AJKDBlog Interviews Editor Timothy Yau (@Maximal_Change) caught up with Kueiyu Joshua Lin to discuss his study recently published in AJKD comparing the safety and effectiveness of warfarin or rivaroxaban versus apixaban in patients with atrial fibrillation (AF) and non–dialysis-dependent CKD stage 4/5.

Dr. Lin is an internist at Massachusetts General Hospital and the Executive Director of the Mass General Brigham (MGB) Center for Integrated Healthcare Data Research. He is currently the Principal Investigator of four large NIH-funded national studies on drug safety and effectiveness in older adults.

AJKDBlog: I remember when the direct oral anticoagulants (DOACs) first came out about 10 years ago and revolutionized A-fib or thrombotic events.  When I was a resident, it was routine to bring someone into the hospital for a heparin drip while bridging them to warfarin, and nowadays there are so many direct oral anticoagulants that I get them confused.  Can you first give an overview of these DOACs like rivaroxaban, apixaban, dabigatran, and edoxaban, if they differ in mechanism of action, and their general use in patients with advanced CKD?

Dr Lin: In the general population, based on randomized controlled trials, compared to warfarin, DOACs are at least as efficacious, with generally lower rates of major bleeding, particularly intracranial bleeds, and fewer drug-drug or drug-food interactions. Another commonly cited advantage of DOACs is that they do not need blood test monitoring and frequent dose adjustments. Thus, DOACs are preferred over warfarin in most populations.

However, the use of DOACs in patients with CKD is challenging given various degrees of renal excretion of specific DOACs (apixaban: ~25%, dabigatran: ~80%, rivaroxaban: ~33%, edoxaban: ~50%). In patients with advanced kidney dysfunction, the anticoagulation effect becomes more unpredictable and the fact they do not have timely and readily available monitoring blood tests becomes a disadvantage (while anti-factor Xa levels can be used for factor Xa inhibitors [i.e., apixaban, rivaroxaban, and edoxaban], and dilute thrombin time [dTT], ecarin chromogenic assay [ECA], and ecarin clotting time [ECT] can be used for direct thrombin inhibitor [dabigatran], they are often not readily available and timely enough for dose titration), which is why warfarin remains an important choice in this vulnerable population. This is also why the risk vs. benefit ratio between DOACs and warfarin may differ in patients with advanced CKD from the general population.

AJKDBlog: You mention in the paper how all of the landmark trials in the above medications excluded patients with CKD and GFRs < 25 ml/min.  Were these contraindicated in these patients when the drugs were approved, and are they still being used in this population despite the lack of inclusion in the original trials?

Dr Lin: Yes, all of the landmark trials excluded patients with CKD and eGFRs < 25 ml/min. Yet, the FDA did not explicitly list “eGFR<25 ml/min” as a contraindication. This is more a question of generalizability where we are not sure if the findings in the trials are applicable in patients with eGFRs < 25 ml/min. Because randomized controlled trials (RCTs) often include patients who are not representative of the frail and complex patients in routine care, it is not uncommon that the approved medicines are used in the patient populations who are severely under-represented or explicitly excluded in the RCTs, under the assumption that the RCT findings are generalizable to patients with different comorbidity profiles. This assumption can only be verified when such “non-evidence-based use” accrues enough numbers to allow an observational study like ours to assess the same safety and effectiveness outcomes as the trials.

AJKDBlog: Can you briefly walk us through your study design and how you created the cohorts for comparison, and what your study outcome was?

Dr Lin: This is a cohort study based on two routinely collected healthcare insurance datasets: 1) Medicare, a federal health insurance program providing healthcare coverage for US residents aged ≥65 years or <65 years with disabilities; 2) Optum’s Clinformatics® Data Mart Database, which includes a national commercially insured US population. Combining the two datasets, we constructed two new-user cohorts: cohort 1 comparing warfarin vs. apixaban, and cohort 2 comparing rivaroxaban vs. apixaban.

“New use” is defined as pharmacy dispensing of the drug of interest without use of any oral anticoagulants in the preceding 365 years. This is what we call “new user, active comparator design” in comparative effectiveness research, which has been demonstrated to have the highest validity and best confounding adjustment results. We identified people with CKD stages 4 or 5 based on a validated claims-based algorithm with a positive predictive value of 76.1% when compared with CKD classification based on serum creatinine results. The primary safety outcome was hospitalization with major bleeding (including gastrointestinal, intracranial, and extracranial bleeding), and the primary effectiveness outcome was ischemic stroke. Secondary outcomes included all-cause mortality and specific types of major bleeding (ie, gastrointestinal bleeding and intracranial bleeding, respectively). Outcomes were defined using previously validated algorithms, with positive predictive values of 85%-90% for ischemic stroke and 86%-96% for bleeding outcomes.

AJKDBlog: Looking at the results, we see the baseline characteristics for your groups were very similar with regard to age, CHADS2VASC scores, and comorbidities.  Did you find a difference between the groups with regard to bleeding or ischemic stroke outcomes?

Dr Lin: After propensity score matching, the patient characteristics were very similar, indicating successful balancing of the confounders. Before matching, we note that in the cohort comparing warfarin to apixaban, warfarin users are slightly younger, with a lower proportion of women, comparable prevalence of cardiovascular comorbidities, and a slightly higher proportion of GI bleeding at baseline. In contrast, before propensity score matching, in the second cohort comparing rivaroxaban to apixaban, rivaroxaban users are slightly younger, with a comparable proportion of women, a lower prevalence of cardiovascular comorbidities, and a comparable proportion of GI bleeding at baseline. Overall, we believe the amount of potential confounding (ie, differential background risks of different users under comparison) is not large, and they are adequately controlled by our propensity score matching.

Compared to apixaban, warfarin was associated with both a higher rate of major bleeding (HR 1.85) and ischemic stroke (HR 1.14).  Compared to apixaban, rivaroxaban was associated with a higher rate of major bleeding, which is driven by extracranial bleeding (eg, GI bleeding). In contrast, apixaban and rivaroxaban users had comparable risks of intracranial bleeding. Also, the risk of ischemic stroke was not significantly different when comparing apixaban to rivaroxaban. One possible contributing factor of the observed differences is that apixaban has lower renal clearance than rivaroxaban.

Cumulative incidence curves for major bleeding (A) and ischemic stroke (B) for warfarin versus apixaban and for major bleeding (C) and ischemic stroke (D) for rivaroxaban versus apixaban. Figure 1 from Fu et al, © National Kidney Foundation.

AJKDBlog:  What I found so unique about this paper is the head-to-head comparison between different DOACs.  Has this been done in the general population, and did we also find superiority with apixaban with regard to bleeding risk?  What are some reasons for this?

Dr Lin: There are a few head-to-head studies in the general population or older adults comparing DOACs but not in patients with advanced CKD. In the general population, compared to apixaban, rivaroxaban was associated with higher rates of GI bleeding, higher risk of ischemic stroke, and comparable risk of intracranial bleeding. Similarly, in patients 65 years or older with frailty, rivaroxaban was associated with a higher risk of extracranial bleeding and a comparable risk of ischemic stroke or intracranial bleeding, compared with apixaban.

AJKDBlog: My big takeaway from this paper is that apixaban should be the preferred anticoagulant in CKD 4-5 patients with A-fib.  Could you generalize this takeaway to patients with PE/DVTs and other thrombotic events?

Dr Lin: I would be cautious about generalizing these findings from AF to PE/DVT population because the prescribing considerations are quite different in the two indications. While lifelong anticoagulation is generally recommended to those with AF with high stroke risk based on CHADS2VASC scores, clinical guidelines recommend clinicians evaluate if the PE/DVT is provoked or not. If provoking factors are reversible, the oral anticoagulants can be discontinued after 3-6 months. Because of the differences in prescribing and deprescribing considerations and the differences in baseline comorbidities associated with AF vs. PE/DVT (e.g., more malignancy and fractures/orthopedic surgery in those with PE/DVT), it may be risky to assume generalizability from AF patients to the PE/DVT population.

AJKDBlog: Thank you for taking the time to do this interview!

To view Fu et al, please visit AJKD.org:
Title: Comparative Safety and Effectiveness of Warfarin or Rivaroxaban Versus Apixaban in Patients With Advanced CKD and Atrial Fibrillation: Nationwide US Cohort Study
Authors: Edouard L. Fu, Rishi J. Desai, Julie M. Paik, Dae Hyun Kim, Yichi Zhang, Julianna M. Mastrorilli, Alexander Cervone, Kueiyu Joshua Lin
DOI: 10.1053/j.ajkd.2023.08.017