November 11, 2019
3 min read

Ascertainment bias could explain paclitaxel mortality signal

You've successfully added to your alerts. You will receive an email when new content is published.

Click Here to Manage Email Alerts

We were unable to process your request. Please try again later. If you continue to have this issue please contact

LAS VEGAS — It is possible that ascertainment bias and other biases from randomized controlled trials could explain the long-term mortality signal observed in paclitaxel-coated devices, according to a speaker at VIVA 19.

During a session on the paclitaxel controversy, which was launched in December 2018 after publication of a summary-level meta-analysis showing elevated long-term mortality risk in patients assigned paclitaxel-coated devices compared with controls, Ramon L. Varcoe, MBBS, MS, FRACS, PhD, director of operating theatres and director of the Vascular Institute at Prince of Wales Hospital, Sydney, and associate professor at the University of New South Wales, said it is possible the signal can be explained by faulty trial designs and follow-up as opposed to the toxicity of the devices.

“It isn’t randomization that prevents bias in outcome assessment, it’s blinding,” he said, noting all 28 studies from the summary-level meta-analysis had some level of lack of blinding.

Performance and detection bias results in “an unconscious or intentional detection of an enhanced treatment effect,” Varcoe said. “But perhaps more relevant in this particular scenario is ascertainment bias. This is when you see a more intense surveillance program for those in the experimental arm of the randomized controlled trial, in this case the drug-coated arm. It is particularly relevant for mortality because it will often lead participants to be uncontactable and lost to follow-up.”

This relates to the tenacity of follow-up, he said. “It’s about the unblinded research team being more tenacious in how they follow up the patients in the experimental arm,” he said. “That’s human nature. We’ve got an experimental new treatment and we want to know if it’s having a good or bad impact on our patients. This may lead to fewer subjects being labeled as missing, with an increased number of mortalities recorded in those experimental arms.”

Missing data ranged from 14% to 46% in the FDA’s as-treated analysis, he noted.

To help determine whether bias played a role in the mortality signal, Varcoe and colleagues performed a meta-analysis of 22 trials of endovascular interventions in the superficial femoral artery that did not involve drug-coated devices, covering 2,451 patients.

The researchers found that the experimental arm had an elevated risk for mortality compared with controls at 1 year (RR = 1.69; 95% CI, 1.11-2.57). The 2-year mortality data, which included 468 patients, also favored the control group but were underpowered (RR = 1.5; 95% CI, 0.48-4.71). The 3-year data, covering 613 patients, showed the same effect but were also underpowered (RR = 1.27; 95% CI, 0.7-2.3).


“Those effects were independent of paclitaxel completely, which in my view casts considerable doubt to the causal link between paclitaxel and mortality,” Varcoe said.

Missing data are very important, Varcoe said, because if 10% of the missing patients from the control arm of the December 2018 meta-analysis had died but were misclassified as missing, the 5-year mortality data are no longer statistically significant.

Varcoe said he knew of no biological mechanisms that would explain the long-term mortality risk associated with paclitaxel. The theories in that area “fail to explain the absence of any one dominant death type,” he said. “You would expect that if there truly were toxicity. It also doesn’t explain why paclitaxel use in high doses in oncology for 27 years actually reduces long-term mortality rather than the opposite, why there’s no dose-response relationship or biological gradient and why, as more and more new data emerge, with subjects lost to follow-up being found, the effect size is being reduced.”

Taken as a whole, the data suggest that it is possible that SFA interventions in general increase mortality “or, more likely, could it be that the association [is] due to the introduction of bias, a more tenacious follow-up in the subjects in the experimental arms and/or increased medical interactions with the health care team and increased medical therapy regimens in the [trial] arms more prone to increased target lesion revascularizations?” Varcoe said. – by Erik Swain


Varcoe R. Is this the end of drug-eluting devices or the beginning? Presented at: VIVA 19; Nov. 4-7, 2019; Las Vegas.

Disclosure: Varcoe reports he consulted for and/or received honoraria from Abbott Vascular, Boston Scientific, Intact Vascular, Medtronic, Shockwave Medical and Surmodics.