Predicting carbapenem-resistant Enterobacteriaceae infections in pediatric liver transplant recipients

The present study complied with the Declaration of Helsinki and was approved by the Ethics Committee of the First Affiliated Hospital, Zhejiang University School of Medicine (FAHZU, IIT20240071A). All organs were donated voluntarily with written informed consent. Medical records of pediatric patients (< 12 years) who underwent liver transplantation between January 2020 and June 2023 in FAHZU (center 1) and those of patients who presented to Tianjin First Central Hospital (center 2) between August 2017 and August 2018 were reviewed retrospectively. Pediatric patients without rectal microbiological cultures before transplantation were excluded. Data for each recipient was included only once, and for patients with two transplantations in the study period, the data from the second transplantation was analyzed.

Baseline data included age (months), weight, gender, primary disease (transplant indications), pediatric end-stage liver disease scores, pre-transplant CRE colonization, pre-transplant intensive care unit (ICU) stay and operative parameters. In addition, post-transplant complications, CRE infection status and clinical outcomes were recorded. Respiratory ribonucleic acid (RNA) virus infection data in transplant recipients was also collected. Respiratory RNA virus infections that occurred before CRE infections were confirmed through nasopharyngeal swabs or secretions from tracheal intubation using polymerase chain reaction (PCR) or high-throughput (next generation) sequencing. The viral spectrum identified by PCR comprised influenza A and B, parainfluenza (I–III), respiratory syncytial viruses and severe acute respiratory syndrome coronavirus.

According to guidelines provided by the Centers for Disease Control and Prevention, CRE are identified as those Enterobacteriaceae non-susceptible to imipenem, meropenem, doripenem or ertapenem, or those strains that have been documented to possess a carbapenemase [13]. Rectal swabs were used to detect intestinal CRE colonization before liver transplantation. CRE colonization was identified according to the guidelines of the 31st edition of the Clinical and Laboratory Standards Institute Performance Standards for Antimicrobial Susceptibility Testing, M100 [14]. Specifically, a minimum inhibitory concentration of imipenem, meropenem or doripenem ≥ 4 mg/L or ertapenem ≥ 2 mg/L was used to confirm CRE. Laboratory detection of CRE along with the appearance of infection symptoms were confirmation of infection. Infections that occurred within 90 days post-transplant were identified as postoperative CRE infections.

Elevated serum transaminase indicated the possibility of acute rejection. Subsequently, experienced pathologists confirmed acute rejection based on the Banff rejection activity index (RAI) [15]; an RAI ≥ 3 identified acute rejection.

Categorical variables were presented using frequencies and percentages and were compared using the Chi-squared or Fisher’s exact tests. Continuous variables were summarized as means or medians as appropriate. Mann–Whitney U test was used to compare continuous variables. Logistic regression analyses were utilized to identify independent risk factors associated with CRE infections after liver transplantation. Factors with P values < 0.01 chosen after univariate analysis were included in subsequent multivariate analysis. Nomograms were constructed based on identified independent factors. Performance of the prediction models were evaluated using receiver operating characteristic curves (ROC) and area under the ROC curve (AUC). All statistical analyses were performed using R 4.1.2, SPSS v26.0 (IBM, Armonk, NY) and Prism 8.0 (GraphPad Soft Inc., San Diego, CA) software programs. A two-sided P value < 0.05 was considered statistically significant.

Data from 778 pediatric patients who underwent liver transplantation between August 2017 and June 2023 was included in this study. Six hundred and twenty-six pediatric patients who presented to FAHZU during the study period were randomly divided into training and internal validation cohorts at a ratio of 7:3, adhering to established machine learning conventions and domain-specific standards (n = 438 and 188, respectively; Fig. 1). This sample size satisfies the 10 events per variable principle, ensuring adequate statistical power for model development [16]. In addition, 152 pediatric patients from the Tianjin First Central Hospital were included as an external validation cohort. More pediatric patients were female and had CRE intestinal colonization in the external validation cohort than those in the training cohort (P = 0.001 and P < 0.001, Table 1). Patients in the external validation cohort exhibited higher PLED scores (P = 0.006). In the training cohort, fifty (11.4%) patients presented with CRE infections after transplantation; the remaining 388 (88.6%) patients were not infected. The number of patients with CRE infections was 23 (12.2%) and 13 (8.6%) in the internal and external validation cohorts, respectively (P = 0.524), demonstrating comparability between cohorts. The majority of baseline characteristics in patients with CRE infections were similar to those not infected (Table 2). However, bile or intestinal leakage, respiratory RNA virus infection and CRE intestinal colonization were observed more frequently in pediatric recipients with CRE infections than those not infected (P < 0.001, P = 0.005, and P < 0.001, respectively). Cefoperazone-sulbactam serves as the first-line agent for perioperative antibiotic prophylaxis, with meropenem used rarely where cefoperazone-sulbactam allergy is confirmed. This selection aligns with established clinical recommendations from prior evidence-based studies [17, 18].

Bild vergrößern

Univariate logistic regression analysis revealed potential risk factors. These included age, weight, CRE rectal colonization before liver transplantation, red blood cell transfusion, plasma transfusion, bile or intestinal leakage and respiratory RNA virus infection (P = 0.057, P = 0.041, P < 0.001, P = 0.021, P = 0.058, P < 0.001 and P = 0.006, respectively; Table 3). Furthermore, multivariate analysis identified CRE intestinal colonization before liver transplantation [odds ratio (OR) = 4.208, 95% confidence interval (CI) = 1.940–9.129; P < 0.001], bile or intestinal leakage (OR = 9.226; 95% CI = 3.774–22.556; P < 0.001) and respiratory RNA virus infection (OR = 2.205, 95% CI = 1.005–4.840; P = 0.049) as independent risk factors associated with post-transplant CRE infections. In addition, variance inflation factors (VIF) confirmed the absence of multicollinearity between clinically correlated variables: specifically, VIF = 1 for both the age-body weight pair and red blood cell-plasma transfusion pair.

We constructed a nomogram (model 1) based on independent risk factors including CRE intestinal colonization before liver transplantation, bile or intestinal leakage and respiratory RNA virus infection to predict CRE infections in pediatric patients after liver transplantation (Fig. 2a). CRE intestinal colonization before liver transplantation and bile/intestinal leakage contributed most to the overall risk score in model 1. Therefore, a second nomogram (model 2) was constructed using only these two factors (Fig. 2b). AUCs of model 1 were 0.724 (95% CI = 0.646–0.802) and 0.738 (95% CI = 0.623–0.852) in the training and internal validation cohorts, respectively (Fig. 3a). Model 2 exhibited AUC values of 0.702 (95% CI = 0.628–0.775) and 0.738 (95% CI = 0.624–0.853) in the training and internal validation cohorts, respectively (Fig. 3b). Furthermore, the performance of model 2 was confirmed using an external validation cohort with an AUC of 0.828 (95% CI = 0.738–0.919, Fig. 3b). Model 1 demonstrated Brier scores of 0.086 and 0.088 in the training and internal validation cohorts, respectively. Model 2 showed Brier scores of 0.088, 0.088, and 0.071 in the training, internal validation, and external validation cohorts, respectively. Brier scores showed generally good alignment between the observed and predicted risks of CRE infection. In addition, we performed subgroup analyses based on operation time, using the median cutoff of 6 hours. In the external validation cohort, model 2 demonstrated superior performance for pediatric patients with prolonged operation time, achieving an AUC of 0.864 (95% CI = 0.840–0.888). In contrast, for shorter procedures, it showed reduced predictive accuracy with an AUC of 0.690 (95% CI = 0.658–0.722). Collectively this indicated that this model may perform better in pediatric patients with longer operative time.

Bild vergrößern

Bild vergrößern

After liver transplantation, there were no significant differences in acute rejection between patients with CRE infections and those not infected (4.0% vs. 6.2%, P = 0.755; Table 4). Patients who developed CRE infections often underwent longer periods of mechanical ventilation and post-transplant ICU stay (both P < 0.001). Seven patients with CRE infections died within 180 days of liver transplantation; eight patients not infected died over the same period (P < 0.001).