The size of the study sample population required to reach 100 qualified participants per decile for Cam-CAN Stage 2 is expected to vary by age when accounting for exclusion and refusal, estimated population data, clinical based experience and estimates of individuals who may refuse to participate in neuroimaging. Numbers are adjusted for the proportion of the general population with exclusion criteria including MR safety contraindications (e.g. pacemakers), learning disability (living at home), cognitive impairment (Mini-Mental State Examination (MMSE) [8 (link)] score of 24 or less) and reduced response from individuals with limited longstanding illness or disability. Proportions are estimated based on data from the Office of National Statistics (ONS), the Medical Research Council Cognitive Function and Ageing Study (MRC-CFAS) [9 ] and the National Health Service (NHS) registrations. We assume that only 30% of the population will undertake the initial interview and of those who do, 40-50% will agree to take part in Stage 2 (age dependent). Numbers predicted to be needed for Stage 1 are shown in Table 2 . The age group above age 88 are recruited to the same population proportion as the 78-87 decile, in order to enable cohort comparison with other population-based studies and investigation of the rare group of oldest old who are experiencing healthy ageing.
The Cam-CAN structure provides sufficient sample size in each decile to separate age-related change from other sources of individual variation. A number of different comparisons can hypothetically be undertaken using this structure. All hypotheses are investigated at a power of 80% and α = 0.05: for linear regression, assuming the continuous data are standardised to a N(0,1) distribution, 100 per decile enables us to investigate i) a linear decline of ±0.04 across the age range; ii) a difference in linear regression slope of size ±0.06 between two risk factor groups with a prevalence of 50% (such as gender); iii) differences in the mean values of two groups (defined with 50% prevalence) of ±0.2; iv) for dichotomous outcomes with prevalence of 0.5 in one group to detect a difference of at least ±0.1. This sample is sufficiently large to be able to detect non-linear change with age, such as a change in rate of decline, and the required size to detect stability with age (to exclude a slope of up to ±0.03 per decile). Multiple hypotheses can also be undertaken, such that linear decline of slope 0.1 can be detected for 100 independent investigations protecting the type I error rate (false positives).
Contact | 750 | 775 | 850 | 950 | 1250 | 1400 | 2850 | 1700 |
Interview | 250 | 250 | 275 | 300 | 400 | 450 | 850 | 500 |
Estimates include numbers per decile to be contacted and interviewed.