- Record: found
- Abstract: found
- Article: found
Distribution of Lutzomyia longipalpis Chemotype Populations in São Paulo State, Brazil
research-article
Author(s):
Claudio Casanova
1
,
* ,
Fernanda E. Colla-Jacques
2 ,
James G. C. Hamilton
3 ,
Reginaldo P. Brazil
4 ,
Jeffrey J. Shaw
5
Publication date (Electronic):
17 March 2015
Journal:
PLoS Neglected Tropical Diseases
Publisher:
Public Library of Science
Read this article at
There is no author summary for this article yet. Authors can add summaries to their articles on ScienceOpen to make them more accessible to a non-specialist audience.
Abstract
Background
American visceral leishmaniasis (AVL) is an emerging disease in the state of São Paulo, Brazil. Its geographical expansion and the increase in the number of human cases has been linked to dispersion of Lutzomyia longipalpis into urban areas. To produce more accurate risk maps we investigated the geographic distribution and routes of expansion of the disease as well as chemotype populations of the vector.
Methodology/Principal Findings
A database, containing the annual records of municipalities which had notified human and canine AVL cases as well as the presence of the vector, was compiled. The chemotypes of L. longipalpis populations from municipalities in different regions of São Paulo State were determined by Coupled Gas Chromatography – Mass Spectrometry. From 1997 to June 2014, L. longipalpis has been reported in 166 municipalities, 148 of them in the Western region. A total of 106 municipalities were identified with transmission and 99 were located in the Western region, where all 2,204 autochthonous human cases occurred. Both the vector and the occurrence of human cases have expanded in a South-easterly direction, from the Western to central region, and from there, a further expansion to the North and the South. The ( S)-9-methylgermacrene-B population of L. longipalpis is widely distributed in the Western region and the cembrene-1 population is restricted to the Eastern region.
Conclusion/Significance
The maps in the present study show that there are two distinct epidemiological patterns of AVL in São Paulo State and that the expansion of human and canine AVL cases through the Western region has followed the same dispersion route of only one of the two species of the L. longipalpis complex, ( S)-9-methylgermacrene-B. Entomological vigilance based on the routes of dispersion and identification of the chemotype population could be used to identify at-risk areas and consequently define the priorities for control measures.
Author Summary
Information on the geographical distribution, dispersal mechanisms and dispersion route of insect-borne diseases can help to identify ongoing transmission areas, new risk areas and guide surveillance and control activities. Lutzomyia longipalpis, the principal vector of American visceral leishmaniasis disease in the Americas, is a group of closely related species that can be separated according to the type of pheromone produced by male individuals. It is still unclear how many members there are in this complex, how they are related and if some are more important vectors than others. In the present study, we show that the expansion of human visceral leishmaniasis in the state of São Paulo, Brazil, has followed the same dispersion route of only one of the two related species of L. longipalpis found in the state. The maps that we produced in the present study will allow us to determine risk areas for the occurrence of human visceral leishmaniasis, and reinforces our hypothesis that within São Paulo state these chemotype populations could have different biological capacities to act as a vector.
Related collections
Most cited references25
- Record: found
- Abstract: found
- Article: found
Control of Visceral Leishmaniasis in Latin America—A Systematic Review
Gustavo A S Romero, Marleen Boelaert (2010)
Introduction Visceral leishmaniasis (VL) in Latin America is a severe systemic disease caused by an intracellular protozoon, Leishmania infantum (syn. L. chagasi). VL is a zoonosis: the domestic dog is the main animal reservoir, while foxes and other wild animals play a role in sylvatic transmission [1]–[5]. The parasite is transmitted by a night-biting sandfly, Lutzomyia longipalpis, a 2 to 3 mm-long insect well adapted to the peri-domestic environment and distributed throughout Latin America [6]–[10]. L. infantum is also transmitted by Lu. cruzi in Brazil [11] and Lu. evansi in Colombia, and Venezuela [12],[13]. Clinically, VL is characterized by prolonged fever, weight loss, hepatomegaly, splenomegaly, hypergammaglobulinemia and pancytopenia and it is usually fatal if not adequately treated [14]. Not all L. infantum infections lead to overt clinical disease: in Brazil ratios of 8–18 incident asymptomatic infections to 1 incident clinical case were described [15]–[17]. Risk factors for the development of clinical disease are only partially understood. Some studies suggest that the susceptibility to VL could be genetically determined [18]–[22]. Malnutrition places children at higher risk [23],[24]. Other studies identified being a young male and the presence of animals in the neighborhood [25], living in houses with a inadequate sewage system and waste collection [26], and residence in an urban slum or in areas with green vegetation [27] as risk factors. The VL disease burden in Latin America is not exactly known because most countries lack effective surveillance systems [28]–[30]. Brazil declared a total of 50,060 clinical VL cases between 1990 and 2006 and this number accounts for 90% of all reported VL cases in the Americas, but is subject to substantial underreporting [29],[31]. The country reported so far 176 HIV-coinfected VL cases [32] but has a significant number of asymptomatic co-infected individuals [33],[34]. Whereas VL was initially concentrated in the poor rural areas in the northeast of the country, since the 1980s epidemics have occurred in major cities such as Belo Horizonte, Campo Grande, Natal, and others [35]–[37]. Some of these urban VL outbreaks were attributed to the migration of families from the rural areas to the peri-urban slums after periods of prolonged drought. Whereas the reported VL incidence in the 1980s averaged at 1,500 cases per year, this figure increased to an average 3,362 per year between 2000 and 2006 [31]. The disease has gradually spread south and eastward and is reported since 1999 from the states of São Paulo and Mato Grosso do Sul [38]. Human VL cases have also been reported from Honduras [39], Venezuela [40], Paraguay [41] and Argentina [42]. Sporadic and/or import human or canine cases were described in Chile [43], Ecuador [44], Bolivia [45], Mexico [46], Costa Rica [47], and French Guyana [48]. A geographically referenced database providing links to published literature about the spatial distribution of VL can be accessed on http://apps.who.int/tools/geoserver/www/ecomp/index.html (Accessed on September 19 2009). Control of VL in the Americas has proved challenging. Early diagnosis and treatment is essential for the patient, but has limited impact on transmission if the main animal reservoir or insect vectors are not tackled [49]. Some studies showed a decreased incidence of VL in both dogs and children following serological screening and culling of seropositive dogs [50],[51], but this control strategy is increasingly debated [52]. Human VL incidence remained high in Brazil despite intensive application of this strategy in recent years [31]. Lack of impact has been attributed to the low sensitivity of the diagnostic tests, the long delay between diagnosis and culling and the low acceptance of culling by dog owners. Mathematical modeling suggests that vector control and vaccination of dogs would be more efficacious than dog culling [49]. Treatment of infected dogs is not an effective strategy as relapses are frequent, and dogs quickly become infectious again [53]. A controlled trial in a different setting of zoonotic VL (Iran) showed how the use of deltamethrin-treated dog collars reduced the risk of infection in dogs (by 54%) and in children (by 43%) [54]. Another controlled trial in Brazil showed only a modest effect on canine seroconversion rates [55] in spite of the proven effect of deltamethrin-impregnated dog collars on vector density [56]. In the Mediterranean region, where VL is also zoonotic with dogs playing a role as main reservoirs, human cases and canine cases are treated with antiparasitic drugs. In Europe, individual measures to protect dogs from sand fly bites using insecticides are common practices, but no public health surveillance and control interventions such as those applied in Brazil are in place [57]. Recently, the governments of India, Bangladesh and Nepal launched a VL elimination initiative, aiming to reduce the annual incidence of VL to less than 1/10,000 population by 2015 [58]. The strategy exploits recent technological developments in diagnosis, drugs and vector control [59]. Though the transmission pattern in this region is totally different, with L. donovani being the causative agent, a different sandfly vector (P. argentipes) and -most importantly- anthroponotic instead of zoonotic transmission, we wanted to examine whether there is a scope for VL elimination or at least improved control in the Americas. Given the heterogeneity in causative species, vector and transmission pattern, evidence on VL control tools from one region cannot be readily extrapolated to another. We report a review of the literature on the effectiveness of novel VL control tools and strategies in Latin-America structured around diagnosis of human and canine VL, treatment of human cases and control of the animal reservatoir and arthropod vectors. Methods The review on VL control interventions was structured around the following topics: (i) Diagnosis of human VL; (ii) Treatment of human VL; (iii) Diagnosis of canine VL; (iv) Control of the animal reservoir and vector. Box 1 shows the Medical Subject Heading (MeSH) terms and keywords used in the search per topic. We searched for English, Portuguese and Spanish–language articles in MEDLINE, LILACS, as well as the Cochrane Central Register of Controlled Trials from 1960 to November 2008. We considered only original research, mainly but not exclusively intervention trials, diagnostic accuracy studies and observational studies, with scope targeted to American VL. Additional articles were obtained through citation tracking of review and original articles. Box 1. Keywords and MESH Headings Used for Literature Searches Diagnosis of human VL: For the PubMed search: (visceral leishmaniasis OR kala-azar OR L.infantum OR L. chagasi OR L.donovani OR Leishmania infantum OR Leishmania chagasi OR Leishmania donovani ) AND (diagnostic accuracy OR diagnostic performance OR sensitivity OR specificity OR validation) AND “Americas” [MeSH]. For the LILACS search the key-words: leishmaniasis AND visceral AND (diagnosis OR DAT OR dipstick) were used. Treatment of human VL: For the PubMed search the following key-words were used: (visceral leishmaniasis OR kala azar OR L. chagasi OR L donovani ) AND (amphotericin b OR glucantime OR sodium stibogluconate OR miltefosine OR sitamaquine OR pentavalent antimonials OR paromomycin) AND “Americas” [MeSH]. For the LILACS search the key-words: leishmaniasis AND visceral AND treatment were used. For the Cochrane Central Register of Controlled Trials we used the term visceral leishmaniasis because the search with the key-words and MeSH terms used for the PubMed searching failed to retrieve any paper. Diagnosis of canine VL: For the PubMed search: (canine visceral leishmaniasis OR L.infantum OR L.chagasi OR L.donovani OR Leishmania infantum OR Leishmania chagasi OR Leishmania donovani ) AND (diagnostic accuracy OR diagnostic performance OR sensitivity OR specificity OR validation) AND “Americas” [MeSH]. For the LILACS search the key-words: canine AND leishmaniasis AND visceral AND diagnosis were used. Control of the animal reservoir and arthropod vector: for the PubMed search: (visceral leishmaniasis OR Leishmania chagasi OR L chagasi OR Kala-azar OR Leishmania infantum ) AND “Americas” [MeSH] AND control. The LILACS search was performed using the term visceral leishmaniasis OR leishmaniose visceral OR leishmaniasis visceral because of the failure to retrieve any paper when using the PubMed approach. In a next step, the titles, abstracts and if necessary the full text of the studies was examined to identify relevant papers for the review. Data were extracted by one researcher directly from the full length articles to structured tables containing all the descriptive variables and relevant outcomes. The inclusion criteria, data extracted for each item and summary measures are listed below. Human diagnosis As we have stated above, we set out to examine whether the existing control tools allow for elimination of VL in the Americas. The goal of elimination requires diagnostic and therapeutic tools that are very easy to use and can be easily decentralized. The World Health Organization now considers two ‘rapid diagnostic tests’ as appropriate for the diagnosis of VL in control programs: the Direct Agglutination Test (DAT) based on whole promastigotes of L. donovani or L. infantum and the rK39-ICT [60]–[62]. As it was not our intention to go into a full review of the available diagnostic tools for VL, we have excluded PCR and serological tests that require substantial laboratory equipment, even though there is extensive experience with the use of IFAT and ELISA tests in the Americas. Moreover, the clinical benefit of antigen-detection and PCR tests still needs to be demonstrated [63],[64]. We therefore limited our systematic review to DAT and rK39-ICT. The eligibility criteria included: original studies evaluating the DAT or the rK39 immunochromatographic test (ICT); clinical visceral leishmaniasis diseases in humans as target condition; adequate reference classification; absolute numbers of true-positive, true-negative, false-positive and false-negative observations available or derivable from the data presented. Accuracy measures were summarized as sensitivity and specificity. Human treatment Clinical trials including uncontrolled and retrospective studies with description of the following characteristics: intervention; case definition; follow-up schedule; therapeutic endpoints; control group; and efficacy measure defined through cure and failure proportions for each treatment. Canine diagnosis Original studies evaluating any diagnostic test for canine leishmaniasis; Leishmania infection and/or VL disease in domestic dogs as target condition; adequate reference classification; absolute numbers of true-positive, true-negative, false-positive and false-negative observations available or derivable from the data presented. Accuracy measures were summarized as sensitivity and specificity. Vector control and animal reservoir control Field trials of control measures (canine culling, impregnated dog collars, canine vaccination, insecticide spraying, insecticide treated bednets, environmental management) evaluating at least one control measure; description of the intervention under analysis; target population, sampling and randomization process; adequate case definitions for asymptomatic infection or VL; definition of outcomes related to humans, dogs or sand flies; at least one effect measure; and at least one point estimation for the magnitude of the expected effect. Results Human diagnosis A Medline search generated 77 papers, and LILACS 179. After screening the titles and abstracts of those papers for evaluations of the DAT or rK39 in human VL, we retrieved eight original papers (Figure 1 and Table 1). We report only descriptive statistics of sensitivity and specificity estimates; without drawing conclusions about differences in these parameters between tests and discuss them in comparison with results of a meta-analysis by Chappuis et al. [65]. 10.1371/journal.pntd.0000584.g001 Figure 1 Flow of inclusion of studies on human VL diagnosis. 10.1371/journal.pntd.0000584.t001 Table 1 Main characteristics of diagnostic accuracy studies reporting on tests for human visceral leishmaniasis in Latin America. Country Type of study Diagnostic test Reference test Number of confirmed VL Sensitivity Number of controls Specificity Ref. Brazil Phase-2 DAT Not described 33 94% 173 OD*** 100% [66] 178 HEC**** 100% Brazil Phase-2 DAT Parasitology or improving after antimonial treatment 16 100% 102 OD 100% [67] 105 HEC 100% Brazil/other Phase-2 FD-DAT* Parasitology 36 100% 42 OD 100% [68] 19 HEC 100% Venezuela Phase-2 FD-DAT Parasitology 30 100% 20 OD 100% [69] 19 HEC 100% Brazil Phase-2 FD-DAT Parasitology 88 96.6% 85 OD 97.6% [70] 20 HEC 100% Venezuela Phase-2 rK39 ICT Composite reference ** 41 87.8% 76 OD 100% [71] Brazil/other Phase-2 rK39 ICT Parasitology 36 85.7% 42 OD 80.9% [68] 19 HEC 84.2% Brazil Phase-2 rK39 ICT Parasitology 128 90% 50 OD 100% [72] 10 HEC 100% Brazil Phase-3 rK39 ICT Parasitology 213 93% 119 OD 97% [73] *FD-DAT: Freeze-dried DAT. **Composite reference: at least 2 positive tests out of 4 (bone marrow, IFAT, CIEP, Western blot). ***OD : patients with other, potentially cross-reacting infectious diseases. ****HEC: Healthy Endemic Controls. Phase 2: Case-Control design, laboratory based study on banked serum samples. Phase 3: Prospective clinical study, recruiting representative patients, all presenting with febrile splenomegaly. I: Direct agglutination test (DAT) for VL Andrade et al (1989) were the first to report a proof-of-principle evaluation of the DAT in Brazil [66]. A recent meta-analysis of the DAT performance showed sensitivity and specificity estimates of 94.8% (95%CI: 92.7–96.4) and 97.1% (95%CI: 93.9–98.7), respectively [65]. The performance of DAT was neither influenced by the region nor by the Leishmania species. However, this meta-analysis included only two studies from Latin –America, both from Brazil, and both with small sample sizes. Garcez et al (1996) reported 100% sensitivity on 16 parasitologically confirmed VL cases and 98.3% specificity on a mixed group of 65 healthy endemic controls and patients with other diseases [67]. Schallig et al (2002) reported 100% sensitivity on 21 confirmed VL cases and 100% specificity on 19 healthy controls and 42 samples of patients with other diseases [68]. More recently, Teran-Angel et al (2007) reported 100% sensitivity on 30 confirmed VL patients in Venezuela and 100% specificity on 39 controls [69]. Pedras et al (2008) compared the freeze-dried DAT (FD-DAT) and a locally produced DAT with 3 other serological tests (rK39 ELISA, ELISA-L. chagasi and IgG-IFAT) and concluded that the FD-DAT was the most efficient, with 96.6% sensitivity (n = 88) and 98.1% specificity (n = 105) [70]. All reported studies are laboratory-based, no large prospective clinical studies evaluating the DAT have been reported from the Americas. II: rK39-based immunochromatographic test (ICT) Delgado et al (2001) evaluated the rK39-ICT in Venezuela, reporting 87.8% sensitivity (36/41 confirmed VL) and a specificity of 100%. The lower sensitivity was attributed to the fact that the false negative sera had been kept at −70° for more than 10 years [71]. A meta-analysis of 13 studies of the rK3 ICT by Chappuis et al (2006) showed sensitivity and specificity estimates of 93.9% (95%CI: 87.7–97.1) and 95.3% (95%CI: 88.8–98.1), respectively, with some regional variation [65]. This meta-analysis included only two studies from Latin-America [68],[72]. De Assis et al (2008) confirmed the excellent diagnostic performance of rK39-ICT in a prospective study in Brazil, with 93% sensitivity on 213 confirmed VL cases and 97% specificity on 119 controls with clinical suspicion of VL but with confirmation of other diseases [73]. On this basis, it seems that rK39 based diagnosis can be adopted in clinical practice, though each new brand put on the market should be evaluated in proper phase-3 designs. III: Key questions for control What should be the diagnostic algorithms for VL for use in primary health care and in active case detection campaigns? How to assure the quality of available VL rapid diagnostic tests? How to define asymptomatic infected individuals (and how to manage them?) How to improve clinician's awareness about the possibility of Leishmania co-infection in HIV/AIDS cases? IV: Questions for research What can be the contribution of novel (molecular) parasite detection tests to clinical diagnosis? What is the performance of diagnostic assays in HIV-Leishmania co-infections? What is the performance of antibody-assays in patients from areas with sympatric circulation of parasites causing cutaneous leishmaniasis? Canine diagnosis Seventy-seven papers were retrieved from Medline/PubMed search and 11 of them were considered relevant. The LILACS database search retrieved 26 papers of which 2 were considered relevant, but 1 was already obtained from the PubMed database (Figure 2). Finally, 12 papers were included in the review, covering 5 serological tests for canine VL: IFAT, ELISA, dot-ELISA, DAT, and rK39-ICT [66], [69], [74]–[83]. IFAT has been the test adopted by the Brazilian Ministry of Health for its dog screening-and-culling campaigns. Published estimates for sensitivity range from 72–100%, for specificity 52–100% (Table 2). The moderate sensitivity and specificity of this test, the long turn-around time between sample taking and culling, and the complexity of its execution have been invoked as one of the reasons for the low effectiveness of the culling campaign. Several ELISA tests have been evaluated, with assays based on homologous antigens usually showing higher sensitivity. Evans et al (1999) showed a higher sensitivity of ELISA compared to IFAT and pleaded for a revision of the screening policy [84]. 10.1371/journal.pntd.0000584.g002 Figure 2 Flow of inclusion of studies on canine VL diagnosis. 10.1371/journal.pntd.0000584.t002 Table 2 Main characteristics of diagnostic tests for canine visceral leishmaniasis in Latin America. Country Diagnostic test Reference test Number of confirmed VL Sensitivity Number and type of controls Specificity Ref. Brazil IFAT Parasitology 46 78% 102 NEC 100% [74] Brazil IFAT Parasitology 21 100% 14 NEC 100% [66] Brazil IFAT Parasitology 112 72% 20 NEC 100% [75] 20 OD 52% Brazil IFAT CRS 36 100% 67 EC 66% [76] Brazil cELISA Parasitology 46 98% 102 NEC 99% [74] Brazil cELISA Parasitology 21 71% 14 NEC 86% [66] Brazil cELISA Parasitology 106 98–100 25 HEC 100% [77] Brazil cELISA Parasitology 112 95% 20 NEC 100% [75] 20 OD 64% Brazil cELISA Parasitology 76 95% 33 NEC 100% [78] Brazil cELISA Parasitology 50 symptomatics 88% 25 NEC 100% [79] 50 asymptomatics 30% 14 OD 64% Brazil rK39 ELISA Parasitology 106 98.1% 25 HEC 100% [77] Brazil rK39ELISA Parasitology 50 symptomatics 100% 25 NEC 100% [79] 50 asymptomatics 66% 14 OD 71% Brazil rK26 ELISA Parasitology 106 99.1% 25 HEC 100% [77] Brazil rK26ELISA Parasitology 50 symptomatics 94% 25 NEC 100% [79] 50 asymptomatics 66% 14 OD 57% Brazil rA2ELISA Parasitology 50 symptomatics 70% 25 NEC 100% [79] 50 asymptomatics 88% 14 OD 93% Brazil Dot-ELISA Parasitology 37 97% 63 HEC 100% [80] 30 NEC 100% Brazil DAT Parasitology 21 71% 14 NEC 71% [66] Brazil DAT Parasitology 112 93% 20 NEC 100% [75] 20 OD 95% Brazil FD-DAT CRS* 36 100% 67 EC 89.5% [76] Venezuela FD-DAT Parasitology 26 85% 16 HEC 100% [69] Brazil rK39 ICT CRS** 74 72 101 HEC 61% [81] Brazil rK39 ICT (6 formats) Clinical + IFAT 50 84–96% 50 HEC 100% [82] 14 OD 100% Brazil rK39 ICT Parasitology 76 83% 33 NEC 100% [78] 25 OD 84% DAT : variable cut-offs were used, and different antigens, see original papers. cELISA: ELISA based on crude soluble antigen; rELISA: ELISA based on recombinant antigens; FD-DAT: Freeze-dried DAT. *CRS: Composite Reference Standard: positive if direct microscopy or culture or PCR positive. **CRS: Composite Reference Standard:Positive if ELISA or PCR positive. NEC: healthy dogs from non-endemic areas. OD: dogs with other, potentially cross-reacting infectious diseases. HEC: healthy dogs from endemic areas. Recently more “user-friendly” diagnostics as the DAT and a canine version of the rK39-ICT were evaluated with good results. For the freeze-dried DAT sensitivity ranged from 85–100%, specificity 89–100% [65],[76],[78] and for the rk39-ICT sensitivity ranged from 72–96%, specificity 62–100% [81],[82]. The main advantage of these rapid tests would be to shorten the delay between diagnosis and culling/treatment. However, the reported estimates of sensitivity in the above studies depend on the type of dogs included in the “true cases” group with higher sensitivity observed in symptomatic than in asymptomatically dogs, and unfortunately, several evaluations failed to include an adequate sample of asymptomatically infected dogs. The sensitivity of the test in asymptomatic dogs is crucial for a control strategy, as those dogs are infectious, and should be targeted by the campaign. Sensitive antigen detection tests as PCR might become a relevant marker of infection in the future with the advantage that they can still be used in vaccinated dogs that will be serologically positive because of the vaccine. However, Quinnell et al (2001) showed in a longitudinal study of naturally infected dogs how the sensitivity of PCR was high early after infection but declined to 50% thereafter. The sensitivity of serology also varied with time, being lowest at the time of infection but clearly superior thereafter (93–100%). They concluded that PCR was most useful for detection of active disease, and considered serology as more adequate for the detection of infection [84]. I: Key questions for control What is the most cost-effective diagnostic strategy for a screen-and-treat or screen-and-cull campaign? Novel screening strategies based on combined, parallel or sequential use of current available tests need to be validated. II: Questions for research How to distinguish an antibody response due to natural infection from that produced after vaccination in dogs? What can be the contribution of novel, molecular, parasite detection tests to clinical diagnosis in dogs? What is the value of the current diagnostic tests in terms of dog infectivity for sandflies? Human treatment Thirty-nine papers were retrieved from Medline/PubMed search and four of them were considered relevant. The LILACS database search retrieved 42 papers of which 24 were not available from the PubMed database. Three of those 24 studies were considered relevant, one of them, was previously identified through the PubMed search. One paper was identified through specific author's name searching in PubMed. The Cochrane Central Register of Controlled Trials search retrieved 103 trials, three of them were conducted in the Americas but all were also identified through the PubMed and LILACS searches. Finally, seven papers were included for review [85]–[91]. Three papers were excluded from further analysis, one because it was a second publication on the same trial [88], one for being a retrospective study with heterogeneous therapeutic interventions with meglumine antimoniate and case definition based on clinical findings plus positive serology without description of the methods and test cut-off. A minority of cases was diagnosed through parasite identification [85], and one paper because it was a case-control study focusing on prognostic factors [87]. The flow for the selection and a summary of the reviewed studies appears in Figure 3 and Table 3. 10.1371/journal.pntd.0000584.g003 Figure 3 Flow of inclusion of studies on VL treatment. 10.1371/journal.pntd.0000584.t003 Table 3 Main characteristics of selected studies reporting treatment of human visceral leishmaniasis in Latin America. Country Type of study Number of subjects Mean patient age (years) Treatment interventions Dose and route Follow-up period Outcomes (%) Ref. Brazil Open-label 10 20.0 Amphotericin B cholesterol dispersion 2.0mg/kg/d for 10 d. I.V. 6–12 months Cure 10/10 (100) [90] Brazil Open-label 10 19.0 Amphotericin B cholesterol dispersion 2.0mg/kg/d for 7 d. I.V. 6–12 months Cure 10/10 (100) [90] Brazil Open-label 10 16.5 Amphotericin B cholesterol dispersion 2.0mg/kg/d for 5 d. I.V. 12 months Cure 9/10 (90) [91] Relapse 1/10 (10) Brazil Open-label Phase II 13 7.6 Liposomal amphotericin B 14mg/kg (total) . I.V. 6 months Cure 8/13 (61) [86] Failure 1/13 (8) Relapse 4/13 (31) Brazil Open-label Phase II 4 7.5 Liposomal amphotericin B 10mg/kg (total) I.V. 6 months Cure 4/4 (100) [86] Brazil Open-label Phase II 15 10.1 Liposomal amphotericin B 20mg/kg (total) I.V. 6 months Cure 13/15 (87) [86] Relapse 2/15 (13) Brazil Open-label, dose-escalating trial 4 19.0 WR6026 (sitamaquine) 1.0mg/kg/d for 28 d. Oral. 12 months Cure 0/4 (0) [89] Brazil Open-label, dose-escalating trial 6 32.8 WR6026 (sitamaquine) 1.5mg/kg/d for 28 d. Oral 12 months Cure 1/6 (17) [89] Brazil Open-label, dose-escalating trial 6 23.8 WR6026 (sitamaquine) 2.0mg/kg/d for 28 d. Oral. 12 months Cure 4/6 (67) [89] Brazil Open-label, dose-escalating trial 5 23.8 WR6026 (sitamaquine) 2.5mg/kg/d for 28 d. Oral 12 months Cure 1/5 (20) [89] Brazil Open-label, dose-escalating trial 1 22.0 WR6026 (sitamaquine) 3.25mg/kg/d for 28 d. Oral 12 months Cure 0/1 (0) [89] Dietze et al (1993) reported an open-label dose-escalating trial with amphotericin B colloidal dispersion (Amphocil) in two small groups of patients who showed similar cure rate suggesting that the 7 days was as effective as the 10 days regimen [90]. In 1995 the same authors reported another open-label trial with Amphocil with a shorter regime of 5 days, observing an episode of relapse [91]. Berman et al (1998) reported the results of an open-label phase II trial with three therapeutic regimens consisting of liposomal amphotericin B 10, 14 or 20 mg/kg total dose; the reported outcomes were cure, failure and relapse and the follow-up period was of six months. This paper suggested that the lower 10mg/kg total dose was less efficacious than the higher 20mg/kg total dose [86]. Dietze et al (2001) concluded from an open-label dose-escalating safety and efficacy trial that sitamaquine was not efficacious for the treatment of VL in young adults. Severe adverse events described as renal toxicity lead to trial interruption when using the higher dose of 3.25mg/kg/d [89]. I: Key questions for control What is the current standard of care for VL treatment in the Americas? What is the case for combination therapy for VL in the Americas? What is the standard of care in VL/HIV co-infection? II: Questions for research What is the current efficacy of pentavalent antimonials, amphotericin B deoxycholate and the liposomal formulations, miltefosine and drug combinations for VL treatment in the Americas? Are there more efficacious, safer, and simpler therapeutic schemes for VL than the current ones? Can a clinical prognostic score for treatment failure be developed to identify those cases most in need for intensive care? What is the role of non-parasite targeted drugs such as immunomodulators, antibiotics and others in VL treatment? Vector and animal reservoir control Incidence and prevalence estimates of canine VL in the Americas have been reported from several foci [2], [40], [92]–[95], but the specific relationship between canine and human VL cases is not well understood. Transmission in the dog population is mainly due to infected sandfly bites but alternative routes have been proposed such as sexual transmission and other potential insect vectors [96]–[98]. The control of the animal reservoir is complex and frequently involves combined interventions. The Brazilian Control Program recommends a strategy based on canine culling and vector control with insecticide spraying. Insecticide-impregnated collars for dogs and canine vaccination are not currently recommended as public health control measures [99]. One-hundred seventy-two papers were retrieved from Medline/PubMed using the search strategy cited above. The LILACS search was performed using the term visceral leishmaniasis because no document was retrieved when using the PubMed approach. The LILACS search was less specific and 519 documents were retrieved; 514 documents comprised an extensive spectrum of research irrelevant for the purpose of this paper and four of the five relevant papers were already identified through the PubMed search. After reading the titles and the abstracts and hand searching reference lists for related papers, fourteen were selected for full text reading because the main subject was at least one intervention for control VL (Figure 4) [50], [55], [100]–[111]. 10.1371/journal.pntd.0000584.g004 Figure 4 Flow of inclusion of studies on VL control. Magalhães et al (1980) published a retrospective –non controlled- study on the impact of a combined intervention consisting of human VL case treatment, culling of seropositive dogs and insecticide spraying with DDT in 19 municipalities of the Rio Doce Valley, State of Minas Gerais, Brazil reporting the disappearance of human symptomatic cases after 15 years of application of this strategy [100]. Dietze et al (1997) reported a field trial of dog screening and culling, based on twice-yearly screening with DOT-ELISA. This trial was conducted in three rural valleys, State of Espirito Santo, Brazil, two benefiting from the intervention and one used as control. At 6-months there was a 16% reduction of seroconversion rate in dogs (36% in the intervention vs. 52% in the control group), but this difference was not significant [101]. Braga et al (1998) reported the comparison of two strategies of dog screening-and-culling: screening by ELISA was compared to IFAT as routinely recommended by the National Control Program. The main difference consisted in the lag times after blood sampling (7 days for ELISA vs. 80 days for IFAT). The trial was conducted in a rural area of Northeastern Brazil where 28 communities were systematically allocated to one of the two groups. In the ELISA arm, reduction of canine seroprevalence was higher, probably due faster dog removal plus higher sensitivity of the ELISA test [102]. Ashford et al (1998) reported a controlled intervention trial of seropositive dog removal in an endemic area of the State of Bahia, Brazil. The intervention area was subjected to screening with FAST-ELISA and removal of seropositive dogs, in the control area no intervention was carried out. A significant reduction of dog seroconversion rate in the intervention area as compared to control was observed, and a significantly lower number of VL cases reported to health facilities in the intervention area [50]. Paranhos-Silva et al (1998) report a follow- up study of several clusters of seronegative dogs in Jequié, State of Bahia, Brazil. The initial prevalence of infection among 1681 dogs was 23.5%. After serological screening every six months for 18 months and removal of the seroconverters, the annual incidence rate of infection was 6.55 cases/100dog-years. The migration of dogs between clusters was 2.3 cases/100 dog-years. This study is relevant because as highlights the challenges posed by dog migration for any control program dealing with the canine reservoir [103]. Da Silva et al (2000) reported a phase III vaccine field trial in seronegative dogs screened with IFAT and FML-ELISA and exposed to fucose-mannose-ligand vaccine in three subcutaneous doses at 21 day intervals. Control arm was treated with saline placebo. Endpoints were symptomatic VL or death, seroconversion rates in FML-ELISA and conversion of leishmanin skin test composed of crude L. donovani antigen. Follow-up evaluations were performed at 2, 7, 13 and 24 months. A significant difference in the three endpoints was observed during the trial. The overall efficacy to prevent symptomatic VL disease was 75% [104]. Giffoni et al (2002) reported the effect of application of a 65% permethrin spot-on formulation on canine VL infection and sandfly abundance. A decrease of canine VL prevalence was observed in the intervention area compared with increased prevalence in the control area. No effect was observed on sandfly population [105]. Feliciangeli et al (2003) described a controlled trial of pyrethroid (λ-cyhalothrin) indoor spraying every 5 month and organophosphate (fenitrothion) ultra-low volume spatial fogging around the houses twice a month for ten months in one intervention compared to one control area. The main vector captured was Lu. longipalpis. A significant decrease of sandfly abundance was observed, with a residual effect of indoor spraying of 3 months. Main limitation of this study was the specific construction style of the houses: completely cemented, plastered and oil-painted walls and zinc roofs, which lowers its external validity [106]. De Oliveira et al (2003) reported the evaluation of routine combined control measures of seropositive dog-culling and insecticide spraying during six years. The intensity of the application of control measures correlated with human VL incidence, the coverage of canine surveys, the number of canine surveyed and the number of buildings submitted to insecticide spraying [107]. Reithinger et al (2004) reported a controlled field trial to evaluate the effectiveness of insecticide impregnated collars to prevent infection detected through serological tests or DNA detection by PCR assay in one intervention compared to one control area. The authors failed to detect a significant difference between groups in the incidence of new infections but they demonstrated a significant reduction of antibody titers in the collar protected dogs. Mathematical modeling using the results obtained in this study suggests that dog collars would be a better alternative than dog culling [55]. Moreira et al (2004) reported the incidence rates of canine Leishmania infection in a cohort of dogs submitted to an optimized culling strategy consisting of: (i) ELISA screening of serum samples; (ii) shortening of the time interval from serodiagnosis to removal of dogs; (iii) screening a high proportion of the dog population. They demonstrated that the incidence of canine infection remained stable through 2.5 years of observation under this strategy but the study had no control arm for comparison. A high replacement rate by susceptible puppies and already infected dogs was observed [108]. Courtenay et al (2007) reported the barrier effect, the 24-h mortality rate and the human landing rates of Lu. longipalpis in households using deltamethrin-impregnated bednets compared others using untreated bed nets. The study described a 39% increase in barrier capacity of the impregnated bednets, 80% reduction in sandfly landing rates on humans and 98% increase in the 24-h sandfly mortality rates. The study was done under field conditions with a small number of observations during a very short period of exposure to the treated bednets (three days) and the residual effect was not measured. However this intervention should be explored further because it could bring an additional benefit in areas where malaria is also endemic [109]. Costa et al (2007) reported a randomized community intervention trial to compare the effect of four strategies on human VL, as follows: (i) spraying houses and animal pens with pyrethroid insecticide; (ii) spraying houses and eliminating seropositive dogs; (iii) combination of spraying houses and animal pens plus eliminating seropositive dogs; and (iv) spraying houses only as the reference comparator. The outcome was evaluated by measuring incidence of seroconversion in humans six months after the application of interventions. The results indicated a positive effect of canine removal on incidence of leishmanial infection in men but surprisingly, the combination of dog culling plus outdoor spraying of peridomestic animal shelters failed to demonstrate any effect. The relevance of this study is that it constitutes the first attempt to measure the effect of combined interventions on human VL incidence [110]. De Souza et al (2008) reported a randomized community intervention trial to compare the effect of (i) pyrethroid insecticide spraying; (ii) pyrethroid insecticide spraying plus culling of seropositive dogs with (iii) no intervention. The interventions were maintained for two years and outcomes were registered every year, insecticide spraying was performed every 6 months. Although a lower incidence was observed in the groups submitted to interventions and that reduction was more intense after two years, the study failed to detect statistically significant differences [111]. The summarized characteristics and main limitations of these studies are shown in Table 4. 10.1371/journal.pntd.0000584.t004 Table 4 Main characteristics of selected studies reporting effects of control interventions against visceral leishmaniasis in Latin America. Country and period Study setting Intervention Comparison Number of subjects in intervention and control arm Follow-up Outcomes (measures) Effect measures Results Main limitations Ref. CULLING Brazil Period: NR Three adjacent rural valleys in the Espírito Santo State Culling of seropositive dogs 0 and 6 months after inclusion 2 intervention valleys vs 1 control valley Intervention valleys – 267 humans 12 months Human infection (seroconversion in Dot-ELISA) Difference in infection rates in humans and dogs in intervention vs control valleys 0% difference in human seroconversion rates i) non-randomized [101] Control valley – 202 humans Canine infection (seroconversion in Dot-ELISA) Not significant difference (4%) in dog seroconversion rate ii)low number of clusters for comparison (2∶1) Dogs – NR iii) 26.5% loss to follow-up in humans iv)small sample of domestic dogs, v)Canine loss to follow-up not described Brazil Period: NR 28 rural villages in the São Luiz do Curu Municipality in the State of Ceará Rapid culling based on ELISA versus conventional culling based on IFAT 1 intervention group vs 1 control group composed Intervention group – 276 dogs 10 months Canine infection (seroconversion in ELISA) Difference of seroprevalence between groups Significant reduction of seroprevalence in the intervention group (27% versus 9%; P = 0.001) i) baseline seroprevalences significantly different [102] Control group – 254 dogs ii) impossibility to disentangle the effect of the time to dog removal from the effect of the lower sensitivity of the IFAT test Brazil 1989–1993 Two neighborhoods of the city of Jequié in the Bahia State Yearly culling of seropositive dogs 1 intervention area vs 1 control Initial number of dogs in the intervention area – 235 5 years Canine infection (seroconversion in FAST-ELISA) Difference in cumulative incidence of canine infection between neighborhoods Canine infection cumulative incidence did not change (P = 0.07) i)small sample size [50] Dogs in the control area – NR Human pediatric VL cases Difference in incidence of pediatric VL Pediatric VL incidence decreased in the intervention area (P<0.01) ii)ineffective dog removal Humans – NR iii) differential losses during follow-up iv) low human VL incidence v) non-randomized vi) areas submitted to heterogeneous follow-up Brazil Period: NR Urban and periurban areas of the Jequié Municipality in the Bahia State Culling of seropositive dogs at baseline and every 6 months Before/after Cohort of 1286 susceptible dogs, no controls 18 month Dog emigration Canine infection (seroconversion in ELISA) evaluated every 6 months Dog emigration rate and canine infection incidence Emigration rate: 2.26 cases/100 dogs-year Intervention of dog culling was not directly evaluated [103] Overall annual incidence of 6.55 cases/100dogs-year Two risk strata for seroconversion rates with higher risk in the periurban versus downtown clusters Brazil 1997–2000 Jequié city, State of Bahia Culling of seropositve dogs at baseline and every 8 months Before/after Dynamic cohort of 447 dogs at study entry 31 months Canine infection (seroconversion in ELISA) evaluated every 8 months Difference in incidence rates every 8 months No significant changes in the incidence rates through the study period i) no control arm [108] INSECTICIDE MATERIALS Venezuela 1999 Two rural villages in the Island of Margarita Pyrethroid λ-cyhalothrin sprayed indoors every 5 months; and organophosphate fenitrothion through peri-domestic fogging 16 times during the year vs control( no intervention) 1 intervention village vs 1 control Five houses in each village (control and intervention) 12 months Plebotomine sandfly density Differences in indoor and outdoor sandfly density between intervention and control groups Significant reduction of the sandfly density in the intervention village (P<0.001) i) small sample size [106] ii) low external validity Brazil 1999 Two localities in the Corumba municipality, State of Mato Grosso 65% permethrin spot-on three times monthly 1 intervention locality vs 1 control Intervention area: 150 dogs 5 months Canine seroconversion in IFAT VL prevalence three months after treatment Reduction of VL prevalence in the intervention area (19.3% to 10.8%) i) non-comparable baseline prevalence [105] Control area: 146 dogs Increase of VL prevalence in the control area (4.1% to 16.8%) ii) low sensitivity of the test used to define infection (IFAT) iii) significant losses during follow-up Brazil 1999–2000 Two neigborhoods in the Capitão Eneas Municipality, State of Minas Gerais Deltamethrin-impregnated dog collars vs none intervention 1 intervention vs 1 control area Intervention area: 251 dogs 5 months Canine infection (conversion in ELISA or peripheral blood PCR-hibridisation assay) Difference in the infection rates between groups 11.9% intervention group vs 17.6% in the control group (P = 0.24) i) the one to one comparison, [55] Control area: 190 dogs ii) non comparable baseline prevalence of VL infection between groups iii) high rate of loss of follow-up, iv) high frequency of collar loss and migration of dogs Brazil 2003 Salvaterra municipality in the Marajó Island, State of Pará Deltamethrin impregnated bednets vs untreated bednets Crossover study Two houses in each group Three consecutive nights Bednet barrier effect, human landing rates and 24h sandfly mortality rates Differences in barrier effect magnitude, landing rates and sandfly mortality rates 39% increasing in barrier effect i) small number of observations [109] 80% reduction in human landing rates ii) short exposure period 98% increasing in sandfly mortality COMBINED INTERVENTIONS Brazil 1965–1979 Mainly rural communities of 19 Municipalities of Minas Gerais State Dog culling + human treatment + DDT spraying of houses Before/after 81,162 dogs, unreported number of human subjects and no control 15 years Human VL (clinical AND/OR positive CFR AND/OR positive parasitology) Incidence of human VL before/after Human VL disappearing ∼0% i) no controls [100] Canine VL (seroconversion in CFR) Incidence of canine VL before/after Canine VL ∼0% ii) intensity and periodicity of intervention poorly described iii)low sensitivity of the complement fixation test to detect canine infection iv) passive reporting cases as the data source for endpoint in humans iv) the human and canine population exposed to the control measures was not reported Brazil 1995–1996 One neighborhood of the city of Teresina, State of Piaui (A)spraying houses and animal pens with insecticide Random allocation of 34 clusters to one of four arms 213 susceptible humans (120 evaluated, numbers of susceptible humans in each intervention were not reported) 12 month Human infection (seroconversion in ELISA) at least 6 months after intervention Difference in incidence rate Significant reduction in incidence in the group exposed to intervention B i) non comparable baseline VL incidence between the house spraying group and the other three groups [110] (B) spraying houses and infected dog-culling Control arm: group submitted to house spraying (D) No significant decrease in incidence in A and C intervention groups ii) high percentage of loss to follow-up of susceptible individuals (44%), (C) combination of (A) and (B) iii) the suboptimal sensitivity and specificity of the method to measure seroconversion (crude antigen-ELISA) (D) spraying houses Brazil 1995–2000 Municipality of Feira de Santana, State of Bahia Culling of seropositive dogs and house and animal shelters pyrethroid insecticide spraying None 124 localities (30 urban and 58 rural with human VL incident cases and 36 localities around them) 6 years Human VL incidence Correlation between measure coverage and frequency with human VL incidence Positive correlation with number canine surveys, coverage of canine surveys and number of sprayed buildings i) secondary source data [107] ii) lack of a control arm Brazil 2004–2006 Two neighborhoods of the Feira de Santana city, State of Bahia (A) No intervention Intervention was randomly allocated to one of 3 areas in each neighborhood Dynamic cohort of 2362 children (688, 782 and 892 allocated to interventions A, B and C, respectively 27 months Human incidence (seroconversion in ELISA) Relative risk for infection every 12 months Lower but not significant incidence decrease in the intervention areas i) low study power, [111] (B) insecticide spraying ii) significant losses during follow-up (C) combination of insecticide spraying and seropositive dog culling DOG VACCINE Brazil Period: NR São Gonçalo do Amaranto Municipality in the Rio Grande do Norte State Vaccination with Fucose-Mannose-ligand antigen, 3 subcutaneous doses at 21 day intervals Intervention arm – FML vaccine Intervention – 58 seronegative healthy dogs (in IFAT and FML –ELISA) 24 months Symptomatic VL at 2, 7, 13 and 24 months Difference in symptomatic VL rate (cumulative at 24 months follow-up) 8% (intervention) vs 67% (placebo) symptomatic VL i) impossibility of accurate evaluation of the infection rate because the vaccine product and probably the repeated leishmanin doses interfered with the serological response with more than half of control subjects showing positive FML-ELISA tests, [104] Control arm –Saline placebo Control – 59 seronegative dogs (in IFAT and FML-ELISA) FML-ELISA seroconversion at 2, 7, 13 and 24 months Differences in seroconversion rates (cumulative at 24 months follow-up) 100% (intervention) vs 68% (placebo) seroconversion rate ii) no random allocation Leishmanin conversion (L. donovani antigen) at 2, 7, 13 and 24 months Difference in leishmanin positive rate (cumulative at 24 months follow-up) 94% (intervention) vs 14% (placebo) leishmanin positive rate iii) lack of baseline data on dog characteristics CFR = complement fixation reaction. NR = Not reported. Key questions for control What is the most cost-effective control strategy for VL? How to conduct a valid impact evaluation? Can general support measures (nutritional rehabilitation and housing improvement) be targeted to VL endemic areas? What is the potential impact of current dog vaccines on transmission? Questions for research What are the determinants of dog infectiousness for the sandfly vector? What are the determinants of dog susceptibility to infection? What is the efficacy of current dog vaccines to prevent disease in dogs and to reduce infectiousness for the sandfly vectors? What is the effectiveness of insecticide- impregnated dog collars to prevent human and canine infection? What is the efficacy/effectiveness of alternative vector control devices (insecticide treated nets, curtains, etc) in the prevention of VL? Discussion Research gaps This review of evidence related to VL control in Latin America revealed that a lot of work remains to be done in order to clarify the dynamics of Leishmania transmission in human, canine and vector populations. The exact burden of disease remains largely unknown. The increasing trend of VL cases observed in Brazil and the spread of transmission to previously not affected areas raise doubts about the impact of ongoing control measures. The determinants of human infection and of symptomatic disease are also poorly understood with the exception of the nutritional status in young children. To diagnose VL in humans the rK39-ICT has clear advantages over the IFAT or ELISA based tests that are widely used in Latin America. The DAT assay has shown similar diagnostic performance but is not as user-friendly as the rK39. The research priorities in this field should be geared towards diagnostic accuracy studies in large prospective trials (phase-3) and to study diagnostic performance in specific groups such as HIV co-infected patients. Current treatment practice in VL in Latin-America is based on rather weak scientific evidence. It is worrisome that case fatality rates remain high and are even increasing, at least in Brazil. The lack of clinical evidence from the region is very worrying. We retrieved not a single phase-3 randomized controlled trial on VL conducted in the Americas. Nowadays, one phase-2 trial with miltefosine is ongoing and two Brazilian large randomized controlled trials with liposomal amphotericin B, amphotericin B deoxycholate and meglumine antimoniate are expected to initiate recruitment in 2009. The research priorities include well-designed clinical trials with pentavalent antimonials, amphotericin B deoxycholate and the liposomal formulations, miltefosine and drug combinations. Although the resistance to antimonials observed in India is less relevant in Latin America, drug combinations are attractive because their potential for shortening treatment schemes and reduction of toxicity. Clinical factors associated with treatment failure should be studied to contribute to the development of a prognostic score that allows early interventions to reduce case fatality rates [14],[87]. Control interventions targeting the dog reservoir for culling/treatment require accurate assays able to detect the asymptomatic infections as well as the symptomatic dogs. Validating such tests is no easy task, as there is no adequate gold-standard for the diagnosis of asymptomatic infection. PCR-assays seem to be very attractive but estimating their accuracy and reproducibility still constitutes a research priority. Moreover, novel screening strategies based on combined, parallel or sequential use of current available tests needs to be validated. Another challenge faced in canine diagnosis is the distinction of positive serology results produced by natural infection from those induced by vaccines. The development and proper validation of tests with capacity to discriminate both phenomena are crucial to avoid interference with concomitant interventions including dog culling and vaccination in the same area. Furthermore, the study of the determinants of dog infectiousness for the sandfly vector is essential to define the best culling strategy [112],[113] and the determinants of dog susceptibility to infection [114] is crucial for the design of canine vaccine trials. Some of the problems with the design of the community intervention trials we reviewed are related to the lack of accurate diagnostic methods to define the relevant outcomes in the human and canine population. Furthermore, the definition of a control group is challenging because of an obvious ethical dilemma. The heterogeneity of disease transmission within the study area often generated imbalances in the baseline comparisons among groups and the random allocation process is also complex because of the mobility of the human, canine and vector population. Most of the reported community trials used a too limited number of clusters for comparison (usually a one to one comparison). In spite of all those limitations a relevant number of reports could be reviewed in detail, showing no strong evidence for a significant impact on VL transmission for any of the interventions reviewed. Canine culling seems to be the least acceptable intervention at community level for obvious reasons and has low efficiency due to high replacement rate of eliminated dogs with susceptible puppies [103],[115],[116]. Vector control interventions are better accepted by the affected populations and mathematical models suggested encouraging efficacy, but they need further study. Better knowledge of vector seasonality and behavior is required for proper timing of these interventions. The current evidence indicates that spatial fogging is useless and that the residual effect of house wall spraying is very short [106],[117]. Insecticide impregnated collars seem to have a longer residual effect [56] and theoretical advantages over the other methods and should be studied in larger and well-designed controlled trials. The potential emergence of resistance to insecticides should also be considered for the long-term planning of any vector control intervention [118]. Canine and human vaccine development needs to be prioritized. The dog vaccines already registered in Brazil have some protective effect against canine VL but none of them were properly evaluated as control measures against human VL [119],[120]. Such evaluation is challenging as field trials should include relevant canine endpoints, related to dog infectiousness for the sandfly vector, as well as relevant human endpoints, that include symptomatic and asymptomatic infections in order to obtain precise estimates of the vaccine effect on transmission rates. Human vaccine development is expected to take at least several years to obtain efficacious and safe candidates for clinical trials. Furthermore, the surrogate markers of the desired protective effect are not well understood and the definition of target population for such products will be a matter of intense debate. The role of sylvatic and peridomestic animals such as foxes, marsupials and rodents in some relevant VL transmission scenarios deserves more specific research [6]. Last but not least, in countries such as Brazil, where the government has put the elimination of hunger as a political priority, targeted nutritional support in VL risk areas would be an interesting and probably cost-effective intervention from a societal perspective. Similarly, schemes for the improving of housing and waste management as well as other general measures involving active community participation should be encouraged [121],[122]. Finally, the strengthening of the surveillance system capacity is essential to avoid the underreporting of human cases [123] and to follow-up the infection behavior in canine population. Strong surveillance will certainly contribute to improve data quality for decision-makers in this complex scenario. Concluding remarks The elimination of zoonotic VL in Latin America is not (yet) a realistic goal taking into consideration the complexity and diversity of its transmission scenarios, the scientific knowledge gaps and the lack of adequate and properly validated interventions. Many countries perceive the burden of leishmaniasis as negligible; there is not much political support nor funding for VL control. The zoonotic nature of transmission is an additional constraint that limits the impact of the few known effective prevention and control interventions. Nonetheless we believe the improved control of VL is possible if the region builds the political will, develops a more coherent regional control policy, and invests in better case management and epidemiological surveillance systems. The implementation of a focused research agenda to support such control initiative is essential. Supporting Information Checklist S1 PRISMA checklist. (0.07 MB DOC) Click here for additional data file.
- Record: found
- Abstract: found
- Article: found
Factors Associated with Visceral Leishmaniasis in the Americas: A Systematic Review and Meta-Analysis
Introduction In the Americas, visceral leishmaniasis (VL) is a zoonosis caused by the protozoan parasite Leishmania infantum (syn. L. chagasi). The disease is transmitted to humans by the bite of female sandflies of the genus Lutzomyia [1]. L. longipalpis is the main vector species in the New World [2] but transmission can also occur through Lu. cruzi and Lu. evansi [3], [4]. Other forms of transmission have already been recorded: congenital, drug injection and blood transfusion [5], [6]. Potential wild disease reservoirs are the fox (Lycalopex vetulus and Cerdocyon thous) and the opossum (Didelphis albiventris) [7]. Dogs (Canis familiaris) are considered the main reservoir in the urban setting [8]. Clinical manifestations of the disease can range from asymptomatic forms to severe visceral involvement [9]. The incubation period lasts around two to six months [10]. The fatal outcome occurs predominantly by co-infection or bleeding complications [11]. Studies in Brazil show a variation in case-fatality rates between 4.2% and 10.2% in treated patients [12]. For the diagnosis of human visceral leishmaniasis, delayed hypersensitivity tests, such as the Montenegro skin test (leishmanin skin test) [13], parasitological, serological (e.g., the Indirect fluorescent antibody test - IFAT, the enzyme-linked immunosorbent assay - ELISA, the direct agglutination test - DAT) and molecular methods are commonly used [14]. Immunological techniques are more often used in epidemiological studies since they are easier to implement. The burden of the disease is not exactly known in the Americas, since there is a lack of effective surveillance systems [15]. Although the Brazilian surveillance system is considered better than in other Latin American countries, VL is also underreported in this country [16]. Previously known as a rural endemic disease, visceral leishmaniasis has become endemic and epidemic in large Brazilian cities since the 1980s [17]. The country accounts for 90% of reported cases in the Americas and is the third largest VL focus globally [15]. In Brazil, the recommended control actions against the disease are the use of insecticides against phlebotomines and culling of seropositive dogs, both strategies with little evidence of effectiveness and operationally costly [18], together with the identification and early treatment of human cases [19]. However, they have not produced effective results, given that neither incidence nor fatality rates were reduced in recent years. Moreover, the disease has spread to all regions of the country, and continues to expand into formerly disease-free cities [20]. Identification and quantification of risk factors for VL, besides being useful to understand the determinants of infection acquisition and disease development, can help develop more effective and well-targeted control measures [21]. There are several factors which may be associated with L. infantum infection in humans and many are considered controversial or only partially understood [18]. In this context of heterogeneity in study results, careful evaluation of its potential causes may, in meta-analysis procedures, be more useful than the mechanical calculation of summary measures [22]. Therefore, when summarizing studies that have addressed factors associated with VL, issues such as the quality of studies, design used, methods of data analysis, control for confounders, the way results were reported, and diagnostic tests employed must be considered in order to obtain the best evidence available. In a systematic review of risk factors for visceral leishmaniasis in South Asia [23], consistent patterns have been described for some variables, such as: spatial clustering of cases, conditions of households, areas of vegetation near homes and poverty. However, due to the peculiar characteristics of the disease in that context (the etiological agent is Leishmania donovani; vector is Phlebotomus argentipes; and infected humans are the only reservoir [23]), results cannot be used to make predictions about other contexts. Specifically with respect to human American visceral leishmaniasis, there is no literature record of systematic reviews and meta-analysis regarding factors associated with infection by L. infantum. Thus, through the analysis of observational studies that investigated such factors, this study aimed to obtain more accurate data about each of the factors studied as well as to analyze the quality of publications and to identify gaps in existing knowledge. Methods Eligibility Criteria Epidemiological cross-sectional, cohort, case-control and ecological studies were included in the review. These should have described associations between socioeconomic, environmental, family and individual variables and the occurrence of any outcome related to the acquisition of infection by L. infantum. This outcome could be the infection itself (as measured primarily in the study, regardless of the diagnostic assay or of the presence or absence of symptoms in the subject), the identification of active patients who had manifested the clinical disease (in the case of case-control studies) or the notification of cases by health services (in the case of ecological studies using secondary data). There was no restriction for age, sex or language. We excluded studies conducted before 1980 or with populations other than those of the American continent, as well as reviews and papers published in scientific conferences or that were purely descriptive with no possibility of obtaining measures of association. With respect to the variables, we excluded those which referred to: a) control actions (not the focus of this review); b) subject's genetic characteristics (more related to differences in clinical prognosis); c) for how long a subject had been living in a household (relevant only in local contexts); d) the categorization of the subject's domicile as urban or rural (due to the lack of standardization in this definition). We maintained, however, those related to the environmental characteristics surrounding the households. We also excluded studies and variables whenever what was being described in the text was impossible to understand or when there were inconsistencies in quantitative data presented or flaws that could invalidate the association described. Search and Extraction of Information All search strategies were conducted independently by two researchers (Belo, VS & Barbosa, DS) between March and June 2011. The terms used and databases consulted are described in Text S1 in the order they were searched. In addition, Brazilian research experts were consulted and searches through the reference lists of each article selected at the initial stage were conducted. All titles and abstracts of identified articles were analyzed, being initially excluded those deemed irrelevant. When the information provided was not sufficient for the decision, or when studies were considered relevant for at least one of the researchers, the full text was analyzed. At this stage, papers were selected to be part of the present review. The extraction of information contained in the publications was conducted by a researcher (Belo, VS) and reviewed by all authors of this article. We contacted authors of primary studies to obtain data necessary to calculate measures of association (and possible inclusion in meta-analysis) when they were not described in the publications. Individual patient data were not requested. Methods for Assessing Risk of Bias in Primary Studies To identify potential risks of bias in studies and limitations in data analysis and in the way results were reported we used the STROBE statement. It issues recommendations that have been developed to improve the presentation of results from observational studies [24] including several items related to the existence of susceptibility to bias [25]. Additionally, questions from the Newcastle-Ottawa Quality Assessment Scale [26] (for assessing quality in observational case-control and cohort studies) and the book of Fischer and Getis [27] (for assessing quality in ecological studies) were consulted. We have not used scales that result in a numerical quality threshold because this procedure involves arbitrary weighting of items, some of which may not to be directly related to the study's validity [25]. Thus, we decided to approach the problem more generally, pointing out the major limitations and susceptibility to bias of the revised studies. Study quality and susceptibility to bias were used to discuss the limitations of existing knowledge, without interference, however, in meta-analysis procedures. There was no exclusion of studies due to those factors. Analysis and Summary of Results Each association between a given exposure variable and a specific outcome was considered a separate and independent meta-analysis. Results of primary studies - except ecological - were described by Odds ratios (OR) and confidence intervals (CI). In cases in which there was information about the probability of significance (p-value), the direction of the association and the sample size of the study, the OR (or correlation coefficient) was estimated by reverse computation. To make decisions about whether to combine or not effect measures, the agreement among the issues added by each study and the differences on their characteristics were evaluated. Whenever a summary measure of effect was obtained, the random effects model was used to pool the data. Under this model it is allowed that the true effect size varies from study to study, thus the overall variability includes the within-studies variance as well as the estimate of the between-studies variance. However, in cases where the number of studies was less than four, due to the lack of precision of the measure of variability [28], we preferred to use the fixed effects model. The Q-Statistical test was used to analyze the occurrence of heterogeneity in effect sizes across studies. The I2 statistic was calculated to determine what proportion of the variance observed represented a real dispersion in the effect sizes (variation between studies), i.e. not due to random error (intra-study variation). Due to the sufficient number of studies and the significant heterogeneity among them, subgroup analyses were carried out for gender, age, presence of dogs in the household and presence of chickens/birds. The following groups were considered: type of study (i. cross-sectional; ii. cohort; iii. case-control), method for measuring the outcome (i. leishmanin skin test (LST); ii. serological; iii. serological and LST; iv. other; v. clinical case), age group of participants (i. children; ii. adults; iii, all ages) and adjustment for confounding (i. yes; ii. no). The Q test was used for comparison of effect sizes and of heterogeneity between subgroups. Those which explain heterogeneity best (higher between subgroups and lower within subgroups) were described. For variables on which subgroup analyses were performed, studies were combined only within the subgroups and an overall measure was not calculated. With regard to studies in which more than one diagnostic method had been performed, results of serological tests (ELISA firstly) were kept in the meta-analysis and results of LST were excluded. However, in analyses where studies were grouped by the method of measuring the outcome, the result of each technique was maintained and used for analyzing the subgroup to which it belonged. The weighted Z-method for combining p-values [29] was applied when the diversity of studies in terms of design, populations or other characteristics became inadequate to obtain a summary-measure based on effect sizes, provided that the question analyzed was similar. We investigated the existence of publication bias among the studies selected for each targeted association (provided that the number of publications is large enough) using the funnel plot, the Egger statistical test and the statistic of “Durval and Tweedie's Trim and Fill”. The last procedure re-computes the effect size imputing studies until the funnel plot is symmetric about the new effect size. Thus the best estimate of the unbiased effect size is calculated [28]. Theoretical discussions analyzing not only statistical significance, but also the strength and the direction of associations were conducted for variables on which it was not possible to use the methods presented. To perform the meta-analysis of p-values, we used the software Meta-P and for remaining analyzes we used the software CMA, version 2.0.057. Results Selection of Publications and Limitations in Studies Included Forty-seven publications (51 studies) were included in the review. Figure 1 shows a flowchart of the search strategy and study selection, along with the reasons for exclusions. 10.1371/journal.pntd.0002182.g001 Figure 1 Flowchart showing the selection of studies analyzed in this review. Regarding the type of study, 24 were cross-sectional, 17 ecological, 7 case-control and 3 were cohorts. Three studies were conducted in Venezuela, two in Colombia, one in Honduras and forty-five in Brazil. In five studies the population consisted of children only. In studies where the infection was measured directly by investigators, the diagnostic test most often used was the LST, followed by ELISA and IFAT. In all case-control studies, cases were patients with clinical disease. The main potential sources of bias in cross-sectional and cohort studies were the eligibility criteria employed and losses due to refusal to participate and during follow-up. In eight studies, the process of selection of participants was not clearly explained. In nine no reference was made to refusals or losses and in sixteen studies, although they have been quantified, there was no discussion about those profiles and the way they could have affected the results. With respect to case-control studies, there were also limitations on the description of the selection of participants and refusals in three studies. In four there was a clear possibility of recall bias. In ecological studies, in addition to limitations inherent to the design and use of secondary data, we detected the possibility of ecological fallacy in five of them and no verification of addresses of cases in six studies. Regarding the limitations of the analyses, only thirteen studies did some kind of control of confounding. Except for two, none of the others described the criteria used for categorizing continuous variables. In ten the strength of the associations was not considered, and only statistical significance was analyzed. In six studies the normality of the outcome variable was not checked, although this was necessary for the statistical tests applied. Specifically for ecological studies, seven did not take into account the possible spatial dependence of data. The main limitations of data presentation were related to the non-description of results for variables that were not statistically significant. In addition, several variables had to be excluded because it was not possible to understand the text description or the analyses performed. In general, there were no differences in the quality of published as compared to unpublished studies. The limitations and susceptibilities to bias identified in both groups were of the same types and occurred at a similar frequency. Table S1 describes the characteristics, susceptibilities to bias and limitations of each study reviewed. Summary of Information I. Gender Eighteen studies were included [13], [30]–[46]. In one of them, data for adults and children were presented separately [46]. Two types of diagnostic tests were used in four studies, and separate analyses were presented for each of one [32], [38], [42]. In the analysis with no division of studies by subgroups, the effect measure was heterogeneous between studies (p = 0.001) and I2 value was 57.73%. When considering the analyses by subgroups of studies, the outcome measurement method was the only grouping characteristic that showed significant differences in the effect measures, significant heterogeneity between subgroups (p<0.001) and lack of heterogeneity within the subgroups (p = 0.625). With the exception of studies where the outcome was the clinical disease, where I2 was 37.13%, in all other subgroups the I2 value was zero. In the forest plot shown in Figure 2 one notices that in the subgroup of studies in which diagnosis was provided by LST and in studies with clinical cases, the male sex was significantly associated with infection, with odds ratios of 1.30 (1.17–1.44) and 2.38 (1.65–3.45), respectively. As for the others subgroups, although the results were not statistically significant, association was in the opposite direction, with male subjects having a smaller chance of infection. 10.1371/journal.pntd.0002182.g002 Figure 2 Forest plot for the variable gender: studies divided into subgroups according method for measuring the outcome. CS-Cross-sectional; *Serological result from study with two tests; #Result in adults; (2) Second study from the same publication; LST- Leishmania skin test; Squares represent the weight of each study; Lozenges represent the summary measure of each subgroup; Reference: Females, OR = 1. II. Age In order to carry out meta-analysis on age we selected the cut-off point of ten years for comparisons, considering only the fact that this was the one used in most of the primary studies. Twelve studies were analyzed [31], [34], [35], [38], [40], [41], [42], [43], [45], [47], [48], [49]. In three of them [38], [42], [49] two different diagnostic tests were used and analyzed separately. In the overall analysis we detected significant heterogeneity between studies (p<0.001), with I2 of 90.34%. In analyses with division of studies by subgroups, statistically significant heterogeneity between subgroups (p<0.001) and not within subgroups (p = 0.867) was found only for the outcome measurement method used. Statistical I2 value was 63.71% in the subgroup of studies for which the outcome was the clinical disease and zero for the others. In Figure 3 one notices that both in serological study subgroup and in the LST study subgroup, greater test positivity was found among individuals aged ten years old or more. However, in the last, associations were stronger and statistically significant. Two studies that used other diagnosis procedure showed the same pattern. However, in both studies in which clinical cases was considered the outcome, the direction of the association was the opposite, that is, those with ≤10 years of age with the higher risk. 10.1371/journal.pntd.0002182.g003 Figure 3 Forest plot for the variable age: studies divided into subgroups according to the method for measuring the outcome. CS-Cross-sectional; *Serological result from study with two tests; #Result in adults; LST- Leishmania skin test; Squares represent the weight of each study; Lozenges represent the summary measure of each subgroup; Reference: Being Over 10 y.o.a., OR = 1. Three papers reporting results of four studies [30], [33], [50] that compared the occurrence of infection in asymptomatic subjects older and younger than 15 years old (by Kalazar detect dipstick rK39 [30], Elisa and LST [33], [50]) detected the same pattern, that is, older subjects were more likely to be positive. In the remaining studies that evaluated age, and for which it was not possible to perform a meta-analysis, overall the number of positive subjects also increased with age in studies which performed LST or serological tests [37], [39], [51]. On the other hand, Viana et al. [13] used both tests and did not observe any differences between the average age of positive and negative subjects. For studies which involved children only [32], [36], [44], [52], meta-analysis were not performed since the upper cut-off point was different between studies. In this case, no consistent patterns were identified in the results. III. Malnutrition In order to evaluate the potential role of malnutrition on the risk of infection by L. infantum we selected four papers reporting results of five studies which compared eutrophic children and those with some degree of malnutrition [32], [36], [48], [52]. Figure 4 shows that the presence of malnutrition (although non-significantly) decreased the likelihood of diagnosis of asymptomatic infection. There was significant heterogeneity in the Q test (p = 0.020), with an I2 value of 65.88%. 10.1371/journal.pntd.0002182.g004 Figure 4 Forest plot: malnutrition. CS-Cross-sectional; *Serological result from study with two tests; (2) According to a study from the same publication; LST-Leishmania skin test; Squares represent the weight of each study; Lozenges represent the summary measure. Reference: Lack of Malnutrition, OR = 1. IV. Presence of dogs in the household The presence of dogs in the household as a risk factor was evaluated in twelve studies [32], [33], [40]–[44], [53]–[55], and three of them used and provided results using two separate diagnostic methods [32], [33]. In the overall analysis without subgroup stratification, no significant heterogeneity was identified by the Q test (p = 0.152). The I2 value was 29.97%. The combined data demonstrated a pattern of increasing likelihood of infection for subjects with dogs in the household (OR = 1.23; 1.07–1.42). However, in both cohort studies analyzed [32], [33] the results were in opposite direction (Figure 5). 10.1371/journal.pntd.0002182.g005 Figure 5 Forest plot for the presence of dogs in the household: studies divided into subgroups according to type of study. CS-Cross-sectional; *Serological result from study with two tests; (2) According to a study from the same publication; LST-Leishmania skin test; Squares represent the weight of each study; Lozenges represent the summary measure; Reference: Not Owning Dogs, OR = 1. V. Seropositive dogs For the association between dog seropositivity and the occurrence of VL, most studies used aggregate data, with the exception of one case-control study [53] in which there was a greater history of canine euthanasia in households of cases. In figure 6 we describe five studies providing correlation coefficients between dog seropositivity (assessed by census or sample serosurveys) and the incidence of human VL (based on notified cases) aggregated by geographic units in a given period of time [56]–[60]. In all studies a positive correlation was described. However, the papers of Bavia et al. [56] and of Carneiro et al. [57] have the limitation of analyzing just the number of human and canine cases, not incidence rates or prevalence. In a study conducted in Teresina [61], the multilevel model that provided the best fit to data included the association between the prevalence of dog seropositivity and the incidence of the disease among humans. In Belo Horizonte, in a Bayesian multivariate model for the relative risk of visceral leishmaniasis, the variable with the greatest power for explaining human VL incidence was the prevalence of canine seropositivity [62]. On the other hand, Oliveira and Araújo showed a weak positive correlation between canine prevalence and the incidence rate of VL in humans in a time series study in the municipality of Feira de Santana, Bahia [63]. However this last study has the limitation of analyzing yearly aggregated data for a short period of time (6 years). 10.1371/journal.pntd.0002182.g006 Figure 6 Forest plot with ecological studies that correlated cases in dogs and in humans in a given analyzed land unit. *Terms used in the primary study (for details, see table S1). VI. Chickens/birds/chicken coops Six papers with eight studies included in the review evaluated at least one variable related to the presence of poultry in the peridomestic environment 32,33,40,42,43,53. Information about the presence of chicken coops or birds was only used when there was no information about the presence of chickens in the primary study. Three studies carried out separate analyses for two diagnostic tests [32], [42], [43]. In the overall analysis the I2 value was 76.53%. In all subgroups analyzed (including when division was made according to the type of poultry) heterogeneity was maintained. Figure 7 shows a forest plot with the summary measure obtained when all studies were pooled. There was no association between the presence of poultries in the households and the occurrence of infection (OR = 0.97; 0.69–1.37). 10.1371/journal.pntd.0002182.g007 Figure 7 Forest plot for the presence of chickens, birds or chicken coops in the household. CS-Cross-sectional; *Serological result from study with two tests; (2) According to a study from the same publication; LST-Leishmania skin test; Squares represent the weight of each study; Lozenges represent the summary measure; Reference: Not Owning Birds, OR = 1. VII. Other animals Given the small number of studies or lack of description of results through measures of association, it was not possible to obtain consistent information on the association between the presence of other specific animals and the occurrence of VL [33], [53], [64]. A study conducted by Gouvea et al. [40] analyzed the association between the presence of animals, with the exception of dogs and birds, and the occurrence of infection. They found a high frequency of infection in the presence of animals; however without statistical significance. Other studies analyzed the overall presence of animals and they identified a trend of increased likelihood of infection in households with animals [33], [44]–[46]. VIII. VL in relatives and neighbors We analyzed six studies with respect to prior occurrence of VL in relatives [32], [40], [44], [64], [65] and three for VL in neighbors [32], [44] (two studies analyzed more than one diagnostic method [32]). Figure 8 shows a forest plot for those associations. The summary measures showed greater chance of infection in subjects who referred VL cases among relatives or among neighbors. However, in addition to being stronger (OR = 2.09) the association was only statistically significant for the occurrence of prior cases among relatives. 10.1371/journal.pntd.0002182.g008 Figure 8 Forest plot: prior cases in relatives or neighbors. CS-Cross-sectional; *Serological result from study with two tests; (2) According to a study from the same publication; LST-Leishmania skin test; #Neighborhood results for studies that also studied relatives; Squares represent the weight of each study; Lozenges represent the summary measure of each subgroup; Reference: Lack of Cases Among Relatives/Neighbors, OR = 1. For the variable “relatives” there was significant heterogeneity of data (p = 0.05 and I2 = 70.57%), which did not occur with the variable “neighbors” (I2 = 0). IX. Socioeconomic variables Three studies at the individual level analyzed the association between the existence of a water network connected to public service and the occurrence of infection [34], [40], [44] and showed there was protection for appropriate supply, with an OR of 0.88 (0.69–1.12) in the summary measure. With respect to ecological studies, Braga [66] demonstrated that census tracts where less than 50% of households had piped water suffered greater increase in the incidence of infections. Cerbino-Neto et al. [67] demonstrated that in neighborhoods where there was a lower proportion of piped water, VL incidence was greater. On the other hand, even if the municipality had been the same and only the period of study had changed, Rodrigues [59] identified an opposite pattern to that of previous results [67]. Araújo [62], in turn, did not identify associations between such variables. With respect to the existence of sewerage system, in three out of four studies which performed analysis at the individual level [33], [34], [40] there was a decrease in the likelihood of infection associated with appropriate sewerage system. The summary odds ratio of data was protective, with a value of 0.78 (0.65–0.95). In ecological studies the association between lack of an adequate sewerage system coverage and an increase in incidence was demonstrated by all four studies analyzed [59], [62], [66], [67], even considering that in only two [66], [67] the results showed statistical significance. For garbage collection the three studies at the individual level [33], [34], [43] showed that appropriate waste collection was associated with a smaller chance of infection, with an OR of 0.63 (0.49–0.80) in the summary measure. For ecological studies, two [64], [67] of the three studies [59] in which geographic units were used for analysis and in which time was aggregated, showed that the incidence was lower in areas with suitable garbage collection. In another ecological study, Araújo [62] did not identify any association between these variables. Finally, a study conducted in Teresina [66] demonstrated that census tracts where less than 50% of households had appropriate waste collection suffered a greater increase in the incidence of infections. A combination of five studies that compared the existence of inadequate floor surfaces in the household (earth or dirt floor) with another type of floor surface considered appropriate [33], [34], [40], [42], [44] showed that the presence of the latter, albeit with significant heterogeneity in data presented (Q test: p<0.001; I2 = 80.51%), was associated with smaller chance of infection, with an OR of 0.50 (0.31–0.80) in the summary measure. In order to analyze the household's finishing, we evaluated eight studies [32]–[34], [40], [43], [44] and the summary measure showed a lower chance of infection when finishing was appropriate (made of bricks and/or concrete, OR = 0.68; 0.47–0.97), albeit also with statistically significant heterogeneity in the Q test (p<0.001) and an I2 value of 78.24%. For roof finishing, we analyzed four studies [33], [34], [40], [44] and also verified the pattern of protection for presence of roof tiles, albeit with no statistical significance (OR = 0.85; 0.55–1.30). In the Q test, no statistically significant heterogeneity was identified (p = 0.051), although the I2 value was 61.33% (Fig. 9g). 10.1371/journal.pntd.0002182.g009 Figure 9 Funnel plot for the variables employed in the detection of publications bias in the review. a-gender; b-age; c-presence of dogs in the household; d-presence of chickens or birds in the household; Dots refer to studies imputed by the Trim and Fill method. Open diamonds represent the original estimate of the Odds ratios while solid diamonds represent Odds rations recomputed by the statistic of Durval and Tweedie's Trim and Fill. When the subjects' income was directly assessed, with the exception of one study [66], in all others [46], [58], [62], [67], [68] an increase in income was associated with a decrease in the occurrence of the disease. Seven studies analyzed the relationship between literacy and the occurrence of VL using the subject as a unit for analysis [31], [32], [34], [40], [46], [63]. Despite them all analyzing a similar issue, studies had different cut-off points and definitions so we chose to combine p-values. The value identified with the combination was lower than 0.001. This demonstrates the association between increase in the level of education and decrease in the likelihood of subjects getting infected. In ecological studies that evaluated issues pertaining to education [62], [67], [68], the same pattern was identified. A meta-analysis of p-values was also performed for the number of people living in a household. Four studies with individual units of analysis were combined [34], [40], [43], [46]. Despite the fact that three studies showed an increase in the likelihood of infection in homes that housed more people, there was no statistical significance in the combination (p = 0.072). The influence of socioeconomic factors in the occurrence of VL was also analyzed by studies that used indicators of social condition including some of the variables analyzed separately in the other studies [46], [69], [70]. All of them demonstrated the association between worse socioeconomic conditions and greater incidence of the disease. Associations between the occurrence of infection by L. infantum and the presence of an automobile or electricity in the household [64] and the subject's knowledge about VL [31] were analyzed in only one study each. X. Environmental and backyard conditions For the following variables we chose not to summarize data using summary measures of association or p-values, considering the difference between the issues analyzed and data available in primary studies. The Normalized Difference Vegetation Index (NDVI) was studied in four publications. The NDVI is a vegetation index extracted from remote sensing images which correlates with precipitation, humidity and the presence of green vegetation. Cerbino-Neto et al. [67] identified an interaction of that with the rate of population growth. In areas of dense vegetation and intense population growth there were higher incidence than what would be expected considering the independent effect of those variables. Those authors also demonstrated the association between a higher NDVI value (stratified in units of analysis as greater or smaller than 0.2) and the incidence of VL. In Teresina a greater NDVI mean value in the unit of analysis was associated with an increase in incidence [70]. In a multilevel analysis [61] the highest minimum NDVI value was the best predictor of increase in incidence. Additionally, an interaction with the level of urbanization was also identified. A study conducted in Belo Horizonte [62] analyzed the median NDVI and did not demonstrate any associations. Population growth was specifically associated with an increase in the incidence of VL [66], as well as the greater population density [42], [64]. Finally, studies that analyzed issues pertaining to urbanization showed greater positivity in households closer to the forest [71]; in regions with a smaller proportion of urban area [59], [65], [72]; and in areas with the worst level of urban development [46], [51]. With respect to backyard features, different issues were analyzed. Caldas et al. [32] showed an association between the presence of trees within a 10-meter radius of households and the occurrence of infection. On the other hand, in a study conducted by Moreno et al. [43] people who lived in homes where rubbish was kept in the backyard had lower prevalence. In a study conducted by Gouvea et al. [40], results showed inconsistent patterns; the presence of rubbish without trees or bushes proved to be a factor associated with an increased prevalence of infection, while the presence of trees, bushes and rubbish in the backyard decreased the prevalence of infection. It should be highlighted that results from those studies were not statistically significant. This did not occur, however, when a score was created in which features of the backyard such as lower frequency of cleaning, presence of trees, of waste, of ants and others, were jointly associated with an increase in the chance of infection in multivariate analysis [46]. Associations between rainfall [64] or periods of El Niño [73] and notifications of human visceral leishmaniasis were analyzed in only one study each. XI. Other variables Two publications [32], [43] detected a greater likelihood of infection for people playing or being outside the home between 18 and 22 hours. The relationship between Phlebotomine sand flies and the incidence of VL was analyzed in three ecological studies only [56], [71], [74] and, as expected, greater abundance of the vector was associated with greater occurrence of the disease. Other variables also analyzed in literature were: ethnic group [43], degree of restriction of dogs [53], presence of plants in the household [40] and prior blood transfusion [75]. Publication Bias The presence of publication bias was analyzed only for four variables since there were not enough publications to discuss its possible influence on the other associations reported. According to the methodology used, no evidence of publication bias was identified for associations regarding gender. In Egger's test, the p-value (one-tailed) was 0.34 and the funnel plot was apparently symmetrical (Figure 9a). The same occurred for age (p = 0.16 in Egger's test) (Figure 9b). For the presence of dogs in the household, however, Egger's test identified the possibility of publication bias (p = 0.042). The analysis of the funnel plot showed that it was slightly asymmetrical (Figure 9c); it was then necessary to impute three studies using the “Durval and Tweedie's Trim and Fill” method to obtain symmetry. The new OR estimated was slightly lower (1.17; 1.01–1.36) but remained statistically significant. Finally, the Egger's test was not significant for the presence of birds (p = 0.06), but the funnel plot was not symmetrical; three studies under “Durval and Tweedie's Trim and Fill” method were imputed in order to obtain symmetry. The recomputed OR was stronger but the association was on the same direction (0.87; 0.65–1.16) and non-significant (Figure 9d). Discussion This review has brought to light several aspects relating to factors associated with L. infantum infection in humans in the Americas, as well as limitations in the literature consulted and gaps in the existing knowledge. For the variable gender we verified that in studies with clinical cases and in studies that used LST, male gender was positively associated with VL. However, with serological tests we verified an association in the opposite direction. Despite the fact that only the first two were statistically significant, the observation of the measures of association and the statistically significant difference in the Q test show that the difference between groups was consistent. In five out of six studies with serological tests, LST were also performed, that is, subjects were the same. Previous studies argued that hormonal or immune issues could affect the progress of infection [76], [77], and male subjects would be more likely to become symptomatic. Such assumptions could justify the results of studies with symptomatic cases; however this would not explain the difference identified between types of diagnostic methods for detecting asymptomatic infection. In a study in Teresina, Costa et al. [78] suggested that infection with L. infantum more often affects men because they are more exposed to vector bites, since they commonly do not wear shirts due to the high temperatures in that region of Brazil. This can lead to greater exposure of the body surface to the vector. Another postulated explanation for the association between male sex and infection is that men engage more frequently in household tasks or leisure-time activities outdoors, especially after normal working hours, during the peak period of vector activity [40]. For age, it became clear that children are less likely to receive a positive diagnosis of asymptomatic infection. Since LST are indicative of infections that have occurred throughout a subject's entire life [79] and serological tests may remain positive even long after the infection [14], results are within what would be expected. One should also take into account whether epidemiological or behavioral factors could be associated with identified patterns in addition to such characteristics of the tests. Results in opposite directions and with strong odds ratios verified in both case-control studies analyzed reveal that children are more prone to develop clinical disease, despite being less likely to get infected. This fact may be verified when one observes the number of symptomatic cases routinely notified by healthcare services [62]. The presence of dogs in the household was positively associated with VL. Among the variables subjected to meta-analysis, this was the one with the lowest heterogeneity. Besides the fact that the dog is considered the most important reservoir of the infection in urban settings [8], its presence is positively correlated with vector abundance, potentially increasing the risk of transmission [80]. Furthermore, even without the possibility of performing meta-analysis, ecological studies showed consistent evidence that dog seropositivity is associated with human VL. This emphasizes the relevance of dogs at least as a marker of the occurrence of visceral leishmaniasis among humans. Regarding the presence of chickens and birds in households, although there is evidence that they might have a role in attracting Phlebotomine sand flies [81], [82], according to the available evidence their presence could not be incriminated as a risk factor for VL. Unlike dogs, chickens and birds are not reservoirs of L. infantum [83] and they might even divert sandflies attention. Indeed the summary measure obtained shows that a protective role of the presence of such animals cannot be discarded, as already discussed in the literature [84]. With respect to the variables for which no subgroup analysis was performed, malnutrition decreased the likelihood of positive results, possibly due to the immunological techniques employed for the diagnosis. It should be noted, however, that these results do not exclude the role of malnutrition in increasing the risk of developing clinical disease, as already shown in other publications [85], [86]. As for the prior occurrence of VL in relatives or neighbors and its relationship with an increase in positivity, the association was stronger for the first one, the same pattern as shown in a study conducted by D'Oliveira et al. [50]. Since most studies considered as relatives only those who lived in the same household, the focal nature of the disease could explain the data presented. However, genetic factors that may be associated with the infection should not be disregarded. With respect to socioeconomic issues, although the strength of evidence was variable depending on the factor analyzed and the number of studies being small, the set of data presented shows that there is a close relationship between visceral leishmaniasis and poor living conditions, lack of basic urban infrastructure services and low levels of education. Such relationship between low socioeconomic status and the risk of VL could be explained to the extent that poor economic conditions are associated with greater abundance of vectors, with the lack of responsible pet ownership practices and with the location of houses in peripheral areas of the cities where vegetation density is favorable for the presence of vectors and, perhaps, of sylvatic reservoirs [61], [80], [87], [88]. Finally, albeit in a small number of studies, variables relating to features of the environment and vegetation (evaluated through the NDVI) demonstrated that visceral leishmaniasis occurs more frequently in areas where the level of urbanization is lower and where vegetation is more abundant, probably creating adequate habitats for breeding of vector population [89]. Although quality analysis performed in this review also has limitations arising from the fact that there is no single standardized tool recommended to evaluate observational studies [25], in particular regarding specific issues of ecological studies, several limitations and susceptibility to bias could be identified in the studies analyzed. Even though some of those might be considered inevitable, such as the lack of quality of secondary data used in ecological studies, many other limitations and biases could have been avoided if researchers had adopted the correct procedures for selecting participants, measuring variables and analyzing data, and also had they used more standardized, appropriate and transparent ways of reporting their studies. Based on data presented, the following should be considered in future studies: a) strategies to minimize refusals or losses should be strengthened, as well as other possible design-specific biases identified; b) standardized procedures to select participants should be adopted, considering the need for this selection to be random and to take into account a sample size that is appropriate so that the study has enough statistical power; c) combinations of diagnostic tests should be used so that results may be more valid and reliable; d) control for confounding should be performed and interactions between variables should be analyzed; e) analyses with continuous data in their original form should be performed, considering their advantages [90] and in cases where stratifications prove to be necessary, these should be explained and justified in terms of analysis and practical applicability and not only aiming to achieve statistical significance; f) studies should use correct statistical procedures which extend beyond the p-value, also discussing the strength and direction of associations; g) studies should use the STROBE statement so that results are better described, clearer and more standardized. A number of limitations in the present study deserve mention. Considering that control for confounding was performed in few studies and considering different factors, there may be a large amount of residual confounding in the data summarized in this review. This could mean that the strength of associations might be overestimated. The same overestimation might also have occurred due to the decision to use odds ratio as a measure of association [91], even if it was necessary so that studies which provided only p-values and sample sizes could be used, as well as for the inclusion of results from logistic regression models. Although results from analyses of publication bias showed no consistent evidence of its occurrence, it should be noted that such analyses were limited considering the existing heterogeneity among the various studies results. Furthermore, they could not be performed for most variables. Despite the fact that an extensive search for theses and dissertations was carried out, we did not include studies published in annals of scientific conferences. In addition, for some of the studies included there was significant loss of information on the association for certain variables because non-significant results were not described. Finally, one should consider the low power of some procedures carried out, as well as the limitations imposed by the fact that it was impossible to perform subgroup analyses for most variables. This caused many summary measures to be obtained for heterogeneous data. Such procedures were followed in order to avoid limited and biased analyses based on statistical significance only, such as ‘vote counting’, which we also tried to avoid when discussing the strength and direction of associations even when data allowed theoretical discussions only [28], [92]. The findings of the present review might influence and improve the design of VL control strategies. For instance, the presence of positive dogs might be a useful indicator for monitoring the force of transmission to humans and thus, areas with high canine prevalence or incidence should be prioritized when delivering interventions. Also concerning the reservoir, it should be stressed the importance of policies to promote responsible dog ownership. Additionally, identification of areas at higher risk for prioritizing intervention should take into account the abundance of green vegetation as well as the prior occurrence of VL cases. Since children are more prone to illness and they have the potential to act as information multipliers [93], they might be the focus of attention in health education initiatives. The results related to socioeconomic conditions showed that inadequate urban infrastructure should also be considered a target for interventions against VL. It is of paramount importance to emphasize the need for the formulation of public policies that will improve living conditions and access to education for the population as a whole, since visceral leishmaniasis is just one among other diseases that are rooted into scarcity and poverty [94]. Our work is the first systematic review about factors associated with visceral leishmaniasis in Latin America. Several relevant aspects that help understand the disease dynamics were clarified and new questions were raised. Patterns identified for all variables, as well as the reason for greater chance of infection among adults and greater likelihood of falling ill among children or reasons why the type of diagnostic tests modified the direction of association between gender and asymptomatic infection should be further evaluated. Future studies should investigate whether these patterns are consistent and how subjects' immune response and those characteristics of the tests employed could explain the observed data. With respect to variables about which there is little or inconsistent knowledge, such as subject's level of knowledge about the disease, ethnic group, backyard conditions and the presence of other animals, especially synanthropic and chickens, future studies should build upon the available information. Additionally, the following should also be encouraged: expanding the use of high definition satellite imaging, taking several environmental aspects into account, from backyard features to broader levels, such as those studied by Franke et al. [73], as well as the association with abundance of Phlebotomine sand flies and with the incidence of infections in human beings. Finally, it is also crucial to conduct research in other geographic areas in Brazil and in other Latin American countries where the disease is endemic. More cohort studies are required and the quality of projects, analyses and publications needs to improve dramatically. Supporting Information Table S1 Main features, risks of bias, and limitations in studies selected for systematic review. (DOC) Click here for additional data file. Text S1 Search strategies for the PubMed, LILACS, Thesis Databank of CAPES and Google Scholar. (DOCX) Click here for additional data file. Text S2 PRISMA checklist. From: Liberati A, Altman DG, Tetzlaff J, Mulrow C, Gotzsche PC, et al. (2009) The PRISMA statement for reporting systematic reviews and meta-analyses of studies that evaluate health care interventions: explanation and elaboration. PLoS Med 6: e1000100. doi:10.1371/journal.pmed1000097. (DOC) Click here for additional data file.
- Record: found
- Abstract: found
- Article: not found
Visceral leishmaniasis in Brazil: trends and challenges.
The urbanization of visceral leishmaniasis in Brazil has been related to environmental changes, migration, interaction and spread of sylvatic reservoirs and infected dogs to areas with no transmission, and adaptation of the vector Lutzomyia longipalpis to the peridomiciliary environment. From 1980 to 2005, Brazil recorded 59,129 cases of visceral leishmaniasis, 82.5% of which in the Northeast region. Visceral leishmaniasis gradually spread to other regions of the country: in 1998 these other regions reported 15% of all cases, but by 2005 this proportion had increased to 44%. From 1998 to 2005, indigenous cases were reported in 1,904 different municipalities of the country (34.2%). Reservoir and vector control pose major challenges for disease control, since there is a need for better knowledge of vector behavior in urban areas, and control activities involve high operational costs. In recent years the Brazilian Ministry of Health has supported research on the laboratory diagnosis of infection and disease in humans and dogs, treatment of patients, evaluation of the effectiveness of control strategies, and development of new technologies that could contribute to the surveillance and control of visceral leishmaniasis in the country.
Author and article information
Contributors
Paul Andrew Bates: Role: Editor
Journal
Journal ID (nlm-ta): PLoS Negl Trop Dis
Journal ID (iso-abbrev): PLoS Negl Trop Dis
Journal ID (publisher-id): plos
Journal ID (pmc): plosntds
Title:
PLoS Neglected Tropical Diseases
Publisher:
Public Library of Science
(San Francisco, CA USA
)
ISSN
(Print):
1935-2727
ISSN
(Electronic):
1935-2735
Publication date
(Electronic):
17
March
2015
Publication date Collection:
March
2015
Volume: 9
Issue: 3
Electronic Location Identifier: e0003620
Affiliations
[1
]Superintendência de Controle de Endemias, Secretaria de Estado da Saúde, Mogi Guaçu,
São Paulo, Brazil
[3
]Centre for Applied Entomology and Parasitology, School of Life Sciences, Keele University,
Keele, United Kingdom
Author notes
Conceived and designed the experiments: CC JJS JGCH. Performed the experiments: CC FECJ JGCH. Analyzed the data: CC JGCH FECJ RPB JJS. Contributed reagents/materials/analysis tools: CC JGCH FECJ. Wrote the paper: CC FECJ JGCH RPB JJS.
* E-mail:
casanovaclaus@
123456gmail.com
Article
Publisher ID: PNTD-D-14-01707DOI: 10.1371/journal.pntd.0003620
PMC ID: 4363371
PubMed ID: 25781320
SO-VID: 0025d820-7d21-46c0-9a46-363ab3260e05
Copyright statement:
Copyright @
2015
License:
This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited
History
Date
received
: 30
September
2014
Date
accepted
: 13
February
2015
Page count
Figures: 6,
Tables: 0,
Pages: 14
Funding
This paper was supported by Superintendência de Controle de Endemias (SUCEN), Coordenação
de Aperfeiçoamento de Pessoal de Nível Superior (CAPES) and Wellcome Trust. The funders
had no role in study design, data collection and analysis, decision to publish, or
preparation of the manuscript.
Categories
Subject:
Research Article
Custom metadata
Data Availability All relevant data are within the paper and its Supporting Information files.
ScienceOpen disciplines: Infectious disease & Microbiology
Data availability:
ScienceOpen disciplines: Infectious disease & Microbiology