The first attempt to functionally inactivate cerebral language regions pharmacologically was made by Gardiner (1), who injected procaine directly into the cerebral cortex before neurosurgical intervention. Harris and Snyder (2) provide a historical review of Gardiner’s contributions to speech localization using this approach. Wada used intracarotid amobarbital injection to lateralize language before electroconvulsive treatment. Originally published in Japanese in 1949, Wada’s seminal paper has been translated into English (3). Wada also describes initially using a carotid injection of amobarbital in a young boy with partial seizures, which is recounted in the historical context of postwar Japan (4). Wada later adapted this procedure to identify cerebral language laterality in candidates for epilepsy surgery while he was a Rockefeller Fellow at the Montreal Neurologic Institute (5). Milner and associates (6) subsequently modified the procedure to include a memory component to assess risk of postoperative amnesia and its relation to seizure onset laterality, likelihood of significant memory loss, and seizure outcome.

Concern about posttemporal lobectomy amnesia arose after experience at the Montreal Neurologic Institute with several early patients, including two who underwent bilateral temporal lobectomy (such as HM), as well as others with contralateral hippocampal lesions who underwent unilateral temporal lobectomy (7,8). The Wada test is used to functionally inactivate a single hemisphere; this allows the assessment of whether language and memory abilities are supported by the nonperfused contralateral hemisphere.

Surveys of epilepsy surgery centers (9,10) suggest great procedural variability. In most approaches, however, a cerebral angiogram is conducted before Wada testing to evaluate cerebral vasculature, assess hemispheric cross-flow, verify a patent posterior communicating artery, and assess perfusion of the posterior cerebral artery. The importance of the angiogram becomes especially clear in the rare instance of a unilateral or bilateral persistent trigeminal artery, which often perfuses vital brainstem structures. Ordinarily, Wada testing should not be conducted when there is evidence of brainstem perfusion.

In most Wada procedures, patients hold their arms straight up in the air and begin counting. A bolus injection of an amobarbital solution, in saline or bacteriostatic water, is then administered into the internal carotid artery over a 4- to 5-second interval through a transfemoral catheter placed above the carotid bifurcation. Onset of contralateral arm paresis is readily evident, and disruption of automatic verbal activity (i.e., counting) typically provides the first indication of language dominance. Interruption of automatic activity, however, may be caused by the acute effect of the amobarbital bolus. Language is assessed and memory items are presented during the hemiparesis. Memory for information presented during the drug effect
is evaluated on return to baseline, which usually occurs after approximately 10 minutes. The second hemisphere is usually perfused in a similar manner, although some centers wait 30 minutes between injections and others perform the second study on a different day.

Beyond these generalities, however, similarities between Wada procedures across various surgery centers are limited. Some centers use preinjection testing similar to that used during the procedure to obtain a detailed baseline. Other centers may use only preinjection hand strength as a means of gauging the hemiparesis. Use of the EEG is variable, and a wide range of stimuli are used for language and memory assessment. Moreover, the Wada test is not performed on all patients in some centers. For example, candidates for right temporal lobectomy who are right-handed and do not have ictal or postictal aphasia during EEG monitoring may not undergo Wada testing at some centers. The amobarbital dose also has been reported to vary from center to center, from about 70 mg to as high as 225 mg per injection (9), although there is a more recent trend toward lower dosing.

Criteria for determining language dominance vary across centers; this limits comparability of study results. Snyder and colleagues (10) found that centers reporting 10% to 60% impairment versus those reporting 0% to 6% used different criteria to detect speech or language impairment. Centers reporting low prevalence of mixed-hemisphere language representation generally did not use criteria involving production of partial phonemes, impairment of serial rote speech, or impaired ability to express familiar words. Ability to name objects, however, was used to determine speech lateralization by more than 90% of the responding centers in the Snyder survey (10). Six or more objects are usually presented for naming during amobarbital-induced hemiparesis. Estimates are in general agreement that 90% or more of right-handed individuals are left-hemisphere language dominant (11,12).

Benson (11) reviewed studies of aphasia in left- and right-handed persons after various types of left- or right-hemisphere insult (e.g., surgical, traumatic). Thirty-two percent of the non-right-handed patients developed aphasia after left-hemisphere insult, whereas 60% of right-handed patients developed aphasia after left-hemisphere insult. Twenty-four percent of the non-right-handed patients developed language disorders after right-hemisphere insult, and only 2% of right-handed patients developed aphasia after right-hemisphere insult. Mixed language dominance has been reported in 0% to 60% of temporal lobectomy patients, with about half of the epilepsy surgery centers in one survey reporting a 0% to 20% incidence (10). In one series, two of 103 patients evaluated with Wada testing had exclusive right-hemisphere language, and another 22 patients had bilateral representation (13). Language representation was asymmetric in 17 of the 22 patients with bilateral language, and 13 of these had more left- than right-hemisphere representation. Exclusive right-hemisphere language, therefore, seems to be relatively rare, a finding supported by functional MRI studies (14). Language representation might best be conceptualized as a continuous rather than a dichotomous variable (13).

Two other studies (15,16) describe the results of Wada language lateralization testing in groups of 368 and 167 epilepsy surgery patients. Risse and coworkers (16) reported that, of 304 right-handed patients, 87% were left-hemisphere language dominant and 13% had right or bilateral representation. Of 64 left-handed patients, 62% were left-hemisphere language dominant and 38% had right or bilateral language representation. Similarly, Helmstaedter and colleagues (15) found that 82% of 137 right-handed patients had left-hemisphere language dominance and 18% had “atypical” dominance. However, only 30% of 30 left-handed patients had left-hemisphere language dominance; the other 70% had atypical language dominance (15). This apparent difference may be caused by sample composition, as Risse and associates may have included more nontemporal surgical candidates in their group; alternatively, it may be related to differences in classification systems (16). Both reports indicated that right-hemisphere language dominance is associated with early onset of seizures, injury, or lesion in the left hemisphere. Women with left-hemisphere ictal onset are more likely to have atypical language representation (17).

There is generally good agreement between language results of Wada testing and operative or extraoperative cortical language mapping (18, 19, 20). Both hemispheres, however, are rarely studied with stimulation mapping, which severely limits opportunities to compare stimulation with amobarbital data.

The rate and pattern of language and memory recovery after sodium amobarbital injection are potential confounding factors that affect interpretation. Ravdin and associates (21) reported on the serial recovery of language in candidates for right (15) and left (22) temporal lobectomy. After injection in the language-dominant hemisphere, “intelligible vocalization” occurred on average at 3 minutes 27 seconds. Paraphasias were common, and language function returned to baseline in a stereotyped order. Average time for return of naming ability was 8 minutes 29 seconds; average time to comprehension return was 9 minutes 58 seconds; and average recovery of repetition ability was 12 minutes 30 seconds. Thirty-six of the 39 patients with left-hemisphere language dominance were initially mute, and 28 of the 39 recovered language in the order presented above, with some variation of order in the other 11 patients.

Morris and colleagues (23) studied language and memory recovery during the Wada test in 49 patients with left-hemisphere language dominance who were candidates for
temporal lobectomy; doses averaged 120 mg. Language began to recover rapidly within about 3 minutes after injection of the language-dominant hemisphere. Language testing included reading words, naming line drawings, and responding to faces. Capacity to encode and later recognize the same items, however, was not associated with the language impairment. The investigators concluded that induced aphasia does not significantly affect recognition memory, even though patients may not be able to name or read stimulus items. A separate study (24) reported similar findings. In general, there is a tendency to perform more poorly when the language-dominant hemisphere is perfused, and some centers employ an “aphasia correction” when calculating memory results in the language-dominant hemisphere. However, this correction does not necessarily improve diagnostic accuracy (25).

Memory testing was added to the Wada test for language lateralization to provide information about the risk of profound postoperative amnesia. The Wada memory test was not originally intended to predict risk for degrees of memory decline after temporal lobectomy, but it has gradually assumed this purpose. Most Wada protocols involve presentation of objects, photographs, or line drawings during the induced hemiparesis; the patient is asked to recall this information after neurologic status has returned to baseline. This usually occurs in about 10 minutes. An exception is the University of Washington protocol, which requires patients to name items and, after a brief interference task, to recall the item that had just been displayed (26).

Wada memory results have been used to aid in confirming seizure onset laterality, to predict postoperative seizure control, and to predict memory outcome. The underlying assumption is that pharmacologic ablation of an abnormal hippocampus, which supports only limited memory function, will result in little change in the ability to recall material presented during hemispheric anesthesia. Chelune (27) has conceptualized this in terms of determining the functional reserve of the contralateral hippocampus and the functional adequacy of the hippocampus ipsilateral to seizure onset in candidates for temporal lobectomy.

The capacity of Wada testing to determine risk for memory decline has been debated. In one series (28), 10 patients who performed poorly on Wada testing underwent anterior temporal lobectomy without becoming amnestic and experienced improved seizure control. This suggests that some patients could needlessly be denied surgery if the Wada test were the only measure of the contralateral hemisphere’s ability to sustain memory function.

The capacity of a Wada procedure to test mesial temporal function has been criticized. Dasheiff and coworkers (29) reported that patients with frontal lobe or generalized epilepsy had poorer Wada performances than temporal lobectomy patients. They suggested that this is contrary to expectation if the Wada test, in fact, measures mesial temporal function. Yet these findings are consistent with temporal lobectomy patients having greater mesial temporal dysfunction than patients with frontal generalized epilepsy. Temporary amobarbital ablation of neurocognitive function supported by abnormal mesial temporal structures in patients with temporal lobe epilepsy should result in little compromise or change of memory function during Wada testing. Ablation of healthier mesial temporal structures in patients with frontal generalized epilepsy would result in greater Wada memory test deficits.

A large number of independently conducted studies support the validity of Wada testing, typically based on comparison of hemispheric memory asymmetry, to lateralize pathology consistent with the hemisphere of ictal onset (24, 25, 26,30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44). Although these reports support the validity of the Wada test in lateralizing cerebral pathology in candidates for epilepsy surgery, particularly temporal lobectomy, differences in Wada protocols involving stimulus presentation and amobarbital dosage confound direct comparison of these studies. Wada testing has been reported to correctly classify 45% to 75% of patients with left temporal epilepsy and 67% to 79% of patients with right temporal epilepsy. Patients with right temporal epilepsy have been more accurately identified in these studies; incorrect classification ranges from about 6% to 11% for patients with left temporal epilepsy and from 4% to 9% for patients with right temporal epilepsy (31,35,36).

Interestingly, two studies address the meaning of nonlateralizing Wada tests. Swearer and associates (44) reported on 23 patients with temporal lobe epilepsy, 19 of whom underwent surgery. The four patients who did not have surgery were considered poor candidates because bihemispheric seizures occurred during intracranial monitoring. These four patients had a significantly lower difference between their Wada memory scores than the other patients, and the authors proposed that lack of a significant Wada memory asymmetry suggests bihemispheric onset. Hamberger and colleagues (34) reported on 20 patients with lateral or neocortical onset and 22 patients with mesiobasal-onset temporal lobectomy. Histopathologic analysis, intracranial EEG, or presence of a temporal neocortical lesion was used to determine region of ictal onset. The investigators correctly classified 18 of the 22 patients with mesial temporal onset and seven of the 10 patients with lateral cortical onset. The patients with lateral onset had a significantly smaller memory asymmetry than the patients with mesial onset. Taken together, these two studies further support the notion that Wada test memory asymmetries are associated with mesial temporal pathology and ictal onset in candidates for temporal lobectomy and
that lack of asymmetry is associated with a nonmesial temporal region of onset.

Several investigations have demonstrated that Wada memory performance is associated with the nature and extent of pathology. Rausch and coworkers (45) found that Wada memory performance was associated with the extent of hippocampal neuronal loss, which was measured as a composite of cell counts across five hippocampal regions. Sass and colleagues (46) demonstrated that temporal lobectomy patients with poor memory performance after injection of the hemisphere contralateral to the seizure focus had lower cell counts in the CA3 hippocampal subfield. CA3 hippocampal cell counts also correlated with the amount of information recalled following contralateral amobarbital injection.

Davies and associates (32) reported on their experience with Wada testing in 30 temporal lobectomy patients with adequate Wada, MRI, and histopathologic results. They demonstrated that recognition memory was asymmetric by comparing right and left hemisphere injections; a difference of at least ±2 (of a possible 8) occurred in 73% of the patients. Lateralization was correct in 91% of the 22 patients; two cases were incorrectly lateralized, and eight patients had indeterminate asymmetry scores. Quantitative MRI hippocampal volumes and T2 values are correlated with Wada memory asymmetry scores (30,47). Baxendale and coworkers (30) studied data from 48 patients with temporal lobe epilepsy and they found that unilateral Wada memory scores were not correlated with individual hippocampal volumes or T2. However, a significant correlation was shown in the difference between hippocampal volumes or T2 and the difference between Wada memory scores.