Letter to the editor in response to Komen et al. 2021.

We read with interest the paper by Komen et al. comparing the effectiveness of Warfarin, antiplatelet therapy (aspirin), and non-vitamin-K oral anticoagulant, constituting a combination of dabigatran, rivaroxaban, apixaban, and edoxaban, as an antithrombotic therapy in Swedish patients with atrial fibrillation (AF) after ischaemic stroke, intracranial haemorrhage, or gastrointestinal (GI) bleed.

A study by Dave et al. examined the different predictors of GI bleeding in AF patients using 1 846 259 hospitalizations in the USA from 2005 to 2015. Similar propensity matching and adjustment for differences in the baseline characteristics of patients were used, and whilst they did not examine anticoagulation therapies used within these patients, they elucidated key details on the baseline characteristics of these patients. They demonstrated that patients who had AF and GI bleeding had higher mortality compared to those with no GI bleeding (7.7% vs. 4.7%, P < 0.001). They also had longer hospital stays [5 days (interquartile range, IQR: 3–9) vs. 4 days (IQR: 2–7), P-value <0.0001] and cost of hospitalization [$9928 ($5916–18665), P < 0.0001]. This analysis did not account for any treatments the patients received, but unlike Komen et al.’s paper where the differences are not elucidated and mortality at 90 days is shown to be lower for AF patients with GI bleeds compared to without (16.1% vs. 25.2%), this showed that AF patients with GI bleeds are older, sicker, and require more medical attention than AF patients. It would be useful to have similar data published from the data from Komen et al.

Other factors predicting GI bleeds in AF patients have also been identified. Tu et al. conducted a systematic review and meta-analysis of the impact of anaemia, which is a common co-morbidity in AF patients, on bleeding, stroke and mortality. They included 28 papers which constituted 365 484 patients, of which 10 studies constituting 75 990 patients examined the association of anaemia with all-cause mortality. Their analysis found a 78% hazard increase [hazard ratio (HR) 1.78; 95% confidence interval (CI) 1.54–2.05] which decreased to 33% (HR 1.33; 95% CI 1.08–1.65) when correcting for publication bias. For GI bleeds, examined in three studies constituting 53 058 patients, there was a 77% increase in hazard ratio (HR 1.77; 95% CI 1.23–2.55). Again it would be useful to have a subgroup analysis based on anaemia in the dataset by Komen et al.

Another factor predicting GI bleeds in AF patients is obesity, which is strongly associated with AF risk and often requires greater doses of anticoagulants. Briasoulis et al. conducted a cohort study of obese patients in the USA from 2010 to 2018, who had AF and were treated with either dabigratran, apixaban, rivaroxaban, or warfarin. Their results demonstrated risk of GI bleed was significantly lower for all three of the anticoagulants compared with warfarin (P < 0.001 for all comparisons). This contrasts with the initial result of no statistical difference between groups by Komen et al. and shows that obese patients benefit from oral anticoagulant therapy.

These factors should be considered in future analysis, to further build on the work by Komen et al. and to provide subgroup analysis for key patient groups that have been shown to have differential outcomes compared to the overall hazard ratios presented and help to yield further translatable clinical findings. We look forward to seeing studies with longer-term follow-up and analysis of differences between the patient characteristics, which will additionally help to clarify the implications of anticoagulant therapy.

Conflict of interest: none declared.