The methodology underpinning the conclusion by Konstantinos Katsanos (Patras, Greece) et al that there is a positive dose-response relationship between paclitaxel and mortality is flawed, argues Andrew Holden (Auckland, New Zealand) and colleagues in a paper published in the Journal of Endovascular Therapy (JEVT). Holden and co-authors list the limitations they see in the summary level meta-analysis published in December 2018 in the Journal of the American Heart Association (JAHA), and call for physicians to “significantly modify the way vascular device trials are performed in the future”.
In the JAHA meta-analysis, Katsanos et al linked paclitaxel-coated balloons and paclitaxel-eluting stents to all-cause mortality, reporting a significant increase in deaths at two and five years in the cohort of femoropopliteal disease patients exposed to the drug compared to those treated who were not. Holden et al write: “Although no plausible mechanism of action was identified, the authors postulated that the excess mortality was caused by paclitaxel”.
However, Holden and colleagues do not find this correlation with paclitaxel levels and mortality, and thus claim there is no causative link between dose and death.
On 15 March 2019, the US Food and Drug Administration (FDA) released an updated letter to healthcare providers detailing that while the results of their preliminary review of the JAHA meta-analysis data was ongoing, they had “identified a potentially concerning signal of increased long-term mortality in study subjects treated with paclitaxel-coated products compared to patients treated with uncoated devices”. The statement also noted that each of the trials used “all-cause” mortality calculation methods that differed significantly and had the potential to influence the way those results were interpreted.
The three pivotal premarket trials evaluating paclitaxel-coated products for the treatment of peripheral artery occlusive disease that informed this statement—Medtronic IN.PACT SFA I and II; Cook Zilver PTX; and BD LEVANT 2 randomised controlled trials—each used a different methodology to calculate mortality. “To facilitate a direct mortality comparison between the three trials”, Holden and colleagues write that they “asked each of the companies to provide five-year mortality using the same proportional methods as used by Medtronic [calculated as the number of patients who died divided by the number of patients followed up to five years]”.
Explaining the rationale behind this strategy, Holden et al say: “This consistent methodology allowed for a direct comparison of product dose to test the dose-response theory proposed by the meta-analysis. The highest five-year mortality was found to be associated with the lowest paclitaxel dose device (Zilver PTX), while the lowest mortality was associated with the highest dose device (IN.PACT). It is our view that this directly refutes the causal relationship theory between paclitaxel dose and all-cause mortality.”
Using this proportional method, the investigators report no significant difference in the five-year mortality between paclitaxel-coated balloons compared to controls in the Medtronic IN.PACT and BD LEVANT 2 trials. The Zilver PTX DES displayed a significant difference (p=0.008) between the arms. When the Zilver PTX randomised controlled trial five-year data are reanalysed by an intention-to-treat analysis, the mortality of the paclitaxel-eluting stent remains the highest (22.7%) of the three paclitaxel devices, but the mortality difference compared to the control arm (18.5%) loses its significance (p=0.45).
As patients previously lost to follow-up are contacted and potentially added into the analysis, the proportional calculation may differ (as the denominator, number of patients follow-up to five years, will increase), though the study authors do not think this will change the outcome. Commenting on these findings, Holden and colleagues say: “We believe it is likely, even with more complete patient follow-up, that the lack of correlation between paclitaxel dose and mortality will persist”. They suggest a Kaplan Meier survival plots could be calculated in the future, if more data were collated.
As Holden et al believe the causation theory behind paclitaxel and mortality “now appears unlikely”, they suggest several alternative hypotheses to explain the original JAHA meta-analysis results, including performance and detection bias, trials being inadequately powered for mortality, significant target lesion crossover from the control to treatment arms, and patients lost to follow-up.
They conclude: “Perhaps the most important lesson from this paclitaxel debate may be the need to reconsider and significantly modify the way vascular device trials are performed in the future. We believe there should be an additional emphasis placed on adherence to guidelines-based medical therapy of all enrolled patients as well as safety endpoints such as mortality. Greater effort must be made to reduce subjects lost to follow-up and properly classify the mortality status of those subjects who withdraw.
“It is only with well-conducted safety evaluations that we will completely clarify this issue regarding paclitaxel. Given the known and undisputed benefits of paclitaxel coating on stents and angioplasty balloons, we owe it to our patients to pursue that truth before we restrict these devices from use.”
The medical community awaits an individual patient data meta-analysis to provide greater clarity regarding the alleged link between paclitaxel and mortality.