Therapeutic drug monitoring for antiepileptic drugs — steady-state timing, range interpretation, phenytoin's nonlinear kinetics, and toxicity-guided dose adjustment
A drug level is only interpretable if it was drawn at the correct time relative to dosing history. For most antiepileptic drugs (AEDs), a level should be obtained at steady state — roughly five half-lives after starting or changing a dose — and, by convention, as a trough level drawn immediately before the next scheduled dose. Ignoring timing is the single most common cause of a "confusing" or misleading level.
After any change in dose (initiation, increase, decrease, or a change in a co-administered interacting drug), serum concentration does not jump instantly to its new equilibrium. It accumulates (or washes out) exponentially, approaching the new steady-state value asymptotically.
After one half-life, concentration has moved about 50% of the way to the new steady state. After two half-lives, ~75%. After roughly five half-lives, accumulation is >95% complete — the pragmatic threshold clinicians use to call a level "at steady state."
For phenytoin, whose half-life ranges widely (roughly 7–42 hours depending on genetics, age, liver function, and interacting drugs), five half-lives can mean anywhere from about 1.5 days to over a week. A level drawn on day 2 after a dose increase may still be climbing — checking it and reacting to a "low" number by increasing the dose again risks overshooting once the earlier increase finally catches up.
Concentration also fluctuates within a dosing interval: it rises after each dose (toward a peak) and falls before the next one (to a trough). For most therapeutic decisions, the trough level — drawn just before the next dose — is preferred because:
• It is the most reproducible point in the cycle across labs and clinics • It represents the lowest concentration the patient experiences, relevant to breakthrough seizures • It avoids the peak-related noise that complicates comparison between visits
Peak levels are occasionally useful when investigating dose-related toxicity that occurs shortly after dosing, but routine monitoring overwhelmingly uses trough sampling.
A level drawn before steady state is reached is not simply "less accurate" — it is systematically biased low relative to where the patient will eventually equilibrate. Reacting to it with another dose increase is a common cause of iatrogenic toxicity days later.
Laboratory-reported therapeutic ranges (for phenytoin, classically 10–20 mcg/mL total drug) were derived from population studies correlating concentration with average seizure control and average toxicity risk. They are statistically useful starting points — but individual patients routinely fall outside them while still being correctly managed.
The 10–20 mcg/mL phenytoin range represents the concentration window within which most patients in historical cohorts achieved seizure control with an acceptable toxicity burden. "Most" is the operative word — it is neither a floor below which the drug cannot work, nor a ceiling above which toxicity is guaranteed for every patient.
Some patients are seizure-free at levels of 6–8 mcg/mL and should not automatically have their dose increased just to reach the "textbook" range if they are already controlled and asymptomatic. Others require levels above 20 mcg/mL to control seizures and tolerate it without clinical toxicity — increasing their dose to a level "in range" may under-treat them.
Most standard TDM assays measure total phenytoin (protein-bound + free/unbound). Only the free fraction (~10% in a typical patient) is pharmacologically active and crosses into the brain.
In conditions that lower albumin or displace phenytoin from albumin — hypoalbuminemia, renal failure, pregnancy, concurrent valproate — the free fraction rises even though the total level may look normal or low. A "subtherapeutic" total level in such a patient can actually correspond to a therapeutic, or even toxic, free level. Free phenytoin levels (or a corrected total, using the Sheiner-Tozer equation) should be requested when these conditions are present.
The clinical rule of thumb for AED level interpretation: use the level to explain and contextualize the clinical picture, not to override it.
• Level low + seizure-free + no toxicity → usually no change needed • Level in range + breakthrough seizures → consider increasing, factoring in kinetics (Stage 3) • Level high + no toxicity signs → consider whether free level correction is needed before reducing dose • Level high + toxicity signs → reduce, regardless of "how close to range" it looks
Each AED (valproate, carbamazepine, levetiracetam, lamotrigine) has its own reference range and its own caveats — phenytoin is used here because its nonlinear kinetics make correct interpretation especially high-stakes.
A therapeutic range is a starting hypothesis for a conversation with the patient, not a laboratory verdict. The best predictor of whether to change a dose is the seizure diary and the exam, with the level explaining why.
Most AEDs follow first-order (linear) elimination: clearance is constant, and steady-state level rises roughly proportionally with dose. Phenytoin is the classic exception. At clinically relevant doses, its hepatic metabolizing enzyme (CYP2C9) becomes saturated, elimination shifts toward zero-order kinetics, and small dose increases can produce disproportionately large — sometimes dangerous — increases in serum level.
For a drug with linear (first-order) kinetics, the rate of elimination is proportional to concentration — clearance (CL) is a constant, and steady-state concentration (Css) scales directly with dosing rate: Css = Dose / CL. Doubling the dose roughly doubles the level. This holds for most AEDs (levetiracetam, lamotrigine, lacosamide) across their clinical dose range.
Phenytoin instead follows Michaelis-Menten kinetics: the rate of elimination is described by Vmax × C / (Km + C), where Vmax is the maximum metabolic capacity of the enzyme system and Km is the concentration at which elimination runs at half its maximum rate. At low concentrations, elimination behaves nearly linearly. But as the dosing rate approaches Vmax, the enzyme system becomes saturated — it simply cannot metabolize drug any faster — and any further dose increase accumulates disproportionately, because clearance itself is falling as concentration rises.
Rearranged for steady state, the relationship becomes approximately: Css = Km × Dose / (Vmax − Dose). As Dose approaches Vmax, the denominator shrinks toward zero and Css rises steeply — theoretically without bound. In practice, this means:
• At low-to-moderate doses, a 10% dose increase produces roughly a 10% level increase — behaves "normally" • Near the top of the therapeutic range, the same 10% dose increase can produce a 50–100%+ level increase • Patients can move from comfortably subtherapeutic to significantly toxic with what looks like a modest dose adjustment
This is precisely why phenytoin dose titration should slow down as the level approaches the therapeutic range — increments of only 30–50 mg, followed by a repeat level after steady state, rather than the larger jumps that are safe at lower doses.
Both Vmax and Km vary substantially between patients due to genetics (CYP2C9 and CYP2C19 polymorphisms), age, liver function, and drug interactions (enzyme inducers like carbamazepine raise apparent Vmax; inhibitors like fluconazole lower it). This means the exact dose at which nonlinearity becomes clinically important differs from patient to patient — there is no single universal "danger dose." The practical takeaway is behavioral, not numerical: titrate cautiously and recheck levels at steady state whenever a patient is near or above the middle of the therapeutic range, rather than relying on a fixed mg-per-kg formula.
Because phenytoin clearance falls as concentration rises, the dose-level relationship is not just nonlinear — it is unstable near saturation. A patient stable for months on one dose can tip into toxicity from a small increase, an added enzyme inhibitor, or even a drop in albumin. This is the single most important pharmacokinetic fact to know before adjusting a phenytoin dose.
Phenytoin toxicity follows a roughly dose-dependent, sequential pattern of neurological signs as serum concentration climbs above the therapeutic range. Recognizing this cascade — and examining for it whenever a level is borderline-high — turns a number on a lab report into an actionable clinical assessment.
Phenytoin toxicity is a textbook example of a concentration-dependent adverse effect gradient, useful precisely because the signs appear in a broadly reproducible order as level rises above the therapeutic ceiling (~20 mcg/mL):
• Nystagmus (involuntary rhythmic eye movement, typically horizontal, on lateral gaze): often the earliest and most sensitive sign, frequently appearing near or just above the upper therapeutic boundary • Ataxia (unsteady, wide-based gait; limb incoordination; dysarthria): appears as levels climb further, reflecting cerebellar dysfunction • Sedation, lethargy, confusion: significant central nervous system depression at higher levels, with cognitive slowing that can be mistaken for other causes • Severe toxicity — stupor, coma, and at very high levels, paradoxically increased seizure activity — occurs at markedly elevated concentrations and constitutes a medical emergency
These thresholds are approximate and vary between patients; some tolerate levels others would find incapacitating, especially with chronic exposure (tolerance) — another reason clinical exam outweighs the number alone.
None of these signs is specific to phenytoin toxicity — nystagmus, ataxia, and sedation can arise from alcohol, other sedating drugs, cerebellar lesions, or the underlying epilepsy syndrome itself. A structured approach:
1. Check timing: are new signs temporally linked to a recent dose increase or a new interacting drug (e.g., an azole antifungal inhibiting metabolism)? 2. Check the level: if drawn appropriately (Stage 1), does it correlate with the exam? 3. Consider free level: hypoalbuminemia or renal disease can produce toxicity signs at a "normal-looking" total level 4. Exclude alternative causes: infection, other CNS-active medications, electrolyte disturbance
When signs and an elevated (or appropriately-corrected) level agree, the diagnosis of phenytoin toxicity is straightforward and should prompt dose reduction.
Because of the steep nonlinear dose-level relationship described in Stage 3, a patient can cross from therapeutic to toxic with a change that looks small on the dosing chart. Toxicity monitoring therefore cannot rely on the level check alone at scheduled intervals — clinicians should have a low threshold to examine for nystagmus and gait instability whenever a dose has recently increased, an interacting drug has been added, or the patient or family reports new unsteadiness, double vision, or excessive drowsiness.
Nystagmus, ataxia, and sedation form a rough dose-response ladder for phenytoin toxicity — but the exact concentration at which each appears varies by patient and by chronicity of exposure. Use the ladder to organize your exam, not as a rigid cutoff to wait for before acting.
A correct dose-adjustment decision synthesizes everything from the previous stages: was the level drawn at the right time, how does it compare to the range in the context of this patient's clinical response, and — critically for phenytoin — how much will a given dose change actually move the level, given where the patient sits on the saturation curve.
A defensible titration decision follows a consistent order of questions:
1. Was this level drawn at steady state, as a trough? If not, its absolute value should not drive a dose change — recheck first. 2. Does the level match the clinical picture? A "toxic" level with no clinical toxicity may reflect free-fraction issues or assay timing; a "subtherapeutic" level with good seizure control may not need correction at all. 3. Where is this dose on the saturation curve? Near the top of the therapeutic range, treat every increment as more powerful than it looks — small steps, then recheck. 4. What is the actionable output? Continue, hold and recheck, or adjust — and by how much.
This sequence prevents the two most common errors in AED titration: reacting to an unsteady-state level, and making a linear-kinetics-sized dose jump on a nonlinear drug.
Because phenytoin's dose-level relationship steepens as saturation is approached, the "right-sized" adjustment is not a fixed percentage of the current dose — it depends on where the patient sits on the curve:
• Well below the therapeutic range: larger increments are reasonably safe, because kinetics there behave closer to linear • Approaching or within the upper half of the range: reduce increments to about 30–50 mg per step, and always recheck a level at steady state before adjusting again • Above the therapeutic range with toxicity: even a modest dose reduction can produce a large drop in level, because the same steep part of the curve works in reverse — patients often improve clinically within days of a small cut
After any change, the plan must include a defined recheck time — generally another ~5 half-lives later — before the next adjustment is made, closing the loop back to Stage 1.
Good TDM-driven prescribing documents the reasoning chain: the level, when and how it was drawn, whether it was at steady state, the clinical correlation (seizure control, exam for toxicity), the kinetic context (where on the curve), and the specific adjustment with its rationale. This turns a single lab value into a reproducible clinical decision that the next clinician — or the same clinician at the next visit — can follow and verify, rather than re-deriving from scratch.
The safest phenytoin titration mindset: think in half-lives, not calendar days; think in saturation curves, not straight lines; and let the exam, not the lab value alone, decide whether the patient actually needs a change.