Ethics

Study Design and Patients

We designed a retrospective multicenter study of the prevalence of infections caused by Pasteurella spp. combined with a meta-analysis to determine the burden of these infections, especially bacteremia. We obtained data for Pasteurella species infections during 2018‒2022 from 7 university hospitals in metropolitan areas (4 in Seoul and 3 in Gyeonggi-do) and 1 reference laboratory, where samples from general and small- and medium-sized hospitals were tested, in South Korea to investigate the overall burden of these infections throughout the country. Patients were included if they had a microbiological examination that was positive for Pasteurella species. We identified isolated species by matrix-assisted laser desorption/ionization time-of-flight mass spectrometry on a Vitek-MS instrument (bioMérieux) or Bruker instrument (Bruker Daltonik), Vitek 2 system (bioMérieux), and the MicroScan Walkaway-96 system (Siemens). We obtained the following clinical variables from patient charts: basic demographics, hospital region of origin, sampling year, isolation site, the presence of polymicrobial infection, any animal contact such as bite or scratch history, and the antibiotics and therapies used. We also collected data for hospitalization and outcomes. From the 316 participants, we excluded 31 whose records lacked basic demographic information, such as age and sex (Figure 1). We included a total of 283 patients after excluding 2 patients whose Pasteurella species were isolated in 2017 and 2023. We obtained only basic data—age, sex, sampling year, isolation site, and region—for 213 participants from the reference laboratory.

Search Strategy and Selection Criteria for Meta-analysis

We performed a systematic review as described in the Cochrane handbook (16) to estimate the global prevalence of bacteremia caused by Pasteurella spp. We referred to the Preferred Reporting Items for Systematic Review and Meta-analysis (17,18) checklist (Appendix Table 1). We included studies that reported prevalence data for patients with Pasteurella species infection and bacteremia based on laboratory results. We included observational cohort studies, regardless of language or publication year. We excluded studies without the necessary data for the calculation of the prevalence of bacteremia (Appendix Table 2).

We performed a comprehensive search of PubMed, Ovid-EMBASE, and the Cochrane Library for articles published through November 1, 2023. The search strategy included use of the keywords “Pasteurella infections,” “bacteremia,” “prevalence,” and “epidemiology.” We included Medical Subject Heading and Emtree terms, text words, and equivalent subject heading and thesaurus terms to ensure inclusivity; we also performed manual searches of the references of relevant articles for completeness (Appendix Table 3). We registered the protocol in the international prospective register for systematic reviews (registration no. CRD42023484039).

Analysis of the Study Population Data and Meta-analysis

For statistical analysis of data from multiple centers in South Korea, we used the Mann-Whitney U test or Pearson’s χ² test to compare groups. We applied multivariate binary logistic regression analyses to investigate variables that correlated independently with the occurrence of bacteremia in patients with Pasteurella spp. infections.

For the meta-analysis, we conducted title and abstract screening of studies selected through the search strategy on the basis of the eligibility criteria (Figure 2). Two reviewers (E.J. and N.L.) independently assessed the full texts of the studies. We settled disagreements by consensus after all reviewers reviewed the data. We extracted the following variables if they were available: demographic information about the study population, collection periods, the presence of animal exposure, the identified species, regions, and outcome measures. We documented he data for hospitalization and death, as well as the prevalence of bacteremia, our primary outcomes. We used the Joanna Briggs Institute checklist (19) to evaluate the quality of the included articles at the study level. We considered scores >70% as high quality. Two reviewers (E.J. and N.L.) assessed the quality of the included studies. S.J. resolved any disagreements.

We calculated the proportion of patients with positive blood cultures for the determining the rate of bacteremia among the total number of patients with bacterial infections. We based the calculation on all types of samples for which culture tests were requested (3,4,10,20). Although simultaneous blood culture is necessary for a more accurate determination of the rates of bacteremia, it was not routinely performed. Blood culture was performed only in 35/283 cases in which the clinician deemed it necessary. We transformed single raw prevalence using the Freeman-Tukey Double arcsine method (21) to stabilize variances. We used the random effects model to calculate the pooled prevalence with 95% CI across studies.

We assessed heterogeneity using Cochran’s Q test and the degree of heterogeneity using the Higgins I² statistic. Indices with values of >75% represented high heterogeneity (22). We performed subgroup analyses to explore potential sources of heterogeneity by study location and period. In addition, we conducted sensitivity analysis to further assess the robustness of the estimates. The software we used for those analyses was the moonBook package in R (The R Project for Statistical Computing), MedCalc software, version 19.8 (MedCalc Software Ltd), Analyze-it Method Evaluation Edition software version 2.26 (Analyze-it Software Ltd), and Stata version 18 (StataCorp LLC). We have deposited the raw data used in this study (Appendix Tables 4, 5) in the Harvard Dataverse (https://doi.org/10.7910/DVN/1QQ9KK).

Prevalence and Clinical Features of Pasteurella Infections

We included a total of 283 cases in the study, 70 from hospital patients with complete data and 213 cases from the main reference laboratory with basic information. We observed an increase in the number of infections caused by Pasteurella spp. from 2018 (n = 46) to 2022 (n = 72) (Figure 3); the increase was significant on the basis of the national population data extracted from the Korean Statistical Information Service (p = 0.012). The median number of cases per year was 55. The predominant species were P. multocida (68.9%, 195/283) and P. canis (25.4%, 72/283), both of which contributed to the increase in infections. The median age of patients with positive culture results was 52.0 years. The number of Pasteurella isolation samples was higher in women (66.4%, 188/283) than in men (33.6%, 95/283). We observed the most cases in patients in the 50‒59-year age group (Figure 3). The ratio of female to male patients was the highest (3.4:1) for patients 20‒29 years of age. Most patients had a history of companion animal exposure (88.6%, 62/70) (Appendix Table 6); half had history of dog exposure and 38.7% (24/62) cat exposure. The rate of polymicrobial infection was 25.7% (18/70). The detected isolates were Staphylococcus aureus (n = 2), Actinomyces spp. (n = 2), Streptococcus sanguinis (n = 1), Proteus mirabilis (n = 1), and Dermabacter hominis (n = 1). The rate of hospitalization was 54.3% (38/70). We noted no significant differences in characteristics between inpatients and outpatients (Appendix Table 7). Among all the patients, most patients were mainly from Seoul (23.0%, 65/283), whereas those included in the reference laboratory study were predominantly from Gyeongsang-do (33.0%, 71/213) and Gyeonggi-do (17.4%, 37/213) (Appendix Figure 1). All patients, except for 3 outpatients, had received antimicrobial therapy. The most frequently prescribed antibiotics were penicillin/β-lactamase inhibitor combinations and first-generation cephalosporins (Appendix Table 4).

The rates of bacteremia were 2.8% (8/283) among all included infections and 7.1% (5/70) among the 7 university-hospital cases (Appendix Table 8). The median age was higher in bacteremia (68.5 years) than that in nonbacteremia (52.0 years). Patients with bacteremia had no animal exposure history. According to the multivariate regression analyses, including sex as the confounder, only increasing age was a significant risk factor for bacteremia (odds ratio 1.05, 95% CI 1.01–1.10; p = 0.024) (Figure 4). Among all bacteremia cases, 1 patient with septic shock caused by P. multocida died after 4 days of hospitalization; that patient had alcoholic liver cirrhosis and asthma as underlying diseases.

Meta-analyses for Global Burden of Pasteurella Bacteremia

The study screening method for the meta-analysis (Figure 2) identified 106 studies, of which 96 articles were relevant and were subsequently screened. A total of 31 reports were available for full-text screening. We excluded papers with insufficient study populations and design and insufficient data for the calculation of the prevalence of Pasteurella spp. bacteremia from the current review (Appendix Table 2). Finally, we included 10 studies in the systematic review (3,4,6,7,10,20,23–26).

We included a total of 2,012 participants from the 10 studies in the meta-analysis (Table). The studies were published during 1985‒2021, half before 2010 (4,10,20,25,26) and half after 2010 (3,6,7,23,24). We observed more than 50% of study participants were female in all the included studies. The occurrence of animal exposure calculated from available study data was 34.3% ‒97.3%. The most commonly isolated species were P. multocida and P. canis. The rates of hospitalization were higher in cases of invasive infection (83.3% [23] and 97.0% [24]) than those in other cases (3,6,7,20). The death rates range was 0.6%‒27.2%. Six studies were conducted in Europe and the United States (3,7,10,20,23–26). Based on the assessment using the Joanna Briggs Institute checklist, all studies showed scores >70%, indicating that they were of high quality. Four studies that included only P. multocida infections had lower scores in the samples frame appropriate to address the target population (3,4,20,25).

The prevalence of bacteremia caused by Pasteurella spp. from 10 studies was 3.4%‒32.5%. The random effects pooled prevalence from the 2,012 cases was 12.4% (95% CI 7.3%–18.6%). The Cochran Q test revealed significant heterogeneity (Q = 52.1, p<0.001). The I² index of the included studies indicated high heterogeneity (I² = 90.7) (Figure 5).

We calculated the pooled prevalence estimates of Pasteurella infection stratified by publication year. The pooled value of 5 studies published before 2010 (4,10,20,25,26) was 12.7% (95% CI 4.1%–24.6%) and of 5 studies published after 2010 (3,6,7,23,24) was 12.1% (95% CI 6.6%–18.9%). Although the I² index for studies published after 2010 (78.9%) was lower than that for studies before 2010 (94.4%), we noted no significant differences in pooled prevalence and heterogeneity. In a subgroup analysis of different regions, the estimated prevalence infection in 6 studies conducted in Europe (9.8%, 95% CI 4.5%–16.6%) was not different from that of 2 studies conducted in the United States (11.6%, 95% CI 3.3%–23.6%). We excluded 1 of the 10 studies from the sensitivity analyses. We compared estimated prevalence to the estimated total prevalence (Appendix Table 9). The pooled estimates from the sensitivity analyses were 9.9%‒13.8%, which was consistent with the total pooled prevalence without statistical differences. The pooled death rate was 8.4% (95% CI 2.7%–16.5%) (Appendix Figure 2).