Decades of Epilepsy Surgery Follow-up: What's Known and What's Not?

Abstract

Long-term outcomes of epilepsy surgery: A 25-year experience from a tertiary referral center

Almeida M, Barros F, Cunha I, Brás A, Teotónio R, Bento C, Sales F. Epileptic Disord. 2025 Dec;27(6):1217-1226. doi: 10.1002/epd2.70101. Epub 2025 Sep 6. PMID: 40913513; PMCID: PMC12747684.

Objective: Despite pharmacological advances in epilepsy treatment, one-third of patients remain pharmacoresistant and may require surgery. Despite extensive literature on epilepsy surgery, studies with follow-ups longer than 5 years are rare. Our goal was to analyze the outcomes of patients undergoing epilepsy surgery at our center, with a minimum follow-up of 15 years. Methods: This was a retrospective study of prospectively collected data. We used the Engel classification to assess seizure freedom, performed univariate descriptive analysis of the variables of interest, and applied appropriate correlation tests for nominal and categorical variables, with statistical significance set at 0.05. Results: We included 160 patients with a minimum follow-up of 15 years. A total of 105 (70%) patients underwent resective surgeries, the most common being lesionectomy (46.7%), followed by anterior temporal lobectomy with amygdalectomy (21.9%). Among resective surgeries, 73.6% used intraoperative ECOG. Most surgeries were in the temporal lobe (68.8%), and mesial sclerosis was the most frequent etiology (33.8%), followed by long-term epilepsy-associated tumors (LEAT) (25.6%). Seizure freedom at 15 years was achieved by 57.5% of patients, and most of the remaining patients (63.2%) had rare disabling seizures. The majority (65%) discontinued at least 1 ASM. Temporal surgeries (χ²(1) = 8.444, P < .05), left-sided surgeries (χ²(1) = 6.436, P  = .04), mesial sclerosis (χ²(1) = 50.870, P  = .024), and the use of intraoperative ECOG (χ²(1) = 23.235, P  < .001) were associated with a better prognosis. No differences in outcome were found between the different temporal lobe surgeries (Fisher's exact test value=0.859, P  = .659). Significance: Appropriate referral to a refractory epilepsy center permits a multidisciplinary approach that can result in long-term seizure freedom for most patients undergoing surgery, especially for left-temporal lobe surgeries performed with the aid of intraoperative monitoring techniques.

Commentary

For focal drug-resistant epilepsy, resective epilepsy surgery is the most effective treatment available and is associated with improved seizure control, improved cognitive outcomes, and enhanced quality of life.^1‐3 However, when counseling patients and caregivers, we often rely on the 1- to 2-year outcome data available in the clinical trial literature. Far less is known, and debate persists, about the durability of these benefits over the long term. ⁴ Existing literature suggests that seizure control declines as a function of time since surgery, following a pattern familiar from survival analysis, steep in the early postoperative years, then flattening. ⁵ But exactly how far does it decline, and when, if ever, does it stop?

Recently, Almeida et al reported cross-sectional, single-center outcomes of 160 patients with a minimum follow-up of 15 years (mean 17.2 years) from a retrospective analysis of prospectively collected data. ⁴ This work is significant because it extends the field's typical outcome window well beyond the 5- to 10-year horizon most commonly reported. ⁶ In this cohort, seizure freedom assessed at this single, distant timepoint was achieved in 57.5% of patients, with the majority (63.2%) of the remaining patients experiencing rare disabling seizures. The majority of patients in this cohort (65%) discontinued at least 1 ASM. Interestingly, better outcomes were associated with temporal lobe surgery, left-sided surgery, mesial sclerosis, and use of intraoperative electrocorticography (ECOG).

In their review on long-term outcomes of epilepsy surgery, Malmgren and Edelvik note the paucity of outcome data beyond 5 to 10 years. ⁶ Almeida et al's work fills this gap directly, showing that seizure control can indeed be durable well past that horizon. Notably, the 15-year follow-up was an inclusion criterion, not a mean reflecting a minority of long-outlier patients that can inflate while most contribute far shorter data; de Tisi et al's cohort, for example, ranges from 1 to 19 years despite a median of 8. ⁷ Even Mohammed et al's 26-year mean follow-up, the longest reported to date, was achieved in only the 32% of patients who could be located and agreed to a survey, which reflects a tradeoff between duration and completeness that Almeida et al's inclusion-criterion approach largely avoids. ⁸ Almeida et al's suggested predictors long suspected from shorter-term literature: that temporal lobe location and MRI-positive lesions such as hippocampal sclerosis predict better long-term outcomes, suggesting that these predictors may extend beyond the timeframes in which they were established.

Several methodological limitations suggest further work is still needed. Importantly, patients who died before reaching 15 years of follow-up were excluded from the study entirely, rather than being censored at the time of death as a survival analysis would allow. This risks survivorship bias: patients with poorly controlled epilepsy face an elevated risk of premature mortality, including SUDEP, so their removal may inflate the reported seizure-freedom rate relative to the true long-term surgical population. A Kaplan-Meier analysis, as used by de Tisi et al and Mohan et al, would have addressed this directly, censoring patients with death or incomplete follow-up at their last known status rather than discarding them outright.^7,9 This is a missed opportunity for a study whose main contribution is long-term follow-up, where such losses are most likely to accumulate. Relatedly, pooling multiple surgical approaches and etiologies, while pragmatic and reflective of real-world practice, introduces clinical heterogeneity that limits the generalizability of the findings.

These types of long-term outcomes studies are crucial to understanding the trajectory of epilepsy surgery and accurate patient counseling. The limitations above point toward possible future directions: prospective registry analyses that censor rather than exclude patients lost to follow-up, survival analyses that capture the full trajectory of remission and relapse, and multivariable models capable of disentangling confounded predictors. Achieving this at scale will require long-term, multicenter registries, since even the largest single-center cohorts in this literature, including Almeida et al's own, are constrained by the patient numbers any 1 center can accrue over a 15- to 25-year window. The Swedish National Epilepsy Surgery Register offers a great example: a population-based, prospective registry that has already produced 10-year outcome data alongside a nonoperative comparison group, providing valuable comparative effectiveness results. ¹⁰ Electronic medical records and cross-institutional data sharing will be essential to building comparable registries elsewhere, tracking individual trajectories over decades rather than reconstructing them retrospectively, as in the present study. Malmgren and Edelvik similarly call for more prospective, longitudinal studies of individual patient trajectories, and this kind of infrastructure will ultimately let us model epilepsy surgery outcomes with the precision our patients deserve, particularly as neuromodulation enters direct comparison with resective surgery's decades-long track record.

Ultimately, Almeida et al have taken an important step forward in our understanding of long-term epilepsy surgery outcomes, even as their limitations illustrate how much we still have to learn. There remains much we do not know about the durability of seizure freedom, the trajectories patients follow between remission and relapse, and the factors that reliably predict which patients will sustain benefit decades after surgery. However, the tools to answer these questions are within reach: electronic infrastructure that consolidates outcomes into large, prospective registries and novel analytic approaches to understand the data. Realizing that potential will take deliberate research investment, not merely accumulating years of follow-up data. Twenty years from now, we should know a great deal more about how durable the benefits of epilepsy surgery truly are, but only if the studies conducted between now and then are designed to answer that question, rather than simply to report it.