1-Year Results of the ZEPHYR Registry (Zilver PTX for the Femoral Artery and Proximal Popliteal Artery)

Predictors of Restenosis

Study population

Patients with symptomatic peripheral artery disease were screened by noninvasive tests to detect limb ischemia and presence of FP lesions. Study enrollment occurred only after satisfying the arteriographic inclusion/exclusion criteria described herein. A total of 831 FP lesions (797 limbs, n = 690) with Zilver PTX implantation were finally enrolled.

Endovascular procedure

The indication for revascularization included symptomatic disease with >50% diameter stenosis determined by angiography and mean pressure gradient >10 mm Hg determined with a 4-F diagnostic catheter. Decision on Zilver PTX implantation was left to the discretion of physicians. For endovascular therapy (EVT), a 6-F sheath was inserted into the femoral artery mostly with a contralateral approach. After 5,000-U heparin infusion, 0.035- or 0.014-inch guidewires were used to cross the lesion. Prior to stent implantation, the guidewire was exchanged for a 0.035-inch one, compatible with the stent delivery system. A final post-dilation was performed for at least 30 s with a noncompliant balloon the size of which was equivalent to distal vessel diameter. Dual antiplatelet therapy with aspirin and either ticlopidine or clopidogrel was recommended to start at least 1 week before and continue for at least 2 months after the stent implantation in accordance with the package insert. Antiplatelet and anticoagulant regimens were used according to the physician’s discretion based on patient’s condition.

In the current study, intravascular ultrasound (IVUS) was performed to record the data on vessel characteristics immediately after wire crossing and at end of the procedure when possible. IVUS was performed with a commercially available IVUS console and phased-array 20-MHz (Eagle Eye Gold, Volcano Corp., Rancho Cordova, California) or mechanical-scan 40-MHz (OptiCross, Boston Scientific, Marlborough, Massachusetts) IVUS catheters. IVUS was manually pushed forward at a constant speed of 10 mm/s. IVUS and fluoroscopy images were recorded simultaneously to link IVUS images with location using ruler for off-line analysis. External elastic membrane (EEM) area before ballooning and minimal stent area (MSA) (i.e., the smallest cross-sectional area within the stent), after post-dilation were assessed. In occluded cases, IVUS was also used if the wire was in a true lumen or subintimal space. In the current study, IVUS data as well as angiographic data at baseline were evaluated at each participating institute and not by the core laboratory review.

Follow-up protocol

All participants were asked to visit centers at 1, 3, 6, and 12 months after EVT. At each visit, ischemic symptoms and ankle-brachial index were evaluated and duplex ultrasonography (DUS) was routinely conducted to evaluate patency. At 12 months after EVT, patients who had not received any reintervention or major amputation were scheduled for readmission at the study institution and underwent follow-up angiography for the treated lesions when possible. The follow-up angiographic data were reviewed at a core laboratory using CAASV software (version 5.7, Pie Medical Imaging, Maastricht, the Netherland). Target lesion revascularization was conducted for ≥50% diameter stenosis within 5 mm of the target lesion after documentation of recurrent clinical symptoms.

Endpoints

The primary endpoint was 12-month restenosis rate assessed by DUS or follow-up angiography, with a tolerance of ±2 months. Restenosis was defined as recurrence of ≥50% diameter stenosis determined by angiography or a peak systolic velocity ratio >2.4 by DUS (3). Requirement of any reintervention or major amputation (defined as surgical limb excision above the ankle) within 1 year was automatically included in restenosis. Secondary endpoints were major adverse limb event (MALE) and stent thrombosis (ST). MALE was defined as major amputation or any reintervention, including both surgical or endovascular reintervention. ST was determined when apparent occlusion met the following criteria: 1) initial procedural success; 2) rapid symptom occurrence; 3) thrombus present at procedure; and 3) lesion resolved with <50% diameter narrowing by thrombolysis therapy.

Statistical analysis

Data are shown as means ± SD for continuous variables or as percentages for dichotomous variables, unless otherwise mentioned. A p of <0.05 was considered significant. One-year restenosis incidence in the overall population was calculated as p_MALE + (1 – p_MALE) × p_Restenosis, where p_MALE and p_Restenosis indicate 1-year MALE prevalence in the overall population and 1-year restenosis prevalence in the MALE-free subgroup, respectively. Because cases exist that completed 1-year follow-up without MALE but failed to have patency assessed at 12 ± 2 months, we estimated p_Restenosis on the basis of the hypothesis that cases with missing data on patency are subject to the same risk as the remaining observed cases, which is similar to the Kaplan-Meier estimation. Therefore, p_Restenosis was calculated to be equal to restenosis prevalence in the population that completed 1-year follow-up without MALE and had patency assessed at 12 ± 2 months. Similarly, p_MALE in the overall population (including cases with death or dropout within 1 year) was calculated to be equal to the MALE prevalence in the population that completed 1-year follow-up without death or dropout within 1 year. The 95% confidence interval (CI) of 1-year endpoint incidence rate was estimated by 100,000-time bootstrap resampling. The association of baseline characteristics with restenosis, as well as MALE, was assessed with their odds ratios, which were derived from the generalized linear mixed model with a logit link function. In the model, the interinstitution variability and the intersubject variability were treated as random effects, whereas the following baseline characteristics were treated as fixed effects: critical limb ischemia; restenotic lesion; reference vessel diameter; lesion length; chronic total occlusion; calcification; IVUS-evaluated EEM area; angiography-evaluated post-treatment stenosis; and IVUS-evaluated MSA. An odds ratio of a baseline characteristic without adjustment for other baseline characteristics, named an “unadjusted” odds ratio, was obtained from the mixed model in which the fixed effect was the covariate alone. We conveniently named this model the “univariate” model. On the other hand, an odds ratio of a baseline characteristic with adjustment for other baseline characteristics, named the “adjusted” odds ratio, was obtained from the model whose fixed effects included the covariate of interest plus other covariates. This model was named the “multivariate” model. The predictive ability was assessed by the area under the receiver-operating characteristic (ROC) curve. For its drawing, sensitivity, p{X ≥ c | D = 1}, and specificity, p{X < c | D = 0}, were calculated as p{X ≥ c} × p{D = 1 | X ≥ c} / p{D = 1} and p{X < c} × p{D = 0 | X < c} / p{D = 0}, respectively, where X denotes a predictive test of interest, c indicates its cutoff value, and D is a binary indicator of restenosis. p{D = 1 | X ≥ c}, p{D = 1}, p{D = 0 | X < c}, and p{D = 0} were estimated as described herein. Monotonicity of the calculated sensitivity and specificity in X was guaranteed by using the nearest neighbor estimation of the bivariate distribution (4). The difference of the area under the ROC curve was tested by 100,000-time bootstrap resampling. Risk stratification analysis was additionally performed according to the number of accumulated risk factors for restenosis. The cumulative incidence of ST was estimated by the Kaplan-Meier method. The association of number of antiplatelet agents (aspirin, thienopyridines, and cilostazol) with ST was evaluated by the Cox regression model with mixed effects. In the model, the interinstitution variability and the intersubject variability were treated as random effects, whereas the number of agents during the follow-up period was entered as a time-dependent covariate with fixed effects. Proportional hazards assumption was made on the basis of the finding that the time-by-covariate interaction term additionally included in the model did not demonstrate statistical significance. Statistical analysis was performed by R (version 3.1.0, R Core Team, Vienna, Austria).