Published by Unseen Progress, an independent publisher of caregiver research. Last reviewed 2026-05-10. Part of the epilepsy caregiver research overview.
Short answer. A breakthrough seizure in clinical use means a seizure occurring during a period of previously stable control. But the literature is clear that not every breakthrough event marks a real change in the underlying condition: epilepsy is intrinsically variable, missed doses produce transient breakthroughs that resolve on their own, and statistical noise alone produces clusters that look like deterioration (Berg et al., 2008; French et al., 2004). Distinguishing a true clinical change from random variation is a 30-day judgement, not a same-day judgement.
Seizure frequency in any individual patient is not a constant — it is a distribution. A patient described as "having about one seizure a month" might go three months with none and then have two in a week, with no underlying change in their condition. The Berg cohort studies of paediatric epilepsy outcomes (Berg et al., 2001; Berg et al., 2008) show that even among patients in long-term remission, occasional events occur in roughly 10–20% within any given two-year window, often without any identifiable cause and without progression. The ILAE's 2014 practical definition (Fisher et al., 2014) treats sustained seizure freedom over a defined window as the meaningful unit, precisely because individual events are too noisy to interpret.
This is the foundation for understanding breakthroughs. The right unit is "what is the seizure rate in this 30-day window compared to the preceding 30-day window?", not "did a seizure happen today?" A single event after a long quiet stretch is a breakthrough by definition, but it is rarely, by itself, evidence that the medication has stopped working.
Across the clinical guidance literature, the most common identified causes of breakthrough seizures in previously controlled epilepsy are, in roughly descending frequency:
When a breakthrough occurs, the first clinical question is which of these explains it? — not should we change the medication? If a transient driver is identified and removed, the breakthrough does not signal a baseline shift.
Even with no real underlying change, seizure counts cluster. Two events in one week, after weeks of none, are entirely consistent with random variation if the long-run rate is non-zero. The literature on epilepsy variability is clear that one event, or even a small cluster, in a previously controlled patient is more likely random variation or a transient driver than a true frequency shift. Acting on a single event by changing medication often introduces avoidable side-effect burden and reshuffles a regimen that was working.
The clinical heuristic that maps to the data is:
The features that the literature treats as evidence of a real change in the underlying condition, rather than variability within the same baseline, are:
These are the qualitative changes that the ILAE recommends triggering a re-evaluation of the diagnosis, the medication choice, or the underlying syndrome — independent of whether the count has shifted.
Caregivers who treat every breakthrough as evidence the medication has failed often push for changes that the data doesn't support. The asymmetry of the situation — every breakthrough is dramatic, every quiet stretch is invisible — pulls toward over-reaction. The research-backed counter is to measure the medication on 30-day windows, not on individual events, and to identify and remove transient drivers before adjusting the regimen.
The neurologist's framing in the AAN/AES guidelines (French et al., 2004) is essentially: a working medication should not be abandoned for a single event with an identifiable transient cause. The cost of an unnecessary medication change — re-titration, new side effects, possible loss of the control the previous drug was providing — usually exceeds the cost of accepting that breakthroughs occur even on effective regimens.
1. When a breakthrough occurs, log the event with the same fields as any other seizure, then run the transient-driver checklist before drawing any conclusion. 2. Do not change medication on the basis of a single event unless a new seizure type or status epilepticus has occurred. 3. Compare the 30-day rolling rate after the event against the 30-day rolling rate before. If they are similar, the breakthrough is variability; if they have shifted, it is a baseline change. 4. For a cluster of events in one week, contact the neurologist same-day for evaluation but distinguish "rescue protocol or short-term adjustment" from "primary medication change." 5. Keep the qualitative changes — new types, lost aura, longer postictal — on a separate tracking line from frequency. A single qualitative change can justify re-evaluation even with a stable count.
---
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.