The association between kidney function and major bleeding in older adults with atrial fibrillation starting warfarin treatment: population based observational study
BMJ 2015; 350 doi: https://doi.org/10.1136/bmj.h246 (Published 03 February 2015) Cite this as: BMJ 2015;350:h246
All rapid responses
Rapid responses are electronic comments to the editor. They enable our users to debate issues raised in articles published on bmj.com. A rapid response is first posted online. If you need the URL (web address) of an individual response, simply click on the response headline and copy the URL from the browser window. A proportion of responses will, after editing, be published online and in the print journal as letters, which are indexed in PubMed. Rapid responses are not indexed in PubMed and they are not journal articles. The BMJ reserves the right to remove responses which are being wilfully misrepresented as published articles or when it is brought to our attention that a response spreads misinformation.
From March 2022, the word limit for rapid responses will be 600 words not including references and author details. We will no longer post responses that exceed this limit.
The word limit for letters selected from posted responses remains 300 words.
Dear Editor,
I would like to thank the authors for an interesting, generally very good and clinically relevant paper. However, I did have some questions about their use of number needed to harm. In their methods they write:
“Based on the adjusted incidence rates of major bleeding in chronic kidney disease (defined as <60 mL/min/1.73m2) and non-chronic kidney disease (≥60 mL/min/1.73m2), we determined the number needed to treat to harm”
Later in their results they write:
“Among participants with chronic kidney disease, the number needed to treat to harm was 22 (95% confidence interval 18 to 27) during the first 30 days of warfarin treatment”
At face value these figures seem quite alarming, given the number of patients with AF and an eGFR<60 one seems to encounter in hospital medicine. However, it seems they derive their number needed to harm from the excess in incidence rate amongst those with an eGFR <60 compared to those with an eGFR >60. If I have read this correctly then isn’t this the number needed to harm for inducing CKD in patients with atrial fibrillation on warfarin? Shouldn’t the number needed to harm of warfarin therapy be calculated based on comparison with a control group who were not started on warfarin?
More generally, I wonder if the authors should be reporting number needed to harm at all. There is a lack of data on crucial confounders such as smoking, no comparison with a control group or with the baseline pre-warfarin hazard of the study participants themselves and those at highest risk will not have been started on warfarin in the first place. The authors correctly never make an outright causal claim as this could not be sustained. However, doesn’t calculating a number needed to harm implicitly involve make a strong causal assumption?
Yours faithfully,
Ben Williams
Competing interests: No competing interests
Re: Numbers needed to harm
We thank the author of the recent rapid response for their interest in our study and for the comment regarding the use of the number needed to treat to harm.
The objective of our study was to determine the relationship between kidney function (assessed as different stages of chronic kidney disease) and the rate of major bleeding (i.e. safety of warfarin therapy) in older adults with atrial fibrillation, specifically in those commencing warfarin therapy. We focused on assessing the risk of major bleeding in this patient group as arguably, the reluctance of warfarin use in patients with chronic kidney disease and atrial fibrillation is largely in relation to uncertainties regarding their safety in this setting and there was limited data assessing this at different stages of chronic kidney disease in the literature. As such, the “control” group (i.e. “non-exposed” group in the treatment exposure sense of the term) used for the calculation of the number needed to treat to harm in our study was the non-chronic kidney disease group (defined as eGFR ≥60 mL/min/1.73m2). In other words, among patients with atrial fibrillation commencing warfarin therapy, for every X number of patients with chronic kidney disease (defined as eGFR <60 mL/min/1.73m2), 1 patient would experience a major bleeding event, compared to patients without chronic kidney disease. In this context, this allows the assessment of the impact of the presence of CKD on the risk of bleeding in atrial fibrillation patients commencing warfarin therapy when estimating absolute measures of treatment effects which are clinically meaningful. The substantial difference in the number needed to treat to harm during the first 30 days of warfarin therapy as compared with the period after, highlights the need to consider the bleeding risk relative to kidney function particularly in the period shortly after warfarin treatment commencement. We do, however, acknowledge that when assessing the number needed to treat to harm in this way, this increased risk of major bleeding with worsening kidney function in patients commencing warfarin therapy should be balanced against the risk of stroke according to chronic kidney disease status. However, based on the rationale described above, our approach was to specifically focus on the safety of warfarin therapy in new users of warfarin.
As the author has correctly pointed out, we do not (and cannot), based on our current data, assert that warfarin “causes” major bleeding in patients with chronic kidney disease and atrial fibrillation. Ultimately, to postulate a causal relationship between warfarin therapy and clinical outcomes in patients with atrial fibrillation (by different levels of kidney function), we agree that the risk of stroke and major bleeding need to be assessed in both the warfarin-exposed and unexposed groups, an assessment that should ideally be made within a randomized controlled trial. Observational studies on pharmacoepidemiology which address potential biases involved in these study types including confounding by indication and immortal time bias can also be considered, however, this was beyond the scope of our current study.
Competing interests: No competing interests