Cancer unmasked by bleeding during anticoagulant therapy: when a problem may become an opportunity

This editorial refers to ‘Gastrointestinal bleeding and the risk of colorectal cancer in anticoagulated patients with atrial fibrillation’, by P.V. Rasmussen et al., doi:10.1093/eurheartj/ehz964. The major reason for anticoagulation worldwide is ischaemic stroke prevention in patients with atrial fibrillation (AF). AF is estimated to affect ∼1% of the general population,1–3 i.e. over 70 million people globally. Anticoagulation promotes bleeding, with ∼3 out of 100 presenting a major bleed per year of treatment.4 More than 80% of major bleeds are extracranial, and approximately half of extracranial bleeds are gastrointestinal.4 In over 50% of cases, gastrointestinal bleeds during anticoagulation are related to pathological lesions that include inflammatory or diverticular disease, haemorrhoids, fissures, ulcers, vascular malformations, radiation enterophathy, adenomatous polyps, or malignancies.5 , 6 Lower gastrointestinal bleeding may unmask an underlying colorectal cancer (CRC), the second leading cause of cancer death after cancer of the lung in men and of the breast in women.7 , 8 Right colon cancer in particular is often poorly differentiated and locally invasive at diagnosis.6 Risk factors for CRC include age, lifestyle, genes, comorbidities, and radiation.7 , 8 Detection and removal of precursor lesions and early cancer diagnosis can reduce CRC incidence and mortality.7 Given the burden of CRC, biochemical and/or visual screening is advised from age 45 for individuals at average risk of disease.7 Precisely how many lower gastrointestinal bleeds are related to CRC in patients started on oral anticoagulation? Should a lower gastrointestinal bleed during anticoagulation be investigated in all cases and all age groups, or can it be dismissed as an incidental manifestation of antithrombotic therapy? At the end of the day, should lower gastrointestinal bleeding be considered only as an adverse reaction of treatment or also as a potential protective factor against CRC occurrence and related mortality? The study by Rasmussen et al. in this issue of the European Heart Journal provides answers to the above questions using Danish nationwide databases of hospitalizations, outpatient visits, and drug prescriptions.9 The authors report the rates of lower gastrointestinal bleeding and subsequent diagnosis of CRC among 125 418 patients with AF started on oral anticoagulation between 1996 and 2014 and followed for up to 3 years.9 Lower gastrointestinal bleeds occurred in 2.1%, mostly during the first month of therapy, affecting 2576 AF patients. Endoscopy was performed in ∼61% of them, and 5.4% of those with bleeding had a subsequent diagnosis of CRC, amounting to 140 cases.9 The age-stratified rate of CRC after bleeding was 3.7% per annum in patients ≤65 years of age and 8.1% per annum in those 76–80 years old. Compared with those without bleeding, the relative risk of CRC after bleeding was 24-fold in patients ≤65 years and 12-fold for those >85 years.9 Thus, in the study by Rasmussen et al., >5% of lower gastrointestinal bleeds were related to CRC. The absolute likelihood of CRC diagnosis was higher among older AF patients, but the relative risk of CRC diagnosis after bleeding (compared with no bleeding) was highest in younger vs. older AF patients, probably owing to more numerous different reasons for bleeding in older patients. The authors conclude that lower gastrointestinal bleeding should be carefully investigated in all anticoagulated AF cases regardless of age.9 The authors should be commended for examining an unselected population with AF and for conducting statistical analyses stratified by age and accounting for time from onset of therapy, competing risks of death, censoring probability, adherence to treatment, and several comorbidities.9 Some limitations are the retrospective design, the non-consideration of CRC risk factors other than age, and the fact that 85% of patients were taking vitamin K antagonists and only 15% direct oral anticoagulants.9 The use of direct oral anticoagulants is known to be associated with an overall 20–25% higher relative rate of gastrointestinal bleeds compared with vitamin K antagonists.4 , 10 Controlled trials of direct oral anticoagulants vs. vitamin K antagonists or vs. placebo report up to 10% of gastrointestinal bleeds attributable to gastrointestinal cancers.11 , 12 In RE-LY, 3% of 18 113 AF patients, randomized to dabigatran or warfarin and followed for a median of 2 years, developed major gastrointestinal bleeds. A total of 44 of the 546 gastrointestinal bleeds (8.1%) were related to gastrointestinal cancers, 80% of which were related to CRC; cancer-associated bleeding was more commonly attributed to CRC with dabigatran and to gastric cancer with warfarin.11 In COMPASS, 9.8% of >27 000 patients with cardiovascular atherothrombotic disease without known AF, randomized to low dose rivaroxaban plus aspirin or to aspirin alone and followed for a mean of 23 months, developed major or minor bleeding; 257 of 2678 cases (9.9%) were subsequently diagnosed with cancer.12 Gastrointestinal bleeding, in particular, was associated with a 20-fold risk of gastrointestinal cancer diagnosis.12 Thus, the data from RE-LY and COMPASS suggest that the findings of Rasmussen et al. would have been amplified by a more prevalent use of direct oral anticoagulants. The data also indicate that, regardless of presence of absence of AF, when using antithrombotic therapy a gastrointestinal bleed should not be overlooked, given the high risk of underlying cancer (Take home figure). Irrespective of the underlying clinical condition and type of study, a gastrointestinal (GI) bleed during antithrombotic therapy is associated with a higher risk of GI cancer, compared with those without bleeding, and warrants specific investigation. CI, confidence interval; RR, relative risk. Bleeding during antithrombotic therapy anticipates the detection of cancer even at sites other than gastrointestinal12 and with therapies other than anticoagulants;13 importantly, early detection of an occult malignancy is expected to reduce cancer-related mortalitygastrointestinal7. In COMPASS, the occurrence of urinary and genitourinary bleeds led to a >30-fold detection of new urinary and genitourinary cancers.12 Among 1532 patients presenting gross haematuria, those on antithrombotic therapy had less advanced stages of bladder cancer than those not on antiplatelet/anticoagulant drugs.13 Taken together, the above data suggest that at least 1 in 1000 anticoagulated AF patients followed for up to 3 years (140 of 125 418 in the study by Rasmussen et al.) may have an anticipated new CRC diagnosis as a consequence of a lower gastrointestinal bleeding complication. What are the implications for clinical practice and for public health? Patients on anticoagulants are usually monitored by blood haemoglobin concentrations followed by faecal occult blood testing when iron deficiency anaemia is found; whether high-sensitivity stool-based tests should become routine initial screening in patients on oral anticoagulants is still debated.7 , 10 , 14 , 15 There is probably room for improvement by expanding endoscopy to all anticoagulated patients with lower gastrointestinal bleeding or positive occult blood test.5 , 7 , 9 If even just 50% of the estimated global burden of AF received anticoagulant therapy, up to 35 000 cases of occult CRC might be unmasked over a 3-year period as a result of properly investigated gastrointestinal bleeding complications. Thus, oral anticoagulation may also be viewed as a stress test to diagnose underlying pathological lesions, including malignancies.16 Several investigators have reported bleeding complications occurring early after treatment initiation.6 , 9 , 10 , 16 , 17 Because early detection of an occult malignancy is expected to improve survival, optimal strategies to detect and remove precursor lesions or early cancers should be clearly outlined not only in guidelines for gastroenterologists,5 , 7 but also in those for cardiologists.