WAIS (-R) WMS (-R)
4. Patients should be matched in the extent and type o f surgery they undergo 5 As much data as possible regarding the structural and functional integrity o f
the mesial and lateral temporal structures o f both temporal lobes should be used to interpret the clinical significance o f both pre- and postoperative neuropsychological deficits.
6. Patients with unilateral TLE are not homogeneous groups. The aetiology and pathology underlying a lateralised seizure focus may influence the nature o f any associated neuropsychological deficit and must be considered in the interpretation o f both pre- and postoperative neuropsychological scores.
Introduction Chapter 3
A Re v i e w o ft h e Me m o r y Te s t s Co m m o n l y Em p l o y e d i n
Ep i l e p s y Su r g e r y St u d i e s
*Ifa picture paints a thousand words.,. \
Throughout the literature the tests used in the studies o f patients with unilateral temporal lobe dysfunction are generally dichotomised into ‘verbal’ and ‘non-verbal’ or ‘visual’ memory tests. As the previous chapter demonstrated, temporal lobe epilepsy patients are not a homogeneous group. In the same way, ‘verbal’ and ‘non verbal’ memory skills are not mutually exclusive, single abilities that can easily be attributed to the function o f single brain structures. Firstly, no test material can be said to be one hundred percent ‘verbal’ or ‘visual’ and secondly, learning, recall and recognition tests all tap ‘memory’ skills. These memory skills dissociate in temporal lobe patients and those with lesions elsewhere in the brain (Delbecq-Derousen et al,
1994; Aggleton & Shaw, 1996; Baxendale, 1997)*'^’^’^^. This chapter examines the nature and sensitivity o f the memory tests used in the studies o f both pre and postoperative temporal lobe epilepsy patients.
What is a Material Specific Memory Deficit?
Material specific memory deficits can be defined in a number o f ways. In many o f the early studies, right and left temporal lobe epilepsy groups were simply compared with each other to detect significant differences in memory test scores (see Table 2.1 ). Whilst this approach may detect relative group strengths and weaknesses, both groups may be impaired when compared to the normal population encouraging misleading conclusions about the lateralised function o f the temporal lobes. In later studies normal control groups were included and the use o f standardised tests with appropriate age-scaled normative data has become more widespread, although
memory tests should be designed to enable the examination o f intra- and inter-patient differences in memory for polarised material, in addition to providing a framework to assess the clinical severity o f a memory deficit for verbal or non verbal material.
Memory Tests
Whilst a number a specialised memory tasks have been developed specifically for use with people with epilepsy (for example, Jones Gotman & Milner, 1978; Jones
Gotman, 1979; Zaidel & Rausch, 1981; Petrides, 1985; Goldstein et al, 1989; Pigott & Milner, 1 9 9 3)191,19 0,4 3 9,3 1 7,1 2 8 . 3 2 0 quoted most frequently in the modem
literature are standardised clinical memory tests. In her 1992 survey o f
neuropsychological services in epilepsy surgery programmes, Jones-Gotman found that the logical memory, paired associate learning and visual reproduction subtests o f the Wechsler Memory Scales (WMS) and the Wechsler Memory Scales - Revised (WMSII) were the most commonly employed clinical tests o f memory function. The delayed recall o f the Rey Osterrieth complex figure with or without an immediate recall condition was the second most popular measure. The Californian Verbal Learning Test and the Rey Auditory Verbal Learning Test are also frequently used in USA and European centres respectively. It is not surprising therefore that these are the tests most frequently cited in the temporal lobe epilepsy literature although the
protocols used frequently deviate from the standardised administration procedures.
Memory deficits in presurgical TLE patients
A number o f studies have reported material specific memory deficits, using these tests in patients with lateralised temporal lobe epilepsy (these studies are summarised in Table 2.1). For example, Delaney et al (1980)*^ examined the WMS scores o f 4 groups: patients with right temporal lobe epilepsy, left temporal lobe epilepsy, frontal
Introduction Chapter 3
temporal groups had significantly lower scores on the immediate and delayed logical memory indices than the other groups. In addition the LTLE group were also impaired relative to the normal and the FLE group on the logical memory percentage retained score. Similarly, the RTLE group performed more poorly than the control group and the FLE group on the percentage retained score from the visual reproduction test. There were no significant differences between the LTLE, FLE and normal control group on any o f the measures from the non-verbal tests. Whilst Goldstein & Polkey (1992)’^^ found no laterality effect on the scores from the logical memory subtest o f the WMS, they did find that their LTLE group performed more poorly than the RTLE group on measures from the paired associate learning task, consistent with a number o f other reports. Although significant RTLE vs. LTLE differences have been reported with the immediate, delayed (Delaney et al, 1980, Moore & Baker, 1996)*^’^^^ and percentage retained scores (Lencz et al, 1992)^*^ from the logical memory subtest and the learning and recall indices from the paired associate learning test, (Davies et al,
1993; Daum et al, 1991; Inoue et al, 1992)^^'^^ '^^ the findings regarding each
individual index are inconsistent. There is no consensus across the studies as to which
is the most sensitive index and scores from the logical memory test and the PALT have been reported to dissociate (Saling et al, 1993; 1993; Goldstein & Polkey,
1 9 9 2)351-352,129
In addition a number o f studies have failed to find right/left differences in
preoperative temporal lobe epilepsy patients’ performance on these tasks (Powell et al, 1985; Mayeux, 1980; Saykin et al, 1989; Goldstein & Polkey, 1992; Breier et al,
1996)^^^’^^^’*^^’'^*. In a recent example Jones Gotman (1996)^^^ examined the preoperative performance o f TLE patients with unilateral hippocampal sclerosis evident on volumetric MRI. The performance measures used were the mean recall score after a 90 minute delay on the WMS logical memory, paired associate learning and visual reproduction subtests and the Rey Osterrieth figure task. There were no significant differences between the left and right HS groups on any o f the measures,
the validity o f the WMS-R in a large sample o f 138 patients with clearly lateralised TLE. The LTLE patients obtained a significantly lower verbal memory index, and poorer scores on the logical memory (immediate and delayed) and digit span tasks compared with the RTLE group. The RTLE group did not differ from the LTLE group on any o f the non-verbal memory scores. Verbal/visual memory discrepancy scores proved highly unreliable in identifying patients with a right temporal lobe seizure focus. Unfortunately the authors did not examine the clinical significance o f the memory deficits in the RTLE and LTLE groups separately. However they do report that the patient group as a whole scored significantly lower than the normative sample across all memory index scores and on most o f the memory subtests. The authors conclude that the scores from the WMS-R are capable o f lateralising left temporal lobe dysfunction, but that the battery has limited value for RTLE patients. Overall this conclusion would be consistent with the collective picture o f the sensitivity and specificity o f the WMS and WMS-R to lateralised temporal lobe dysfunction emerging from the literature.
A somewhat more consistent pattern has emerged from studies employing the Rey Auditoiy Verbal Learning Test (RAVLT) and the Californian Verbal Learning Test (CVLT). Inoue et al (1992)’^', found that their LTLE group obtained significantly lower scores on the RAVLT compared to the RTLE group. Similarly Helmstaedter et al (1991)’"^^, reported the ‘traditional’ material specific memory deficits in a sample o f 52 presurgical TLE patients. The LTLE group performed more poorly than the RTL group on the RAVLT, whilst the RTLE group obtained significantly lower scores than the LTLE group on a visual analogue o f the task (Diagnosticum fur cerebralshaden). In a replication o f their earlier findings, Helmstaedter et al (1996)*^^^ reported deficits in the LTLE group compared to the RTLE group on the immediate and delayed recall indices from the RAVLT. Although there were no significant differences between the two groups on the immediate span and learning scores, both groups were impaired
Introduction Chapter 3
significantly poorer CVLT scores in the group o f LTLE patients compared with a matched RTLE and normal control group. Selective verbal memory deficits in LTLE patients have been reported on similar measures from the CVLT (Hermann et al,
1987)'".
However, as with the memory indices from the WMS/WMS-R, some inconsistencies do exist. The Helmstaedter et al group themselves report no significant preoperative differences between their 1996 RTL and LTL sample on any o f their measures from the RAVLT, although both groups obtained significantly lower scores than a normal control group. The RAVLT and the CVLT do not yield a single score for 'verbal memory' but rather a number o f scores, reflecting learning, free recall, recognition etc. Again as with the WMS, different scores from the tests have been found to
differentiate RTLE from LTLE patients in each study. In addition to the standard scores a number o f groups have re-interpreted the test data to create a large number o f permutations, which can make it difficult to compare the results from two studies ostensibly studying similar groups using the same test. For example, using the CVLT, Mungas et al (1985)^^^, found that the performance o f their LTLE group averaged
across three delayed recall trials (cued recall, phonemic recall and semantic cued recall), was significantly poorer than that o f the RTLE and control groups, whilst immediate and delayed free recall and phonemic and semantic cued recall for the distracter list, did not discriminate the two groups. Hermann et al (1992)*^^ presented RTLE vs. LTLE group differences on over thirty indices gleaned from the CVLT, ranging from ' the total number o f List A words recalled across trials 1 -5' to 'the slope o f a least-squares regression line calculated to fit changes in correct response scores across trials 1-5'. However, despite the very large number o f indices in this latter study, the measures used by Mungas et al (1985)^*^ are not included. There are both theoretical and statistical dangers in the generation o f a large number o f scores from a single test. Firstly it is unlikely that a single test can provide enough quality
avoid the impression o f a 'fishing trip' for significant results. Secondly, whilst a number o f statistical techniques are available to guard against the increased likelihood o f a type I error (see Dunn, 1961)*^* it is not uncommon for researchers to present interesting trends 'that would have been significant if....' (see Hermann et al, 1992)^^^. This introduces an element o f subjectivity to the reporting o f results given the
inevitable bias towards supporting current hypotheses.
Despite these problems, overall it would appear that these verbal learning tests with integral recall measures are more sensitive to left temporal lobe dysfunction than the individual verbal memory subtests o f the WMS-R. There may be a number o f reasons for this. On a practical level it may simply reflect the large number o f reports from a relatively small number o f centres, in effect multiple reports on an expanding core patient sample. However it is likely that the nature o f these learning tasks, with multiple trials, measures o f learning, immediate and delayed, cued and free recall, and pro-active and retro-active interference, may make them more sensitive to disturbance in verbal memory processes than the standard administration o f the WMS PALT. Goldstein et al (1988)*^^ examined the sensitivity o f each item in the WMS PALT to lateralised temporal lobe damage and found that only four o f the ten word pairs (three hard & one easy) produced a significant difference between the scores o f the LTL group and controls and between the LTL and RTL group. Jones-Gotman (1996)*^^ demonstrated many o f the factors that may impair recall following a single exposure to test material. She found that attention, comprehension and individual strategies were most variable on the first or only trial, but that these factors became less intrusive if a learning component was included in the test design.
In contrast to the hazy picture o f verbal memory deficits associated with a left
temporal lobe seizure focus, no clear picture has emerged o f the nature o f the deficits associated with a right temporal lobe epilepsy focus. A number o f studies have
Introduction Chapter 3
1989; Ribbler & Rausch, 1990; Saling et al, 1 9 9 3)3 0 4,34 6 , 3 5 1 some studies report a selective deficit on Maze tasks (Corkin, 1965; Milner, 1965)^^’^^^ and spatial
learning tests (Petrides, 1985; Martin et al, 1991; Helmstaedter et al, 1 9 9 1 ) 317,245,146
others have failed to find any deficits on a variety o f tests designed to tap visual memory skills including the Benton Visual Retention Test, the Rey Osterrieth figure recall, and the WMS figure recall (Loring et al, 1990; Jones Gotman, 1996)^^^’*^^. Barr (1996)^' recently reported the preliminary findings from an analysis o f over three thousand presurgical patients who had completed the delayed recall o f the Rey Osterrieth figure in epilepsy centres across North America. He found no significant difference between the RTLE and LTLE patients. Interestingly, there was a significant effect o f Epilepsy Surgery Centre suggesting significant discrepancies between the scoring systems used. Helmstaedter et al (1991)*'*^ do report promising findings regarding the specificity and sensitivity o f their version o f the ‘diagnosticum flir cerebralshaden’, a visual analogue o f the Rey Auditory Verbal Learning Test, but it has yet to be applied in an international context.
Similarly, Jones Gotman (1996)'^^ reported a significant difference between her right HS (hippocampal sclerosis) and LHS groups on an abstract design learning task with the RHS group performing significantly worse than the LHS group in the learning phase o f the task.
One possible explanation for some o f the apparent discrepancies in the un-operated TLE literature is that whilst patients with a left temporal lobe focus are consistently impaired on verbal memory tasks, RTLE patients show a more variable pattern on verbal memory tests and may, or may not be impaired. Non significant group
differences probably reflect the poor performance o f the RTL group, rather than intact skills in the LTL group. This highlights the necessity for a reference to normative data, either in the form o f standardised test norms or data from an experimental control group to make sense o f these findings. Studies that have failed to find RTLE
impaired compared to normals (Oxbury & Oxbury, 1989; Helmstaedter et al, 1996; Jones Gotman & Milner, 1978; Cavazutti et al, 1980; Prevey et al, 1988; Delaney et al, although others report normal memory functioning in both groups (Powell et al, 1985)^^^. As already discussed, the poor sensitivity and specificity o f the non-verbal memory tasks to lateralised right temporal lobe dysfunction is frequently attributed to duel encoding o f the stimuli.
Memory deficits in TLE patients following a temporal lobe resection
Whilst the lateralising significance o f deficits on some o f these verbal tests is unclear in unoperated TLE patients, these tasks do appear to be more sensitive in
discriminating left from right TLE patients postoperatively (Noveliy et al, 1984; Goldstein et al, 1988; Saykin et al, 1989; Loring et al, 1991, Saling et al, 1993, Inoue et al, 1992)^^^' i .351. i ?i although again, there is some inconsistency regarding the nature o f the verbal memory deficit, with some authors emphasising poor learning (Inoue et al, 1992; Helmstaedter & Eiger, 1996; Goldstein et al, 1988,; Ribbler & Rausch, 1990; Loring et al, 1991; Hermann et al, 1992; Saling et al,
1993)171,142,127,346,231,150,351 others stress impairments in immediate and delayed recall (Powell et al, 1985; Channon et al, 1989; Frisk & Milner, 1990;1991; Noveliy et al, 1994; Helmstaedter et al, 1 9 9 5.1 9 9 6)32 3,59,118,117,2 9 5,144,142 ^i^^ reports o f visual memory deficits in right temporal lobectomy patients are somewhat more consistent but by no means unanimous. Highly specific memory deficits have also been reported in postoperative patients following the removal o f temporal lobe tumours (Cavazutti etal, 1980/^
In the light o f this, at best equivocal evidence, why do these old tests continue to be clinically employed on such a grand scale, particularly in the latéralisation o f
Introduction Chapter 3
using these tests. Although new and imaginative tests have evolved, these are
necessarily developed in a research setting and only rarely seem to make the transition from research tool to wide clinical use (but see Helmstaedter
. As a result there is often sparse data on the preoperative sensitivity o f many o f the more innovative memory tests that have been shown to be sensitive to lateralised temporal lobe damage in postoperative patients.
Overall the results from these studies suggest that tests employing strongly polarised (verbal vs. non verbal) test materials appear to have the most sensitivity in localising memory deficits in both the pre- and postoperative patient populations. Learning tasks that incorporate an immediate and delayed recall condition appear to be most sensitive to left temporal lobe dysfunction when verbal material is employed and right temporal lobe dysfunction when visual material is used. The strategy o f using standardised
matched tasks to study analogous brain regions in the right and left hemispheres adds
power to the neuropsychologists’ ability to localise dysfunction and to analyse the nature o f a patients deficits.
Recognition Memory
The Warrington Recognition Memory Test (RMT)"^^^ is probably the most widely used test o f recognition memory for verbal (visually presented words) and non-verbal material (faces), although new tests that overcome some o f the difficulties o f the WRMT have recently been developed, for example the Doors and Faces Test (Baddeley et al, 1995)*^. Patients with left temporal lobe lesions (primarily infarcts and neoplasms) have been shown to be impaired on the verbal component (RMW) whilst patients with right temporal lobe lesions are impaired on the visual component (RMF) (Warrington, 1984)"*^^ . These deficits have also been reported in postoperative temporal lobectomy patients (Morris and Abrahams, 1995; Hermann et al,
TLE patients preoperatively is poor. Hermann et al (1995)^"^^ examined the pre- and postoperative RMT performance o f 77 temporal lobectomy patients. Patients with structural lesions (other than hippocampal sclerosis) evident on MRI were excluded from the study. Preoperatively, no aspect o f the RMT could reliably distinguish the right from left TLE groups. Examination o f pre-postoperative memory change showed a decline in verbal recognition memory after left temporal lobectomy and a