Data sources

We aggregate data from five primary source types, each with different reporting conventions and authority signals:

  1. Government testing programs. FDA Total Diet Study, EPA UCMR 5 drinking water testing, NOAA marine environmental monitoring. These are authoritative, free, and consistent over time.
  2. Peer-reviewed academic research. Studies indexed in PubMed and similar databases, with documented methodology and replication. We prioritize meta-analyses and large-N studies over single-lab work.
  3. Independent lab investigations. Commissioned testing by Consumer Reports, the PlasticList project, Environmental Working Group, and Mamavation. Quality varies; we note methodology limitations where they exist.
  4. Regulatory action and policy. Federal and state laws, EPA rulemaking, EFSA opinions, WHO guidelines. Tracked separately from contamination data.
  5. Journalism. Investigative reporting that includes original testing or document review. Cited but treated as secondary to underlying sources where available.
We do not invent or interpret results. Every value displayed traces to its source study with full citation.

Ingestion process

Source data arrives in many formats: FDA publishes XLSX spreadsheets, academic studies publish CSV supplements (or none at all), advocacy organizations publish HTML tables. Our ingestion process normalizes all of these into one consistent schema:

  1. Each source is parsed into row-per-measurement format, with one row per (sample × contaminant × study).
  2. Sample identifiers are linked to canonical food entries via FDA Total Diet Study naming or manual mapping.
  3. Below-detection-limit reporting is preserved as a first-class value rather than substituted with zero — distinguishing "tested and not found" from "not tested."
  4. Method detection limits and limits of quantitation are stored with the study, not the value, since they vary by analyte and matrix.

Unit normalization

Different sources report concentrations in different units. PFAS is typically reported in parts per trillion (ppt, equivalent to ng/kg). Phthalates and bisphenols are reported in nanograms per gram or per serving. Microplastics are reported in particles per gram, particles per serving, or by mass. Total organic fluorine (a PFAS screening measurement) is reported in parts per million.

We do not auto-convert across measurement methods. Cross-class comparisons — "salmon has more PFAS than tea has microplastics" — would require assumptions about particle size, mass, and exposure pathway that source studies rarely justify. On display, we group results by their original unit and compare only within compatible measurement methods.

Detection status

Every measurement has one of five status values:

  • Detected: numeric value at or above the limit of quantitation.
  • Trace: numeric value between the method detection limit and limit of quantitation. Reported by some sources as italicized values.
  • Below MDL: tested for, but no detectable concentration found.
  • Not tested: analyte not included in this study's panel. Distinct from "below MDL."
  • Invalid: contamination, instrument error, or other quality issue reported by the source.

Limitations

Three things this database does not do:

  1. We do not estimate population-level exposure. Detection counts represent samples tested, not foods in the supply. A food appearing in many studies has been investigated, not necessarily contaminated more frequently in everyday consumption.
  2. We do not provide medical or dietary advice. Whether a detected concentration matters for an individual depends on diet, age, health status, and pathways outside food. Consult a healthcare provider for personal exposure questions.
  3. We do not compare values across testing methods that are not directly comparable. Particle counts and mass concentrations measure different things even when both are called "microplastics."

Corrections and updates

If you find an error, please email us with the study citation and the corrected value. We maintain an audit trail: corrected values are versioned, not silently overwritten. Original source documents are linked from every value, so anyone can verify our parsing.

The database is updated on a rolling basis as new datasets become available. Most major sources update annually; FDA's Total Diet Study and EPA UCMR 5 publish in batches.