15.3 Therapeutic Drug Monitoring
Key Takeaways
- Therapeutic drug monitoring (TDM) adds value for drugs with a narrow therapeutic index, no easy clinical endpoint, nonlinear kinetics, or when toxicity or adherence is in question.
- Digoxin samples must be drawn at least 6 hours post-dose (ideally 12–24 h); target 0.5–2.0 ng/mL, and hypokalaemia potentiates toxicity even at 'normal' levels.
- Phenytoin (10–20 mg/L) is nonlinear and highly protein-bound — correct the measured level for low albumin before interpreting it.
- Lithium is sampled as a standardised 12-hour post-dose level (target 0.6–1.0 mmol/L for maintenance); aminoglycosides use peak/trough or nomogram-based timing, and vancomycin is best monitored by AUC/MIC 400–600 rather than trough alone.
- The commonest TDM disasters are sampling errors — wrong timing, lines contaminated with the drug, and adsorption to gel separator tubes — so always interpret a level against the dose history and the clinical picture.
When TDM Adds Value — and When It Does Not
Therapeutic drug monitoring (TDM) is the use of measured drug concentrations to individualise dosing. It is worth doing only when the result will change management. The classic indications are:
- Narrow therapeutic index — the gap between effective and toxic concentrations is small (digoxin, lithium, phenytoin, theophylline, aminoglycosides, ciclosporin, tacrolimus)
- No easy clinical endpoint — you cannot titrate seizure prophylaxis or transplant rejection prophylaxis by observation the way you titrate an antihypertensive to blood pressure
- Nonlinear or unpredictable kinetics — phenytoin's saturation kinetics make dose–level relationships counter-intuitive
- Suspected toxicity or non-adherence — a level distinguishes under-dosing from non-adherence and confirms or refutes toxicity
- Major physiological change — renal failure, pregnancy, critical illness, or interacting drug started/stopped
Conversely, TDM is wasted effort when a reliable clinical or surrogate endpoint exists (blood pressure for antihypertensives, INR for warfarin, glucose for antidiabetics) or when the assay cannot be timed or interpreted correctly. A level drawn at the wrong time is worse than no level, because it invites a confident wrong decision.
Reference Targets and Sample Timing
| Drug | Target range | Sample timing | Key notes |
|---|---|---|---|
| Digoxin | 0.5–2.0 ng/mL (lower end, 0.5–0.9, preferred in heart failure) | ≥6 h post-dose; ideally 12–24 h | Early samples read falsely high — distribution into tissue is incomplete; hypokalaemia potentiates toxicity even at in-range levels |
| Lithium | 0.6–1.0 mmol/L maintenance (up to 1.2 acutely) | Standardised 12-h post-dose sample | Check renal and thyroid function; dehydration and NSAIDs raise levels |
| Phenytoin | 10–20 mg/L | Trough at steady state (may take 2+ weeks) | Nonlinear; 90% albumin-bound — correct for hypoalbuminaemia |
| Valproate | 50–100 mg/L | Trough | Poor level–effect correlation; interpret with the clinical picture |
| Carbamazepine | 4–12 mg/L | Trough | Autoinduction lowers levels over the first weeks of therapy |
| Theophylline | 10–20 mg/L | Steady-state sample (timing depends on formulation) | Nonlinear at higher levels; smoking induces, ciprofloxacin inhibits clearance |
| Aminoglycosides (gentamicin/tobramycin) | Conventional: peak 5–10 mg/L, trough <2 mg/L | Peak ~30–60 min after infusion ends; trough pre-dose | Once-daily regimens use a single timed level interpreted against the Hartford nomogram (e.g. 6–14 h post-dose) rather than classic peak/trough targets |
| Vancomycin | AUC/MIC 400–600 (MIC assumed 1 mg/L) | Two levels at steady state for AUC estimation, or Bayesian software | AUC-guided dosing is preferred over the older trough-only target of 15–20 mg/L |
| Ciclosporin / tacrolimus | Assay- and indication-specific | C0 trough (pre-dose, 12 h) | CYP3A substrates — levels soar with azole antifungals and crash with enzyme inducers; grapefruit juice raises levels |
Sampling Errors That Ruin Good Assays
- Wrong timing — the single most frequent error. A digoxin level drawn 2 hours post-dose can read twice the interpretable value because distribution is incomplete; an aminoglycoside 'trough' drawn mid-interval is meaningless. Document the exact dose and sample times on the request.
- Contaminated line — blood drawn from the same cannula used to infuse the drug (classic with vancomycin and aminoglycosides) reads astronomically high. Draw from the opposite limb or after an adequate flush.
- Gel separator tubes — some drugs adsorb to the serum-separator gel over time, falsely lowering results (phenytoin is the textbook concern); use the tube type the laboratory specifies and separate promptly.
- Sampling before steady state — a level taken two half-lives into a new regimen reflects a concentration still climbing; the apparent 'subtherapeutic' result tempts an unnecessary dose increase.
Interpreting a Level in Clinical Context
A concentration is a number in search of a story. Ask: was it steady state? Was timing correct? Does the patient look toxic or controlled? Two worked situations illustrate this. First, digoxin with hypokalaemia: a level of 1.8 ng/mL — technically in range — in a patient with potassium 2.8 mmol/L, nausea and new arrhythmia is toxicity; hypokalaemia increases digoxin's binding to Na⁺/K⁺-ATPase, so the tissue effect exceeds what the number suggests. Correct the potassium and treat the patient, not the assay. Second, phenytoin with hypoalbuminaemia: only unbound drug is active, and in low-albumin states (uraemia, pregnancy, critical illness) the measured total level underestimates the active fraction. The Sheiner–Tozer correction adjusts it:
C_adjusted = C_measured / (0.2 × albumin + 0.1)
with albumin in g/dL. Worked example: measured total phenytoin 8 mg/L with albumin 2.0 g/dL gives 8 / (0.2 × 2.0 + 0.1) = 8 / 0.5 = 16 mg/L — an apparently subtherapeutic level is actually mid-therapeutic, and a knee-jerk dose increase could have caused toxicity.
Communicating and Acting on Dose Changes
Make one change at a time, in proportion for linear drugs and in small increments for nonlinear ones, and state the rationale in the record: current regimen, measured level with sample time, target, new regimen, and the date for the repeat level — after 4–5 half-lives for linear drugs, and after two or more weeks for phenytoin. Tell the patient exactly what changed and which toxicity or under-treatment symptoms should trigger early review. TDM only improves outcomes when the loop — sample, interpret, adjust, re-sample — is closed.
When should a blood sample for digoxin therapeutic drug monitoring be drawn?
A patient with albumin 2.0 g/dL has a measured total phenytoin level of 8 mg/L. Using the Sheiner–Tozer correction, what is the albumin-adjusted level?
Which approach is currently preferred for monitoring vancomycin efficacy and safety in serious infections?