This review was performed in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-analyzes (PRISMA) guidelines (PRISMA Transparent Reporting of Systematic Reviews and Meta-Analysis) (11). We searched the PubMed ¹ and Scopus ² databases from January 1, 2000, to December 31, 2022 to retrieve relevant studies. Search terms included a combination of words “bovine” or “cattle” and “abortion.”

A series of inclusion and exclusion criteria were established, and the reference lists of all included articles were manually searched for potentially eligible literature. The inclusion criteria were as follows: 1) full-text articles available online in English or Spanish languages, 2) studies examining infectious causes of abortion, 3) studies examining at least one infectious agent as a possible foetal death cause, 4) beginning and end date of study implementation should be mentioned, and 5) number of abortion cases in the study is specified. On the other hand, all the studies that did not include the above-mentioned criteria, review articles, systematic analyzes, and experimental infection studies were excluded. In addition, seroprevalence studies in which only serological tests were performed or studies of an outbreak case were also excluded. Furthermore, duplicates and articles that did not include in the title the words “bovine,” “cattle” or “ruminant,” and “abortion/aborted” or “reproductive failure” were eliminated. Then, the abstracts of the remaining studies were screened. Subsequently, the full texts of the potentially eligible studies were screened against the inclusion criteria.

Full text articles were examined and assessed for eligibility by two independent researchers (YPH and SGO). Discrepancies between the authors at any stage of the selection process were resolved by consulting a third and fourth reviewer (PHI, LMOM). Relevant data were extracted from text and placed into purpose-built tables using Microsoft Excel 2019 (Microsoft Corp., Redmond, WA, United States). The required information was extracted for each study, including name of the first author, country of the study, publication year, abortion moment (1st, 2nd, or 3rd trimester); type of production system (dairy or beef); number of analyzed farms; cattle management system (intensive or extensive); history of reproductive failures; type of samples analyzed and sample size (the number of examined animals), infectious pathogens analyzed, diagnostic methods and results, final diagnosis; and any additional data that could be of interest for the study.

In the present study, the quality of articles was assessed using a modified version of the Newcastle–Ottawa Scale (NOS) (12). This quality scale ranges from 0 to 9 points, and higher scores indicate better quality studies. Only articles with acceptable quality (≥4) were included in this study.

All analyzes were carried out using the “SPSS v. 28” and “Open Meta-Analyst 10.12” programs. First, an analysis of the global prevalence of infectious agents related to bovine abortion was performed. All studies that diagnosed a wide panel of agents (viruses, bacteria, protozoa and/or fungi) were included in this analysis. Using the website http://mapinseconds.com, a map that represents the number of cases analyzed in the different countries was created. Second, a meta-analysis was carried out to determine the prevalence of the main infectious agent in bovine abortion. In all cases a random effects model, since high heterogeneity between studies was expected, and confidence levels of 95% were applied.

The degree of heterogeneity between the analyzed studies was determined using the I² statistic. Heterogeneity was considered low if the I² value was less than 25%; average heterogeneity was indicated by an I² value of approximately 50%, and high heterogeneity was indicated by an I² value above 75% (13). Forest plots were used to show the results of each study and the heterogeneity among studies. For the analysis of the prevalence of infectious agents related to bovine abortion, due to the high rate of heterogeneity obtained, sensitivity analysis was performed using the LOOCV (leave-one-out cross-validation) method. This was performed to exclude small studies with extreme effect sizes that could affect the overall results. To determine if publication bias could affect the estimated prevalence, a funnel plot and Egger’s test were used.

To determine the influence of the diagnostic method on the high heterogeneity, subgroup analyzes were performed, and significant differences between groups were analyzed by using a chi-square test. p values <0.05 indicated statistical significance and the existence of a relationship between the diagnostic method used and the prevalence obtained (13). Sensitivity analysis was also performed using the LOOCV (leave-one-out cross-validation) method. Finally, we performed a meta-regression to evaluate the effect of the publication year of studies on the prevalence estimates by adding the publication year as a variable to the simple model regression.

The search of two databases (PubMed and Scopus) yielded 6,539 articles; 3,739 articles remained after removing duplicates. Following an initial screening based on titles and abstracts, 3,206 studies were excluded because they did not meet the inclusion criteria. Next, the remaining 533 full-text articles were assessed for eligibility. Ultimately, 76 of these articles met the inclusion criteria and were included in the meta-analysis (Figure 1). Information and characteristics of the included articles are presented in Table 1 (4, 7, 14–87). The quality assessment of studies with the NOS checklist indicated that the articles included in this meta-analysis were of acceptable quality (≥4 for each study) (Supplementary Table S1).

The included articles were published from 2000 to 2022, and all articles and all studies that examined one or more infectious agents in bovine abortion cases during this period were included. Overall, the selected works came from 4 continents, although not in the same proportion (Supplementary Figure S1). There were 32 studies from Europe (Belgium:4; Cyprus:1; Denmark:1; Finland:1; Hungary:4; Italy:2; Latvia:1; Netherland:1; Portugal:1; Scotland:2; Serbia:1; Spain:2; Switzerland:4; Turkey:5; United Kingdom:2), 23 studies from North and South America (Argentina:4; Brazil:10; Canada:2; Chile:1; Mexico:2; Peru:1; Uruguay:1; United States:2), 16 studies from Asia (China:3; Iran:11; Israel:1; Republic of Korea:1), and 5 studies from Africa (Algeria:2; Rwanda:1; South Africa:1; Tanzania:1). In 30 of the 76 articles, information about the time of gestation in which reproductive failure occurred was provided. Out of 11,376 aborted foetuses, 4.9% examined cases from 1st–2nd trimester of pregnancy (n = 553), 72.4% from the 2nd–3rd trimester (n = 8,232) and 22.8% from the 1st–2nd–3rd trimester (n = 2,591), without specifying the exact number in each trimester. Moreover, 32 (n = 8,576) of the 76 articles included the information about the type of herd production system (dairy or beef herds) from which the analyzed samples proceeded. Of these studies, 24.1% of cases (n = 2,064) were collected exclusively from dairy herds and 75.9% (n = 6,512) collected from dairy and beef herds without specifying the number of aborted specimens from each farm.

In relation to the analyzed samples, most of them were composed only of foetal tissues (41 studies with 10,700 cases). Only 26 studies (6,905 cases) included foetuses and some of their placentas for the diagnosis of abortion, although the exact number of placentas was not specified. In 5 studies (426 cases), the placenta was the only sample analyzed, in 4 studies (790 cases) only abomasal fluid was studied, and in one study, both placenta and abomasal fluid samples (249 cases) were investigated. A high or moderate grade of autolysis was mentioned in 17 studies in which some foetal tissues or placenta were not available for histopathological examination. In the same way, secondary contaminations that made the diagnosis difficult were mentioned in three studies.

Although 76 articles were included in this meta-analysis, only 27 were selected to determine the prevalence of infectious agents related to bovine abortion. These articles included in their analysis the detection of a wide panel of viral, bacterial, protozoan and/or fungal agents to reach the diagnosis of abortion. The other studies (n = 49) were excluded from this analysis because their aims were to study the implication of some specific agents in bovine abortion. A total of 10,667 aborted bovine foetuses from 17 countries from America, Europe, Africa, and Asia were examined to detect infections that caused abortion.

The global estimated prevalence of infectious agents that caused bovine abortion based on the random effects model was 45.7% (95% CI, 38.6–52.8%) (Figure 2). The heterogeneity among studies was high (I² = 99.2%, p < 0.001); however, Egger’s test showed that there was no publication bias in the overall prevalence estimated (p = 0.239). This fact could be graphically observed in the funnel plot (Figure 3). In addition, the results of the LOOCV method showed that the overall estimated prevalence was not influenced by any specific study (Figure 4). Finally, we assessed the effect of the year of publication on the estimated prevalence of infectious agents related to bovine abortion, and although there was not a significant effect (coefficient = 0.006; p = 0.199), an increasing trend was observed over time (Figure 5).

To determine the prevalence of the different transmissible agents in bovine abortion, studies in which some laboratory tests were performed (culture or polymerase chain reaction -PCR- or foetal serology or immunohistochemistry -IHC-) to detect a specific agent (on aborted foetus or placenta) were selected. However, in this analysis were excluded studies in which diagnosis of the abortion cause was made using only maternal serology. On the other hand, the detection of an agent in abortion samples does not confirm its participation in the production of abortion. Therefore, in this study only in the cases in which specific histopathological changes were reported, it was the cause of the abortion considered confirmed. From the 19,070 cases of bovine abortion included in the 76 selected studies, in 7,319 specimens some infectious agents were detected, but only 54.4% (3,977 cases) were confirmed as the cause of foetal death.

According to this meta-analysis, Neospora caninum was the most common infectious agent found in cases of bovine abortion, with an estimated prevalence of 22.2% (95% CI: 17.6–26.8) from 9,164 aborted bovine foetuses examined (Supplementary Figure S2a). However, the contribution of this protozoan parasite to foetal death was only confirmed in 16.7% (95% CI: 12.3–21.1) of the analyzed specimens when histopathological changes were found (Table 2).

The presence of sporadically isolated or opportunistic bacteria (Salmonella spp. Escherichia coli, Trueperella pyogenes, Bacillus licheniformis, Pajaroellobacter abortibovis, Acinetobacter spp., Histophilus somni, Actinomyces spp., Aeromona spp., Bordetella spp., Cardiobacterium spp., Klebsiella spp., Enterobacter spp., Pasteurella spp., Staphylococcus spp., Streptococcus spp., Pseudomona spp., and Yersinia paratuberculosis) was observed in 21.4% (95% CI: 10.4–32.5) of the 9,824 analyzed cases from 18 articles (Supplementary Figure S2b). In twelve of these studies (4,997 cases), the authors also searched for histopathological compatible changes, confirming some of these bacteria as causes of foetal death in 12.6% (95% CI: 7.9–17.3) (Table 2).

In third place, the prevalence of the Chlamydiaceae family was 10.9% (95% CI: 4.2–17.7) of 2,651 aborted foetuses (Supplementary Figure S2c). The detection of its infection was mostly carried out by PCR, and only compatible lesions were reported in 6.8% (95% CI: 0.0–13.5) of the cases (Table 2). The specific Chlamydia species were only reported in 10 studies (1,817 cases) mentioning the presence of C. abortus (90 cases), C. psittaci (10 cases), C. suis (1 case) and C. pecorum (1 case). In addition, five studies mentioned the detection of Parachlamydia spp., although its role as an abortigenic agent in these studies could not be confirmed (absence of compatible lesions). On the other hand, the estimated prevalence of Coxiella burnetii in bovine abortion was 9.5% (95% CI: 3.7–15.3) in the analyzed cases (n = 7,987) (Supplementary Figure S2d), and this agent was detected mostly by PCR (Table 2). Interestingly, this bacterium was confirmed as the cause of abortion (compatible histopathological changes) in only 1.1% (95% CI: −0.1–2.3) of the analyzed cases (n = 2,008).

Infection with Leptospira spp. was detected in 5.2% (95% CI: 1.7–8.6) of 5,522 aborted foetuses from 17 studies (Supplementary Figure S2e), but in any selected study was mentioned the Leptospira species found in the specimen analyzed. This spirochete was detected mostly by PCR and histopathology (HP) and was confirmed as a cause of death in 5.2% (95% CI: 0.5–9.9) of the cases (Table 2). In addition, the search of Brucella spp. was made in 4,238 specimens from 15 studies, yielding a prevalence rate of 5.1% (95% CI: 1.5–8.6) (Supplementary Figure S2f), although only it was confirmed as a cause of abortion in 3.2% (95% CI: 0.2–6.2). Only five studies (556 cases) detected Brucella spp., including the presence of B. abortus (39/556) and B. melitensis (22/556). Remarkably, the presence of Campylobacter spp. was detected in 1.3% (95% CI: 0.2–2.3) of 5,312 abortion cases (Supplementary Figure S2g) and only in 7 studies (1,805 cases) was the species of Campylobacter specified. The species found were C. fetus (21/1,805), C. fetus subsp. fetus (17/1,805) and C. fetus subsp. venerealis (10/1,805) and C. jejuni (4/1,805). The final diagnosis of campylobacteriosis was confirmed in 2.3% (95% CI: 0.3–4.3) of foetal tissues referred for pathological diagnosis (4,218 cases). Listeria monocytogenes was detected in 2.6% (95% CI: 0.5–5.2) of the 9,103 analyzed cases (Supplementary Figure S2h). On the other hand, Tritrichomonas foetus was detected in 2.3% (95% CI: −0.1–4.7) of 608 cases, although only five included studies reported the presence of this protozoa (Supplementary Figure S2i). Finally, the diagnosis of fungi as causal agents of bovine reproductive failure was investigated in 10,216 cases from 19 studies. The participation of these agents in reproductive bovine failure was 3.4% (95% CI: 1.5–5.3) (Supplementary Figure S2j), with different species of Aspergillus being the most commonly diagnosed. In fourteen of these studies (5,341 cases), the authors also searched for histopathological compatible changes, confirming some fungi as causes of foetal death in 1.4% (95% CI: 0.7–2.0) (Table 2).

In the case of viral pathogens, the most commonly involved in bovine reproductive failure were bovine herpes virus type 1 (BoHV-1) and bovine viral diarrhea virus (BVDV) (Table 2). The estimated prevalence of BoHV-1 was 5.9% (95% CI: 1.5–10.2) from 6,006 total cases (Supplementary Figure S2k) and histopathological changes were found in 0.9% (95% CI: 0.2–1.5) of the cases submitted for histopathological examination (n = 5,496). In the case of BVDV, the prevalence rate was 4.7% (95% CI: 2.7–6.6) among the 11,348 analyzed cases (Supplementary Figure S2l) and the diagnosis was confirmed in 1.9% (95% CI: 1.0–2.9) of foetal samples analyzed by histopathology (n = 5,232). For both viral pathogens, the viral antigen was detected mostly using PCR or direct ELISA or direct fluorescent antibody tests (DFAT). In addition, in four studies, the presence of BoHV-4 DNA was reported in aborted foetuses, although a clear link between this viral infection and abortion was not provided.

For all pathogens, the heterogeneity among studies was high (Table 2). However, the subgroup analysis based on the diagnostic method showed that regardless of the technique used for its analysis, the heterogeneity remained high, discarding this as the cause of such heterogeneity (Supplementary Figure S3). In addition, the results of the LOOCV method showed that the overall estimated prevalence for each infectious agent was not influenced by any specific study (Supplementary Figure S4). Finally, there was no effect of study year on the prevalence of all pathogens (p > 0.05) (Supplementary Figure S5).