Introduction Leprosy is a chronic granulomatous infectious disease caused by the obligate intracellular organism Mycobacterium leprae that affects mainly the skin and peripheral nerves, which can lead to severe physical disabilities and deformities if not diagnosed and appropriately treated with multidrug therapy (MDT) in its early stages. Evidences suggest that M. leprae can spread from person to person through nasal and oral droplets and this is considered to be the main route of transmission, especially among household contacts of untreated multibacillary (MB) patients. M. leprae multiplies very slowly (12–14 days) and the mean incubation period of the disease is about three to five years, but symptoms can take as long as 30 years to appear. Early detection and properly MDT treatment are the key elements of leprosy control strategy [1]. Although leprosy has been successfully suppressed in developed countries, 219,075 new cases in 105 countries were detected in 2011, as reported to the World Health Organization (WHO), with India, Brazil and Indonesia contributing 83% of all new cases [2]. Brazil, with 33,955 new cases detected in 2011 (according to the official numbers of the Brazilian Ministry of Health), has one of the highest annual case detection rates in the world (17.65/100,000 people), and the prevalence rate has yet to be reduced below the threshold of 1/10,000 people – the level at which leprosy would be considered “eliminated” as a public health problem [2]. The spatial distribution of leprosy in Brazil is heterogeneous: the more socioeconomically developed states in the south have achieved the elimination target, though high-disease burden pockets still remain in North, Central-West and Northeast Brazil [3]. These high-burden areas encompass 1,173 municipalities (21% of all Brazilian municipalities), approximately 17% of the total national population and 53.5% of all Brazilian leprosy cases detected between 2005 and 2007 [4]. Most of the areas with spatial clusters of cases are in the Brazilian Amazon, long recognized as a highly endemic leprosy area [3]–[6]. More than 7.5 million people live in the state of Pará, located in the Amazon region. This state is hyperendemic for leprosy both among the general population (51.1/100,000 people) and among children 1.000), similar to that observed in multibacillary patients, were dwelling within 100 meters of at least one leprosy case, consistent with the uncovered and upcoming spatio-temporal associations. 10.1371/journal.pntd.0002665.g004 Figure 4 Space-time links among cases and proximity to students. An expanded view of a specific region identified as a cluster of leprosy (see Figure 2C, Kulldorff's spatial scan statistics), showing the space-time links among cases and the spatial relationship with a surveyed school and seropositive students. 10.1371/journal.pntd.0002665.t002 Table 2 Knox space-time clustering analysis for leprosy cases.* Space-time lag (meter-years) Number of space-time links Number of cases p-value (999 Monte Carlo simulations) 50 - 1 56 91 0.013 50 - 2 69 108 0.012 100 - 1 176 226 0.010 100 - 2 224 259 0.012 100 - 3 270 289 0.019 100 - 4 296 307 0.011 200 - 2 663 406 0.009 * Only statistically significant space-time lags are shown here (p<0.05). Total number of analyzed cases = 499. Discussion The pattern of leprosy cases reported from 2004 to 2010 in Castanhal showed significant spatio-temporal heterogeneity, and we found spatial clusters of high and low detection rates in the urban area. Using spatial global tests, we were also able to determine that the spatial autocorrelation of both the raw detection rate at the census tract level and of individual cases occurred at fine temporal and spatial scales. According to an analysis of the spatial pattern of serological data obtained by testing students, we ascertained that children with a high serological titer of anti-PGL-I were in close proximity to spatial-temporal clusters of leprosy cases. These findings can be applied to guide leprosy control programs to target intervention to locations with the highest risk of leprosy. De Souza Dias and colleagues [20] described the first application of GIS tools to direct active case-finding campaigns at a fine geographic scale in Brazil [20] and were able to target hot spots, resulting in the enhanced detection of new cases in addition to realizing important cost reductions for leprosy control activities. The surprisingly high previously undiagnosed prevalence of leprosy and of subclinical infection with M. leprae among school children can be explained by the close proximity of these students' homes to detected cases. It has been shown that, in addition to household contacts, people living in the vicinity of a leprosy case and their social contacts have a higher risk of infection [18], [26], [37]. In fact, because M. leprae is highly infective but has a low pathogenicity, most people who harbor a subclinical infection will never develop clinical signs and symptoms of leprosy; indeed, only about 10% of all infected individuals eventually develop leprosy symptoms [38]. Due to the slow doubling time (13 days) and long incubation period prior to the onset of frank disease symptoms (3–5 years or longer), it is likely that many hidden cases exist, although serological responses to some protein antigens have been shown to predict disease progression up to a year prior to diagnosis [39]–[43]. It has been well-established that the titer of anti-PGL-I IgM antibody is directly correlated to the bacillary index, and that very high titers to PGL-I and certain protein antigens, such as LID-1 and Ag85B (ML2028) indicate a greater risk of developing disease [27], [40], [43]. The main challenge is to discover which biomarkers of infection serve as the best predictors of who will succumb to disease. Accordingly, performing targeted surveillance on individuals living in high endemic areas and following individuals with a high titer of anti-PGL-I is a strategy that must be implemented to perform early diagnosis, prevent physical disabilities and break the chain of transmission. A number of serological surveys have shown that the rate of anti-PGL-I seropositivity in endemic settings correlates well with leprosy incidence in the community [44], [45]. All of the surveyed schools in this study were located in the hyperendemic census tracts of the city. This finding explains the absence of significant differences in the seroprevalence or in the titer of antibodies in the students based on a geographic location, given that nearly all (95%) of them were living within 200 meters of a detected leprosy case. As observed for the students, there were no differences in the titer of anti-PGL-I or seroprevalence among the household contacts living inside or outside a cluster of cases. This is also not surprising, given that, even outside a cluster, all household contacts were living in very high or hyperendemic areas and that the most likely source of M. leprae is a close contact that shares the same house or room. Indeed, when 942 students and 58 teachers from Castanhal were asked if they knew a person affected by leprosy, 17.7% of the students and 53.4% of the teachers answered in the affirmative. In addition to this proximity, those harboring a subclinical infection could be a potential source of contamination to others [46], rendering such frequent-, intensive- and close-social-contact environments, such as households and schools, as locations that are favorable for M. leprae transmission. Considering its total area, the Brazilian Amazon region has the lowest population density (4.12 individuals/km2) in the country but the highest number of people per household (3.97). This is a direct result of poverty, which compels relatives and others to live together for long periods of time, especially young married couples and their children, typically under precarious sanitation conditions. Furthermore, the average household density was even higher in the residences with a leprosy case (5.0), and, for purpose of comparison, this population density per square kilometer within the cluster of leprosy (9,536/km2 – Figure 2C) was as high as New York City (10,429/km2 - http://www.census.gov). Within the context of the wide recognition that high levels of crowding facilitate the transmission of infectious disease [47], it is reasonable to suggest that improvements in the socioeconomic status and living conditions should be part of the overall leprosy control strategy. The introduction of GIS to leprosy epidemiology brought new insight to the concept of defining contacts based on relative distance. The importance of performing periodic surveillance among household contacts and including different classes of social and neighboring contacts has been highlighted by several authors [33], [37], [48]. Bakker and colleagues [18] observed increased subclinical infection for contact groups living ≤75 meters of anti-PGL-I-positive leprosy patients. Another report described that 92% of the dwellings of contacts were within a distance of 100 meters of the index patient [33]. For this study, we selected radii of 50, 100 and 200 meters and observed significant space-time clusters within all of these distances. Leprosy was also found to exhibit a clustered spatio-temporal pattern in an analysis of more than 11,000 cases for a period of 15 years in Bangladesh [49], with most clusters having a duration of 1 or 2 years and one cluster a 4-year time span. In our study, we observed significant spatio-temporal clustering, even within a very fine geographic scale, which is compatible with direct human-to-human transmission. Most of the students diagnosed with leprosy (8 of 9) lived in close proximity to previously detected cases. A spatially empirical Bayes smoothed case detection rate has been used in leprosy studies to smooth the random variations in small areas with few people (where small variations in the number of cases results in dramatic changes in disease rates) and to enhance the visualization of spatial patterns [17], [50]–[52]. Smoothing is also a way to estimate uncertain values for areas with no registered cases, areas where disease is not necessarily absent but may not have been detected due to operational limitations. Smoothing produced a clearer map of leprosy in Castanhal but increased the estimate of the number of people to be followed to detect one case. We agree with Odoi and colleagues [23] that the results obtained using spatial smoothing need to be treated with caution because they can mask large differences between neighboring regions. Given that 71 (12.5%) cases in the urban area were not mapped and analyzed in this study and considering the high prevalence of undiagnosed cases in Castanhal, our data strongly supports the notion that many more individuals than those presented here, including many children <15 years old, are currently infected with M. leprae. In the last decade, spatial analysis and GIS have become important tools for understanding leprosy transmission dynamics in resource-poor countries. Different spatial statistical methods have been applied, including Kulldorff's spatial scan statistics [53] and global and local Moran's I indices of spatial autocorrelation [54]. However, because all spatial statistics have advantages and disadvantages, more than one method may be necessary to analyze the data and to enable decision makers to determine the priority areas for targeting control activities. Overlaying individual case point maps over high-resolution satellite images from high-risk areas (not shown here to protect the individual addresses) provides a clear visualization of the leprosy problem and can help to optimize active case-finding strategies and plan further clinical, epidemiological and prophylactic studies. Additionally, combining clinical, epidemiological, serological and spatial data provided a better understanding of the transmission dynamics of leprosy at fine spatial scales and indicated high rates of childhood leprosy transmission within hyperendemic cities of the Brazilian Amazon region. Supporting Information Checklist S1 STROBE checklist. (PDF) Click here for additional data file. Figure S1 Correlogram of global Moran's I for the detection rates of leprosy by census tract in the urban area. Significant (p<0.01) spatial autocorrelation of the census tracts with the high or low raw detection rate of leprosy per 100,000 people. Taking into account the location of the census tract centroids, the most significant (p<0.01) clustering distance was between 1 and 2 km (peaking at 1.5 km). (TIF) Click here for additional data file. Figure S2 Multi-distance spatial cluster analysis (Ripley's k-function). There is significant clustering of individual cases starting at a distance of 50 meters (p<0.01), indicating that cases tend to be detected in close spatial proximity. (TIF) Click here for additional data file.
