The Prevalence and Incidence of Frontotemporal Dementia: a Systematic Review

Abstract Background Population-based prevalence and incidence studies are essential for understanding the burden of frontotemporal dementia (FTD). Methods The MEDLINE and EMBASE databases were searched to identify population-based publications from 1985 to 2012, addressing the incidence and/or prevalence of FTD. 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-six studies were included. Methodological limitations led to wide ranges in the estimates for prevalence (point prevalence 0.01-4.6 per 1000 persons; period prevalence 0.16-31.04 per 1000 persons) and incidence (0.0-0.3 per 1000 person-years). FTD accounted for an average of 2.7% (range 0-9.1%) of all dementia cases among prevalence studies that included subjects 65 and older compared to 10.2% (range 2.8-15.7%) in studies restricted to those aged less than 65. The cumulative numbers of male (373 [52.5%]) and female (338 [47.5%]) cases from studies reporting this information were nearly equal (p=0.18). The behavioural variant FTD (bvFTD) was almost four times as common as the primary progressive aphasias. Conclusions Population-based estimates for the epidemiology of FTD varied widely in the included studies. Refinements in the diagnostic process, possibly by the use of validated biomarkers or limiting case ascertainment to specialty services, are needed to obtain more precise estimates of the prevalence and incidence of FTD. RÉSUMÉ Prévalence et incidence de la démence fronto-temporale : une revue systématique du sujet. Contexte : Les études de population sur la prévalence et l’incidence sont essentielles à la compréhension du fardeau associé à la démence fronto-temporale (DFT). Méthodologie : Nous avons cherché dans les bases de données MEDLINE et EMBASE les articles publiés entre 2000 et 2012 portant sur l’incidence et/ou la prévalence de la DFT dans la population. Nous avons également examiné les références des articles inclus dans notre étude ainsi que celles des revues systématiques antérieures. Deux évaluateurs ont examiné tous les résumés et le texte intégral des publications et l’extraction des données, et ils en ont évalué la qualité. Résultats : Vingt-six études ont été retenues. Des limites méthodologiques expliquent les écarts dans les estimations de prévalence (prévalence ponctuelle de 0,01 à 4,6 par 1 000 ; prévalence d’une période donnée de 0,16 à 31,04 par 1 000) et incidence (0,0 à 0,3 par 1 000 personnes-années). La DFT constituait en moyenne 2,7% (écart de 0 à 9,1%) de tous les cas de démence dans les études de prévalence qui incluaient des sujets de 65 ans et plus par rapport à 10,2% (2,8 à 15,7%) dans les études portant sur des sujets âgés de moins de 65 ans. Les nombres cumulatifs d’hommes (373 [52,5%]) et de femmes (338 [47,5%]) tirés des études dans lesquelles cette information était mentionnée étaient pratiquement égaux (p=0,18). La variante comportementale DFT était presque quatre fois plus fréquente que les aphasies progressives primaires. Conclusions : Les estimations basées sur la population en ce qui concerne l’épidémiologie de la DFT étaient très variables dans les études que nous avons retenues. Il faudra raffiner le processus diagnostique, possiblement par l’utilisation de biomarqueurs validés ou limitant la constatation des cas à ceux confirmés par des services spécialisés, pour obtenir des estimations plus précises de la prévalence et de l’incidence de la DFT.


INTRODUCTION
Frontotemporal dementia (FTD) is a clinically and pathologically heterogeneous group of non-Alzheimer neurodegenerative dementias characterized by progressive decline in behaviour (behavioural variant FTD [bvFTD]) and/or language (primary progressive aphasias, such as semantic dementia and progressive non-fluent aphasia [PNFA]) associated with degeneration of the frontal and anterior temporal lobes. 1,2 These presentations can overlap with atypical parkinsonian disorders (i.e., corticobasal syndrome [CBS], progressive supranuclear palsy [PSP]) and amyotrophic lateral sclerosis [ALS]). Though a variety of terms have been employed to describe these conditions (e.g., frontotemporal degeneration, frontotemporal lobar degeneration [which some restrict to pathologically confirmed cases], dementia of the frontal type, Pick complex disorder), in this paper we will use FTD. First described in the late 19th century, 2 FTD is generally held to be a relatively common cause of early-onset (<65 years) dementia. 3 Dementia clinic-based studies suggest that FTD may account for approximately 5% of all dementia cases, 4 but this may represent an inaccurate estimate of the proportion of dementia cases arising from this condition because of selection or referral bias. 5 In the absence of autopsy confirmation, the diagnosis of FTD is dependent on searching for clinical diagnostic features. In research and practice, the most commonly used criteria for detection of FTD over the last 20 years were those developed by the Lund and Manchester Groups, 6 Gregory and Hodges, 7 Neary et al. 8

and the Work Group of Frontotemporal Dementia and
Pick's Disease (also referred to as the McKhann criteria). 9 Revised diagnostic criteria have been recently developed. 10,11 Population-based prevalence and incidence studies are essential for understanding the societal burden of FTD and planning for the range of healthcare services needed for those with this condition. In this paper, we report on a systematic review of population-based prevalence and incidence studies of FTD.

METHODS
This is one in a series of systematic reviews on the prevalence and incidence of priority neurological conditions identified by the Public Health Agency of Canada and the Neurological Health Charities Canada as part of the National Population Health Study of Neurological Conditions. 12

Search Strategy
The systematic review was conducted according to a predetermined protocol based on the PRISMA Statement for systematic reviews and meta-analyses. 13 The search strategy (see Appendix A) was developed by the study authors, who possess extensive expertise in dementia and/or epidemiology, in consultation with a research librarian experienced in the performance of systematic reviews. The initial MEDLINE and EMBASE search was conducted in February of 2011 and then updated in May of 2012. The review was restricted to studies written in English or French and published from the year 2000 or later for international studies and 1985 or later for Canadian studies. References were exported and managed using EndNote X5.

Study Selection
Two reviewers independently screened all abstracts in order to identify original research that appeared to be reporting on the prevalence or incidence of dementia. These papers were selected for full-text review. Studies were excluded at this stage if the abstract clearly indicated that the study was not population-based.
Two reviewers independently performed the full-text reviews. Articles were included in this systematic review if they met the following criteria: (1) represented original research; (2) were population-based (i.e., involved a defined "general population" as opposed to a specific hospital-or clinic-based population); and (3) reported an incidence and/or prevalence estimate of FTD. English and French articles were screened and reviewed in a similar fashion by reviewers fluent in English and French, respectively. The references of included articles were hand searched for additional articles. All additional articles were evaluated in a manner identical to what has been previously outlined. The references of systematic reviews and literature reviews on the epidemiology of dementia were also hand searched. Disagreements were resolved by consensus, with involvement of a third party if necessary (this step was never required).

Data Extraction and Study Quality
Two reviewers extracted data from included articles using a standard data collection form. Agreement was reached on all items. If multiple articles reported data on the same study population, the most comprehensive data were utilized. In cases where the studies reported on different time frames or subgroups (e.g., by sex and/or age), all data were included. Demographic data retrieved included age, sex and study location. Source/type of clinical data and the definition/diagnostic criteria used for the diagnosis of FTD were noted. Incidence and prevalence estimates of dementia from each study were recorded, along with any stratification by age, sex or year of data collection. The quality of the included studies was evaluated using an assessment instrument 14,15 (see Appendix B). This instrument included an evaluation of sample representativeness, condition assessment and statistical methods. Each study was given a quality score that ranged from 0 to 8 (higher being better).

Data Analysis
Pooled meta-analyses were not done due to significant betweenstudy heterogeneity and small sample size. Forest plots presenting the distribution of study estimates were produced. As it is held that FTD disproportionately affects middle-aged individuals, 2 we compared studies that included older participants (65 + ) with those that restricted themselves to younger (<65) individuals. All statistical analyses were carried out in R version 2.14. The meta package was employed to produce the forest plots. Depending on the methodology of the study, incidence proportion, incidence rate, period prevalence and point prevalence are provided.

Identification and Description of Studies
The search strategy yielded 16,066 citations (8743 from MED-LINE, 7323 from EMBASE, with 7923 remaining after the removal of duplicates) (Figure 1). A total of 707 articles were selected for full-text review. Of the 176 studies (total of the original and updated searches plus the hand search of included articles) that presented data on the prevalence and/or incidence of dementia and its subtypes, 26 reported on FTD and were included in our systematic review.
The characteristics of the 26 included studies are summarized in Tables 1-3. Nineteen 16-34 reported on prevalence, six 35-40 on incidence and one study provided data on both the incidence and prevalence of FTD. 41 Sixteen studies presented data from Europe, seven from Asia, two from South America, and one from North America. While a variety of approaches was employed, most studies were either surveys of residents of a specific location (n = 12) or based on cases identified by specialty services serving a defined catchment area (n = 8). The Neary 8 (n = 16) and/or Lund and Manchester 6 (n = 9) criteria were the diagnostic criteria most commonly used, with diagnosis typically based on an assessment by a healthcare professional (n = 22), often coupled with imaging studies (n = 18) and/or laboratory investigations (n = 15), and/or health record review (n = 10).
Among studies restricted to individuals less than 65 years of age, 17,23,27,30 point prevalence was in a narrower range (0.07-0.30 per 1000). Studies that surveyed a defined population 22,24,26,[32][33][34] generally reported higher estimates than those based on enumerating cases identified by specialty services 17,18,20,30 (range of estimates in surveys 0.13-4.61 per 1000 compared to 0.15-0.40 per 1000 for specialty services), but the latter were in a narrower range.
Six studies 16,19,21,25,29,41 reported on period prevalence ( Figure 3). Period prevalence estimates ranged from 0.16 to 31.04 per 1000. The Gislason study 21 reported a strikingly high prevalence. A number of features of this study are unique: a single-phase survey approach was used; only a relatively small number of participants were examined (n = 494); the age range studied (85-86) was both narrow and quite advanced; and the focus of the study was detecting what was termed a frontal lobe syndrome, with these individuals diagnosed as having bvFTD if they did not meet exclusionary criteria, irrespective of whether they met DSM-III-R criteria for a dementia (9/14 cases did not). Estimates derived from surveys 16,19,21,29 tended to be higher     FTD accounted for 2.7% (range 0-9.1%) of all dementia cases in the prevalence studies that included individuals older than 65 years of age 16

Incidence of Frontotemporal Dementia
Only one study reported on incidence proportion. 41 This community-based study of subjects between 42 to 92 years of age consisted of 29,357 persons followed for 6 years. Incidence proportion was estimated to be 0.11 per 1000.

Sex
In the 13 studies providing information on the sex of those with FTD, 17

Type of Frontotemporal Dementia
Among studies that supplied information on the frequency of the type of FTD detected, 18,20,30,[35][36][37] bvFTD (n = 299, 79.7% of the total of 375 cases, which included one classified as FTD with ALS) was approximately four times as common as the primary progressive aphasias (n = 75, 20%; 32 semantic dementia, 43 PNFA). A study that provided population estimates based on FTD cases admitted to state psychiatric hospitals serving defined catchment areas 25 not surprisingly reported that all had behavioural symptoms (61% had behavioural only and 39% both behavioural and language symptoms). Two studies 31,37 reporting on a total of 356 individuals with FTD noted that 12 (4.6% of the total number of FTD cases) also had a diagnosis of ALS (note that individuals with ALS were included in the FTD total). Five studies 23,[33][34][35]37 reporting on a total of 95 individuals with FTD found an additional 5, and 17 study participants diagnosed with CBS and PSP, respectively (note that individuals with these conditions were not included in the FTD total).

DISCUSSION
We found that population-based prevalence and incidence estimates for FTD were generally low and varied widely, especially among older individuals. FTD accounted for a higher proportion of dementia cases among younger (<65) individuals. In the studies reviewed, there was no apparent predisposition based on sex, and bvFTD was the most common form encountered, and relatively few individuals were felt to have CBS, PSP or ALS.
Two previous systematic reviews 42,43 based on a smaller number of studies reported similar incidence and prevalence estimates. Knopman and Roberts 42 reviewed five prevalence and three incidence studies. For those 45-64 years of age, point prevalence estimates varied tenfold, from 0.02 to 0.22 per 1000, while incidence rates were between 0.027 and 0.041 per 1000. Onyike and Diehl-Schmid 43 identified seven prevalence and three incidence studies. Their reported prevalence (0.02-0.31 per 1000) and incidence (0.013 and 0.167 per 1000) rates also ranged widely. After completion of our systematic review, three otherwise eligible studies 44-46 that reported on the incidence and/or prevalence of FTD were published. Their estimates fall within the range we report and do not change our main findings.
We suspect that the limitations of the clinical diagnostic criteria used in included studies [6][7][8][9] and concerns about how the criteria were operationalized partially explain the wide ranges in the estimates of prevalence and incidence. These criteria have been criticized for (among other things): (1) their large number of features (some of which were very rare or of debatable validity); (2) lack of guidance as to the number of features required for diagnosis and the relative importance of symptoms; (3) placing greater emphasis on behavioural manifestations compared to language (one of the studies reviewed 19 only considered behavioural symptoms, while the five studies that solely utilized the Lund and Manchester criteria 23,24,27,35,41 would be expected to preferentially detect bvFTD); (4) ambiguity in the description and time frame of behavioural manifestations; (5) need to infer some aspects of the person's state (e.g., assessment of the lack of insight); (6) uncertainty on how best to assess cognitive (traditional executive measures vs. social cognition and decision-making tasks) and behavioural (e.g., objective vs. subjective) characteristics; (7) rigidity in how criteria were applied; (8) limited role for supportive features; (9) no estimate for level of diagnostic certainty (e.g., probable or possible); (10) impact of the exclusionary criteria; (11) relative neglect of imaging and genetic characteristics; and (12) their insensitivity for early disease. 11,[47][48][49] Patients with bvFTD are frequently misdiagnosed as suffering from a psychiatric illness early in the course of their illness 50,51 and referred to mental health services. 52 As an accurate clinical diagnosis of FTD can be   very difficult to make upon presentation, long-term follow-up may be needed to establish its presence. 51,53,54 Autopsy studies indicate that the criteria used in our included studies lack sensitivity. For example, Neary criteria 8 were positive in only 79 of 152 (52%) autopsy-confirmed cases of bvFTD, 11 with particular issues among those over the age of 65. The exclusionary features (e.g., early severe amnesia, spatial disorientation) eliminated 26 confirmed cases. Another study 55 of these criteria found that they had a low sensitivity (36.5%) at the time of the person's initial presentation compared to their final clinical diagnosis. The revised diagnostic criteria 10,11 will likely have improved sensitivity, 11,56 but possibly at the cost of worse specificity and a heightened risk of misclassifying individuals as suffering from FTD when they in fact have a frontal variant of Alzheimer's disease, 57 other neurological causes or a psychiatric diagnosis. 58 Further validation of these revised criteria is qrequired. The newer criteria were not utilized in the studies we reviewed.
Experienced clinicians working in specialty clinics are able to accurately diagnose FTD. 59,60 Clinical acumen can be supplemented by appropriate use of investigations. For example, finding frontotemporal abnormalities without corresponding ones in more posterior brain areas on functional (e.g., single-photon emission computed tomography, positron emission tomography) or structural (e.g., magnetic resonance imaging) neuroimaging studies can improve on the sensitivity of clinical criteria in detecting FTD. 61 Basing estimates on patients seen by specialty services who are comprehensively investigated would partially address concerns about the validity of the diagnosis, but at the potential cost of missing cases. A proportion of affected individuals in the catchment area of the study may not be seen by the services utilized in identifying cases, as they may be referred elsewhere (though some authors minimize this possibility because of their conviction that the high regard local practitioners have for their service means that suspected cases will be referred to them 17,18,37 ), are not experiencing the types of symptoms that would lead to a referral, or are unwilling to be seen.
We did not find a lower FTD prevalence or incidence among older (65 + ) compared to younger (<65) individuals. Five studies 18,20,30,31,36 reporting estimates by age subgroups found the highest prevalence and incidence rates between 60 and the mid-70s. The study 21 with the oldest subjects included in our systematic review reported the highest prevalence of FTD, but this investigation had other unique features that may have contributed to the high estimate. Studies published subsequent to our review have reported high prevalence 44 and incidence 45 rates in older populations. While we found that FTD made up a larger proportion of dementia cases among those less than 65 compared to older patients, this appeared to be driven more by the exponential increase in prevalence and incidence with advancing age of other neurodegenerative causes of dementia, in particular Alzheimer's disease (AD), than a higher incidence of FTD among those less than 65 compared to older individuals. Differentiating FTD from AD can be particularly challenging at more advanced ages. Compared to younger individuals with FTD, older (65+ ) persons with this condition tend to have more memory and visuospatial deficits suggestive of AD while showing less pronounced frontal and temporal lobar atrophy on imaging studies. 62,63 The perception that FTD becomes less common as we age may be due to the increasing difficulty in differentiating it from other forms of dementia.
Our systematic review was not restricted to population-based studies with autopsy confirmation of the clinical diagnosis.
There are but few of these studies-a systematic review published in 2006 64 could only identify six. Their restriction to high-income countries, questions about the generalizability of their results and the relative rarity of FTD coupled with the limited number of brains being collected mean that these extremely valuable studies cannot fully address questions about the population-based prevalence and incidence of FTD. As well, in an era without specific disease-modifying therapies, the patient's clinical profile rather than their underlying pathology will be driving service provision.
Notwithstanding the limitations noted above, our systematic review of the incidence and prevalence of FTD updates and expands on prior work. Because of the nature of their symptoms, it has been argued that individuals with FTD, especially early-onset cases, 17,35 will be referred to assessment and management services. While relying on figures from these sources will underestimate the overall prevalence and incidence of this condition, collating data from specialty services might still be an efficient way of capturing data on patients requiring assistance  from the healthcare system. The Cambridgeshire studies of early-onset dementia found that all those with FTD in contact with the healthcare system were known to local specialist services, 30,37 while in Sweden 45 diagnoses of FTD were almost exclusively made by specialist clinics. Standardization of methods and refinements in the diagnostic process, possibly by the use of validated biomarkers, will hopefully improve the precision of prevalence and incidence estimates of this challenging condition.