Surgical treatment of epilepsy in polymicrogyria: A subject‐level meta‐analysis and decision‐making framework

Abstract

Objective

Polymicrogyria (PMG) presents a complex challenge in epilepsy surgery. The optimal surgical strategy and extent of resection, from hemispheric to more limited approaches, remain debated. We aimed to summarize subject‐level surgical outcomes and identify factors informing procedure selection.

Methods

We conducted a subject‐level pooled analysis of 161 patients across 20 retrospective studies, evaluating surgical outcomes (Engel classification), in relation to anatomic extent, surgical procedure, use of intracranial electroencephalography (ICEEG), and other decision‐influencing factors.

Results

At ≥12 months follow‐up, ~70% of patients achieved seizure freedom (Engel Class I). In our univariate analysis Engel Class I outcomes were associated with shorter epilepsy duration (8.05 vs 11.92 years, p = 0.009). Hemispheric PMG was linked to earlier seizure onset (p = 0.02) and a higher incidence of epileptic encephalopathy with spike–wave activation in sleep (p < 0.0005). Among unilateral non‐hemispheric and bilateral PMG cases, seizure‐freedom rates were similar between hemispheric and more limited resections, but the latter were associated with a lower incidence of new or worsened motor deficits. Mixed‐effects logistic regression (n = 160) showed that hemispheric surgery increased the odds of seizure freedom but without statistical significance (odds ratio [OR] = 3.52, p = 0.055). ICEEG did not significantly influence seizure outcomes but may play a key role in identifying eloquent cortex and guiding safer, tailored resections.

Significance

In PMG‐related epilepsy, surgical strategy must balance seizure control with preservation of function. ICEEG (especially stereo‐EEG [SEEG]) remains a valuable tool for functional mapping and tailored resections.

Key points.

• Epilepsy surgery in polymicrogyria achieves seizure freedom in ~70% of patients at ≥12 months of follow‐up.

• Limited resections may preserve function with comparable seizure outcomes to hemispheric surgery.

• Intracranial electroencephalography (ICEEG)/stereo‐EEG (SEEG) enables tailored resections (more unilobar/non‐hemispheric) and may reduce functional morbidity.

1. INTRODUCTION

Polymicrogyria (PMG) is among the most common malformations of cortical development (MCDs), and results from abnormal neuronal migration during fetal brain development. ¹ It is characterized by an excessive number of small, disorganized gyri intermixed with shallow sulci, leading to a thickened and irregular cortical surface. ² Histopathologically, PMG is defined by abnormal cortical lamination, which may be unlayered or, in some cases, show a simplified four‐layered structure. ³ The extent and distribution of PMG are highly variable with unilateral or bilateral and hemispheric or sublobar patterns; however, the perisylvian region is most commonly involved. ⁴ PMG etiology is heterogenous and may remain unclear in many cases. Both genetic forms (syndromic or not) and acquired factors (such as ischemic insults early in pregnancy or prenatal infections) have been reported. ⁵

From a clinical perspective, PMG is associated with a broad phenotypic spectrum, including intellectual disability, language impairment, and sensorimotor dysfunction. ⁴ Epilepsy is one of the most common clinical features, typically with onset in early childhood. ⁴ Notably, subjects frequently develop an early drug‐resistant epilepsy (DRE), which complicates seizure management and strongly affects both neurodevelopmental and cognitive trajectories. ⁴ , ⁶ , ⁷

Moreover, children with PMG have an increased risk of developing Developmental and epileptic encephalopathy with spike–wave activation in sleep (D/EE‐SWAS), a severe electroclinical syndrome with impact on neurocognitive development. ⁸ , ⁹ , ¹⁰ Antiseizure medications (ASMs), including benzodiazepines and corticosteroids, can be effective, particularly in non‐structural etiologies within the D/EE‐SWAS spectrum. In PMG‐associated D/EE‐SWAS, corticosteroids may provide a prompt electroclinical benefit, yet relapses after tapering/withdrawal have been reported, and the age‐limited natural history makes sustained treatment effects difficult to interpret. ¹¹ , ¹² Accordingly, although pharmacologic management remains first line, epilepsy surgery may be considered in a selected subset of children with PMG and DRE, with decisions individualized to anatomy, electroclinical data, and functional risk. ¹³ , ¹⁴

In this study, we conducted a systematic review and meta‐analysis of the available literature to assess and stratify the risks and benefits of the various surgical strategies and approaches employed in patients with PMG. Our analysis focuses on the relationship between clinical outcomes after surgery and key neurophysiological, epileptological, anatomic, and functional factors relevant to surgical decision‐making.

2. METHODS

Literature review was conducted according to the Cochrane Handbook for Systematic Reviews of Interventions and presented based on PRISMA (Preferred Reporting Items for Systematic reviews and Meta‐Analyses) guidance. ¹⁵

2.1. Search strategy

To identify relevant studies, we systematically searched PubMed/Medline, Scopus, and EMBASE databases. The search strategy was carried out up to April 23, 2025. The search terms were as follows: (A) “Polymicrogyria” and (B) “epilepsy surgery”.

2.2. Inclusion and exclusion criteria

We included studies published in peer‐reviewed journals, regardless of language.

Inclusion criteria were as follows:

• A confirmed diagnosis of PMG based on magnetic resonance imaging (MRI) findings or histopathological examination;

• A clearly defined extent of PMG involvement (hemispheric, unilateral non‐hemispheric or bilateral) in patients who have undergone surgical treatment. A diagnosis of epilepsy related to PMG;

• A clearly reported postoperative seizure‐outcome either (i) explicitly reported using Engel classification, or (ii) inferable as Engel Class I vs non–Engel I from an unambiguous narrative description of seizure status at ≥12 months of follow‐up.

Exclusion criteria included:

• Studies not available in full text online;

• Review articles, letters to the editor, and conference abstracts.

To minimize attribution bias, we recorded whether cohorts included patients with double etiologies (polymicrogyria co‐occurring with another potentially epileptogenic malformation such as schizencephaly, periventricular heterotopia, or focal cortical dysplasia [FCD]). When individual‐level data were available, such cases were flagged at data extraction.

2.3. Study selection

Two independent reviewers (L.B. and C.V.) screened the titles and abstracts of all retrieved records. The objective was to identify all reported cases of epilepsy associated with PMG that underwent surgical treatment. To ensure comprehensive coverage, case reports were also included. In instances where multiple publications referred to the same patient cohort, preference was given to the study with the largest sample size.

Studies that were not accessible online, or in which critical data were unavailable or could not be extracted, were excluded. Manuscripts published in languages other than English were translated by two medically trained reviewers with language proficiency; disagreements were adjudicated by a senior reviewer. When helpful, we used machine‐assisted translation.

Study eligibility was assessed according to predefined inclusion and exclusion criteria. All extracted data were cross‐checked, and any discrepancies were resolved through consensus with a third reviewer (S.F.).

2.4. Data extraction and synthesis

Following the screening of 382 publications, 20 manuscripts met the eligibility criteria and were included in our review (Figure S1). For each patient, we extracted the following data: age at seizure onset, age at surgery, duration of epilepsy prior to surgery, sex, extent and distribution of PMG, intellectual functioning or cognitive status before and after surgery, neurological deficits before and after surgery, presence of EE‐SWAS, presence of mixed etiologies, type of surgical procedure performed (unilobar, multilobar, non‐hemispheric, or hemispheric resection), use of intracranial‐electroencephalography (ICEEG), cases in which radiofrequency thermocoagulation (rf‐THC) was performed via stereo‐electroencephalography (SEEG) without subsequent resective surgery and postoperative seizure outcome, assessed according to the Engel classification (with a minimum follow‐up of 12 months).

Across studies, we recorded any new or worsened postoperative neurological deficit that persisted at the last follow‐up (≥12 months) as a binary variable (present/absent). Early transient postoperative deficits that fully resolved by follow‐up were not counted. When available, domain‐level descriptors (e.g., motor, visual, sensory, language, bulbar/cranial‐nerve) were retained for descriptive reporting.

We classified the PMG topography based on hierarchical, mutually exclusive rules: we divided PMG into three groups: (1) hemispheric, (2) unilateral non‐hemispheric, and (3) bilateral. We also provided a subclassification of the unilateral non‐hemispheric lesions in: multilobar, perisylvian, and unilobar. Perisylvian was assigned only when the malformation was confined to the perisylvian network (opercular/insular/peri‐Rolandic regions) without extension beyond this territory. When available, MRI or schematic lesion maps dictated the assignment; otherwise, coding was based on the narrative clinical/imaging description. If reports suggested clear multilobar involvement without explicit perisylvian localization, cases were coded multilobar. When Engel classification was not explicitly reported, but sufficient clinical information was available, we inferred the class based on the described outcomes.

2.5. Outcomes

The primary objective of this study was to evaluate postoperative seizure outcomes, as measured by the Engel classification, in relation to the type of surgical procedure performed and the extent of PMG involvement.

We also defined several secondary objectives, including:

• Assessing the role of ICEEG (SEEG and/or subdural strip) in accurately delineating the epileptogenic zone (EZ) and in mapping eloquent cortical areas in patients with PMG.

• Determining the incidence and characteristics of postsurgical neurological deficits in relation to both the surgical approach and the extent of cortical malformation.

• Evaluating cognitive outcomes following surgery, with a particular focus on comparing results between hemispheric and non‐hemispheric resections.

2.6. Quality assessment

The quality of evidence for each included study was evaluated using a modified version of the Newcastle–Ottawa Scale (NOS), adapted specifically for the assessment of non‐comparative studies. ¹⁶ Each article was assessed based on a predefined set of five criteria, tailored to the objectives of our study. A binary scoring system was applied to each item: 1 (fulfilled) or 0 (not fulfilled).

The five criteria were as follows:

1. Was the diagnosis (PMG) accurate?

2. Did the included patients represent the entire cohort treated at the medical center during the study period (i.e., were all eligible patients managed within the defined timeframe included)?

3. Was the postoperative follow‐up period sufficient to allow for a reliable outcome evaluation (i.e., minimum follow‐up of ≥12 months)?

4. Was the clinical outcome after surgery appropriately defined, specifically using validated criteria or scales (e.g., Engel classification)?

5. Were all relevant clinical and demographic data adequately reported in the study?

2.7. Statistical analysis

We provided descriptive statistics for the entire cohort: continuous variables were expressed as mean, median, and range (min–max); whereas categorical variables were expressed as absolute or relative frequencies.

The Mann–Whitney U test was performed to conduct the non‐parametric analysis for continuous variables. The test was implemented to assess differences between groups, ensuring robust and reliable results in line with the objectives of this study. Fisher's exact test for categorical variables was used to measure differences between groups.

Seizure outcome data were estimated using the Engel classification. Relative epilepsy duration was computed as (epilepsy duration/age at surgery) × 100 (% of life). For exploratory univariable regression analyses (n = 69), this predictor was scaled per 10%‐of‐life increments; we report odds ratios (ORs) per +10% with 95% confidence intervals (CIs) (two‐sided α = 0.05). ¹⁷

We pooled all studies with five or fewer patients into a single study to avoid the loss of valuable information considering the rarity of the disease.

A proportional meta‐analysis was conducted to assess the proportion of patients in Engel Class I across multiple studies. For each measure of the effect, the 95% CIs were calculated.

Two models of analysis were applied: a fixed‐effects model and, when heterogeneity was substantial, a random‐effects model. For outcome measures, we assessed the statistical heterogeneity between data using the I ². We used the following cutoffs for heterogeneity: <25%, no heterogeneity; 25%–49%, low heterogeneity; 50%–74%, moderate heterogeneity; and ≥75%, high heterogeneity. I ² >50% was interpreted as significant heterogeneity. We found a low heterogeneity (I ² = 41.1%), so we combined studies results using the fixed‐effects model. ¹⁸ Forest plot was generated to visualize the individual study proportions. Statistical analyses were performed using Python (v3.10) with standard statistical packages and R (v4.5.0) with lme4 (v1.1–37) as specified below. As sensitivity analyses, we repeated the proportional meta‐analysis excluding very small series (n ≤ 5) and excluding subjects with double etiologies to address potential confounding of outcome attribution. Heterogeneity was assessed with Cochran's Q and I ² (Tables S1, S2; Figures S2, S3).

A subject‐wise heatmap was created to visualize individual outcomes (Engel class, motor deficits, type of surgery) across all included patients, grouped by study. Color coding was used to reflect presence/absence of each characteristic.

To identify predictors of surgical outcomes, we analyzed, using logistic regression (logit link), both the entire cohort and a sub‐cohort excluding PMG with hemispheric extension. In the full cohort, we fitted mixed‐effects logistic models with a random intercept to account for residual clustering. Outcomes were seizure freedom (Engel Class I vs non‐I; n = 160) and postoperative motor deficit (yes/no; n = 89). Predictors were surgical procedure (hemispheric vs non‐hemispheric), PMG extent (hemispheric vs unilateral non‐hemispheric or bilateral), and ICEEG use (yes/no); for the motor‐deficit model, Engel class was additionally included as a covariate. In the combined unilateral non‐hemispheric and bilateral PMG cohort, we used fixed‐effects logistic models. Outcomes were seizure freedom, postoperative motor deficit, and surgical extent (unilobar vs non‐unilobar). Predictors were PMG extent (unilobar, multilobar, perisylvian, bilateral), ICEEG use (yes/no), and surgical procedure (unilobar, multilobar, hemispheric). Results are reported as ORs with 95% CIs and two‐sided p‐values (α = 0.05).

3. RESULTS

3.1. Clinical and anatomic features

Across the 20 selected manuscripts, we identified a total of 161 subjects who met the inclusion criteria ¹³ , ¹⁴ , ¹⁹ , ²⁰ , ²¹ , ²² , ²³ , ²⁴ , ²⁵ , ²⁶ , ²⁷ , ²⁸ , ²⁹ , ³⁰ , ³¹ , ³² , ³³ , ³⁴ , ³⁵ , ³⁶ (Table S3).

Of the 96 patients for whom sex was reported, 61% (59/96) were male. Among those with available data, the mean age at seizure onset was 5.91 years (range: 0–32), the mean age at surgery was 12.91 years (range: 0.5–44), and the mean duration of epilepsy prior to surgery was 8.88 years (range: 0.2–42).

Thirty‐five subjects had hemispheric PMG, whereas 112 presented with unilateral non‐hemispheric PMG (29 unilobar, 43 multilobar, 40 perisylvian), and 14 had bilateral PMG. Among those for whom data were available, 67% (44/66) had cognitive impairment, 63% (40/64) had a motor deficit, and 42% (36/85) had a history of EE‐SWAS. Notably, hemispheric PMG was significantly more prevalent among EE‐SWAS patients compared to those without EE‐SWAS: 69% (25/36) vs 8% (4/49), p < 0.0005. In addition, EE‐SWAS was not significantly associated with age at seizure onset (p = 0.40). Patients with hemispheric PMG had earlier seizure onset compared to those with unilateral non‐hemispheric and bilateral PMG (mean 2.48 vs 7.59 years, p = 0.02). All the data are provided in Tables 1 and 2 and Figure 1A,B.

TABLE 1.

Clinical, neurophysiological and surgical characteristics of the entire patient cohort.

Total subjects selected	161
Male	59/96
Mean age at seizure onset, years	5.91 (0–32)
Median age at seizure onset, years	2.7
Mean age at surgery, years	12.91 (0.5–44)
Median age at surgery, years	8.9
Mean disease duration before surgery, years	8.88 (0.2–42)
Median disease duration before surgery, years	6.45
Mean relative epilepsy duration, % of life	68.75 (27.59–98.86)
PMG Unilobar	29
PMG Multilobar	43
PMG Perisylvian	40
PMG Bilateral	14
PMG Hemispheric	35
PMG Unilateral Non‐hemispheric	112
Cognitive impairment before surgery	44/66
Neurological deficit before surgery	42/65
Motor deficit before surgery	40/64
EE‐SWAS	36/85
ICEEG (SEEG or subdural strip) performed	85/160
Unilobar surgery	36
Multilobar surgery	39
Non‐hemispheric surgery (including SEEG rf‐THC)	105
Hemispheric surgery	56
SEEG‐rf‐THC without surgery	6
Engel Class I	114/161
Cognitive improvement after surgery	21/25
New post‐surgical neurological deficit	38/83
New or worsening of motor deficit post‐surgery	31/89

Note: Denominators differ across variables because percentages were calculated on available cases for each outcome (i.e., only subjects with explicit reporting for that specific variable). Abbreviations: EE‐SWAS, epileptic encephalopathy with spike‐wave activation in sleep; ICEEG, intracranial electroencephalography; PMG, polymicrogyria; rf‐THC, radiofrequency thermocoagulation; SEEG, stereoelectroencephalography.

TABLE 2.

Clinical correlates and prognostic indicators. ^Note:

Pre‐surgical correlates
Overall	Hemispheric PMG	p‐value (Fisher test or Mann–Whitney U test)
EE‐SWAS	69% (25/36)	<0.0005
Non‐EE‐SWAS	8% (4/49)
Overall	Mean age at seizure onset (years)
Hemispheric PMG	2.48	0.02
Non‐hemispheric PMG	7.59
Prognostic indicators
Overall	Relative disease duration before surgery (%)
Engel Class1	66%	0.048
Non‐Engel Class 1	79%
Unilateral non‐hemispheric PMG cohort	Engel class 1
EE‐SWAS	90% (9/10)	0.14
Non‐EE‐SWAS	63% (26/41)
Engel Class 1 Unilateral non‐hemispheric PMG cohort	Non‐hemispheric surgery
ICEEG recording performed	98% (48/49)	<0.0001
ICEEG recording not performed	41% (11/27)
Hemispheric surgery cohort	Engel class 1
Mean IQ pre‐surgery	46.91	0.09
Mean IQ post‐surgery	57.09

Note: Boldface indicates p < 0.05.

Abbreviations: EE‐SWAS, epileptic encephalopathy with spike‐wave activation in sleep; IQ, intelligence quotient; PMG, polymicrogyria. Boldface indicates p < 0.05.

FIGURE 1.

Box plots illustrating: (A) The distribution of age at seizure onset based on the anatomic extent of polymicrogyria; and (B) the distribution of age at seizure onset in relation to the occurrence of EE‐SWAS. The horizontal line represents the median, and the dot indicates the mean value. (C) The distribution of epilepsy duration in patients who did or did not achieve Engel Class I. (D) The distribution of epilepsy duration by type of surgical procedure. (E) The distribution of epilepsy duration according to the presence or worsening of postoperative motor deficits. Abbreviations: EESWAS, epileptic encephalophaty with spike‐wave activation in sleep; PMG, polymicrogyria.

3.2. Surgical approaches and prognostic correlates

Among the 161 patients included, 56 underwent anatomic or functional hemispherectomy, whereas 105 received non‐hemispheric procedures, including 36 unilobar and 39 multilobar resections, and 6 cases of SEEG‐guided rf‐THC not followed by open resection. ICEEG was employed in 53% of cases (85/160), mostly through SEEG, and was used almost exclusively in patients with unilateral non‐hemispheric or bilateral PMG (98%, 83/85) (Table S4).

At a minimum follow‐up of 12 months, 71% (114/161) of patients achieved seizure freedom (Engel Class I). Postoperative cognitive improvement was documented in 84% (21/25). Persistent (≥12 months) new or worsened motor deficits occurred in 31 of 89 (35%), whereas persistent neurological deficits (any domain) were reported in 38 of 83 (46%) (Table 1). The different denominators reflect variable outcome reporting across studies, with motor deficits more consistently reported than non‐motor neurological deficits. In patients with bilateral PMG, only 50% achieved seizure freedom. Across the entire cohort, patients classified in Engel Class I had a significantly shorter mean duration of epilepsy compared to non–Engel Class I patients at univariate analysis (8.05 vs 11.92 years, p = 0.009) (Figure 1C). Consistently, relative epilepsy duration (% of life) was associated with higher odds of non‐Engel Class I (n = 69; OR per +10% = 1.37; 95% CI 1.00–1.88; p = 0.048). Absolute duration was shorter in hemispheric vs non‐hemispheric cases (p < 0.0005) (Figure 1D), but this difference disappeared when using relative duration (percent of life with epilepsy, p = 0.998), indicating an age‐scaling effect. No statistically significant correlations were found between epilepsy duration and the likelihood of developing postoperative motor deficits (p = 0.44) (Figure 1E). Within hemispheric cases, relative duration was higher in non–Engel Class I vs Engel Class I (median 86.0% vs 70.4%; p = 0.036), but in univariable logistic regression (event = non–Engel Class I) the association did not reach statistical significance (n = 29; OR per +10% = 2.59; 95% CI 0.82–8.67; p = 0.103). Accordingly, in the combined unilateral non‐hemispheric and bilateral cohort, neither relative nor absolute epilepsy duration was significantly associated with worse seizure outcome.

In patients with unilateral non‐hemispheric PMG, Engel Class I outcome was achieved in 90% (9/10) of those with previous EE‐SWAS, compared to 63% (26/41) in the non‐EE‐SWAS group (p = 0.14); however, hemispheric surgery was performed more frequently in the EE‐SWAS subgroup than in those without (80% and 20%), indicating substantial confounding by surgical extent. In seizure‐free subjects of the unilateral non‐hemispheric extent subgroup, a prior ICEEG exploration correlated with a smaller resection (non‐hemispheric vs hemispheric procedure, 98%, 48/49) compared to those who did not (41%, 11/27; p < 0.0001). Although data on cognitive outcome were very limited, especially for non‐hemispheric procedures, patients undergoing hemispheric surgery who achieved Engel Class I showed a mean increase of 10.18 IQ points postoperatively (from 46.91 to 57.09), although this improvement did not reach statistical significance (p = 0.09). All data are provided in Table 2.

To complement these individual‐level findings, we performed a proportional meta‐analysis (Figure 2A) across the studies, yielding a pooled estimate of Engel Class I outcome of ~70% (95% CI: 62%–76%). Excluding case series with fewer than five patients, the pooled Engel Class I proportion was 69.7% (95% CI 58.8–78.8), and excluding double etiologies across cohorts, the pooled estimate was 72.0% (95% CI 59.7–81.7) (Figures S2, S3). These results are consistent with the 71% seizure‐freedom rate observed across the entire subject‐level dataset (114/161). Taken together, both the individual‐level analysis and the meta‐analytic aggregation confirm the overall effectiveness of epilepsy surgery in PMG.

FIGURE 2.

(A) Fixed‐effects Forest plot of clinical benefits (expressed as Engel Class I) of surgical treatment in polymicrogyria (PMG). Studies with fewer than five patients were grouped under “Others”. (B) Subject‐level heatmap displaying Engel outcome (green), surgical approach (blue), and presence of postoperative motor deficits (red) for each individual, grouped by study. (C) Multivariate mixed‐effects logistic regression model predicting Engel Class I outcome based on surgical strategy (hemispheric vs non‐hemispheric), ICEEG use, and PMG extension (hemispheric vs both unilateral non‐hemispheric and bilateral PMG). (D) Multivariate mixed‐effects logistic regression model predicting the occurrence of postoperative motor deficits based on Engel outcome, surgical strategy, ICEEG use, and PMG extension. Abbreviations: ICEEG, intracranial electroencephalography; OR; odds ratio; PMG, polymicrogyria.

In the full‐cohort (n = 160) multivariable analysis, modeling the odds of non–Engel Class I (vs Engel I) and using hemispheric PMG as the reference, bilateral PMG was associated with higher odds of an unfavorable seizure outcome (OR = 6.00; 95% CI 1.46–24.63; p = 0.013). Multilobar and perisylvian PMG did not reach statistical significance but showed higher odds of worse outcomes (multilobar: OR = 3.00; 95% CI 0.97–9.30; p = 0.057; perisylvian: OR = 2.77; 95% CI 0.86–8.91; p = 0.087), whereas unilobar PMG did not differ significantly (OR = 2.29; 95% CI 0.66–7.97; p = 0.194). Restricting the analysis to the combined unilateral non‐hemispheric and bilateral cluster and using unilobar PMG as the reference, we found no significant differences in seizure freedom when comparing multilobar, perisylvian, or bilateral PMG with unilobar extent. Similarly, no statistically significant associations emerged between PMG extent and postoperative motor deficit in this sub‐cohort. In contrast, PMG extent was strongly related to surgical extent: compared with unilobar PMG, multilobar and perisylvian patterns were much less likely to result in an unilobar resection (multilobar: OR = 0.20; 95% CI 0.07–0.56; p = 0.002; perisylvian: OR = 0.02; 95% CI 0.01–0.18; p < 0.001), whereas the bilateral pattern showed a nonsignificant reduction (OR = 0.23; 95% CI 0.03–1.46; p = 0.12) (Table 3).

TABLE 3.

Association between PMG extent, surgical extent, and ICEEG use with clinical outcomes in the combined unilateral non‐hemispheric and bilateral PMG cohort (excluding patients undergoing only rf‐THC).

Unilateral non‐hemispheric and bilateral PMG cohort	Seizure freedom, OR (CI)	p‐ value	Motor deficit OR (CI)	p‐ value	Unilobar resection OR (CI)	p‐ value
PMG extent	N = 120		N = 56		N = 96
Multilobar vs Unilobar	0.81 (0.29–2.29)	0.70	2.24 (0.51–9.85)	0.29	0.20 (0.07–0.56)	0.002
Perisylvian vs Unilobar	0.80 (0.27–2.36)	0.68	5.50 (0.91–33.18)	0.06	0.02 (0.01–0.18)	<0.001
Bilateral vs Unilobar	0.44 (0.11–1.73)	0.24	1.83 (0.12–27.79)	0.66	0.23 (0.03–1.46)	0.12
Surgical extent	N = 96		N = 56
Unilobar vs Hemispheric	0.56 (0.18–7.56)	0.33	0.26 (0.06–1.13)	0.07	/	/
Multilobar vs Hemispheric	1.21 (0.36–4.06)	0.76	0.21 (0.06–0.79)	0.02	/	/
ICEEG recording	N = 119		N = 56		N = 95
Yes – no	1.14 (0.51–2.55)	0.76	0.41 (0.14–1.25)	0.11	3.61 (1.41–9.21)	0.007

Note: The table reports odds ratios (OR) with 95% confidence intervals (CI) and p‐values from eight distinct binomial logistic regression models, each fitted independently with a single outcome: I. PMG extent vs 1 Seizure Freedom, 2 Postoperative motor deficit, and 3 unilobar resection; II. Surgical extent vs 4 Seizure Freedom, and 5 Postoperative motor deficit; III. ICEEG vs 6 Seizure Freedom, 7 Postoperative motor deficit, and 8 unilobar resection. Boldface indicates p < 0.05.

Abbreviations: CI, confidence interval; ICEEG, intracranial electroencephalography; OR, odds ratio; PMG, Polymicrogyria; rf‐THC, radiofrequency thermocoagulation.

3.3. Impact of surgical extent on seizure freedom and motor outcomes

In the overall cohort, patients who underwent hemispheric surgery showed a higher crude rate of seizure freedom compared with sub‐hemispheric resections (absolute difference ≈ 23%; Fisher's exact test p = 0.003). However, in both unilateral non‐hemispheric and bilateral PMG sub‐cohort (excluding rf‐THC‐only cases) and using hemispheric surgery as the reference, no statistically significant differences were observed in the odds of achieving Engel Class I across surgical categories in the logistic models. Among the six patients treated exclusively with SEEG‐guided rf‐THC, three (50%) achieved seizure freedom.

In contrast to seizure outcomes, new or worsened postoperative motor deficits were more frequent after hemispheric resections. In the overall cohort, the crude rate of motor deficits was higher by ~23% in hemispheric vs non‐hemispheric surgeries (Fisher's exact test p = 0.044). This pattern was consistent in the logistic analysis (unilateral non‐hemispheric and bilateral cohort only): compared with unilobar surgery, hemispheric procedures showed a trend toward higher odds of motor deficit (OR 3.93; p = 0.07), and the association was significant when hemispheric surgery was compared with multilobar resections (OR 4.71; p = 0.02). Detailed estimates are reported in Figure 3A,B and Table 3. Moreover, a subject‐level heatmap (Figure 2B) was used to summarize Engel outcomes, extent of resection, and postoperative motor deficits. This visualization highlighted substantial variability in treatment patterns and outcomes, even within the same study. In the mixed‐effects logistic regression model of the full dataset (Figure 2C) evaluating seizure outcome, hemispheric surgery showed higher odds of Engel Class I, but the association did not reach statistical significance (OR 3.52; 95% CI 0.97–12.79; p = 0.055). ICEEG use and PMG hemispheric extent were not significantly associated with seizure freedom. In the second model (Figure 2D) examining predictors of postoperative motor deficits, hemispheric surgery was associated with higher odds of postoperative motor impairment, but this association was not statistically significant (OR 6.55; 95% CI 0.79–54.06; p = 0.081). Engel class, ICEEG use, and PMG extent were not significantly associated with postoperative motor deficits.

FIGURE 3.

Relationship between polymicrogyria extent, surgical approach, ICEEG use, and clinical outcomes. Estimated marginal means (EMMs) and 95% confidence intervals (CIs) from binomial logistic regression models are shown for: (A) probability of Engel Class I outcome according to surgical procedure (unilobar, multilobar, hemispheric); (B) probability of postoperative motor deficit by surgical extent; (C) probability of Engel Class I outcome; (D) probability of postoperative motor deficit; and € probability of unilobar resection according to ICEEG use. Bars represent EMM ± 95% CI. Abbreviations: ICEEG, intracranial electroencephalography; PMG, polymicrogyria.

3.4. The role of ICEEG in surgical decision‐making and functional prognosis

In the overall cohort, patients without ICEEG showed an ~15% higher crude rate of seizure freedom (Fisher's exact test p = 0.039). This difference is likely confounded by indication, given that patients undergoing direct hemispheric resections for hemispheric lesions (who less often required ICEEG) had favorable seizure outcomes. Consistently, in the combined unilateral non‐hemispheric and bilateral PMG cohort, ICEEG use was not associated with seizure freedom in the logistic model (OR 1.14; p = 0.76).

Conversely, in the overall cohort, patients without ICEEG had an ~17% higher crude rate of postoperative motor deficits (Fisher's exact test p = 0.109), which did not reach statistical significance. In the non‐hemispheric analysis, the absence of ICEEG was associated with higher, but non‐significant, odds of motor deficits (OR 2.44; p = 0.11; reference = ICEEG).

Finally, in both the overall and the combined unilateral non‐hemispheric and bilateral PMG cohorts, patients who underwent ICEEG were significantly more likely to receive unilobar resections than those who did not (overall: Fisher's exact test, p < 0.005; unilateral non‐hemispheric and bilateral cohort: OR 3.61, p = 0.007). All data are provided in Figure 3C–E and Table 3.

4. DISCUSSION

The management of epilepsy in patients with PMG remains among the most complex therapeutic challenges in clinical practice. ¹⁹ , ³⁷ Surgical intervention has emerged as a potentially effective strategy for achieving seizure freedom in this population. ¹³ However, there are currently no established guidelines to support clinicians and surgeons in determining the optimal timing for surgery or the most appropriate approach based on a comprehensive risk–benefit evaluation. Although several studies suggest that extensive procedures, such as anatomic or functional hemispherectomy, can provide long‐term seizure control, ²⁵ , ²⁷ these interventions are often accompanied by significant functional consequences that must be carefully considered when evaluating surgical candidacy. ³⁸

In this review, we aimed to synthesize and compare the current knowledge and most commonly adopted surgical approaches, considering key disease‐related features, including anatomic extent, cognitive impairment, presence of EE‐SWAS, and pre‐existing neurological deficits.

With respect to the primary outcome, we found that ~70% of patients achieved Engel Class I at ≥12 months. Although both higher absolute and relative (% of life) epilepsy duration showed associations with outcome in entire cohort (univariate analyses), when clustering the groups in hemispheric PMG or non‐hemispheric and bilateral PMG, these signals did not remain significant: in a hemispheric‐only sensitivity analysis, relative duration was higher among non–Engel Class I than Engel Class I patients (p = 0.036), yet the corresponding logistic estimate was borderline and imprecise (n = 29; OR per +10% = 2.59; p = 0.103), underscoring limited power rather than a robust duration effect. Accordingly, our data do not establish a causal link between pre‐surgical duration and seizure freedom.

Consistent with current clinical practice, ⁴ , ¹³ , ³⁹ we suggest early surgical evaluation in well‐lateralized hemispheric PMG, particularly when preoperative deficits are present, while emphasizing that prospective, adequately powered studies, stratified by PMG‐related syndromes (including D/EE‐SWAS) and using standardized neurocognitive measures, are needed to clarify whether timing independently modifies outcomes.

In addition, we found no statistically robust evidence that EE‐SWAS per se modifies postsurgical seizure outcomes; therefore, EE‐SWAS should not be used as an exclusion criterion for surgical candidacy. Across the D/EE‐SWAS spectrum, benzodiazepines, ASMs, and corticosteroids can lead to electroclinical improvement, and pharmacological therapy remains first‐line. Recent cohort data suggest that self‐limited focal epilepsies tend to show shorter SWAS duration and more favorable outcomes than structural etiologies overall ¹² ; however, extrapolation to specific malformations such as PMG is limited by small numbers. ¹² Even in PMG‐associated SWAS, corticosteroids may provide a prompt electroclinical benefit, yet drug resistance and relapses after tapering/withdrawal have been reported. ¹¹ Of note, available long‐term follow‐up data suggest a progressive attenuation of SWAS activity over time, with remission typically occurring before adolescence, which complicates attribution of long‐term outcomes to a given treatment. ¹⁰ , ¹¹ Moreover, neuropsychological outcome may be substantially worsened by years of SWAS over what expected by the malformation per se, making the specific contribution of SWAS difficult to disentangle. ¹¹ In this context, surgery should not be viewed as a substitute for medical therapy for SWAS; rather, in selected PMG patients with drug‐resistant epilepsy, early surgical evaluation remains appropriate including those with D/EE‐SWAS, when a well‐lateralized epileptogenic network is identified after adequate medical trials, with decisions individualized to anatomy, electroclinical data, and cognitive/functional risk, rather than based on EE‐SWAS status alone. ⁴⁰

In the overall cohort, hemispheric surgery showed a higher crude rate of seizure freedom than sub‐hemispheric resections (Fisher's exact test, p = 0.003), but it also carried a higher crude rate of new postoperative motor deficits (~23%; p = 0.044). However, in the multivariable (mixed‐effects) models for the full cohort, surgical extent was not independently associated with either seizure freedom or motor deficit. In the combined unilateral non‐hemispheric and bilateral PMG cohort, hemispheric surgery was associated with a greater odds of postoperative motor deficit compared with multilobar resections (OR 4.71; p = 0.02) and showed a non‐significant trend vs unilobar resections (OR 3.93; p = 0.07). By contrast, seizure freedom did not differ by surgical extent in the combined unilateral non‐hemispheric and bilateral models (unilobar vs hemispheric: OR 0.56, p = 0.33; multilobar vs hemispheric: OR 1.21, p = 0.76).

This attenuation suggests that the apparent advantage of hemispherectomy may be partly explained by confounding factors, such as the shorter epilepsy duration typically observed in these patients. Accordingly, and despite ~20% of the combined unilateral non‐hemispheric and bilateral PMG cases undergoing hemispherectomy in our series, this strategy should not be considered the default in non‐hemispheric PMG. Rather, surgical decisions should be individualized, carefully weighing seizure control against cognitive development and the risk of postoperative neurological deficits, two of the secondary outcomes assessed in our study. Cognitive data were sparse and did not allow a robust multivariable comparison between hemispheric and non‐hemispheric surgeries. In an exploratory univariate analysis limited to patients who underwent hemispherectomy and achieved seizure freedom, we observed an average increase of ~10 IQ points; however, this difference did not reach statistical significance and should be interpreted with caution given the small sample size. Overall, our dataset does not support firm conclusions regarding the impact of hemispherectomy on cognitive trajectory in PMG. Furthermore, we also explored the role of ICEEG in surgical decision‐making. Although most patients underwent SEEG, a few were evaluated using subdural strip electrodes. Of interest, in the overall cohort, patients who did not undergo ICEEG achieved better seizure outcomes. However, this likely reflects the inclusion of patients with hemispheric lesions who went directly to hemispherectomy. In fact, considering only patients with non‐hemispheric or bilateral PMG, no statistically significant difference was observed in seizure outcomes based on ICEEG performance. Notably, among seizure‐free patients with unilateral non‐hemispheric PMG, 98% of those who underwent ICEEG received a non‐hemispheric surgery compared with 41% of those without ICEEG (p < 0.001). Although statistical significance was not achieved, the logistic model, applied to the combined unilateral non‐hemispheric and bilateral cohort, suggested a directional association indicating that the use of ICEEG may contribute to reducing the incidence of postoperative motor deficits. Indeed, by enabling precise mapping of eloquent cortical areas and more accurate delineation of the epileptogenic zone, ICEEG appears to support more tailored resections that could minimize functional disruption. We therefore view ICEEG/SEEG chiefly as an enabling tool for tailored, function‐sparing resections, with a possible reduction in postoperative motor morbidity and no signal of advantage or disadvantage on seizure freedom in this cohort. This balance (comparable seizure outcomes with fewer deficits and smaller resection) depends on rigorous anatomo‐electro‐clinical concordance and its careful interpretation.

Finally, based on integration of our findings with current clinical practice, we propose a conceptual decision‐making framework for the surgical management of drug‐resistant epilepsy in patients with PMG (Figure 4). Two distinctive features of PMG make such a tailored approach necessary: functional preservation can be observed even within structurally abnormal cortex, and not all the PMG areas are necessarily epileptogenic. For hemispheric PMG, early surgical consideration, typically involving hemispherectomy or hemispherotomy, should be considered in the presence of hemispheric PMG extent and non‐unifocal electroclinical features, particularly when developmental regression is significant or in presence of a major functional deficit. In unilobar PMG, where the lesion is clearly delineated, concordant with seizure onset, and located outside eloquent cortex, resection without invasive monitoring may be appropriate. In contrast, multilobar, bilateral, or perisylvian PMG often require SEEG to define the EZ and to map the functional regions. This algorithm is not intended as a prescriptive guideline or a validated tool but as a conceptual framework to support individualized surgical planning, especially in complex presentations. It also underscores the importance of considering functional outcomes, particularly the risk of new motor deficits, when tailoring surgical strategies. This study has several limitations. Given the rarity of PMG, we included small case series and case reports to maximize the number of analyzed cases. In addition, heterogeneity across retrospective series and case reports may have influenced outcome reporting, and cognitive data were too sparse to allow robust conclusions. Moreover, even the “non‐hemispheric” category remains heterogeneous in both topography and extent (e.g., unilobar, multilobar or perisylvian); future prospective registries should, therefore, capture more granular, ideally continuous, measures of lesion extent rather than relying solely on categorical bins. Some cohorts included patients with additional malformations (e.g., schizencephaly, heterotopia, FCD), which could confound outcome attribution. However, our sensitivity analysis excluding double etiologies produced pooled estimates almost identical to the primary analysis, thereby mitigating this concern. All studies were retrospective in nature, and variables such as the surgeon's experience, institutional volume, and technical variations in the procedure may have influenced outcomes. Such variability may affect completeness of resection, the ability to spare eloquent cortex, peri‐operative morbidity, and, ultimately, seizure freedom. Although our pooled estimate was robust in sensitivity analyses, we cannot disentangle the contribution of operator/center effects or specific technical steps (e.g., extent of disconnection or intraoperative monitoring) because these data were incompletely and inconsistently reported. This limitation underscores the need for standardized reporting of procedural details and completeness of resection in future series. This observation, however, reinforces our conclusions: even though hemispherectomy should theoretically provide near‐universal seizure control in PMG, factors, such as neurophysiological and surgical expertise, can influence outcomes, narrowing the gap between Engel outcomes for hemispheric vs sub‐hemispheric procedures. Finally, the proposed algorithm is a conceptual framework derived from aggregated data rather than a validated protocol and should be adapted to individual patients.

FIGURE 4.

Surgical decision‐making algorithm for epilepsy surgery in PMG, based on lesion extent, seizure features, and functional considerations. Developed from the current subject‐level meta‐analytic findings. Abbreviations: EZ, epileptogenic zone, PMG, polymicrogyria; SEEG, stereoelectroencephalography.

5. CONCLUSION

Our findings challenge the assumption that hemispherectomy is superior to more limited resections in patients with unilateral non‐hemispheric PMG. Although seizure outcomes were comparable, hemispheric surgery carried a significantly higher risk of motor deficits. ICEEG appeared valuable in guiding tailored and safer resections and detecting the EZ. Tailored surgical strategies, based on lesion extent, ICEEG data, and preoperative function, are crucial to optimize both seizure control and neurological outcomes in PMG‐related epilepsy.

AUTHOR CONTRIBUTIONS

Luca Bosisio: conception and design of the study, literature review and article selection, manuscript writing, and realization of figures and tables. Carolina Viglietti: literature review and article selection and manuscript writing. Alessandro Consales: review, editing, validation, and supervision. Giulia Nobile: participation in writing of the manuscript. Mattia Pacetti: participation in writing of the manuscript. Maria Margherita Mancardi: review, editing, validation, and supervision. Domenico Tortora: validation and supervision. Lino Nobili: review, editing, validation, and supervision. Stefano Francione: conception of the study, review, editing, validation, and supervision. All authors contributed to the article and approved the submitted version.

FUNDING INFORMATION

This research did not receive any specific grant from funding agencies in the public, commercial, or not‐for‐profit sectors.

CONFLICT OF INTEREST STATEMENT

None of the authors has any conflict of interest to disclose.

ETHICS STATEMENT

We confirm that we have read the Journal's position on issues involved in ethical publication and affirm that this report is consistent with those guidelines.

Supporting information

Figure S1:

Figure S2:

Figure S3:

Table S1:

Table S2:

Table S3:

Table S4:

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the corresponding author upon reasonable request.