Epidemiology of early onset dementia and its clinical presentations in the province of Modena, Italy

Patients with early onset dementia (EOD), defined as dementia with symptom onset at age <65, frequently present with atypical syndromes. However, the epidemiology of different EOD presentations, including variants of Alzheimer's disease (AD) and frontotemporal dementia (FTD), has never been investigated all together in a population‐based study. Epidemiologic data of all‐cause EOD are also scarce.

tailored to older patients with different social and family situations. 3 In addition, patients with EOD wait longer than patients with late-onset dementia before receiving a correct diagnosis after symptom onset, probably because they are not referred to dementia centers soon enough and because reaching a correct diagnosis in these patients is more challenging, 4 since they frequently present with atypical manifestations of dementia syndromes. 5 As an example, in young patients, Alzheimer's disease frequently presents with a number of non-amnestic variants, including posterior cortical atrophy (PCA) and the logopenic variant of primary progressive aphasia (lvPPA), compared to the more common amnestic presentation of dementia due to Alzheimer's disease (AD) in older patients. 6 EOD prevalence has been variably reported ranging between 38 and 420 EOD cases per 100,000 in the age group 30 to 64, 7 whereas EOD incidence has been reported to vary between 2.4 and 22.6 new cases per 100,000 per year. 8,9 However, no previous population-based studies on EOD have reported the epidemiology of different presentations or phenotypes of AD and frontotemporal dementia (FTD) spectrum.
Providing the epidemiology of all the different clinical presentations of EOD would not only benefit medical professionals in their diagnostic reasoning when faced with young patients with cognitive symptoms.
More importantly, it would also allow a better understanding of the impact of different presentations on society and an improved planning of dementia services and resource allocation.
We aimed to establish the prevalence and incidence of all-cause EOD as well as of EOD presentations in a Northern Italy community by studying demographic and clinical features of patients with dementia symptom onset before the age of 65.

METHODS
We conducted an epidemiological study in the province of Modena, are followed periodically in their CDCD at least every 6 months (this is also related to the Italian regulation regarding prescription of cholinesterase inhibitor and neuroleptic medications that requires periodic medical checks). Because it is not infrequent that young patients with cognitive disorders and other neurological accompanying symptoms may be first referred to movement disorders or motor neuron disease clinics, all these clinics from the two neurology services of the province were also involved in the recruitment.
All these services extensively cover the entire province and are part of the Italian National Health System (Sistema Sanitario Nazionale).
Patients with cognitive symptoms who are younger than 65 years are usually referred to neurologic CDCDs, whereas older patients are referred to geriatric CDCDs. Referrals to CDCDs can be made by either general practitioners or specialists such as Psychiatrists (as frequently is the case for patients presenting with behavioral symptoms). The dementia care network is organized so that patients 65 and older can also be referred to neurologic CDCD (either by general practitioners or geriatric CDCD) in case of diagnostic uncertainty, whereas EOD patients with severe functional impairment or living in nursing homes are referred to geriatric CDCD.
We identified retrospectively all EOD patients from January 1,2006 to December 31, 2016, and prospectively all new EOD patients from January 1, 2017 to June 30, 2019. We adopted a mixed recruitment strategy because we were interested not only in EOD epidemiology, but also wanted to estimate the needs of all the living patients and their families with the ultimate aim of optimizing resource allocation. Therefore, after having received ethical approval for the current study, we commenced prospective recruitment of all new EOD cases but also searched retrospectively for all the living cases diagnosed in the previ- (CADASIL). 23 The diagnosis of AD was supported by at least one biomarker suggestive of amyloid deposition (either CSF or amyloid-PET) in the large majority of cases. The macro-classification of different syndromes was based on Elahi and Miller. 24 Because clinical criteria may have evolved over time, for the retrospective part of the study the diagnoses were systematically assessed and harmonized according to the most recent criteria, upon consensus between the neurologists reviewing the cases. As an example, living cases of the logopenic variant of PPA (or lvPPA), which would have been included in the FTD spectrum prior to 2011, upon review were reclassified as a variant of dementia due to AD. Dementia cases that did not meet criteria for a specific type of dementia were classified as "not otherwise specified" (or NOS).
We collected demographic and clinical characteristics for each EOD case. Education was defined as the number of years of education.
Age at symptom onset was defined as the referred age (by patient or caregiver) when the first cognitive or behavioral symptom had been observed by the patient themselves or by the caregiver. Age at diagnosis was defined as the age when the patient was diagnosed with dementia for the first time, irrespective of the specific type of dementia syndrome. For patients with a clinical diagnosis of mild cognitive impairment (MCI) or mild neurocognitive disorder (ie, presence of measurable deficit on at least one cognitive domain that does not affect everyday functioning) the age at diagnosis was the age when they converted to dementia, regardless of the fact that biomarkers suggestive of a specific underlying neurodegenerative pathophysiology may have also been available in the MCI phase (eg, MCI with evidence of underlying AD pathophysiology). 25 We computed both crude and age-and sex-adjusted incidence rates with reference to the entire period January 1, 2016 through June 30, 2019, and to each of these years. For both calculations, we used the Modena province resident population on January 1 of each year from 2016 to 2019, 26 after exclusion of previously diagnosed EOD cases from the denominators. The rate was then directly standardized to the 2013 European standard population, 27 taking into account age and sex.
We focused on the most recent years to avoid biases due to possible changes in diagnostic sensitivity over time.
We computed the crude and age-and sex-adjusted prevalence rate at census date (June 30, 2019), using as denominator the residing population on January 1, 2019, which was the most recent available.
To allow comparison with previous studies, we computed incidence and prevalence rates for the age groups 30 to 44 and 45 to 64 years, and for the entire population at risk, 30 to 64 years.
When reporting incidence and prevalence rates on 100,000 inhabitants, we considered as denominator the entire population of Modena province from age 0, whereas when reporting incidence and prevalence rates on 100,000 persons at risk, we used as denominators the resident population in the corresponding age subgroups (  Notably, the rate increased with age from 1.0 in the 40 to 44 age group, to 6.7 in the 50 to 54 group, to 59.8 in the 60 to 64 group.

EOD prevalence
We

Clinical variants
AD was the most frequent clinical diagnosis, followed by clinical syndromes of the FTD spectrum, VaD, and LBD (

DISCUSSION
We report the incidence and prevalence of all the phenotypes of EOD   to 64. 8 They are also comparable but slightly smaller to those obtained from the Rochester Epidemiology Project, 28 which also included cases with cognitive impairment secondary to brain tumors or chronic mental illness that, instead, we purposefully excluded from our study. Our crude incidence for the 45 to 64 age group (22.1/100,000) is almost twofold higher than the incidence of 11.5/100,000 reported by a study of all cases seen in an hospital-based study in the UK, 29 possibly reflecting a higher rate of physician consultation and specialist referral in Italy compared to other European countries. 29 We found a prevalence of all-cause EOD of 74.3/100,000 persons at risk in the age range 30 to 64 and 119.9/100,000 persons at risk in the age range 45 to 64. This prevalence overlaps with those obtained in two recent studies conducted in Norway 7 and Australia 30 but is greater than those reported in older studies conducted in the UK 31 and Japan. 32 Of interest, even when considering the diagnoses of AD and FTD dementia only, we found a greater prevalence compared to that reported by the only prevalence study on EOD conducted in Italy, in the Brescia province. 33 That we found greater prevalence relative to older studies may reflect the general improvement in dementia identification and diagnosis seen over the past decade worldwide, which is a consequence not only of educational interventions to improve primary care practice, 34 but also of the general improvement in the diagnostic ability through the use of biomarkers, which were not easily available in clinical practices a decade ago. In addition, the high prevalence reported in the present study may be related to the high level of diagnostic accuracy and to the widespread recruitment based on an inclusive and easily accessible dementia care network. Alternatively, the greater prevalence may reflect a really larger incidence and/or survival in our population.
We found that the most frequent cause of EOD is AD dementia. This is consistent with previous epidemiological studies on EOD incidence 8,35 and with most studies on EOD prevalence. 7,36 The few prevalence studies that found that AD dementia was the second prevalent cause of EOD after VaD were either not population-based but conducted on hospital-based cohorts, 37,38 or carried out in countries such as Japan, 32 with a known greater incidence of stroke relative to the Caucasian population in presenile ages. 39 A recent EOD prevalence study, which found AD second to alcohol-related dementia, may have been biased by over-diagnosis of the latter by clinicians who were not dementia specialists, as stated by the authors. 30 Of interest, in our series, there were no differences in diagnostic delay between the different clinical syndromes, suggesting that even challenging cases such as those presenting with behavioral symptoms were identified promptly by the study network.
With regard to the clinical variants of AD dementia, their incidence or prevalence has not been investigated in population-based studies yet, possibly because their characterization is relatively recent and only subsequent to the advent of AD biomarkers. 15,16 In our population-based study the non-amnestic variants represented 34% of all AD dementia cases, with lvPPA being the most frequent. This is comparable to one of the two previous hospital-based studies conducted on cohorts of patients referred to dementia clinics, which reported a relative proportion of non-amnestic presentations of 32%, 40 but is lower than the one conducted on consecutive cases seen at a specialist dementia center, which reported a proportion of nonamnestic presentations of 64% 6 and might have been biased by a focus on atypical presentations. As for the FTD spectrum, one previous study conducted in [2007][2008][2009] reported the incidence of bvFTD, svPPA, CBS, and PSP, but not of nfvPPA and ALS-FTD. 8 Our incidence values are consistent for CBS and PSP, but higher for bvFTD and svPPA, possibly reflecting the fact that the refinement of diagnostic criteria for the latter variants was subsequent to that study and that our diagnosis was based on an in-depth clinical assessment and supported by the use of biomarkers. There are no comprehensive studies on prevalence of the clinical presentations of the FTD spectrum.
A major strength of our study was the ability to merge several sources of health data from a capillary network of centers for the diag- Finally, we are immensely grateful to Prof Paolo Nichelli, who mentored us throughout the study and provided insightful comments on the study design and manuscript.

FUNDING INFORMATION
The study was supported by a grant "Dipartimenti di eccellenza 2018-

CONFLICTS OF INTEREST
The authors report no conflicts of interest.