Early Post-Transplant Bacterial Infections Predict Long-Term Mortality in Kidney Transplant Recipients: A 10-Year Retrospective Cohort Study

Introduction

Kidney transplantation remains the most effective therapeutic option for patients with end-stage kidney disease, yielding improved survival and life expectancy in this population [1,2]. However, lifelong immunosuppression increases vulnerability to infectious complications, which continue to be a leading cause of both morbidity and mortality. Among these, bacterial infections – particularly urinary tract infections (UTIs) – are the most prevalent, affecting up to 70% of recipients within the first 3 years after transplant [3–5]. Although many of these infections are mild or self-limited, severe presentations may trigger systemic inflammatory responses, predispose to sepsis, compromise graft viability, and negatively impact overall survival [6].

Despite advances in antimicrobial prophylaxis and tailored immunosuppressive regimens, post-transplant bacterial infections remain a persistent clinical challenge, particularly in the context of rising antimicrobial resistance worldwide [7]. The role of antibiogram-guided therapy in reducing multidrug resistance (MDR), optimizing efficacy, and promoting microbiological stewardship has been proposed, but remains insufficiently validated in transplant-specific cohorts [8,9]. In addition, non-immunological contributors – including malnutrition, anemia, and urological interventions such as ureteral stent placement – can further increase susceptibility to infection, yet these variables are underexplored in longitudinal studies [6].

The early post-transplant period (≤6 months) is a critical window characterized by intensive immunosuppression and high infection risk [5]. Although the link between infection and mortality is recognized, few studies have systematically examined early bacterial infections as independent predictors of long-term mortality. Furthermore, the contribution of temporary urological devices to adverse outcomes remains a subject of clinical interest.

We hypothesized that early bacterial infections during the first 6 months after transplantation are independently associated with long-term mortality and are influenced by both immunological and non-immunological clinical factors.

Material and Methods

STUDY POPULATION AND PATIENT SELECTION:

The present retrospective cohort included kidney transplant recipients who underwent transplantation at Çanakkale 18 Mart University Organ Transplant Center between January 2014 and January 2024. Patients with a minimum follow-up of 12 months were included, except for those who died in the early post-transplant period (defined as ≤6 months), as early deaths would preclude proper classification of the exposure of interest – early bacterial infection (≤6 months). For mortality analysis using Cox regression, only patients with a follow-up duration of at least 30 months were further evaluated. The study conforms to the STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) guidelines for reporting observational research, ensuring a methodologically rigorous and transparent research design [10]. The overall patient selection process, exclusions, and infection distribution are summarized in Figure 1A.

The study was conducted in accordance with the ethical standards of the institutional and national research committees and with the most recent version of the Declaration of Helsinki. The study protocol was approved by the Çanakkale Onsekiz Mart University Clinical Research Ethics Committee (approval number: 2024/08–13). As this was a retrospective analysis of anonymized clinical data, the requirement for informed consent was waived by the ethics committee.

The first 6 months after kidney transplantation is a critical period characterized by intensified immunosuppression and increased susceptibility to infections. Studies have suggested that early infections are associated with higher mortality and graft dysfunction, whereas later infections often develop under a more stable immunosuppressive regimen [11]. To ensure a focused analysis, the present study primarily examined early bacterial infections (≤6 months). Patients who developed infections in both early and late periods were classified based on their first documented infection, as the primary objective was to assess the impact of early infections on outcomes. This classification aligns with previous reports and reflects the progressive tapering of immunosuppression, distinguishing infections with potentially different prognostic implications.

Patients with asymptomatic bacteriuria (ABU) were not included in the analysis, as the study focused on clinically significant bacterial infections requiring targeted treatment. Only 1 patient met the criteria for ABU and was excluded accordingly.

To enhance the reliability of mortality estimates, patients with less than 30 months of follow-up who remained alive were excluded. This decision minimized the risk of overestimating survival duration in patients with shorter follow-up periods. To further ensure temporal clarity, only patients who survived the first 6 months after transplant were included in the mortality analysis. This approach also served as a built-in sensitivity analysis, helping to mitigate immortal time bias and strengthen the temporal validity of the mortality analysis. As such, all deaths occurred after the early infection window, supporting a temporal sequence between early infections and subsequent mortality. This exclusion strategy effectively functioned as a sensitivity analysis to mitigate immortal time bias by ensuring that mortality events could be temporally attributed to early infections.

Baseline characteristics were initially assessed using univariable analyses, followed by logistic regression for infection risk factors and Cox regression for mortality analysis. To maintain model parsimony and prevent overfitting, only clinically relevant and statistically significant variables from univariable analyses were included in the final multivariable models. Variables that showed no meaningful association in univariable analyses were excluded to improve interpretability and robustness.

STUDY AND CONTROL GROUPS:

Recipients with microbiologically confirmed bacterial infections constituted the study cohort, whereas the control group consisted exclusively of recipients who remained entirely free of microbiologically confirmed bacterial infections throughout the follow-up period. Hospitalizations for any non-infectious reason were not considered in group classification, as only the presence or absence of bacterial infection determined group allocation. To ensure a targeted evaluation, cases involving fungal or viral infections were systematically excluded from the analysis.

EXCLUSION CRITERIA:

The following criteria led to exclusion from the study: age below 18 years, pregnancy during transplantation or throughout the follow-up period, documented history of malignancy, receipt of simultaneous multi-organ transplantation, and mortality occurring in the early post-transplant phase.

CLINICAL AND LABORATORY DATA COLLECTION:

Clinical and microbiological information was retrospectively obtained from institutional electronic health records and laboratory databases. Bacterial infections were verified through standard microbiological techniques, including urine and blood cultures. Antimicrobial susceptibility profiles of isolated organisms were assessed via standardized sensitivity assays, and resistance classifications were made according to the presence of extended-spectrum beta-lactamase (ESBL) production and carbapenem resistance.

For each patient, the following laboratory parameters were documented: complete blood count (hemoglobin, leukocyte, and platelet counts), serum creatinine, blood urea nitrogen (BUN), estimated glomerular filtration rate (eGFR) using the CKD-EPI equation, serum albumin, uric acid, liver enzymes (aspartate aminotransferase [AST], alanine aminotransferase [ALT]), bilirubin levels (total and direct), alkaline phosphatase (ALP), gamma-glutamyl transferase (GGT), inflammatory markers including C-reactive protein (CRP) and, for infected individuals, procalcitonin. Tacrolimus trough levels were also recorded to assess immunosuppressive exposure.

IMMUNOSUPPRESSIVE THERAPY AND INFECTION PROPHYLAXIS:

Immunosuppressive and prophylactic regimens were retrospectively reviewed based on the standard clinical protocols in use at the time of transplantation. Most patients had received triple immunosuppressive therapy consisting of tacrolimus (target trough levels: 8–10 ng/mL in the first month, 6–8 ng/mL thereafter), mycophenolate mofetil (750–1000 mg twice daily), and corticosteroids tapered to 5 mg/day.

Infection prophylaxis typically included trimethoprim-sulfamethoxazole for Pneumocystis jirovecii pneumonia (6–12 months), valganciclovir for cytomegalovirus (3–6 months), and nystatin for fungal prophylaxis (3–6 months), in accordance with KDIGO guidelines [12,13].

Adjustments to immunosuppression, including dose modifications or drug discontinuation, were made as clinically indicated (eg, neutropenia, infections), but were not included in the analysis due to their individualized nature.

STATISTICAL ANALYSIS:

Statistical analyses were conducted using Jamovi v2.3.28.0. Kaplan-Meier estimators and Cox proportional hazards models were used to assess survival outcomes, while logistic regression was applied to identify early infection predictors. Model variables were selected based on clinical relevance and univariable screening (P<0.10), ensuring that clinically important predictors were retained even if they were not statistically significant in univariable analyses, while keeping the model sufficiently parsimonious given the limited number of mortality events.

SAMPLE SIZE JUSTIFICATION:

A priori assessment of sample adequacy was performed before model development. The initial power estimation indicated that a cohort of approximately 96 recipients would provide 80% power (α=0.05) to detect a large effect for the association between early infection and mortality. For the Cox regression model, sample adequacy was further evaluated in line with widely accepted methodological recommendations for survival analysis, which emphasize maintaining an appropriate relationship between the number of observed events and the number of variables included in the model. Accordingly, the final multivariable model was intentionally kept parsimonious and restricted to clinically relevant predictors. Sample size estimation was performed using G*Power version 3.1 (Heinrich-Heine-Universität Düsseldorf, Germany).

EARLY INFECTION RISK MODELING (≤6 MONTHS):

To determine independent variables associated with early post-transplant infections (≤6 months), a multivariable logistic regression model was fitted. Covariates included were selected based on clinical plausibility and statistical significance (P<0.10) in univariable analyses. Model discrimination was quantified by the area under the receiver operating characteristic curve (AUC). Classification performance was further characterized through calculated values for accuracy, sensitivity, and specificity. The model’s explanatory strength was quantified using McFadden’s pseudo-R² coefficient.

MORTALITY RISK MODELING (≥30-MONTH FOLLOW-UP):

This analysis aimed to identify independent predictors of post-transplant mortality using a Cox proportional hazards regression model. The model included patients who had at least 30 months of follow-up. Variables were selected based on clinical relevance and statistical significance observed in univariable analyses (P<0.10). Model discrimination was assessed by the concordance index (C-index). The likelihood ratio test was used to evaluate model fit, and the R² statistic was calculated to quantify the proportion of explained variance.

Results

PATIENT CHARACTERISTICS AND INFECTION DISTRIBUTION:

A total of 123 individuals who underwent kidney transplantation were retrospectively reviewed. Of these, 27 were excluded due to inadequate follow-up (n=22) or early post-transplant mortality (n=5), yielding a final analytic cohort of 96 patients. Most were male (n=63, 66%), and most grafts were from deceased donors (n=85, 88%). The median follow-up duration was 52.5 months (interquartile range [IQR]: 38.0–74.8).

Over the course of follow-up, 47 patients (49%) experienced at least 1 episode of culture-confirmed bacterial infection, totaling 60 distinct events. Of these, 58% (n=35) occurred within the first 6 months after transplantation – a period characterized by heightened immunosuppressive intensity. Recurrent urinary tract infections (UTIs), defined as ≥2 episodes within 1 year, were documented in 8 patients.

Bacterial infections imposed a notable burden among kidney transplant recipients. Urinary tract infections (UTIs) were the most frequent, comprising 67% of all infection episodes, with cystitis accounting for the majority (66.7%). Non-urinary infections included pyelonephritis (18.3%), pneumonia (10.0%), and sepsis (1.7%). The distribution of infection types is presented in Figure 1B. Despite their lower incidence, non-urinary infections contributed substantially to clinical complexity and necessitated organism-specific antimicrobial treatment strategies (Table 1, Figure 1B). These findings highlight the predominance of gram-negative pathogens during the early post-transplant period.

RISK FACTORS FOR EARLY POST-TRANSPLANT BACTERIAL INFECTIONS:

Patients who developed early bacterial infections (n=34) had significantly lower eGFR (30.5 vs 41.0 mL/min/1.73 m2, P=0.04), lower hemoglobin levels (9.45 vs 12.5 g/dL, P<0.001), and lower albumin levels (3.62 vs 4.5 g/dL, P<0.001) compared to patients without early infections (n=62) (Table 2).

The presence of a urethral stent was strongly associated with infection risk (50.5% vs 11.3%, P<0.001). Diabetes mellitus showed a borderline association (47.0% vs 27.9%, P=0.06). No significant differences were observed in tacrolimus levels, donor age, or liver function tests between groups (Table 2).

BACTERIAL PATHOGENS AND ANTIBIOTIC USE:

Most isolates were gram-negative (69%), with Escherichia coli, Enterococcus spp., and Klebsiella pneumoniae as the most common urinary pathogens. Bloodstream infections frequently involved ESBL-producing organisms. Detailed distributions of cultured bacteria and antibiotic regimens are presented in Table 3.

MULTIVARIABLE ANALYSIS AND RISK FACTORS FOR EARLY POST-TRANSPLANT INFECTIONS:

Multivariable logistic regression analysis identified 3 independent predictors of early post-transplant infections (≤6 months): low serum albumin (OR=0.078, P<0.001), low hemoglobin (OR=0.696, P=0.04), and presence of a ureteral stent (OR=11.626, P=0.009). Recipient age, donor age, and tacrolimus trough levels did not show statistical significance. The model demonstrated excellent discriminative performance (AUC=0.905, accuracy=83.3%). Figure 2 illustrates these results in a forest plot highlighting significant predictors in red. Additional regression details, including full coefficient estimates and confidence intervals, are presented in Table 4.

INDEPENDENT PREDICTORS OF MORTALITY:

A multivariable Cox regression model identified early bacterial infections (≤6 months) as a significant independent predictor of all-cause mortality (adjusted hazard ratio [aHR]=3.03; 95% confidence interval [CI]: 1.63–6.98; P=0.004). End-stage kidney disease due to diabetic nephropathy was also independently associated with increased mortality risk (aHR=4.14; 95% CI: 1.61–10.64; P=0.003).

Older recipient age had a strong and incremental association with mortality (aHR=1.10 per year; 95% CI: 1.06–1.14; P<0.001), while donor age was not significantly associated (aHR=1.01; 95% CI: 0.99–1.03; P=0.169). Chronic obstructive pulmonary disease (COPD) was significant in univariable analysis but lost significance after adjustment.

Bacterial urinary tract infections were not significantly associated with mortality, suggesting that more severe systemic infections – such as bacteremia or pneumonia – drive the observed outcomes. Model performance was strong, with a concordance index of 0.821 and R2=0.462. Further details are provided in Figure 3 and Table 5.

Kaplan-Meier survival curves demonstrated significantly reduced survival probabilities in recipients with early bacterial infections compared to those without (log-rank P=0.01; Figure 4). A detailed comparison of clinical and laboratory parameters between survivors and non-survivors is provided in Table 6.

SUMMARY OF RESULTS:

Early bacterial infections were common and represented 58% of all infection episodes. Low albumin, low hemoglobin, and ureteral stent placement were the strongest predictors of early infection. Early bacterial infections, diabetic nephropathy-related ESKD, and older recipient age independently predicted long-term mortality. The final Cox model demonstrated high discriminative performance (C-index: 0.821). No multidrug-resistant organisms were identified.

Discussion

EARLY POST-TRANSPLANT PERIOD AS A CRITICAL WINDOW FOR BACTERIAL INFECTIONS:

In our cohort, 60 bacterial infection episodes were recorded, of which 58% occurred within the first 6 months. This aligns with the well-established susceptibility associated with early intense immunosuppression, surgical exposure, and urological instrumentation. Consistent with the literature, this time frame remains the most vulnerable for opportunistic and nosocomial infections [14–17]. Interventions in this period, such as immunosuppressive adjustment and urological device protocols, can significantly affect patient outcomes [18].

SUSCEPTIBILITY FACTORS FOR EARLY INFECTION:

Multivariable analysis identified low serum albumin, low hemoglobin, and ureteral stent placement as independent predictors of early infection. In our multivariable model, each 1 g/dL increase in serum albumin was associated with a 92% reduction in the odds of early infection (OR=0.08), underscoring the clinical importance of nutritional status in early susceptibility. These findings reflect the impact of nutritional and hematologic compromise on host defense, as supported by earlier reports [6].

The significant infection risk associated with ureteral stents (OR=11.63, P=0.009) likely stems from biofilm formation and ascending infections despite protocolized early removal, a concern echoed in prior studies [8]. In our center, ureteral stents are routinely placed in all kidney transplant recipients as part of the standard surgical protocol; therefore, the observed association is unlikely to reflect selective use in anatomically complex cases. The markedly elevated odds ratio is thus more plausibly explained by the inherent susceptibility of indwelling stents to bacterial colonization and biofilm formation under early post-transplant immunosuppression.

Tacrolimus levels were not significantly associated with early infection (P=0.988), paralleling earlier mixed findings on this subject [19,20]. While our infected patients showed a non-significant trend toward higher tacrolimus exposure, the overall data support continued therapeutic drug monitoring to balance infection and rejection risks.

Notably, predictors of early infection included non-immunologic factors such as hypoalbuminemia and ureteral stent presence, suggesting that susceptibility to infection was not solely driven by overall illness severity or baseline frailty. Baseline serum creatinine, liver function tests, and comorbidity burden did not differ significantly between patients with and without early infections, indicating that infection development was not merely a surrogate for pre-existing clinical deterioration.

INFECTION BURDEN AND PROGNOSTIC IMPLICATIONS:

Our data demonstrate that early bacterial infections independently predict long-term mortality (aHR=3.03, P=0.004), highlighting their role beyond immediate clinical morbidity. This observation adds to earlier evidence linking infection-related complications to graft dysfunction and patient death [21]. While some studies have focused on recurrent UTIs as predictors of poor graft outcomes [22], we did not observe a significant association between isolated UTIs and mortality. Instead, bloodstream infections and pneumonia likely drove the observed mortality risk. These results emphasize the importance of early etiology-based therapy and close clinical monitoring in the early post-transplant period.

A growing body of evidence suggests several biological mechanisms through which early severe infections – particularly bloodstream infections and pneumonia – can exert a lasting impact on post-transplant survival. Acute systemic inflammation can trigger irreversible alloreactive priming, thereby accelerating chronic immune activation. In parallel, infection-related inflammatory cascades can contribute to endothelial dysfunction and long-term cardiovascular risk, which remains a leading cause of mortality after transplantation. Additionally, severe early infections can induce persistent subclinical graft injury through ischemic or inflammatory pathways, potentially reducing long-term physiological reserve even in patients who experience apparent clinical recovery. These mechanisms may collectively explain why early infectious events translate into excess long-term mortality.

MORTALITY PREDICTORS: DIABETES, AGE, AND COMORBIDITIES:

Diabetic nephropathy as the cause of ESKD emerged as a strong independent mortality predictor (aHR=4.14, P=0.003), in line with its known systemic effects, including endothelial dysfunction, immune compromise, and elevated cardiovascular burden [23,24]. Age was similarly associated with mortality (aHR=1.10, P<0.001), reflecting the growing importance of frailty and immunosenescence in transplant outcomes [25]. Although COPD appeared significant in univariate analysis, it lost significance after adjustment, suggesting confounding by age or cumulative comorbidity.

MICROBIAL PATTERNS AND MULTIDRUG RESISTANCE:

Escherichia coli, Klebsiella spp., and Enterococcus spp. accounted for most urinary pathogens, while Pseudomonas aeruginosa and Enterococcus spp. dominated bloodstream infections. These findings reflect broader transplant microbiology trends [25]. Importantly, our cohort showed no MDR isolates, contrasting with global trends but aligning with recent local data [26–28]. We attribute this favorable finding to a strict antibiogram-guided approach and restraint in empirical broad-spectrum antibiotic use. Other studies emphasizing antimicrobial stewardship similarly reported lower MDR rates [29,30].

The antibiotics used – primarily ceftriaxone, meropenem, vancomycin, and imipenem – were selected based on local resistance patterns and severity of infection. This supports targeted therapy as a strategy to reduce MDR emergence while preserving efficacy [31,32].

Several limitations merit consideration. First, the single-center design may limit generalizability to different transplant populations. Second, the retrospective design carries inherent bias risks, despite adherence to standardized data procedures; reliance on electronic health records may have introduced misclassification of infection episodes or clinical variables. Third, although a uniform immunosuppressive protocol was used, individualized dose adjustments were not captured. Fourth, the absence of longitudinal resistance monitoring limited our ability to evaluate MDR evolution. Although a separate sensitivity analysis was not performed, excluding patients with less than 30 months of follow-up helped reduce immortal time bias.

Another limitation relates to the sample size and number of mortality events, which may restrict the stability and precision of the Cox model estimates. Although the final model was intentionally kept parsimonious to comply with recommended event-per-variable considerations in survival analysis, some uncertainty in effect estimates cannot be fully excluded.

Finally, our variable-selection strategy – based on clinical relevance combined with univariable screening (P<0.10) – may still result in residual confounding. This approach was chosen to avoid overparameterization given the limited number of events, but it is possible that some clinically important covariates that were non-significant in univariable screening may exert confounding effects only in multivariable contexts. This should be considered when interpreting our findings.

Conclusions

Although the retrospective design precludes definitive causal inference, the consistent association between early infections and mortality highlights a clinically actionable window. Enhanced infection surveillance and individualized prophylaxis during the early post-transplant period may help reduce long-term risk. This study identified early bacterial infections as strong independent predictors of long-term mortality in kidney transplant recipients. Diabetic nephropathy and recipient age also conferred excess mortality risk, while donor characteristics showed no significant impact. The role of ureteral stents in early infection development deserves further exploration and standardized protocols. Our findings underscore the importance of early risk stratification, vigilant infection control, and individualized immunosuppressive and antimicrobial strategies to optimize transplant outcomes. Based on these findings, several practical implications emerge. First, patients with hypoalbuminemia or anemia may benefit from early nutritional optimization and more frequent laboratory monitoring in the initial post-transplant period. Second, the strong association between ureteral stent placement and early infection supports consideration of shorter indwelling times and closer surveillance for stent-related complications. Third, implementing antibiogram-guided empirical therapy protocols – particularly in high-risk recipients – may further reduce severe infection episodes. Together, these approaches may enhance early risk stratification and enable individualized prophylaxis and follow-up pathways.