Prevalence and risk factors for Q fever, spotted fever group rickettsioses, and typhus group rickettsioses in a pastoralist community of northern Tanzania, 2016–2017

In northern Tanzania, Q fever, spotted fever group (SFG) rickettsioses, and typhus group (TG) rickettsioses are common causes of febrile illness. We sought to describe the prevalence and risk factors for these zoonoses in a pastoralist community.


INTRODUCTION
Q fever and spotted fever group (SFG) rickettsioses are important causes of febrile illness in Tanzania [1,2].Typhus group (TG) rickettsioses are also implicated in febrile illness, although less frequently [1].Coxiella burnetii is the causative agent of Q fever and human transmission occurs predominantly through inhalation of contaminated aerosols or consumption of infected, unpasteurized dairy products [3].The role of ticks in the transmission of C. burnetii to humans is ambiguous [2].Rickettsia africae and Rickettsia conorii are the causative agents of SFG rickettsioses.Rickettsia africae is transmitted by Amblyomma spp.ticks that feed on domestic livestock whereas R. conorii is transmitted by Rhipicephalus sanguineus ticks that are frequently found on dogs; both can infect humans directly through bites [4].Rodents are the hosts for Rickettsia typhi and the rat flea, Xenopsylla cheopis, is the primary vector mediating human infection with TG rickettsioses [5].
There are gaps in the understanding of the epidemiology, risk factors for acute illness and serologic exposure, and clinical impact of these zoonotic diseases in sub-Saharan Africa [2,6,7].Q fever, SFG rickettsioses, and TG rickettsioses share non-specific presenting symptoms and signs including fever, headache, and myalgia [8,9].Limited diagnostic capacity and low awareness of zoonoses among healthcare providers contribute to underdiagnosis [10].
Studies from sub-Saharan Africa indicate highly variable levels of serologic exposure to C. burnetii and SFG Rickettsia by location [1,6,11].Frequent, close contact with livestock and their tick ectoparasites has been associated with the risk of Q fever and SFG rickettsioses [6,7,12].Handling animal abortion materials, slaughtering animals, and consuming raw or locally fermented dairy products have been associated with Q fever [13][14][15].There have been few studies of seroprevalence or acute illness in populations that are likely to be highly exposed to these infections such as livestock-dependent pastoralists.
We describe the prevalence and identify factors associated with serologic evidence of exposure and acute illness due to C. burnetii, SFG Rickettsia, and TG Rickettsia among individuals presenting to a rural hospital in northern Tanzania that serves a predominantly pastoralist population.

Study site
Participants were recruited at Endulen Hospital, a 110-bed facility serving the predominantly pastoralist population in the Ngorongoro Conservation Area (NCA) of northern Tanzania.The NCA is a multiple land use area for conservation of wildlife, tourism, and livestock-keeping (predominantly cattle, sheep, and goats) by the local, pastoralist community that consists predominantly of individuals identifying as Maasai [16].There is bimodal seasonal variability with wet seasons typically occurring October through December and March through May [17].Further details of the foundational study focused on brucellosis including the study site and participant population have been published previously [18].

Enrolment
Patients seeking care in the outpatient department of Endulen Hospital were screened for eligibility from August 2016 through October 2017.Screening occurred $4-5 days each week on a total of 259 (61.4%) of 422 days in the study period.Eligible individuals were aged ≥2 years with reported fever within the past 72 h or with a tympanic temperature of ≥38.0 C at presentation.Eligible patients were approached by a study team member to obtain written informed consent to participate in the study.After enrolment, blood was drawn for serology and a study team member administered a structured questionnaire including closed-ended questions related to demographic data, clinical characterisation of illness, recent illness, occupation, and livestock-related activities, but not including capture of direct dog or tick exposures, during the past month and past year (Supplementary Methods S1).Four to six weeks after enrolment, study team members conducted home visits to collect convalescent-phase blood samples from participants.

Laboratory testing
Serum was separated and stored at 4 C at Endulen Hospital prior to transport to Kilimanjaro Clinical Research Institute (KCRI) the next day at 4-10 C. At KCRI, sera were stored at À70 C then shipped on dry ice to the Rickettsial Zoonoses Branch, US Centers for Disease Control and Prevention (US CDC) for immunofluorescent antibody (IFA) testing.All sera were tested by IFA IgG assays using C. burnetii (Nine Mile strain) Phase I and Phase II antigens for Q fever, R. africae (Z9-Hu strain) antigen for SFG rickettsioses, and R. typhi (Wilmington strain) for TG rickettsioses.Paired acute and convalescent serum samples were tested concurrently using the same reagent lot.Samples were tested using doubling dilutions to determine the last dilution at which antibody was detected or a result of <1:32 recorded if no antibody was detected.

Outcome definitions
US Council of State and Territorial Epidemiologists case definitions for Q fever and SFG rickettsioses were used with the modification that all participants were considered to have a clinically compatible syndrome based upon their presentation with fever [19,20].Henceforth, exposure is defined as serologic evidence of exposure to a pathogen of interest.Exposure to C. burnetii was defined by a single IFA titre of ≥1:64 to Phase II antigen in either the acute or convalescent sample [1].Probable acute Q fever was defined by a single IFA titre of ≥1:128 to phase II antigen [19].Confirmed acute Q fever was defined by a ≥4-fold rise in IFA titre to C. burnetii Phase II antigen between acute and convalescent serum samples [19].Chronic Q fever was defined by an IFA titre of ≥1:1024 to C. burnetii Phase I antigen [21].
For SFG Rickettsia and TG Rickettsia, exposure was defined by an IFA titre to ≥1:64 in either the acute or convalescent sample using R. africae and R. typhi antigens, respectively [1].Probable illness was defined by an IFA titre ≥1:128 in either the acute or convalescent sample [20].Confirmed illness was defined by a ≥ 4-fold rise in IFA titre to SFG rickettsioses antigen between acute and convalescent serum specimens [1,20].

Statistical analysis
Data were entered using the OpenText Teleform System (Open Text, Waterloo, Ontario, Canada) into an Access database (Microsoft Corporation, Redmond, WA, USA).Data were analysed using R version 4.1.1(R Foundation for Statistical Computing, Vienna, Austria).We chose independent variables for analyses through literature review and based on dataset variables from the primary study.Logistic regression models were used to identify associations between selected independent variables and the outcome variables of exposure to C. burnetii, acute Q fever, exposure to SFG Rickettsia, and SFG rickettsioses.For the risk factor analyses for acute Q fever, individuals with illness that met probable and confirmed case definitions were considered as cases.For the risk factor analyses for SFG rickettsioses, only individuals with illness that met confirmed case definitions were considered as cases and analyses were restricted to individuals with paired sera.
For models of exposure to C. burnetii and SFG Rickettsia, variables defining animal-related activities such as direct animal contact, contact with animal products, or consumption of animal products in the preceding year were analysed (Tables 2 and 4).For analyses of acute Q fever and SFG rickettsioses, variables describing animal-related activities in the preceding month were analysed, aligning with the 7-32 day incubation period for Q fever and the 6-to 10-day incubation period for SFG rickettsioses [22,23] (Tables 3 and 5).
Data regarding participants' consumption of raw dairy products were only available for the preceding month, and data regarding exposure to livestock abortions were only available for the preceding year.These variables and time periods were used in all analyses.Data on participant occupation were grouped to compare high-and low-risk occupations based on likelihood of livestock and tick exposure.Farmers, livestock attendants, or persons who worked with wildlife were classified as engaging in high-risk occupations and all other reported occupations were classified as low risk.January, February, June, July, August, and September were classified as dry season months, and all other months as wet season [17].
Univariable models were explored for all outcomes modelled.Variables with likelihood ratio test (LRT) p ≤ 0.2 in the univariable model were considered for inclusion in multivariable models.Correlations between plausibly correlated independent variables (e.g., milking cattle, milking goats, milking sheep) were assessed using a matrix of Pearson's product moment correlation coefficients (Tables S1-S3).For variable pairs with Pearson's correlation coefficients >0.5, a single variable from the set of correlated variables with univariable LRT p ≤ 0.2 was selected to represent an animal-related exposure or activity in multivariable models.Multivariable models were created by initially fitting maximal models and simplifying by removing variables with LRT p > 0.05, with terms with the largest LRT p-values removed first.Interactions between season and abortion associated risks were considered.Variables were excluded from multivariable models if a small number of observations in any category caused convergence problems.

Research ethics
This study was conducted in accordance with the Declaration of Helsinki.It was approved by the Tanzania National Institute for Medical Research National Health Research Ethics Review Committee (NIMRlHQ/R.8cIV0111/708), Kilimanjaro Christian Medical University College Research Ethics Committee (698), University of Glasgow 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 participants ≥18 years old or their guardians if <18 years old; minors aged 13-17 years also provided assent.

Sociodemographic and characteristics of study participants
A total of 228 participants with complete questionnaires and diagnostic test data were included in analyses (Figure 1).The median (interquartile range) age of participants was 27 (16-41) years and 99 (43.4%) were male.Of all participants, 146 (62.9%) had results of paired acute and convalescent serum sample testing.Further details of the study population and the frequency of potential risk factors for study outcomes are given in Tables 1-5.Distributions of age, sex, and season of presentation are shown in Figures S1-S3.

Univariable and multivariable logistic regression
Logistic regression models are presented for the outcome variables of exposure to C. burnetii, acute Q fever (probable and confirmed cases), exposure to SFG Rickettsia, and SFG Note: Bold indicates significant variables included in multivariable analyses.Asterisk denotes question answered as yes or no.All variables with (y) notation refer to the performance of or exposure to the stated activity in the 12 months prior to presentation.All variables with (m) notation refer to the performance of or exposure to the stated activity in the 1 month prior to presentation.High risk occupation includes farmers, livestock attendants, or worked with wildlife.All animal related activities include cattle, sheep and goats in all cases and other species for specific questions.Abbreviations: aOR, adjusted odds ratio; CI, confidence interval.
rickettsioses.Regression models of exposure to TG Rickettsia and TG rickettsioses were not performed due to small numbers of outcomes.

Exposure to C. burnetii and Q fever
The results of univariable and multivariable risk factor analyses for exposure to C. burnetii and acute Q fever are given in Tables 2 and 3. Pearson's correlation coefficients between animal-related activities undertaken within the past year and past month that were evaluated in C. burnetii relevant models are given in Tables 2, S1, and S2.The final model for exposure to C. burnetii identified increased odds of exposure among those who reported handling the products of livestock abortion events within the prior year compared with those who did not (adjusted odds ratio [aOR] 6.36, 95% confidence interval [CI] 2.05-28.02)and among those who presented to the hospital during the dry season compared with the wet season (aOR 2.16, 95% CI 1. 25-3.77).The final model for acute Q fever, which included probable and confirmed cases, showed increased odds of disease among those who slaughtered animals within the prior month compared with those who did not (aOR 2.54, 95% CI 1.38-4.76)and among those who presented to the hospital during the dry season compared with the wet season (aOR 3.14, CI 1.70-6.02).There were decreased odds of acute Q fever among those who milked cattle within the prior month compared with those who did not (aOR 0.43, CI 0.17-0.97).

Exposure to SFG Rickettsia and SFG rickettsioses
The results of univariable and multivariable risk factor analyses for exposure to SFG Rickettsia and SFG rickettsioses are given in Tables 4 and 5. Pearson's correlation coefficients between animal-related activities undertaken within the past year that were evaluated in SFG Rickettsia relevant models are given in Table S3.The multivariable logistic regression model of exposure to SFG Rickettsia identified increased odds of exposure among those who herded cattle within the prior year compared with those who did not (aOR 4.12, 95% CI 1.74-11.47)and with age, with increased odds of exposure per year of age (aOR 1.04, 95% CI 1.02-1.06;Table 4).No multivariable model is presented for the SFG rickettsioses outcome as none were a better fit than univariable models.In univariable models of SFG rickettsioses, the only statistically significant association (p < 0.05) identified was slaughtering sheep (OR 4.63, 95% CI 1.08-23.50;Table 5).No significant associations were identified between herding livestock with dogs and exposure to SFG Rickettsia or SFG rickettsioses.

DISCUSSION
We found high levels of exposure to C. burnetii and SFG Rickettsia among febrile patients seeking outpatient care at a hospital that serves a predominantly pastoralist population in northern Tanzania.For acute Q fever and SFG rickettsioses, large proportions of the study population met probable case definitions, although confirmed case numbers were smaller.In contrast, exposure to TG Rickettsia was low, observed in <5% of participants.Livestock-related activities were associated with increased odds of exposure to C. burnetii and SFG Rickettsia.
Hospital presentation during the dry season was associated with increased odds of exposure to C. burnetii and acute Q fever.Our findings add to the growing evidence that C. burnetii and SFG Rickettsia are important causes of disease in northern Tanzania that warrant further study particularly among livestock-keeping populations.Livestock-related activities were associated with exposure to C. burnetii, and SFG Rickettsia.Handling aborted livestock products was associated with exposure to C. burnetii in this population, likely due to the localization of the bacteria to the uterus and products of conception in infected animals [9].Participants who herded cattle had higher odds of exposure to SFG Rickettsia compared with those who did not herd cattle.This is consistent with associations between human SFG Rickettsia seropositivity and cattle contact described in southern Tanzania and plausibly explained by a cattle ectoparasite acting as source of infection for humans in these Tanzanian contexts [12].The odds of exposure to SFG Rickettsia increased with each year of participant age (aOR 1.04, CI 1.02-1.06).This could be due to the accumulation of exposure probability over time.Slaughtering animals was associated with increased odds of acute Q fever in this study population.This association is consistent with the findings from studies of nomadic pastoralists in northeast Kenya and abattoir workers in Australia [14,24].Previous studies have shown specific increased risk associated with slaughter of pregnant animals due to release of C. burnetii from the animal's uterus [23,25].In this study population, milking animals was protective against acute Q fever; this association merits further study.
Many of the livestock-related activities associated with exposure to C. burnetii or SFG Rickettsia could also be plausibly associated with human tick-contact from close and frequent contact with animals.IFA testing is a group-specific tool detecting antibody to antigens of all SFG Rickettsia, thus we cannot differentiate which Rickettsia species was associated with exposure in this study population.Most cases of SFG rickettsioses in returning travellers from sub-Saharan Africa are attributed to R. africae that is transmitted by livestock-associated ticks.In this study dataset, the only variable measuring contact with dogs, or their ticks, related to the use of dogs for herding.Herding with dogs was not associated with exposure to either C. burnetii or SFG Rickettsia or on outcomes of acute Q fever or SFG rickettsioses.Previous work in Tanzania found no association between dogs and SFG Rickettsia exposure, but did find an association between SFG Rickettsia exposure and livestock density within a community [12].Together, these findings suggest that R. africae rather than R. conorii may be the more frequent cause of SFG Rickettsia exposure in Tanzania.In many cases, human interactions with livestock and dogs may be correlated due to the use of dogs for livestock herding and the co-distribution of these animals at the population and activity level.Further study is needed to disentangle the specific SFG Rickettsia pathogens in Tanzania and further understand risk factors for human exposure to tick vectors.
Presentation to the hospital during the dry season was associated with increased odds of exposure to C. burnetii and of acute Q fever.Coxiella burnetii is highly stable in the environment and infectious aerosols may be created from environmentally persisting bacteria long after the release of bacteria from an infected animal [26].Aerosolized transmission of C. burnetii between infected herds, and from livestock or their environments to humans is known to be associated with windborne transmission and open landscapes where aerosols or spore-like particles are more likely to be dispersed during dry seasons [27].The risk of C. burnetii infection in rural areas is highest within 5-10 km of farms with infected animals [28].Understanding the mechanics of transmission is key to inform context-specific approaches to preventing human cases and minimise disease.
The proportions of participants seropositive for C. burnetii and SFG Rickettsia in this predominantly pastoralist population were 51.3% and 77.2%, respectively.The prevalence of serologic exposure to C. burnetii was higher in this northern Tanzanian population compared with other studies in sub-Saharan Africa that have demonstrated seroprevalence of 1%-32% [6].Exposure to SFG Rickettsia in this study population was similar to that seen in previous studies in rural areas of southwestern Tanzania and Senegal [12,29].Further understanding of the immune response to and longterm persistence of antibodies against C. burnetii and SFG Rickettsia is necessary to maximise epidemiologic and clinical inference from antibody detection.There are limited data on long-term antibody persistence after natural infection by C. burnetii, and the role of antibodies in protection against C. burnetii disease is unknown.In rickettsial infections, it is thought that antibodies play a minor role in clearance of primary infection [30].Host factors, pathogen factors, and exposure profiles including the timing and frequency of pathogen exposure may all play a role in the progression from pathogen exposure to acute illness with C. burnetii and SFG Rickettsia, and further study of these determinants is needed.
Laboratory diagnosis of Q fever and SFG rickettsioses is challenging due to the need for both acute and convalescent sera to confirm diagnosis.In Tanzania, patients with unexplained fever rarely receive empiric treatment for Q fever or rickettsial infections [5,30].Doxycycline, an accessible antimicrobial in many low-and middle-income countries, including Tanzania, effectively treats acute Q fever and SFG rickettsioses [1].However, clinical management algorithms for fever applied in Tanzania focus predominantly on malaria and invasive bacterial disease.Tetracyclines are not included in local or national guidelines for febrile illness, and receive scant attention in international guidelines for the empiric treatment of severe febrile illness or sepsis [32][33][34].Untreated Q fever and rickettsial disease are not without risk of severe illness and death [3,8,11,35].Our findings support re-evaluation of guidelines and algorithms for the assessment and management of patients with severe febrile illness in this setting [1,31].
We acknowledge the limitations to our study.We were unable to directly measure exposure to ectoparasites and used proxy variables for vector exposure such as activities involving close animal contact.We also had little data on the interactions of participants with rodents and dogs, animals known to harbour ectoparasites capable of transmitting SFG and TG Rickettsia, limiting our ability to explore any risks associated with these animals.We lacked clinical data on many symptoms and signs associated with the diseases investigated, potentially contributing to misclassification of cases.Our study population is biased towards those with increased hospital-based health-seeking behaviour and may not be generalizable to the wider population.The case definitions used in this study were adapted from those used principally for diagnosis in the United States and these definitions may not be appropriate for populations with more frequent serologic exposure to the study pathogens.It is plausible that high levels of serologic exposure to C. burnetii and Rickettsia spp. in this population reduced the specificity of probable acute disease case definitions, leading to over-estimation of the prevalence of probable cases.Delayed hospital presentation for acute illness may have reduced the probability of detecting seroconversion and thus a likely under-estimation of the prevalence of confirmed illness.

CONCLUSIONS
We demonstrate that exposure to C. burnetii and SFG Rickettsia was common among patients seeking care for fever in a predominantly pastoralist area of rural northern Tanzania, and that substantial proportions of participants met criteria for probable acute illness due to these two pathogens.Direct interactions with livestock conferred increased odds of both exposure and acute illness with C. burnetii and SFG Rickettsia.Further understanding of the complex interactions between humans, livestock, and ectoparasites is needed to disentangle the epidemiology of these pathogens, improve clinical management, and ultimately develop disease prevention strategies for Q fever and SFG rickettsioses.
Demographic and clinical characteristics of study participants, Endulen Hospital, Tanzania, 2016-2017.Flow chart showing the steps in the screening, enrolment, and data collection from the study population of individuals with febrile illness seeking care in the outpatient department of Endulen Hospital from August 2016 through October 2017.T A B L E 2 Univariable and multivariable logistic regression analyses for factors associated with C. burnetii exposure among febrile study participants, Endulen Hospital, Tanzania, 2016-17.Univariable and multivariable logistic regression analyses for factors associated with acute Q fever among febrile study participants, Endulen Hospital, Tanzania, 2016-17.Univariable and multivariable logistic regression analyses for factors associated with spotted fever group (SFG) Rickettsia exposure among febrile study participants, Endulen Hospital, Tanzania, 2016-17.
T A B L E 1Abbreviations: IQR, interquartile range; SFG, spotted fever group; TG, typhus group.Note: Bold indicates variables included in multivariable analyses.Asterisk denotes question answered as yes or no.All variables with (y) notation refer to the performance of or exposure to the stated activity in the 12 months prior to presentation.All variables with (m) notation refer to the performance of or exposure to the stated activity in the 1 month prior to presentation.High risk occupation includes farmers, livestock attendants and those who worked with wildlife.All animal-related activities include cattle, sheep and goats in all cases and other species for specific questions.Abbreviations: aOR, adjusted odds ratio; CI, confidence interval.T A B L E 3Note: Bold indicates significant variables included in multivariable analyses.Asterisk denotes question answered as yes or no.All variables with (y) notation refer to the performance of or exposure to the stated activity in the 12 months prior to presentation.All variables with (m) notation refer to the performance of or exposure to the stated activity in the 1 month prior to presentation.Acute Q fever includes those with probable and confirmed cases.High risk occupation includes farmers, livestock attendants, or worked with wildlife.All animal related activities include cattle, sheep and goats in all cases and other species for specific questions.Abbreviations: aOR, adjusted odds ratio; CI, confidence interval.T A B L E 4