Initial SAFE-PAD Data Find No Mortality Difference Between Paclitaxel-Coated Devices and Non–Drug-Coated Devices for All-Cause Mortality in Fempop Revascularizations

Key Findings

• Paclitaxel-coated devices used for peripheral revascularization (median follow-up, 2.72 years) showed no increased mortality compared with non–drug-coated devices.
• Data collected so far and during long-term follow-up (when median follow-up exceeds 5 years for all patients studied), offer a mechanism for the continued safety evaluation of paclitaxel-coated devices.

ENDOVASCULAR TODAY ASKS….

We spoke with Eric A. Secemsky, MD, lead author for the SAFE-PAD study and Director of Vascular Intervention at Beth Israel Deaconess Medical Center in Boston, Massachusetts about some additional aspects of the SAFE-PAD study.

With all the data published and presented regarding the mortality question, including those you’ve published from the Medicare database, how would you summarize what these findings add to our understanding?

Thank you for the question. Thus far, the mortality signal associated with paclitaxel-coated peripheral devices have really been restricted to the pivotal trial data used for device approval. We first saw this signal in the Katsanos et al publication, with only three trials having data at 4 to 5 years. In the FDA internal meta-analysis as well as the VIVA-NAMSA meta-analysis, similar trial data were analyzed. Each time these data were examined, the mortality signal was attenuated and, at least in part, this was due to the inclusion of missing follow-up data.

However, outside of these studies, no other analysis to my knowledge has demonstrated this signal of harm. In regard to observational data, we have now looked at three different Medicare populations—the most comprehensive in our recent SAFE-PAD publication—and we have not been able to reproduce this signal of harm. The SAFE-PAD study included > 30,000 patients who had follow-up of > 4 years, thus providing one of the longest-term survival assessments to date. Furthermore, this analysis included key subgroups, including those treated as outpatients and by specific device types, which prior analyses have not been able to perform due to small sample sizes. Overall, we found no evidence of harm no matter how we analyzed the data.

These data add to the plethora of observational data that has emerged since the safety signal was first reported. For instance, data from the VQI, NCDR PVI, and the Veteran Affairs health care system all similarly demonstrated no safety concerns with paclitaxel-coated devices. There have also been large publications from outside the United States, including from the German BARMER health insurance database, which did not find harm with PCDs. In addition, we now have a prospective randomized controlled trial in SWEDEPAD, as well as the sub-analysis of the VOYAGER PAD trial, again demonstrating safety with these devices.

All things considered, it appears the pivotal trial data, which were not designed to look at long-term mortality, are now the outlier. Furthermore, we still lack a clear mechanism as to how isolated lower extremity paclitaxel exposure causes increased death, as no cause of death has been identified and increasing paclitaxel exposure has not correlated with a greater risk of death. I think we are very close if not at the point of confirming that the initial finding of harm was more a signal than a true relationship.

In the discussion section, it is noted that most procedures indexed in SAFE-PAD are retrospective to before the initial meta-analysis raising the safety signal was published and, therefore, likely differ from patients treated after that publication. Can you expand on why these specific real-world data from before that initial publication will help inform the discussion of safety going forward?

This is a really important point. We previously used peripheral PCDs liberally in clinical practice across all patient subsets, as these were considered important first-line therapies to prevent restenosis. However, after the Katsanos et al meta-analysis and the recommendation from the FDA to restrict use to only those at highest risk, national practice patterns changed. Use of these devices dropped dramatically, and patients being treated were primarily those at very high risk and likely with more guarded long-term prognoses. As such, it became harder to determine the safety of these devices in observational data when the patients being treated with these devices differed substantially than before. By restricting our study to before the Katsanos et al meta-analysis was published, we avoided the influence of this bias.

What are the benefits to having large data sets of real-world data to analyze a safety signal? What limitations does this type of data set present?

One of the main strengths of SAFE-PAD and the potential for observational data is the ability to perform timely safety evaluations. For instance, we were able to publish our first study on PCD safety among Medicare beneficiaries within about 6 weeks of the publication of the Katsanos et al meta-analysis. With SAFE-PAD, we also now have a cohort of patients that is 160 times larger than the population of patients available at 4 to 5 years in the meta-analysis. These patients also more closely approximate the patients we treat in clinical practice, as enrollment in randomized trials is often restricted to those with strict clinical and anatomic criteria.

On the other hand, not all safety assessments will have available real-world data that are able to address these concerns. For instance, this safety concern was unique in that it involved the endpoint of all-cause mortality, which is collected with high fidelity and does not require adjudication for specific causes. However, other endpoints may be more challenging to evaluate. Furthermore, treatment selection bias remains an important concern with observational studies. We are particularly proud of SAFE-PAD as we used several causal inference methods to evaluate the strength of the statistical signal observed in our primary analysis. Nonetheless, there may be scenarios where only randomization can overcome biases such as confounding by indication.

Drs. Rita F. Redberg and Mary M. McDermott published a commentary focusing in part on the high mortality seen in Medicare PAD patients regardless of device type. What are your thoughts on this finding, as well as their call to action to focus on conservative treatments?

I think their commentary is important—these patients suffer very high death rates and this needs to be considered. Death among patients with PAD is not often driven by limb events, however, but noncardiovascular death, in particular malignancy, and cardiovascular events, such as stroke and myocardial infarction. As such, this is not unique to those undergoing revascularization, but to all patients with PAD. I do think revascularization remains an important treatment tool irrespective of this. We know that peripheral intervention increases quality of life measures for patients with life-limiting claudication and improves rates of limb salvage for those with critical limb ischemia. As such, for patients who fail conservative measures or are at risk of limb loss, we should not withhold these procedures to those who may benefit.