select search filters
briefings
roundups & rapid reactions
Fiona fox's blog
A study published in The BMJ assesses direct-to-consumer self-tests sold in the UK.
Prof Amitava Banerjee, Professor of Clinical Data Science and Honorary Consultant Cardiologist, Institute of Health Informatics, UCL, said:
“Direct-to-consumer, self-tests are increasingly used by people with and without disease for screening and are widely available from high street vendors. In these rigorous, real-world studies led by the University of Birmingham, we see two main findings. First, across 30 self-tests in 19 conditions from infertility and menopause to raised cholesterol and anaemia, there is a not enough information for consumers to judge when and why to do the test, and how to interpret or how to act on the results. Second, the evidence and the support from clinical guidelines to use these tests is often lacking, suggesting that regulatory oversight needs to be improved.
“Sometimes people use self-tests because they “feel it is better to know” and they are trying to inform their health and healthcare decisions. This research shows that these self-tests are often not providing relevant knowledge or information and they are not informing decisions in the right way. Therefore, all stakeholders need to consider the quality of self-tests and information available to members of the public or health professionals before recommending their use, whether in the health and wellness space or in diagnosis, treatment and prevention of disease.”
Rachel Richardson, Acting Head of Methods Support, Evidence Production and Methods Directorate, The Cochrane Collaboration, said:
“This well-conducted research shines a welcome light on an area of healthcare which appears to be inadequately regulated.”
Prof Kevin McConway, Emeritus Professor of Applied Statistics, The Open University, said:
“I think the findings of these new studies on self-tests for health conditions, available (at a cost) in supermarkets, high street chemists and online, are scary and concerning. I don’t doubt the findings of the researchers, that many of the available tests don’t make it clear who could make good use of them, how accurate the results might be, or what someone should do in the light of their results.
“These are good studies in my view. The researchers do list some limitations in the discussion sections of the papers, in particular that their samples of tests were obtained two years ago and were not specifically intended to be a sample of what was available across the country, but given what they do say about where they got the tests, I’d be surprised if they aren’t pretty much the same anywhere nowadays. Also, the researchers didn’t check with representatives of the public whether the instructions were as unhelpful to understanding as they believe they were, but I don’t think this affects their conclusions.
“I’m certainly not saying that tests like this should be banned, or even radically discouraged. The authors of these research papers aren’t saying that either. Experience during the heights of the Covid pandemic showed how useful home testing could be, particularly when access to other information about one’s health might not be easily available (as can still be the case at some GP practices, for instance). And, generally as a default position, I don’t like telling people they can’t do something that they want to do – though only in the light of clear, transparent and easily available information on the pros and cons, and in the presence of adequate regulation. These studies make it clear that users of many self-tests aren’t given easy access to relevant information, and that the regulation isn’t appropriate at present.
“I’ll just mention one particular aspect, because it’s one that I have studied and written about myself. This is about why the findings are important, not about the quality of the research. No diagnostic or screening test for a health condition can be 100% accurate. There will inevitably be false positives – people with a positive test result for the condition who actually don’t have the condition – and false negatives – people with a negative test result for a condition who actually do have the condition. These are aspects of accuracy, though discussions of that word don’t always make it clear enough that there are two different ways in which a test result can be wrong.
“You probably recall some of the interest and media discussion about these things in relation to Covid testing. Not all of the discussion was logical or well argued, but it clearly and correctly drew attention to the fact that test results can be wrong sometimes.
“Fewer than half of the self-tests examined by the researchers gave any information at all on the box about accuracy of the results. Even when they did give information about accuracy on the box or in the instructions inside, the information was sometimes itself not accurate, or was based on the results of laboratory studies under careful conditions, not on findings on use of the tests by people who are not health professionals.
“But even if all the tests had given information about accuracy, and all that information was reliable, there can still be problems. I’ll describe how.
“Because there are two kinds of wrong results from tests – false positives and false negatives – we need to look at two aspects of the chance of making an error. One common way of doing this, that was used in some of the self-test instructions, is as follows. Findings from the development and use of the test can estimate the probability that someone, who is known to have the health condition in question, will have a true positive test result rather than a false negative result. (In the jargon, that probability is called the test sensitivity – but trust me, knowing the jargon doesn’t help understanding.) Another finding from test development and use is an estimate of the chance that a person, who is known not to have the condition on question, will have a true negative test result rather than a false positive result. (That’s called the test specificity.)
“The trouble is that these two probabilities are the probability of the person having a positive or a negative test result, in the position where we know whether they really have the health condition. But you don’t do these tests if you know already whether you have the health condition. So these probabilities are the wrong way round. What people (and health professionals) want to know is, for example, if we know someone has a positive test result, what’s the chance that they really have the health condition that is being tested for. Or, if we know someone has a negative test result, what’s the chance that they really don’t have the health condition? (There are jargon names for those too – the positive predictive value and the negative predictive value, but again I don’t think those names help much, as there’s too much risk of confusion.) And I’m sure that’s the kind of thing someone would want to know if they buy a self-test and see what result it gives for them.
“However, the first lot of probabilities, the sensitivity and specificity, are different from the second lot, the predictive values. If I tell you that the chance that a person, known already to have the health condition, will have positive test result is 98%, that doesn’t tell you what the chance is that a person, who has a positive test result, actually has the health condition. That second probability is almost certainly not 98%, and in many circumstances it would be very much less than 98%. To get from one set of probabilities to the other, you would need more information, such as how likely it is that the person has the condition if we don’t yet know the test result.
“Just to rub in that these two probabilities aren’t the same, consider the following silly story. You find a man in the street in London. You happen to know he is the Pope. What’s the chance that he is a Roman Catholic? Obviously, 100%. But now suppose the thing you know ,and the thing you want to know the chance of, are the other way round. You know, somehow, that a different man in the London street is a Roman Catholic. What’s the chance that he is the Pope? Well, very much less than 100%. It matters, a lot, which thing you already know and which thing you want the probability for.
“So, in testing you get different probabilities if you know whether the person being tested has the health conditions, and want the probability that the test will be positive, from if you know what the person’s test result is, and want the probability that they have the health condition. And only one of these probabilities – the second one – tells you what a test result is really saying about the chance of having the health condition.
“There has been a lot of research in the past on how people, including health professionals and also non-professionals that might buy one of these self-tests, understand the findings, when they are given some information about the probabilities. Several studies, for instance, found that many doctors and health professionals weren’t using the information on probabilities when the person’s health status is already known (the sensitivity and specificity) properly in trying to answer the question of how likely it is that someone, with a positive test result, actually has the health condition. And if doctors might not be getting it right, how could a non-expert be expected to interpret their own test results properly?
“The position on that maybe isn’t as grim as it sounds, though. Other research has indicated that there are ways of getting the information across so that it’s useable by non-experts. That has been done by several groups, including the Winton Centre for Risk and Evidence Communication in Cambridge (which has now closed, though its findings are still available), groups led by the psychologist Gerd Gigerenzer in Berlin, and many others. Somehow, those communication findings need to be incorporated, as well as they can be, in the instructions for these tests. But that will require more and better regulation.
“Also, some doctors in primary health, including Jessica Watson and Margaret McCartney, who wrote the editorial accompanying these two new research papers in the BMJ, have worked on ways of helping people to understand test results – though you’d need to ask them how much of their findings could transfer easily to something that could be written clearly in test instructions rather than used in direct communication between health professionals and patients.”
Paper 1: ‘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’ by Clare Davenport et al. was published in the BMJ at 23:30 UK time on Wednesday 23 July 2025.
DOI: 10.1136/bmj-2025-085546
Paper 2: ‘Direct-to-consumer self-tests sold in the UK in 2023: cross sectional review of regulation and evidence of performance’ by Bethany Hillier et al. was published in the BMJ at 23:30 UK time on Wednesday 23 July 2025.
DOI: 10.1136/bmj-2025-085547
Declared interests
Prof Amitava Banerjee: “AB declares no relevant conflicts of interest.”
Prof Kevin McConway: “I have no conflicts of interest to declare.”
Rachel Richardson: “I have no interests to declare.”
This Roundup was accompanied by an SMC Briefing.