PREDICTion of GastroIntestinal malignancy in patients with IRON Deficiency (PREDICT-GI-IRON)

Research question: What is the probability that a patient with an iron deficiency (ID) has an undiagnosed gastrointestinal malignancy?

Background: Gastrointestinal tract cancers (i.e. oesophagus, stomach and bowel) are the second most common cancer in the UK with approximately a third of these patients observed to have an ID prior to their diagnosis of cancer.

Current UK guidelines therefore recommend urgent evaluation of the gastrointestinal tract in patients with a new diagnosis of iron deficiency anaemia (IDA), the sequela of ID.

However, ID is a common finding on 1 in 14 routine blood tests and the risk of cancer in different patient groups with ID is unclear.

Endoscopy is the gold standard diagnostic test for gastrointestinal malignancy, however it is invasive and linked with uncommon but potentially serious complications.

As a result, deciding which patients should be referred for investigation with endoscopy (and whether urgently) is challenging as these decisions are likely to impact on the long-term health of patients.

NHS endoscopy services are significantly overstretched and there is a pressing need for more judicious use of limited endoscopic resource.

Aim: to quantify the relative and absolute risk of gastrointestinal malignancy in patients with ID based on their symptoms, blood results and other information in their GP medical record.

Specific objectives are to: 1) estimate the 5-year cumulative incidence of gastrointestinal malignancy separately for adults with IDA and ID without anaemia; 2) Estimate the association between ID (separately with and without anaemia) and gastrointestinal malignancy; 3) Develop and validate a risk prediction model (PREDICT-GI-IRON) for malignant GI disease in patients diagnosed with IDA and ID without anaemia; 4) Develop a web-based calculator for clinical use; 5) Develop a protocol to evaluate the clinical and cost effectiveness of PREDICT-GI-IRON.

Methods: A prospective open population-based cohort study using data from The Clinical Practitioners Research Datalink (CPRD) AURUM dataset linked to hospitalisation, mortality records and national cancer registration data.

The study will span from 1st January 2010 to 30th September 2023 and participants are those aged ≥18-years with a first diagnosis of ID with and without anaemia and no prior diagnosis of gastrointestinal cancer.

Five-year cumulative gastrointestinal cancer incidence for IDA and ID without anaemia will be estimated, stratified by sex and age group. English practices will be divided by region into derivation (85%) and validation (15%) cohorts.

Multivariable models for the outcome of gastrointestinal cancer by 5-years will be developed using proportional subhazards and Cox proportional hazard regression modelling. Model discrimination and calibration will be evaluated and models will be validated. Decision curve analysis will evaluate the clinical utility of models.

Active involvement of key stakeholders will ensure the web-based calculator and protocol to evaluate clinical and cost-effectiveness is GP, patient- and public-centred.

Anticipated impact: This work is likely to lead to revision of current guidance with more judicious use of constrained endoscopy resources, while supporting primary care clinicians in improving clinical and patient-centred decision making regarding the need for further investigation of ID.