Introduction

Brucellosis is a significant zoonotic disease that is a severe health hazard to humans and can result in substantial economic losses to the livestock industry (Wang et al. 2021; Peng et al. 2020). Brucella abortus (B. abortus), the primary causative agent of bovine brucellosis, affects dairy cows' reproductive performance and productivity (Seleem et al. 2010; Godfroid et al. 2011). Infected cattle commonly display abortion, retained placenta, orchitis or epididymitis, and a reduced milk yield. However a subset of infected cattle are asymptomatic and latent, seroconverting only after calving or aborting (Bercovich 1998), making the early diagnosis of infected animals challenging. Following calving or abortion, infected cattle shed large numbers of B. abortus resulting in heavy environmental contamination, especially in intensively housed animals, with subsequent rapid transmission to susceptible animals leading to an abortion outbreak (Bercovich 1998). Although the infectious secretions gradually decrease 2–3 months after calving or abortion, some cattle can intermittently excrete B. abortus lifelong (Bercovich 1998; Godfroid et al. 2011). Therefore, once an outbreak of brucellosis occurs on a dairy farm, it is difficult to control and eliminate B. abortus from that farm.

Many countries have successfully controlled or eliminated B. abortus by implementing intervention measures, including test-and-slaughter, mass vaccination, and improved farm biosecurity (Zhang et al. 2018; Robertson 2020). However, test-and-slaughter of cattle in China is not practical or economical due to the large number of infected animals and the high cost of culling. Although farm biosecurity is generally poor on farms in China, this is one area that could be significantly improved in the future (Chen et al. 2021); however this alone cannot eliminate the disease from infected herds. The B. abortus A19 vaccine was introduced to China from the former USSR in the 1950s and has been identified to be 99.9% homologous with the internationally accepted and widely used B. abortus S19 Strain (Wang et al. 2020a, b; Cheng et al. 2021; WOAH 2019). The primary difference between these two strains is that the A19 strain is erythritol-resistant compared with the erythritol-sensitive S19 strain (Thomas et al. 1981; Wang et al. 2020a, b). Although the A19 vaccine is the bovine Brucella vaccine used in China, information on the exact origin of this strain has not been published and its protective effect and efficacy have never been thoroughly investigated. Notwithstanding this, vaccination of calves between 6 and 12 months of age with the A19 vaccine has been adopted in China to prevent infection and abortion, although the overall vaccination coverage in Chinese farms is reportedly low (Chen et al. 2021). This low vaccination coverage is probably attributable to the current veterinary policies in China, which discourage vaccination on dairy farms, and only allow vaccination if the within-herd seroprevalence is more than 1% (Ministry of Agriculture and Rural Affairs of the People's Republic of China 2016). Vaccination of adult cows is also generally not adopted in China as this may induce abortions post-vaccination for pregnant cows, especially given the erythritol-resistant nature of A19 (Hou et al. 2019). However, vaccination of adult animals with a reduced dose of the S19 vaccine was adopted in England and the United States of America (USA) in the early stage of their respective brucellosis control programs when culling infected animals was not economically feasible and test-and-slaughter alone was proved to be inadequate to eliminate the disease from infected herds (Lawson 1950; Barton and Lomme 1980). To date, vaccination with a reduced dose of A19 has not been practised in China.

In the face of an outbreak of brucellosis, rapid removal of all infected animals from the population is desirable, but again this is often not feasible or practical because of economic reasons and the lack of accurate diagnostic assays with perfect diagnostic sensitivity (Gall and Nielsen 2005; McDermott, Grace, and Zinsstag 2013). The alternative is to conduct emergency vaccination of the entire cattle population and gradually cull infected individuals. This control strategy using the S19 vaccine has been successfully adopted in high incidence areas in Portugal compared to only adopting a test-and-slaughter measure (Caetano et al. 2016). Herrera et al. (2008) reported that implementing a S19 vaccination program also increased the milk yield of dairy cows on a farm in Mexico where brucellosis was endemic. Although previous studies reported the role of A19 vaccination in the control of brucellosis (Liu et al. 2019; Tang 2017), its quantitative effects on abortion rate and milk production remain unknown in Chinese dairy cows.

The current study was designed to (1) identify the associated risk factors for brucellosis seropositivity during an abortion storm on a dairy farm in Hubei Province, China; (2) evaluate the effects of emergency vaccination with A19 on subsequent abortions and milk production on this dairy farm one-year after vaccination.

Results

Outbreak investigation

Of the 169 serum samples tested for antibodies against brucellosis in December 2018, 41 were positive on the RBT of which 39 were positive to the C-ELISA, resulting in an animal-level seroprevalence of 23.1% (95% confidence intervals (CI): 17.0, 30.2) when tests were interpreted in series. The monthly numbers of normal calving and aborting events from January 2017 to December 2020 are displayed in Fig. 1. Cows that aborted had a median pregnancy duration of 220 days (Inter Quartile Range (IQR): 145.5, 241.5), compared with 275 days (IQR: 271.0, 277.0) for those that calved normally in 2018. The majority of abortions in 2018 occurred during the last trimester of gestation (63.3%, 19/30) prior to implementing vaccination. The annual incidence of abortions in 2018 (pre-vaccination) and 2019 (post-vaccination) were 22.1% (95% CI: 15.4, 30.0, p < 0.001) and 34.0% (95% CI: 24.6, 44.5, p < 0.001), respectively, which were both significantly higher than that in 2017 (5.7%, 95% CI: 2.3, 11.4), with corresponding RRs of 3.9 (95% CI: 1.8, 8.5) and 6.0 (95% CI: 2.8, 13.0), respectively compared to 2017. Thirty-two females aborted within one-year after vaccination, and the abortion rate in the low-dose immunised group (32.9%, 25/76) was not significantly different from that in the standard-dose group (41.2%, 7/17) (p = 0.52).

Of 84 females in 2018 prior to emergency vaccination, 14 had abortions and 24 were seropositive to Brucella. Notably, most aborting animals (78.6%, 11/14) were seropositive against Brucella. Seropositive animals were 9.2 times (95% CI: 2.8, 30.0, p < 0.001) more likely to have aborted than seronegative animals. The attributable fraction (AF) for abortion arising from being seropositive to Brucella was estimated to be 89.1% (95% CI: 64.3, 96.7). Three factors, including specific sheds, animals purchased from outside the herd, and the presence of abortion, had a p < 0.20 in the univariable analyses and were included in the saturated multivariable logistic regression model (Table 1). Only the presence of abortion and specific sheds were significantly associated with seropositivity at the animal level in the final model. Cows that aborted had a greater odds of being seropositive (Odds ratios (OR) = 21.4, 95% CI: 4.4, 168.4) than those that calved normally. Cows in sheds A2 and C1 were significantly more likely to be seropositive compared to those from shed B1 (OR = 10.2, 95% CI: 1.4, 128.0, OR = 17.0, 95% CI: 2.8, 190.3, respectively, Table 2). The Akaike information criterion (AIC) value of the final model was 78.76 with a Hosmer–Lemeshow goodness-of-fit value (χ2 = 1.36, df = 8, p = 0.995), indicating that the model was a good fit of the data. The area under the receiver operating characteristic (ROC) curve was 0.827 (95% CI: 0.727, 0.927), suggesting that the model had a good predictive ability.

Table 1 Univariable analysis of animal level risk factors associated with brucellosis seropositivity
Full size table
Table 2 Multivariable logistic regression analysis of risk factors associated with brucellosis seropositivity
Full size table

Emergency vaccination

In Table 3 the extra serological results from one week pre-vaccination to 24 months post-vaccination (standard-dose vaccine group) are summarised for 33 heifers that received the full dose of the A19 vaccine. Two of these 33 heifers (6.1%) tested positive one week before vaccination. Forty-two percent tested positive at 8 months post-vaccination, and then only one heifer (3.0%) tested positive at 12 months post-vaccination (Table 3). The percentage of test positive increased to 24.2% (8/33) at 20 months post-vaccination and then decreased to 15.2% (5/33) at 24 months post-vaccination. Interestingly, six of the eight animals that tested positive at 20 months post-vaccination were seronegative at 24 months post-vaccination. Only one animal was test-positive at all sampling time points. One of the 33 cattle aborted in the first trimester but remained seronegative against Brucella throughout the observation period.

Table 3 Serological follow-up results post-vaccination for 33 heifers vaccinated with a standard dose of A19
Full size table

After implementing emergency vaccination against brucellosis in the herd in January 2019, the incidence of abortion remained high and declined from eight months post-vaccination before returning to levels comparable to those in 2017 (Fig. 1). Only 10 of 101 pregnant animals aborted in 2020 (annual incidence of abortion = 9.9%; 95% CI: 4.9, 17.5). This annual incidence in 2020 was significantly lower than that reported in 2018 (RR(relative risk)  = 0.45, 95% CI: 0.23, 0.88, p = 0.014) and 2019 (RR = 0.29, 95% CI: 0.15, 0.56, p < 0.001) but was comparable to that of 2017 (RR = 1.74, 95% CI: 0.69, 4.41, p = 0.237) prior to the introduction of replacement heifers from northern China. The overall herd protection effectiveness against abortion was 56.8% (95% CI: 15.8, 77.8) (Table 4). When stratified by parity, the estimates increased to 86.7% (95% CI: 4.4, 98.1) and 63.3% (95% CI: 0.0, 87.3) for 1st parity and 2nd parity females, respectively (Table 4). However, vaccination did not offer sufficient protection against abortion for cows that were ≥ 3rd parity, as the incidence of abortions for this latter group did not significantly change between 2017 and 2020 (p = 0.686, Table 4).

Table 4 The influence of parity on estimates of vaccine (Brucella abortus A19) effectiveness for preventing abortions
Full size table

During 2018 (pre-vaccination), although 150 cows were being milked, only half of them (n = 76) were still on the farm and available for sampling in December 2018. The mean 305-day milk yield of the seropositive group was slightly, but not significantly, lower than that of the seronegative group in 2018 (5215.3 vs 5578.0 kg, respectively—Table 5 and Fig. 2 Panel A) (Kruskal–Wallis χ2 = 1.152, df = 1, p = 0.283). After implementing emergency vaccination in January 2019, the 305-day milk yield in 2019 for the whole milking herd was similar to that in 2017 and 2018 (Table 5 and Fig. 2 Panel B). However the milk yield in 2020 was significantly higher than that in 2018 and 2019 (p < 0.05), and marginally higher than that in 2017 (p = 0.063) (Table 5). No significant difference in milk yield was identified between other years.

Table 5 Effects of brucellosis seropositivity and emergency vaccination with A19 vaccine on 305-day milk yield during 2017–2020
Full size table
Fig. 1
figure 1

An epidemiological curve for the monthly number of normal calvings and abortions from 2017 to 2020

Full size image

Discussion

Although the A19 strain has been registered for use in dairy cattle in China, few studies quantify its effects on reducing abortions and increasing milk production (Liu et al. 2016). Since this farm had not been vaccinated against Brucella before, these animals (39/169) can be diagnosed with Brucella infection through combined serological tests recommended by WOAH (2019). Brucella infection accounted for 90% of abortions, and abortions from other causes were minimal. Finally, a few animal-level risk factors were identified, and more herd-level management practices and biosecurity measures should be investigated in future studies. These factors, such as the live cattle movement, are commonly considered to be one of the critical drivers of infectious disease transmission between farms, and risk-based control measures can only be developed if these are thoroughly understood.

To control or eradicate the disease in this dairy herd, vaccination alone is unlikely to be sufficient, and a combination of identification and removal of infected animals based on culture is needed, along with regular environmental disinfection. However this study highlights the value of using emergency vaccination with A19 in containing an abortion storm.

Conclusions

In conclusion, this study first demonstrated that the most likely cause of the abortion storm was Brucella infection with an attributable fraction of 89.1%. Administering emergency vaccination with A19 on this farm yielded good outcomes with: protection effectiveness of 56.8% and 86.7% for the entire population and 1st parity cows, respectively; reduced incidence of abortions; and increased milk yield one-year after vaccination. A rigorous clinical trial should be conducted to confirm the safety and efficacy of A19 to promote its broader use. In conjunction with vaccination, improved management practices and biosecurity levels should be coordinated and included in all future disease control programs.

Materials and methods

Dairy farm background, intervention and sampling design

A dairy farm located in the Yichang administrative area of Hubei Province participated in a cross-sectional study on brucellosis between January 2018 and April 2018 (Wang et al. 2021). The owner reported that the introduced cattle had not been tested for any diseases before introduction and incorporation into the replacement heifer population in shed A2. Abortions were first observed in March 2018. These abortions occurred in the middle and last trimester of gestation (Fig. 1).

At the end of December 2018, 169 serum samples were collected from all animals and sent to the State Key Laboratory of Agricultural Microbiology at Huazhong Agricultural University. These animals had not been vaccinated against brucellosis. Given that our previous study reported a true prevalence increased from 9.3% to 16.4% with a lactoconversion rate of 10.3% per three months between January and April 2018 using a commercial milk I-ELISA kit (Wang et al. 2020c), infection with Brucella spp. was initially suspected and investigated. Despite the high lactoconversion rate of brucellosis in milk samples, it cannot be excluded that the dairy farm had been infected prior to this study (Wang et al. 2020c). Following the National Standard Diagnostic Techniques for Animal Brucellosis (GB/T 18646–2018) and World Organisation for Animal Health (WOAH) manual of diagnostic tests and vaccines for terrestrial animals (WOAH 2019), combined serological tests were performed to detect antibodies against Brucella spp. This diagnostic strategy was chosen due to its extensive use in epidemiological studies of brucellosis and only serum samples available (Zhang et al. 2018). Unfortunately no tissues were cultured for Brucella spp or tested for DNA as only serum samples were available at that time. The serological test results (39/169), the typical clinical signs of abortions described by the farmer and the high lactoconversion rate in the early of 2018 from a previous study (Wang et al. 2020c), indicate that this abortion outbreak was highly likely caused by infection with Brucella spp. The dairy farm lacked facilities to separate seropositive and seronegative animals and could not afford to cull the large number of seropositive animals (39/169) in a short time. Emergency vaccination of all animals (seropositive and seronegative animals) with A19 was undertaken (outlined below), in conjunction with environmental cleaning and disinfection with 0.02% sodium hypochlorite solution. Subsequently seropositive animals that had aborted were removed from the herd over the animal’s next production or lactation year.

Emergency vaccination of animals

In the first week of January 2019 all cattle on the farm were vaccinated with the A19 vaccine. Due to the ongoing abortions on this dairy farm (Fig. 1), having a non-vaccinated control group was not practical or ethical. Calves and non-pregnant replacement heifers (up to 15 months of age) were inoculated with 6 × 1010—12 × 1010 colony forming units—CFU (a standard dose of A19), and the adults (≥ 2 years of age) or pregnant replacement heifers with a reduced dose of A19 (6 × 108—12 × 108 CFU, 1/100 dilution). The rationale of this reduced dose is to minimise vaccine side effects, referring to studies reporting that a reduced dose of immunisation of adult and pregnant heifers does not induce abortion (Chand et al. 2015; Geong and Robertson 2000). A19 and S19 are highly homologous strains, but no studies have reported the safety and efficacy of a reduced dose of A19 in pregnant cows. Therefore, this study used a 100-fold reduced dose of A19 to immunize pregnant adults and heifers, as was done in the S19 studies (Chand et al. 2015; Geong and Robertson 2000). The vaccine was inoculated subcutaneously behind the shoulder. The vaccine (Reference number: CVCC 70202) was purchased from Spirit **yu Biological Pharmaceutical Co., LTD, China, and stored at -20℃ until use. Sterile vaccine diluent (phosphate-buffered-saline) was used to prepare the standard and reduced dose vaccine according to the manufacturer’s instructions. The identification of each vaccinated animal was recorded. In addition, calves born subsequently were also vaccinated with a standard dose of the A19 vaccine when they were six months old. No additional booster vaccination was performed as the A19 vaccine has been reported to induce a lifelong antibody response (Anniwaer et al. 2020; Qiao et al. 2019).

Blood sampling and serological tests

Blood samples (5 mL) were collected from the coccygeal vein in the last week of December 2018 (samples collected for diagnosing the cause of abortions). An extra sampling of 33 available heifers vaccinated with a standard dose of A19 was undertaken at one week pre-vaccination and 8, 12, 20 and 24 months post-vaccination. As the cows in this case study had been exposed to a potentially heavily Brucella-contaminated environment, they were excluded from the serological follow-up sampling. Each serum sample was tested for antibodies with a commercial Rose Bengal Test (RBT) (IDEXX, USA) and a C-ELISA (Ingezim Brucella Compac 2.0, Spain) using the test protocols recommended by the manufacturers. The test results were interpreted in series where a sample was defined as seropositive only if both the RBT and C-ELISA were test-positive. The reported sensitivity and specificity of the RBT are 81.2% and 86.3%, respectively, and for the C-ELISA, 98.0% and 99.9%, respectively (Gall and Nielsen 2005). The sensitivity and specificity of the combined tests interpreted in series were 79.6% and 99.9%, respectively.

Ethical statement

The animal study was reviewed and approved by the institutional ethics committee on animal experimentation of the Laboratory Animal Centre of Huazhong Agricultural University (Protocol No: HZAUCA-2019–006). All animal manipulation in cattle and experiments were strictly performed following the Guidance for the Use and Care of Laboratory Animals, Hubei Province, China. Data were obtained and published with the informed consent of the farm owner. It was considered unethical to include a non-vaccinated group in this study because of the severe impact of brucellosis on cows.

Data collection and analysis

Animal demographic and management data (age, date of calving or abortion, duration of pregnancy, parity, the shed where each animal was located), and animal purchase histories were obtained from the farm owner for all cattle between 2017 and 2020. Data on a history of abortion and calving within each calendar year were collected for the pregnant females. The results of milk testing for brucellosis in January and April 2018, using an individual milk-based indirect ELISA, were sourced from a previous study (Wang et al. 2020c; Robertson et al. 2020). Data on 305-day milk yield for each available lactating animal from 2017 and 2020 were obtained from the Hubei Dairy Herd Improvement (DHI) Centre.

All data were entered into a spreadsheet in Microsoft Excel (Microsoft Excel 2019, Redmond, USA) and then imported into the statistical software R (V. 4.1.2) for further data manipulation and analysis (R Core Team 2021).

An epidemiological curve displaying monthly aborting and calving numbers, animal purchasing histories and emergency vaccination intervention during 2017–2020 was generated to visualise the timeline of the abortion outbreak and management practices in the dairy herd (Fig. 2). The AF of brucellosis seropositivity for abortion was calculated based on the serological test results for December 2018 and the presence of abortion in 2018. The AF was calculated using the following formula:

$$AF= \frac{{Incidence}_{seropositive}- {Incidence}_{seronegative}}{{Incidence}_{seropositive}} \times 100\%$$

where incidence represents the frequency of abortion in each serological group in 2018. The AF is a relative measure of the importance of a risk factor (i.e. Brucella seropositivity in this study), and it expresses the proportion of total risk in exposed animals which is due to the risk factor (Dohoo et al. 2009).

Totally 131 animals had at least one abortion or normal calving event in 2018, of which five animals had two records, resulting in a total of 136 records. However, of the 131 cattle, only 84 (64.1%) were available for sampling in December 2018, with the remainder culled mainly due to mastitis (51.7%), low milk yield (20.0%) or abortion (13.3%). Only data from these 84 animals were included in the risk factor analysis and the estimation of AF. To identify the risk factors associated with the presence of abortion at the animal level, univariable and multivariable logistic regression models were developed. Variables with a p ≤ 0.20 on the univariable analyses were initially offered to a multivariable logistic regression model. The model was generated using a backward stepwise process, and the most appropriate model was identified through assessing the likelihood ratio test and the minimum AIC value (Hosmer and Sturdivant 2013). The model was further assessed with the Hosmer–Lemeshow goodness-of-fit test. A ROC curve was created to display the predictive accuracy of the model using the R package 'pROC' (Robin et al. 2011). OR and their 95% confidence intervals (95% CI) were also calculated to estimate the degree of association between different variables and abortion.

Abortion and calving events in 2019 were not included in the analysis to evaluate the effect of emergency vaccination on reducing the incidence of abortion in subsequent pregnancy because of concerns over the impact of potentially latent infections on the results. RR was computed to determine the degree of association between pre- and post-vaccination on the incidence of abortion. The protection effectiveness of the vaccine was calculated as described by van Straten et al. (2016) using the formula (1– RR) × 100% based on the incidence of abortion. The AF and RR and their respective 95% CIs were calculated using 'epiR' package (Stevenson et al. 2022). The continuous and categorical variables were presented as median (IQR) and number (percent, %), respectively. The Pearson’s Chi-square test and Fisher’s exact test were used to compare the difference in the incidence of abortion between test positive and test negative groups, where appropriate. The Shapiro–Wilk test was used to evaluate the normality of the continuous data, indicating that the animal-level 305-day milk yield did not follow a normal distribution (p < 0.05) and therefore, the Wilcoxon Rank Sum test and Kruskal–Wallis Rank Sum test were used to compare the statistical difference between groups for these data, depending on the number of groups. If significant, we used the Dunn’s test with Holm-Bonferroni correction to determine the difference in the milk yield between individual years. A raincloud plot was created to summarise the distribution, median value and variability of the 305-day milk yield in each group using the R package 'ggplot2' (Wickham 2016).