Ethics statement

We obtained ethical approval for human and livestock sampling from the Institutional Research Ethics Committee at the International Livestock Research Institute (reference number: ILRI-IREC2020-07). The household heads provided written consent for livestock sampling, and written consent was obtained from all selected humans for sampling. The study clinicians informed all participants that their participation was voluntary, and they could decline without any consequences. For adults (aged >18 years), written consents were provided; for children aged between 13–17 years, written assent forms and parental/guardian permissions were obtained. For children under 13 years, only written parental/guardian permissions were obtained.

Study area

The study was conducted in Garbatula subcounty in Isiolo County. Isiolo County is in the northern region of Kenya (Fig 1). The subcounty is subdivided into three administrative wards: Garbatulla, Kinna and Sericho. The study area is in a relatively low-lying region, with Kinna ward having the highest altitude of 589 meters (m) and the lowest area being Sericho at 459m. The area receives an average of 580 millimetres of rainfall per annum, with November and April being the wettest months. The area is characterized by hot and dry climate for most of the year, with a mean annual temperature of 29°C, with variations due to differences in altitude [30].

Garbatula subcounty is sparsely populated, with an average population density of nine people per square kilometre [31]. The Borana tribe is the most predominant, and livestock is a source of livelihood for over 80% of the inhabitants [32]. Cattle, sheep, goats, and camels are the main livestock species raised in the area, although goats and sheep have the highest population. The subcounty shares a border with Meru National Park and the Bisanadi Game Reserve, two unconfined sanctuaries for wild animals. As a result, there are frequent interactions between domestic livestock and wild animals in certain regions of the subcounty.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Map of Africa and Kenya showing sampling sites within Garbatula subcounty (The underlying map layer used to generate this figure was obtained from https://www.diva-gis.org/gdata).

https://doi.org/10.1371/journal.pntd.0011682.g001

Sample size determination

The sample size (n) for humans and animals (camel, cattle, sheep, and goats combined) was estimated using the formula; [33]. The expected (a priori) seroprevalence estimate (p) of Brucella species was assumed to be 50% in both humans and animals since there were no studies on Brucella burden from the area. It was further assumed that the study would estimate this outcome with an error margin (d) of 0.05.

Previous studies have shown that Brucella spp. seroprevalence clusters in humans and animals at the household and herd levels. A design effect (Deff) was therefore determined for adjusting the naïve sample sizes by assuming that an average of twenty animals per herd and three humans per household would be sampled and that Brucella spp. seroprevalences have intracluster correlation coefficients (ICC) of 0.2 in animals and 0.21 in humans, respectively [24]. DEFF estimates of 4.8 and 1.42 were derived for animals and humans, respectively. Minimum sample sizes of 546 humans and 1844 animals were therefore estimated based on multiplying the DEFF estimates with the naïve sample sizes.

Sampling design for livestock and humans

Livestock and human data were collected in July-August 2021 in a cross-sectional, two-stage random sampling design. The primary sampling units were individual livestock and humans within randomly selected households. Households were eligible for selection if they kept or owned any of the four livestock species and had at least one consenting adult human. To identify study households, 250 random geographical coordinates (RGC) were generated using Quantum Global Information System (QGIS) version 3.4.4 software (https://qgis.org), for Garbatula Subcounty and households closest to an RGC were selected for sampling. For the coordinates that fell near a cultural homestead (Manyatta) where livestock from different households are kept in one enclosure, one household was randomly selected, and only humans and animals owned by the household were eligible for sampling.

A household was defined as a group of people using a common cooking area. A total of five animals were selected per species (camel, sheep, cattle, and goats) in each herd using systematic random sampling. Animals under one month and the very sick were excluded from the study. In herds with fewer than five animals per species, all animals were selected. All household members were eligible for enrolment, excluding children under two years since the ethical approval did not cover them.

Sample collection in livestock and humans

Qualified veterinarians collected livestock samples. During sampling, animals were restrained by trained animal handlers. Using aseptic techniques, blood was collected from the jugular vein of each animal into 10 ml plain vacutainers with unique barcode labels. The vacutainers were stored in cool boxes with icepacks to maintain temperatures of 4–8°C in the field before transfer to the local field veterinary laboratory for serum extraction and temporary storage at -20°C. The samples were centrifuged at 2500 gravitational force for 15 minutes, and the serum was stored in 2 ml barcoded cryovials. Extracted serum samples were transported at -20°C using motorized freezers to the International Livestock Research Institute (ILRI) Laboratories in Kenya every five days, where they were first stored at -20°C before testing for antibodies to Brucella species and then transferred to -80°C for long-term storage. Qualified clinicians conducted the sampling in humans. Each participant had 5 mL of blood drawn from their median cubital vein on the left hand using plain vacutainers with unique barcode labels. All blood samples were immediately placed in a cool box and maintained at 4–8°C in the field before being transferred to nearby health centres, within four hours. The procedures for extracting serum, storing and transporting samples were identical to those used for animal samples. Testing of human samples was conducted at ILRI.

Data collection

Data was collected by the trained study staff using standardized questionnaires installed in Open Data Kit (ODK) electronic data collection tool. The questionnaires were uploaded to smartphones and were pretested in the field before deployment. Clinicians administered household and individual human questionnaires, while veterinarians administered the herd-level and individual animal questionnaire. The first questionnaire was administered to household heads to collect herd-level animal husbandry practices linked with the transmission of Brucella spp. such as herd size, breeding between herds, sharing of grazing areas and sale or introduction of new animals into the herd. Individual-level animal data of the recruited animals were then collected, including age, sex, species (cattle, camel, goat, or sheep) and history of reproductive syndromes (such as abortion, weak birth, stillbirth, infertility, and swollen testis).

Livestock species were categorized into different age groups based on their age range. The first category included suckling animals such as sheep and goats (0–3 months old) and cattle and camels (0–6 months old),- and these animals usually stay close to households. The second category included weaners or young adults—no longer dependent on milk and often separated from their mothers. This category had goats and sheep that were 7–12 months old and cattle and camels that were seven months to two years old (for cattle) and up to three years old (for camels). Animals that were above these age ranges were categorized as adult animals that are of reproductive age.

Finally, human subject-level data such as demographic information, history of febrile illness, consumption of raw milk and milk products, handling of animals and animal products and occupational practices (which may include herding, milking, slaughtering, and assisting in animal births) were collected.

Laboratory testing

Livestock serum samples were tested using the ID Screen Brucellosis Serum Indirect Multi-species kit (Grabels, France) for the detection of anti-Brucella spp. IgG antibodies in bovine, ovine, caprine, and porcine samples [34]. Briefly, the 10 μl of serum samples were diluted using 190 μl of dilution buffer two and loaded onto the pre-coated ELISA reaction plate. The test plates were incubated at room temperature for 45 minutes, after which they were washed three times. This was followed by the addition of 100 μl of the conjugate solution before the plate was incubated at room temperature for 30 minutes. After three washes, 100 μl of the substrate solution was added, and the plate was incubated in darkness for 15 minutes. The reaction was stopped using 100 μl of the stop solution, and immediately, the optical densities (ODs) of each well were determined at 450 nm wavelength. All the samples were run in duplicates with each plate comprising positive and negative control. The results for each plate were considered valid if the mean OD for the positive control was greater than 0.35 and the ratio of the OD values between the positive and negative controls was greater than 3. The percentage S/P was calculated for each sample as described by the manufacturer. Following the manufacturer’s recommended cut-off values, all samples whose percentage S/P was greater than or equal to 120% were considered positive, negative if less than or equal to 110% and doubtful if greater than 110% but less than 120%.

All human serum samples were tested using IBL-America IgG enzyme-linked immunosorbent assay (ELISA) kits (IBL-America, Minneapolis, USA) according to the manufacturer’s recommendation [35]. Briefly, the sera were diluted at 1:100 using sample diluent and added onto antigen pre-coated microtitre plates followed by incubation for 1 hour at 37° C. Thereafter, 100 μl of ready to use conjugate was added and incubated for 30 minutes at 37° C before adding para-Nitrophenyl Phosphate (pNPP) substrate and further incubated at 37°C for 30 minutes. The conjugate–substrate reaction was conducted by adding 100 μl of ready-to-use stop solution included in the kit. The test sera ODs were read at 405 nm against a substrate blank and 630 nm was used as a reference wavelength to correct for the background fluorescence. Wash steps preceded both conjugate and substrate addition steps, while all sera and controls were run in duplicates. The results for each plate were valid when the substrate blank OD was less than 0.25 and the negative control gave negative results as per the interpretations. We first calculated the upper and lower cut-off limits to interpret our results. The upper cut-off limit was calculated as a product of the serum OD, the kit-specific correction factor (0.666) and the kit-specific grey zone given as 14% of the mean standard OD divided by Positive control OD. To calculate the lower cut-off limit, the kit specific grey zone was subtracted from the product of sample OD and the kit specific correction factor. Samples were interpreted as positive if their mean OD fell above the upper cut-off limit, doubtful if between the upper and lower cut-off limits, and negative if below the lower cut-off limit. Doubtful samples were considered as negative in subsequent analyses.

Data management and statistical analyses

The serological and questionnaire data were first combined into one comma-delimited value (CSV) file and cleaned prior to analysis. We performed all analyses using the R statistical environment, version 3.6.0 [36], with descriptive statistics being analyzed using the gmodels [37] and epiR [37] packages. The first analysis was the calculation of animal/subject and household/herd levels overall seroprevalence estimates of Brucella spp., with a 95% confidence interval (CI) in humans and livestock. Seroprevalence was assessed at the individual, household, and herd levels. A household or herd was considered seropositive if at least one human or animal, respectively, tested positive. Additional analysis of livestock data included the estimation of animal/subject seroprevalence by categorical variables, which included animal species (camel, cattle, goat, and sheep), animal sex (male and female), animal age (suckling/lamb/calf, weaner/youngling, and adult) and herd size (<200, 200–400,>400). For human data, seroprevalence was estimated for gender (male and female), age (4–20,21–40,41–60,>60), and the highest level of education completed (no education, primary, secondary, and above). Herd size and age for humans were considered as quantitative discontinuous variables. To determine the normality of our sample data, we initially performed the Shapiro-Wilk test to check for the normality of residuals. These two variables violated the linearity assumption and were transformed into both log-transformed and categorical variables and then compared with the outputs. We also determined the correlation between herd level exposure to Brucella spp. with human exposure. We defined an exposed herd as any herd that had at least one seropositive animal.

To start the statistical analyses, we employed Chi-square (χ2) test to assess the independence of categorical variables. Additionally, we conducted univariable mixed-effects logistic regression models to investigate the association between independent variables and the outcome variable. We used the generalised linear mixed models to conduct both univariable and multivariable analyses. All variables with a p-value lower than 0.20 from the univariable analysis were selected for the multivariable analysis [38]. The lme4 package was used to fit the human and livestock data using the glmer function with random effects for household ID [38]. The final model was done through backward elimination. Firstly, unconditional models without independent variables were fitted for the livestock and human data, and then global models for variables selected from the univariable data. We systematically eliminated variables with p values lower than 0.05 until we achieved the lowest Akaike’s information criteria (AIC) in the multivariable model. The final models from this procedure were then assessed for the goodness of fit through inspection of residual plots against the fitted values from the model [39]. Finally, we tested for potential interaction effects within the selected variables by conducting a likelihood ratio test (LRT) that was used to determine if there were potential interaction effects among product pairs of the factors in the final multivariable model [40]. Furthermore, we estimated the ICC for within-herd and within-household clustering of animal and human brucellosis using the icc function available in the performance package [41]

Descriptive statistics

A total of 2,157 animals (258 cattle, 953 goats, 849 sheep and 97 camels) were sampled from 231 households and tested for anti-Brucella antibodies against Brucella spp. In eleven households (3.0%), animals were not available for sampling. In Garbatula, Sericho and Kinna wards, 977 (45.1%), 551 (25.5%) and 629 (29.2%) animals were sampled, respectively. Of the total sample, 1660 (77.0%) were female and 497 (23.0%) were males. The median herd size, combining cattle, sheep, goats, and camels was 176 (range: 8–1113), while the median herd/flock sizes for cattle, goats, sheep, and camels were 80 (range: 1–460), 80 (range: 8–546), 79 (range: 4–626) and 50 (range: 19–100), respectively. The median number of total animals sampled per household was 10 (range: 1–16).

This study also sampled 683 humans from 242 households. Most humans were sampled in Garbatula ward (39.1%, n = 267), followed by Sericho ward (34.8%, n = 238) and Kinna ward (26.1%, n=178). More males (83.0%, n = 567) were sampled than females (17.0%, n = 116). The median age of sampled subjects was 34 (range: 4–82), while the median number of individuals sampled in each household was 3 (range: 1–8).

Seroprevalence of Brucella spp., in livestock and humans and univariable mixed-effects logistic regression model results from the analysis of livestock and human data

The univariable mixed-effects logistic regression model results from the analysis of livestock data, with households as random effects, are shown in Table 1. The overall animal and herd-level seroprevalence estimates of Brucella spp. in livestock were 10.4% (95% CI: 9.2–11.7) and 51.7% (95% CI: 47.9–55.7), respectively. The animal-level seroprevalence differed among livestock species. Camels had the highest seroprevalence of 19.6% (95% CI: 12.4–27.3), followed by goats at 13.2% (95% CI: 9.3–17.1), cattle at 13.1% (95% CI: 11.1–15.3) and sheep at 5.4% (95% CI: 4.0–6.9) in decreasing order (Table 1). Female animals had higher seroprevalence of 11.6% (95% CI: 10.2–13.2) compared to male animals with a prevalence of 6.2% (95% CI: 4.4–8.4). When stratified by age, the seroprevalence increased with age with adult animals having a higher seroprevalence of 11.9% (95% CI: 10.4–13.4) compared to young adults at 3.1% (95% CI: 1.4–4.8) and calves 2.8% (95% CI: 0–5.7). Animals from Sericho ward had a higher seroprevalence estimate of 12.9% (95% CI: 10.3–15.7) compared to Kinna at 12.7% (95% CI: 10.3–15.4) and least in Garbatula at 7.5% (95% CI: 5.9–9.1). Herds with more than 400 animals had higher odds of seropositivity compared to animals from smaller herds. Additionally, herds, where new animals were introduced, reported higher odds of being seropositive compared to herds that did not introduce new animals. Finally, herds with at least two animal species had significantly higher odds of exposure to Brucella spp. compared to those with one species. The results are given in Table 1.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Univariable mixed-effects logistic regression analysis of the factors associated with livestock brucellosis seropositivity.

https://doi.org/10.1371/journal.pntd.0011682.t001

Results for the human univariable models are shown in Table 2. The individual-level seroprevalence of Brucella spp. in humans was 54.0% (95% CI: 50.2–58.0) while the household seroprevalence was 86.4% (95% CI: 84.0–89.0). Exposure to Brucella spp. was significantly higher in males 56.8% (95% CI: 52.7–61.2) compared to females 40.5% (95% CI: 31.9–49.9). When data were stratified by age, the individual human seroprevalence estimate increased with age. On education level, higher seroprevalence was among those who had no formal education at 57.5% (95% CI: 53.3–62.1) followed by those who completed primary education at 47.0% (95% CI: 38.5–56.8) and it was least among those attained secondary or higher education at 31.6% (95% CI: 21.7–47.6). Residents of Sericho had a higher seroprevalence of 60.9% (95% CI: 54.6–67.2) compared to those in Garbatula at 52.4% (95% CI: 46.4–58.8) and Kinna at 47.2% (95% CI: 39.9–54.9). Among the risk factors for exposure to Brucella spp. analyzed, consuming raw milk from any species had higher odds of seropositivity than consuming pasteurized milk. Additionally, handling livestock was identified as a significant risk factor, but we found no significant association with Brucella spp. exposure among those who handled wild animals or consumed bushmeat.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Univariable mixed-effects logistic regression analysis of the factors associated with human brucellosis seropositivity.

https://doi.org/10.1371/journal.pntd.0011682.t002

Multivariable analyses

The results from the multivariable mixed-effects model for animal data are summarized in Table 3. Four fixed effects were significant predictors of Brucella spp. exposure among animals, including animal sex, age, species, and herd size. Specifically, female animals had 1.7 times higher odds of being seropositive compared to male animals. The odds of being seropositive increased with increasing animal age, and adult animals were seven times more likely to be exposed to Brucella spp. than calves. Sheep had significantly lower odds of being seropositive compared to goats. However, cattle or camel did not have a significant effect on the likelihood of exposure when compared to goats. Animals in larger herd sizes (above 400 animals) had significantly higher odds of Brucella spp. exposure compared to animals in smaller herds, with the likelihood of exposure increasing with the increase in the size of the herds. The history of abortion variable in sampled animals was excluded from the multivariable models because it only affected adult females and would not improve the model’s predictability, despite its significant association with exposure to brucellosis in the univariable analysis. The ICC estimate for the household random effect in the final model was 0.138 (95% CI: 0.554–1.046), while the LRT χ2 for the product pairs created for the independent variables in the final models was not significant (p ≤0.05), suggesting that the model’s predictors were not significantly collinear.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Multivariable mixed-effects logistic regression analysis of the factors associated with livestock brucellosis seropositivity.

https://doi.org/10.1371/journal.pntd.0011682.t003

The outputs from the multivariable mixed-effects model for human data are summarized in Table 4. Three variables were significant: gender, education level, and herd exposure. Males had twice the odds of exposure compared to females, while individuals who did not complete any formal education had three times higher odds of exposure compared to those who completed secondary education or higher. The odds of exposure decreased with an increase in educational attainment.

Consumption of raw milk did not significantly affect the likelihood of exposure. There was a strong correlation between human seropositivity and livestock seropositivity for Brucella spp., with almost two-fold higher odds of exposure to people from households with at least one seropositive animal than those without. The ICC value calculated from the variance estimates of the final multivariable model was 0.03 (95% CI; 0.01–0.74).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Multivariable mixed-effects logistic regression analysis of the factors associated with human brucellosis seropositivity.

https://doi.org/10.1371/journal.pntd.0011682.t004

Limitations

Our study had a few limitations. The study was conducted during a hot and dry season, a period often characterized by adaptation strategies, including an increase in droughts-adapting livestock, herd splitting, household splitting and clustering of herds for security reasons and in areas proximal to pasture and water for their livestock. [56, 57]. Additionally, the study included a few animals below six months of age, which have the potential to carry maternal antibodies. This introduced a small risk of obtaining false positive results. Finally, the study employed ELISA exclusively for serological testing, despite its sensitivity and specificity being subject to fluctuations, particularly in endemic areas. Moreover, ELISA is susceptible to cross-reactivity with closely related pathogens, potentially resulting in false-positive outcomes. Additionally, the test may fail to detect active infections, thereby resulting in false-negative results [35, 58, 59]. To enhance accuracy, it is advisable to combine ELISA with other serological tests, such as the Rose Bengal test (RBT) and the serum (tube) agglutination test (SAT), along with molecular techniques [6]

Conclusion and recommendations

Our study reports a high prevalence of brucellosis in humans and livestock in the study area, underscoring the need for targeted prevention and control measures to curb its spread. The study revealed several factors associated with Brucella spp. seropositivity in humans and livestock, including animal age, herd size, livestock species, gender, education level, and herd exposure. Enhancing animal health management procedures, encouraging appropriate hygiene standards when handling animals, and promoting the consumption of safe animal products are three vital ways of mitigating disease transmission. This can involve creating more public awareness about the disease to minimize the risk of disease transmission and increasing veterinary outreach services by the County Government of Isiolo to pastoralists, such as vaccination. In addition, the One Health approach applied in this study emphasizes the significance of understanding the interactions between humans and animals in transmitting and controlling zoonotic diseases, showing the need to establish an integrated One Health surveillance system in such high-risk populations. Future research should examine the efficacy of measures to mitigate the spread of brucellosis and the socio-economic impacts of brucellosis in the region.