- Top of page
- Conflict of interest
Dementia with Lewy bodies (DLB) and dementia associated with Parkinson's disease (PDD) are considered to share major similarities and together account for up to 25% of the global incidence of dementia (Auning et al., 2012). The complex courses of DLB and PDD include cognitive, attentional, motor and psychiatric symptoms; making these forms of dementia particularly challenging in terms of quality of life (QOL) for patients and carers, admission to nursing homes and health-related costs (Lippa et al., 2007).
DLB patients have been shown to have greater attentional deficits than disease severity-matched Alzheimer's disease (AD) patients, but with smaller deficits to episodic memory (Ayre, 1998; Calderon et al., 2001). Since Donders (1868) developed automated tests of simple reaction time (SRT) and choice reaction time (CRT), these tests have been central to attention testing in psychology and clinical research. The CDR System is an automated set of tests of attention (including SRT and CRT) and memory, developed for repeated use in clinical trials and validated for use in dementia (Simpson et al., 1989, 1991; Nicholl et al., 1995). Using the CDR System, Ayre (1998) identified impairments in AD patients on SRT and CRT compared with controls, but established greater impairment to both tasks in DLB patients, despite the DLB patients having smaller deficits to episodic memory than the AD patients. Further, he established that the DLB to AD deficit on CRT was significantly greater than that on SRT, showing a greater impairment to information processing over and above the increased disruption to the ability to focus attention. A widely prevalent symptom in DLB is fluctuating cognition, identified in 80–90% of patients (Byrne et al., 1989; McKeith et al., 1992). Walker in a series of studies demonstrated that in DLB patients, clinically rated cognitive fluctuations correlated strongly with increased variability of the reaction times on the CDR System attention tests and that such increased variability could reliably distinguish DLB from AD and vascular dementia patients (Walker et al., 1999, 2000a, 2000b). Ballard et al. (2002) replicated the profile of poorer focussed attention, disrupted information processing and greater variability in reaction times in DLB using the CDR System; but importantly demonstrated that the same profile could be seen in PDD patients, compared with AD patients and controls. This characteristic profile of impairment to attentional processes has been used in current consensus guidelines for DLB (McKeith et al., 2005) and PDD (Emre et al., 2007) as evidence to support these impairments as core clinical features of both dementias.
Memantine is an N-methyl-D-aspartate receptor (NMDA) receptor antagonist that affects glutamatergic neuronal transmission and prevents the toxic effects of raised concentrations of the excitatory neurotransmitter glutamate. Memantine has proven effective in treating the symptoms of AD (McShane et al., 2006).
The first randomised, double-blind, placebo-controlled trial of memantine in DLB and PDD involved 72 patients (32 DLB and 40 PDD) recruited from four sites who were studied over 24 weeks (Aarsland et al., 2009). The primary outcome was the Alzheimer's Disease Cooperative Study Clinical Impression of Change (ADCS-CGIC), secondary outcomes including mini mental state examination (MMSE), a quick test of cognitive speed (AQT; Andersson et al., 2007), the Neuropsychiatric Inventory (NPI) and the disability assessment for dementia (DAD; Gelinas et al., 1999). At 24 weeks, a significant improvement was seen with memantine in the study population on the ADCS-CGIC, no differences in response being seen between the DLB and PDD patients. Whereas none of the other clinical scales showed effects, one of the three AQT subtests (the time taken to name 40 shapes) showed a significant benefit with memantine.
Emre et al. (2010) reported a second randomised, double-blind, placebo-controlled, 30 site trial of memantine in 78 patients with DLB and 121 with PDD. There were no pre-specified primary end points, the outcome measures including the ADCS-CGIC, the ADCS-activities of daily living scale, Zarit burden interview, NPI and 20 measures from 16 cognitive tests. At 24 weeks, memantine significantly improved the ADCS-CGIC and the NPI in the DLB patients, but not in the PDD patients or in the total population. There were no effects seen on the majority of the other assessments. For the 20 cognitive measures, memantine had no effects in the PDD population, but in DLB improved a single measure, the congruent part of the Stroop test.
Thus to date, there is evidence from two randomised double-blind trials that DLB patients treated with memantine for 24 weeks have superior ADCS-CGIC scores to controls, and in one trial PDD patients also. Matsunaga et al. (2013) reviewed the area and identified these two double-blind studies and five unblinded trials, concluding that memantine does not benefit cognition or motor function, but does improve the ‘overall impression’ of the disorders. There is clearly a paucity of evidence of improvements to objective tests of cognitive function; the only reliable effect so far reported being to one of three measures from the AQT in both DLB and PDD (Aarsland et al., 2009). Thus, there is evidence of change with memantine from these studies which the clinicians are able to detect, but which is largely unsupported by measurable changes in cognitive function. The study of Aarsland et al. (2009) also used cognitive tests from the CDR System, but the analyses of data from these tests were not available at the time of publication. As these tests have contributed to the consensus criteria for both dementias and have additionally proven sensitive in identifying cognitive enhancements in the major randomised double-blind, placebo-controlled trials with rivastigmine in DLB (McKeith et al., 2000) and PDD (Emre et al., 2004); the data from the CDR System tests are reported here for the first time to determine if they can also identify beneficial cognitive effects in these two dementias with memantine.
- Top of page
- Conflict of interest
Fifty-nine patients performed the computerised tests during the study, and both baseline and post-dosing data were available for 51 patients (21 DLB and 30 PDD). All 51 patients had test data available at week 12, and 48 had data available at week 24. The groups were balanced as to clinical and demographic characteristics at baseline (Table 1), no significant differences being detected for any measure.
Table 1. Baseline characteristics of study population
| ||Placebo (n = 24)||Memantine (n = 27)|
|Age (years)||76.2 (4.8)||76.1 (6.5)|
|Men : Women||18:6||22:5|
|Duration of disease (years)||8.5 (4.7)||6.6 (5.3)|
|PDD : DLB||16:8||14:13|
|Duration of PD before onset of dementia||5.2 (3.7)||4.6 (3.7)|
|MMSE||20.8 (3.7)||20.4 (3.6)|
|NPI total score||13.0 (10)||15.1 (14.6)|
|Cholinesterase inhibitors||16 (67%)||11 (41%)|
|Antiparkinson medication||22 (92%)||21 (78%)|
|Levodopa dose (mg/day)a||437 (255)||574 (366)|
The cognitive data and results of the ANCOVAs are summarised in Table 2. It can be seen that prior to dosing, there were numeric differences between the groups on some measures, but none were significant (all p > 0.1). However, baseline scores were included as covariates in the analyses to adjust for these numeric differences. None of the interaction terms from the ANCOVAs was significant (all p > 0.1), and thus, the combined effects of the two dementias over the post-dosing study period became the foci of interest. Compared with placebo, memantine produced favourable effects on all the measures from all four tests. Although the numeric benefit on SRT with memantine was not statistically significant, the benefits of the compound over placebo on CRT (Figure 1) and the accuracy scores from the two-word recognition tests were statistically reliable (Figure 2). Further, CRT declined significantly from baseline under placebo but not memantine; whereas recognition accuracy for both the IWR and DWR tests was significantly improved over the baseline level with memantine, but not placebo. The supportive measures from these tests provided support for the benefits of memantine, accuracy on CRT showing a numeric improvement and recognition times on the two memory tasks that both became numerically faster. Cohen classified effect sizes of 0.5 and 0.8 to be medium and large, respectively (Cohen, 1988). The effect sizes of the three statistically significant improvements ranged from 0.75 to 0.79, and thus approached large effects.
Table 2. Results of the efficacy analysis
|Measures||Treatment||n||Baseline||Change at 12 weeks||Change at 24 weeks||Overall change||Overall difference||Treatment||Type||Treatment * Week||Treatment * Type||Treatment * Week * Type||Cohen's d effect size|
|Mean (SD)||LS means (SEM)||LS means (SEM)||LS means (CI)||LS means (CI)||p||p||p||p||p|| |
|SRT RT (ms)||Placebo||24||563 (298)||41 (110)||187 (110)||114 (−43 to 272)||−41 (−255 to 173)||0.7019||0.7723||0.4019||0.6964||0.8033||0.11|
|Memantine||27||715 (447)||89 (97)||58 (101)||73 (−68 to 215)|
|CRT accuracy (%)||Placebo||24||84.0 (12.2)||−0.11 (2.7)||−0.82 (2.7)||−0.46 (−4.39 to 3.47)||3.44 (−1.91 to 8.79)||0.2017||0.5385||0.8305||0.4672||0.4707||0.37|
|Memantine||27||89.4 (12.4)||3.89 (2.4)||2.1 (2.5)||2.98 (−0.56 to 6.51)|
|CRT RT (ms)||Placebo||24||874 (335)||71 (80)||217 (80)||144 (31 to 258)a||−209 (−373 to −81)||0.0086||0.4323||0.3182||0.6138||0.1135||0.79|
|Memantine||27||995 (527)||−61 (70)||−68 (74)||−65 (−167 to 38)|
|IWR accuracy (%)||Placebo||22||51.1 (29.7)||−3.87 (5.9)||−7.09 (6)||−5.48 (−14 to 3.1)||14.2 (2.6 to 25.8)||0.0176||0.3574||0.7250||0.3645||0.9914||0.75|
|Memantine||25||46.7 (34.4)||8.3 (5.4)||9.15 (5.6)||8.73 (0.9 to 16.6)a|
|IWR RT (ms)||Placebo||22||1374 (405)||238 (256)||468 (260)||353 (−17 to 724)||−236 (−744 to 273)||0.3546||0.6592||0.8868||0.5636||0.8678||0.28|
|Memantine||25||2078 (1952)||−33 (234)||268 (240)||117 (−222 to 456)|
|DWR accuracy (%)||Placebo||22||50.0 (33.8)||3.23 (6.5)||−1.69 (6.4)||0.77 (−8.5 to 10)||15.6 (3.1 to 28.2)||0.0161||0.7410||0.4720||0.6066||0.9534||0.76|
|Memantine||25||38.3 (32.2)||14.4 (5.7)||18.41 (6)||16.41 (8 to 24.8)a|
|DWR RT (ms)||Placebo||22||1483 (510)||−74 (210)||343 (210)||135 (−167 to 436)||−261 (−672 to 149)||0.2059||0.2996||0.3852||0.3357||0.7053||0.39|
|Memantine||25||2204 (2340)||−159 (189)||−94 (194)||−127 (−400 to 147)|
Figure 1. Effects of memantine on simple reaction time and choice reaction time. Scores are least squares means from the analyses of covariances of the change from baseline scores over the 24 weeks of the study. Descending scores reflect impairments. The error bars are 95% confidence intervals.
Download figure to PowerPoint
Figure 2. Effects of memantine on immediate and delayed word recognition accuracy scores. Scores are least squares means from the analyses of covariances of the change from baseline scores over the 24 weeks of the study. Ascending scores reflect improvements. The error bars are 95% confidence intervals.
Download figure to PowerPoint
The significant improvements to attention with memantine as assessed by CRT in this study may have contributed to the improved recognition accuracy in the two memory tasks. To examine this, the major ANCOVAs for IWR and DWR accuracy were re-run, fitting CRT reaction time as an additional covariate. For IWR, this moved the previously significant main effect to a trend (p = 0.0645) and reduced the overall benefit slightly (13.3% from 14.2%). However, for DWR, the significant difference remained (p = 0.0031), and the beneficial effect of memantine actually improved (20.2% from 16.6%). This suggests that the beneficial effects of memantine in DLB and PDD on the delayed recognition task were largely independent to the benefits seen to attention.
From Table 3, it can be seen that at the start of the study, all measures from the four tasks were impaired in both dementias compared with age-matched normals, with large effect sizes (0.95–2.7). The profiles of impairment over the measures were fairly consistent between DLB and PDD. Further, the significant impairments to CRT and reaction time variability support previous findings that these are characteristic deficits of the two dementias (e.g. Ballard et al., 2002).
Table 3. Comparisons to normal controls
| ||Controls||Dementia with Lewy bodies||Parkinson's disease dementia|
|LS mean (SEM)||LS mean (SEM)||Impairment LS mean (95% CI)||% impairment||Cohen's d||LS mean (SEM)||Impairment LS mean (95% CI)||% impairment||Cohen's d|
|SRT (ms)||316 (8)||678 (31)||362 (300 to 424)||114||2.40||621 (26)||305 (251 to 359)||96||2.05|
|CRT accuracy %||96.6 (0.4)||85.8 (1.4)||−10.8 (−8 to −11)||−11||−1.62||86.4 (1.2)||−10.2 (−7.7 to −12.7)||−11||−1.53|
|CRT (ms)||501 (11)||1075 (42)||574 (115 to 662)||115||2.70||920 (37)||419 (343 to 495)||84||2.00|
|Immediate word recognition (%)||86.7 (1.2)||49.5 (3.7)||−37.2 (−29.5 to −43)||−43||−2.04||46 (3.4)||−40.7 (−33.6 to −47.9)||−47||−2.24|
|Immediate word recognition time (ms)||897 (55)||2164 (169)||1266 (141 to 1617)||141||1.53||1684 (156)||786 (460 to 1113)||88||0.95|
|Delayed word recognition (%)||80.4 (1.4)||45.5 (4.3)||−35 (−26 to −43)||−43||−1.65||39.4 (4)||−41.1 (−32.7 to −49.4)||51||−1.94|
|Delayed word recognition time (ms)||881 (55)||2181 (167)||1300 (148 to 1646)||148||1.59||1822 (154)||941 (619 to 1264)||107||1.15|
|Cognitive reaction time (ms)||184 (9)||396 (35)||212 (115 to 283)||115||1.26||299 (30)||114 (53 to 175)||62||0.69|
|Reaction time variability (CV%)||35.4 (1.7)||77.5 (44)||42.1 (119 to 51.9)||119||1.85||77.6 (3.7)||42.1 (33.8 to 50.5)||119||1.94|
An advantage of having normative data on tests used in dementia trials is that it is possible to estimate the degree to which treatment has moved patients towards ‘normal’ levels (Wesnes et al., 2002, 2005; Vellas et al., 2005). In this trial, over the 24-week treatment period, memantine moved patients towards controls by 43% on CRT (33% on accuracy), 36% on IWR accuracy (23% on speed) and 41% on DWR accuracy (24% on speed).
The CDR System scores at baseline were correlated with those of four clinical scales used in the study (Table 4). For the MMSE, all CDR System measures except accuracy on the CRT task correlated significantly, and superior scores on each measure correlated with higher MMSE scores. The same pattern was seen on DAD, all CDR System measures except CRT accuracy correlated significantly, and superior scores on each measure correlated with higher DAD scores. A comparable pattern was also seen on a 13-item self-rating of QOL (Logsdon et al., 1999); all measures except CRT accuracy and SRT correlated with the ratings and superior scores on each performance measure correlated with higher QOL scores. No significant correlations between the cognitive scores and the NPI were identified.
Table 4. Pearson's correlation coefficients between the CDR System scores of the study population at baseline with four clinical scales
|CDR System measure||MMSE||DAD||13-QOL||NPI|
|Mean 20.6, SD 3.6||Mean 12.5, SD 4.6||Mean 33.7, SD 6.7||Mean 14.1, SD 12.6|
|CRT accuracy %||0.2||−0.02||0.11||0.24|
|Immediate word recognition (%)||0.49*||0.31*||0.45*||0.04|
|Immediate word recognition time(ms)||−0.43*||−0.6*||−0.31*||0.07|
|Delayed word recognition (%)||0.51*||0.35*||0.38*||−0.01|
|Delayed word recognition time (ms)||−0.37||−0.46*||−0.41*||0.16|
To examine the relationship between the improvements on the three cognitive tests and the ADCS-CGIC scores reported previously (Aarsland et al., 2009), correlations were performed at week 24. CRT alone was found to show a correlation with ADCS-CGIC at this time (r = 0.33, p < 0.025). To further examine the relationship, three groupings of ADCS-CGIC scores were created: ‘Improved’ (scores 1–3; ‘substantial to minimum improvement’), ‘No change’ (score 4) and ‘Worsened’ (scores 5–7, ‘minimum to substantial worsening’). These three groupings were used instead of treatment in the same ANCOVA model used to assess treatment efficacy. A statistically significant main effect of ADCS-CGIC groupings was identified (p = 0.0057), but no simple or higher order interactions were seen (all p > 0.27). Speed on CRT was non-significantly enhanced in the ‘Improved’ (−52 ms) and ‘No change’ groupings (−34 ms). However, in the ‘Worsened’ grouping, CRT declined significantly (p = 0.0013; 265 ms, 95% CI 116–413). Thus, changes in ADCS-CGIC ratings were associated meaningfully with changes to CRTs in this study.
- Top of page
- Conflict of interest
Although other data from this study have already been published (Aarsland et al., 2009), this is the first reporting of the data from the CDR System. The major finding from this analysis was that memantine produced significant improvements on three of four tests of cognitive function in both DLB and PDD. SRT was not statistically significantly improved, although there was a small numeric benefit. CRTs showing a significant shortening combined with a numeric pattern of improved accuracy, suggesting that attentional testing with greater information processing requirement benefits from memantine. This effect is consistent with the improvement in the AQT with memantine in this study (Aarsland et al., 2009). Memantine also statistically significantly improved accuracy on immediate and DWR tests, with speed of recognition showing numerical improvements on both. The improvements on the three tasks approached large effect sizes and are thus likely to be of clinical and everyday relevance. They also compare favourably with the benefits of anticholinesterases in AD, a systematic review of 20 trials (including all pivotal studies) identifying Cohen's d effect sizes on the Alzheimer's Disease Assessment Scale Cognitive subscale which ranged from 0.01 to 0.3 (Rockwood, 2004).
The findings of this study are in agreement with previous efficacy trials with rivastigmine and donepezil in both dementias using the same CDR System tests, attention being improved by the two anticholinesterases in DLB (McKeith et al., 2000; Wesnes et al., 2002; Rowan et al., 2007) and also in PDD (Emre et al., 2004; Wesnes et al., 2005; Rowan et al., 2007). Further, a score comprising the accuracy scores from IWR and DWR plus a picture recognition task (not used in this study) showed significant improvements with rivastigmine in DLB (Wesnes et al., 2002).
Previous work with the CDR System in therapeutic dementia trials has employed the technique of evaluating treatment response in terms of the degree that pre-dosing deficits are reversed towards the levels of normals. In two randomised trials with rivastigmine, a CDR System measure, which combines speed on three attention tests (including SRT and CRT), was moved towards the scores for normals by 50% in DLB (Wesnes et al., 2002) and 24% in PDD (Wesnes et al., 2005). In a trial of galantamine in AD, CRT was moved 30% towards normal scores after 12 weeks of treatment (Vellas et al., 2005). Overall, the 43% identified in the present study with CRT is in the same range as these effects. Further, in the present study, the movements towards control levels on the IWR and DWR accuracy of 40% and 43%, respectively, are also treatment effects which are arguably clinically relevant.
The relationship established in the present study between the improved attention with memantine (as assessed by CRT) and the ADCS-CGIC at 24 weeks suggests that enhanced attention played a role in the improved clinical ratings. As the ADCS-CGIC raters did not have access to the data from the computerised tests, this strengthens the possible association. Finally, the relationships identified at baseline between CRT (and the other cognitive tests) and MMSE, QOL and activities of daily living also support the possibility that the cognitive benefits contributed to the improved ADCS-CGIC scores with memantine in this study. Further CDR System attention tests have been shown to be strong predictors of activity of daily living in both DLB (Ballard et al., 2001) and PDD (Bronnick et al., 2006).
The larger trial of Emre et al. (2010) used SRT and CRT tasks from the CogState battery; for both tests, the mean (standard error of the mean) changes from baseline after 24 weeks were 0 (0) for both the placebo and memantine groups. This contrasts notably with the effects seen in the present study, particularly on CRT. Possible explanations for this difference could relate to differences in the two study populations, for example, the patients in the study of Emre et al. were not taking anticholinesterases, whereas 70% of those in the present study were on stable therapy. Another could be that not all SRT and CRT tasks have equivalent sensitivity in dementia trials.
The mechanisms by which memantine may have produced the improvements to attention and episodic memory in the present study deserve consideration. The additional ANCOVAs of the memory scores fitting the improved CRTs as covariates strongly suggested the benefits on DWR to be largely independent of attentional improvements. NMDA antagonists can promote hippocampal neurogenesis (Nacher et al., 2003), which may be a potential mechanism for the improvements in episodic memory in this study and also in that seen previous AD trials. It is relevant to such a mechanism that memantine improved memory above baseline levels in this study. The major effect of memantine on CRT in this study was to prevent the decline which occurred in placebo-treated patients over 24 weeks. Such an effect may be more consistent with the neuroprotective properties of memantine. Of possible relevance is a previous finding in first-time diagnosed previously unmediated epilepsy patients treated with remacemide (Wesnes et al., 2009), which, like memantine, is a low-affinity NMDA receptor antagonist. In that study, the CDR System SRT and CRT tests were employed, and remacemide significantly prevented declines on these measures over 48 weeks, which occurred in patients randomised to carbamazepine. Further, at baseline, attention was identified to be compromised in the epilepsy patients compared with healthy controls, potentially because of the negative neural effects of seizures, and remacemide actually improved attentional performance above the baseline levels in the study. This suggests that low-affinity NMDA antagonists can protect against neurotoxicity to favourably influence aspects of attention; which could be a possible mechanism for the effects of memantine seen in the present study.
In conclusion, the present analysis suggests that memantine produces worthwhile cognitive benefits in DLB and PDD; evidenced by the large effect sizes, the movement of the test scores of the patients towards normal levels and the prior established relevance of the attention tests to everyday activities. This possibility is further supported by the association of enhanced attention to the improvement in ADCS-CGIC detected in the study (Aarsland et al., 2009). Furthermore, the effects seen in the two dementias with memantine are comparable in magnitude to the cognitive benefits in these conditions identified previously with anticholinesterase treatment. Overall, these findings indicate that larger future trials of memantine and related compounds are warranted both in DLB and PDD.