USDA NUTRIENT COMPOSITION DATA: ACCURACY "MAY BE IN QUESTION"

USDA NUTRIENT COMPOSITION DATA: ACCURACY "MAY BE IN QUESTION" because of the methods used by the Department of Agriculture to assess the nutrient value of foods, the General Accounting Office concluded in a report entitled "Better Guidance to Improve Reliability of USDA's Food Composition Data," which was released on Nov. 22. The questionable reliability of the USDA data, GAO determined, could have "implications for users, such as the federal agencies that use the data for food consumption studies." The report was requested by House Science, Space, and Technology Committee Chairman George Brown (D-Calif.) to evaluate the reliability of the Human Nutrition Information Service's (HNIS) food composition data base, known as "Handbook 8." In addition to the "considerable latitude" with which HNIS staff determine the "amount and type of scientific information needed to qualify data for entry into the Handbook 8 data base," the report comments, "data have been accepted into the data base with little or no supporting information on the testing and quality-assurance procedures used to develop the data." In the report, GAO listed three reasons for the unreliability of the USDA data: First, the report notes, HNIS does not set down specific criteria against which the principal investigators of the HNIS' Nutrient Data Research Branch can evaluate "food composition data obtained from industry and scientific literature." The House committee, in a Nov. 22 release, said that often "this information [from the industry] is not accompanied by quality-assurance documentation." In addition, the two HNIS documents used to guide the evaluation process "are so general," the report maintains, that HNIS "cannot ensure that investigators will apply the same standards in reviewing data for inclusion into Handbook 8." Second, GAO determined that the HNIS investigators have used data that "did not have sufficient supporting information on the testing and quality-assurance procedures used to develop the data." GAO noted that officials from several federal agencies and private laboratories suggested that scientific validity of nutrient composition data "requires a review of [five] quality assurance measures used to produce that data." And last, GAO found that HNIS investigators "do not appropriately direct contracted studies." Investigators do not "regularly visit the laboratories to ensure that the required procedures and methods are being used," the report notes. In addition, control samples are not properly disguised, "thereby reducing the usefulness of this quality control effort," GAO said. The report points out that HNIS-contracted studies only require that two samples be analyzed for food composition data. FDA and Food Safety and Inspection Service officials, GAO said, have determined that "two samples are not sufficient for inclusion in Handbook 8." Laboratory experts interviewed by GAO stated that "at least six samples should be analyzed before data are accepted," according to the report. GAO submitted recommendations to the committee suggesting that: (1) specific quality-assurance criteria be established to evaluate data obtained from other sources; and (2) procedures be developed to better direct the nutrient analyses provided by HNIS' contracts. In response to a draft form of the report, HNIS Acting Administrator Janice Lilja concurred with the two recommendations suggested by GAO and promised that HNIS will "move forward as quickly as possible to implement them." Lilja "generally agreed" with GAO's findings, the report notes, but cited a lack of funding as the primary reason for the inadequacies in the methods used to develop food composition data. "To generate the amount of completely documented data that GAO suggests is necessary is well beyond the HNIS budget," Lilja told GAO. She said that average sample analyzed to HNIS standards costs $ 2,000 and that to analyze six samples, as recommended by GAO, would cost $ 12,000 per food item. She noted that there are currently 5,300 separate food items in Handbook 8. The Nutrient Data Research Branch received only $ 200,000 to support laboratory analysis of specific food items in FY 1993 and HNIS' budget for FY 1994 has been cut from $ 13 mil. in FY 1993 to $ 11.1 mil., Lilja pointed out.