ColonFlag for identifying people at risk of colorectal cancer

Level of innovation

Two experts agreed that the technology application is innovative but that the technology itself is not. Two experts thought that the machine learning component is innovative but another thought that more advanced machine learning techniques are available. Two experts believed that the concept is novel but unproven. Five out of 6 experts said there are no technologies that predicted risk in asymptomatic patients but 1 stated that there are. This commentator was not convinced that machine learning techniques are better at predicting risk than other methods.

Three experts were aware of technologies that are similar to ColonFlag, but 2 added that they may be better for diagnostic rather than prognostic purposes because the risk factors assessed are different. One expert highlighted that in primary care, risk-prediction algorithms are already used for colon cancer and other long-term conditions to help patients and clinicians with decision-making.

Potential patient impact

Four experts thought that using ColonFlag may lead to earlier diagnosis and treatment for colorectal cancer, which would improve patient survival and quality of life. Two experts thought ColonFlag being non-invasive could lead to fewer anxious patients. One expert thought that ColonFlag would personalise patient care. However, 1 expert thought these benefits would only happen if every patient in a GP practice had a complete blood count (CBC) test each year. One expert thought that using ColonFlag may lead to more guaiac faecal occult blood test (gFOBT) or faecal immunochemical test (FIT) tests, and another thought that it would lead to more colonoscopies.

Three experts thought that people outside the age range for screening but who are at high risk of colorectal cancer would benefit most from ColonFlag. One expert believed that all at‑risk groups for colorectal cancer would benefit, whereas another expert said that older people would benefit. One expert thought that all people over 40 would potentially benefit but particularly people without symptoms. Two of the experts thought that up to half of the population of England would be eligible for ColonFlag but that its use would depend on the proportion of people with CBC results.

Potential system impact

Three experts thought that ColonFlag had the potential to reduce costs for the NHS by avoiding unnecessary investigations, decreasing treatment costs and providing more efficient care. Another expert thought that ColonFlag could lead to more consistency in gFOBT/FIT testing in primary care.

Two experts thought that ColonFlag would represent an additional cost to the NHS. Three experts felt that ColonFlag was expensive with 1 expert stating that gFOBT/FIT testing was cheaper and 2 stating the maintenance charge of £40,000 was excessive. One expert thought that the cost would be around the same as current practice but with improved targeting of resources. One expert thought that the overall cost effect was unknown because more patients would need treatment but fewer patients would need later-stage treatments.

One expert thought that ColonFlag would move care and resources upstream by identifying cancer earlier. Another expert noted that more staff would be needed for consultations with patients at high risk. Further, more resources would be needed for the increased use of the next test in the pathway. One expert thought that the resource impact would depend on the new care pathway; they stated that if gFOBT/FIT testing was the next step on the care pathway, the effect would be a small extra cost. If colonoscopy was the next step, there would be a larger increase in cost. Another expert agreed that there would be considerable resource consequences in secondary care to provide the extra investigations. Another expert remarked that adopting ColonFlag would need resources in primary care, such as extra GP time or hiring an administrator to operate the system. One expert did not foresee any changes to resource use from adopting ColonFlag.

The experts largely agreed that there would be little need for change to infrastructure to use ColonFlag. Three experts thought that only minimal training would be needed. One expert thought that there may need to be changes to infrastructure if there was insufficient capacity to deal with increased demand for colonoscopies. Another expert thought that some changes to infrastructure may be needed to incorporate the technology with electronic health records.

Two experts expressed concerns about patient confidentiality and stressed the importance of keeping patient details secure. Another expert thought that the technology itself had not been evaluated thoroughly and independently and could miss patients at high risk. ColonFlag would be an addition to the current standard of care but 1 expert said that there is a lot of variation in how frequently GPs use risk assessment tools.

General comments

One expert was unsatisfied with ColonFlag's reporting, particularly that statistical assessment of the tool was impossible because it is a closed-source software. The expert added that ColonFlag's reporting does not follow the standard transparent reporting of a multivariable prediction model for individual prognosis or diagnosis (TRIPOD) guidelines and there was no comparison of the algorithm to other methods like logistic regression. Furthermore, the expert had concerns about the applicability of the algorithm to the UK population. They thought that recalibrating the model to UK data would improve outcomes.

Three experts raised concerns about the cost of ColonFlag as a barrier to its adoption, with one also considering time constraints to be an obstacle. Another expert thought that there was a need to discuss ColonFlag's place in the patient pathway. ColonFlag's incorporation within the software systems currently used in the NHS was also considered to be a potential issue by 1 expert.

All of the experts felt that more research was needed to address the uncertainties in the evidence base.