Solving the mystery of excessive warfarin-induced bleeding: a personal historical perspective.

http://dx.doi.org/10.1160/TH14-05-0433 Thromb Haemost 2014; 112: 853–856 For the first four decades after their introduction in 1941, the clinical use of vitamin K antagonists (VKAs) was hampered by haphazard laboratory control, lack of rigorous evidence for efficacy, and bleeding complications. In this narrative historical perspective, I discuss my personal experience in providing evidence for their clinical efficacy and in making oral anticoagulation with VKAs safer. I was joined in my efforts by key members of the McMaster University Thrombosis Program. Coumarins were first used clinically in 1941 (1, 2). For the next four decades, their laboratory control was haphazard, evidence for their efficacy was weak and physicians were reluctant to use them because of the fear of bleeding. Three advances have made vitamin K antagonists (VKAs) safer: lowering the intensity of their effect, introducing the international normalized ratio (INR), and improving anticoagulant monitoring. I was first alerted to the problem by Russell Hull in 1977. At that time, our approach was to treat patients who had acute venous thrombosis with unfractionated heparin for about 10 days, start warfarin on day 5 and then discharge them on adjusted dose warfarin targeted to a prothrombin time (PT) ratio of 1.5 to 2.5. One of the senior physicians refused to allow Russell to prescribe warfarin for his patients because he said that warfarin caused his patients to bleed. Since low-dose heparin (5,000 U twice daily subcutaneously) was effective for primary prevention of venous thrombosis, we wondered if we could replace warfarin with low-dose heparin. To test this hypothesis, we performed a randomised controlled trial in which we compared our standard approach of adjusted dose warfarin with low-dose heparin. The results showed that patients in the warfarin arm had a high rate of bleeding (21%) and that low-dose heparin was ineffective (▶ Table 1) (3). We presented the results to our colleagues in Hamilton who did not believe that warfarin adjusted to a PT ratio of 1.5 to 2.5 caused so much bleeding. We decided to perform a second clinical trial with an identical design, except that the maintenance dose of heparin given after discharge from hospital was adjusted to an activated partial thromboplastin time (APTT) ratio of 1.5 (4). Heparin and warfarin were both effective (recurrent thromboembolism rates of 3.8% and 1.9%, respectively), but as in the first study, warfarin was associated with a higher rate of bleeding than adjusted dose heparin (17% and 1.8%, respectively) (▶ Table 2). This second study convinced us that our warfarin regimen was associated with an unacceptably high risk of bleeding. Although adjusted-dose heparin was as effective as warfarin, and much safer, this heparin regimen was impractical for outpatient use. In retrospect, we should have lowered the intensity of warfarin. It was not until I presented our results at a meeting in London in 1980, that the penny dropped. My presentation was attended by a knowledgeable audience that was vocal in their disbelief of our findings. Our high rates of bleeding just did not match their experience. The most vocal dissenter was Leon Poller, a pioneer in the field of PT reagent development and VKA monitoring. For over a decade, he had been promoting a national PT reagent standard in the UK and by the early 1980s, his laboratory was distributing their human brain thromboplastin to clinical laboratories throughout the UK. We were using Simplastin, a PT reagent that was commonly used in North America. I met with Leon after my presentation and we continued our debate over a drink. I asked him what PT ratio he targeted, and he said 2.0 to 3.0, which was slightly higher than ours. Further questioning revealed that the mean dose of warfarin used in his patients was 4 mg per day; ours was over 5 mg per day. Leon said that he had just finished a warfarin dosage survey in seven countries (UK, US, Canada, Sweden, Australia, Hong Kong, and Zimbabwe) and found that the average daily doses varied from a low of about 2 mg to a high of over 8 mg (5). It was now clear that we needed to do a third clinical trial comparing Poller’s warfarin regimen with ours. Leon was enthusiastic. In our third randomized trial, patients were assigned to our standard regimen, warfarin dose-adjusted to a mean PT ratio of 1.8 using the Simplastin reagent (which corresponded to a PT ratio of 3.2 with Poller’s Manchester reagent), or a less intense regimen wherein the warfarin was dose-adjusted to a mean Simplastin PT ratio of 1.3, which corresponded to a mean PT ratio of 2.0 with the Manchester reagent (6). Two patients (4%) in the less intense group had haemorrhagic complications, compared with 11 (22%) in the more intensely anticoagulated group. The frequency of recurrent venous thromboembolism was 2% in both groups (▶ Table 3). Subsequently, we performed a fourth trial in which we evaluated the less intense warfarin regimen in patients with bioprosthetic heart valves and showed that the less intense regimen was effective but safer than the higher intensity regimen (▶ Table 4) (7).