Risk of Arterial Thrombosis Not Increased by Sorafenib or Sunitinib

Abstract

TO THE EDITOR: In an analysis assessing the risk of arterial thrombotic events (ATEs) with sunitinib or sorafenib in patients with cancer, Choueiri et al reviewed multiple studies using these agents. However, their finding of an increased risk of ATEs rests on their analysis of only three randomized controlled trials comparing sunitinib or sorafenib therapy with placebos, because the expected incidence of ATEs in uncontrolled trials is uncertain. They calculate a relative risk for thrombosis of 2.39 with sunitinib; however, this was not statistically significant (95% CI, 0.12 to 49.41). Their calculated relative risk in one of the three trials, that by Escudier et al, comparing sorafenib with a placebo for renal cell cancer, was statistically significant. However, because the authors failed to distinguish between incidence per treatment episode and incidence per unit of time (the more relevant statistic), their conclusions may be erroneous. In oncology clinical trials such as those reviewed, patients are generally not treated for a predetermined, fixed period of time; rather, they are treated until their cancer shows evidence of progression on computed tomography scans, at which time the patient is removed from the trial, and further adverse events are not recorded beyond 30 days after the end of treatment. Because sunitinib and sorafenib are beneficial oncology drugs, the patients in these trials often continued treatment with the active drug much longer than patients given a placebo; therefore, patients receiving the active drug had more time to develop conditions leading to adverse events, such as arterial thrombi. For example, in the study by Escudier et al, the median treatment duration with sorafenib was 5.8 months, versus 3 months with the placebo. This would translate into an average observation time of 6.8 months for patients treated with sorafenib, versus 4 months for patients who received the placebo.Twelve ATEs were recorded in patients receiving sorafenib during 3,067 patient-months of observation, versus two events in patients receiving the placebo during 1,804 patient-months of observation. The difference here is not significant (Fisher’s exact test, two-tailed, P .097). These three randomized trials show no clear evidence of an increased rate of significant arterial thrombi per patient per month with either sorafenib or sunitinib, compared with placebos. The same authors made the same error of confusing adverse event per treatment episode with the more important statistic of adverse event per unit of time in their article on bleeding with sunitinib and sorafenib, as was described in my letter. In their reply, the authors asserted that adverse effects, such as bleeding or thrombosis, occur early after drug exposure; however, I am unaware of any studies or clinical experience to support that hypothesis. A similar error was also made by Nalluri et al in a meta-analysis of venous thrombosis with bevacizumab, and when they calculated adverse events per unit of time, their results were no longer significant.