Letter: Incidence and Predictors of Major Gastrointestinal Bleeding in Patients on Aspirin, Low-Dose Rivaroxaban or the Combination: Secondary Analysis of the COMPASS Randomised Controlled Trial. Authors' Reply

Abstract

We thank Dr. Yang et al. for their interest in our study [1]. In their letter [2], they suggested that our analysis of the incidence and predictors of major gastrointestinal bleeding in patients on aspirin 100 mg once daily, rivaroxaban 5 mg twice daily or the combination of aspirin 100 mg once daily and rivaroxaban 2.5 mg twice daily had ‘critical’ limitations [1].

First, they suggested that we ‘neglected’ to report long-term outcomes after major acute bleeding. While we agree that an analysis of outcomes after gastrointestinal bleeding is of interest, our goal—as clearly described in our original manuscript—was to describe the incidence and predictors of bleeding on aspirin, low-dose rivaroxaban or both in patients with chronic coronary and/or peripheral artery disease, an issue that had not been previously addressed.

Second, they claimed that the heterogeneity of our patient population limited generalisability. While this may be true in analyses involving small numbers of patients and events, we included 27,395 patients, of whom almost 300 experienced a first gastrointestinal bleeding event. Including a population that is more representative of the target population enhances rather than limits generalisability [3].

Third, they suggested that a ‘significant limitation’ of our analyses was the inadequate evaluation of proton pump inhibitor (PPI) use. COMPASS was a partial factorial trial that also randomised patients not already taking a PPI at baseline to pantoprazole or placebo. Involving 17,598 patients, this is the largest trial to date examining the effects of a PPI in patients treated with antithrombotic drugs [4]. Given this robust sample size, our detailed multivariable analyses did account for PPI use, as was clearly indicated in our methods and results, and as we subsequently discussed at considerable length.

Fourth, they incorrectly suggested that the fixed-dose regimens of rivaroxaban evaluated in COMPASS do not reflect real-world practice. When used for vascular protection (as tested in COMPASS), rivaroxaban was given at a dose of 2.5 or 5 mg twice daily. Only the 2.5 mg dose was subsequently approved for clinical use, without dose adjustment for age, body weight or renal function. There is no evidence that personalised dosing of either rivaroxaban or aspirin for this indication provides efficacy or safety advantages.

Fifth, it is theoretically possible that socio-economic status or genetic makeup could contribute to bleeding risk. However, this is unproven, and we did not collect DNA samples in COMPASS.

Finally, it appears that Yang et al. overlooked that, in addition to approximately 30 other baseline variables, we considered alcohol use in our multivariable analyses.

Thoughtful critiques of published work can help to advance science by identifying potential deficiencies in methodological approaches and/or interpretations of findings. However, in our manuscript, we clearly stated our goals and considered important covariates, including several that Yang et al. suggested that we missed. Letters to the editor should serve to clarify, or put into context, the information provided.

Nauzer Forbes: writing – original draft; writing – review & editing. Jackie Bosch: writing – original draft; writing – review & editing. John W. Eikelboom: writing – original draft; writing – review & editing.

The authors declare no conflicts of interest.

This article is linked to Forbes et al papers. To view these articles, visit https://doi.org/10.1111/apt.18139 and https://doi.org/10.1111/apt.18306.