Figure 1. Group mean ± SEM scores on long-term retrieval for subjects who were positive (ε4 positive) and negative (ε4 negative) for the apolipoprotein E ε4 allele in all conditions. The scores represent the number of words recalled without reminding. There was a significant effect of dose in that fewer items were recalled in the high-dose lorazepam condition compared with the low-dose lorazepam and placebo conditions. There was also a significant effect of time with performance highest at baseline, lowest at 1 hour, and increasing at 2.5 and 5 hours. There was no main effect of apolipoprotein E ε4 allele status, but this status interacted with dose and time.
Figure 2. Group mean ± SEM scores on short-term retrieval for subjects who were positive (ε4 positive) and negative (ε4 negative) for the apolipoprotein E ε4 allele in all conditions. The scores represent the number of words recalled immediately after reminding. The results are complementary to those observed for long-term retrieval.
Figure 3. Group mean ± SEM scores on long-term retrieval in the high-dose condition as a function of apolipoprotein E ε4 allele status and baseline performance. Subjects were divided into those with high (above the median) and those with low (below the median) baseline performance. The scores represent the number of words recalled without reminding. The delayed-recovery effect at 5 hours was only observed in the low-baseline subgroup that was positive for the allele (ε4-positive). ε4-Negative indicates the group that was negative for the allele.
Pomara N, Willoughby L, Wesnes K, Greenblatt DJ, Sidtis JJ. Apolipoprotein E ε4 Allele and Lorazepam Effects on Memory in High-Functioning Older Adults. Arch Gen Psychiatry. 2005;62(2):209-216. doi:10.1001/archpsyc.62.2.209
Copyright 2005 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2005
The apolipoprotein E (APOE) ε4 allele has been implicated as a significant risk factor in the development of late-onset Alzheimer disease, but the evidence of cognitive sequelae in healthy individuals has been mixed.
To determine if the APOE ε4 allele increases susceptibility to lorazepam-induced verbal learning impairment in nondemented older adults.
A placebo-controlled crossover design.
A community-based sample of subjects.
Sixty-four cognitively intact and highly educated (>12 years) adults. Twenty-four subjects (mean age, 66.3 years) were carriers of an APOE ε4 allele (ε4 positive) and 40 (mean age, 66.0 years) were not (ε4 negative).
All subjects received a single oral dose of placebo and lorazepam (0.5 and 1.0 mg) 1 week apart.
Main Outcome Measure
We used the Buschke Selective Reminding Test to assess verbal learning during a 5-hour period after placebo or lorazepam administration.
We found a time-related, dose-dependent effect of lorazepam, with long-term recall generally decreasing with higher doses of lorazepam at up to 2.5 hours. At 5 hours, the ε4-negative group showed significant improvement in long-term memory, but the ε4-positive group demonstrated a persistent deficit. Subsequent analysis revealed that the poor performance at 5 hours was found in an ε4-positive subgroup with lower baseline performance.
In cognitively intact, older adults, the effect of the APOE ε4 allele is not necessarily seen in the immediate response to benzodiazepine challenge. Rather, the APOE ε4 allele appears to affect the carrier’s ability to recover from a cognitive challenge in a normal fashion, at least in a subgroup of subjects with relatively low baseline performance. This suggests that although carrying an APOE ε4 allele increases the risk for cognitive toxic effects, allele status alone is not a sufficient predictor of such effects. Studying the response to and the recovery from cognitive challenges may provide insights into the role of the APOE ε4 allele and its interaction with other factors in the development of Alzheimer disease and other age-related cognitive problems.
The identification of susceptibility factors for drug-induced cognitive toxic effects in older adults is of considerable significance, because centrally acting medications such as the benzodiazepines continue to be widely prescribed in this population, and older individuals may be more sensitive to some of the adverse central effects.1,2 Moreover, normal age-related decline in cognitive functions might accentuate drug-induced deficits. In the case of benzodiazepines, other studies from our group have demonstrated that the correlations between drug-induced cognitive toxic effects and plasma drug levels, although statistically significant, were modest, suggesting that subject characteristics such as genetic differences may play a role in determining sensitivity to drug effects.3,4
One of the likely genetic factors in the sensitivity to drug effects is the apolipoprotein E (APOE) ε4 allele. This particular form of the APOE gene has been demonstrated to be a significant risk factor for the development of sporadic and late-onset familial Alzheimer disease (AD).5 In addition, in normal or in minimally impaired older adults, the APOE ε4 allele has been associated with poor performance on delayed verbal recall tests (the best discriminator of early AD from normal aging),6- 9 greater sensitivity to anticholinergic challenge,10 decreased cerebral glucose metabolism on positron emission tomography findings,11,12 greater hippocampal atrophy,13- 15 increased mortality,16- 21 reduced recovery of neurological function after various metabolic and traumatic brain insults,22- 25 more persistent memory deficits after cardiopulmonary bypass surgery,26 and chronic daytime somnolence.27
Although it is generally thought that the presence of an APOE ε4 allele has a negative impact on cognitive function, especially memory, there is variability in the literature.28,29 Although some studies have shown that the presence of an ε4 allele predicts cognitive decline,8,28,30- 34 others have shown that the APOE ε4 allele predicts changes only on certain tests7,35,36 or only when the Mini-Mental Status Examination score is below a certain level.37 No differences as a function of APOE ε4 status have been reported,38,39 and lower scores but not progressive cognitive decline as a function of APOE ε440 have also been reported. The APOE ε4 allele appears to be a risk factor for cognitive loss, but its effects appear to be subtle in many situations, especially with nondemented subjects. The use of pharmacological challenges to study the effects of an APOE ε4 allele has the potential to enhance otherwise subtle differences between subjects with (ε4-positive subjects) and without (ε4-negative subjects) the APOE ε4 allele.
A study from our group previously examined the relationship between the APOE ε4 allele and sensitivity to benzodiazepine-induced cognitive toxic effects in older adults.41 Whereas normal, cognitively intact ε4-positive subjects showed impaired performance on immediate and delayed verbal recall at 1 and 2.5 hours after an acute, single 1.0-mg oral dose of lorazepam, ε4-negative subjects experienced impairment only at 1 hour after drug intake. That study suggested that the APOE ε4 allele may increase susceptibility to drug-induced impairment, but it had a relatively small number of subjects, and APOE ε4 status was determined by means of phenotyping, which is known to be less accurate than genotyping. The present study was designed to extend this observation in a larger group of normal, cognitively intact, high-functioning, older subjects whose APOE ε4 status was determined by genotyping.
The study was approved by the institutional review board at the Nathan S. Kline Institute for Psychiatric Research, Orangeburg, NY, and performed according to the Declaration of Helsinki. All participants provided written consent, and confidentiality was strictly enforced. Sixty-four cognitively intact older adults meeting the educational requirements were recruited through advertisement. Subjects had to have had at least some college education and be free of cognitive impairment. All subjects were free of significant neurological and medical illnesses as determined by results of a medical history, physical examination, and routine laboratory tests that were conducted 1 week before the beginning of the experiment. No subjects reported using neuroleptics, antidepressants, or any other medications known to affect cognitive functioning for at least 2 weeks before beginning the study. Results of a urine toxicology screen were negative for all participants. Psychiatric evaluations were conducted, and none of the subjects met DSM-IV criteria for a psychiatric disorder.42 Subjects were free of significant anxiety or depressive symptoms on the Hamilton Anxiety Scale43 and Hamilton Rating Scale for Depression.44 Subjects were also free of significant cognitive and memory impairments, as determined by the general memory index score of the Wechsler Memory Scale–Revised,45 vocabulary subtest of the Wechsler Adult Intelligence Scale–Revised,46 and the Mini-Mental Status Examination.47 Ages ranged from 60 to 75 years. The demographic characteristics and prestudy screening measures are summarized in Table 1. All subjects were paid $200 at the end of the study.
The APOE genotyping was performed at the Northwest Research Laboratories at the University of Washington, Seattle, using the restriction enzyme isoform genotyping method.48 Twenty-four subjects were ε4 positive, and 40 were ε4 negative. The ε4-positive group consisted of 3 ε4/ε2, 18 ε4/ε3, and 3 ε4/ε4 carriers; the ε4-negative group, 10 ε3/ε2 and 30 ε3/ε3 carriers. Subjects were not informed of their genotype.
We found no significant differences in levels of high- or low-density lipoprotein cholesterol, diastolic or systolic blood pressure, or white blood cell count between the APOE ε4 groups. Triglyceride levels (mean ± SD) were greater in the ε4-positive group (214.4 ± 158.4 mg/dL [2.4 ± 1.8 mmol/L]) than in the ε4-negative group (139.9 ± 61.6 mg/dL [1.6 ± 0.7 mmol/L]) (t62 = −2.67 [P = .001]).
After completion of the diagnostic evaluation, individuals participated in 3 sessions, each 1 week apart, during which they were administered oral doses of placebo or lorazepam (0.5 or 1 mg) in a double-blind, randomized design. The order of the doses was counterbalanced. Each test session began at approximately 9 AM under nonfasting conditions after a brief history and vital signs were obtained.
A repeated battery of psychometric tests, which included the Buschke Selective Reminding Test,49,50 was administered at baseline and 1, 2.5, and 5 hours after drug administration. Blood samples for quantitation of drug level were collected before each psychometric assessment and were determined by means of electron-capture gas chromatography.51,52
The Buschke Selective Reminding Test assesses a wide range of memory functioning. Several measures were examined, including total recall, long- and short-term retrieval, and delayed recall. Our Buschke Selective Reminding Test consists of a 16-noun list read to subjects at a rate of 1 word every 2 seconds. Subjects were instructed to recall as many words as possible. When no more words could be recalled, the subject was read the words that were omitted. Subjects were then instructed to recall the entire list of 16 words. This procedure was repeated until a maximum of 7 list recall trials had been completed. Total recall was the total number of words correctly recalled across learning trials. Short-term retrieval was the number of words recalled immediately after reminding. All other recalled words made up the long-term retrieval score. Delayed recall was the number of words recalled 15 minutes after the initial learning phase.
The study compared the time course of the effects of a single 0.5- or 1.0-mg dose of lorazepam or placebo on verbal memory. Each subject participated in all drug conditions. The presence or absence of an APOE ε4 allele was the between-subjects factor. Dose (placebo or 0.5 or 1.0 mg of lorazepam) and time (baseline and 1, 2.5, and 5 hours after the dose) were repeated measures.
The 12 versions of the Bushke Selective Reminding Test were normalized using the mean values for each of the versions in the placebo conditions. The highest mean value was used as the numerator, and the mean scores for each of the versions were used as denominators for the normalization.53 These normalization values were then applied to each version across treatment conditions. Recall scores were subjected to pairwise t test comparisons, and multiple factors were analyzed using repeated-measures analyses of variance, with dose and time serving as repeated measures and APOE ε4 status serving as the between-subject factor for all comparisons. P<.05 was considered significant.
There were no significant differences in lorazepam levels between the ε4-negative and ε4-positive groups across dose and time conditions. The mean levels for 1, 2.5, and 5 hours for each condition are as follows: 5.5 ± 3.0, 5.8 ± 1.5, and 4.5 ± 1.2 ng/mL, respectively, for the ε4-negative low-dose group; 9.1 ± 5.7, 11.8 ± 2.2, and 9.2 ± 1.6 ng/mL, respectively, for the ε4-negative high-dose group; 5.2 ± 3.3, 6.2 ± 2.4, and 4.5 ± 1.9 ng/mL, respectively, for the ε4-positive low-dose group; and 10.4 ± 5.2, 12.8 ± 3.1, and 9.3 ± 3.4 ng/mL, respectively, for the ε4-positive high-dose group (Unless otherwise indicated, all data are expressed as mean ± SD).
Baseline values for the ε4-negative and ε4-positive groups for each drug condition were compared by means of independent-groups t tests. Although there was a tendency toward lower baseline performance by the ε4-positive group for total recall in the low-dose condition and for total recall and delayed recall in the high-dose condition, none of the ε4 group baseline comparisons were significant. Placebo condition scores were also evaluated using a repeated-measures analysis of variance examining APOE ε4 status and time as factors. There were no significant main effects of APOE ε4 status on any measure during the placebo condition. However, we found the following significant main effects of time for each measure in the placebo condition: total recall (P<.001), long-term retrieval (P = .002), short-term retrieval (P = .03), and delayed recall (P<.001). For total recall, long-term retrieval, and delayed recall, this effect reflected relatively higher performance at baseline and lower performance at subsequent time points in the placebo condition. For short-term retrieval, the opposite effect was observed. Time and APOE ε4 group status interacted on total recall (P = .046) and delayed recall (P = .007). These interactions reflected greater declines from baseline during subsequent time points for the ε4-positive group compared with the ε4-negative group in the placebo condition.
There were significant effects of dose (F1,62 = 24.96 [P<.001]) and time (linear F1,62 = 56.506 [P<.001]; quadratic F1,62 = 137.00 [P<.001]) such that total recall scores were lowest in the high-dose condition, intermediate in the low-dose condition, and highest in the placebo condition. With respect to time, performance was highest at baseline, lower at 1 and 2.5 hours, and increasing at 5 hours. Total recall scores are presented in Table 2. There was also an interaction between dose and time (linear F1,62 = 12.49 [P<.001]; quadratic F1,62 = 15.04 [P<.001]) such that placebo performance was fairly flat between 1 and 5 hours, but high-dose performance dropped until 2.5 hours and started to improve by 5 hours. There was no significant main effect of APOE ε4 status, but APOE ε4 status interacted with dose and time (quadratic F1,62 = 6.85 [P = .01]), reflecting a trend toward poorer performance by the ε4-positive group (46.21 ± 11.61) compared with the ε4-negative group (52.50 ± 13.15) in the high-dose condition at 5 hours (t53.41 = 2.00 [P = .05]).
The long-term retrieval results are presented in Figure 1. As with total recall, there were significant effects of dose (F1,62 = 32.61 [P<.001]) and time (linear F1,62 = 62.38 [P<.001]; quadratic F1,62 = 80.46 [P<.001]; cubic F1,62 = 12.50 [P = .001]) such that long-term retrieval scores were lowest in the high-dose condition, intermediate in the low-dose condition, and highest in the placebo condition. With respect to time, long-term retrieval scores were highest at baseline, lowest at 1 hour, and increased at 2.5 and 5 hours. The interaction between dose and time (linear F1,62 = 17.37 [P<.001]; quadratic F1,62 = 11.94 [P = .001]) was similar to that observed for total recall. There was no significant main effect of APOE ε4 status, but APOE ε4 status interacted with dose and time (quadratic F1,62 = 6.476 [P = .01]). As with total recall, the interaction reflected an APOE ε4 effect at 5 hours in the high-dose condition. However, this interaction with the long-term retrieval measure was more reliable than with total recall. The ε4-positive group had significantly lower long-term retrieval scores (18.2 ± 15.2) (t49.31 = 2.09 [P = .04]) than the ε4-negative group (26.43 ± 15.50) in the high-dose condition at 5 hours.
The short-term retrieval results are presented in Figure 2. The pattern of results is complementary to that observed for long-term retrieval. There were significant effects of dose (F1,62 = 10.69 [P = .002]) and time (linear F1,62 = 43.56 [P<.001]; quadratic F1,62 = 10.52 [P = .002]; and cubic F1,62 = 6.81 [P = .01]) such that short-term retrieval scores were lowest in the placebo condition (22.5 ± 7.4), intermediate in the low-dose condition (23.7 ± 6.8), and highest in the high-dose condition (24.4 ± 6.4). Performance was lowest at baseline (20.7 ± 7.4), higher at 1 (24.1 ± 7.0) and 2.5 hours (24.2 ± 6.8), and highest at 5 hours (25.2 ± 6.4). Dose and time also interacted (F1,62 = 7.097 [P = .01]) such that performance was fairly flat over time in the placebo condition but increased over time in the high-dose condition. There was no significant main effect of APOE ε4 status, but as with the other measures, APOE ε4 status interacted with dose and time (quadratic F1,62 = 7.10 [P = .01]). As a complement to the long-term retrieval measure, this interaction reflected a trend toward higher scores by the ε4-positive group (27.9 ± 5.8) compared with the ε4-negative group (26.1 ± 4.7) in the high-dose condition at 5 hours. However, no results of the individual comparisons were significant.
The delayed-recall results were similar to those observed for total recall and long-term retrieval and are presented in Table 3. There were significant dose (F1,62 = 26.91 [P<.001]) and time effects (linear F1,62 = 148.82 [P<.001]; quadratic F1,62 = 190.26 [P<.001]) and their interaction (linear F1,62 = 9.16 [P = .004]; quadratic F1,62 = 7.16 [P = .01]). There was no significant main effect of APOE ε4 status, but APOE ε4 status interacted with dose and time (quadratic F1,62 = 7.16 [P = .01]). None of the individual APOE ε4 comparisons in this interaction were significant.
Given the tendency of ε4-positive subjects to have lower baseline performance, an additional comparison was performed on the long-term retrieval results. The question was whether the poor performance at 5 hours in the high-dose condition was a function of APOE ε4 status or low baseline function. Subjects were classified as low (<42 items recalled) or high (≥42 items recalled) baseline performers according to their baseline scores on long-term retrieval, independent of APOE ε4 status. For the high-baseline group, there was no difference in long-term retrieval scores 5 hours after the high dose when ε4-negative subjects (33.0 ± 15.3; n = 24) were compared with ε4-positive subjects (32.6 ± 13.6; n = 9). In contrast, for the low-baseline group, the ε4-negative subjects (16.5 ± 9.4; n = 16) performed significantly better (t28.61 = 2.23 [P = .03]) than the ε4-positive subjects (9.5 ± 7.9; n = 15). This comparison is presented in Figure 3. There were no significant differences between the high- and low-baseline ε4-positive subjects with respect to age, education, or Mini-Mental Status Examination or vocabulary scores.
The results of this study demonstrate a time-related, dose-dependent effect of lorazepam on memory function. Recall decreased with higher doses of lorazepam, and the decrease was greatest at 2.5 hours after the administration of the drug. Five hours after lorazepam was taken, recall was significantly improved compared with 2.5 hours. The effect of lorazepam was most evident when recall was temporally separated from reminding (long-term retrieval and delayed recall). Words recalled closer to reminding (short-term retrieval) demonstrated a complementary effect, demonstrating that the benzodiazepine effect was not simply globally depressed performance. This inverse relationship between long-term and short-term recall has been reported after benzodiazepine challenge and is similar to the relationship between these memory processes observed in progressive dementia.54 The presence of an APOE ε4 allele modulated this general pattern.
Comparisons of ε4-positive and ε4-negative subjects indicated that the dose × time interactions differed significantly between these groups. The most consistent APOE ε4 effect was found in long-term recall performance at 5 hours in the high-dose condition. At that time point, the ε4-negative group showed significant improvement after the maximum drug impairment 2.5 hours earlier, but the ε4-positive group demonstrated a persistent deficit. Short-term recall actually increased in the high-dose condition, and the increase was greatest at 5 hours for the ε4-positive subjects.
An additional analysis compared the effects of baseline levels of performance and APOE ε4 status. The results suggested that persistent impairment was a function of both the APOE ε4 allele and low baseline performance. The remaining difference between APOE ε4 groups was a subtle one; ie, performance declined slightly during the course of the day in the placebo condition in the ε4-positive but not the ε4-negative group.
The neurobiological mechanism by which the APOE ε4 allele places an individual at risk for cognitive loss is unknown. However, with the previously noted associations between the APOE ε4 allele and increased mortality, reduced recovery of neurological function after various metabolic and traumatic brain insults, and more persistent memory deficits after cardiopulmonary bypass surgery, the present results suggest an abnormality in the ability of the central nervous system to respond to insult. In the present study, the benzodiazepine challenge served as a temporary, reversible, central nervous system insult that produced impaired long-term memory function in the ε4-positive and ε4-negative subjects. The effect of an APOE ε4 allele was not seen in the initial response to the benzodiazepine, but rather in a retarded recovery. Whether due to reduced plasticity55 or some other process, a reduced capacity for recovery could certainly lead to progressive functional decline. The consequences of such a reduced capacity would also be variably expressed in the APOE ε4 population because its expression would also be the result of the number, nature, and extent of the physiological challenges previously experienced by an individual’s central nervous system. The low-baseline ε4-positive subgroup may represent individuals who have a particularly increased risk for cognitive decline for reasons as yet unknown.
These results also demonstrate another potential difficulty in studying a group at risk for diminished cognitive function with a demanding protocol. During the course of the daylong evaluation, the ε4-positive group’s performance declined at 1 and 2.5 hours but returned to baseline by 5 hours in the placebo condition. Although this effect complicated the analyses, it did not confound the main finding that the ε4-positive group failed to improve by 5 hours in the high-dose condition. However, it suggests that at least in some individuals, the APOE ε4 allele may be associated with reduced cognitive reserve. This temporal effect may well reflect the same process that was reported in the study of chronic daytime somnolence and APOE ε4 status.27 The earliest effects of the APOE ε4 allele may not be cognitive, and the earliest cognitive effects may require pharmacological challenge to be revealed. General temporal effects like chronic daytime somnolence, together with specific dose × time × APOE ε4 status interactions, may also account for some of the variability in the literature. In cognitively intact individuals, the recovery from a pharmacological cognitive challenge may be a more sensitive measure of the effect of an APOE ε4 allele than the initial response to the challenge itself.
Finally, the specificity of the delayed-recovery APOE ε4 effect should be addressed. A recent report found a similar effect after an anticholinergic challenge.10 One of the many questions that remains is whether delayed recovery after pharmacological challenge is a general characteristic of ε4-positive subjects. If so, is there a common neurobiological mechanism or is recovery slowed by a more systemic process? In either case, studies of the response to pharmacological challenges may provide insights into what makes the APOE ε4 allele a risk factor for cognitive decline.
These results raise a number of important questions. Does the low baseline performance associated with poor recovery in the ε4-positive subgroup represent a recent or long-standing problem? Does the combination of low baseline performance and an APOE ε4 allele predict progression to mild cognitive impairment or AD? Do the results of pharmacological challenges have greater prognostic value than the results of baseline testing? The present results do not answer these questions. We can speculate that the low baseline performance in the poor-recovery ε4-positive subgroup represents an insidious process, because the groups were well matched for education and basic intellectual attainment (arguing against a long-standing problem), and none complained of recent cognitive changes. The relative prognostic values of APOE ε4 status, response to this pharmacological challenge, and baseline performance remain to be established. Given the difficulty of predicting cognitive decline, it is likely that each of these factors may provide useful information. Moreover, additional study is necessary to determine if the cognitive toxic effects associated with other pharmacological challenges such as that produced by anticholinergic drugs, which we have also found to be modulated by the APOE ε4 allele, can identify elderly individuals at increased risk for subsequent cognitive decline and individuals with mild cognitive impairment at increased risk for conversion to AD.
This study demonstrated a time- and dose-dependent negative effect of lorazepam on memory function, as has been reported previously. More importantly, this study demonstrated a significant difference between ε4-positive and ε4-negative subjects in their recovery time from this cognitive challenge. This was especially true in a subset of ε4-positive subjects with low baseline performance on memory testing. These results suggest that before the onset of cognitive problems in older adults, the presence of an APOE ε4 allele may reduce the capacity of the central nervous system to respond to insult. Studying the time course of the response to pharmacological cognitive challenges may provide insight into the preclinical consequences of an ε4 allele in older adults.
Correspondence: Nunzio Pomara, MD, Geriatric Psychiatry Program, Nathan S. Kline Institute for Psychiatric Research, 140 Old Orangeburg Rd, Bldg 35, Orangeburg, NY 10962 (Pomara@nki.rfmh.org).
Submitted for Publication: August 4, 2003; final revision received September 7, 2004; accepted September 16, 2004.
Funding/Support: This study was supported by grant MH056994 from the National Institute of Mental Health, Rockville, Md.