ABC1921 participants were all born in 1921 and underwent an IQ-type test (the Moray House Test No 12) in a Scottish-wide survey of schoolchildren in June 1932. Full details of the ABC1921 study design and test battery are published elsewhere [2.9,10,11] but in 1997, with permission from Grampian Research Ethics Committee, at the start of "Wave 1" of the ABC1921 study, 297 individuals living in or near Aberdeen were identified and 234 agreed to be assessed. By Wave 5, which was conducted between March 2003 and March 2004, 107 individuals were still in touch with the study. The median dates (participant numbers) for the first, second, third, fourth and fifth Waves respectively were May 1998 (234), September 1999 (207), March 2001 (162), March 2002 (146) and August 2003 (107). It should however be noted that 47 individuals were recruited to the ABC1921 cohort for the first time in Wave 2 and a further 16 were recruited in Wave 3 as refreshment of the sample in early Waves.
The 1932 Scottish Mental Survey  provides unique pre-morbid mental ability data relevant to a cohort at risk of age-related diseases. Under the auspices of the Scottish Council for Research in Education (SCRE), all children at school in Scotland on June 1st 1932 and born in the calendar year 1921 undertook a group-administered mental ability test, including some practice items. 87,498 (44,210 boys and 43,288 girls) were tested on this Moray House Test (MHT) that comprised a wide range of items with a maximum score of 76 and includes verbal reasoning, numerical, spatial and other items . The scores on the 1932 Moray House Test were validated by individually re-testing a representative sample of 1000 of the children (500 boys, 500 girls) on the Stanford Revision of the Binet-Simon test. In addition, at first attendance as adults the following cognitive tests were administered: Mini-Mental State Examination (MMSE – a general cognitive screening test), Raven's Progressive Matrices (RPM – a measure of non-verbal reasoning), Rey Auditory Verbal Learning Test (AVLT – a measure of verbal declarative memory), Wechsler Adult Intelligence Scale subtests Block Design (BD – a measure of visuo-spatial ability) and Digit Symbol coding (DS – a measure of processing speed), and Uses of Common Objects Test (UCO – a measure of executive function) as detailed previously .
We recorded data on marital status, living group, postal address, usual occupation (before retirement) and years of education. Occupations were classified using the UK Registrar General's Classification of Occupations  (1990). We used the ecological method of Carstairs and Morris  (1990) to assign a socio-economic deprivation index to each postal address. This method has advantages over occupational classification especially among older women. The usual paternal occupation (or breadwinner) in 1932 was recorded and classified as either (1) professional administrative; (2) skilled manual or (3) unskilled manual or unemployed .
The number of persons living in the home at age 11, the number of public rooms in the home, whether at age 11 the participant was routinely required to share a bed (BS), indoor or outdoor sanitation (SN) and how many people shared this toilet were recorded (SS). An "overcrowding index" (OI) was derived by dividing the total number of usual residents in the house by the number of public rooms.
Information on disease history and prescribed medication was recorded at interview. The research nurse completed a clinical examination that included pulse, systolic and diastolic BP (mean of three occasions, sitting), height, weight, best of three of Forced Expiratory Volume in one second (FEV1), Forced Vital Capacity (FVC) and Peak Expiratory Flow Rate (PEFR) measured using a microspirometer. Subjects were classified as never smoker, past smoker, or current smoker. In addition we administered the Hospital Anxiety and Depression scales (HADS) scoring on both anxiety and depression subscales .
This was determined for a census date of 12 May 2007 from Scottish Community Health Indices (CHIs). Deaths are flagged on each CHI from the General Register Office for Scotland via computerised links provided by ISD Scotland. However the vital status of participants no longer listed on Scottish CHIs remained undetermined.
After initial data checking and description, univariate associations with mortality were explored using Chi-square test (with continuity correction where appropriate) or Fisher's exact test for categorical variables and analysis of variance or Wilcoxon's test for non-categorical variables. Following this Cox's proportional hazards modelling was performed. As a first step the effect of age was tested for in this narrow-age cohort. Then models were built according to the proportion of participants with available data for each variable, entering those variables with least missing data first. This resulted in socio-demographic variables being examined prior to cognitive variables. At each stage we applied backward elimination with p > .1 criterion for variable removal. All analyses were undertaken using the SPSS 14.0 statistical package.