From tests to truth: A misclassification-aware machine learning framework for estimating brucellosis seroprevalence in wild canids

Nahal Sarvestani; Farzane Shams; Armin Mirshahi; Mobina Pato; Aria Javani Farbod; Armina Khayatderafshi; Mobina Payami; Arman Abdous

doi:10.1371/journal.pntd.0014029

Abstract

Brucellosis is an important zoonotic disease affecting humans, livestock, and wildlife, yet prevalence estimates in wild species are often underestimated due to limited attention to surveillance, as well as insufficient and biased sampling. To clarify exposure patterns in wild canids, we searched PubMed, Scopus, Web of Science, SciELO, and Google Scholar (1962–2025) for primary prevalence studies of Brucella species (spp.) in free-ranging canids. Serological data were analyzed using a misclassification-aware, multi-assay model that corrects for imperfect test sensitivity and specificity. Confirmatory polymerase chain reaction (PCR) and culture results were considered separately from serological data, as each offers a different perspective on disease status. Across 48 wild serology populations (n = 3,925 animals), the global misclassification-adjusted true seroprevalence was 8.2% (95% confidence interval [CI]: 5.1–11.3%). Confirmed active infection, based on PCR or culture, was uncommon (3.9%; 95% CI: 3.0–5.1%). Exposure levels varied across continents, with higher estimates in South America (approximately 18%) and lower levels in Europe (approximately 0.8%) and North America (approximately 4.1%). Data from Africa were limited, and Asian estimates were based on sparse wild samples, leading to wide uncertainty. Seroprevalence was consistently influenced by assay type, host species, and region. Overall, wild canids exhibit modest but widespread serological exposure to Brucella spp., whereas confirmed active infection remains rare. Because evidence quality and diagnostic rigor varied considerably across regions, especially in Africa and parts of Asia, results from these areas should be interpreted with caution. By correcting for imperfect tests and explicitly accounting for study heterogeneity, this framework provides more reliable and transparent prevalence estimates and highlights geographic gaps where improved, targeted One Health surveillance is most urgently needed.

Author summary

Brucellosis is a bacterial disease that affects livestock, wildlife, and people. While surveillance usually focuses on domestic animals, far less is known about infection in wild carnivores such as wolves, foxes, and jackals, even though these species often live near farms and may signal when Brucella is present in the environment. Research on wild canids is difficult to compare because studies use many different diagnostic tests, some of which can give false-positive or false-negative results. We reviewed all available studies of Brucella in wild canids worldwide and combined the data using a modeling approach that adjusts for imperfect diagnostic accuracy. We found that exposure is widespread but unevenly documented, with major gaps in Africa and much of Asia. After accounting for test performance, wild canids showed modest levels of exposure overall, and confirmed infections were rare. Studies using older or less reliable tests tended to report inflated prevalence. By correcting for diagnostic limitations, our study provides more reliable estimates and highlights where improved wildlife surveillance is most needed to support One Health monitoring.

Citation: Sarvestani N, Shams F, Mirshahi A, Pato M, Farbod AJ, Khayatderafshi A, et al. (2026) From tests to truth: A misclassification-aware machine learning framework for estimating brucellosis seroprevalence in wild canids. PLoS Negl Trop Dis 20(3): e0014029. https://doi.org/10.1371/journal.pntd.0014029

Editor: Richard A. Bowen, Colorado State University, UNITED STATES OF AMERICA

Received: November 5, 2025; Accepted: February 11, 2026; Published: March 6, 2026

Copyright: © 2026 Sarvestani et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Data Availability: All analyses were performed using Python. The complete analysis scripts and associated datasets are publicly available at the ShamsFarzane GitHub repository: https://github.com/ShamsFarzane/brucellosis-canids-seroprevalence.git.

Funding: The author(s) received no specific funding for this work.

Competing interests: The authors have declared that no competing interests exist.

Introduction

Brucellosis is a globally distributed zoonotic disease of major veterinary and public health concern, caused by Gram-negative bacteria of the genus Brucella [1]. These pathogens infect a wide range of mammals, including livestock, humans, and wildlife. While most control and surveillance programs have primarily focused on livestock and human infections, wildlife, especially free-ranging canids such as wolves (Canis lupus), foxes (Vulpes spp., Lycalopex spp.), jackals (Canis aureus), and coyotes (Canis latrans), may also contribute to the ecology and transmission of Brucella species [2]. As predators and scavengers, wild canids may act as incidental hosts, spillover recipients, or mechanical vectors, and in certain contexts, could act as local reservoirs. These dynamics are most relevant in ecotones and rural or peri-urban landscapes, where land-use change increases contact among wildlife, livestock, and humans.

Despite the zoonotic importance of Brucella, surveillance in wild canids remains uneven. Most published studies originate from the Americas, while large data gaps persist in Africa, Asia, and Eastern Europe [3]. Much of the existing field research relies on serological tests such as the Standard Agglutination Test (SAT), Rose Bengal Test (RBT), and enzyme-linked immunosorbent assay (ELISA) because they are affordable and practical, yet these methods primarily reflect past exposure rather than active infection [4]. In contrast, bacterial culture and polymerase chain reaction (PCR) provide stronger evidence of active infection but are costly and logistically challenging, which has limited their routine use in wildlife studies [5]. Distinguishing clearly between seroprevalence (exposure) and PCR/culture-confirmed infection is therefore essential for understanding transmission potential and designing targeted surveillance systems.

Previous reviews of wildlife brucellosis by Dadar et al. and Kosoy and Goodrich have synthesized data across species and diagnostic methods, but they often lacked species-specific analyses, formal quality assessment, and statistical correction for imperfect diagnostic performance [3,6]. These limitations, combined with sparse and geographically uneven datasets, make it difficult to draw reliable conclusions about exposure patterns in wild canids.

To address these gaps, we focused specifically on the family Canidae (globally distributed, mobile species that commonly interact with livestock) and applied a reproducible, quantitative framework that builds on established misclassification and latent-class approaches. We extended these approaches to handle challenges such as multi-assay testing within the same sampled populations, heterogeneous study designs, and variable diagnostic quality. Using a PRISMA-ScR-guided scoping review and a structured quality assessment adapted from the Newcastle–Ottawa Scale, we first characterized the global evidence base and identified taxonomic and geographic gaps. Second, we implemented a misclassification-aware model that integrates results from multiple serological assays to estimate true population-level seroprevalence, rather than treating each assay arm independently, and we analyzed PCR/culture data separately. Third, we applied an exploratory machine learning meta-regression to identify consistent predictors of seroprevalence such as assay type, species, and region to support practical surveillance design tasks, including sample size estimation under imperfect testing conditions. Together, these components synthesize existing methodological principles into an integrated, wildlife-focused framework that produces uncertainty-aware, species- and region-specific prevalence estimates to inform One Health surveillance and control strategies at the wildlife–livestock–human interface.

Methods

Bibliographic search strategy

We conducted a global scoping review and quantitative synthesis of Brucella exposure and infection in wild canids (family Canidae), following the PRISMA Extension for Scoping Reviews (PRISMA-ScR) guidelines [7]. The study protocol defined eligibility, screening, data extraction, and analysis steps a priori and was piloted before full implementation. A PRISMA flow diagram summarizes record identification, screening, and full-text assessment (Fig 1), with detailed reasons for full-text exclusion provided in S1 Table. The completed PRISMA-ScR checklist is provided in S1 File.

Download:

Fig 1. PRISMA-ScR flow diagram of study selection for the global scoping review of Brucella exposure and infection in wild canids (1962–2025).

https://doi.org/10.1371/journal.pntd.0014029.g001

A comprehensive literature search was conducted in July 2025 across PubMed, Scopus, Web of Science Core Collection, SciELO, and Google Scholar. The search strategy combined controlled vocabulary, where available, and free-text keywords related to Brucella infection, wild canids, epidemiology, and methods. Based on pilot testing, the first 200 Google Scholar records were screened for relevance. Detailed search strings, filters, and database-specific syntax are provided in S2 File. As a representative example, the PubMed search strategy was: (“Brucellosis”[Mesh] OR “Brucella abortus”[Mesh] OR “Brucella suis”[Mesh] OR “Brucella canis”[Mesh] OR brucellosis OR Brucella) AND (“Canidae”[Mesh] OR canids OR “wild canids” OR foxes OR wolves OR jackals OR coyotes OR “wild dogs”) AND (prevalence OR seroprevalence OR serology OR ELISA OR PCR OR culture OR diagnostics). All retrieved records were de-duplicated in a unified library before screening.

Two reviewers independently screened titles and abstracts, followed by full-text assessment using prespecified eligibility criteria. Disagreements were resolved by consensus, and inter-rater agreement was assessed during a calibration exercise on a random subset of studies using Cohen’s kappa (κ).

Data extraction and synthesis

A piloted extraction form was used to record metadata from each included study, including author, sampling year(s), country and continent, host species (common and scientific names), diagnostic methods, life condition (wild or captive), sample size (N), number of positives (n), and the Brucella species reported, when available. Species names were standardized according to current Canidae taxonomy, and country names were harmonized using International Organization for Standardization (ISO) formats to ensure consistency across studies.

Each study was assigned a unique Study ID composed of the first author’s surname and publication year, for example Davis_1979. Records sharing the same Study ID and sampling year(s) were treated as the same study, and duplicate records retrieved from multiple databases were merged. When a single publication reported results for multiple species, diagnostic methods, or sampling periods, each unique combination was treated as a distinct study arm.

Study arms were collapsed only when they clearly referred to the same animals, indicated by identical species, country, sampling period, and total sample size (N). This occurred primarily when multiple serological assays such as RBT, SAT, or ELISA were applied to the same individuals. In such cases, the number of positives and total sample size were summed and treated as a multi-test population. Study arms differing in species, country, sampling year(s), or N, or where overlap of individuals was uncertain, were retained as separate populations.

A strict separation was maintained between diagnostic tracks. Serological assays were used to assess exposure, whereas PCR and bacterial culture were used to identify active infection. These diagnostics were applied to different individuals in nearly all included studies and were therefore never combined. Each population was assigned exclusively to either the serology track or the PCR and culture track. Study-level prevalence data derived from this process are summarized in Table 1.

Download:

Table 1. Study-level brucellosis prevalence in wild canids worldwide (1962–2025).

https://doi.org/10.1371/journal.pntd.0014029.t001

Inclusion and exclusion criteria

Records were eligible if they reported primary prevalence data for wild-caught canids, including wolves (Canis lupus), foxes (Vulpes spp., Lycalopex spp.), jackals (Canis aureus), coyotes (Canis latrans), African wild dogs (Lycaon pictus), and used validated diagnostic methods. No restrictions on publication year or language were applied during database searches, and the final body of eligible evidence spanned publications from 1962 to 2025. All retrieved records were screened regardless of date, language, or region of origin.

Studies were eligible if they reported primary prevalence data for wild-caught canids and provided sufficient methodological detail to confirm host identity and diagnostic procedures.Diagnostics were categorized a priori into two groups:

Serology (exposure): Standard Agglutination Test (SAT), Rose Bengal Test (RBT), Complement Fixation Test (CFT), agar gel immunodiffusion (AGID), enzyme-linked immunosorbent assay (ELISA), Fluorescence Polarization Assay (FPA), or equivalent antibody-based assays.
Active infection (confirmation): Polymerase chain reaction (PCR), including quantitative PCR, and/or bacterial culture.

Studies involving zoo or laboratory animals were excluded, as were reports lacking primary numerators and denominators or sufficient information to verify host species identity or diagnostic methodology.

Quality appraisal (QA)

Study quality was evaluated using an adapted Newcastle–Ottawa Scale for prevalence studies [45].

Because the original NOS was developed for cohort and case–control designs, the scale was adapted to address sources of bias relevant to wildlife serosurveys and molecular diagnostic studies. The adapted NOS included seven binary items grouped into three domains: selection (S1–S3), comparability/diagnostic quality (C1–C2), and outcome reporting (O1–O2).

Each item was scored as 1 (criterion met) or 0 (criterion not met), giving a total score ranging from 0 to 7, consistent with standard NOS-based scoring approaches for prevalence studies. The summed score for each study is reported in the final “QA” column, where higher values indicate lower risk of bias. Based on total scores, studies were categorized as having low (5–7 points), moderate (3–4 points), or high (0–2 points) risk of bias, reflecting increasing concern about internal validity.

Study quality was not used as an inclusion criterion but was used to characterize the overall risk-of-bias profile of the evidence base and to contextualize regional and diagnostic-specific findings, and it was explicitly examined in sensitivity analyses. Detailed item-level scores for each NOS domain are provided in S2 Table.

Outcomes and unit of analysis

Two outcomes were analyzed on separate diagnostic tracks:

Exposure (serology): misclassification-adjusted true seroprevalence.
Active infection: prevalence confirmed by PCR and/or culture (descriptive only, as confirmatory data were too limited for integration into the Bayesian modeling framework).

The unit of analysis was a “population” defined as the unique combination of Study ID (first author and publication year) × species × country × year × total sample size (N). Rows sharing identical parameters across all of these dimensions and N were treated as the same animals tested by different assays (multi-test populations). Rows differing in N, year, or species were treated as distinct populations. PCR/culture data were never combined with serology.

Analyses were performed for (i) wild-only datasets (primary) and (ii) wild plus farm/captive datasets (sensitivity analysis). “Farm/captive” status was based on the original study context.

Serology track: misclassification-adjusted true seroprevalence

For each population j tested with one or more serological assays t, we estimated a single true seroprevalence (θⱼ). Observed positives (yⱼt) were modeled as:

y_jt ~ Binomial(N_j, q_jt), where q_jt = Se_t·θ_j + (1−Sp_t)(1−θ_j).

where Se_t and Sp_t are assay-specific sensitivity and specificity [46]. When studies reported serial or parallel testing algorithms (for example “RBT screen followed by CFT confirm”), we used the implied combined sensitivity/specificity (serial: Se = Se₁Se₂, Sp = 1 − (1 − Sp₁)(1 − Sp₂); parallel: Se = 1 − (1 − Se₁)(1 − Se₂), Sp = Sp₁Sp₂). Otherwise, each assay contributed its own likelihood term.

True seroprevalence (θⱼ) was estimated using grid-based maximum likelihood estimation (MLE) with 95% profile-likelihood confidence intervals. Continent-level summaries were calculated as N-weighted means of θ̂ⱼ with conservative weighted-variance approximations. Estimates are presented separately for wild-only and sensitivity datasets [47].

Infection track: PCR/culture prevalence

Because confirmatory data were sparse, PCR and culture prevalence were summarized descriptively as pooled proportions (total positives/ total tested) within species, country, and continent strata, with 95% Wilson confidence intervals. Zero-event studies were retained.

Machine learning Bayesian meta-analysis (serology)

To account for imperfect test performance and heterogeneity, we applied a hierarchical Bayesian measurement-error model. Observed positives per arm were modeled as binomial outcomes with apparent probabilities derived from assay Se/Sp applied to an underlying latent true prevalence. True prevalence was modeled on the logit scale with random effects for study and continent; species effects were added in stratified models.

When available, study-reported Se/Sp were used directly; otherwise, method-specific priors centered on literature-typical performance were applied. A beta–binomial extension addressed over-dispersion where necessary. Models were fitted using Hamiltonian Monte Carlo (No-U-Turn Sampler [NUTS]) with four chains, adaptive warm-up, and conservative target acceptance. Convergence was verified via Gelman–Rubin statistics (R̂ ≈ 1.00), effective sample sizes, and visual inspection of trace, rank, autocorrelation, and energy/ Bayesian fraction of missing information (BFMI) plots [47].

Model performance was assessed through posterior predictive checks and study-level calibration plots. Comparisons were made using Pareto-smoothed importance sampling leave-one-out cross-validation (PSIS-LOO) and widely applicable information criterion (WAIC); study influence was summarized using Pareto-k values. To enhance robustness, two safeguards were applied: (i) a Trim-K influence-weighted likelihood, and (ii) heavy-tailed priors on variance components. The combined robust+Trim-K model was reported alongside the baseline.

As a transparency check, we also generated Monte Carlo Rogan–Gladen estimates of true prevalence that propagate uncertainty in assay performance; these were consistent with Bayesian posteriors.

Software and reproducibility

All analyses were scripted in Python and world heat map figure was constructed using R [48] Reproducible code and documentation are available via the public ShamsFarzane GitHub repository (https://github.com/ShamsFarzane/brucellosis-canids-seroprevalence.git).

Results

Search results

The database search yielded 4,120 records, including PubMed (n = 1,020), Scopus (n = 985), Web of Science (n = 910), Google Scholar (n = 720), and SciELO (n = 485). An additional 87 records were identified through reference list screening. After removal of 307 duplicates, 3,900 records remained for title and abstract screening. Of these, 3,810 were excluded because they did not report primary prevalence data, focused on domestic or captive animals, or involved non-canid host species. Ninety full-text articles were assessed for eligibility. Fifty-three articles were excluded for lack of original prevalence data (n = 18), focus on domestic or zoo animals (n = 13), unclear or invalid diagnostic methods (n = 9), unrepresentative or insufficient sample sizes (n = 8), or incomplete methodological reporting (n = 5). A detailed exclusion log is provided in S1 Table. A total of 37 studies met the inclusion criteria and were incorporated into the scoping review and quantitative synthesis. The study selection process is summarized in Fig 1. Inter-rater agreement during the calibration exercise was high (κ = 0.86).

Characteristics of included studies

The evidence base comprised 37 studies conducted between 1962 and 2025, representing 48 wild canid populations from 16 countries across five continents. Geographic representation was uneven, with most data originating from North America, South America, and Europe, and more limited sampling in Asia and Africa. Sample sizes ranged from 10 to 997 individuals. Twelve wild canid species were represented, with the most frequently sampled taxa being Canis lupus, Vulpes vulpes, Canis latrans, Cerdocyon thous, Vulpes lagopus, and Lycalopex culpaeus.

Serological diagnostics included the Rose Bengal Test (RBT), Standard Agglutination Test (SAT), Buffered Plate Agglutination Test (BPAT), Card Agglutination Test (CARD), Complement Fixation Test (CFT), enzyme-linked immunosorbent assay (ELISA), agar gel immunodiffusion (AGID), counterimmunoelectrophoresis (CIE), fluorescence polarization assay (FPA), and immunochromatographic test (ICT). PCR and culture data were available from a smaller subset of studies. Study quality, assessed using the adapted Newcastle–Ottawa Scale, ranged from 1 to 7.

Risk of bias assessment

Using predefined thresholds based on total Newcastle–Ottawa Scale scores, 21 studies were classified as low risk of bias, 12 as moderate risk, and 4 as high risk. Risk-of-bias patterns varied across regions, diagnostic approaches, and reporting practices. Studies classified as moderate or high risk most frequently lacked explicit inclusion or exclusion criteria, sample-size justification, or detailed reporting of diagnostic cut-off values. High-risk studies commonly relied exclusively on older agglutination-based assays without confirmatory testing.

Geographically, lower-risk studies were more common in South America and North America, whereas studies from Africa and parts of Asia more often exhibited methodological limitations. Diagnostic modality was also associated with study quality, with investigations using ELISA, PCR, or CFT generally receiving higher appraisal scores than studies based solely on legacy agglutination tests. Quality scores for individual studies are reported in Tables 1 and S2. Full item-level scoring and a graphical summary of risk-of-bias domains are provided in S1 Fig.

Global exposure is widespread, whereas confirmed active infection is uncommon

Using a misclassification-aware model that integrates all serological assays performed on the same animals, we estimated the global true seroprevalence (θ) of Brucella exposure in wild canids. Across 48 wild populations (N = 3,925 animals), the N-weighted global true seroprevalence was 8.2% (95% CI: 5.1–11.3%). Estimates varied widely across populations and continents, reflecting substantial heterogeneity in diagnostic methods, sampling frames, and time periods (S3 Table).

Confirmed active infection was rare. Pooled PCR and culture data across studies yielded 3.9% (95% Wilson CI: 3.0–5.1%; 52/1,325 positives). Because infection data were sparse and unevenly distributed (e.g., culture positives from Russia for B. suis biovar 4), these results are reported separately from serology.

When captive or farmed canids were included as a sensitivity analysis, the overall misclassification-adjusted seroprevalence increased to 15.5% (95% CI: 9.4–21.6%; 51 populations; N = 5,079), with high prevalence observed in blue-fox breeding farms. The infection proportion remained 3.9% since confirmatory tests were limited to wild settings.

Hierarchical Bayesian models correcting for imperfect diagnostic performance produced higher estimates. The baseline model estimated 23.1% (95% CrI: 19.4–27.1%); with influence trimming (Trim-K = 5), 15.8% (12.2–19.5%); with robust priors, 27.3% (22.6–32.5%); and with the combined robust+Trim-K specification, 21.7% (17.5–27.2%). Across model specifications, the estimated global true seroprevalence range was 16–27%, centered around 21–23%.

Posterior predictive checks and calibration plots demonstrated excellent model fit and convergence (no divergences; R̂ ≈ 1.00; large effective sample sizes). Influence diagnostics confirmed that no single study dominated inference (Fig 2).

Download:

Fig 2. Global true seroprevalence and model calibration.

(a) Posterior distribution of the misclassification-adjusted hierarchical Bayesian model for global Brucella seroprevalence in wild canids; the vertical line marks the posterior median. (b) Study-level calibration (observed vs. predicted) aligns with the 1:1 line, indicating good model fit. Chains converged (R̂ ≈ 1.00; no divergences), and posterior predictive checks reproduced the empirical distribution. (c) Sensitivity of pooled true seroprevalence to modeling choices (baseline, Trim-K, robust priors, robust + Trim-K). Influence trimming lowers estimates, while robust priors raise them; the combined model yields a stable mid-range (~22%). (d) Forest plot showing observed (×) versus misclassification-adjusted (●) prevalence with 95% confidence intervals (CIs) by diagnostic test. Marked heterogeneity is evident, with older, lower-specificity assays producing higher apparent prevalence, while PCR/culture arms remain low.

https://doi.org/10.1371/journal.pntd.0014029.g002

Asia: sparse wild data, localized infection

After excluding captive data, Asian wild canid data were extremely limited. Only one wild population (N = 56 golden jackals, Israel) was identified with 0/56 seropositive results. The corresponding misclassification-adjusted estimate was near zero, with a binomial 95% upper bound of ≈5.2%.

In contrast, culture-confirmed infection was documented in Russia, with pooled data showing 48/1,031 = 4.66% (95% Wilson CI: 3.53–6.12%) across wolves and Arctic foxes. Including captive populations (notably blue-fox farms in China) increased seroprevalence substantially but did not represent free-ranging wildlife.

The Bayesian model estimated a true seroprevalence of 4.6% (95% CrI: 4.0–5.5%). Diagnostic plots (energy/BFMI ≈ 0.92–0.96) and posterior predictive checks indicated good convergence. Estimates are shown in Fig 3.

Download:

Fig 3. True seroprevalence of brucellosis in Asian wild canids.

(a) Posterior distribution from the misclassification-adjusted hierarchical Bayesian model; the dashed line marks the posterior median (≈4.6%, 95% CrI 4.0–5.5%). (b) Posterior predictive check comparing replicated (blue) and observed (black) prevalence distributions. The model reproduces the central mass of the empirical data but slightly under-captures the tails, indicating an overall adequate though imperfect fit. Sparse data—driven mainly by one golden-jackal series—limit precision and highlight the need for additional wild sampling in Asia.

https://doi.org/10.1371/journal.pntd.0014029.g003

Europe: lowest exposure and limited infection evidence

Europe showed the lowest exposure levels. Using the misclassification-aware model, the N-weighted true seroprevalence was ~ 0.8% (95% CI: 0.0–4.5%; 4 populations; N = 437).

PCR/culture data were scarce but present, yielding 4/254 = 1.6% (95% Wilson CI: 0.61–3.98%). The Bayesian model estimated a true seroprevalence of 1.6% (95% CrI: 1.3–1.9%) with strong convergence (R̂ ≈ 1.00; energy/BFMI ≈ 0.86–0.92) and accurate posterior predictive fit (Fig 4). Positive detections were reported among wolves, Arctic foxes, and jackals, and were also observed in a subset of red fox studies.

Download:

Fig 4. Very low pooled prevalence with localized outliers in Europe.

(a) Posterior of true seroprevalence in Europe, with a median of ≈1.6% (95% CrI 1.3–1.9%). (b) Observed (×) versus misclassification-adjusted (●) prevalence by study. Most species are near zero; a few red-fox datasets drive the right-tail, reflecting local outliers rather than widespread exposure.

https://doi.org/10.1371/journal.pntd.0014029.g004

North America: low-to-moderate exposure, scarce confirmation

Across 22 North American wild populations (N = 2,066), the misclassification-aware model estimated ~4.1% true seroprevalence (95% CI: 0.5–7.8%). Exposure levels varied among species but remained generally low to moderate.

No PCR or culture-confirmed infections were found in the wild-only dataset. Bayesian models yielded pooled true seroprevalence estimates around 10%, depending on model specification, with satisfactory calibration and posterior predictive performance (Fig 5). These estimates indicate modest levels of serological exposure in North American wild canids.

Download:

Fig 5. Pooled true seroprevalence and study-level estimates in North America.

(a) Posterior distribution from the misclassification-adjusted hierarchical model; median true seroprevalence ≈10.0% (95% CrI 7.0–13.0%). The unimodal and moderately narrow posterior indicates intermediate exposure relative to other continents and good model pooling across studies. (b) Study-level apparent (×) and adjusted (●) prevalence with 95% CIs, labeled by diagnostic test. Coyote and red-fox series contribute most of the signal; older SAT/RBT assays tend to report higher apparent values, while adjusted estimates cluster lower, reflecting diagnostic-driven heterogeneity.

https://doi.org/10.1371/journal.pntd.0014029.g005

South America: higher exposure, limited confirmatory data

South America exhibited the highest apparent exposure. The misclassification-aware model estimated ~18% true seroprevalence (95% CI: 15–22%), consistently higher than in other continents. Confirmatory data were scarce: pooled infection results showed 0/40 PCR/culture positives (95% upper bound ≈7%).

The Bayesian baseline estimated true seroprevalence at 20.0% (95% CrI: 13.5–27.0%), and robust variants stabilized near 21.5% (18.0–26.5%). Diagnostic and influence analyses showed stable results without domination by individual studies (Fig 6). These results indicate higher serological prevalence estimates in South America relative to other regions, alongside limited confirmatory testing data.

Download:

Fig 6. Elevated true seroprevalence and diagnostic variability in South America.

(a) Posterior distribution from the misclassification-adjusted hierarchical model; median ≈21.5% (95% CrI ~ 18–26.5%). The unimodal, moderately tight posterior indicates consistently higher exposure than in North America or Europe. (b) Study-level apparent (×) and adjusted (●) prevalence with 95% CIs, labeled by diagnostic test. Crab-eating fox and culpeo series dominate the signal; SAT/RBT assays tend to overestimate compared to confirmatory tests. (c) Posterior predictive check for observed positives across study arms. Simulated (blue) distributions closely envelop empirical counts (black), and the posterior predictive mean (dashed) tracks the data, indicating good model fit.

https://doi.org/10.1371/journal.pntd.0014029.g006

Africa: under-sampled and uncertain

African data comprised only two populations from Tanzania (black-backed jackal and African wild dog). The estimated N-weighted true seroprevalence was ~ 6.0% (95% CI: 0.0–24.6%; k = 2; N = 180). The wide confidence interval reflects the small number of populations and limited sample size. All serological data were derived from SAT testing, and no PCR or culture-confirmed infections were reported for Africa. Despite adjustment for diagnostic performance, uncertainty remained high due to sparse data.

Species patterns and diagnostic effects

Among species, coyotes (Canis latrans) showed the highest weighted seroprevalence (≈15.6%; 85/545).Higher seroprevalence estimates were observed for coyotes compared with other canid species.

Diagnostic usage strongly shaped observed patterns. SAT was the most common assay (~2,510 animals tested across 15 arms), while ELISA was used less frequently (~630 tested, 5 arms). Screening tests (RBT, SAT) consistently produced higher positivity than confirmatory tests (CFT, ELISA) across matched populations (Fig 7A).

Download:

Fig 7. Geographic clustering of seropositivity and assay-driven differences.

(a) For identical populations, bars show apparent seroprevalence (95% CIs); asterisks (*) denote significant differences. Screening assays (RBT, CARD, BPAT) consistently yield higher positivity than confirmatory tests (CFT, ELISA), underscoring diagnostic misclassification and the need for confirmatory validation. (B) Choropleth map of the total number of animals tested, aggregated by country across included studies (blue = lower totals; red = higher totals; grey = no data). (C) Choropleth map of apparent seroprevalence by country from serological studies (seropositive/total tested; blue = lower; red = higher; grey = no data). (D) Choropleth map of confirmed infection by country based on PCR and/or culture (blue = lower; red = higher; grey = no data). Maps were created in R using public-domain basemap shapefiles from Natural Earth (https://www.naturalearthdata.com).

https://doi.org/10.1371/journal.pntd.0014029.g007

Geographically, apparent seroprevalence in Asia was disproportionately influenced by data from farmed blue-fox populations in China (Fig 7C), reflecting captive rather than free-ranging systems, alongside concentrated sampling effort (Fig 7B). Confirmed infection remained limited and geographically restricted (Fig 7D).

Discussion

Wild canids exhibit widespread serological exposure to Brucella spp. but active infection appears to be underestimated [3,46]. In our analysis among wild canids, integrating all serological assays and correcting for imperfect test performance, the estimated global true seroprevalence was 8.2% (95% CI: 5.1–11.3%). In contrast, only a small number of cases were confirmed by PCR or culture (3.9%; 95% CI: 3.0–5.1%), probably reflecting both the rarity of true infection and differences in testing availability between regions. Including farmed or captive populations raised seroprevalence to 15.5% (9.4–21.6%), largely due to outbreaks in blue-fox breeding facilities, demonstrating that captive amplification can distort regional patterns if not analyzed separately. Collectively, these findings indicate that Brucella exposure in wild canids is globally distributed but primarily localized around wildlife–livestock interfaces. Importantly, our use of a misclassification-aware, multi-assay modeling framework yields calibrated, decision-relevant prevalence estimates that reconcile heterogeneous diagnostic data, thereby addressing a persistent limitation in prior wildlife disease studies.

Seroprevalence differed across continents in consistent and biologically plausible ways. Europe showed the lowest exposure (~0.8%), with most wolves, jackals, and arctic foxes seronegative, while older red-fox data accounted for isolated outliers [3]. North America exhibited low-to-moderate exposure (~4.1%), consistent with localized transmission at livestock interfaces rather than broad enzootic circulation [3]. South America had higher values (~18%), largely in fox species from agro-pastoral landscapes, though confirmatory data (0/40 PCR or culture) remain absent [3]. Africa was severely under-sampled (~6%, N = 180; wide CI), limiting precision, [3] and Asia, when farm data were excluded, showed sparse wild serology (0/56 in jackals) but detectable infection in Russian culture data (4.66%) among wolves and arctic foxes, highlighting the danger of extrapolating from captive outbreaks [3,41]. Species composition also contributed: synanthropic canids, particularly coyotes (Canis latrans), showed the highest weighted seroprevalence (~15.6%), likely due to frequent contact with livestock environments, feeding on infected tissues, and shared resource use in peri-urban or agricultural landscapes [3,6]. Together, these patterns suggest that Brucella exposure in wild canids is regionally heterogeneous and often reflects ecological proximity to livestock rather than long-term maintenance within wild populations.

These patterns should be interpreted in light of the risk-of-bias assessment, which indicated greater methodological limitations in under-sampled regions and in studies relying exclusively on legacy serological assays. Much of the observed heterogeneity arises from methodological rather than biological causes. Agglutination-based screening assays such as RBT and SAT, consistently overestimate seropositivity relative to confirmatory tests such as CFT or ELISA [5,46]. Within-study comparisons revealed diagnostic inflation, with higher positivity reported for RBT than for CFT or AGID, and for CARD than for SAT or Rapid Immunoversion Test (RIV). After correcting misclassification and combining data from multiple assays, extreme values were reduced and regional estimates became more consistent. This suggests that apparent differences in naïve pooled results are driven mainly by variation in assay methods and cut-off thresholds, rather than by pathogen ecology alone [5]. The hierarchical Bayesian models,which explicitly correct for imperfect sensitivity and specificity and share information across studies, produced higher global seroprevalence estimates (approximately 16–27%,centered around 21–23%), as expected under diagnostic uncertainty and high heterogeneity. Posterior predictive checks, convergence diagnostics, and influence assessments all supported model stability, and robust or influence-trimmed variants yielded consistent results. The overall agreement between these Bayesian results and the misclassification-aware frequentist estimates suggests that much of the observed variation in seroprevalence is driven by differences in diagnostic methods rather than true ecological variation in Brucella circulation, resulting in more calibrated and decision-relevant estimates of global exposure in wild canids [3,46].

For One Health surveillance, programs should prioritize targeted sampling at livestock–wildlife interfaces, especially foxes and coyotes in agro-pastoral landscapes, and combine screening with confirmatory testing while explicitly modeling diagnostic misclassification to avoid bias [4,5]. Regions with elevated serology but little confirmatory data, including South America and parts of Asia, warrant targeted PCR and culture investigations, and captive systems, including fox farms in China, require strengthened biosecurity to prevent amplification and spillover [40]. By quantifying diagnostic uncertainty and translating results into a surveillance design framework, this study provides a reproducible model for decision-grade wildlife disease monitoring applicable across taxa and regions.

The main limitations of this study are uneven geographic and species coverage, particularly in Africa and parts of Asia; reliance in many studies on legacy agglutination assays with variable thresholds; and the scarcity of paired serology and PCR data from the same individuals. Standardized metadata collection, including age, carcass condition, vaccination history, and habitat context, alongside harmonized confirmatory testing protocols, would improve bias adjustment and strengthen inference on transmission potential. Future research should extend this framework to other wildlife reservoirs and zoonotic pathogens, integrating serological, molecular, and ecological data under unified Bayesian evidence synthesis. Such approaches will be critical for identifying true maintenance hosts and informing effective One Health interventions.

Conclusion

Wild canids show widespread exposure to Brucella spp. while confirmed active infection is reported much less often, a pattern shaped not only by biology but also by important differences in diagnostic methods and geographic coverage. Our risk-of-bias assessment indicates that much of the apparent variation among regions and species reflects these methodological differences particularly the reliance on older assays and limited data from Africa and parts of Asia, rather than true ecological contrasts alone. After accounting for diagnostic error and study-level uncertainty through misclassification-aware hierarchical modeling, exposure appears globally distributed but uneven, with higher levels concentrated near wildlife–livestock interfaces and in intensively monitored settings. Overall, the evidence is most consistent with wild canids acting primarily as sentinels of environmental and livestock-associated Brucella exposure, while still allowing for localized maintenance in specific ecological or management contexts, especially in captive or high-contact systems. By integrating diverse diagnostic data and explicitly addressing bias, this study provides more reliable, decision-relevant estimates of global exposure and a transferable framework for improving wildlife disease surveillance and One Health monitoring.

AI assistance disclosure

The authors used ChatGPT-5 (OpenAI) solely to assist with grammar, sentence structure, and clarity during manuscript revision. All scientific content, data analysis, interpretation, and conclusions were generated entirely by the authors. The authors reviewed and edited all AI-suggested language and take full responsibility for the final text.

Supporting information

S1 File. PRISMA-ScR Checklist.

Attribution: This PRISMA-ScR checklist is reproduced from the PRISMA-ScR statement materials and is licensed under the Creative Commons Attribution 4.0 International License (CC BY 4.0). The original guideline and supporting materials are available at https://www.prisma-statement.org/. The checklist is cited as: Tricco AC, Lillie E, Zarin W, O’Brien KK, Colquhoun H, Levac D, et al. PRISMA Extension for Scoping Reviews (PRISMA-ScR): Checklist and Explanation. Ann Intern Med. 2018;169(7):467–473. https://doi.org/10.7326/M18-0850.

https://doi.org/10.1371/journal.pntd.0014029.s001

(DOCX)

S1 Fig. Risk-of-bias heat-map figure.

https://doi.org/10.1371/journal.pntd.0014029.s002

(DOCX)

S2 File. Full database search strings used across all bibliographic sources.

https://doi.org/10.1371/journal.pntd.0014029.s003

(DOCX)

S1 Table. Full-text exclusion log.

https://doi.org/10.1371/journal.pntd.0014029.s004

(DOCX)

S2 Table. Newcastle–Ottawa Quality Appraisal Scores for included studies.

https://doi.org/10.1371/journal.pntd.0014029.s005

(DOCX)

S3 Table. Study-level meta-analysis data for Brucella exposure and infection in wild canids.

https://doi.org/10.1371/journal.pntd.0014029.s006

(DOCX)

References

1. Qureshi KA, Parvez A, Fahmy NA, Abdel Hady BH, Kumar S, Ganguly A, et al. Brucellosis: epidemiology, pathogenesis, diagnosis and treatment-a comprehensive review. Ann Med. 2023;55(2):2295398. pmid:38165919
- View Article
- PubMed/NCBI
- Google Scholar
2. Sambu RM, Mathew C, Nonga HE, Lukambagire AS, Yapi RB, Akoko J, et al. Circulating Brucella species in wild animals of the Serengeti ecosystem, Tanzania. One Health Outlook. 2021;3(1):15. pmid:34425909
- View Article
- PubMed/NCBI
- Google Scholar
3. Dadar M, Shahali Y, Fakhri Y, Godfroid J. The global epidemiology of Brucella infections in terrestrial wildlife: A meta-analysis. Transbound Emerg Dis. 2021;68(2):715–29. pmid:32679611
- View Article
- PubMed/NCBI
- Google Scholar
4. Michel AL, Van Heerden H, Crossley BM, Al Dahouk S, Prasse D, Rutten V. Pathogen detection and disease diagnosis in wildlife: challenges and opportunities. Rev Sci Tech. 2021;40(1):105–18. pmid:34140737
- View Article
- PubMed/NCBI
- Google Scholar
5. Godfroid J, Nielsen K, Saegerman C. Diagnosis of brucellosis in livestock and wildlife. Croat Med J. 2010;51(4):296–305. pmid:20718082
- View Article
- PubMed/NCBI
- Google Scholar
6. Kosoy M, Goodrich I. Comparative ecology of Bartonella and Brucella infections in wild carnivores. Front Vet Sci. 2019;6:322.
- View Article
- Google Scholar
7. Tricco AC, Lillie E, Zarin W, O’Brien KK, Colquhoun H, Levac D, et al. PRISMA Extension for Scoping Reviews (PRISMA-ScR): Checklist and Explanation. Ann Intern Med. 2018;169(7):467–73. pmid:30178033
- View Article
- PubMed/NCBI
- Google Scholar
8. Proença LM, Silva JCR, Galera PD, Lion MB, Marinho-Filho JS, Ragozo AMA, et al. Serologic survey of infectious diseases in populations of maned wolf (Chrysocyon brachyurus) and crab-eating fox (Cerdocyon thous) from Aguas Emendadas Ecological Station, Brazil. J Zoo Wildl Med. 2013;44(1):152–5. pmid:23505716
- View Article
- PubMed/NCBI
- Google Scholar
9. Martino PE, Montenegro JL, Preziosi JA, Venturini C, Bacigalupe D, Stanchi NO. Serological survey of selected pathogens of free-ranging foxes in southern Argentina, 1998–2001. Rev Sci Tech. 2004;23:801–6.
- View Article
- Google Scholar
10. Oliveira-Filho EF, Pinheiro JWJ, Souza MMA, Santana VLA, Silva JCR, Mota RA, et al. Serologic survey of brucellosis in captive neotropical wild carnivores in northeast Brazil. J Zoo Wildl Med. 2012;43(2):384–7. pmid:22779245
- View Article
- PubMed/NCBI
- Google Scholar
11. Dorneles EMS, Pellegrin AO, Péres IAHFS, Mathias LA, Mourão G, Bianchi R de C, et al. Serology for brucellosis in free-ranging crab-eating foxes (Cerdocyon thous) and brown-nosed coatis (Nasua nasua) from Brazilian Pantanal. Cienc Rural. 2014;44(12):2193–6.
- View Article
- Google Scholar
12. Uzai GJS, Monteiro CP, Soares R, Silva MA, Oliveira AR, Santos DO. Morphological and molecular diagnosis of diseases of free-ranging crab-eating foxes (Cerdocyon thous). Arq Bras Med Vet Zootec. 2021;73(3):583–8.
- View Article
- Google Scholar
13. Azevedo SS de, Silva MLCR, Batista C de SA, Gomes AA de B, Vasconcellos SA, Alves CJ. Anticorpos anti Brucella abortus, anti Brucella canis e anti Leptospira spp. em raposas (Pseudalopex vetulus) do semiárido paraibano, Nordeste do Brasil. Cienc Rural. 2009;40(1):190–2.
- View Article
- Google Scholar
14. Galarce N, de la Fuente S, Escobar B, Dettleff P, Abalos P, Hormazábal JC, et al. Survey of Zoonotic Bacterial Pathogens in Native Foxes in Central Chile: First Record of Brucella canis Exposure. Animals (Basel). 2021;11(7):1980. pmid:34359107
- View Article
- PubMed/NCBI
- Google Scholar
15. Fuente González SI de la. Detección de la infección por Brucella abortus y Brucella canis en cánidos no domésticos de centro de rehabilitación y exhibición de la zona central de Chile. Santiago: Universidad de Chile; 2021.
16. Moya S, Oettinger S, Borie C, Flores R, Abalos P, Briceño C. Serologic Survey of Brucella canis and Leptospira spp. in Free-Ranging Wild and Domestic Canids from Tierra del Fuego, Chile. J Wildl Dis. 2019;55(3):713–6. pmid:30566382
- View Article
- PubMed/NCBI
- Google Scholar
17. Hidalgo-Hermoso E, Cabello J, Verasay J, Moreira-Arce D, Hidalgo M, Abalos P, et al. Serosurvey for selected parasitic and bacterial pathogens in darwin’s fox (lycalopex Fulvipes): Not only dog diseases are a threat. J Wildl Dis. 2022;58(1):76–85. pmid:34714909
- View Article
- PubMed/NCBI
- Google Scholar
18. Fiorello CV, Noss AJ, Deem SL, Maffei L, Dubovi EJ. Serosurvey of small carnivores in the Bolivian Chaco. J Wildl Dis. 2007;43(3):551–7. pmid:17699100
- View Article
- PubMed/NCBI
- Google Scholar
19. Szyfres B, González Tomé J. Natural Brucella infection in Argentine wild foxes. Bull World Health Organ. 1966;34(6):919–23. pmid:5296540
- View Article
- PubMed/NCBI
- Google Scholar
20. de Macedo GC, Herrera HM, de Oliveira Porfírio GE, Santos FM, de Assis WO, de Andrade GB, et al. Brucellosis in the Brazilian Pantanal wetland: threat to animal production and wildlife conservation. Braz J Microbiol. 2022;53(4):2287–97. pmid:36269553
- View Article
- PubMed/NCBI
- Google Scholar
21. Chitwood MC, Swingen MB, Lashley MA, Flowers JR, Palamar MB, Apperson CS, et al. Parasitology and serology of free-ranging coyotes (canis Latrans) in North Carolina, USA. J Wildl Dis. 2015;51(3):664–9. pmid:25984773
- View Article
- PubMed/NCBI
- Google Scholar
22. Williams JD, Heck FC, Davis DS, Adams LG. Comparison of results from five serologic methods used for detecting Brucella abortus antibody activity in coyote sera. Vet Immunol Immunopathol. 1991;29(1–2):79–87. pmid:1949585
- View Article
- PubMed/NCBI
- Google Scholar
23. Zarnke RL, Ver Hoef JM, DeLong RA. Geographic pattern of serum antibody prevalence for Brucella spp. in caribou, grizzly bears, and wolves from Alaska, 1975–1998. J Wildl Dis. 2006;42(3):570–7.
- View Article
- Google Scholar
24. Hoq MA. A serologic survey of Brucella agglutinins in wildlife and sheep. Calif Vet. 1978;32(3):15–7.
- View Article
- Google Scholar
25. Hoff GL, Bigler WJ, Trainer DO, Debbie JG, Brown GM, Winkler WG, et al. Survey of selected carnivore and opossum serums for agglutinins to Brucella canis. J Am Vet Med Assoc. 1974;165(9):830–1. pmid:4426867
- View Article
- PubMed/NCBI
- Google Scholar
26. Randhawa AS, Kelly VP, Baker EF Jr. Agglutinins to Coxiella burnetii and Brucella spp, with particular reference to Brucella canis, in wild animals of southern Texas. J Am Vet Med Assoc. 1977;171(9):939–42. pmid:924870
- View Article
- PubMed/NCBI
- Google Scholar
27. Davis DS, Boeer WJ, Mims JP, Heck FC, Adams LG. Brucella abortus in coyotes. I. A serologic and bacteriologic survey in eastern Texas. J Wildl Dis. 1979;15(3):367–72. pmid:116012
- View Article
- PubMed/NCBI
- Google Scholar
28. Neiland KA. Rangiferine brucellosis in Alaskan canids. J Wildl Dis. 1970;6(3):136–9. pmid:5424432
- View Article
- PubMed/NCBI
- Google Scholar
29. Tessaro SV. A descriptive and epizootiologic study of brucellosis and tuberculosis in bison in northern Canada. Saskatoon: University of Saskatchewan. 1987.
30. Neiland KA. Further observations on rangiferine brucellosis in Alaskan carnivores. J Wildl Dis. 1975;11(1):45–53. pmid:1113440
- View Article
- PubMed/NCBI
- Google Scholar
31. Schnurrenberger PR, Brown RR, Hill EP, Scanlan CM, Altiere JA, Wykoff JT. Brucella abortus in wildlife on selected cattle farms in Alabama. J Wildl Dis. 1985;21(2):132–6. pmid:3923213
- View Article
- PubMed/NCBI
- Google Scholar
32. Morton JK. Brucella suis type 4 in foxes and their role as reservoirs and vectors among reindeer. Fairbanks (AK): University of Alaska Fairbanks. 1989.
33. McCue PM, O’Farrell TP. Serological survey for selected diseases in the endangered San Joaquin kit fox (Vulpes macrotis mutica). J Wildl Dis. 1988;24(2):274–81.
- View Article
- Google Scholar
34. Cirović D, Chochlakis D, Tomanović S, Sukara R, Penezić A, Tselentis Y, et al. Presence of Leishmania and Brucella species in the golden jackal Canis aureus in Serbia. Biomed Res Int. 2014;2014:728516. pmid:24967397
- View Article
- PubMed/NCBI
- Google Scholar
35. Nakonechnyi IV, Perots’ka LV, Pyvovarova IV, Chornyi VA. Ecological and epizootic roles of Golden jackal, genus Canis aureus in the Northwest of Black Sea coast. SMLNUVMB. 2019;21(94):37–43.
- View Article
- Google Scholar
36. Bertelloni F, Cagnoli G, Ebani VV. Survey on the Occurrence of Zoonotic Bacterial Pathogens in the Feces of Wolves (Canis lupus italicus) Collected in a Protected Area in Central Italy. Microorganisms. 2024;12(11):2367. pmid:39597755
- View Article
- PubMed/NCBI
- Google Scholar
37. Nymo IH, Fuglei E, Mørk T, Breines EM, Holmgren KE, Davidson RK. Why are Svalbard Arctic foxes Brucella spp. seronegative?. Polar Research. 2022;41:7867.
- View Article
- Google Scholar
38. Ebani VV, Trebino C, Guardone L, Bertelloni F, Cagnoli G, Nardoni S, et al. Occurrence of Bacterial and Protozoan Pathogens in Red Foxes (Vulpes vulpes) in Central Italy. Animals (Basel). 2022;12(20):2891. pmid:36290277
- View Article
- PubMed/NCBI
- Google Scholar
39. Minichino A, Ciuca L, Dipineto L, Rinaldi L, Montagnaro S, Borrelli L, et al. Exposure to selected pathogens in wild mammals from a rescue and rehabilitation center in southern Italy. One Health. 2025;20:101049. pmid:40959801
- View Article
- PubMed/NCBI
- Google Scholar
40. Zhou Y, Meng Y, Ren Y, Liu Z, Li Z. A Retrospective survey of the abortion outbreak event caused by brucellosis at a blue fox breeding farm in heilongjiang Province, China. Front Vet Sci. 2021;8:666254. pmid:34212019
- View Article
- PubMed/NCBI
- Google Scholar
41. Egorov EA, Kalinovski AI, Maramovich AS, Cherniavski VF. Problems of epidemiologic surveillance of brucellosis in reindeer breeding. Epidemiol Infect Dis. 1997;2:18–21.
- View Article
- Google Scholar
42. Lapid R, Motro Y, Craddock H, Khalfin B, King R, Bar-Gal GK, et al. Fecal microbiota of the synanthropic golden jackal (Canis aureus). Anim Microbiome. 2023;5(1):37. pmid:37542305
- View Article
- PubMed/NCBI
- Google Scholar
43. Pinigin AF, Zabrodin VA, Nikulina VI. Brucellosis in arctic foxes (Alopex lagopus). Moscow: Academy of Sciences of the USSR. 1970.
44. Sachs R, Staak C, Groocock CM. Serological investigation of brucellosis in game animals in Tanzania. Dar es Salaam: Veterinary Research Institute; 1968.
45. Wells GA, Shea B, O’Connell D, Peterson J, Welch V, Losos M. The Newcastle–Ottawa Scale (NOS) for assessing the quality of nonrandomised studies in meta-analyses. 2000. [Cited 2026 January 20]. https://ohri.ca/en/who-we-are/core-facilities-and-platforms/ottawa-methods-centre/newcastle-ottawa-scale
- View Article
- Google Scholar
46. Greiner M, Gardner IA. Epidemiologic issues in the validation of veterinary diagnostic tests. Prev Vet Med. 2000;45(1–2):3–22. pmid:10802331
- View Article
- PubMed/NCBI
- Google Scholar
47. Hoffman MD, Gelman A. The No-U-Turn Sampler: Adaptively Setting Path Lengths in Hamiltonian Monte Carlo. J Mach Learn Res. 2014;15:1593–623.
- View Article
- Google Scholar
48. Hyppönen V. A comparative analysis of migration tools for Python 3 in medical device software. Tampere: Tampere University; 2025.

[ref1] 1. Qureshi KA, Parvez A, Fahmy NA, Abdel Hady BH, Kumar S, Ganguly A, et al. Brucellosis: epidemiology, pathogenesis, diagnosis and treatment-a comprehensive review. Ann Med. 2023;55(2):2295398. pmid:38165919
View Article
PubMed/NCBI
Google Scholar

[2] View Article

[3] PubMed/NCBI

[4] Google Scholar

[ref2] 2. Sambu RM, Mathew C, Nonga HE, Lukambagire AS, Yapi RB, Akoko J, et al. Circulating Brucella species in wild animals of the Serengeti ecosystem, Tanzania. One Health Outlook. 2021;3(1):15. pmid:34425909
View Article
PubMed/NCBI
Google Scholar

[6] View Article

[7] PubMed/NCBI

[8] Google Scholar

[ref3] 3. Dadar M, Shahali Y, Fakhri Y, Godfroid J. The global epidemiology of Brucella infections in terrestrial wildlife: A meta-analysis. Transbound Emerg Dis. 2021;68(2):715–29. pmid:32679611
View Article
PubMed/NCBI
Google Scholar

[10] View Article

[11] PubMed/NCBI

[12] Google Scholar

[ref4] 4. Michel AL, Van Heerden H, Crossley BM, Al Dahouk S, Prasse D, Rutten V. Pathogen detection and disease diagnosis in wildlife: challenges and opportunities. Rev Sci Tech. 2021;40(1):105–18. pmid:34140737
View Article
PubMed/NCBI
Google Scholar

[14] View Article

[15] PubMed/NCBI

[16] Google Scholar

[ref5] 5. Godfroid J, Nielsen K, Saegerman C. Diagnosis of brucellosis in livestock and wildlife. Croat Med J. 2010;51(4):296–305. pmid:20718082
View Article
PubMed/NCBI
Google Scholar

[18] View Article

[19] PubMed/NCBI

[20] Google Scholar

[ref6] 6. Kosoy M, Goodrich I. Comparative ecology of Bartonella and Brucella infections in wild carnivores. Front Vet Sci. 2019;6:322.
View Article
Google Scholar

[22] View Article

[23] Google Scholar

[ref7] 7. Tricco AC, Lillie E, Zarin W, O’Brien KK, Colquhoun H, Levac D, et al. PRISMA Extension for Scoping Reviews (PRISMA-ScR): Checklist and Explanation. Ann Intern Med. 2018;169(7):467–73. pmid:30178033
View Article
PubMed/NCBI
Google Scholar

[25] View Article

[26] PubMed/NCBI

[27] Google Scholar

[ref8] 8. Proença LM, Silva JCR, Galera PD, Lion MB, Marinho-Filho JS, Ragozo AMA, et al. Serologic survey of infectious diseases in populations of maned wolf (Chrysocyon brachyurus) and crab-eating fox (Cerdocyon thous) from Aguas Emendadas Ecological Station, Brazil. J Zoo Wildl Med. 2013;44(1):152–5. pmid:23505716
View Article
PubMed/NCBI
Google Scholar

[29] View Article

[30] PubMed/NCBI

[31] Google Scholar

[ref9] 9. Martino PE, Montenegro JL, Preziosi JA, Venturini C, Bacigalupe D, Stanchi NO. Serological survey of selected pathogens of free-ranging foxes in southern Argentina, 1998–2001. Rev Sci Tech. 2004;23:801–6.
View Article
Google Scholar

[33] View Article

[34] Google Scholar

[ref10] 10. Oliveira-Filho EF, Pinheiro JWJ, Souza MMA, Santana VLA, Silva JCR, Mota RA, et al. Serologic survey of brucellosis in captive neotropical wild carnivores in northeast Brazil. J Zoo Wildl Med. 2012;43(2):384–7. pmid:22779245
View Article
PubMed/NCBI
Google Scholar

[36] View Article

[37] PubMed/NCBI

[38] Google Scholar

[ref11] 11. Dorneles EMS, Pellegrin AO, Péres IAHFS, Mathias LA, Mourão G, Bianchi R de C, et al. Serology for brucellosis in free-ranging crab-eating foxes (Cerdocyon thous) and brown-nosed coatis (Nasua nasua) from Brazilian Pantanal. Cienc Rural. 2014;44(12):2193–6.
View Article
Google Scholar

[40] View Article

[41] Google Scholar

[ref12] 12. Uzai GJS, Monteiro CP, Soares R, Silva MA, Oliveira AR, Santos DO. Morphological and molecular diagnosis of diseases of free-ranging crab-eating foxes (Cerdocyon thous). Arq Bras Med Vet Zootec. 2021;73(3):583–8.
View Article
Google Scholar

[43] View Article

[44] Google Scholar

[ref13] 13. Azevedo SS de, Silva MLCR, Batista C de SA, Gomes AA de B, Vasconcellos SA, Alves CJ. Anticorpos anti Brucella abortus, anti Brucella canis e anti Leptospira spp. em raposas (Pseudalopex vetulus) do semiárido paraibano, Nordeste do Brasil. Cienc Rural. 2009;40(1):190–2.
View Article
Google Scholar

[46] View Article

[47] Google Scholar

[ref14] 14. Galarce N, de la Fuente S, Escobar B, Dettleff P, Abalos P, Hormazábal JC, et al. Survey of Zoonotic Bacterial Pathogens in Native Foxes in Central Chile: First Record of Brucella canis Exposure. Animals (Basel). 2021;11(7):1980. pmid:34359107
View Article
PubMed/NCBI
Google Scholar

[49] View Article

[50] PubMed/NCBI

[51] Google Scholar

[ref15] 15. Fuente González SI de la. Detección de la infección por Brucella abortus y Brucella canis en cánidos no domésticos de centro de rehabilitación y exhibición de la zona central de Chile. Santiago: Universidad de Chile; 2021.

[ref16] 16. Moya S, Oettinger S, Borie C, Flores R, Abalos P, Briceño C. Serologic Survey of Brucella canis and Leptospira spp. in Free-Ranging Wild and Domestic Canids from Tierra del Fuego, Chile. J Wildl Dis. 2019;55(3):713–6. pmid:30566382
View Article
PubMed/NCBI
Google Scholar

[54] View Article

[55] PubMed/NCBI

[56] Google Scholar

[ref17] 17. Hidalgo-Hermoso E, Cabello J, Verasay J, Moreira-Arce D, Hidalgo M, Abalos P, et al. Serosurvey for selected parasitic and bacterial pathogens in darwin’s fox (lycalopex Fulvipes): Not only dog diseases are a threat. J Wildl Dis. 2022;58(1):76–85. pmid:34714909
View Article
PubMed/NCBI
Google Scholar

[58] View Article

[59] PubMed/NCBI

[60] Google Scholar

[ref18] 18. Fiorello CV, Noss AJ, Deem SL, Maffei L, Dubovi EJ. Serosurvey of small carnivores in the Bolivian Chaco. J Wildl Dis. 2007;43(3):551–7. pmid:17699100
View Article
PubMed/NCBI
Google Scholar

[62] View Article

[63] PubMed/NCBI

[64] Google Scholar

[ref19] 19. Szyfres B, González Tomé J. Natural Brucella infection in Argentine wild foxes. Bull World Health Organ. 1966;34(6):919–23. pmid:5296540
View Article
PubMed/NCBI
Google Scholar

[66] View Article

[67] PubMed/NCBI

[68] Google Scholar

[ref20] 20. de Macedo GC, Herrera HM, de Oliveira Porfírio GE, Santos FM, de Assis WO, de Andrade GB, et al. Brucellosis in the Brazilian Pantanal wetland: threat to animal production and wildlife conservation. Braz J Microbiol. 2022;53(4):2287–97. pmid:36269553
View Article
PubMed/NCBI
Google Scholar

[70] View Article

[71] PubMed/NCBI

[72] Google Scholar

[ref21] 21. Chitwood MC, Swingen MB, Lashley MA, Flowers JR, Palamar MB, Apperson CS, et al. Parasitology and serology of free-ranging coyotes (canis Latrans) in North Carolina, USA. J Wildl Dis. 2015;51(3):664–9. pmid:25984773
View Article
PubMed/NCBI
Google Scholar

[74] View Article

[75] PubMed/NCBI

[76] Google Scholar

[ref22] 22. Williams JD, Heck FC, Davis DS, Adams LG. Comparison of results from five serologic methods used for detecting Brucella abortus antibody activity in coyote sera. Vet Immunol Immunopathol. 1991;29(1–2):79–87. pmid:1949585
View Article
PubMed/NCBI
Google Scholar

[78] View Article

[79] PubMed/NCBI

[80] Google Scholar

[ref23] 23. Zarnke RL, Ver Hoef JM, DeLong RA. Geographic pattern of serum antibody prevalence for Brucella spp. in caribou, grizzly bears, and wolves from Alaska, 1975–1998. J Wildl Dis. 2006;42(3):570–7.
View Article
Google Scholar

[82] View Article

[83] Google Scholar

[ref24] 24. Hoq MA. A serologic survey of Brucella agglutinins in wildlife and sheep. Calif Vet. 1978;32(3):15–7.
View Article
Google Scholar

[85] View Article

[86] Google Scholar

[ref25] 25. Hoff GL, Bigler WJ, Trainer DO, Debbie JG, Brown GM, Winkler WG, et al. Survey of selected carnivore and opossum serums for agglutinins to Brucella canis. J Am Vet Med Assoc. 1974;165(9):830–1. pmid:4426867
View Article
PubMed/NCBI
Google Scholar

[88] View Article

[89] PubMed/NCBI

[90] Google Scholar

[ref26] 26. Randhawa AS, Kelly VP, Baker EF Jr. Agglutinins to Coxiella burnetii and Brucella spp, with particular reference to Brucella canis, in wild animals of southern Texas. J Am Vet Med Assoc. 1977;171(9):939–42. pmid:924870
View Article
PubMed/NCBI
Google Scholar

[92] View Article

[93] PubMed/NCBI

[94] Google Scholar

[ref27] 27. Davis DS, Boeer WJ, Mims JP, Heck FC, Adams LG. Brucella abortus in coyotes. I. A serologic and bacteriologic survey in eastern Texas. J Wildl Dis. 1979;15(3):367–72. pmid:116012
View Article
PubMed/NCBI
Google Scholar

[96] View Article

[97] PubMed/NCBI

[98] Google Scholar

[ref28] 28. Neiland KA. Rangiferine brucellosis in Alaskan canids. J Wildl Dis. 1970;6(3):136–9. pmid:5424432
View Article
PubMed/NCBI
Google Scholar

[100] View Article

[101] PubMed/NCBI

[102] Google Scholar

[ref29] 29. Tessaro SV. A descriptive and epizootiologic study of brucellosis and tuberculosis in bison in northern Canada. Saskatoon: University of Saskatchewan. 1987.

[ref30] 30. Neiland KA. Further observations on rangiferine brucellosis in Alaskan carnivores. J Wildl Dis. 1975;11(1):45–53. pmid:1113440
View Article
PubMed/NCBI
Google Scholar

[105] View Article

[106] PubMed/NCBI

[107] Google Scholar

[ref31] 31. Schnurrenberger PR, Brown RR, Hill EP, Scanlan CM, Altiere JA, Wykoff JT. Brucella abortus in wildlife on selected cattle farms in Alabama. J Wildl Dis. 1985;21(2):132–6. pmid:3923213
View Article
PubMed/NCBI
Google Scholar

[109] View Article

[110] PubMed/NCBI

[111] Google Scholar

[ref32] 32. Morton JK. Brucella suis type 4 in foxes and their role as reservoirs and vectors among reindeer. Fairbanks (AK): University of Alaska Fairbanks. 1989.

[ref33] 33. McCue PM, O’Farrell TP. Serological survey for selected diseases in the endangered San Joaquin kit fox (Vulpes macrotis mutica). J Wildl Dis. 1988;24(2):274–81.
View Article
Google Scholar

[114] View Article

[115] Google Scholar

[ref34] 34. Cirović D, Chochlakis D, Tomanović S, Sukara R, Penezić A, Tselentis Y, et al. Presence of Leishmania and Brucella species in the golden jackal Canis aureus in Serbia. Biomed Res Int. 2014;2014:728516. pmid:24967397
View Article
PubMed/NCBI
Google Scholar

[117] View Article

[118] PubMed/NCBI

[119] Google Scholar

[ref35] 35. Nakonechnyi IV, Perots’ka LV, Pyvovarova IV, Chornyi VA. Ecological and epizootic roles of Golden jackal, genus Canis aureus in the Northwest of Black Sea coast. SMLNUVMB. 2019;21(94):37–43.
View Article
Google Scholar

[121] View Article

[122] Google Scholar

[ref36] 36. Bertelloni F, Cagnoli G, Ebani VV. Survey on the Occurrence of Zoonotic Bacterial Pathogens in the Feces of Wolves (Canis lupus italicus) Collected in a Protected Area in Central Italy. Microorganisms. 2024;12(11):2367. pmid:39597755
View Article
PubMed/NCBI
Google Scholar

[124] View Article

[125] PubMed/NCBI

[126] Google Scholar

[ref37] 37. Nymo IH, Fuglei E, Mørk T, Breines EM, Holmgren KE, Davidson RK. Why are Svalbard Arctic foxes Brucella spp. seronegative?. Polar Research. 2022;41:7867.
View Article
Google Scholar

[128] View Article

[129] Google Scholar

[ref38] 38. Ebani VV, Trebino C, Guardone L, Bertelloni F, Cagnoli G, Nardoni S, et al. Occurrence of Bacterial and Protozoan Pathogens in Red Foxes (Vulpes vulpes) in Central Italy. Animals (Basel). 2022;12(20):2891. pmid:36290277
View Article
PubMed/NCBI
Google Scholar

[131] View Article

[132] PubMed/NCBI

[133] Google Scholar

[ref39] 39. Minichino A, Ciuca L, Dipineto L, Rinaldi L, Montagnaro S, Borrelli L, et al. Exposure to selected pathogens in wild mammals from a rescue and rehabilitation center in southern Italy. One Health. 2025;20:101049. pmid:40959801
View Article
PubMed/NCBI
Google Scholar

[135] View Article

[136] PubMed/NCBI

[137] Google Scholar

[ref40] 40. Zhou Y, Meng Y, Ren Y, Liu Z, Li Z. A Retrospective survey of the abortion outbreak event caused by brucellosis at a blue fox breeding farm in heilongjiang Province, China. Front Vet Sci. 2021;8:666254. pmid:34212019
View Article
PubMed/NCBI
Google Scholar

[139] View Article

[140] PubMed/NCBI

[141] Google Scholar

[ref41] 41. Egorov EA, Kalinovski AI, Maramovich AS, Cherniavski VF. Problems of epidemiologic surveillance of brucellosis in reindeer breeding. Epidemiol Infect Dis. 1997;2:18–21.
View Article
Google Scholar

[143] View Article

[144] Google Scholar

[ref42] 42. Lapid R, Motro Y, Craddock H, Khalfin B, King R, Bar-Gal GK, et al. Fecal microbiota of the synanthropic golden jackal (Canis aureus). Anim Microbiome. 2023;5(1):37. pmid:37542305
View Article
PubMed/NCBI
Google Scholar

[146] View Article

[147] PubMed/NCBI

[148] Google Scholar

[ref43] 43. Pinigin AF, Zabrodin VA, Nikulina VI. Brucellosis in arctic foxes (Alopex lagopus). Moscow: Academy of Sciences of the USSR. 1970.

[ref44] 44. Sachs R, Staak C, Groocock CM. Serological investigation of brucellosis in game animals in Tanzania. Dar es Salaam: Veterinary Research Institute; 1968.

[ref45] 45. Wells GA, Shea B, O’Connell D, Peterson J, Welch V, Losos M. The Newcastle–Ottawa Scale (NOS) for assessing the quality of nonrandomised studies in meta-analyses. 2000. [Cited 2026 January 20]. https://ohri.ca/en/who-we-are/core-facilities-and-platforms/ottawa-methods-centre/newcastle-ottawa-scale
View Article
Google Scholar

[152] View Article

[153] Google Scholar

[ref46] 46. Greiner M, Gardner IA. Epidemiologic issues in the validation of veterinary diagnostic tests. Prev Vet Med. 2000;45(1–2):3–22. pmid:10802331
View Article
PubMed/NCBI
Google Scholar

[155] View Article

[156] PubMed/NCBI

[157] Google Scholar

[ref47] 47. Hoffman MD, Gelman A. The No-U-Turn Sampler: Adaptively Setting Path Lengths in Hamiltonian Monte Carlo. J Mach Learn Res. 2014;15:1593–623.
View Article
Google Scholar

[159] View Article

[160] Google Scholar

[ref48] 48. Hyppönen V. A comparative analysis of migration tools for Python 3 in medical device software. Tampere: Tampere University; 2025.

Figures

Abstract

Author summary

Introduction

Methods

Bibliographic search strategy

Data extraction and synthesis

Inclusion and exclusion criteria

Quality appraisal (QA)

Outcomes and unit of analysis

Serology track: misclassification-adjusted true seroprevalence

Infection track: PCR/culture prevalence

Machine learning Bayesian meta-analysis (serology)

Software and reproducibility

Results

Search results

Characteristics of included studies

Risk of bias assessment

Global exposure is widespread, whereas confirmed active infection is uncommon

Asia: sparse wild data, localized infection

Europe: lowest exposure and limited infection evidence

North America: low-to-moderate exposure, scarce confirmation

South America: higher exposure, limited confirmatory data

Africa: under-sampled and uncertain

Species patterns and diagnostic effects

Discussion

Conclusion

AI assistance disclosure

Supporting information

S1 File. PRISMA-ScR Checklist.

S1 Fig. Risk-of-bias heat-map figure.

S2 File. Full database search strings used across all bibliographic sources.

S1 Table. Full-text exclusion log.

S2 Table. Newcastle–Ottawa Quality Appraisal Scores for included studies.

S3 Table. Study-level meta-analysis data for Brucella exposure and infection in wild canids.

References