Abstract: Objectives. Paclitaxel drug-eluting stents (DESs) have been shown to improve primary patency of femoropopliteal lesions compared to plain balloon angioplasty with provisional bare-metal stents (BMSs) in randomized controlled studies. However, data are lacking on patency outcomes of real-world DES use relative to BMS use. This study compared clinically driven target-lesion revascularization (TLR), target-vessel revascularization (TVR), and target-limb revascularization outcomes at 1 year between DES and BMS treatments in a real-world setting. Methods. The study identified 174 DES (Zilver PTX; Cook Medical) and 784 BMS femoropopliteal interventions from the available 969 Excellence in Peripheral Artery Disease (XLPAD) registry patients between October 2013 and December 2016. We analyzed both unmatched (174 DES and 784 BMS) and propensity score (PS)-matched datasets (174 for each). Results. This study found that patients who underwent DES femoropopliteal endovascular revascularization had significantly lower TLR rates in both unmatched (hazard ratio [HR], 0.57; 95% confidence interval [CI], 0.35-0.91; P=.02) and matched data (HR, 0.50 95% CI, 0.27-0.91; P=.02). The DES group had a 43% lower TVR risk than the BMS group in the PS matched cohort (HR, 0.57; 95% CI, 0.33-0.98; P=.04). Mortality rate in the DES group (5%) was significantly higher than the BMS group in both unmatched (2%; P=.04) and matched groups (1%; P=.046) at 1 year. Conclusions. Patients treated with DES had higher lesion and vessel patency than BMS after adjusting for confounding, which included complexity of lesion characteristics and operators’ clinical decision-making regarding selection of treatment modalities, in femoropopliteal endovascular interventions in a real-world registry.
REPRINTED WITH PERMISSION FROM: J INVASIVE CARDIOL 2018;30(10):373-379.
Key words: endovascular revascularization, drug-eluting stent, peripheral artery disease, propensity score matching, stent, Zilver PTX
Endovascular intervention is a common treatment for patients presenting with femoropopliteal peripheral artery disease (PAD), and it is associated with a lower mortality, shorter recovery time, and non-inferior procedural success rate compared to surgical alternatives.1 However, endovascular intervention in the femoropopliteal artery has been plagued by lower patency rates secondary to in-stent restenosis (ISR). Zilver PTX stent (Cook Medical) is the only approved drug-eluting stent (DES) indicated for treatment of the femoropopliteal artery. Zilver PTX is a nitinol, self-expanding stent coated with paclitaxel, an antiproliferative agent that limits neointimal smooth muscle cell proliferation,2 and is associated with a lower rate of ISR.3 Data from the Zilver PTX Randomized Controlled Trial of Paclitaxel-Eluting Stents for Femoropopliteal Disease demonstrated an improved 5-year primary patency rate in superficial femoral artery (SFA) lesions treated with Zilver PTX compared to balloon angioplasty with provisional bare-metal stent (BMS) placement (66.4% vs 43.4%).4 The Zilver PTX Randomized Trial was a United States Food and Drug Administration (FDA)-approval study for the treatment of intermediate-complexity lesions. In real-world practice, the Zilver PTX stent is used for treatment of more complex lesions and patients with greater burden of comorbidities. However, adjudicated outcomes of Zilver PTX stent applications are lacking.
We conducted a comparative effectiveness analysis of Zilver PTX and contemporary BMS for femoropopliteal artery endovascular intervention using the multicenter Excellence in Peripheral Artery Disease (XLPAD) registry (NCT01904851) data. Details of the XLPAD registry data have been previously published.5
This study compared clinically driven 1-year target-lesion revascularization (TLR), target-vessel revascularization (TVR), and target-limb revascularization rates between Zilver PTX stent and BMS treatments in femoropopliteal distribution. We also compared 1-year all-cause mortality in the respective treatment groups.
Propensity-score matching method was used to match patient and lesion characteristics, and sample sizes between Zilver PTX and BMS groups. Core-laboratory adjudicated lesion characteristics included lesion length, vessel diameter, lesion calcification, ISR, chronic total occlusion, and diffuse (≥20 cm in length) disease. In addition, duration of procedure as reported by the operating physician was included. The propensity-score matching method has been utilized and published previously.6
The study extracted data from the available 969 XLPAD registry patients who underwent femoropopliteal stent interventions (unmatched data) between October 2013 and December 2016 at eleven XLPAD participating sites in the United States.
One-year all-cause mortality and three revascularization outcomes – clinically driven TLR, TVR, and target-limb revascularization – were included in the analysis. Target-limb revascularization was defined as the need for revascularization to the index limb, TVR as revascularization to the index femoropopliteal vessel, and similarly TLR as revascularization to the index Zilver PTX- or BMS-treated lesion as determined by the VA North Texas Angiography and Ultrasound Core Laboratory analysis of the index procedure and 1-year follow-up angiograms. Clinical indication for TLR was reported by the operators. Location of ISR (within stent or involving the stented segment, within 5 mm from the proximal or distal stent edge)7 was also examined.
Propensity-score matching method was performed to address the selection bias of this observational dataset.8 The propensity-score matching method with a 1:1 ratio of Zilver PTX and BMS was used to create the propensity-score matched cohorts.3
The propensity scores were created from a logistic model predicting the use of Zilver PTX stent relative to BMS. The final logistic model included clinical and lesion characteristics at baseline. Clinical characteristics included in the final model were participating institutions, diabetes mellitus (DM), chronic kidney disorder (CKD), previous congestive heart failure (CHF), coronary artery disease (CAD), hyperlipidemia, hypertension, body mass index, and Rutherford categories. Lesion characteristics were lesion length, reference vessel diameters, presence of heavy calcification, diffuse disease, chronic total occlusion (CTO), and ISR at baseline. This study identified BMS interventions matched to Zilver PTX interventions using a closest distance between Zilver PTX and BMS groups in calculated propensity scores (SAS %PSmatch_multi).9
This study conducted an extensive sensitivity analysis in finalizing the model specification for predicting Zilver PTX use relative to BMS. A sensitivity analysis examined the “goodness of fit” of the model based on missing data on covariates, calculability of propensity scores, and Akaike Information Criterion (AIC). In addition, the standard differences of <15% in baseline covariates between the two groups were used to select an optimal propensity score matching algorithm. The standard differences for continuous covariates were calculated as:
and for categorical covariates were calculated as differences in proportions for each level of the measured covariate divided by the standard deviation in the BMS group. Standard deviation
where p = proportion with N = the entire sample size. The current study model discriminated Zilver PTX femoropopliteal procedures from BMS femoropopliteal procedures with 83% accuracy according to receiver operating characteristic (ROC) curve analysis.
The study compared group differences in demographic, clinical, and lesion characteristics at baseline and procedural characteristics using both unmatched and propensity-score matched data. The demographic variables included gender, age, and race groups. Clinical characteristics included baseline co-morbid conditions, lesion characteristics, and laboratory variables at baseline. The procedural characteristics included duration of the procedure and number of lesions treated.
Cochran-Mantel-Haenszel (CMH) statistical analysis was conducted to examine group differences for categorical variables and two-sided Wilcoxon rank-sum statistics for continuous variables between Zilver PTX stent and BMS groups.
To address a competing risk between mortality and repeat revascularization events, the study conducted a time-to-event analysis using cumulative incidence function (CIF). The CIF analyses were executed using both unmatched and propensity-score matched data. The CIF calculated the marginal probability for each TLR, TVR, and target-limb revascularization event, regardless of whether they were censored or failed from death event occurrence in the previous months. The Gray test10 was conducted to examine group differences in 1-year outcomes, and CIF hazard ratio (HR) and 95% confidence interval (CI) were reported using the Fine and Gray subdistribution hazard functions for TLR, TVR, and target-limb revascularization. The subdistribution hazard functions estimate hazard rates for TLR, TVR, and target-limb revascularization at each month (t) based on the risk set that remains at the month t after accounting for all previously occurring death events.11 Also, the CIF analysis was adjusted for a procedural duration and adjusted CIF HRs and 95% CIs were presented.
SAS 9.4 (SAS Institute) was used for all data analyses and P<.05 was set for a statistical significance.
The unmatched data included 174 femoropopliteal interventions that employed the Zilver PTX stent and 784 that used BMS, and the propensity-score matched data included 174 Zilver PTXs and 174 BMSs. The BMS group included SMART (n = 378; 48%), Absolute (n = 123; 16%), Lifestent (n = 100; 13%), Everflex (n = 94; 12%), Zilver bare metal (n = 53; 6.8%), Epic (n = 46; 6%), Sentinel (n = 14; 1.8%), and Vision RX stents (n = 1; 0.13%). Supera stents (n = 14) were used adjunctively with other stents. The propensity-score matched data constituted a total of 311 unique patients with femoropopliteal disease treated with either Zilver PTX or BMS. The BMS types included SMART (42%), Everflex (21%), Lifestent (19%), Epic (9%), Absolute (8%), Zilver bare-metal (4%), and 1 use of Supera stent in adjunction with Zilver bare-metal stent.
In the unmatched cohort, the Zilver PTX stent was used less frequently in treating patients with diabetes mellitus compared with BMS. The Zilver PTX stent was used more frequently in treating ISR lesions, and less frequently in treating diffuse and heavily calcified lesions. The propensity-score matched data showed no significant group differences (Table 1).
For the unmatched cohort, the duration of the Zilver PTX stent procedure was significantly shorter than the BMS procedure duration (P<.01), but the statistical significance of the group difference in the procedure duration was attenuated in the propensity-score matched data (P=.09) (Table 1).
In the unmatched cohort, the Zilver PTX procedures did not significantly differ in TLR and TVR rates at 1 year from BMS procedures (TLR: 9.2% vs 13.8% [P=.10]; TVR: 9.8% vs 13.8% [P=.16]). In the propensity-score matched data, Zilver PTX procedures had significantly lower TLR and TVR rates compared to BMS procedures (TLR: 9.2% vs 18.4% [P=.01]; TVR: 9.8% vs 18.4% [P=.02]).
Time-to-event analyses were performed to address mortality and repeat revascularization competing risks using both unmatched and propensity-score matched data. Using unmatched data, the Zilver PTX group had a significantly lower risk of TLR than the BMS group (43% lower risk; CIF HR, 0.57; 95% CI, 0.35-0.91; P=.02). Although not statistically significant under P<.05 criterion, the Zilver PTX group had a 37% lower risk of TVR (CIF HR, 0.63; 95% CI, 0.39-1.01; P=.05) and a 32% lower risk of target-limb revascularization (CIF HR, 0.68; 95% CI, 0.44-1.07; P=.10) than the BMS group under unmatched data.
A subgroup analysis of unmatched data excluding ISR lesions (128 Zilver PTX, 698 BMS) showed a significantly lower TLR risk in the Zilver PTX group (CIF HR, 0.48; 95% CI, 0.26-0.89; P=.02) than the BMS group, while TVR risk was lower in the Zilver PTX group without meeting a statistical significance under P<.05 criterion (CIF HR, 0.58; 95% CI, 0.33-1.03; P=.06). There was no significant difference in target-limb revascularization between Zilver PTX and BMS groups using a subgroup dataset excluding ISR interventions (CIF HR, 0.66; 95% CIs, 0.38-1.13; P=.13).
Using propensity-score matched data, patients who underwent Zilver PTX femoropopliteal endovascular revascularization had a 50% lower risk of TLR (CIF HR, 0.50; 95% CI, 0.27-0.91; P=.02) (Figure 2), a 43% lower risk of TVR (CIF HR, 0.57; 95% CI, 0.33-0.987; P=.04) (Figure 3), and a non-statistically significant 41% lower risk of target-limb revascularization (CIF HR, 0.59; 95% CI, 0.35-1.01; P=.05) than patients who underwent BMS implantation.
Using the propensity-score matched data, the CIF analysis with adjustment for the duration of the procedure – a surrogate for the complexity of the endovascular intervention – confirmed a statistically significant lower risk of 1-year TLR with Zilver PTX stent placement compared to BMS (CIF HR, 0.53; 95% CI, 0.30-0.94; P=.03), but a lower 1-year TVR risk in the Zilver PTX stent (CIF HR, 0.60; 95% CI, 0.35-1.06) with an attenuated statistical significance (P=.08). The Zilver PTX group no longer had a significantly lower risk of target-limb revascularization than the BMS groups at 1 year (CIF HR, 0.73; 95% CI, 0.41-1.31; P= .29).
Meanwhile, the Zilver PTX group had a significantly higher all-cause mortality rate than the BMS group in both unmatched (5.3% vs 1.8%; P=.03) and propensity-score matched data (5.3% vs 1.3%; P=.04). Per sensitivity analysis of propensity-score matched BMS group to the Zilver PTX group, 1-year mortality rate in the BMS group ranged between 1.1% and 1.8% per model specification. The cumulative death incident rates were 5.3% and 1.1% for the Zilver PTX stent and BMS groups, respectively, in the propensity-score matched data (P=.047) (Figure 1) and 1.8% in the unmatched BMS group (P=.07).
The 1-year death events in the Zilver PTX group included 2 cardiovascular, 1 lung cancer, 1 renal failure, 1 accident, and 3 additional undetermined causes of death. The 1-year death events in the matched BMS group included 1 cardiovascular event and 1 unknown (Table 2).
This study provides real-world data on the 1-year patency of the Zilver PTX DES compared to BMS in femoropopliteal endovascular intervention. There was a significantly higher 1-year lesion patency rate using the Zilver PTX than BMS.
Considering that death and repeated revascularization are competing events, a higher death rate in the Zilver PTX group may be a possible reason for lower rates of 1-year TLR and TVR in the Zilver PTX group compared to the BMS group. The Zilver PTX group had a fewer number of patients that survived the 1-year follow up compared with the BMS group. To address this competing event issue, this study conducted a cumulative incidence function (CIF) method for a time-to event analysis. The lower rates of TLR and TVR in the Zilver PTX group remained significant in the CIF analysis.
This study adds to the existing comparative effectiveness literature of the Zilver PTX stent on 1-year patency using real world data. It provides insights to the current practice of PAD treatment of complex femoropopliteal disease. The registry data reflect the current and up-to-date clinical practice in employing newer devices in treating more complex and severe femoropopliteal diseases, a population that is not well represented in randomized controlled trials.
A potential pitfall of registry data, however, is selection bias due to unbalanced data. A propensity-score matching method was used to address this potential selection bias. This method creates a dataset close to randomized controlled trial data by matching the BMS procedure with the Zilver PTX procedure in baseline characteristics and sample sizes.
Clinical decisions on selecting revascularization treatment modalities in the real-world setting take into consideration the complexity and treatment difficulty of the diseased lesions to achieve optimal stent patency. As such, the unmatched data provide evidence of selection bias in lesion characteristics between the Zilver PTX and BMS groups in this study. The Zilver PTX has been shown to have a greater patency compared to BMS in ISR lesions in a randomized controlled study.12 This could explain the higher baseline ISR rates reported in Zilver PTX (26%) than BMS (11%) in our study, with a lower rate of Zilver PTX use in treating heavily calcified (32% vs 54%) and diffuse lesions (53% vs 64%).
TLR rate in the propensity-score matched BMS group increased to 18% from 13% in the unmatched BMS group. This could be due to a higher rate of ISR in the matched group compared with the unmatched group. Despite lower rates of diabetes mellitus, heavy calcification, and diffuse disease, the matched BMS group had a higher TLR rate than the unmatched BMS procedures. This was explored further by conducting both a subgroup analysis excluding ISR procedures, and a logistic regression model to predict the 12-month TLR rate adjusted for risk factors in the BMS group.
Our study’s subgroups analysis excluding ISR procedures found a similar result – a higher TLR rate in the Zilver PTX group compared to the BMS group. The logistic regression analysis showed no significant associations of diabetes mellitus, heavily calcification, and diffuse disease with a higher TLR rate, while only ISR was significantly associated with a higher TLR rate at 1 year (Figure 4).
This real-world study demonstrated a favorable lesion patency with the use of Zilver PTX compared to BMS. While the findings in this study are consistent with the previous Zilver PTX RCTs,13,14 this is the first study to report a significantly better patency rate in Zilver PTX compared to BMS using real-world data. A single-center study using real-world data found no superiority of Zilver PTX to BMS in femoropopliteal endovascular revascularization; however, the study was under-powered, with a small sample size of <50 in the Zilver PTX treatment arm.7
The current study included both de novo and repeat femoropopliteal interventions. While a 12-month patency rate may differ between de novo and repeat interventions, including both procedures in the current study did not create a noise in estimation of 1-year patency outcomes due to a selection bias. This is supported by the consistent findings from the subgroup analysis excluding ISR procedures, in addition to the existing literature showing that Zilver PTX clinical effectiveness in de novo and repeat revascularization was similar in mid-term patency.15,16
While lesion patency of treated lesion and vessel patency are relevant outcomes of comparative studies of two different stent types in femoropopliteal disease interventions, the causes of death were further explored in this study to address the higher mortality rate in the Zilver PTX group relative to the BMS group. Among the known causes of death in the Zilver PTX group, none were related to PAD, while causes of 3 death events were unknown.
Previous studies have also reported similar mid-term mortality rates in the Zilver PTX procedures. Although not statistically significant, Dake and colleagues (2011)13 reported a higher mortality rate in the Zilver PTX arm (n = 9; 3.7%) compared to the plain balloon angioplasty (PTA) arm (n = 4; 1.7%; P=.17) at 1-year follow-up. A single-arm study of Zilver PTX14 reported a 5.2% 2-year mortality rate, while a 7.6% mortality rate is reported from RCT data. All-cause 1-year mortality rate was 5.1% in a post-market surveillance study of Zilver PTX.17 The higher mortality rate could be the result of unmeasured confounding effects of the use of the Zilver PTX stent in patients with more severe disease and unaccounted comorbidities. There was no group difference in cardiovascular deaths, the most common cause of death in PAD patients, between Zilver PTX (2/174; 0.6%) and BMS groups (1/174; 1%; P=.90).
This study presented both TLR and TVR as patency outcomes. While TLR is a more relevant patency measure in a comparative analysis of devices, the study finding of a greater rate of TVR than TLR in the Zilver PTX group may warrant a further investigation.
Study limitations. Potential limitations to our study deserve mention. The majority of study data were retrospectively collected (approximately 75%), and factors determining operators’ decision-making of stent choices are not known. Moreover, no data were collected on clinical indications of repeat revascularization. Despite the propensity-score matching technique that partially addresses selection bias, a prospective study is needed to further evaluate clinical indications for stent selection. A cost-effectiveness analysis of the Zilver PTX stent compared to BMS use should be performed. PAD-related cost effectiveness analysis of the Zilver PTX stent relative to other revascularization strategies is also warranted in the future.
Femoropopliteal artery endovascular interventions employing the Zilver PTX stent have been shown clinically effective in 1-year target lesion patency rates compared to BMS. This study demonstrated a higher clinically driven primary lesion patency at 1 year with the use of the Zilver PTX stent compared with BMS for endovascular treatment of femoropopliteal artery in a real-world registry.
Utilizing the core-laboratory adjudicated XLPAD data with the propensity-score matching technique, this study attempted to disentangle confounding effects between complexity of lesion characteristics and operators’ clinical decision making on treatment modalities in femoropopliteal artery endovascular interventions in the real world. The study findings provide valuable insights on clinical efficacy of the Zilver PTX stent in femoropopliteal artery endovascular intervention. Further studies are warranted to examine the use of different modalities for treatment of TLR, and the associated procedural and limb outcomes.
Acknowledgment. The authors of this manuscript would like to acknowledge the contributions of Mazin Foteh, MD, Mujtaba Ali, MD, Ian Cawich, MD, Gerardo Rodriguez, MD, PhD, Dharam Kumbhani, MD, Michael Luna, MD, and Tony S. Das, MD for their contributions to the XLPAD registry. We also acknowledge the support of the University of Texas Southwestern Medical Center for their support in establishing and managing the REDCap database software utilized in the XLPAD registry (Academic Information Systems NIH grant UL1-RR024982).
1. Hirsch AT, Criqui MH, Treat-Jacobson D, et al. Peripheral arterial disease detection, awareness, and treatment in primary care. JAMA. 2001;286:1317-1324.
2. Axel DI, Kunert W, Goggelmann C, et al. Paclitaxel inhibits arterial smooth muscle cell proliferation and migration in vitro and in vivo using local drug delivery. Circulation. 1997;96:636-645.
3. Herdeg C, Oberhoff M, Baumbach A, et al. Local paclitaxel delivery for the prevention of restenosis: biological effects and efficacy in vivo. J Am Coll Cardiol. 2000;35:1969-1976.
4. Dake MD, Ansel GM, Jaff MR, et al. Durable clinical effectiveness with paclitaxel-eluting stents in the femoropopliteal artery: 5-Year results of the Zilver PTX randomized trial. Circulation. 2016;133:1472-1483; discussion 1483.
5. Banerjee S, Sarode K, Mohammad A, et al. Femoropopliteal artery stent thrombosis: report from the excellence in peripheral artery disease registry. Circ Cardiovasc Interv. 2016;9:e002730.
6. Jeon-Slaughter H, Tsai S, Kamath P, Shammas NW, Brilakis ES, Banerjee S. Comparison of lower extremity endovascular intervention outcomes in women versus men. Am J Cardiol. 2017;119:490-496.
7. Vent PA, Kaladji A, Davaine JM, et al. Bare metal versus paclitaxel-eluting stents for long femoropopliteal lesions: prospective cohorts comparison using a propensity score-matched analysis. Ann Vasc Surg. 2017;43:166-175.
8. Rosenbaum PR, Rubin DB. The central role of the propensity score in observational studies for causal effects. Biometrika. 1983;70:41-55.
9. Fraeman KH. An introduction to implementing propensity score matching with SAS applications development. NESUG. 2010. United BioSource Corporation, Bethesda, Maryland.
10. Gray RJ. A class of K-sample tests for comparing the cumulative incidence of a competing risk. Ann Stat. 1988:1141-1154.
11. Fine JP, Gray RJ. A proportional hazards model for the subdistribution of a competing risk. J Am Stat Assoc. 1999;94:496-509.
12. Zeller T, Dake MD, Tepe G, et al. Treatment of femoropopliteal in-stent restenosis with paclitaxel-eluting stents. JACC Cardiovasc Interv. 2013;6:274-281.
13. Dake MD, Ansel GM, Jaff MR, et al. Paclitaxel-eluting stents show superiority to balloon angioplasty and bare metal stents in femoropopliteal disease: twelve-month Zilver PTX randomized study results. Circ Cardiovasc Interv. 2011;4:495-504.
14. Dake MD, Ansel GM, Jaff MR, et al. Sustained safety and effectiveness of paclitaxel-eluting stents for femoropopliteal lesions: 2-year follow-up from the Zilver PTX randomized and single-arm clinical studies. J Am Coll Cardiol. 2013;61:2417-2427.
15. Tran K, Ullery BW, Kret MR, Lee JT. Real-world performance of paclitaxel drug-eluting bare metal stenting (Zilver PTX) for the treatment of femoropopliteal occlusive disease. Ann Vasc Surg. 2017;38:90-98.
16. Iida O, Takahara M, Soga Y, et al. 1-year results of the ZEPHYR registry (Zilver PTX for the femoral artery and proximal popliteal artery): predictors of restenosis. JACC Cardiovasc Interv. 2015;8:1105-1112.
17. Yokoi H, Ohki T, Kichikawa K, et al. Zilver PTX post-market surveillance study of paclitaxel-eluting stents for treating femoropopliteal artery disease in Japan: 12-month results. JACC Cardiovasc Interv. 2016;9:271-277.
REPRINTED WITH PERMISSION FROM: J INVASIVE CARDIOL 2018;30(10):373-379.