To assess the reproducibility of a 135-item self-administered semi-quantitative FFQ.

Control subjects who had previously completed an FFQ relating to usual dietary intake in a nationwide case–control study of cancer between November 2003 and April 2004 were randomly selected, re-contacted, and invited to complete the same FFQ a second time approximately one year later (between January and April 2005). Agreement between the two FFQ was compared using weighted kappa statistics and intraclass correlation coefficients (ICC) for food groups and nutrients. Summary questions, included in the FFQ, were used to assess overall intakes of vegetables, fruits and meat.

General community in Australia.

One hundred men and women aged 22–79 years, randomly selected from the previous control population.

The weighted κ and ICC measures of agreement for food groups were moderate to substantial for seventeen of the eighteen food groups. For nutrients, weighted κ ranged from 0·44 for starch to 0·83 for alcohol while ICC ranged from 0·51 to 0·91 for the same nutrients. Estimates of meat, fruit and vegetable intake using summary questions were similar for both survey periods, but were significantly lower than estimates from summed individual food items.

The FFQ produced reproducible results and is reasonable in assessing the usual intake of various foods and nutrients among an Australian adult population.