Use of quantified and frequency indices of vitamin A intake in a case-control study of lung cancer.

Recent publications have examined the extent of food frequency data needed to estimate nutrient intakes in epidemiological studies. The need for amount (usual portion size) data to supplement information on average frequency of intake has been questioned. In a case-control study of risk factors for lung cancer, we have collected data on frequency, amount and past pattern of intake for common or rich sources of vitamin A. These data have been combined with standard content information to calculate three different types of indices: one based on frequency alone, a quantitative index that included both frequency and amount, and a past-weighted index that combined the frequency, amount and data on relative consumption in the past. Odds ratios by tercile of carotene, retinol and total vitamin A intake varied little between the frequency index and the quantified and past-weighted indices. Higher intakes of carotene and total vitamin A were associated with a lower risk for lung cancer regardless of index type. This consistency of results is explained by a stability of relative intakes regardless of index type. Spearman rank order correlation coefficients between the frequency and quantitative indices exceeded 0.90 for total vitamin A, carotene, and retinol. This stability is attributable to parallel trends of increased frequency of consumption and portion size at higher levels of nutrient intake. Because similar trends were observed in five main food groups included in these analyses, these findings may be generalizable to other nutrients.