In this study the association was assessed between T. gondii seroprevalence and potential risk factors for T. gondii infections in finishing pig herds in the Netherlands. Twelve out of 30 variables were identified as potential risk factors. Most of these 12 potential risk factors are already well known for T. gondii and in general related to the presence of cats, presence of other animals, the accessibility of cats, rodents and birds to the stables and feeds and rodent control . To determine the association, the seroprevalence in the selected herds was calculated on the basis of a serological surveillance system for which from every delivery of finishing pigs to the slaughterhouse one or six serum samples were taken and tested for anti T. gondii antibodies . Because of the significant association with known farm risk factors, it can be concluded that this serological surveillance system can be used to identify finishing pig farms where the typical T. gondii risk factors are present. This finding emphasizes that determination of the within-herd T. gondii seroprevalence is valuable to guide and monitor the control of T. gondii in pig herds. Recently, we performed an intervention study on five pig farms in which the within-herd T. gondii seroprevalence was successfully used to evaluate the effectiveness of the interventions on T. gondii risk factors . These results confirm that determination of within-herd T. gondii seroprevalence is a useful part of a surveillance system based on serology for detection of T. gondii infections in pigs.
As in other studies, in our study the presence of cats at the barnyard or in the pig stables was associated with a significantly increased seroprevalence of T. gondii in pigs. Pigs can get infected by uptake of soil, feed and water contaminated with oocysts shed by cats in the environment, or by ingestion of cysts in the tissues of infected intermediate hosts (e.g. rodents, birds, meat and cannibalism) .
Our results also showed that not just the presence of cats on pig farms is a significant risk factor but that this significance increased when kittens were present. Kittens pose the highest risk of spreading oocysts in the environment, because most cats are infected with T. gondii as juveniles  or even as suckling kittens . Cats only spread T. gondii in their feces for 1–3 weeks following the first episode of infection and they become immune to re-shedding of oocysts . Neutering adult cats to prevent kittens to be born was found to be a successful intervention to achieve a significant reduction in T. gondii seroprevalence in a pig herd . On farms, cats are often used to control rats and mice surrounding the pig stable which is also important in preventing a T. gondii infection and in improving biosecurity. Thus, many pig farmers might not want to remove all cats from the farm. Instead, the advice can be given not to keep kittens on the farm and that neutering of cats is a suitable approach to achieve this.
Our questionnaire included several questions about feed-related variables, because uptake by pigs of sporulated oocysts of T. gondii in animal feed represents an important route by which pigs can be infected. Open or less confined feed storage or feeding area represent an increased risk for exposure of livestock to the parasite . However, most of the feed-related variables could not be analyzed in our multivariable analysis due to collinearity with other variables or due to missing values. The only feed-related variable which we could analyze was the use of heated feed for feeding the pigs, and this was found to be significantly reduce T. gondii seroprevalence. High temperatures during the production of pig feed can inactivate the parasite. More research is needed to analyze the impact of other feed-related variables.
We found that feeding of goat whey is associated with a high seroprevalence. Although there were only four farms where whey was fed, the difference in seroprevalence with an OR of 11.30 between these four farms and the other 63 farms was considerable. This is in line with other studies that showed that feeding of pigs with raw milk goat whey is an important risk factor for infection with T. gondii [20, 31].
As in other studies, rodent control was identified as a risk factor for T. gondii infections in pigs [14, 17]. Besides that, in this study we found that rodent control using a combination of poison and traps has a higher OR than the use of poison and traps separately. It could be that simultaneous application of the two approaches for rodent control is more effective than each single approach. As in other studies, we identified shielding of birds as a preventive factor for T. gondii infections in pig herds . Birds can acquire T. gondii infection through ingestion of oocysts from the ground or through ingestion of tissue cysts present in infected prey. Like rodents, birds are incidentally caught and eaten by pigs.
In our study, presence of other farm animals (cattle, sheep and/or goats) on the farm was found to be a risk factor for T. gondii infection in finishing pigs, while in other studies it does not seem to be a risk factor [19, 32]. However, in line with our findings, a recent review  suggested that the presence of multiple animal species on a farm could serve as an indicator of low farming intensity and that this low intensity was often related to a higher risk of T. gondii seropositivity.
In our study presence of dogs on the farm was found to be a preventive factor for T. gondii infection in pigs (OR = 0.5). Hill et al. (2010) also found that the presence of dogs was significantly associated with a reduced number of T. gondii seropositive samples on surveyed farms and the explanation is that dogs could be used for rodent control . In contrast, other studies identified presence of dogs as a significant risk factor for T. gondii infection in pigs [14, 33] or did not find a significant effect .
The use of boots only in the stables was identified as potential risk factor, although the crude percentage of positive samples was lower for this category compared to the reference. A similar apparent mismatch was observed for the variable ‘type of farms’. Additional modelling (forward multivariable selection, bi- and trivariable logistic regression; data not included), showed that confounding and effect modification were unlikely to explain this observation. We hypothesize that these observations result from the effect explained by other variables in the multivariable model. The remaining effect attributed to the two mentioned risk factors is thus with an sign opposite to what one would initially expect. We hypothesize that these observations result from the combined effect explained by all other variables in the multivariable model.
A further difference was found for the water source as a risk factor for T. gondii infection in pigs. In our study, use of well water was found as a preventive factor for T. gondii infection compared to tap water. A recent review  concluded that it is hard to quantify the risk for a T. gondii infection of pigs through water, because in some studies well water was associated with an increased risk, while it seemed to have a protective statistical effect in others. A potential reason for these differences could be that in some studies cats had access to the water at any stage before it reached the pigs and contaminated it with oocysts, whereas in other studies they did not. Water can be supplied to the pigs from a variety of sources and on different ways, which may depend on different production systems and regional differences. It should be noted that changing from water source might not be a possible intervention for all pig farmers, because the production system might prescribe a certain source or regional circumstances prevent implementation of certain sources.