The MAVIS trial design and recruitment has been described in more detail elsewhere . 910 men and women aged 65 years and over who had not taken vitamin, mineral or fish oil supplements within three months of recruitment (one month for supplements of water-soluble vitamins other than vitamin B12) were recruited from six primary care health centres in North-East Scotland between February and December 2002. 97% of participants in both supplemented and placebo groups were living in the community at recruitment and all were considered by their family doctor to be suitable for the study. We did not exclude any participants on the basis of impaired cognitive function, though those with dementia were unlikely to volunteer or would have been excluded by their doctor. We did not collect information on the educational status of participants but this age group would all have attended formal education for at least seven years.
Participants were randomly allocated to receive either a mineral and vitamin supplement or matching placebo for 12 months, with minimisation on health centre, age, gender and place of residence (living in the community or in care). 456 participants (239 men, 217 women) received the active supplement and 454 participants (240 men, 214 women) received the placebo. The supplement contained 800 μg vitamin A, 60 mg vitamin C, 5 μg vitamin D, 10 mg vitamin E, 1.4 mg thiamin, 1.6 mg riboflavin, 18 mg niacin, 6 mg pantothenic acid, 2 mg pyridoxine, 1 μg vitamin B12, 200 μg folic acid, 14 mg iron, 150 μg iodine, 0.75 mg copper, 15 mg zinc, and 1 mg manganese. The participants were asked to take one tablet daily for a 12-month period, with compliance checked by reported consumption at monthly intervals. The nutrient content of the tablet was chosen to be similar to most other commonly used supplements purchased over-the-counter and the dosage was according to the manufacturers' instructions. A 10% random sample provided tablets for tablet counting. Risk of iron, folate, vitamin C or vitamin D deficiency at baseline was assessed using a simple 17 item nutrition risk questionnaire for assessing risk of micronutrient deficiency in older people .
Cognitive function was assessed in all participants by digit span forward and verbal fluency tests carried out face-to-face at recruitment and over the telephone at the end of the intervention period. The digit span forward test requires the participant to repeat a sequence of random digits pronounced at a rate of approximately one per second, beginning with two sequences of two digits and continuing with two sequences of increasing length up to a sequence of nine digits or to failures on both tests of a shorter sequence. One point is scored for every correct sequence repeated with a maximum score of 16 points . The verbal fluency test requires the participant to name words beginning with each of three separate letters in an exactly-timed one-minute period. The three letters used (either P, R and W or C, F and L) were randomly allocated at the baseline assessment and the other three letters used at the one year assessment. The score is the total number of words named correctly, excluding proper nouns and words with the same start but different endings .
Analyses were conducted with adjustment for baseline values and the trial minimisation factors where appropriate. Intention-to-treat analyses were conducted for trial group comparisons. Statistical significance was based upon the 95% confidence interval. For the two pre-planned trial sub-group analyses (those aged 75 years and over and those at increased risk of deficiency assessed by the nutrition risk questionnaire) the 99% confidence interval was used to give the correct width of the 95% confidence interval.
The study protocol was approved by Grampian Research Ethics Committee and all participants gave written consent to take part.