We hypothesized that being seropositive in pregnancy would be associated with adverse pregnancy outcome, potentially mediated by reactivation of a latent infection [9–11]. We also hypothesized that acute infection during pregnancy would be related to adverse pregnancy outcome. Neither of these hypotheses were confirmed as no increased risk of adverse pregnancy outcome was found in women with verified exposure to C. burnetii.
To our knowledge, this is the first population-based seroepidemiologic study evaluating pregnancy outcome in women with serologically verified exposure to C. burnetii against a comparable reference group of seronegative women.
When diagnosing Q fever, a variety of serological methods are available; the Panbio ELISA kit has previously been showed to be superior to other methods  and suitable for large-scale screening [17, 19]. The micro immunofluorescence antibody test (IFA) is regarded as the gold standard  because it is capable of determining both phase I and II antibodies simultaneously by the use of two different antigens in a single sample. We have previously demonstrated coherence between ELISA and IFA .
Villumsen et al. established a national, very restrictive cutoff in order to obtain a high specificity and a high predictive value of a positive result ; this decision was based on the assumption that Q fever was sporadic in Denmark. However, particularly in rural populations of Denmark, Q fever is more widespread than previously considered [7, 8] and one may now argue that the cutoff may be too conservative.
Consequently, in the present study, we decided to use a modified version of the Danish cutoff. A more conservative interpretation of the serological values (theoretically leading to a lower positive prevalence and higher predictive value) did not reveal any associations between seropositivity and adverse outcome of pregnancy.
Finally, we also acknowledge that the cutoff applied in our study is high compared with some other studies. However, in a seroepidemiologic study including healthy individuals, our priority was to maintain a high predictive value for a positive result. The application of a lower cutoff would have falsely classified additional women as seropositive and lead to misclassification and thus a higher risk of overlooking a potential association between (true) seropositivity and adverse outcome of pregnancy.
Most of the seropositive women had markers of previous infections, but ten met the criteria for IFA seroconversion. It is worth to note that two out of these women gave birth to infants that were SGA. We cannot draw any conclusions on the risk of adverse pregnancy outcome from 10 cases and the low number of seroconverters is a limitation to this study. Hence, we cannot make an inference with respect to pregnancy outcome in women with acute and, in particular, symptomatic infections.
The risk of reactivation of latent infection leading to adverse pregnancy outcome has been reported [9, 10]. However, the IgG positive women in our study had a similar proportion of previous spontaneous abortions as the seronegative women, and overall, reactivation of latent infections leading to adverse pregnancy outcomes was not observed in this population.
Detailed information on previous preterm births was not available, and we chose adjustment for prior pregnancies regardless of pregnancy outcome.
In women with contact to livestock, we had the opportunity to evaluate seroconversion throughout pregnancy; in women with no contact to livestock we only had blood samples from beginning of the pregnancy. This could potentially bias data as the women without animal contact were assumed to be negative throughout pregnancy when, theoretically, they could be infected later in their pregnancy. This is why women with seroconversion as well as women who were seronegative in the midpregnancy or in the umbilical cord sample were classified as seronegative in analyses of pregnancy outcome. Also, stratified analysis on contact to livestock and pregnancy outcome (spontaneous abortion and preterm birth), irrespective of titer status, showed no significant difference between the groups (results not shown).
A high seroprevalence of C.burnetii accompanied by few clinical symptoms in farmers and veterinarians has been found in Denmark as well as abroad [7, 8, 22]. We evaluated pregnancy outcome in seropositive versus seronegative women who had occupational, domestic, or no exposure to livestock (as stated in the methods section). The vast majority of the seropositive women were exposed to animals (Table 1). Due to few unexposed, seropositive women we are unable to study adverse pregnancy outcome in this group of women or clarify whether the dynamics of infection differ in unexposed women compared to women heavily exposed to C.burnetii.
The evidence of the impact of Q fever on pregnancy outcome mainly originates from French case studies of referred infected pregnant patients and pregnancies with Q fever diagnosed retrospectively after an adverse pregnancy outcome [7, 8]. The authors conclude that there is a link between placentitis and obstetric complications. However, in a recent study by Angelakis et al.,  a study of 30 pregnant women with acute infection in pregnancy, no placentitis or isolation of C.burnetii is found in 14 available biopsies. 17 of the women were asymptomatic, but only two of these had an uncomplicated pregnancy illustrating the difficulty in segregating harmless seroconversion from infection threatening maternal and foetal health. In that study, genotyping showed that QpDV plasmid was present in 4 of 7 C. burnetii strains isolated from infected women with miscarriage. Apart from differences in study design, numbers of pregnancies included, selection bias and cutoffs, the disagreements between the French, the Dutch and our studies could be related to strain specificity. Risk assessment and management of Q fever in pregnancy may therefore benefit from further clarification of the role of strain differences and virulence factors.
The present study is subject to some limitations.
Due to the design of the study, it was not possible to include early miscarriage as an outcome. Only few participants were included prior to 8 weeks of gestational age (Table 1). It is possible that the study population is biased towards a “healthy pregnant population”. An increased risk in early pregnancy may in our study be reflected by a “protective” effect in later pregnancy.
Also, maternal IgM cannot be detected in umbilical cord blood, meaning that theoretically we could miss a narrow window of acute infections in very late pregnancy with positive IgM but before IgG phase II elevation; the potential effect on pregnancy outcome from this is, however, speculative.
The French recommendation regarding treatment with cotrimoxazole throughout pregnancy in seropositive women [9, 10, 23] is widely practiced, but has recently been questioned . However, the number of acute infections in our study is too small to impact these recommendations.
Overall, our findings are in line with two new studies from The Netherlands, a country that recently saw the world’s largest Q fever outbreak . One study included serum samples from early pregnancy of 1174 pregnant women living in the high-risk area and found no association between positive Q fever serology and adverse pregnancy outcome . The other study was a randomized controlled trial with 1229 women split into a screening group and a control group; no difference in pregnancy outcome was found between the two groups .