Circulation Journaldoi: 10.1253/circj.CJ-21-0171

M edical therapies and a supervised exercise pro-gram confer a mortality benefit on patients with peripheral artery disease (PAD).1–4 Thus, maxi-

mizing medication and participation in an exercise program has been recommended in guidelines.5,6 In contrast, the endovascular first approach for femoropopliteal disease has emerged as a safe and effective treatment for symptomatic PAD, as reflected by studies showing improved quality of life, reduced symptoms, and a lower incidence of procedural complications.7 At present, peripheral revascularization is becoming an increasingly common procedure.8 Despite the treatments available, endovascular management of symp-tomatic PAD remains challenging and is often burdened by restenosis that results in treatment failure. Therefore, given the consistent results of trials demonstrating superiority of

paclitaxel-containing devices (PTXD) over traditional devices in terms of efficacy endpoints,9–15 treatment with PTXD has become the first-line approach for symptomatic PAD.5,6 However, in December 2018, a study-level sum-mary meta-analysis by Katsanos et al indicated that PTXD has a negative effect on mortality (93% increase of relative risk of all-cause death).16 In addition, it found dose-depen-dent adverse effects of PTXD. Later, the US Food and Drug Administration (FDA) and the organization, Vascu-lar Interventional Advances (VIVA), analyzed the individ-ual data sets; the FDA reported that the use of PTXD was associated with an excess 5-year mortality risk of 57%, and

Editorial p ????

Received March 2, 2021; revised manuscript received March 8, 2021; accepted March 16, 2021; J-STAGE Advance Publication released online April 29, 2021 Time for primary review: 4 days

Division of Cardiovascular Medicine, Toho University, Ohashi Medical Center, Tokyo (M.N.); Clinical Research Data Center (M.T., T.Y.), Clinical Research, Innovation and Education Center (Y.S., K.I.), Tohoku University Hospital, Sendai; Division of Biostatistics, Tohoku University Graduate School of Medicine, Sendai (M.T., T.Y.); Cardiovascular Center, Fukuoka Sanno Hospital, International University of Health and Welfare, Fukuoka (H.Y.); and Cardiology, Fukuoka Kinen Hospital, Fukuoka (T.U.), Japan

Mailing address: Masato Nakamura, MD, PhD, Division of Cardiovascular Medicine, Toho University, Ohashi Medical Center, 2-22-36 Ohashi, Meguro-ku, Tokyo 153-8515, Japan. E-mail: masato@oha.toho-u.ac.jp

All rights are reserved to the Japanese Circulation Society. For permissions, please e-mail: cj@j-circ.or.jpISSN-1346-9843

An Individual-Level Meta-Analysis Using Real-World and Pivotal Studies on Mortality From the Use of

Paclitaxel-Containing Devices in Japanese Femoropopliteal Disease Patients

Masato Nakamura, MD, PhD; Munenori Takata, MD, PhD; Hiroyoshi Yokoi, MD; Takafumi Ueno, MD, PhD; Yuka Suzuki, PhD; Koji Ikeda, PhD; Takuhiro Yamaguchi, PhD

Background: The effect of treatment with paclitaxel-containing devices (PTXD) on mortality in patients with peripheral artery disease remains controversial.

Methods and Results: An independent patient-level meta-analysis of 12 clinical trials (1,389 PTXD patients and 1,192 non-PTXD patients) was conducted. This study included 7 pivotal trials and 5 post-marketing surveillance studies on endovascular treatment for femoropopliteal artery by 6 companies. The primary endpoint was all-cause death, and 5-year cumulative mortality was estimated by a Kaplan-Meier curve. Cox proportional hazard model was used to calculate the hazard ratio (HR) and confidential interval (CI). During the median follow up of 3.0 years, 459 patients died. The cumulative 5-year mortality for the entire cohort was significantly lower in the PTXD than in the non-PTXD group (24.4% vs. 27.4%, respectively; HR, 0.81; 95% CI, 0.67–0.97; P=0.023), but this difference was no longer significant after adjustment for relevant covariates (HR, 1.01; 95% CI, 0.39–2.58; P=0.987). The Cox pro-portional hazard model revealed that sex, hyperlipidemia, Type 2 diabetes, hemodialysis, Rutherford category, and age above 75 years were significantly associated with 5-year mortality, but treatment with PTXD was not.

Conclusions: This large individual meta-analysis of patients with femoropopliteal artery disease found that the use of PTXD does not have a negative effect on 5-year mortality.

Key Words: Femoropopliteal artery; Meta-analysis; Paclitaxel; Post-marketing surveillance; Randomized controlled trial

ORIGINAL ARTICLE

Advance Publication

2 NAKAMURA M et al.

pharmaceutical and medical devices agency. Trials were eligible for inclusion in our analysis if they studied devices approved as of March 2019 for use in the femoropopliteal artery in Japan and had a follow-up period of at least 1 year. We requested collaboration with all 8 manufacturers and signed a confidentiality agreement with 6 of them; these 6 companies had performed a total of 12 eligible trials (Figure 1): 6 randomized controlled trials (RCTs) and 6 single-arm prospective studies. All 6 RCTs were pivotal studies performed to obtain approval of the following devices for treating PAD: IN.PACT Admiral (Medtronic, Santa Rosa, CA, USA); Lutonix Paclitaxel-Coated Balloon (Bard/Becton Dickinson, Covington, GA, USA); Zilver PTX Drug-Eluting Peripheral Stent (Cook Medical, Bloomington, IN, USA); Eluvia Drug-Eluting Vascular Stent System (Boston Scientific, Marlborough, MA, USA); Misago (Terumo, Tokyo, Japan); and Smart Vascular Stent System (Cordis, Cardinal Health, Santa Clara, CA, USA). The 6 single-arm studies comprised 1 pivotal study on the Life Stent (Bard/Becton Dickinson, Covington, GA, USA) and 5 PMS studies on the IN. PACT Admiral; Zilver PTX and Zilver Flex Vascular Self-Expanding Stent (Cook Medical, Bloomington, IN, USA); Misago; and Smart Stent. Subsequently, we signed contracts with the manufac-turers to obtain the blank case report forms. After confirm-ing case report forms included the items we required for our analysis, we finalized the statistical analysis plan (SAP), which was approved by the 6 companies. Then, each com-pany provided anonymized data to the Clinical Research Data Center at Tohoku University, which created a data set in a format compliant with the FDA patient-level data request table,17 and analyzed the data. To achieve com-monality of variables, after mapping the variables with SAS version 9.4 (SAS Institute, Cary, NC, USA), the data sets from individual trials were combined into analytic

the VIVA group found an excess risk of 38%.17,18 These reports indicated a late all-cause mortality signal. However, it was also found that the hazard risk ratio for mortality gradually decreased as efforts were made to minimize miss-ing data.17,18 In contrast, recent detailed, product-specific, patient-level analyses found no relationship between the use of PTXD and increased mortality;19–21 however, these anal-yses were imbalanced because of the low number of control cases. Various other efforts have been made to overcome these shortcomings, but this issue still remains controversial.

A better strategy to address the urgent question of whether PTXD increases mortality risk is for manufactur-ing companies to share all the clinical trial data they hold and for researchers to validate these individual data in a robust statistical analysis. Despite the inherent perceived limitations of an observation study, non-randomized data, such as from post-marketing surveillance (PMS), are likely to play an important role in identifying late mortality sig-nals (PMS of devices represents long-term assessment of open-label treatment to identify malfunctioning and poten-tial harmful effects). Therefore, we collaborated with the 6 companies that conducted pivotal studies or PMS on PTXD in Japan to obtain their patient-level data and performed an independent, patient-level meta-analysis under a grant from the Japanese Ministry of Health, Labour and Welfare (MHLW). We aimed to evaluate whether PTXD have a favorable or harmful effect on mortality.

MethodsPooled Analysis Study DesignWe identified 15 trials on PAD conducted by 8 manufactur-ing companies according to the principles of good clinical practice (GCP) or good post-marketing surveillance prac-tice (GPSP), based on the information from the Japanese

Figure 1. Flow chart of the study. GCP, good clinical practice; GPMS, good postmarketing surveillance; FEMOROPOPLITEAL ARTERY, superficial femoral artery; F/U, follow up.

Advance Publication

3Meta-Analysis of Paclitaxel-Containing Devices

analyzed together. The RCTs were analyzed on an inten-tion-to-treat basis. In this analysis, cross-over to treatment with PTXD during follow up was not captured. Survival was estimated by Kaplan-Meier curves and hazard ratios (HR), and 95% confidence intervals (CI) were calculated by using the Cox proportional hazard model. The model was adjusted for the following clinically relevant covariate fac-tors that were expected to have a potential effect on mortal-ity: study, age, sex, hypertension, hyperlipidemia, Type 2 diabetes, smoking, hemodialysis, and Rutherford category >3. The level of statistical significance was set at P<0.05. Statistical analyses were performed with SAS version 9.4 (SAS Institute, Cary, NC, USA).

ResultsRisk of Bias AssessmentAll of the included studies were at high and consistent risk for performance bias from the open label nature of long-term assessments of vital status, as well as the extent of missing data. Conversely, the risk of detection bias was low as outcome assessment was carried out by other blinded investigators, with the exception of the PMS of IN.PACT Admiral. In addition, all the RCTs and single-arm studies received industry sponsorship.

Demographic and Baseline CharacteristicsThe study characteristics are summarized in Table 1. Two RCTs evaluated a drug-coated balloon (DCB); 2, a drug-eluting stent (DES); and 2, a bare-metal stent (BMS). The control arm of all RCTs except one was balloon angio-plasty; 1 RCT on a DES compared the DES with another type of DES. The RCT cohort consisted of 249 cases treated with PTXD and 303 cases treated with non-PTXD. The median follow up was 3.0 years (range, 0.1–5.3 years). An overview of the devices included in the RCTs is shown in Supplementary Table 1.

The 6 single-arm studies consisted of 4 studies on a BMS, 1 on a DES study, and 1 on a DCB. One BMS study was a pivotal study for approval from the MHLW, and the others were PMS studies of the devices. The median follow up was 4.0 years (range, 0–5.6 years).

The entire cohort included 2,581 cases, 1,389 of whom were treated with PTXD (DCB, n=388; DES, n=1,001) and

databases. The manufacturers were responsible for data monitoring, review, and quality during the respective clin-ical trials. Patient-level data were reviewed for integrity and recoded as necessary to facilitate harmonization of variables and definitions across data sets. This meta-analysis was approved by the ethics committee of Toho University School of Medicine, but institutional review board approval was not required for this individual patient-level meta-analysis.

Endpoint DefinitionEach study was conducted prospectively, and this meta-analysis was a post-hoc retrospective analysis that used all deaths as the primary endpoints. Data were included for the longest available follow up. Except for the PMS of IN. PACT Admiral, all major adverse events, deaths, and tar-get limb amputations in each study follow-up period were independently adjudicated at the time of the respective study by blinded specialists with no conflicts of interest; therefore, no new re-adjudication was performed for the meta-analysis. Deaths were categorized and grouped as cardiovascular and non-cardiovascular deaths. Non-car-diovascular death was further classified into malignancy, infection, and others on the basis of the original assessment in each study. Clinical staging for PAD was evaluated by using Rutherford classification. It includes 7 categories: 0-asymptomitic, 1-mild claudication, 2-moderate classifi-cation, 3-severe claudication, 4-rest pain, 5-ischemic ulcer-ation, and 6-severe ischemic ulcer or frank gangrene.

Statistical AnalysisAll analyses were prespecified in the SAP. All baseline demographics and clinical characteristics were summarized on a patient basis, and lesion characteristics were summa-rized on a lesion basis. Continuous variables were described as mean ± SD and were compared by using a Student’s t-test. Dichotomous and categorical variables were described as counts and proportions and were compared by using the chi-squared test. Median follow-up time was estimated by using the reverse Kaplan-Meier method. To evaluate the effects of paclitaxel, endovascular treatment (EVT) devices were categorized into 2 groups: PTXD and non-PTXD. Data from the RCTs and single-arm studies were analyzed separately, and then the data from both study types were

Table 1. Study Characteristics

Name of trial Device tested Allocation of study Enrollment period Study design

AllocationSubjects included

Median F/U, months

Total patient years of F/U

No. of deaths

NCT02574481 Eluvia/Zilver PTX DES/DES 　Dec. 2015–Feb. 2017 RCT 2:1 56/28 35.8 　　　233.2 　　　　6NCT00120406

Zilver PTXDES/Balloon 　Jun. 2007–Aug. 2008 RCT 1:1 26/27 60.4 　　　251.0 　　　　7

NCT02254837 DES May. 2012–Feb. 2013 PMS 891 58.5 3,113.7 181

NCT01947478IN.PACT Admiral

DCB/Balloon 　Sep. 2013–Feb. 2015 RCT 2:1 68/32 36.2 　　　282.1 　　　　7UMIN000030540 DCB 　Dec. 2017–July. 2018 PMS 249 12.0 　　　233.7 24

NCT01816412 Lutonix DCB/Balloon 　Nov. 2012–May. 2015 RCT 2:1 71/38 26.3 　　　247.3 　　　　5UMIN000003291

MisagoBMS/Balloon 　Apr. 2010–Jun. 2011 RCT 1:1 50/51 35.9 　　　275.3 　　　　7

UMIN000026875 BMS 　Jan. 2013–Feb. 2013 PMS 292 59.7 　　　956.7 　　90

UMIN000003928Smart

BMS/Balloon 　Aug. 2010–Aug. 2011 RCT 1:1 52/53 35.4 　　　284.4 　　10

NA BMS May. 2013–July. 2013 PMS 317 58.8 1,021.1 　　88

NCT02254356 Zilver Flex BMS May. 2012–Feb. 2013 PMS 206 35.5 　　　435.5 　　24

NCT01746550 Life BMS 　Sep. 2012–Sep. 2013 Pivotal 　　74 42.3 　　　220.9 　　10

BMS, bare-metal stent; DCB, drug coated balloon; DES, drug eluting stent; F/U, follow up; NA, not available; PMS, post marketing surveil-lance; RCT, random controlled study.

Advance Publication

4 NAKAMURA M et al.

Data From RCTs The sociodemographic character-istics of the patients in the RCTs are presented in Supplementary Table 2. Significant differences between the patients treated by PTXD and non-PTXD were found only in history of smoking, and Rutherford category. The non-PTXD studies included more cases with a Rutherford cate-gory of 4, but any patients with a Rutherford category 5 or 6 were not included. The cumulative 5-year mortality was numerically higher in the non-PTXD studies than in the PTXD studies, but the difference was not statistically sig-nificant (15.4% vs. 11.7%, respectively; HR, 0.81; 95% CI, 0.44–1.51; P=0.511) (Figure 2). After adjustment for covari-ates, the HR increased to 1.01 (95% CI, 0.39–2.61; P=0.979) (Supplementary Table 3).

Data From Single-Arm Studies The demographics of

1,192 with non-PTXD (BMS, n=991; balloon angioplasty, n=201). The baseline demographics of the entire cohort are summarized in Table 2. Among the whole group of patients, 47.9% were aged ≥75 years and 70.3% were male. Type 2 diabetes was present in 57.6%, 26.3% were receiving hemo-dialysis, and 26.4% were in a Rutherford category >3. Com-pared with patients treated with non-PTXD, those treated with PTXD had more comorbidities, including hyperten-sion and hyperlipidemia, received hemodialysis more fre-quently and had longer lesions. However, the 2 groups showed no difference in Rutherford category.

Survival AnalysesThe median follow up was 3 years (range, 0–5.6 years). During follow up, 459 patients died.

Table 2. Patient Demographics and Baseline Lesion Characteristics

All, n (%) or mean ± SD

PTXD, n (%) or mean ± SD

Non PTXD, n (%) or mean ± SD P value

No. of cases 2,581 1,389 1,192

Age (years) 　0.622

<65 　　　383 (14.9) 　　　198 (14.3) 185 (15.6)

65–74 　　　957 (37.2) 　　　521 (37.5) 436 (36.8)

>74 1,233 (47.9) 　　　670 (48.2) 563 (47.6)

Male 1,815 (70.3) 　　　961 (69.2) 854 (71.6) 　0.173

Comorbidities

Diabetes mellitus 1,472 (57.6) 　　　800 (58.4) 672 (56.7) 　0.365

Hypertension 2,127 (82.8) 1,176 (85.2) 951 (80.1) <0.001

Hyperlipidemia 1,490 (58.1) 　　　842 (61.1) 648 (54.6) <0.001

Smoking 1,657 (66.3) 　　　911 (66.1) 746 (66.7) 　0.708

Hemodialysis 　　　657 (26.3) 　　　402 (30.5) 255 (21.7) <0.001

Rutherford category 　0.115

R1–3 1,818 (73.6) 　　　991 (74.9) 827 (72.1)

R4–6 　　　652 (26.4) 　　　332 (25.1) 320 (27.9)

Lesion length (mm) 138.5±103.1 149.9±109.9 125.2±92.8 <0.001

PTXD, paclitaxel-containing device; non-PTXD, non-paclitaxel-containing device; R, Rutherford.

Figure 2. Kaplan-Meier curves of cumulative mortality in randomized controlled trials (RCTs) (A), and single-arm studies (B).

Advance Publication

5Meta-Analysis of Paclitaxel-Containing Devices

sis and smoking were removed from the adjustment (HR, 0.88; 95% CI, 0.72–1.08; P=0.231) (Supplementary Table 3).

Data From the Entire Cohort In the entire cohort, the estimated 5-year mortality was 24.4% with PTXD and 27.4% with non-PTXD. The 5-year mortality was signifi-cantly lower with PTXD (HR, 0.81; 95% CI, 0.67–0.97; P=0.023), but the difference was no longer statistically sig-nificant after adjusting for clinically relevant covariates (HR, 1.01; 95% CI, 0.39–2.57; P=0.987) (Supplementary Table 3). The Kaplan-Meier curve of the cumulative incidence of mortality before and after adjustment is shown in Figure 3. The Cox proportional hazard model for all-cause mortality revealed that dialysis, Rutherford category, and age >75 years were significantly associated with 5-year mortality, but treatment with PTXD was not (Table 3).

the patients in the single-arm studies are shown in Supplementary Table 2, and the Kaplan-Meier curve of the single-arm studies is depicted in Figure 2. We found a sig-nificant difference in patient backgrounds, including the incidence of comorbidities, but no difference in the inci-dence of a Rutherford category >3 between PTXD and non-PTXD studies. In the crude analysis, the cumulative 5-year mortality for PTXD was statistically lower than that for non-PTXD (26.4% vs. 31.0%, respectively; HR, 0.77; 95% CI, 0.63–0.93; P=0.007). The difference in mor-tality was still significant after adjustment for age, sex, hypertension, hyperlipidemia, Type 2 diabetes, smoking, hemodialysis, and Rutherford category >3 (HR, 0.74; 95% CI, 0.60–0.92; P=0.006); however, the difference in mortal-ity was no longer statistically significant when hemodialy-

Figure 3. Kaplan-Meier curve of 5-year cumulative mortality in the entire cohort. (A) before adjustment, (B) after adjustment for relevant covariates. In (B), the two lines were so close that they almost overlapped and were difficult to distinguish.

Table 3. Estimate of Each Variable From the Adjusted Model for All-Cause Death

Variables Hazard ratio 95% CI P value

PTXD vs. Non-PTXD 1.01 0.39 2.58 　0.987　　Age >74 vs. 65–74 years 1.65 1.32 2.06 <0.001　　Age <65 vs. 65–74 years 0.66 0.47 0.93 　0.017　　Sex female vs. male 0.63 0.49 0.81 <0.001　　Hypertension 0.84 0.66 1.08 　0.174　　Hyperlipidemia 0.72 0.59 0.89 　0.003　　Type 2 diabetes 1.27 1.02 1.57 　0.031　　Smoking active or previous vs. never 0.79 0.63 1.00 　0.051　　Hemodialysis 2.79 2.23 3.49 <0.001　　Rutherford category 1–3 vs. 4–6 0.35 0.28 0.44 <0.0001

CI, confidence interval. Other abbreviations as in Table 2.

Advance Publication

6 NAKAMURA M et al.

The applicability to daily practice of the results of RCTs needs to be confirmed by real-world data. Therefore, the present study may provide deep insights for physicians caring for patients with PAD that were not sufficiently clear in recently published meta-analyses. Despite the care-ful design of the present analysis, its results should be interpreted with caution because of the disparate study designs and the use of mixed data sources.

The outcomes of our meta-analysis of RCTs were con-sistent with those of a device-specific meta-analysis, but not with those of the recently reported VIVA group analy-sis.18 Product-specific studies found no harmful effects of PTXD,19–21 but they were limited by a large imbalance in the number of control arms, which may hamper accurate statistical analysis of the potential mortality risk of PTXD. As mentioned above, the inclusion of BMS trials in the present analysis ensured that the PTXD and non-PTXD groups were more balanced and thus strengthened the analytical assessment.

This study may highlight the clinical implications of missing data for the assessment of mortality risk. Approx-imately 25% of data were missing in the VIVA group meta-analysis, and when recovered vital status data were included, a secondary analysis of the HR for mortality over time decreased from 38% to 27%.18 Fewer than 10% of the RCT participants were lost to follow up, and the meta-analysis showed no difference in mortality between PTXD and non-PTXD studies. More recently, an interim

Causes of DeathThe causes of death are summarized in Table 4 as simple unadjusted percentages. The high crude non-PTXD mor-tality rate across the single-arm studies and the entire cohort was due to the high incidence of non-cardiovascular death (Figure 4).

DiscussionThis meta-analysis indicates that use of PTXD in symp-tomatic femoropopliteal artery disease does not have a negative effect on 5-year mortality. First, in the entire cohort, we found no difference in the adjusted 5-year mor-tality of patients treated with PTXD; second, the effect of PTXD on mortality was different in RCTs and single-arm studies; and third, in the adjusted model, the HRs of all-cause mortality for each variable were significantly greater than for PTXD.

This study is unique and may offer some advantages over previous patient-level meta-analyses. We obtained all the clinical trial data held by 6 sponsors and validated them individually. Integrating studies conducted in accor-dance with the principles of GCP or GPSP allowed us to include studies on BMS–which eliminated the imbalance in control arms between the PTXD and non-PTXD that was present in previous meta-analyses–and a wide range of clinical entities, such as chronic limb-threatening ischemia, and thus enabled robust and reliable statistical analyses.

Table 4. Crude Mortality by Cause of Death Through 5 years

StudyCV death

Non-CV deathAll-cause deathDeath related

to cancerDeath related to infection

Death related to the other

PTXD Non-PTXD PTXD Non-PTXD PTXD Non-PTXD PTXD Non-PTXD PTXD Non-PTXD

Entire cohort 8.5 7.4 3.8 3.5 4.8 8.3 3.6 4.4 24.4 27.4

RCTs 0.8 2.6 2.8 4.7 1.5 5.7 6.6 1.8 11.7 15.4

Single-arm study 9.8 8.5 3.7 3.5 5.2 9.6 3.4 5.1 26.4 31.0

CV, cardiovascular. Other abbreviations as in Tables 1,2.

Figure 4. Kaplan-Meier curves of cumulative estimated (A) cardiovascular death and (B) non-cardiovascular death (entire study).

Advance Publication

7Meta-Analysis of Paclitaxel-Containing Devices

have a slightly favorable effect on mortality. The exact rea-sons are difficult to explain, but may be due to differences in definitions between studies, inadequate adjustment, or what we consider to be the result of chance. Thus, consis-tent with other device-specific meta-analyses19,20 and the VIVA group analysis,18 this meta-analysis does not indicate any potential specific cause of death from PTXD use.

Clinical ImplicationsAlthough we must take responsibility for these new treat-ment technologies and provide a balance between effective treatment and protection from harmful signals, we believe that this study will alleviate the concerns of patients and physicians about the long-term safety of PTXD in daily practice. This study will play an important role in resolving them and bringing this controversy to an end. Given the apparent clinical benefit of PTXD in reducing restenosis and revascularization of target lesions after EVT, it seems appro-priate to offer PTXD to patients with intermittent claudi-cation and chronic limb-threatening ischemia to improve their symptoms and quality of life. Additionally, this study provided important information for informed consent.

Study LimitationsThis study has several important limitations. First, although it was larger than previous studies, the number of cases may still be insufficient for evaluating an unplanned safety end-point. Second, the data were obtained by combining data from RCTs and non-randomized observational cohort stud-ies. As previously mentioned, observational studies have inherent limitations because they do not address cofounding factors and may be of inadequate quality. However, this meta-analysis included prospective multicenter clinical tri-als conducted in accordance with the principles of GCP or GPSP, suggesting a reliable data source. Furthermore, we included data from almost all studies on PTXD conducted in Japan, which could prevent a bias in study selection. Thus, we believe that this study provides a high level of evidence that PTXD do not affect mortality. Third, some data were missing. The difference in mortality between PTXD and non-PTXD found in the above-mentioned sum-mary meta-analysis by Katsanos et al was not confirmed by subsequent analyses, including the present analysis. Never-theless, the missing data in our study may have introduced a bias that could affect the validity of the statistical results. Fourth, although we adjusted for patient demographics, an unknown cofounder may have affected the results. Fifth, cross-over to PTXD treatment, including treatment of the contralateral limb, was limited to 1.9% in the Zilver PTX or Zilver Flex stent trials, but cross-over to PTXD in cases initially treated with BMS and balloon angioplasty in other trials was not fully captured. Finally, the devices and data included in our analysis were limited to those approved and commercially available in Japan. Thus, a direct com-parison with other meta-analyses may not be possible.

ConclusionsThis patient-level meta-analysis of data from 2,581 Japanese cases obtained in RCTs and single-arm studies found no harmful effect on 5-year mortality with the use of PTXD for symptomatic femoropopliteal artery disease across a variety of patient backgrounds. However, this study used a mixed data source from studies with different designs, so its find-ings must be interpreted with caution.

analysis of the randomized Swedish Drug-elution Trial in Peripheral Arterial Disease (SWEDEPAD), with a mean follow up of 2.49 years, reported no missing data and found no sign of an adverse effect of PTXD use on mortal-ity.22 In contrast, the current meta-analysis of single-arm trials, in which ~35% of the data were missing, suggested a potential benefit of PTXD use on mortality, although this was no longer present after adjusting for covariates. Taken together, these findings clearly show that missing data may result in analyses both overestimating and under-estimating the risk of mortality from PTXD use.

The finding of a potential benefit of PTXD use on mor-tality in the meta-analysis of single-arm studies seems to be in line with recent studies that used real-world data.23–26 Compared with RCTs, such studies are more representative of patients treated in standard clinical practice. Recent insurance claims analyses emphasized the differences in outcomes between population-based evidence and meta-analyses of RCTs.23–25 Although analyzing claim data has advantages regarding the integrity of the follow-up data, it has the potential shortcoming of not controlling for resid-ual confounders. A recent retrospective mortality analysis by Böhme et al found that long-term mortality was lower after DCB angioplasty than after plain old balloon angio-plasty of femoropopliteal lesions, and that the use of PTXD was not an independent risk factor of mortality.26 However, other researchers have suggested that inherent selection bias can obscure and overestimate the real effect of the use of PTXD.27 In the present study, the inconsistent HRs in the different adjustment models probably suggest a large selec-tion bias and imperfections in adjustment. This situation is in contrast to a pragmatic trial such as SWEDEPAD.22 Although SWEDEPAD included patients with a broad range of clinical presentations, it found no increase in mor-tality with PTXD use in patients with chronic limb-threat-ening ischemia or in those with intermittent claudication.22 Another possible reason that may explain our findings from the single-arm study data may be the difference in patient management, including medication and follow up. The single-arm studies were conducted during a similar period, but differences in patient management after EVT, such as prolonged double antiplatelet therapy and careful follow up, may have contributed to the current findings. Patients treated with new drug technology devices generally tend to receive intensive management. Indeed, Wang et al reported that being lost to follow up after EVT led to a 6-fold increase in mortality at 1 year,28 and reintervention itself has been reported to protect against mortality irrespective of whether patients are treated with PTXD.20

Potential specific mechanisms by which PTXD may cause death include increased risk of infection, pulmonary disease, and malignancy,29 although the actual cause of mortality after the use of PTXD still has not been elucidated. The present study found that non-cardiovascular death occurred more frequently with the use of non-PTXD. In this study, the use of PTXD was not a significant factor in the adjusted model for all-cause mortality, but other well-known vari-ables, such as hemodialysis, Rutherford category, diabetes mellitus, and advanced age were. Although the risk was lower in cases of dyslipidemia, patients treated with statin were defined as hyperlipidemic, so it would be reasonable to assume that this finding is consistent with previous stud-ies showing a beneficial effect of statin use on outcomes such as worsening symptoms, peripheral revascularization, and mortality.3,30 In addition, smoking was suggested to

Advance Publication

8 NAKAMURA M et al.

first or a balloon angioplasty-first revascularization strategy. J Vasc Surg 2010; 51(5 Suppl): 5S – 17S.

8. Agarwal S, Sud K, Shishehbor MH. Nationwide trends of hospital admission and outcomes among critical limb ischemia patients: From 2003 – 2011. J Am Coll Cardiol 2016; 67: 1901 – 1913.

9. Dake MD, Ansel GM, Jaff MR, Ohki T, Saxon RR, Smouse HB, et al. Paclitaxel-eluting stents show superiority to balloon angio-plasty and bare metal stents in femoropopliteal disease: Twelve-month Zilver PTX randomized study results. Circ Cardiovasc Interv 2011; 4: 495 – 504.

10. Dake MD, Ansel GM, Jaff MR, Ohki T, Saxon RR, Smouse HB, 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.

11. Tepe G, Schnorr B, Albrecht T, Brechtel K, Claussen CD, Scheller B, et al. Angioplasty of femoral-popliteal arteries with drug-coated balloons: 5-year follow-up of the THUNDER trial. J Am Coll Cardiol Interv 2015; 8: 102 – 108.

12. Rosenfield K, Jaff MR, White CJ, Rocha-Singh K, Mena-Hurtado C, Metzger DC, et al. Trial of a paclitaxel-coated balloon for femoropopliteal artery disease. N Engl J Med 2015; 373: 145 – 153.

13. Tepe G, Laird J, Schneider P, Brodmann M, Krishnan P, Micari A, et al. Drug-coated balloon versus standard percutaneous transluminal angioplasty for the treatment of superficial femoral and popliteal peripheral artery disease: 12-month results from the IN.PACT SFA randomized trial. Circulation 2015; 131: 495 – 502.

14. Brodmann M, Werner M, Meyer DR, Reimer P, Krüger K, Granada JF, et al. Sustainable antirestenosis effect with a low-dose drug-coated balloon: The ILLUMENATE European randomized clinical trial 2-year results. JACC Cardiovasc Interv 2018; 11: 2357 – 2364.

15. Steiner S, Willfort-Ehringer A, Sievert H, Geist V, Lichtenberg M, Del Giudice C, et al. 12-Month results from the first-in-human randomized study of the Ranger paclitaxel-coated balloon for femoropopliteal treatment. JACC Cardiovasc Interv 2018; 11: 934 – 941.

16. Katsanos K, Spiliopoulos S, Kitrou P, Krokidis M, Karnabatidis D. Risk of death following application of paclitaxel-coated bal-loons and stents in the femoropopliteal artery of the leg: A sys-tematic review and meta-analysis of randomized controlled trials. J Am Heart Assoc 2018; 7: e011245.

17. US Food and Drug Administration. Paclitaxel-Coated Drug-Coated Balloon (DCB) and Drug-Eluting Stent (DES) Late Mortality Panel. 2019. https://www.fda.gov/media/128140/download (accessed January 13, 2020).

18. Rocha-Singh KJ, Duval S, Jaff MR, Schneider PA, Ansel GM, Lyden SP, et al. Mortality and paclitaxel-coated devices an individual patient data meta-analysis. Circulation 2020; 141: 1859 – 1869.

19. Schneider PA, Laird JR, Doros G, Gao Q, Ansel G, Brodmann M, et al. Mortality not correlated with paclitaxel exposure: An independent patient-level meta-analysis of a drug-coated bal-loon. J Am Coll Cardiol 2019; 73: 2550 – 2563.

20. Ouriel K, Adelman MA, Rosenfield K, Scheinert D, Brodmann M, Peña C, et al. Safety of paclitaxel-coated balloon angioplasty for femoropopliteal peripheral artery disease. JACC Cardiovasc Interv 2019; 12: 2015 – 2524.

21. Dake MD, Ansel GM, Bosiers M, Holden A, Iida O, Jaff MR, et al. Paclitaxel-coated zilver PTX drug-eluting stent treatment does not result in increased long-term all-cause mortality compared to uncoated devices. Cardiovasc Intervent Radiol 2020; 43: 8 – 19.

22. Nordanstig J, James S, Andersson M, Andersson M, Danielsson P, Gillgren P, et al. Mortality with paclitaxel-coated devices in peripheral artery disease. NEJM 2020; 383: 2538 – 2546.

23. Secemsky EA, Kundi H, Weinberg I, Jaff MR, Krawisz A, Parikh SA, et al. Association of survival with femoropopliteal artery revascularization with drug-coated devices. JAMA Cardiol 2019; 4: 332 – 340.

24. Freisinger E, Koeppe J, Gerss J, Goerlich D, Malyar NM, Marschall U, et al. Mortality after use of paclitaxel-based devices in peripheral arteries: A real-world safety analysis. Eur Heart J 2020; 41: 3732 – 3739.

25. Behrendt CA, Sedrakyan A, Peters F, Kreutzburg T, Schermerhorn M, Bertges DJ, et al. Long term survival after femoropopliteal artery revascularisation with paclitaxel coated devices: A propen-sity score matched cohort analysis. Eur J Vasc Endovasc Surg 2020; 59: 587 – 596.

26. Böhme T, Noory E, Beschorner U, Jacques B, Bürgelin K, Macharzina

Author ContributionsM.N. conceived the idea for the manuscript. M.T., T.Y. conducted and verified analysis. All authors discussed the results and contributed to the final manuscript.

AcknowledgmentsThe authors express their thanks to Dr. Chiu Shih-Wei of the Clinical Research Data Center, Tohoku University Hospital, Sendai, Japan for his assistance with the analysis of this study, Kaori Shimogama from the General Incorporated Association Japan Endovascular Treatment Conference, Fukuoka, Japan for serving as the secretariat and Yamada Translation Bureau (www.ytrans.com/translation.html) for English-language editing.

Source of FundingThis study was supported by a grant from the Japanese Ministry of Health, Labour and Welfare (19CA2018).

DisclosuresM.N. reports receiving personal fees from Boston Scientific Japan K.K., Medtronic Japan Co., Ltd, Medicon, Inc. Cook Medical Japan G.K. and Terumo Corporation.

H.Y. reports receiving personal fees from Boston Scientific Japan K.K., Medtronic Japan Co., Ltd, Medicon, Inc., Terumo Corporation, Cook Medical Japan G.K. and Cardinal Health Japan.

T.U. reports receiving personal fees from Terumo Corporation, and is a member of Circulation Journal’s Editorial Team.

All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.

IRB InformationThe study was approved by the ethics committee of Toho University School of Medicine on 3 December 2019 (Reference code: A19047). Clinical Trial Registration: URL: http://www.umin.ac.jp, UMIN000041208.

Data AvailabilityThe deidentified participant data will not be shared.

References 1. Eikelboom JW, Connolly SJ, Bosch J, Dagenais GR, Hart RG,

Shestakovska O, et al. Rivaroxaban with or without aspirin in stable cardiovascular disease. N Engl J Med 2017; 377: 1319 – 1330.

2. Sabatine MS, Giugliano RP, Keech AC, Honarpour N, Wiviott SD, Murphy SA, et al. Evolocumab and clinical outcomes in patients with cardiovascular disease. N Engl J Med 2017; 376: 1713 – 1722.

3. Zinman B, Wanner C, Lachin JM, Fitchett D, Bluhmki E, Hantel S, et al. Empagliflozin, cardiovascular outcomes, and mortality in type 2 diabetes. N Engl J Med 2015; 373: 2117 – 2128.

4. Murphy TP, Cutlip DE, Regensteiner JG, Mohler ER, Cohen DJ, Reynolds MR, et al. Supervised exercise, stent revascularization, or medical therapy for claudication due to aortoiliac peripheral artery disease: The CLEVER study. J Am Coll Cardiol 2015; 65: 999 – 1009.

5. Bailey SR, Beckman JA, Dao TD, Misra S, Sobieszczyk PS, White CJ, et al. ACC/AHA/SCAI/SIR/SVM 2018 appropriate use criteria for peripheral artery intervention: A report of the American College of Cardiology Appropriate Use Criteria Task Force, American Heart Association, Society for Cardiovascular Angiography and Interventions, Society of Interventional Radiology, and Society for Vascular Medicine. J Am Coll Cardiol 2019; 73: 214 – 237.

6. Aboyans V, Ricco JB, Bartelink MEL, Björck M, Brodmann M, Cohnert T, et al. 2017 ESC Guidelines on the diagnosis and treat-ment of peripheral arterial diseases, in collaboration with the European Society for Vascular Surgery (ESVS): Document cov-ering atherosclerotic disease of extracranial carotid and vertebral, mesenteric, renal, upper and lower extremity arteries. Eur Heart J 2018; 39: 763 – 816.

7. Bradbury AW, Adam DJ, Bell J, Forbes JF, Fowkes FG, Gillespie I, et al. Bypass versus angioplasty in severe ischaemia of the leg (BASIL) trial: An intention-to-treat analysis of amputation-free and overall survival in patients randomized to a bypass surgery-

Advance Publication

9Meta-Analysis of Paclitaxel-Containing Devices

loon for the treatment of femoropopliteal lesions. Circ Cardiovasc Interv 2019; 12: e007702.

30. Kumbhani DJ, Steg PG, Cannon CP, Eagle KA, Smith SC Jr, Hoffman E, et al. Statin therapy and long-term adverse limb out-comes in patients with peripheral artery disease: Insights from the REACH Registry. Eur Heart J 2014; 35: 2864 – 2872.

Supplementary Files

Please find supplementary file(s);http://dx.doi.org/10.1253/circj.CJ-21-0171

R, et al. Evaluation of mortality following paclitaxel drug-coated balloon angioplasty of femoropopliteal lesions in the real world. JACC Cardiovasc Interv 2020; 13: 2052 – 2061.

27. Rocha-Singh KJ. Retrospective real-world studies of paclitaxel and mortality defining the many faces of bias. J Am Coll Cardiol Intv 2020; 13: 2062 – 2064.

28. Wang GJ, Judelson DR, Goodney PP, Bertges DJ. Loss to fol-low-up 1 year after lower extremity peripheral vascular interven-tion is associated with worse survival. Vasc Med 2019; 24: 332 – 338.

29. Laird JA, Schneider PA, Jaff MR, Brodmann M, Zeller T, Metzger DC, et al. Long-term clinical effectiveness of a drug-coated bal-

Advance Publication