Published by Unseen Progress, an independent publisher of caregiver research. Last reviewed 2026-05-10. Part of the epilepsy caregiver research overview.
Short answer. Anti-seizure medications (ASMs) need 4–8 weeks of titration to reach the target dose, another 4–8 weeks at that dose to reach steady-state pharmacokinetic effect, and typically 3–6 months total before a meaningful change in seizure frequency can be reliably distinguished from normal variation (French et al., 2004; Kwan et al., ILAE, 2010). Caregivers who decide a medication isn't working in the first few weeks are almost always evaluating before the drug has had a chance to declare itself.
The Kwan and Brodie cohort study (Kwan & Brodie, 2000) — still the foundational reference for ASM response — defined a successful trial as at least 12 months of seizure freedom at an adequate dose. That definition exists because seizure frequency is intrinsically noisy: a patient with one seizure a month can go three months without one and then have two in a week, without any underlying change in their condition. Distinguishing genuine response from random quiet stretches requires enough time for the noise to average out. The literature converges on three time-scales that matter for this distinction:
These are not arbitrary numbers. They reflect both the pharmacology of the drug class and the statistics of seizure counting.
The titration period itself is rarely at therapeutic dose. ASMs are introduced slowly to limit side effects — typical titration ladders take 2–6 weeks to reach the target dose for drugs like lamotrigine (mandatorily slow due to rash risk), levetiracetam (faster), or valproate (intermediate). Seizures during titration are not evidence the drug failed; they are evidence the drug has not yet been tried. French and colleagues' AAN/AES guidelines (French et al., 2004) emphasise this point repeatedly, because the most common failure mode in clinical practice is concluding non-response and switching before the drug has had its chance.
A second source of misleading first-weeks data is that caregivers' attention sharpens during the titration window, and seizures that would otherwise have been missed start being captured. The frequency in the diary appears to rise even when underlying frequency is flat or falling. This is the same Hoppe-style under-counting that affects baseline data, now resolving in the wrong direction relative to perception.
The ILAE consensus paper (Kwan et al., 2010) operationalises an adequate trial as:
1. Appropriately chosen for seizure type and syndrome. Carbamazepine is wrong for absence seizures; ethosuximide is wrong for tonic-clonic alone. The drug must match the diagnosis. 2. Tolerated. If side effects forced a dose drop below therapeutic, the drug has not been adequately tried — it has been intolerable. This is a different clinical picture than non-response. 3. Adequately dosed. Reached the target therapeutic dose, sustained for long enough to evaluate. 4. Adequate duration. Sustained at target dose for at least three half-lives plus the clinical response window — practically, 3–6 months minimum for most ASMs at typical seizure frequencies.
A trial that fails on criterion 1, 2, or 3 is not informative about whether the drug works for that patient. It only counts toward the two-drug rule when all four are satisfied.
Practical clinical guidance in Wilner (2008) and the AAN/AES guidelines suggests a measurable evaluation framework that maps to the pharmacology:
This is the framework the neurologist's decision actually runs on. Caregivers who measure the medication this way usually find that the felt-sense of "not working" at week 6 was premature, and that by week 12 the data has either confirmed or denied the early reading.
Across the trial-and-error phase of epilepsy, the most common reason the two-drug rule is reached without surgical evaluation is that none of the early drugs was given an adequate trial (French et al., 2004; Kwan & Brodie, 2000). Families switch at week 4, switch again at week 6, conclude after the third switch that the patient is drug-resistant — when in fact none of the three has been tested under the ILAE criteria. Premature switching costs the patient months of seizure exposure and delays the conversation about comprehensive epilepsy center referral, because the failures don't count.
The asymmetric incentive matters here. Continuing a drug that turns out not to work costs weeks of seizures. Switching prematurely costs the same weeks plus the ability to learn anything from the trial. The research consistently favours patience to the point of an adequate trial.
1. Write down the date the dose change started, the target dose, and the day-30 / day-60 / day-90 evaluation dates before the trial begins. 2. At day 30, evaluate tolerability only. If it is intolerable, address the side effects or change drugs — not because of seizure data, but because criterion 2 has failed. 3. At day 60, compare 30-day rolling frequency against the pre-trial 30 days. Note the direction; do not act on it yet. 4. At day 90, run the same comparison with the most recent 30 days. A clear directional shift is interpretable; a flat or rising trend at maximum tolerated dose justifies the next conversation. 5. Bring the four ILAE criteria explicitly to any "is this working?" discussion with the neurologist. If any of them is unmet, the trial is not yet a trial.
---
Unseen Progress publishes long-form caregiver research and builds research-backed daily trackers for the families covered. See the full epilepsy caregiver research overview for the complete framework.