England’s STRIDE-RCT will study multi-cancer blood tests in people who consult a GP with abdominal symptoms. The trial puts a practical question at the centre of AI-assisted detection: can an additional signal help patients reach the right investigation sooner?
The first decision in a cancer investigation can arise in a conversation that is quite ordinary. Consider an illustrative consultation in which a person describes abdominal discomfort that has persisted long enough to disrupt work or sleep. The symptom has several possible explanations, and the clinician must decide which questions and investigations could narrow them down. A blood test that suggests a possible site of cancer would introduce new information into that conversation. Its significance would depend on what the patient and clinician could do with the result, how quickly they could do it and what would happen if the signal turned out to be misleading.
On 4 November 2026, the UK Department of Health and Social Care announced STRIDE-RCT, a trial intended to involve more than 20,000 patients at at least 200 GP practices in England. The announcement describes participation for eligible patients aged 45 and over who present with abdominal symptoms. The tests under study are designed to look for signals associated with several cancers and suggest where in the body a signal may originate. The department describes a variety of methods, including artificial intelligence and measurement of blood proteins. The University of Birmingham lists the cluster platform trial as active (Department of Health and Social Care announcement and University of Birmingham project record)
The announcement concerns an evaluation in symptomatic primary care. That setting matters because the patient is already seeking help and the test would sit alongside a clinical assessment. A trade report in The Pharmacist describes the same eligibility and the intention to support referral decisions. Neither the report nor the launch announcement supplies results from the trial. The question opened by the study is whether the additional information can improve the path to diagnosis for the patients invited to participate, under the conditions in which their practices and referral services actually operate.
From a measurement to a decision
A blood sample can contain many measurable features. As a general principle, a statistical model can combine information from several measurements to estimate whether their pattern resembles one associated with disease. The model may detect a combination that would be hard to recognise by inspecting each measure separately. How well the association transfers depends on the development data, the way samples are collected and the patients on whom the test is used. That is a general account of how an AI component can contribute. The announcement describes multiple tests and provides no basis for attributing the same inputs, model or decision threshold to every one of them.
Once a result arrives, its interpretation depends partly on what was plausible beforehand. The same test can produce different proportions of genuinely informative positive results in populations where the condition is more or less common. This is one reason to keep evidence about symptomatic patients separate from evidence about screening people who feel well. The US National Cancer Institute, in its discussion of multi-cancer detection for asymptomatic screening, explains that clinical usefulness depends on what happens during diagnostic investigation and on the availability of effective care. That programme studies a different population. The importance of the steps after a positive result is relevant in both settings.
The UK announcement describes a positive result as information that could help a GP and patient decide whether urgent referral is appropriate. Confirmation is a separate clinical step. A result that suggests a site could help focus an investigation, but its value would depend on whether the suspected site is correct and whether the appropriate next service can act. If the site remains uncertain, the patient may still need more than one investigation. The test therefore changes the information available at a decision point. The trial has to establish whether changing that information makes the sequence of decisions work better.
Earlier for whom, and earlier by what measure?
A quicker laboratory result, a quicker referral and a quicker diagnosis are different outcomes. A useful signal might reach a practice promptly and still leave the patient waiting for imaging, endoscopy or a specialist appointment. Equally, a test that adds a short step in primary care could improve the overall sequence if it helps clinicians choose a more appropriate investigation. Looking at time from the first relevant consultation through to a confirmed answer would capture more of the patient’s experience than measuring how long the laboratory took. It would also reveal where any delay has moved if one part of the pathway speeds up and another slows down.
A comparison between groups can address questions that striking individual cases leave open. People with easier access to appointments or with more readily recognised symptoms may reach a diagnosis sooner for reasons that precede the blood test. When comparing routes, researchers need to account for differences in the patients and practices involved. The Birmingham project record describes a cluster platform design. The public summary establishes the type of study, while the eventual methods and results will be needed to judge who entered the comparisons, how the groups differed and which outcomes the researchers were able to measure (STRIDE-RCT research record)
Time to diagnosis also has to be interpreted alongside the treatment and health outcomes that follow. A diagnosis recorded earlier can lengthen the measured interval between diagnosis and death even if the patient’s lifespan remains the same. Conversely, faster clarification can have value to a patient even before long-term effects are known, for example by reducing the period in which nobody can explain their symptoms. Those are different possible benefits. An evaluation can make progress by reporting them separately and by showing whose uncertainty was shortened, whose treatment changed and which questions still need longer follow-up.
What a result sets in motion
Consider an illustrative patient who receives a result suggesting a cancer signal but whose first investigation finds no explanation. They may face a second appointment, an invasive procedure or a period of waiting for the clinical team to interpret conflicting findings. That experience cannot be represented fully by the number of positive tests. A study of the pathway would be more informative if it could show how many people reached a clear conclusion, how much investigation that took and how often a signal remained unresolved. The National Cancer Institute treats the work needed to confirm or rule out disease as central to evaluating multi-cancer tests.
A result with no signal presents a different difficulty. In the illustrative consultation, the clinician may still have good reason to investigate or to review symptoms that persist. A reassuring result can change how the patient hears the rest of the conversation. If the limits of the test are poorly explained, an intention to continue assessment may sound like a minor precaution that can be put off. It would be useful to see whether services can communicate the next step clearly, including who will make contact and what happens when a result and the rest of the assessment point in different directions.
Those tasks create a new kind of coordination work. Someone has to notice a result, judge its urgency and ensure it is linked to the right consultation. If the patient is referred, the receiving service needs enough context to know why it is being asked to investigate. If two services are involved, each needs to know which one will communicate with the patient. These are practical questions for any proposed rollout, and they are particularly important when a test is designed to suggest several possible routes into specialist care. An extra result can be valuable only if it reliably reaches a person able to act on it.
Whose pathway gets easier?
The announcement says the trial will include a range of settings, including rural, inner-city, coastal and deprived communities. That breadth creates the opportunity to examine how the same proposal fits different circumstances. A practice may be able to add the sample to a routine blood draw, but nearby services can still differ in availability and in the journeys patients must make. Two people offered the same next test could face very different costs in travel, time off work or childcare. Those differences are relevant to whether people can complete an investigation, even when the first step is equally accessible.
There are also questions about who participates and whose results can be interpreted with confidence. Overall performance can conceal different experiences across age groups, patterns of symptoms and clinical settings. A useful analysis would report the denominators as well as the eventual diagnoses, including how many people accepted a test, obtained a usable result and completed the recommended investigation. Small groups may yield uncertain estimates. Reporting those uncertainties helps a reader distinguish a reason for further study from a claim that a gap has already been found. It also helps services decide whether further evaluation is needed before extending the approach.
The same evidence would matter to other patients who rely on diagnostic capacity. If a new test focuses referrals well, it could reduce some avoidable investigations or direct patients more effectively. If it generates additional referrals that are difficult to resolve, it could increase demand on the same services. Neither effect follows automatically from better performance on a laboratory sample. Assessment at the level of a service would need to examine the volume and mix of subsequent investigations, the time staff spend resolving uncertain cases and the effects on people referred through existing routes.
A trial of care as well as technology
For health services considering a future decision, the useful result would connect the stages that patients actually pass through. It would explain which tests added information to clinical judgement, how that information changed referrals and what kinds of diagnostic work followed. It would also be possible for one test to work better for a particular purpose than another, or for the same result to be easier to act on in one setting than another. Knowing where the approach falls short can guide a narrower and better supported use, rather than leaving a service to infer that the average result applies everywhere.
The promise of adding AI to a blood test can be assessed in that wider frame. A better way to detect patterns may be an important contribution, particularly where familiar symptoms are difficult to interpret. The patient experiences the value through a chain of events that extends well beyond the model, from the first conversation to an explanation they can understand and an investigation they can complete. STRIDE-RCT gives researchers an opportunity to study that chain in primary care. The evidence that matters will be what the additional signal enabled clinicians and patients to do, and what it took from the rest of the service to make that possible.
Image credit: Vacutainer with blood collection tube by Whispyhistory, CC BY 3.0 US. Illustrative photograph unrelated to STRIDE-RCT. The source file is unedited. WordPress may resize or crop display versions.
