Updated Summary-Level Analysis of Paclitaxel-Coated Devices Shows No Evidence of Increased All-Cause Mortality Risk

Key Findings

• Quantitative analysis included 34 studies involving 7,654 patients (PCD, n = 4,147; uncoated controls, n = 3,507).
• Duration of follow-up ranged from 12 to 60 months; all 34 studies reported at least 12-month follow-up, 20 studies reported 24-month follow-up, one study reported 36-month follow-up, and nine studies reported 60-month follow-up.
• Overall, there were 622 (15%) deaths in the PCD group and 475 (13.5%) deaths in the uncoated control group.
• All-cause mortality was similar between groups at 12, 24, and 60 months.

ENDOVASCULAR TODAY ASKS…

We asked study investigator Ramon L. Varcoe, MBBS, with Prince of Wales Hospital in Sydney, Australia, to elaborate on the results:

How would you briefly summarize the ways in which this new meta-analysis is both similar to and unique from the 2018 model on which it is based?

The methodology is near identical to the Journal of the American Heart Association (JAHA) meta-analysis from 2018. We did this intentionally. The JAHA meta-analysis produced findings that no one could explain. We felt a duty to investigate the possibility that those findings were due to chance, bias, or both. So, we decided to collect additional data and retrace the steps of that study. We all know that the included RCTs were powered for 1-year patency, not for mortality, let alone long-term mortality. Our rationale was that with additional numbers at longer-term follow-up and vital-status adjudication of the withdrawn patients, the additional power achieved would reduce the influence of bias/chance and produce findings that were closer to the truth.

There were two major differences between our study and the one published in 2018. The first was the number of subjects in the quantitative analysis, which amounted to almost three times as many patients with 5-year follow-up. The second was that we chose not to attempt a repeat drug dose analysis, as we felt the original methodology was flawed, and dose has subsequently been shown to be unrelated to mortality in more sophisticated patient-level studies.

Dose response calculations differ from study to study but are critical in determining causality. Based on your findings and others, what are the knowns and unknowns of paclitaxel dose response, both in terms of safety and efficacy?

It is our view that the assumptions required to estimate dose in a summary-level meta-analysis introduce a fundamental flaw into the analysis that makes the result impossible to interpret. They use mean values for lesion length and reference vessel diameter taken over the entire study. However, the use of lesion length is imprecise because it is common practice to both overlap multiple drug-coated balloons and include adjacent normal vessel wall to avoid geographic miss. Furthermore, a fundamentally misleading aspect of the original dose response calculation was the introduction of time as a factor in the assessment of overall dose. Introducing time to the exposure equation assumes that the dose is stable and remains constant over the follow-up period. However, there is no evidence from any clinical or preclinical study that this is true and no clear explanation has been given for why that assumption was made. The time parameter used in this way penalizes older studies with longer follow-up. Shorter follow-up studies will be biased toward the null, and longer follow-up studies will be biased toward finding a stronger association between PCDs and increased mortality. Patient-level data, with precise measurement of actual device length, dose, and vessel diameter is the only accurate method of overcoming those shortcomings, and multiple well-designed studies using those methods have definitively ruled out such a dose-mortality relationship.

Matching the methods to the 2018 meta-analysis, your study analyzed results on an ITT basis but noted that its safety findings would have been even more compelling if analyzed “as treated.” Can you elaborate on the as-treated scenario described?

One response that was common after the 2018 publication was for the investigators of the included RCTs to have reanalyzed their data based on whether subjects actually received a PCD. It turns out that there was significant crossover in most of the larger trials. In contacting the authors and trial investigators to gather the most accurate and updated mortality figures, it was common for that “as-treated” data to be supplied to us. There was a clear trend toward even less difference in mortality numbers when we examined the data in this way; however, we decided to remain true to the conventional ITT analysis to both maintain the anticonfounding effect of randomization and be consistent/comparable with the 2018 study.

The safety concerns raised by the initial 2018 meta-analysis resulted in regulatory warnings and other local restrictions on use, some of which remain in place, and in their conclusion, the authors state that limitations on the use of these devices are no longer justified. What do you feel is the appropriate and optimal regulatory stance or action going forward?

It is our view that the signal that first led to these warnings was a type I error (false positive due to a combination of bias and chance) and not a true danger to our patients. The inherent disadvantages of a summary-level meta-analysis performed with inadequate patient numbers at long-term follow-up led to an erroneous conclusion. Numerous subsequent studies, including SWEDEPAD, VOYAGER, and large real-world registries from numerous geographic regions, support this point of view. Given what is known about the patient and economic benefits that come with the use of PCDs, it is no longer justified to limit their use.