This study is a longitudinal prospective cohort analysis of a group of patients identified in a neurodegenerative disorders research clinic in Subiaco, Western Australia, from 1 January 1994 until 31 January 2011: The Artemis Project. The Artemis Project is an attempt to study the neurobiology of early onset dementia of all causes in Western Australia. Participants consisted of 155 patients with clinically confirmed sporadic early-onset AD or FTD, with onset prior to 65 years, who gave informed consent and attended the clinic on more than one occasion.
Patients with early onset AD were diagnosed using NINCDS-ADRA criteria and supported by structural and functional imaging (FDG PET) and neuropsychometry.3
Patients were diagnosed with FTD using existing published criteria7
and refined as new technology (such as FDG PET) became available6
and at annual review using contemporary international guidelines.3
With this approach, no patients have been excluded and our neuropathological sensitivity and specificity for the diagnosis of AD and FTD=100%.
The FTD group comprised only patients with the behavioural variant of FTD for uniformity of analysis; patients identified with semantic dementia (n=1), primary progressive aphasia (n=8) and motor neuron disease–FTD complex (n=2) were not included.
A detailed history was recorded from the patient at baseline, including education, family history and medical history. This information was then corroborated with the spouse or primary carer, and the general practitioner. Additionally, during follow-up this information was updated if the patient characteristics changed; for example, if a patient developed cancer.
Age at onset was determined from self and carer/spouse report of the onset of symptoms. Education is defined as self-reported years of formal education. Family history is defined as a self-reported family history of dementia in first-degree and second-degree relatives. Cerebrovascular risk factors are hypertension, hypercholesterolaemia, increased girth, obesity, diabetes, smoking, excessive alcohol consumption, coronary heart disease and peripheral vascular disease, the measurement of which has been described elsewhere.9
The presence of at least one of these risk factors was determined from self-report and medical notes. The presence of mental illness (depression or psychosis) or other comorbidities was also determined through self-report, carer information and medical notes.
Mini mental state exam (MMSE) was performed at each visit as a measure of cognition.10
Global Deterioration Scale (GDS) and Total Functional Capacity (TFC) staging were determined for each visit as markers of the severity of the dementia and abilities to perform acts of daily living.11
The ratings were performed by the same trained team of cognitive assessors.
DNA screening was performed on our patients with an autosomal-dominant pattern of inheritance—nucleotide changes in the coding sequences of the amyloid precursor protein, progranulin gene, presenilin-1 and presenilin-2 genes, Tau and SIGMAR 1 gene were performed by direct sequencing of polymerosechoid reaction (PCR) products derived from genomic DNA. Nucleotide sequence information from each PCR product was obtained from both strands and possible mutations were verified by an independent amplification of the PCR product and resequenced. ApoE genotypes were determined by restriction fragment polymorphism analysis of PCR amplified products. Individuals identified with presenilin-1 mutation (Q222H), progranulin mutations and SIGMAR 1 mutations are not included in this study and will be reported elsewhere.
Statistical analyses, including χ2 tests for difference in proportions, t tests for difference in means and Kaplan-Meier survival analysis, were performed using Stata VII. As the MMSE and other variables are not normally distributed, the t test was not used and the Kolmorgorov-Smirnov test employed to test for differences in means. No assumptions were violated for the log-rank test, nor drawing of the Kaplan-Meier curves—this is a non-parametric method. The Proportional Hazard Model was not used, which expects the baseline hazards for the two groups to be proportional.