Randomized controlled trial of 2.5-cm versus 3.5-cm mesial temporal resection—part 2: volumetric resection extent and subgroup analyses

Abstract

This paper is addressing outcome differences in interesting subgroups from a previous randomized controlled trial of the extent of mesial temporal lobe resection (TLR) for drug-resistant epilepsy, by looking at effects of randomization, intended resection group, center, and true resection extent on seizure outcome. One hundred and seventy-nine cases with volumetrically assessed resection extent were used. Analyses of the extent of resection and subgroups and within subgroups for the two treatment arms will be performed, looking for confounding factors and using statistical methods (chi-square test, logistic regression analysis, and two-factorial ANOVA). True resection extent varied considerably. Outcome comparison for right versus left resections, subgroups with mesial temporal sclerosis (MTS), or largest and smallest resections revealed no remarkable difference, compared to overall class I outcome. The intent-to-treat analyses within these subgroups revealed differences for class I outcome, albeit lacking in significance, except for better TLR outcome. Small true resection volume differences or randomization into the two resection groups could not explain the outcome differences between the selective amygdalohippocampectomy (SAH) and TLR subgroups. Logistic regression analysis showed an interaction between intended resection length and surgery type, confirming the impression of different impacts of the intended resection length under the two surgery types. The outcome difference between SAH and TLR was more likely explained by a center effect. In a two-factorial ANOVA for resected hippocampal volume, Engel outcome class I, and resection type, the outcome was not found to be correlated with true resection volume. A multifactorial logistic regression showed a mild interaction between the resection type with center on the Engel outcome class, extent of resection, and surgery type interacted, as did the extent of resection and center. Patients with quite similar extent of resection can be seizure free or non-seizure free. In this cohort, seizure freedom rates fell again when the extent of mesial resection was maximized. Differences in class I outcome for SAH and TLR were not due to erroneous randomization, true resection extent, or presence of MTS, but were influenced by a center effect. Subgroup analyses did not help to provide arguments to favor one surgery type over the other.