Adequate Antimicrobial Dosing in Critically Ill Patients Receiving Extracorporeal Membrane Oxygenation: Where to Go from Here?

Adequate antimicrobial use is key in times of increasing antimicrobial resistance, particularly when antibiotic drug development has virtually been abandoned (1). Antimicrobial dose optimization is trending in intensive care medicine; large variability in antimicrobial exposure has repeatedly been demonstrated, potentially leading to underdosing and therapeutic failure or to overdosing and toxicity (2). Patients treated with extracorporeal membrane oxygenation (ECMO) are at increased risk of acquiring nosocomial infections requiring antimicrobial treatment (3). ECMO is increasingly used in the management of cardiac or respiratory failure and also in patients admitted with influenza or coronavirus disease (COVID-19) (4).

Antimicrobial dosing is challenging in patients receiving ECMO, as pharmacokinetics (PK) might be impacted. The volume needed to prime the circuit might increase the volume of distribution of (mainly hydrophilic) antimicrobials. In addition, drug loss due to adsorption to the circuit has been described for lipophilic and highly protein-bound antimicrobials (5). Both may lead to underdosing, potentially impacting clinical outcome, especially because in patients receiving ECMO, often less-susceptible pathogens lead to infection for which antibiotics are often a last-resort therapeutic option (3).

To date, guidelines on antimicrobial dosing in patients receiving ECMO are still lacking, because of the heterogeneous and often conflicting body of evidence. For many antimicrobials, substantial PK alterations due to drug loss have been shown in ex vivo models; however, these results are not always reproducible in clinical studies (6, 7). Clinical studies often have methodological issues. Some are retrospective evaluations of, for example, results of therapeutic drug monitoring; others are rigorous prospective PK studies based on rich sampling, but sample sizes are often too limited to draw definitive conclusions (5). Therefore, the key question of whether antimicrobial PK changes are due to ECMO or rather critical illness itself remains unanswered.

In this issue of the Journal, Shekar and colleagues (pp. 704–720) report the impact of ECMO on antimicrobial exposure (8). The Antibiotic, Sedative and Analgesic Pharmacokinetics during Extracorporeal Membrane Oxygenation (ASAP-ECMO) study is the largest on this topic; they enrolled 85 adult patients receiving ECMO in six ICUs in whom total and free antimicrobial concentrations were measured during a single dosing interval. Eleven relevant antibiotics were studied, all after intravenous intermittent administration except for oseltamivir, which was administered enterally. It is important to note that the study was set up as a pragmatic real-world study; antimicrobial dosing and ECMO modalities were left to the clinician’s discretion. Almost half of patients (44.7%) required both renal replacement therapy (RRT) and ECMO when sampled, with the majority of patients (68.4%) receiving continuous veno-venous hemodiafiltration. To determine the adequacy of antimicrobial dosing, measured exposures were compared with predefined PK/pharmacodynamic targets, retrieved from the literature.

As expected, a large variability in all PK parameters, with coefficients of variation > 30% for all antimicrobials, was found. In patients sampled while receiving both ECMO and RRT, variability in exposure and PK was even more pronounced, with coefficients of variation of 63% up to 99% reported for meropenem, piperacillin, and vancomycin, respectively, with a 200-fold variation in trough concentrations. PK/pharmacodynamic targets were not attained in about half of the patients (44%), with subtherapeutic exposure as the main problem.

The results of the ASAP-ECMO study are in line with previously published clinical studies and are both worrying and reassuring (5, 6). The study once again confirmed that currently applied dosing regimens lead to poor target attainment in many patients. This has been associated with therapeutic failure, antimicrobial resistance development, and toxicity. Poor target attainment was most frequently found for oseltamivir, piperacillin, and vancomycin, which is striking, as optimized dosing for piperacillin-tazobactam (4,000/500 mg every 6 h) and therapeutic drug monitoring for vancomycin are often applied. However, PK estimates of most antimicrobials were comparable to previously published data in ICU patients without ECMO support. Based on previous ex vivo results, the authors expected to find considerable differences in the impact of ECMO on the exposure of studied antimicrobials, especially as physicochemical characteristics of the drugs (e.g., caspofungin vs. meropenem) were highly different (9). These results, conflicting with ex vivo data, are in line with what has been discussed in earlier clinical studies, confirming that ex vivo experiments are not necessarily reflective of drug behavior in ICU patients on ECMO (7). It is not clear whether this is related to differences in ex vivo blood composition or whether an increase in PK variability related to ECMO cannot be differentiated from the large PK variability associated with critical illness itself.

Some methodological considerations are worth mentioning. First, although the high variability in exposure is evident, and not explained by variability in dosing, other factors, not captured in the current study, could have contributed. Starting antibiotics before or during ECMO and differences in ECMO and/or RRT modalities among centers may result in different exposures (10). Second, for some of the antimicrobials studied, the number of patients studied was lower than anticipated, but the overall message remains that variability is high. Including more patients in the analysis would probably not have changed that conclusion. The same holds for the low target attainment, although other strategies to optimize exposures, such as prolonged infusion, were not considered in this study. Finally, outcomes such as clinical cure and mortality are often considered the only relevant outcome in clinical studies, but admittedly this would be very difficult to assess in these patients; clinical cure is difficult to assess in patients with these levels of organ support, in whom outcomes are largely determined by the underlying disease.

It is clear that antimicrobial dosing in these challenging patients deserves our continued attention. Ideally, one would like to identify all contributors to the variability described, but the number of patients and observations required to develop robust models would be very high, compared with the frequency with which ECMO is used. While awaiting evidence-based guidance on antimicrobial dosing, we therefore propose a pragmatic approach to antimicrobial dosing in these patients (Figure 1). This includes high initial dosing, particularly when antimicrobials are started while patients are on ECMO (with or without RRT), optimized maintenance dosing (taking into account residual kidney function), and routine use of therapeutic drug monitoring. In this way, variability in exposure will be monitored, and both suboptimal and excessive dosing can be avoided. Reliable and accessible therapeutic drug monitoring is therefore urgently needed for antimicrobials commonly used in these patients.

Figure 1.

Antimicrobial dosing strategy in patients treated with ECMO with or without RRT. AKI = acute kidney injury; CKD = chronic kidney disease; ECMO = extracorporeal membrane oxygenation; mCrCI = measured creatinine clearance; RRT = renal replacement therapy.