What is the clinical-effectiveness and cost-effectiveness of embedded risk-of-cancer assessment of patients in primary care: the ERICA trial


An estimated 10,000 annual UK cancer deaths would not occur if the UK matched the outcomes of other European countries. Most of the difference relates to diagnostic delay. The NHS Long Term plan, published in January, 2019, specifically targets an increase in the percentage of cancer patients whose cancer is stage 1 or 2 (curable) to rise from the current 54% to 75% by 2028.

This requires major changes in the way cancer is diagnosed. Within general practice, where the core problem exists of identifying the patient who may have a cancer explaining their symptoms, there has been an enormous increase in research efforts, much by the Exeter team. These studies were aimed at identifying which symptoms of possible cancer actually matter. The main outputs from them have been Risk Assessment Tools (generally abbreviated to RATs); these give precise estimates of the chance of an underlying cancer as a percentage figure. This figure can be calculated for single symptoms (e.g. the risk of cancer of the lung with coughing blood is 2.4%), as pairs of symptoms (coughing blood accompanied by loss of weight is 9.2%) or as repeated symptoms (a re-attendance with coughing blood is 17%).

We have published RATs for the 18 most common cancers, accounting for nearly 90% of the total cancer burden. These papers have been highly influential: two have won Research Paper of the Year; more importantly, they strongly underpinned the NICE 2015 guideline revision. Indeed, 100 of the 220 recommendations in NICE 2015 can be traced back to RATs studies.

RATs themselves have evolved. The initial versions were in paper, mouse mat, calendar, or web-based forms; they increased cancer diagnostic activity, but the formats did not allow close integration with GPs’ clinical systems. In partnership with Macmillan, the UK cancer charity, electronic versions for seven major cancers (lung, colorectal, pancreas, oesophago-gastric, bladder and kidney and ovary) have been developed. These eRATs have been integrated into the GPs’ clinical software, for two of the main UK GP software providers, with two other major providers almost ready to host them. The eRATs prompt the GP when the risk of one or more of these cancers is above 2%. They have been downloaded by 1,000 of the UK’s 10,000 practices, though actual usage is largely unknown. Macmillan commissioned two internal studies of eRATs, which examined practitioner acceptability: these reports were positive, but did not address important clinical outcomes, such as stage at diagnosis or survival.

A pilot feasibility trial of the oesophago-gastric eRAT is ongoing. Although important, findings from this study will be restricted to the single eRAT. Crucial questions regarding the clinical effectiveness and usability of multiple eRATs remain.

One major aspect of the eRAT research that is outstanding relates to cost-effectiveness, there is very little health economic evidence to date. This is crucial: annual NHS spending on cancer diagnosis is approximately £1bn. Observational data suggests that expedited cancer diagnosis may be clinically effective, but there is no data to examine cost-effectiveness. What work has been published has been judged to be of very poor quality, or simplistic. A carefully designed clinical trial is now needed to address the following question: what is the clinical effectiveness and cost-effectiveness of electronic decision support embedded within primary care IT systems in the early diagnosis of symptomatic cancer?

Project Team


The Dennis and Mireille Gillings Foundation: £2,000,000.00
The University of Exeter: £200,000
Cancer Research UK: £25,000

Aims & Objectives

The overarching aim of the trial is to assess the clinical and cost effectiveness of electronic Risk Assessment Tools (eRATs) for cancer compared to usual care for patients in general practice. The specific objectives of this study are to compare the effects of eRATs (vs usual care) on: cancer staging at time of diagnosis, cost to the NHS, patient experience of care, and service delivery.

Methodology & activity

Main RCT only: This is a pragmatic, cluster RCT. The 530 practices will be randomised using a 1:1 ratio into one of two trial arms: usual diagnostic practice (control) and usual diagnostic practice plus access to the suite of Macmillan electronic risk assessment tools (eRATs) (as the intervention), for a total of 265 practices per arm. Randomisation will be remote and web-based, conducted by an independent member of the data team at the Exeter Clinical Trials Unit, overseen by the CTU statistician (not the trial statistician). The sequence of randomisation will be computer generated. To ensure there is balance between the trial arms regarding practices’ propensity to refer patients for cancer investigation, we will minimise the randomisation by two week wait referral rate (the best available proxy) in national tertiles. We will use simple randomisation to allocate the first 50 practices (~10% of the total target), and then apply minimisation by two week wait referral rate, taking into account the previous allocations to inform the minimisation algorithm. To promote allocation concealment, all allocations using the minimisation algorithm will retain a stochastic element.

Given the nature of the interventions, it is not possible to blind GPs or the GP practice to treatment allocation. Similarly, the trial team overseeing the day-to-day management of the trial will not be blinded. The data analysis will be carried out by the trial statistician who is blind to treatment allocation and all primary outcome data are objective assessments of clinical outcome.

The primary outcome is whether a patient diagnosed with one of the six cancers during the follow-up period is diagnosed at stage 1/2 (early – cure likely) or stage 3/4 (late – cure not likely). Secondary outcomes include route to diagnosis, 30-day survival, 1-year survival, and adverse event (an estimate any change in the expected number of adverse events from imaging investigations (colonoscopies, sigmoidoscopies, upper gastro-intestinal endoscopies, chest x-rays, abdominal ultrasounds, and abdominal CT scans)).

Outputs & impact

Key outputs from the trial will be presentations at national and international conferences, seminars, PPI events and dissemination through internationally recognised peer reviewed journal publications (including open access web sources), newsletters and media releases.

The eRATs are, in principle, already available to a large number of GP practices. Outcomes indicating that the eRATs help GPs catch cancer sooner will likely lead to an increase in the uptake of clinical decision support tools. They may even be endorsed by the Department of Health. However, this may only happen to the extent that there are shown to be financially economical.

Next steps

The RCT is due to commence in June 2019 with practice recruitment from this period until December 2019. Practices will be in the trial for approximately 2 years; with the data collection phase closing in December 2021. Final data should be available from the Cancer Registry around December 2022 with analyses and write up planned for the majority of 2023 up until the project end of August 2023.

Back to Projects