Paclitaxel-Coated or Uncoated Devices: Significant Differences in Patient Populations and Mortality Led to Study Incomparability

The SWEDEPAD trial reported an unplanned interim analysis to show no difference in the mortality rate between the paclitaxel-coated and uncoated groups (Nordanstig et al., 2020), which contradicts the long-term risk of paclitaxel-coated devices claimed by a meta-analysis (Katsanos et al., 2018). However, there existed significant differences in mortality rates between the SWEDEPAD trial and the trials included in the meta-analysis, which were caused by significant differences in the patient populations. As a result, the SWEDEPAD trial and meta-analysis results are not directly comparable. An updated meta-analysis including the SWEDPEPAD trial and all studies in the meta-analysis (Katsanos et al., 2018) shows marginal differences in mortality rates between the paclitaxel-coated and control groups at two years with Bayesian relative risk (RR) 1.39 (95% credible interval (CrI) [1.01, 2.39]) and frequentist RR 1.16 (95% confidence interval (CI) [0.99, 1.36]) and differences in mortality rates during the entire follow-up period with Bayesian RR 1.29 (95% CrI [1.01, 1.72]) and frequentist RR 1.13 (95% CI [0.99, 1.28]) under random-effects models. Given the relatively short follow-up thus far in the SWEDEPAD trial (with a mean follow-up of 2.49 years) and the paclitaxel-coated risk being long-term (e.g., 4 or 5 years), the interim results on the risk of paclitaxel-coated devices reported by the SWEDEPAD trial warrant further investigation.

The SWEDEPAD trial 1 , which are investigating paclitaxel-coated versus uncoated 2 devices in patients with peripheral artery disease (PAD), was temporarily suspended on 3 December 2018 due to the potential long-term risk of paclitaxel-coated devices reported 4 by a meta-analysis. 2 An unplanned interim analysis of the SWEDEPAD trial showed no 5 difference in the mortality rate between the paclitaxel-coated and uncoated groups at one 6 year with a hazard ratio (HR) of 1.03 (95% CI [0.77, 1.37]) or during the entire follow-up 7 period with an HR of 1.06 (95% CI [0.92,1.22]). Based on these results, it was decided 8 to resume enrollment in the SWEDEPAD trial on March 23, 2020. However, Figure 2 of 9 the SWEDEPAD trial publication shows that the cumulative incidence curves of drug-10 coated devices constantly stayed above those of uncoated either for the overall population 11 or subgroups of chronic limb threatening ischemia or intermittent claudication. 1 The risk 12 of paclitaxel-coated devices is known to be of long term, 2,3 while the SWEDEPAD trial 13 has so far had a mean follow-up of 2.49 years which is still relatively short for evaluating 14 long-term risk. 15 16 2

18
By pooling the numbers of deaths and patients from all studies included in the meta-19 analysis, we treat the combined data as a mega-trial, 2 and compare the differences in 20 mortality between the SWEDEPAD trial 1 and mega-trial. For the paclitaxel-coated group, 21 we found significant differences in the all-cause death rates at one year with relative risk 22 (RR) 4.4 (95% CI [3.2, 6.0]) and during the entire study period with RR 3.7 (95% CI [3.1, 23 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted February 1, 2021. ; https://doi.org/10.1101/2021.01.29.21250732 doi: medRxiv preprint 4.5]). For the uncoated group (control), the mortality rate differences between the 1 SWEDEPAD trial 1 and mega-trial were also significant with RR 4.2 (95% CI [3.0, 5.9]) 2 at one year, and RR 5.7 (95% CI [4.5, 7.2]) during the entire study period. All RRs 3 comparing the SWEDEPAD trial 1 and mega-trial 2 are larger than 3.5 with p-values 4 smaller than 0.0001, which could have attributed to incomparability between the 5 SWEDEPAD trial 1 and the trials included in the meta-analysis 2 . There are significant 6 differences at the baseline in the patient populations between the SWEDEPAD trial and 7 trials in meta-analysis: the former had 35% intermittent claudication (IC) and 65% 8 chronic limb threatening ischemia (CLTI) while the latter had 89% IC and 11% CLTI. 9 CLTI is an advanced stage of PAD. Compared with intermittent claudication, CLTI has a 10 negative prognosis within a year after the initial diagnosis, with a 1-year amputation rate 11 of approximately 12% and mortality rates of 50% at 5 years and 70% at 10 years. 4 The 12 baseline prognosis difference could be a main cause of observed differences in mortality 13 rates between the studies. 14 15 2.2 Updated Meta-Analysis 16 We conducted an updated meta-analysis which included the SWEDEPAD trial 1 and all 17 studies in the meta-analysis 2 . The number of deaths by two years of the SWEDEPAD 18 trial 1 was estimated by multiplying the estimated two-year cumulative incidence rate and 19 the total number of patients. For each study in the meta-analysis 2 , we took the results 20 with the longest follow-up as those for the entire study period. The R packages 'meta' 21 and 'bayesmeta' were used to formulate frequentist and Bayesian random-effects models, 22 respectively. 23 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted February 1, 2021. ; https://doi.org/10.1101/2021.01.29.21250732 doi: medRxiv preprint 1 As shown in Table 1, both frequentist and Bayesian random-effects models demonstrate 2 no difference in one-year all-cause death rates between paclitaxel-coated and uncoated 3 arms, indicating no short-term risk of paclitaxel-coated devices. At two years or during 4 the entire follow-up period, the frequentist random-effects model yields marginally 5 insignificant differences in all-cause death rates with two-year RR 1. 16  the entire follow-up period, which suggests significant risk of paclitaxel-coated devices 10 with longer-term follow-ups. The posterior probabilities for the death rate of the 11 paclitaxel-coated arm being higher than that of the control arm were around 0.98 at two 12 years and during the entire follow-ups, with Bayes factors larger than 40. This provides 13 strong evidence for the long-term mortality risk of paclitaxel-coated devices. 14 15

16
The SWEDEPAD trial 1 enrolled 2289 patients, which is about half of the total number of 17 patients included in the meta-analysis 2 . Although one study with such a large sample size 18 and insignificant results was added, the new meta-analysis still yields marginally 19 significant differences in the mortality rates at two years and during the entire follow-up. 20 Moreover, with a mean follow-up of only 2.49 years, the results of the SWEDEPAD trial 21 warrant further investigation, because the risk from paclitaxel-coated devices is long-22 term. 5 23 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted February 1, 2021. ; https://doi.org/10.1101/2021.01.29.21250732 doi: medRxiv preprint . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted February 1, 2021. ; https://doi.org/10. 1101/2021