. The other arboviruses identified, including VEEV, MAYV, and OROV, in total were associated with approximately 8% of febrile episodes. Our study has provided source material for various phylogenetic analyses [12,14,19,23,26] and will provide important baseline data for monitoring changes in arbovirus ecology, epidemiology, and genetics.
There were several significant limitations to our study. First, with the exception of Peru, the number of study sites in the other countries was quite limited. Even in Peru, it is unclear whether these results are indicative of arbovirus circulation in other regions of the country. Another shortcoming was the focus on arboviruses. While clearly these are important pathogens in the tropical rainforest regions, along the desert coast (Piura) and in the highlands (Cusco), other types of pathogens will need to be given greater consideration. Another limitation of our study design is the passive surveillance strategy employed. Clinic-based surveillance is likely to significantly underestimate true arbovirus circulation, as those with milder disease manifestations are less likely to visit a health center. In studies of DENV transmission in Iquitos we have observed that incidence rates calculated from community-based active surveillance are several times higher than those calculated based on passive surveillance (TJK, ACM, and BMF, unpublished results). Accordingly the incidence rates presented here should be interpreted carefully and considered a conservative estimate of the true number of febrile episodes caused by each virus. Another shortcoming of clinic-based surveillance is the difficulty of extrapolating the data to the entire population. As we show here in Iquitos, those who present to the health centers and those that are willing to participate in these studies are often not fully representative of the population at-large, which may lead to biases in age-dependent incidence rates. In addition, with the exception of Iquitos, we did not collect sufficient data from non-participants to fully contextualize these results. Our data suggest that malaria may contribute to approximately 30% of acute febrile illnesses in Iquitos, a figure that would not be apparent based solely on those who enrolled in the study. One advantage of clinic-based passive surveillance is expanded geographic coverage and more limited costs relative to other surveillance strategies, which is critically important when studying the relatively obscure arboviruses described here. Only through the large number of participants presented here were we able to detect sufficient cases of VEEV, MAYV, and OROV for further epidemiological analysis.
Over the course of the study, DENV serotypes were by far the most common arboviruses associated with febrile disease, accounting for 26% of febrile participants. DENV serotypes have  [27,28]. Here we demonstrate that DENV-3 (previously identified as subtype III [26]) was the predominant serotype in the region between 2001-2007, although we also observed significant DENV-1 and DENV-2 transmission in certain regions. Not surprisingly DENV circulation was found to be more regiondependent than country-dependent. Specifically, Tumbes and Piura along the coast of northern Peru share common trends with Guayaquil in Ecuador (Figure 2A), while DENV circulation in Puerto Maldonado in southern Peru is more closely tied to trends observed in Bolivia ( Figure 2C). More recently we observed that a  (2007), before spreading to the tropical rainforests of northeastern Peru (Iquitos and Yurimaguas; 2008) [23]. As multiple serotypes have been circulating in the region severe disease resulting from heterologous secondary infection is increasingly likely to occur [27,29]. In this study we did not distinguish between primary and secondary infection, and thus further analysis will be needed to identify the genotypes [30] and prior DENV immune status associated with more severe disease outcomes in the region. Regardless, the data described here will provide a springboard for future studies of regional DENV maintenance and dispersion patterns [31] as well as analysis of genetic adaptation and selective pressures. Other than DENV, the only other flavivirus isolated was YFV. One well-documented hindrance to study flavivirus is the crossreaction observed among even disparate species [32,33]. Similarly here we observed significant cross-reaction between DENV and YFV antigen in serum from patients with defined DENV infection, thus there is a possibility that some of the cases have been misclassified. For YFV, we only considered those instances where there was no DENV IgM detected. Furthermore, there was a strong correlation between participants reporting recent YF vaccination and having YFV-reactive IgM, suggesting that these results were not due to cross-reactivity with other flaviviruses circulating in South America. Low grade fever and headache are not uncommon outcomes within the two weeks following YF vaccination [34,35], so it is possible that these cases are due to the vaccination. It should be noted, however, that flavivirus IgM can be long-lived [32], and thus many of these febrile episodes classified as ''YFV infection'' may not represent the true etiologic agent. In addition, in our study we only rarely observed severe disease associated with YFV-reactive IgM, suggesting that these cases largely do not reflect natural infection and thus should be interpreted with caution.
In addition to DENV and YFV, there are other flaviviruses circulating in the region that need to be considered, including WNV, Rocio virus (ROCV), Ilheus virus (ILHV), and St. Louis encephalitis virus (SLEV). These flaviviruses have been isolated either from mosquitoes [36], birds [37], or mammals [38], including humans [39,40], in parts of South America. More closely related to our study areas, ILHV has been isolated from a febrile patient in Ecuador [41], and ILHV and SLEV have been isolated from mosquitoes in Iquitos [36], clearly demonstrating that these viruses are circulating near human populations in the region. None of these viruses were isolated from participants in our study, suggesting that human infection is uncommon. However, in a preliminary survey of a subset of our participants we have identified cases where ROCV, ILHV, SLEV, or WNV IgM was detected, with no reactivity with DENV or YFV antigen (data not shown), with confirmation by virus neutralization assay, considered the most specific tool for flavivirus serology [42]. Overall, the cross-reactivity reported here and elsewhere [32] and the longevity of flavivirus IgM underscore the complications of flavivirus serodiagnosis, which represents a great hindrance for epidemiological surveillance.
The most common Alphavirus species identified were VEEV and MAYV. Scant evidence for human infection with EEEV was identified, consistent with previous reports [43], despite evidence of EEEV circulation in mosquitoes near Iquitos [36,44], for example. In light of the recent emergence of another alphavirus, CHIKV, in the Indian Ocean region, VEEV and MAYV represent interesting cases to consider with regards to potential for urban emergence. In laboratory studies Aedes spp., the primary vectors for DENV, have been shown to be a competent vector for VEEV [45] and MAYV [46]. Even without adapting to human-Aedes-human cycles, epizootic VEEV subtypes have been associated with large outbreaks of human disease across South America [6]. As recently as 1995 a VEEV outbreak was responsible for nearly 100,000 febrile cases in Venezuela and Colombia [47,48]. While the VEEV strains isolated in our study all belong to enzootic genotypes of the virus complex [12,14,24], genetic studies have demonstrated that enzootic and epizootic subtypes are closely related. A modest number of amino acid changes can alter the viral phenotype dramatically, converting an enzootic strain to an epizootic strain [49][50][51][52]. Similarly, amino acid variants in the CHIKV E1 protein have been associated with increased epidemic potential [53][54][55]. Several other factors further suggest that potential for neotropical alphavirus emergence is high. In the Iquitos area, while we found that VEEV was more commonly associated with rural clinics ( Table 7), many of the participants with confirmed VEEV infection lived within the city and did not report leaving the urban area during the month prior to the febrile illness [24](data not shown). This data is corroborated by a previous study of healthy participants, in which we found that nearly 25% of the urban population carries VEEV-neutralizing antibodies [24]. In addition, based on data collected through this program the geographic range of MAYV and VEEV is wider than had been previously demonstrated, extending to southern Peru and Bolivia [14,19]. Taken together these factors suggest that the potential establishment of the neotropical alphaviruses as urban pathogens should be closely monitored.
In addition to the flaviviruses and alphaviruses, orthobunyaviruses were significant sources of febrile illness in the study, accounting for 2.5% of febrile episodes analyzed. While all orthobunyavirus isolates came from patients in Peruvian rainforest sites during the course of this study, we did find serological evidence for OROV and Group C viruses in Ecuador and Bolivia. More recently (2008) we have definitively identified Group C viruses in Bolivia, isolated from two participants in the Cochabamba region (data not shown). Like VEEV and MAYV, OROV is an interesting case study with regards to potential for broader emergence. OROV has been associated with numerous outbreaks in Brazil, infecting approximately 500,000 people in South America over the past 45 years [56][57][58]. Two distinct transmission cycles have been proposed, a poorly-defined sylvatic cycle and an urban cycle where OROV is transmitted among humans by the biting midge Culicoides paraensis [58,59]. In Iquitos, we found that unlike the Group C viruses, VEEV, and MAYV, evidence of recent OROV infection (based on both IgM and virus isolation data) showed no significant bias towards rural clinics, suggesting similar transmission levels between urban and rural sites, consistent with results from an earlier survey of healthy participants in the region [10]. This pattern may reflect a periurban transmission cycle, as the majority of the OROV isolates were detected in both the rural sites and an urban site located towards the periphery of the city in 2005 and 2006 during a period of markedly increased transmission (Table 6). OROV isolates from previous Iquitos studies (prior to 1998) were found to belong to lineage II, similar to strains associated with Brazilian OROV outbreaks [56,60]. Future sequence analysis of the more recent isolates described in this current study from Iquitos, Yurimaguas, and Puerto Maldonado, will provide a more complete description of OROV geographic and temporal genetic variability.
Considering the association of arboviral pathogens with human disease and the potential for wider-scale emergence, disease surveillance is an integral component of public health planning, disease control, and analysis of potential intervention strategies. Unfortunately, for the arboviruses described here syndromic surveillance is complicated by the lack of pathogen-specific signs and symptoms [5], particularly early in disease progression. As with other reports [7,9], our study underscores the need for laboratory-based support of febrile surveillance studies. Even within our study other pathogens clearly need to be considered, as the majority of febrile episodes in this study were not associated with an arboviral etiology. In Iquitos past studies have linked both Leptospira spp. and Rickettsia spp. with a significant percentage of febrile illnesses [61][62][63]. To-date, solid data are lacking for the other study sites included in this study, although our preliminary results suggest that Leptospira spp. and Rickettsia spp. are common human pathogens in these locations as well (TJK, unpublished results). Admittedly our studies provide little in the way of guidance for patient care but do point toward the need for the development of pharmaceutical therapies for the treatment of a variety of viral infections. In addition the development of rapid and inexpensive diagnostic tools should be given high research priority, particularly to distinguish arbovirus infection from other pathogens where effective and inexpensive pharmaceutical treatment is already available, such as for Rickettsia spp. and Leptospira spp.