June 20, 2019
FDA Paclitaxel Advisory Meeting Day 1: Panel Addresses Updated Data, Signal Presence, Class Effect, Missing Data, Subgroups, and Causes of Death
June 19, 2019—Among efforts to explore and address concerns raised by recent meta-analysis data that found an increased mortality signal in patients treated with lower extremity paclitaxel delivery devices, the US Food & Drug Administration convened an advisory committee to be held June 19–20, 2019, in Gaithersburg, Maryland.
The original meta-analysis was authored by Katsanos et al and published in the Journal of the American Heart Association in December 2018. Since then, industry, clinical investigators, and regulatory bodies have initiated additional reviews of all available data, resulting in numerous presentations and publications, as well as the halt (and now planned resumption) of several large trials.
The advisory committee meeting's first day was highlighted by the presentation and scrutiny of updated data sets, followed by a public comment period and ultimately the panel's address of five specific questions posed by the FDA. (Questions can be reviewed in full here.)
DAY 1 SUMMARY
The interventional physician and industry presentations were largely favorable and consistent in their message as to a lack of causal connection between the respective devices and increased late mortality, as well as any relationship between increased dose and higher mortality as suggested by the Katsanos et al findings.
Practicing vascular specialists spoke to the benefits of the therapies, including vascular surgeon Daniel Clair, MD, who said the value of these devices is indisputable. In their responding to the questions raised by the Katsanos et al meta-analysis, "Industry competitors have become partners," he said, describing a level of collaboration he has not seen before. "I'm proud to stand with them in this effort," asserted Dr. Clair, voicing concern as to the effect of the FDA's communication suggesting limited use of paclitaxel devices for the time being.
Gary Ansel, MD, agreed, stating that, "Until a relationship between drug-based devices and mortality is definitive, withholding this therapy places patients at undue risk."
Beyond the data presented by the five companies with approved paclitaxel devices in the United States, additional data sets and perspectives were shared. For example, Eric Secemsky, MD, offered more data from his group's analysis (current N = 152,473) of the Medicare database through a median 799 days of follow-up (longest, 1,573 days). The study has not yet found evidence of harm from paclitaxel devices, and sensitivity analyses found no differences in subgroups or across devices. Data will continue to be collected to examine the potential connection.
Dr. Andrew Holden presented a just-published data set exploring perceived errors in dose/time calculations in the original Katsanos et al paper. The authors asked for data from each company that presently has 5-year follow-up and conducted a uniform analysis, which ultimately showed that the device with the highest paclitaxel dose had the lowest mortality rate, and vice-versa. The full article is available here.
Raised and debated in previous forums, challenges remain in how to address whether control patients were ever treated with paclitaxel. The relative benefits of intention-to-treat (ITT), as-treated (AT), and modified as-treated (mAT) were discussed at several points during the panel session. Presenting the updated Cook Medical Zilver PTX data set, Michael Dake, MD, suggested that ITT analysis is not an appropriate method to assess for paclitaxel-related mortality. He shared an updated mAT analysis that included protocol-allowed crossover from control to PTX later in the trial, which was not included in the Katsanos et al meta-analysis, and concluded that if Zilver PTX patients are analyzed based on how they were actually treated, the results for the entire meta-analysis are not significant.
Konstantinos Katsanos, MD, also shared additional data on behalf of the original meta-analysis authors. "We have presented an updated version of our meta-analysis with increased sample size through 5 years," said Dr. Katsanos, summarizing his presentation in comments to Endovascular Today. "We have also performed a risk-benefit analysis illustrating that different devices demonstrate different risk-benefit patterns. However, the mortality signal remains strong and consistent in the long-term as previously published. In addition, we explored potential heterogeneity of the treatment effect (ie, the increased paclitaxel-associated mortality risk). Based on aggregate trial-level data and the ZILVER PTX individual patient data, we have shown that the observed mortality hazard seems to vary according to background patient risk. In detail, there is an inverse association between them—the mortality hazard is getting smaller as the number of patient risk factors is increasing. We do not know whether this is because of competing risks or some other unknown biological factor. However, we propose that patient-risk stratified analyses of individual patient data are performed to look into this further."
Overall, there seemed to be a discrepancy between the sentiments and data shared by the presenters representing industry trials and most of the open public session commentators in comparison to the later responses of the panel. As panelist Joaquin Cigarroa, MD, pointed out, FDA's finding that a signal continues to be present conflicts with the presented industry findings, a challenge fellow panelist John Hirshfeld, Jr, MD, also voiced. "We're being subjected to a forest of numbers, and those presented by industry are not the same as those presented by FDA," he said.
The agency was asked why this is, and Eleni Whatley, PhD, explained that the FDA and the companies reviewed different data sets in different ways. The CMS database analyses, those done by VIVA Physicians (to be covered later in greater detail), and Holden et al, as well as the original and updated Katsanos et al meta-analyses, were also each unique in their own populations and/or methods. The discrepancies and seeming preference for the FDA and Katsanos et al data factored into the panel's day 1 deliberation.
PANEL QUESTIONS, DELIBERATIONS, AND ANSWERS
Signal Presence and Strength
The first element the panel was asked to address was the potential presence and magnitude of a late mortality signal associated with paclitaxel devices, based on their interpretation of the data presented. Ultimately, the panel reached consensus on the presence of a signal based on the aggregate data presented, but they lacked clarity as to the signal's strength, citing differences in each data set presented. They were also unable to definitively determine a causal mechanism, certain only as to the statistical finding, as was the case with the original meta-analysis and FDA analysis that followed. One panelist, vascular surgeon Col. Todd Rasmussen, MD, believed the findings to be statistically significant, but not "practically significant," a distinction he feels is important given the therapies' ability to improve patient quality of life.
During deliberation of this first question, the panel focused significant attention on the need for more rigor in the conduct of peripheral clinical trials—specifically with regard to what it considered high numbers of patients lost to long-term follow-up. Although the United States investigational device exemption (IDE) trials were designed to follow the mortality endpoint only out to 1 year, having a more complete data set would have likely provided a more definitive picture.
Panelist David Kandzari, MD, an interventional cardiologist, reflected on the previous experience assessing paclitaxel-eluting stents in the coronary arteries, a landscape he opined had a higher level of data than that seen in the lower extremity trials. Dr. Kandzari wondered if the plausible mechanism might not be the drug/device combination but rather the way the trials were conducted. A point of consensus among panelists and presenters throughout the day was the opportunity to learn more about the ideal means of conducting and following up peripheral trials going forward.
The second question posed to the panel regarded class effect and the strength of evidence as to whether all United States–approved paclitaxel femoropopliteal devices carried a late mortality risk, regardless of dose, platform (stent or balloon), and drug formation. The dose component of this question was frequently discussed throughout the day and focal in the industry presentations, as well as an update from the meta-analysis authors by Dr. Katsanos.
In general, the panel did not think there were sufficient data to definitively rule in or out a class effect. Philips' Stellarex device, which showed no significant difference and even a numerical decrease in mortality compared with control at the available 3-year follow-up was specifically asked about, as was Boston Scientific Corporation's Eluvia drug-eluting stent (DES), which was compared with another DES rather than a bare-metal device in its IDE trial (and also not included in the original meta-analysis). However, the panel did not see reason to separate the platforms at this time, believing the industry should continue to work together to address the question.
This addressed the spirit of the FDA's question, which was whether a potential future trial to further examine any mortality increase would be viewed as valid if it commingled technologies in its study arm. To power a trial for a mortality signal, a vastly larger population size would be needed, likely requiring multiple device platforms to be included.
Question 3 asked the panelists to weigh the effects of missing data on the clarity of the meta-analysis and subsequent data reviews. In the wake of the meta-analysis, industry made strides in collecting as much vital status data as possible to close the gap, though not all patients could be tracked. The panel felt that the overall data remain incomplete, challenging their evaluation.
Echoing previous discussions, the panel referred to the need for more thorough follow-up in such trials, as did Bram Zuckerman, MD, Director of the FDA's Office of Cardiovascular Devices. Dr. Zuckerman said that the emphasis on increasing the quality of trials and their follow-up will not stop after the current evaluation, describing the situation as a "very disappointing experience."
The panel did not identify specific subgroup signals that identified which patients might have increased risk or benefit from paclitaxel therapies, citing a need for more data and more heterogeneity among them in order to do so. Cardiologist Mitchell Krucoff, MD, weighed in that the available data were insufficient overall, much less to determine subgroup effects requiring reduction of a denominator that is already "torturously small." Dr. Krucoff also pointed out the challenge posed by many patients having multiple comorbidities.
Causes of Death
The final question of the day was whether the cause of death data support the presence of a late mortality signal associated with paclitaxel treatment, and whether the data suggest a mechanism for this signal. Overall, the panel lacked confidence as to the ascertainment of the correct causes of death. Earlier in the sessions, the challenges in determining causes of death in patients with multiple comorbid conditions were discussed, as was that of in linking late deaths to specific procedures performed years earlier. Also described in detail were the somewhat widespread causes of death reported and the difficulties in evaluating causes of death deemed at the time to be "unknown" in this setting.
There was no consensus as a presence of a data suggesting a clear mechanism, but the possibility was raised that just because a causal link is not currently known, this does not mean one may not be determined with future understanding.
DAY 2 PREVIEW
The June 20 session will further review:
- the paclitaxel dose/mortality relationship
- insights from preclinical studies
- benefit-risk profiles
- objectives of any new postmarket studies
- potential labeling modifications
- changes to future study designs
- risk-benefit considerations in other indications (ie, arteriovenous access, critical limb ischemia)
To continue reading coverage of Day 2 of the hearing, please click here.