Advertisement

Quantification of risk factors for bovine viral diarrhea virus in cattle herds: A systematic search and meta-analysis of observational studies

Open AccessPublished:July 31, 2020DOI:https://doi.org/10.3168/jds.2020-18193

      ABSTRACT

      Bovine viral diarrhea virus (BVDV) is endemic in many parts of the world, and multiple countries have implemented surveillance activities for disease control or eradication. In such control programs, the disease-free status can be compromised by factors that pose risks for introduction or persistence of the virus. The aim of the present study was to gain a comprehensive overview of possible risk factors for BVDV infection in cattle herds in Europe and to assess their importance. Papers that considered risk factors for BVDV infection in cattle were identified through a systematic search. Further selection of papers eligible for quantitative analysis was performed using a predefined checklist, including (1) appropriate region (i.e., studies performed in Europe), (2) representativeness of the study population, (3) quality of statistical analysis, and (4) availability of sufficient quantitative data. In total, 18 observational studies were selected. Data were analyzed by a random-effects meta-analysis to obtain pooled estimates of the odds of BVDV infection. Meta-analyses were performed on 6 risk factors: herd type, herd size, participation in shows or markets, introduction of cattle, grazing, and contact with other cattle herds on pasture. Significant higher odds were found for dairy herds (odds ratio, OR = 1.63, 95% confidence interval, CI: 1.06–2.50) compared with beef herds, for larger herds (OR = 1.04 for every 10 extra animals in the herd, 95% CI: 1.02–1.06), for herds that participate in shows or markets (OR = 1.45, 95% CI: 1.10–1.91), for herds that introduced cattle into the herd (OR = 1.41, 95% CI: 1.18–1.69), and for herds that share pasture or have direct contact with cattle of other herds at pasture (OR = 1.32, 95% CI: 1.07–1.63). These pooled values must be interpreted with care, as there was a high level of heterogeneity between studies. However, they do give an indication of the importance of the most frequently studied risk factors and can therefore assist in the development, evaluation, and optimization of BVDV control programs.

      Key words

      INTRODUCTION

      Bovine viral diarrhea (BVD) virus (BVDV) is a pestivirus belonging to the Flaviviridae family (
      • Olafson P.
      • Rickard C.G.
      Further observations on the virus diarrhea (new transmissible disease) of cattle.
      ). It is one of the most common viral diseases in cattle and endemic in many parts of the world (
      • Scharnböck B.
      • Roch F.-F.
      • Richter V.
      • Funke C.
      • Firth C.L.
      • Obritzhauser W.
      • Baumgartner W.
      • Käsbohrer A.
      • Pinior B.
      A meta-analysis of bovine viral diarrhoea virus (BVDV) prevalences in the global cattle population.
      ). Bovine viral diarrhea virus is mainly spread by persistently infected (PI) cattle, which were infected in utero between 40 and 120 d of gestation and shed large amounts of virus into the environment after birth (
      • McClurkin A.W.
      • Littledike E.T.
      • Cutlip R.C.
      • Frank G.H.
      • Coria M.F.
      • Bolin S.R.
      Production of cattle immunotolerant to bovine viral diarrhea virus.
      ). Bovine viral diarrhea virus can be transmitted directly through nose-to-nose contact between cattle or indirectly through contaminated materials (
      • Tråvén M.
      • Alenius S.
      • Fossum C.
      • Larsson B.
      Primary bovine viral diarrhoea virus infection in calves following direct contact with a persistently viraemic calf.
      ;
      • Niskanen R.
      • Lindberg A.
      Transmission of bovine viral diarrhoea virus by unhygienic vaccination procedures, ambient air, and from contaminated pens.
      ). Infections with BVDV can lead to respiratory and reproductive issues, causing major economic losses (
      • Houe H.
      Economic impact of BVDV infection in dairies.
      ). Many European countries have implemented BVDV control or eradication programs, and some have already successfully eradicated the virus or reached a herd-level prevalence below 1.5% (Sweden, Norway, Finland, Denmark, Germany, Austria, Switzerland, and Ireland;
      • Nuotio L.
      • Juvonen M.
      • Neuvonen E.
      • Sihvonen L.
      • Husu-Kallio J.
      Prevalence and geographic distribution of bovine viral diarrhoea (BVD) infection in Finland 1993–1997.
      ;
      • Bitsch V.
      • Hansen K.E.
      • Rønsholt L.
      Experiences from the Danish programme for eradication of bovine virus diarrhoea (BVD) 1994–1998 with special reference to legislation and causes of infection.
      ;
      • Hult L.
      • Lindberg A.
      Experiences from BVDV control in Sweden.
      ;
      • Rikula U.
      • Nuotio L.
      • Aaltonen T.
      • Ruoho O.
      Bovine viral diarrhoea virus control in Finland 1998–2004.
      ;
      • Rossmanith W.
      • Deinhofer M.
      • Janacek R.
      • Trampler R.
      • Wilhelm E.
      Voluntary and compulsory eradication of bovine viral diarrhoea virus in Lower Austria.
      ;
      • Presi P.
      • Struchen R.
      • Knight-Jones T.
      • Scholl S.
      • Heim D.
      Bovine viral diarrhea (BVD) eradication in Switzerland—Experiences of the first two years.
      ;
      • Norström M.
      • Jonsson M.E.
      • Åkerstedt J.
      • Whist A.C.
      • Kristoffersen A.B.
      • Sviland S.
      • Hopp P.
      • Wahlström H.
      Estimation of the probability of freedom from bovine virus diarrhoea virus in Norway using scenario tree modelling.
      ;
      • Foddai A.
      • Stockmarr A.
      • Boklund A.
      Evaluation of temporal surveillance system sensitivity and freedom from bovine viral diarrhea in Danish dairy herds using scenario tree modelling.
      ;
      • AHI (Animal Health Ireland)
      Programme results.
      ). Within those control programs, animals, herds, regions, or the country are ascribed a BVDV-free status that is subsequently monitored.
      The probability that a herd categorized as free within a control program is truly free of infection will be influenced by risk factors for introduction of the virus (i.e., the probability that the virus is introduced or reintroduced into the herd between test moments) and factors that cause delayed detection of the virus after introduction or reintroduction (i.e., the probability that the virus had been introduced but not yet detected). The effectiveness of surveillance relies on an understanding of these risk factors. Delayed detection of the virus can be associated with herd management, control program design (e.g., test population, test frequency, sample size, test validity), and test performance. Risk factors for introduction depend on the contact structure between herds, such as purchase or contact with cattle from neighboring herds. The introduction of purchased animals is a well-known risk factor. However, an overview of the magnitude of the risk, and of country-level differences, is lacking.
      Risk factors for introduction and delayed detection of BVDV are not easily studied in isolation due to the difficulty of determining exactly when the virus is introduced into a herd. Risk factors for the presence of infection are more often reported (e.g.,
      • Graham D.A.
      • Clegg T.A.
      • Lynch M.
      • More S.J.
      Herd-level factors associated with the presence of bovine viral diarrhoea virus in herds participating in the voluntary phase of the Irish national eradication programme.
      ;
      • Byrne A.W.
      • Guelbenzu-Gonzalo M.
      • Strain S.A.J.
      • McBride S.
      • Graham J.
      • Lahuerta-Marin A.
      • Harwood R.
      • Graham D.A.
      • McDowell S.
      Assessment of concurrent infection with bovine viral diarrhoea virus (BVDV) and Mycobacterium bovis: A herd-level risk factor analysis from Northern Ireland.
      ;
      • Amelung S.
      • Hartmann M.
      • Haas L.
      • Kreienbrock L.
      Factors associated with the bovine viral diarrhoea (BVD) status in cattle herds in Northwest Germany.
      ) and could serve as a proxy for introduction and delayed detection. In this study, we have conducted a systematic literature search, seeking to gain a comprehensive overview of possible risk factors for the presence of BVDV infection in cattle herds in Europe. We aimed to assess the importance of the most frequently studied risk factors and, depending on study quality and the availability of quantitative data, to perform meta-analyses to obtain pooled values. This information is critical for the development, evaluation, and optimization of BVDV control programs. Control program managers can list and prioritize risk factors in their country based on the pooled values or choose the results from countries most comparable with their situation.

      MATERIALS AND METHODS

      This systematic review was reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (
      • Moher D.
      • Liberati A.
      • Tetzlaff J.
      • Altman D.G.
      Preferred Reporting Items for Systematic Reviews and Meta-Analyses: The PRISMA statement.
      ) with the PRISMA 2009 Checklist (Supplemental File S1, https://doi.org/10.3168/jds.2020-18193).

       Search Strategy

      Three databases (PubMed, CAB Abstracts, and Scopus) were interrogated using the search terms defined below. The final complete data search in all 3 databases was performed on September 21, 2018. An additional search was performed after the full-text screening and before data analysis on July 15, 2019. This additional search was performed only in PubMed because Scopus and CAB Abstracts do not allow selection for specific publication dates, only per year.
      The research questions include 4 key aspects: BVDV, risk factors, introduction, and delayed detection. The BVDV search terms included the following: BVD, BVDV, bovine viral diarrh(o)ea, bovine viral diarrh(o)ea, and bovine viral diarrh(o)ea virus. Risk factor search terms included the following: risk factor, purchase, import, trade, market, grazing, nose-to-nose contact, direct contact, over the fence contact, density, contact structure, herd, herd size, seasonal calving, calving pattern, housing system, management, biosecurity, vaccination, artificial insemination, embryo transfer, PI, persistent infection, and persistently infected. Introduction search terms included the following (where * indicates a wildcard): introduction, pathway, epidemio*, incidence, prevalence, and contamin*. Finally, delayed detection search terms included the following: diagnostic test, persist*, delayed detection, test strategy, test scheme, test performance, test characteristics, sensitivity, control program*, eradication program*, surveillance, false negative, free, freedom, transmission, and spread. The full electronic search strategy is included in Supplemental File S2 (https://doi.org/10.3168/jds.2020-18193).

       Study Selection

      Studies published in peer-reviewed journals with full text available were considered. They reported either risk factors for introduction of BVDV in cattle herds or risk factors for the presence of BVDV from which risk factors for introduction could be inferred. During the initial screening, studies were also included from which risk factors for delayed detection could be inferred (e.g., studies reporting test characteristics). In a later stage, it was decided to focus on risk factors for introduction and presence of BVDV to narrow down the search. Only studies with a cross-sectional, cohort, case-control, or randomized controlled trial study design were considered. Languages that were accepted were English, Dutch, French, Spanish, and German. Studies published since 1980 were included to focus on modern farm management systems.
      The search in PubMed, CAB Abstracts, and Scopus was carried out by one researcher (AvR). The researcher imported all references into the online systematic review management tool Covidence (Veritas Health Innovation, Melbourne, VIC, Australia). In Covidence, duplicates were deleted automatically or following a manual review. Two researchers (AvR and MM) both went independently through the following consecutive phases of the review: (1) screening titles and abstracts based on the inclusion criteria described above, and (2) reviewing full-text articles based on the inclusion criteria described above. After these review steps, conflicting opinions on papers were discussed with the other co-authors to reach consensus on inclusion or exclusion.
      All full-text studies that were selected based on the inclusion criteria were further assessed for their appropriateness for meta-analyses by one researcher (AvR). This was done using the approach presented in Table 1. This checklist consists of 4 questions regarding internal validity (how well is the study conducted?) and external validity (generalizability). As no generic tool is available for appraisal of observational studies for meta-analysis (
      • Sanderson S.
      • Tatt I.D.
      • Higgins J.
      Tools for assessing quality and susceptibility to bias in observational studies in epidemiology: A systematic review and annotated bibliography.
      ), we created our own checklist with relevant checkpoints based on our own observations and in alignment with the methods used in previous studies (
      • National Institutes of Health
      Study quality assessment tools.
      ;
      • Downes M.J.
      • Brennan M.L.
      • Williams H.C.
      • Dean R.S.
      Development of a critical appraisal tool to assess the quality of cross-sectional studies (AXIS).
      ).
      Table 1Checklist study appraisal for quantitative analysis
      ItemNot appropriate for meta-analysisAppropriate for meta-analysis
      External validity
       1. Is the cattle production system comparable with the European situation?Studies were performed outside of Europe.Studies were performed in Europe.
       2. Are the selected animals or herds representative of the target population (commercial cattle herds in Europe)?No, with high possibility of selection bias. Animals or herds are selected purposively.Yes, with low or medium possibility of selection bias. Animals or herds are selected randomly or in a way that represents the target population.
      Internal validity
       3. Was the unit of interest appropriate for a herd-level risk factor study?Animal-level data were used without correction for within-herd correlation.Herd-level data or animal-level data that were corrected for clustering were used.
       4. Are quantitative data available?No, there are only descriptive studies, or some quantitative data but no odds ratios or data from which odds ratios could be derived.Yes, there are quantitative data (odds ratios or data to derive odds ratios) of univariable or multivariable analysis.
      On several occasions, multiple studies were described in a single paper (so-called split studies)—for example, if a risk factor study was performed on different outcome variables (e.g., antibody or virus) or different types of cattle (e.g., beef or dairy) or if more than one final risk factor model was developed. We decided to include both split studies where beef and dairy herds were analyzed separately because these risk factor analyses were performed on different populations (e.g.,
      • Gates M.C.
      • Woolhouse M.E.J.
      • Gunn G.J.
      • Humphry R.W.
      Relative associations of cattle movements, local spread, and biosecurity with bovine viral diarrhoea virus (BVDV) seropositivity in beef and dairy herds.
      ,
      • Gates M.C.
      • Humphry R.W.
      • Gunn G.J.
      • Woolhouse M.E.
      Not all cows are epidemiologically equal: Quantifying the risks of bovine viral diarrhoea virus (BVDV) transmission through cattle movements.
      ). When studies concluded with more than one final model, the model indicated by the authors as best describing the data was included. If no choice was made between the different final models, we selected the model that took into account the full data set. Risk factor analyses performed on subsets of the data were excluded.

       Data Collection

      Data were extracted from all selected studies using an Excel (Microsoft Corp., Redmond, WA) form that was prepared in advance. Data were extracted by one researcher (AvR) and checked by the other researcher (MM). A pilot test of the Excel form was conducted by these 2 researchers working together on 3 selected papers to increase uniformity in extracting the data.
      For each selected study, detailed data were extracted regarding study type, location, size of the study population, diagnostic tests used, risk factors studied in univariable and multivariable analysis, the effect size (odds ratio, OR; relative risk), confidence intervals, and the statistical analysis that was performed.

       Meta-Analysis

      All risk factors from the studies that were selected for quantitative analysis were listed and combined into groups of similar risk factors. Per group, OR reported in at least 2 independent studies were analyzed by a random-effects meta-analysis to obtain pooled estimates of the odds of BVDV infection. In some cases, variables first had to be restructured to be able to include them in the meta-analysis. For example, this was the case with introduction of cattle where we wanted to combine variables with “yes introduction” versus “no introduction” with categorical variables where different numbers of introduced cattle were compared with zero introduction. In this case, we first performed a within-study fixed-effects meta-analysis on the different categories of this variable to obtain a summary estimate across all categories. This summary estimate could subsequently be included in the overall meta-analysis for introduction of cattle.
      A random-effects approach is considered the default method in meta-analysis of observational studies (
      • Mueller M.
      • D'Addario M.
      • Egger M.
      • Cevallos M.
      • Dekkers O.
      • Mugglin C.
      • Scott P.
      Methods to systematically review and meta-analyse observational studies: A systematic scoping review of recommendations.
      ). This approach accounts for the fact that the study effect estimates are not drawn from a single population, which would be the case when using a fixed-effects approach (
      • Harrer M.
      • Cuijpers P.
      • Furukawa T.A.
      • Ebert D.D.
      Doing Meta-Analysis in R: A Hands-on Guide.
      ). The random-effects models were fitted in a 2-step approach. First, between-study variance, represented by the distribution of the true study effects, was estimated with the DerSimonian-Laird approach. Then, weights were assigned to all included studies based on the inverse of the variance as in general the population size between observational studies is not equal and pooled OR were estimated (
      • Viechtbauer W.
      Conducting meta-analyses in R with the metafor package.
      ). In this process, the OR and their 95% confidence intervals (CI) as reported in the individual studies were log-transformed; therefore, due to rounding errors, the 95% CI in our results might differ slightly from the data reported in the individual studies. Preferably, adjusted OR that resulted from multivariable analysis were used. When no multivariable results were available, crude OR that resulted from univariable analysis were included. If no OR were available but frequencies were reported, OR were calculated. In each forest plot, the univariable results were marked. Also, subanalyses were performed in which univariable and multivariable results were analyzed separately.
      Heterogeneity between studies was studied by the I2 statistic. The I2 statistic shows what proportion of the variance is due to heterogeneity in true effects rather than sampling error (
      • Borenstein M.
      • Higgins J.P.
      • Hedges L.V.
      • Rothstein H.R.
      Basics of meta-analysis: I2 is not an absolute measure of heterogeneity.
      ). To identify studies with the greatest influence on the results, an influential case analysis was performed with cut-off values proposed by
      • Viechtbauer W.
      • Cheung M.W.L.
      Outlier and influence diagnostics for meta-analysis.
      . The studies indicated as outliers were marked in each forest plot. The change in the summary estimates and I2 statistic when retaining or removing outliers was of minor importance. Publication bias could not be properly assessed due to the low number of studies included in our meta-analyses (n < 10;
      • Higgins J.P.T.
      • Thomas J.
      • Chandler J.
      • Cumpston M.
      • Li T.
      • Page M.J.
      • Welch V.A.
      Cochrane Handbook for Systematic Reviews of Interventions. Version 6.0.
      ). Funnel plots were checked for asymmetry, with some indication of publication bias, but these plots are not reported as it was not possible to determine whether this was by chance or real asymmetry due to the low number of studies. Meta-analyses were performed using R statistical software (
      • R Core Team
      R: A Language and Environment for Statistical Computing..
      ) and the metafor package (
      • Viechtbauer W.
      Conducting meta-analyses in R with the metafor package.
      ).

      RESULTS

       Literature Search

      The original searches revealed 12,028 papers, of which ultimately 259 papers were full-text screened and narrowed down to 51 papers (Figure 1). Based on Table 1, all 51 papers were screened for their appropriateness for quantitative analyses (Supplemental File S3, https://doi.org/10.3168/jds.2020-18193). Eventually, 18 papers (20 studies) were selected for inclusion in the meta-analysis (Table 2).
      Figure thumbnail gr1
      Figure 1Flow diagram showing the total number of papers identified and excluded per stage of the selection process. At the eligibility stage, we decided to exclude papers that were initially selected for delayed detection. The 175 “no risk factor(s) for BVDV studied” papers were about BVDV test characteristics. BVDV = bovine viral diarrhea virus.
      Table 2Studies selected for the meta-analyses
      Study ID
      Study ID were assigned to the 51 papers that were selected in the second-last selection step (Supplemental File S3, https://doi.org/10.3168/jds.2020-18193).
      StudyCountryStudy designUnit of interest in risk factor analysisType of cattle studiedOutcome measureDiagnostic test
      BTM = bulk tank milk; VNT = virus neutralization test.
      2
      • Amelung S.
      • Hartmann M.
      • Haas L.
      • Kreienbrock L.
      Factors associated with the bovine viral diarrhoea (BVD) status in cattle herds in Northwest Germany.
      GermanyCross-sectional2,542 herdsCombinationVirusELISA on ear notch followed by PCR on ear notch
      5B
      These rows represent one of 2 or 3 studies presented in a single paper. Each of these studies was chosen for inclusion in further analyses because they either present the best final model or were performed on the full data set. Excluded split studies can be found in Supplemental File S3.
      • Barrett D.
      • Parr M.
      • Fagan J.
      • Johnson A.
      • Tratalos J.
      • Lively F.
      • Diskin M.
      • Kenny D.
      Prevalence of bovine viral diarrhoea virus (BVDV), bovine herpes virus 1 (BHV 1), leptospirosis and neosporosis, and associated risk factors in 161 Irish beef herds.
      IrelandCross-sectional139 herdsBeefVirusOn ear notch
      7
      • Bishop H.
      • Erkelens J.
      • Van Winden S.
      Indications of a relationship between buying-in policy and infectious diseases on dairy farms in Wales.
      WalesCross-sectional36 herdsDairyAntibodiesELISA on BTM
      9B
      These rows represent one of 2 or 3 studies presented in a single paper. Each of these studies was chosen for inclusion in further analyses because they either present the best final model or were performed on the full data set. Excluded split studies can be found in Supplemental File S3.
      • Byrne A.W.
      • Guelbenzu-Gonzalo M.
      • Strain S.A.J.
      • McBride S.
      • Graham J.
      • Lahuerta-Marin A.
      • Harwood R.
      • Graham D.A.
      • McDowell S.
      Assessment of concurrent infection with bovine viral diarrhoea virus (BVDV) and Mycobacterium bovis: A herd-level risk factor analysis from Northern Ireland.
      Northern IrelandCross-sectional2,827 herdsCombinationVirusPCR on ear notch
      11
      • Charoenlarp W.
      • Frankena K.
      • Strain S.A.J.
      • Guelbenzu-Gonzalo M.
      • Graham J.
      • Byrne A.W.
      Spatial and risk factor analysis of bovine viral diarrhoea (BVD) virus after the first-year compulsory phase of BVD eradication programme in Northern Ireland.
      Northern IrelandCross-sectional17,186 herdsCombinationVirusELISA, PCR, or both on ear notch
      15
      • Ersbøll A.K.
      • Ersbøll B.K.
      • Houe H.
      • Alban L.
      • Kjeldsen A.M.
      Spatial modelling of the between-herd infection dynamics of bovine virus diarrhoea virus (BVDV) in dairy herds in Denmark.
      DenmarkCohort7,921 herdsDairyVirusELISA on BTM and blood
      18A
      These rows represent one of 2 split studies presented in a single paper. Each of these studies had been conducted on different herds (beef or dairy) and has been analyzed separately.
      • Gates M.C.
      • Humphry R.W.
      • Gunn G.J.
      • Woolhouse M.E.
      Not all cows are epidemiologically equal: Quantifying the risks of bovine viral diarrhoea virus (BVDV) transmission through cattle movements.
      ScotlandCross-sectional255 herdsBeefAntibodiesELISA on blood
      18B
      These rows represent one of 2 split studies presented in a single paper. Each of these studies had been conducted on different herds (beef or dairy) and has been analyzed separately.
      • Gates M.C.
      • Humphry R.W.
      • Gunn G.J.
      • Woolhouse M.E.
      Not all cows are epidemiologically equal: Quantifying the risks of bovine viral diarrhoea virus (BVDV) transmission through cattle movements.
      ScotlandCross-sectional189 herdsDairyAntibodiesELISA on blood
      19A
      These rows represent one of 2 split studies presented in a single paper. Each of these studies had been conducted on different herds (beef or dairy) and has been analyzed separately.
      • Gates M.C.
      • Woolhouse M.E.J.
      • Gunn G.J.
      • Humphry R.W.
      Relative associations of cattle movements, local spread, and biosecurity with bovine viral diarrhoea virus (BVDV) seropositivity in beef and dairy herds.
      ScotlandCase-control249 herds (65 cases and 184 controls)BeefAntibodiesELISA on blood
      19B
      These rows represent one of 2 split studies presented in a single paper. Each of these studies had been conducted on different herds (beef or dairy) and has been analyzed separately.
      • Gates M.C.
      • Woolhouse M.E.J.
      • Gunn G.J.
      • Humphry R.W.
      Relative associations of cattle movements, local spread, and biosecurity with bovine viral diarrhoea virus (BVDV) seropositivity in beef and dairy herds.
      ScotlandCase-control185 herds (119 cases and 66 controls)DairyAntibodiesELISA on blood
      20A
      These rows represent one of 2 or 3 studies presented in a single paper. Each of these studies was chosen for inclusion in further analyses because they either present the best final model or were performed on the full data set. Excluded split studies can be found in Supplemental File S3.
      • Graham D.A.
      • Clegg T.A.
      • Lynch M.
      • More S.J.
      Herd-level factors associated with the presence of bovine viral diarrhoea virus in herds participating in the voluntary phase of the Irish national eradication programme.
      IrelandCross-sectional3,894 herdsCombinationVirusELISA or PCR on ear notch
      21A
      These rows represent one of 2 or 3 studies presented in a single paper. Each of these studies was chosen for inclusion in further analyses because they either present the best final model or were performed on the full data set. Excluded split studies can be found in Supplemental File S3.
      • Graham D.A.
      • Clegg T.A.
      • Thulke H.H.
      • O'sullivan P.
      • McGrath G.
      • More S.J.
      Quantifying the risk of spread of bovine viral diarrhoea virus (BVDV) between contiguous herds in Ireland.
      IrelandCross-sectional58,479 herdsCombinationVirusUnknown
      22
      • Hanon J.B.
      • De Baere M.
      • de la Ferté C.
      • Roelandt S.
      • Guillot G.
      • Van der Stede Y.
      • Cay B.
      Serological monitoring on milk and serum samples in a BVD eradication program: A field study in Belgium showing antibody ELISA performances and epidemiological aspects.
      BelgiumCross-sectional51 herds and 3,017 cattleCombinationAntibodiesDifferent ELISA and VNT on blood and milk
      24A
      These rows represent one of 2 or 3 studies presented in a single paper. Each of these studies was chosen for inclusion in further analyses because they either present the best final model or were performed on the full data set. Excluded split studies can be found in Supplemental File S3.
      • Houe H.
      • Baker J.C.
      • Maes R.K.
      • Lloyd J.W.
      • Enevoldsen C.
      Comparison of the prevalence and incidence of infection with bovine virus diarrhoea virus (BVDV) in Denmark and Michigan and association with possible risk factors.
      ,b)
      DenmarkCross-sectional19 herdsDairyVirusVirus isolation and virus neutralization on blood
      30
      • Mainar-Jaime R.C.
      • Berzal-Herranz B.
      • Arias P.
      • Rojo-Vázquez F.A.
      Epidemiological pattern and risk factors associated with bovine viral-diarrhoea virus (BVDV) infection in a non-vaccinated dairy-cattle population from the Asturias region of Spain.
      SpainCross-sectional529 cattleDairyAntibodiesELISA on blood
      31
      • Martinez-Ibeas A.M.
      • Power C.
      • McClure J.
      • Sayers R.G.
      Prevalence of BoHV-1 seropositive and BVD virus positive bulls on Irish dairy farms and associations between bull purchase and herd status.
      Republic of IrelandCross-sectional305 herdsDairyAntibodiesELISA on BTM and blood
      35
      • Presi P.
      • Struchen R.
      • Knight-Jones T.
      • Scholl S.
      • Heim D.
      Bovine viral diarrhea (BVD) eradication in Switzerland—Experiences of the first two years.
      SwitzerlandCross-sectional33,188 herdsCombinationVirusELISA or PCR on ear notch
      40A
      These rows represent one of 2 or 3 studies presented in a single paper. Each of these studies was chosen for inclusion in further analyses because they either present the best final model or were performed on the full data set. Excluded split studies can be found in Supplemental File S3.
      • Sarrazin S.
      • Veldhuis A.
      • Méroc E.
      • Vangeel I.
      • Laureyns J.
      • Dewulf J.
      • Caij A.B.
      • Piepers S.
      • Hooyberghs J.
      • Ribbens S.
      • Van Der Stede Y.
      Serological and virological BVDV prevalence and risk factor analysis for herds to be BVDV seropositive in Belgian cattle herds.
      BelgiumCross-sectional664 herdsCombinationAntibodies and virusELISA on blood
      49A
      These rows represent one of 2 or 3 studies presented in a single paper. Each of these studies was chosen for inclusion in further analyses because they either present the best final model or were performed on the full data set. Excluded split studies can be found in Supplemental File S3.
      • Valle P.S.
      • Martin S.W.
      • Tremblay R.
      • Bateman K.
      Factors associated with being a bovine-virus diarrhoea (BVD) seropositive dairy herd in the Møre and Romsdal County of Norway.
      NorwayCase-control314 herds (162 cases and 152 controls)DairyAntibodiesBTM screening and pooled milk sample followed by ELISA on blood
      50
      • Williams D.
      • Winden S.V.
      Risk factors associated with high bulk milk antibody levels to common pathogens in UK dairies.
      United KingdomCross-sectional1,088 herdsDairyAntibodiesELISA on BTM
      1 Study ID were assigned to the 51 papers that were selected in the second-last selection step (Supplemental File S3, https://doi.org/10.3168/jds.2020-18193).
      2 BTM = bulk tank milk; VNT = virus neutralization test.
      3 These rows represent one of 2 or 3 studies presented in a single paper. Each of these studies was chosen for inclusion in further analyses because they either present the best final model or were performed on the full data set. Excluded split studies can be found in Supplemental File S3.
      4 These rows represent one of 2 split studies presented in a single paper. Each of these studies had been conducted on different herds (beef or dairy) and has been analyzed separately.

       Overview of Risk Factors

      All risk factors that were studied in the final 18 papers were grouped into 6 risk factor categories: (1) herd and animal characteristics, (2) cattle movement, (3) reproduction, (4) neighborhood risk, (5) farm management and biosecurity, and (6) diagnostic testing and control programs.

       Description of Risk Factors

       Herd and Animal Characteristics

      Herd and animal characteristics that were studied included milk yield, sex, age, infection with other pathogens, mortality, region, herd type, and herd size. Of all herd characteristics, variables describing herd size, herd type, and region were included most frequently (Table 3).
      Table 3Overview of the number of risk factor studies (out of the selected 18 papers on 20 studies) that included herd and animal characteristics and the availability of quantitative data
      FactorNo. of studiesNo. of variablesStudy ID
      Study ID were assigned to the 51 papers that were selected in the second-last selection step (see Supplemental File S3, https://doi.org/10.3168/jds.2020-18193).
      No. of variables with quantitative data
      Milk yield222, 301
      Sex11221
      Age2222, 302
      Infection with other pathogens365B, 9B, 303
      Mortality575B, 9B, 20A, 30, 355
      Region882, 9B, 11, 15, 20A, 21A, 30, 317
      Herd type
      Included in the meta-analysis.
      9112, 9B, 11, 20A, 21A, 22, 30, 35, 40A9
      Herd size
      Included in the meta-analysis.
      14202, 5B, 9B, 11, 15, 20A, 21A, 22, 24A, 30, 31, 35, 40A, 5013
      1 Study ID were assigned to the 51 papers that were selected in the second-last selection step (see Supplemental File S3, https://doi.org/10.3168/jds.2020-18193).
      2 Included in the meta-analysis.
      No further analysis could be performed on milk yield and sex as for both there was only 1 study with quantitative data. Age was included as a categorical variable in 2 studies (
      • Mainar-Jaime R.C.
      • Berzal-Herranz B.
      • Arias P.
      • Rojo-Vázquez F.A.
      Epidemiological pattern and risk factors associated with bovine viral-diarrhoea virus (BVDV) infection in a non-vaccinated dairy-cattle population from the Asturias region of Spain.
      ;
      • Hanon J.B.
      • De Baere M.
      • de la Ferté C.
      • Roelandt S.
      • Guillot G.
      • Van der Stede Y.
      • Cay B.
      Serological monitoring on milk and serum samples in a BVD eradication program: A field study in Belgium showing antibody ELISA performances and epidemiological aspects.
      ), both with higher OR for the presence of BVD antibodies in higher age classes. However, the age categories within those 2 studies were not comparable and therefore were unsuitable for meta-analysis. Infection with other pathogens associated with BVD infection was considered in 3 studies but could not be compared because different pathogens were studied (i.e., Neospora caninum, bovine herpes virus-1, and bovine tuberculosis). Mortality was considered in 5 studies, but as this was regarded more an outcome than a risk factor for BVDV, it was not included in the meta-analysis. Finally, region was not included in the meta-analysis even though this was one of the most studied risk factors within the herd and animal characteristics group. Because different regions were included in different studies, comparison of the risk estimates between regions was impossible. Nevertheless, most studies found significant differences between regions, which makes this an important risk factor to consider. Meta-analysis was performed on herd type and herd size.

       Cattle Movement

      Movement characteristics that were studied included introduction of cattle, cattle shows or markets, and other movements (e.g., sale and exchange of calves). Of all cattle movement risk factors, variables describing introduction of cattle into a herd were included most frequently (Table 4). We considered studies on introduction of cattle into a herd and on purchase, where the latter assumes monetary transfer, which is not necessarily the case with introduction. In this paper, we use “introduction,” which also covers purchase.
      Table 4Overview of the number of risk factor studies (out of the selected 18 papers on 20 studies) that included cattle movement variables and the availability of quantitative data
      FactorNo. of studiesNo. of variablesStudy ID
      Study ID were assigned to the 51 papers that were selected in the second-last selection step (see Supplemental File S3, https://doi.org/10.3168/jds.2020-18193).
      No. of variables with quantitative data
      Other movement222, 352
      Cattle shows or markets
      Included in the meta-analysis.
      552, 19A, 19B, 22, 355
      Introduction of cattle
      Included in the meta-analysis.
      17622, 5B, 7, 9B, 18A, 18B, 19A, 19B, 20A, 21A, 22, 24A, 30, 31, 35, 49A, 5048
      1 Study ID were assigned to the 51 papers that were selected in the second-last selection step (see Supplemental File S3, https://doi.org/10.3168/jds.2020-18193).
      2 Included in the meta-analysis.
      Other types of cattle movements were studied by
      • Valle P.S.
      • Martin S.W.
      • Tremblay R.
      • Bateman K.
      Factors associated with being a bovine-virus diarrhoea (BVD) seropositive dairy herd in the Møre and Romsdal County of Norway.
      and
      • Amelung S.
      • Hartmann M.
      • Haas L.
      • Kreienbrock L.
      Factors associated with the bovine viral diarrhoea (BVD) status in cattle herds in Northwest Germany.
      .
      • Valle P.S.
      • Martin S.W.
      • Tremblay R.
      • Bateman K.
      Factors associated with being a bovine-virus diarrhoea (BVD) seropositive dairy herd in the Møre and Romsdal County of Norway.
      looked at “other animal traffic,” combining mainly exchange of calves and sharing of cattle housing with other farmers during summer. They found a very high OR of 28.60 (95% CI: 3.23–252.22).
      • Amelung S.
      • Hartmann M.
      • Haas L.
      • Kreienbrock L.
      Factors associated with the bovine viral diarrhoea (BVD) status in cattle herds in Northwest Germany.
      studied sale of cattle, which was not comparable with the cattle movement studied in
      • Valle P.S.
      • Martin S.W.
      • Tremblay R.
      • Bateman K.
      Factors associated with being a bovine-virus diarrhoea (BVD) seropositive dairy herd in the Møre and Romsdal County of Norway.
      . Meta-analysis was performed on cattle shows or markets and introduction of cattle.

       Reproduction

      Reproduction variables that were studied included AI versus use of bulls and calving pattern (Table 5). The number of studies was too small or the definition of the variables varied too much between studies to enable a meta-analysis to be conducted.
      Table 5Overview of the number of risk factor studies (out of the selected 18 papers on 20 studies) that included reproduction variables and the availability of quantitative data
      FactorNo. of studiesNo. of variablesStudy ID
      Study ID were assigned to the 51 papers that were selected in the second-last selection step (see Supplemental File S3, https://doi.org/10.3168/jds.2020-18193).
      No. of variables with quantitative data
      Calving pattern11501
      AI/use of bulls342, 7, 503
      1 Study ID were assigned to the 51 papers that were selected in the second-last selection step (see Supplemental File S3, https://doi.org/10.3168/jds.2020-18193).
      Variables regarding AI or the use of bulls were included only in univariable analyses. In
      • Amelung S.
      • Hartmann M.
      • Haas L.
      • Kreienbrock L.
      Factors associated with the bovine viral diarrhoea (BVD) status in cattle herds in Northwest Germany.
      , higher but nonsignificant OR were found for BVD infection in herds with AI (OR = 1.28, 95% CI: 0.96–1.71) compared with herds without AI but also in herds with a bull for insemination (OR = 1.17, 95% CI: 0.93–1.48) compared with herds without a bull.
      • Williams D.
      • Winden S.V.
      Risk factors associated with high bulk milk antibody levels to common pathogens in UK dairies.
      compared herds with a bull present on the farm with herds with AI only and found that herds with a bull present on the farm had higher but nonsignificant odds of infection with BVD (OR = 1.16, 95% CI: 0.90–1.49). Calving pattern was found only once in a univariable risk factor analysis and showed higher odds of infection (OR = 1.80, 95% CI: 1.22–2.67) in herds with year-round calving compared with seasonal calving (
      • Williams D.
      • Winden S.V.
      Risk factors associated with high bulk milk antibody levels to common pathogens in UK dairies.
      ).

       Neighborhood Risk

      Variables related to neighborhood risk included farm fragmentation, environment, cattle density, BVD-positive neighbor herds, contact with other animal species, and pasture. Of all neighborhood risk factors, variables describing cattle density, contact with other animal species, and pasture were included most frequently (Table 6).
      Table 6Overview of the number of risk factor studies (out of the selected 18 papers on 20 studies) that included neighborhood variables and the availability of quantitative data
      FactorNo. of studiesNo. of variablesStudy ID
      Study ID were assigned to the 51 papers that were selected in the second-last selection step (Supplemental File S3, https://doi.org/10.3168/jds.2020-18193).
      No. of variables with quantitative data
      Farm fragmentation1120A1
      Environment14114
      Cattle density6911, 15, 19A, 19B, 21A, 307
      BVD
      Bovine viral diarrhea.
      -positive neighbor herds
      31111, 15, 21A8
      Contact with other animal species5102, 19A, 19B, 20A, 49A8
      Pasture
      Included in the meta-analysis.
      8202, 11, 19A, 19B, 22, 24A, 35, 49A14
      1 Study ID were assigned to the 51 papers that were selected in the second-last selection step (Supplemental File S3, https://doi.org/10.3168/jds.2020-18193).
      2 Bovine viral diarrhea.
      3 Included in the meta-analysis.
      Farm fragmentation (number of individual noncontiguous parcels of land associated with the herd) and environment (i.e., natural grassland, forest) were both studied only once; therefore, no meta-analysis could be performed. Cattle density and BVD-positive neighbor herds were studied more frequently but in such different ways that meta-analysis was not possible. Both variables describe in different ways the distance to (positive) neighboring herds or the number of (positive) neighboring herds contiguous to the farm or in a 5- or 10-km radius. They are continuous or categorical. Most studies showed higher odds of BVD infection when the distance to (positive) neighbors is shorter, when there are more (positive or unknown status) neighbors close by, or when BVD-positive animals are retained for a longer period. One study found that seropositivity increased with a larger distance (in km) to the closest dairy farm (
      • Mainar-Jaime R.C.
      • Berzal-Herranz B.
      • Arias P.
      • Rojo-Vázquez F.A.
      Epidemiological pattern and risk factors associated with bovine viral-diarrhoea virus (BVDV) infection in a non-vaccinated dairy-cattle population from the Asturias region of Spain.
      ). Variables regarding contact with other animal species included the presence of, contact with, close proximity of, or grazing with sheep, pigs, deer, or wildlife. No meta-analysis could be performed on contact with other animal species.

       Farm Management and Biosecurity

      Variables included were quarantine, vaccination, mixed beef and dairy farm, type of housing, shared equipment, people on farm, and other biosecurity. None of these variables were suitable for meta-analysis because of noncomparable definitions or the low number of studies in which these factors were studied (Table 7).
      Table 7Overview of the number of risk factor studies (out of the selected 18 papers on 20 studies) that included farm management and biosecurity variables and the availability of quantitative data
      FactorNo. of studiesNo. of variablesStudy ID
      Study ID were assigned to the 51 papers that were selected in the second-last selection step (Supplemental File S3, https://doi.org/10.3168/jds.2020-18193).
      No. of variables with quantitative data
      Other biosecurity2219A, 19B2
      Hygiene2219A, 19B2
      Quarantine337, 19A, 19B2
      Mixed beef and dairy farm3319A, 19B, 49A2
      Vaccination3422, 31, 40A2
      Housing242, 224
      Shared equipment3519A, 19B, 49A4
      People on farm2819A, 19B8
      1 Study ID were assigned to the 51 papers that were selected in the second-last selection step (Supplemental File S3, https://doi.org/10.3168/jds.2020-18193).
      Most farm management and biosecurity variables were studied by
      • Gates M.C.
      • Woolhouse M.E.J.
      • Gunn G.J.
      • Humphry R.W.
      Relative associations of cattle movements, local spread, and biosecurity with bovine viral diarrhoea virus (BVDV) seropositivity in beef and dairy herds.
      . They studied the relative influence of cattle movements, local spread, and biosecurity on BVDV seropositivity. The variables we included in the farm management and biosecurity group were not exactly identical to the classification of biosecurity variables in the study of
      • Gates M.C.
      • Woolhouse M.E.J.
      • Gunn G.J.
      • Humphry R.W.
      Relative associations of cattle movements, local spread, and biosecurity with bovine viral diarrhoea virus (BVDV) seropositivity in beef and dairy herds.
      , but especially for beef herds, cattle movement had the greatest influence on BVDV seropositivity. Also, in the other studies included in Table 7, most biosecurity variables were nonsignificant.

       Diagnostics Testing and Control Programs

      Multiple papers studied variables related to diagnostic testing and control programs that we grouped into BVDV testing, farmer behavior, control program, and other (Table 8). However, either the number of studies was too small or the definition of these variables varied too much between studies to enable a meta-analysis to be conducted.
      Table 8Overview of the number of risk factor studies (out of the selected 18 papers on 20 studies) that included diagnostic testing and control program variables and the availability of quantitative data
      FactorNo. of studiesNo. of variablesStudy ID
      Study ID were assigned to the 51 papers that were selected in the second-last selection step (Supplemental File S3, https://doi.org/10.3168/jds.2020-18193).
      No. of variables with quantitative data
      Other3311, 40A, 21A2
      Farmer (behavior)242, 49A3
      Control program332, 11, 223
      BVDV
      Bovine viral diarrhea virus.
      testing
      789B, 19A, 19B, 20A, 30, 31, 40A6
      1 Study ID were assigned to the 51 papers that were selected in the second-last selection step (Supplemental File S3, https://doi.org/10.3168/jds.2020-18193).
      2 Bovine viral diarrhea virus.
      Within the diagnostic testing and control programs group, BVDV testing was studied most. Examples of variables studied are the total number of BVDV tests undertaken and detection of PI animals in the past. One study (
      • Amelung S.
      • Hartmann M.
      • Haas L.
      • Kreienbrock L.
      Factors associated with the bovine viral diarrhoea (BVD) status in cattle herds in Northwest Germany.
      ) found that herds that participated in a control program has slightly higher odds (OR = 1.28, 95% CI: 1.01–1.64) for BVDV infection in univariable analysis than herds that do not participate. One of the studies looking at farmer behavior showed that the age of farmers was associated with the BVD status. Herds of farmers younger than 40 yr were more often infected than herds of farmers between 50 and 60 yr.

       Meta-Analyses

       Herd and Animal Characteristics

      Herd type was studied frequently and was always included as a categorical variable (i.e., dairy, beef, mixed, beef breeding; Supplemental File S4, section 4.1.1, https://doi.org/10.3168/jds.2020-18193). A meta-analysis was conducted on the 6 studies that compared dairy versus beef herds (reference category; Supplemental File S4, section 4.1.2). We found a combined effect estimate of 1.63 higher odds (95% CI: 1.06–2.50) of BVDV infection in dairy herds compared with beef herds (Figure 2). The heterogeneity between studies (I2) was 97.30% (95% CI: 91.87–99.47).
      Figure thumbnail gr2
      Figure 2Forest plot of the effect of herd type with beef herds as reference category (upper plot) and herd size per additional animal in the herd (lower plot) on bovine viral diarrhea virus infection. *Univariable result; #study indicated as outlier in the influential case analysis.
      Herd size was studied frequently and was always included as an either categorical or continuous variable (Supplemental File S4, section 4.1.1). However, very few variables were comparable; therefore, meta-analysis was conducted on the 4 studies with OR per additional cow (Supplemental File S4, section 4.1.2). Other variables showing the log number of cows or different herd size categories were not included because they were not comparable.
      For every extra animal in the herd, we found a combined effect estimate of 1.004 higher odds (95% CI: 1.002–1.006) of BVDV infection (Figure 2). For every 10 extra animals in the herd, this would be 1.04 higher odds of BVDV infection (95% CI: 1.02–1.06). The results of
      • Presi P.
      • Struchen R.
      • Knight-Jones T.
      • Scholl S.
      • Heim D.
      Bovine viral diarrhea (BVD) eradication in Switzerland—Experiences of the first two years.
      could not be included in the pooled estimate because weights are assigned to all factors based on the inverse of the variance and these results had a variance of zero. The heterogeneity between studies (I2) was 55.96% (95% CI: 0.00–99.98).

       Cattle Movement

      In all studies, participation in cattle shows or markets was included as a yes–no variable (Supplemental File S4, section 4.2.1) and therefore they could all be included in meta-analysis (Supplemental File S4, section 4.2.2). We found a combined effect estimate of 1.45 higher odds (95% CI: 1.10–1.91) of BVDV infection in herds that participated in shows or markets compared with herds that did not (Figure 3). The heterogeneity between studies (I2) was 61.70% (95% CI: 0.00–96.60).
      Figure thumbnail gr3
      Figure 3Forest plot of the effect of participation in shows or markets (upper plot) and introduction of cattle (lower plot) on bovine viral diarrhea virus infection. Gates et al., 2013 (A) and (B), refers to substudies, as indicated in . *Univariable result; #study indicated as outlier in the influential case analysis.
      Introduction of cattle was the most often studied movement variable but was not easily compared between studies because of the many different ways in which introduction of cattle was coded (i.e., introduction yes−no, source of introduced animals, continuous variables, and introduction of different types of cattle). We decided to focus further meta-analysis on introduction yes−no because these variables were most comparable (Supplemental File S4, section 4.2.2). In 2 studies (
      • Graham D.A.
      • Clegg T.A.
      • Lynch M.
      • More S.J.
      Herd-level factors associated with the presence of bovine viral diarrhoea virus in herds participating in the voluntary phase of the Irish national eradication programme.
      ,
      • Graham D.A.
      • Clegg T.A.
      • Thulke H.H.
      • O'sullivan P.
      • McGrath G.
      • More S.J.
      Quantifying the risk of spread of bovine viral diarrhoea virus (BVDV) between contiguous herds in Ireland.
      ), a sub-meta-analysis was first performed to obtain pooled estimates comparable with the estimates of the yes−no variables (Supplemental File S5, https://doi.org/10.3168/jds.2020-18193). We found a combined effect estimate of 1.41 higher odds (95% CI: 1.18–1.69) of BVDV infection in herds that introduce cattle into the herd compared with herds that do not (Figure 3). The heterogeneity between studies (I2) was 82.98% (95% CI: 71.48–99.47).

       Neighborhood Risk

      Pasturing of cattle was the most often studied neighborhood risk variable. Variables described whether cattle had access to pasture, the possibility of contact with cattle from other herds at pasture, and shared pasture (Supplemental File S4, section 4.3.1). First studies were compared that looked at the presence versus absence of pasture (Supplemental File S4) followed by contact between cattle on pasture (Supplemental File S4, section 4.3.2).
      We found a nonsignificant combined effect estimate of 1.10 higher odds (95% CI: 0.62–1.97) of BVDV infection in herds that graze their cattle compared with herds that do not (Figure 4). The heterogeneity between studies (I2) was 73.30% (95% CI: 0.83–99.80). Studies on contact between cattle at pasture were divided into shared pasture and the possibility of contact with cattle from other herds at pasture (e.g., contact over the fence) but were also analyzed together (Figure 4).
      Figure thumbnail gr4
      Figure 4Forest plot of the effect of herds grazing (upper plot) and contact between cattle on pasture by either shared pasture or over-the-fence contact (lower plot) on bovine viral diarrhea virus infection. Gates et al., 2013 (A) and (B), refers to substudies, as indicated in . *Univariable result; #study indicated as outlier in the influential case analysis.
      For both shared pasture and contact at pasture, we found nonsignificant odds of BVDV infection: 1.34 (95% CI: 0.85–2.10) and 1.33 (95% CI: 0.99–1.78), respectively (Figure 4). However, we found an overall significant combined effect estimate of 1.32 higher odds (95% CI: 1.07–1.63) of BVDV infection in herds where contact between cattle at pasture is possible either because different herds share pasture or because of contact between herds in contiguous pastures (Figure 4). The heterogeneity between studies (I2) was 53.90% (95% CI: 0.00–97.70).

      DISCUSSION

      By conducting this systematic literature search we have gained a comprehensive overview of potential risk factors for the presence of BVD in cattle herds. We decided to focus on studies performed in Europe in an attempt to reduce heterogeneity between results caused by different cattle production systems on different continents. However, the results could be generalized to areas outside Europe where there are similar cattle production systems (e.g., areas in the United States). The 18 European publications that were included in this study showed a wide range of potential risk factors that were grouped into 6 categories with similar characteristics: (1) herd and animal characteristics, (2) cattle movement, (3) reproduction, (4) neighborhood risk, (5) farm management and biosecurity, and (6) diagnostic testing and control programs. Although there was a lot of variation in risk factors between studies, we performed several meta-analyses and obtained pooled estimates for several frequently found risk factors.
      Two herd characteristics that were frequently studied were herd size and herd type. Most studies found that larger herds were associated with higher odds of BVD infection. Only
      • Hanon J.B.
      • De Baere M.
      • de la Ferté C.
      • Roelandt S.
      • Guillot G.
      • Van der Stede Y.
      • Cay B.
      Serological monitoring on milk and serum samples in a BVD eradication program: A field study in Belgium showing antibody ELISA performances and epidemiological aspects.
      found the highest seroprevalence in the smallest herds (<100 cattle). They did find a higher seroprevalence in farms with a higher number of stables (>3). The pooled estimate in our meta-analysis showed a significantly higher risk of infection per extra 10 animals in the herd (OR = 1.04, 95% CI: 1.02–1.06). This could be explained by the tendency for larger herds to have a decreased probability of self-clearance of infection and to be more likely to contain a higher number of pregnant cattle and purchased cattle, increasing the risk of introduction of PI into the herd (
      • Lindberg A.
      • Houe H.
      Characteristics in the epidemiology of bovine viral diarrhea virus (BVDV) of relevance to control.
      ;
      • Sarrazin S.
      • Veldhuis A.
      • Méroc E.
      • Vangeel I.
      • Laureyns J.
      • Dewulf J.
      • Caij A.B.
      • Piepers S.
      • Hooyberghs J.
      • Ribbens S.
      • Van Der Stede Y.
      Serological and virological BVDV prevalence and risk factor analysis for herds to be BVDV seropositive in Belgian cattle herds.
      ;
      • Barrett D.
      • Parr M.
      • Fagan J.
      • Johnson A.
      • Tratalos J.
      • Lively F.
      • Diskin M.
      • Kenny D.
      Prevalence of bovine viral diarrhoea virus (BVDV), bovine herpes virus 1 (BHV 1), leptospirosis and neosporosis, and associated risk factors in 161 Irish beef herds.
      ). In our meta-analysis, dairy herds were also found to be at higher risk of infection than beef herds (OR = 1.63, 95% CI: 1.06–2.50). It has been suggested that this is related to the higher number of contacts between cattle and people and traffic on dairy farms compared with beef farms (
      • Amelung S.
      • Hartmann M.
      • Haas L.
      • Kreienbrock L.
      Factors associated with the bovine viral diarrhoea (BVD) status in cattle herds in Northwest Germany.
      ).
      Movement of cattle is considered one of the most important risk factors for BVD infection, especially purchase (
      • Courcoul A.
      • Ezanno P.
      Modelling the spread of bovine viral diarrhoea virus (BVDV) in a managed metapopulation of cattle herds.
      ;
      • Gates M.C.
      • Woolhouse M.E.J.
      • Gunn G.J.
      • Humphry R.W.
      Relative associations of cattle movements, local spread, and biosecurity with bovine viral diarrhoea virus (BVDV) seropositivity in beef and dairy herds.
      ;
      • Qi L.
      • Beaunée G.
      • Arnoux S.
      • Dutta B.L.
      • Joly A.
      • Vergu E.
      • Ezanno P.
      Neighbourhood contacts and trade movements drive the regional spread of bovine viral diarrhoea virus (BVDV).
      ). Our meta-analysis showed higher odds (OR = 1.41, 95% CI: 1.18–1.69) for herds that introduced cattle into the herd in the previous year compared with herds that did not. However,
      • Gates M.C.
      • Humphry R.W.
      • Gunn G.J.
      • Woolhouse M.E.
      Not all cows are epidemiologically equal: Quantifying the risks of bovine viral diarrhoea virus (BVDV) transmission through cattle movements.
      illustrated that not all purchased cattle pose the same risk. They found that purchase of pregnant heifers and open cows with a calf at foot are associated with a higher risk of BVDV infection in beef herds, with OR of 2.18 (95% CI: 1.17–4.08) and 2.09 (95% CI: 1.13–3.88), respectively. The number of cattle introduced was also studied several times, generally showing increasing odds with increasing numbers of introduced cattle (
      • Gates M.C.
      • Woolhouse M.E.J.
      • Gunn G.J.
      • Humphry R.W.
      Relative associations of cattle movements, local spread, and biosecurity with bovine viral diarrhoea virus (BVDV) seropositivity in beef and dairy herds.
      ;
      • Graham D.A.
      • Clegg T.A.
      • Lynch M.
      • More S.J.
      Herd-level factors associated with the presence of bovine viral diarrhoea virus in herds participating in the voluntary phase of the Irish national eradication programme.
      ,
      • Graham D.A.
      • Clegg T.A.
      • Thulke H.H.
      • O'sullivan P.
      • McGrath G.
      • More S.J.
      Quantifying the risk of spread of bovine viral diarrhoea virus (BVDV) between contiguous herds in Ireland.
      ;
      • Byrne A.W.
      • Guelbenzu-Gonzalo M.
      • Strain S.A.J.
      • McBride S.
      • Graham J.
      • Lahuerta-Marin A.
      • Harwood R.
      • Graham D.A.
      • McDowell S.
      Assessment of concurrent infection with bovine viral diarrhoea virus (BVDV) and Mycobacterium bovis: A herd-level risk factor analysis from Northern Ireland.
      ). It was, however, suggested that the number of cattle introduced is related to herd size (
      • Graham D.A.
      • Clegg T.A.
      • Thulke H.H.
      • O'sullivan P.
      • McGrath G.
      • More S.J.
      Quantifying the risk of spread of bovine viral diarrhoea virus (BVDV) between contiguous herds in Ireland.
      ;
      • Byrne A.W.
      • Guelbenzu-Gonzalo M.
      • Strain S.A.J.
      • McBride S.
      • Graham J.
      • Lahuerta-Marin A.
      • Harwood R.
      • Graham D.A.
      • McDowell S.
      Assessment of concurrent infection with bovine viral diarrhoea virus (BVDV) and Mycobacterium bovis: A herd-level risk factor analysis from Northern Ireland.
      ), indicating the importance of correcting for herd size when studying purchase. A different way to study the risk of introduction is to look at the number of source herds.
      • Gates M.C.
      • Woolhouse M.E.J.
      • Gunn G.J.
      • Humphry R.W.
      Relative associations of cattle movements, local spread, and biosecurity with bovine viral diarrhoea virus (BVDV) seropositivity in beef and dairy herds.
      found a significant association between BVDV infection and a larger number of source herds in dairy herds (OR = 4.42 in units of 10 farms, 95% CI: 1.86–10.00) and beef herds (OR = 10.60 in units of 10 farms, 95% CI: 3.91–31.00). However, there was strong correlation between the number of cattle introduced and the number of source herds (
      • Gates M.C.
      • Woolhouse M.E.J.
      • Gunn G.J.
      • Humphry R.W.
      Relative associations of cattle movements, local spread, and biosecurity with bovine viral diarrhoea virus (BVDV) seropositivity in beef and dairy herds.
      ).
      Another risk factor related to cattle movement that was studied frequently is participation in shows or markets. Our pooled estimate shows significant higher odds of infection for herds that visit cattle shows or markets (OR = 1.45, 95% CI: 1.10–1.91) compared with herds that do not. This could be explained by the possibility that cattle come in contact with BVDV-infected cattle at the show or market and infect the herd upon returning or because of infection during transport.
      No meta-analysis could be performed on any of the reproduction variables because of the low number of comparable studies. However, concerns have been raised about transmission of BVDV by AI (
      • Gard J.A.
      • Givens M.D.
      • Stringfellow D.A.
      Bovine viral diarrhea virus (BVDV): Epidemiologic concerns relative to semen and embryos.
      ;
      • Rikula U.
      • Nuotio L.
      • Laamanen U.I.
      • Sihvonen L.
      Transmission of bovine viral diarrhoea virus through the semen of acutely infected bulls under field conditions.
      ). This may be prevented by regular testing of bulls at AI centers and testing of imported semen (
      • Eaglesome M.D.
      • Garcia M.M.
      Disease risks to animal health from artificial insemination with bovine semen.
      ;
      • Wentink G.H.
      • Frankena K.
      • Bosch J.C.
      • Vandehoek J.E.D.
      • Van Den Berg T.
      Prevention of disease transmission by semen in cattle.
      ;
      • Lindberg A.
      • Brownlie J.
      • Gunn G.
      • Houe H.
      • Moening V.
      • Saatkamp H.W.
      • Sandvik T.
      • Valle P.S.
      The control of bovine viral diarrhoea virus in Europe: Today and in the future.
      ). Also, the within-herd calving pattern could not be compared between studies, but
      • Williams D.
      • Winden S.V.
      Risk factors associated with high bulk milk antibody levels to common pathogens in UK dairies.
      found an increased likelihood of BVDV presence with year-round calving compared with seasonal calving. They indicated that this could be related to the fact that with year-round calving there are almost always pregnant cows present within the susceptible window for BVDV infection of the fetus. When developing or optimizing BVD control programs, calving pattern could be an important factor to consider. In block calving systems, tissue tag testing of newborn calves provides the opportunity to identify and remove the majority of PI calves before the breeding season commences, reducing the risk of establishing more PI calves to be born the following season. In year-round calving systems, spot testing could be a cost-effective option to monitor new infections (
      • Tratalos J.A.
      • Graham D.A.
      • More S.J.
      Patterns of calving and young stock movement in Ireland and their implications for BVD serosurveillance.
      ).
      Bovine viral diarrhea can easily spread between herds direct contact is possible between cattle (
      • Tråvén M.
      • Alenius S.
      • Fossum C.
      • Larsson B.
      Primary bovine viral diarrhoea virus infection in calves following direct contact with a persistently viraemic calf.
      ). Therefore, grazing is considered a risk factor for BVD as nose-to-nose contact between cattle of different herds may occur. However, our pooled estimate did not show significant odds (OR = 1.10, 95% CI: 0.62–1.97) for BVD infection for herds that graze compared with herds that do not. When results that indicated shared pasture were separated from results that indicated whether contact between cattle at pasture could occur (e.g., over-the-fence contact), our pooled estimates were nonsignificant, but when taken altogether and thus increasing statistical power, we found a significant effect indicating that contact between cattle at pasture had a higher odds of BVD infection (OR = 1.32, 95% CI: 1.07–1.63). The risk of grazing is likely influenced by many factors, such as cattle density and the prevalence of BVDV in the area (
      • Houe H.
      • Baker J.C.
      • Maes R.K.
      • Lloyd J.W.
      • Enevoldsen C.
      Comparison of the prevalence and incidence of infection with bovine virus diarrhoea virus (BVDV) in Denmark and Michigan and association with possible risk factors.
      ), regulations around communal grazing (
      • Rossmanith W.
      • Janacek R.
      • Wilhelm E.
      Control of BVDV-infection on common grassland—The key for successful BVDV-eradication in Lower Austria.
      ), the number of cattle and herds sharing pasture (
      • Presi P.
      • Struchen R.
      • Knight-Jones T.
      • Scholl S.
      • Heim D.
      Bovine viral diarrhea (BVD) eradication in Switzerland—Experiences of the first two years.
      ), and the number of neighbors.
      In the current study, no meta-analysis was performed on any of the farm management and biosecurity variables due to the low number of studies and the differing ways in which biosecurity was measured. It was unexpected that most studies did not find a significant association between biosecurity measures and BVDV infection because biosecurity is considered an important aspect of BVDV control (
      • Moennig V.
      • Houe H.
      • Lindberg A.
      BVD control in Europe: Current status and perspectives.
      ;
      • Lindberg A.
      • Brownlie J.
      • Gunn G.
      • Houe H.
      • Moening V.
      • Saatkamp H.W.
      • Sandvik T.
      • Valle P.S.
      The control of bovine viral diarrhoea virus in Europe: Today and in the future.
      ).
      • Gates M.C.
      • Woolhouse M.E.J.
      • Gunn G.J.
      • Humphry R.W.
      Relative associations of cattle movements, local spread, and biosecurity with bovine viral diarrhoea virus (BVDV) seropositivity in beef and dairy herds.
      suggested that this could be related to the design of questionnaires (e.g., questionnaires that primarily use closed yes–no questions, which forces farmers to choose one of the options even if neither is completely true). Farmers could also give socially desirable answers because they fear possible consequences. Farmer behavior is another factor for which there were not enough quantitative data for meta-analysis. This lack of quantitative data does not necessarily mean that farmer behavior and biosecurity are not important factors for BVD, but they are more often studied qualitatively, which made it impossible to include them in the meta-analysis. Qualitative research into farmer behavior and biosecurity related to BVD stresses the importance of addressing farmer attitudes toward BVD control (
      • Heffernan C.
      • Azbel-Jackson L.
      • Brownlie J.
      • Gunn G.
      Farmer attitudes and livestock disease: Exploring citizenship behaviour and peer monitoring across two BVD control schemes in the UK.
      ;
      • Azbel-Jackson L.
      • Heffernan C.
      • Gunn G.
      • Brownlie J.
      Exploring the role of voluntary disease schemes on UK farmer bio-security behaviours: Findings from the Norfolk-Suffolk Bovine Viral Diarrhoea control scheme.
      ). A meta-analysis on epidemiological and mitigation measures that influence production losses in cattle due to BVDV has been reported (
      • Pinior B.
      • Garcia S.
      • Minviel J.J.
      • Raboisson D.
      Epidemiological factors and mitigation measures influencing production losses in cattle due to bovine viral diarrhoea virus infection: A meta-analysis.
      ). These authors found that vaccination and biosecurity had a positive influence on the annual BVDV production losses per animal. We agree that farmers' attitudes toward BVD control and biosecurity-related measures are important and influence the effect of the risk factors we found in this paper. For example, when a new cow is kept in quarantine and tested for BVD before its introduction in the herd, the risk of introduction will be lower compared with new cows that are directly introduced in the herd. Therefore, we recomm­end further study of the quantitative association between BVD control and biosecurity and farmer behavior.
      No meta-analysis could be performed on any of the diagnostic testing and control program variables because of both the small number of studies and the large variation between variables. One study found slightly higher odds for presence of BVDV when participating in control programs in univariable analysis (
      • Amelung S.
      • Hartmann M.
      • Haas L.
      • Kreienbrock L.
      Factors associated with the bovine viral diarrhoea (BVD) status in cattle herds in Northwest Germany.
      ), which could probably be explained by the assumption that farms with BVDV problems are more likely to participate in a control program. Another interesting result was that herds of farmers younger than 40 yr were more often infected than herds of older farmers (
      • Valle P.S.
      • Martin S.W.
      • Tremblay R.
      • Bateman K.
      Factors associated with being a bovine-virus diarrhoea (BVD) seropositive dairy herd in the Møre and Romsdal County of Norway.
      ). According to
      • Valle P.S.
      • Martin S.W.
      • Tremblay R.
      • Bateman K.
      Factors associated with being a bovine-virus diarrhoea (BVD) seropositive dairy herd in the Møre and Romsdal County of Norway.
      , this is probably due to different attitudes and management practices of younger farmers, such as not asking for health certificates when purchasing animals. This would be an interesting factor to consider in future quantitative studies about BVDV infection and farmer behavior.
      In our meta-analyses, several pooled estimates were significant. However, the results could be biased because most studies looked at the presence of BVDV and not introduction of the virus. With presence of infection, it is unknown when the actual infection happened, which complicates finding direct associations between infection and risk factors. However, this would probably be less influential when considering risk factors that do not change much over time, such as whether herds graze at pasture, herd type, and herd size. When studying the introduction of BVDV, it is possible that there is a delay between introduction and detection. For example, a PI calf introduced on a farm that monitors by bulk milk testing is unlikely to be promptly detected unless individual animal testing is also conducted on newly imported animals on the farm. Such situations complicate efforts to identify direct associations between infection and risk factors. Therefore, we think that the presence of BVDV is a reasonable proxy for introduction of the virus. In addition, the presence of risk factors does not often change as they are part of regular farm management.
      Another complicating factor in comparing different studies was the way in which herds were categorized as infected or not infected (e.g., based on antibodies or virus) using different sample types, tests, and strategies to confirm the infection status. These differences could be considered by performing a formal assessment of risk of bias. However, because we already had a low number of studies per meta-analysis, we did not want to exclude any more studies and decided to include only the most important internal and external validity checkpoints (Table 1). Also, not all information was available in each publication for a proper bias risk assessment.
      For several risk factors, it was not appropriate to perform a meta-analysis given that there were not enough comparable studies with sufficient quantitative data. For the risk factors with sufficient data, the meta-analyses indicated high levels of heterogeneity. This was expected as all papers included in our meta-analyses were observational studies with different objectives, study designs, and context. For that reason, performing meta-analysis on observational studies and obtaining pooled estimates have been extensively debated (
      • Egger M.
      • Schneider M.
      • Smith G.D.
      Spurious precision? Meta-analysis of observational studies.
      ;
      • Blettner M.
      • Sauerbrei W.
      • Schlehofer B.
      • Scheuchenpflug T.
      • Friedenreich C.
      Traditional reviews, meta-analyses and pooled analyses in epidemiology.
      ;
      • Ioannidis J.P.
      • Patsopoulos N.A.
      • Rothstein H.R.
      Reasons or excuses for avoiding meta-analysis in forest plots.
      ). However, the number of published meta-analyses on observational data has substantially increased, and the need for guidelines for performing meta-analysis on observational data is emphasized (
      • Mueller M.
      • D'Addario M.
      • Egger M.
      • Cevallos M.
      • Dekkers O.
      • Mugglin C.
      • Scott P.
      Methods to systematically review and meta-analyse observational studies: A systematic scoping review of recommendations.
      ;
      • Dekkers O.M.
      • Vandenbroucke J.P.
      • Cevallos M.
      • Renehan A.G.
      • Altman D.G.
      • Egger M.
      COSMOS-E: Guidance on conducting systematic reviews and meta-analyses of observational studies of etiology.
      ). In the current study, we decided to perform meta-analyses on observational studies to provide an overview of available quantitative data, including a weighted average estimate. In this subject area, quantitative risk factor information is available only from observational studies. A key principle underpinning this study is the potential for countries without local knowledge of risk factors for BVDV to learn from those countries where data are available. In our view, weighted average estimates have the potential to be more helpful to readers while being cognizant of heterogeneity between studies rather than being solely a listing of all available quantitative results.
      In our study, we tried to control for heterogeneity and bias as much as possible through the checklist of study appraisal for quantitative analysis (Table 1) and by very carefully choosing the factors that could be compared. The I2 statistics still showed a very high level of heterogeneity for all factors, but it is known to be not very accurate when only a small number of studies (n < 20) are available (
      • Huedo-Medina T.B.
      • Sánchez-Meca J.
      • Marín-Martínez F.
      • Botella J.
      Assessing heterogeneity in meta-analysis: Q statistic or I2 index?.
      ). Also, the very wide 95% CI of the I2 statistic we observed show the degree of uncertainty about the heterogeneity estimations. The influential case analyses showed that the I2 estimate was often lower when removing outliers from the meta-analyses; however, CI remained wide. Given this result, and because I2 is unreliable when few studies are available, we elected to retain the outliers but to show the summary estimates and I2 of each meta-analysis when excluding the outliers (Supplemental File S6, https://doi.org/10.3168/jds.2020-18193).
      To maximize the amount of quantitative data, we decided to include both univariable and multivariable OR in our analyses. Therefore, in 3 of the 6 meta-analyses we combined univariable and multivariable results. The rationale behind this is that in different studies the multivariable OR were adjusted for different factors and referred to different reference situations and are therefore not necessarily more comparable than unadjusted univariable results. On the other hand, univariable OR can under- or overestimate the strength of association. As there is not yet a uniform approach regarding the use of univariable and multivariable results in meta-analysis, adjusted and unadjusted OR often are combined (
      • Liu T.
      • Nie X.
      • Wu Z.
      • Zhang Y.
      • Feng G.
      • Cai S.
      • Lv Y.
      • Peng X.
      Can statistic adjustment of OR minimize the potential confounding bias for meta-analysis of case-control study? A secondary data analysis.
      ). As we decided to combine adjusted and unadjusted OR, we have performed subanalyses in which we compared the results when including only the univariable results or the multivariable results. In most cases we observed only minor differences. In the meta-analyses on herd type and introduction of cattle, we did see a substantial decrease in heterogeneity (I2). However, keeping in mind that the I2 statistic becomes increasingly unreliable when even fewer studies are included and because the summary estimates did not change that much, we decided to combine univariable and multivariable results. The results of the subnalyses are reported in Supplemental File S7 (https://doi.org/10.3168/jds.2020-18193). We also selected different observational study designs to maximize the number of studies in our meta-analyses. Therefore, in 2 of the 6 meta-analyses (participation in cattle shows and markets and contact between cattle at pasture) we combined cross-sectional studies with case-control studies. In the scientific literature, there is disagreement about whether different study designs can be combined (
      • Mueller M.
      • D'Addario M.
      • Egger M.
      • Cevallos M.
      • Dekkers O.
      • Mugglin C.
      • Scott P.
      Methods to systematically review and meta-analyse observational studies: A systematic scoping review of recommendations.
      ). The influential case analysis was conducted to determine whether the case-control studies (only 3 out of 20 studies) were indicated as outliers, which they were not. Consequently, leaving them out would not make much difference, and therefore we decided to retain both study designs. We note that these 2 study designs are differing types of observational studies and use OR as outcome.
      All studies included in our meta-analyses used OR to show the strength of association between risk factors and BVD infection. It should be kept in mind that these OR are based on a certain reference population and are therefore sensitive to how the reference category is defined. For this reason, it can be questioned whether OR are the right means to compare studies. It would have been better to obtain probabilities of infection and risk factor occurrence. However, given that these were often not reported and the fact that OR do provide a rough risk estimate, it was decided to conduct the meta-analysis on OR. This should be considered when interpreting the results of this study.

      CONCLUSIONS

      In this study, we found a wide range of potential risk factors and performed meta-analyses on 6 risk factors for BVDV: herd size, herd type, participation in shows or markets, introduction of cattle, pasture, and contact at pasture. We did not find any unexpected risk factors, and the pooled estimates can help guide advice to farmers and assist in the development, evaluation, and optimization of BVD control programs. The results of the meta-analyses must be interpreted with care due to a high level of study heterogeneity but can assist in the development, evaluation, and optimization of BVD control programs. They can also be used as input for BVDV modeling studies in herds that are comparable with the European cattle production systems. It was challenging to combine estimates of different studies due to heterogeneity between studies (e.g., study design, data analysis, data reporting), showing the need for more standardized methodologies in risk factor studies.

      ACKNOWLEDGMENTS

      This work was carried out with the financial support of the Dutch Ministry of Agriculture, Nature and Food Quality (the Hague, the Netherlands) and is part of the STOC free project that was awarded a grant by the European Food Safety Authority (EFSA, Parma, Italy) and was co-financed by public organizations in the countries participating in the study. The authors have not stated any conflicts of interest.

      Supplementary Material

      REFERENCES

        • AHI (Animal Health Ireland)
        Programme results.
        http://animalhealthireland.ie/?page_id=229
        Date: 2019
        Date accessed: September 17, 2019
        • Amelung S.
        • Hartmann M.
        • Haas L.
        • Kreienbrock L.
        Factors associated with the bovine viral diarrhoea (BVD) status in cattle herds in Northwest Germany.
        Vet. Microbiol. 2018; 216 (29519519): 212-217
        • Azbel-Jackson L.
        • Heffernan C.
        • Gunn G.
        • Brownlie J.
        Exploring the role of voluntary disease schemes on UK farmer bio-security behaviours: Findings from the Norfolk-Suffolk Bovine Viral Diarrhoea control scheme.
        PLoS One. 2018; 13 (29432435)e0179877
        • Barrett D.
        • Parr M.
        • Fagan J.
        • Johnson A.
        • Tratalos J.
        • Lively F.
        • Diskin M.
        • Kenny D.
        Prevalence of bovine viral diarrhoea virus (BVDV), bovine herpes virus 1 (BHV 1), leptospirosis and neosporosis, and associated risk factors in 161 Irish beef herds.
        BMC Vet. Res. 2018; 14 (29304782): 8
        • Bishop H.
        • Erkelens J.
        • Van Winden S.
        Indications of a relationship between buying-in policy and infectious diseases on dairy farms in Wales.
        Vet. Rec. 2010; 167 (21257463): 644-647
        • Bitsch V.
        • Hansen K.E.
        • Rønsholt L.
        Experiences from the Danish programme for eradication of bovine virus diarrhoea (BVD) 1994–1998 with special reference to legislation and causes of infection.
        Vet. Microbiol. 2000; 77 (11042407): 137-143
        • Blettner M.
        • Sauerbrei W.
        • Schlehofer B.
        • Scheuchenpflug T.
        • Friedenreich C.
        Traditional reviews, meta-analyses and pooled analyses in epidemiology.
        Int. J. Epidemiol. 1999; 28 (10195657): 1-9
        • Borenstein M.
        • Higgins J.P.
        • Hedges L.V.
        • Rothstein H.R.
        Basics of meta-analysis: I2 is not an absolute measure of heterogeneity.
        Res. Synth. Methods. 2017; 8: 5-18
        • Byrne A.W.
        • Guelbenzu-Gonzalo M.
        • Strain S.A.J.
        • McBride S.
        • Graham J.
        • Lahuerta-Marin A.
        • Harwood R.
        • Graham D.A.
        • McDowell S.
        Assessment of concurrent infection with bovine viral diarrhoea virus (BVDV) and Mycobacterium bovis: A herd-level risk factor analysis from Northern Ireland.
        Prev. Vet. Med. 2017; 141 (28532992): 38-47
        • Charoenlarp W.
        • Frankena K.
        • Strain S.A.J.
        • Guelbenzu-Gonzalo M.
        • Graham J.
        • Byrne A.W.
        Spatial and risk factor analysis of bovine viral diarrhoea (BVD) virus after the first-year compulsory phase of BVD eradication programme in Northern Ireland.
        Prev. Vet. Med. 2018; 157 (30086847): 34-43
        • Courcoul A.
        • Ezanno P.
        Modelling the spread of bovine viral diarrhoea virus (BVDV) in a managed metapopulation of cattle herds.
        Vet. Microbiol. 2010; 142 (19875250): 119-128
        • Dekkers O.M.
        • Vandenbroucke J.P.
        • Cevallos M.
        • Renehan A.G.
        • Altman D.G.
        • Egger M.
        COSMOS-E: Guidance on conducting systematic reviews and meta-analyses of observational studies of etiology.
        PLoS Med. 2019; 16 (30789892)e1002742
        • Downes M.J.
        • Brennan M.L.
        • Williams H.C.
        • Dean R.S.
        Development of a critical appraisal tool to assess the quality of cross-sectional studies (AXIS).
        BMJ Open. 2016; 6 (27932337)e011458
        • Eaglesome M.D.
        • Garcia M.M.
        Disease risks to animal health from artificial insemination with bovine semen.
        Rev. Sci. Tech. Off. Int. Epiz. 1997; 16 (9329119): 215-225
        • Egger M.
        • Schneider M.
        • Smith G.D.
        Spurious precision? Meta-analysis of observational studies.
        BMJ. 1998; 316 (9462324): 140-144
        • Ersbøll A.K.
        • Ersbøll B.K.
        • Houe H.
        • Alban L.
        • Kjeldsen A.M.
        Spatial modelling of the between-herd infection dynamics of bovine virus diarrhoea virus (BVDV) in dairy herds in Denmark.
        Prev. Vet. Med. 2010; 97 (20850882): 83-89
        • Foddai A.
        • Stockmarr A.
        • Boklund A.
        Evaluation of temporal surveillance system sensitivity and freedom from bovine viral diarrhea in Danish dairy herds using scenario tree modelling.
        BMC Vet. Res. 2016; 12 (27323903): 118
        • Gard J.A.
        • Givens M.D.
        • Stringfellow D.A.
        Bovine viral diarrhea virus (BVDV): Epidemiologic concerns relative to semen and embryos.
        Theriogenology. 2007; 68 (17588651): 434-442
        • Gates M.C.
        • Humphry R.W.
        • Gunn G.J.
        • Woolhouse M.E.
        Not all cows are epidemiologically equal: Quantifying the risks of bovine viral diarrhoea virus (BVDV) transmission through cattle movements.
        Vet. Res. 2014; 45 (25323831): 110
        • Gates M.C.
        • Woolhouse M.E.J.
        • Gunn G.J.
        • Humphry R.W.
        Relative associations of cattle movements, local spread, and biosecurity with bovine viral diarrhoea virus (BVDV) seropositivity in beef and dairy herds.
        Prev. Vet. Med. 2013; 112 (24012354): 285-295
        • Graham D.A.
        • Clegg T.A.
        • Lynch M.
        • More S.J.
        Herd-level factors associated with the presence of bovine viral diarrhoea virus in herds participating in the voluntary phase of the Irish national eradication programme.
        Prev. Vet. Med. 2013; 112 (23932098): 99-108
        • Graham D.A.
        • Clegg T.A.
        • Thulke H.H.
        • O'sullivan P.
        • McGrath G.
        • More S.J.
        Quantifying the risk of spread of bovine viral diarrhoea virus (BVDV) between contiguous herds in Ireland.
        Prev. Vet. Med. 2016; 126 (26850846): 30-38
        • Hanon J.B.
        • De Baere M.
        • de la Ferté C.
        • Roelandt S.
        • Guillot G.
        • Van der Stede Y.
        • Cay B.
        Serological monitoring on milk and serum samples in a BVD eradication program: A field study in Belgium showing antibody ELISA performances and epidemiological aspects.
        Prev. Vet. Med. 2018; 160 (30054018): 136-144
        • Harrer M.
        • Cuijpers P.
        • Furukawa T.A.
        • Ebert D.D.
        Doing Meta-Analysis in R: A Hands-on Guide.
        https://bookdown.org/MathiasHarrer/Doing_Meta_Analysis_in_R/
        Date: 2019
        Date accessed: November 20, 2019
        • Heffernan C.
        • Azbel-Jackson L.
        • Brownlie J.
        • Gunn G.
        Farmer attitudes and livestock disease: Exploring citizenship behaviour and peer monitoring across two BVD control schemes in the UK.
        PLoS One. 2016; 11 (27023269)e0152295
        • Higgins J.P.T.
        • Thomas J.
        • Chandler J.
        • Cumpston M.
        • Li T.
        • Page M.J.
        • Welch V.A.
        Cochrane Handbook for Systematic Reviews of Interventions. Version 6.0.
        www.training.cochrane.org/handbook
        Date: 2019
        Date accessed: July 7, 2019
        • Houe H.
        Economic impact of BVDV infection in dairies.
        Biologicals. 2003; 31 (12770546): 137-143
        • Houe H.
        • Baker J.C.
        • Maes R.K.
        • Lloyd J.W.
        • Enevoldsen C.
        Comparison of the prevalence and incidence of infection with bovine virus diarrhoea virus (BVDV) in Denmark and Michigan and association with possible risk factors.
        Acta Vet. Scand. 1995; 36 (8669379): 521-531
        • Huedo-Medina T.B.
        • Sánchez-Meca J.
        • Marín-Martínez F.
        • Botella J.
        Assessing heterogeneity in meta-analysis: Q statistic or I2 index?.
        Psychol. Methods. 2006; 11 (16784338): 193-206
        • Hult L.
        • Lindberg A.
        Experiences from BVDV control in Sweden.
        Prev. Vet. Med. 2005; 72 (16213610): 143-148
        • Ioannidis J.P.
        • Patsopoulos N.A.
        • Rothstein H.R.
        Reasons or excuses for avoiding meta-analysis in forest plots.
        BMJ. 2008; 336 (18566080): 1413-1415
        • Lindberg A.
        • Brownlie J.
        • Gunn G.
        • Houe H.
        • Moening V.
        • Saatkamp H.W.
        • Sandvik T.
        • Valle P.S.
        The control of bovine viral diarrhoea virus in Europe: Today and in the future.
        Rev. Sci. Tech. Off. Int. Epiz. 2006; 25 (17361763): 961-979
        • Lindberg A.
        • Houe H.
        Characteristics in the epidemiology of bovine viral diarrhea virus (BVDV) of relevance to control.
        Prev. Vet. Med. 2005; 72 (16214248): 55-73
        • Liu T.
        • Nie X.
        • Wu Z.
        • Zhang Y.
        • Feng G.
        • Cai S.
        • Lv Y.
        • Peng X.
        Can statistic adjustment of OR minimize the potential confounding bias for meta-analysis of case-control study? A secondary data analysis.
        BMC Med. Res. Methodol. 2017; 17 (29284414): 179
        • Mainar-Jaime R.C.
        • Berzal-Herranz B.
        • Arias P.
        • Rojo-Vázquez F.A.
        Epidemiological pattern and risk factors associated with bovine viral-diarrhoea virus (BVDV) infection in a non-vaccinated dairy-cattle population from the Asturias region of Spain.
        Prev. Vet. Med. 2001; 52 (11566379): 63-73
        • Martinez-Ibeas A.M.
        • Power C.
        • McClure J.
        • Sayers R.G.
        Prevalence of BoHV-1 seropositive and BVD virus positive bulls on Irish dairy farms and associations between bull purchase and herd status.
        Ir. Vet. J. 2015; 68 (26664696): 28
        • McClurkin A.W.
        • Littledike E.T.
        • Cutlip R.C.
        • Frank G.H.
        • Coria M.F.
        • Bolin S.R.
        Production of cattle immunotolerant to bovine viral diarrhea virus.
        Can. J. Comp. Med. 1984; 48 (6326980): 156
        • Moennig V.
        • Houe H.
        • Lindberg A.
        BVD control in Europe: Current status and perspectives.
        Anim. Health Res. Rev. 2005; 6 (16164009): 63-74
        • Moher D.
        • Liberati A.
        • Tetzlaff J.
        • Altman D.G.
        Preferred Reporting Items for Systematic Reviews and Meta-Analyses: The PRISMA statement.
        PLoS Med. 2009; 6 (19621072)e1000097
        • Mueller M.
        • D'Addario M.
        • Egger M.
        • Cevallos M.
        • Dekkers O.
        • Mugglin C.
        • Scott P.
        Methods to systematically review and meta-analyse observational studies: A systematic scoping review of recommendations.
        BMC Med. Res. Methodol. 2018; 18 (29783954): 44
        • National Institutes of Health
        Study quality assessment tools.
        • Niskanen R.
        • Lindberg A.
        Transmission of bovine viral diarrhoea virus by unhygienic vaccination procedures, ambient air, and from contaminated pens.
        Vet. J. 2003; 165 (12573600): 125-130
        • Norström M.
        • Jonsson M.E.
        • Åkerstedt J.
        • Whist A.C.
        • Kristoffersen A.B.
        • Sviland S.
        • Hopp P.
        • Wahlström H.
        Estimation of the probability of freedom from bovine virus diarrhoea virus in Norway using scenario tree modelling.
        Prev. Vet. Med. 2014; 116 (25037848): 37-46
        • Nuotio L.
        • Juvonen M.
        • Neuvonen E.
        • Sihvonen L.
        • Husu-Kallio J.
        Prevalence and geographic distribution of bovine viral diarrhoea (BVD) infection in Finland 1993–1997.
        Vet. Microbiol. 1999; 64 (10028175): 231-235
        • Olafson P.
        • Rickard C.G.
        Further observations on the virus diarrhea (new transmissible disease) of cattle.
        Cornell Vet. 1947; 37 (20243965): 104-106
        • Pinior B.
        • Garcia S.
        • Minviel J.J.
        • Raboisson D.
        Epidemiological factors and mitigation measures influencing production losses in cattle due to bovine viral diarrhoea virus infection: A meta-analysis.
        Transbound. Emerg. Dis. 2019; 66 (31328411): 2426-2439
        • Presi P.
        • Struchen R.
        • Knight-Jones T.
        • Scholl S.
        • Heim D.
        Bovine viral diarrhea (BVD) eradication in Switzerland—Experiences of the first two years.
        Prev. Vet. Med. 2011; 99 (21371766): 112-121
        • Qi L.
        • Beaunée G.
        • Arnoux S.
        • Dutta B.L.
        • Joly A.
        • Vergu E.
        • Ezanno P.
        Neighbourhood contacts and trade movements drive the regional spread of bovine viral diarrhoea virus (BVDV).
        Vet. Res. 2019; 50 (31036076): 30
        • R Core Team
        R: A Language and Environment for Statistical Computing..
        R Foundation for Statistical Computing, Vienna, Austria2019
        • Rikula U.
        • Nuotio L.
        • Aaltonen T.
        • Ruoho O.
        Bovine viral diarrhoea virus control in Finland 1998–2004.
        Prev. Vet. Med. 2005; 72 (16182393): 139-142
        • Rikula U.
        • Nuotio L.
        • Laamanen U.I.
        • Sihvonen L.
        Transmission of bovine viral diarrhoea virus through the semen of acutely infected bulls under field conditions.
        Vet. Rec. 2008; 162 (18204031): 79-82
        • Rossmanith W.
        • Deinhofer M.
        • Janacek R.
        • Trampler R.
        • Wilhelm E.
        Voluntary and compulsory eradication of bovine viral diarrhoea virus in Lower Austria.
        Vet. Microbiol. 2010; 142 (19931989): 143-149
        • Rossmanith W.
        • Janacek R.
        • Wilhelm E.
        Control of BVDV-infection on common grassland—The key for successful BVDV-eradication in Lower Austria.
        Prev. Vet. Med. 2005; 72 (16213041): 133-137
        • Sanderson S.
        • Tatt I.D.
        • Higgins J.
        Tools for assessing quality and susceptibility to bias in observational studies in epidemiology: A systematic review and annotated bibliography.
        Int. J. Epidemiol. 2007; 36: 666-676
        • Sarrazin S.
        • Veldhuis A.
        • Méroc E.
        • Vangeel I.
        • Laureyns J.
        • Dewulf J.
        • Caij A.B.
        • Piepers S.
        • Hooyberghs J.
        • Ribbens S.
        • Van Der Stede Y.
        Serological and virological BVDV prevalence and risk factor analysis for herds to be BVDV seropositive in Belgian cattle herds.
        Prev. Vet. Med. 2013; 108 (22878124): 28-37
        • Scharnböck B.
        • Roch F.-F.
        • Richter V.
        • Funke C.
        • Firth C.L.
        • Obritzhauser W.
        • Baumgartner W.
        • Käsbohrer A.
        • Pinior B.
        A meta-analysis of bovine viral diarrhoea virus (BVDV) prevalences in the global cattle population.
        Sci. Rep. 2018; 8 (30258185)14420
        • Tratalos J.A.
        • Graham D.A.
        • More S.J.
        Patterns of calving and young stock movement in Ireland and their implications for BVD serosurveillance.
        Prev. Vet. Med. 2017; 142 (28606363): 30-38
        • Tråvén M.
        • Alenius S.
        • Fossum C.
        • Larsson B.
        Primary bovine viral diarrhoea virus infection in calves following direct contact with a persistently viraemic calf.
        J. Vet. Med. Series B. 1991; 38 (1719713): 453-462
        • Valle P.S.
        • Martin S.W.
        • Tremblay R.
        • Bateman K.
        Factors associated with being a bovine-virus diarrhoea (BVD) seropositive dairy herd in the Møre and Romsdal County of Norway.
        Prev. Vet. Med. 1999; 40 (10423772): 165-177
        • Viechtbauer W.
        Conducting meta-analyses in R with the metafor package.
        J. Stat. Softw. 2010; 36: 1-48
        • Viechtbauer W.
        • Cheung M.W.L.
        Outlier and influence diagnostics for meta-analysis.
        Res. Synth. Methods. 2010; 1 (26061377): 112-125
        • Wentink G.H.
        • Frankena K.
        • Bosch J.C.
        • Vandehoek J.E.D.
        • Van Den Berg T.
        Prevention of disease transmission by semen in cattle.
        Livest. Prod. Sci. 2000; 62: 207-220
        • Williams D.
        • Winden S.V.
        Risk factors associated with high bulk milk antibody levels to common pathogens in UK dairies.
        Vet. Rec. 2014; 174 (24789852): 580