Abstract
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.
1 Introduction
The SWEDEPAD trial1, which are investigating paclitaxel-coated versus uncoated devices in patients with peripheral artery disease (PAD), was temporarily suspended on December 2018 due to the potential long-term risk of paclitaxel-coated devices reported by a meta-analysis.2 An unplanned interim analysis of the SWEDEPAD trial showed no difference in the mortality rate between the paclitaxel-coated and uncoated groups at one year with a hazard ratio (HR) of 1.03 (95% CI [0.77, 1.37]) or during the entire follow-up period with an HR of 1.06 (95% CI [0.92, 1.22]). Based on these results, it was decided to resume enrollment in the SWEDEPAD trial on March 23, 2020. However, Figure 2 of the SWEDEPAD trial publication shows that the cumulative incidence curves of drug-coated devices constantly stayed above those of uncoated either for the overall population or subgroups of chronic limb threatening ischemia or intermittent claudication.1 The risk of paclitaxel-coated devices is known to be of long term,2,3 while the SWEDEPAD trial has so far had a mean follow-up of 2.49 years which is still relatively short for evaluating long-term risk.
2 Methods and Results
2.1 Baseline Differences in Patient Populations
By pooling the numbers of deaths and patients from all studies included in the meta-analysis, we treat the combined data as a mega-trial,2 and compare the differences in mortality between the SWEDEPAD trial1 and mega-trial. For the paclitaxel-coated group, we found significant differences in the all-cause death rates at one year with relative risk (RR) 4.4 (95% CI [3.2, 6.0]) and during the entire study period with RR 3.7 (95% CI [3.1, 4.5]). For the uncoated group (control), the mortality rate differences between the SWEDEPAD trial1 and mega-trial were also significant with RR 4.2 (95% CI [3.0, 5.9]) at one year, and RR 5.7 (95% CI [4.5, 7.2]) during the entire study period. All RRs comparing the SWEDEPAD trial1 and mega-trial2 are larger than 3.5 with p-values smaller than 0.0001, which could have attributed to incomparability between the SWEDEPAD trial1 and the trials included in the meta-analysis2. There are significant differences at the baseline in the patient populations between the SWEDEPAD trial and trials in meta-analysis: the former had 35% intermittent claudication (IC) and 65% chronic limb threatening ischemia (CLTI) while the latter had 89% IC and 11% CLTI. CLTI is an advanced stage of PAD. Compared with intermittent claudication, CLTI has a negative prognosis within a year after the initial diagnosis, with a 1-year amputation rate of approximately 12% and mortality rates of 50% at 5 years and 70% at 10 years.4 The baseline prognosis difference could be a main cause of observed differences in mortality rates between the studies.
2.2 Updated Meta-Analysis
We conducted an updated meta-analysis which included the SWEDEPAD trial1 and all studies in the meta-analysis2. The number of deaths by two years of the SWEDEPAD trial1 was estimated by multiplying the estimated two-year cumulative incidence rate and the total number of patients. For each study in the meta-analysis2, we took the results with the longest follow-up as those for the entire study period. The R packages ‘meta’ and ‘bayesmeta’ were used to formulate frequentist and Bayesian random-effects models, respectively.
As shown in Table 1, both frequentist and Bayesian random-effects models demonstrate no difference in one-year all-cause death rates between paclitaxel-coated and uncoated arms, indicating no short-term risk of paclitaxel-coated devices. At two years or during the entire follow-up period, the frequentist random-effects model yields marginally insignificant differences in all-cause death rates with two-year RR 1.16 (95% CI [0.99, 1.36], P=0.063), and the entire follow-up RR 1.13 (95% CI [0.99, 1.28], P=0.065). However, the Bayesian random-effects model yields an RR of 1.39 (95% credible interval (CrI) [1.01, 2.39]) at two years and an RR of 1.29 (95% CrI [1.01, 1.72]) during the entire follow-up period, which suggests significant risk of paclitaxel-coated devices with longer-term follow-ups. The posterior probabilities for the death rate of the paclitaxel-coated arm being higher than that of the control arm were around 0.98 at two years and during the entire follow-ups, with Bayes factors larger than 40. This provides strong evidence for the long-term mortality risk of paclitaxel-coated devices.
3. Conclusion
The SWEDEPAD trial1 enrolled 2289 patients, which is about half of the total number of patients included in the meta-analysis2. Although one study with such a large sample size and insignificant results was added, the new meta-analysis still yields marginally significant differences in the mortality rates at two years and during the entire follow-up. Moreover, with a mean follow-up of only 2.49 years, the results of the SWEDEPAD trial warrant further investigation, because the risk from paclitaxel-coated devices is long-term.5
Data Availability
The data that support the findings of this study are available from the corresponding author, G Yin, upon reasonable request.
Footnotes
Conflicts of interest: The authors declare no potential conflict of interest.