Response

Extract

We thank Dr. Nyberg for his interest in our recent correspondence to the Journal (1). In 1993, we reported a pronounced effect of saturated fat intake on risk of lung cancer (2). In that analysis, we entered both saturated fat and total calories as categoric variables in a standard multivariate model. Subsequently, we were persuaded that the standard multivariate model exaggerates the true variation in fat intake. Either the nutrient residual or the nutrient density approach provides a better estimate of the true variation in fat intake when data are modeled as quantile-categoric variables (3,4). Nyberg has asked us to address two questions. Was energy adjustment necessary for dietary constituents other than saturated fat? If we had not adjusted for total calories or used a different method of adjustment, would we have arrived at different conclusions?

Nyberg suggested that we examine risk estimates unadjusted for energy for a variety of dietary variables presented in two tables. The data are provided in Table 1 for selected dietary constituents. First, we examined energy-providing nutrients, all of which were highly correlated with total calories (Spearman correlations between .8 and .9). Unadjusted for total calories, both fats and carbohydrate were directly related to risk of lung cancer. When we examined fat and carbohydrate as a percent of total calories consumed, the fat and saturated fat associations persisted, but carbohydrate was now inversely associated with risk of the disease (odds ratios across increasing quintiles 1.0, 1.19, 0.87, 0.80, 0.75; two-sided P for trend .09), suggesting that these macronutrients should be energy adjusted. By use of the standard multivariate approach, risk associated with fat was increased, whereas risk associated with intake of carbohydrate disappeared (2). Two food group variables— 1) beans and peas and 2) citrus fruit and juice—were independently associated with risk of the disease; neither was affected by energy adjustment. Spearman correlations with total calories were .2 for beans and peas and .3 for citrus fruit and juice. In the present study, we also included yellow and green leafy vegetables, because the protective effect of this food group is well established for lung cancer. We saw no clear benefit of frequent consumption of these foods either with or without energy adjustment. Similarly, b-carotene was not protective or affected by method of energy adjustment.