Comparative clinical outcomes of pandrug-resistant Acinetobacter baumannii and Klebsiella pneumoniae infections in children

Abstract

Background

The global rise of antimicrobial resistance has become a major public health concern. However, data on the clinical course and outcomes of infections caused by pandrug-resistant (PDR) bacteria in children remain limited. This study aimed to compare the clinical characteristics and treatment outcomes of PDR Acinetobacter baumannii and Klebsiella pneumoniae infections in pediatric patients.

Methods

We retrospectively analyzed pediatric cases with PDR A. baumannii and K. pneumoniae infections diagnosed between 2020 and 2023 in a tertiary-care center. Demographic characteristics, infection sites, clinical findings, and outcomes were compared between the two groups using appropriate statistical tests.

Results

PDR K. pneumoniae infections occurred in significantly younger children (p = 0.004). PDR K. pneumoniae episodes most commonly involved pneumonia (47.4%) or bloodstream infection (42.1%), with mortality rates of 15.8%. PDR A. baumannii episodes were most frequently associated with pneumonia (65.3%), with mortality rates of 7.7%. Recurrence rates were similar between the two groups (36.8% vs. 26.9%; p = 0.478), and none of the recurrent infections resulted in death. Fever duration was longer in the PDR A. baumannii group (p = 0.009), whereas co-infections and complications were numerically more frequent but did not reach statistical significance (respectively; p = 0.06 and p = 0.08). Ceftazidime-avibactam (CZA) resistance was observed in 63.2% of PDR K. pneumoniae isolates, and all recurrent PDR K. pneumoniae episodes were CZA-resistant; no deaths occurred among recurrent episodes in either group.

Conclusions

PDR A. baumannii and K. pneumoniae infections showed comparable severity and mortality in children, although K. pneumoniae affected younger patients and A. baumannii was associated with longer fever duration. Both pathogens were typically identified after prolonged hospitalization and were closely linked to intensive care exposure. The high rate of CZA resistance among PDR K. pneumoniae isolates, including all recurrent PDR K. pneumoniae episodes, was a clinically important finding that may further limit therapeutic options.

Clinical trial number

Not applicable.

Background

The global burden of antimicrobial resistance (AMR) has increased substantially and now represents one of the most important public health challenges worldwide. In 2019, bacterial AMR was estimated to be associated with 4.95 million deaths globally, including 1.27 million deaths directly attributable to resistant bacterial infections. Pandrug resistance refers to non-susceptibility to all agents in all antimicrobial categories relevant to the organism, leaving very limited or no effective therapeutic options [1]. Acinetobacter baumannii and Klebsiella pneumoniae are the most important gram-negative microorganisms that can rapidly develop AMR and tend to spread in hospital environments [2]. The World Health Organization (WHO) has defined multidrug-resistant (MDR) K. pneumoniae and A. baumannii infections as a critical priority global health problem due to the increasing prevalence and difficulties in their treatment. In addition, both the Centers for Disease Control and Prevention (CDC) and the WHO have designated A. baumannii as a primary research target [3–6].

Previous reports have shown that approximately 45% of all A. baumannii isolates are MDR, with rates reaching 90% in countries such as Türkiye, Greece, and the Middle East region [7, 8]. Similarly, carbapenem resistance rates up to 67% were reported for K. pneumoniae in Türkiye [9, 10]. There is limited information on the clinical course and outcomes of infections caused by pandrug-resistant (PDR) bacteria in children [11]. Although some evidence on PDR bacterial infections is available, pediatric studies remain scarce [10, 12, 13]. In addition, it remains unclear whether PDR A. baumannii and K. pneumoniae infections differ in their clinical outcomes in children.

In this study, we aimed to investigate the clinical characteristics and treatment outcomes of PDR A. baumannii and K. pneumoniae infections in pediatric patients and to compare these two groups.

Methods

Study design and setting

This retrospective observational study was conducted at Dokuz Eylül University Hospital, a tertiary-care university referral hospital in İzmir, western Türkiye. The hospital provides care for a broad pediatric population, including critically ill patients requiring pediatric intensive care and children with complex underlying conditions such as hematologic/oncologic diseases and immunocompromising disorders. Pediatric patients diagnosed with pandrug-resistant (PDR) A. baumannii or K. pneumoniae infections between January 2020 and September 2023 were evaluated. Clinical, microbiological, and laboratory data were retrieved from electronic medical records and microbiology laboratory databases. Data collected included demographic characteristics, underlying medical conditions, recent hospitalization history, colonization status, presence of a central venous catheter, laboratory findings, infection sites, coinfections, surgical interventions, treatment regimens and durations, clinical outcomes, and mortality.

Study population

Patients younger than 18 years of age with microbiologically confirmed PDR A. baumannii or K. pneumoniae infection during hospitalization were eligible for inclusion. Only infection episodes meeting predefined clinical and microbiological criteria were analyzed.

Inclusion criteria

Patients were included if they met all of the following:

• Age < 18 years

• Hospitalization during the study period

• Culture-proven infection caused by PDR A. baumannii or K. pneumoniae

• Clinical signs consistent with infection requiring antimicrobial therapy

• Availability of complete clinical, laboratory, and treatment data

Exclusion criteria

Patients were excluded if they had:

• Age ≥ 18 years

• Colonization without clinical evidence of infection

• Incomplete medical records

• Duplicate isolates from the same infection episode

• Polymicrobial infection in which the PDR organism was not considered the primary pathogen

For patients with multiple isolates during a single infection episode, only the first isolate was included in the analysis.

Microbiological methods

Clinical specimens were collected using sterile techniques and processed according to standard microbiological procedures. Samples were inoculated onto appropriate culture media and incubated aerobically at 35–37 °C. Bacterial identification was performed using matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF MS) according to the manufacturer’s instructions.

Antimicrobial susceptibility testing was performed using an automated system and/or the broth microdilution method when necessary. Minimum inhibitory concentrations (MICs) were interpreted according to the European Committee on Antimicrobial Susceptibility Testing (EUCAST) clinical breakpoints valid at the time of testing [14]. The classification of pandrug resistance was based on the standard international definitions proposed by Magiorakos et al. [15]. Accordingly, isolates were classified as pandrug-resistant when they showed non-susceptibility to all agents in all antimicrobial categories relevant to the organism [15].

Quality control procedures were conducted regularly using reference strains in accordance with EUCAST recommendations.

Definitions

Bloodstream infection (BSI) was defined as isolation of a pathogenic organism from at least one blood culture in a patient with compatible clinical signs. Central-line associated bloodstream infection (CLABSI) was defined according to standard diagnostic criteria, including differential time to positivity or matching cultures from catheter and peripheral blood [16].

Ventilator-associated pneumonia (VAP) was defined according to pediatric CDC/NHSN pneumonia/VAP surveillance criteria as pneumonia occurring ≥ 48 h after initiation of mechanical ventilation, with new or progressive radiologic infiltrates and compatible clinical and microbiological findings [17].

Urinary tract infection (UTI) was defined as significant bacterial growth in urine culture accompanied by clinical or laboratory evidence of infection. Central nervous system (CNS) infection was defined by pathogen isolation from cerebrospinal fluid or compatible clinical findings supported by microbiological or radiologic evidence.

Recurrent infection was defined as a subsequent episode caused by the same pathogen following clinical resolution of the index infection, completion of antimicrobial therapy, and documentation of at least two consecutive negative cultures. Separate infection episodes were determined based on clinical and microbiological criteria and analyzed independently.

Fever duration was defined as the number of days with a body temperature ≥ 38.0 °C. Time to bacterial growth was defined as the interval between hospital admission and the first positive culture. Infections occurring 48 h after hospitalization were classified as nosocomial [18]. Co-infection was defined as simultaneous infection of a host by more than one pathogen species. Mortality was defined as all-cause death occurring during the infection episode.

Outcome measures

Primary outcomes included mortality, recurrence, and clinical response to therapy. Secondary outcomes included duration of fever, treatment duration, and need for surgical intervention.

Demographic variables, including age and sex, were analyzed at the patient level. Infection-related characteristics, treatment variables, recurrence, and outcomes were analyzed at the infection-episode level. For patients with multiple isolates during the same infection episode, only the first isolate was included.

Statistical analysis

Statistical analyses were performed using IBM SPSS Statistics for Windows, version 26.0 (IBM Corp., Armonk, NY, USA). The Shapiro-Wilk test was used to investigate whether the continuous variables were normally distributed. Descriptive statistics were presented as medians (25th − 75th p) regarding the distribution of variables. If variables were normally distributed, an independent samples t-test was used, and if they were not normally distributed, the Mann-Whitney U test was used to compare two independent groups. Chi-square or Fisher exact tests were used for comparisons between categorical variables. Values of p < 0.05 were considered statistically significant. For categorical variables, odds ratios (ORs) and their 95% confidence intervals (CIs) were calculated from 2 × 2 contingency tables using the same software. Fisher’s exact test was used when expected cell counts were small. Post-hoc power analysis was performed using the observed effect sizes, total sample size, and a two-sided alpha level of 0.05. Because of the limited sample size, these estimates were interpreted descriptively.

An exploratory multivariable analysis was performed to assess the potential influence of age and major clinical covariates on between-group comparisons. Because of the small sample size and limited number of outcome events, only a restricted number of clinically relevant variables were included to reduce the risk of overfitting. Variables with p < 0.10 in univariate analyses were considered for inclusion. Variables causing complete or quasi-complete separation, as well as clinically inappropriate post-infection variables, were not entered into the final model. Adjusted odds ratios (aORs) with 95% confidence intervals (CIs) were reported.

Results

Pandrug-resistant Klebsiella pneumoniae

We diagnosed a total of 19 episodes from 12 patients with PDR K. pneumoniae infections. The median age was 0.38 (0.14–5.59) years, and 58.3% (n = 7) of patients were female. All of the patients except one had underlying comorbidities. The most common of these were 33.3% (n = 4) neurometabolic disease, 25% (n = 3) congenital heart disease, 16.7% (n = 2) malignancy, 8.3% (n = 1) prematurity, and 8.3% (n = 1) hydronephrosis. The primary infection sites were pneumonia (47.4%, n = 9), BSI (42.1%, n = 8), UTI (21.1%, n = 4), and central nervous system involvement (5.3%, n = 1). Because some episodes involved more than one infection focus, the percentages of primary infection sites are not mutually exclusive and may exceed 100%. There was more than one site of infection in 21.1% (n = 4) of the episodes. 37.5% (n = 3) of BSIs were catheter-related, and all pneumonia episodes were VAP. Most patients (63.2%, n = 12) were admitted to the pediatric intensive care unit (PICU). Bacterial growth was identified at a median of 21 (14–46) days after hospitalization.

Ceftazidime-avibactam (CZA) resistance was observed in 63.2% (n = 12) of isolates; whereas this could not be evaluated in 3 (15.8%) isolates. In the first-line treatment, a high-dose meropenem plus colistin combination was administered in 63.2% (n = 12) of episodes, CZA in 15.8% (n = 3), high-dose meropenem plus tigecycline in two (10.5%), high-dose meropenem plus colistin plus amikacin in 5.3% (n = 1), and colistin plus ciprofloxacin in 5.3% (n = 1). A second-line treatment was required in two infection episodes, both of which were included high-dose meropenem plus colistin plus tigecycline combination. Surgical intervention was performed in 10.5% (n = 2) of episodes [5.3% (n = 1) thoracotomy, and 5.3% (n = 1) removal of a stent from the bile duct]. Mortality occurred in 15.8% (n = 3) of episodes. A 2-month-old female patient (who had a history of premature birth) died on the day 10 of BSI treatment. A 5.5-year-old male (who had a congenital heart disease) and a 2.5-month-old female patient (who had a neurometabolic disease) with both BSI and pneumonia died on the 16th and 7th days of their treatment, respectively. When infection episodes with and without mortality were compared, total leukocyte count was lower (8.000/mm3 vs. 12.750/mm3, p = 0.014), platelet count was lower (77.000/mm3 vs. 305.500/mm3, p = 0.019), and C-reactive protein was higher (228 vs. 97.7 mg/L, p = 0.044) in the mortality group (Table 1).

Table 1.

Comparison of clinical characteristics, laboratory findings, complications, and mortality rates in patients with pandrug-resistant Klebsiella pneumoniae, regarding recurrence and mortality

	Recurrence (n = 7)	Without recurrence (n = 12)	p-value	Mortality (n = 3)	Without mortality (n = 16)	p-value
CRP (mg/L)	129.4 (108.1-189.3)	85.1 (57.3-203.5)	0.398	228 (189.3–228)	97.7 (57.3–130)	0.044
TLC/mm3	9.800 (8.900–18.600)	12.400 (9.275–14.375)	0.735	8.000 (600-8.000)	12.750 (9.575–16.450)	0.014
ANC/mm3	8.800 (5.800–18.000)	7.050 (3.600–8.300)	0.236	5.800 (0-5.800)	7.950 (6.650-11.725)	0.093
ALC/mm3	800 (500 − 3.800)	4.300 (1.350–6.300)	0.042	600 (500–600)	4.000 (925-6.075)	0.064
Hemoglobin (g/dL)	9 (6.6–10.2)	9.1 (7.6–9.5)	0.735	8.8 (7.5–8.8)	9.2 (7.5–10.1)	0.342
Platelets/mm3	254.000 (201.000-326.000)	305.500 (261.000-455.000)	0.271	77.000 (22.000–77.000)	305.500 (261.000-455.000)	0.019
Treatment duration (days)	16 (10–32)	14 (10-14.7)	0.155	10 (7–10)	14 (10-17.5)	0.30
Surgery (%, n)	14.3% (n = 1)	8.3% (n = 1)	1.00	0% (n = 0)	12.5% (n = 2)	1.00
BSI (%, n)	57.1% (n = 4)	33.3% (n = 4)	0.377	100% (n = 3)	31.3% (n = 5)	0.058
Pneumonia (%, n)	28.6% (n = 2)	58.3% (n = 7)	0.350	66.7% (n = 2)	43.8% (n = 7)	0.582
Complications (%, n)	57.1% (n = 4)	50.0% (n = 6)	1.00	66.7% (n = 2)	50% (n = 8)	1.00
Mortality (%, n)	0% (n = 0)	25.0% (n = 3)	1.00	-	-	-

(CRP: C-reactive protein, TLC: Total leukocyte count, ANC: Absolute neutrophil count, ALC: Absolute lymphocyte count, BSI: Bloodstream infection)

Recurrent infections occurred in 36.8% (n = 7) of episodes. Four of them occurred in the blood, two in the lungs, and one in both the blood and the CNS. All of the recurrent infections were resistant to CZA. When comparing those with and without recurrent infections, we found that the duration of treatment was similar (16 vs. 14 days, p = 0.155), and absolute lymphocyte count was lower (800/mm3 vs. 4.300/mm3, p = 0.042) in recurrent infections. No mortality was observed among recurrent infections (Table 1).

Pandrug-resistant Acinetobacter baumannii

We diagnosed a total of 26 episodes from 19 patients with PDR A. baumannii infection. The median age was 8.9 (4.43–15.74) years, and 52.6% (n = 10) of patients were male. All of the patients had underlying comorbidities. The most common of these were 31.6% (n = 6) neurometabolic disease, 26.3% (n = 5) post-trauma, 15.8% (n = 3) had malignancy, 10.5% (n = 2) had congenital heart disease, 5.3% (n = 1) had prematurity, 5.3% (n = 1) had primary immunodeficiency, and 5.3% (n = 1) had brain abscess. The most common focus of infection was pneumonia in 65.3% (n = 17), BSI in 34.6% (n = 9), CNS in 15.3% (n = 4), UTI in 3.8% (n = 1), thrombophlebitis in 3.8% (n = 1), and surgical site infection in 3.8% (n = 1). In 19.2% (n = 5) of episodes, patients had more than one focus of infection. 87.5% (n = 7) of the BSIs were catheter-related, and all pneumonia episodes were VAP. In 65.4% (n = 17) of the episodes, patients were hospitalized in the PICU. Bacterial growth was identified at a median of 17 (9.2–51.7) days after hospitalization. Co-infections occurred in 19.2% (n = 5) of the episodes [3.8% (n = 1) Candida tropicalis, 3.8% (n = 1) Candida krusei, 3.8% (n = 1) Aspergillus flavus, 3.8% (n = 1) Enterococcus faecium, and 3.8% (n = 1) Klebsiella pneumoniae].

In the first-line treatment, a high-dose ampicillin-sulbactam (SAM) plus tigecycline combination was administered in 46.2% (n = 12) of infection episodes, high-dose meropenem plus colistin in 23.1% (n = 6), high-dose SAM plus colistin in 15.4% (n = 4), high-dose meropenem plus tigecycline in 7.7% (n = 2), high-dose SAM plus tigecycline plus colistin in 3.8% (n = 1), and high-dose SAM plus meropenem plus colistin in 3.8% (n = 1). Antimicrobial therapy was switched to second-line treatment in 34.6% (n = 9) of PDR A. baumannii infection episodes. In 23% (n = 6) of the infection episodes, patients required surgical interventions [11.5% (n = 3) extraventricular shunt revision, 7.7% (n = 2) thoracotomy, 3.8% (n = 1) debridement surgery]. The mortality rate was 7.7% (n = 2) in episodes. An 8.4-year-old male patient with Langerhans cell histiocytosis with lung involvement died on the 4th day of pneumonia treatment (high-dose SAM plus tigecycline). A 15.7-year-old male patient who had been in PICU for a long time died due to CLABSI and septic shock on the 4th day of his treatment. There was no significant difference in laboratory findings when infection episodes with and without mortality were compared (p > 0.05) (Table 2).

Table 2.

Comparison of clinical characteristics, laboratory findings, complications, and mortality rates in patients with pandrug-resistant Acinetobacter baumannii, regarding recurrence and mortality

	Recurrence (n = 7)	Without recurrence (n = 19)	p-value	Mortality (n = 2)	Without mortality (n = 24)	p-value
CRP (mg/L)	59.6 (46.9-206.6)	88.9 (23.3-141.4)	0.506	103.7 (67.9-103.7)	84.6 (28.3–150)	0.773
TLC/mm3	14.700 (11.200–19.900)	11.600 (8.900–15.700)	0.488	9.600 (9.000-9.600)	12.600 (9.200-17.250)	0.290
ANC/mm3	13.300 (8.500–16.200)	8.400 (5.600–12.600)	0.326	7.300 (5.800–7.300)	9.800 (6.200-14.275)	0.386
ALC/mm3	1.500 (1.100–1.700)	1.500 (1.000-1.900)	0.885	1.800 (1.500–1.800)	1.450 (1.025–1.875)	0.385
Hemoglobin (g/dL)	8.6 (8-11.7)	9.6 (8.6–11.1)	0.469	7.3 (5.5–7.3)	9.8 (8.4–11.4)	0.248
Platelets/mm3	349.000 (174.000-595.000)	299.000 (96.000-408.000)	0.312	78.000 (28.000–78.000)	336.000 (157.500-426.500)	0.102
Treatment duration (days)	27 (14–34)	15 (12–28)	0.353	4 (4–4)	16 (14–31)	0.020
Surgery (%, n)	28.6% (n = 2)	21.1% (n = 4)	1.00	50.0% (n = 1)	20.8% (n = 5)	0.415
BSI (%, n)	28.6% (n = 2)	36.8% (n = 7)	1.00	50% (n = 1)	33.3% (n = 8)	1.00
Pneumonia (%, n)	57.1% (n = 4)	63.2% (n = 12)	1.00	50% (n = 1)	62.5% (n = 15)	1.00
Complications (%, n)	57.1% (n = 4)	84.2% (n = 16)	0.293	100.0% (n = 2)	75.0% (n = 18)	1.00
Mortality (%, n)	0% (n = 0)	10.5% (n = 2)	1.00	-	-	-

(CRP: C-reactive protein, TLC: Total leukocyte count, ANC: Absolute neutrophil count, ALC: Absolute lymphocyte count, BSI: Bloodstream infection)

26.9% (n = 7) of the episodes were recurrent infections. Three of these occurred in the lungs, two were in the CNS, one in the blood, and one in both the blood and lungs. There was no significant difference in laboratory findings and treatment durations when comparing those with and without recurrent infections (p > 0.05). The shorter treatment duration in fatal episodes reflected early death during therapy rather than a planned shorter treatment course. No mortality was observed among recurrent infections (Table 2).

Comparison of pandrug-resistant Acinetobacter baumannii and pandrug-resistant Klebsiella pneumoniae infections

Patients with PDR K. pneumoniae infection were younger than those with PDR A. baumannii infection (0.38 vs. 8.9 years, p = 0.004). There was no difference between the two groups in terms of gender (p = 0.550). The duration of fever before bacterial growth and duration of fever after bacterial growth were longer in PDR A. baumannii infections than in PDR K. pneumoniae infections (p = 0.009 and p = 0.04, respectively). Although co-infections and complications were more frequent in PDR A. baumannii infections, we did not find a statistically significant difference (p = 0.06 and 0.08, respectively). When the infection sites were compared, pneumonia was frequently seen in PDR A. baumannii infections (65.3% vs. 47.4%; p = 0.227), and BSI was frequently seen in K. pneumoniae infections (34.6% vs. 42.1%; p = 0.609). However, these differences did not reach statistical significance. The comparison of demographic and clinical characteristics, laboratory findings, recurrence and mortality rates of the patients with PDR A. baumannii and K. pneumoniae infections was presented in Table 3.

Table 3.

The comparison of demographic features, clinical characteristics, laboratory findings, complications, and mortality rates of the patients with pandrug-resistant Klebsiella pneumoniae and pandrug-resistant Acinetobacter baumannii infections

	Pandrug-resistant Acinetobacter baumannii	Pandrug-resistant Klebsiella pneumoniae	p-value	Effect size	95% CI	Post-hoc power (%)
Age (years)	8.9 (4.43–15.74)	0.38 (0.14–5.59)	0.004	r = 0.70	0.36–0.88	93
Male sex	52.6% (n = 10)	41.7% (n = 5)	0.550	OR = 1.55	0.41–5.89	11
Any underlying comorbidity	100% (n = 19)	91.7% (n = 11)	1.000	OR = 1.56	0.06–41.9	5
Neurometabolic disease	31.6% (n = 6)	33.3% (n = 4)	1.000	OR = 1.13	0.17–2.50	10
Post-trauma	26.3% (n = 5)	0% (n = 0)	0.058	OR = 10.1	0.52–196.2	58
Congenital heart disease	10.5% (n = 2)	25.0% (n = 3)	0.625	OR = 0.42	0.06–3.17	10
Malignancy	15.8% (n = 3)	16.7% (n = 2)	1.000	OR = 1.10	0.16–7.56	5
Prematurity	5.3% (n = 1)	8.3% (n = 1)	1.000	OR = 0.71	0.04–12.6	5
Primary immunodeficiency	5.3% (n = 1)	0% (n = 0)	1.000	OR = 2.26	0.09–59.8	5
Hydronephrosis	0% (n = 0)	8.3% (n = 1)	0.420	OR = 0.30	0.01–7.93	8
Brain abscess	5.3% (n = 1)	0% (n = 0)	1.000	OR = 2.26	0.09–59.8	5
Hospitalization within the last 6 months	46.2% (n = 12)	52.6% (n = 10)	0.660	OR = 0.78	0.23–2.68	7
Hospitalization within the last month	23.1% (n = 6)	42.1% (n = 8)	0.170	OR = 0.32	0.08–1.26	37
Colonization	30.8% (n = 8)	26.3% (n = 5)	0.740	OR = 1.25	0.31–5.00	6
Presence of CVC	65.4% (n = 17)	52.6% (n = 10)	0.380	OR = 1.70	0.50–5.75	18
Recurrent infection	26.9% (n = 7)	36.8% (n = 7)	0.478	OR = 0.63	0.17–2.31	13
PICU	65.4% (n = 17)	63.2% (n = 12)	0.870	OR = 1.10	0.31–3.94	5
Co-infections	19.2% (n = 5)	0% (n = 0)	0.060	OR = 10.1	0.52–196.2	58
Complications	76.9% (n = 20)	52.6% (n = 10)	0.080	OR = 3.00	0.80–11.2	50
Septic shock	11.5% (n = 3)	21.1% (n = 4)	0.430	OR = 0.49	0.09–2.65	17
Mortality	7.7% (n = 2)	15.8% (n = 3)	0.630	OR = 0.44	0.06–3.17	10
Surgery	23.1% (n = 6)	10.5% (n = 2)	0.430	OR = 2.57	0.45–14.6	18
Duration of fever before bacterial growth (days)	2 (1-5.25)	1 (0–2)	0.009	r = 0.45	0.12–0.70	80
Duration of fever after bacterial growth (days)	3.5 (1–10)	1 (0–5)	0.040	r = 0.36	0.03–0.62	64
Treatment duration (days)	15 (13.5–29)	14 (10–16)	0.130	r = 0.20	–0.10–0.47	28
CRP (mg/L)	84.6 (31.3-150.2)	117.3 (67-189.3)	0.500	r = 0.12	–0.18–0.39	10
TLC/mm3	12.100 (8.975–16.550)	11.800 (9.200–14.500)	0.720	r = 0.06	–0.24–0.34	6
ANC/mm3	8.700 (6.050-14.225)	7.400 (5.800–10.600)	0.200	r = 0.22	–0.08–0.50	33
ALC/mm3	1.500 (1.075-1.900)	3.100 (800 − 5.400)	0.220	r = 0.22	–0.08–0.50	34
Hemoglobin (g/dL)	9.4 (8.3–11.2)	9.1 (7.5–9.8)	0.450	r = 0.13	–0.17–0.40	13
Platelets/mm3	315.000 (120.000-417.500)	285.000 (235.000-395.000)	0.970	r = 0.01	–0.28–0.30	5

CVC: Central venous catheter, PICU: Pediatric intensive care unit, TLC: Total leukocyte count, ANC: Absolute neutrophil count, ALC: Absolute lymphocyte count, CRP: C-reactive protein, CI: confidence intervals. Values are presented as median (interquartile range) or n (%). Effect sizes are expressed as rank-biserial correlation coefficients (r) for continuous variables and odds ratios (OR) for categorical variables. Post-hoc power values were calculated based on the observed effect sizes

Age, sex, and underlying comorbidities were analyzed at the patient level (PDR A. baumannii, n = 19 patients; PDR K. pneumoniae, n = 12 patients). Infection-related characteristics, treatment variables, recurrence, and outcomes were analyzed at the episode level (PDR A. baumannii, n = 26 episodes; PDR K. pneumoniae, n = 19 episodes)

In the exploratory multivariable logistic regression analysis, younger age remained independently associated with PDR K. pneumoniae infection. Increasing age was associated with lower odds of PDR K. pneumoniae infection compared with PDR A. baumannii infection (aOR: 0.746; 95% CI: 0.613–0.906; p = 0.003). Duration of fever before bacterial growth and complications were not significant in the adjusted model (Table 4).

Table 4.

Multivariable logistic regression analysis for factors associated with pandrug-resistant Klebsiella pneumoniae infection

Variable	Adjusted OR	95% CI	p-value
Age (years)	0.746	0.613–0.906	0.003
Duration of fever before bacterial growth (days)	0.771	0.518–1.149	0.202
Complications	3.937	0.708–21.878	0.117

Variables with a p-value < 0.10 in the univariate analyses and clinically relevant variables were entered into the multivariable binary logistic regression model. The dependent variable was bacterial species (0 = pandrug-resistant Acinetobacter baumannii, 1 = pandrug-resistant Klebsiella pneumoniae). Adjusted odds ratios (aORs) with corresponding 95% confidence intervals (95% CIs) are presented. Model fit statistics were as follows: omnibus χ² = 23.102 (df = 3, p < 0.001), Cox and Snell R² = 0.402, and Nagelkerke R² = 0.540

Discussion

Both A. baumannii and K. pneumoniae are well-recognized causes of invasive nosocomial infections, most commonly VAP and CLABSI [7, 19]. Castillo Bejarano et al. reported that VAP and CLABSI were the leading infection foci in pediatric patients with carbapenem-resistant A. baumannii, diagnosed at a mean of 20 days after hospitalization [20]. Similarly, Havan et al. showed that VAP and CLABSI were the predominant infection sites in PICU patients with PDR K. pneumoniae [10]. Consistent with these reports, VAP and bloodstream infection were the most frequent clinical presentations in our cohort, with median diagnostic times of 17 and 21 days after hospitalization for A. baumannii and K. pneumoniae, respectively. These findings emphasize that resistant nosocomial pathogens should be strongly suspected in critically ill children with prolonged hospitalization (> 14 days), and that early empiric therapy targeting these organisms may be warranted in high-risk settings.

Huang et al., in a comprehensive pediatric study on respiratory tract infections, showed that C. albicans was usually detected together with other pathogens rather than as a single organism, and that A. baumannii was prominent among the co-pathogens identified in PICU cases [21]. Xiaojiang Tan et al. demonstrated that VAP caused by A. baumannii was more frequent in patients whose respiratory tract was colonized with Candida species [22]. Although co-infections and complications were numerically more frequent in the PDR A. baumannii group, these comparisons did not reach statistical significance and should be interpreted with caution because of the small sample size and limited statistical power. Notably, 60% of these co-infections were fungal. While this observation is clinically suggestive, larger prospective studies are required to determine whether PDR A. baumannii infection truly increases susceptibility to invasive fungal disease or merely reflects host severity and antibiotic exposure.

Data on CZA resistance in pediatric K. pneumoniae infections remain limited, and reported rates vary considerably according to center, age group, and carbapenemase type. Most available evidence comes from carbapenem-resistant K. pneumoniae (CRKP) cohorts rather than specifically from pandrug-resistant isolates [23]. In the largest pediatric series reported from China, KPC-producing isolates were fully susceptible to CZA, whereas aztreonam-avibactam rather than CZA showed activity against NDM-producing isolates [23]. In contrast, in a neonatal intensive care unit study from Egypt, 32% of K. pneumoniae isolates were CRKP, and CZA resistance was reported in 77.8% of serine carbapenemase-producing isolates and 100% of metallo-β-lactamase-producing isolates [24]. Additional pediatric evidence is mostly limited to case-based data; for example, Zhou et al. reported that two of 12 initially KPC-positive and CZA-susceptible K. pneumoniae isolates in a PICU developed CZA resistance during treatment [25]. Adult and surveillance data also show variable resistance rates: CZA susceptibility among CRKP isolates was 63.3% in a global surveillance study, whereas Turkish CRKP studies reported CZA resistance rates of 42.8% among bloodstream isolates and 21.5% among hospital isolates [26–28]. Against this background, the 63.2% CZA resistance rate observed in our PDR K. pneumoniae isolates appears particularly high and clinically concerning. Moreover, all recurrent PDR K. pneumoniae episodes in our cohort were CZA-resistant, suggesting that limited therapeutic options and possible acquired or persistent resistance may contribute to recurrence in this vulnerable population. Potential mechanisms of CZA resistance include metallo-β-lactamase production, which is not inhibited by avibactam, as well as altered blaKPC variants, increased β-lactamase expression, and reduced outer-membrane permeability related to porin alterations. Local carbapenemase epidemiology may therefore have substantially influenced CZA activity in our cohort. However, because sequential isolates were not subjected to molecular characterization or clonal relatedness analysis, it is not possible to determine whether these episodes represented treatment-emergent resistance, persistence of a pre-existing CZA-resistant strain, or reinfection with a related resistant clone. Because CZA susceptibility could not be evaluated in three isolates and molecular characterization was not available, the mechanisms underlying CZA resistance and its relationship with recurrence should be interpreted cautiously.

Mortality rates reported for pediatric A. baumannii infections range widely, from 17% to 63% [29–34]. Whether carbapenem resistance independently increases mortality risk remains controversial [35]. Differences in patient populations, comorbidity burden, and treatment strategies likely contribute to these discrepancies. Experimental data suggest that specific antimicrobial resistance genes and virulence determinants correlate with mortality [36]. Conversely, some highly resistant strains may exhibit reduced virulence or slower growth, potentially limiting their clinical impact [11]. The mortality rate in our study was 7.7%, which is lower than most published pediatric series. The relatively low mortality observed in our cohort is noteworthy. This may partly reflect early targeted empirical combination therapy guided by local resistance epidemiology, close pediatric infectious diseases follow up, and timely source control when required. Nevertheless, this interpretation should be cautious, as the small sample size, retrospective design, and lack of molecular characterization prevent firm conclusions regarding the relative contribution of treatment strategy. Furthermore, although all isolates were phenotypically classified as PDR, molecular characterization was not available; therefore, we could not determine whether the isolates carried specific resistance genes, virulence determinants, or clonal features associated with more severe clinical outcomes. The absence of molecular data also limits our ability to distinguish whether the relatively low mortality reflected pathogen-related factors, such as reduced virulence or strain-specific characteristics, or host and treatment-related factors. Future studies integrating molecular typing, resistance gene analysis, and virulence profiling are needed to better clarify the relationship between bacterial genotype and clinical outcomes in pediatric PDR A. baumannii infections.

For carbapenem-resistant Enterobacteriaceae infections, a meta-analysis reported mortality rates of 26–44% [37]. Havan et al. documented a mortality rate of 14.2% in pediatric PDR K. pneumoniae infections [10]. Our observed rate of 15.8% is consistent with these findings. Although mortality appeared numerically higher in PDR K. pneumoniae than in PDR A. baumannii, the difference was not statistically significant, suggesting comparable overall severity between these pathogens in pediatric populations.

Recurrence is an increasingly important clinical challenge, particularly in patients with chronic disease and long-term invasive devices. The risk depends on pathogen characteristics, antimicrobial resistance profiles, and host factors [38]. Studies in adults with multidrug-resistant A. baumannii report recurrence rates between 19.2% and 50% [39, 40], although pediatric outcome data remain scarce [41–46]. We observed a recurrence rate of 26.9% in PDR A. baumannii infections, with no significant differences in laboratory parameters or treatment duration between recurrent and non-recurrent cases. For K. pneumoniae, Huang et al. reported a recurrence rate of 4.2% for bacteremia in adults [47], and Giannella et al. reported 16.6% for carbapenem-resistant strains [48]. Our recurrence rate of 36.8% in pediatric PDR K. pneumoniae infections is notably higher than adult estimates. This finding suggests that children may represent a particularly vulnerable population and should be closely monitored after treatment completion. The recurrence rate was substantial in both groups, particularly in PDR K. pneumoniae infections. This may reflect persistent colonization pressure, repeated antimicrobial exposure, prolonged hospitalization, and ongoing risks related to invasive devices such as central venous catheters or mechanical ventilation. The high burden of underlying comorbidities may also contribute to impaired bacterial clearance and vulnerability to recurrent infection. These findings suggest that recurrence may be influenced not only by treatment duration but also by host fragility and persistent healthcare-associated exposure. Importantly, all patients with recurrent infections survived in our cohort, which may reflect host factors, pathogen virulence characteristics, or prompt re-treatment with effective combinations. Treatment duration did not differ in recurrent cases, indicating that prolonged therapy alone may not prevent relapse.

To our knowledge, this is the first study to directly compare clinical outcomes of PDR A. baumannii and PDR K. pneumoniae infections in children. The principal limitations are the relatively small sample size and retrospective design. However, such infections remain rare in pediatric populations, making large prospective studies difficult. In addition, molecular characterization of resistance genes and virulence determinants was not available, which limits mechanistic interpretation. Despite these constraints, our findings provide clinically relevant comparative data and may inform empirical management strategies and future research on pediatric PDR infections. Another methodological consideration is that some patients contributed more than one infection episode. Although demographic variables were analyzed at the patient level and duplicate isolates from the same episode were excluded, episode-level analyses may still be affected by residual within-patient correlation. Because of the small sample size, clustered or mixed-effects analyses were not feasible; therefore, these comparisons should be interpreted as exploratory. In addition, mortality was assessed as all-cause mortality during the infection episode. Because of the retrospective design and the high burden of underlying comorbidities, infection-attributable mortality could not be reliably distinguished from mortality related to underlying disease or other concurrent clinical conditions. Although some laboratory parameters differed significantly between fatal and non-fatal episodes in both of the PDR K. pneumoniae group and the PDR A. baumannii group, these findings should be interpreted cautiously because they were based on the limited sample size. These analyses should be regarded as exploratory and should not be interpreted as evidence of independent predictors of mortality. In addition, duration of PICU stay and mechanical ventilation, as well as validated clinical severity scores, were not evaluated in this study. Consequently, unmeasured differences in baseline clinical severity between the groups cannot be excluded. Because of the small sample size, the exploratory multivariable model included only a limited number of covariates. Therefore, residual confounding from unmeasured factors might had an impact on the outcomes.

Conclusions

Pandrug-resistant A. baumannii and K. pneumoniae infections were associated with considerable clinical severity and all-cause mortality in children. K. pneumoniae affected younger patients, whereas A. baumannii infections were associated with longer fever duration and a tendency toward more co-infections. Both pathogens were typically detected after prolonged hospitalization and were strongly linked to intensive care exposure. Recurrence rates were substantial for both organisms, underscoring the need for close follow-up after treatment. These findings highlight the importance of early suspicion and timely targeted therapy for pandrug-resistant infections in high-risk pediatric patients. The high rate of CZA resistance among PDR K. pneumoniae isolates, including all recurrent PDR K. pneumoniae episodes, was a clinically important finding that may further limit therapeutic options.

Abbreviations

AMR

Antimicrobial resistance

BSI

Bloodstream infection

CLABSI

Central line-associated bloodstream infection

CNS

Central nervous system

CZA

Ceftazidime-avibactam

EUCAST

European Committee on Antimicrobial Susceptibility Testing

MALDI-TOF MS

Matrix-assisted laser desorption/ionization time-of-flight mass spectrometry

MDR

Multidrug resistant

MIC

Minimum inhibitory concentration

PDR

Pandrug resistant

PICU

Pediatric intensive care unit

SAM

Ampicillin-sulbactam

UTI

Urinary tract infection

VAP

Ventilator-associated pneumonia

Author contributions

ICE and NB conceived the study. ICE drafted the manuscript. ŞÖA, AÇG, HKA, EA, AHB, and MCE contributed to data acquisition, microbiological evaluation, and interpretation of the findings. NB supervised the study. All authors critically revised the manuscript, approved the final version, and agree to be accountable for all aspects of the work.

Funding

This study received no specific funding.

Data availability

The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

The study was approved by the Ethics Committee of Dokuz Eylül University (protocol number: 2023/31-17). Due to the retrospective design, the requirement for informed consent was waived. All procedures were conducted in accordance with the principles of the Declaration of Helsinki.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.