The Reply

Abstract

Glasheen and Prochazka highlight some limitations that they suggest explain our study’s negative findings.1Stroud L.F. Mamdami M.M. Kopp A. Bell C.M. The safety of levofloxacin in elderly patients on warfarin.Am J Med. 2005; 118 (e7-1417.e12): 1417Abstract Full Text Full Text PDF PubMed Scopus (12) Google Scholar Specifically, they describe the absence of clinical information and our inability to document warfarin dose changes. They posit that co-existing medication changes may have had a protective effect on patients and our failure to include less severe patient outcomes could have biased the findings. The administrative data we used was detailed, but we were unable to control for changes in drug dosage or patient monitoring. Indeed, significant elevations in international normalized ratio (INR) could have occurred in many patients. However, the association between INR and hemorrhage is not perfect. First, assessing the effect of INR level is problematic unless data are available for the time period just preceding and at the time of the bleed. This is often not feasible, even in prospective trials.2Douketis J.D. Arneklev K. Goldhaber S.Z. et al.Comparison of bleeding in patients with nonvalvular atrial fibrillation treated with ximelagatran or warfarin: assessment of incidence, case-fatality rate, time course and sites of bleeding, and risk factors for bleeding.Arch Intern Med. 2006; 166: 853-859Crossref PubMed Scopus (87) Google Scholar Second, the INR level may increase as a consequence of bleeding. Therefore, data at the time of the hemorrhage may not reflect the effect of over-anticoagulation. Third, bleeding can occur irrespective of the INR level, whether supra-therapeutic (>4.5) or sub-therapeutic (<2.0).3Kucher N. Connolly S. Beckman J.A. et al.International normalized ratio increase before warfarin-associated hemorrhage: brief and subtle.Arch Intern Med. 2004; 164: 2176-2179Crossref PubMed Scopus (42) Google Scholar, 4Fanikos J. Grasso-Correnti H. Shah R. et al.Major bleeding complications in a specialized anticoagulation service.Am J Cardiol. 2005; 96: 595-598Abstract Full Text Full Text PDF PubMed Scopus (63) Google Scholar Moreover, our model incorporated drugs interacting with warfarin. Finally, our analysis did control for physician consultations to address the intensity of anticoagulation monitoring. Including less severe outcomes such as clinic appointments or emergency department visits may have captured more adverse events and increased statistical power. However, we could then be criticized for heterogeneous composite endpoints that include diverse severity levels ranging from an added physician visit to death (considerably different outcomes when considering a patient’s perspective). Had we demonstrated an adverse effect in this circumstance, others might argue for more severe clinically relevant endpoints. Glasheen and Prochazka also compare their study results with our work.5Glasheen J.J. Fugit R.V. Prochazka A.V. The risk of overanticoagulation with antibiotic use in outpatients on stable warfarin regimens.J Gen Intern Med. 2005; 20: 653-656Crossref PubMed Google Scholar Some of the variation in findings may relate to the different designs, choice of comparators, and statistical power. Our study used a population-based, nested case-control method. There were 4269 people who were on chronic warfarin therapy matched to 17,048 control patients requiring treatment with an antibiotic for similar disease indications. We accounted for many covariates. In contrast, Glasheen et al used a cohort from a single Veteran’s Affairs Center. They assessed INR levels in 27 patients on warfarin also prescribed levofloxacin. They compared their findings with 3 other antibiotics that also demonstrated significant INR elevations (possibly implicating the effect of the illness). Further, most patients had no significant INR elevation. No hemorrhagic outcomes were observed. We are unsure whether the main reason both our studies did not demonstrate increased hemorrhagic events for patients prescribed levofloxacin while on warfarin is due to small sample size. Indeed, it is difficult to detect these events in population-based research. However, we were able to demonstrate an effect in patients prescribed cefuroxime, a drug with no previous reports of warfarin interaction. Much of the criticism raised by Glasheen and Prochazka, such as warfarin dosing, INR monitoring, and interacting drugs, are part of usual clinical practice. By including them in a real-world analysis, we maximize the generalizability of our findings. Still, we agree with their caution that patients on warfarin should be closely monitored after the addition of any medication. The Safety of Levofloxacin in Patients on WarfarinThe American Journal of MedicineVol. 120Issue 4PreviewWe would like to commend Stroud and colleagues for their excellent work regarding the safety of levofloxacin use in elderly patients on warfarin.1 Their large population-based study attempts to rectify the conflicting evidence regarding the warfarin-levofloxacin interaction. They correctly note that prior case reports and case series appear to contradict data from more controlled trials. They also note a paucity of studies evaluating this interaction in acutely ill individuals, their so-called “real-world practice.” Full-Text PDF