This was a single-center PROBE (Prospective Randomized Open Blinded Endpoint) trial conducted at The Brain Hospital of Guangxi Zhuang Autonomous Region from March 2022 to March 2025. Randomization was performed using a computer-generated block randomization sequence (block size = 4) stratified by baseline Sequential Organ Failure Assessment (SOFA) score (≤8 vs. >8). Participants, clinicians, and data collectors were aware of treatment allocation, but endpoint adjudicators (e.g., radiological and microbiological assessors) were blinded to group assignments to minimize outcome assessment bias. Radiological improvement was assessed independently by two blinded radiologists using the 2016 IDSA/ATS criteria, with discrepancies resolved by a third expert. Improvement was defined as ≥50% reduction in infiltrate size on chest X-ray/CT, measured semi-quantitatively with standardized lung zone grids.

Age ≥ 18 years.

Severe pneumonia meeting IDSA/ATS 2016 criteria (e.g., requiring invasive mechanical ventilation, septic shock, or multilobar infiltrates).

Confirmed multidrug-resistant Gram-negative bacilli (MDR-GNB) infection (per EUCAST 2023 breakpoints) isolated from lower respiratory tract cultures (bronchoalveolar lavage or sputum).

Polymicrobial infection involving fungi or Gram-positive pathogens.

Hypersensitivity to β-lactam antibiotics.

Concomitant aminoglycosides were prohibited from isolating the intervention effect, as synergistic combinations could confound survival outcomes.

Pregnancy or lactation.

Participants received imipenem/cilastatin 1 g every 8 hours via 30-minute intravenous infusion, prepared in 100 mL 0.9% saline.

Participants received the same dose (1 g q8 h) but administered as a 3-hour prolonged infusion, diluted in 100 mL 0.9% saline.

Infusion protocols followed consensus recommendations for prolonged β-lactam administration, including pump calibration and stability validation [12] [13] . Concomitant antibiotics (e.g., aminoglycosides) were prohibited to isolate the intervention effect [14] . Concomitant antibiotics (e.g., aminoglycosides) were prohibited.

Clinical response rate at day 8: Defined as resolution of systemic inflammatory response syndrome (SIRS) criteria (temperature ≤ 37.6˚C, heart rate ≤ 90 bpm, respiratory rate ≤ 20 breaths/min, WBC ≤ 12 × 10⁹/L) and radiological improvement (≥50% reduction in infiltrate size on chest X-ray/CT).

Time to microbial eradication: Daily quantitative cultures from bronchoalveolar lavage fluid (BALF) or sputum until negative conversion.

90-day all-cause mortality.

Nephrotoxicity: Defined by KDIGO criteria (serum creatinine increase ≥ 26.5 μmol/L within 48 hours, ≥1.5 × baseline within 7 days, or urine output < 0.5 mL/kg/h for ≥ 6 hours).

Hepatic injury: Alanine aminotransferase (ALT) or aspartate aminotransferase (AST) > 3 × upper limit of normal (ULN).

Neurological adverse events: Seizures, encephalopathy, or delirium.

Based on prior studies, we hypothesized a 60% clinical response rate in the short-term infusion group versus 80% in the prolonged infusion group. With α = 0.05 (two-tailed), β = 0.2 (80% power), and an anticipated 10% dropout rate, a minimum of 44 participants per group (total N = 88) was initially estimated. However, due to resource constraints and preliminary data suggesting a larger effect size, the target enrollment was pragmatically set to 41 participants per group (total N = 82). An interim analysis conducted at 82 participants (41 per group) demonstrated statistically significant superiority of prolonged infusion in the primary outcome (P = 0.002), leading to early trial termination as per pre-specified stopping rules for efficacy.

Between-group differences in clinical response rates on Day 8 were assessed using Pearson’s chi-square test (or Fisher’s exact test for expected cell counts < 5). Univariable analyses of baseline characteristics and day 8 biomarkers (e.g., CRP, AST) were performed using Independent t-tests or Mann-Whitney U tests for continuous variables (parametric/nonparametric based on distribution). Chi-square/Fisher’s exact tests for categorical variables. Kaplan-Meier curves with log-rank tests compared 90-day survival between groups. Univariable Cox proportional hazards models identified candidate predictors of survival (variables with P < 0.20 retained for multivariable analysis). A stepwise backward Cox regression (likelihood ratio test, removal threshold P ≥ 0.10) was applied to adjust for confounders. The proportional hazards assumption was verified using Schoenfeld residuals and log-log plots. Multiple imputation with chained equations (MICE) under the missing at random (MAR) assumption addressed missing primary outcome data. An interim analysis was pre-planned at 50% enrollment (n = 41/group). Using the O’Brien-Fleming stopping boundary (α = 0.005), the observed effect size (P = 0.002) crossed the efficacy threshold, prompting early termination.

Software: All analyses were performed using SPSS version 27.0 (IBM Corp, Armonk, NY, USA) and R 4.43 version.

A total of 82 participants were randomized to either prolonged infusion (n = 41) or short-term infusion (n = 41). Baseline demographic, clinical, and laboratory characteristics were balanced between groups ( Table 1 ). No significant differences were observed in age, sex, BMI, vital signs, inflammatory biomarkers (CRP, procalcitonin), or organ function parameters (all P > 0.05), confirming successful randomization.

By day 8, the Prolonged infusion group exhibited significantly fewer secondary fungal infections compared to the short-term group (7.3% vs. 29.3%, P = 0.010). Other adverse events, including hepatic dysfunction, renal impairment, and seizures, showed no statistically significant differences (all P > 0.05) ( Table 2 ). Notably, neurological adverse events (e.g., seizures) occurred in 2.4% of the Prolonged group versus 9.8% of the short-term group, though this difference did not reach statistical significance (P = 0.17). The reduced incidence of secondary fungal infections (7.3% vs. 29.3%) aligns with evidence linking optimized β-lactam exposure to decreased ecological disruption [15] .

Notes: Data presented as mean ± SD for normally distributed variables (Age, BMI), median (IQR) for non-normal variables, or n (%) for categorical variables. P values from independent t-test (Age, BMI), Mann-Whitney U test (non-normal variables), or χ²/Fisher exact test (categorical variables). Abbreviations: BMI, body mass index; WBC, white blood cell; CRP, C-reactive protein; AST, aspartate aminotransferase; ALT, alanine aminotransferase.

Prolonged infusion was associated with an increase in clinical response (82.9% vs. 58.5%; P = 0.015), consistent with prior trials targeting ƒT > MIC > 90% [16] [17] . At day 8, the clinical response rate was 82.9% in the prolonged group versus 58.5% in the short-term group (P = 0.015). Survival analysis revealed a significant advantage for prolonged infusion, with a 90-day survival probability of 95.1% compared to 68.3% (P = 0.02, log-rank test) ( Figure 1 ). Total hospital stay and 30-day recurrence rates did not differ significantly between groups. Prolonged infusion significantly accelerated microbial eradication (median 6.0 [IQR 4.5 - 7.5] vs. 7.0 [6.0 - 9.0] days ( Table 3 ).

Notes: ULN (upper limit of normal): AST = 40 U/L, ALT = 40 U/L. Significant P values in bold (P < 0.05).

Notes: Microbiological eradication time is defined as the first of two consecutive negative cultures. P value from Mann-Whitney U test.

Both groups received identical total daily doses (3.00 ± 0.00 g, P = 1.00). No differences were observed in total drug consumption or defined daily doses (DDDs), indicating comparable direct medication costs ( Table 4 ).

Univariate Cox regression and K-M survival analysis identified prolonged infusion as a protective factor for survival (HR = 0.134, 95% CI: 0.030 - 0.595; P = 0.008), while older age, longer treatment duration, higher CRP levels, and delayed microbial eradication were associated with increased mortality (all P < 0.05) ( Table 5 ; Figure 1 ). In the final multivariable model, prolonged infusion remained independently protective (adjusted HR = 0.08, 95% CI: 0.01 - 0.51; P = 0.008), alongside peak AST levels and treatment duration as risk factors ( Table 6 ).

Clinical Impact: Prolonged infusion of imipenem/cilastatin significantly improved clinical response rates and survival while reducing secondary fungal infections.

Safety: No increased risk of hepatic/renal toxicity or neurological events.

Cost-Effectiveness: Equivalent drug utilization supports the economic feasibility of prolonged infusion.

Our results corroborate prior observational studies suggesting that prolonged infusion of carbapenems improves clinical efficacy in critically ill patients. For pathogens with elevated MICs (e.g., Pseudomonas aeruginosa), short-term infusions may fail to achieve adequate ƒT > MIC thresholds, leading to therapeutic failure. By contrast, the 3-hour infusion strategy likely optimized drug exposure, as evidenced by faster microbial eradication ( Table 3 ) and reduced recurrence rates. The lower incidence of secondary fungal infections in the prolonged group further supports its role in minimizing collateral damage from broad-spectrum therapy, potentially by shortening the duration of systemic inflammation and mucosal disruption. These findings support incorporating prolonged infusion into antimicrobial stewardship programs for MDR-GNB pneumonia. Clinical Practice should prioritize this strategy in critical care formularies, particularly for regions with high carbapenem resistance rates. Logistical barriers, such as infusion pump availability and nursing workflow adjustments, may hinder adoption in resource-limited settings. Simulation-based training and phased implementation could mitigate these challenges.

While our findings are consistent with meta-analyses favoring prolonged infusion, they contrast with smaller RCTs reporting neutral effects. This discrepancy may stem from differences in patient selection—our trial exclusively enrolled microbiologically confirmed MDR-GNB pneumonia, whereas prior studies included mixed infections or retrospective cohorts [4] . Additionally, our use of blinded endpoint adjudication reduced outcome assessment bias, a limitation in earlier open-label trials.

Despite superior efficacy, prolonged infusion did not increase direct medication costs, as total drug consumption and treatment duration were equivalent between groups ( Table 4 ). This cost neutrality, combined with reduced hospital stays (25 vs. 33 days, P = 0.35), suggests prolonged infusion is economically viable. However, logistical challenges (e.g., infusion pump availability) may hinder implementation in resource-limited settings. Future guidelines should weigh these practical barriers against the clear survival benefits.