Mem Inst Oswaldo Cruz, Rio de Janeiro, 113(1) January 2018
Assessing the risk zones of Chagas’ disease in Chile, in a world marked by global climatic change
1Universidad de Chile, Facultad de Medicina, Escuela de Salud Pública, Programa de Salud Ambiental, Santiago, Chile
2Universidad de Chile, Facultad de Ciencias Veterinarias y Pecuarias, Departamento de Ciencias Biológicas Animales, Santiago, Chile
3Universidad Metropolitana de Ciencias de la Educación, Departamento de Entomología, Santiago, Chile
4Ministerio de Salud, Control de Vectores, Santiago, Chile
5Universidad de Chile, Facultad de Medicina, Laboratorio de Parasitología, Santiago, Chile
6Universidad de Chile, Facultad de Medicina, Departamento de Medicina, Santiago, Chile
BACKGROUND Vector transmission of Trypanosoma cruzi appears to be interrupted in Chile; however, data show increasing incidence of Chagasu2019 disease, raising concerns that there may be a reemerging problem.
OBJECTIVE To estimate the actual risk in a changing world it is necessary to consider the historical vector distribution and correlate this distribution with the presence of cases and climate change.
METHODS Potential distribution models of Triatoma infestans and Chagas disease were performed using Maxent, a machine-learning method.
FINDINGS Climate change appears to play a major role in the reemergence of Chagasu2019 disease and T. infestans in Chile. The distribution of both T. infestans and Chagasu2019 disease correlated with maximum temperature, and the precipitation during the driest month. The overlap of Chagasu2019 disease and T. infestans distribution areas was high. The distribution of T. infestans, under two global change scenarios, showed a minimal reduction tendency in suitable areas.
MAIN CONCLUSION The impact of temperature and precipitation on the distribution of T. infestans, as shown by the models, indicates the need for aggressive control efforts; the current control measures, including T. infestans control campaigns, should be maintained with the same intensity as they have at present, avoiding sylvatic foci, intrusions, and recolonisation of human dwellings.
Chagas' diseaseis one of the most prevalent, yet neglected, diseases in the Americas and anemerging disease in other locations throughout America and Europe. It has beencompared to the early stages of the HIV/SIDA pandemic. Annual incidence variesfrom 28,000 to 56,000 individuals, with 10,000 to 14,000 annual deaths (Hotezet al. 2012), affecting 6-11 million individuals (Cucunubá et al. 2016 12 ),but with 65 to 100 million at risk (MINSAL 2014 23 , 2016). In Chile, the endemicarea is located between the Arica-Parinacota (18º30'S) and Libertador BernardoO'Higgins (34º36'S) regions, where approximately 900,000 individuals remainat risk (MINSAL 2014 23 ).
Chagas' diseaseis a protozoan infection caused by Trypanosoma cruzi, transmitted inChile by the kissing bug vectors Triatoma infestans, Mepraia spinolai,Mepraia gajardoi and Mepraia parapatrica (Hemiptera, Reduviidae,Triatominae); T. infestans is the domiciliary vector (Apt & Reyes1986a, b, Canals et al. 1992 9 , Botto-Mahan et al. 2005 5 , Frías-Laserreet al. 2017) and the main species responsible for the prevalence of this diseasein Chile and in the Americas (Canals et al. 1993 10 ). The protozoan can also betransmitted via transfusion and congenital, oral, and accidental routes; however,these are of minor importance in Chile (Nóbrega et al. 2009 24 , MINSAL 2014 23 ).
The last nationalhealth report (ENS) indicated a population prevalence of 0.7%, with 1.5% and0.6% in rural and urban zones, respectively (MINSAL 2009-2010 21 , 2016), and avery low domiciliary infestation rate (MINSAL 2016 22 ). The data in these reportsare in contrast with those reported in the 1980s and 1990s. For example, between1937-1980 the prevalence was 16.7% in rural endemic zones with a maximum of43.6% in the Coquimbo region (Schenone et al. 1980 30 ). This did not vary in the1982-1985 period (Schenone et al. 1985 28 ), while between 1982-1989, a prevalenceof 1.9% was reported in urban zones (Schenone & Rojas 1989). Previouslyreported domiciliary infestation data indicated rates between 26.8% and 33.2%between the Arica and Libertador Bernardo O'Higgins regions (Schenone et al.1980, MINSAL 2016 22 ).
Since 1999 in Chile,1997 in Uruguay, 2006 in Brazil, and recently (since 2016) in Paraguay, vectortransmission by T. infestans has been interrupted as a consequence ofefficient eradication campaigns (MINSAL 2014 23 , Rojas de Arias 2016 27 ). This interruptioncould have changed the dynamics of Chagas' disease from vector to congenitaltransmission, with consequences in the reproductive number (R0), prevalence,incidence and trypano-triatomine indices (Massad 2008 20 , Rojas de Arias 2016 27 ,Canals et al. 2017a 8 , b). This creates a false impression that Chagas' diseaseis not a problem in these countries, which has consequences on efforts madetoward prevention and control. Thus, this disease is neglected (Hotez et al.2012, Rojas de Arias 2016 27 ).
There are somedata in Chile that show increasing incidence (MINSAL 2016 22 ) and reports of sylvaticfoci of the main vector, T. infestans (Bacigalupo et al. 2010 3 , Canalset al. 2017b). In this scenario, it is difficult to obtain an accurate impressionof the real risk of Chagas' disease in this country.
Moreover, the worldis experiencing social, political and climatic changes that may have consequenceson the distribution of vectors and the prevalence of infectious diseases. Aparadigmatic example is the variation in the distribution of Aedes aegyptiand, consequently, on the distribution of dengue between 1930 and the present.This mosquito initially inhabited a large region throughout the Americas andthe Caribbean; its distribution was reduced to some Caribbean zones in 1970.However, currently, the distribution is similar to, or greater than, the originaldistribution (Gubler 2008 17 ), including northern Chile, with an explosive incrementin dengue in the last few years. For this reason, it is necessary to considerthe historical distribution of the vectors (ecological maps) and correlate thesedistributions with cases (incidence-based maps) to estimate the risk of vector-bornediseases in a changing world.
In this article,we assess the historical distribution of T. infestans, the main and domesticvector of Chagas' disease in Chile, comparing current distribution with theoriginal distribution where the dominant transmission form was vectors. We thenexplore the changes in the distribution of vectors under two climate changescenarios.
Vector occurrencedata - Data on the occurrence of T. infestans in Chile were obtainedfrom the National Museum of Natural History, Entomology Institute of MetropolitanUniversity of Educational Sciences, Health Ministry of Chile, Public HealthInstitute and literature reports, covering the time-period from 1943, onwards.Duplicate and erroneous records were excluded. Records in oversampled locationswere also excluded based on subsampling among very close pairs of points, toreduce sampling bias (Peterson et al. 2008 26 ). Points of occurrence were separatedby at least 1 km. Data were filtered following the criteria: (1) the informationmust be accurately geo-referenced (2) the data must include the name of thezoologist who determined the species, to avoid taxonomic problems. A total of222 occurrences were considered and after filtering we obtained a total of 110points of occurrence.
Chagas' diseaseoccurrence data - We obtained occurrence data for Chagas' disease from thearchives of the Parasitology Laboratory of the Medicine Faculty of the Universityof Chile. Records between 1939-1965 were considered because, during this period,the transmission was mainly by T. infestans (Schenone et al. 1980 30 ). Thedata were organised in an Excel file, geo-referencing the location (if the informationwas available). In Chile, there are regional administrative divisions; eachregion is divided into communes. When the exact address was not recorded, thegeometric centre of the residence commune was geo-referenced. Geo-referencingwas done using geographic coordinates Datum WGS 84 (World Geodetic System).This method considers some spatial uncertainty (sensu Peterson &Samy 2016). To estimate this, we used the average radius over all communes Su= ?ri/nwith ri the radius of a circumference, with equal area of each commune"i". This was a broad estimate of the spatial uncertainty. Also, similarly tovector occurrence data, error and oversampled locations were avoided. A totalof 3395 cases were considered, and after filtering, 219 cases were includedin the model.
Climatic andenvironmental data - The dataset of environmental variables was composedof proxy bioclimatic variables obtained from the Worldclim database (http://www.worldclim.org/),with two spatial resolutions: 3 arc-sec (?1km2); this dataset included a total of 19 bioclimatic variables thatsummarised temperature and precipitation data. Seven variables were selectedfrom this set, considering their relationship to the distribution of Triatomines:B1 = mean annual temperature; B4 = temperature seasonality (standard deviation*100);B5 = maximum temperature in the warmest month; B6 = minimum temperature in thecoldest month; B12 = annual precipitation; B14 = precipitation during the driestmonth and B15 = precipitation seasonality (coefficient of variation). Thesevariables were chosen considering that T. infestans populations are affectedby temperature and precipitation, including two variables that consider loads(B1 and B12) and information regarding the deviations of these variables (B4and B15). We also included extreme temperature variables (B5 and B6) becauseextreme temperatures exert known effects on development, dispersion, and mortalityof T. infestans (Canals et al. 1992 9 , 2016).
Global changescenarios - The same variables used previously, but for 2070 (average between2061-2080), were obtained from the Worldclim database for global climate model(GCM) in two global change scenarios: optimistic (RCP: 2.6) and pessimistic(RCP: 8.5), performed by the Instituto Nacional de Pesquisas Espaciais (INPE)of Brazil with a resolution of 2.5 arc-min (?5km2).
Analyses- Spatial distribution models were constructed for occurrences of the vectorand Chagas' disease cases using Maxent, a machine-learning method that assessesthe distribution probability of a case or species by estimating the maximumentropy probability distribution; it is a proven method with very good results.We used bootstrap subsampling with 30 replicated and random seeds, and the meanof replicates. The model was smoothed to avoid over-parametrisation (Petersonet al. 2008). The Maxent output was converted to binary maps using an errorrate of 10%.
Considering that:(1) The southern boundary of the distribution of T. infestans in Chileis not well known and was only established in a single, dated study; (2) Humansaid this insect's dispersal capacity, and there are reports of colonies of theseinsects in trains (Faundez 2016); (3) The aim of this study was to compare thedistribution of patients with Chagas's disease; there are permanent ministerialreports of cases of Chagas' disease in latitudes further to the South (Bio-Bio,Araucania and De los Ríos Regions: 36º-40º33'S). The modelswere not calibrated based on hypotheses of accessible areas "M" for these vectors(Barve et al. 2011 4 ). Thus, for comparative purposes, we extended the model tothe whole Chilean national territory. To study the model goodness-of-fit, weused the area under the curve (AUC) in the receiver operating characteristic(ROC) analysis.
To compare thecurrent distribution model of T. infestans to the distribution modelfor cases of Chagas' disease and the distribution model of T. infestans,under the optimistic and pessimistic global change scenarios, a reclassificationfunction was used in DIVA-GIS: 1 for a probability of occurrence > 0.2 and0 otherwise. This procedure allowed estimation of the suitable areas (km2),and the areas of superposition.
The distributionmodels for Chagas' disease and for T. infestans showed a good fit (AUC= 0.957 ± 0.005 and AUC = 0.954 ± 0.010, respectively). For bothmodels the maximum temperature in the warmest month and precipitation in thedriest month contributed considerably to the distribution. For the Chagas' diseasemap, annual precipitation, temperature seasonality, and average temperaturewere also relevant (Table). A broad estimationof the spatial uncertainty was Su = 17.7 ± 18.0 km.
The jackknife methodshowed that, for Chagas' disease, precipitation seasonality and maximum temperatureduring the warmest month were the best predictor variables. For T. infestansdistribution, these were maximum temperature in the warmest month and the meanannual temperature (Fig. 1), accounting for more than 0.85in the AUC. The potential distributions of Chagas' disease and T. infestanswere similar (Fig. 2). The area occupied by Chagas'disease was 109,034 km2 and the T. infestans distributionwas 90,829 km2 with an overlap of 67796 km2, representing51.33% of the total area (Fig. 3).
The distributionof T. infestans under the two climatic change scenarios studied showedlow variation with a minimal reduction tendency in suitable areas (Fig.4). In the benign scenario, the suitable area was 99.33% of the currentarea, and in the pessimistic scenario it was 93.64% of the current area, withoverlap percentages of 92.01% and 91.67%, respectively.
The distributionof cases of Chagas' disease covers an area slightly larger than the distributionarea of T. infestans. This is an expected result because of internalpopulation migratory movements. The distribution of T. infestans is consistentwith that usually reported for this species with a southern limit in the O'Higgins region. For Chagas' disease cases and for T. infestans, the zoneswith high presence probability were Antofagasta, Coquimbo, Valparaiso, and theMetropolitan region (Santiago), in agreement with frequently-reported data (Schenoneet al. 1980, Apt & Reyes 1986a, b, Schenone & Rojas 1989) and prevailingin Mediterranean zones in the interior valleys, which feature arid and semi-aridweather. For example, recent data show that the highest incidence was reportedin Coquimbo, which along with Antofagasta, Coquimbo, and the Metropolitan regionencompass approximately two-thirds of the notified cases (MINSAL 2014 23 , 2016).Although the Ministry of health reported a prevalence of 0.1 to 0.5/100000 habitantsin latitudes that are farther south (Bio-Bio, Araucania and De los RíosRegions: 36ºS-40º33'S) where T. infestans is not reported,our data and suitable distribution maps do not support this report, suggestingthat those cases likely represent immigrants from endemic zones.
Our results agreewith the potential distribution maps reported for T. infestans, basedon surveys (Hernández et al. 18 2013). Hernández et al. 18 (2013) reportedthat the presence probability distribution increases towards the North (CoquimboRegion), tending to cover the central-coastal region, and avoiding areas ofthe Andes Range. Also, they proposed a tendency towards lower probabilitiesin the South, near the Pacific coast. Our results are consistent with observedtendencies for avoiding the Andes Range; although the Coquimbo, Valparaiso andMetropolitan regions had the most suitable areas, the presence probability wasmore concentrated in the central zone. This has two possible explanations. Thefirst is that Hernández et al. 18 (2013) worked with surveys exploring thepresence of the two insect vectors (T. infestans and M. spinolai),and we only were interested in T. infestans, the domestic vector. Thesecond is that there may be a sample bias in our occurrence data, since thecentral zone of Chile is always the most studied due to the good climatic conditionsand, in contrast to the amplitude of the northern desert, prevents good samplingof the area.
The overlap betweenT. infestans and the Chagas' disease cases was 51.33% considering areaswith a presence probability greater than 0.2 (threshold) as "suitable". Thisestimation is sensitive to the threshold value used; the lower the threshold,the higher the overload percentage. Comparing the distribution maps of Chagas'disease cases and T. infestans with empirically reported occurrences(Fig. 2), the threshold value 0.2 appears tobe reasonable because the occurrences were distributed mainly in zones withpresence probabilities greater than 0.2 (green, orange and red areas in Fig.2). There was an area of complete overlap (51.33%), areas where T. infestansis probable but there are no cases of Chagas' disease (17.44%), and areas whereChagas' disease cases are probable but the areas are not suitable for T.infestans (31.22%). The suitable areas for T. infestans, but notfor Chagas' disease, are zones with probable underreporting, with low humandensity, or with low bug population density. For example, the Sierra Gorda andCalama communes where the reports of domiciliary infestation were fewer than1% during a period where there were few control efforts (Burchard et al. 6 1984).The zones with Chagas' disease cases, but without T. infestans, are probablyzones with subsampling of bugs. The zones of least coincidence were Tarapacáand Antofagasta, particularly the latter which is a desert zone of low populationdensity with undersampling of bugs and underreporting of Chagas cases.
The climate changepredictions in Chile include an increase of temperature over the entire nation,with a gradient of higher to lower temperatures, from north to south and fromthe Andes to the Pacific Ocean. This increase in temperature is less than expectedconsidering the predicted rates of mean global warming. During the period of2011-2030, the temperature increase would be about 0.5ºC in the south zoneand 1.5ºC for the north and the Chilean Altiplano. During the period of2031-2050, the warming pattern would be maintained; however, the incrementswould be greater (up to 2ºC). It is expected that the greatest warmingwould be at high altitudes in the Andes Range of northern Chile. In the coast,warming will be modest (0.5 to 1ºC) and may increase up to 5ºC inthe Andes (Garreaud 2011). A decrease in precipitation, between 5% and 15%,is expected during the period of 2011-2030 between the basins of the Copiapoand the Aysen Rivers (27ºS-45ºS). This decrease in precipitation willintensify during the 2031-2050 period (Garreaud 2011). Our results suggest thatthese changes will only produce a small effect on the distribution of T.infestans, with a slight reduction in suitable areas. This is consistentwith the decrease in suitable areas proposed for the species M. spinolaiin the same area, and is in contrast with the high impact on the distributionof M. gajardoi, a species with a small distribution in the coast of northernChile (Garrido 2017 16 ). Under the assumption of niche conservatism, the latterspecies would suffer disappearance of its habitat, while M. gajardoi,a species with distributions similar to T. infestans and with similarpreferred environmental conditions, would decrease its distribution area inthe interior valleys while increasing its distribution on the coast (Garrido2017), like T. infestans. Since T. infestans is a species residingin arid and semi-arid habitats, its distribution area would not be affectedsignificantly, maintaining the transmission risk of Chagas' disease in thiszone. Thus, suitable areas for the development of sylvatic foci and human dwellingintrusions in Chile will be maintained under climate change. The campaigns forcontrol of T. infestans should be maintained with the same intensityas they exhibit at present, avoiding sylvatic foci, intrusions, and re-colonisationof human dwellings.
VT-G - Definitionof problem, design, statistical analyses and discussion; DPF, SA and DC - statisticalanalyses and discussion; AM - statistical analyses, georeferentiation and discussion;DF-L, CRG and AP - data collection Triatoma infestans; LC - data collectionChagas disease and digitalisation; WA - data collection Chagas disease and discussion;MC - definition of problem, design, statistical analyses and discussion.