Challenges for malaria elimination in Brazil

One of the first reports of asymptomatic malaria infections in Brazil came from surveys by Avery-Jones and Ferreira Neto [60] in the costal belt of Santa Catarina State (Southern Region), where P. vivax was the dominant species. In this typical “bromeliad-malaria” setting covered by Atlantic rainforest (see below), the only known malaria vectors were Anopheles (Kerteszia) cruzi and Anopheles (K.) bellator, which breed in water trapped in bromeliad plants [61]. Using conventional microscopy, the investigators found 39 individuals to be parasitaemic at one or more monthly house-to-house surveys, but only one of them had a malaria-related illness at the time of diagnosis. Asymptomatic parasite carriage persisted for up to 6 months. Moreover, nine symptomless relapses were diagnosed following CQ treatment of 13 symptomatic infections [60]. Similarly, low-density asymptomatic P. vivax infections, some of them initially missed by conventional microscopy but indirectly diagnosed by serology [62, 63], have also been reported in other bromeliad-malaria settings in southeast Brazil [16]. Subclinical infections were later described in Amazonians living in traditional riverine communities and exposed to malaria since birth [64‐66]. These symptomless infections were found to be four to five times more prevalent than the symptomatic ones in riverine communities in Rondônia, with the risk for developing symptomatic malaria being inversely correlated with the subjects’ age [64].

In contrast, all malaria infections, even those with very low parasite densities, were thought to elicit clinical disease in malaria-naïve migrants settled on the Amazon frontier [67‐70]. Molecular diagnosis, however, has often revealed entirely symptomless, low-density infections in migrant subjects living in typical frontier settlements in the Amazon [71‐73]. For example, 56.6 % of all P. vivax infections diagnosed over 3 years in newly occupied farming settlements in Amazonas State were found to be asymptomatic; one-third of them were both submicroscopic (i.e., missed by microscopy) and asymptomatic [70]. Apparently healthy subjects accounted for half of the total P. vivax biomass found in the local population and nearly all asymptomatic P. vivax carriers had mature gametocytes detected with molecular techniques. Only 17.0 % of the asymptomatic P. vivax carriers developed clinical symptoms over 6 weeks of follow-up, becoming detectable by routine surveillance of febrile illnesses [73]. Similar findings were reported during the follow-up of asymptomatic parasite carriers in another Amazonian settlement [74]. After five to 8 years of continuous exposure to low/moderate levels of malaria transmission, the prevalence or incidence of both infection and malaria-related disease decreased steadily among native Amazonians and migrants, suggesting the acquisition of not only anti-disease immunity but also some degree of anti-parasite immunity [65, 73].

The infectiousness of gametocytaemia in asymptomatic carriers remains largely unknown. Avery-Jones and Ferreira-Neto [60] suggested that even a small but constantly present number of symptomless individuals could result in a large number of infective man-days. Specifically, if four infective asymptomatic gametocyte carriers were present during all 365 days of the year, a total of 1460 infective man-days would be observed. Similarly, if 10 asymptomatic gametocyte carriers were present during all 365 days of the year, a total of 3650 infective man-days would be observed. However, assuming that individuals would remain infective for 6 days, and considering all 170 laboratory-confirmed symptomatic malaria cases that had been diagnosed over a period of 12 months in their study site in Santa Catarina, only 1020 infective man-days would be observed. Nevertheless, the infectiousness of asymptomatic carriers of low-grade parasitaemia to Amazonian malaria vectors remains understudied; one study described an infection rate of 1.2 % after feeding An. darlingi with blood from asymptomatic P. vivax carriers, and of 22 % for the mosquitoes fed with blood from symptomatic carriers [75].

With regards to control activities, the challenge is how to detect asymptomatic infections in order to minimize the number of new infections originating from them. Mass blood surveys can detect asymptomatic malaria, but as transmission declines large populations would have to be screened to diagnose relatively few asymptomatic carriers, reducing the cost-effectiveness of this approach [60]. An alternative is to conduct reactive active case detection so that a confirmed malaria infection triggers the screening of individuals in a defined neighbourhood [76]. Regardless of the detection strategy, the diagnostic techniques available for large-scale use, such as microscopy and RDT, may not be sensitive enough to detect low-grade infections that are typical of residual malaria settings [77].

A reactive active case detection strategy with sensitive molecular diagnostic methods to detect new P. vivax infections in the neighbourhood of malaria cases diagnosed by routine surveillance (index cases) is currently being tested in Brazil. This strategy has been tailored for P. vivax infections, which tend to be maintained at low parasite densities and may relapse despite radical treatment. Four case-detection rounds (0, 30, 60 and 180 days after the index case diagnosis) were carried out within a 3 km radius around the index case to capture individuals during their primary infections and relapses. Molecular barcoding using single-nucleotide polymorphisms (SNPs) [78] or microsatellites [79] will facilitate the analysis of genetic relatedness between patient-derived P. vivax samples to elucidate malaria transmission networks.

A small proportion (about 0.05 %) of all autochthonous malaria cases currently recorded in Brazil originate from areas along the Serra do Mar mountain range that extends across the states of Rio de Janeiro, Minas Gerais, Espírito Santo, São Paulo, Paraná, and Santa Catarina, in the Southeast and South regions, and is covered with by the Atlantic rainforest biome (Fig. 4) [16]. In these areas, the main (or only) malaria vectors are An. cruzi and An. bellator, which breed in water trapped by the leaf axils of bromeliad plants that are particularly abundant in the Atlantic rainforest [61]. Initially called “forest malaria” [80], and later described as “bromeliad malaria” [81, 82], malaria transmitted by Kerteszia anophelines was historically endemic in coastal areas of Southeast and South Brazil [83]. Bromeliad malaria is currently uncommon in most of the Atlantic forest biome of Brazil, but eliminating residual transmission in remaining foci, which involve non-human primates as parasite reservoirs, has been a major challenge for several decades.

Only P. vivax and P. malariae infections are currently reported in humans in bromeliad malaria settings in Brazil, where monkeys of the genus Alouatta (howler monkeys) and Cebus (capuchin monkeys) are naturally infected with these malaria parasites and may represent major animal reservoirs of parasites [84]. However, P. falciparum was nearly as prevalent as P. vivax in bromeliad malaria areas of Santa Catarina State in the late 1940s [60]. Furthermore, P. falciparum infection was recently reported in An. cruzi vectors from São Paulo State [85], suggesting that Kerteszia anophelines may also transmit P. falciparum.

Bromeliad malaria constituted a major public health challenge in Santa Catarina, with over 26,000 laboratory-confirmed infections diagnosed in 1947 in a population of 670,000 [60]. Local transmission of P. falciparum and P. malariae ceased in 1956 and 1961, respectively, as a result of control activities that included tree removal near major towns and replacement with eucalyptus trees (in which bromeliads do not grow) and vegetable/flowering plants, manual removal of bromeliad plants, intensive DDT spraying, and prompt diagnosis and CQ treatment of laboratory-confirmed infections [86, 87]. Nevertheless, P. vivax transmission persisted in the northern coast of Santa Catarina until the mid-1980s, where subpatent and asymptomatic human infections remained relatively common.

Prior to the molecular diagnosis era, serology was used to identify subjects with evidence of current or recent infections in bromeliad malaria foci along the coast of northern Santa Catarina [18]. In 1980–1982, over 20,000 subjects living in endemic sites in the towns of São Francisco do Sul and Araquari had finger-prick blood samples spotted onto filter papers and examined for malarial antibodies with the indirect fluorescent antibody test (IFAT), using P. vivax-infected red blood cells as capture antigens. The nearly 500 subjects with antibody titers ≥64, regardless of clinical symptoms, received a full course of CQ and PQ. No further autochthonous malaria cases have been reported in these two towns since 1986 [86]. Similarly, over 11,000 samples were tested for antibodies by IFAT in three consecutive population-based surveys carried out between 1984 and 1984 in rural sites surrounding the town of Peruíbe, southern São Paulo. Overall, 146 (1.3 %) samples tested positive (antibody titers ≥ 16). The 70 (0.6 %) subjects with antibody titers ≥64 were treated with CQ and PQ; 40 infections were further confirmed by thick-smear microscopy [63].

Interestingly, serology has the potential to detect not only ongoing blood-stage infections missed by conventional microscopy [62], but also recent infections, either symptomatic or not, that might have led to P. vivax hypnozoite formation in the liver. Despite this potential, the use of serology for targeting asymptomatic carriers of hypnozoites and blood-stage infections in areas approaching P. vivax malaria elimination has been rarely reported.

Average parasite densities tend to decrease, with higher proportions of P. vivax infections being missed by microscopy but detected by PCR as malaria prevalence decreases in Amazonian communities [73]. Accordingly, between half and three-fourths of all PCR-diagnosed infections are typically missed by microscopy in hypo- and mesoendemic settings in Brazil [64, 65, 72, 73]. One-fourth of all P. vivax infections diagnosed in one of these settings had densities <10 parasites per microlitre of blood [73], being likely to be missed even by experienced microscopists. Similarly, the proportion of P. falciparum infections that are submicroscopic but still detectable by nucleic acid amplification (NAA) increases in Africa, from 20 % in areas with the highest transmission to 70–80 % in areas with the lowest transmission [88].

From a clinical perspective, these chronic, low-density infections may not be entirely harmless. For example, subpatent parasitaemia has been associated with increased risk of anemia in riverine Amazonian populations [65]. From a public health perspective, the most important question is to what extent submicroscopic infections contribute to residual malaria transmission, arguing for the use of more sensitive diagnostic techniques in the context of malaria elimination [89]. In low-prevalence settings in Africa, carriers of a submicroscopic P. falciparum parasitaemia were estimated to be the source of 20–50 % of all mosquito infections [88]. However, no comparable estimates are currently available for P. vivax in the Americas. Accordingly, WHO acknowledges the need for “more research to understand better the contribution of submicroscopic infections in malaria transmission in low endemic settings and to identify which diagnostic strategies and NAA-based diagnostic techniques are most cost-effective in accelerating malaria elimination, compared to conventional malaria elimination methods” [89].

Serological tests that detect malaria-specific antibodies are currently unable to differentiate between recent and old infections. Therefore, they play no major role in routine case finding and management. However, in very low transmission settings serology helps to identify subjects who have been recently exposed to P. vivax and may still carry hypnozoites or even a low-grade (mostly asymptomatic) asexual blood-stage parasitaemia, that are missed by microscopy [62, 63]. The bromeliad-malaria elimination campaigns in southern Brazil, as discussed before, illustrate how serology can be useful for targeting human reservoirs of P. vivax in areas of residual transmission. Moreover, in communities where malaria has been eliminated, serology may help to identify resurgent malaria, especially when children and young adults, who were born after transmission was interrupted, are found to have specific antibodies.

In 1825, physician José Maria Bomtempo reported that quinine was used indiscriminately and at very high dosages in Rio de Janeiro [90]. Indeed, the first evidence of malaria parasite’s resistance to quinine came from Brazil in the early 1900s. Arthur Neiva, in 1907, diagnosed malaria infections in individuals under compulsory prophylaxis with quinine (500 mg every 2 or 3 days) who worked in highly malarious marshlands 60 km away from Rio de Janeiro, the country’s capital at that time [91]. He thus recommended daily 500 mg doses of quinine as the only effective chemoprophylaxis in that area [91]. Miguel Couto [92] reported the use of intravenous methylene blue to treat quinine-resistant malaria infections acquired in Rio de Janeiro as well as in the Amazon, but he did not describe the quinine dose initially used. The finding by Oswaldo Cruz of quinine-resistant malaria among workers in the Madeira-Mamoré railway construction site located in Rondônia state, western Amazon [93], was soon confirmed by Bernhard Nocht and Heinrich Werner, who documented quinine failures during the treatment of German workers returning from that region to Hamburg in 1910 [94]. Interestingly, despite these early reports, quinine remained widely used in Brazil until the mid-1990s, usually in association with tetracycline or doxycycline, to treat uncomplicated P. falciparum infections. Relatively little evidence of in vitro [95, 96] or in vivo resistance to quinine [97, 98] has been published since then.

Emerging resistance to CQ and a fixed-dose sulfadoxine–pyrimethamine (SP) combination may have greatly contributed to explosive P. falciparum malaria outbreaks across the Amazon in the 1980s. The first report of CQ failure to clear P. falciparum infections in Brazil, presented at a national medical conference in 1954, drew relatively little attention [99], but later publications documented CQ resistance in Brazil [100, 101] and Colombia [102] in the early 1960s. At least two CQ-resistant founder populations have been identified in this region, one circulating in the Amazon Basin of Brazil and Peru, and the second in Ecuador and Colombia [103]. CQ resistance became widespread across the Amazon Basin by the mid-1980s [104], when malaria cases were on the rise associated with colonization projects and environmental changes in the region (Fig. 3).

SP had been available in Brazil since the 1960s to treat CQ-resistant falciparum malaria [105]. SP-resistant P. falciparum strains emerged within a few years, being first documented in the Centre-West region in 1972 [106]. By the end of the 1980s, SP failed to cure >90 % of P. falciparum infections in the Amazon [99], and since 1990, SP is no longer recommended for malaria chemotherapy in Brazil [99]. Before the introduction of ACT in 2006 [107], the first-line regimens for uncomplicated falciparum malaria recommended by the Ministry of Health of Brazil were quinine plus doxycycline (formerly tetracycline) for 7 days or a single dose of mefloquine. Case reports of mefloquine (e.g., [108]) and quinine–tetracycline [e.g., 109] failure in falciparum malaria have been published since then. The few clinical trials that evaluated the therapeutic efficacy of these regimens reported cure rates ranging between 98.9 [110] and 100 % [111] for mefloquine, and between 77.3 [98] and 100 % [97] for quinine plus tetracycline.

ACT has gradually replaced quinine–doxycycline and mefloquine as the first-line regimen for uncomplicated falciparum malaria in Brazil in the late 2000s. Nearly 24,000 P. falciparum infections were treated with a fixed-dose artesunate–mefloquine (ASMQ) combination in Juruá Valley, Acre, between July 2006 and December 2008. Following this large-scale intervention, P. falciparum malaria incidence rates decreased substantially, with lower hospital admission rates, and a reduced proportion of P. falciparum to P. vivax infections in this region [112]. A recently completed trial showed no evidence, either clinical or molecular, for emerging P. falciparum resistance to the fixed-dose ASMQ combination in Juruá Valley [113]. More data on ACT efficacy in other areas of Brazil are expected to be made publicly available by the Amazon Network for the Surveillance of Antimalarial Drug Resistance (RAVREDA), a network organized in 2001 by the Amazonian countries of South America, with PAHO support, to respond to the challenge of anti-malarial drug resistance in the region. RAVREDA operates in close association with the Amazonian Malaria Initiative (AMI), a partnership of the United States Agency for International Development with PAHO, the Centers for Disease Control and Prevention (CDC), the Management Sciences for Health’s Rational Pharmaceutical Management Plus (MSH/RPM Plus) programme, the United States Pharmacopoeia, and Links Media, Inc.

Resistance to artemisinin derivatives has not been documented in Brazil, but there is some concern that resistant parasites may circulate in areas of Suriname, Guyana and French Guyana bordering Brazil. Suriname has recently attracted a large number of immigrants, mainly from Brazil, to work in gold mines in the eastern and central parts of the country. An estimated 15,000 (mostly illegal) miners currently live in settlements where health infrastructure and compliance with national malaria treatment policies are typically very poor [114]. A recent study in Suriname has documented a large proportion (31 %) of P. falciparum-infected patients treated with artemether–lumefantrine who remained parasitaemic by day 3, consistent with some degree of artemether resistance in Suriname [115], although these findings have not been confirmed in other areas of the country. Nevertheless, Suriname is complying with WHO recommendations for malaria control and P. falciparum malaria is now rarely diagnosed in the country [116]. In Guyana, the kelch propeller domain (k13) gene mutation C580Y, associated with artemisinin resistance, has been found in 5 % of 98 P. falciparum isolates analysed [117]. Interestingly, an analysis of polymorphisms upstream and downstream the k13 gene suggest that the C580Y mutation has emerged independently in Guyana and Southeast Asia [117]; these mutant parasites, however, did not have their drug sensitivity patterns analyzed so far. In French Guyana, up to 15,000 illegal gold prospectors live in high-endemic areas along with 30,000 legal residents. They typically use erratic ACT-based regimens, potentially including fake or substandard drugs, to treat P. falciparum infections, favouring the selection of drug-resistant strains. Artemisinin resistance had not been characterized so far in French Guyana, but the country fails to comply with WHO recommendations for anti-malarial treatment [116]. Therefore, unequal cross-border efforts to control malaria and contain the emergence of anti-malarial drug resistance pose potential threats to eliminate the disease from French Guyana, Brazil and Suriname [116].

The global spread of CQ-resistant P. vivax strains [118] can further complicate malaria elimination efforts in Brazil. Radical cure of P. vivax malaria is achieved with 25 mg/kg of CQ base over 3 days (maximum adult dose, 1.5 g over 3 days), combined with a short hypnozoitocidal regimen of 0.5 mg/kg/day of PQ base (maximum daily dose, 30 mg/day) over 7 days in patients that weight under 70 kg. Since subtherapeutic PQ doses may lead to relapses in overweight patients [119], weight-adjusted PQ doses are now recommended in Brazil for patients over 70 kg [120].

CQ-resistant vivax malaria has been reported in Colombia and Brazil since in the late 1980s, but anti-malarial treatment had not been supervised, and drug levels were not measured, in these early studies [reviewed by 121]. The first solid evidence of CQ-resistant P. vivax infections in South America was published in 1996. Two Canadian patients returning from Guyana had received CQ and PQ, with therapeutic CQ levels documented at the time of parasite recurrence [122]. Further evidence of P. vivax resistance, confirmed by CQ level measurements, originated from patients from Peru given CQ alone in 2003 [123]. Two studies described P. vivax resistance to CQ in Manaus, the major port city in the Amazon Basin of Brazil, in the 2000s. The first study documented P. vivax recrudescence, despite adequate plasma levels of CQ, in 11 of 109 (10.1 %) patients treated with CQ alone who were followed up for 28 days [124]. The second study reported P. vivax recurrences in seven of 135 (5.2 %) patients treated with CQ and PQ who were followed up for 28 days; all of them had CQ levels above 100 ng/mL of whole blood at the time of recurrence [125]. It remains to be determined whether CQ resistance has spread to other areas in Brazil; a very recent clinical trial has detected little P. vivax resistance to CQ alone (2 %), and none to CQ–PQ combination, in major malaria hotspots in Juruá Valley (MUF, unpublished observations). Although ACT has been used to treat CQ-resistant P. vivax infections worldwide [126], its efficacy as an alternative to CQ remains unknown in Brazil.

The finding that PQ reverses CQ resistance in P. falciparum suggests that a similar effect might occur in P. vivax isolates simultaneously exposed to both drugs [reviewed by 127]. In fact, recent meta-analyses showed a lower efficacy of CQ alone against P. vivax asexual blood stages compared to CQ co-administered with PQ [118, 128]. Over the past 60 years, CQ and PQ have been routinely combined for the radical cure of P. vivax infections in most of South America, but not in Melanesia and Southeast Asia, where highly prevalent severe G6PD deficiency may lead to PQ-induced haemolysis. Indeed, CQ-resistant P. vivax emerged in regions where PQ is not widely used and remains less common in countries where CQ and PQ are routinely coadministered [118, 121], indicating that CQ–PQ combination therapy may have delayed the emergence of CQ-resistant P. vivax strains in selected settings.

PQ safety is an open question, since there is no practical, field-deployable rapid diagnosis test for G6PD deficiency screening prior to PQ administration [129]. Severe G6PD deficiency is uncommon across the Amazon Basin [130, 131], but severe haemolysis has been occasionally reported following PQ treatment [132]. Moreover, PQ cannot be administered to pregnant and breastfeeding women and to children less than 6 months of age, because the risk of haemolysis. PQ efficacy also remains largely undetermined in Brazil. Relapse rates between 14.0 and 24.5 % have been described following supervised CQ–PQ treatment of imported P. vivax infections in non-endemic sites [119, 133]. Furthermore, an imported P. vivax infection acquired in Brazil was found to relapse despite the administration of 900 mg of PQ over 30 days [134]. The extent to which PQ metabolism in Amazonian populations is affected by polymorphisms in cytochrome P450 (CYP) family members, such as CYP 2D6, remains unknown. In fact, reported failures of PQ may be partially associated with poor PQ metabolism in patients, rather than true drug resistance. However, the local monkey-adapted P. vivax strain Brazil I is PQ-resistant in monkeys [135]. If PQ is failing in Brazil, for whatever reason, a substantial proportion of P. vivax infections may actually be relapses, placing improved anti-relapse treatment as a top research priority for guiding malaria elimination in this country.

The finding that high-dose PQ regimens are safe and effective among G6PD-normal subjects led the US Centers for Disease Control and Prevention (CDC) to recommend 420 mg of PQ over 14 days as the standard anti-relapse regimen for adults in areas where standard PQ treatment fails [136]. WHO recommends a PQ dose of 0.5 mg/kg of body weight over 14 days to prevent relapses of infections acquired in Southeast Asia and Oceania [137], but this high-dose PQ regimen has not been evaluated in Brazil. Tafenoquine (TQ) is an alternative to PQ that requires shorter treatment regimens with improved compliance; a single dose of 300 mg TQ coadministered with CQ is effective to prevent P. vivax relapses in Brazil, Peru, India, and Thailand [138]. However, treatment with this long-lasting hypnozoitocidal drug does require prior screening for G6PD deficiency, an additional challenge due to the lack of a rapid screening test. Moreover, TQ is neither licensed nor available commercially in Brazil.

Malaria infection during pregnancy is associated with substantial risks for the mother, her fetus, and the neonate. Stillbirth, intrauterine growth retardation, prematurity, low birth weight, spontaneous abortion, increased neonatal and maternal mortality, and reduced neurocognitive function later in childhood are documented complications of malaria in pregnancy (MiP) [139, 140].

Between 6000 and 9000 microscopy-confirmed MiP cases are recorded in Brazil each year [141]. This represents 4–6 % of all malaria cases in the country. However, since conventional microscopy and RDTs fail to detect a substantial proportion of MiP episodes that are later diagnosed by PCR [142], these figures are likely to be underestimated. Health professionals involved in antenatal care, such as nurses and obstetricians, usually do not perceive MiP as a major preventable and treatable cause of morbidity in pregnant women and their offspring. This is partially due to the fact that infections with P. vivax are associated with less severe clinical consequences in pregnant women than those with P. falciparum, although an increased risk of low birth weight and anemia was documented in a large cohort study in Southeast Asia [143]. Erythrocytes parasitized with the former species do not sequester massively in placental capillaries [144, 145], but some placental changes associated with MiP, such as syncytial knotting and increased thickness of the placental barrier, have been recently documented in Brazil and may affect fetal growth [145, 146]. Because PQ cannot be administered during pregnancy, relapses are common. Overall compliance with a weekly CQ regimen for suppressing relapses following P. vivax infection during pregnancy is usually poor in endemic areas of Brazil.

Almost one-third of the MiP cases recorded in Brazil between 2003 and 2012 were due to P. falciparum, although this species accounted for only one-fourth of the malaria burden in non-pregnant women during the same period [141]. Indeed, a cross-sectional survey of 1699 febrile women at childbearing age found a significantly greater P. falciparum to P. vivax ratio in pregnant women (1:2.3) than in their non-pregnant counterparts (1:5.6) living in Manaus, Brazil [147]. Also, a cohort study in the Peruvian Amazon found pregnant women to be twice more likely to have P. falciparum infections than non-pregnant women living in the same community, but no pregnancy-associated difference was found in the risk of P. vivax infections [148].

Pregnant women living in malaria-endemic areas of Brazil must be tested for malaria parasites at every antenatal care visit and receive supervised anti-malarial treatment whenever MiP is confirmed [141]. In areas with stable transmission of P. falciparum in Africa, pregnant women are encouraged to sleep under insecticide treated bed nets (ITNs), and at least two doses of intermittent preventive treatment (IPTp) with SP should be administered during antenatal care visits to prevent MiP [149]. Although sub-Saharan Africa has high rates of antenatal care attendance, current IPTp coverage is still low [150], despite the fact that IPTp-SP is highly cost-effective for prevention of MiP and reduction of neonatal mortality, even in areas with some molecular evidence of SP resistance in P. falciparum populations [151]. Currently, there is no evidence to support the use of IPTp-SP to prevent MiP in areas with substantially lower levels of malaria transmission, such as those in Brazil.

Research conducted in the Amazon in 1931 showed that An. darlingi was found in much greater numbers indoors (endophilic) than outdoors (exophilic). Data collected between July 1942 and June 1946 in Rondônia (then called Guaporé territory) indicated that 93 % of the mosquitoes captured were present inside the houses. For the whole Amazon, this percentage was 88 % [152]. In 1977, another study showed that An. darlingi would enter houses that had been sprayed with DDT, but only to feed, and not to rest, as they would do in unsprayed houses [153]. Early records also show that An. darlingi was to some extent exophilic before the use of DDT [154]. Studies in Rondônia during the 1980s, however, found An. darlingi mainly outdoors, in the vicinity of houses [69, 153, 155, 156]. In addition to differences in biting location, An. darlingi also shows different patterns of peak biting time. For example, bimodal biting curves have been reported, with peaks in the early evening and at dawn, hours at which workers could be carrying out outdoor activities [156‐158].

The exophilic behaviour is a point of discussion. Some argue that DDT use could have contributed to this transformation, selecting the exophilic strains to be more prevalent. Others consider that extensive deforestation and disorganized occupation could be indirectly responsible for modifications in the mosquito behaviour, subtracting its former sources of food (wild animals, who were scared away by the new settlers) and bringing man closer to its breeding places [3]. Population and mutation differences have been documented in South America (where An. darlingi is the dominant vector), with a marked north–south divide [159]. Although chromosomal and isoenzymatic studies show a high degree of heterogeneity, no morphological characteristics for separation of sub-species have been found [156].

This diversity in An. darlingi behaviour has important implications for vector control. First, an exophilic behaviour and peak biting times limit the impact that using ITNs could have in malaria transmission. Despite the evidence on ITN effectiveness [160], no large scale trial of bed nets have been conducted in the Amazon. However, there is evidence that targeting ITNs to specific areas based on vector behaviour can be an important strategy to add to a package of interventions to reduce malaria transmission. Two studies conducted in specific areas in Venezuela and Colombia showed efficacy in ITNs to reduce malaria risk [161, 162]. In Brazil, LLINs impregnated with permethrin were distributed across Juruá Valley (see Fig. 4) starting in late-2006 [163]. Although malaria incidence fell by 32 % over the next months, the intervention did not follow a systematic design to allow a proper impact evaluation, controlling for other ongoing interventions. Also, two additional rounds of LLIN distribution (2010, 2014) led to little changes in malaria rates. Second, with regard to IRS, limiting factors include An. darlingi behaviour and the quality of houses, particularly in rural areas—these are often made with poor construction materials, with partial walls and no windows [56]. Although IRS was largely used in Brazil starting in the mid-1940s, during the first elimination campaign (Fig. 1), currently it is restricted to areas that match certain technical and operational requirements defined by the Ministry of Health [164]. Third, the use of larvicides is restricted to specific water habitats that can be easily identified and reached, such as fish ponds [165].

In summary, vector ecology and behaviour coupled with the local environmental conditions in the Amazon bring about major barriers for the adoption of large-scale vector control strategies. Indeed, in 2014, about 3 % of the population in the Amazon was covered by treated nets, and 1 % covered by IRS [166]. Yet, targeted interventions should be considered as important additions to a range of control strategies to reduce malaria transmission, as long as decisions are based on local knowledge of vector behaviour.

Approximately 20,000 P. falciparum malaria cases were recorded in the Amazon Basin of Brazil in 2014. They were mostly restricted to ten high-risk areas, which account for three-fourths of the cases recorded countrywide (Fig. 5). Four municipalities in Juruá Valley (in the Western portion of the Amazon, close to the border with Peru), with a combined population of nearly 120,000 inhabitants, reported 46 % of all P. falciparum infections diagnosed in Brazil in 2014. The highly focal nature of P. falciparum malaria indicates that eliminating transmission of this species may be technically feasible by targeting a limited number of transmission pockets. Moreover, the potential of falciparum malaria for high morbidity and mortality and multidrug resistance, as well as its comparatively greater susceptibility to existing control measures, provide compelling arguments for prioritizing this species in malaria elimination strategies in Brazil. Indeed, the recently launched Plan for Elimination of Malaria in Brazil devotes special attention to P. falciparum malaria, and has the ultimate goal of eliminating the disease by 2030 [167].

As argued for the Greater Mekong Subregion in Asia [168], the window of opportunity for P. falciparum elimination may be rather short. First, an elimination strategy must be implemented before widespread anti-malarial drug resistance (more specifically, resistance to artemisinin derivatives) makes P. falciparum infections in Brazil untreatable with currently available drugs. In other words, “multidrug resistance is both an impediment to elimination and a reason for pursuing it” [168]. Second, financing and political commitment to malaria elimination tends to wave once P. falciparum becomes widely perceived as a relatively minor public health problem and other competing priorities emerge countrywide.

Eliminating residual P. falciparum transmission will require a comprehensive assessment of the factors that favor malaria transmission in well-defined foci across the Amazon. Improved local knowledge regarding the relative contribution of vectors (e.g., species distribution, abundance, behaviour, and insecticide resistance), parasites (e.g., drug resistance and virulence), humans (e.g., behaviour, health infrastructure, and patterns of mobility), and the environment (e.g., land use and land change) to persisting P. falciparum transmission will be crucial, especially since underlying drivers of transmission are not likely to be same across foci. Indeed, human-made vector breeding sites (e.g., fish pounds opened for commercial aquaculture) are a major factor leading to the current increase in P. falciparum malaria incidence in urban and peri-urban areas of Juruá Valley (Acre State, Western Amazon). Mining activities favor transmission in the outskirts of Itaituba (Pará State, Eastern Amazon), while in Barcelos (northern Amazonas State, Western Amazon) the majority of malaria cases is recorded in remote protected areas (mostly indigenous reserves). Finally, the main endemic sites in Lábrea (southern Amazonas State) are newly opened farming settlements. Thus, different strategies will be needed for controlling and eventually eliminating P. falciparum in each area.

At present, there is no evidence that artemisinin-resistant P. falciparum strains have been imported into Brazil. No data are available on anti-malarial drug resistance patterns of P. falciparum infecting Brazilians returning from mining camps in the Guiana Shield (a region comprising Guyana, French Guiana, Suriname and parts of Colombia, Venezuela and the northernmost tip of Brazil), where artemisinin resistance is suspected to be emerging [115].

The natural environmental characteristics of the Amazon Basin offer suitable conditions for malaria transmission: temperature, humidity and the local vegetation guarantee a large population of vectors year-round, with less seasonal variation when compared to other areas [26]. Also, the water level of rivers increases dramatically during the rainy season, flooding areas immediately proximal to the margins, and as the water level decreases with the ending of the rainy season, pools of water suitable for mosquito breeding proliferate [169].

The massive deforestation observed in the Brazilian Amazon in the 1970s and 1980s brought about significant environmental changes and was associated with increases in malaria transmission, as previously discussed. In addition, large-scale development projects (e.g., road and dam construction) often produce environmental disturbances and social conditions that can be conducive to increases in vector-borne diseases, including malaria [170]. While expanding areas protected under environmental conservation policies would be beneficial to prevent an increase in malaria transmission [171], current trends make it reasonable to expect that deforestation in the Amazon will continue to be a reason for concern [172] driven by varied economic pressures (e.g., cattle ranching, agriculture, mining, and large-scale development projects). To mitigate some of those problems, since 2001 the Brazilian legislation requires that, as part of the licensing of development projects located in malaria endemic areas, an evaluation should be conducted by the Ministry of Health [173].

Considered as the biggest global health threat of the twenty first century [174‐176], climate change is likely to have an impact on malaria, despite contrasting results of global models [177‐182]. Climatic effects on malaria can occur directly, through extreme events (e.g., drought, flooding), increases in average temperature, and changes in precipitation patterns, but also indirectly, by population displacement, and water and food insecurity, which impact individuals’ exposure and vulnerability to infections. Despite the uncertainty embedded in climate change scenarios [183], it is expected that impacts on malaria will be observed, although the impacts and the responses are unlikely to be uniform across and within regions [184]. The magnitude of the impact will depend on: (i) temperature variability [185]; (ii) countries’ capacity to anticipate future changes; (iii) adaptation strategies adopted at varied scales; (iv) ongoing malaria control interventions; (v) the possibility that local characteristics that affect malaria risk can be augmented by climatic changes; and (vi) the level of clinical immunity among the population [178]. Since 1996, more than a dozen atypical years (El Niño/La Niña), and at least one extreme drought not at all related to El Niño/La Niña, were observed in the Amazon. Yet, their potential effect on the patterns and level of malaria transmission remains largely unknown.

Surveillance can be defined in different ways [186]. Considering it as an intervention through which data are systematically collected, analysed, and interpreted, so that both symptomatic and asymptomatic cases are identified, triggering quick action to prevent further transmission [187], then the main challenge is how to properly identify all the cases. Among the potential problems in achieving proper surveillance are human mobility, border issues, asymptomatic infections, and weak health and information systems.

With respect to information systems, Brazil is in a privileged position. Since malaria is one of the diseases that require compulsory notification in the country, a national system gathers information on all malaria tests performed across the country (microscopy and RDT), with demographic information on patients, and with detailed test results. Data are entered locally, and available real-time at the municipal, state and federal levels.

Migratory movements have important malaria implications depending on the levels of immunity among migrants, the underlying patterns of transmission in sending and receiving areas, the local characteristics observed in areas of origin and destination, the individual knowledge about the disease, and the pattern of mobility [188, 189]. Human mobility is intense in the Brazilian Amazon driven by new economic opportunities (e.g., mining, land availability) or by the need to obtain goods and/or services [190]. In a context of intense human mobility, the challenge of asymptomatic infections is likely to be augmented, since symptomless migrants represent a mobile and silent reservoir of malaria infections. Moreover, human movements can facilitate the reintroduction of malaria in areas that have significantly reduced (or even eliminated) malaria. Critically to that is when unequal control efforts are adopted by neighbouring countries, potentially resulting in a high prevalence of imported cases near the borders [191, 192]. In that regard, one of the critical challenges in the Amazon region is the border between French Guiana, Suriname, and Brazil, where intense mobility (and importation of malaria cases) is fueled by gold mining in French Guiana and Suriname [116, 193].