Abstract
Introduction
Memory impairment is prevalent in multiple sclerosis (MS), 1 but there are no approved drugs to treat these memory problems. Animal research has identified l- and d-isomers of amphetamine as effective stimulators of norepinephrine release 2 and memory consolidation, 3 and the l-isomer achieves these effects at lower doses and with fewer adverse side effects. 3 In a recent clinical trial, l-amphetamine surpassed placebo in improving auditory/verbal and visual/spatial memory in MS. 4 However, given that processing speed was the primary outcome (based on pilot work 5 ), the trial was not optimized to detect improvements in memory. That is, while all patients enrolled in the study were defective on processing speed tests, many patients without memory deficits were enrolled, thus potentially diluting the positive effect of l-amphetamine on memory. With this in mind, we reanalyzed Morrow and colleagues’ 4 data to evaluate the effect of l-amphetamine on memory in MS patients with and without memory impairment.
Methods
Subject enrollment
Subject recruitment and random assignment are described in detail elsewhere. 4 Briefly, 151 persons with MS 6 were randomly assigned to treatment (30 mg l-amphetamine, N = 108) or placebo (N = 43) in a four-week, double-blind, parallel group, dose titration trial (ClinicalTrials.gov identifier: NCT00529581). The treatment group received an initial l-amphetamine dose of 5 mg, which increased to 15 mg after seven days, and 30 mg after another seven days. This 30 mg dose was maintained for 14 days. Memory testing occurred at baseline (day 0) and follow-up (day 29). Attrition was comparable across groups (final sample: 136 patients; treatment = 99; placebo = 37). Final sample characteristics: age (48.5 ± 8.5 years), education (14.7 ± 2.5 years), gender (105 females, 31 males), MS type (89 relapsing–remitting, 47 secondary-progressive). There were no demographic or disease differences between treatment and placebo groups (all p > 0.10).
This study was conducted across 36 sites within the USA, with approval by either an internal or central institutional review board, depending on the individual requirements of each site. Written informed consent was obtained from all subjects prior to participation.
Memory assessment
The California Verbal Learning Test, Second Edition (CVLT-II 7 ) assessed recall of 16 words across five learning trials (Total Learning) and again after a 25-minute delay (Delayed Recall). CVLT-II Delayed Recall was our primary measure of auditory/verbal memory. The Brief Visuospatial Memory Test, Revised (BVMT-R 8 ) assessed recall of six figures (and their spatial locations) across three learning trials (Total Learning) and again after a 25-minute delay (Delayed Recall). BVMT-R Delayed Recall was our primary measure of visual/spatial memory.
Statistical analyses
A median split of baseline CVLT-II Delayed Recall raw performance divided the sample into subgroups with lower and higher baseline auditory/verbal memory performance. A 2 (Lower Memory, Higher Memory) × 2 (l-amphetamine, placebo) × 2 (baseline, follow-up) ANOVA evaluated the effect of l-amphetamine on CVLT-II Delayed Recall. A secondary analysis was performed for CVLT-II Total Learning. Identical procedures evaluated the effect of l-amphetamine on visual/spatial memory, beginning with a median split of baseline BVMT-R Delayed Recall raw performance and a subsequent 2 × 2 × 2 ANOVA. A secondary analysis examined BVMT-R Total Learning. Thus, post-hoc subject assignment differed between these analyses.
Results
Baseline memory
There were no differences between l-amphetamine and placebo groups in baseline memory performance (p > 0.10).
Auditory/verbal memory
The median split divided the sample into subgroups with impaired auditory/verbal memory (N = 74, raw = 5.2 ± 2.1, normative z = −2.2 ± 1.0) or intact auditory/verbal memory (N = 62; raw = 11.6 ± 2.3; normative z = 0.0 ± 0.9). There were no subgroup differences in treatment group membership, age, education, MS subtype (p > 0.1) or gender (p = 0.09). ANOVA revealed a 2 × 2 × 2 interaction (F[1, 132] = 5.52, p = 0.02, η2 = 0.04) whereby l-amphetamine improved CVLT-II Delayed Recall performance more than placebo, but this effect was specific to patients with baseline memory impairment (Figure 1A; Table 1). In fact, memory-impaired patients showed a 49% improvement on l-amphetamine, compared with 7% on placebo. Indeed, among memory-impaired patients (N = 74), the 2 (l-amphetamine, placebo) × 2 (baseline, follow-up) interaction was very large (F[1, 72] = 10.45, p = 0.002, η2 = 0.13).
Effect of l-amphetamine versus placebo on (A) auditory/verbal memory, and (B) auditory/verbal learning from baseline (light gray) to follow-up (dark gray) for MS patients with baseline memory impairment (left) and MS patients with normal baseline memory (right). CVLT-II and BVMT-R raw score performance by treatment group (placebo, l-amphetamine) and baseline memory group (intact, impaired)
A secondary analysis performed for CVLT-II Total Learning (memory impaired: raw = 34.6 ± 8.3; normative z = −1.5 ± 1.0; memory intact: raw = 52.9 ± 7.9; normative z = 0.3 ± 0.9) did not find a 2 × 2 × 2 interaction (F[1, 132] = 2.43, p = 0.12, η2 = 0.02). There was no effect of l-amphetamine on Total Learning regardless of baseline memory functioning (Figure 1B; Table 1).
Visual/spatial memory
A median split divided the sample into those with impaired visual/spatial memory (N = 73, raw = 3.9 ± 1.9, normative z = −2.4 ± 0.7) and intact visual/spatial memory (N = 63, raw = 9.3 ± 1.7, normative z = 0.0 ± 0.9). There were no subgroup differences in treatment group, age, education, or gender (p > 0.1), but patients with secondary-progressive MS were more likely to be memory impaired (p = 0.005). Similar to auditory/verbal memory results, there was a 2 × 2 × 2 interaction (F[1, 132] = 6.51, p = 0.012, η2 = 0.05) such that l-amphetamine improved BVMT-R Delayed Recall performance more than placebo, and this effect was specific to patients with baseline memory impairment (Figure 2A; Table 1). In fact, memory-impaired patients showed a 48% improvement on l-amphetamine, but no improvement on placebo. Among memory-impaired patients (N = 73), the 2 (l-amphetamine, placebo) × 2 (baseline, follow-up) interaction was very large (F[1, 71] = 9.83, p = 0.002, η2 = 0.12).
Effect of l-amphetamine versus placebo on (A) visual/spatial memory, and (B) visual/spatial learning, from baseline (light gray) to follow-up (dark gray) for MS patients with baseline memory impairment (left) and MS patients with normal baseline memory (right).
Unlike auditory/verbal learning, a secondary analysis performed for BVMT-R Total Learning (memory impaired: raw = 11.0 ± 4.4; normative z = −2.4 ± 0.6; memory intact: raw = 22.6 ± 5.8; normative z = −0.4 ± 1.2) revealed a 2 × 2 × 2 interaction (F[1, 132] = 5.98, p = 0.016, η2 = 0.04). L-amphetamine improved BVMT-R Total Learning more than placebo, and this effect was specific to patients with baseline memory impairment (Figure 2B; Table 1).
Supplemental analysis
We also evaluated whether the positive impact of l-amphetamine on memory among MS patients with baseline memory impairment was due to changes in arousal/fatigue. Self-reported fatigue was assessed at baseline and follow-up with the Fatigue Severity Scale (FSS 9 ) in 133 of the 136 subjects. We conducted the same 2 (l-amphetamine, placebo) × 2 (baseline, follow-up) × 2 (lower verbal memory, higher verbal memory) ANOVA described previously, with FSS scores as our dependent measure. Unlike results for memory performance, the 2 × 2 × 2 interaction was not significant (F[1, 129] = 0.60, p = 0.440, η2 = 0.005), indicating no difference in arousal/fatigue from baseline to follow-up due to treatment group (l-amphetamine, placebo) or baseline verbal memory functioning (intact, impaired). Moreover, there was no 2 (l-amphetamine, placebo) × 2 (baseline, follow-up) interaction across the whole sample (F[1, 129] = 0.75, p = 0.387, η2 = 0.006), nor was there a 2 (l-amphetamine, placebo) × 2 (baseline, follow-up) interaction among subjects with baseline verbal memory impairment (F[1, 70] = 0.00, p = 0.950, η2 = 0.000). (Note that similar results were found when dichotomizing the sample based on baseline visual/spatial memory.) Taken together, l-amphetamine did not impact arousal/fatigue, and the positive impact of l-amphetamine on memory appears unrelated to changes in arousal/fatigue.
Discussion
In this post-hoc analysis of a previously published clinical trial, 4 we found large beneficial effects of l-amphetamine on both auditory/verbal memory and visual/spatial memory among memory-impaired MS patients, but no effect for patients with normal memory performance. As such, the effects of l-amphetamine were specific to patients most in need of treatment. Among memory-impaired patients, delayed recall improved about 48.5% for those on l-amphetamine, compared with only 1.0% on placebo.
These findings are consistent with animal research linking l-amphetamine to norepinephrine release 2 and improved memory consolidation, 3 and linking l-methamphetamine to learning-induced Arc/Arg3.1 protein synthesis within the hippocampus. 3 The notion that l-amphetamine improves memory by enhancing hippocampal function is consistent with the pattern of verbal learning and memory results in the current study: improved delayed recall of a word list (consolidation) without improved immediate memory (learning and/or working memory). As an alternative to this hippocampal-enhancement hypothesis, l-amphetamine may improve attention/working memory capacity by supporting frontal-striatal networks, which subsequently leads to improvements in delayed recall. We reject this second explanation because l-amphetamine failed to improve attention/working memory capacity within the current sample of MS patients, 4 nor did l-amphetamine increase arousal/reduce fatigue within this sample. Also, l-amphetamine is inferior to d-amphetamine in the treatment of attention deficits more generally (likely due to the lesser dopinergic effect of l-amphetamine relative to d-amphetamine 2 ). Furthermore, if l-amphetamine improved attention/working memory in the current study, then auditory/verbal learning (CVLT-II Total Learning) should have also improved with treatment, which it did not.
In contrast with auditory/verbal learning, treatment with l-amphetamine did improve performance on visual/spatial learning. That is, MS patients taking l-amphetamine were better able to immediately recall visual objects and their locations, suggesting improved visual/spatial working memory. Consistent with the notion that l-amphetamine enhances hippocampal function (rather than frontal-striatal circuits), previous research has demonstrated selective hippocampal activation during immediate memory for object–location pairings, but not for object–color pairings or single objects. 10 Given that (a) hippocampal function contributes to visual/spatial working memory of object–location pairings, 10 and (b) our hypothesis that l-amphetamine improves hippocampal functioning, then it is not surprising that l-amphetamine selectively improved immediate memory for object–location pairings (but not words) in the current study.
The current results are timely, as there are no drugs approved to treat memory problems in MS patients. Moreover, although preliminary findings for donepezil were promising, a recent phase III clinical trial failed to support donepezil as a memory treatment in MS. 11 Of course, a phase III clinical trial of l-amphetamine designed to evaluate memory benefits is needed before clinical recommendations should be offered. Also, future research is needed to evaluate the effects of l-amphetamine on patient reports of everyday memory functioning, as our treatments should aim to improve quality of life in naturalistic settings. Finally, future research should also investigate the effect of l-amphetamine on long-term memory outcomes assessed well after the completion of the trial. For now, however, the large effects of l-amphetamine on memory in the current study are encouraging. These effects are, however, limited by the re-analytical nature of this study.
Footnotes
Funding
This work was supported by Memen Pharmaceuticals, LLC.
Conflict of interest statement
JFS, NC and JD received salary support through compensation to the Kessler Foundation Research Center from the study’s sponsor, Memen Pharmaceuticals, LLC. DE is a shareholder of the sponsor company, and TK receives financial compensation from the sponsor. RHBB had previously received financial support from the sponsor prior to this study.