FDA Panel Talks Labeling, Urges Cooperation for Paclitaxel Mortality Signal in PAD
The data are a mess, and more uniformity is needed in clinical trials going forward. So concludes some “inconclusive” debate.
(UPDATED) In an underwhelming finale to 2 days of debate and discussion over a long-term mortality signal associated with paclitaxel-based devices for PAD, the Circulatory System Devices panel of the US Food and Drug Administration concluded that more cooperation among all stakeholders is urgently needed to deal with the current mess of data that proved insurmountable during the session. There’s also an equally pressing need for uniformity in trial design and reporting accountability, panel members said here.
Speaking with TCTMD, panel chair Richard Lange, MD (Texas Tech University Health Sciences Center, El Paso), said the likely next step for the agency is to ask for additional studies to get more conclusive date on the mortality signal. Throughout this week’s meeting, the panel heard from FDA statisticians and others about how best to do that, as well as how to clean up the existing data from the pivotal device RCTs in which a wide range of patients are still missing or lost to follow-up despite months of industry’s attempts to reconcile it.
Lange added that his sense is that the agency "absolutely" will move toward some type of device labeling modification to reflect the issues raised since the publication of the Katsanos meta-analysis that first raised the alarm of a potential increased mortality signal beginning at around 2 years in patients treated with paclitaxel-based DES or drug-coated balloons (DCB) across a range of large trials and small studies.
"I think they'll be working to make it simple enough that patients can understand it and it's not confusing, but also make it true to the available data," he said. “It should describe our current understanding as simply as possible.”
Over the course of the meeting, the panel addressed 12 questions posed by the FDA related to their need for more information or input. On the afternoon of the second day they began with one of the trickier subjects: whether there is a dose-response relationship between paclitaxel and mortality. During the discussion, “inconclusive” came up multiple times and was the overwhelming answer by all.
I think they'll be working to make [labeling] simple enough that patients can understand it and it's not confusing, but also make it true to the available data. Richard Lange
When the panelists were asked if available preclinical animal studies provide mechanistic insight into the signal, the answer was no. Some on the panel, however, felt that new animal data could be helpful and discussed the possibility of studying paclitaxel devices in aging rabbit models, animals exposed to cigarette smoke, and animals with cancer. Only one member seemed to feel that new animal models were not necessary, saying it would likely be a step backwards and unproductive.
The panel was overwhelmingly unanimous that paclitaxel-based devices have demonstrated a benefit for many patients but struggled with issues of risk and how it should be conveyed to patients. Mitchell W. Krucoff, MD (Duke University Medical Center, Durham, NC), said: “We need to understand how to communicate short-term benefit in the face of long-term risk, which is uncertain.” As a group, the panel was fairly uniform in believing risk-benefit ratio is individual and that patient preference and informed consent is paramount.
When the panel came to the discussion about what modifications, if any, should be made to the labeling of currently approved paclitaxel-coated devices it was clear that all seemed to feel that patients need to be informed about the signal in some way through the labeling. Several members noted that the meta-analysis and the discussions that have been held regarding it are far too complicated to try to explain to patients.
John C. Somberg, MD (Rush University in conjunction with the University of Chicago, Lake Bluff, IL), volunteered that the labeling should reference the Katsanos meta-analysis and say that it shows that there may be late mortality that must be balanced against early and sustained benefit of the device for the individual patient. While most seemed in agreement, the industry rep on the panel asked whether devices that are approved but were not included in the meta-analysis should be subjected to the same label warning. Panel members, invoking the totality of the data, said that they should all be subjected to the same type of warning on the label, with Lange noting that all patients need the information to make an informed decision.
Some of the most informative discussion came with regard to how the Katsanos meta-analysis and its fallout will affect future studies. At least one panel member implied that the FDA should force the sponsors of the pivotal trials to find their missing data. The FDA’s Bram Zuckerman, MD, responded that the agency has limited authority, adding, “We need to work as a community.”
Echoing that sentiment, other panel members discussed the need for some sort of consortium of manufacturers and trialists that would work with the FDA to ensure appropriate and timely data collection in the long-term follow-up periods. They also suggested that a uniform case report form, which is used in other medical fields such as oncology and heart failure, should be adopted so that all ongoing registries are collecting data in the same way. This would potentially make it much easier to look retrospectively at common elements.
We need to understand how to communicate short-term benefit in the face of long-term risk, which is uncertain. Mitchell W. Krucoff
In the end, Zuckerman told the panel that the suggestions of the last 2 days had been helpful and the agency will continue to work toward analyzing the data they have and consider creative ways to move forward. There was no discussion of removing the devices from the market but the panel members made it clear that they want to see some of their suggestions implemented and not have to revisit the situation in coming years, with Lange saying it is “morally imperative” that industry understands their obligation to long-term data collection, ideally through 5 years of follow up. John W. Hirshfeld Jr, MD (University of Pennsylvania Medical Center, Philadelphia, PA), said the last 2 days had left him with “a déjà vu moment,” referencing the coronary DES firestorm of 2006 that raised the issue of needing better postmarket surveillance for devices.
“That was 13 years ago and we’re having the same discussion today,” Hirshfeld said. “This should be a message to everyone here that this is something that will happen again if we don’t close the loop on this process.”
Some ‘Jaw-dropping’ Data
The panel began the day with a return of the FDA’s reviewers and statisticians who presented reevaluations of preclinical animal data on paclitaxel. They concluded that the long-term effects of treatment with paclitaxel-coated devices in the peripherals are unclear and that even at very low drug concentrations, they cannot rule out an association with adverse effects. In the FDA’s dose analysis, which looked at the as-treated populations from the individual pivotal trials with 5-year mortality, the only trial that showed a possible trend for higher mortality with increasing dose was LEVANT 2. However, the confidence intervals overlapped, resulting in a P value of 0.038.
FDA medical officer Donna Buckley, MD, reviewed the known incidence of clinically driven TLR at 5 years in the paclitaxel-treated populations. The rate of TLR was 21% lower in the drug-treated than in the uncoated device group based on an averaging of data from LEVANT 2, IN.PACT SFA I and II, and ZILVER PTX. According to Buckley, the number-needed-to-treat (NNT) to avoid one TLR was 13 at 5 years. However, the number-needed-to-harm (NNH) was almost identical, 14 at 5 years.
Somberg called the NNT and NNH data “jaw-dropping,” adding that these numbers have to be conveyed to patients because they may be trading convenience for enhanced mortality if the numbers are correct.
Later in the day, the individual sponsors presented their own NNT and NNH data, most which showed lower NNTs and higher NNHs than the aggregated FDA data.
Addressing the question of the feasibility of new RCTs of paclitaxel-coated versus uncoated devices, Buckley’s prepared statistical analyses suggested that to ideally control for mortality at 3 years, a sample size of 43,276 patients would be necessary. For 5-year mortality, patient size would need to be over 20,000 to have confidence in both relative risks and hazard ratios.
Somberg responded that the numbers are substantial and pose challenges. He then questioned if the sample sizes could be culled down by increasing the number of high-risk patients.
To TCTMD, Lange said the least burdensome way to do further analyses is to draw from the existing pool of patients with complete data, then determine how many more patients need to be added and what length of follow-up is sufficient. The number of new patients that need to be followed could be as low as 2,000 to 4,000. “If each of the companies step up to the plate together, such a study could be done very quickly,” he added.
Somberg also criticized the FDA’s dose-response analysis. In one of the agency’s slides in which patients were separated into seven different dose categories, the highest-dose categories each had only two patients while lower doses had several dozen each. Somberg called it “distorted” and “off balance” as an agency statistician struggled to explain that it again came down to them not having enough data with which to work.
Despite advice from committee member Joaquin E. Cigarroa, MD (Knight Cardiovascular Institute, Portland, OR), that “we have to think differently” and look beyond the existing data for further clarity, the open public hearing portion of the morning saw multiple presentations from researchers reviewing animal data and other long-published literature on paclitaxel-treated PAD patients showing no evidence of a mortality signal.
But toward the end, new data did come from Robert W. Yeh, MD (Beth Israel Deaconess Medical Center, Boston, MA), who showed an analysis of real-world deidentified patients who underwent revascularization with a paclitaxel DCB or DES, BMS, or uncoated balloon between 2015 and 2017. Adjusted all-cause mortality was the same—14.9% at a median follow-up of 763 days (ranging from 522 to 1,028 days)—in the drug-treated groups compared with the BMS or uncoated balloon groups (P = 0.11). These data bolster those of Medicare analyses of over 152,000 patients that were reviewed and updated yesterday and that show no mortality signal among the paclitaxel patients. Eric A. Secemsky, MD (Beth Israel Deaconess Medical Center), who led the Medicare analysis, reiterated today that the RCT populations amount to “a sliver” of the overall PAD patients needing treatment in daily clinical practice.