Research report: Open-source tool helps researchers study unprocessed and processed pork and beef separately

An open-source R package called PFclassify produced processed-meat intake estimates comparable to an established National Cancer Institute classification method while enabling researchers to distinguish unprocessed and processed pork, beef, chicken, turkey and other meat and poultry. The peer-reviewed study evaluated the software with nationally representative dietary data from the 2017–2020 pre-pandemic National Health and Nutrition Examination Survey, including 11,816 people aged 2 years or older. This was a descriptive methods-validation study, not a clinical trial, and it did not test changes in diet or health outcomes.

The package and its analytical code are publicly available. Nutrition studies frequently combine red and processed meats into broad groups, which can prevent investigators from examining intake by individual meat subtype. PFclassify addresses that analytical problem with a standardized workflow for classifying and quantifying meats and poultry in dietary data recorded with United States Department of Agriculture food codes.

The software separates unprocessed red meat into pork, beef and other red meat, while processed red meat is divided into processed pork, processed beef and other processed red meat. It also classifies unprocessed poultry and divides processed poultry into chicken, turkey and other processed poultry. Researchers can therefore calculate subtype-specific intake measures rather than relying only on totals for red meat, poultry or processed meat.

The approach builds on a standardized method previously developed by the National Cancer Institute to separate processed meat from processed red meat and processed poultry. PFclassify extends that framework by splitting the broad processed red-meat category into pork, beef and other red meat, and by separating processed poultry into chicken, turkey and other processed poultry. Its contribution is thus a more detailed classification within an established structure.

The study describes the software as an R package designed to give researchers an automated and reproducible workflow for red-meat and poultry intake in NHANES and other datasets that use USDA food codes. For the demonstration, the researchers analyzed dietary information from 11,816 NHANES participants aged 2 or older in the 2017–2020 pre-pandemic survey cycle. The investigators used these nationally representative data for descriptive analyses of meat intake.

At the broader-category level, they compared PFclassify with the classification method established by the National Cancer Institute. This comparison assessed whether the new package could reproduce established processed-meat estimates while adding a finer breakdown by subtype. The published results concern classification and intake estimation; the research did not evaluate whether any classification leads to different nutritional recommendations, clinical interventions or disease-risk estimates.

PFclassify and the earlier National Cancer Institute method produced close agreement in the reported NHANES demonstration. Differences were less than 5% for meat-category composition and less than 0.5 ounce equivalents per day for absolute intake estimates. These figures apply to the reported comparison in the 2017–2020 pre-pandemic sample.

They show that the software’s broad estimates were comparable to those from the established method, while its additional purpose was to allow processed and unprocessed pork and beef, among other subtypes, to be examined independently. The numerical agreement does not by itself establish that every subtype assignment is equally precise in every dataset, nor does the study assess any health effect associated with a meat category. PFclassify, its documentation, example data and the analytical code used in the study are available through a public GitHub repository.

This makes the classification workflow accessible to investigators working with NHANES or other USDA-coded datasets and allows the implemented rules and analyses to be inspected. The authors present the package as a framework that researchers can apply, test and refine. Its open availability is intended to support more standardized, reproducible analysis across studies, while the quality of any resulting estimate remains tied to the data and the classification workflow used.

PFclassify was developed with funding from the National Pork Board and the National Cattlemen’s Beef Association, which is a contractor to the Beef Checkoff. The study’s disclosure statement also reports research grants, editorial roles, royalties and advisory positions held by some authors, including relationships involving the National Pork Board. The paper itself was published in the Journal of Food Composition and Analysis, and the supplied conflict-of-interest statement identifies competing interests held by individual authors.

They do not, in the supplied evidence, establish that the organizations altered the results; the reported validation consists of the package demonstration, the comparison with the National Cancer Institute method and the numerical agreement in the study. The research was classified as data and statistical analysis and used one nationally representative, pre-pandemic NHANES cycle for descriptive purposes. It did not involve an intervention, a randomized exposure or follow-up designed to measure health outcomes.

The agreement of less than 5% in category composition and less than 0.5 ounce equivalents per day in absolute intake is a specific result for the reported comparison, not a universal error rate for all USDA-coded datasets. Its function is to standardize the classification and quantification step for meat and poultry subtypes. PFclassify’s main reported result is therefore both technical and practical: in nationally representative dietary data from 11,816 people aged 2 or older, the package matched the established National Cancer Institute method at the broad processed-meat level within the reported tolerances, while providing separate estimates for unprocessed and processed pork, beef, chicken, turkey and other categories.

The study does not show that the tool changes the foods people eat, improves dietary quality or alters measured health risk. Instead, it demonstrates a public, reproducible software workflow for a more specific treatment of meat and poultry intake in research that uses USDA food codes. By making pork and beef available as distinct analytical categories rather than components of a single pooled group, PFclassify offers investigators a way to pose and compare more detailed dietary questions within the limits imposed by the underlying data and the classification method.

Subject of Research: Validation of open-source software for classifying processed and unprocessed meat and poultry intake in USDA-coded dietary data

Article Title: Open-source tool helps researchers study unprocessed and processed pork and beef separately

Article References: Open-source tool helps researchers study unprocessed and processed pork and beef separately. Original publication (Bibliographic details unavailable.)

Image Credits: AI Generated

DOI: Not provided

Keywords: PFclassify, R package, dietary classification, processed meat, unprocessed meat, pork, beef, poultry, NHANES, USDA food codes, open-source software, methods validation