Skip to main content
Browse Subject Areas

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

Impact of the dog population and household environment for the maintenance of natural foci of Leishmania infantum transmission to human and animal hosts in endemic areas for visceral leishmaniasis in Sao Paulo state, Brazil

  • Patricia Sayuri Silvestre Matsumoto ,

    Contributed equally to this work with: Patricia Sayuri Silvestre Matsumoto, Elivelton da Silva Fonseca, José Eduardo Tolezano

    Roles Conceptualization, Formal analysis, Funding acquisition, Investigation, Methodology, Resources, Software, Validation, Visualization, Writing – original draft, Writing – review & editing

    Current address: Av. Dr. Arnaldo, São Paulo, SP, Brazil

    Affiliation Parasitology and Mycology Center, Adolfo Lutz Institute (IAL), Sao Paulo, Sao Paulo, Brazil

  • Roberto Mitsuyoshi Hiramoto ,

    Roles Methodology, Writing – review & editing

    ‡ These authors also contributed equally to this work.

    Affiliation Parasitology and Mycology Center, Adolfo Lutz Institute (IAL), Sao Paulo, Sao Paulo, Brazil

  • Virgínia Bodelão Richini Pereira ,

    Roles Investigation, Methodology, Writing – review & editing

    ‡ These authors also contributed equally to this work.

    Affiliation Adolfo Lutz Institute, Regional Laboratories Center II Bauru, Bauru, Sao Paulo, Brazil

  • Valéria Medina Camprigher ,

    Roles Methodology, Writing – review & editing

    ‡ These authors also contributed equally to this work.

    Affiliation Center for Zoonoses Control of Bauru, Health Secretariat of Bauru, Bauru, Brazil

  • Helena Hilomi Taniguchi ,

    Roles Methodology, Writing – review & editing

    ‡ These authors also contributed equally to this work.

    Affiliation Parasitology and Mycology Center, Adolfo Lutz Institute (IAL), Sao Paulo, Sao Paulo, Brazil

  • José Eduardo de Raeffray Barbosa ,

    Roles Methodology, Writing – review & editing

    ‡ These authors also contributed equally to this work.

    Affiliation Bioterium nucleos, Adolfo Lutz Institute (IAL), Sao Paulo, Sao Paulo, Brazil

  • Luiz Ricardo Paes de Barros Cortez ,

    Roles Methodology, Writing – review & editing

    ‡ These authors also contributed equally to this work.

    Affiliation Center for Zoonoses Control of Bauru, Health Secretariat of Bauru, Bauru, Brazil

  • Elivelton da Silva Fonseca ,

    Contributed equally to this work with: Patricia Sayuri Silvestre Matsumoto, Elivelton da Silva Fonseca, José Eduardo Tolezano

    Roles Validation, Writing – original draft, Writing – review & editing

    Affiliation Institute of Geography, Federal University of Uberlândia, Uberlândia, Minas Gerais, Brazil

  • Raul Borges Guimarães ,

    Roles Validation, Writing – review & editing

    ‡ These authors also contributed equally to this work.

    Affiliation Department of Geography, Sao Paulo State University/Faculty of Sciences and Technology (FCT/UNESP), Presidente Prudente, Sao Paulo, Brazil

  • José Eduardo Tolezano

    Contributed equally to this work with: Patricia Sayuri Silvestre Matsumoto, Elivelton da Silva Fonseca, José Eduardo Tolezano

    Roles Conceptualization, Funding acquisition, Investigation, Project administration, Resources, Supervision, Visualization, Writing – original draft, Writing – review & editing

    Affiliation Parasitology and Mycology Center, Adolfo Lutz Institute (IAL), Sao Paulo, Sao Paulo, Brazil


When it comes to visceral leishmaniasis (VL) in Brazil, one of the main targets of public health policies of surveillance is the control of domestic canine reservoirs of Leishmania infantum. This paper aims to evaluate the effect of the dog population and household environment for the maintenance of natural foci in the transmission to human and animal hosts in an endemic city for VL, Bauru, in Brazil. We collected 6,578 blood samples of dogs living in 3,916 households from Nov.2019 to Mar.2020 and applied geospatial models to predict the disease risk based on the canine population. We used Kernel density estimation, cluster analysis, geostatistics, and Generalized Additive Models (GAM). To validate our models, we used cross-validation and created a receiver operating characteristic (ROC) curve. We found an overall canine VL (CVL) seroprevalence of 5.6% for the sampled dogs, while for the households, the positivity rate was 8.7%. Odds ratios (OR) for CVL increased progressively according to the number of canines for >2 dogs (OR 2.70); households that already had CVL in the past increased the chances for CVL currently (OR 2.73); and the cases of CVL increase the chances for human VL cases (OR 1.16). Our models were statistically significant and demonstrated a spatial association between canine and human disease cases, mainly in VL foci that remain endemic. Although the Kernel density ratio map had the best performance (AUC = 82), all the models showed high risk in the city’s northwest area. Canine population dynamics must be considered in public policies, and geospatial methods may help target priority areas and planning VL surveillance in low and middle-income countries.

1. Introduction

Leishmaniasis is a group of infectious diseases caused by a protozoan of the Leishmania genus that affects humans and animals. The transmission occurs by the bite of the dipterous of the subfamily Phlebotominae, the sand flies. It is considered one of the most widely distributed neglected diseases worldwide [1], being a health problem in North and East Africa, West and East Asia, and the Americas [2].

Leishmaniasis is a focal disease, and its epidemiology differs according to nosogeography entity, which means that different spatial patterns for each species of Leishmania occur, and different strategies for the control of leishmaniasis are demanded. Visceral leishmaniasis (VL), for instance, is one of the most severe leishmaniasis. It threatens more than one billion and a half persons living in at-risk areas around the world. In 2015, between 50 and 90 thousand new cases were estimated per year, with an incidence rate of 2.27 per 100 thousand inhabitants. Only six countries, including Brazil, accounted for about 90% of all new cases [3, 4]. In the last update in 2021, Brazil notified 51,931 cases from 2005 to 2019, a mean of 3,462 per year [2].

The main reservoir of L. infantum in domestic habitats is the domestic dog [5, 6]. However, for control programs, an integrated knowledge about the ecological niche of the vector and environmental conditions is fundamental to address effective measures [7]. For this reason, geospatial modeling is a valuable instrument to target interventions of control programs [8].

In Brazil, there is a great difficulty for the effective implementation and operation of the VL control programs [9]. Overall, the Brazilian Visceral Leishmaniasis Control Program (VLCP) is based on canine reservoirs’ control, consisting of serosurvey and culling of dogs; control of the vector spraying insecticides inside the households and environmental management; and early diagnosis and treatment of human cases [8].

The first evidence of VL in Sao Paulo state was the discovery of Lutzomyia longipalpis in the urban area of Araçatuba municipality in 1997 [10]. In 1998, autochthonous canine cases of VL were reported, and for the first time in the state, the sand fly was suspected as the vector. In 1999 autochthonous human cases were reported for the first time [11, 12].

Regarding public health, several factors may be responsible for the persistence of the number of cases and deaths in Sao Paulo state, such as difficulty for early diagnosis and specific treatment in humans; difficulty in the correct identification and control of domestic reservoirs; and difficulty in controlling vector population [6, 13]. In addition, there is unclear knowledge about other determinants that may influence the design of novel strategies for the control and prevention of VL [14].

In Sao Paulo state, the municipality of Bauru had the first evidence of VL in 2002, when the sandflies were found, and the first autochthonous infection in a dog was reported. The first human records occurred in 2003. Since that time, there have been 580 cases and 46 deaths, a lethality rate of 7.9% from 2003 to 2020 [15]. Bauru was chosen to perform this research because of its high number of cases and endemicity in the region. In Bauru, there is a lack of information about the spatial distribution and a long-term follow-up of CVL, which could aid in the global understanding of the problem. Mapping the exact occurrence of the human or canine cases may help better understand the disease and plan public policies regarding VL.

This study aimed to calculate the impact of the household environment and canine population for visceral leishmaniasis risk through geospatial methods. We hypothesize that: a) in endemic areas for VL, a higher number of dogs in the households increases human or canine VL cases. b) the urban area is stratified by different geographical profiles that allow the remaining endemicity, needing targeted strategies as control measures. Using spatial analysis and statistical approaches, we constructed a space framework based on a large serosurvey conducted between November 2019 and March 2020 in the urban area of Bauru.

2. Materials and methods

Bauru is a central municipality in the state of Sao Paulo (22°18’52" S, 49°03’31" W). It is crossed by important highways: SP-300 Marechal Cândido Rondon Highway, SP-294 Comandante João Ribeiro de Barros Highway, SP-321 Cezário José de Castilho Highway, and SP-225 Engenheiro João Batista Cabral Rennó Highway, giving access to the countryside cities of the state and the capital Sao Paulo (Fig 1).

Fig 1. Geospatial location of human cases of VL and the serosurvey in Bauru’s urban area.

A total of 6,578 blood samples of dogs were analyzed. Points represent each dog’s address. Positive dogs for CVL are represented by red dots and negative by blue. Black dots are HVL cases. Points are overlapped because of the spatial resolution of the cartographic scale.

According to the Köppen-Geiger climate classification updated system [16, 17], Bauru climate is classified as Cfa, which means temperate, without dry season, and with hot Summer. The soil is unsaturated, reddish and dark brown, fine clay sand texture, underlain by sandstone of the Bauru group. The vegetation of the urban area is Tropical Semideciduous Forest mixed with Cerrado, highly impacted by urbanization variable patchy pattern. The average altitude of Bauru is 527.4 m.

The population comprises an estimative of 379,297 inhabitants [18]. A research was conducted in 20,958 households in Sao Paulo state countryside, in which 52.6% proven dog ownership, with an average of 1.6 dogs at home and an inhabitant ratio of 1:4 dogs per person [19]. Following this study, we estimated the dog population at 99,815, according to the Brazilian Geographic and Statistics Institute [18].

2.1. Study design

The canine serosurvey was conducted from December 2019 to March 2020. Agents of the Center for Zoonoses Control visited 3,916 households to collect the blood samples of 6,578 canines. In addition, a short survey was applied (S1 Fig) to the dog’s guardian to identify the previous presence of an infected dog in the household (Fig 2A). We tested and mapped the samples (Fig 2B and 2C), and then we used spatial analysis to prepare data for creating thematic maps (Fig 2F) and statistical models (Fig 2D–2I).

Fig 2. Synthesis of the performed methodology.

The methodological proceedings were performed according to (a) Collecting samples and guardian’s survey. (b) Analyzing the samples. (c) Mapping the samples of dogs and human cases of VL. (d) Changing the scale of the dogs into households. (e) Applying statistical models. (f) Performing thematic maps. (g) Preparing data for the spatial model. (h) Preparing training data for the spatial model. (i) Validation of the models. (j) Using acquired knowledge for decision-making.

We applied a framework of starting with a statistical to select variables most influential in running the spatial models. The chosen method of statistics was the binary logistic regression, used to associate the size of the canine population related to the odds ratio of having VL cases. As the results of the statistical model were statistically significant, we ran the spatial models. The generalized additive model—GAM, geostatistic model, and Kernel density ratio were chosen to identify the spatial dependence of the cases and their spatial association with the number of the canine population, emphasizing hotspots of CVL. We finally validated our model using cross-validation (Fig 2F and 2J). The methodology performed here will be detailed in the following sections.

2.1.1 Definition of cases.

Dual-Path Platform rapid test (TR-DPP, Biomanguinhos®, Rio de Janeiro, Brazil,) is used by the Brazilian VLCP to test the samples. The TR-DPP® is a test for Leishmania infantum based on the reaction of IgG to the antigen K28. Enzyme Linked Immunosorbent Assay ELISA—Biomanguinhos® is used to confirm the positive diagnoses. ELISA is characterized by the reaction of soluble and purified antigens of Leishmania promastigotes, obtained from cultures and adsorbed in microtiter wells with Leishmania-specific antibodies present in serum samples. The diagnostic was run in a Multiscan spectrophotometer using a 450 nm filter and cutoff values ("Cutoff" = CO): CO = average negative controls x 2. The diagnoses were performed according to the manufacturer’s instructions and the directions of the VLCP.

A combination of TR DPP® and ELISA reagent was considered a positive result according to the Brazilian VLCP recommendations for canine diagnoses, routinely used by the Centers for Zoonoses Control in Sao Paulo [20]. TR DPP® non-reagent was considered a negative result–Fig 2A. The prevalence was calculated based on the outcome, being a proportion of a dog found positive for CVL divided by the sampled dog population. The guardian of the dog provided consent for sample collection involving domestic dogs in the areas surveyed. All serosurvey was supervised by the veterinary group of the Center for Zoonoses Control in Bauru municipality. Households without dogs, closed or that refused to give consent were excluded from the analysis.

The human laboratory diagnoses are based mainly on serological methods and microscopic diagnoses (parasitological). When amastigotes are identified, it is considered a certainty diagnostic. Patients with clinical manifestation and reagent rapid immunochromatographic test rK39 and/or Indirect immunofluorescence with titers equal to or greater than 80 are considered positive for VL [20]. Human cases addresses come from the epidemiological surveillance center (CVE), State Health Secretariat [15] and comprehend secondary databases from the Brazilian System for Notifiable Diseases (SINAN) in the period from 2003 to 2020.

This study was approved under number 03/2000 in the Ethics Committee on the Use of Animals in Research at the Adolfo Lutz Institute, Sao Paulo-SP, Brazil.

2.1.2. Mapping.

The addresses of dogs and the human cases were geocoded by an application programming interface (API) of Google Maps (Google®), based on the municipality’s cartographic street map. A lower score of geocoding was topologically adjusted to ensure the correction of georeferencing. Point features were plotted in a Geographic Information Systems (GIS) ArcGIS 10.2.2 (ESRI, Imagem).

Canine data and the households were mapped as point data; however, they have different cartographic scales once the household (mapped by address) is the boundary, besides dogs are also represented by geocoded addresses. It means that more than one dog in a household is visualized as one point, but there are overlapped(s) dog(s) (points) in there. Bearing this in mind, the dog layer was transformed into the household layer with the number of dogs (Fig 2C and 2D) and categorized as negative or positive for VL in each survey. Fig 3 shows the mapped data.

Fig 3. Geocoded households of Bauru, stratified by dog population, positive dogs, and buffer zones.

The households are identified according to the conducted surveys. Yellow, orange, or red symbols represent the dog’s households sampled. Yellow had not an infected dog; orange had an infected dog (in the past); red had and currently has an infected dog. Proportional circles represent the number of dogs in each household. We created a buffer of 100m in each sampled household to calculate the number of dogs, positive or negative dogs in this area.

To run the Generalized Additive Model (GAM), we created a fishnet grid of 1000x1000 of 50m containing the estimated number of dogs per domicile in each coordinated. We calculated the number of dogs based on the human population by census tracts [21], in a ratio of 1:4 dogs/inhabitants [19]. The mean number of dogs was divided by the mean number of households in each census tract. We then created a centroid and calculated the Inverse distance weighted (IDW) interpolation, considering the mean number of dogs per domicile. The grid values were extracted from the raster surface generated by the IDW (S7 and S8 Figs).

To analyze the area of influence of households with infected dogs in the environment, we created buffer zones of 100m (Fig 3). We then calculated the number of dogs, negative dogs, and positive dogs using spatial analysis tools. Finally, we aggregated features of point data into polygons, using the census tracts database, to stratify the prevalence spatially.

2.2. Statistical and spatial analysis

For all the performed calculations, we considered a significant value at p≤0.05. We used the geographic information system (GIS) ArcGIS 10.2.2 and R language, with several packages described in the sections below.

The number of dogs per household was categorized as binary to verify each group, for instance, households that have only one dog (1 = 1 dog and 0 > 2 dogs); two dogs (1 = 2 dogs; 0 = 1 or > 2 dogs); and two or more dogs (1 = > 2 and 0 = 1 or 2 dogs). Households that already had a positive dog or a human case were also categorized as binary, e.g., 1 for cases and 0 for non-cases.

2.2.1. Pearson’s correlation.

We calculated Pearson’s correlation to identify a possible association between the number of cases of CVL and: i) the number of investigated samples or ii) the number of households that already had an infected dog, or iii) the number of households that already had and currently have an infected dog/dogs.

2.2.2. Binary logistic regression.

We tested if the households with an infected dog (outcome = 0 for a household with no infected dog/dogs or outcome = 1 for a household with an infected dog/dogs) or an area of influence of household (outcome = 0 for areas of influence of household with no infected dog/dogs or outcome = 1 for an area of influence of household with infected dog/dogs) could increase the chances to have cases of the disease. We used ’oddsratio’ package in RStudio (4.0.0).

2.2.3. K-function.

Being aware of spatial dependence of CVL promoting different risks or protection, we evaluate, locally, the spatial interactions in the urban neighborhoods. Ripley’s K-function with 999 permutations was applied to identify households’ spatial patterns at distances [22]. In this function, K(t) is the number of events within a distance of an arbitrary event, divided by the overall density of events. We plotted maximum and minimal envelopes of K(t) simulated values, giving the statistical significance for clustered or dispersed patterns (S2 Fig).

2.2.4. Cluster analysis.

We used cluster analysis to detect significant concentrations of CVL within Confidence Intervals (CI) of 90%, 95%, and 99%—S5 Fig. Clusters were calculated using the Getis-Ord Gi statistic, which identifies features with high or low values of a spatial cluster. The pattern can be expressed by clustered, dispersed, or random features representing a measurable spatial aggregation unit.

2.2.5. Kernel density.

Using the K-function dependence, we chose the minimal distance of concentration of our data, 0.5 km, to set the bandwidth. We used the quartic kernel function [23], which is given by (1) where:

  1. i = 1,,n are the input points.
  2. di is the distance between the point s and the observed event in location,
  3. si and τ is the radius centered on s.

We plotted Kernel density maps for CVL cases (S3 Fig) and canine samples (S4 Fig). A Kernel density ratio map was then performed (CLV: samples), which gives a visualization of the risk for the disease.

2.2.6. Geostatistical approach.

According to the number of dogs, a geostatistical approach was performed to predict the higher risk areas for CVL. We used the Ordinary Kriging method and select two datasets: cases of CVL and number of dogs. We adjusted data in a stable model in a semivariogram, in which for a set of experimental values z(x) and Z (x1+h), separated by h distance, is defined by the Eq 2: (2) Where,

  1. N(h) is the number of experimental pairs;
  2. h is the regular interval that separates z(xi) e z(xi+h)

Geostatistics parameters were adjusted as follow: number maximum and minimal neighbours = 5 and 2, respectively; lags = 12; lag size = 0.64; nugget = 0.46; range = 3.8974; sill = 0.062 and 45 degrees–S6 Fig and S1 File.

2.2.7. Generalized additive model.

We run a GAM according to an approach reported for case-control data [24, 25], in which we considered Yi = 1 (cases) and Yi = 0 (non-cases), d is the number of dogs in location i, and P(Yi = 1|di Si) is calculated according to Eq 3: (3) where,

  1. a is the ratio of cases to non-cases,
  2. b is the coefficient for the number of dogs per household,
  3. Si is a function of the residual spatial variation after accounting for the effect of the number of dogs.

We model Si by a locally estimated scatterplot smoothing (LOESS) regression smoother against the Universal Transverse Mercator coordinates. We choose the optimal smoother parameter of the models based on Akaike’s Information Criterion (AIC) [24] after testing multiple bandwidths (S2 File). We predicted the adjusted log odds for each location and omitted the covariate and smoothing terms through a null model. GAM was run in RStudio using the ’gam’ package.

2.3. Cross-validation

For further analysis, we validated our models using cross-validation. We created random samples in ArcGIS and split our database into training (75%, 2,937 points) and testing (25%, 979 points). Spatial models were created using the training dataset to predict the risk for the testing dataset. For each model, the best threshold was chosen, and we calculated specificity, sensitivity, and accuracy for correctly predicting the observed value of a case or non-case at the testing coordinates. To sum up, we calculated the area under the receiver operating characteristic (ROC) curve (AUC) with 95% confidence interval, which plots the true positive rate versus false positive rate, allowing identifying the performance of the models. We used the ’pROC’ and ’ggplot2’ packages in RStudio.

3. Results

We investigated 6,578 dogs, in which we found Anti-Leishmania spp. antibodies in 8.1% of TR DPP® (535/6,578) and 5.6% in both TR DPP® and ELISA (369/6,578). We found different spatial prevalence in the city, ranging from 0 to 50%. The mean prevalence was 2.67%, and the higher prevalence (>7.5%) was regularly distributed in the sampled area (Fig 4).

Fig 4. Geocoded households of Bauru, stratified by prevalence and cluster analysis.

Each black dot is a household with no identified cluster. Magenta dots are the clusters of domiciles with infected dogs; green dots are the clusters of households that already had infected dogs; and orange dots represent the clusters of households that had and currently have infected dogs. Different size symbols and opacity households were set to ensure the spatial visualization of overlapped households.

3.1. Cluster analysis

We identified a clustered pattern of households with CVL with statistical significance from approximately 0.5 to 6.5 km, and a clustered pattern of human cases from 0.5 to 4 km (S2 Fig). We found spatial clusters of high values (hot spots) in the west, north, east, south, northeast, southwest, southeast, and in Tibiriçá, a municipality district (Fig 4).

3.2. Kernel’s map

The ratio of cases per sample concentration represented in the Kernel map shows high-risk areas in the Pq. Jaraguá, Pq. Santa Edwiges and Vila Nipônica neighborhoods (Fig 5). Other high-risk concentrations in Kernel’s map represent the border effect.

Fig 5. Kernel density ratio for canine visceral leishmaniasis.

Kernel density ratio map ranging from 0 (blue) to 0.7 (red), which gives a visualization of the risk dividing the concentration of cases of CVL (S3 Fig) by the concentration of dog samples (S4 Fig). The areas of higher risk are in the west and southwest.

3.3. Pearson correlation

Pearson’s correlation was positive and moderate considering the number of infected dogs and the households investigated (0.508, p-value = 0.000); positive low for infected dogs and the households that already had an infected dog/dogs (0.240, p-value = 0.000); and positive and low for infected dogs and the households that already had and currently have a dog/dogs with VL (0.129, p-value = 0.000). All conditions were statistically significant.

3.4. Binary logistic regression for visceral leishmaniasis

We investigated 3,916 households, in which 16.7% (656/3,916) already had a positive dog—independently when it was (Table 1). Nowadays, 8,7% (341/3,916) of the households have at least one positive dog. From the households with positive dog/dogs in the past, 17.0% (112/656) still have positive dog/dogs currently.

Table 1. CVL diagnostic, dog count in the households, and buffer zone extraction versus the number of sampled dogs in Bauru.

The maximum number of dogs per household was 17, but the mode was 1, and the mean was 1.67, with a standard deviation of 1.08. In an area of influence of a household (a = 31,374m2), the maximum number of dogs was 58, the mode 11, mean 16, and the standard deviation 11.37. Households that contained only one dog represent almost 60% of the domiciles, two dogs 27.6%, and more than three 14%.

Analyzing the census tracts, the odds ratio (OR) for the number of CVL and the examined dogs was 1.37 (Table 2). OR for CVL increased proportionally to the number of dogs. For households that contained only one dog was 0.40 and increased 242% for those with two dogs (OR = 1.39); and 97% when more than two dogs (OR = 2.70). For households that already had a positive dog, the OR was 2.73.

Table 2. Binary logistic regression of canine visceral leishmaniasis diagnostic, dog count in the buffer zone extraction versus the number of positive dogs and human cases in Bauru.

Similarly, OR for the area of influence of household (buffer of 100m) also increased according to the dogs’ population. From 10 to 20 dogs, OR was 1.25 and increased 120% for 21 to 30 dogs (OR = 2.76). The influence area of households with more than 30 dogs increased by more than 150% (OR>7). In an area of influence of household, households that already had a positive dog/dogs with VL increase the chances of CVL 299% (OR = 2.99), analogous to the condition of households that already had dog/dogs with VL, in which the OR was 2.73.

Considering the human cases in an area of influence of household, the number of dogs increased the chances 102%, and the number of positive dogs 116%, demonstrating the association between canine and human VL. The number of dogs increased the chances 261% for more than 40 dogs.

3.5. Spatial risk

Considering high OR for CVL according to the number of dogs, we created the spatial models using dogs as a predictor. Fig 6 shows that both models (geostatistical and GAM) were considerable commonality in the spatial pattern regarding the higher odds ratio areas. The higher risk is in the city’s borders, especially in the northwest and in the southeast. The last one can be a border effect. Overall, both models are spatially consistent with the Kernel density ratio map (Fig 5).

Fig 6. Predicting the risk for canine visceral leishmaniasis using geospatial methods.

Spatial prediction models using the dog population. Risk is scaled from low (blue) to high (red), as shown by the legends. (a) Geostatistical approach using the ordinary Kriging method. (b) Generalized additive model.

3.6. Cross-validation

We plotted the Kernel density ratio, Geostatistical, and GAM models in a receiver operating characteristic (ROC) curve (Fig 7). The Kernel density ratio map presented the best threshold of 0.059, a sensitivity of 88%, a specificity of 62%, and an accuracy of 64%. The geostatistical model presented the best threshold of 0.075, a sensitivity of 65%, a specificity of 52%, and an accuracy of 53%. The GAM model presented the best threshold of 0.076, a sensitivity of 88%, a specificity of 19%, and an accuracy of 25%. The first had an AUC of 0.81 (CI 0.76–0.85), the second of 0.59 (CI 0.53–0.66) and the third of 0.54 (CI 0.47–0.60). The Kernel density ratio map presented the best performance in the ROC curve (Fig 7).

Fig 7. Area under the receiver operating characteristic (ROC) for canine visceral leishmaniasis models.

For each model, the AUC was calculated with 95% confidence intervals. The best model in predicting canine risk disease was the Kernel density ratio map.

4. Discussion

In the current study, we found a CVL TR DPP® sero-reaction rate of 8.1% (535/6578) and 5.6% in TR DPP® confirmed by ELISA (369/6578), likely consistent with an endemic area of Sao Paulo state, Araçatuba, where the average prevalence between 2010 and 2015 was 6.8% [26]. Studies in other states of Brazil found a variable range of values of prevalence, for instance, 4.38% in Londrina, in Paraná state [27]; 8.1% in the communities of Cuiabá river [28], and 19.2% in Rondonópolis [29], both in the state of Mato Grosso; 4.16% in Belo Horizonte, in the state of Minas Gerais [30]; and a high prevalence of 50.3% in Buerarema, in the state of Bahia [31]. In Bauru, we found different spatial prevalences, ranging from 0 to 50%. Lamattina et al. (2019) found prevalences per site varying from 0 to 80% [32] and Carvalho et al. (2018) from 0 to 35% [29].

Particularly, prevalence can reveal bias once it may not represent the real number of canines. Overall, the serosurveys are directed to human case areas or areas of a suspect or identified CVL case [6, 20]. Historically, in Bauru, some neighborhoods have never performed a serosurvey before this study. On the other hand, some neighborhoods were investigated more than once since the first human case appearance, recognized by its recurrence of CVL. Our study planned the serosurvey to collect a large number of dog samples in different neighborhoods, giving a panorama of VL’s endemicity and spatial epidemiological profile in a short time. Nevertheless, sampled canines comprised less than 7% of the estimated dog population (6,578/99,815 dogs).

In the present study, our scale is the household instead of only the dogs, identifying spatial characteristics regarding the domiciles and dog population. We highlight that on the household scale, the positivity rate of domiciles that contains infected dogs (8.7%) is superior to the sampled prevalence of CVL (5.6%), which emphasizes the importance of the household environment in the disease context. Clusters of households that already had CVL can point out the areas that remain a source of infection and are unnoticed. Almost all investigated areas had these clusters. Additionally, asymptomatic dogs can be highly competent [33] and remain a source of infection without being identified. They contribute to the silent endemic areas. It can turn out into highly endemic areas or possibly a human case site.

The recent expansion of VL to new endemic areas has been attributed to the adaptation of L. longipalpis (sandflies) to naïve ecological niches. The risk of expansion of VL increases in areas identified as migratory poles of attraction. Moreover, CVL has been highlighted as the primary cause of outbreaks [34]. In these areas, canine enzootic disease precedes the appearance of human cases. In our study, human cases can increase 102% according to the dog population and 116% by the CVL, demonstrating the spatial association between canine and human VL. Other studies found that the risk increased substantially for individuals when the presence of seropositive dogs [35] or previous cases of CVL in the household [36, 37]. Furthermore, we identified a clustered pattern for both human and canine cases from approximately 500m.

As we identified, households that contain CVL and the dog population can increase the odds ratio for VL. They may influence the natural foci of Leishmania infantum transmission to human and animal hosts, which urges specific public policies focused on education in animal health, especially in areas target as critical. We found the same mean number of dogs per household (1.6), as reported in previous research [19]. However, about 15% of the domiciles contain more than three dogs. They should be monitored and investigated as possible infection sites. In the neighborhoods where the canine population is large, animal health assistance is required. Therefore, canine population dynamics must be considered in public policies.

Our results revealed that the risk of increasing CVL or human cases oscillated by areas. Of note, kernel maps studies have used the total number of cases or applied a constant [30]. Our study used the number of cases and samples, which gives a visualization of the risk. In accordance, a study [29] found a similar pattern of critical areas in the city’s borders. This peripheral spatial pattern seems to be expected in small and medium-sized cities of similar urbanization processes in low and middle-income countries where VL is endemic. The Kernel density ratio map was the best in the ROC curve, showing the potential of spatial analysis tools. Studies that use spatial models predicting disease risk are promising for decision-making in the control of VL.

Regarding leishmaniasis, such studies use machine learning for cutaneous leishmaniasis vectors prediction [38] or human cases prediction [39]. Nonetheless, studies using a machine learning-based approach of CVL risk prediction considering the number of dogs are still incipient. Bi et al. stress that future research about VL should focus on spatial simulation and agent-based simulation [40]. Machine learning is a novel approach that allows the forecast of disease risk. It can anticipate disease transmission dynamics and identify disease control strategies to fight endemic and emerging diseases [40]. Our models bring new insights for thinking VL through dogs from a social perspective, which has been one of the most debatable points of control programs and tends to be of low priority in the context of general public health.

It is a time of changing public policies in relation to VL. The general principles that guided the past control programs are now questionable. Guided by the VLCP, Brazilian municipalities have presented operational difficulties in executing VL control strategies [41]. Additionally, we emphasize the unavailability of the proven effectiveness of technical alternatives for laboratory diagnosis, identification, and elimination or protection of reservoirs [42].

In Brazilian cities, culling dogs has been recommended as a control measuring to reduce VL [20, 43], which creates a dog stigma. Nonetheless, culling dogs is highly controversial, considering the time between diagnosis and action. Also, the rapid replacement of euthanized dogs, entrance of new animals into the households [4446], and the persistent disease spread challenges public policies [13, 42, 47, 48]. On the other hand, other strategies, such as the use of dog collars with insecticides for sand flies, or vaccination [4951], have been highly encouraged due to their effectiveness in reducing the population of vector, parasitic load, and potentially the VL transmission [5255].

Furthermore, vast territorial areas should be treated by priority order, emphasizing different profiles of VL. The decision-making should be supported by an integrated approach, considering the genetic diversity of vectors [56, 57] and the protozoa [58, 59] but integrated with education, health, and environment, including vectors, causal agents, canines, households, population density, urbanization, industries, and environmental factors, such as vegetation, water bodies, temperature, and precipitation. Animal health needs to be discussed in public policies without its stigma. Furthermore, VL should be addressed in the context of One Health [42].

To conclude, this paper had several limitations that should be recognized. Firstly, we had to use the census tract information based on the human population to calculate the dog population grid because of the lack of animal information. This could be solved with an updated canine census, hardly achieved in low and medium-income countries. Secondly, the performance of our spatial models had medium and low accuracy, although the critical areas being commonly similar to the Kernel density ratio map of higher performance. A better model’s performance could be improved with an updated census and adding real-world covariates when data become available.

Yet, there are research gaps concerning VL, and many areas of study remain unexplored. It remains the question of balancing the effectiveness and costs involved in such a VL control plan [40]. As future work, the next step of our study is to analyze the canines’ role with new insights of controlling VL, for instance, the canine cohort of insecticide-impregnated collars, vaccination, and treatment in different areas of this endemic site, as an individual and collective measure in the environment.

5. Conclusions

As a rule of thumb, one can say that the number of dogs and the households impact the risk for maintenance of natural foci of Leishmania infantum transmission to human and animal hosts in endemic areas for VL. Overall, this investigation serves as a case study for regional and global applications. It reveals the importance of canines on the household scale in low and middle-income countries. It is time for changing VL public policies using a targeted plan of priority through spatial analysis. This statement invites further investigations regarding VL characteristics involving socioeconomic and environmental variables in the context of One Health.

Supporting information

S1 Fig. Survey applied in the collection of dog’s blood.

The survey was applied at the moment of the collection of dog´s blood to detect anti-leishmania antibodies. It is a short inquiry once the study covered a large number of dogs, and it was conducted by agents of the center for zoonoses control (not specialists). After the diagnose results, the guardians of the positive dogs were notified to schedule an appointment with a veterinarian.


S2 Fig. K-function for visceral leishmaniases at distances.

The red line is the observed values. The blue line is the expected for a random sample. Dashed lines represent the superior and inferior envelopes for statistical significance. (a) households with CVL currently; (b) human cases (2003–2019) of VL in Bauru, São Paulo, Brazil.


S3 Fig. Kernel map for canine cases of visceral leishmaniasis.

We performed a Kernel density map for the total number of canine cases using a bandwidth of 500m (approximately the minimal concentration of K-function). We select the default cells and the output in meters square.


S4 Fig. Kernel map for canine samples.

We performed a Kernel density map for the total number of dog samples using a bandwidth of 500m (approximately the minimal concentration of K-function). We select the default cells and the output in meters square.


S5 Fig. Cluster map for households that have CVL currently.

This is an example of a cluster map for the households that have CVL currently. For each category, cluster maps were created: i)households that have CVL; ii) households that already had CVL; iii) households that already had and currently have CVL. The coldspots and the non-significant data were excluded in the final cartographic representation (Fig 4).


S6 Fig. Semivariogram for canine visceral leishmaniasis and the number of canines.

The stable theoretical model was adjusted to the points according to the parameters described below.


S7 Fig. Features of the grid of the GAM model.

To calculate the number of dogs per domiciles, we used the study of Alves et al. 2005, an investigation conducted in the cities of the state of São Paulo, considering a ratio of 1:4 dogs/persons. We calculated the number of dogs based on the human population census tract (Matsumoto et al., 2021). We then used the number of households (IBGE,2010) to find the number of dogs at that point (centroid). Finally, a fishnet of 1000 cells versus 1000 cells was created to extract the point value of the number of dogs interpolation.


S8 Fig. Interpolation of the number of dogs using Iverse Distance Weighted (IDW).

This method interpolates the estimative of the cell values using the average of the points in each region. We used the census tract data and the estimative of dogs according to Alves et al. 2005. The map shows a higher number of dogs per domicile (brown to white) in the city’s outskirts and fewer dogs in the central areas (green to yellow). The grid (S7 Fig) extracted the IDW values of the correspondent location of each point. The grid can not be seen on the cartographic scale of 1:50,000, but it is visible on the scale of 1:5,000.


S1 File. Parameters of the semivariogram.

We adjusted both datasets (CVL cases and the number of dogs) using a stable model.


S2 File. AIC results for each span.

We select the best (minimal) Akaike information criterion (AIC) for choosing the span function of our GAM model. The best AIC = 1704, span = 0.15.



We thank the Bauru Center for Zoonoses Control staff: Aline Fernanda Peral Cano, Cláudia Cilene Barbosa Gomieri, Josiane Silva Cano, Maria Emília Bodini Santiago, Roldão Antônio Puci Neto; Adolfo Lutz Institute students and staff: Aghata Regina de Oliveira Alves Palmeira; Amanda Gonçalves Martins da Costa; Alessandra Ventura Santos, Luana Ribeiro Manzi, Naíra Ruiz Lenharo, and Maria Cristina de Carvalho.


  1. 1. WHO. Control of the Leishmaniases. WHO Expert Comm Control Leishmaniases, Geneva, 22–26 March 2010. 2010; 22–26. Available:
  2. 2. WHO. Leishmaniasis, number of cases of visceral leishmaniasis reported. 2021 pp. 27–28. Available:
  3. 3. Wamai RG, Kahn J, McGloin J, Ziaggi G. Visceral leishmaniasis: a global overview. J Glob Heal Sci. 2020;2: 1–22.
  4. 4. Pasquali AKS, Baggio RA, Boeger WA, González-Britez N, Guedes DC, Chaves EC, et al. Dispersion of Leishmania (Leishmania) infantum in central-southern Brazil: Evidence from an integrative approach. PLoS Negl Trop Dis. 2019;13: 1–20. pmid:31465422
  5. 5. Dantas-Torres F, Solano-Gallego L, Baneth G, Ribeiro VM, de Paiva-Cavalcanti M, Otranto D. Canine leishmaniosis in the Old and New Worlds: Unveiled similarities and differences. Trends Parasitol. 2012;28: 531–538. pmid:22995719
  6. 6. BRASIL. Manual de Vigilância e Controle da Leishmaniose Visceral. Brasília; 2014 [cited 20 Jan 2015] p. 120. Available:
  7. 7. Peterson AT, Pereira RS, Camargo-Neves VF. Using epidemiological survey data to infer geographic distributions of leishmaniasis vector species. Rev Soc Bras Med Trop. 2004;37: 10–14. pmid:15042174
  8. 8. Pigott DM, Golding N, Messina JP, Battle KE, Duda KA, Balard Y, et al. Global database of leishmaniasis occurrence locations, 1960–2012. Sci Data. 2014;1: 1–7. pmid:25984344
  9. 9. Harhay MO, Olliaro PL, Costa DL, Costa CHN. Urban parasitology: Visceral leishmaniasis in Brazil. Trends Parasitol. 2011;27: 403–409. pmid:21596622
  10. 10. Costa AIP da, Casanova C, Rodas LAC, Galati EAB. Atualização da distribuição geográfica e primeiro encontro de Lutzomyia longipalpis em área urbana no Estado de São Paulo, Brasil. J Public Health (Bangkok). 1997;31: 535–541. pmid:9430928
  11. 11. Camargo-Neves VLF De, Katz G, Rodas LAC, Poletto DW, Lage LC, Spínola RMF, et al. Utilização de ferramentas de análise espacial na vigilância epidemiológica de leishmaniose visceral americana—Araçatuba, São Paulo, Brasil, 1998–1999. Cad Saude Publica. 2001;17: 1263–1267. pmid:11679900
  12. 12. Tolezano JE, Taniguchi HH, Araújo MDFL, Barbosa JADR, Barbosa JEDR, Luvizotto MC, et al. Leishmaniose Visceral Americana em Araçatuba, região oeste do Estado de são Paulo. Investigações laboratoriais e diagnóstico etiológico de uma doença emergente no estado de São Paulo. 35 Congresso da sociedade Brasileira de Medicina Tropical, 1999, Guarapari Revista da sociedade Brasileira de Medicina Tropical, 1999. 1999. pp. 9–39.
  13. 13. Dantas-Torres F, Miró G, Bowman DD, Gradoni L, Otranto D. Culling Dogs for Zoonotic Visceral Leishmaniasis Control: The Wind of Change. Trends Parasitol. 2019;35: 97–101. pmid:30554966
  14. 14. Moreira ED, De Souza VMM, Sreenivasan M, Lopes NL, Barreto RB, De Carvalho LP. Peridomestic risk factors for canine leishmaniasis in urban dwellings: New findings from a prospective study in Brazil. Am J Trop Med Hyg. 2003;69: 393–397. pmid:14640499
  15. 15. CVE. Centro de Vigilância Epidemiológica “Prof. Alexandre Vranjac.” 2020 [cited 20 Jan 2017]. Available:
  16. 16. Peel MC, Finlayson BL, McMahon TA. Updated world map of the Köppen-Geiger climate classification. Hydrol Earth Syst Sci. 2007;11: 1633–1644.
  17. 17. Alvares CA, Stape JL, Sentelhas PC, De Moraes Gonçalves JL, Sparovek G. Köppen’s climate classification map for Brazil. Meteorol Zeitschrift. 2013;22: 711–728.
  18. 18. IBGE. Brazilian Institute of Geography and Statistics. 2021. Available:
  19. 19. Alves MCGP, Matos MR de, Reichmann MDL, Dominguez MH. Estimation of the dog and cat population in the State of São Paulo. Rev Saude Publica. 2005;39: 891–897. doi:/S0034-89102005000600004 pmid:16341397
  20. 20. Ministério da Saúde do Brasil. Guia de vigilância em saúde 3a edição. BRASIL, Ministério da Saúde.Secretaria de Vigilância em Saúde. Coordenação-Geral de Desenvolvimento da Epidemiologia em Serviços. 2019. Available:érios+Diagnósticos+de+Infecção+Associada+à+Assistência+à+Saúde+Neonatologia/9fa7d9be-6d35-42ea-ab48-bb1e068e5a7d
  21. 21. Matsumoto PSS, Flores EF, Barbosa JS, Pessoto UC, Tolezano , Hiramoto José Eduardo, et al. The path of canine visceral leishmaniasis versus the path of the Center for Zoonoses Control: contributions of spatial analysis to health. Reports Public Heal. 2021;printing p.
  22. 22. Dixon PM. Ripley ‘ s K function Theoretical K (t) function. Statistics (Ber). 2014; 1–16.
  23. 23. Bailey TC, Gatrell AC. Interactive spatial data analysis. New York: Longman; 1995.
  24. 24. Baker S, Holt KE, Clements ACA, Karkey A, Arjyal A, Boni MF, et al. Combined high-resolution genotyping and geospatial analysis reveals modes of endemic urban typhoid fever transmission. Open Biol. 2011;1. pmid:22645647
  25. 25. Webster T, Vieira V, Weinberg J, Aschengrau A. Method for mapping population-based case-control studies: An application using generalized additive models. Int J Health Geogr. 2006;5: 1–10. pmid:16390549
  26. 26. Costa DNCC, Bermudi PMM, Rodas LAC, Nunes CM, Hiramoto RM, Tolezano JE, et al. Human visceral leishmaniasis and relationship with vector and canine control measures. Cad Saude Publica. 2018;52: 1–11. pmid:30484481
  27. 27. Benitez A do N, Martins FDC, Mareze M, Nino B de SL, Caldart ET, Ferreira FP, et al. Spatial and simultaneous seroepidemiology of anti-Leishmania spp. antibodies in dog owners and their dogs from randomly selected households in a major city of southern Brazil. Prev Vet Med. 2018;154: 47–53. pmid:29685444
  28. 28. Franco Sousa VR, De Lima Ruy Dias ÁF, Rodrigues JY, De Medeiros Torres M, Assunção Rosa Moreira JM, Nakazato L, et al. Canine visceral leishmaniasis in riverside communities of the Cuiabá river watershed. Semin Agrar. 2019;40: 3313–3319.
  29. 29. Carvalho AG de, Luz JGG, Rodrigues LD, Dias JVL, Fontes CJF. High seroprevalence and peripheral spatial distribution of visceral leishmaniasis among domestic dogs in an emerging urban focus in Central Brazil: a cross-sectional study. Pathog Glob Health. 2018;112: 29–36. pmid:29460695
  30. 30. de Arruda RMF, Cardoso DT, Teixeira-Neto RG, Barbosa DS, Ferraz RK, Morais MHF, et al. Space-time analysis of the incidence of human visceral leishmaniasis (VL) and prevalence of canine VL in a municipality of southeastern Brazil: Identification of priority areas for surveillance and control. Acta Trop. 2019;197: 105052. pmid:31233726
  31. 31. Leça Júnior NF, Guedes PEB, Santana LN, Almeida V dos A, Carvalho FS, Albuquerque GR, et al. Epidemiology of canine leishmaniasis in southern Bahia, Brazil. Acta Trop. 2015;148: 115–119. pmid:25917715
  32. 32. Lamattina D, Berrozpe PE, Casas N, Moya SL, Giuliani MG, Costa SA, et al. Twice upon a time: The progression of canine visceral leishmaniasis in an Argentinean city. PLoS One. 2019;14: 1–18. pmid:31276573
  33. 33. Laurenti MD, Rossi CN, Matta VLR da, Tomokane TY, Corbett CEP, Secundino NFC, et al. Asymptomatic dogs are highly competent to transmit Leishmania (Leishmania) infantum chagasi to the natural vector. Vet Parasitol. 2013;196: 296–300. pmid:23562649
  34. 34. Teixeira-Neto R, da Silva E, Nascimento R, Belo V, de Oliveira C, Pinheiro L, et al. Canine visceral leishmaniasis in an urban setting of Southeastern Brazil: an ecological study involving spatial analysis. Parasit Vectors. 2014;7: 485. pmid:25326767
  35. 35. Costa DNCC, Blangiardo M, Rodas LAC, Nunes CM, Hiramoto RM, Tolezano JE, et al. Canine visceral leishmaniasis in Araçatuba, state of São Paulo, Brazil, and its relationship with characteristics of dogs and their owners: A cross-sectional and spatial analysis using a geostatistical approach. BMC Vet Res. 2018;14: 1–13. pmid:29291752
  36. 36. Coura-Vital W, Reis AB, Reis LES, Braga SLÔ, Roatt BM, Aguiar-Soares RD de O, et al. Canine visceral leishmaniasis: Incidence and risk factors for infection in a cohort study in Brazil. Vet Parasitol. 2013;197: 411–417. pmid:23941965
  37. 37. Chapman LAC, Jewell CP, Spencer SEF, Pellis L, Datta S, Chowdhury R, et al. The role of case proximity in transmission of visceral leishmaniasis in a highly endemic village in Bangladesh. PLoS Negl Trop Dis. 2018;12: e0006453. pmid:30296295
  38. 38. Mollalo A, Sadeghian A, Israel GD, Rashidi P, Sofizadeh A, Glass GE. Machine learning approaches in GIS-based ecological modeling of the sand fly Phlebotomus papatasi, a vector of zoonotic cutaneous leishmaniasis in Golestan province, Iran. Acta Trop. 2018;188: 187–194. pmid:30201488
  39. 39. Ding F, Wang Q, Fu J, Chen S, Hao M, Ma T, et al. Risk factors and predicted distribution of visceral leishmaniasis in the Xinjiang Uygur Autonomous Region, China, 2005–2015. Parasites and Vectors. 2019;12: 1–10. pmid:30606222
  40. 40. Bi K, Chen Y, Zhao S, Kuang Y, John Wu CH. Current Visceral Leishmaniasis Research: A Research Review to Inspire Future Study. Biomed Res Int. 2018;2018. pmid:30105272
  41. 41. Von Zuben APB, Donalísio MR. Dificuldades na execução das diretrizes do Programa de Vigilância e Controle da Leishmaniose Visceral em grandes municípios brasileiros. Cad Saude Publica. 2016;32: 1–11. pmid:27333135
  42. 42. Dantas-Torres F, Miró G, Baneth G, Bourdeau P, Breitschwerdt E, Capelli G, et al. Canine leishmaniasis control in the context of one health. Emerg Infect Dis. 2019;25: E1–E4. pmid:31742505
  43. 43. Brasil. Manual de Vigilância e Controle da Leishmaniose Visceral—Ministério da Saúde. Brasília; 2006. Available:
  44. 44. Nunes CM, Lima VMF de, Paula HB de, Perri SHV, Andrade AM de, Dias FEF, et al. Dog culling and replacement in an area endemic for visceral leishmaniasis in Brazil. Vet Parasitol. 2008;153: 19–23. pmid:18314275
  45. 45. Costa DNCC, Codeço CT, Silva MA, Werneck GL. Culling Dogs in Scenarios of Imperfect Control: Realistic Impact on the Prevalence of Canine Visceral Leishmaniasis. PLoS Negl Trop Dis. 2013;7. pmid:23951375
  46. 46. Wilke VML. Avaliação das atividades de controle da Leishmaniose visceral canina no município de Sabará, Minas Gerais, 1995 a 2000. Universidade Federal de Minas Gerais. 2005.
  47. 47. Machado CJS, Silva EG, Vilani RM. Use of an instrument of controversial public health policy: Euthanasia of dogs contaminated by leishmaniasis in Brazil. Saude e Soc. 2016;25: 247–258.
  48. 48. Donato LE, Lima Júnior FEF de, Alburquerque R, Gomes MLS. Vigilância e controle de reservatórios da leishmaniose visceral no Brasil: aspectos técnicos e jurídicos. Rev Educ Contin em Med Veterinária e Zootec do CRMV-SP. 2013;11: 18–23.
  49. 49. Palatnik-de-Sousa CB, Day MJ. One Health: The global challenge of epidemic and endemic leishmaniasis. Parasit Vectors. 2011;4. pmid:21214898
  50. 50. Tolezano JE, Matsumoto PSS, Taniguchi HH, Bertollo DMB, Pierre MK, Barbosa JEDR, et al. Evaluation of the effectiveness of using deltamethrin- impregnated collars to control visceral leishmaniasis in the municipality of Votuporanga, State of São Paulo, Brazil, 2014–2016. Rev Inst Adolfo Lutz. 2018; 1–10.
  51. 51. Dantas-Torres F, Nogueira F dos S, Menz I, Tabanez P, da Silva SM, Ribeiro VM, et al. Vaccination against canine leishmaniasis in Brazil. Int J Parasitol. 2020;50: 171–176. pmid:32126240
  52. 52. Gálvez R, Montoya A, Fontal F, Martínez De Murguía L, Miró G. Controlling phlebotomine sand flies to prevent canine Leishmania infantum infection: A case of knowing your enemy. Res Vet Sci. 2018;121: 94–103. pmid:30366124
  53. 53. Mazloumi Gavgani AS, Hodjati MH, Mohite H, Davies CR. Effect of insecticide-impregnated dog collars on incidence of zoonotic visceral leishmaniasis in Iranian children: A matched-cluster randomised trial. Lancet. 2002;360: 374–379. pmid:12241778
  54. 54. Yimam Y, Mohebali M. Effectiveness of insecticide-impregnated dog collars in reducing incidence rate of canine visceral leishmaniasis: A systematic review and meta-analysis. PLoS One. 2020;15: 1–15. pmid:32881961
  55. 55. Coura-Vital W, Gomes de Almeida Leal G, Marques LA, Da Costa Pinheiro A, Carneiro M, Reis AB. Effectiveness of deltamethrin-impregnated dog collars on the incidence of canine infection by Leishmania infantum: A large scale intervention study in an endemic area in Brazil. PLoS One. 2018;13: 1–17. pmid:30532255
  56. 56. De Souza NA, Brazil RP, Araki AS. The current status of the Lutzomyia longipalpis (Diptera: Psychodidae: Phlebotominae) species complex. Mem Inst Oswaldo Cruz. 2017;112: 161–174. pmid:28225906
  57. 57. Casanova C, Colla-Jacques FE, Hamilton JGC, Brazil RP, Shaw JJ. Distribution of Lutzomyia longipalpis chemotype populations in São Paulo state, Brazil. PLoS Negl Trop Dis. 2015;9: e0003620. pmid:25781320
  58. 58. Motoie G, Ferreira GEM, Cupolillo E, Canavez F, Pereira-Chioccola VL. Spatial distribution and population genetics of Leishmania infantum genotypes in São Paulo State, Brazil, employing multilocus microsatellite typing directly in dog infected tissues. Infect Genet Evol. 2013;18: 48–59. pmid:23665466
  59. 59. Schwabl P, Boité MC, Bussotti G, Jacobs A, Andersson B, Moreira O, et al. Colonization and genetic diversification processes of Leishmania infantum in the Americas. Commun Biol. 2021;4: 1–13. pmid:33398033