(C) PLOS One
This story was originally published by PLOS One and is unaltered.
 . . . . . . . . . .



Climate-driven models of leptospirosis dynamics in tropical islands from three oceanic basins [1]

['Léa Douchet', 'Entropie', 'Ird', 'Univ Reunion', 'Cnrs', 'Ifremer', 'Univ Nouvelle Calédonie', 'Nouméa', 'New Caledonia', 'Espace-Dev']

Date: 2024-05 

Abstract Background Leptospirosis is a neglected zoonosis which remains poorly known despite its epidemic potential, especially in tropical islands where outdoor lifestyle, vulnerability to invasive reservoir species and hot and rainy climate constitute higher risks for infections. Burden remains poorly documented while outbreaks can easily overflow health systems of these isolated and poorly populated areas. Identification of generic patterns driving leptospirosis dynamics across tropical islands would help understand its epidemiology for better preparedness of communities. In this study, we aim to model leptospirosis seasonality and outbreaks in tropical islands based on precipitation and temperature indicators. Methodology/Principal findings We adjusted machine learning models on leptospirosis surveillance data from seven tropical islands (Guadeloupe, Reunion Island, Fiji, Futuna, New Caledonia, and Tahiti) to investigate 1) the effect of climate on the disease’s seasonal dynamic, i.e., the centered seasonal profile and 2) inter-annual anomalies, i.e., the incidence deviations from the seasonal profile. The model was then used to estimate seasonal dynamics of leptospirosis in Vanuatu and Puerto Rico where disease incidence data were not available. A robust model, validated across different islands with leave-island-out cross-validation and based on current and 2-month lagged precipitation and current and 1-month lagged temperature, can be constructed to estimate the seasonal dynamic of leptospirosis. In opposition, climate determinants and their importance in estimating inter-annual anomalies highly differed across islands. Conclusions/Significance Climate appears as a strong determinant of leptospirosis seasonality in tropical islands regardless of the diversity of the considered environments and the different lifestyles across the islands. However, predictive and expandable abilities from climate indicators weaken when estimating inter-annual outbreaks and emphasize the importance of these local characteristics in the occurrence of outbreaks.

Author summary Tropical islands are particularly vulnerable to leptospirosis outbreaks. Hot and rainy climate, abundance of reservoir species and outdoor lifestyle contribute to the high risk for human infection. These isolated areas also deal with difficulties associated with diagnosis because of low awareness of the medical staff, non-specific symptoms of leptospirosis and limited availability of laboratory testing. Leptospirosis remains poorly documented, and a better understanding of its dynamics and its climate drivers would help improve awareness and preparedness of the public health services. In this study, we provide a climate-based model of leptospirosis seasonal dynamics in 7 tropical islands. The use of climate variables from publicly available satellite data makes the model expandable to predict leptospirosis seasonal dynamics in other tropical islands where the disease is not routinely monitored. This study emphasizes the importance of rainfall and temperature in driving the seasonality of leptospirosis in tropical islands. However, climate alone did not appear to not be a sufficient indicator to predict interannual variations, suggesting that the risk of leptospirosis outbreaks must be refined, considering local specificities as the lifestyle and the very local environment.

Citation: Douchet L, Menkes C, Herbreteau V, Larrieu J, Bador M, Goarant C, et al. (2024) Climate-driven models of leptospirosis dynamics in tropical islands from three oceanic basins. PLoS Negl Trop Dis 18(4): e0011717. https://doi.org/10.1371/journal.pntd.0011717 Editor: Dileepa Ediriweera, University of Kelaniya Faculty of Medicine, SRI LANKA Received: October 12, 2023; Accepted: April 5, 2024; Published: April 25, 2024 Copyright: © 2024 Douchet et al. 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. Data Availability: Precipitation and land surface temperature data are publicly available and can respectively be found at https://gpm.nasa.gov/missions/GPM and https://modis.gsfc.nasa.gov/data/dataprod/mod11.php. The leptospirosis raw data were provided by several countries who have not provided permission for it to be publicly shared. For New-Caledonia, the leptospirosis data can be requested by writing to the department of Health and Social affairs of New Caledonia (dass@gouv.nc). For Guadeloupe, a demand should be addressed to the Regional Office of the French Institute for Public Health surveillance Antilles-Guyane (antilles@santepubliquefrance.fr). Data from Reunion Island and Mayotte can be requested through the the French National Reference Center for leptospirosis (spiroc@pasteur.fr). Data for Tahiti can be requested to the Direction of Health of French Polynesia,Health surveillance Office (veille.sanitaire@administration.gov.pf). For Futuna, data may be asked to the Wallis & Futuna health authorities upon request to Deputy Director Public Health, Agence de Santé et l'Hopital de Sia, B.P. 4G Mata'Utu, 98600 UVEA, Wallis (da.sp@adswf.fr). Leptospirosis data for Fiji can be requested to the Ministry of Health by filling a data request form (https://www.health.gov.fj/wp-content/uploads/2014/05/Data-Request-Form.pdf) and sending it to the contact provided on the form. Funding: This study was funded by Pacific Fund (Fonds de coopération économique, sociale et culturelle pour le Pacifique, french government)(M.M., C.M.) and ECOMORE 2 (funded by the Agence Française de Développement and coordinated by Institut Pasteur)(V.H., M.M., C.M.). This project has received funding from the European Union's Horizon 2020 research and innovation program under the Marie Sklodowska-Curie Grant agreement No101027577 (M.B.). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Competing interests: The authors have declared that no competing interests exist.

Introduction Leptospirosis, a bacterial infection caused by a spirochete of the genus Leptospira, is one of the most widespread zoonotic diseases causing over 1 million cases yearly [1]. Most cases are reported in tropical regions with about 70% of the annual cases under these latitudes [1]. Despite a burden comparable to schistosomiasis, leishmaniasis or lymphatic filariasis, leptospirosis remains neglected and receives insufficient research attention [2]. Asymptomatic carriers, commonly rodents and livestock, shed pathogenic leptospires through their urine after they multiply in kidney tubules for their entire lifespan [3]. Once released into water and soil, leptospires can survive from one week to months in the environment [4]. Although humans can get infected through direct contact with a reservoir animal, most contaminations occur indirectly, after contact with a previously contaminated environment. The bacteria get entry into the body through mucous membranes as nose, mouth, and eyes and through cuts and abrasions of the skin. The time between exposure and the onset of symptoms ranges from 2 days to 4 weeks [3]. Although leptospirosis is commonly benign with few or no clinical manifestations, patients may develop a flu-like illness difficult to diagnose clinically that can drift to severe and potentially fatal Weil’s disease or the severe pulmonary hemorrhage [3]. The mean case fatality rate was estimated to be 6.85% globally, but can reach 30% in some developing countries [1]. While modelling the global burden of leptospirosis, Costa et al. identified tropical islands as a particular high-risk setting for human infections [1]. Tropical islands are indeed more vulnerable to invasive species like rodents as they are delicate ecosystems with a generally lower biodiversity but higher levels of endemism [5]. Living outdoors also increases exposure to leptospirosis in the islands. As a water-borne disease, leptospirosis is affected by climate conditions. Tropical hot and humid environments favor the survival of pathogenic leptospires [3], while heavy rainfall and flooding lead to a greater exposure to contaminated water [4,6] and increased contacts between rodents and human populations. Although the effect of floods is not well characterized and depends on the study setting, the considered territory and the scale of the study, a review encompassing 14 case-control and cohort studies identified floodings as a significant risk factor for leptospirosis with an overall odds ratio of 2.19 [7]. Many studies evidenced rainfall as the main climate contributor to leptospirosis outbreaks worldwide [6], including in tropical islands [8–12]. However, no consensus was reached on the time lag between a rain event and an increase in the number of cases reported by the health system, or on the direction of the association (positive or negative). For example, a long-term association with lags of 8 to 10 months was demonstrated in Thailand [13], while rainfall is expected to increase incidence with a lag of one week in Colombia [14]. In the wet zone of Sri Lanka, incidence was positively associated with rainfall with a lag of 5 months, but negatively associated with this same indicator with a lag of 2 and 3 months [15]. In tropical islands, a strong relationship between leptospirosis incidence and rainfall with a two-month time lag was evidenced in Reunion Island [10], French Polynesia [8] and in Futuna [11]. Studies in Mayotte supported a 3-month lag between cases and rainfall and the intensity of outbreaks was associated with the number of consecutive rainy months rather than rainfall amount [9]. In Fiji, the maximum rainfall in the wettest month contributed to model the spatial distribution of leptospirosis seroprevalence [12]. Temperature is also considered an important climatic factor for leptospirosis [3].A warm environment can increase the risk of exposure by attracting animals and humans to the same water sources and by encouraging water-based activities [6,16]. In Reunion Island, temperature of the current month appeared as a great explanatory variable of leptospirosis [10]. Inter-annual increases in incidence in New Caledonia were linked with La Niña phases of the ENSO (El Niño Southern Oscillation), characterized locally by a local hotter and wetter climate [17]. Similarly, in Guadeloupe, a four-fold increase in incidence between 2002 and 2004 occurred simultaneously with two El Niño events [18] that brought unusual weather conditions. The population of the islands are more at risk, as they live in a tropical climate and generally have difficult access to healthcare (e.g., due to hilly topography and poverty) [6]. Therefore, outbreaks can rapidly overflow health systems. In addition, leptospirosis is poorly notified and poorly documented in many tropical countries, and in tropical islands in particular, mainly because of a lack of medical awareness and difficulties associated in diagnosis as the unavailability of laboratory testing and the non-specific symptoms confound with other tropical infections [19]. Better understanding of leptospirosis’ dynamics and environmental triggers would promote the preparedness and awareness of communities and public health services. Studies of leptospirosis in tropical islands, conducted independently, generally differ in terms of available data and methods of investigations, making them difficult to compare and preventing any extension to other territories for which leptospirosis is poorly documented [10,20]. A generic model, i.e., a model validated across different islands, is therefore needed not only to help identify common features that favor infections, but also to highlight differences that can lead to unexpected local outbreaks. Such a model also could be used to extrapolate predictions and therefore inform on the disease burden in areas where it is not monitored. In this study, we investigated the relationship between climate and leptospirosis incidence in seven tropical islands from three oceanic basins: namely the South-Pacific (New Caledonia, Fiji, Tahiti and Futuna), the Indian Ocean (Reunion and Mayotte) and the North Atlantic (Guadeloupe in the West Indies). These islands are of various scales and are differently affected by leptospirosis. However, they all have tropical to subtropical climates with heavy rainfall during their respective warm seasons. We used satellite data for precipitation and temperature to study the relationship between climate and leptospirosis in tropical island settings. The developed model assisted in providing estimates of leptospirosis seasonal profile in Vanuatu (South-Pacific Ocean) and Puerto Rico (Caribbean Sea), two tropical islands where the disease is recognized as a health threat, but its dynamic remains poorly documented [21,22].

Discussion In this study, we provided for the first time a global insight of the leptospirosis’ dynamic over tropical islands. Our study encompassed Reunion Island, Mayotte, Guadeloupe, New Caledonia, Fiji, Futuna, and Tahiti into a global analysis to identify seasonal and inter-annual climate determinants of leptospirosis. We adjusted two types of models, a global seasonal model, and an inter-annual model specific to each island. These models helped identify common weather patterns leading to human disease cases in tropical islands and brought into light specific characteristics that impact the inter-annual variations of incidence. Moreover, our models provided the first estimates of leptospirosis’ seasonal profiles in Vanuatu and Puerto Rico, where disease incidence data was not available, even though the disease has been recognized as a public health issue [21,22]. Time series of leptospirosis data brought into light a strong seasonality with higher incidence occurring between March and April for islands located in the southern hemisphere and between March and November for Guadeloupe. As commonly described in tropical areas, peak of incidence of this seasonal dynamic occurred during the hot rainy season [3,8–10,12]. The seasonal profile computed as the median incidence of each month and observed in Futuna did not draw a clear seasonal pattern. This island was therefore not included in the global seasonal model. A recent study conducted in Futuna drew the monthly cumulated number of cases of a 10-years period and revealed a seasonal pattern with highest number of infections occurring during the rainy season, the first half of the year [11], similarly to the dynamics observed in New Caledonia. We could not observe this pattern while using median profile because of the strong inter-annual variability of leptospirosis in this poorly populated island [11] that might have hidden any relationship with climate determinants. As expected in a tropical setting [6], we identified rainfall as the main climate determinant of leptospirosis seasonality in the islands studied. In this model, rainfall of the current months and with a lag of 2 months was associated with an increase in human infections with a nonlinear relationship that was also evidenced in Salvador, Brazil [24]. Runoff caused by heavy rainfall are prone to spread the bacteria throughout territories by washing contaminated soil and draining pathogenic leptospires into freshwater [4]. It can also favor the movement of reservoir animals increasing contacts with the human population [6]. Despite leptospirosis outbreaks have been monitored after extreme rainfall events and floodings [25], we did not identify rainfall as an important climate determinant of inter-annual anomalies in Reunion Island and in Guadeloupe. Large outbreaks usually occur during the rainy season. It was therefore possible that for these islands the environment was already at such a high risk for infections that more rainfall would not expand the contaminated environment. This effect was hypothesized in American Samoa, one of the wettest inhabited places of the world, where no association was found between local rainfall and leptospirosis seroprevalence [26]. In Mayotte, rainfall was expected to reduce incidence after a certain threshold. This effect was also observed in Thailand [27] and in Colombia [28] and suggests that extreme rainfall could also dilute and wash away the leptospires. Temperature of the current and of the previous month also participated to model seasonal leptospirosis. In Reunion Island, the temperature of the current month has been shown to be positively correlated with the incidence [10]. However, while encompassing all islands into a single model to estimate seasonality, the marginal effect of temperature brought into light a negative relationship with the temperature of the current months. Surprisingly, temperature appeared as the most important climate determinant of leptospirosis inter-annual anomalies with a negative relationship with incidence in Reunion Island and Guadeloupe while both negative and positive anomalies increase leptospirosis incidence in New-Caledonia. As observed in Thailand, lower temperatures seemed more suitable for leptospirosis [27]. Optimal temperatures can favor leptospires growth [3]. Although desiccation induced by hot climate is not suitable for leptospires survival, it could promote water-based activity of both human and animal as bathing and drinking and intensify the sharing of persisting water sources increasing the probability of contamination [6,16]. Our models showed that climate lags ranging from 0 to 2 months prior to the onset of symptoms were important climate determinants of leptospirosis burden. These lags were consistent temporally with the leptospires survival into flooded lands and water-soaked soils [4] followed by the incubation period [3]. In our model, we focus on the short-term effect of climate events. Previous study identified ENSO-related long-lasting effects of climate that could impact leptospirosis inter-annually by building up the rodent reservoir resulting in an increased contamination of the environment and an increased seasonal rain intensity [17]. Although we achieved good predictions of incidence seasonal dynamics by building a global model of leptospirosis in tropical islands, we showed that 1) the incidence rate was difficult to capture in models and 2) climate determinants of inter-annual variability differed across islands, compromising the adjustment of a generic model. This suggested specific characteristics of islands cannot be neglected while studying leptospirosis dynamics. Our model did not include environmental indicators as water pH, soil type [4] and land use [6] that affect leptospires survival in the environment nor any consideration of the reservoir animals as rodents [29] and livestock [30] species both affecting the spatial distribution of the disease. Pathogenic leptospires have been detected in many mammalian species including domestic animals, livestock and wild animals [31]. In many islands the identification of the most common serotypes suggested that rats are the main reservoir of pathogenic leptospires [31]. In Reunion island, the dogs were also identified as a major carrier of the bacteria [10]. Futuna’s economy relies on pig farming and taro cultivation and the agricultural practice increases human exposure to leptospires, with both rats and humans spreading the bacteria in the environment [11]. In Fiji the presence of rats, mice and mongoose at home did not appear as a significant risk factor while the presence of pigs in the community and cattle in the district significantly increased the odds of being infected [30]. Livestock and domestic animals are restricted to specific areas (as farms, crops, backyard and private houses) and have an increased contact with the population as farmers, breeders, veterinary and pets sitters. In contrast, the spread of wild animals such as rodents is more difficult to control and contributes to the spread of bacteria in many different environments, from highly urbanized and populated areas to cultivated fields and very remote areas, potentially exposing a wide range of populations. Indication of the land use could therefore inform on the population exposure. The specificities of Leptospira animal reservoirs across islands is therefore likely to modify the local epidemiology of the disease temporally and spatially [31]. In addition, leptospirosis is recognized as a social disease whose occurrence depends on the behavior of the population, during professional or leisure activities, or in relation to cultural attitudes [3]. Our study area encompassed seven islands from different countries distributed around the world with different living conditions. In Mayotte, a large proportion of farmers allow their cattle to roam freely along dirt roads, fields, and rivers. In combination with poor living conditions and a lack of treated water at home, a quarter of the population turns to public fountains and rivers, which increases the risk of leptospirosis transmission [32]. Although no study has been conducted in these islands to link behavioral factors and infection, it is well established that impoverished rural-subsistence farmers and urban slum dwellers constitute the most vulnerable populations for leptospirosis [30]. Other risk factors recurrently reported include walking barefoot, swimming in streams and consuming water from different sources [16]. Agricultural activity either private or occupational appeared as the main risk factor of infection in French Polynesia, closely followed by the freshwater leisure activities [8]. Leptospirosis in Fiji was associated with outdoor occupations and poverty [30] whereas in New Caledonia most infections occurred during leisure or subsistence activities as hunting, fishing, bathing or swimming [33] and earlier peaks of incidence, caused by activities carried out during summer holidays, occurred in young age classes [34]. Leptospirosis is indeed increasingly associated with outdoor recreational exposure in developed countries. This change in exposure was expected to modify leptospirosis epidemiology in Reunion Island [10] and Guadeloupe [4]. The different lifestyles among tropical islands likely explained a part of the incidence not captured by our model solely based on climate indicators. Our model accurately identified the high season of leptospirosis and the epidemic years. However, peaks of incidence remained underestimated. We used climate data issued from satellites images that, despite providing a reliable gridded estimation of climate over the territories, remain an indirect measure of the in-situ data (fragmented data). In addition, we spatially aggregated the climate data to match the resolution of the leptospirosis records available (monthly data per island) and estimate leptospirosis burden at the island scale. Highly localized and intense climate conditions not captured by our averaged proxy values could be related to an increase in infections. In Reunion Island, prevalence spatially varied within the island according to the annual rainfall level [10]. The use of in situ climate data could reveal particular conditions that increase incidence locally. In addition, as a recreational disease, epidemics can occur during punctual gathering events as the well documented outbreaks of the Lake Springfield Triathlon in 1998 [35] and more recently a triathlon in Reunion Island in 2013 [36]. Such events are likely not captured by general models and must be kept in mind while interpreting model’s estimates, especially when working on small territories. Moreover, the detection of leptospirosis cases relies on the surveillance system implemented which can strongly vary across countries and over time. Although our data solely included laboratory confirmed cases, leptospirosis cases must, first, be suspected clinically. Diagnosis therefore depends on the awareness of medical staff. The switch from a passive to an active surveillance of the disease multiplied the number of detected cases by 5 in Hawaii, when testing patients presenting with at least 2 symptoms among fever, headache, myalgias, and nausea/vomiting [37] and by 8 in Futuna, when testing all acute fever [11]. It might partly explain the high incidence monitored in Futuna compared to the other islands. Increased communication and awareness might occur when an epidemic trend is discovered that leads to over-testing compared to the non-epidemics periods [8,25]. Symptoms of leptospirosis overlap with other tropical diseases, especially arboviral diseases, rendering diagnosis difficult. In 2018, a dengue outbreak occurred in Reunion Island [38] simultaneously with a peak in leptospirosis incidence. As observed in Hawaii during the period 2001–2002 [39] and in Puerto Rico in 1996 [40], dengue outbreaks might have contributed to increase detection of leptospirosis with dengue negative samples being tested for leptospirosis as differential diagnosis. In addition, periods with no or low incidence of dengue might promote leptospirosis as primary diagnosis and increase detection. Recorded incidence therefore strongly depends on the surveillance system implemented. Variations in case detection among islands and throughout time probably participated in the underestimation of outbreaks and impeded the adjustment of global models.

Conclusion Despite the complexity of the relationship between humans, environment, and reservoir animals, our results emphasize the importance of past and current climate factors, such as, temperature, and precipitation (lagged from 0 to 2 months), in the dynamics of leptospirosis. The key finding of this study is the identification of a unique climate-based model capable of describing the seasonal dynamics of leptospirosis in tropical islands across three oceanic basins on a global scale. This suggests that the seasonality of leptospirosis can be captured solely by climate indicators. However, at the local scale, accurately estimating leptospirosis burden in tropical islands remains a significant challenge due to the diversity of lifestyles and of environments that influence the ability of climate to trigger outbreaks. While climate factors can easily estimate the seasonality of leptospirosis in tropical islands, a deeper understanding of local specificities, such as human behavior and environment characteristics should be investigated to improve our understanding of the occurrence of outbreaks inter-annually. The identification of climate indicators linked to leptospirosis dynamics represents a crucial initial step toward developing forecasting models. Future research should involve the investigation of a model at finer scale to capture the relationship between climate and disease dynamics at the local level and lay the foundations for the development of early warning systems for leptospirosis epidemics. Such models would inform decision-makers in the health sector and contribute to epidemic preparedness and management.

Supporting information S1 Fig. Leptospirosis, precipitation and temperature seasonal profiles of the studied Islands. The seasonal profile for Reunion Island (a), Guadeloupe (b), Tahiti (c), Fiji (d), New Caledonia (e), Mayotte (f) and Futuna (g) are defined by the mean temperature (°C—red line) and the mean precipitation rate (mm/hr—grey bar) of each month (left panel). Leptospirosis profile is given in incidence per 100,000 inhabitants and was defined as the median value of incidence of each month (blue line). https://doi.org/10.1371/journal.pntd.0011717.s001 (TIF) S2 Fig. Leave-island-out MAE scores for the tested global seasonal models. Leptospirosis normalized seasonal profile was estimated based on the normalized seasonal profile of temperature (S(T)) and precipitation (S(P)) with a lag ranging from 0 to 2 months. We performed normalization by removing the mean of the variables in each island. The black dotted boxes frame the best model selected for modelling the leptospirosis seasonal dynamics. https://doi.org/10.1371/journal.pntd.0011717.s002 (TIF) S3 Fig. Leave-year-out ΔMAE scores of the per island inter-annual models. Models estimate the anomalies of log incidence based on precipitation and temperature anomalies (respectively A(P) and A(T)) with a lag ranging from 0 to 2 months. ΔMAE compares the accuracy of the predicted incidence to the seasonal profile. The black dotted boxes frame the best models selected for modelling the leptospirosis anomalies in each island. https://doi.org/10.1371/journal.pntd.0011717.s003 (TIF) S4 Fig. Third dimension of the principal component analysis plotted against dimension 1 and 2 for Tahiti and Futuna. https://doi.org/10.1371/journal.pntd.0011717.s004 (TIF)

Acknowledgments We thank the various stakeholders that provided leptospirosis data: the Institut Pasteur and the department of Health and Social affairs of New Caledonia, the Regional Office of the French Institute for Public Health surveillance Antilles-Guyane (Cire), the French National Reference Center for leptospirosis, the Direction of Health of French Polynesia, Dr Mike Kama from the Fiji ministry of Health, Clément Couteaux from Agence de Santé de Wallis & Futuna.

[END]
---
[1] Url: https://journals.plos.org/plosntds/article?id=10.1371/journal.pntd.0011717

Published and (C) by PLOS One
Content appears here under this condition or license: Creative Commons - Attribution BY 4.0.

via Magical.Fish Gopher News Feeds:
gopher://magical.fish/1/feeds/news/plosone/