Recovered Paclitaxel PAD Data Shift the Mortality Picture

Since the publication in 2018 of a meta-analysis showing higher rates of mortality among patients with PAD who received paclitaxel-based balloons or stents versus placebo, the hunt has been on to find follow-up data that were missing from the pivotal clinical trials. In a newly published paper, a collaboration of researchers and industry say the recovery of the lost data changes the picture and considerably reduces the mortality signal.

The meta-analysis that kicked off the controversy was led by Konstantinos Katsanos, MD, PhD (Patras University Hospital, Rion, Greece). In a pooled numbers from 28 randomized femoropopliteal trials of DES and drug-coated balloons, he and his colleagues found a whopping 93% increased relative risk of all-cause mortality and a 6.6% increased absolute risk in paclitaxel-treated patients over 4 to 5 years. However, it quickly became apparent as the troubling data were discussed at annual society meetings and at a 2-day US Food and Drug Administration advisory panel session that a significant contingent of patients—up to 30%—had been lost to follow-up, partly because the studies were not designed to track long-term mortality.

VIVA Physicians then announced that the organization would work with industry to find as much of the missing data as possible and conduct another meta-analysis, which has just been published online ahead of print in Circulation. These data show a 38% relative increase in risk of death in paclitaxel-treated patients relative to controls, and an absolute risk increase of 4.6% over 5 years. When recovered supplemental vital status data were included, the relative increase in death further declined to 27% at 5 years.

Overall, mortality was 18.3% in the paclitaxel group and 13.7% in the control group (HR 1.38; 95% CI 1.06-1.80). Both the unadjusted and adjusted results of as-treated analyses were similar at HR 1.36 (95% CI 1.04-1.78) and HR 1.37 (95% CI 1.04-1.80), respectively.

A “tipping point analysis” revealed the sensitivity of the results according to the potential pattern of vital status among subjects with missing data. Further efforts to include vital status through direct queries to clinical sites and searches of the National Death Index brought the loss to follow-up down to 9.8% for controls (from 24%) and to 9.3% for paclitaxel patients (from 23%).

Commenting on the study for TCTMD, Eric Secemsky, MD (Beth Israel Deaconess Medical Center, Boston, MA), said the it shows how important those missing patients were to the overall discussion.

“I think that this is the clearest picture we're probably going to get in terms of the most up-to-date mortality data. That said, there's still about 10% of follow-up data missing, and you have to wonder what adding that 10% back in would do to further decrease that signal of harm,” he said. “There are some limitations to going back and reopening everything and trying to track everyone down. It would be nice if we had a more systematic way of doing that, but I think the effort was there in order to go from such a high [percentage of] loss to follow-up [down] to under 10% percent. Obviously in a perfect world, we want to see zero loss to follow-up and know exactly where everybody stands at 4 or 5 years.”

No Dose-Response Relationship

For the VIVA analysis, led by Krishna J. Rocha-Singh, MD (Prairie Heart Institute, Springfield, IL), de-identified individual data (baseline characteristics and mortality) were obtained on 2,185 patients (n = 1,382 paclitaxel; n = 803 controls). Only trials with at least 2 years of follow-up were included: ILLUMENATE, ILLUMENATE EU, IN.PACT SFA I/II, IN.PACT SFA Japan, LEVANT I, LEVANT II, LUTONIX Japan, and ZILVER PTX. The shortest follow-up was for LEVANT I at 2.0 years, and the longest were IN.PACT SFA I/II, LEVANT II, and ZILVER PTX at 4.9 years each.

I don't think we expected the signal to disappear in this meta-analysis, but I think it’s surprising just how much it attenuated. Eric Secemsky

In addition to the mortality signal, the analysis addresses the paclitaxel dose-mortality relationship, which in the Katsanos analysis showed mortality and amputations correlated with higher paclitaxel doses. Rather than analyzing summary data of dosing, Rocha-Singh and colleagues obtained index procedure nominal drug load data from each study sponsor and standardized the distribution. They also accounted for assessments of exposure categories based on tertiles of exposure dose. By this method, they found no increased mortality with paclitaxel versus placebo regardless of whether patients received high, medium, or low doses.

“Compared to non-paclitaxel-exposed subjects, the HR was 1.30 (95% CI, 0.92 to 1.82) for the low exposure group, 1.23 (95% CI, 0.87 to 1.73) for the medium exposure group, and 1.41 (95% CI, 0.96 to 2.07) for the high exposure group, showing that mortality risk did not increase with exposure,” the study authors write. They also were unable to identify a specific cause of death unique to the paclitaxel group.

“This meta-analysis gives us one more piece of data saying that we really cannot find one of the main pathways through which we would think paclitaxel could cause harm,” Secemsky said. “We can't find that in the dose relationship, and we also still don't understand the signal completely when it continues to attenuate with more follow-up data and a more complete patient-level meta-analysis.”

Last year, Secemsky published two large analyses of Medicare patients that also showed no increase in mortality in patients treated with paclitaxel balloons or stents compared with non-paclitaxel devices. Those were followed by the German BARMER Health Insurance study showing no increased mortality in patients receiving paclitaxel devices out to 11 years.

Nevertheless, Rocha-Singh and colleagues acknowledge that while the signal has weakened, “the issue of risk versus benefit must be considered given the increased mortality we are reporting.” For physicians like Secemsky, the need to have these conversations with patients continues.

“I try to be incredibly direct and transparent and true to the data with patients. I don't think we expected the signal to disappear in this meta-analysis, but I think it’s surprising just how much it attenuated,” he said. Taken together, the attenuated mortality signal and the lack of dose-response relationship “gives us support to continue to discuss these therapies, continue to support them remaining on the shelf, and continue to explore a better understanding of why we're seeing this relationship.”