A systematic review of cognitive decline in dementia with Lewy bodies versus Alzheimer’s disease

Introduction The aim of this review was to investigate whether there is a faster cognitive decline in dementia with Lewy bodies (DLB) than in Alzheimer’s disease (AD) over time. Methods PsycINFO and Medline were searched from 1946 to February 2013. A quality rating from 1 to 15 (best) was applied to the included studies. A quantitative meta-analysis was done on studies with mini mental state examination (MMSE) as the outcome measure. Results A total of 18 studies were included. Of these, six (36%) reported significant differences in the rate of cognitive decline. Three studies reported a faster cognitive decline on MMSE in patients with mixed DLB and AD compared to pure forms, whereas two studies reported a faster decline on delayed recall and recognition in AD and one in DLB on verbal fluency. Mean quality scores for studies that did or did not differ were not significantly different. Six studies reported MMSE scores and were included in the meta-analysis, which showed no significant difference in annual decline on MMSE between DLB (mean 3.4) and AD (mean 3.3). Conclusions Our findings do not support the hypothesis of a faster rate of cognitive decline in DLB compared to AD. Future studies should apply recent diagnostic criteria, as well as extensive diagnostic evaluation and ideally autopsy diagnosis. Studies with large enough samples, detailed cognitive tests, at least two years follow up and multivariate statistical analysis are also needed.


Introduction
Dementia with Lewy bodies (DLB) and Alzheimer's disease (AD) are the two most common subtypes of neurodegenerative dementia, representing 15 to 20% and 65% of all dementia cases, respectively [1]. DLB is characterized clinically by symptoms such as visual hallucinations, Parkinsonism and fluctuating cognition in addition to cognitive impairment with typically more visuospatial and executive impairment relative to memory impairment [2]. There is some evidence that DLB patients have more rapidly progressing dementia compared to AD [3], and more recent studies also reported a more severe course with shorter survival [4], higher rate of nursing home admissions [5] and higher costs in DLB as compared to AD [6].
An overlap in neuropathology between AD and DLB has been noted [7]. Parkinson's disease (PD) and DLB also share some clinical and pathological features [8]. Subgroups with different cognitive profiles have been described in patients with PD [9], and there is evidence that this differentiation is related to the rate of cognitive decline [10]. Similar neuropsychologically defined subgroups may exist also in DLB [8], which could also predict differences in the rate of progression to end-stage dementia. Data supports accelerated disease progression when AD and DLB pathologies are present together [11].
To our knowledge, no systematic review has compared rate of cognitive decline in DLB versus AD. We therefore systematically reviewed the literature to find studies assessing overall cognitive decline in DLB and AD. We specifically noted studies that had investigated the potential differences in cognitive decline in subgroups with DLB and the effect of employing different diagnostic criteria.

Methods
PsycINFO and Medline were searched in February 2013, using key words listed in Table 1. References from reviewed articles were also searched for relevant studies. The following inclusion criteria were used: a) paper published in a peer-reviewed journal; b) written in English; c) DLB or mixed AD/DLB compared with AD; d) application of at least one neuropsychological test, and e) at least 6 months follow up. The following exclusion criteria were used: a) drug trials, and b) survival studies with death as the only outcome.

Quality assessment
Two independent raters rated all studies with a selfdesigned quality scale and arrived at the same result. The domains, a) number of patients included; b) follow-up time; c) clinical criteria; d) autopsy, and e) neuropsychological tests) were rated on a four-point scale adapted from Aarsland et al. (2005) [12]: 0 (none), 1 (poor), 2 (fair) and 3 (good). See Table 2. Studies could be assigned 1 to 15 points.

Statistical analysis
For studies reporting mini mental state examination (MMSE) results, standardized mean difference in annual progression between DLB and AD was calculated as the difference between annual progression between the DLB and AD groups divided by the pooled standard deviation across groups in each included study. The standardized mean differences were combined in a random-effects model to obtain summary estimates of the effect in each study. The overall results from each trial were then combined using a random-effects model to obtain a pooled summary estimate of effect across all trials [13]. To assess heterogeneity, the I 2 as proposed by Higgins and colleagues [14] was chosen, indicating the percentage of total variation across studies due to heterogeneity.

Results
Of the 18 studies included in this review (see Table 3), six (36%) reported a statistically significant difference in cognitive decline over time between AD and DLB (see Table 4). Three studies reported a faster cognitive decline on cognitive screening tests in the neuropathologically mixed AD/DLB group [3,15,16] compared to those with pure AD or DLB. One study reported a faster decline in DLB than in AD on verbal fluency [17], and two in AD compared to DLB on memory [18,19]. For a full description of neuropsychological tests used in included studies, see Table 3.
Six studies either reported annual decline in MMSE scores, or included data enabling calculation of annual decline based on reported scores. In AD, mean annual decline was 3.3 (SD 1.7, range 1.8 to 4.9), and in DLB 3.4 (SD 1.4, range 1.8 to 5.8). One study also reported annual decline of 5.0 in AD/DLB (see Figure 1). The random-effects meta-analysis revealed an overall effectsize of −0.035 (negative sign indicates faster progression in DLB) (P = 0.764; 95% CI = 0.261, 0.192). I 2 was 50.3, which is considered to represent moderate heterogeneity [14].

Cognitive domains
Six studies measured memory, and two reported differences in memory over time, both a faster decline in AD. Delayed recall was found to have a faster decline in AD compared to AD/DLB when measured with the Consortium to Establish a Registry for Alzheimer's Disease (CERAD) evaluation, with 15% of patients with AD versus 32% of patients with AD/DLB remembering any item at the last evaluation [17]. Recognition was found to have a faster decline in AD compared to DLB as measured with Hopkins verbal learning test-revised (HVLT-R) (scores not available) [19]. Eight studies measuring language and ten studies measuring visuospatial ability reported no differences in rate of decline. Seven studies measured explicit executive functions, and one reported differences over time. In that study, verbal fluency was found to have a more rapid decline in DLB compared to AD, measured with the Cambride cognitive examination (CAMCOG) (subscores not available) [17].

Subgroups
Two studies [28,30] divided patients into two groups according to high or low visuospatial functioning. In the first study, DLB patients with a low baseline score (<20) on the Wechsler intelligence scale for children-revised, block design (WISC-R) and impaired clock drawing test (CDT) had a faster decline on the dementia rating scale (DRS), compared to DLB patients with a high baseline score. In the latter study, DLB patients with a low baseline score on the Newcastle visual perception battery (NEVIP) had a faster decline in activities of daily living (ADL) than those with higher score, but no difference on any of the cognitive tests. There were no differences in the AD groups.

Quality assessment
The mean quality score for all the included studies was 9.4 points (SD 2.5, range 5 to 14) (see Table 5). Only two studies were rated fair or good on all quality measures [26,27]. Three studies were rated poor on one variable, but fair and good on the others [15,16,22]. Mean quality scores for studies that found any differences in cognitive decline was 9.8 points (SD 2.4, range 5 to 11) compared to 9.3 points (SD 2.6, range 5 to 14) in the group with no differences (P = 0.335).

Clinical and neuropathological diagnostic criteria
There were no systematical differences in clinical or neuropathological criteria between studies that found differences in cognitive decline and those who did not (see Table 6

Discussion
In the 18 studies included in this review, no consistent faster rate of decline in DLB as compared to AD on cognitive screening tests was found. When combining studies that used MMSE, the most frequently used scale, a meta-analysis revealed no difference in the annual rate of cognitive decline. There were mixed findings on decline in specific cognitive domains. Two of six studies of memory found a more rapid decline in AD. Only one of seven studies of executive function found a more rapid decline in DLB, and differences in visuospatial or language tests were not found. The hypothesis of a more rapid cognitive decline in autopsied patients with both AD and DLB pathology was supported in three studies. However, findings were inconsistent and other studies did not find differences. Differences in methods such as selection criteria, design, neuropsychological tests, dementia severity, diagnostic procedures and criteria can explain the diverse findings and lack of firm conclusions. However, quality assessment did not reveal any systematic differences between studies with high or low quality scores. There were large differences in sample sizes (n = 28 to 315), and the studies that could not be included in the metaanalysis or used other tests than MMSE, thus, may have had varying statistical power to detect significant differences between groups. To be able to compare the overall results and draw some general conclusions it would have been ideal that uniform diagnostic criteria had been used in all the studies. Some of the studies initially included patients with a clinical diagnosis of AD only, where analyses were based on autopsy diagnosis which included both AD and DLB.
A common weakness in the included studies was the choice of neuropsychological measures. When studying         Screening tests or batteries that use a total score only, often designed for purposes other than research are less suitable. In this review, the MMSE was the most used test, either alone, or in combination with others. The MMSE may not be an optimal measure, especially when using only the total score and not separate subscores for different cognitive domains, as AD and DLB have different cognitive profiles at onset [32]. This difference in cognitive profile leads to difficulties in choosing an optimal cognitive screening instrument to compare AD and DLB. The MMSE is heavily based on memory and language and is thus more sensitive to changes in AD than in DLB [33].
DLB is associated with a more severe visuospatial deficit than AD [32,34], but only 1 of 30 points on the MMSE comes from a measure of visuospatial functioning. MMSE may also be less than optimal because of the ceiling and floor effect [35], which refers to a test being too easy or too difficult to discriminate below or above a certain point, which is a common problem when testing people with dementia. In one of the reviewed studies the children's version of the Wechsler intelligence scale was used to avoid this. The test then lacks age adjusted norms, but it gains a wider range in scores, and therefore can monitor the cognitive decline over a longer period of time. Studies differed also with regard to the time period of observation, from 1 to    [36] have argued that MMSE registrations need to be separated by at least three years in order to be a reliable measure of cognitive decline in AD.

Std diff Standard
Only few studies investigated, or reported, subgroups with different cognitive profiles in DLB. It could be due to a low number of cases in several studies, and subsequent low statistical power. People die from dementia or reach an endpoint where they are not capable of performing cognitive tests, and therefore in several studies there was a lower number of patients towards the end of the study. This is challenging when performing statistical analysis. Our search did not cover the issue of subgroups with different cognitive profiles thoroughly, as we only included studies comparing DLB with AD, and not studies describing cognitive decline in DLB and potential subgroups alone. However, there are some data that support the hypothesis that there are subgroups in DLB with different cognitive profiles, and subgroups with poor initial visuospatial function may have a more rapid decline than DLB with good visuospatial function [28].
Due to overlapping symptoms, it can be difficult to determine the correct diagnosis ante mortem between the pure form of AD, mixed AD/DLB and the pure form of DLB. Because clinical criteria cannot distinguish with certainty the individual pathology, the gold standard for validating the clinical assessment is neuropathological diagnosis. Clinical criteria may have a low sensitivity in particular for DLB, which could have been a source of bias in studies that did not include a neuropathological validation of the diagnosis. However, dementia is a clinical diagnosis and both AD and DLB pathology can be found also in cognitively normal elderly subjects. In one study with autopsy, 50% of cases with widespread αsynucleinopathy did not show any clinical signs of dementia [37].
In most studies with autopsy, consensus neuropathological criteria were used. Even though not all included studies used consistent and the same neuropathological methods and criteria, and many also used varying combinations, use of post-mortem verification at least increases the validity of the clinical diagnosis.
It is also important to mention that the sensitivity for detecting Lewy bodies has increased with anti-ubiquitin immunostaining, where tau-positive samples indicate Alzheimer's pathology. Anti-α-synuclein immunostaining has been incorporated in the assessment, which is most sensitive for Lewy body pathology [2]. Thus, the neuropathological identification of cases may have been less accurate before the new methods were established, and more reliable staging strategies have been developed [38].
A complicating issue is the frequent occurrence of mixed pathology [39], and to underline the complexity of dementia and its pathology, at least four distinct pathological phenotypes have been identified between AD and DLB [40]. According to Schneider et al. (2012) [7], the locus of neuropathology is associated with a faster decline in cognition. A neocortical type of Lewy body pathology is associated with increased odds of dementia and a faster decline in episodic, semantic and working memory. The limbic-type is more associated with more rapid decline in visuospatial function. Olichney et al. (1998) [3], concluded that patients with Lewy body variant decline faster than patients with Alzheimer's disease. This statement has often been used with reference to rapid progression in DLB, but it actually refers to an AD variant with Lewy body pathology, not to pure DLB. It should be emphasized that it is still uncertain whether AD and DLB are two independent pathologies that may coexist, or the pathologies are related, or one of them is a consequence of the other.

Conclusion
Only 6 of the 18 included studies in this review found some differences in cognitive decline between DLB and AD over time, and only one of them found a faster decline in DLB. It is difficult to draw firm conclusions based on available studies, since the results are contradictory. Future studies will need to apply recent diagnostic criteria, as well as extensive diagnostic evaluation and autopsy to confirm the diagnosis. Studies with large enough samples, adapted cognitive tests, more than one year of follow up and multivariate statistical analysis are also needed. Inclusion of mild cognitive impairment patients, with subclinical manifestations and an increased risk of developing DLB (for example, who present rapid eye-movement (REM) sleep behavior disorder) could also strengthen the studies. Our final conclusion is that the studies in this review support neither the hypothesis of a faster cognitive decline in DLB, nor in AD. Authors' contributions MHB and LJC have made the conception and design, data acquisition, analysis and interpretation of data, and drafted the manuscript. AR, MJH, CJ, KB and DA have contributed to the analysis and interpretation of data, and revised the manuscript critically for important intellectual content. KB also performed the meta-analysis. All authors have read and approved the final version of the manuscript.