The Prevalence and Incidence of Dementia with Lewy Bodies: a Systematic Review

ABSTRACT Background Population-based prevalence and incidence studies are essential for understanding the societal burden of dementia with Lewy bodies (DLB). Methods The MEDLINE and EMBASE databases were searched to identify publications addressing the incidence and/or prevalence of DLB. References of included articles and prior systematic reviews were searched for additional studies. Two reviewers screened all abstracts and full-text reviews, abstracted data and performed quality assessments. Results Twenty-two studies were included. Incidence rates ranged from 0.5 to 1.6 per 1000 person-years. DLB accounted for 3.2-7.1% of all dementia cases in the incidence studies. Point and period prevalence estimates ranged from 0.02 to 63.5 per 1000 persons. Increasing prevalence estimates were reported with increasing age. DLB accounted for from 0.3 to 24.4% of all cases of dementia in the prevalence studies. Conclusions DLB becomes more common with increasing age and accounts for about 5% of all dementia cases in older populations. RÉSUMÉ Prévalence et incidence de la démence à corps de Lewy : revue systématique du sujet. Contexte: Les études de population sur sa prévalence et son incidence sont essentielles à la compréhension du fardeau social de la démence à corps de Lewy (DCL). Méthodologie: Nous avons identifié les publications sur l’incidence et/ou la prévalence de la DCL dans les bases de données MEDLINE et EMBASE. Nous avons recherché des études supplémentaires dans les articles et les revues systématiques antérieures cités à titre de références. Deux évaluateurs ont examiné tous les résumés et les évaluations de textes intégraux et l’extraction des données, et ils ont évalué la qualité des publications. Résultats: Vingt-deux études ont été incluses. Les taux d’incidence allaient de 0,5 à 1,6 par 1 000 personnes-années. La DCL constituait 3,2 à 7,1% de tous les cas de démence dans les études sur l’incidence. Les estimés de prévalence ponctuelle et par période allaient de 0,02 à 63,5 par 1 000 personnes. Des taux croissants de prévalence étaient rapportés en association avec l’augmentation de l’âge des sujets. La DCL était responsable de 0,3 à 24,4% de tous les cas de démence dans les études de prévalence. Conclusions: La DCL est plus fréquente avec l’âge et constitue environ 5% de tous les cas de démence dans les populations plus âgées.


INTRODUCTION
Dementia with Lewy bodies (DLB) is one of the commonest neurodegenerative dementias encountered in older patients. 1 While recent clinic-based studies suggest that it accounts for up to 20% of all dementia cases, 1-5 this may represent an overestimate because of selection or referral bias. 6 Population-based estimates of attributable risk derived from autopsy studies suggest that Lewy body disease accounts for 3 to 10% of the burden of dementia in older populations. 7,8 Building on clinicopathological work, [9][10][11] consensus criteria for the diagnosis of DLB have been developed. [12][13][14] Their central feature is the presence of progressive cognitive decline of sufficient magnitude to interfere with normal social or occupational functioning. During the early stages of the illness, deficits in attention, executive function and visuospatial ability are often more striking than those with memory. The core clinical features of DLB are fluctuations in cognition with pronounced variations in attention and alertness, recurrent detailed visual hallucinations and spontaneous parkinsonism. The latest revision 14 added three suggestive features (neuroleptic sensitivity, reduced basal ganglia dopamine uptake on single-photon emission computed tomography [SPECT] or PET imaging, REM sleep behaviour disorder). Probable DLB is diagnosed if two core features or one core and at least one suggestive feature are present. Possible DLB requires the presence of either one core or suggestive feature.
Population-based prevalence and incidence studies are essential for understanding the societal burden of DLB and planning for the range of healthcare services needed for these individuals. In this paper, we report on a systematic review of population-based prevalence and/or incidence studies of DLB. This is a relatively understudied area, with only two prior systematic reviews on the epidemiology of DLB published. 15,16 Since the appearance of the first one in 2005, 15 awareness of DLB has increased with both revised diagnostic criteria 14 and additional studies reporting primary data published. The earlier systematic review 15 was based on seven studies compared to the 22 that we included. The second systematic review 16 is more current, but there were significant differences in the methodology used by the authors compared to our work. For example, their literature search was limited to one database (PubMed), and different criteria were utilized for the selection of articles. Of the 32 population-based studies included in either systematic review, only 11 [34.4%] were common to both. As our systematic review includes studies not incorporated in the previously published systematic reviews and utilizes different study methods, we feel it offers an additional perspective for the literature on this topic.

METHODS
This is one of a series of systematic reviews on the incidence and prevalence of priority neurological conditions identified by the Public Health Agency of Canada and Neurological Health Charities Canada as part of the National Population Health Study of Neurological Conditions. These conditions included dementia.

Search Strategy and Study Selection
The MEDLINE and EMBASE databases were searched using terms and approaches developed by the authors in consultation with an academic librarian to identify publications dealing with the incidence and/or prevalence of dementia and the commoner dementia aetiologies in middle-aged and older adults (i.e., Alzheimer's disease, vascular dementia, frontotemporal dementia, DLB) (see Appendix A for detailed search terms). Population-based studies that reported on the incidence and/or prevalence of dementia and the commoner aetiologies were included provided they: (1) were published in English or French; (2) reported primary or original data; and (3) had a publication date of 2000 or later (international studies) or a publication date of 1985 or later (Canadian studies). The 2000-or-later publication date was used to avoid overlap with previous systematic reviews of the epidemiology of dementia, while the earlier date for Canadian studies was chosen to ensure that all relevant national reports were included (note: no Canadian studies on DLB were found).
Two reviewers independently screened all retrieved abstracts and performed full-text reviews on selected papers. Disagreements were resolved by consensus with the option of involving a third reviewer if one could not be reached (which was never required). The reviewers also searched the references of included articles and prior systematic reviews in order to detect additional eligible studies. Only studies that reported on the incidence and/or prevalence of DLB (or provided the necessary information to calculate an estimate) were included.
As more than a year elapsed between the initial literature search and completion of data abstraction, the EMBASE and MEDLINE database searches were updated (July of 2012). Search terms dealing with dementia aetiologies other than DLB were removed (see Appendix B for detailed search terms); otherwise, the methodology employed (including inclusion criteria and review procedures) was identical to that used in the original search.

Data Abstraction and Study Quality
The two reviewers independently abstracted data using a standard form. Overall prevalence and/or incidence values for DLB were recorded along with age-and sex-stratified data when available. Information regarding study location, age range of the study population (i.e., <65, 60+, 65+, 70+, 75+, 81+), case ascertainment methods, diagnostic criteria and definition of DLB were collected. The reviewers independently completed a quality assessment for each included study using an instrument specifically designed for this series of systematic reviews. Based on prior recommendations, 17,18 studies were assessed on eight characteristics (an affirmative response was given one point, while all others were scored zero): (1) target population was clearly defined; (2) cases were ascertained by either a survey of the entire population or probability sampling; (3) the response rate was >70%; (4) nonresponders were described; (5) the sample was representative of the population; (6) data collection methods were standardized; (7) validated diagnostic criteria were used; and (8) presentation of incidence and/or prevalence estimates was reported with confidence intervals or by subgroup. Possible scores ranged from 0 to 8, with higher ones indicating a better quality assessment. For studies where the estimates were based on information provided by practitioners and/or health administration data, reviewers responded in the affirmative for the third, fifth and sixth points, and "not applicable" for the fourth. Any discrepancies were resolved by consensus or involvement of a third reviewer, if necessary.

Data Analysis
A pooled meta-analysis was not done due to significant between-study heterogeneity and the small sample size. Forest plots presenting the distribution of study estimates were produced. All statistical analyses were carried out in R version 2.14, and the Meta package was employed to produce forest plots. Because of higher diagnostic specificity, 12 the most restrictive definition for DLB that could be obtained from the reviewed papers (e.g., probable DLB rather than possible DLB) was used in our primary analyses of incidence and/or prevalence. Depending on study methodology, incidence proportion (or cumulative incidence, the proportion of the population at risk who develop DLB within a defined period), incidence rate (the rate at which new cases of DLB occur in a population during a defined period), period prevalence (total number of persons known to have DLB during a specified period) and point prevalence (the number of persons with DLB at a specified point in time) were provided.

Search Strategy
In the initial search, 7923 unique references were identified on MEDLINE and/or EMBASE (see Figure 1). A total of 707 articles were selected for full-text review. Of these, 230 were excluded on the basis of publication date, 164 for not reporting on the incidence and/or prevalence of dementia, 116 as they were not population-based and 39 because they did not report original data. The updated MEDLINE and EMBASE search yielded a further 334 references. Eleven were selected for full-text review, and four met our inclusion criteria. An additional 12 articles were identified from the references of selected articles. A total of 174 articles met our inclusion criteria for dementia. Twenty-two of them reported on DLB and were included in our systematic review.  Of these papers, four focused solely on DLB, 19,34,35,40 while 18 dealt with some other aspect of the epidemiology of dementia but included data on DLB. [20][21][22][23][24][25][26][27][28][29][30][31][32][33][36][37][38][39]
Please refer to Tables 1 and 2 for a summary of study characteristics. Eight originated from the Western Pacific (Australia 1, Japan 6, Korea 1), 7 from Europe (Finland 1, France 1, Italy 1, Spain 1, Turkey 1, United Kingdom 2), 6 from the Americas (Brazil 3, Cuba 1, United States 2) and 1 from Southeast Asia (Sri Lanka). Most (n = 21) of the studies reported on the incidence and/or prevalence in a defined geographic area. The exception examined prevalence among residents of 22 randomly selected assisted-living facilities in the city of Baltimore and several Maryland counties. 36 Eight studies provided information on sex, with four reporting that it was more common among women 24,29,34,35 and four among men. 33 Excluded as did not report on DLB incidence or prevalence (n = 152) Figure 1: Flowchart of studies.    reported on study populations <65 years of age. 25,28 Of those dealing with older populations, the lower age thresholds were 60+, 20,37 65+, 21,22,26,27,[29][30][31][32][33][34][38][39][40] 70+, 23,24 75+ 35 and 81+. 19 The final study did not employ an age range but dealt with an institutional population (mean age 85.6 years). 36 Fourteen studies used a multistage approach (i.e., initial screening assessment followed by an in-depth evaluation of those who screened positive for dementia ± a random sample of those screening negative) to detect cases. Most (n = 20) studies used either DSM-III-R 41 or DSM-IV 42 criteria for the diagnosis of dementia. The two exceptions were one 19 that utilized Clinical Dementia Rating (CDR) 43 scores and another 38 that used both DSM-IV and ICD-10 criteria. 44 The disciplinary backgrounds of those performing clinical assessments were diverse and included (listed alphabetically) geriatric psychiatry, geriatrics, internal medicine, neurology (including residents in neurology), neuropsychology, neuropsychiatry, nursing, physiatry/disability medicine, psychiatry, physician (no specialty specified), psychology and psychometry. A consensus approach (i.e., more than one person involved) was used to make final diagnoses in 11 studies. 20,21,23,24,26,[32][33][34][35][36]38 Where stated (three studies did not provide information), 20,32,37 various versions of criteria developed by McKeith et al. [10][11][12][13][14] were used for diagnosing DLB. Fourteen studies did not indicate whether probable or both possible and probable DLB cases were being enumerated. 20,21,[23][24][25][26][27][28][31][32][33]36,37,39 An incidence study 30 and two prevalence studies 29,40 reported on the number of possible and probable DLB cases, while an incidence study 34 and four prevalence ones 19,22,35,38 gave the number of probable DLB cases (four of these studies also provided data on possible DLB cases). The specific methods used to identify hallucinations were provided by ten studies 19,23,25,27,29,[32][33][34][35][36] (in seven, this was by the administration of a standardized behavioural instrument such as the Neuropsychiatric Inventory 45 ). Four studies gave information on detection of parkinsonism 19,34,35,38 (in three, this was by detailing components of the clinical assessment), and three provided the approach used for fluctuations (principally by asking subjects and informants). 19,34,35 While four studies 30,31,34,39 used the latest revision of the consensus criteria 16 that included suggestive features, no study described the methods used for their detection.
Seventeen studies did not address possible overlap between Alzheimer's disease (AD) and DLB. The five that did used dissimilar methodologies and reported on differing aspects of that relationship. Though the total number of cases per study was small, it was reported that a third 23 to a half 38 of DLB cases also met AD criteria, that the ratio of "combined" AD and DLB to "pure" DLB ranged from 0.75 30 to 7.0, 32 and that the ratio of AD with parkinsonism to probable DLB was 0.69. 34 Eleven studies gave data on Parkinson's disease dementia (PDD) as well as DLB, 20,21,25,26,28,32,33,35,[38][39][40] with seven providing the criteria used to diagnose PDD (i.e., six 20,21,25,28,35,38 utilized DSM-IV criteria 42 and another 39 used published PDD criteria 46 ). In one report, 35 the authors adhered to the "one-year rule" in distinguishing DLB from PDD, 14,16 while in another, 38 no restriction on the duration of parkinsonism prior to the onset of dementia was used (the remaining studies did not comment on this issue). In three studies, 35,38,40 DLB was more common than PDD, while PDD was more common in five 21,26,32,39 (the conditions in two studies 20,25 were equally common, and the final one 28 made no attempt to distinguish between the two).
Among those aged 65+ residing in the community, DLB incidence rates ranged from 0.5 to 1.6 per 1000 person-years. Little could be said about the influence of age on incidence rates, as all incidence studies dealt with similarly aged individuals (i.e., 65+). One study reported that no increase in incidence was observed with increasing age. 30 All incidence studies scored 5+ on the quality assessment.
DLB accounted for 3.2 to 7.1% of incident dementia cases (mean 4.6%, SD 1.5). The percentages ascribed to DLB were similar in the study 31 that used 1996 consensus criteria 12 compared to the three 28,29,32 that utilized the 2005 revision 14 (4 and 5.3%, respectively). Likewise, the percentages were similar when the two studies published in 2005 or earlier (3.6%) were compared to the three published in 2006 or later (5.3%). The one study that reported on both probable and possible cases 34 attributed 4.5% of dementia cases to DLB if both were included, compared to 2.0% if cases were restricted to those with probable DLB.
Point and period prevalence estimates varied more widely: from a low of 0.02 in a community and institutional study from Japan to a high of 63.5 per 1000 in a community study from Australia (Figures 2 and 3). If analysis was limited to the 12 studies scoring 5+ on the quality assessment, prevalence per 1000 varied from 0.02 to 33.3 per 1000 (Table 3). [20][21][22][23][24]28,29,[35][36][37][38][39] Estimates rose with increasing age. In studies restricted to participants less than 65 years of age, prevalence was 0.02-0.06 per 1000 compared to 0.3-6.45 when participants were 60 or 65+, 8.6-33.3 among those 70+, and 63.5 in the one study limited to those 81+. This latter study 19 had the highest reported prevalence of DLB and a high proportion of dementia cases attributed to this condition (probable and possible DLB accounted for 53/78 [67.9%] of all dementia cases). Other unique features of the study included the use of CDR criteria to detect dementia and the severity of dementia among participants diagnosed with DLB (mean Mini-Mental State Examination [MMSE] score of 12.6). This study was assigned a quality score of 4 (see Table 3) because of methodological concerns (i.e., a single non-examining physician made the diagnosis based on a review of assessment information). 19 The study with the second highest prevalence reported on the second oldest group of participants, 35 and, again, when present, dementia was of moderate severity (mean MMSE score of 14.1).
DLB made up between 0.3 and 24.4% of all cases of dementia in the prevalence studies when the most restrictive definition for DLB provided was used (mean 6.4%, SD 6.1). The proportions were similar in the 11 studies 19,[22][23][24][25][26][27][28][29]35,40 using 1996 consensus criteria 12 and the one 39 that utilized the 2005 revision 14 (7.8 and 5.3%, respectively). Likewise, the percentages were similar when the nine studies published in 2005 or earlier (7.5%) were compared with the eight reports published in 2006 or later (4.4%). Three studies provided information on both probable and possible DLB cases. 19,35,38 Compared to restricting case definition to only probable DLB, including possible and probable cases increased by 1.5-3.2 fold the proportion of dementia cases attributed to DLB.

DISCUSSION
We updated and/or expanded on prior systematic reviews on the incidence and prevalence of DLB. 15,16 This allowed us to highlight a number of methodological issues that need to be addressed in order to derive more precise estimates of the true burden of this form of dementia. We could not perform a pooled meta-analysis due to the significant between-study heterogeneity and small sample size. While we found that DLB incidence rates ranged from 0.5 to 1.6 per 1000 person-years, prevalence varied more widely and ranged from 0.02 to 33.3 per 1000 even when restricted to higher-quality studies. The two other DLB systematic reviews 15,16 reported similar results, while one that dealt with PDD estimated that in the general population in the 65+ group its prevalence was between 2 and 5 per 1000. 47 The uncertainty about the prevalence of DLB is at least partially due to methodological issues. Most of the included studies (n = 18) were not designed to specifically investigate the incidence and/or prevalence of DLB. The cognitive testing used to screen for cases had been primarily selected to detect AD (see Tables 1 and 2). How well they worked in detecting potential cases of DLB is unknown. The differing diagnostic criteria for dementia used would influence how often dementia would be detected. 48 While a clinic-based study reported that the 2005 consensus criteria for DLB identified more individuals with a dementia as suffering from probable DLB compared to the 1996 version (essentially by classifying more cases as probable and fewer as possible DLB), 3 it is unknown whether they are more accurate. There is a suggestion that the diagnostic error rate for DLB has recently increased. 49 We did not find marked differences between the two versions of diagnostic criteria in terms of the  proportion of those with dementia identified as having DLB. How diagnostic criteria are operationalized is likely of more importance than which specific one is employed. Most of the studies we examined (n = 14) did not specify whether they were reporting on probable DLB or possible as well, and they provided little if any information on how core and suggestive features of DLB were detected. Challenges in assessing these features (especially fluctuations) contribute to the difficulties encountered in accurately diagnosing DLB. 13 The age ranges studied varied markedly. With an age-associated condition like DLB, this would have an effect on the estimates provided, which is what we found. Finally, the number of cases upon which estimates were based was often small (see Tables 1 and 2), thus decreasing their precision. 50 Issues exist with the consensus clinical criteria used for diagnosis, as the core and suggestive features are not specific to DLB. For example, parkinsonism, especially if mild, can be found with normal aging, idiopathic Parkinson's disease and AD. 51 Autopsy-based validation studies of these diagnostic criteria have shown variable but often low sensitivity and high specificity with stage-dependent false negative (usually early in the course of the illness) and false positive errors (especially with advanced dementia). 49 Particular challenges arise when there are coexisting pathologies such as AD. 52,53 The complex relationship between DLB and AD is still being elucidated. Patients with DLB commonly have β-amyloid deposition and diffuse plaque formation without a significant burden of neocortical neurofibrillary tangles. 1,14 Concomitant Alzheimer's pathology is not excluded by a diagnosis of either probable or possible DLB. Another issue is the relationship between PDD and DLB. They are typically differentiated by the arbitrary "one-year rule" (i.e., DLB is diagnosed if dementia occurs within 12 months of the onset of extrapyramidal motor symptoms, while the diagnosis is PDD if the clinical history of parkinsonism is longer than a year). 12,14 The clinical and pathological overlaps with both Alzheimer's and Parkinson's raises uncertainty in the diagnosis of DLB. The term DLB, while acknowledging the presence of Lewy bodies, does not specify their relative importance compared to other brain pathologies in explaining a patient's cognitive symptoms. It is not clear how cases with presumed "mixed" aetiologies, which are common with Lewy body disease, should be dealt with. As a minimum, the approach taken by the investigators should be described. This was often not mentioned in the studies reviewed. Validated biomarkers for DLB could improve on the accuracy of diagnosing DLB but are not generally available for clinical use. 54 Biomarkers may not be practical for population-based studies, and difficulties would remain in dealing with cases having more than one contributing aetiology and distinguishing DLB from PDD and atypical parkinsonian syndromes.
Notwithstanding these limitations, we can conclude that clinically defined DLB becomes more common with increasing age and accounts for about 5% of all dementia cases encountered in older populations. This updated information can be used in the planning of healthcare services and educational programs for practitioners. DLB presents unique diagnostic challenges that likely contributed to the wide variation in reported rates found in the literature reviewed. Standardization in the collection and reporting of data, how diagnostic criteria are operationalized and the approach taken in dealing with the borders between DLB and both AD and PDD would improve on the reliability and comparability of future epidemiologic studies of this condition.