A common finding, but not a universal rate
The review pooled results from 10 studies and found 3,699 positive cultures among 5,147 sampled people. That produced a patient-level bile-culture yield of 77.3%, meaning the share of sampled people whose ductal-bile culture was positive in the included data. It is a sampling yield, not an estimate of how common clinical biliary infection is.
The estimate came with a 95% confidence interval of 67.8% to 84.6%. A 95% prediction interval, which shows how widely results may vary in another study, ran from 36.0% to 95.4%. The reported between-study heterogeneity was 97.7%, a sign that the studies did not produce closely similar results.
An exploratory, post hoc comparison put the yield at 82.1% in clinically defined infection reports and 63.3% in broader or unselected sampling. The interaction P value was .014, but the comparison was post hoc and heterogeneous, so it should be read as a clue for further study rather than a settled difference.
The cultures were dominated by Gram-negative bacteria
At the isolate level, Gram-negative bacteria made up 69.8% of isolates. E. coli accounted for 27.4% and K. pneumoniae for 14.1%. The corresponding 95% confidence intervals were 65.6% to 73.7%, 23.5% to 31.7%, and 12.3% to 16.2%.
Those proportions also varied substantially between studies. Heterogeneity was 94.9% for Gram-negative bacteria, 96.5% for E. coli and 86.0% for K. pneumoniae. The isolate totals were heavily influenced by the Li et al. dataset, which contributed 25,573 of 35,267 source-resolved organism observations, or 72.5%, and had no patient denominator. Because report denominators were not comparable, the review did not calculate one combined patient total.
Cultures described as polymicrobial, meaning they contained more than one organism, were estimated at 43.7% across 12 studies, with a 95% confidence interval of 26.4% to 62.7%. The result was highly unstable across studies: heterogeneity was 99.2%, and the 95% prediction interval ran from 3.2% to 94.7%. When the analysis was limited to studies reporting exact counts, the estimate was 32.7% across seven studies. These were descriptive estimates, not a universal prevalence.
Resistance could not be reduced to one number
The review could not combine the resistance findings into a valid overall estimate. The reports used different organism denominators, drug panels, laboratory breakpoints, units based on patients or isolates, and definitions of multidrug resistance. No reproducible estimate was available for extensively drug-resistant, or XDR, findings.
Some individual results illustrate the problem without resolving it. The ESBL resistance phenotype was reported in 12 of 85 E. coli isolates, or 14.1%, and in 3 of 47 K. pneumoniae isolates, or 6.4%. Multidrug-resistant pathogens were reported in 24 of 83 patients, or 28.9%. These figures used different denominators and cannot be treated as one pooled resistance rate.
Why the result needs careful reading
The evidence base was 97% retrospective and 3% prospective. The 16 included reports came from nine countries and were published from 2015 to 2025. This was an observational evidence base drawn from separate reports, not a single patient population.
Selection limits matter. The search screened 2,551 records. Of 66 reports sought, 18 could not be retrieved; 48 were assessed in full, 32 were excluded, and 16 were included. Eligible evidence had to come from a retrievable open-access full text, a restriction that leaves room for selection bias.
Fourteen reports received low-risk prevalence ratings and two moderate-risk ratings. Even so, the appraisal did not remove review-level bias linked to the single-database, open-access selection. Retrospective single-centre sampling and incomplete coverage reporting were common issues.
One robustness check removed the Li dataset. The estimates changed only modestly: Gram-negative bacteria moved from 69.8% to 70.2%, E. coli from 27.4% to 26.9%, and K. pneumoniae from 14.1% to 14.8%. These checks were not significance tests, and substantial heterogeneity remained when moderate-risk reports were excluded.
The review followed PRISMA 2020 and was registered retrospectively on the Open Science Framework under OSF 89wqs. It reported no funding.
The central message is therefore narrower than a simple headline percentage. Positive ductal-bile cultures were common among the sampled people in the included reports, while the wide variation between studies limits any universal yield estimate. The review also cannot supply a single resistance burden. It offers a map of the available microbiology, with important gaps in how samples, patients and resistance are counted.
Paper data and sources
Original title: Microbiology and antimicrobial resistance in biliary tract infections including acute and post-ERCP cholangitis: a systematic review and meta-analysis.
Authors: Salman A, Salman MA, Abdallah A et al.
Journal/Repository: Annals of Saudi medicine
Status: Peer-reviewed
First online: 2026-10-01
DOI: 10.5144/0256-4947.2026.388
Original paper · Full text