Predictive factors for Leishmania infantum infection in dogs examined at a veterinary teaching hospital in Teresina, State of Piauí, Brazil

Abstract:

INTRODUCTION:

In Brazil, culling of seropositive dogs is one of the recommended strategies to control visceral leishmaniasis. Since infectiousness is correlated with clinical signs, control measures targeting symptomatic dogs could be more effective.

METHODS:

A cross-sectional study was carried out among 1,410 dogs, predictive models were developed based on clinical signs and an indirect immunofluorescence antibody test.

RESULTS:

The validated predictive model showed sensitivity and specificity of 86.5% and 70.0%, respectively.

CONCLUSIONS:

Predictive models could be used as tools to aid control programs in focusing on a smaller fraction of dogs contributing more to infection dissemination.

Keywords:
Canine; Control measures; Visceral leishmaniasis

Visceral leishmaniasis (VL) is a severe disease affecting thousands of people worldwide, mainly those living in poverty in developing countries ¹. In Brazil, the disease is caused by the protozoan parasite Leishmania infantum(syn = Leishmania chagasi), transmitted by the bites of female sandflies from the genus Lutzomyia, and dogs are considered the main source of infection in urban areas.

Infected dogs may not exhibit clinical signs, but many will develop signs of the disease during the course of infection ². Both clinically and subclinically affected dogs may transmit Leishmaniato sandflies, but clinically affected dogs transmit the disease much more efficiently ^{2) (3}.

As dog infectiousness is correlated with clinical signs ³, the development of a predictive system for canine visceral leishmaniasis (CVL) combining serological results with clinical information might be an alternative for improving the effectiveness of VL control strategies.

We report a cross-sectional study carried out among 1,410 dogs (from 1 month to 13 years of age) from January 2003 to December 2004. These animals, brought by their owners, were examined by veterinarians at the university veterinary hospital of the Federal University of Piauí (FUFPI) and sent to the Animal Health Laboratory (FUFPI) for serological and parasitological testing.

A diagnosis of CVL (the primary outcome measure) was confirmed when Leishmaniaspp. amastigotes could be detected in bone marrow aspirates, lymph node aspirates, or skin samples. Bone marrow biopsy was performed on all dogs > 3 months of age, while lymph node smears were obtained only from those with lymphadenopathy, and skin samples were obtained from those showing skin lesions.

Serological testing was performed by means of an indirect immunofluorescence antibody test (IFAT) using a canine leishmaniasis kit supplied by Bio-Manguinhos, FIOCRUZ, Rio de Janeiro, Brazil. Five veterinarians were responsible for clinical evaluations of the dogs upon arrival at the hospital, and the findings were recorded on a standardized clinical evaluation form.

The predictive variables considered were based on clinical signs of the disease (presence or absence), sex (male/female), age (< 1, 1, 2, 3, 4-5, and > 5 years; used as an ordinal variable in the analyses) and IFAT results. We used a split-sample validation approach in which the original dataset was divided into two groups: the test sample(dogs evaluated in 2003, n = 713) to develop the models, and the validation sample(dogs evaluated in 2004, n = 697) for model validation ⁴.

To develop predictive models, we initially employed simple logistic regression models to estimate odds ratios (ORs) and respective 95% confidence intervals. Variables showing a univariate association with CVL at a p-value  0.20 were selected for the multivariate analyses. Only variables with an association with a p-value  0.05 remained in the final model. Three predictive models were developed: one with only demographic and clinical variables (the clinical model), one based only on IFAT results (the IFAT model), and one with both IFAT results and demographic and clinical variables (the clinical + IFAT model).

A predictive model based on a scoring system, with points allocated to each predictive variable, was created from each of the three final regression models run on the test sample ⁵. A final score for each model was obtained through the sum of the points attributed to the presence of each predictive variable that remained in the final model and from the results of the IFAT. These models were then validated using the validation sample.

The area under the receiver operating characteristic curve (auROC), sensitivity, and specificity were used to evaluate the performances of the models. Sensitivity and specificity were obtained considering a predicted probability of CVL of 40% (> 40% indicating the presence of CVL), a value close to the prevalence of CVL in both the test and validation samples (38.2% in 2003 and 43.6% in 2004). The calibration of the model with the higher discriminatory ability was evaluated using the Hosmer-Lemeshow statistic ⁶. Statistical analyses were performed using Stata 9.2 (STATA Corp., College Station, TX, USA).

From all predictive demographic and clinical variables selected for multivariate analyses, only variables listed in Table 1and Table 2remained significantly associated with CVL in the final clinical model.

Table 1:
Odds ratios and 95% confidence intervals for CVL for variables retained in the final models generated from the test sample.

• CVL: canine visceral leishmaniasis; IFAT: immunofluorescence antibody test; OR: odds ratios adjusted for other variables in the table; CI: confidence interval; #p-value for the trend test.

Table 2:
Odds ratios and 95% confidence intervals for CVL for variables retained in the final models generated from the test sample.

• CVL: canine visceral leishmaniasis; IFAT: immunofluorescence antibody test; OR: odds ratios adjusted for other variables in the table; CI: confidence interval.

Table 3shows the scores for each clinical variable and IFAT result obtained from models developed from the test sample. Considering a predicted probability of CVL of 40%, the cutoff for the scoring system indicating CVL was ≥ 2 for the clinical model, and ≥ 9 for the clinical + IFAT model. For the IFAT model, the cutoff was ≥ 2, corresponding to an IFAT reaction ≥ 1:80.

Table 3:
Scores attributed to categories of predictive variables associated with CVL obtained from the test sample.

• CVL: canine visceral leishmaniasis; IFAT: immunofluorescence antibody test.

The scoring system was then applied to the validation sample to evaluate its performance, and the following results were obtained for the clinical model, IFAT model, and IFAT + clinical model: 1) sensitivity: 75.3%, 87.8%, and 86.5%; and 2) specificity: 65.9%, 63.6%, and 70%, respectively. The auROC curve of the clinical + IFAT model (87.9%) was significantly higher than that of the IFAT model (84.2%) (p = 0.001). Model calibration indicated no statistically significant difference between the predicted and observed outcomes for the clinical + IFAT model in the validation sample (p = 0.313).

Since both clinical status and levels of antibodies are associated with infectiousness ^{2) (3}, the use of such a model would be in line with the idea that focused interventions oriented towards dogs that contribute more to transmission tend to be more efficient ^{7) (8}. Although removing only highly infectious dogs might not be sufficient to interrupt transmission, since asymptomatic dogs can also transmit Leishmania, it has been suggested that interventions targeting subgroups that contribute most to transmission will have a higher impact than non-targeted interventions ⁷⁾.

A critical point to be highlighted is that a predictive model using IFAT might not be useful in practice, since the Brazilian Ministry of Health changed the serological test used for detecting seropositive dogs in the field. However, one may consider this an initial effort that demonstrates the potential of a control strategy focusing on more highly infectious dogs. As a matter of fact, the use of IFAT as a serological test does not invalidate the study, and actually indicates that the model may have even better performance if used with other more accurate serological tests. Therefore, the present study, being the first to use this approach for visceral leishmaniasis, contributes new ideas and opportunities for research focusing on the development of a more rational control strategy for VL in Brazil.

Other points regarding the importance of an approach focusing on clinical signs combined with serological tests should be emphasized. First, if this model was applied in practice, a dog would only be culled if it scored ≥ 9 points, which is a score difficult to obtain only on the basis of an IFAT result, unless there is IFAT positivity at a 1:160 dilution, which is much more specific than the cutoff commonly used in Brazil (1:40). This indicates the importance of clinical signs in association with serological tests to optimize disease control. Second, selecting only a fraction of seropositive dogs to cull from the environment would help increase the ability of health services to sustain control actions. For instance, using the common IFAT cutoff of 1:40, 499 dogs from the validation sample would be culled. However, in the clinical + IFAT model, only 380 would be culled, resulting in a relative decrease of 23.8% in the total population of potentially culled dogs. Finally, applying such a model in canine surveys could be useful, even if dog culling is abandoned as a public health intervention. Detecting the most infectious dogs could be used as a marker of a higher risk of transmission to humans, and could contribute to identifying areas of priority to which other types of interventions should be allocated. All of these considerations would apply broadly to any predictive model based on a combination of clinical signs and any available serological test.

A particular limitation of the study is the distinctive feature of the sample (dogs examined at a veterinary hospital), which presented a higher prevalence of canine infection than most settings in which VL transmission occur. Therefore, one would expect lower positive predictive values if any of the models was applied in field conditions.

The use of parasitological tests as a reference standard for CVL diagnosis is another potential limitation of this study. The well-known imperfect sensitivity of parasitological tests, and in particular, for asymptomatic dogs, leads to many truly infected dogs being classified as not infected (false negatives) ^{9) (10}. This might have contributed to underestimations of their specificities. Not considering the severity of clinical signs, but only their presence or absence, is an additional limitation, since the disease severity is directly correlated with infectiousness ¹¹.

Chronic infections, such as those caused by Leishmaniaspp., Trypanosoma cruzi, Ehrlichia canis, and Babesia canis, may coexist in the same dog, and cross-reactions cannot be excluded as a source of error in this study ¹². Although serological cross-reactivity with Ehrlichia canisand Babesia canismight not be common ^{13) (14}, the burden of ehrlichiosis and babesiosis in this population should be investigated, since the clinical presentations of both diseases might be similar to that of CVL, which would affect the specificity of the model using clinical signs in particular. The use of serological testing based on recombinant antigens could be a solution to reduce the threat of serological cross-reactions ^{15) (12}.

Considering all the limitations described, one may view the predictive models developed and validated in our study more as examples of how to approach the problem by focusing on clinically affected animals to help improve the efficiency of VL control programs rather than to properly define a method that could be used promptly in different settings. Clinical signs together with the results of a serological test could be used to select a smaller fraction of the dog population to be culled from the environment, or alternatively, to be used as markers of high transmission areas when prioritizing interventions.

Ethical considerations

This study is part of a research proposal approved by the Ethics Committee for Research with Animals / Federal University of Piauí (026/2008).

• ¹
  Alvar J, Vélez ID, Bern C, Herrero M, Desjeux P, Cano J et al. Leishmaniasis worldwide and global estimates of its incidence. PLoS One 2012; 7: e35671.
• ²
  Courtenay O, Quinnell RJ, Garcez LM, Shaw JJ, Dye C. Infectiousness in a cohort of brazilian dogs: why culling fails to control visceral leishmaniasis in areas of high transmission. J Infect Dis 2002; 186: 1314-1320.
• ³
  Quinnell RJ, Courtenay O. Transmission, reservoir hosts and control of zoonotic visceral leishmaniasis. Parasitology 2009; 136:1915-1934.
• ⁴
  Bouwmeester W, Zuithoff NP, Mallett S, Geerlings MI, Vergouwe Y, Steyerberg EW et al. Reporting and methods in clinical prediction research: a systematic review. PLoS Med 2012; 9:1-12.
• ⁵
  Barquet N, Domingo P, Cayla JA, González J, Rodrigo C, Fernandéz-Villadrich P, et al. Prognostic Factors in Meningococcal Disease. Development of a Bedside Predictive Model and Scoring System. JAMA 1997; 278:491-496.
• ⁶
  Hosmer DW, Lemeshow S. Applied logistic regression. 2nd ed. New York: John Wiley & Sons Inc; 2000.
• ⁷
  Koopman JS, Simon CP, Riolo CP. When to control endemic infections by focusing on high-risk groups. Epidemiology 2005; 16: 621-627.
• ⁸
  Woolhouse ME, Dye C, Etard JF, Smith T, Charlwood JD, Garnett GP et al. Heterogeneities in the transmission of infectious agents: implications for the design of control programs. Proc Natl Acad Sci USA. 1997; 94:338-342.
• ⁹
  Reis AB, Martins-Filho OA, Carvalho AT, Carvalho MG, Mayrink W, França-Silva JC, et al. Parasite density and impaired biochemical/hematological status are associated with severe clinical aspects of canine visceral leishmaniasis. Res Vet Sci 2006; 81:68-75.
• ¹⁰
  Saridomichelakis MN, Mylonakis ME, Leontides LS, Koutinas AF, Billinis C, Kontos VI. Evaluation of lymph node and bone marrow cytology in the diagnosis of canine leishmaniasis ( Leishmania infantum) in symptomatic and asymptomatic dogs. Am J Trop Med Hyg 2005; 73:82-86.
• ¹¹
  Travi BL, Tabares CJ, Cadena H, Ferro C, Osorio Y. Canine visceral leishmaniasis in Colombia: relationship between clinical and parasitologic status and infectivity for sand flies. Am J Trop Med Hyg2001; 64:119-124.
• ¹²
  Rosário EY, Genaro O, França-Silva JC, Costa RT, Mayrink W, Reis AB, et al. Evaluation of enzyme-linked immunosorbent assay using crude Leishmaniaand recombinant antigens as a diagnostic marker for canine visceral leishmaniasis. Mem Inst Oswaldo Cruz 2005; 100:197-203.
• ¹³
  Guillén+Llera JL, López Garcia ML, Martin Reinoso E, De Vivar González R. Differential serological testing by simultaneous indirect immunofluorescent antibody test in canine leishmaniosis and ehrlichiosis. Vet Parasitol 2002; 109:185-190.
• ¹⁴
  Oliveira TMFS, Furuta P, Carvalho D, Machado RZ. A study of cross-reactivity in serum samples from dogs positive for Leishmaniasp., Babesia canisand Ehrlichia canisin enzyme-linked immunosorbent assay and indirect fluorescent antibody test. Rev Bras Parasitol Vet 2008; 17:7-11.
• ¹⁵
  Boarino A, Scalone A, Gradoni L, Ferroglio E, Vitale F, Zanatta R, et al. Development of recombinant chimeric antigen expressing immunodominant B epitopes of Leishmania infantumfor serodiagnosis of visceral leishmaniasis. Clin Diagn Lab Immun 2005; 12:647-653.

• Partially funded by CNPq (#306267/2010-1; #483378/2010-0; #311507/2014-0).

Publication Dates

• Publication in this collection
  Jan-Feb 2016

History

• Received
  15 June 2015
• Accepted
  16 Sept 2015