Transcript Tularemia in Georgia - CLAS Users

Tularemia in Georgia :1946-2009
Nikoloz Tsertsvadze1, Lela Bakanidze1, Paata Imnadze1, Shota Tsanava1, Julieta Manvelyan1, Ian
Kracalik2, S. Elizabeth Rácz3, and Jason K. Blackburn2
1National
Center for Disease Control of Georgia, 9, M.Asatiani str. 0177, Tbilisi, Georgia; 2 Spatial Epidemiology and Ecology Research Lab, Department of Geography and
Emerging Pathogens Institute, University of Florida, Gainesville, FL, 32611, U.S.A.; 3 W-529 Nebraska Hall, H.W. Manter Laboratory, University of Nebraska-Lincoln, Lincoln,
Nebraska, 65855 U.S.A.
Introduction
Tularemia is a disease considered to be widely distributed across the Northern Hemisphere (Petersen 2005). Tularemia, caused by the gramnegative bacterium Francisella tularensis, is endemic in small mammals (rodents, insectivores, rabbits and hares) and transmitted by arthropod vectors
(ticks, fleas) in Georgia. Additionally, surveillance data also shown isolates from water, wheat, and avian emeses. Tularemia investigations in Georgia
began in the mid-1940's, and initially helped to establish the existence of autonomous natural foci: one in the mountainous region of Meskhet-Javakheti
and a second in the Kartl-Kakheti valley. The main reservoir is Microtus arvalis transcaucasicus (common vole) and the principal arthropod vector is
Dermacentor marginatus (sheep tick). In order to increase the capacity and efficiency of public health surveillance in Georgia geographic information
systems (GIS) were used to help identify spatial epidemiological characteristics of tularemia.
Uncovering factors related to the transmission and epidemiology of tularemia may provide critical information regarding control of the disease
to public health officials and the general population. Incorporating ecological niche models (ENMs) and spatial statistical models into investigations of
disease occurrence is one way of elucidating aspects related to a disease’s presence. Several studies have used ENMs to describe the ecological
distribution of disease causing agents (Blackburn et al. 2007, Joyner et al. 2010). One such modeling technique known as the Genetic Algorithm for Rule
Set Production (GARP) uses known species occurrence data in conjunction with environmental and climatic data to predict areas on the landscape that
may potentially support the presence of a species. Recent research in the United States has attempted to map and describe the potential geographic
distribution of F. tularensis using an ENM known as the Genetic Algorithm for Rule Set Production (GARP) and environmental data to identify shifting
patterns of the predicted distribution in relation to climatic change (Nakazawa et al. 2007). In this study researches found that climatic and
environmental characteristics may be strong predictors of the presence of F. tularensis. In addition to ENMs spatial statistical methodlogies may also be
used to identify areas that may have a higher presence of a disease. The SaTScan spatial statistic developed by Kulldorff (1997) has been used in
numerous studies to locate clusters or a high occurrence of disease (Coleman et al. 2009). Areas with a higher than expected distribution of a
pathogen may represent areas with an increased risk for infection therefore, may aid in directing public health officials in establishing control and
prevention guidelines. The purpose of this study was to describe spatial epidemiological characteristics of F. tularensis in Georgia.
Results
Spatial Analysis
The space-time permutation model in the SaTScan software package (Kulldorff et al. 2005) was used to analyze the
presence of clustering among the isolate samples collected. This method analyzes the distribution of events retrospectively to
identify space-time clustering of events in the absence of population data. SaTScan uses a series of thousands or millions of
overlapping circles up to a predetermined size to statistically identify areas, which given a certain likelihood contain a higher
number of cases in side a circle compared to outside of the circle.
Table 1. Environmental variables used during the model-building process
Environmental Variables
Annual Mean Temperature
Temperature Annual Range
Annual Precipitation
Precipitation of Wettest Month
Precipitation of Driest Month
Elevation (Altitude)
Mean Annual NDVI*
Annual NDVI Amplitude
Name
BIO1
BIO7
BIO12
BIO13
BIO14
ALT
wd0114a0
wd0114a1
* Normalized Difference Vegetation Index
Source
WorldClim (www. worldclim.org)
WorldClim (www. worldclim.org)
WorldClim (www. worldclim.org)
WorldClim (www. worldclim.org)
WorldClim (www. worldclim.org)
WorldClim (www. worldclim.org)
TALA (Hay et al. 2006)
TALA (Hay et al. 2006)
Table 2. Accuracy metrics for the potential distribution of
Dermacentor marginatus based on the GARP experiment
Metric
Current Scenario
N to build models
20†
N to test models
7
Total Omission
14.3%
Average Omission
6.0%
Total Commission
7.01%
Average Commission
3.14%
AUC*
0.9038 (z=6.02§, SE=0.0768)
The location of training and independent data is shown in figure 2. Training data were used for modelbuilding and independent data were used for model validation. The current potential distribution of F. tularensis
is shown in figure 3 and it was created in GARP using tularemia isolate data obtained from Dermacentor
marginatus and eight environmental variables shown in Table 1. Accuracy metrics were created after the
model-building process using independent data and are shown in Table 2. Areas of highest model agreement
shown in red were located in the in the central portion of the country.
The SaTScan space-time permutation model indicated the presence of statistically significant space-time
clusters (Figure 4). In total three clusters were identified: one primary cluster and two secondary clusters. The
primary cluster located in the center part of the country consisted of a radius of 7.83Km ranging in time from
2007 to 2010. The two secondary clusters were 4.85km ranging in time from 1990 to 1991 and 15.26Km
ranging in time from 1981 to 1982 respectively.
Discussion
* AUC = area under curve
† N was divided into 50% training/50% testing at each model iteration
§ p < 0.001
Francisella tularensis, while rare in humans, has an established natural ecology within Georgia. A
species-specific ecological niche model of Dermacentor marginatus, a primary vector of the pathogen, indicates
the potential geographic locations where the bacterium and vector maybe present on the landscape (Figure 3).
This modeling experiment represents only one of the potential vectors for the disease. However, this may allow
for more direct intervention strategies by targeting a single vector/host. Areas indicated by the model that
promote the occurrence of the vector and pathogen could be used by public health officials to inform vector
control and the public of potential risks. In addition to providing an increased awareness of the potential
presence of the vector/pathogen, the ENM can be utilized to direct further surveillance efforts. Areas not
previously sampled, but identified in the model as potentially harboring the vector and bacterium can be used to
direct targeted sampling/collection activities in the future.
Space-time patterns of tularemia isolates indicate relatively short temporal and distinct spatial cluster
locations durations during the 54-year study period. Clusters in the center of the country while temporally
divergent were located adjacent to each other revealing a potential localized focus of tularemia (Figure 4).
Although clusters were statistically significant they could have been a result of directed sampling efforts or
fluctuations in specific vector/host populations. The space-time permutation does not take into account the
Figure 3. The potential geographic distribution of Dermacentor marginatus in Georgia based on an
eight variable ecological niche model. The color ramp represents model agreement from the best
subset routine, with darker red colors reflecting higher model agreement, or greater confidence in the
prediction of actual bacteria habitat. Yellow and green dots represent model training and testing data,
respectively.
underlying population therefore clusters are calculated in proportion to the entire sample distribution. Despite
the lack of population estimates the identification of clusters may help target high risk areas.
Using niche modeling strategies in conjunction with spatial clustering techniques could help in identify
areas of concern. In this study SaTScan and GARP both identified similar areas of interest in the center of the
country. This combination of spatial methodologies can aid in the validation of mapping high risk areas. In
Figure 1. The distribution of F. tularensis isolates identified by the
sample origin.
addition to modeling more vectors of the bacterium, future studies could examine the potential effects of climate
Figure 2. Testing and training data of F. tularensis isolates used in the GARP
modeling algorithm.
change on the distribution of Dermacentor marginatus. Additionally, further studies should focus on modeling the
larger group of host and vector species to develop a more complete understanding of the distribution of this
pathogen and its transmission dynamics in Georgia.
References
Materials and Methods
Brotons L, Thuiller W, Araujo MB, Hirzel AH (2004) Presence-absence versus presence-only modelling methods for predicting bird habitat suitability.
Ecography 27: 437-448.
Coleman, Marlize, Michael Coleman, Aaron M. Mabuza, et al. 2009. Using SaTscan method to detect local malaria clusters for guiding control programs.
Malaria Journal. 68(8).
Hay, S.I., A.J. Tatem, A.J. Graham, S.J. Goetz, and David Rogers. (2006).Global environmental data for mapping infectious disease distribution. Advances in
Parasitology. 62:37-77.
Hijmans RJ, Cameron SE, Parra JL, Jones PG, Jarvis A (2005) Very high resolution interpolated climate surfaces for global land areas. International Journal of
Climatology 25: 1965-1978.
Joyner, T.A., L. Lukhnova, Y. Pazilov, G. Temiralyeva, M.E. Hugh-Jones, A. Aikimbayev, and J.K. Blackburn. 2010. Modeling the potential distribution of Bacillus
anthracis under multiple climate change scenarios for Kazakhstan. PLoS One.
Kulldorff, M., R. Heffernan, J. Hartman, R. Assuncao, and F. Mostashari. 2005. A space-time permutation scan statistic for disease outbreak detection. PLoS
Tularemia Occurrence Data
A database totaling 93 strains was constructed from records between 1946 and 2010 archived at the National Centers for Disease Control (NCDC),
Tbilisi, Georgia. Samples were collected across 11 rayons and obtained from several sources (Figure 1). Field sites were surveyed by scientists from NCDC
and samples collected via X,Y latitudinal and longitudinal pairs georeferenced to the nearest village. Diagnoses were carried out serologically, specifically, by
identifying an antibody titer in the blood serum of a patient. Isolation of tularemia causative agent was performed, using bioprobes and on McCoy egg-yolk
medium. Improvements in diagnostics now include isolation of strains with cysteine-added chocolate agar, ELISA, and RT-PCR techniques.
Medicine.
Environmental and Climate Variables data
Nakazawa Y, Williams R, Peterson AT, Mead P, Staples E, et al. (2007) Climate change effects on plague and tularemia in the United States. Vector-Borne and
Zoonotic Diseases 7(4): 529-540.
Petersen, Jeannine M., and Martin E. Schriefer. 2005. Tularemia: emergence/re-emergence. Vet Res. 36:455-467.
Peterson AT, Sanchez-Cordero V, Beard CB, Ramsey JM (2002) Ecologic niche modeling and potential reservoirs for chagas disease, Mexico. Emerging
Infectious Diseases 8(7): 662-667.
Smith KL, DeVos V, Bryden H, Price LB, Hugh-Jones ME, et al. (2000) Bacillus anthracis diversity in Kruger National Park. Journal of Clinical Microbiology
38(10): 3780-3784.
Stockwell DRB, Peters D (1999) The GARP modelling system: problems and solutions to automated spatial prediction. International Journal of Geographical
Information Science 13(2): 143-158.
Sweeney AW, Beebe NW, Cooper RD, Bauer JT, Peterson AT (2006) Environmental factors associated with distribution and range limits of malaria vector
Anopheles farauti in Australia. Journal of Medical Entomology 43(5): 1068-1075.
Current climate grid data were freely downloadable (www.worldclim.org) on the WORLDCLIM website (Hijmans et al. 2005) or provided by (Hay et
al. 2006). A resolution of 8 km was utilized for this study because village latitude and longitude coordinates were occasionally estimated to be greater than 1
km away from farms where a tularemia isolate was obtained.
Ecological Niche Modeling and Spatial Modeling
For this study, one modeling scenario was employed to examine the current geographic distribution of F. tularensis. The scenario contained eight
environmental variables that described temperature, precipitation, vegetation, and elevation to create a model of the potential current distribution of
F. tularensis.
We used the GARP ecological niche modeling program to build the current predictive model for Georgia. GARP is a presence-only genetic algorithm that
models species’ potential geographic distributions through an iterative process of training and testing that occurs through resampling and replacement of
input data (Stockwell and Peters 1999). A pattern matching process is applied that finds non-random relationships between species localities and specific
variables that describe the environment. These relationships are written as a series of if/then logic statements (known as rules) that define whether
conditions within the rule are defining presence or absence.
Figure 4 SaTScan space-time permutation model identifying significantly high space-time
clusters. The red circle represents the primary cluster identified by SaTScan with a radius of
7.83Km and lasting from 2007 to 2010. Secondary clusters are shown in blue 15.26Km (1981
to 1982) and in black 4.85Km (1990 to 1991).
Acknowledgements
This Cooperative Biological Research project was funded by the United States Defense Threat Reduction Agency (DTRA) as part of the
Biological Threat Reduction Program in Georgia. UF funding is administered through the Joint University Partnership under the
University of New Mexico.