Test Kit Research & Comparison Center
Evidence methodology and corrections
How each claim in the guide is graded, checked and corrected. Last reviewed 21 October 2026.
Principles
- Every figure is read from the primary source by a person, on the stated page and table, before it is published. Figures are never copied from secondary summaries, retailer pages or automated metadata.
- Claims are kept separate by device, analyte, cutoff, sample type and study design, with their denominator and context.
- Grades apply to individual claims. There is no overall "most accurate" ranking, and none will be added.
- Where a source's own figures disagree, we show what it printed alongside what its counts give, and do not silently correct it.
- What we could not verify is marked "not verified" or left out. We do not estimate.
- P in the Cup does not perform the testing or certify these products.
Evidence tiers
How claims are graded
Shown next to each claim, never as a global ranking.
| Tier | Definition | Used today for |
|---|---|---|
| A | Independently published head-to-head comparative study for the identical product, analyte, cutoff and matrix, with documented reference method and limitations; no extrapolation across products. | The 2022 study of three point-of-care devices, for those devices only |
| B | FDA review or decision summary, or published validated laboratory data, for the identical device or variant; we specify if the underlying tests were conducted by the sponsor or manufacturer. | K091588 and K091612 figures (sponsor studies reviewed by FDA) |
| C | Manufacturer package insert or supplier study with complete methodology, not independently verified. | None yet |
| D | Product specification sheet or retailer marketing claim only; never shown as a numerical accuracy figure without robust underlying support. | None |
| E | Consumer purchase review: experience only, not a performance grade. | None |
| Unverified | Missing exact device match, insufficient design or sample documentation, or unclear result interpretation. | K221871 performance (no data in the FDA review; applicant summary not yet extracted); the 2014 on-site study (product not named in the abstract) |
A tier describes how strong the evidence for one claim is. It is not a score for a product, and a tier A claim about one device says nothing about another.
Reviewer checklist
Completed for every source before any of its figures is published:
- Source link opened and working; publisher is the regulator, journal or manufacturer it claims to be.
- Exact device or study identified: name, applicant or authors, model or format, and version where stated.
- Intended use recorded from the source: over the counter, prescription or point of care, laboratory, or research.
- For each figure: analyte, matrix, cutoff, sample size, study design, reference method, denominator, and the PDF page and table label.
- Who ran the study: manufacturer or sponsor, independent researchers, or FDA review of sponsor data.
- Raw agreement kept distinct from sensitivity, specificity and predictive values; reading usability kept distinct from analytical performance.
- Data not reported in the source labelled "not reported" or "not available", never estimated.
- Subscriber-only or copyrighted tables linked, not reproduced.
- No retailer, white-label or resold product linked to a clearance without documentary evidence of an exact match.
- Source checked for revisions or recalls; verification date recorded.
Link checks
Source links are checked monthly; the first check was on 21 October 2026, when every link on these pages opened. If a link breaks, we show the citation with an archived or updated registry link and note the change in the corrections log; we never substitute a document about a different model. The evidence data behind the table is published at /assets/data/evidence.json, so anyone can audit it.
What the catalog contains
Public regulatory and scientific information only. It contains no participant data, purchase identities, test results, group memberships or behavioural marketing identifiers.
Corrections log
What changed, and when
Significant corrections stay listed here rather than being silently edited away.
| Date | Change |
|---|---|
| 21 October 2026 | Guide launched with five records (FDA 510(k) K091588, K091612 and K221871; two independent studies). All figures checked against the primary sources. Noted in the source, not corrected: seven printed percentages in the K091588 tables differ slightly from the table's own counts (shown on the record page); the K091612 buprenorphine table labels its 7.5 ng/mL row "25%" although 7.5 ng/mL is 75% of the cutoff; the K091612 decision summary spells the applicant "UPC BioSciences, Inc.". The 2014 on-site study was graded Unverified because its abstract does not name the product. The test kit guide moved from /test-kits/ to /test-kit-guide/. |
To report an error: A contact address will be published here before launch. Please do not send test results, health information or anything personal.
Page last updated