Safety and Efficacy Study Comparing 3 New Types of Coronary Stents

Background and Study Design

The study is a multicentre, prospective, randomized, open-label trial comparing safety and efficacy of the Nobori® drug-eluting stent (DES), the Xience Prime® DES, and the ProKinetic® bare-metal stent (BMS) in patients at low risk of restenosis, i.e. receiving stents >=3.0mm diameter only, on the background of contemporary antiplatelet therapy.

2289 patients were recruited at 8 centres in 5 countries and randomized 2:1 to DES or BMS, and 1:1 to either DES subgroup. Randomization was stratified according to centre.

Analysis Datasets

The full analysis set (FAS) will include all randomized patients of whom written informed consent was obtained. Patients needing an urgent PCI were asked for oral consent prior to the percutaneous coronary intervention (PCI) and for written informed consent afterwards. This means that some randomized patients gave oral but not written informed consent. Patients who gave oral informed consent and died before written informed consent could be obtained will be included in the FAS. All other patients without signed informed consent will be excluded regardless of oral informed consent. In accordance with the intention-to-treat principle all patients will be analysed according to the allocated treatment group.

The per-protocol set (PPS) will include only those patients from the FAS who did not have one of the following major protocol violations:

• inclusion criteria not met
• exclusion criteria met
• procedures performed that were not approved by Institutional Review Board (IRB)/Ethics Committee (EC)
• no study stent (Nobori,Xience Prime or ProKinetic) received
• only stents < 3mm in nominal size, provided that the maximum vessel size measured during the procedure was <2.75mm
• undergoing multiple step procedure with a different stent received than randomised for the second or subsequent step.

Inclusion of patients older than 75 years was not approved by the local ethics committee of the Canton St.Gallen, Switzerland, and consequently, age > 75 years was added to the exclusion criteria for the center Cantonal Hospital St. Gallen. Patients who did not receive the randomly allocated stent, but another study stent according to the study protocol, will be included and analysed according to the stent received (per protocol analysis).

Demographic and Baseline Characteristics

Demographic and baseline data will be presented for each treatment arm using the FAS. Continuous variables will be presented as mean, standard deviation, median, 1st and 3rd quartile. Categorical variables will be presented as frequencies and percentages. Comparisons of continuous data will be done using a non-parametric Kruskal-Wallis rank sum test. Comparisons of categorical data will be done using a chi-squared test if the expected number of observations in each cell exceeds 5 - otherwise Fisher's exact test will be applied.

Primary Objective

The primary objective is to demonstrate non-inferiority of the Nobori DES stent compared to the Xience Prime DES stent on safety and efficacy in patients requiring stents >=3.0mm in diameter on the background of contemporary dual antiplatelet therapy (DAPT) with prasugrel and aspirin.

Primary Endpoint

The primary endpoint is the time to the first major adverse cardiac event (MACE) observed within 24 months. MACE is a composite endpoint including cardiac death, myocardial infarction (MI) and target-vessel revascularization (TVR).

Statistical Hypothesis, Model and Method of Analysis

Statistical Hypothesis

The statistical null-hypothesis is that Nobori DES is inferior to Xience Prime DES regarding the MACE rate (pi) at 24 months when using a prespecified non-inferiority margin (delta), in terms of absolute risk difference:

H0 : pi_Nobori - pi_XiencePrime >= delta

The alternate hypothesis is that the e ect of Nobori DES is non-inferior to Xience Prime DES using the same non-inferiority margin as above.

alternate hypothesis (HA) : pi_Nobori - pi_XiencePrime < delta

We will use a non-inferiority margin of 3.8 % absolute risk difference. The sample size calculation was based on delta = 3.8 %, assuming a MACE rate in the Xience Prime arm of 7.6 %, as observed in the BASKET-PROVE (BP) trial, for both DES (Kaiser et al., 2010). The stated non-inferiority margin considers a 50 % relative excess of events or more in the Nobori DES arm as inferior. This threshold is similar to that of the LEADERS trial (Windecker et al., 2008).

Statistical Model and Method of Analysis

The absolute risk difference of MACE at 24 months between the Nobori DES and the Xience Prime DES will be compared to the non-inferiority margin, using a two-sided 95 % confidence interval (CI) applying a continuity-corrected modification of the Wilson's score method (Newcombe, 1998). Non-inferiority will be declared if the upper limit of the 95 % CI of the absolute risk difference does not exceed delta. The probability of MACE within 24 months will be calculated and graphically visualized using the Kaplan-Meier estimator.

We will estimate the hazard ratio of the Nobori DES versus the Xience Prime DES using Cox proportional hazards (PH) survival analysis, allowing to test for superiority if non-inferiority can be established. The model will contain the factor "stent type" (Nobori DES vs. Xience Prime DES) and be stratified according to the variable "centre". Hazard ratios will be presented with the corresponding 95 % CI.

The proportional hazards assumption will be assessed in two steps: 1) Visually through log-log curves and 2) Testing Schoenfeld's residuals for time-dependency. In case of non-proportional hazards a time-independent logistic regression model will be used to analyze the incidence of the primary endpoint within 24 months as binary variable.

An additional landmark analysis will be performed for the time-intervals 0-12 months and 12-24 months. We will t a time-strati ed Cox PH model, including an interaction term between the factor "stent type" and time-interval (0-12 vs. 12-24 months). The interaction term will be compared to the null hypothesis of no-interaction using a log-likelihood ratio test.

The analyses are performed on the PPS (but see sensitivity analyses). All tests will have a two-sided significance level, alpha, of 0.05.

Handling of Data and Missing Values

Data will be analysed for potential extreme outliers. If present, each outlier will be investigated. However, outliers will be included in the analysis, except if a clinical cause can be excluded. There will be no missing values for the primary endpoint since losses to follow-up can be incorporated in the Cox proportional hazards models by censoring the time observed at the last follow-up/last contact date (right censoring).

Sensitivity Analyses

S1: All analyses will be repeated for the FAS. It is important to note that for non-inferiority analyses, estimates from intention-to-treat analyses are less conservative, i.e., more likely to establish non-inferiority than those from per-protocol analyses, since differences may be blurred by including patients with major protocol violations.

S2: In the Cox PH model described for the primary endpoint, we will test for an interaction between "stent type" and "centre" using a log-likelihood ratio test. A significant interaction means differences in treatment-effect across centres. This will determine whether the treatment-effect was homogeneous, which is important for the interpretation of results.

S3: In case of non-proportional hazards, time-independent analyses using logistic regression will be performed for the prespecified time-periods 0-12 months and 0-24 months.

Subgroup Analyses

We will investigate the effect of Nobori DES versus Xience Prime DES across the following prespecified subgroups:

1. Diabetes (yes vs. no).
2. Acute coronary syndrome (ACS, ACS vs. stable coronary artery disease).
3. Stent length per segment.
4. At least one stent <3.0mm received (yes vs. no).
5. More than one stented segment (yes vs. no).
6. Prior myocardial infarction (yes vs. no).
7. Smoking current (yes vs. no).
8. Gender (female vs. male)

Interaction terms between the factor "stent type" and each of the above factors will be included in the Cox regression model, together with the main effects of "stent type" and the respective factor, and compared to the null hypothesis of no-interaction using a log-likelihood ratio test. A significant interaction means that the effect of treatment differs between subgroups.

Secondary Objectives

A1: To demonstrate superiority of the Nobori DES over the ProKinetic BMS in terms of MACE within 24 months. This comparison will also act as an internal control for the non-inferiority analysis (primary objective), with the ProKinetic BMS as a putative placebo.

A2: To compare the Nobori DES to the ProKinetic BMS, and the Nobori DES to the Xience Prime DES in terms of late harm (cardiac death, MI and stent thrombosis from month 12-24).

B: To compare the Nobori DES to the Xience Prime DES, and the Nobori DES to the ProKinetic BMS in terms of additional secondary endpoints.

C: To investigate the safety - or possible harm - of DAPT with prasugrel and aspirin in patients with stable coronary artery disease (CAD) versus acute coronary syndromes (ACS) regarding bleeding events.

D: To assess the effect of DAPT with aspirin plus prasugrel versus aspirin plus clopidogrel as used in a historical comparator cohort with BASKET-PROVE (BP).

Secondary Endpoints

1. Components of the primary endpoint.

• Cardiac death
• Myocardial infarction (MI)
• Any MI
• Non-fatal MI
• Target vessel revascularization (TVR)
• Any TVR
• Non-MI related TVR

1. Composite safety endpoint of cardiac death and non-fatal MI.
2. Stent thrombosis according to ARC definitions.

• Definite
• Definite or probable
• Definite, probable or possible

1. Major bleeding including fatal bleeding, i.e., BARC >=3.
2. All cause death.
3. Net clinical benefit = Primary endpoint plus major bleeding.

Statistical Hypothesis, Model and Method of Analysis

[A1] Nobori DES versus ProKinetic BMS: MACE within 24 months

The statistical null hypothesis is that there is no difference between Nobori DES and ProKinetic BMS in terms of MACE. The Cox PH model described for the primary end-point which includes the factor "stent type" now compares Nobori DES to ProKinetic BMS. In case of non-proportional hazards a logistic regression model will be used to analyze the incidence of MACE within 24 months.

[A2] Nobori DES versus ProKinetic BMS, and Nobori DES versus Xience Prime DES: Late harm

The statistical null hypothesis is that there is no difference between Nobori DES and Prokinetic BMS, and no difference between Nobori DES versus Xience Prime DES in terms of late harm. We will use a Cox PH model, landmark analysis and a time-strati ed Cox PH model (for the time-intervals 0-12 months and 12-24 months), as described in for the primary objective, for the endpoints cardiac death, non-fatal MI and stent thrombosis. Assessment of the proportional hazards assumption will be performed in accordance to the method described for the primary endpoint. In case of non-proportional hazards, logistic regression models will be used to analyze the incidence of events.

[B] Nobori DES versus ProKinetic BMS, and Nobori DES versus Xience Prime DES: Other secondary endpoints

The statistical null hypothesis is that there is no difference between Nobori DES and ProKinetic BMS, and Nobori DES and Xience Prime DES in terms of secondary endpoints. All secondary endpoints 1-6 will be analysed using Cox PH models including the factor "stent type" and strati ed according to "centre" for events within 24 months - as specified for the primary endpoint. Assessment of the proportional hazards assumption will be performed in accordance to the method described for the primary endpoint. In case of non-proportional hazards, logistic regression models will be used to analyze the incidence of events. Landmark analyses at 0-12 and 12-4 months will be performed. Absolute probabilities of the event of interest will be calculated and graphically visualized using the Kaplan-Meier estimator. Subgroup analyses as specified for the primary endpoint will be performed for all secondary endpoints.

[C] Safety of DAPT with prasugrel and aspirin in patients with stable CAD regarding bleeding events

For this analysis we will generate four subgroups based on stent type (DES vs. BMS) and indication: 1) DES+Stable CAD, 2) DES+ACS, 3) BMS+Stable CAD, and 4) BMS+ACS. The endpoints major bleeding (BARC >=3), MACE (including its components), and stent thrombosis (see definitions in Section 5.1) will be analysed using Kaplan-Meier curves and a Cox PH model including the above groups as the 4-level factor "indication-by-stent type" and the stratifying factor "centre". Thereby, we will make the following comparisons: a) DES-ACS vs. DES-stable CAD, b) BMS-ACS vs. BMS-stable CAD, and c) DES-stable CAD vs. BMS-stable CAD. In case of non-proportional hazards, the survival analysis will be complemented by a time-independent logistic regression analysis on major bleeding events at 24 months as well as at a cut-o point of 12 months.

[D] Comparison between BP II and BP regarding DAPT with prasugrel plus aspirin versus clopidogrel plus aspirin

Two separate comparisons between trials will be made in terms of major bleeding events (BARC >=3), MACE (including its components), and stent thrombosis (see definitions in Section 5.1), using the FAS from both trials (BP: as presented in Kaiser et al. (2010)):

Comparison 1

DAPT (12 months duration) with prasugrel and aspirin in DES patients (BP II) versus DAPT (12 months duration) with clopidogrel and aspirin in DES patients (BP).

The statistical null-hypothesis is that there are no differences in rates of major bleeding events between DAPT treatment with prasugrel plus aspirin (as in BP II) versus clopidogrel plus aspirin (as in BP) in patients treated with DES. The endpoint major bleeding (BARC >=3) will be compared between all DES patients in BP II and all DES patients in

BP using a Cox PH model. Because patient populations come from two different trials rather than from one single randomised trial, a propensity score weighted analysis will be used to balance the data for confounding effects. The predictors for the calculation of propensity scores will include potential confounders, i.e., variables that may affect the treatment (DAPT as determined by trial) as well as the outcome (occurrence of bleeding events) such as: age, sex, body weight, renal function, hypertension, diabetes, prior MI, etc. We will use inverse probability weighting (Hernan \& Robins, 2006) to estimate the e ect of "DAPT" on the hazard of bleeding. The Cox PH model will include "DAPT" (prasugrel vs. clopidogrel) and "indication" (stable CAD vs. ACS) as predictors. In addition, the model will include the interaction between "DAPT" and "indication", to test whether the difference due to type of DAPT depends on indication. Hazard ratios along with corresponding 95 % CI will be presented. If the two DES in BP II (Nobori and Xience Prime) should differ with regard to bleeding events, MACE and stent thrombosis, we will also include "stent type" as predictor (Cypher and Xience in BP were very similar).

Comparison 2

DAPT with prasugrel and aspirin (1 month duration) (BP II) versus DAPT with clopidogrel and aspirin (12 months duration) (BP) in stable CAD patients with BMS.

The statistical null-hypothesis is that there are no differences in rates of major bleeding events between DAPT treatment with prasugrel plus aspirin (as in BP II) versus clopidogrel plus aspirin (as in BP) in stable CAD patients treated with BMS. The endpoint major bleeding (BARC >=3) will be compared between stable CAD patients with BMS from BP II and stable CAD patients with BMS from BP using a Cox PH model. As for comparison 1 a propensity score weighted analysis will be performed, and an inverse probability weighting used to estimate the effect of "DAPT" on the hazard of bleeding. Hazard ratios along with corresponding 95 % CI will be presented.

General

In case of non-proportional hazards, the survival analysis will be complemented by a time-independent logistic regression analysis on bleeding events at 12 and 24 months, estimating odds ratios instead of hazard ratios.

5.3 Handling of Data and Missing Values

Data will be analysed for potential extreme outliers. If present, each outlier will be investigated. However, outliers will be included in the analysis, except if a clinical cause can be excluded. In case of missing values for secondary endpoints, all available values will be used and no imputation of missing values will be performed.

References

Hernan, M. A. \& Robins, J. M. 2006: Estimating causal e ects from epidemiological data. J Epidemiol Community Health 60:578-586.

Kaiser, C.; Galatius, S.; Erne, P.; Eberli, F.; Alber, H.; Rickli, H.; Pedrazzini, G.; Hornig, B.; Bertel, O.; Bonetti, P.; De Servi, S.; Brunner-La Rocca, H.-P.; Ricard, I. \& Pfisterer, M. 2010: Drug-eluting versus bare-metal stents in large coronary arteries. New England Journal of Medicine 363:2310-2319.

Newcombe, R. G. 1998: Interval estimation for the difference between independent pro-portions: Comparison of eleven methods. Statistics in Medicine 17:873-890.

Windecker, S.; Serruys, P. W.; Wandel, S.; Buszman, P.; Trznadel, S.; Linke, A.; Lenk, K.; Ischinger, T.; Klauss, V.; Eberli, F. et al. 2008: Biolimus-eluting stent with biodegrad-able polymer versus sirolimus-eluting stent with durable polymer for coronary revascularisation (leaders): a randomised non-inferiority trial. The Lancet 372(9644):1163-1173.

Clinical Relevance:

The results of BPII will provide evidence for the performance of the newest-generation stents on the market in daily practice and insights about the efficacy and safety of current stent designs in an all-comer population. Moreover, BPII will assess the performance of DAPT with aspirin and prasugrel compared with aspirin plus clopidogrel in terms of ischemic and bleeding endpoints. - Thus, these findings should have major impact on the current use of coronary stents, medical antiplatelet therapy regimes and our understanding of possible reasons for late stent thrombosis.