Klebsiella pneumoniae as leading cause of pyogenic liver abscess: Three years study in Southern Vietnam

Abstract

BACKGROUND

Pyogenic liver abscess (PLA) is increasingly recognized worldwide. In Asia, Klebsiella pneumoniae (K. pneumoniae) has emerged as the predominant pathogen, yet contemporary data from Vietnam remain limited.

AIM

To determine the microbial spectrum of PLA and compare clinical, computed tomography (CT), management, and outcomes between K. pneumoniae and non-K. pneumoniae cases in Southern Vietnam.

METHODS

This retrospective cohort included adults with PLA managed at Nhan Dan Gia Dinh Hospital from June 2021 to June 2024. Of 123 cases, 17 were excluded (8 with unspecified Gram-negative bacilli, 9 without CT), leaving 106 patients (83 K. pneumoniae, 23 non-K. pneumoniae). Data on demographics, comorbidities, presentation, laboratory, imaging, microbiology, treatment, and outcomes were analyzed using standard parametric/nonparametric and χ²/Fisher tests (two-sided α = 0.05).

RESULTS

Mean age was 59.2 years, and 67.0% were male. Diabetes was more frequent in K. pneumoniae (55.4% vs 30.4%; P = 0.034). C-reactive protein was higher in K. pneumoniae but not significant (229.9 mg/L vs 185.0 mg/L; P = 0.069). Aspartate aminotransferase was significantly elevated (P = 0.048) and alanine aminotransferase borderline (P = 0.065). On CT, K. pneumoniae abscesses more often had irregular margins (P = 0.038) and heterogeneous architecture (P = 0.003). K. pneumoniae predominated in pus (71.9%) and blood (78.6%) cultures. Source-control and hospital stay were similar. Mortality was 10.4% overall, occurring only in K. pneumoniae (13.3% vs 0%). Older age predicted poorer survival (odds ratio = 0.95; P = 0.029).

CONCLUSION

In Southern Vietnam, K. pneumoniae predominates in PLA, characterized by distinctive CT features and higher mortality, emphasizing early recognition, pathogen-directed therapy, and timely image-guided drainage.

Key Words: Pyogenic liver abscess; Klebsiella pneumoniae; Diabetes mellitus; Computed tomography; Vietnam

Core Tip: This three-year study from a tertiary hospital in Southern Vietnam shows that Klebsiella pneumoniae (K. pneumoniae) is now the leading cause of pyogenic liver abscess, responsible for nearly 80 percent of cases. Patients with K. pneumoniae infection were more likely to have diabetes, higher inflammatory markers, and typical computed tomography findings with irregular margins and heterogeneous internal structure. All deaths in this cohort occurred in the K. pneumoniae group. These results highlight the importance of early antibiotic coverage active against K. pneumoniae, timely image-guided drainage, and close clinical monitoring in areas where this organism is common.

INTRODUCTION

Liver abscesses are a serious global health problem with rising incidence. Reported rates are 2.3-3.6 per 100000 in Europe and the Americas but reach 14-18 per 100000 in Asia, where the condition is endemic[1-3]. Mortality remains substantial, ranging from 2% to 31%[4]. If untreated, abscesses can lead to sepsis, multiorgan failure, and death. The rising prevalence of diabetes mellitus, malignancy, chronic kidney disease, and widespread use of immunosuppressants has contributed to the increasing disease burden.

Pyogenic liver abscess (PLA), which accounts for about 80% of liver abscesses, is predominantly bacterial. Historically, Escherichia coli (E. coli) and polymicrobial infections secondary to intra-abdominal pathology such as appendicitis or diverticulitis were common causes, particularly in Western countries[4]. In recent decades, however, hepatobiliary diseases and interventional procedures have become increasingly important risk factors. Advances in microbiology have also reshaped understanding of the causative spectrum: In Asia, Klebsiella pneumoniae (K. pneumoniae) has overtaken E. coli as the predominant pathogen[4]. Unlike most bacteria associated with biliary tract disease, K. pneumoniae abscesses are often cryptogenic, larger in size, and strongly associated with diabetes[5,6].

The emergence of hypervirulent K. pneumoniae strains further complicates the clinical picture[7,8]. These strains harbor enhanced virulence factors that predispose to invasive disease and metastatic complications such as pneumonia, meningitis, and endophthalmitis, sometimes in otherwise healthy hosts. Rising antimicrobial resistance among K. pneumoniae isolates adds another challenge, complicating empirical treatment and raising concerns for patient outcomes.

While numerous studies from Taiwan, Hong Kong, China, and Southeast Asia have confirmed the predominance of K. pneumoniae, data from Vietnam remain limited[9-11]. Most reports have been restricted by small sample sizes, pediatric populations, or single-center design, and few have specifically examined Southern Vietnam. The present study was therefore undertaken to define the current microbial spectrum of PLA and to compare the clinical characteristics and outcomes of K. pneumoniae vs non-K. pneumoniae cases at a large tertiary referral hospital in Ho Chi Minh City. These findings are expected to provide updated insights for empirical therapy and more effective management strategies in this high-burden setting.

MATERIALS AND METHODS

This study was designed as a retrospective cohort, conducted at Nhan Dan Gia Dinh Hospital, a tertiary referral center in Southern Vietnam. The manuscript adheres to the STrengthening the Reporting of OBservational studies in Epidemiology guidelines for cohort studies. The study protocol was approved by the Institutional Review Board of Nhan Dan Gia Dinh Hospital (Approval No. 63/NDGĐ-HĐĐĐ), and all procedures complied with the ethical standards outlined in the Declaration of Helsinki.

Study setting and participants

All adult patients (aged 18 years and older) diagnosed with PLA between June 1, 2021 and June 1, 2024 were retrospectively identified through the hospital’s electronic medical record system using the International Classification of Diseases, Tenth Revision code K75.0. The diagnosis of PLA was established when there was a localized collection of pus within the liver parenchyma confirmed by imaging and/or microbiological testing, with clinical signs of systemic infection and response to antimicrobial therapy. Patients were excluded if they had incomplete records or if the etiology of the liver abscess was non-bacterial. Specifically, we excluded cases with features suggestive of amoebic liver abscess, parasitic liver abscess caused by Fasciola hepatica or other flukes, eosinophilic abscesses characterized by peripheral eosinophilia, and abscesses secondary to hepatic or metastatic malignancy. Potential contaminants were excluded according to Centers for Disease Control and Prevention/National Healthcare Safety Network criteria (single positive culture without clinical correlation).

Isolates were classified as K. pneumoniae-positive when K. pneumoniae was detected in pus and/or blood cultures, irrespective of concomitant organisms. Cases without K. pneumoniae isolation were coded as non-K. pneumoniae. Patients with polymicrobial cultures were retained and classified in the non-K. pneumoniae arm unless K. pneumoniae was co-isolated.

Data collection and variables

Medical records were reviewed by two independent investigators. Data extracted included demographic information (age, sex), comorbidities (notably diabetes mellitus and biliary tract disease), and clinical symptoms at presentation (fever, right upper quadrant pain, jaundice). Per institutional policy, only patients with persistent high-grade fever > 38.5 °C were indicated for blood culture testing. Cultures were processed at the microbiology laboratory using the Vitek^® automated system (bioMérieux, France). Although isolation and capsular typing of K. pneumoniae strains were not performed - consistent with the current limitations across most Vietnamese hospitals - when K. pneumoniae was identified in positive cultures, antibiotic susceptibility testing (antibiogram) was conducted using the Vitek^® system.

We extracted laboratory parameters including white blood cell count, C-reactive protein, liver function tests, renal function indicators, and coagulation profiles [international normalized ratio (INR)]. Variables such as hypotension and procalcitonin were initially planned for collection but were excluded due to incomplete or inconsistent documentation in the medical records. Data on treatment modalities, including empirical and definitive antibiotic regimens, drainage techniques (percutaneous vs surgical), and supportive care, were recorded.

Regarding outcomes, only length of hospital stay and in-hospital mortality were consistently available and analyzed. Length of stay was treated as a continuous variable without categorization. Since the primary objective of this study was to describe the distribution of bacterial etiologies of PLA and to compare clinical characteristics between the K. pneumoniae and non-K. pneumoniae groups, other outcomes such as treatment failure or complications were not included in the final analysis.

Pulmonary infection was defined radiologically based on the presence of new parenchymal opacities, including consolidation or ground-glass opacities on multislice chest computed tomography (CT), or homogeneous infiltrates on standard posteroanterior chest X-ray, as reported by board-certified radiologists. All patients underwent multislice contrast-enhanced CT of the abdomen at the time of hospital admission. Scans were performed using Siemens SOMATOM Definition (Siemens Healthineers Ltd., Germany) and Philips Incisive 64-detector CT systems [Philips Healthcare (Suzhou) Co., Ltd., China]. Dynamic contrast-enhanced protocols included non-contrast, arterial, portal venous, and delayed phases, with tube voltage 120 kVp, tube current 200-250 mAs, collimation 64 mm × 0.6 mm, slice thickness 5 mm, and reconstruction interval 3 mm. A nonionic iodinated contrast medium (1.5 mL/kg, maximum 120 mL) was administered intravenously at 3.0-3.5 mL/second, followed by a 30 mL saline flush. Non-contrast images were obtained before injection, arterial phase images at 25-30 seconds, portal venous phase at 60-70 seconds, and delayed phase at 3-5 minutes after injection. Imaging characteristics recorded included number, size, location (right or left lobe), presence of gas, multiloculated appearance, and rim enhancement. These features were compared between patients with K. pneumoniae-associated abscesses and those with non-K. pneumoniae or culture-negative results.

Statistical analysis

Data were entered into a secure database and analyzed using R version 4.4.2 (R Foundation for Statistical Computing, Vienna, Austria). Descriptive statistics for continuous variables were presented as mean ± SD or median with interquartile range, depending on distribution. Categorical variables were expressed as n (%). Comparative analyses between K. pneumoniae and non-K. pneumoniae groups were performed using the Student’s t-test or Mann-Whitney U test for continuous variables, and χ² or Fisher’s exact test for categorical variables. A P-value < 0.05 was considered statistically significant. Missing data were managed by listwise deletion. If clinically appropriate, multivariate logistic regression will be conducted to identify independent predictors of adverse outcomes such as in-hospital mortality or treatment failure.

Study objectives

The primary objective of this study was to determine the spectrum and frequency of bacterial pathogens in PLA at Nhan Dan Gia Dinh Hospital and identify the dominant species, hypothesized to be K. pneumoniae. Secondary objectives included comparison of clinical presentation, CT imaging features, and outcomes between K. pneumoniae-associated and non-K. pneumoniae cases.

RESULTS

Study cohort

During the study period (June 1, 2021 to June 1, 2024), we analyzed 106 culture-positive cases of PLA with complete clinical and imaging data, as shown in Figure 1. Of these, 83 (78.3%) were associated with K. pneumoniae (K. pneumoniae group) and 23 (21.7%) were non-K. pneumoniae (including other organisms and polymicrobial isolates). The mean age was 59.2 years (SD 15.1), and 71/106 (67.0%) were male. Diabetes mellitus was present in 53/106 (50.0%) overall and was more frequent in the K. pneumoniae group (46/83, 55.4%) than in the non-K. pneumoniae group (7/23, 30.4%; P = 0.034).

Figure 1 Flow diagram showing the selection of reports included in the review. CT: Computed tomography; K. pneumoniae: Klebsiella pneumoniae.

Clinical presentation and laboratory findings

Fever was documented in 86/106 (81.1%), right upper quadrant pain in 65/106 (61.3%), and jaundice in 9/106 (8.5%), without significant between-group differences. Mean C-reactive protein was higher in the K. pneumoniae group than in the non-K. pneumoniae group (229.9 mg/L vs 185.0 mg/L), but this difference was not statistically significant (P = 0.069). Aspartate aminotransferase was also higher in the K. pneumoniae group (median 78 U/L vs 37 U/L; P = 0.048), whereas alanine aminotransferase showed a borderline difference (P = 0.065). Other laboratory parameters, including white blood cell count, creatinine, and INR, did not differ significantly between groups. Pulmonary infections were observed in 29 patients (27.4%) overall, with comparable prevalence in K. pneumoniae and non-K. pneumoniae groups (25.3% vs 34.8%; P = 0.367). Detailed results are provided in Table 1.

Table 1 Baseline characteristics and imaging features of liver abscess by Klebsiella pneumoniae vs non-Klebsiella pneumoniae, mean ± SD/n (%)/median (interquartile range).

Characteristic	Overall (n = 106)	K. pneumoniae (n = 83)	Non-K. pneumoniae (n = 23)	P value1
Demographics
Age	59.2 ± 15.1	60.3 ± 13.8	55.3 ± 18.9	0.248
Male	71 (67.0)	56 (67.5)	15 (65.2)	0.839
Diabetes mellitus	53 (50.0)	46 (55.4)	7 (30.4)	0.034
Pulmonary infection	29 (27.4)	21 (25.3)	8 (34.8)	0.367
Clinical presentation
Fever	86 (81.1)	70 (84.3)	16 (69.6)	0.134
Right upper quadrant pain	65 (61.3)	52 (62.7)	13 (56.5)	0.593
Jaundice	9 (8.5)	7 (8.4)	2 (8.7)	> 0.999
Laboratory tests at baseline
WBC, × 109/L	14.9 (11.4-20.4)	14.8 (10.7-20.9)	15.2 (13.6-18.8)	0.457
CRP, mg/L	219.0 ± 96.2	229.9 ± 99.4	185.0 ± 78.4	0.069
Missing (CRP)2, n	24	21	3
AST, U/L	71.2 (37.6-128.8)	78.3 (41.9-131.2)	36.9 (30.0-126.3)	0.048
ALT, U/L	70.1 (40.9-118.2)	76.8 (45.3-126.0)	42.7 (32.3-107.6)	0.065
Creatinine, μmol/L	96.5 (76.7-134.0)	100.1 (77.1-133.2)	89.1 (76.7-134.6)	0.577
Missing (creatinine), n	15	15	0
INR	1.3 (1.2-1.4)	1.3 (1.2-1.4)	1.3 (1.2-1.5)	0.332
Missing (INR)2, n	17	17	0
Imaging characteristics
Abscess size, mm	65.0 (53.0-87.0)	65.0 (52.0-87.0)	69.0 (55.0-91.0)	0.401
Number of abscesses				> 0.999
1	84 (79.2)	65 (78.3)	19 (82.6)
2	13 (12.3)	10 (12.1)	3 (13.0)
3	6 (5.7)	5 (6.0)	1 (4.4)
≥ 4	3 (2.8)	3 (3.6)	0 (0)
Abscess location				0.305
Left lobe	31 (29.2)	23 (27.7)	8 (34.8)
Right lobe	60 (56.6)	46 (55.4)	14 (60.9)
Bilateral	15 (14.2)	14 (16.9)	1 (4.3)
Multiplicity				0.777
Single	84 (79.2)	65 (78.3)	19 (82.6)
Multiple	22 (20.8)	18 (21.7)	4 (17.4)
Perilesional enhancement	65 (61.3)	52 (62.7)	13 (56.5)	0.593
Abscess margin				0.038
Smooth	7 (6.6)	3 (3.6)	4 (17.4)
Irregular	99 (93.4)	80 (96.4)	19 (82.6)
Abscess homogeneity				0.003
Homogeneous	18 (17.0)	9 (10.8)	9 (39.1)
Heterogeneous	88 (83.0)	74 (89.2)	14 (60.9)
Intralesional gas	12 (11.3)	10 (12.0)	2 (8.7)	> 0.999
Biliary gas	8 (7.5)	7 (8.4)	1 (4.3)	> 0.999

¹P-values were calculated using the Student’s t test or Mann-Whitney U test for continuous variables and the χ² test or Fisher exact test for categorical variables, as appropriate.

²Missing values (rows labeled “missing, n”) were excluded from comparative tests.

WBC: White blood cell count; CRP: C-reactive protein; AST: Aspartate aminotransferase; ALT: Alanine aminotransferase; INR: International normalized ratio; K. pneumoniae: Klebsiella pneumoniae.

Microbiological spectrum

Among pus isolates (n = 89), K. pneumoniae accounted for 64/89 (71.9%), followed by Streptococcus species (aggregated across species) at 12/89 (13.5%) and E. coli at 6/89 (6.7%); mixed K. pneumoniae-E. coli accounted for 2/89 (2.2%), and each remaining single organism represented ≤ 1.1% of isolates. In blood isolates (n = 42), K. pneumoniae predominated (33/42, 78.6%), streptococci collectively accounted for 2/42 (4.8%), and each remaining single or mixed organism represented 1/42 (2.4%). The overall distribution of bacterial isolates is shown in Figure 2A. Antibiotic resistance rates of K. pneumoniae from both pus (n = 55) and blood (n = 26) isolates are presented in Figure 2B. All values refer to isolates rather than unique patients; blood cultures were performed per institutional policy as described in the methods section.

Figure 2 Etiology and resistance profile of pyogenic liver abscess. A: Spectrum of bacterial isolates in pus and blood among patients with pyogenic liver abscess, 2021-2024. Bars show columnbased percentages. Counts refer to isolates, not unique patients. Data show isolates (not unique patients). Column totals: Pus isolates n = 89; Blood isolates n = 42. Percentages are column-based. Species names standardized from raw abbreviations present in the source comma-separated values; B: Antibiotic resistance rates of Klebsiella pneumoniae isolates from pyogenic liver abscess, comparing pus (blue) and blood (orange) cultures. Antibiotic susceptibility testing was performed in 55 pus isolates and 26 blood isolates. Percentages are column-based. K. pneumoniae: Klebsiella pneumoniae; E. coli: Escherichia coli.

Imaging characteristics

The median maximal abscess diameter at baseline was 65 mm and did not differ significantly between groups (P = 0.401). Most patients presented with solitary abscesses (84/106, 79.2%), while multiple lesions were observed in 22/106 (20.8%; P = 0.777). The right hepatic lobe was the most common location (60/106, 56.6%), followed by the left lobe (31/106, 29.2%) and bilateral involvement (15/106, 14.2%; P = 0.305). Two qualitative imaging features were significantly more frequent in the K. pneumoniae group: Irregular or lobulated margins (80/83, 96.4% vs 19/23, 82.6%; P = 0.038) and heterogeneous internal architecture (74/83, 89.2% vs 14/23, 60.9%; P = 0.003). The presence of intralesional gas, biliary gas, or portal venous gas was uncommon and showed no between-group differences. These findings are summarized in Table 2. Representative examples of K. pneumoniae abscesses with irregular and heterogeneous appearance are shown in Figure 3. In contrast, Figure 4 illustrates a case of PLA caused by Streptococcus species, which displayed slightly irregular margins but a homogeneous, non-septate internal architecture with rim enhancement, underscoring the radiological differences between K. pneumoniae and non-K. pneumoniae etiologies.

Figure 3 Characteristics of Klebsiella pneumoniae liver abscess on multislice computed tomography scan. The abscesses (white asterisk) appear as multiple lesions with irregular and lobulated margins. The internal architecture of the abscesses is heterogeneous, reflecting varying degrees of necrosis, fluid content, and inflammatory debris. A: Non-contrast phase; B: Arterial phase; C: Portal venous phase; D: Delayed phase.

Figure 4 Multiphase computed tomography images of a pyogenic liver abscess caused by Streptococcus species (white asterisks). The cavity shows slightly irregular margins but homogeneous, non-septated internal architecture with rim enhancement. This contrasts with most Klebsiella pneumoniae liver abscesses, which typically demonstrate heterogeneous internal architecture. A: Non-contrast phase; B: Arterial phase; C: Portal venous phase; D: Delayed phase.

Table 2 Management and inhospital outcomes of liver abscesses by Klebsiella pneumoniae vs non-Klebsiella pneumoniae.

Characteristic	Overall (n = 106)	K. pneumoniae (n = 83)	Non-K. pneumoniae (n = 23)	P value1
Management during index admission
Drainage at admission	41 (38.7)	32 (38.6)	9 (39.1)	0.960
Drainage after medical therapy	44 (41.5)	33 (39.8)	11 (47.8)	0.487
Surgery at admission	4 (3.8)	3 (3.6)	1 (4.3)	> 0.999
Surgery after medical therapy	4 (3.8)	4 (4.8)	0 (0)	0.575
Outcomes
LOS (days)	9.0 (6.0-14.0)	10.0 (6.0-15.0)	8.0 (6.0-12.0)	0.324
In-hospital mortality	11 (10.4)	11 (13.3)	0 (0)	0.116

¹P-values were calculated using the Student’s t test or Mann-Whitney U test for continuous variables and the χ² test or Fisher exact test for categorical variables, as appropriate.

LOS: Length of stay; K. pneumoniae: Klebsiella pneumoniae.

DISCUSSION

In this three-year cohort, K. pneumoniae emerged as the predominant cause of PLA in Southern Vietnam, accounting for nearly four out of five culture-confirmed cases. This observation aligns with the epidemiological shift reported across East and Southeast Asia, where K. pneumoniae has overtaken E. coli as the leading pathogen, contrasting with the predominance of E. coli in Western cohorts[9-11]. Reports from Taiwan and China established this shift more than a decade ago and highlighted the clinical salience of hypervirulent lineages capable of metastatic complications in otherwise healthy hosts[12-14]. Recent Vietnamese series reported similar organism distributions, supporting the notion that K. pneumoniae now accounts for a substantial share of PLA across the country[15]. The predominance of K. pneumoniae has important clinical implications: It shapes empirical antibiotic choices, highlights the need for early recognition of invasive disease phenotypes, and underscores regional differences in microbial epidemiology. Our study therefore contributes locally relevant data to the global understanding of PLA and provides a framework for tailoring management strategies in Vietnam.

The association between K. pneumoniae and diabetes observed here is biologically plausible. Hyperglycemia impairs neutrophil chemotaxis, phagocytosis, and intracellular killing, and experimental data suggest that host susceptibility interacts with capsular and siderophore determinants that characterize hypervirulent K. pneumoniae[16]. Clinical reviews and mechanistic studies describe a pathotype that preferentially causes communityacquired disease, often with hematogenous seeding of the eye and central nervous system. The epidemiology remains most prominent along the Asian Pacific Rim but is now reported globally. Clinicians in K. pneumoniae-predominant settings should therefore maintain a low threshold to evaluate for metastatic infection when bacteremia persists or focal symptoms arise[16,17]. Pulmonary involvement was also observed in a substantial subset of patients, but its prevalence did not differ between K. pneumoniae and non-K. pneumoniae PLA. While pulmonary complications such as septic emboli and pneumonia are well recognized in invasive K. pneumoniae infections, our findings suggest that concomitant pulmonary disease is not specific to K. pneumoniae and should be systematically assessed in all PLA patients[4,10,18]. Serum creatinine and INR were analyzed as markers of systemic severity rather than pathogen specificity. Creatinine reflects renal impairment from sepsis or comorbidities, while INR indicates coagulation and hepatic dysfunction that may affect drainage risks. Although no significant differences were found, including these variables provides a more complete clinical profile of PLA patients.

The radiologic observations in this cohort have practical implications. Heterogeneous internal architecture and lobulated or illdefined margins on CT likely reflect multi-loculation, necrotic debris, and viscosity gradients that can defeat singlepass aspiration. Contemporary imaging reviews and multidetector CT-based studies emphasize that morphology informs drainability and that early catheter placement should be considered when collections are large, complex, or nonliquefied[19,20]. In parallel, randomized and pooled evidence indicates that percutaneous catheter drainage provides higher clinical success and faster clinical improvement than needle aspiration for many patients with PLA[21]. These considerations align with an interventional strategy that favors early catheter drainage or planned repeat aspirations when K. pneumoniae abscesses present with the complex morphology described here[22]. The presence of gas within a liver abscess on CT is a common feature that may fluctuate depending on the timing of imaging relative to the disease course[23]. In our cohort, however, all patients underwent multislice CT imaging at the time of admission using standardized dynamic contrast-enhanced protocols. As such, variation in gas appearance over the natural course of disease could not be evaluated. This represents a limitation of our study, as we were unable to assess temporal changes in gas formation. Prospective studies with serial CT follow-up could better clarify the evolution of intralesional gas in PLA.

Our organism pattern dovetails with the Asia-Pacific literature, yet it contrasts with several European cohorts in which Enterobacterales remain common but K. pneumoniae is less dominant and enterococci are proportionally more frequent. In a recent German tertiarycare series, positive blood cultures and growth of Enterobacterales were each associated with higher 30-day mortality, underscoring that microbiology and bacteremia status carry prognostic information independent of geography[24]. Populationbased data from Canada also linked poorer outcomes to polymicrobial bacteremia and to absence of drainage[25]. These external data help reconcile our finding of numerically higher mortality limited to K. pneumoniae cases, despite the metaanalytic observation that K. pneumoniae-PLA can have lower crude mortality than non-K. pneumoniae-PLA at the global level; case mix, source control strategy, and comorbidity likely explain some of the heterogeneity[26].

Therapeutic implications follow. Empirical regimens in K. pneumoniae-predominant regions should ensure early, adequate Gram-negative coverage active against K. pneumoniae while preserving anaerobic activity where biliary pathology or polymicrobial infection is suspected[27]. Regional reports from Vietnam note high K. pneumoniae yields in pus or blood, reinforcing the importance of early cultures to narrow therapy. Once microbiology is available, deescalation should be coupled with timely, imageguided source control. In centers where capsular serotyping or virulence profiling is not yet routine, structured screening for ocular or central nervous system complications is reasonable for patients with persistent bacteremia, severe sepsis, or focal neurologic or visual complaints[28,29].

Several cautions apply when interpreting these results. The retrospective singlecenter design constrains causal inference and limits generalizability. Culture practices that favored blood cultures in patients with persistent fever may have introduced ascertainment bias in the comparison of bloodstream isolates. Missingness in laboratory variables reduced power for betweengroup differences. We did not perform capsular typing or assess hypervirulence markers such as regulator of mucoid phenotype A/regulator of mucoid phenotype A2 or aerobactin, so any inference about hypervirulent K. pneumoniae remains indirect. Finally, relatively few deaths occurred, which limits precision around organismspecific risk estimates and precluded robust multivariable adjustment in this cohort. These limitations are common to many observational PLA series and argue for coordinated prospective work[30].

Priorities for Vietnam include multicenter surveillance using standardized indications for blood culture, harmonized definitions for radiologic features, and organism characterization that incorporates capsular typing and hypervirulence markers. Given the imaging phenotype observed here, pragmatic trials or registries that compare initial needle aspiration with upfront catheter drainage in complex collections would be valuable, with stratification by suspected organism at presentation. Antimicrobial stewardship programs should integrate local susceptibility patterns while preserving early broad coverage for highrisk presentations. The accumulating regional literature, including Vietnam-based cohorts, provides a foundation for such efforts and should inform guideline development tailored to K. pneumoniae-predominant settings[15,29].

CONCLUSION

In a K. pneumoniae-predominant setting, patients with PLA displayed a characteristic clinical-radiologic profile and a mortality signal concentrated in the K. pneumoniae group. These data, generated from Southern Vietnam, corroborate the regional shift toward K. pneumoniae as the leading etiology and support early organism-focused management pathways that integrate prompt, tailored source control with appropriate empirical antimicrobial coverage.

ACKNOWLEDGEMENTS

The authors would like to express their sincere gratitude to the Department of Hepato-Pancreato-Biliary at Nhan Dan Gia Dinh Hospital, Ho Chi Minh City, Viet Nam for providing invaluable support throughout the study. We are deeply grateful to all patients and their families whose medical data contributed to this research. Without their cooperation, this study could not have been conducted.