Vaccination against Staphylococcus aureus mastitis in two Swedish dairy herds

Herds

Two dairy herds (A and B) with S. aureus mastitis problems for at least 5 years according to herd veterinarians and owners
were enrolled in the study. Traditional control measures consisting of identification
and segregation of infected cows, culling of chronically infected cows, selective
dry cow therapy, control of milking equipment and improvement of hygienic measures
had been performed but were considered unsuccessful.

Both herds were situated in the southern one-third of Sweden, had warm free-stall
housing systems and were enrolled in the Swedish Official Milk Recording Scheme (SOMRS,
Växa Sverige, Stockholm, Sweden). Herd A milked their cows in a milking parlor, while
herd B used a milking rotary. Information on numbers of cows, milk production, bulk
milk SCC, and proportion of cows veterinary-treated for clinical mastitis (VTCM) the
year before the start of the study are given in Table 1. In herd A, the distribution of breeds among all cows in the herd was 21 % Swedish
Red (SR), 72 % Swedish Holstein (SH) and 7 % other breeds, while the corresponding
distribution in herd B was 63 % SR, 35 % SH and 2 % other breeds.

Table 1. Descriptive statistics of herd data for the two herds included in a study on vaccination
with Startvac
^®
for the 12-months period preceding the start of the trial

Study design including farm data registration and milk sampling

Vaccinations were performed during a 12-months period starting in October 2010. Cows
with uneven eartag numbers were vaccinated three times according to the manufacturer’s
instructions, i.e. 45 days before expected calving, 35 days after the first vaccination,
and 62 days after the second vaccination. Cows with even eartag numbers were not vaccinated
and constituted the control group. For practical reasons a deviation from the vaccination
protocol was done in herd A, the first vaccination was performed during the period
45–60 days before expected calving. On both herds vaccination was performed on one
or two specific weekdays for practical reasons.

All heifers and cows present in the herds 6 months before the start of the trial were
included in the study. The farmers registered all animals in an Excel document at
the latest 45 days before expected calving. Information on group (vaccinated or control),
expected calving day, day of actual calving and milk SCC at the first 4 monthly milk
recordings after calving was registered for each cow. When occurring, day of VTCM,
and day and cause of culling, was also registered. The file was sent every 3 months
to the first author via e-mail. From all cows included in the study aseptic quarter
milk samples were taken from affected udder quarters when CM or SCM were suspected
during the first 4 months after calving. The samples were sent the same day via postal
mail to the National Veterinary Institute (SVA), Uppsala. After culturing and identification
of bacterial growth according to accredited routines at the laboratory, the results
were registered in the SOMRS database.

Collection of additional data

For cows included in the study, individual cow data (such as breed, lactation number,
calving dates, genetic merit for milk production, monthly milk recording data on daily
milk yield and SCC, culling, results from culturing of milk samples) from 6 months
before the start of vaccination until 1 year after the end of the vaccination period
(November 2012) was collected from the SOMRS.

Data editing and statistical analyses

Data editing and statistical analyses were performed using Stata (Stata Statistical
Software: Release 9.2; College Station, TX, USA: StataCorp LP).

The effects of vaccination on milk production, udder health [cow composite milk SCC,
SCM, and mastitis cases with specific growth of S. aureus, CNS, streptococci or coliforms (Escherichia coli, Klebsiella spp)] and survival (slaughter, death) were evaluated. For milk production (kg milk/day)
and SCC, data from the first 4 monthly milk recordings after calving were used. The
occurrence of SCM was defined as SCC ? 200,000 cells/ml at any of the first 4 monthly
milk recordings. Only cows with SCC data for all four milk recordings were used. Growth
of the above-mentioned specific udder pathogens in tested milk samples originating
in udder quarters with CM or SCM was registered when at least one milk sample with
growth of one udder pathogen was found during the follow-up period.

Differences in SCC and milk production between vaccinated and unvaccinated cows at
the first four milk recordings were analyzed using mixed-effects linear regression
models, including cow identity as random effect. SCC and milk production was transformed
using the natural logarithm. The effect of vaccination on the prevalence of SCM and
mastitis due to S. aureus, CNS, streptococci or coliforms was analyzed using logistic regression models. Two
models for SCM were run; (1) including all cows, and (2) only including primiparous
cows and multiparous cows without SCM before dry-off (new SCM). The effect of vaccination
on survival was analyzed using survival analysis (Cox proportional hazard models).
The number of days from calving to exit from the herd (culling/death/euthanasia due
to udder disease) was calculated. A cow was right censored if she had not calved again,
was culled due to other reasons than udder disease or was still in the herd at the
end of the follow-up period.

In all models vaccination (yes/no), breed (SR, SH, other), parity (1, 2, ?3), SCM
status at dry-off (1st parity and healthy, 2nd parity and healthy, 2nd parity and
SCM, ?3rd parity and healthy, ?3rd parity and SCM), herd (A, B), and proportion of
cows vaccinated (10, 10–19, 20–29, 30–39, ?40 %) were included as explanatory variables.
In the analyses of SCC and milk production days in milk (DIM) were included as a categorical
fixed effect (categorized into quartiles).

The fit of the linear regression models was evaluated by visual inspection of plots
of standardized residuals vs predicted values, and Q–Q plot of standardized residuals.
The fit of the logistic regression models and Cox proportional hazard models was evaluated
by goodness-of-fit tests.