Incidence of Brucella spp. in various livestock species raised under the pastoral production system in Isiolo County, Kenya

Background Brucellosis is an important zoonosis with a worldwide distribution. The disease is caused by multiple species of Brucella that can infect a wide range of mammalian hosts. In the sub-Saharan Africa, many studies have been implemented to determine the prevalence of the disease in livestock, but not much is known about its incidence. We implemented a longitudinal study to determine the incidence of Brucella spp. infection in cattle, camels, sheep and goats that were being raised in a pastoral area in Isiolo County, northern Kenya. Methods An initial cross-sectional survey was implemented to identify unexposed animals for follow up; that survey used 141 camels, 216 cattle, 208 sheep and 161 goats. A subsequent longitudinal study recruited 31 cattle, 22 sheep, 32 goats and 30 camels for follow up. All the samples collected were screened for Brucella spp. using the Rose Bengal Plate test (RBPT), a modified RBPT, and an indirect multispecies Enzyme Linked Immunosorbent Assay (iELISA) kit. Samples that tested positive by any of these serological tests were further tested using real-time PCR-based assays to detect genus Brucella DNA and identify Brucella species. These analyses targeted the alkB and BMEI1162 genes for B. abortus, and B. melitensis, respectively. The longitudinal study took 12 months and data were analysed using Cox proportional hazards model that accounted for clustering of observations within herds. Changes in anti-Brucella IgG optical values between successive sampling periods were determined to confirm primary exposures. Results The mean incidence rate of Brucella spp. was 0.024 (95% confidence interval [CI]: 0.014 – 0.037) cases per animal-months at risk. Brucella spp. incidence in camels, cattle, goats and sheep were 0.053 (0.022 – 0.104), 0.028 (0.010 – 0.061), 0.013 (0.003 – 0.036) and 0.006 (0.0002 – 0.034) cases per animal-month at risk, respectively. A higher incidence rate of Brucella spp. was found among females (0.020, 0.009 – 0.036) than males (0.016, 0.004 – 0.091), while young animals (0.026, 95% CI; 0.003 – 0.097) had a slightly higher incidence rate compared to adults (0.019, 95% CI; 0.009 – 0.034). RT PCR analyses showed that B. abortus was more prevalent than B. melitensis in the area. The results of multivariable Cox regression analysis identified species (camels and cattle) as an important predictor of Brucella spp. exposure in animals. On the diagnostic tests, modified RBPT provided similar findings as the iELISA test. Conclusions Our findings indicated that camels and cattle have a higher incidence of Brucella spp. exposure than the other livestock species. This could be due to the higher prevalence of B. abortus, which readily infects these species, than B. melitensis. More studies are underway to identify ecological factors that influence the persistence of the key Brucella species in the area. The study further concluded that modified RBPT test can give reliable results as those of a formal iELISA test, and it can therefore be used for routine surveillance in the region. Author summary Brucellosis is a neglected disease that is endemic in many pastoral areas. This study describes the incidence patterns of Brucella spp. in various livestock species (cattle, camels, sheep and goats) in Kinna in Isiolo County, northern Kenya. We also evaluated the diagnostic sensitivity of three serological tests; RBPT, a modified RBPT and an iELISA test in the diagnosis of brucellosis in animals that were suspected to be naturally exposed. Results from this study showed that both cattle and camels had a significantly higher incidence of Brucella spp. compared to sheep and goats. The number of animals found to be seropositive for Brucella spp. by the modified RBPT and iELISA did not differ significantly. Both tests also detected a significantly higher number of seropositive animals than RBPT. This finding confirms that the modified RBPT provides comparable results as iELISA, which is known to have higher sensitivity and specificity, and therefore the former can be used for more surveillance activities in pastoral areas.


74
Brucellosis is an important zoonotic disease that affects a great variety of hosts such as 75 livestock (cattle, sheep, goats and camels), humans and wildlife [1]. Whereas this disease has 76 been successfully controlled or eradicated in livestock populations in many developed 77 countries including New Zealand, Japan and Australia [2], it remains a major problem affecting 78 both livestock production and humans in Kenya [3], and also other parts of Africa [4]. 79 Brucellosis causes direct production losses resulting from abortions, stillbirths, infertility, the 80 mortality of calves/kids/lambs, longer calving intervals, reduced draught power, poor weight 81 gain, and reduced milk production [5]. The etiological agent of this disease is an intracellular  There are limited studies that have been carried out to understand the epidemiology of 92 Brucella spp. in Kenya [3] even though this pathogen is known to be endemic in pastoral areas 124 125

126
Study area 127 This study was conducted in Kinna ward in Isiolo County, northern Kenya (Fig 1). The 128 area was selected purposively due to good accessibility and reliable security. In addition, a 129 previous survey (in press) that involved the screening of milk for Brucella spp. using milk ring 130 test and real time PCR indicated that Kinna had a higher prevalence of Brucella spp. compared 131 to other areas that were surveyed in Isiolo and the neighbouring Marsabit counties. Pastoral 132 livestock production system is the main cultural and economic activity for the local people 133 because the area is semi-arid [16]. The average annual rainfall is 580 mm [16], and ranges 134 between 350 and 600 mm [17]. Rainfall in the area has a bimodal distribution; long rains occur  Fisher's exact test was also used to determine the association between categorical predictors 245 (animal's sex, age, pregnancy status, and species) and the outcome variable. The aggregated 246 data from all the animals was also further subset by the livestock species and the above 247 categorical predictors assessed for their association with the outcome variable.

249
For the cross sectional data, risk factor analysis was done at the animal-level. We did For the analysis of the longitudinal data, we first removed cases that were classified as 264 being positive during the cross-sectional study to remain with uninfected animals. We also did 265 a secondary analysis to detect the change in OD levels from the iELISA test in newly infected 266 animals between the first and subsequent sampling periods over the longitudinal phase.
Animals were followed on monthly basis until they got exposed. Animal-time (in 269 months) at risk for each animal was generated and aggregated to obtain the denominator for 270 the overall Brucella infection incidence. The numerator was the total number of new Brucella 271 spp. cases recorded during the follow-up period. The estimation of the mean incidence rates 272 with their respective 95% CI (overall mean as well as by livestock species, sex and age) were 273 performed using the epi.conf function in epiR package [26].

274
This study also determined the hazard rate ratio for the above categorical variables 275 using univariable and multivariable Cox proportional hazards models. In these analyses,

276
Brucella spp. exposure in animals and time at which the exposure was detected, were both 277 included in the Cox regression models as the outcome of interest. We fitted data to these models proportion of the sampled animals, 309 (36.7%) were not included in the analysis due to either 289 missing epidemiological data or they were not tested using RBPT and mRBPT due to logistical 290 constraints. The overall seroprevalence of Brucella spp. at animal-level was 11.3% (95% CI; 8.6-14.0, n = 532) based on the parallel interpreted results of the three serological tests (Table   292 2 species grouped by animal sex, age and pregnancy status are summarized in Table 3.   using McNemar's χ 2 showed significant differences between RBPT and mRBT (P < 0.001),

Conflict of interest 479
The authors have declared that no competing interests exist. All relevant data are within the paper and its supporting information files.