This study provides the first real-life pharmacokinetic analysis of meropenem in critically ill patients undergoing ADVOS therapy and compares it with CVVHDF and non-extracorporeal treatments. It is also the first study to measure pre- and post-filter levels of meropenem to calculate machine-specific clearance in these patient groups.

The key findings indicate that ADVOS therapy is safe for critically ill patients treated with meropenem, as meropenem clearance by ADVOS (CLADVOS) was lower than anticipated, and no trough levels fell below the therapeutic target of 2–8 mg/L. The ADVOS machine clearance (CLADVOS) of 3.5 L/h, measured in this study, is comparable to the documented in vitro values (3.4 L/h) [12]. This is only approximately 30% of the normal intrinsic meropenem clearance in healthy adults (12.5 L/h [14]) or 40% of the clearance observed by Jaruratanasirikul et al. in non-renal insufficient intensive care patients (7.8 L/h) [15]. This clearance in critically ill patients is consistent to the clearance of 8 L/h measured in our control group (CLpatient). Importantly, only 3 trough levels below 2 mg/L were observed, all in the control group without extracorporeal therapy. However, a significant number of trough levels exceeded the optimal range of 2–8 mg/L, particularly in the CVVHDF group (77.8%), compared to the ADVOS (57.2%) and control groups (52.5%).

These results align with previous in vitro studies and case reports on ADVOS König et al. reported meropenem ADVOS clearances between 3.4 and 6.08 L/h in an in vitro model [12], depending on flow rates. Similarly, she described a dialyzer clearance of 6.3 L/h in a patient treated with ADVOS at a higher blood flow rate (250 mL/min). Both findings support our observed ADVOS clearance of 3.5 L/h at lower flow rates (100 mL/min)​​. All our patients were treated with 100 mL/min blood flow to ensure hemodynamic stability and to avoid disturbances of the blood flow. Higher blood flow rates to treat severe acidosis more effectively were not necessary in our patient cohort.

These findings demonstrate that the limited ADVOS machine clearance of 3.5 L/h at a blood flow of 100 mL/min cannot compensate for the complete loss of renal function, with an assumed renal meropenem clearance in ICU patients of approximately 8 L/h. Hence, meropenem underdosing in patients with renal failure on ADVOS therapy receiving a high dose meropenem regimen (2 g q8h) is unlikely with these flows. Nonetheless, the ADVOS clearance is much more effective than the CVVHDF clearance (3.5 vs 1.8 L/h) and this must be considered in dosing. In CVVHDF, higher trough levels are consistent with existing studies that show impaired intrinsic clearance in patients with renal failure, even with extracorporeal therapy [16]. Prior studies on CVVHDF also highlight the unpredictability of antibiotic clearance due to variability in flow rates, membrane permeability, and patient-specific factors [16, 17]. The findings from this study further underscore the necessity of therapeutic drug monitoring (TDM) in CVVHDF to avoid overdosing, as the use of tools like MeroEasy, which relies on population estimates, may overestimate clearance without accounting for real-world machine maintenance pauses or individual variations. We recommend using the MeroEasy tool, when using continuous renal replacement therapies (CRRT, ADVOS), to adjust the dosage based on measured trough levels. Similarly, a recent study investigating hemoadsorption with CytoSorb® coupled post-filter to continuous renal replacement therapy reported no significant impact of this device in meropenem or piperacillin levels [18], emphasizing that the impact of extracorporeal therapies on antibiotic clearance is highly modality-specific and must be evaluated on a case-by-case basis.

In our study, the moderate correlation between estimated and measured meropenem clearance (r = 0.46, Graph 3C) highlights the limitations of creatinine-based estimation tools like MeroEasy in critically ill patients, particularly those undergoing extracorporeal therapies. These tools rely on stable renal function and demographic parameters, but do not account for the clearance contributed by ADVOS or CVVHDF. Moreover, serum creatinine often poorly reflects acute changes in kidney function and is affected by fluid shifts, inflammation, and muscle catabolism common in ICU patients [19]. Additionally, therapy interruptions, individual pharmacokinetics, and machine-specific settings (e.g., dialysate flow, membrane properties) introduce variability not captured in the model. This explains the observed overestimation of clearance and further supports the necessity of therapeutic drug monitoring to guide individualized dosing in this complex population.

The safety of the dosing regimen employed (2 g q8h) in ADVOS and CVVHDF was supported by the absence of levels below therapeutic thresholds. However, a substantial proportion of trough levels exceeded 8 mg/L, raising concerns about potential toxicity. Neurotoxic and nephrotoxic effects have been reported at meropenem trough levels exceeding 44.45 mg/L and 64.2 mg/L, respectively (6), which was exceeded by one of our measurements in the ADVOS group. This high proportion of supra-therapeutic levels warrants further evaluation, especially in patients with limited residual renal function. As potential signs of neurotoxicity, delirium was documented in 10 out of 16 patients, and seizures occurred in two patients, both with underlying neurological disorders. It remains challenging to differentiate potential neurotoxicity from the consequences of critical illness. The patient with 62.8 mg/L meropenem trough level had severe pneumococcal pneumonia and acute renal failure and developed delirium but no convulsions.

The high variation of trough levels in the ADVOS group from 2.1 to 62.8 mg/L may be explained by several factors: the highly variable residual kidney function, the variable duration of treatment pauses of the ADVOS machine which might have caused accumulation of meropenem, different distribution volumes because of anasarca or capillary leak and dilution by a high fluid load are all possible explanations.

Generalizability is supported by the inclusion of a significant number of measurements (159), covering a range of clinical conditions, but caution is warranted in extrapolating the findings to antibiotics with different pharmacokinetic profiles, such as high protein binding or low intrinsic clearance agents (e.g. ceftriaxone, daptomycin).

This study has several limitations. First, the retrospective, single-center design limits its external validity. Second, the small sample size within subgroups, particularly for ADVOS therapy, reduces the power to detect statistically significant differences in meropenem clearance, especially in comparative analyses (e.g., t-tests). Third, the study did not account for interindividual variability in patient factors such as fluid status, inflammation, and other pharmacokinetic modifiers, which may influence meropenem levels. Together with the low number of patients, this may limit its generalizability and statistical interpretation. Two patients contributed data to more than one treatment group, this might introduce bias in between-group comparisons.

Moreover, the reliance on TDM performed under routine clinical conditions introduces potential inconsistencies in sample timing and handling. Since only values after dosing intervals without ADVOS interruptions were included, the total daily clearance may be overestimated since the daily pause to set up and exchange the ADVOS machine is not accounted for.

Future studies should investigate the pharmacokinetics of antibiotics with high protein binding and low intrinsic clearance during ADVOS therapy. Prospective, multicenter studies could provide more robust data and validate these findings across different ICUs. Moreover, detailed analyses of the impact of therapy interruptions and machine configurations on drug clearance are necessary to refine dosing strategies.

Pharmacokinetic modeling incorporating real-world TDM data could improve tools like MeroEasy to account for individual variations and therapy-specific factors. Additionally, studies comparing ADVOS with other extracorporeal modalities in critically ill patients with different organ dysfunction profiles (e.g., liver failure, intoxication) could help optimize dosing protocols for a broader range of clinical scenarios.

In conclusion, this study shows that patients with renal failure on ADVOS therapy at low blood flows of 100 mL/min receiving a high dose meropenem regimen (2 g q8h) have no risk of antibiotic underdosing and ADVOS therapy is safe for patients with severe infections. The study highlights the importance of TDM in extracorporeal therapies and critically ill patients, as the ideal dosing is unpredictable due to variable factors influencing drug levels. Our findings emphasize the need for personalized dosing regimens to balance therapeutic efficacy and safety and to impede overdosing.