Learning and Recall in Subjects at Genetic Risk for Alzheimers Disease

J. Gene Chen, B.S. Christopher L. Edwards, Ph.D. Suman Vidyarthi, B.S. Suresh Pitchumoni, B.S. Sara Tabrizi, B.S. Dan Barboriak, M.D. H. Cecil Charles, Ph.D. P. Murali Doraiswamy, M.D. Decits in delayed recall of learned information may be an early marker of Alzheimers disease (AD). The apolipoprotein E E4 allele and a positive family history (FH) are both genetic risk factors for AD. The authors cross-sectionally compared performance on the California Verbal Learning Test (CVLT) in 153 prospectively recruited normal elderly subjects (mean age 67 years, mean MMSE 28) stratied by genetic risk into four groups (E4 /FH , E4 /FH, E4/FH , E4/FH). Neither FH nor E4 status affected performance, except on List B (a distraction word list), on which the FH group performed worse. The high-risk group (E4 /FH ) also performed worse on List B than the low-risk group (E4/FH) but did not differ on other measures. Memory impairments associated with genetic or family history risk may not manifest until the person is much closer to the onset age of AD.
(The Journal of Neuropsychiatry and Clinical Neurosciences 2002; 14:5863)

here is great interest in studying cognitive function in subjects at risk for Alzheimers disease (AD).1 The apolipoprotein E E4 (APOE E4) allele is a risk factor for the development of AD, and subjects carrying the E4 allele have an earlier onset of AD.2 Decits in learning and delayed recall are reported to be among the most sensitive clinical markers of the early stages of AD.36 For example, a 13-year prospective study of 1,045 subjects reported that measures of verbal memory were predictive of the development of AD.5 Measures of delayed recall are currently being used as one of the criteria to select individuals for dementia prevention trials. The California Verbal Learning Test (CVLT) is a psychometric assessment designed to measure list learning and verbal memory ability.7,8 Unlike tests that ask a subject to recall stories or remember certain tasks, the CVLT does not provide a semantic or procedural context in which the subject can organize the information presented. Thus it measures explicit or episodic memory9 in adolescent or adult subjects of any age. In addition to simple recall indices, CVLT results can reveal the underlying mechanisms that lead to changes in memory performance, including learning strategies, error types, and learning processes.10,11 CVLT scores correlate highly with both the Verbal Comprehension Index of the Wechsler Adult Intelligence Scale-III and the Auditory
Received June 8, 2000; revised February 21, 2001; accepted March 7, 2001. From the Departments of Psychiatry and Medicine (Geriatrics), Center for the Study of Aging and Human Development, Duke University Medical Center, Durham, North Carolina. Address correspondence to Dr. Doraiswamy, Box 3018, Duke University Medical Center, Durham, NC 27710. E-mail: dorai001@mc.duke.edu. Copyright 2002 American Psychiatric Publishing, Inc.

58

J Neuropsychiatry Clin Neurosci 14:1, Winter 2002

CHEN et al. Memory and General Memory indices of the Wechsler Memory Scale-III.12 Normative CVLT data have been presented on healthy young,13 depressed,14 and elderly individuals.11 Paolo et al.11 studied a large sample of elderly subjects (N 212, age range 5594) and examined norms broken down by gender, age, and education, but not by APOE genotype. Studies of normative data on list-learning tests show that gender accounts for a considerable portion of the variance, with age and education accounting for much less variance in the prediction of cognitive performance.15 Although the APOE E4 allele itself is not necessarily associated with poorer memory performance,16,17 some studies have demonstrated that seemingly normal individuals with the E4 allele who are on the verge of developing AD show memory decits as much as two years before the onset of dementia.18,19 The purpose of this study was to compare CVLT performance cross-sectionally in an elderly sample prospectively stratied by two risk factors for dementia: APOE and family history. ple would have resulted in far fewer subjects with E4 allele and very few subjects E4 /FH or vice versa. A sample size of 160 was our target. APOE genotyping was performed with DNA extracted from buffy coat, using polymerase chain reaction (PCR) amplication and the HhaI restriction digest method of Hixson and Vernier.20 Assessment Instruments During each visit an experienced rater administered the Mini-Mental State Examination (MMSE), to screen for cognitive impairment, and the CVLT, a measure of list learning and verbal memory. The rater then scored the CVLT with computer software designed by the tests developers.7 The CVLT centers around a 16-item list, composed of four semantic categories of four words each. Adjacent words on the list are from different categories. Subjects may adopt a serial learning strategy by simply memorizing the words in order, or a semantic learning strategy by assigning the words to their respective semantic categories. The fact that the words are not prototypic examples of their categories increases the difculty (for example, slacks is in the clothing category, but pants is not). During a CVLT administration, the main 16-item list (List A) is presented ve times, generating a summary learning measure (the sum of scores of the ve trials), a recall consistency measure (the same words remembered across the trials), a learning rate (the number of words remembered plotted against trial number), and learning strategy measures (semantic and serial clustering ratios, and the percentage of words recalled from the primacy, middle, and recency regions). A second, different list (List B) designed to interfere with List A is then presented. Afterwards, the subject is asked to name the items from List A freely (short-delay free recall) and by category (short-delay cued recall). The subject then is distracted for 20 minutes, after which he or she is asked again to name the List A words in any order (long-delay free recall), and by category (long-delay cued recall). During any of the recall stages, the subject may name the same word more than twice (a perseveration) or name a word that was not on the original list (an intrusion). Finally, the administrator tests recognition memory by reading a longer list and asking the subject whether each word was on List A or not, generating measures of correct answers (recognition hits), identication of words absent from the list (false positives), and the ability to tell apart valid list words from nonlist words (discriminability). From these data, scoring software generates a list of contrast measures that indicate the rate of forgetting

METHODS
Patient Population Normal elderly subjects ages 55 and older were recruited from a registry of elderly volunteers interested in research, and by newspaper and yer advertisement, and gave written informed consent. The primary goal of the study was to examine the effects of APOE and family history status on memory and brain chemistry as measured by magnetic resonance (MR) spectroscopy. Respondents underwent a screening and a brain magnetic resonance imaging scan to rule out signicant neurological and psychiatric disorders. The MR scans are currently still being analyzed. All subjects possessed adequate vision and hearing, although some wore glasses and/or hearing aids. Individuals with dementia, depression, and past strokes were excluded from the study. Subjects with pacemakers or other metallic implants that prohibited them from undergoing an MRI were also excluded. Within the pool of subjects meeting our initial study criteria, emphasis was placed on recruiting equal numbers of subjects with and without the APOE E4 allele, and with and without a positive family history of dementia. Our goal was to recruit at least 25 subjects in each of the above four cells, with a target of about 40 per cell, so that we would have sufcient sample sizes in each cell to make comparisons of the memory and brain measures. This stratication required a two-step recruitment because the APOE alleles and FH are not evenly distributed in the population and a random sam-

J Neuropsychiatry Clin Neurosci 14:1, Winter 2002

59

SUBJECTS AT GENETIC RISK FOR AD across various stages during the test. We did not use the last contrast measure, percentage change from discriminability to long-delayed free recall. All in all, one CVLT administration generates 28 indices of verbal memory, learning ability, learning strategy, and forgetfulness. Statistical Methods To test the effects of demographic variables, subjects were divided into two groups around the median age, 65; two groups based on gender; three groups based on education level (completed high school education or less, completed or attempted college education, and schooling beyond college); two groups based on the presence or absence of a family history of dementia; and two groups based on the presence (E2/E4, n 4; E3/E4, n 59; E4/E4, n 9) or absence (E2/E2, n 0; E2/E3, n 9; E3/E3, n 72) of the APOE E4 allele. In accordance with the ndings of Small et al.19 and Bondi et al.,10 we also hypothesized that subjects without E4 alleles would have higher scores than participants with at least one E4 allele. Age, education, and gender effects were also predicted from the literature. General linear models, t-tests, and chi-square tests on SAS software were done. All analyses were two-tailed. All of our subjects were elderly and nondemented. Because any group differences were likely to be small, a testwise alpha level of 0.05 was used to determine signicance for the a priori hypotheses. sure, including the summary learning measure (t 4.4, df 152, P 0.001). Women also used semantic clustering more than men (t 2.5, df 152, P 0.05) and were better at discriminating words (t 2.3, df 152, P 0.05) and rejecting false positives (t 2.5, df 152, P 0.05) on the recognition portion of the test. CVLT performance among the education groups differed signicantly on only one measure, percent recall from the recency region (F 4.3, df 2,152, P 0.05). Table 1 displays CVLT scores grouped by APOE E4 and FH status. There were no signicant differences on any measures between the APOE E4 and non-APOE E4 groups. FH subjects were younger than FH subjects (Table 1). The only signicant CVLT difference between the FH and FH groups was on List B scores (t 2.6, df 151, P 0.05). Table 2 summarizes test performance for subjects grouped by both FH and APOE status. There were no differences overall. The only variable that demonstrated a signicant difference was once again List B, on which the E4/FH group performed better than the E4 /FH group (t 2.6, df 93, P 0.05). We then used a nonhierarchical linear model to examine the effects of age, gender, education, and genetic status on CVLT measures. After adjusting for the effects of the other variables, the effect of age or E4 status was not signicant on any of the recall indices. The effect of gender remained signicant (P 0.01) in this model for all recall indices as well as four derived measures. Five recall measures, including the summary learning measure (F 3.8, df 2, 152, P 0.05), the short delayed free recall measure (F 3.1, df 2, 152, P 0.05), and the long delayed free recall measure (F 4.0, df 2, 152, P 0.05) showed signicant differences by education status. Female gender and higher education levels were related to better performance.

RESULTS
In total, 153 subjects (77 males and 76 females) participated in the study. The age range was 55 to 85, with a mean ( SD) of 66.7 7 years. The mean number of years of education was 15.6 3 and the mean MMSE score was 28.4 1. Of all the subjects, 87 were family history negative (FH) and 66 were family history positive (FH ). In addition, 81 had no APOE E4 (APOE E4) alleles, and 72 had either one or two APOE E4 alleles (APOE E4 ). Subjects with ages greater than 65 years demonstrated a greater decline from List A Trial 1 to List B (t 2.1, df 152, P 0.05) and a greater increase from the short delayed free recall measure to the long delayed free recall measure (t 2.0, df 152, P 0.05) than younger subjects. The effect of age as a continuous variable on CVLT performance was also examined. Both the summary learning measure (F 4.8, df 1,152, P 0.05) and the percent recall from the recency region (F 4.1, df 1,152, P 0.05) declined with age. Women scored higher than men on every recall mea-

DISCUSSION
We present CVLT data for a normal, nondemented elderly population stratied by two genetic risk factors for AD. This is also the rst report to systematically examine the effects of genetic risk status on CVLT performance in a relatively large aging sample. Aside from the CVLT manual,7 only one study, Paolo et al.,11 has presented normative CVLT data for elderly persons. Our ndings conrm this report. Examining age as a continuous variable, Paolo et al. reported an age effect on all eight CVLT recall indices. Although we found a statistically signicant negative age effect only on the CVLT summary learning measure, Paolo and colleagues group was older (mean age 70.6 7 years) than ours (65.8 7 years). Since we had a more limited age range,

60

J Neuropsychiatry Clin Neurosci 14:1, Winter 2002

CHEN et al. we analyzed our data with age split at age 75 (not shown), and these analyses tended to conrm the ndings by Paolo et al. Like those authors, we found that women consistently scored higher than men on both the recall and recognition measures. Likewise, we conrmed their ndings that higher education is associated with better performance on many recall indices. Thus, these ndings emphasize that the CVLT scores in elderly subjects, like many other cognitive tests, must also be interpreted in the context of age, education, and gender. This study extends the report of Paolo et al. by also simultaneously examining the effects of family history of dementia and APOE genotype, using a planned stratied recruitment design. In our population, subjects with a family history of dementia performed worse on only one CVLT measure (List B), and subjects with at least one E4 allele did not perform worse on any measures than those without such risk factors. List B is a distraction list, and hence the clinical signicance of this nding is puzzling. It could reect a spurious nding
TABLE 1. Variable Age, years, mean SD Male/Female CVLT measures, mean SD Recall measures (number correct) List A trials 15 total List A trial 1 List A trial 5 List B List A short-delay free recall List A short-delay cued recall List A long-delay free recall List A long-delay cued recall Learning characteristics, List A trials 15 Semantic cluster ratio (observed/expected) Serial cluster ratio (observed/expected) % correct recall from primacy region % correct recall from middle region % correct recall from recency region Slope (increments in words recalled per trial) % recall consistency across trials 15 Recall errors, Lists A and B Perseverations (free and cued recall total) Free recall intrusions (total) Cued recall instrusions (total) Intrusions (free and cued recall total) Recognition measures Recognition hits Discriminability False positives Contrast measures List B compared with List A, trial 1 (% change) Short-delay free recall vs. List A, trial 5 (% change) Long-delay free recall vs. short-delay free recall (% change) Recognition hits vs. long-delay free recall (% change) *FH vs. FH , P 0.05.

(due to multiple comparisons) or could reect a true subtle difference in cognitive capacity. In a prior study, Bondi et al.10 reported that E4-positive elderly subjects tested worse on nine different CVLT measures. Bondi and colleagues subject pool was smaller (17 E4 , 35 E4 in Bondi; 72 E4 , 81 E4 in our study) and 6 years older on average (mean age 71 in Bondi study; mean age 66 in our study). Age differences may explain both our results, since E4-positive subjects would be expected to perform worse if tested closer to the onset of dementia. Alternatively, our sample consisted of healthy volunteers and the differences may reect a sampling effect. There is interest in using tests of word learning and word recall to detect the early stages of dementia. In this respect, our results provide additional data that may enhance the interpretation of these tests in cognitively impaired patients. However, there are also some limitations to this report. We selected a group of subjects from an aging centers registry of volunteers and by advertisement. Our sample size may have been too small to

California Verbal Learning Test (CVLT) scores by presence or absence of APOE E4 alleles and family history (FH) No E4 Alleles (n 81) 68.29 6.85 38/43 49.38 11.58 7.01 1.96 11.60 2.73 6.84 2.47 9.93 3.19 11.21 2.80 10.37 3.11 11.01 2.85 2.24 1.94 29.86 41.62 28.54 1.08 79.64 5.52 2.23 2.15 4.38 0.94 1.51 5.58 7.08 6.37 0.55 11.79 4.95 2.35 2.81 4.33 1 E4 Allele (n 72) 65.20 7.09 39/33 48.81 6.69 11.49 6.15 9.75 11.00 10.31 10.81 2.08 2.15 30.08 41.37 28.51 1.09 80.75 9.82 1.66 2.33 1.97 2.77 2.51 2.89 2.61 0.87 1.98 6.41 8.65 6.26 0.54 10.75 FH (n 87) 66.98 7.01 45/42 49.26 6.84 11.70 6.92 9.99 11.16 10.34 10.90 2.24 2.01 30.51 41.49 27.97 1.12 80.66 11.72 2.13 2.62 2.48 3.01 2.73 2.96 2.80 0.89 1.76 6.35 7.69 5.74 0.51 11.13 FH (n 66) 64.18 6.61* 32/34 48.91 6.89 11.36 5.98 9.65 11.05 10.33 10.94 2.06 2.07 29.25 41.51 29.28 1.05 79.49 9.41 1.34 2.44 1.84* 2.99 2.59 3.08 2.67 0.92 1.73 5.37 8.06 6.96 0.60 11.55

5.82 4.98 2.45 2.61 1.89 1.80 4.34 3.87 14.49 1.91 92.35 8.43 1.87 3.06 0.01 0.37 7.77 27.43 15.54 17.38 6.67 22.66 59.29 89.38

5.23 4.50 2.33 2.66 2.30 2.74 4.63 4.61 14.39 1.73 92.83 6.56 1.54 1.90 0.01 0.32 3.71 28.30 16.20 17.60 6.70 22.95 54.51 50.89

6.23 5.48 2.34 2.22 1.66 1.78 4.00 3.30 14.73 1.84 92.95 7.81 1.84 3.04 0.04 0.37 11.07 29.99 16.02 19.96 12.05 46.72 62.96 91.83

14.58 1.66 93.35 5.66 1.49 1.73 0.01 0.32 6.08 30.75 16.64 19.66 11.02 43.21 57.02 49.96

J Neuropsychiatry Clin Neurosci 14:1, Winter 2002

61

SUBJECTS AT GENETIC RISK FOR AD detect subtle differences, and the higher education level may also have obscured genetic differences on test scores. We ran multiple analyses and did not adjust for multiple comparisons. Most of these analyses were preplanned, and the P-values reported should be interpreted with that in mind. The strengths of this study are the careful selection of subjects and the availability of APOE genotyping. These results must be interpreted accordingly. In summary, age, gender, and educational level have signicant, but variable, effects on the CVLT. Neither family history of dementia nor APOE E4 status had a
TABLE 2. Variable Age, years, mean SD Male/Female CVLT measures (mean SD) Recall measures (number correct) List A trials 15 total List A trial 1 List A trial 5 List B List A short-delay free recall List A short-delay cued recall List A long-delay free recall List A long-delay cued recall Learning characteristics, List A trials 15 Semantic cluster ratio (observed/expected) Serial cluster ratio (observed/expected) % correct recall from primacy region % correct recall from middle region % correct recall from recency region Slope (increments in words recalled per trial) % recall consistency across trials 15 Recall errors, Lists A and B Perseverations (free and cued recall total) Free recall intrusions (total) Cued recall instrusions (total) Intrusions (free and cued recall total) Recognition measures Recognition hits Discriminability False positives Contrast measures List B vs. List A trial 1 (% change) Short-delay free recall vs. List A trial 5 (% change) Long-delay free recall vs. short-delay free recall (% change) Recognition hits vs. long-delay free recall (% change) * P 0.05.

detectable consistent effect on any learning or memory measure. This nding may reect either the relative insensitivity of the CVLT or the fact that memory impairments in at-risk subjects may not occur until the person is much closer to the onset of AD. Community-based studies of subjects followed longitudinally will be needed to examine these issues further. This work was supported by the Paul Beeson award from the American Federation for Aging Research and the Novartis Scholar award to Dr. Doraiswamy.

California Verbal Learning Test (CVLT) scores across four groups broken down by APOE E4 status and family history (FH) E4/FH (n 55) 66.65 6.68 24/31 50.71 7.04 11.93 7.36 10.24 11.27 10.40 10.96 2.32 1.91 30.20 42.04 27.78 1.11 81.35 5.22 2.24 2.33 4.56 14.58 93.33 1.49 0.01 12.39 2.24 2.77 2.55 3.16 2.87 3.15 2.92 0.91 1.56 5.84 6.91 5.41 0.51 10.80 4.28 2.45 3.12 4.71 1.36 5.19 1.74 0.32 E4 /FH (n 40) 63.32 6.09* 18/22 50.43 6.85 11.61 6.15 9.90 11.03 10.35 10.83 2.05 2.12 29.31 42.03 28.69 1.06 81.79 6.28 2.41 1.59 4.00 14.85 92.67 2.08 0.06 8.21 15.71 5.74 68.23 9.32 1.42 2.32 1.81* 2.84 2.55 3.10 2.65 0.88 1.93 5.66 8.49 6.35 0.57 9.88 5.05 2.27 1.62 3.27 1.57 8.51 3.64 0.40 28.20 18.47 22.50 113.51

1.69 29.13 16.71 18.46 6.07 23.02 58.06 54.12

References

1. Doraiswamy PM, Steffens DS, Tabrizi S, et al: Early recognition of Alzheimers disease: Whats consensual? Whats controversial? Whats practical? J Clin Psychiatry 1998; 59(suppl 13):618 2. National Institute on Aging/Alzheimers Association Working Group: Apolipoprotein E genotyping in Alzheimers disease. Lancet 1996; 347:10911095

3. Welsh K, Butters N, Hughes J, et al: Detection of abnormal memory decline in mild cases of Alzheimers disease using CERAD neuropsychological measures. Arch Neurol 1991; 48:278281 4. Howieson DB, Dame A, Camocioli R, et al: Cognitive markers preceding Alzheimers dementia in the healthy oldest old. J Am Geriatr Soc 1997; 45:584589

62

J Neuropsychiatry Clin Neurosci 14:1, Winter 2002

CHEN et al.
5. Linn RT, Wolf PA, Bachman DL, et al: The preclinical phase of probable Alzheimers disease. Arch Neurol 1995; 52:485490 6. Peterson RC, Smith GE, Ivnic RJ, et al: Apolipoprotein E status as a predictor of the development of Alzheimers disease in memory-impaired individuals. JAMA 1995; 273:12741278 7. Delis DC, Kramer JH, Kaplan E, et al: The California Verbal Learning Test. New York, The Psychological Corporation, 1987 8. Fox LS, Olin JT, Erblich J, et al: Severity of cognitive impairment in Alzheimers disease affects list learning using the California Verbal Learning Test. Int J Geriatr Psychiatry 1998; 13:544549 9. Stout JC, Bondi MW, Jernigan TL, et al: Regional cerebral volume loss associated with verbal learning and memory in dementia of the Alzheimer type. Neuropsychology 1999; 13:188197 10. Bondi MW, Monsch AU, Glasko D, et al: Preclinical cognitive markers of dementia of the Alzheimer type. Neuropsychology 1994; 8:374384 11. Paolo AM, Troster AI, Ryan JJ: California Verbal Learning Test: Normative data for the elderly. J Clin Exp Neuropsychol 1997; 19:220234 12. The Psychological Corporation: WAIS-IIIWMS-III Technical Manual. San Antonio, TX, The Psychological Corporation, Harcourt Brace and Company, 1997 13. Wiens AN, Tindall AG, Croseen JR: California Verbal Learning Test: a normative data study. The Clinical Neuropsychologist 1994; 8:7590 14. Otto MW, Bruder GE, Fava M, et al: Norms for depressed patients for the California Verbal Learning Test: associations with depression severity and self-report of cognitive difculties. Arch Clin Neuropsychol 1994; 9:8188 15. Bolla-Wilson K, Bleecker ML: Inuence of verbal intelligence, sex, age, and education on the Rey Auditory Verbal Learning Test. Dev Neuropsychol 1986; 2:203211 16. Smith GE, Bohac DL, Waring SC, et al: Apolipoprotein e genotype inuences cognitive phenotype in patients with Alzheimers disease but not in healthy control subjects. Neurology 1997; 50:355362 17. OHara R, Yesavage JA, Kraemer HC, et al: The APOe e4 allele is associated with decline on delayed recall performance in community-dwelling older adults. J Am Geriatr Soc 1998; 46:1493 1498 18. Bondi MW, Salmon DP, Monsch AU, et al: Episodic memory changes are associated with the Apoe-eE4 allele in nondemented older adults. Neurology 1995; 45:22032206 19. Small BJ, Basun H, B ckman L: Three-year changes in cognitive performance as a function of apolipoprotein e genotype: evidence from very old adults without dementia. Psychol Aging 1998; 13:8087 20. Hixson JE, Vernier DT: Restriction isotyping of human apolipoprotein E by gene amplication and cleavage with Hha I. J Lipid Res 1990; 31:545548

J Neuropsychiatry Clin Neurosci 14:1, Winter 2002

63