Prevalence and risk factors for human leptospirosis at a hospital serving a pastoralist community, Endulen, Tanzania

Background Leptospirosis is suspected to be a major cause of illness in rural Tanzania associated with close contact with livestock. We sought to determine leptospirosis prevalence, identify infecting Leptospira serogroups, and investigate risk factors for leptospirosis in a rural area of Tanzania where pastoralist animal husbandry practices and sustained livestock contact are common. Methods We enrolled participants at Endulen Hospital, Tanzania. Patients with a history of fever within 72 hours, or a tympanic temperature of ≥38.0°C were eligible. Serum samples were collected at presentation and 4–6 weeks later. Sera were tested using microscopic agglutination testing with 20 Leptospira serovars from 17 serogroups. Acute leptospirosis cases were defined by a ≥four-fold rise in antibody titre between acute and convalescent serum samples or a reciprocal titre ≥400 in either sample. Leptospira seropositivity was defined by a single reciprocal antibody titre ≥100 in either sample. We defined the predominant reactive serogroup as that with the highest titre. We explored risk factors for acute leptospirosis and Leptospira seropositivity using logistic regression modelling. Results Of 229 participants, 99 (43.2%) were male and the median (range) age was 27 (0, 78) years. Participation in at least one animal husbandry practice was reported by 160 (69.9%). We identified 18 (7.9%) cases of acute leptospirosis, with Djasiman 8 (44.4%) and Australis 7 (38.9%) the most common predominant reactive serogroups. Overall, 69 (30.1%) participants were Leptospira seropositive and the most common predominant reactive serogroups were Icterohaemorrhagiae (n = 20, 29.0%), Djasiman (n = 19, 27.5%), and Australis (n = 17, 24.6%). Milking cattle (OR 6.27, 95% CI 2.24–7.52) was a risk factor for acute leptospirosis, and milking goats (OR 2.35, 95% CI 1.07–5.16) was a risk factor for Leptospira seropositivity. Conclusions We identified leptospirosis in approximately one in twelve patients attending hospital with fever from this rural community. Interventions that reduce risks associated with milking livestock may reduce human infections.


Introduction
Leptospirosis is a major cause of human illness in sub-Saharan Africa [1,2].However, diagnostic tests are rarely available in countries in sub-Saharan Africa and leptospirosis is under-recognised by clinicians [3,4].In Tanzania, outbreaks of leptospirosis are reported [5], and endemic disease also causes hospitalisations [6].Gaps in understanding of the epidemiology of leptospirosis hinder control efforts.Data from a small number of predominately urban hospitals inform estimates of disease incidence in Tanzania [3,7,8] and indicate that a substantial proportion of human leptospirosis is acquired through transmission from livestock to humans [9][10][11].Communities that practice pastoralism live in close contact with livestock and might have a higher incidence of leptospirosis than urban communities or districts where smallholder farming practices predominate [12].Estimates of the prevalence of leptospirosis in patients presenting to hospitals provide insights into the frequency of severe disease and can estimate incidence when combined with information on healthcare utilisation [13].Therefore, determining the prevalence of acute cases among hospitalised patients is a critical step towards understanding the impact of leptospirosis.
Leptospirosis is a zoonotic disease with a wide variety of animals acting as potential sources of human infection.Identifying the principle sources of human infection is vital for control efforts but also challenging as there are at least 40 pathogenic species and >250 pathogenic serovars of Leptospira that can infect a large number of animals [14].As reservoir hosts vary by infecting serovar and this impacts transmission pathways, understanding which Leptospira serovars cause human disease locally is critical to planning effective control measures.Leptospira serovars are organized into serogroups, which contain a limited number of antigenically similar serovars.While micro-agglutination testing (MAT) serology cannot establish the infecting Leptospira serovar because of cross-reactivity between serovars within serogroups [15], it can be used to determine infecting serogroup.When considered in conjunction with serologic data from livestock, interpretation of serogroups infecting people can suggest likely animal sources of human infection and thus have implications for disease control efforts.Data gathered at northern Tanzania show that Leptospira serovars from similar serogroups, particularly the Australis serogroup, infect humans and livestock but that infecting serogroups vary by location and over time [11,[16][17][18][19].For example, in Kilimanjaro Region, serogroup Mini was frequently the predominant reactive serogroup among people hospitalised with leptospirosis during 2008-2009, but was rarely identified in 2012-2014 [3,6].
Humans are incidental hosts of Leptospira and can be infected through exposure to either infected animals or to a contaminated environment [20].Knowledge of specific risk behaviours or activities is essential for interventions to reduce human disease.Although risk factors for leptospirosis are broadly established [21], and include diverse exposures to rodents, livestock, and floodwaters, the importance of specific risk factors varies substantially by context due to variation in animal reservoirs and human activities.In areas of Tanzania where smallholder farming predominates [22], acute leptospirosis has been associated with activities such as cleaning cattle waste and feeding cattle [10].Exploration of livestock-associated risk factors among pastoralist communities may provide actionable public health interventions to reduce the incidence of leptospirosis among high risk communities.
We sought to address three knowledge gaps to inform the control of leptospirosis in Tanzania.We aimed to estimate the prevalence of leptospirosis among patients attending a rural hospital serving a pastoralist community, identify the predominant reactive serogroups, and investigate behavioural risk factors for human disease in this population.

Ethics statement
This study was conducted in accordance with the Declaration of Helsinki.It was approved by the Tanzania National Institutes for Medical Research National Ethics Coordinating Committee (NIMRlHQ/R.8cIV0111/708), Kilimanjaro Christian Medical University College Research Ethics Committee (698), Glasgow University, College of Medical, Veterinary and Life Sciences Ethics Committee (200150140), and the University of Otago Human Ethics Committee (Health) (H17/052).Written informed consent was obtained from all adult participants, with forms translated into Kiswahili and verbal Maa translation also provided as needed.Informed written consent for minors (under 18 years old) was provided by their parent or guardian and assent was additionally obtained from all participants aged 13-17 years old.Sample shipments between Tanzania and the United States were performed in accordance with the Material Transfer Agreement between Kilimanjaro Christian Medical Centre and the US Centers for Disease Control and Prevention (NCEZID-136933-00).

Study site
We recruited participants at Endulen Hospital in the Ngorongoro Conservation Area (NCA) of northern Tanzania (Fig 1) [23].Endulen hospital is a 110-bed hospital and the only hospital in the NCA.The elevation of the NCA ranges from approximately 1,000m through 3,600m above sea level.Climate is characterised by bimodal seasonal variability with two wet seasons from October through December and March through May [24].The NCA is a multiple land use area, designated for the pastoralist activities of the local community, the conservation of wildlife, and tourism.Pastoralist livestock-keeping with cattle, sheep and goats managed extensively in mixed herds is commonly practiced.Further details of the study site and population are reported elsewhere [25].

Enrolment
Patients seeking care at the outpatient department of Endulen Hospital from August 2016 through October 2017 were eligible for screening.Screening was performed on 259 of 422 days in the screening period (4-5 days in each week).Patients aged �2 years were eligible to enrol if they had a history of fever within 72 hours, or a measured tympanic temperature of �38.0˚C at presentation.After clinical assessment by Endulen Hospital staff, eligible patients were approached by a study team member for informed written consent to participate in the study.Enrolled patients underwent phlebotomy, following skin sterilisation with isopropyl alcohol and povidine iodine.Blood samples were prioritised for culture and malaria testing.A target volume of 10 mL blood was then placed into a plain vacutainer (BD, Franklin Lakes, NJ, USA) for serology.At enrolment a member of the study project team administered a structured questionnaire comprising closed-ended questions [25].Question topics included demographic characteristics, symptoms of the current illness and within the preceding two months, animal-related activities undertaken during the preceding month and the preceding year, and Administrative boundary shapefile sourced from the Tanzania National Bureau of Statistics, https://www.nbs.go.tz/index.php/en/census-surveys/populationand-housing-census/172-2012-phc-shapefiles-level-one-and-two[43].Map image is the intellectual property of Esri and is used herein under license.
Copyright 2020 Esri and its licensors.All rights reserved.Maps drawn using R [23], and Quantum Geographic Information System (QGIS) open access software [44].Shapefiles were obtained from Tanzania National Bureau of Statistics [43]. https://doi.org/10.1371/journal.pntd.0011855.g001 animal health (S1 File).Study team members visited participants at their homes 4-6 weeks after enrolment for collection of a convalescent serum sample.

Case definitions
We defined acute leptospirosis as a clinically compatible illness plus either a �four-fold rise in agglutinating antibody titres between acute and convalescent serum or a single reciprocal titre of �400 for one or more Leptospira serovars [27].Seropositive participants were defined by a single reciprocal titre of �100 in either the acute or convalescent sample.Seronegative participants were those with reciprocal titres <100 for all serovars tested in their available samples.The predominant reactive serogroup for acute cases and seropositive participants was defined as the serogroup containing the serovar with the highest titre, with multiple serogroups recorded in the event of equal highest titres.

Statistical analysis
Data were collated in Microsoft Excel (Microsoft Corporation, Redmond, WA, USA), and analysed using Stata, version 16.0 (StataCorp, College Station, TX, USA).Logistic regression was used to explore associations between independent variables and two outcome measures: acute leptospirosis and Leptospira seropositivity.For models of acute leptospirosis, only data from acute cases and seronegative participants were included.For models of seropositivity, all data were included with comparison of data from seronegative participants to combined data from acute cases and seropositive participants.We assessed activities undertaken within the preceding month in analyses relating to acute leptospirosis as we considered this time frame most relevant to the 3-30 day incubation period of leptospirosis [20].We assessed activities undertaken within the preceding year in analyses relating to Leptospira seropositivity due to the long persistence of anti-Leptospira antibodies [28].Independent variables included in both multivariable models (Tables 1 and 2) were selected a priori through review of relevant literature and construction of a directed acyclic graph (Fig 2).For both analyses, univariable models were constructed for all candidate variables and variables for which the univariable P values were �0.2 were carried forward for inclusion in multivariable models.We assessed correlation using a correlation matrix and combined variables that had a correlation >0.7.For example, highly correlated variables recording slaughter of two livestock species (cattle and goats) in the past year were used to create a combined variable recording slaughter of cattle and/or goats.Interactions between a history of livestock death within the household herd (within the preceding month for acute leptospirosis and preceding year for Leptospira seropositivity) and livestock behaviour variables were evaluated for both outcomes.We employed backwards selection to determine the final model for each outcome, selecting models with the smallest Akaike Information Co-efficient (AIC) at each step.Logistic regression P values are two sided, and significance was set at 0.05.

Results
Of 234 participants enrolled in the study, 229 (97.9%) completed questionnaires and provided serum samples, and were included in this analysis.Among participants, the median (range) age was 27 (0, 78) years and 99 (43.1%) were male (Table 3).The exact duration of illness prior to presentation was not recorded, but 202 (88.2%) participants reported the duration of their fever as lasting days, as opposed to weeks or months.One hundred and thirty five (59.0%) participants reported participation in at least one cattle-husbandry activity within the last month; 138 (60.3%) reported participation in a goat-husbandry activity; and 115 (50.2%) reported participation in a sheep-husbandry activity.The proportion of participants who reported that they had performed specific activities are shown in Tables 1 and 2. Some 134 (58.1%) participants reported performing husbandry activities for more than one livestock species, with correlations between specific activities shown in S2 File.We identified 12 (8.2%)cases of acute leptospirosis among 146 participants providing paired serum samples, and an additional 6 (7.2%) cases among 83 participants providing only a single sample.The most common predominant reactive serogroups among cases (Table 4) were Djasiman (n = 8, 44.4%) and Australis (n = 7, 38.9%).In total, 69 (31.1%) of 229 participants were seropositive to Leptospira.The most common predominant reactive serogroups among seropositive participants were Icterohaemorrhagiae (n = 20, 30.0%),Djasiman (n = 19, 27.1%), and Australis (n = 17, 24.3%).The remaining 160 (69.9%) participants were classified as seronegative.
The final logistic regression model of acute leptospirosis identified having milked cattle within the preceding month (OR 5.42, 95% CI 1.96, 14.99) as a risk factor.Correlations

Discussion
Our study of patients presenting with fever at a rural hospital identified that leptospirosis is a frequent cause of illness, present in nearly in approximately one in twelve patients.Leptospira serovars from serogroups Djasiman and Australis were most frequently the predominant reactive serogroup.Activities placing people in close contact with livestock, notably milking, are risk factors for disease transmission.Approximately one in twelve patients presenting with a febrile illness met our definition for acute leptospirosis.We consider this prevalence estimate to be an underestimate due to our reliance on MAT serology.MAT is considered to have imperfect sensitivity, particularly when diagnosis rests on testing of a single serum sample [29].The seroprevalence of 30.1% estimated in this population is greater than estimates from studies of patients hospitalised in urban and small-holder farming systems in Tanzania [9], but similar to an estimate from the semi-arid Dodoma Region [30].
In our study, the most frequent predominant reactive serogroups in participants with acute leptospirosis were Australis and Djasiman.While serovars from these serogroups have rarely been isolated in samples from East Africa, they have frequently been identified as the predominant reactive serogroups in previous reports from people in Tanzania [9].Similarly, a recent report from Uganda identified Djasiman as the most commonly reactive serogroup among general and renal clinic patients [31].Human seropositivity to Leptospira serovars from the  Australis serogroup has been a common finding across multiple time periods and Regions of Tanzania [3,6,30,32,33], and has been detected in livestock, particularly cattle, as well as rodents [19,33].Animal hosts of Leptospira serovars from the Djasiman serogroup have not been determined in Africa [2,34].Therefore further work to isolate Leptospira from animal populations, particularly livestock is needed.Seropositivity to Leptospira serogroup Icterohaemorrhagiae was common in this population, but the serogroup was predominantly reactive for only a single case of acute leptospirosis.This is consistent with findings in Kilimanjaro Region at a similar time period, from 2012 through 2014) [6].The discrepancy between acute cases and seropositivity may indicate that a Leptospira serovar from Icterohaemorrhagiae serogroup was previously a more common cause of acute disease, and is in keeping with other evidence suggesting that the relative importance of serovars may change over time [6].
Our study supports previous work in northern Tanzania that suggested an important role for livestock in the transmission of leptospirosis to humans and provides insights into activities leading to human infection.In this study, milking goats and cattle were identified as risk factors for Leptospira seropositivity and acute leptospirosis respectively.In both regression analyses several activities showed association with the outcome of acute leptospirosis or seropositivity in the univariable analyses.However, many of the behavioural variables were clearly correlated (S2 File) and only a single variable was identified as a significant predictor of participant leptospirosis status in each final model.This demonstrates the challenges of differentiating the contribution of different activities to leptospirosis transmission in contexts where individuals undertake many potential risk behaviours and interact with multiple livestock species.This link of leptospirosis with behaviours exposing people to livestock urine is consistent with research from the neighbouring Kilimanjaro Region, activities that exposed participants to cattle urine, such as cleaning cattle waste, have been identified as risk factors for leptospirosis [10].In addition, molecular and serologic data highlighted genomic and serologic similarity between Leptospira shed in the urine of livestock and those detected in patients with leptospirosis in Kilimanjaro region [11].Similar work to link Leptospira from animals and humans is important step for future research in our study area.
We used hospital-based surveillance in order to detect acute leptospirosis.A limitation of this approach is that seroprevalence of the study population may differ from the community and care must be taken if comparing our seroprevalence estimate to population-based serosurveys.Further, leptospirosis testing and analysis was a secondary aim of the study and was not subject to a priori sample size calculations for leptospirosis prevalence estimation or risk factor identification specifically.The foundation study was powered to investigated the epidemiology of brucellosis in this hospital population.One component of the original study was powered with required sample size of 126 to estimate prevalence, assuming a prevalence of 9%, desired precision of 5% and power of 80% [25].Given plausibly similar expected prevalence of brucellosis and leptospirosis in this population [35], and availability of data from of 229 participants it is unlikely that this study was underpowered for the prevalence estimation goal but it may be underpowered to identify some of the associations assessed in the risk factor analysis.Strengths of our study include the systematic testing of patients using reference standard diagnostics.However, limitations of test sensitivity mean we are likely to have underestimated the true prevalence of acute leptospirosis and Leptospira seroprevalence.The collection of convalescent serum samples from approximately two thirds of participants in this study is consistent with the proportion collected in similar febrile illness studies [35][36][37].The lack of convalescent serum from one third of participants likely contributes further to underestimation of the prevalence of acute leptospirosis due to the lower sensitivity of MAT when testing acute serum samples only [29].We also acknowledge that due to the variable time-course of anti-Leptospira antibody decay and the high proportion of participants with Leptospira exposure, some leptospirosis cases defined by a single antibody titre �400 may reflect recent rather than current infection.We used a broad array of serovars covering all pathogenic serogroups, but sensitivity could have been further enhanced by inclusion of local isolates among the tested antigens [38].Despite previous studies in Tanzania finding Leptospira nucleic acid amplification testing (NAAT) of blood and urine of infected patients had low sensitivity of to diagnose leptospirosis [11], NAAT may be a useful addition to MAT serology in future studies estimating leptospirosis prevalence [39].Our risk factor analyses provide useful insights into livestock related risk factors, but the absence of data regarding exposure to wild animals, particularly rodents, means that we cannot make conclusions about their role as potential source populations in Leptospira transmission in this context.
Our work has important public health and clinical implications.The high prevalence of acute leptospirosis in a population presenting to hospital with febrile illness suggests that resources are needed to reduce the burden of leptospirosis in this population.A recent outbreak of leptospirosis has brought attention to leptospirosis in Tanzania [5], but clinician awareness is limited [4], and leptospirosis is not included in the Standard Treatment Guidelines and Essential Medicines List for Tanzania Mainland [40].Clinical care would be aided by improved recognition through availability of accurate diagnostic tests and inclusion in future clinical guidelines.There is increasing evidence of a central role for livestock as a source of human leptospirosis in Tanzania [10,11,18,19,30].In the broader context of febrile illness, leptospirosis is one of many zoonotic diseases that is prevalent among rural communities across Africa.Previous work has also indicated a high prevalence of brucellosis in the Endulen study population [25].Q fever, spotted fever group rickettsioses, and Rift Valley fever, among other livestock associated zoonoses, have also been identified as important causes of human febrile illness in East Africa [41,42].Strategies are needed to reduce the prevalence of multiple zoonoses in livestock and to reduce their transmission to people.For leptospirosis, providing information about transmission routes and discussing feasible behavioural preventive strategies with communities living in close connection with their livestock might reduce the risk of disease.For example, changes to milking practices to avoid gross exposure to urine, might reduce risk and their feasibility and effectiveness should be explored.In addition, identification of which serovars are present in livestock might enable the future use of livestock vaccination as a strategy to improve both livestock health and human health.
In conclusion, our study highlights the important role of leptospirosis as a cause of febrile illness among a pastoralist community in Tanzania.Further, we have added to the evidence base highlighting livestock as an important source of human leptospirosis in East Africa.In the context of increasing evidence of the high prevalence of several zoonotic diseases in livestock, these findings further emphasise the need for an integrated approach to zoonoses to improve both human and animal health.and Ngorongoro Conservation Area Authority (NCAA) for approvals to conduct this project within the Ngorongoro Conservation Area.
Disclaimer: The findings and conclusions in this report are those of the author(s) and do not necessarily represent the official position of the Centers for Disease Control and Prevention.

Fig 1 .
Fig 1. Location of Endulen Hospital and administrative wards that study participants presented from, Arusha Region, Tanzania.Map showing the administrative wards of the Ngorongoro, Karatu and Meatu districts in northern Tanzania (polygons) and location of Endulen hospital (black dot).Administrative boundary shapefile sourced from the Tanzania National Bureau of Statistics, https://www.nbs.go.tz/index.php/en/census-surveys/populationand-housing-census/172-2012-phc-shapefiles-level-one-and-two[43].Map image is the intellectual property of Esri and is used herein under license.Copyright 2020 Esri and its licensors.All rights reserved.Maps drawn using R[23], and Quantum Geographic Information System (QGIS) open access software[44].Shapefiles were obtained from Tanzania National Bureau of Statistics[43].
between livestock-related activities undertaken within the preceding month are shown in S2 File.No significant interaction terms were identified.The final model of Leptospira seropositivity included age (OR 1.01, 95% CI 1.00, 1.03 per year) and identified having milked goats (OR 2.35, 95% CI 1.07, 5.16) as a risk factor.Correlations between livestock-related activities undertaken within the preceding year are shown in S2 File.No significant interaction terms were identified.