Bias and error in the determination of common macronutrients in foods: interlaboratory trial.

Chemical analyses of nutrient values in foods form the basis of much of the science and practice of nutrition and dietetics, but little is known about the accuracy and precision of common macronutrient analyses. Therefore, an interlaboratory study was set up. One batch each of egg powder, full-fat milk powder, whole rye flour, whole wheat flour, biscuits, and French beans (snap beans, "haricots verts") was thoroughly homogenized. Samples were sent to 19 leading food analysis laboratories in Europe and the U.S., and each performed analyses of macronutrients by its own routine methods. Most were government or semi-government laboratories and major contributors to national nutrient data banks. The results for dry matter content and ash agreed well between laboratories. For protein, the coefficient of variation between laboratories (CV between) ranged from 2.8% to 6.4%. The reproducibility within laboratories was sometimes quite poor. The CV between for total fat ranged from 5.4% to 54%. For "available" carbohydrates, the CV between ranged from 9% to 27%. The CV between for total dietary fiber ranged from 23% to 84%. Only part of the variability could be explained by the use of methods of different principle. It is concluded that leading laboratories produce widely different values for macronutrients in common foods. Quality control programs and reference materials of certified nutrient concentration are urgently needed.