This study was carried out in Nakuru County, Kenya. It is one of the main dairy farming zones in Kenya and is also the main catchment area for dairy cattle breeding stock in Kenya and the East African region. The dairy cattle population in this area ranges from 100,000-120,000, most of which are large herds on large-scale farms. Though Nakuru is a large-scale farming district in terms of land area, with many large-scale farms averaging 1100 acres, many small-scale farms with average sizes of between 0.3–10 acres do exist [31,32,33].
Study farms and animals and data collection for the survey
This phase of the study was carried out between January and May 2010. A list of more than 300 dairy cattle farms was collected from the local animal production office and dairy societies (the sampling frame), and divided into small- and large-scale farms, with large-scale farms having > 30 dairy cattle and small-scale farms having ≤29 cattle. A stratified random sampling procedure was used: 50 farms having < 29 dairy cattle and 20 farms having > 30 dairy cattle were randomly selected into the study. Herd sample size was determined based on the number of farms that could logistically and financially be handled for the study. Of the 70 farms, 64 agreed to participate.
For the prevalence survey, blood samples were collected from 398 randomly selected dairy cattle (pregnant and non-pregnant; over 6 months old) of the estimated 200,000 dairy cattle on the participating farms, based on the following formula
$$ n=\frac{4\times \mathrm{P}\left(1-\mathrm{P}\right)}{L^2} $$
where: n = sample size, L = Precision (0.15) and P = incidence risk estimate (25%), using 95% confidence levels [35].
These selected cattle were restrained in a crush, and the ventral aspect of the tail disinfected with ethyl alcohol-soaked cotton swabs. Vacutainer needles (BD Vacutainer® blood collection needle) and serum tubes (BD Vacutainer® blood collection tubes) with clot activator were used to collect the blood samples. The blood samples were centrifuged at 2500 rotations per minute for 15 min, and serum was separated and frozen at -20 °C until all samples were collected, at which time, laboratory tests were conducted.
Data collection on the participating farms and cattle was conducted through a face-to-face interview with the farm owner or manager, obtaining information on cattle age, breed, parity, and breeding at the cow level, and at grazing and vaccination practices at the herd level.
Study farms and animals and data collection for the prospective study
This phase of the study was carried out between January 2010 and May 2011. On monthly visits to the same 64 farms, 279 dairy cattle that were confirmed pregnant by rectal palpation (40–60 days post service) were selected for the prospective study. However, 19 dairy cattle were lost to follow-up due to sales of the animals from the farms, leaving 260 pregnant cows for the analyses.
Rectal palpation was performed monthly to test for continued pregnancy in cows, and blood samples were collected monthly until the cow calved or the pregnancy was lost (abortion/EED/mummification). The time of pregnancy loss was when the foetal loss was detected, by the veterinarian through rectal palpations and/or by farmers finding a foetus or vulvar foetal membranes, and the date of an abortion was estimated retrospectively.
Again, blood samples were centrifuged, and serum was separated and frozen at -20 °C until all samples were collected, at which time, laboratory tests were conducted. Data collection on the participating cows was again conducted through a face-to-face interview with the farm owner or manager.
Laboratory analyses
For the survey, commercial Enzyme-Linked Immunosorbent Assays (ELISA) were used to screen for antibodies against bovine viral diarrhoea virus, Neospora caninum and Brucella abortus antibodies (IDEXX Laboratories, Switzerland AG). For the prospective study, the same blood testing was done monthly to monitor changes in antibody titres (to BVDV, Brucella abortus and Neospora caninum). This method of analysis was selected due to its high sensitivity and specifity as well as its ability to test large numbers of samples at one time. The intra- and inter-test variability was minimized by using the same laboratory and technologist to perform all the testing.
Statistical analysis
Data from the serological survey and prospective study were entered and stored in Microsoft Office Excel 2007 (Microsoft Corporation, 2007). The data were imported into Genstat® 13th edition, service pack two, for analysis (VSN international).
Descriptive statistics, including prevalence and incidence risk, were computed for the serological survey and abortion parameters. To calculate cause-specific incidence risks of abortion, monthly antibody levels to BVDV, BA and NC were examined from samples before and after the reported abortions, with a four-fold increase in a titre for a specific pathogen indicating the likely aetiology of the abortion by that pathogen [36].
While the purpose of the study was not to identify risk factors to foetal loss, we did conduct Pearson’s Chi-square tests to determine significant differences in dichotomous predictor variables between outcome groups (e.g. those that aborted versus those that didn’t abort). Multivariable logistic regression was carried out to model the incidence risk of abortion in dairy cattle in Nakuru County, and odds ratios, as a measure of strength of association between the significant model variables (P < 0.05) and the outcome, foetal loss, were calculated. The backward elimination procedure was used for regression, and factors that were significant (P < 0.05) were retained in the final model. Potential clustering of animals within farms was controlled for by including farm as a random effect in the modelling.
