Ecological setting of study area
The study was conducted in Niger State in the Southern Guinea Savannah zone of Nigeria, between latitudes 8° 20′ N and 11° 30′ N, and longitudes 3° 30′ E and 7° 20′ E. The state serves as transit routes for pastoral herds on seasonal transhumance movements between northern and southern parts of Nigeria. It has three designated Agro-ecological zones: southern, eastern and northern zones, with variable climatic conditions (Fig. 2). These zones are characterized by many rivers, streams and ponds, fadamas for rice farming and four hydroelectric dams. There are also Kainji National Game Reserve and many transnational stock routes.
The state experiences two distinct seasons: rainy season (April to October) and dry season (November to March), with mean annual rainfall of about 150 cm spanning for a period of approximately 180 days. It has average annual temperature range of 22°C to 39°C and relative humidity of about 58.6%. These ecological variables predispose the state to annual flooding and consequently provide suitable breeding environments for vectors of vector-borne diseases, such as RVF, in water-filled topographic depressions, the ‘dambos’. The state has an estimated cattle population of 2.5 million, which are mostly in the custody of pastoralists .
Study design and target population
A cross-sectional study was conducted in pastoral herds, herding local breeds of cattle in two production systems in North-central Nigeria, between October 2017 and September 2018. Both serological and questionnaires tools were used for sample collection. For seropositivity investigation, average nomadic and agro pastoral herd sizes were 50 and 25 cattle, respectively. Sampled cattle were of both sexes, aged at least one year to exclude the effect of colostral antibodies and no vaccination history. Accessibility of herds was also considered, with insecure areas being excluded. Using questionnaire tool, age eligibility for pastoral household heads that participated was 20 years or above. They were expected at these ages to possess existing veterinary knowledge on livestock health management and diseases risk factors .
For the purpose of this research, a nomadic production system was defined as management that kept mainly cattle and took part in year-round movements of herds over large ranges for grazing without a permanent homestead. An agro-pastoral production system was a semi-settled herd with small number of cattle, cultivating few crops and having limited movements on low range grazing within their environments.
Sample size and sampling procedure
For animal seroprevalence, sample size was determined using random sampling for finite population, with power set at 11.3% , 6% desired absolute precision at 95% confidence level. A sample size of 107 cattle was obtained. Sample size of households for questionnaire administration was determined using the same approach, with power set at 50% frequency of response; margin of error was 5% at 95% confidence level. A sample size of 384 households was obtained. To take care of non-response, a 5% contingency was added. Thus, 403 household heads were targeted for data collection.
A multi-stage sampling method was carried out. For questionnaire administration, three agro-ecological zones were purposively considered in the first stage. In the second stage, 15 settlements were selected for each production system (30 pastoral settlements in all) across the State, with five from either nomadic or agro pastoral herds in each Agro-zone. In the final stage, 134 pastoral households (67 from either group) were randomly selected in each zone. A total sample of 403 respondents, made up of 202 nomadic and 201 agro-pastoralists, were selected. For the seropositivity, 10 herds (5 nomadic and 5 agro-pastorals) were purposively selected in each zone. Also, a minimum of 3 cattle were randomly selected proportionately to the size of each herd. Agro-ecological zones A and C: 35 cattle each; and zone B: 37 cattle.
Data collection: serum samples and serological analysis
Blood samples were collected from 107 cattle in the three agro-ecological zones of Niger State. During sampling, ages of the animals were recorded, and if the herder was not aware of an animal’s age, the dentition of the cattle was used to estimate the age. Sampled cattle were classified according to age: 1–3 years and >3 years. Blood samples (5mls) were collected in dry vacutainer tubes from the jugular vein of each animal. Each vacutainer tube was labeled and individual animal information recorded. The collected blood samples were kept at 4oC for 12 hours to allow blood clot, centrifuged at 3,000 g for 10 minutes for erythrocytes sedimentation and serum formation. The sera were transferred into new vials and labeled before being stored at -20oC until further processing.
Serological analysis using IgM capture ELISA was conducted. To detect recent infection (IgM), all samples were tested using ID Screen RVF IgM ELISA (ID-Vet Innovative Diagnostics, Grabels, France) according to the manufacturer’s instructions. The test was considered valid when the mean value of the positive control OD (ODPC) was greater than 0.35 and the ratio of mean values of the positive and negative control ODs (ODPC and ODNC) was greater than 3. The sample was considered positive when the competition percentage was greater than or equal to 50%, doubtful when between 40% and 50%, and negative when ≤40%. All doubtful samples were considered as negative in this study.
Data collection: questionnaire administration
We developed a structured questionnaire with mostly categorical questions to ease data processing and improve precision of responses. It was interviewer-administered by eight trained animal health technicians and supervised by the authors. The questionnaire consisted of four sections: demographic characteristics of respondents (6 questions); herd biodata (4 questions); existing knowledge about RVF (9 questions); and socio-ecological predisposing factors of RVFV occurrence in cattle herds (9 questions). The questionnaire was originally designed in English and verbally translated to local Hausa language during administration for respondents without formal education.
Questionnaire was pre-tested on 15 pastoral cattle herds’ settlements before final administration, and identified problems were eliminated and final high quality data collected. To achieve maximum response, advocacy visit was made to the leader (Dikkos) of each pastoral settlement a week prior to data collection and permission obtained. Respondents were assured of voluntary participation, confidentiality of responses and the opportunity to withdraw at any time without prejudice in line with the World Medical Association Declaration of Helsinki . Informed consent was obtained either by signatures (for literates) or thumb-printings (for illiterates) on a sheet before questionnaire administration and none declined to participate.
Data management and statistical analysis
Data from the field and laboratory were summarized into Microsoft Excel 7 (Microsoft Corporation, Redmond, WA, USA) spreadsheets and stored. Descriptive and analytical statistics were used. Frequencies and proportions were used for descriptive analysis. Categorical variables were presented as proportions and their associations determined by bivariate analysis using Chi-square tests. Associations were analyzed by univariable tests and multivariable logistic regressions analysis.
RVFV seropositivity in animals was measured as the proportion of animals presenting antibodies against RVFV to the total number of animals in the target population. To assess associations, demographic characteristics of animals and socio-ecological factors were the independent (explanatory) variables. Identified seropositivity and seronegativity as well as pastoralists’ categorical responses to questions in questionnaire formed the dependent (outcome) variables. All explanatory and outcome variables were initially screened by univariable analysis using Chi-square tests  or Fisher’s exact test, where appropriate. Likelihood stepwise backward multivariable logistic regressions model was built by adding variables in a backward selection process in order to start with those with significant p-value from the univariable analysis. This was used to control for confounding and test for effect modification. Variables with a p-value more than 0.05 on the univariable analysis were not included in the final model. The EpiInfo 3.4.3 (CDC, Atlanta, GA, USA) and OpenEpi version 2.3.1  statistical packages were used for statistical analyses. A p<0.05 was considered statistically significant in all analyses.