The overall animal and farm seroprevalence of C. burnetii in dairy goats farms with ≥ 100 dairy goats observed in this study was 21.4% and 43.1% respectively. These seroprevalence estimates increased compared to the seroprevalence measured in 2008, when 14.7% of individual dairy goats were serologically positive and 17.9% of farms tested positive. The within-herd prevalence on positive dairy goat farms in our study was 46.6% compared to 32.1% (95%CI 28.4%-35.9%) in 2008 (van den Brom R, Moll L, Vellema P: Q fever seroprevalence in sheep and goats in the Netherlands in 2008, submitted). This study demonstrates substantial transmission of C. burnetii within and between dairy goat farms in recent years prior to the mandatory vaccination campaign in the Netherlands.
The relatively low overall participation rate of 37% probably reflects the reluctance to take part in the study at the same time as control measures increased, including finally the culling of pregnant goats at bulk milk PCR-positive farms. The overrepresentation of farms located in the mandatory vaccination area probably reflects that the risk perception of the farmers played a role. Because of the higher participation rate in this area, we might have overestimated the overall seroprevalence in eligible dairy goat farms in the Netherlands. As expected, dairy goat farms located in the mandatory vaccination area were more often seropositive in our study, as was previously observed in 2008 (van den Brom R, Moll L, Vellema P: Q fever seroprevalence in sheep and goats in the Netherlands in 2008, submitted). In contrast, the estimated seroprevalence might have been underestimated as the non-eligible farms probably overrepresented positive farms, such as farms with a clinical history of Q fever. These were prioritized for vaccination early 2009, and probably positive and suspected farms relatively more often volunteered for vaccination in the 2008 campaign. Nevertheless, the net effect of these biases are thought to be limited, as bulk milk-positive farms were equally represented among participating and non-participating farms. As the diversity in farms, also outside the vaccination area, was still large, effect on the risk factor analyses is considered limited and results are considered generalizable to all commercial dairy goat farms in the Netherlands.
Small ruminant studies have shown that goats test significantly more often serologically positive during pregnancy and in the periparturient period compared to early pregnancy or non-pregnant period (van den Brom R, Moll L, Vellema P: Q fever seroprevalance in sheep and goats in the Netherlands in 2008, submitted), . Different sampling periods in our study, mainly at the end of the lambing season in 2009 inside the vaccination area and in the beginning of the lambing season 2010 outside the vaccination area, make it difficult to disentangle the possible effect of seasonal sampling on the observed significant regional differences. We think this study shows a true higher seroprevalence in the mandatory vaccination area as it (1) confirms the significant difference already observed in 2008 in the south-eastern part of the country and (2) reflects the major burden of human and veterinary clinical Q fever cases that occurred in the south-eastern part of the country [1, 2, 10]. A distinction with ELISA between IgG phase 1 and 2 antibodies might have helped to distinguish more recent infection and older infections in animals to assess such a sampling effect .
Seroprevalence of C. burnetii in goats has been studied in several countries . Comparison should be done with caution as study populations and study years vary and different serological assays with different performance are used [23, 24]. The goat seroprevalence was 8.8% in Albania  and 6.5% in Northern Greece . In Spain, the goat and farm prevalence was 8.7% and 45%, respectively, based on ELISA  similar to results of a smaller study from Northern Ireland (goat seroprevalence 9.3%, farm prevalence 42.9%) , and in Sardinia in 1999-2002: goat prevalence 13% and farm prevalence 47% using an alternative criterion of two or more seropositive animals per farm . In our study, the farm seroprevalence was 38.2% (95%CI 29.6%-46.8%) using the same criterion, so only slightly lower than with at least one positive goat as criterion for a positive farm. Clearly different goat prevalences were observed in Poland, with the absence of C. burnetii IgG phase 2 antibodies in 918 goats from 48 herds  while high estimates were observed in Cyprus (48.2% in 420 random goats)  and in Gran Canaria island, Spain (60.4% in 733 goats) . Ignoring these last exceptions, the overall goat prevalence of 21.4% observed in our study was relatively high compared to other European seroprevalence studies (6.5%-13%), while the farm prevalence falls within the range of farm prevalences (43-47%) in other European countries. The within-herd prevalence of 46.6% among the positive farms indicates strong circulation of the bacterium within the herds, suggesting farm conditions or practices favoring spread, such as a relatively large number of goats per farm, year-round housing in deep litter stables or reflects circulation of a unique efficiently spreading strain. In France, goat herds with a within-herd prevalence over 40% had the highest proportion of shedder goats and highest averages of shedding quantities as determined by real-time qPCR on vaginal swabs, representing a high risk level for environmental contamination and by that transmission within farms . At about one quarter of the bulk milk PCR-negative farms, on average 37% of the goats tested seropositive. This might be explained by the fact that not all seropositive goats shed the bacterium in milk and that excretion of the bacterium is intermittently [7, 15]. Besides, antibodies persist in goats , finding still positive serology but no actual excretion of DNA which is measured in the bulk milk monitoring program.
Considering the risk factors analyses, the exposure information collected in the farm questionnaire is not necessarily related to the relevant time period for seroconversion as we do not know at what moment the actual infection with C. burnetii occurred in serologically positive goats. However, since the median age of tested goats was 2.3 years, and infections especially occur during the first pregnancy of nulliparous goats (between 1-2 years of age) it is plausible that infection in the majority of goats occurred during the periods covered in the questionnaire. It is most likely that goats on these farms get infected the same way as humans, i.e. by inhalation of C. burnetii infected aerosols , as indicated by the increased risk of a farm location within 8 km of a bulk milk-positive small ruminant farm. From literature we know that herd size and high farm and animal densities can augment the risk for acquisition of (respiratory) zoonoses, for example in swine diseases and avian influenza in poultry [30, 31]. In our study, we found that farms with more than 800 goats had a higher risk to be positive than smaller farms. This corresponds with Rupanner et al. who observed in the 1970s an increased infection risk of goats with C. burnetii with increasing herd size . Similar associations with herd size were found for Q fever in dairy cattle [17, 33]. This can be explained by a larger population at risk, an increased risk of introduction and transmission of pathogens within and between herds for instance by larger amounts of feed, animal supply and more professionals working at or visiting the farm. In addition, farm management practices or environmental characteristics related to large farms but not covered in our questionnaire might play a role in the observed increased risk. As about 35% of the unexplained variance in the model was explained by the farm-effect, relevant underlying factors might have been missed. Therefore, an advice to limit the herd size without further changes in farm management does not necessarily guarantee a reduction in infection risk. Artificial insemination was an independent risk factor at animal level and found to be related to farms with a herd size over 800 goats. Artificial insemination can therefore be an indirect marker of farm management practices in larger farms that were not covered in the questionnaire. From cattle studies, it is known that viable C. burnetii is detected in semen of seropositive bulls indicating the possibility of sexual transmission . Between 3000 and 4000 inseminations each year are carried out by the main goat artificial insemination (AI) cooperative using fresh semen from the Netherlands and frozen semen from French or Dutch origin. Since end of 2008, AI bucks are routinely screened for presence of C. burnetii. In a targeted survey, so far, goat semen samples from 300 bucks present on bulk milk-positive farms were all negative (personal communication, P. Vellema, Animal Health Service). High cattle density in the municipality where the farm was located was also an independent risk factor, indicating the presence of one or several cattle farms in the same municipality as the goat farm. A recent review on C. burnetii infection in domestic ruminants suggested a higher seroprevalence in cattle compared to goats and sheep . In the Netherlands, a prevalence of bacterial DNA of 56.6% in cattle bulk tank milk was found as compared to 24.4% bacterial DNA in goat bulk tank milk among participating farms in our study, confirming widespread circulation of the bacterium among cattle . However, an association with cattle density was not observed when the outcome variable 'bulk milk PCR-positivity' was used instead of ELISA-seropositivity (data not shown). Therefore, it is hypothesized that cattle especially played a role in the more historical infections in goats, while spread between dairy goat herds is responsible for the more recent infections and a large part of the epidemic observed since 2007. The serological status of cattle and foremost comparison of C. burnetii isolates by subtyping in different ruminant species might help to elucidate the transmission pathways between different species of ruminants and to humans. So far, one unique genotype predominated in dairy goats herds, although at 50% of the farms at least one additional genotype was observed . Very sparse data on cattle isolates in the Netherlands suggest different subtypes from those found in goats, sheep and humans . More and nationwide representative data are urgently needed to confirm these distinct types for cattle, and to study if some cattle types match with the non-dominant genotypes regularly observed at dairy goat farms.
Previous ruminant studies have shown that farm management practices can influence the seroprevalence of C. burnetii [17, 38]. Straw, used widely as bedding in deep litter stables, could be a way in which Q fever was introduced in the Dutch dairy goat farms as import of straw from abroad or unknown origin was an independent risk factor. Farmers indicated straw was most often imported from Germany and France, which are endemic countries for Q fever. Microbiological examination from straw originating from France showed presence of C. burnetii by PCR, although the method of sampling does not exclude contamination at the farm . Contact with straw and other farm products was also a risk factor for humans in the first documented outbreak in 2007 and in international outbreak studies [40–42]. The presence of dogs and cats in the goat stable was related to a seropositive Q fever status of dairy goat farms. Furthermore, the seven farms without companion animals were all seronegative. This suggests introduction of C. burnetii or facilitation of within farm-spread by infected companion animals. In a study in Cyprus, risk factors for Q fever abortions compared to abortions of other causes were studied in a convenience sample of ruminant farms including only two goat farms; among others presence of dogs and cats were on farm risk factors . Pets, especially during kidding, have been associated with outbreaks in the past [5, 6]. In a Dutch study in the early 1990s, 13.2% of dogs and 10.4% of cats tested positive for C. burnetii by ELISA . To study the role of companion animals in current transmission, an update of this study, ideally also looking at shedding by PCR, is needed.
Covering airspaces in the stable to control nuisance animals, such as wild birds, unexpectedly was an independent risk factor. As wild birds may play a role in the transmission within and between farms and were the cause of a familial Q fever outbreak  a protective effect was expected if any. However, presence of wild birds in the stable was not a risk factor in the multivariable analysis but mainly indicated the farmer actively controlled nuisance animals such as by covering airspaces. In addition, we found at animal level an increased risk for farms that use windbreak curtain, sometimes in combination with windshields. These two risk factors could point at a more air-locked stable, facilitating accumulation of C. burnetii inside the stable, which may promote spread within the herd. This accumulation risk was indirectly shown in the study from Cyprus where a high frequency of litter cleaning was found to be a protective factor . Although a less confined farm might limit the within-herd spread, such open constructions can be a risk for aerosol spread to other farms and persons in the near environment. Presence of mice and rats in the stable was not found to be a risk factor in our study, although a recent study showed presence of C. burnetii in rats at livestock farms in the Netherlands . Whether vermin are able to maintain the transmission cycle and are able to (re)introduce Q fever at farms is currently under investigation.