Diagnostic utility of Wada Memory Asymmetries: Sensitivity, specificity, and likelihood ratio characterization.

Abstract

The authors used logistic regression, dichotomous and multiple level likelihood ratios, and receiver operating characteristic (ROC) analyses to examine Wada Memory Asymmetries (WMAs) in 324 patients who subsequently underwent temporal lobe (TL) surgery (left TL surgery = 172; right TL surgery = 152) using the Medical College of Georgia Wada protocol. Logistic regression correctly classified 84% of left TL patients and 77% of right TL patients using WMA. Corresponding dichotomous likelihood ratios (LRs) were: LR+ = 3.64; LR- = 0.21. The area under the curve using ROC was similarly highly significant (.886, standard error = .018, p < .001). When classifying patients using multiple level LRs, 40 left TL patients (23.3%) obtained asymmetry scores greater than +4, whereas no right TL patients obtained asymmetry scores in this range. No left TL patients obtained a WMA of -8 or less, although 12 right TL patients (7.9%) obtained a difference score of -8. Multiple level LRs indicate impressive diagnostic sensitivity for certain WMA ranges, greatly increasing the probability of undergoing either left or right TL surgery depending on WMA magnitude.