Abstracts

APPLICATION OF SIGNAL DETECTION THEORY TO RECOGNITION MEMORY TESTING FOR THE DIFFERENTIAL DIAGNOSIS OF PSYCHOGENIC NONEPILEPTIC AND EPILEPTIC SEIZURES

Abstract number : 1.158
Submission category :
Year : 2005
Submission ID : 5210
Source : www.aesnet.org
Presentation date : 12/3/2005 12:00:00 AM
Published date : Dec 2, 2005, 06:00 AM

Authors :
1Kelly A. McNally, 1Steven Howe, 1Bruce K. Schefft, and 2Michael D. Privitera

Neuropsychological test results are often used in the differential diagnosis of epileptic seizures (ES) and psychogenic non-epileptic seizures (PNES). However, since deficits have been reported in both groups, simply investigating overall performance may not be sufficient. It may be important for neuropsychologists to shift their attention to the qualitative differences in test responses between patients with PNES and ES. Signal Detection Theory (SDT) can be used to investigate differential response patterns on recognition memory tasks. SDT posits that performance is dependent on two independent factors: 1) ability to discriminate between targets and distractors (sensitivity, or actual memory ability) and 2) propensity to respond in a particular manner (response bias). One hundred patients diagnosed with PNES and 108 patients with temporal lobe ES, 62 left temporal (LTLE) and 46 right temporal (RTLE), were studied. Each participant was administered the Wechsler Memory Scale (WMS-III). Scores on the recognition subtest of the List Learning task were analyzed with the traditional method of scoring and also subjected to SDT analytic decomposition into indices of sensitivity (d[apos]) and bias (c). Logistic regression analysis was conducted to investigate the diagnostic utility of raw scores and SDT measures. Raw scores on the recognition memory task were found to significantly discriminate between ES and PNES ([chi]2 = 6.01 p = .014, Area under ROC curve, or AUC = .57) however, analysis of SDT measures slightly improved the diagnostic utility ([chi]2 = 16.95, p [lt] .001, AUC = .62). Both sensitivity (Wald [chi]2 = 8.43, p =.004) and bias (Wald [chi]2 = 7.92, p =.005) significantly contributed to the discrimination of ES and PNES. Increased sensitivity and a more negative response bias increased the odds of PNES (odds ratio, OR, sensitivity = 1.84, bias = 0.09). Decomposition into SDT measures did not significantly improve the discrimination between LTLE and RTLE (raw scores [chi]2 = 6.63, p = .01., AUC = .66; SDT measures [chi]2 = 14.42, p [lt].001, AUC = .68). Sensitivity (Wald [chi]2 = 8.17, p = .004) but not response bias (Wald [chi]2 = 2.9, p = .09) was found to be a significant predictor of left versus right TLE, with decreased sensitivity increasing the odds of LTLE (OR = 0.45). This study provides evidence that SDT measures of sensitivity and response bias may be useful tools to aid in the differential diagnosis of ES and PNES. Patients with PNES were found to have a negative response bias as compared to patients with ES. The negative response bias in the PNES group may be related to depression, inadequate effort or low motivation on the diagnostic testing. These results suggest that, not only is overall performance on neuropsychological testing important, but differences in response patterns should also be considered.