Duke Epilepsy · Retrospective outcome analysis

Surgical Decision Support for Epilepsy Outcomes

I led the statistical analysis of whether the 5-SENSE focality score is associated with seizure freedom after resection or laser interstitial thermal therapy. The results support interpreting the score as one component of a broader clinical assessment, not as a stand-alone threshold rule.

BiostatisticsEpilepsy surgeryModel evaluationSensitivity analysis

Clinical question

How much prognostic information does a focality score add for postoperative seizure freedom?

The 5-SENSE score summarizes focality from stereoelectroencephalography and related clinical evidence. We evaluated whether the continuous score was associated with Engel IA seizure freedom after resection or LITT and examined why some patients with high scores did not achieve the expected outcome.

I led cohort reconciliation, statistical analysis, sensitivity checks, and interpretation. Neuromodulation was analyzed separately because its responder endpoint is not equivalent to postoperative seizure freedom.

110Treated patients

→

88Resection or LITT

→

42Engel IA

46Not Engel IA

Distribution of 5-SENSE scores by Engel IA seizure-freedom outcome — **5-SENSE score by outcome.** Scores were higher among seizure-free patients in the primary surgery/LITT cohort. The continuous comparison supports an association, but the overlap between groups argues against interpreting the score in isolation.

ROC curve for the continuous 5-SENSE score — **Continuous-score discrimination.** The AUC was 0.71, with a bootstrap 95% confidence interval of 0.60 to 0.81. This indicates moderate, not definitive, discrimination.

AUC0.71Bootstrap 95% CI: 0.60 to 0.81

Per 10-point increase1.35×Higher odds of Engel IA

Threshold near 4195.2%Sensitivity

Threshold near 4130.4%Specificity

Threshold interpretation

The transferred threshold retained high sensitivity but generated many false positives.

The threshold near 41 came from prior work on SEEG focality, not from validation against postoperative seizure freedom. In this cohort, it retained nearly all favorable outcomes, but its low specificity means that it should not be treated as a validated surgical cutoff.

The continuous score carries more information than a single transferred threshold. Its practical value is likely to depend on pathology, imaging concordance, treatment context, and other clinical evidence.

Failure analysis

Histology was associated with divergent outcomes among high-score cases.

Among eligible high-score surgery cases, positive histology was present in 81.8% of true-positive cases and 42.9% of false-positive cases. The exact odds ratio was 5.77. This was the largest explanatory difference in the updated analysis and remained evident in an exploratory model adjusted for SEEG.

The result supports interpreting the score together with pathology. It does not establish histology as a causal modifier, and the subset size limits precision.

Histology subtype counts among high-score true-positive and false-positive surgery cases — **Histology among high-score surgery cases.** Curated pathology showed a higher proportion of lesional findings among true-positive cases and more negative or no-lesion findings among false-positive cases.

Takeaway

The score provides prognostic information, but not a stand-alone decision rule.

The continuous 5-SENSE score was associated with Engel IA and showed moderate discrimination. Its interpretation is strongest when combined with pathology and other clinical evidence. Follow-up uncertainty, transfer of the proposed threshold, and small treatment-specific subsets remain important limitations.

View Duke epilepsy role