This systematic review and meta-analysis was reviewed by the Human Research Ethics Committee, Faculty of Public Health, Mahidol University (Protocol No. MUPH 45/2026), which determined that the study protocol complies with the criteria for the “research with exemption category”. As the study synthesized and analyzed data from previously published studies and did not involve direct recruitment of human participants, collection of identifiable personal information, or direct experimentation on animals, additional consent was not required.

Data extraction and statistical analyses were conducted from April to May 2026. The review included eligible studies published from January 1, 2012, to December 31, 2024. The electronic literature search was conducted using PubMed, Scopus, and Google Scholar to identify eligible articles published within the predefined study period. In addition, the reference lists of all included studies were manually screened to identify potentially eligible publications that were not retrieved through the electronic database searches.

This study was conducted as a systematic review and meta-analysis in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 guidelines [25]. The review protocol was prospectively registered in the International Prospective Register of Systematic Reviews (PROSPERO; registration No. CRD420251114318). During the review process, the PROSPERO record was amended to include the additional keywords without any substantive modification of the registered methodological protocol.

The literature search strategy was developed using a modified Population, Exposure, Comparator, and Outcome (PECO) framework [26]. The strategy was designed to identify both phenotypic and genotypic evidence of extended-spectrum β-lactamase (ESBL)-producing Vibrio spp. across human, animal/aquaculture, food, and environmental sources. Controlled vocabulary and free-text terms related to Vibrio spp. and ESBL production were combined to maximize search sensitivity.

The core Boolean search strategy was structured as follows: “Vibrio” AND (“ESBL” OR “extended-spectrum β-lactamases” OR “extended-spectrum beta lactamase” OR “bla_CTX-M” OR “bla_TEM” OR “bla_SHV”). Additional species-specific terms, including Vibrio parahaemolyticus, Vibrio vulnificus, and Vibrio cholerae, were incorporated where appropriate, with the search syntax and fields adapted to the requirements of each database. Searches were restricted to original research articles published in English during the predefined publication period. Complete database-specific search strategies, including Boolean operators, search fields, and applied filters, are provided in Table S1.

A Vibrio isolate was classified as ESBL-producing when phenotypic ESBL production was confirmed and/or an ESBL-associated gene was detected using a molecular method, regardless of whether phenotypic testing was performed. Phenotypic confirmation included standard ESBL confirmatory methods reported in the original studies, such as the double-disk synergy test, combined-disk test using third-generation cephalosporins with and without clavulanic acid, ESBL E-test, or minimum inhibitory concentration (MIC)-based confirmation demon-strating reduced cephalosporin susceptibility with clavulanate inhibition. Where specified by the original studies, interpretation was based on recognized criteria, including those of the Clinical and Laboratory Standards Institute (CLSI) or European Committee on Antimicrobial Susceptibility Testing (EUCAST). Molecular confirmation included the detection of ESBL-associated β-lactamase genes, including bla_CTX-M, bla_TEM, bla_SHV, and other ESBL-associated genes reported by the included studies.

To evaluate the potential influence of this broader ESBL-definition on the pooled prevalence estimate, a sensitivity analysis was performed by restricting the analysis to studies reporting phenotypically confirmed ESBL-producing isolates.

Studies were eligible if they investigated Vibrio isolates obtained from clinical cases or healthy humans; animals, including aquaculture species and terrestrial livestock; food or seafood products; or environmental sources, including water and sediment. Studies were required to report phenotypic confirmation of ESBL production using recognized methods and/or molecular detection of ESBL-associated genes or their variants. No geographical restrictions were applied, but only original research articles published in English were eligible.

Review articles, meta-analyses, editorials, conference abstracts, short communications without sufficient primary data, and case reports lacking data suitable for prevalence estimation were excluded.

The primary outcome was the prevalence of ESBL-producing Vibrio spp. Secondary outcomes included the prevalence and distribution of Vibrio spp., distribution of ESBL-associated genes and their variants, phenotypic AMR profiles, and MDR, where reported. Geographic, source-specific, and species-specific patterns were also evaluated where sufficient comparable data were available.

All retrieved records were imported into reference-management software, and duplicate records were removed. Titles and abstracts were independently screened by two reviewers (VT and SA), followed by full-text assessment of potentially eligible articles. Disagreements were resolved through discussion and consensus, with consultation with a third reviewer (SJ) when necessary. Reasons for exclusion at the full-text stage were documented, and the overall study selection process was summarized using a PRISMA flow diagram.

Data were independently extracted by two reviewers using a standardized and piloted extraction form. The variables collected included publication details, geographical location and continent, study design, sampling strategy, sample source, Vibrio species, species identification methods, total number of samples tested, total number of Vibrio isolates tested, number of ESBL-positive isolates, phenotypic and molecular ESBL-detection methods, ESBL-associated genes and variants, antimicrobial susceptibility testing (AST) methods, and interpretative standards, including CLSI and EUCAST criteria where reported. A detailed list of the extracted variables is provided in Supplementary File S2. When relevant information was missing or unclear, attempts were made to contact the corresponding authors for clarification.

The methodological quality and risk-of-bias of the included studies were independently assessed by VT and SA using the Joanna Briggs Institute Critical Appraisal Checklist for Studies Reporting Prevalence Data [27]. The checklist assesses the appropriateness of the sampling frame and sampling method, adequacy of sample size, description of study participants and setting, coverage of the identified sample in the analysis, validity of methods used to identify the condition, reliability of measurement, appropriateness of statistical analysis, and adequacy of the response rate or management of incomplete data. Each domain was rated as “yes,” “no,” “unclear,” or “not applicable.”

For descriptive classification, the overall risk-of-bias was categorized according to the number of applicable domains rated “yes”: low risk (7–9 points), moderate risk (4–6 points), and high risk (0–3 points). Risk-of-bias findings were incorporated into sensitivity analyses. Where a meta-analysis included studies classified as having a high risk of bias, the analysis was repeated after excluding these studies. The resulting pooled prevalence estimates, 95% confidence intervals (CIs), and heterogeneity statistics were compared with those from the primary analyses to assess the robustness of the findings. Additional sensitivity analyses excluding studies with small sample sizes were performed where applicable.

A narrative synthesis was initially performed to summarize study characteristics, geographical distribution, Vibrio species, ESBL-associated gene profiles, and AMR patterns. Where sufficiently comparable quantitative data were available, meta-analyses were conducted to generate pooled prevalence estimates.

The denominator was defined according to the outcome being analyzed. For the pooled prevalence of Vibrio spp., the numerator was the number of Vibrio-positive samples and the denominator was the total number of samples tested. ESBL production and AMR were analyzed as isolate-based outcomes. For ESBL prevalence, the numerator was the number of ESBL-positive Vibrio isolates and the denominator was the total number of Vibrio isolates tested for ESBL production. For individual antimicrobial agents, the numerator was the number of resistant Vibrio isolates and the denominator was the number of Vibrio isolates tested against the corresponding antimicrobial agent.

Prevalence proportions were logit-transformed to stabilize variances and pooled using a random-effects model with restricted maximum likelihood (REML) estimation of between-study variance. No continuity correction was applied to studies reporting zero events. Pooled estimates are presented with 95% CIs.

Statistical heterogeneity among studies was assessed using Cochran’s Q test and quantified using the I² statistic. I² values of approximately 25%, 50%, and 75% were interpreted as indicating low, moderate, and high heterogeneity, respectively.

Pre-specified subgroup analyses were conducted to investigate potential sources of heterogeneity. Subgroup meta-analyses by continent were performed for the pooled prevalence of Vibrio spp. and ESBL-producing Vibrio spp., and additional analyses were stratified by Vibrio species. To minimize unstable estimates, subgroup meta-analysis was undertaken only when at least two studies were available within a subgroup and the numerator and denominator were clearly reported. Sensitivity analyses were also conducted by sequentially excluding studies with high prevalence estimates to evaluate the robustness of the pooled results.

Potential publication bias was evaluated separately for each meta-analyzed outcome. Funnel plots, Galbraith plots, and Egger’s regression test were used when an outcome included at least three studies. All statistical analyses were performed using Stata version 19 (StataCorp LLC, College Station, TX, USA). A two-sided p < 0.05 was considered statistically significant.

A total of 1,120 records were identified through PubMed (n = 44), Scopus (n = 80), and Google Scholar (n = 996). After removal of 40 duplicate records, 1,080 records remained for title and abstract screening. Of these, 847 were excluded because they were irrelevant (n = 810) or published in languages other than English (n = 37). Thirteen records could not be retrieved, leaving 220 full-text articles for eligibility assessment.

Of the 220 full-text articles assessed, 189 were excluded for the following reasons: outcomes were irrelevant (n = 48), studies were unrelated to ESBL or Vibrio spp. (n = 99), theses (n = 15), reviews or notes (n = 9), letters or conference papers (n = 9), case reports (n = 5), book chapters (n = 1), short communications (n = 1), and studies involving only one isolate (n = 2). Finally, 31 studies were included in the systematic review and meta-analysis (Figure 1).

The 31 included studies were published between 2012 and 2024. The most were published in 2022 (n = 8), followed by 2023 (n = 4) and 2024 (n = 4). Geographically, studies were concentrated in Asia (n = 19) and Africa (n = 8), with fewer studies conducted in Europe (n = 2), North America (n = 1), and South America (n = 2). Overall, the included studies represented 19 countries. Thailand (n = 5), China (n = 4), and India (n = 3) contributed the most studies, while most other countries were represented by a single study.

Only studies reporting both numerator and denominator data were included in the respective prevalence meta-analyses. Specifically, 11 studies comprising 4,486 samples reported the exact numbers of Vibrio-positive samples and total samples examined and were included in the pooled prevalence analysis of Vibrio spp. (Table 1) [8, 28–57]. Studies contributing to the pooled prevalence analysis of ESBL-producing Vibrio spp. similarly reported the exact number of ESBL-positive Vibrio isolates and the corresponding number of Vibrio isolates tested for ESBL production (Table S2). Among studies with quantitative prevalence data, the number of samples examined varied considerably, ranging from 16 to 2,051.

Among the 4,486 samples contributing to the pooled prevalence analysis, sample sources were grouped into three One Health sectors: animal-derived samples (n = 2,899), human samples (n = 1,301), and environmental samples (n = 286). Animal-derived samples included aquatic animals, seafood, meat, and livestock products, whereas environmental samples primarily comprised water samples. Mixed-source studies were classified according to the source-specific sample numbers reported by the original investigators.

Most studies isolated Vibrio spp. using procedures consistent with the U.S. Food and Drug Administration Bacteriological Analytical Manual, typically involving enrichment in alkaline peptone water followed by selective culture on thiosulfate-citrate-bile salts-sucrose (TCBS) agar. Alternative selective media included chromogenic agar, marine agar, and thiosulfate-taurocholate-gelatin agar (TTGA).

Species identification was performed using one or more confirmatory methods. Conventional biochemical testing, either alone or in combination with commercial identification systems such as Analytical Profile Index, was the most frequently reported approach (n = 15). Polymerase chain reaction (PCR)-based molecular confirmation targeting species-specific genes, including toxR, tl, and hsp60, was reported in nine studies. Matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF) mass spectrometry was used in six studies, whereas VITEK-based automated identification systems were used in two studies. Two studies did not clearly report the species-confirmation method.

For AST, the Kirby–Bauer disk diffusion method was the most frequently used approach (n = 22), whereas agar dilution and broth microdilution were each reported in four studies. Multiple editions of CLSI guidelines were used, with the 2015 edition cited most frequently (n = 5). Two studies applied interpretative criteria based on EUCAST guidelines. The most frequently reported quality-control strain was Escherichia coli American Type Culture Collection (ATCC) 25922. Additional control strains included Staphylococcus aureus ATCC 29213, Pseudomonas aeruginosa ATCC 27853, and Klebsiella pneumoniae ATCC 700603. However, 17 studies did not report the quality-control strains used.

The Vibrio species identified among the included studies were diverse (Table S2). The most frequently reported species was V. parahaemolyticus (n = 10 studies), followed by V. cholerae (n = 9), unspecified Vibrio spp. (n = 5), V. vulnificus (n = 3), and Vibrio alginolyticus (n = 3). Less frequently reported species included Vibrio diabolicus, Vibrio mimicus, Vibrio fluvialis, Vibrio navarrensis, and Vibrio scophthalmi.

Of the 31 included studies, 15 were classified as having a low risk-of-bias and 16 as having a moderate risk-of-bias; none were classified as having a high overall risk-of-bias (Table S3). At the domain-level, favorable ratings were observed for an appropriate sampling frame (Q1; 83.9%), adequate description of study participants and settings (Q4; 67.7%), adequate coverage of the identified sample in the analysis (Q5; 100%), valid methods for identification of the condition (Q6; 96.8%), reliable outcome measurement (Q7; 93.5%), and appropriate statistical analysis (Q8; 100%) (Figure 2).

In contrast, high risk ratings were more frequent for sampling method (Q2; 58.1%) and sample size adequacy (Q3; 45.2%), while 12.9% of studies had an unclear risk-of-bias for sample size adequacy. The response-rate domain (Q9) was considered not applicable to all included studies. Moderate overall risk-of-bias classifications were therefore primarily associated with limitations in sampling strategy, sample size adequacy, and representativeness.

The pooled prevalence of Vibrio spp. was estimated from 11 studies, comprising 4,486 samples. The overall pooled prevalence was 48.3% (95% CI: 34.0–62.7%), with substantial between-study heterogeneity (I² = 99.37%, p < 0.001; Figure 3A).

Geographically, the pooled prevalence was higher in Asia at 56.7% (95% CI: 40.9–72.5%) than in Africa at 21.8% (95% CI: 6.1–37.6%). Considerable heterogeneity persisted within both continental subgroups, with I² values of 98.72% for Asia and 97.58% for Africa. The test for subgroup differences was statistically significant (Qb = 9.39, p < 0.001), indicating significant variation in pooled prevalence between continents.

The funnel plot showed an asymmetric distribution of study estimates around the pooled effect estimate, suggesting possible publication bias or small-study effects (Figure 3B). Several studies with relatively large standard errors were distributed away from the central estimate, and some estimates fell outside the pseudo 95% confidence limits. Thus, smaller studies appeared to show greater variability in prevalence estimates.

Leave-one-out sensitivity analysis showed that exclusion of Yano et al. [54] produced no meaningful change in pooled prevalence, which remained at 48.3% (Supplementary File S4). The τ² value decreased slightly from 0.087 to 0.082, while I² remained essentially unchanged (99.37% vs. 99.4%), indicating that this study did not substantially influence the overall pooled estimate.

Based on five studies comprising 388 isolates, the pooled prevalence of ESBL-producing Vibrio spp. was 19.4% (95% CI: 3.2–43.7%) using a random-effects model (Figure 4A). Study-specific estimates ranged from 0.0% to 74.1%, with the highest prevalence reported by Narayanan et al. [45], whereas two datasets reported no ESBL-producing Vibrio isolates. Because of the small number of contributing studies and their geographical concentration, including three studies from Asia, continent-specific subgroup analysis was not performed.

Substantial heterogeneity was observed, with τ² = 0.46, I² = 95.87%, and H² = 24.20. Cochran's Q test indicated statistically significant heterogeneity (Q(6) = 119.96, p < 0.001). The overall effect was also statistically significant (z = 2.93, p < 0.01).

The Galbraith plot (Figure 4B) showed broad dispersion of study estimates around the regression line, further demonstrating substantial between-study variability. However, all study estimates remained within the 95% confidence bounds, indicating no extreme outliers.

Sensitivity analysis restricted to phenotypically confirmed ESBL-producing Vibrio isolates included three studies comprising four species-level datasets and 212 isolates tested using the combined-disk method or double-disk synergy test. The pooled prevalence decreased to 12.5% (95% CI: 3.9–33.3%) compared with the overall estimate of 19.4% (95% CI: 3.2–43.7%).

AST covered multiple antimicrobial classes, including aminoglycosides, β-lactams with or without β-lactamase inhibitors, carbapenems, macrolides, quinolones, tetracyclines, polymyxins, sulfonamides, phenicols, and nitrofurans (Table S4).

Among β-lactams, pooled resistance estimates were highest for ampicillin/sulbactam (74.5%, 95% CI: 37.5–98.3%), ampicillin (73.6%, 95% CI: 58.4–86.4%), and amoxicillin (60.9%, 95% CI: 12.0–98.8%). Resistance to amoxicillin/clavulanic acid was substantially lower at 8.3% (95% CI: 2.2–17.0%). Among cephalosporins, cefuroxime showed the highest pooled resistance estimate at 51.1% (95% CI: 15.8–85.8%), whereas overall pooled resistance to cefotaxime, cefepime, ceftazidime, and ceftriaxone remained below 20%.

Among aminoglycosides, pooled resistance was highest for kanamycin (47.8%, 95% CI: 24.2–71.8%) and streptomycin (32.6%, 95% CI: 10.0–60.6%), whereas lower estimates were observed for amikacin (8.6%, 95% CI: 1.6–19.3%) and gentamicin (1.6%, 95% CI: 0–5.2%). For macrolides, pooled resistance was 28.5% (95% CI: 3.9–63.0%) for azithromycin and 23.0% (95% CI: 6.1–45.9%) for erythromycin.

Overall resistance to quinolones was comparatively low, including ciprofloxacin (5.3%, 95% CI: 2.0–9.9%), norfloxacin (2.8%, 95% CI: 0–9.1%), and ofloxacin (13.5%, 95% CI: 3.4–27.9%). No levofloxacin resistance was detected in the available datasets. Among tetracyclines, pooled resistance was 8.0% (95% CI: 2.7–15.3%) for tetracycline and 4.1% (95% CI: 0–13.1%) for oxytetracycline, while no minocycline resistance was detected. Nitrofurantoin resistance was estimated at 79.6% (95% CI: 34.2–100.0%) based on African studies only. Among polymyxins, pooled resistance was 66.4% (95% CI: 7.7–100.0%) for polymyxin B and 57.7% (95% CI: 15.2–94.4%) for colistin.

The overall pooled MDR prevalence was 31.5% (95% CI: 6.1–64.7%). Continent-specific estimates were 100.0% (95% CI: 95.6–100.0%) in Africa and 23.3% (95% CI: 3.0–53.7%) in Asia, although the number of contributing studies differed among strata.

Among β-lactamase resistance determinants, bla_TEM was the most frequently detected gene, with a pooled prevalence of 33.8% (95% CI: 7.0–67.4%). Carbapenemase-associated genes were also detected, including bla_VIM (15.5%, 95% CI: 0.7–41.7%) and bla_IMP (9.1%, 95% CI: 4.3–15.2%). Pooled prevalence estimates were lower for blaOXA-48 (4.4%, 95% CI: 1.3–8.9%), bla_KPC (1.7%, 95% CI: 0–10.4%), and bla_SHV (1.6%, 95% CI: 0–9.2%). Among non-β-lactam resistance determinants, aadA was detected in 18.7% of isolates (95% CI: 2.7–43.1%), whereas strA and sul2 were each detected in 4.9% (95% CI: 0.6–12.0%). The pooled prevalence of tetA was 6.8% (95% CI: 2.2–13.1%).

Virulence-associated genes were also identified. Their pooled prevalence estimates were 25.8% (95% CI: 0–99.9%) for ctxA, 26.3% (95% CI: 0–99.9%) for tcpA, 11.5% (95% CI: 0–57.6%) for tdh, and 0.3% (95% CI: 0–2.1%) for trh. Substantial between-study heterogeneity was observed for many phenotypic and genotypic outcomes (Table S4).

Resistance patterns differed markedly among Vibrio species (Table S5). Ampicillin showed an overall pooled resistance prevalence of 73.6% (95% CI: 58.4–86.4%). Species-specific resistance was particularly high in V. alginolyticus (97.8%, n = 68), V. mimicus (95.0%, n = 40), V. fluvialis (87.8%, n = 41), and V. parahaemolyticus (84.1%, n = 974) (Figure 5), whereas the pooled estimate for V. cholerae was lower at 24.1% (n = 705).

Resistance to amoxicillin/clavulanic acid remained low overall (8.3%, 95% CI: 2.2–17.0%). Species-specific estimates were higher in V. navarrensis (17.4%, n = 46) and V. parahaemolyticus (16.3%, n = 203; 9.4%, n = 64), whereas no resistance was reported in V. alginolyticus (n = 21), V. cholerae (n = 20), or V. vulnificus (n = 31).

Among third- and fourth-generation cephalosporins, overall pooled resistance to cefotaxime was 16.2% (95% CI: 5.9–29.9%; n = 2,273). Higher species-specific estimates were observed for unspecified Vibrio spp. (36.6%, n = 643) and V. scophthalmi (57.6%, n = 33). Overall ceftazidime resistance was 7.2% (95% CI: 1.7–15.6%), although resistance reached 82.1% in V. cholerae O1 (n = 39). Ceftriaxone resistance was 13.2% overall (95% CI: 0.6–35.5%), with higher estimates in unspecified Vibrio spp. (62.3%, n = 59) and V. scophthalmi (33.3%, n = 33), whereas resistance in V. parahaemolyticus was 0.4% (n = 214).

Macrolide resistance also varied among species. Azithromycin resistance was 28.5% overall (95% CI: 3.9–63.0%) and reached 100.0% in V. fluvialis, V. mimicus, and V. vulnificus. Pooled erythromycin resistance was 23.0% (95% CI: 6.1–45.9%), with a higher estimate in V. alginolyticus (76.9%).

For quinolones, pooled ciprofloxacin resistance was 5.3% (95% CI: 2.0–9.9%), with higher estimates in V. alginolyticus (34.1%) and V. fluvialis (26.8%) than in V. cholerae (0.8%) and unspecified Vibrio spp. (2.0%). Nalidixic acid resistance was 31.7% (95% CI: 7.4–62.8%) overall and ranged from 95.2%–100.0% in V. cholerae O1 to 0.4% in V. parahaemolyticus. No levofloxacin resistance was detected (0%, 95% CI: 0–0.4%).

Among polymyxins, colistin resistance was 57.7% (95% CI: 15.2–94.4%), including 93.5% in V. vulnificus and 74.1% in V. diabolicus, compared with 7.0% in unspecified Vibrio spp. Polymyxin B resistance was 66.4% (95% CI: 7.7–100.0%) overall and reached 100.0% in V. fluvialis, V. mimicus, and V. vulnificus, whereas no resistance was detected in V. cholerae.

The overall pooled prevalence of MDR was 31.5% (95% CI: 6.1–64.7%). No MDR was detected in V. alginolyticus (0%, n = 21) or V. navarrensis (0%, n = 23), whereas MDR prevalence reached 100.0% in two V. cholerae O1 datasets (n = 39 and n = 22) and ranged from 18.3% to 28.1% in V. parahaemolyticus.

The prevalence of resistance-associated genes also varied among species. bla_TEM had an overall pooled prevalence of 33.8% (95% CI: 7.0–67.4%), with high estimates in V. fluvialis (85.4%) and V. mimicus (95.0%), whereas one V. parahaemolyticus stratum reported no detection. blaOXA-48 remained infrequent overall (4.4%, 95% CI: 1.3–8.9%), while bla_PER was not detected. Among other resistance determinants, strA and sul2 each had an overall pooled prevalence of 4.9%, and tetA was detected in 6.8% of isolates (95% CI: 2.2–13.1%).

This systematic review and meta-analysis provides a quantitative One Health synthesis of Vibrio spp. occurrence across human, animal, and environmental sources, extending previous evidence that was more restricted in scope and highlighting substantial geographic and species-level heterogeneity. The detection of Vibrio spp. across all three One Health sectors emphasizes their importance at the interface of aquatic ecosystems, food production, animal health, and human health. Most samples originated from animal sources, particularly aquatic animals, consistent with the close ecological association of Vibrio spp. with aquatic environments and the potential role of seafood and other animal-derived products in human exposure [58, 59]. The inclusion of human isolates further demonstrates the clinical relevance of antimicrobial-resistant Vibrio spp. [59, 60]. In contrast, the comparatively limited representation of environmental samples identifies an important surveillance gap and supports more balanced monitoring across human, animal, and environmental sectors.

The pooled prevalence of Vibrio spp. was 48.3%, although substantial between-study heterogeneity was observed (I² = 99.37%, p < 0.001). This heterogeneity likely reflects differences in geographic region, sample source, aquatic production systems, target Vibrio species, sampling periods, environmental conditions, and laboratory detection methods among the included studies. Prevalence differed significantly by continent, with a higher pooled estimate in Asia than in Africa. The overall pooled estimate was slightly higher than previously reported species-specific prevalence estimates, including 46.0% for V. parahaemolyticus in shrimp and 10.5% for V. vulnificus in seafood in Asia [3, 61]. These differences may partly reflect the broader scope of the present analysis, which combined multiple Vibrio species and heterogeneous sample sources rather than focusing on a single species or commodity.

Accordingly, the pooled prevalence should be interpreted as a descriptive summary of the available evidence rather than as a universally representative estimate of Vibrio occurrence. Greater emphasis should be placed on study-specific and subgroup estimates because of the substantial heterogeneity and geographic imbalance in the evidence base. The leave-one-out sensitivity analysis showed that no single influential study drove the pooled prevalence. Although funnel plot asymmetry suggested possible publication bias or small-study effects, persistent heterogeneity after sensitivity analysis indicates that the asymmetry may also reflect genuine differences among study populations and methodologies. Publication bias therefore cannot be excluded, but its interpretation should remain cautious. Future studies using standardized sampling frameworks, species identification procedures, and microbiological methods would improve comparability among geographic regions and strengthen pooled prevalence estimates.

The pooled prevalence of ESBL-producing Vibrio spp. was 19.4% (95% CI: 3.2–43.7%), indicating that ESBL-associated resistance occurs in a notable proportion of the isolates represented in the available literature. This finding is important because Vibrio spp. have historically been regarded as comparatively susceptible to several extended-spectrum cephalosporins, whereas increasing detection of acquired β-lactam resistance determinants may compromise this susceptibility [7, 20, 21]. Nevertheless, the estimate should not be interpreted as evidence that approximately one-fifth of all Vibrio isolates globally are ESBL producers because the evidence base was limited and highly heterogeneous.

Substantial heterogeneity was observed for ESBL prevalence (I² = 95.87%), indicating pronounced variation among studies. The pooled estimate therefore represents a context-dependent summary rather than a stable global prevalence. Differences in geographic setting, sample source, antimicrobial use practices, environmental contamination, study population, AST methodology, and criteria for defining ESBL production may contribute to this variation [14, 21, 62]. The concentration of available studies in Asia and parts of Africa further limits global representativeness.

The sensitivity analysis provided additional evidence of the influence of ESBL classification methods. When the meta-analysis was restricted to phenotypically confirmed ESBL-producing Vibrio isolates, pooled prevalence decreased from 19.4% to 12.5%. This finding suggests that inclusion of molecularly detected ESBL-associated genes without corresponding phenotypic confirmation can influence pooled prevalence estimates. Detection of an ESBL-associated gene does not necessarily establish phenotypic expression under the testing conditions used. Consequently, consider differences in ESBL definitions, confirmatory procedures, and reporting practices when comparing prevalence estimates across studies.

The geographic concentration of studies in Asia and Africa may reflect, at least in part, the importance of aquaculture intensity, urbanization, wastewater discharge, and other anthropogenic pressures in shaping aquatic AMR epidemiology [16, 18, 63]. Aquatic environments exposed to antimicrobial residues and resistant bacteria can provide ecological opportunities for persistence and horizontal transfer of resistance determinants. Mobile genetic elements, including plasmids, integrons, and SXT/R391-related elements, have been implicated in the dissemination of AMR among aquatic bacteria [18, 22, 64].

Among the resistance determinants evaluated, bla_TEM showed the highest pooled prevalence among the reported β-lactamase genes at 33.8%. Other clinically important determinants, including bla_VIM, bla_IMP, bla_OXA-48, bla_KPC, and bla_SHV, were also identified, although generally at lower pooled prevalences. These findings support the potential role of aquatic Vibrio populations as reservoirs of clinically relevant resistance determinants. The detection of resistant Vibrio spp. in aquatic animals, seafood, environmental samples, and human clinical sources further demonstrates the interconnected nature of resistance across the One Health interface [15, 23, 58].

The present meta-analysis also identified virulence-associated genes, including ctxA, tcpA, tdh, and trh. Their occurrence alongside resistance determinants in the overall evidence base is of concern because pathogenicity and AMR can coexist within Vibrio populations. However, the present pooled analyses do not establish that resistance and virulence genes were necessarily carried by the same isolates unless the original studies explicitly demonstrated this. Therefore, the evidence should not be interpreted as proving the widespread emergence of simultaneously hypervirulent and multidrug-resistant lineages.

The ecological relationships among antimicrobial exposure, mobile genetic elements, aquatic microbial communities, food production, and human exposure are summarized conceptually in Figure 6.

The phenotypic resistance profile varied markedly across antimicrobial classes. Ampicillin resistance was particularly high, with a pooled estimate of 73.6%. This finding is consistent with the frequent occurrence of β-lactam resistance among Vibrio spp. and may reflect both intrinsic and acquired resistance mechanisms [7, 8]. The pooled prevalence of ESBL-producing isolates and detection of ESBL-associated genes indicate that acquired β-lactam resistance mechanisms also contribute to the resistance landscape.

Resistance to cefuroxime was also relatively high at 51.1%, whereas overall resistance to cefotaxime, cefepime, ceftazidime, and ceftriaxone remained below 20%. These differences among β-lactam agents show that resistance cannot be generalized across an entire antimicrobial class and should be interpreted by individual antimicrobial agent and Vibrio species.

In contrast, overall ciprofloxacin resistance was comparatively low at 5.3%. Resistance to other quinolones was also generally lower than resistance to several β-lactams. However, the included studies did not directly evaluate the mechanisms behind these differences as part of the meta-analysis. Therefore, the observed disparity should not be attributed solely to differences between chromosomal mutation and horizontal gene transfer without supporting molecular evidence.

The pooled MDR prevalence was 31.5%, although substantial regional variation was observed. The African estimate was particularly high; however, this result should be interpreted cautiously because of the limited number of contributing studies and the wide differences in sampling and study characteristics. Regional differences in antimicrobial use, wastewater management, aquaculture practices, environmental contamination, and surveillance intensity may contribute to variation in MDR occurrence [8, 21].

The detection of carbapenemase-associated genes such as bla_VIM and bla_IMP is particularly noteworthy because these determinants are clinically important and may circulate through mobile genetic elements [17, 24]. Their identification in aquatic Vibrio spp. supports continued monitoring of aquatic ecosystems and seafood production systems as potential reservoirs and transmission pathways for clinically important AMR determinants [15, 23]. Nevertheless, the present findings demonstrate occurrence rather than direct transmission of these genes between environmental, animal, and human populations.

Marked interspecies variation was observed across multiple antimicrobial classes, emphasizing the limitations of genus-level generalization. Ampicillin resistance was particularly high in V. alginolyticus, V. mimicus, V. fluvialis, and V. parahaemolyticus, whereas the pooled estimate for V. cholerae was substantially lower. Similarly, ceftazidime resistance reached 82.1% in V. cholerae O1 in one dataset, whereas resistance remained much lower in V. parahaemolyticus. Cefotaxime and ceftriaxone resistance were also comparatively high in V. scophthalmi relative to several other species.

Species-specific differences were also evident for macrolides, quinolones, and polymyxins. For example, azithromycin resistance reached 100% in some datasets for V. fluvialis, V. mimicus, and V. vulnificus, whereas pooled ciprofloxacin resistance remained low overall but was substantially higher in V. alginolyticus and V. fluvialis. These patterns indicate that AMR within the genus Vibrio is highly species dependent and may additionally vary according to geographic setting, source population, and antimicrobial exposure.

These findings support species-level AMR surveillance rather than reliance solely on genus-level summaries. From veterinary and public health perspectives, improved species identification, standardized AST, and molecular characterization would provide more informative data for antimicrobial stewardship, aquaculture biosecurity, food safety surveillance, and clinical risk assessment [11, 15, 17].

This systematic review and meta-analysis integrates epidemiological, phenotypic, and genotypic evidence on ESBL-producing and antimicrobial-resistant Vibrio spp. across human, animal, food, and environmental sources. The pooled ESBL prevalence of 19.4%, high resistance to several β-lactam agents, overall MDR prevalence of 31.5%, and detection of clinically important resistance determinants demonstrate that AMR in Vibrio spp. warrants continued attention within a One Health framework.

The findings have practical implications for both clinical and veterinary settings. Empirical antimicrobial treatment of vibriosis should account for local susceptibility patterns, and AST-guided therapy should be used where feasible. In aquaculture, antimicrobial stewardship should emphasize appropriate antimicrobial selection, avoiding unnecessary exposure, biosecurity, disease prevention, and resistance surveillance within production systems.

A major contribution of the present study is the demonstration that resistance patterns are highly heterogeneous across geographic regions and Vibrio species. This heterogeneity indicates that a single global pooled estimate cannot adequately describe the complexity of AMR across aquatic systems. Surveillance programs should therefore integrate species-level identification, standardized phenotypic testing, molecular detection of resistance determinants, and source-specific epidemiological information.

The One Health perspective matters because aquatic environments connect antimicrobial use, wastewater contamination, food production, animal populations, and human exposure. The available evidence supports strengthening coordinated surveillance across these interfaces rather than monitoring clinical isolates in isolation. Future surveillance incorporating genomic approaches could further clarify the relatedness of resistant isolates, mobile genetic elements, and potential transmission pathways among environmental, animal, food, and human reservoirs.

This study provides a comprehensive synthesis of the available evidence on ESBL-producing Vibrio spp. across environmental, seafood, aquaculture, animal, and clinical sources. A major strength is integrating phenotypic AMR profiles with genotypic data on ESBL-associated and other resistance determinants, enabling a broader assessment of resistance characteristics across diverse Vibrio species and source populations. The study also evaluates species-specific AMR patterns and geographic variation, which are important for identifying areas and taxa in which resistance may be more frequently reported.

Another strength is the application of a One Health framework to integrate evidence across aquatic environments, food production systems, animals, and humans. This approach provides a more comprehensive perspective on the epidemiology of antimicrobial-resistant Vibrio spp. than analyses restricted to a single host or environmental compartment. In addition, subgroup analyses, sensitivity analyses, risk-of-bias assessment, and separate evaluation of phenotypic and genotypic outcomes strengthen interpretation of the pooled estimates and highlight how study design, geographic setting, species distribution, and ESBL-definition criteria may influence the findings.

This study has several limitations. First, substantial heterogeneity was observed across many pooled outcomes. This heterogeneity may reflect differences in geographic location, sample source, target Vibrio species, sampling strategy, laboratory detection procedures, AST methods, and interpretative criteria. In particular, antimicrobial susceptibility classification may have varied across studies because different editions of CLSI and EUCAST guidelines, as well as human- or veterinary-oriented interpretative standards, were used. Such methodological variability limits direct comparability among studies and may contribute to uncertainty in pooled estimates.

Second, only English-language publications were included, which may have introduced language bias and excluded relevant evidence published in other languages. Third, although multiple information sources were searched, the inclusion of Google Scholar may have introduced additional variability because its indexing, ranking, and reproducibility differ from those of conventional bibliographic databases.

Fourth, geographic representation was uneven. Most available evidence originated from Asia and Africa, whereas Europe, North America, South America, and other regions were comparatively underrepresented. Consequently, regional pooled estimates should not be interpreted as equally representative across continents, and the overall findings should not be generalized uncritically to regions with limited data.

Fifth, selective reporting of resistant isolates, specific antimicrobial agents, ESBL-associated genes, or statistically notable findings may have influenced the available evidence base and could have contributed to overestimation of some pooled resistance or prevalence estimates. Formal assessment of publication bias was also constrained by the small number of contributing studies for several outcomes, reducing the reliability of funnel plot interpretation and the statistical power of Egger's regression test. Potential selective reporting therefore had to be considered qualitatively when interpreting the findings.

Finally, the included studies differed in their definitions and confirmation methods for ESBL production. Some studies relied on phenotypic confirmation, whereas others reported molecular detection of ESBL-associated genes. Because the presence of a resistance-associated gene does not necessarily indicate phenotypic expression, combining these approaches may have influenced pooled prevalence estimates. This concern was partly addressed through sensitivity analysis restricted to phenotypically confirmed ESBL-producing isolates, but methodological heterogeneity remains an important limitation. Overall, the findings should therefore be interpreted cautiously, particularly when making subgroup comparisons or extrapolating the pooled estimates to a global population.