- Top of page
Summary: Purpose: Although anterior temporal lobectomy (ATL) is an effective treatment for many patients with medically refractory temporal lobe epilepsy (TLE), one risk associated with this procedure is postsurgical decline in memory. A substantial number of past studies examined factors that predict memory decline after surgery, but few have investigated multiple predictors simultaneously or considered measures that are currently in use.
Methods: This study compared the relative contributions made by presurgical neuropsychological test scores, MRI-based hippocampal volumetric analysis, and Wada test results to predicting memory outcome after ATL in a group of 87 patients.
Results: Logistic regression analyses indicated that noninvasive procedures (neuropsychological testing and MRI) made significant contributions to improving the prediction of memory outcome in this sample. The results from the Wada procedure did not significantly improve prediction once these other factors were considered. The only exception was in predicting memory for visual information after a delay, in which Wada results improved prediction accuracy from 78% to 81%.
Conclusions: Current neuropsychological tests and MRI volumetric measures predict changes in verbal and visual memory after ATL. The relatively small change in correct classification rates when Wada memory scores are considered calls into question the benefits of using Wada test results to predict memory outcome when the results of noninvasive procedures are available.
Anterior temporal lobectomy with resection of the mesial structures (ATL) is known to be an effective intervention for patients with temporal lobe epilepsy (TLE). However, one risk associated with this procedure is postoperative decline in cognitive abilities, and particularly in episodic memory (Ivnik et al., 1987; Hermann and Wyler, 1988; Ivnik et al., 1988; Naugle et al., 1993; Martin et al., 1998). Across studies, from 25 to 75% of left ATL (LATL) patients and 22 to 60% of right ATL (RATL) patients experience some degree of decline in their memory abilities after surgery (Ivnik et al., 1988; Chelune et al., 1991; Phillips and McGlone, 1995; Martin et al., 1998). Thus the ability to predict postoperative memory outcome has become a highly relevant issue for patients with medication-resistant TLE who are considering surgical intervention as a treatment option.
Researchers have made considerable strides over the last 20 years in their attempts to delineate individual factors that are associated with memory decline after ATL. One of the most robust findings is that temporal lobe resections involving the speech-dominant left hemisphere are more consistently associated with memory loss than are those confined to the non–speech-dominant right hemisphere (Ivnik et al., 1987; Ivnik et al., 1988; Chelune et al., 1991; Chelune et al., 1993; Naugle et al., 1993; Trenerry et al., 1993; Phillips and McGlone, 1995; Baxendale et al., 1998; Martin et al., 1998). Second, an absence of cell loss in the hippocampus before surgery (hippocampal sclerosis: HS) is associated with a higher incidence of memory decline after surgery, regardless of whether the hippocampal damage is documented histologically (Hermann et al., 1992; Hermann et al., 1994; Seidenberg et al., 1996; Baxendale et al., 1998; Seidenberg et al., 1998) or on presurgical MRI scans (Trenerry et al., 1993). A third factor known to share a strong relation with postoperative changes in memory capacity is preoperative memory ability, with stronger baseline memory abilities predicting greater memory decline after surgery (Ivnik et al., 1988; Chelune et al., 1991; Jokeit et al., 1997; Davies et al., 1998), presumably because removing a functional hippocampus negatively affects memory (Chelune, 1995). Finally, demographic characteristics of epilepsy patients (age, sex, level of general intellectual functioning) and aspects of their seizure disorders (age at seizure onset, age at first risk for developing seizures) also share significant relations with memory outcome after ATL (Jokeit et al., 1997; Davies et al., 1998; Seidenberg et al., 1998; Chelune and Najm, 2000). However, these variables account for little variance in memory change once other predictors are taken into account in multifactorial models (Jokeit et al., 1997; Chelune and Najm, 2000).
In addition to these factors, performance on the Wada test is commonly used in clinical settings to evaluate a patient's risk of memory decline after ATL. The Wada test (Wada, 1960), which was traditionally used to determine speech lateralization in surgical candidates, was later adapted to evaluate the propensity of the nonaffected hemisphere to support memory (Milner et al., 1962) to identify patients at risk for postoperative global amnesia (Scoville and Milner, 1957). Because recent research has called into question the validity of the Wada for this purpose (Loring et al., 1990; Lee et al., 1995; Kubu et al., 2000), attention has shifted to using the Wada test to predict more subtle preoperative-to-postoperative memory changes. Both low memory scores after left injection (i.e., poor right hemisphere–mediated memory) (Davies et al., 1998; Bell et al., 2000) and high memory scores after right injection (i.e., good memory abilities sustained by the to-be-resected left hemisphere) (Kneebone et al., 1995) have been shown to correspond to poor memory outcome after surgery in LATL patients. However, direct comparisons of the relations between absolute and relative memory scores on the Wada and memory outcome suggest that relative memory performances correlate more strongly with postoperative memory decline than with the absolute level of performance achieved by either hemisphere alone (Wyllie et al., 1991; Loring et al., 1995).
The majority of the published studies in the literature have focused on single predictors of memory outcome. The few multifactorial studies that have been published have typically investigated demographic characteristics of patients in combination with preoperative memory performances, with or without taking into consideration Wada memory test scores (Loring et al., 1995; Jokeit et al., 1997; Davies et al., 1998). Two recently published studies have examined more comprehensive multivariate models for predicting memory outcome after ATL (Chelune and Najm, 2000; Stroup et al., 2003). Chelune and Najm (2000) found that a combination of side of surgery, demographic characteristics, baseline memory abilities, presence or absence of hippocampal atrophy, and Wada test scores accounted for 39% of the variance in memory outcome in 72 TLE patients. Side of surgery, hippocampal atrophy, and baseline memory scores were the primary predictors of memory outcome, and other variables (including Wada test scores) did not significantly improve the regression equation. A logistic regression analysis then demonstrated that these three factors correctly classified 71% of the patients as experiencing stable or diminished memory after surgery. Although these results offer some preliminary evidence that Wada test scores do not improve clinicians' ability to predict memory outcome, these scores were considered only at the group, not at the individual, prediction level. Thus the possibility remains that the Wada test contributes to predicting memory outcome for some, but not all, patients.
The second published comprehensive multivariate study was conducted by Stroup and colleagues (2003). This study found that resection of the speech-dominant hemisphere, MRI findings inconsistent with strictly unilateral mesial temporal sclerosis in the resected hemisphere, poor scores on immediate and good scores on delayed memory measures, and good performance on a Wada memory test after contralateral injection were all significant, unique, independent predictors of postsurgical verbal memory decline in a sample of 132 TLE patients. Although scores on the Wada test were significant predictors of outcome, they were relatively weak when compared with side of resection, absence of unilateral mesial temporal sclerosis, and baseline delayed verbal memory.
The previously reviewed studies suggest that side of surgery (speech-dominant vs. nondominant hemisphere), hippocampal atrophy on MRI scans, baseline neuropsychological performance, and (possibly) Wada test results each contributes to predicting memory outcome after surgery. However, none of the multifactorial studies examined memory outcome on visuospatial memory measures. In addition, in all studies to date, both preoperative and postoperative memory abilities were documented by using previous versions of standard neuropsychological tests (Wechsler, 1987) rather than more current measures (Wechsler, 1997).
With this in mind, the current study had two goals: (a) to study the value of new neuropsychological assessment techniques in the prediction and evaluation of memory outcome after ATL in patients with medically refractory TLE, and (b) to compare the relative contributions made by neuropsychological assessment, MRI-based hippocampal volumetric analysis, and Wada test results to the prediction of memory outcome after ATL in patients with TLE. The primary hypothesis of this study was that side of resection, MRI variables, and presurgical memory abilities would be the best predictors of memory decline, with scores derived from the Wada test failing to improve significantly the prediction of outcome once these other factors were taken into account.
- Top of page
Addressing the issue of whether the relative contribution Wada test results make to predicting memory outcome is beneficial enough to justify the risks and costs inherent in the procedure is essential. The possibility of reducing or eliminating reliance on Wada test results when predicting memory outcome is a particularly important consideration due to the invasive nature of the Wada test and the corresponding risk associated with cerebral angiography, including an estimated morbidity of 1 to 3% (Wada et al., 1960; Branch et al., 1964; Serafetinides et al., 1965). As alternative techniques for lateralizing and localizing speech, such as functional MRI of the brain (Binder et al., 1996) and repetitive transcranial magnetic stimulation (Jennum et al., 1994; Epstein et al., 1999), are developed and improved, the necessity of completing a Wada test to document speech dominance may be reduced or eliminated. If new, state-of-the-art, noninvasive techniques are also sufficient to maximize prediction of memory outcome after ATL, routine use of the Wada test for this purpose may be unnecessary. As a result, the risks and costs inherent in evaluating the surgical candidacy of patients with medically refractory temporal lobe epilepsy may be drastically reduced.
Although other studies have suggested that the Wada test contributes only limited information to the prediction of outcome after the results of noninvasive procedures are considered (Chelune and Najm, 2000; Stroup et al., 2003), the current study expands the literature on multivariate prediction of memory outcome after anterior temporal lobectomy by evaluating predictors of both auditory (verbal) and visual memory decline. It also uses the most recent version of commonly used memory measures. Thus these results should apply more directly to current clinical practice than those of previous studies.
Perhaps one of the most surprising findings of this study was that fewer patients than expected demonstrated statistically significant postsurgical memory decline. Whereas prior studies using similar methods (but previous versions of memory tests) reported postsurgical memory decline in 35–45% of ATL patients, in the current sample, estimates of memory decline ranged from merely 18% for auditory immediate memory to 28% for visual delayed memory. Because regression-based z-scores at the 15th percentile were used to define operationally postoperative memory decline, memory decline would be expected to occur at a base rate of 15% in a control group of medically managed epilepsy patients. Thus the portion of this sample that evidenced decline (18–28%) was not that much greater than would be expected in a group of epilepsy patients tested twice without undergoing surgical intervention in the interim.
An important limitation to these results warrants mention. The patients who were selected for this study were a highly screened sample of epilepsy patients. The preoperative tests that were used in this study were also available to and used by the neurologists managing the patient's care when reaching a decision about treatment options and recommendations. Although the data considered in this study may have been more precise than those available to the clinicians working with these patients (e.g., quantitative versus qualitative MRI data), a large amount of overlap existed between the factors used in this study and those used to counsel these patients about surgery. Patients determined to be at high risk for memory decline may have been discouraged from surgery or may have decided against surgery when offered. As a result, a possible explanation for the lower-than-expected incidence of memory decline in this sample is that, as clinicians have gained a better understanding of the variables that affect memory decline after surgery in recent years, patients are being selected more carefully to minimize the negative cognitive outcomes that previously accompanied surgical intervention.
An alternate explanation for this pattern of results may be a lower level of sensitivity on the part of current memory measures to detect changes over time relative to those used in past studies. For example, the WMS-III auditory indices do not require word-list learning and recall, a task commonly thought to be more sensitive to verbal memory decline in epilepsy patients than more context-rich verbal memory measures. Although the patients in this study did complete a word-list learning and memory test, regression-based norms are not currently available to apply to these measures to determine a meaningful change after surgery, rendering them inappropriate for analysis with the techniques used in this study. If test–retest data with nonsurgical epilepsy control patients becomes available for word-list learning tasks, the current techniques and analyses could be applied to investigate empirically the role of the Wada test for predicting outcome on these types of memory measures.
The logistic regression analyses suggested that baseline scores derived from preoperative procedures were better predictors of decline than were asymmetry scores. This result was somewhat surprising, given previous findings in the literature that suggest that asymmetry scores are strong predictors of outcome (Wyllie et al., 1991; Loring et al., 1995). However, this may be due to the inclusion of bilateral indices (e.g., both left and right hippocampal volumes, both auditory and visual memory) in the analyses. Including bilateral indicators allowed the relation between these variables to be considered when the best-fitting logistic regression model was determined. Thus these results do not suggest that unilateral predictors are adequate for assessing risk of memory decline after surgery; instead, they indicate that the relation between ipsilateral and contralateral measures may be more complex than simple difference scores or ratio measures can capture.
In general, side of surgery in and of itself did not significantly predict memory decline, but it did interact with neuropsychological and MRI variables to improve the accurate classification of outcome. Both baseline memory scores and MRI volumes contributed to predicting outcome across the four memory measures, with the nature of their influence dependent on the hemisphere undergoing surgery and the type of memory being considered. Both left and right hippocampal volumes tended to contribute equally to the prediction equations, and MRI results played a role in predicting auditory memory outcome, but not visual memory outcome. Interestingly, only baseline visual memory (not baseline auditory memory) was important to consider when predicting postsurgical memory decline, with its influence being stronger for delayed than for immediate memory measures. The findings that both MRI and neuropsychology make independent contributions to predicting outcome is consistent with other recent findings in the literature that suggest that the functional status of the hippocampus may be somewhat independent of its structural status (LoGalbo et al., 2005).
Consistent with expectations, Wada test results did not typically emerge as significant predictors of outcome in this series of logistic regression analyses once side of surgery, neuropsychological variables, and results from MRI volumetric analysis were taken into account. This appears to support the conclusions of the two earlier studies that addressed this issue with previous versions of verbal memory tests (Chelune and Najm, 2000; Stroup et al., 2003) and different versions of the Wada (Stroup et al., 2003). Perhaps this finding is not surprising, given that the Wada test has not been designed specifically to predict subtle declines in material-specific memory. In addition, it may be unreasonable to expect a brief 12-item or 16-item memory test to demonstrate the equivalent predictive power of a much lengthier standardized neuropsychological test, especially when memory outcome is also being assessed on the standardized measure.
That being said, Wada test results did predict outcome on delayed visual memory measures. One reason Wada test results may have predicted delayed visual memory is that the stimuli used in the Wada procedure were presented visually rather than in an auditory modality, even when they were verbal in nature (i.e., words), making the type of encoding similar for these two types of tests. However, even in this case, the improvement in prediction accuracy was relatively small (78 to 81%), calling into question whether the benefits of considering Wada test results when predicting postsurgical memory outcome outweigh the risks inherent in that procedure.
Of course, the results of this study may or may not generalize to other Wada procedures. Approaches to memory testing during the Wada differ across centers, and other techniques may offer more toward the prediction of outcome. However, the fact that Stroup and colleagues (2003), working at a different epilepsy center, also found little incremental predictive power associated with the Wada after side of surgery, neuropsychological performance, and MRI findings were taken into account, lends support to the recommendation that risks and benefits be carefully considered before administering the Wada test.
Because Wada procedures differ across centers, one question that might arise when interpreting these results is the validity of the Wada procedure used in this study. To assure that the lack of contribution of the Wada test to predicting memory outcome was not due to invalid Wada memory test scores, additional logistic regression equation analyses were conducted with Wada percentage-correct scores (after left and right injection) and their interaction with side of surgery as the only predictors of memory decline. For both the Auditory Immediate and the Auditory Delayed Index, using Wada memory scores to predict verbal memory outcome resulted in a statistically significant improvement in the logistic regression model. Thus in and of themselves, Wada results did predict postsurgical verbal memory in support of previous findings in the literature (Loring et al., 1990; Wyllie et al., 1991; Kneebone et al., 1995; Loring et al., 1995; Bell et al., 2000). It is only when other noninvasive procedures are considered first and are allowed to account for as much variance in outcome as possible that Wada memory scores do not significantly contribute to prediction equations. This could be explained by a large amount of overlap in the prognostic information contributed by the Wada test and by neuropsychological or neuroimaging techniques. For example, a previous study demonstrated high correlations (r= 0.51) between left hemisphere and right hemisphere Wada memory test scores and ipsilateral hippocampal volumes (Cohen-Gadol et al., 2004). Although both Wada scores and hippocampal volumes showed significant inverse relations with memory outcome when considered separately, the high correlations among these two variables suggest that they may be measuring a common factor (“functional integrity,” according to the authors). If this be the case, having two measures of the functional integrity of the to-be-resected hippocampus may not improve the predictability of outcome over having a single measure of this construct, making the Wada a weak predictor once hippocampal volumes have been taken into account.
To assure that the data from the small number of patients evaluated with methohexital (n = 10) did not unduly influence the results of this study, the logistic regression analyses reported in the Results section were rerun for the subset of patients who received amobarbital as the anesthetic agent during the Wada procedure (n = 77). The previously reported results were essentially unchanged in these follow-up analyses. The only exception was that the significant improvement in predicting outcome on the Visual Immediate Memory Index based on neuropsychological and MRI variables no longer reached statistical significance (p = 0.055) because of the decrease in statistical power that resulted from the smaller number of participants.
One scenario that often poses difficulty for clinicians involves patients with left TLE whose left hippocampus is similar in volume or larger than the right hippocampus and whose verbal memory abilities appear intact or stronger than visual memory abilities on neuropsychological assessment. This situation would result in a prediction of high risk for memory decline, but whether Wada results might aid decision making in this context is worth exploring. Five patients in the current sample fit this description, and of these, all five did show significant decline on auditory memory tests. In three of these cases, the Wada test results supported the prediction of memory decline or were inconclusive. However, in two cases, the Wada test would have falsely predicted a positive memory outcome because memory performance after left injection (right hemisphere–mediated memory: in both cases, 88% correct) was good and was significantly better than after right injection (left hemisphere–mediated memory: in both cases, 50% correct). Thus in these two patients, Wada test results might have incorrectly led clinicians to conclude a lower risk of memory decline, when in fact memory decline was the final outcome.
Accurately predicting change in memory as measured by psychometric tests does not necessarily reflect the impact this memory change will have on an individual patient. As has been previously demonstrated in the literature, epilepsy patients' self-assessments of changes in their memory from before to after surgery often differ from actual changes documented on standardized memory tests (Vermeulen et al., 1993; McGlone, 1994; Sawrie et al., 1999). However, some recent evidence has suggested that patients are, indeed, sensitive to changes in their memory after surgery (Gleisner et al., 1998; Lineweaver et al., 2004), lending support to the extensive work aimed at predicting psychometric change to provide patients with as much relevant information as possible when they are reaching a decision about whether to pursue surgery as a treatment option.