E L S E V I E R Epilepsy Research 25 (1996) 243-248

EPILEPSY RESEARCH

Memory performance on the intracarotid amobarbital procedure as a predicator of seizure focus

Deborah D. Roman a,*, Thomas E. Beniak b, Sean Nugent c

a Departments of Physical Medicine and Rehabilitation, Neurosurgery, and Psychiatry, UniL,ersity of Minnesota, Box 390, 420 Delaware St. S.E., and MINCEP Epilepsy Care, Minneapolis, MN 55455, USA

b Department of Neurosurgery, University of Minnesota and MINCEP Epilepsy Care, 5775 Wayzata Bh, d., Minneapolis, MN 55416, USA c Psychiatry Service (l16A), VA Medical Center, 1 Veterans Drive, Minneapolis, MN 55417, USA

Received 5 October 1995; revised 23 May 1996; accepted 4 June 1996

Abstract

The intracarotid amobarbital procedure (IAP) was performed on 56 patients with intractable complex-partial epilepsy who were candidates for temporal lobectomy. Seizure focus was lateralized to one hemisphere, as determined by surface EEG recordings and MRI evidence of temporal lobe disease. IAP memory items were presented following injection of 125 mg of sodium amytal into the internal carotid artery. Verbal, Nonverbal, Design, Pictorial, and Total memory scores were calculated based on recall/recognition of memory stimuli following drug recovery. Poorer memory was observed in the hemisphere ipsilateral to seizure focus on all memory scores. The Total Memory Score was the best memory measure, correctly classifying the largest number of patients. Using optimal cut-off scores on this measure, 75% of the patients with left hemisphere seizure focus and 79% of the patients with right seizure focus were correctly classified. There was a definite tendency for the dominant hemisphere to outperform the non-dominant. This must be taken into account in arriving at optimal cut-off points.

Keywords: Wada; Intracarotid amobarbital procedure; Neuropsychological assessment; Epilepsy

1. Introduct ion

The intracarotid amobarbital procedure (IAP) has gained acceptance as a means of assessing temporal lobe functional integrity. Testing conducted during the IAP is used to make inferences about the relative contribution of each hemisphere to memory function- ing. If memory is not well mediated by the hemi- sphere contralateral to intended surgery, the patient is considered to be at risk for post-surgical signifi-

* Corresponding author.

cant memory decrement or amnesia. Also, since h ippocampal dysfunct ion and complex-part ial epilepsy are closely associated, this procedure may be useful in lateralizing seizure focus. As such, the IAP has become an important part of the presurgical work-up [6,10].

Despite its widespread use, the validity of the IAP in lateralizing memory functions has been questioned on several grounds. First, the brevity of the proce- dure limits the scope of memory assessment and may result in spurious, non-replicable outcomes. Second, excessive sedation, neglect, visual field cuts, adverse

0920-1211/96/$15.00 Copyright © 1996 Elsevier Science B.V. All rights reserved. Pll S0920-1211(96)00064-2

244 D.D. Roman et a l . /Ep i l ep sy Research 25 (1996) 243 248

emotional responses, and aphasia may interfere with memory assessment. Third, the hippocampus is pri- marily served by the posterior cerebral artery and would not ordinarily be fully perfused by an intrac- arotid amobarbital injection. If the hippocampus is not rendered dysfunctional during the lAP, infer- ences about the relative contribution of each hemi- sphere to memory functions are not possible.

A number of studies have investigated the utility of the lAP in assessing memory and determining seizure focus. Christianson et al. [2] assessed lAP memory functioning in 12 patients with left temporal lobe lesions, 8 patients with right temporal lesions, and 7 patients with left frontal lesions. Lesion local- ization was based on clinical evidence using surface and invasive EEG recordings. Memory stimuli in- cluded concrete and abstract nouns, pictured objects, abstract designs, and faces. Word stimuli, either concrete or abstract, did not effectively classify the left and right temporal lobe epilepsy patients, with both patient groups performing better when the right (non-dominant) hemisphere was injected. In contrast, recall of the pictured common objects was better when the hemisphere ipsilateral to seizure focus was injected. A similar but less robust finding was ob- served using geometric designs as the memory stim- uli. The groups did not differ significantly on facial recognition.

Perrine et al. [9] examined IAP memory function- ing in 57 patients with temporal lobe complex-partial epilepsy. Eight memory items, including verbal and non-verbal stimuli, were used. Left versus right hemisphere memory performances were compared and 83% of the patients were correctly classified (with poorer performance in the hemisphere ipsilat- eral to the seizure focus).

In contrast, Risse et al. [11] found dominant (left) hemisphere superiority for verbal material, regardless of lesion lateralization. When memory performance of "non-les ioned" hemispheres was compared, memory for words and abstract designs was substan- tially better mediated by the dominant hemisphere. The authors concluded that there was a tendency for the dominant hemisphere to outperform the non- dominant, regardless of the type of memory stimulus used.

The current study examines the efficacy of vari- ous IAP memory items in determining seizure later-

alization in a group of patients with complex-partial epilepsy.

2. Method

Subjects were 56 patients (27 males, 29 females) with intractable epilepsy who were candidates for temporal lobectomy. The mean age for the group was 32 (range: 15-52). Seizure semiology, surface EEG recordings and MRI abnormalities suggested unilateral seizure focus: 28 individuals had a left hemisphere focus and 28 had a right hemisphere focus. All subjects were left hemisphere dominant for speech and language, as determined by the IAP. The groups were comparable in age, gender compo- sition, IQ, age at seizure onset, and seizure disorder duration. Of the 18 left temporal focus patients who had MRI evidence of focal left temporal disease, 12 had temporal atrophy or sclerosis, 2 had gliomas, 2 had other local lesions, 1 had an arteriovenous mal- formation (AVM), and 1 had an internal capsule infarct. Of the 18 right temporal focus patients, 13 had evidence of temporal atrophy/sclerosis, 1 had a cavernous angioma, 1 had encephalomalacia, 1 had a questionable AVM, 1 had a tumor, and 1 had a questionable hemorrhage. Table 1 summarizes the characteristics of the two groups.

Baseline EEGs were obtained and baseline grip strength was assessed for both hands using a dy- namometer. Test items were presented following in- jection of 125 mg of sodium amytal, delivered manu- ally in a single bolus into the internal carotid artery via a transfemoral approach. Both hemispheres were assessed on the same day. The order of the injections was randomized.

The psychometric battery was designed to assess language and verbal and non-verbal memory. Verbal items were selected with the aim of minimizing the subject's ability to encode the items non-verbally. Accordingly, some verbal items were abstract words which would be difficult to visualize. Non-verbal items were selected with the aim of minimizing the subject's ability to successfully remember the items using verbal encoding. For the pictured objects, the /'oils used for the recognition phase of the test were from the same semantic class as the target items (e.g., 4 fish). For the abstract designs, the 3 foils contained similar elements, making it difficult to

D.D. Roman et a l . / Epilepsy Research 25 (1996) 243-248 245

Table 1 Descriptive characteristics of left and right temporal lobe epilepsy patients

Left TL group Right TL group

Age Mean (S.D.) 31 (9.19) 32 (8.86) Median 30 31.5 Range 16-57 15-54

Gender Male/Female 12 / 16 15 / 13

1Q Mean (S.D.) 96.5 (12.62) 101 (10.61) Median 95.5 100 Range 77-130 84-121

Seizure duration in years Mean (S.D.) 20.6(11.7) 21.1 (11.2) Median 20.5 20 Range 3-54 < 1-47

Age of seizure onset Mean (S.D.) 10.2 (7.55) 11.2 (8.89) Median 8 11 Range 1-27 < 1-37

Number with focal MRl lesions 18/28 18/28

distinguish the targets from the foils using verbal means (both the target stimuli and the foils consisted of similar geometric elements).

Memory items were 4 printed words (2 concrete nouns, 2 abstract nouns), a nursery rhyme phrase, 4 geometric figures, and 3 line-drawings of common objects. Geometric figures were symmetrical to mini- mize the adverse effects of hemispatial neglect. All memory items were presented regardless of whether the patient was aphasic. The words were presented visually and orally. The first memory item was presented within approximately 15 s of the injection. Exposure time was approximately 10 s per item. Additional language items included 4 objects to be named, word and phrase repetitions, and 5 oral com- mands using token test stimuli.

Following return of EEG and contralateral grip strength to baseline levels, memory was assessed. Free recall was attempted for most items. Recogni- tion was assessed by presenting the target stimuli among an array of 3 foils. For the 4 words, foils

included semantically or phonemically similar words of comparable complexity [13]. The target nursery rhyme was presented among 3 other nursery rhymes. The pictured object foils included items from the same semantic class. The abstract design foils were also symmetric designs similar to the target items.

The protocol yields 3 memory scores: a Verbal Memory score (word and nursery rhyme recall/re- cognition), Non-verbal Memory score (recognition of designs and pictured objects), and Total Memory score (a composite of verbal and non-verbal memory scores). In addition, for the purpose of this study, non-verbal memory was further divided into memory for geometric designs and memory for pictured ob- jects.

Receiver operating characteristic (ROC) analysis was employed to determine the extent to which IAP scores (Verbal, Non-verbal, Design, Pictorial, and Total Memory scores) lateralize seizure focus in this sample. ROC analysis evolved from decision theory and was derived by its use in signal detection theory [3,12] as a means to determine an electronic receiver's ability to distinguish a true radio signal from noise.

ROC analysis has proven useful with data col- lected in the behavioral and social sciences [5,8]. Unlike Chi-square or the Fisher's exact tests, ROC analysis simultaneously provides a method for choosing optimal cut-off points and statistically com- paring diagnostic efficiency on dimensional (i.e., continuous) measures that establish a diagnostic marker/disease relationship.

In the current study, we test the ability of the various IAP memory scores to differentiate between patients with left vs. right seizure focus by choosing cut-off scores on each of the 5 memory measures and determining the sensitivity as well as the speci- ficity. Using ROC terminology, sensitivity in this analysis will represent those patients with left lateral- ized seizure focus who have scores above the chosen cut-off point. The specificity will be represented by the proportion of patients with a right lateralized seizure focus who have scores below the chosen cut-off point. The relationship between the sensitiv- ity and specificity values will vary, depending on the cut-off scores. For example, a low cut-off score will increase sensitivity at the cost of decreasing speci- ficity. Thus, more left temporal patients will be correctly classified but more fight temporal patients

246 D.D. Roman et al. / Epilepsy Research 25 (1996) 243 248

will be misclassified. Conversely, if specificity is increased sensitivity decreases. In this case, more right temporal patients are correctly classified but more left temporal patients are misclassified.

By plotting the true positive fraction vs. the false positive fraction we can obtain the ROC curve which represents the balance between the sensitivity and specificity or the differential overlap for the given measure. Calculating the area under the curve (AUC) provides an interpretable statistic with regard to the performance of a given measure. Potential range of values for this AUC fall between 1.0 and 0.5. In this instance, a memory measure that is perfectly able to distinguish left from right seizure focus patients (100% correct classification of both groups) would have an AUC value of 1.0. Conversely, if the groups overlap completely for the varying ranges of cut-off points the measure would have a value of 0.5. A value of 0.5 would be represented by a diagonal or "a line of no information." In summary, a test will show greater performance if it has less differential

overlap which will in turn be represented as a larger value of AUC and a curve that is farther away from the diagonal. The AUC value can then be interpreted as an estimate of the probability that a randomly chosen patient with the condition (e.g., a left tempo- ral seizure focus) will have a higher score on the measure than a randomly chosen patient without the condition (e.g., a patient with fight seizure locus).

Statistical analysis was performed using LABROC and CLABROC programs [7]. LABROC provides maximum likelihood estimates of a binormal ROC curve and its associated parameters from a set of continuously distributed data.

3. R e s u l t s

ROC analysis was employed using Total Mem- ory, Verbal Memory, Non-verbal Memory, Design Memory, and Pictorial Memory scores. The ROC analysis deternlines the ability of each measure in

>

'S<

0 °~

100

90

80

70

60

50

40

30

20

10

0

" a::'~': A... ~ ' ~ " e = - - l ~ . _

':;~::':i~:'." ........ "'". . . . . . . . .

-? \ ,

Total Memory

......... ~' ....... Verbal Memory

. . . . o--- Non Verbal Memory

...... × .... Design Memory

........... i ...... Pictorial Memory

0 l 0 20 30 40 50 60 70 80 90

P r o p o r t i o n o f R i g h t T e m p o r a l P a t i e n t s Be low the Cutof f Point

(Specif icity)

Fig. 1. Proportion of patients correctly classified at various cut-off points for the 5 memory measures.

100

D.D. Roman et a l . / Epilepsy Research 25 (1996) 243-248 247

differentiating between left vs. right lateralized seizure focus compared to chance alone. The analy- sis also tests whether the performance of a given measure differs significantly from another. Lastly, it determines the optimal cut-off points which correctly classifies the largest number of patients from both groups.

Fig. 1 displays the ROC curves for each of the 5 memory measures, along with the line of no informa- tion represented by the diagonal.

Table 2 shows the values of the area under the curve (AUC) with standard deviation and 95% confi- dence intervals for the 5 memory measures.

Each of the 5 memory measures differed signifi- cantly from the line of no information as indicated by the 95% confidence intervals, none of which contain 0.50 within the interval. The Total Memory score differed significantly from the Verbal, Design, and Pictorial Memory scores at the 0.05 level. These differences are apparent from their 95% confidence intervals. If the values contained within the confi- dence interval from one score do not overlap the values contained within the confidence interval from another score, then these scores differ significantly. Total and Non-verbal Memory scores did not differ significantly from each other. Addit ional ly, Verbal, Design, and Pictorial Memory scores did not differ significantly from each other. Of the 5 measures, Total Memory and Non-verbal Memory scores cor- rectly classified the largest number of patients from each of the two groups.

Table 3 shows the optimal cut-off points on each measure that yields the highest sensitivity (correct classification of left temporal patients) and speci- ficity (correct classification of patients with right seizure focus). Note that there is a tendency for patients to score higher when using the left hemi- sphere.

Table 2 AUC values and confidence intervals for 5 memory measures

Memory measure AUC value 95% Confidence interval

Total memory 0.8099 _+ 0.0573 0.7949-0.8249 Non-verbal memory 0.7942 _+ 0.0591 0.7787-0.8097 Verbal memory 0.7733 +0.0621 0.7570-0.7896 Design memory 0.7431 _+0.0629 0.7265-0.7597 Pictorial memory 0.7444 _+ 0.0668 0.7267 0.7621

Table 3 Optimal cut-off points and percent of patients with left and right seizure focus correctly classified

Memory measure Optimal % of left % of right cut-off * TL patients TL patients

correctly correctly classified classified (sensitivity) (specificity)

Total - 6.0 75.0 78.6 Non-verbal - 3.0 78.6 71.4 Verbal - 1.0 78.6 64.3 Design - 1.0 66.7 75.0 Pictorial - 1.0 74.1 57.2

* The cut-off scores represent right hemisphere performance mi- nus left hemisphere performance.

4. Discussion

These findings suggest that IAP memory testing is a valid means of assessing hippocampal integrity and has utility in lateralizing seizure focus in patients with complex-part ial epilepsy. This is consistent with a recent depth electrode study which demonstrated delta slowing in the hippocampus following intrac- arotid injection of sodium amytal [4]. This suggests that the hippocampus is rendered dysfunctional, pre- sumably through deafferentation, even though it may not be directly perfused by an internal carotid injec- tion.

Despite its value in lateralizing temporal lesions, use of IAP test results in ascertaining the relative contribution of the two temporal lobes in mediating memory functioning must be undertaken with cau- tion. Our findings are consistent with those of others [2,11] in showing a dominant hemisphere advantage for many types of traditional IAP memory items, including seemingly " non -ve rba l " measures. Addi- tionally, preliminary results from a recent multicen- ter project [1], which studied 419-535 epilepsy pa- tients with presumed focal, unilateral temporal lobe disease, failed to demonstrate a right hemisphere advantage for figural memory on baseline testing. Non-verbal memory was assessed using the Wech- sler Memory Scale Visual Reproduction subtest and Rey-Oster r ie th . Patients with left and right seizure foci performed similarly on these measures, contrary to the expectation that right temporal lesions would result in poorer performance on these measures.

248 D.D. Roman et al. / Epilepsy Research 25 (1996) 243-248

These findings raise doubts about presumed non- dominant hemisphere superiority in mediating figural memory and suggest that traditional IAP procedures may not be adequately assessing right temporal lobe functions.

There are two possible explanations for the ob- served dominant hemisphere advantage on IAP memory testing. First, aphasia may interfere with the effective processing of many types of test stimulus. Since adequate processing is an essential prerequisite to memorization, poorer memory performance occurs following the dominant hemisphere injection, regard- less of the nature of the to-be-remembered material. Alternately, the dominant hemisphere superiority may be an artifact of the test items. In designing this battery, we endeavored to select fairly simple items, in the hope that the stimuli could be adequately processed by either hemisphere. But this IAP battery, and others with similar item content, may not suffi- ciently tap memory functions unique to the non- dominant lobe.

Further refinement of the IAP battery may im- prove its utility in lateralizing temporal lobe lesions and predicting post-lobectomy outcomes, especially if better tests of non-dominant temporal lobe func- tioning can be incorporated. In the meantime, it is necessary to adjust cut-off scores to correct for the left (dominant) hemisphere advantage. Given the strong tendency for the dominant hemisphere to out- perform the non-dominant hemisphere, equivalent or better memory performance by the non-dominant hemisphere seems to be a strong indication of domi- nant temporal lobe disease. In contrast, a slight dominant hemisphere advantage on IAP testing should not be considered compelling evidence of a non-dominant hemisphere lesion or an indication that the non-dominant hemisphere contributes minimally in day-to-day memory functioning.

Acknowledgements

This research was supported in part by NIH- NINDS Grant PSONS16308. The authors wish to

thank Elizabeth Destafney and John Rarick for their help with data collection and data base setup. We are grateful to Dr. Michael Risinger for his insightful comments on an earlier draft of this paper.

References

[1] Barr, W.B. and The Bozeman Epilepsy Consortium, The right temporal lobe and memory, J. Int. Neuropsyehiat. So<'., 1 (1995) 139-140.

[2] Christianson, S.A., Salsa, J. and Silfvenius, H., Hemisphere memory differences in sodium amytal testing of epileptic patients, J. Clin. Exp. Neuropsychol., 12 (1990) 681 694.

[3] Egan, J., Signal detection theory and the ROC analysis. In: E. Carterette and M. Friedman (Eds.), Depth Perception Through Motion, Academic Press, New York. 1975, pp. 6 91).

[4] Gotman, J., Bonwer, M.S. and Jones-Gotman, M., lntracra- nial EEG study of brain structures affected by internal carotid injection of amobarbital, Neurology, 42 (1992) 2136 2143.

[5] Hsiao, J.K.. Bartko, J.J. and Potter, W.Z., Diagnosing diag- noses, Arch. Gen. Psyehiat., 46 (1989) 664-667.

[6] Loring, D.W., Meador, K.J., Lee, G.P. and King, D.W., Amobarbital Effects and Lateralized Brain Function: the Wada Test, Springer, New York, 1992.

[7] Metz, C., Shen, J., Wang, P. and Kronman, H., CLABROC, LABROC, University of Chicago, Department of Radiology, Chicago, IL, 1988.

[8] Murphy, J., Berwick, D., Weinstein, M., Borus, J., Budman. S. and Klennan, G., Performance of screening and diagnostic tests, Arch. Gen. Psychiat., 44 (1987) 550 555.

[9] Perrine, K., Sass, K.J., Choi, I.S., Brown, E.R., Luciano, D.J. and Devinsky, O., Interhemispheric difference scores in lAP memory testing, J. Clin. Exp. Neuropsy~hol., 15 (1993) 24.

[10] Rausch, R. and Risinger, M.. Intracarotid sodium amobarbi- tal procedure. In: A.A. Boulton, G.B. Baker and M. Hiscock (Eds.), Neuromethods. Vol. IZ Neuropsyehology, Humana Press, Clifton, NJ, 1990.

[11] Risse, G.L., Fangman, M.C. and Gates, J.R., Left hemisphere superiority for verbal and visual recognition memory in patients undergoing the intracarotid sodium amytal test. J. Clin. Erp. Neuropsyschol., 15 (1993) 100.

[12] Swets, J. and Pickett, R.. Evaluation qf Diagnostic Systems, Academic Press, Orlando, FL, 1982.

[13] Thomdike, E.L. and Lorge, 1.. Teacher's Word Book qf 30,000 Words, Teacher's College, Columbia University, New York, 1944.