Researchers bought 30 self-tests off UK shelves. Most could not justify their claims

Home test kits have moved from pharmacy counters to supermarket shelves, and the category now covers far more than pregnancy and Covid-19: vitamin levels, thyroid function, menopause, bowel cancer, fertility. Two companion papers published in The BMJ asked a simple question about that market — do the boxes tell a buyer enough to use the test sensibly, and is there evidence the tests work?

The researchers answered it by shopping. They identified self-tests in supermarkets, pharmacies and health and wellbeing shops within a 10-mile radius of the University of Birmingham's Edgbaston campus during 2023, finding 35 and obtaining 30 [s2].

Those 30 tests used seven different sample types and measured 20 different biomarkers across 19 conditions [s1][s2].

What the packaging tells you

The first paper assessed the information on boxes and in instructions-for-use leaflets, judged against what a buyer would need for an informed decision [s1].

Information to guide purchase was mostly absent. Only 8 of 30 boxes (27%) said who should use the test and when; 7 (23%) said what to do about the result; and 10 (33%) gave any numerical measure of test performance [s1].

From the box or the leaflet combined, 21 tests (70%) were judged to be intended for diagnosis and 15 (50%) for screening [s1]. Of the diagnostic ones, 3 of 21 (14%) gave no information about symptoms; of the screening ones, 10 of 15 (67%) gave no information about risk factors to guide who should test [s1]. Those are the two pieces of information that determine whether a test result means anything for a given person.

Follow-up advice was asymmetric. Twenty-seven tests (90%) recommended seeing a healthcare professional after a positive or abnormal result; only 14 (47%) did so after a negative or normal one [s1].

The most consequential finding is the last. For 11 of the 19 conditions covered (58%), the researchers judged the intended use to run contrary to evidence-based guidance in at least one respect — the population being tested, the frequency of testing, the threshold used, or the investigative approach appropriate for that condition [s1].

What the evidence behind them looks like

The second paper went after performance claims and the studies supporting them [s2].

Accuracy claims appeared in the instructions-for-use documents of 24 of 30 tests: accuracy for 19, sensitivity for 17, specificity for 16 [s2]. The claims were confident. Performance of 98% or better was claimed on accuracy by 53% (10 of 19), on sensitivity by 41% (7 of 17) and on specificity by 63% (10 of 16) [s2].

Where a reference standard was reported, it was often not much of one: 29% (5 of 17) evaluated the self-test's accuracy against a similar rapid test [s2]. Comparing a rapid test against another rapid test does not establish accuracy against the underlying condition.

The researchers then sought the underlying study documents. They obtained nine sets of clinical and lay-person study reports, covering 12 tests [s2]. Assessed with QUADAS-2, a standard tool for diagnostic accuracy studies, 73% of domains across those documents were rated at unclear risk of bias because of poor reporting, and 58% were rated as having high applicability concerns because of inappropriate study designs [s2].

Descriptions of participants were particularly inadequate in the clinical study reports, and even the lay-person reports presented few demographics — up to four [s2]. Some populations were unrepresentative of the intended user, inappropriate reference standards and thresholds were used, and mentions of blinding were scarce [s2].

The authors also note that manufacturers were unwilling to provide public access to study documents, which they describe as raising ethical concerns [s2].

Usability, which is where errors happen

Eighteen of the 30 tests carried at least one high-risk usability or safety concern [s2]. Eleven of those concerns arose from the equipment, 10 from the sampling process, and 15 from the instructions or the interpretation of results [s2]. (The categories overlap; a single test could raise concerns in more than one.)

This is the part of the analysis most likely to be underweighted by buyers. A test with genuinely good analytical performance in a laboratory can still produce a wrong answer at a kitchen table if the sample is collected badly or the result is read wrongly — and the second paper's conclusion says exactly that: concerns about ergonomics and usability might lead to test errors [s2].

What the papers conclude

The first paper's conclusion is that the current self-test market does not support consumers in making informed decisions about use, interpretation and subsequent action, and that clinicians working downstream of self-tests are likely to face important challenges incorporating the results into practice [s1]. The second concludes that improved regulatory oversight and clearer standards are urgently needed to protect the public from poorly performing diagnostic self-tests [s2].

Both papers use the word "urgently", and both frame the reason the same way: use of self-tests is rising [s1][s2].

Limits worth stating

This is a snapshot of one retail geography in one year. Tests were identified within 10 miles of a single university campus in 2023 [s1][s2], and the market changes quickly — products are added, withdrawn and reformulated, and regulatory requirements can change.

The judgments about intended use running contrary to evidence-based guidance are assessments made by the review team against guidance documents, not empirical measurements of harm [s1]. Nobody in these studies was tested, misdiagnosed or followed up. What the papers establish is that the information supplied and the evidence available are inadequate for the purpose the products are sold for — not that a given number of people were harmed.

And the performance claims were not independently verified. The researchers assessed what manufacturers claimed and what documentation supported it [s2]; a claim resting on a poorly reported study is not thereby shown to be false.

The practical point

A self-test result is an input, not an answer, and the two questions that determine what it means — should this person have been tested at all, and what should follow from the result — are the two the packaging most often fails to address [s1].

This article is informational and does not constitute medical advice.

Sources

• [s1] Direct-to-consumer self-tests sold in the UK in 2023: cross sectional review of information on intended use, instructions for use, and post-test decision making — The BMJ, published online 23 July 2025. https://doi.org/10.1136/bmj-2025-085546
• [s2] Direct-to-consumer self-tests sold in the UK in 2023: cross sectional review of regulation and evidence of performance — The BMJ, published online 23 July 2025. https://doi.org/10.1136/bmj-2025-085547