Therapeutic drug monitoring in ICU — vancomycin AUC24/MIC, aminoglycoside levels, phenytoin and immunosuppressants

Table of Contents

1. Physiological and Pharmacological Framework
2. Why ICU Patients Are Different
3. Vancomycin TDM, AUC/MIC Strategy
4. Aminoglycosides, Extended-Interval Gentamicin
5. Phenytoin, Protein Binding and Free-Fraction
6. Levetiracetam
7. Lithium
8. Tacrolimus and Cyclosporin in Solid Organ Transplant
9. TDM in Special ICU Contexts
10. Evidence Base
11. Practical ANZ ICU Application
12. Key Numbers
13. Viva Points

1. Physiological and Pharmacological Framework {#framework}

Therapeutic drug monitoring (TDM) exists because the relationship between administered dose and achieved plasma concentration is unpredictable, particularly in the critically ill. TDM closes the loop by using measured concentrations to infer pharmacokinetic parameters and adjust dosing so that the target exposure is achieved.

The PK/PD Conceptual Scaffold

Every drug-monitoring decision rests on the pharmacokinetic-pharmacodynamic (PK/PD) axis:

The three PD exposure indices that determine efficacy for anti-infectives are:

• fAUC/MIC, area under the free-drug concentration-time curve relative to MIC (concentration-independent, total-exposure driver). Vancomycin is the prototype.
• Cmax/MIC, peak-to-MIC ratio (concentration-dependent killing). Aminoglycosides are the prototype.
• fT>MIC, time free drug concentration remains above MIC (time-dependent killing). Beta-lactams are the prototype.

For non-antimicrobials (phenytoin, levetiracetam, lithium, calcineurin inhibitors) the PD target is a therapeutic window defined by efficacy and toxicity thresholds established in clinical populations, though the ICU frequently distorts the PK so that "standard" targets no longer reliably follow from standard doses.

Core Equations Underlying TDM

Volume of distribution: $$V_d = \frac{\text{Dose}}{C_0}$$

First-order elimination half-life: $$t_{1/2} = \frac{0.693 \times V_d}{CL}$$

Steady-state concentration (one-compartment, intermittent dosing): $$C_{ss,avg} = \frac{F \times \text{Dose}}{CL \times \tau}$$

AUC (trapezoidal, two-point): $$AUC_{0-\infty} = \frac{C_1 - C_2}{k_e} + \frac{C_2}{k_e}$$

where $k_e = \frac{\ln(C_1/C_2)}{t_2 - t_1}$.

Bayesian estimation integrates population priors with individual measured concentrations to derive individual PK parameters, this is the method underpinning modern vancomycin AUC software tools (e.g. DoseMeRx, InsightRx, Tucuxi).

2. Why ICU Patients Are Different {#icu-different}

The critically ill patient is a PK outlier in almost every domain. Understanding the mechanisms prevents algorithmic errors.

Expanded Volume of Distribution

• Systemic inflammation increases endothelial permeability → third-spacing of fluid → $V_d$ for hydrophilic drugs (vancomycin, aminoglycosides, beta-lactams) increases markedly. A 70 kg patient in septic shock may have an effective $V_d$ for vancomycin of 0.9-1.2 L/kg rather than the population norm of 0.5-0.7 L/kg.
• Clinical consequence: loading doses must be weight-based and generous; initial trough or AUC targets will be missed if doses are conservative.

Altered Clearance

• Augmented renal clearance (ARC): Hyperdynamic physiology in early sepsis, traumatic brain injury, burns, and young patients drives GFR well above 130 mL/min/1.73 m². Renally cleared drugs (vancomycin, gentamicin, levetiracetam, lithium) are eliminated faster than predicted from serum creatinine (which is artefactually low in the presence of reduced muscle mass or rapid fluid dilution). ARC causes subtherapeutic concentrations.
• Acute kidney injury (AKI): Drastically reduces clearance of renally eliminated drugs → accumulation, toxicity.
• Continuous renal replacement therapy (CRRT): Adds a clearance pathway. Drug removal depends on membrane sieving coefficient, molecular weight, protein binding, and effluent flow rate. CRRT substantially alters vancomycin, levetiracetam, and other drug clearance.
• Hepatic dysfunction: Reduces clearance of highly extracted drugs (phenytoin, calcineurin inhibitors). Hypoalbuminaemia alters free fraction of highly protein-bound drugs.

Altered Protein Binding

• Hypoalbuminaemia (almost universal in ICU) reduces protein binding of acidic drugs: phenytoin (~90% bound), valproate, diazepam. Free (pharmacologically active) fraction rises even when total concentration appears low or normal.
• Uraemia causes accumulation of endogenous organic acids that displace drugs from binding sites, worsened in AKI, particularly for phenytoin.

Enteral Absorption Variability

• Impaired gastric emptying, ileus, altered gut blood flow, and drug-drug interactions reduce oral bioavailability unpredictably. Tacrolimus oral absorption varies enormously in the post-operative period.

3. Vancomycin TDM, AUC/MIC Strategy {#vancomycin}

Mechanism of Action and Relevant PD

Vancomycin is a glycopeptide that binds the D-Ala-D-Ala terminus of peptidoglycan precursors in gram-positive organisms (MRSA, MRSE, Enterococcus), preventing cross-linking and causing osmotic lysis. Its killing is concentration-independent but total-exposure dependent: the PD driver is AUC₂₄/MIC.

The MRSA target: AUC₂₄/MIC of 400-600 mg·h/L (most MRSA strains have MIC ≤ 1 mg/L by broth microdilution). This target was codified in the 2020 ASHP/IDSA/SIDP Vancomycin Monitoring Consensus Guidelines.

Why Trough-Only Monitoring Is Obsolete

Legacy practice used a trough target of 15-20 mg/L for serious MRSA infection. Problems:

1. Trough alone does not estimate AUC accurately, it only reflects one point on the concentration-time curve and assumes a fixed relationship that does not hold across the range of $V_d$ and $CL$ seen in ICU patients.
2. Trough-guided therapy to achieve AUC 400-600 required troughs of 15-20 mg/L in many patients, but this drove nephrotoxicity without necessarily achieving AUC target.
3. AUC-guided dosing achieves target exposure at lower trough concentrations in patients with high clearance (ARC), reducing nephrotoxicity.

Key data: A retrospective study by Neely et al. demonstrated that Bayesian AUC-guided vancomycin achieved target AUC₂₄/MIC more reliably than trough-guided methods while reducing nephrotoxicity rates (approximately 15% vs 25%).

Practical AUC Calculation

Two clinically used methods:

1. Two-point Bayesian estimation (preferred): Obtain two concentrations at defined intervals around a dose (e.g. 1-2 h post-infusion and a trough), input into validated Bayesian software (DoseMeRx, InsightRx). Software calculates individual $CL$ and $V_d$, then calculates AUC₂₄.

2. Two-point non-Bayesian (trapezoidal): Using $C_{peak}$ (at end of infusion) and $C_{trough}$: $$AUC_{\tau} = \frac{(C_{peak} + C_{trough})}{2} \times \tau$$ This approximation is less accurate but acceptable when Bayesian software is unavailable.

Dosing Strategy

Monitoring Schedule

• First TDM sample: after 4th-5th dose (approximately steady-state for average CL) or sooner in AKI or suspected ARC.
• With ARC or changing renal function: monitor every 24-48 h.
• Renal function monitoring: serum creatinine every 48 h; daily in those with pre-existing CKD or receiving concurrent nephrotoxins.

Nephrotoxicity

Defined as ≥2 consecutive rises in creatinine ≥0.3 mg/dL (26.5 µmol/L) or ≥50% rise from baseline. Vancomycin-associated AKI risk is potentiated by concurrent piperacillin-tazobactam, loop diuretics, contrast, and pre-existing CKD. Consider switching to daptomycin or teicoplanin in high-risk patients.

4. Aminoglycosides, Extended-Interval Gentamicin {#gentamicin}

Mechanism and PD Driver

Gentamicin (and tobramycin) are aminoglycosides that bind the 30S ribosomal subunit, causing misreading of mRNA and insertion of incorrect amino acids, leading to dysfunctional proteins and membrane disruption. Killing is concentration-dependent: the PD driver is Cmax/MIC.

Aminoglycosides also exhibit:

• Post-antibiotic effect (PAE): Continued bacterial suppression after concentrations fall below MIC. Duration is directly proportional to the Cmax/MIC ratio.
• Adaptive resistance: Repeated low-concentration exposure to aminoglycosides upregulates efflux pumps. Extended-interval dosing (EID) exploits PAE and avoids the concentration window that triggers adaptive resistance.

Extended-Interval Dosing Rationale

EID (once-daily or extended-interval) gives a single large dose to maximise Cmax/MIC, allows trough to fall to near-zero, exploits PAE, and reduces nephrotoxicity compared with multiple daily dosing. Nephrotoxicity with aminoglycosides is caused by accumulation in proximal tubular cells, a saturable process. The trough-free interval allows renal tubular drug egress and partial cellular recovery.

Dosing Protocol

Hartford nomogram: An alternative monitoring approach. One concentration is measured at 6-14 h post-dose and plotted on the Hartford nomogram to determine whether the interval (q24, q36, or q48 h) is appropriate. This approach is designed for the general hospital population; in ICU the variable $V_d$ means Bayesian methods are more reliable.

Toxicities and TDM Rationale

• Nephrotoxicity: Tubular cell accumulation. Prevented by ensuring trough < 1 mg/L.
• Ototoxicity (cochlear): Irreversible sensorineural hearing loss. Correlated with cumulative exposure and sustained high troughs. Less responsive to TDM adjustment than nephrotoxicity.
• Vestibulotoxicity: Also cumulative, assess for oscillopsia, Romberg, dynamic visual acuity in longer courses.

Special ICU Populations

• ARC: Cmax may be achieved only transiently or not at all if clearance is very high. EID may need to use upper end of dose range (7 mg/kg) or empirically check Cmax.
• Obesity: Use adjusted body weight: $ABW = IBW + 0.4 \times (TBW - IBW)$.
• Cystic fibrosis, burns: Significantly expanded $V_d$ and increased CL, may need 10 mg/kg.
• CRRT: Significant drug removal. Check post-filter concentration; consider supplemental doses after CRRT sessions.

5. Phenytoin, Protein Binding and Free-Fraction {#phenytoin}

Mechanism

Phenytoin stabilises neuronal membranes by blocking voltage-gated sodium channels in their inactivated state, reducing repetitive high-frequency neuronal firing. It does not suppress normal neuronal function at therapeutic concentrations.

PK Complexity: Non-Linear Saturable Metabolism

Phenytoin is metabolised hepatically by CYP2C9 (primarily) and CYP2C19. The metabolic pathway is saturable (Michaelis-Menten kinetics), at therapeutic concentrations the enzyme is nearly saturated, meaning: $$\text{Rate of elimination} = \frac{V_{max} \times C}{K_m + C}$$

Small dose increases produce disproportionately large rises in plasma concentration. This is not true first-order kinetics, small dosing errors cause toxicity or subtherapeutic levels unpredictably.

Protein Binding and the Free-Fraction Problem

Phenytoin is ~90% protein-bound to albumin in a healthy adult. Only the free (unbound) fraction is pharmacologically active. The therapeutic range of 10-20 mg/L applies to total phenytoin in a patient with normal albumin (~40 g/L).

In ICU:

• Hypoalbuminaemia: Less binding protein → higher free fraction at any given total concentration. A total phenytoin of 10 mg/L may represent a free concentration of 3 mg/L (therapeutic target for free phenytoin: 1-2 mg/L), risking toxicity.
• Uraemia/AKI: Accumulation of endogenous displacing organic acids → further increases free fraction even when albumin is near-normal.

Winterbottom/Sheiner Correction for Hypoalbuminaemia

Albumin only (no renal failure): $$C_{corrected} = \frac{C_{measured}}{(0.2 \times \text{albumin g/dL}) + 0.1}$$

Combined hypoalbuminaemia and renal failure (GFR < 10 mL/min) or dialysis: $$C_{corrected} = \frac{C_{measured}}{(0.1 \times \text{albumin g/dL}) + 0.1}$$

These formulae estimate what the total concentration would be if albumin were normal (4 g/dL). The corrected value is then compared with the 10-20 mg/L target. If albumin = 2 g/dL (20 g/L), a total phenytoin of 8 mg/L corrects to ~16 mg/L, within range despite a seemingly low total.

Direct free-fraction measurement is preferred in ICU where resources allow, many ANZ laboratories can measure free phenytoin (reference range 1-2 mg/L). This bypasses the correction formula assumptions entirely.

Loading, Maintenance, and Monitoring

6. Levetiracetam {#levetiracetam}

Mechanism

Levetiracetam binds synaptic vesicle glycoprotein 2A (SV2A), modulating vesicular neurotransmitter release. It also inhibits calcium channels and reduces zinc-mediated inhibition of GABA receptors. Its exact mechanism is incompletely characterised but it is highly effective for focal and generalised seizures.

PK Rationale for TDM

Levetiracetam is ~95% renally cleared unchanged, has minimal protein binding (~10%), and a $V_d$ of ~0.5-0.7 L/kg. These properties make it:

• Susceptible to ARC: Young ICU patients with high GFR may have rapid clearance and subtherapeutic concentrations despite standard doses.
• Susceptible to AKI: Accumulates substantially in AKI; dose reduction needed.
• Affected by CRRT: Significantly removed by CRRT, supplemental dosing required.

Routine TDM Controversy

Levetiracetam has a wide therapeutic index and there is no strong evidence that routine TDM improves clinical outcomes in the general hospital population. The therapeutic reference range is typically 12-46 mg/L (trough). TDM is most useful in:

• AKI or CRRT, to avoid accumulation
• ARC or augmented clearance, to ensure adequacy
• Suspected treatment failure despite adequate dosing
• Status epilepticus managed with high-dose levetiracetam

Dosing

No hepatic dose adjustment needed (minimal hepatic metabolism). Psychiatric adverse effects (aggression, psychosis) correlate loosely with high concentrations but not linearly enough to reliably guide TDM.

7. Lithium {#lithium}

Mechanism

Lithium (Li⁺) is a monovalent cation with a poorly characterised but likely multifactorial mood-stabilising mechanism, it inhibits inositol monophosphatase (disrupting phosphatidylinositol signalling), inhibits glycogen synthase kinase-3β (GSK-3β), and modulates monoamine neurotransmitter systems. It has a very narrow therapeutic index.

PK in the ICU

Lithium is eliminated entirely by the kidney with pharmacokinetics similar to sodium, volume of distribution ~0.6-0.9 L/kg, filtered and reabsorbed in the proximal tubule alongside sodium. Conditions that increase sodium reabsorption also increase lithium reabsorption:

• Volume depletion / dehydration, classic cause of lithium toxicity in the community (gastroenteritis, aggressive diuretic use, hot weather).
• AKI, dramatic reduction in clearance.
• NSAIDs, reduce renal prostaglandin synthesis → increase proximal tubular reabsorption of sodium (and lithium).
• ACE inhibitors / ARBs, reduce GFR and aldosterone-mediated distal effects → increase lithium reabsorption.
• Thiazide diuretics, increase proximal reabsorption of sodium (and lithium) as a compensatory response to distal blockade. Particularly dangerous. Loop diuretics are safer but still require monitoring.

Lithium Toxicity

SILENT (syndrome of irreversible lithium-effectuated neurotoxicity): Persistent cerebellar syndrome, cognitive impairment, and brainstem dysfunction even after lithium normalisation, occurs after severe or prolonged toxicity.

ICU Management and TDM

• Timing of lithium levels: Sample ≥ 12 h post last dose (12-h post-dose trough is the standard). Levels taken sooner reflect distribution phase and are unreliable.
• Haemodialysis: Effective at removing lithium. Indicated for level > 3.5-4 mmol/L, severe toxicity (coma, seizures, arrhythmias), or AKI preventing excretion. Rebound rise in concentration post-dialysis is expected (distribution from CNS compartment back into plasma), repeated sessions or prolonged CRRT may be needed.
• CRRT: Less efficient than intermittent haemodialysis but can be used for haemodynamically unstable patients.
• Volume resuscitation: Restoring normovolaemia improves lithium clearance. Do not restrict sodium, lithium clearance depends on normal sodium delivery.

8. Tacrolimus and Cyclosporin in Solid Organ Transplant {#calcineurin}

Mechanism

Both agents are calcineurin inhibitors (CNIs):

• Tacrolimus binds FKBP-12 → complex inhibits calcineurin phosphatase → prevents dephosphorylation of NFAT (nuclear factor of activated T cells) → reduced IL-2 transcription → T-cell suppression.
• Cyclosporin binds cyclophilin → complex inhibits calcineurin → same downstream pathway.

Why TDM Is Critical for CNIs

The transplant ICU intensivist must understand CNI TDM because:

1. The therapeutic window is extremely narrow, the difference between rejection and nephrotoxicity is often a 2-fold concentration change.
2. CNIs are CYP3A4 and P-glycoprotein substrates, drug interactions with azoles (voriconazole, fluconazole), macrolides, rifampicin, phenytoin, and many others cause dramatic concentration changes.
3. Bioavailability in the post-operative period is erratic, gastric dysmotility, altered gut absorption, and biliary diversion (especially liver transplant) alter absorption.
4. Critically ill transplant patients often require drugs that interact with CYP3A4, any infection treatment in this population must prompt CNI level checking.

Pharmacokinetics

Target Concentrations

Targets are organ- and protocol-specific, vary by transplant centre, and change with time post-transplant. The following are indicative, always defer to the transplant team's protocol:

Cyclosporin C2 monitoring: 2-h post-dose concentrations (C2) correlate better with AUC than C0 in some centres. C2 targets: early 1200-1500 ng/mL (renal transplant), gradually reducing to 600-800 ng/mL at 12 months.

Toxicity

• Nephrotoxicity: Afferent arteriolar vasoconstriction → reduced GFR. May be reversible (acute calcineurin inhibitor nephropathy) or progress to chronic interstitial fibrosis and arteriolar hyalinosis.
• Neurotoxicity: Posterior reversible encephalopathy syndrome (PRES), tremor, seizures, more common with tacrolimus, associated with supratherapeutic levels and hypomagnesaemia.
• Metabolic: Hypertension, dyslipidaemia (more cyclosporin), new-onset diabetes post-transplant, NODAT (more tacrolimus), gingival hypertrophy, hirsutism (cyclosporin).
• Hyperkalaemia/hypomagnesaemia: Through renal tubular effects.

Critical Drug Interactions in ICU

9. TDM in Special ICU Contexts {#special}

CRRT and Drug Removal

The clearance of a drug by CRRT depends on: $$CL_{CRRT} = SC \times Q_{eff}$$

where $SC$ (sieving coefficient) $\approx$ fraction unbound, and $Q_{eff}$ is effluent flow rate (typically 20-35 mL/kg/h for standard CRRT).

Drugs with low protein binding and small molecular weight are efficiently removed. Implications:

ARC Recognition

Suspect ARC when:

• Age < 50 years
• SOFA ≤ 10 (early phase sepsis)
• Trauma, burns, haematological malignancy
• Serum creatinine < 60 µmol/L (despite ICU stay)

Confirm with 8- or 24-h measured urine creatinine clearance. Measured CrCl > 130 mL/min/1.73 m² defines ARC. Do not use Cockcroft-Gault or CKD-EPI in this context, both underestimate GFR in ARC because they rely on creatinine generation assumptions that do not hold in ICU.

10. Evidence Base {#evidence}

Vancomycin

• 2020 ASHP/IDSA/SIDP Vancomycin Monitoring Consensus Guidelines: Landmark document formally replacing trough-only monitoring with AUC/MIC-guided dosing (AUC₂₄ 400-600 mg·h/L for MRSA). Endorsed Bayesian software-assisted AUC calculation.
• CAMERA2 trial (NEJM 2019): Combination vancomycin plus flucloxacillin for MRSA bacteraemia, the combination arm had higher rates of AKI compared with vancomycin monotherapy, highlighting nephrotoxicity as a key outcome in vancomycin TDM research.
• Retrospective Neely et al. data: Bayesian AUC-guided therapy resulted in higher rates of target attainment with lower rates of nephrotoxicity compared

Sources

• ANZCOR / ARC Resuscitation Guidelines
• ANZICS guidelines & CORE registry
• Surviving Sepsis Campaign