Design and primary findings of the Minimizing Adverse Haemorrhagic Events by Transradial Access Site and Systemic Implementation of Angiox (MATRIX) trial have been previously reported.6–8 Briefly, MATRIX (NCT01433627) was a program of 3 independent randomized controlled trials in an all-comers population with ACS. The first trial, MATRIX-Access Site, compared TRA with TFA in 8404 ACS patients. The second trial, MATRIX-Antithrombin (n=7213), was a randomized comparison of 2 antithrombotic strategies: bivalirudin with use of glycoprotein IIb/IIIa inhibitors (GPI) restricted to angiographic complications (no-reflow or giant thrombus) compared with UFH with use of GPI left to the investigator's discretion. The third trial, MATRIX-Treatment-Duration, was a randomized comparison within patients assigned to bivalirudin, comparing prolonged (after PCI) with short-term (during PCI only) bivalirudin administration. VPs were those presenting with acute HF (Killip class II to IV)9–11 or OHCA at the time of randomization. The trial was approved by the institutional review board at each participating center, and all patients gave written informed consent.

Patients were randomized to receive treatments in a 1:1 ratio, with a random block size stratified by type of ACS, intended or ongoing use of P2Y12 inhibitor (clopidogrel vs ticagrelor or prasugrel), and study site. All interventions were administered in an open-label fashion. Access site management during and after the diagnosis or therapeutic procedure was left to the discretion of the treating physician, and closure devices were allowed as per local practice. Bivalirudin was administered according to the product labeling, with a 0.75mg/kg bolus, followed immediately by a 1.75mg/kg/h infusion until completion of the PCI. In those assigned to bivalirudin prolongation, the choice between 2 regimens (full dose for up to 4hours or a reduced dose of 0.25mg/kg/h for at least 6hours) was made at the discretion of the treating physicians. UFH was administered at a dose of 70 to 100 units or 50 to 70 units/kg in patients not receiving or receiving GPI, respectively. Subsequent UFH dose adjustment on the basis of the activated clotting time was left to the discretion of the treating physicians. A GPI could be administered before PCI in all patients in the UFH group on the basis of the treating physician's judgment, but the drug was to be administered in the bivalirudin group only in patients who had periprocedural ischemic complications after PCI. The use of other medications was allowed according to professional guidelines.

Clinical follow-up was performed at 30 days. The 2 coprimary 30-day composite outcomes of the MATRIX-Access and Antithrombin trials were major adverse cardiovascular events (MACE), defined as the composite of all-cause mortality, MI, or stroke; and net adverse clinical events (NACE), defined as the composite of MACE or noncoronary artery bypass grafting (CABG)-related major bleeding (Bleeding Academic Research Consortium [BARC] type 3 or 5). The primary outcome for MATRIX-Treatment-Duration was a composite of urgent target vessel revascularization (TVR), definite stent thrombosis, or NACE. Secondary outcomes included each component of the composite outcomes, cardiovascular mortality, and stent thrombosis (defined as the definite or probable occurrence of a stent-related thrombotic event according to the Academic Research Consortium classification). Bleeding was also assessed and adjudicated on the basis of the Thrombosis in Myocardial Infarction (TIMI) and Global Utilization of Streptokinase and Tissue Plasminogen Activator for Occluded Arteries (GUSTO) scales. All outcomes were prespecified. An independent clinical events committee blinded to treatment allocation adjudicated all suspected events.

The MATRIX-Access and Antithrombin trials were designed as superiority studies on 2 coprimary outcomes at 30 days, expecting a rate reduction of 30%, corresponding to a rate ratio of 0.70. All analyses were performed per intention-to-treat principle, including all patients in the analysis based on the allocated treatment. Events up to 30 days postrandomization were considered. We analyzed primary and secondary outcomes separately for VP and non-VP patients as time to first event using the Mantel-Cox method, accompanied by log-rank tests to calculate corresponding 2-sided P values. We did not perform any adjustments for multiple comparisons but set the alpha error at 2.5% to correct for the 2 coprimary outcomes. We analyzed secondary outcomes with a 2-sided α set at 5% to allow conventional interpretation of results. Survival curves were constructed using Kaplan–Meier estimates. Absolute risk differences and number needed to treat/harm (NNT/NNH) were also calculated. We performed secondary analyses in the VP group to separately assess clinical outcomes in the HF and OHCA subgroups. We also performed stratified analyses according to prespecified subgroups (the center's annual volume of PCI, the center's proportion of radial PCI, age, sex, type of ACS, body mass index, intended start or continuation of prasugrel or ticagrelor, diabetes, estimated glomerular filtration rate, history of peripheral vascular disease, previous heparin, and randomization to access site/antithrombin type) and estimated possible effect modifications using interaction terms or tests for trend across ordered groups separately for the VP and non-VP study populations. We also performed sensitivity analyses by using Cox regression analysis (unadjusted and adjusted) for coprimary endpoints and all-cause mortality and competing risk analysis (for death) for individual components of coprimary endpoints (MI, stroke and BARC 3 or 5). All analyses were performed using the statistical package Stata 15.1 and R 3.4.4.

Rates of MACE and NACE were higher in VP than in non-VP, as were nearly all secondary outcomes (table 5 of the supplementary data).

No significant interaction was noted between access site and VP criteria with respect to 30-day MACE and NACE coprimary endpoints (interaction P=.77 and 0.89, respectively; figure 1 and table 3; figure 1 and table 6 of the supplementary data). MACE and NACE were similarly reduced with radial compared with femoral in VP (MACE: 14.9% vs 16.5%; RR, 0.89; 95%CI, 0.64-1.25; P = .51; NACE: 15.8% vs 18.9%; RR, 0.82; 95%CI, 0.59-1.13; P = .22) or in non-VP (MACE: 8.1% vs 9.5%; RR, 0.85; 95%CI: 0.72-0.99; P = .039; NACE: 9.0% vs 10.7%; RR, 0.84; 95%CI: 0.72-0.97; P = .022) patients (figure 1, table 3; table 6 of the supplementary data). TRA provided consistent relative benefits over TFA in terms of individual endpoints (table 3; figure 2 and table 6 of the supplementary data), including all-cause mortality (interaction P = .55), cardiovascular mortality (interaction P = .46), and BARC 3 or 5 bleeding (interaction P = .30). Absolute benefits in favor of TRA over TFA were at least four-fold greater in VP compared with non-VP (absolute risk difference of −1.7%, −1.5% and −2.7% in VP and −0.4%, −0.4% and −0.6% in non-VP for all-cause mortality, cardiovascular mortality and BARC 3 or 5 bleeding respectively; as shown on table 3, and on table 6 of the supplementary data).

Figure 1.

Main outcomes of radial vs femoral access in VP and non-VP. Radial and femoral access were compared on the basis of hemodynamic/electric vulnerability, with rate ratios and 95% confidence intervals (95%CI), for the coprimary endpoints and their components (death, myocardial infarction, stroke, BARC 3 or 5). BARC, Bleeding Academic Research Consortium; MI, myocardial infarction; VP, vulnerable patients.

(0.17MB).

There was also no interaction between allocation to antithrombin treatment and VP criteria for MACE or NACE (interaction P=.43 and.17, respectively; see figure 2 and table 3; figure 3 and table 7 of the supplementary data). Bivalirudin was associated with a nominally significant reduction in all-cause and cardiovascular mortality compared with UFH in VP (all-cause mortality: RR, 0.51; 95%CI, 0.31-0.84; P = .007; cardiovascular mortality: RR, 0.50; 95%CI, 0.30-0.83; P = .006; with risk differences of −5.3% for both), but not in non-VP (all-cause mortality: RR, 0.99; 95%CI, 0.62-1.58; P = .97; cardiovascular mortality: RR, 0.99; 95%CI: 0.60-1.65; P = .97; with risk differences of 0% for both; see table 3; and table 7 of the supplementary data). However, interaction testing for both endpoints did not reach statistical significance (interaction P = .056 and .060 respectively, figure 4 of the supplementary data). Bivalirudin reduced BARC 3 or 5 bleeding rates in both VP (RR, 0.30; 95%CI, 0.13-0.66; P = .0015) and non-VP (RR 0.65; 95%CI, 0.45-0.95; P = .023) groups compared with UFH (interaction P = .073), with somewhat greater absolute benefit in the former (absolute risk difference −4.4%) compared with the latter group (absolute risk difference −0.8%; table 3; table 7 of the supplementary data).

Figure 2.

Main outcomes of bivalirudin vs UFH in VP and non-VP. Bivalirudin and UFH were compared on the basis of hemodynamic/electric vulnerability, with rate ratios and 95% confidence intervals (95%CI), for the coprimary endpoints and their components (death, myocardial infarction, stroke, BARC 3 or 5). BARC, Bleeding Academic Research Consortium; MI, myocardial infarction; UFH, unfractionated heparin; VP, vulnerable patients.

(0.17MB).

Overall findings were largely consistent when the VP group was stratified according to the presence, absence or severity of advanced Killip class or OHCA at presentation as well as according to prespecified subgroups (data not shown) or alternative statistical methods were applied (table 8 of the supplementary data).

Patients with ACS presenting with hemodynamic or electrical vulnerability are frequent in daily practice and suffer from a higher risk of morbidity and mortality. European Society of Cardiology guidelines underline that high-risk ACS patients with acute HF, cardiogenic shock or OHCA are those who benefit the most from expediting all steps of the care pathway.1,2 Nevertheless, there is no specific recommendation concerning the preferable access site or antithrombotic treatment combination for angiography and/or PCI, if clinically indicated. This reflects the paucity of randomized data on this high-risk ACS patient population undergoing invasive management.

TRA is recommended over TFA in ACS patients across the board in the European guidelines but not in the ACC/AHA guidelines. In VP with hemodynamic or electrical instability, TFA is more frequently undertaken, as low systolic and mean arterial pressure hampers radial artery accessibility, coronary intervention is typically more complex and the need for concomitant hemodynamic support devices is more frequent. Advanced Killip class has been repeatedly identified among the main causes of radial failure.12–14 However, over the last few years, experience and emerging techniques have facilitated the use of TRA. A large analysis of the NCDR CathPCI Registry in 692 433 STEMI patients undergoing primary PCI explored the temporal trends of TRA, which increased from 2% in 2009 to 23% in 2015, with significant geographic variation.15 Age, sex, cardiogenic shock, cardiac arrest, operators entering practice before 2012, and nonacademically affiliated institutions were all associated with lower rates of TRA.15 Among the 7231 patients with advanced Killip class in the British Cardiovascular Intervention Society database, TFA was used in 5354 and TRA in 1877 patients. TRA was independently associated with lower 30-day mortality, in-hospital MACE and major bleeding.16 In the present study, we observed that TRA remained associated with consistent benefits in VP compared with non-VP, with a treatment effect on absolute basis being larger in the former compared with the latter group. A reasonable interpretation of our findings is that use of TRA does not seem to be associated with specific penalties in VP, who suffer from greater absolute risks and derive a higher absolute risk reduction from this intervention. Our observation, therefore, supports the use of TRA as the default access in all ACS patients undergoing invasive management. At subgroup analysis, the effect of TRA vs TFA remained consistent throughout all prespecified covariates with the notable exception of center proportion of radial PCI. Both VP and non-VP groups allocated to TRA in centers with the highest proportion of radial PCI experienced a clinically meaningful reduction in MACE or NACE endpoints compared with TFA. In contrast, in centers with a low or average proportion of radial PCI, TRA was apparently associated with somewhat smaller benefits in non-VP or even a slightly increased risk, especially for MACE, in VP. The current findings should be interpreted by taking into account that operators enrolling in MATRIX had to be adequately trained in both TRA and TFA. This observation has important clinical implications,17reinforcing the notion that especially for VP undergoing invasive management, TRA should be the default access site only if performed by routine radial operators. Conversely, less expert centers/operators might further expand their training by selecting TRA in less vulnerable patients.

Data comparing bivalirudin with UFH in ACS patients with HF and/or OHCA are limited.18 Some evidence suggests that chronic HF, independent of atherosclerosis and ACS, is associated with elevated thrombin levels and faster formation of compact, resistant fibrin clots, and therefore bivalirudin might be even more beneficial in these patients.4,5 In the EUROMAX, there was no significant interaction for the primary endpoint across patients with Killip class I vs II-IV, but this latter group was small (77 and 69 in the bivalirudin and heparin groups, respectively).19 In the HEAT-PPCI, the primary endpoint was consistent across stratification according to left ventricular function impairment (defined by left ventricle ejection fraction< 55%).20 Pinto et al.5 reported an analysis of the Premier Hospital Database comparing the use of bivalirudin and heparin in more than 116 000 congestive HF patients undergoing PCI. In-hospital mortality, which was the primary outcome of the study, was lower for bivalirudin monotherapy (2.3%) compared with heparin monotherapy (4.8%). In a matched propensity-score analysis, a mortality benefit remained associated with the use of bivalirudin compared with heparin. Bivalirudin therapy was also associated with lower bleeding or transfusion rates, as well as shorter hospitalizations.5

In the MATRIX trial, bivalirudin failed to significantly reduce the composite coprimary endpoints compared with UFH, but was associated with lower rates of major bleeding and all-cause mortality, irrespective of GPI use in the comparator arm.7,21 In the current analysis, we observed that VP showed greater absolute benefits from bivalirudin compared with UFH with respect to both coprimary endpoints, likely reflecting the higher event rates observed in these patients. The trends in favor of bivalirudin in VP at interaction testing for all-cause mortality, cardiovascular fatalities or major bleeding might suggest that a direct compared with an indirect thrombin inhibitor might be particularly beneficial in this selected population. The significant reduction of mortality and cardiovascular mortality, despite a trend toward higher MI, with bivalirudin observed in VP, might be attributed to the greater benefit derived by these patients in terms of major bleeding, and partially to the trend in lower stroke rates compared with the non-VP group. However, our analyses remain inconclusive and at best hypothesis generating. Subgroup analyses showed that prior administration of UFH but not ticagrelor or prasugrel might further optimize the MACE or NACE endpoints compared with UFH in VP, which is at variance with the corresponding observation in the non-VP group. The notion that the uptake of even newer P2Y12 oral inhibitors is particularly delayed in VP22 may provide a possible mechanistic explanation for our current findings, which altogether reinforce the message that parenteral strategies (ie, cangrelor or short glycoprotein IIb/IIIa inhibitors infusion or prolonged bivalirudin at full PCI dose)22–24 more than oral antiplatelet agents might be prioritized in VP.

This is a post hoc analysis of the MATRIX trial, which was not powered to investigate the effects of the experimental treatment strategies in the VP subgroup. Our results should be interpreted in the context of uncontrolled Type I and Type II errors and regarded as hypothesis generating. We did not adjust for multiple comparisons, increasing the risk of type I error. The MATRIX study did not exclude patients based on advanced Killip class or OHCA at presentation. Nevertheless, it remains unknown if and how much this high-risk patient category was consecutively included in the study. The requirement of written informed consent before patient participation obviously skewed inclusion toward conscious and collaborative patients only, to whom our results should apply. Even so, the proportion of VP (11.1%) compares favorably with many other previous ACS studies, which almost completely excluded them from inclusion.

Our definition of VP was not prespecified and encompasses a heterogeneous patient population and only a few patients with overt cardiogenic shock (Killip class IV) were included at the time of PCI.