ECMO drug dosing — circuit sequestration, increased Vd, dosing strategies for fentanyl, propofol and antimicrobials

Overview and Clinical Relevance

Extracorporeal membrane oxygenation (ECMO) profoundly alters the pharmacokinetic (PK) behaviour of virtually every drug administered in the ICU. The consequence is predictable: standard dosing regimens derived from non-ECMO populations are frequently inadequate, producing drug concentrations that are sub-therapeutic, unexpectedly toxic, or wildly variable. For the CICM Fellowship candidate, the intellectual task is not to memorise a revised drug schedule but to understand the mechanistic reasons why ECMO disrupts PK, and from that framework derive rational, adaptable dosing strategies.

The four dominant mechanisms, circuit sequestration, expanded volume of distribution ($V_d$), altered protein binding, and modified clearance, interact with each other and with the underlying critical illness that prompted ECMO in the first place. Critical illness alone causes PK chaos; ECMO superimposes an additional and partially independent layer of disturbance.

Physiological and Pharmacological Framework

1. Circuit Sequestration

ECMO circuits introduce a large foreign surface area into the patient's circulation. Tubing, oxygenator membrane, connectors, and heat exchanger present several square metres of polyvinylchloride (PVC), silicone, polymethylpentene (PMP), or polyurethane to flowing blood. Drugs adsorb to these surfaces through non-covalent hydrophobic and electrostatic interactions.

The physicochemical predictors of sequestration are:

Lipophilic drugs partition into the PVC tubing itself, this is true adsorption in the chemical sense, where the hydrophobic polymer interior acts as a lipid sink. Highly protein-bound drugs are sequestered partly via protein-surface interaction, and partly because changes in free-fraction alter apparent distribution. In both cases, the circuit acts as a "drug sponge," reducing the fraction reaching the patient.

Sequestration is most pronounced:

• In the first few hours after drug administration (especially loading doses)
• With new circuits (older circuits have partially saturated binding sites)
• With lower-volume oxygenators (older hollow-fibre vs. newer PMP oxygenators differ substantially, PMP oxygenators sequester significantly less lipophilic drug than older silicone membrane types)

Oxygenator type matters: PMP (e.g. Quadrox-iD) is the current ANZ standard and sequesters considerably less propofol and fentanyl than older silicone oxygenators. However, clinically meaningful sequestration still occurs.

2. Increased Volume of Distribution

$V_d$ describes how widely a drug distributes throughout body compartments relative to plasma. ECMO patients typically have a markedly expanded $V_d$ for several overlapping reasons:

$$V_d = \frac{\text{Dose}}{\text{Plasma concentration at time zero}}$$

• Priming volume: The circuit prime (typically 300-500 mL in adults, 100-300 mL in neonates/children) dilutes the first dose, a direct, immediate reduction in plasma concentration.
• Systemic inflammation: Critical illness and ECMO-related inflammation increase capillary permeability, shifting drug into interstitial spaces.
• Fluid overload: Common in ECMO patients; third-spacing expands interstitial compartments.
• Hypoalbuminaemia: Reduces protein binding, paradoxically increasing free (unbound) drug and redistributing it to peripheral compartments.

The clinical consequence: for drugs with a narrow therapeutic window, a loading dose calculated on published $V_d$ data from healthy volunteers will produce sub-therapeutic concentrations after circuit sequestration and dilutional effects combine.

$$\text{Loading dose} = V_d \times C_{target}$$

If $V_d$ is expanded by 40-80% (as observed for midazolam and fentanyl in ECMO) and a standard loading dose is given, the resulting peak concentration will be proportionally lower.

3. Altered Protein Binding

Drugs distribute according to their free (unbound) fraction, not total plasma concentration. ECMO alters protein binding through:

• Hypoalbuminaemia: Reduces albumin-bound drug fraction (acidic drugs: phenytoin, warfarin, most beta-lactams).
• Haemodilution: Reduces both albumin and alpha-1-acid glycoprotein (AAG) concentrations.
• Competitive displacement: Heparin infusions (universal in ECMO) release free fatty acids from albumin, which compete with drug-binding sites.
• Hypothermia: Alters protein conformation and binding affinity.
• pH changes: Alters ionisation state and therefore protein binding.

For drugs with high protein binding (e.g. propofol 97-99%, fentanyl ~84%, midazolam 94-97%, voriconazole ~58%), even a small increase in free fraction amplifies both pharmacodynamic effect and redistribution into peripheral compartments.

4. Altered Clearance

ECMO modifies both hepatic and renal clearance:

Hepatic clearance: ECMO can cause hepatic congestion (especially VA-ECMO with elevated right atrial pressures) or hypoperfusion during cannulation/initiation. Cytochrome P450 (CYP) activity, which governs the metabolism of drugs including midazolam, fentanyl, propofol, and voriconazole, is altered by:

• Reduced hepatic blood flow (VA-ECMO)
• Systemic inflammatory mediators downregulating CYP expression
• Concurrent medications (azole antifungals as potent CYP3A4 inhibitors themselves)

Renal clearance: Acute kidney injury is near-universal in ECMO; continuous renal replacement therapy (CRRT) adds further complexity. For renally cleared drugs (many antibiotics), AKI reduces elimination; CRRT partially restores it but adds filter adsorption as an additional loss mechanism.

Non-biological clearance: Technically, circuit sequestration is a form of drug loss that mimics, but is not, clearance. It reduces circulating drug concentrations without producing metabolites. It is important not to conflate this with true clearance when constructing pharmacokinetic models.

Drug-Specific Details

Sedation and Analgesia

Fentanyl

Fentanyl is the archetype ECMO-problematic opioid:

• $\log P = 4.05$, extremely lipophilic
• Protein binding ~84% (predominantly AAG)
• $V_d$ 4-6 L/kg in healthy adults; expanded further in ECMO

Sequestration evidence: Ex vivo ECMO circuit studies (Shekar et al. published in Critical Care) demonstrated that up to 50-70% of a fentanyl dose can be sequestered by a PVC/silicone circuit within 60 minutes of administration. PMP circuits sequester less, approximately 25-40%, but clinically meaningful sequestration persists.

Wildschut et al. studying neonatal ECMO, demonstrated profound fentanyl sequestration with recovery of less than 30% of administered dose in plasma during the first hour.

Clinical implication: The loading dose required to achieve target plasma concentration is substantially higher on ECMO. Titration to effect is mandatory. Remifentanil is an alternative (ester hydrolysis in plasma, not circuit-sequestered) but its ultra-short action makes it logistically difficult on prolonged ECMO runs.

Morphine

• $\log P = 0.9$, relatively hydrophilic
• Protein binding ~35%
• Much lower circuit sequestration than fentanyl
• Active metabolite morphine-6-glucuronide (M6G) accumulates in renal failure, relevant given near-universal AKI in ECMO

Morphine is often a more predictable choice for baseline analgesia on ECMO, though accumulation of M6G limits its use in severe AKI.

Propofol

Propofol is the extreme case:

• $\log P = 3.8$, highly lipophilic
• Protein binding 97-99%
• Formulated in a lipid emulsion that interacts directly with PVC tubing

Circuit data: Multiple in vitro studies demonstrate near-complete sequestration of propofol by PVC tubing. Ex vivo data show that circuits can sequester 40-80% of propofol within 30 minutes. This creates a situation where dose titration is unreliable at low infusion rates, the circuit acts as a buffer, and clinical effect lags dramatically behind dose adjustments.

Propofol infusion syndrome (PRIS) risk is heightened in ECMO because high infusion rates are required to compensate for sequestration. ANZ practice increasingly favours midazolam or ketamine as first-line sedative agents on prolonged ECMO runs, reserving propofol for short-term procedural sedation.

Midazolam

• $\log P = 3.4$, lipophilic
• Protein binding 94-97%
• CYP3A4-dependent metabolism

Midazolam is highly sequestered. Shekar et al. in vitro data show approximately 55-70% sequestration over 24 hours in PVC circuits. Despite this, midazolam remains commonly used in ANZ practice because:

1. It is familiar and titratable
2. Flumazenil reversal is available
3. Ketamine is often co-administered to reduce midazolam requirements

Practical reality: Midazolam doses of 10-20 mg/hour (or higher) are not unusual in adult ECMO patients, compared to typical ICU doses of 2-10 mg/hour. Accumulation becomes a concern once the circuit binding sites saturate, typically after several days.

Ketamine

• $\log P = 3.1$, moderately lipophilic
• Protein binding ~45-50%
• Metabolism via CYP3A4 and CYP2B6

Ketamine is increasingly favoured as an ECMO sedation adjunct. It has moderate lipophilicity (less sequestration than fentanyl or propofol), analgesic and sedative properties, and bronchodilatory effects. It does not cause vasodilation, which is advantageous in haemodynamically unstable patients. Emergence reactions are minimised by concurrent benzodiazepine use (which is already the norm).

Dexmedetomidine

• $\log P = 2.89$
• Protein binding ~94%
• Moderate sequestration predicted

Limited ECMO-specific data exist. Dexmedetomidine is hepatically cleared and may accumulate with hepatic congestion. It can exacerbate the bradycardia associated with ECMO, particularly relevant in VA-ECMO where heart rate is a key determinant of native cardiac output. Use with caution.

Antiinfectives

Voriconazole

Voriconazole is the most pharmacokinetically problematic antifungal on ECMO:

• $\log P = 1.8$ (moderate lipophilicity)
• Protein binding ~58%
• Non-linear pharmacokinetics, saturable CYP2C19 metabolism
• Oral bioavailability ~96% in healthy patients (but reduced by gut oedema in ECMO)
• Potent inhibitor of CYP3A4, CYP2C19, CYP2C9, major drug interactions

ECMO-specific data: Ex vivo circuit studies (Shekar, UQ/CONECT-4U group) show approximately 60-71% of voriconazole can be sequestered by PVC circuits. This is clinically devastating given the narrow therapeutic index (target trough 1-6 mg/L; toxicity including hepatotoxicity and visual disturbance above 5-6 mg/L).

Therapeutic drug monitoring (TDM) is mandatory for voriconazole on ECMO. The unpredictable combination of circuit sequestration, non-linear kinetics, and CYP variability (CYP2C19 polymorphism affects 15-20% of Asian populations, relevant in the ANZ context) makes empirical dosing unreliable.

Alternative strategy: Isavuconazole (isavuconazonium sulfate) has more predictable linear pharmacokinetics, less CYP3A4 inhibition, and emerging ECMO data suggesting less circuit sequestration (though evidence is limited). Anidulafungin or caspofungin may be preferred for Candida while awaiting TDM results.

Fluconazole

• $\log P = 0.5$, hydrophilic
• Protein binding ~12%
• Renal clearance (unchanged drug)
• Minimal circuit sequestration

Fluconazole is an example of a drug that behaves relatively normally on ECMO from a sequestration perspective. However, AKI with concurrent CRRT requires dose adjustment for renal clearance. Fluconazole remains a reasonable choice for susceptible Candida species.

Anidulafungin and Caspofungin (Echinocandins)

• Large molecular weight proteins
• High protein binding (anidulafungin ~99%, caspofungin ~97%)
• Not renally cleared, less affected by AKI/CRRT

Paradoxically, despite high protein binding, echinocandins show minimal clinically significant sequestration in ex vivo models (Shekar group). This is likely because:

1. Their large molecular size limits diffusion into tubing walls
2. They are present in blood at low concentrations relative to their high $V_d$
3. Their dose-response relationship is linked to AUC/MIC, not peak or trough alone

Echinocandins are generally preferred as first-line antifungal therapy in moderate-to-severe systemic candidiasis on ECMO.

Beta-Lactam Antibiotics

• Low lipophilicity ($\log P < 1$ for most agents)
• Variable protein binding (piperacillin ~30%, meropenem ~2%, ceftriaxone ~90%)
• Hydrophilic, minimal circuit sequestration

Beta-lactam pharmacokinetics on ECMO are primarily affected by:

• Expanded $V_d$ (first dose dilution)
• Altered renal clearance (AKI, CRRT)
• Haemofilter adsorption (for CRRT)
• Hypoalbuminaemia altering free fraction of highly protein-bound agents (ceftriaxone, oxacillin)

Key principle for beta-lactams: Efficacy is time-dependent (% time free drug > MIC). Strategies include:

• Extended/continuous infusions, strongly supported for piperacillin-tazobactam, meropenem, cefepime
• Larger loading doses to account for expanded $V_d$
• TDM where available (meropenem, piperacillin TDM assays are available in major ANZ centres)

Vancomycin

• Low lipophilicity
• Protein binding ~50%
• Renal clearance
• Minimal direct circuit sequestration

Vancomycin dosing on ECMO is complicated by:

• AKI (reduced clearance, toxicity risk)
• CRRT (increased clearance, under-dosing risk)
• Expanded $V_d$ (requires higher loading doses)
• Concurrent nephrotoxins (antifungals, contrast, haemolysis from circuit)

AUC/MIC-guided vancomycin dosing (target AUC 400-600 mg·h/L for MRSA bacteraemia) is the current ANZ/ASHP-endorsed approach. Bayesian dosing software (e.g. DoseMeRx, InsightRx) outperforms trough-only strategies in ECMO patients with dynamic renal function.

Anticoagulation

Unfractionated Heparin (UFH)

UFH is the standard anticoagulant for ECMO in ANZ centres:

• Highly protein-bound (including to antithrombin III, platelet factor 4, endothelial cell surface proteins)
• Binds to ECMO circuit surfaces, both adsorption and consumption
• Heparin-coated circuits (Carmeda, Bioline) reduce but do not eliminate this

Monitoring: Anti-Xa levels (target typically 0.3-0.7 IU/mL for ECMO, centre-specific), with APTT as a secondary monitor. Anti-Xa is preferred because heparin binding proteins (acutely elevated in critical illness) falsely prolong APTT without reflecting true heparin activity.

Bivalirudin

Used when heparin-induced thrombocytopaenia (HIT) is suspected or confirmed:

• Direct thrombin inhibitor, predominantly cleared by plasma proteolytic degradation (~80%) with minor renal contribution
• Less affected by ECMO circuit than UFH
• Short half-life (25 minutes), advantageous if rapid reversal needed
• Monitored by APTT (target 50-90 s) or ecarin clotting time (ECT)
• Note: bivalirudin degrades in blood within circuits during standstill, relevant for circuit storage samples

Antiepileptics and Neurological Drugs

Phenytoin / Fosphenytoin

• High protein binding (phenytoin ~90%, predominantly albumin)
• Hypoalbuminaemia falsely lowers total phenytoin concentrations without reducing free (active) fraction
• Free phenytoin monitoring is recommended in ECMO (target free phenytoin 1-2 mg/L)

Levetiracetam

• Low protein binding (~10%)
• Renal clearance
• Minimal circuit sequestration
• Dose reduce in AKI; supplement after CRRT sessions

Levetiracetam is increasingly preferred over phenytoin in ECMO-associated seizures due to its predictable PK.

Drug Summary Table

Evidence Base

Shekar et al. (ECMO PK Studies, UQ/CONECT-4U Group)

Kiran Shekar and colleagues at the Prince Charles Hospital/University of Queensland have produced the most comprehensive body of in vitro and ex vivo ECMO pharmacokinetic data available. Key publications include:

Shekar K et al. "Sequestration of drugs in the circuit may lead to therapeutic failure during extracorporeal membrane oxygenation" (Critical Care, 2012): This landmark paper systematically characterised the sequestration behaviour of commonly used ICU drugs in ex vivo ECMO circuits. Key findings:

• Fentanyl, midazolam, and voriconazole showed substantial sequestration (>50% within 24 hours in PVC circuits)
• Circuit type significantly affected sequestration, PMP oxygenators sequestered less than silicone membrane oxygenators
• Drug physicochemical properties (lipophilicity, protein binding) were the strongest predictors of sequestration
• This paper established the mechanistic framework that underpins current ECMO dosing practice

Shekar K et al. "Altered antibiotic pharmacokinetics during extracorporeal membrane oxygenation" (Journal of Antimicrobial Chemotherapy, 2012 and subsequent): Demonstrated that while most beta-lactams are not significantly sequestered, the combination of expanded $V_d$, haemofilter adsorption (during concurrent CRRT), and altered renal clearance produces clinically relevant PK changes. Extended/continuous infusion strategies are mechanistically justified.

Shekar K et al. (ASAP ECMO studies): A series of prospective pharmacokinetic studies in adult ECMO patients quantifying in vivo PK changes for specific drug classes.

Wildschut et al. (Neonatal/Paediatric ECMO)

Wildschut ED et al. "Determinants of drug absorption in different ECMO circuits" (Intensive Care Medicine, 2010): This paper examined drug sequestration in neonatal ECMO circuits, which use smaller but proportionally more surface-heavy tubing. Key contributions:

• Demonstrated that drug lipophilicity and protein binding predicted sequestration across all circuit types tested
• Fentanyl sequestration was profound in all circuit types
• Proposed that higher loading doses are required to achieve target concentrations in neonates on ECMO, a finding with direct paediatric ECMO dosing implications
• Highlighted that circuit age (older circuits with partially saturated binding sites) altered subsequent drug sequestration, a dynamic process, not static

Wildschut et al. (subsequent neonatal PK studies): Extended these findings to chloral hydrate, morphine, and lorazepam in neonatal ECMO, broadly confirming lipophilicity as the dominant predictor.

Additional Key Literature

• Raffaeli G et al. (systematic reviews of ECMO PK) have synthesised available evidence confirming the mechanistic framework, though large randomised trials assessing outcomes based on ECMO-adjusted dosing do not yet exist.
• Preston TJ et al. demonstrated ECMO circuit sequestration for propofol in paediatric circuits, supporting concerns about PRIS risk when high doses are needed to overcome sequestration.
• Hahn J et al. and multiple groups have published meropenem, piperacillin, and vancomycin PK studies in adult ECMO showing expanded $V_d$ and the value of extended infusions.
• ELSO (Extracorporeal Life Support Organisation) guidelines acknowledge altered PK in ECMO but stop short of specific dosing recommendations, deferring to institutional protocols and TDM.
• ANZICS ECMO Special Interest Group guidance echoes ELSO recommendations regarding TDM and dose titration, published position statements are available but lack drug-specific dosing tables.

Practical ANZ ICU Application

General Dosing Strategy Framework

The CICM trainee should approach ECMO drug dosing through a systematic checklist:

Step 1: Classify the drug by physicochemical properties

• $\log P > 2$ or protein binding $> 85%$ → anticipate significant sequestration → higher loading dose, titrate to effect
• $\log P < 1$ and protein binding $< 40%$ → less sequestration → standard loading dose, adjust for renal/hepatic function

Step 2: Consider circuit age and type

• New PMP circuit (first 24-48 hours): highest sequestration risk, give generous loading doses
• Old PVC circuit (> 5 days): binding sites partially saturated, risk of sudden increase in plasma concentration if infusion rate is maintained when circuit changed

Step 3: Assess circuit change timing

• Sedation/analgesic reassessment is mandatory after every circuit change, binding site re-establishment can cause abrupt under-dosing

Step 4: Concurrent renal replacement therapy

• Add haemofilter adsorption and altered clearance to the model
• Check whether drug is cleared by CRRT (hydrophilic, small molecular weight drugs are; lipophilic drugs are not, but are already sequestered)

Step 5: Therapeutic drug monitoring

• Mandatory for: voriconazole, vancomycin (AUC-guided), aminoglycosides, phenytoin (free levels), meropenem/pip-tazo where assays available
• Consider for: midazolam, fentanyl where plasma assays are accessible (limited in Australian routine clinical practice)

Sedation and Analgesia: ANZ Approach

Target: RASS −1 to −2 for most ECMO patients (deeper for VV-ECMO patients with severe ARDS dyssynchrony; lighter if attempting liberation or spontaneous breathing trial)

Analgesic-first approach: Consistent with PADIS guidelines, establish analgesia before sedation. However, fentanyl titration is profoundly challenging, consider:

• Hydromorphone as an alternative (moderately lipophilic, less sequestration data but less problematic than fentanyl in practice)
• Morphine for background analgesia with low-dose PRN opioid for procedural pain
• Paracetamol (IV or enteral) as opioid-sparing, minimal PK effect from ECMO

Sedation agents in ANZ practice:

• Midazolam is most widely used despite sequestration, doses of 5-30 mg/hour are documented in practice
• Ketamine is an excellent adjunct (50-200 mg/hour infusion; or 1-2 mg/kg/hour)
• Propofol use is reduced to short-term procedural sedation or ECMO weaning trials
• Clonidine (enteral or IV) as an adjunct, low lipophilicity, less sequestration, reduces opioid and benzodiazepine requirements

Daily sedation interruption (DSI) on ECMO: controversial. ACURASYS demonstrated benefit of DSI in general ICU but ECMO patients were excluded. ROSE (2019) found no benefit of neuromuscular blockade on mortality in ARDS; by extension, deep sedation strategies lack outcome evidence in ECMO-specific populations. SPICE-III demonstrated that lighter sedation with propofol (dexmedetomidine equivalent) did not improve outcomes in general ICU. In ECMO practice, light sedation is theoretically preferable (PADIS framework) but circuit flow requirements and dyssynchrony may mandate deeper sedation in early VV-ECMO.

Antifungal Management: Practical Implications

In an ANZ ICU managing a patient on VV-ECMO for severe ARDS with suspected invasive candidiasis:

1. Start an echinocandin (anidulafungin 200 mg loading, then 100 mg daily; or caspofungin 70 mg loading, then 50-70 mg daily), minimal circuit sequestration, good bioavailability, reliable kinetics
2. Avoid voriconazole as empiric therapy on ECMO unless there is a specific indication (e.g. Aspergillus) and TDM is immediately available
3. If voriconazole is required: load with IV formulation (6 mg/kg Q12h × 2 doses), switch to oral (which has near-complete bioavailability in non-ECMO patients, but note gut oedema reduces enteral absorption in ECMO), and check TDM trough at 48-72 hours (target 1-5 mg/L for invasive aspergillosis)
4. Isavuconazole (isavuconazonium sulfate, Cresemba, TGA-approved in Australia): emerging alternative with more linear PK; limited ECMO data but preferred over voriconazole by some ANZ centres when mould coverage is required

Antibiotic Stewardship on ECMO

• All empiric antibiotic decisions should still follow local antimicrobial stewardship protocols and be reviewed at 48-72 hours with culture data
• Extended/continuous beta-lactam infusions are standard of care in most ANZ ICUs, ECMO does not change this recommendation, it reinforces it
• Vancomycin AUC-guided dosing with Bayesian software is recommended, dynamic renal function in ECMO makes empiric fixed-interval dosing unreliable

Dealing with Circuit Changes

A circuit change represents a pharmacokinetic reset. Management:

• Anticipate decreased drug exposure in the first 4-6 hours after circuit change (new circuit has unsaturated binding sites)
• Administer loading doses of sedation/analgesia agents at the time of circuit change
• Increase infusion rates prophylactically and titrate to effect over the following hours
• Communicate to nursing staff that the patient may appear lighter post-circuit change and require urgent dose adjustment

Key Numbers

KEY NUMBERS, ECMO Drug Dosing

• 50-70% of a fentanyl dose sequestered by PVC circuits within 60 minutes (Shekar)
• 55-70% midazolam sequestration over 24 hours in PVC circuits
• 60-71% voriconazole sequestration in ex vivo PVC ECMO circuits
• <30% of administered fentanyl recoverable in plasma in first hour (Wildschut neonatal data)
• $\log P > 2$ threshold above which clinically significant circuit sequestration is likely
• $>85\%$ protein binding, additional risk factor for sequestration/altered free fraction
• 300-500 mL adult ECMO circuit prime volume, causes immediate dilutional drop in loading dose plasma concentration
• 2-5× typical midazolam dose increase required on ECMO vs. standard ICU dosing
• AUC 400-600 mg·h/L, vancomycin target for MRSA bacteraemia (ASHP/ANZ endorsed)
• Voriconazole trough 1-5 mg/L, therapeutic range (toxicity risk above 5-6 mg/L)
• RASS −1 to −2, recommended sedation depth for most VV-ECMO patients
• ~80% bivalirudin clearance by plasma proteolytic degradation (not renal), relatively preserved in AKI

Viva Points

VIVA POINTS, ECMO Drug Dosing

Q: Why do lipophilic drugs sequester in ECMO circuits? Lipophilic drugs partition into the hydrophobic polymer matrix of PVC tubing and oxygenator components through non-covalent hydrophobic interactions. The polymer interior acts as a lipid sink. This is true adsorption, drug is physically absorbed into the tubing wall, not just surface-adherent.

Q: Why does PMP oxygenator sequester less drug than silicone membrane oxygenators? PMP (polymethylpentene) has lower lipid sol

Sources

• PADIS guidelines (SCCM Pain/Agitation/Delirium/Immobility/Sleep)
• ANZICS guidelines & CORE registry