A Day of Data: Signal of Mortality Increase With Paclitaxel Devices in PAD Continues to Vex

It has been nearly 3 months since a controversial meta-analysis suggested that paclitaxel-based balloons and stents may be associated with higher risk of death than nondrug devices in patients with femoropopliteal artery disease. On the first day of a meeting dedicated to the issue at the VIVA Physicians’ Vascular Leaders Forum (VLF), experts continued to express bewilderment at the findings and what they might mean.

Prior to the sessions, steering committee member Christopher J. White, MD (Ochsner Medical Center, New Orleans, LA), told TCTMD that there is still a heavy emphasis on trying to understand and not “sweep anything under the rug” in terms of exploring the correlation suggested by the meta-analysis of Konstantinos Katsanos, MD, PhD (Patras University Hospital, Rion, Greece). In the paper, Katsanos and colleagues called for urgent review of the findings, which has been ongoing in the form of special sessions at various meetings and reanalysis of some of the large RCTs with the inclusion of patient-level data that were not available to Katsanos for the meta-analysis.

“The goal here is to provide a safe space for an honest discussion about the safety signal and about the strengths and weaknesses of Katsanos' methodology,” White commented.

Early in the day, multiple researchers presented basic science data, background information on the history and biology of paclitaxel, oncologic viewpoints on how paclitaxel is successfully used in cancer patients, and theories on trial design.

At one point, panelist Barry T. Katzen, MD (Miami Cardiac & Vascular Institute, FL), questioned whether the basic science being presented was in sync with the fundamental concerns that practitioners and patients have regarding whether the mortality signal for paclitaxel devices is real.

Others, like Peter Schneider, MD (Hawaii Permanente Group, Honolulu), and Raghu Kolluri, MD (OhioHealth Riverside Methodist Hospital Columbus), noted that mortality has not traditionally been a focus of any of the studies supporting paclitaxel-based balloons and stents for use in the peripheral arteries, and that the mechanism that would explain an impact of tiny doses of paclitaxel on long-term mortality is difficult to grasp even after months of discussions and data presentations.

In a highly anticipated presentation via Skype, Katsanos showed new data that restricted the analysis to 15 studies for the TLR endpoint and 16 studies for the mortality endpoint—both at 2 years. For TLR, paclitaxel was associated with a significant -19% absolute risk difference compared with uncoated balloon or stent and a risk ratio of 0.45 (95% CI 0.36-0.55). In contrast, the mortality difference at 2 years for paclitaxel compared with uncoated balloon or stent amounted to 3%, with a risk ratio of 1.48 (95% CI 1.05-2.07).

Katsanos said the numbers indicate that the efficacy and safety endpoints are coming together from opposite directions. In further Bayesian analysis of the highest-quality RCTs and registries from the original meta-analysis, he showed that those data confirm the original findings with an excess mortality of 56% at 2 years for paclitaxel versus control.

But ultimately, Katsanos acknowledged that in terms of understanding the signal “everything is incomplete and we need to . . . take actions.”

In a critical appraisal of the original meta-analysis, William A. Gray, MD (Lankenau Heart Institute, Wynnewood, PA), pointed out multiple issues including distribution assumptions, which when corrected for show no significant mortality differences between paclitaxel devices and percutaneous transluminal angioplasty. Additionally, Gray noted mathematical errors, selection bias due to lack of complete follow-up, problems with application of the dose-time product, and assumptions about uniform delivery of paclitaxel. Importantly, Gray said discussions he has had with statisticians suggest that to adequately assess mortality it would be necessary to enroll 2,000 to 6,000 patients.

At one point, Gray asked Katsanos if he would agree that any dose-response relationship to mortality is too confounded by a variety of factors to make definitive statements about it.

“I agree that there is heterogeneity, there is confounding that we may have missed,” Katsanos said. “I have shown, however, that the dose-response model is very valid for the TLR and it seems to be following the same lines in the opposite direction for mortality risk. So, again, I put it to you: why does everybody accept the benefit so easily and we don’t really accept the side effect?”

Gray maintained that the TLR is a local effect of paclitaxel, while the mortality is a systemic effect. “The two are not linked or married in any meaningful way,” he said.

Panelist Koen Deloose, MD (AZ Sint Blasius Dendermonde, Belgium), said he continues to believe that the Katsanos meta-analysis represents a statistical difference that has not been born out in independent patient-level data.

“From a clinical point of view, I don’t see, at the moment, any association or any cause between paclitaxel and mortality rate,” he said.