Assessing malaria transmission in a low endemicity area of north-western Peru

Despite recent reduction in its estimated incidence, malaria remains the most important human arthropod-borne disease, worldwide and in the Americas [1]. In Peru, malaria incidence has fluctuated dramatically over the last 50 years. Between 1954 and 1967, malaria was well under control by coordinated eradication efforts (resulting in only 1,500 cases in 1965) [2], but returned to high levels when eradication campaigns ceased. Following the occurrence of drug-resistant Plasmodium falciparum strains to chloroquine (CQ) and sulphadoxine-pyrimethamine (SP) [3, 4] and the spread of Anopheles darlingi in the Amazon Region [5], malaria re-emerged in the 1990s, reaching a peak of more than 200,000 cases in 1998 after the El Niño Southern Oscillation (ENSO) climatologic phenomenon [6]. Between 2000 and 2005, the annual incidence fluctuated between 70,000 and 80,000 cases followed by a steady decrease until 2011, when 22,877 cases were reported [7]. This achievement can be attributed to the strengthening of the National Malaria Control Programme (NMCP) and to the implementation of comprehensive interventions such as the use of artemisinin-based combination therapy (ACT), the distribution of long-lasting insecticidal mosquito nets (LLINs), and health education campaigns with the strong support of international donors, e g, US Agency for International Development (USAID), and the Global Fund for Aids, Tuberculosis and Malaria (GFATM) [8‐10].

In Peru, the significant decline in malaria incidence over the past decade has modified its epidemiological profile and consequently calls for adapted control strategies. Currently, there is the need for targeting interventions and surveillance to foci of residual transmission in the north-west coast as well as in the Peruvian Amazon region. A key element for this re-orientation will be the availability of accurate measurement of malaria transmission intensity (MTI) and its evolution in space and time [11, 12]. However, it is not clear how best to monitor changes in transmission and disease burden in low endemicity areas [13].

Traditional methods to measure MTI include the collection of entomological and parasitological parameters. However, in areas of low endemicity, these measures are often subject to fluctuations for reasons other than true changes in transmission. For example, the improvement of case detection methods usually leads to an artificial increase in malaria incidence, which makes difficult the comparison between pre- and post-intervention data. Furthermore, entomological (entomological inoculation rate (EIR)) and parasitological measures (infections detected by microscopy) estimated through community surveys require very large sample sizes and are therefore extremely time and money consuming, while not able to catch seasonal variations [14, 15]. Moreover, parasitological surveys using microscopy cannot detect subpatent infections which are commonly reported in areas of low transmission [16, 17]. Molecular tests such as polymerase chain reaction (PCR) are much more sensitive for the detection of low parasite-density infections, and will become increasingly important for malaria elimination programmes [18, 19].

Serological markers have recently been used in several endemic areas across the world to estimate MTI [20‐24] and monitor its changes over time following interventions [25]. Serological techniques are particularly suited to low endemic areas as antibodies remain in the blood longer than malaria parasites and are thus easier to detect and less subject to seasonal variations. This paper reports on population-based estimates of P. falciparum and Plasmodium vivax prevalence and seroprevalence in an area of low endemicity in the north-west coast of Peru. The combination of molecular and serological tools is aimed at improving the detection of current malaria infections as well as recent exposure, in order to get insights into the residual transmission dynamics.

A total of 996 households and 4,650 individuals were identified during the census, and were distributed over the three study settlements as follows: PAV (433 houses; 2,163 individuals), JCM (312; 1,358) and NP (251; 1,129). The overall sex ratio was 0.94, and the age distribution showed that half of the population was <20 years old. A total 2,733 individuals (58.8%) living in 871 households (87.4%) were available at the time of the survey and 97.6% (2,667/2,733) of them accepted to participate. Females (61%) slightly outnumbered males, but the age distribution was very similar to the one in the total population. Among adults aged ≥18 years old, 60% had completed secondary education level. Only about 20% of the respondents had a regular income, almost all below the minimum national vital wage. The most commonly reported occupations were housewife (27.1%), labourer (9.0%) and trader (6.4%), while preprofessionals (children, students) and retired represented half of the participants. Very few respondents reported having been working as farmer (2.4%) or moto-taxi driver (2.6%) in the past year. More than half (52.6%) of the participants stated they were regularly sleeping under an untreated net, and 11.5% had experienced at least one confirmed P. vivax episode during the previous year (Table 1).

Parasite prevalence determined by microscopy was 0.3% (total eight P. vivax infections) while by PCR it was estimated at 1% including one P. falciparum and 25 P. vivax infections (Table 2). Malaria seroprevalence for P. vivax and P. falciparum was estimated at 13.6 and 9.8%, respectively, with 9.0% of the total seropositive individuals having antibodies against both antigens. A very weak correlation between antibody responses (percent positivity = PP) to the two species were found in individuals older than ten years (r = 0.1, p < 0.001), but not in younger children (p > 0.1). Most of the P. vivax PCR-positive cases were submicroscopic (17/25 = 68.0%) and asymptomatic (22/25 = 88.0%) at the time of the survey (Table 3). PCR positivity to P. vivax was significantly associated with the presence of P. vivax parasites (by microscopy) (p < 0.001) or antibodies (p < 0.001), as well as with symptoms (p < 0.001). Association between PCR and sero-positivity to P. falciparum could not be explored since only one P. falciparum infection was detected by PCR. This infection was asymptomatic and submicroscopic, and was found in a 16-year-old girl who also had antibodies against P. falciparum.

Given the small number of PCR-positive individuals, a multivariate adjusted risk factor analysis was not carried out, and the univariate analysis showed that distance to water drain was significantly associated with the risk of P. vivax infection (no association could be found with age, gender, income, education, bed net, or previous travelling outside the community). Individuals living at <100 m from the water drain had increased odds of P. vivax infection than those living farther (≥100 m) (OR = 3.4, 95% CI [1.3–9.0]).

Distance to water drain, age, education, income, occupation as farmer in the previous year, and history of P. vivax malaria in the previous year were significantly associated to P. vivax exposure in the univariate analysis (Table 4). The multivariate analysis (excluding history of malaria in the previous year) showed that only age and distance to water drain remained independently associated with P. vivax seropositivity. Individuals older than ten years had increasingly higher odds of P. vivax seropositivity compared to younger children (test for trend p < 0.001), while those living at closer distance to the water drain had increasingly higher odds of exposure to P. vivax compared to those living farther than 250 m. A significant interaction (OR = 5.7; p = 0.014) between age groups > ten years and distance <100 m was found, showing that the effect of age (>ten versus ≤ ten years) was significantly higher in those living closer (<100 m) to the drain (OR = 11.7; 95 CI [3.0; 45.1]; p < 0.001) compared to those living farther (OR = 2.0; 95 CI [1.5; 2.7], p < 0.001). Conversely, the effect of distance to water (<100 m versus ≥100 m) was significant only in the older age group (OR = 2.4, 95 CI [1.6; 3.7], p < 0.001).

Significant risk factors for P. falciparum exposure identified in the univariate analysis were the following: settlement, age, education, prior work as farmer or moto-taxi driver, and bed net use (Table 5). The multivariate-adjusted analysis showed that only settlement, age and previous work as moto-taxi driver remained independently associated with exposure to P. falciparum (no interaction was found). Individuals living in JCM (AOR: 1.5, 95% CI [1.02–2.1]) and in NP (AOR: 2.2, 95% CI [1.6–3.2]) were more likely to be exposed to P. falciparum compared to those living in PAV. In addition, individuals older than five years had increasingly higher odds (test for trend p < 0.001) of P. falciparum seropositivity compared to younger children, and those having worked as moto-taxi driver in the previous year were two times more exposed to P. falciparum (AOR: 2.1, 95% CI [1.1–3.8]) compared to those who did not.

Age seroprevalence plots for PvMSP1₁₉ and PfGLURP are shown in Figure 2 (A and B), and confirm previous results: increasing P. vivax and P. falciparum seroprevalence with age. Likelihood ratio (LR) tests supported models with two SCR (forces of infection) rather than one SCR for both P. vivax (p = 0.028) and P. falciparum (p = 0.046), indicating significant changes over time in the MTI for both species. For PvMSP1₁₉, best estimates indicated a significant four-fold increase in the SCR approximately four years prior to the survey (year 2006): post-2006 λ ₁  = 0.024 (95% CI [0.018–0.032]) and pre-2006 λ ₂  = 0.006 (95% CI [0.002–0.017]). For PfGLURP, best estimates indicated a 19–fold drop in the SCR approximately eight years prior to the survey (year 2002): post-2002 λ ₁  = 0.022 (95% CI [0.014–0.033]), and pre-2002 λ ₂  = 0.431 (95% CI [0.041–4.54]).

The retrospective analysis of 20-year annual malaria incidence and rainfall from 1990 to 2010 in Piura department showed that after the El Niño-Southern Oscillation (ENSO) phenomenon in 1997–1998 (Figure 3), when nearly 50,000 P. falciparum cases were recorded (1998), P. falciparum incidence dropped and stabilized around 4,000–5,000 cases annually in 2000–2001. This was followed by a steady decrease, and since 2006, autochthonous P. falciparum reported cases have become scarce (six and three reported cases in 2006 and 2008, respectively; and no cases in 2007, 2009 and 2010). Similarly to P. falciparum, P. vivax incidence decreased gradually after the ENSO (around 23,000 cases in 1998) reaching very low levels during the four-year period of intense droughts, with only 700, 315, 168, and 606 reported cases, respectively from 2004 to 2007. However, since then, P. vivax cases have increased with 4,185, 2,735 and 2,153 cases in 2008, 2009 and 2010, respectively. During the same period, annual rainfall in Piura dramatically dropped from 1998 (1,687 mm) to minimum values (between 14.3 and 59.4 mm) during the drought period, but subsequently increased since 2008 (193.5 mm in 2008).

The combination of molecular and serological tools, in addition to standard microscopy, allowed for an in-depth characterization of the current malaria transmission pattern as well as recent changes in species-specific MTI in this peri-urban area of the Peruvian northern coast. Most of the malaria infections were submicroscopic (70%), as well as asymptomatic (89%) despite the low level of transmission. Though P. vivax was predominant, species-specific seroprevalence showed that exposure to P. falciparum still occurs in the study area. Interestingly, over the past ten years, both species experienced significant but opposite changes in MTI: while P. falciparum MTI dropped after 2002, the one of P. vivax significantly increased after 2006.

Despite having been conducted at the end of rainy season, when the health information system usually records the highest malaria incidence [7], few malaria-infected individuals were identified and most of them had not had malaria-related symptoms within the two days prior to the survey. These results are in line with findings from other areas in the northern coast [39] and the Peruvian Amazon region [40, 41] and support the hypothesis that this asymptomatic and submicroscopic reservoir of infections contributes to malaria transmission [42].

In areas of low malaria transmission with seasonal patterns affected by climatic conditions, such as the Peruvian northern coast region, parasitological indicators determined by microscopy or PCR can be insensitive. In these areas, serological measures can be used as a proxy for malaria transmission intensity [20, 23, 25, 36]. Indeed, seroprevalence reflects the cumulative exposure to malaria and, since antibodies persist longer in the blood than parasites, it is more sensitive and less affected by seasonality or unstable transmission [22, 23, 43]. Though it may be limited in detecting discrete seasonal variations in transmission, it is a good indicator of long-term transmission potential [22, 24].

Even though P. falciparum reported cases in Piura department have been extremely rare in the past five years, P. vivax and P. falciparum seroprevalence indicated ongoing exposure to both species. The low PfGLURP seropositivity in young children and the only one P. falciparum case detected by PCR suggest that a residual P. falciparum transmission is still present in the study area. This residual transmission may be due to asymptomatic P. falciparum infections that remain undetected, unless a population-wide survey (such as the present study) is carried out. However, the implementation of mass screening as a public health strategy in low transmission settings remains a matter of debate [18, 40, 41], as the cost-effectiveness of such an intervention remains difficult to establish. The role of asymptomatic infections in maintaining malaria transmission as well as the impact of mass screening and treatment should be further evaluated through modelling techniques [44].

One might argue that an alternative explanation for the observed P. falciparum seropositivity could be the presence of antibodies cross-reacting with P. vivax or other infectious agents’ antigens. Limited information is available about the cross-reactivity between P. falciparum and P. vivax antigens [35, 45], and this has not been reported for PfGLURP. In addition, the lack of association between confirmed P. vivax infections by PCR and PfGLURP seropositivity, and the absence of correlation between antibody responses to the two species suggest that cross-reactivity is probably minimal. A few studies have suggested the cross-reactive potential of PfGLURP with Schistosoma haematobium[46], which is unlikely in the study area since the latter has not been reported in Piura.

Serologic markers can also be used to determine risk factors for malaria. This is particularly important in low endemicity areas as the one reported here, where the low malaria prevalence jeopardizes the risk factor analysis. In the study area, age was the only independent risk factor that was associated to both species. The age-dependent PvMSP119 and PfGLURP seropositivity indicates cumulative P. vivax and P. falciparum exposure with increasing age. As antibody responses will persist for prolonged periods, the rate of acquisition in young age groups can be used to determine current malaria transmission intensity [22‐25] as well as to identify recent changes in MTI.

The higher seroprevalence in individuals with several P. vivax malaria episodes also confirmed that PvMSP119 antibody detection measures cumulative exposure. However, the history of P. vivax malaria in the past year was not included in the multivariate risk factor analysis for P. vivax exposure, since it is not a risk factor for P. vivax infections per se but rather a consequence of previous exposure to a given risk factor. Plasmodium vivax infections are on the pathway between risk factors and the presence of P. vivax antibodies.

Differences in the risk factors for P. vivax and P. falciparum seropositivity reflect the heterogeneity of malaria transmission intensity between species. Age and distance to the water drain were the only two independent risk factors significantly associated with P. vivax exposure. As mentioned above, the effect of age on seroprevalence usually reflects the cumulative exposure to malaria infections occurring with age rather than a higher risk of exposure in adults as compared to children. However, the significant interaction found between age and house distance to the drain, beside the cumulative effect of age on seroprevalence, indicates a significant lower exposure in children aged < ten years. Indeed, the effect of distance (<100 m versus ≥100 m) on P. vivax seropositivity was only significant in adults, indicating that children living close to the drain were not exposed any more to infectious bites compared to children living farther. Water accumulation in the drain provides breeding habitats for Anopheles malaria vectors after rains [30, 39], resulting in a higher risk of exposure to infectious vectors for people living close to the drain. This is also supported by the results of univariate analysis for the risk of P. vivax infection. However, since adults are more likely to have outdoor evening activities, this could explain why children below ten, were no more at risk of malaria while living close to the drain.

Besides age, settlement and prior work as moto-taxi driver in the previous year, there were two additional independent risk factors for P. falciparum seropositivity. Historically, the three study settlements were important foci of P. falciparum transmission during the epidemic of 1998 and the early 2000s [47, 48]. Therefore, higher seroprevalence in JCM and NP, compared to PAV again suggest differential reduction rates in MTI between villages (heterogeneity). Occupational activities outside settlements (as moto-taxi driver) that were associated with increased P. falciparum exposure support the hypothesis that transmission occurs not only in the immediate house environment as previously described by Guthmann et al. [29], but outside the community as well. Moto-taxi drivers in the study area usually work in evening times, when exposure to An. albimanus bites increases. Occupational risk for malaria has also been found in the Peruvian Amazon Region, where malaria morbidity has been clearly associated with forest-related activities such as farming, land clearing and wood extraction [49, 50].

The catalytic model used to generate PfGLURP seroprevalence curves suggested that a significant decline in P. falciparum transmission occurred in the study area approximately eight years before the current study (around 2002). This supports the idea that the wide scale implementation of ACT since 2001 has significantly contributed to the reduction of P. falciparum MTI. Indeed, following the 1998 epidemic and the subsequent intensified control interventions implemented [9], P. falciparum incidence returned to the pre-epidemic levels (1997) by 2000–2001 (around 5,000 cases), and since 2002 it steadily decreased to reach almost undetected levels from 2007 onwards (Figure 3). In addition, the absence of P. falciparum surge after the drought period in 2007, as observed with P. vivax, advocates for the additional significant effect of the ACT deployment on P falciparum MTI in the northern coast region [8].

Since the end of the drought period and the resuming of the rains, P. vivax incidence substantially increased [7] (Figure 3) which is concordant with the fitted results from the model, i.e., significant increase in P. vivax MTI after 2006. Plasmodium vivax transmission in this area seems to be more sensitive than P. falciparum to changes in climatological and environmental conditions, especially since the implementation of ACT. The recent changes in species specific MTIs are also in line with the results of the multivariate risk factor analysis. While children under-five were significantly at lower risk of P. falciparum exposure compared to older children age 6-10, this was not the case for the P. vivax exposure where the under five were at similar risk of exposure compared to the 6-10 years old.

The number of antigens used in the two serological tests could be a limitation of the study. Due to the degree of sequence variability, it is recommended to use multiple antigens for each species in order to optimize the sensitivity of the antibody detection and to improve the identification of changes in malaria transmission over time in low transmission settings [20, 38]. However, the observation that the antigens used in the study are relatively conserved [51] combined with results from recent studies performed in Cambodia, Vietnam and Peru suggests that PfGLURP and PvMSP1₁₉ are highly suitable for the detection of antibodies against P. falciparum and P. vivax, respectively [33, 52]. Multiplexing the serological tests as well as standardizing the antigens used is of crucial importance for monitoring elimination efforts and comparing trends across countries and regions.