The current study, comparing two historical cohorts of patients with claudication and de novo femoropopliteal artery lesions treated with and without paclitaxel coated devices (PCD), failed to show any excess mortality with the use of PCD. On the contrary, there was a statistically significant excess mortality at one year of follow-up in the non PCD group. This excess mortality was not maintained over time at 2 and 5 years of follow-up. Those results are consistent with two recent studies that also failed to show any significant increase in mortality when using paclitaxel coated devices(5,6).
Despite its retrospective design, this work has several strengths. The availability of a national mortality registry limited the potential lost to follow-up bias. This allowed to draw reliable conclusions regarding the mortality of the patients included in this work. Furthermore, it was chosen to only include the patients who presented for their very first endovascular treatment of femoropopliteal artery related disabling claudication. It is important to note that the patients in the POBA group had significantly higher cardiovascular risk factors such as tobacco use, hypertension, dyslipidemia and diabetes. Nevertheless, after adjustment for potential confounders and accounting for cross-over, PCD didn’t show any significant association with all-cause mortality.
On the other hand, patients in the PCD group had significantly longer lesions and more occlusions. These differences are very likely explained by the fact that our institutional technology evaluation unit recommended the use of PCD for TASC C and D femoropopliteal lesions but the choice between POBA and PCD was left at the operator’s discretion for TASC A and B lesions (4).
The question of how a supposedly local treatment may be responsible for an excess mortality 2 years and 5 years later is currently a conundrum. Several factors influence the fraction of drug released into the arterial wall and the fraction released in the systemic circulation, such as the drug carrier and the arterial lesion complexity(8,9). Though there is an initial burst of serum paclitaxel following DCB or some of the DES delivery, this is a transient phenomenon with the drug being cleared during the following days(10,11). The rest of paclitaxel is receptor bound, and there is no evidence for a paclitaxel reservoir in the organism(12). A dose-effect relationship had been initially suspected but never confirmed in accordance with the current study. Performance and detection biases had been suspected since interventional studies are difficult to blind(13). One of the hypothesis was that the patients in the control arm, with more restenosis, might have had more visits with study investigators at which secondary prevention medical therapies could have been properly adjusted. Another hypothesis could be that the patients treated with PCD having a better ambulatory capability could have been more active, raising the risk of stress-induced myocardial infarction. Nevertheless, as of today, no study has demonstrated a significant increase in cardiovascular events in the paclitaxel group. It is also important to realize that the mechanism of action, if mechanism there is, could impact the way we should analyze the mortality. If the phenomenon precipitates some already preexisting conditions that were about to decompensate in the near future, then this premature death will be noticed only during a certain period of time. This mortality displacement or “harvesting effect” implies that after some periods with excess mortality, there is a decrease in overall mortality during a subsequent period of time(14). On the other hand if this is a toxic phenomenon, occurring randomly in patients or without regard to the underlying conditions, then the excess mortality should be maintained over a longer period of time. The current study refutes both hypotheses with on the contrary an excess mortality in the POBA group at 1 year of follow-up and almost no difference at 2 and 5 years of follow-up.
Patients with claudication are expected to benefit the most from drug coated devices. The longer patency rates associated with PCD are synonym with a better ambulatory capability and better quality of life. Both Katsanos et al and Rocha-Singh et al studies included mostly patients with claudication but also 11.4% and 6.1% of patients with CLI respectively. It is important to note that patients with critical ischemia have a higher mortality rate than patients with claudication. When combined together, the mortality rate of patients with claudication and critical ischemia does not follow a normal distribution, but a bimodal distribution since mortality has been reported to be up to 36% at two years in CLI (7) while ranging between 4 and 10% in patients with claudication (2,5). Combining both groups together may have an important impact on mortality studies. The probability of death has a binomial distribution (dead or alive). The standard deviation of a binomial distribution is estimated to be with p: probability to be dead at two years, q: probability to be alive at two years and n: number of patients. Following this definition, the closer the probability of an event is to 50%, the larger is its standard deviation and its confidence interval. If it is assumed that the respective mortality rates of patients with claudication and CLI are 4% and 36%, using a similar number of patients as in the meta-analysis by Katsanos et al (4133 patients with claudication and 530 with CLI), the standard deviation of the mortality rates at two years is 0.3% in patients with claudication and 2.08% in patients with CLI. The respective confidence intervals are then [3.41 - 4.59] (4±1.96x0.3) and [31.9 - 40.1] (36±1.96x2.08). Among the 4133 patients with claudication, at two years the 95% confidence interval of deaths ranges between 141 and 190 patients while it could vary between 169 and 213 for the 530 patients with CLI. This example shows how an apparently small proportion of patients with CLI may have a huge impact on mortality rates. On a same note, to show an excess mortality rate of 3.4%, as reported by Katsanos et al at 2 years, 1618 patients are necessary (809 per arm) when using a mortality risk of 4%, a risk alpha of 5% and a risk beta of 20%. This number becomes 7192 (3596 per arm) when using a mortality risk of 36%. Consequently, both groups of patients with claudication and CLI should rather be studied separately as in the study by Nordanstig et al. It could be relevant for both the VIVA group and Katsanos et al to report the respective mortality rates of patients with de novo claudication, recurrent disease and critical limb ischemia to better appreciate the impact of each group on their mortality analyses.
Given our large use of this very promising technology with excellent patency rates, the study by Katsanos et al has had a profound impact upon our practice. Although claudication may be a severely debilitating disease with dire consequences for the quality of life of patients, the possibility of increasing the risk of death dramatically lowered its benefit to risk ratio. On the other hand, given the lack of physiopathological mechanism, and knowing that the meta-analysis combined studies that were not powered to study mortality, it was very difficult to offer a second class treatment with all its limitations instead of drug eluting devices that would provide improved patency rates, lower reinterventions and better quality of life. In order to help us with this ethically demanding task, the FDA recently asked the Multi-Specialty and Multi-Society Coalition for Patient Safety with Paclitaxel Technologies to develop a set of talking points concerning the risks and benefits of using paclitaxel devices(15). Among others, the committee indicated that in individual patients judged to be at particularly high risk for restenosis and repeat femoropopliteal interventions, clinicians may determine that the benefits of using a paclitaxel-coated device may outweigh the risk of late mortality. The results of the current study, that failed to show any increase in mortality in patients with de novo claudication treated with paclitaxel coated devices, support the need for additional prospective randomized studies properly powered to study mortality.