TITLE PAGE

Full Title: BIVALIRUDIN, GLYCOPROTEIN INHIBITOR AND

HEPARIN USE AND ASSOCIATION WITH OUTCOMES

OF PRIMARY PERCUTANEOUS CORONARY

INTERVENTION IN THE UNITED KINGDOM

Short title : Primary PCI – outcomes by anti-thrombotic choice

Authors names, degrees and affiliations :

1) Alex Sirker, MA (Cantab), MB BChir, PhD; University College London Hospitals,

UK

2) Mamas Mamas MA (Oxon), BM BCh, DPhil; Keele University and University of

Manchester, UK

3) Derek Robinson MA (Cantab), MSc, DPhil; University of Sussex, UK

4) Simon Anderson MSc, MB BCh, PhD; University of Manchester, UK

5) Tim Kinnaird MB BCh, MD; University Hospital of Wales, UK

6) Rod Stables MA (Cantab), BM BCh, DM: Liverpool Heart and Chest Hospital, UK

7) Mark A de Belder MA (Cantab), MD; The James Cook University Hospital,

Middlesborough, UK

8) Peter Ludman MA (Cantab), MD; Queen Elizabeth Hospital, Birmingham, UK

9) David Hildick-Smith MA (Cantab), MD; Brighton and Sussex University

Hospitals, UK

Address for correspondence and reprints

Dr. David Hildick-Smith, Department of Cardiology, Royal Sussex County Hospital, Eastern Road, Brighton, UK. BN2 5BEEmail: david.hildick-smith@bsuh.nhs.uk Telephone: +44 (0)1273 664494 ; Fax: +44 (0)1273 684554

Total word count : 4869

1

ABSTRACT

Aims

The HORIZONS trial reported a survival advantage for bivalirudin over heparin-with-

glycoprotein inhibitors (GPI) in primary PCI for ST elevation myocardial infarction.

This drove an international shift in clinical practice. Subsequent studies have

produced divergent findings on mortality benefits with bivalirudin. We investigated

this issue in a larger population than studied in any of these trials, using the United

Kingdom national PCI registry.

Methods and Results

61,136 primary PCI procedures were performed between January 2008 and January

2012. Demographic and procedural data were obtained from the registry. Mortality

information was obtained through the UK Office of National Statistics. Multivariable

logistic regression and propensity analysis modelling were utilized to study the

association of different antithrombotic strategies with outcomes. Unadjusted data

demonstrated near-identical survival curves for bivalirudin and heparin-plus-GPI

groups. Significantly higher early and late mortality was found in patients treated with

heparin alone (p<0.0001) but this group had a markedly higher baseline risk. After

propensity matching, the bivalirudin versus heparin-plus-GPI groups still

demonstrated very similar adjusted mortality (odds ratio 1.00 at 30 days, and 0.96 at 1

year). Patients treated with heparin alone continued to show higher mortality after

adjustment, although effect size was considerably diminished (odds ratio versus other

groups 1.17-1.24 at 30 days).

2

Conclusions

Analysis of recent UK data showed no significant difference in short or medium term

mortality between STEMI patients treated with bivalirudin versus heparin-plus-GPI at

primary PCI.

KEYWORDS

heparin, bivalirudin, glycoprotein inhibitor, primary PCI, STEMI, antithrombotic

Abbreviations in manuscript:

CABG – coronary artery bypass grafting

DAPT – dual antiplatelet therapy

GPI – glycoprotein 2b3a inhibitor

PCI – percutaneous coronary intervention

PPCI – primary PCI

RCT – randomized controlled trial

STEMI – ST elevation myocardial infarction

MACCE – major adverse cardiac and cerebrovascular events

NACE – net adverse clinical events

OR – odds ratio

CI – confidence interval

3

BIVALIRUDIN, GLYCOPROTEIN INHIBITORS AND

HEPARIN USE AND ASSOCIATION WITH OUTCOMES OF

PRIMARY PERCUTANEOUS CORONARY INTERVENTION

IN THE UNITED KINGDOM

Introduction

Primary PCI (PPCI) has become the treatment of choice for ST elevation myocardial

infarction (STEMI). However the optimal antithrombotic regimen remains

controversial. Evidence for benefit of GPIs came from trials such as ADMIRAL,

CADILLAC and ISAR 2. (1) A meta-analysis of over 27,000 patients treated with

abciximab in PPCI RCTs showed significant short- and long- term mortality benefits.

(2) Studies on other GPI agents suggested similar benefits.(3-5) However, subsequent

studies challenged their ongoing relevance. For example, BRAVE-3 suggested that

benefits of abciximab in STEMI were negated when higher loading doses of

clopidogrel were used.(6) The place of GPIs was further challenged by the

HORIZONS trial.(7) Early and late mortality benefits were found with bivalirudin,

compared to a heparin-plus-routine GPI combination, with a 30-day absolute

mortality difference of 1% (relative risk 0.66). The mechanisms by which bivalirudin

achieved these mortality benefits remain unclear and several issues are noteworthy.

(a) The earlier incidence of stent thrombosis with bivalirudin (with later catch-up in

the other treatment arm) was postulated to be relevant (since in-hospital complications

might be more rapidly and hence successfully addressed) – however a recent

HORIZONS substudy indicated this theory was not the explanation.(8) (b) The other

4

main candidate mechanism is via the demonstrably lower rate of bleeding in patients

receiving bivalirudin. However the means by which early bleeding problems led to

survival curves that were still diverging beyond 1 year is uncertain.(9) (c) Even more

intriguingly, a recently published analysis from HORIZONS indicates that the

mortality benefits of bivalirudin over heparin-plus-GPI were evident even when

patients with major bleeds were excluded.(10) This led to speculation about

additional, poorly defined pleiotropic actions of bivalirudin. This mortality advantage

appeared to be restricted to patients with the highest blood leucocyte counts, for

unknown reasons (11). Subsequent studies that compared bivalirudin with other

antithrombotic strategies (including EuroMax, BRIGHT, HEAT-PPCI and most

recently MATRIX Antithrombin), have produced inconsistent mortality findings

(covered in greater detail in the Discussion). This has led to considerable uncertainty

in terms of potential survival benefits with bivalirudin use compared to other

therapies in PPCI.

The diversity of clinical practice in the UK permits the study of 3 broad

antithrombotic combinations in primary PCI - bivalirudin, heparin-with-GPI, and

heparin-alone. We sought to investigate patterns of use and their association with

outcomes, using the United Kingdom National Registries for PCI procedures and

mortality.

5

Methods

UK National PCI Database

Data on PCI in the UK are compiled under the auspices of the National Institute of

Cardiovascular Outcomes Research (NICOR) and the British Cardiovascular

Intervention Society (BCIS). The dataset is composed of 113 variables, including

patient demographics and procedural details. Mortality tracking is undertaken

annually using the patients’ unique National Health Service numbers to link the

BCIS-NICOR database to that held by the UK Office of National Statistics (ONS), on

which all UK deaths are recorded. This tracking is not available for patients in

Scotland or Northern Ireland. Details on data capture, handling and validation in the

registry have been published previously (18).

Study population

Study subjects were those patients who underwent a PPCI procedure for STEMI in

the UK between 1st January 2008 and 1st January 2012. This time period was chosen

since it reflects an era of “contemporary STEMI management” in which, for example,

high dose oral antiplatelet loading was routinely used pre-procedure and radial access

was utilized in a significant proportion of cases.

Outcome measures

6

The primary outcome measure for this study was 30-day mortality. Secondary

endpoints were 1-year mortality, time to death (survival curve analysis), and in-

hospital major bleeding. Major bleeds comprised gastrointestinal bleeds, intracerebral

bleeds, retroperitoneal bleeds, blood or platelet transfusion, or an arterial access site

complication requiring surgery. Major bleeding events were obtained from recorded

entries on the BCIS-NICOR database - recording is dependent on appropriate

completion of the relevant field on the database (concerning PCI complications) by

clinical staff, either at the time of the PPCI procedure (for early bleeds) or

subsequently (for delayed bleeds). This field is not cross-populated from another

database (in contrast to mortality data).

Study definitions

Patients were coded into 3 treatment groups based on physician-allocated therapy at

time of STEMI. These groups were i) bivalirudin, ii) heparin-alone (i.e. without

accompanying GPI or bivalirudin), iii) heparin-plus-GPI. Patients who received

bivalirudin plus heparin were coded within the bivalirudin group. Patients who

received both bivalirudin and a GPI were also coded within the bivalirudin group

(since they were considered likely to represent an initial bivalirudin strategy with bail-

out GPI).

Statistical methods

7

a) Descriptive statistics

For basic analysis of patient demographics, procedural characteristics and unadjusted

outcomes, continuous variables were evaluated as means ± SD. Means, standard

deviations and percentages quoted refer to numbers within the cohort where data were

available. Chi-square tests were used to assess the significance of differences in

proportions between groups for categorical variables. Student t-testing was used for

continuous variables. Log-rank testing was used for survival curve analysis. All

statistical tests were two-tailed and p<0.05 was taken as significant.

b) Imputation

To reduce inherent bias from complete case-only analysis, missing values were

addressed by a standard technique of multiple imputation, performed using the MI

and MIANALYZE procedures of the statistical software SAS (version 9.2).

c) Multivariable logistic regression modelling

Since the primary endpoint was binary, a standard logistic regression model was

employed. The model was constructed by fitting it first to treatment allocation alone

and then by adding each of the variables of interest to the model individually.

Variables of interest were selected to encompass known or potential predictors of 30

day mortality, and included: age, gender, weight, diabetes, hypertension, smoking,

cerebrovascular disease (TIA or stroke), peripheral vascular disease, renal impairment

(defined as serum creatinine>200 micromoles/litre or on dialysis), warfarin use,

cardiogenic shock, previous MI, previous PCI, previous CABG, Q wave on ECG, and

procedural details (including year of procedure, route of access, use of a novel potent

oral antiplatelet agent (i.e. prasugrel or ticagrelor), left main stem culprit lesion, use of

8

thrombectomy, and IABP use. Type of stent (DES versus BMS) was also included

since it typically appears as an independent predictor of mortality in registry studies

of PCI, almost certainly reflecting selection bias on the part of operators. Hence stent

type may be considered a surrogate marker for other unrecorded or unquantified

comorbidities (including frailty) and its use as regressor within the model should help

to mitigate the influence of unmeasured confounders. Variables found to be

significant (p<0.05) on univariate analysis were put forward into a multivariable

logistic regression.

d) Propensity analysis

As an alternative method for addressing the potential confounding factors due to non-

randomized treatment allocation, propensity matching was employed in a separate

analysis. Propensity scores were derived for each patient, based on the estimated

probabilities of that patient being assigned to each of the treatment groups, given the

values of various predictor variables for treatment allocation. The propensity score

was then obtained using a multinomial regression model. The predictors for treatment

allocation were chosen as those measurable variables that were likely to influence

which of the 3 treatments under consideration was chosen by the operator. Hence

factors associated with increased bleeding risk (and therefore appearing on standard

bleeding risk scores (19)) were included, with the full list of variables used as shown

in for the multivariate analysis, (section c) above).

In order to undertake propensity matching involving more than 2 treatment groups,

the method described previously by Spreeuwenberg et al was employed. (20). This

used logistic regression with 30-day mortality as the response, with regressors

9

consisting of the propensity scores p1, p2 and their product p1p2, as well as the three

treatments. Standard checks for overlap of propensity scores between the treatment

arms, and subsequently for the balance of covariates, were performed.

Multivariable logistic regression modelling and propensity analysis modelling were

performed at 30 day and 1 year post PPCI time points, involving those patients on

whom there was available mortality data at the respective time points.

Results

Patient numbers, baseline demographics and procedural details

The summary of patient numbers is shown in Figure 1. The lower proportion of

tracked mortality data for the heparin-plus-GPI group reflected high use of this

strategy in Scotland, where database linkage was not possible. Baseline clinical

demographics are displayed in Table 1. It was noted that the heparin-only treatment

group contained patients with a significantly greater incidence of risk factors for

adverse outcome compared to other 2 treatment arms – specifically, greater age; more

frequent diabetes, hypertension, smoking history, peripheral vascular disease,

cerebrovascular disease and significant renal impairment; poorer NYHA class at

baseline; more prior IHD and prior CABG, and greater concurrent use of warfarin.

Procedural details are given in Table 2. Patients presenting with shock were found in

highest proportion in the heparin-plus-GPI group. In contrast, use of radial access, a

10

thrombectomy device, drug-eluting stents and novel oral antiplatelet drugs (prasugrel

or ticagrelor) was found in the highest frequency in the bivalirudin group.

Unadjusted mortality

Unadjusted mortality data for the 3 groups during hospital admission, at 30 days and

at 1 year, are shown in Table 3. Short- and medium- term mortality were significantly

higher in the heparin-only treatment groups. This is also shown in Figure 2, where log

rank testing demonstrated poorer unadjusted survival for heparin-only patients

(p<0.0001 for heparin-only versus other groups; for bivalirudin versus heparin-plus-

GPI, p=0.63).

Major bleeding

Database-recorded rates of in-hospital major bleeding were 38/4158 (0.91%) for

bivalirudin, 273/35129 (0.78%) for heparin-plus-GPI, and 110/21849 (0.5%) for

heparin alone. Of the 38 patients on bivalirudin who had a major bleed, 3 patients had

received a GPI and 35 patients received heparin. Of the 4120 bivalirudin patients who

did not have a major bleed, 600 were on a GPI and 3520 were on heparin. GPI use

was not significantly different between those in the bivalirudin group who did and did

not have a major bleed (p=0.25, chi-square test).

Multivariable logistic regression model for 30-day mortality and 1 year mortality

30-day mortality data were available for 54,514 patients (breakdown by treatment

shown in Figure 1). Following univariate logistic regression and backward stepwise

11

elimination, the following variables remained significant and were thus included in

the multivariable regression model: age, gender, diabetes, hypertension, smoker/ex-

smoker, cerebrovascular disease, peripheral vascular disease, renal impairment, year

of procedure, previous PCI, Q waves, access site, shock, left main lesion, stent type

(DES or BMS), IABP use, use of prasugrel or ticagrelor, and treatment allocation

groups (bivalirudin, heparin plus GPI, and heparin-only, with one of these acting as

the reference level).

The association of treatment allocation with 30-day mortality is indicated by the

Forest plot in Figure 3a. The bivalirudin and heparin-plus-GPI groups showed very

similar results, whilst heparin-only patients showed significantly higher mortality.

The odds ratios and 95% CI boundaries were as follows (comparisons shown in the

form x:y, with y acting as the reference) bivalirudin : heparin/GPI - OR 1.01 (95% CI

0.84-1.21), p=0.92; heparin-only : heparin/GPI - OR 1.24 (95% CI 1.13-1.35),

p<0.0001 and heparin-only : bivalirudin – OR 1.23 (95% CI 1.02-1.47), p=0.026.

Results 1-year post-PPCI are shown graphically in Figure 4a. The parsimonious

model contained similar variables to that derived for 30-day mortality, with the

addition of weight and previous MI, and the exclusion of smoking, hypertension and

gender. Patterns remained essentially unchanged from those seen at 30 days. The odds

ratios and 95% CI boundaries at 1 year were bivalirudin : heparin/GPI – OR 0.96

(95% CI 0.83-1.11), p=0.59; heparin-only : heparin/GPI – OR 1.28 (95% CI 1.19-

1.37), p<0.0001; and heparin-only : bivalirudin – OR 1.33 (95% CI 1.15-1.54),

p<0.0001.

12

Propensity analysis modelling for 30-day mortality and 1 year mortality

Propensity analysis based comparison of treatment groups is shown in Figure 3b and

4b. Consistent with the findings from multivariable analysis, there was no significant

evidence for a difference between bivalirudin and GPI in terms of 30 day mortality.

The odds ratios were bivalirudin : GPI – OR 1.00 (95% CI 0.84-1.18), p=0.98;

heparin : GPI – OR 1.17 (95% CI 1.08-1.27), p<0.0001; and heparin : bivalirudin –

OR 1.18 (95% CI 1.00-1.38), p=0.05. When comparison is made to Figure 3a, it is

seen that these 2 distinct modelling methods produced strikingly similar findings.

Patterns seen at 30 days post PPCI remain evident at 1 year. The odds ratios and 95%

CI boundaries at 1 year were bivalirudin : heparin/GPI - OR 0.96 (95% CI 0.84-1.10),

p=0.57; heparin-only : heparin/GPI - OR 1.24 (95% CI 1.16-1.32), p<0.0001; and

heparin-only : bivalirudin - OR 1.29 (95% CI 1.13-1.47), p=0.0002.

13

Discussion

Rationale for present study

Our study employed the UK national registry of PCI to analyse patterns of practice

and address areas of continuing controversy in STEMI management. HORIZONS had

a profound impact on the management of STEMI patients, in part due to its findings

of reduced mortality with bivalirudin. This led to changes in European and American

guidelines (21, 22). The HORIZONS findings on mortality were supported by an

analysis of STEMI patients in the US PREMIER registry but follow-up in this latter

work only covered index hospital admissions and hence early post-discharge events

would have been missed.(12) In contrast, the EuroMax trial studied 30-day outcomes,

finding a significant reduction in the composite of death plus major bleeding (driven

by the latter), favouring bivalirudin over a strategy in which heparin was used in

combination with a GPI in the majority (69%) of cases.(13) A patient-level pooled

analysis of HORIZONS and EuroMax did not demonstrate any significant reduction

in all-cause mortality with bivalirudin use (with bail-out GPI) versus heparin +/- GPIs

(14). Consistent findings followed from the multicentre Chinese BRIGHT study, with

non-significant changes in mortality despite significant reductions in major bleeding

with bivalirudin versus heparin-plus-GPI (15). No study yet has been adequately

14

powered to definitively address mortality effects and this question remains

unresolved.

Another contentious issue is the role for a ‘heparin-only’ strategy (i.e. without

planned concomitant GPI or bivalirudin) in primary PCI. The HEAT PPCI trial

recently addressed this in an all-comers RCT (using GPIs only for bail out, in both

arms) and found heparin yielded similar bleeding rates but significantly fewer

ischaemic events (stent thrombosis or reinfarction) compared to bivalirudin at 30 days

(16). The impact of these findings on mortality remains unclear, although a non-

significant absolute mortality difference of 0.8% (favouring the heparin group) was

noted. In contrast, BRIGHT reported a borderline significant reduction in NACE with

bivalirudin (driven by reduced bleeding) but no difference in mortality between

heparin-only and bivalirudin treatment groups in BRIGHT. Most recently, data has

been presented from the European MATRIX study (17) CHANGE REF in over 7000

patients with acute coronary syndromes (including STEMI). There was no evidence of

significant reduction in a composite MACE primary endpoint with bivalirudin versus

heparin, +/- GPI (used in 4.6% and 26% of cases respectively). It did however find

significant reductions in bleeding and in all-cause mortality. MATRIX is the first

RCT since HORIZONS to suggest an all-cause mortality benefit with bivalirudin use.

However, as with HORIZONS, an important caveat is that the apparent advantage

with bivalirudin over heparin, in terms of bleeding, is likely driven (at least in part) by

the differential GPI use between comparator arms. However, whether this explains the

mortality signal favouring bivalirudin in these two studies remains unclear. We

therefore analysed the UK national database of primary PCI to explore this specific

issue further.

15

Patient demographics and treatment allocation

Review of patient demographics and treatment allocation in our study population

showed that many patients receive heparin alone. These patients tended to be older,

and have more comorbidities. This may reflect the view that giving heparin alone

would be considered a safer option than use of heparin-plus-GPI in patients felt to be

at higher risk of bleeding (particularly relevant to operators who do not use

bivalirudin). The greater incidence of unfavourable patient characteristics in the

heparin-only group is likely to be (at least) partly responsible for the poorer

unadjusted outcomes of this group - these adverse traits are only partially quantifiable

and incompletely captured in the BCIS dataset (in common with other registries),

hence important differences (such as patient frailty) might not be reconciled even

when advanced statistical methods were employed in adjusted analyses.

Heparin-plus-GPI tended to be used in patients at relatively lower risk of bleeding

(lower age, less renal dysfunction, fewer females etc.). Some differences were also

seen in baseline demographics and procedural characteristics between the bivalirudin

and heparin-plus-GPI cohorts. It is noteworthy that the use of either prasugrel or

ticagrelor was most common in the bivalirudin group, and then more common in the

heparin only group compared to those treated with both heparin and a GPI. It is

conceivable that operators felt the benefits of a GPI were fewer when these new

antiplatelet drugs (rather than clopidogrel) were used. Operators who elected to use

bivalirudin were also more likely to incorporate other recent developments in PCI,

such as radial access and thrombectomy devices. Hence it is certainly plausible that

operators choosing bivalirudin might be ‘early adopters’ of other PPCI innovations.

16

However, an alternative and equally plausible explanation relates to the higher

proportion of bivalirudin use during the latter part of the time period under

consideration, when these other PCI developments were also being incorporated more

frequently into clinical practice. Of course, both explanations may be valid, since they

are not mutually exclusive. Importantly however, the influence of all these potentially

confounding factors (i.e. other aspects of PPCI treatment and the year of treatment

itself) should be mitigated through their inclusion as covariates in both our logistic

regression and propensity-based analyses.

Association of anti-thrombotic choice and outcomes

There was a close similarity between the unadjusted survival curves for the

bivalirudin and heparin-plus-GPI groups. 30-day mortality was 5.1% and 4.6%

respectively from our data – if there had been an underlying absolute difference of 1%

in mortality between these two received-treatment groupings, we calculate that our

sample size gave 78-81% power to detect this as a significant difference between

groups (at p<0.05). Of course, the unmatched nature of these non-randomly assigned

groups makes it impossible to directly infer the impact of antithrombotic therapy from

this non-significant difference. In order to gain insight into the potential underlying

effects on mortality of these alternative received-treatments, two different forms of

statistical modelling were employed. The findings of the two different models

strongly concurred. Neither model provided evidence for a mortality difference at 30

days between bivalirudin and heparin-plus-GPI inhibitor groups. This signal is thus

consistent with that from the strategy comparisons performed in the RCTs EuroMax,

BRIGHT and also ISAR REACT 4 (which examined a high-risk NSTEMI setting)

17

(13, 15, 24). One of the most potent messages from HORIZONS was an improved

survival with bivalirudin in STEMI, which helped to drive practice change

internationally (towards greater use of this drug). However this does not appear

consistent with the overall published experience to date and our findings are

consistent with a pooled analysis of recent randomized studies indicating no survival

difference with these contrasting antithrombotic options (25).

Even after statistical adjustment, the heparin-only group in our work fared less well.

As discussed above, it is certainly possible that adjustments (based, by necessity, only

on captured and quantified variables) may not compensate for all differences between

groups – hence an incompletely reconciled higher baseline risk in the heparin-only

group might explain the higher mortality seen. However, the results do raise an

alternative possibility, namely that a decision to use heparin on its own (possibly due

to fear of bleeding problems) might increase the risk for an adverse cardiovascular

outcome. This result would of course be contrary to the message emerging from

HEAT PPCI and hence it is worth considering why this might be so (26). In this

regard, it might be highly relevant that over 90% of patients in HEAT PPCI received a

potent novel oral antiplatelet agent (prasugrel or ticagrelor) compared with a far lower

proportion (13.8%) in our heparin-only group. It is thus plausible that the lack of

concomitant use of either a novel oral antiplatelet agent or an intravenous GPI in the

vast majority of these patients leaves them with insufficient platelet inhibition during

this early window i.e. during, and in the hours immediately after, PPCI. This might, in

turn, translate into increased adverse events with a consequent impact on survival.

Limitations

18

Being derived from a real-world PCI registry, this paper incorporates “all-comers”

undergoing primary PCI for STEMI, avoiding selection and screening bias against

high-risk patients (as is frequently encountered in RCTs). The registry represents

daily practice and contributes to the overall knowledge base, alongside data from

carefully controlled trials. However, a number of caveats and limitations must be

borne in mind to allow appropriate reader interpretation of our presented findings.

1) Common to all registry-based work, it is recognized that no statistical modelling

can fully compensate for unidentified or unmeasured confounding factors that may be

imbalanced between non-randomly allocated treatment groups. We recognize that the

presence of such confounders may therefore also have potentially influenced

comparisons in our work.

2) A second noteworthy issue is that comparisons were made between groups based

on received antithrombotic treatments. This is intrinsically different to a comparison

of treatment protocols – for example, the database does not allow differentiation

between GPI use being planned versus bail-out. Hence the heparin-plus-GPI group in

our work will include patients managed with both of these strategies (planned use and

bail-out use) and hence is somewhat heterogeneous. A further source of heterogeneity

arises from between-centre and between-operator differences in protocols for the use

of alternative antithrombotics – for example, in centres where bivalirudin is available,

it may be the default choice for some operators, whilst being used more selectively

(only in higher bleeding risk cases) by others. In contrast, at centres where bivalirudin

is not available, heparin-only is likely to be the selected strategy for high bleeding risk

cases. Nevertheless, we strongly feel that our analysis by received-treatment provides

useful new information, but it should not be misinterpreted as a comparison of upfront

treatment strategies.

19

3) As the BCIS registry is not monitored at the point of data entry, mis- or under-

recording of the antithrombotic used is, of course, possible. However, we believe that

any such under-recording would be entirely unintentional and very unlikely to

introduce any important systematic bias in favour of any particular antithrombotic

treatment.

4) The low recorded major bleeding rates reflect a combination of i) the BCIS

definition of major bleeding (see Methods), which differs from contemporary trial

definitions, making direct comparisons difficult, and ii) the possibility of under-

recording of bleeding problems by operators, since this requires retrospective logging

of events. This bleeding event registration contrasts with the robustness of our

mortality data taken from the UK Government’s Office of National Statistics.

5) Neither dosing of antithrombotic used nor ACT values achieved are recorded in the

BCIS-NICOR registry. Usual UK practice is consistent with European and US

guidelines with usual initial unfractionated heparin dose in the range 70-100 units

heparin/kg body weight without GPI and 50-70 units heparin/kg body weight when

used with GPI (reference to ESC/EACTS guidelines on myocardial revascularisation).

Conclusions

In summary, our analysis of the UK National PCI database from 2008-2012 shows no

significant difference in short or medium term mortality between groups treated with

bivalirudin versus heparin-plus-GPI in primary PCI. Our statistical modelling

suggests that previously described mortality benefits of bivalirudin over heparin-plus-

GPI are not evident in contemporary practice. Further RCTs of different regimens in

the current era are warranted, but our data suggest that the differences in mortality

20

outcomes associated with different regimens might be small. Thus, very large studies

might be needed to conclusively settle these issues.

Funding Sources None

Disclosures/Conflicts of Interest None

21

References

1. Kandzari DE, Hasselblad V, Tcheng JE, Stone GW, Califf RM, Kastrati A, et al. Improved clinical outcomes with abciximab therapy in acute myocardial infarction: a systematic overview of randomized clinical trials. Am Heart J 2004;147(3):457-62.2. De Luca G, Suryapranata H, Stone GW, Antoniucci D, Tcheng JE, Neumann FJ, et al. Abciximab as adjunctive therapy to reperfusion in acute ST-segment elevation myocardial infarction: a meta-analysis of randomized trials. Jama 2005;293(14):1759-65.3. De Luca G, Ucci G, Cassetti E, Marino P. Benefits from small molecule administration as compared with abciximab among patients with ST-segment elevation myocardial infarction treated with primary angioplasty: a meta-analysis. J Am Coll Cardiol 2009;53(18):1668-73.4. Zeymer U, Margenet A, Haude M, Bode C, Lablanche JM, Heuer H, et al. Randomized comparison of eptifibatide versus abciximab in primary percutaneous coronary intervention in patients with acute ST-segment elevation myocardial infarction: results of the EVA-AMI Trial. J Am Coll Cardiol 2010;56(6):463-9.5. Akerblom A, James SK, Koutouzis M, Lagerqvist B, Stenestrand U, Svennblad B, et al. Eptifibatide is noninferior to abciximab in primary percutaneous coronary intervention: results from the SCAAR (Swedish Coronary Angiography and Angioplasty Registry). J Am Coll Cardiol 2010;56(6):470-5.6. Mehilli J, Kastrati A, Schulz S, Frungel S, Nekolla SG, Moshage W, et al. Abciximab in patients with acute ST-segment-elevation myocardial infarction undergoing primary percutaneous coronary intervention after clopidogrel loading: a randomized double-blind trial. Circulation 2009;119(14):1933-40.7. Stone GW, Witzenbichler B, Guagliumi G, Peruga JZ, Brodie BR, Dudek D, et al. Bivalirudin during primary PCI in acute myocardial infarction. N Engl J Med 2008;358(21):2218-30.8. Dangas GD, Claessen BE, Mehran R, Brener S, Brodie BR, Dudek D, et al. Clinical outcomes following stent thrombosis occurring in-hospital versus out-of-hospital: results from the HORIZONS-AMI (Harmonizing Outcomes with Revascularization and Stents in Acute Myocardial Infarction) trial. J Am Coll Cardiol 2012;59(20):1752-9.9. Stone GW, Witzenbichler B, Guagliumi G, Peruga JZ, Brodie BR, Dudek D, et al. Heparin plus a glycoprotein IIb/IIIa inhibitor versus bivalirudin monotherapy and paclitaxel-eluting stents versus bare-metal stents in acute myocardial infarction (HORIZONS-AMI): final 3-year results from a multicentre, randomised controlled trial. Lancet 2011;377(9784):2193-204.10. Stone GW, Clayton T, Deliargyris EN, Prats J, Mehran R, Pocock SJ. Reduction in Cardiac Mortality With Bivalirudin in Patients With and Without Major Bleeding: The HORIZONS-AMI Trial (Harmonizing Outcomes with Revascularization and Stents in Acute Myocardial Infarction). J Am Coll Cardiol 2014;63(1):15-20.11. Palmerini T, Brener SJ, Mehran R, Dangas G, Genereux P, Riva DD, et al. Leukocyte count is a modulating factor for the mortality benefit of bivalirudin in ST-segment-elevation acute myocardial infarction: the HORIZONS-AMI trial. Circ Cardiovasc Interv 2013;6(5):518-26.

22

12. Pinto DS, Ogbonnaya A, Sherman SA, Tung P, Normand SL. Bivalirudin therapy is associated with improved clinical and economic outcomes in ST-elevation myocardial infarction patients undergoing percutaneous coronary intervention: results from an observational database. Circ Cardiovasc Qual Outcomes 2012;5(1):52-61.13. Steg PG, van 't Hof A, Hamm CW, Clemmensen P, Lapostolle F, Coste P, et al. Bivalirudin started during emergency transport for primary PCI. N Engl J Med 2013;369(23):2207-17.14. Stone GW, Mehran R, Goldstein P, Witzenbichler B, Van't Hof A, Guagliumi G, et al. Bivalirudin versus heparin with or without glycoprotein IIb/IIIa inhibitors in patients with STEMI undergoing primary percutaneous coronary intervention: pooled patient-level analysis from the HORIZONS-AMI and EUROMAX trials. J Am Coll Cardiol 2015;65(1):27-38.15. Han Y, Guo J, Zheng Y, Zang H, Su X, Wang Y, et al. Bivalirudin vs heparin with or without tirofiban during primary percutaneous coronary intervention in acute myocardial infarction: the BRIGHT randomized clinical trial. Jama 2015;313(13):1336-46.16. Shahzad A. How Effective are Antithrombotic Therapies in PPCI (HEAT PPCI). In: American College of Cardiology/i2 Annual Scientific Session.; 2014; Washington DC; 2014.17. Valgimigli M, Calabro P, Cortese B, Frigoli E, Garducci S, Rubartelli P, et al. Scientific foundation and possible implications for practice of the Minimizing Adverse Haemorrhagic Events by Transradial Access Site andSystemic Implementation of AngioX (MATRIX) trial. J Cardiovasc Transl Res 2014;7(1):101-11.18. Ludman PF. British Cardiovascular Intervention Society Registry for audit and quality assessment of percutaneous coronary interventions in the United Kingdom. Heart 2011;97(16):1293-7.19. Flores-Rios X, Couto-Mallon D, Rodriguez-Garrido J, Garcia-Guimaraes M, Gargallo-Fernandez P, Pinon-Esteban P, et al. Comparison of the performance of the CRUSADE, ACUITY-HORIZONS, and ACTION bleeding risk scores in STEMI undergoing primary PCI: insights from a cohort of 1391 patients. Eur Heart J Acute Cardiovasc Care 2013;2(1):19-26.20. Spreeuwenberg MD, Bartak A, Croon MA, Hagenaars JA, Busschbach JJ, Andrea H, et al. The multiple propensity score as control for bias in the comparison of more than two treatment arms: an introduction from a case study in mental health. Med Care 2010;48(2):166-74.21. O'Gara PT, Kushner FG, Ascheim DD, Casey DE, Jr., Chung MK, de Lemos JA, et al. 2013 ACCF/AHA guideline for the management of ST-elevation myocardial infarction: a report of the American College of Cardiology Foundation/American Heart Association Task Force on Practice Guidelines. J Am Coll Cardiol 2013;61(4):e78-140.22. Steg PG, James SK, Atar D, Badano LP, Blomstrom-Lundqvist C, Borger MA, et al. ESC Guidelines for the management of acute myocardial infarction in patients presenting with ST-segment elevation. Eur Heart J 2012;33(20):2569-619.23. Stone GW, McLaurin BT, Cox DA, Bertrand ME, Lincoff AM, Moses JW, et al. Bivalirudin for patients with acute coronary syndromes. N Engl J Med 2006;355(21):2203-16.24. Kastrati A, Neumann FJ, Schulz S, Massberg S, Byrne RA, Ferenc M, et al. Abciximab and heparin versus bivalirudin for non-ST-elevation myocardial infarction. N Engl J Med 2011;365(21):1980-9.

23

25. Ferrante G, Valgimigli M, Pagnotta P, Presbitero P. Bivalirudin versus heparin in patients with acute myocardial infarction: A meta-analysis of randomized trials. Catheter Cardiovasc Interv 2015.26. Shahzad A, Kemp I, Mars C, Wilson K, Roome C, Cooper R, et al. Unfractionated heparin versus bivalirudin in primary percutaneous coronary intervention (HEAT-PPCI): an open-label, single centre, randomised controlled trial. Lancet 2014;384(9957):1849-58.

24

FIGURE LEGENDS

Figure 1 : Summary of antithrombotic treatment allocation and procedural

numbers per group

Figure 2 : Kaplan-Meier curves for survival (i.e. freedom from death), censored

at 1 year post primary PCI - Probability of survival is shown on the vertical axis.

Table below graph indicates numbers at risk in each group at 100, 200 and 300 days.

Figure 3 : Forest plot indicating effect of treatment allocation on 30-day mortality

from a) multivariable logistic regression model and b) propensity matching

model - Comparisons are shown as x : y (y being the reference), such that a position

left of OR 1 indicates greater survival with x and a position right of OR 1 greater

survival with y.

Figure 4 : Forest plot indicating effect of treatment allocation on 1-year mortality

from a) multivariable logistic regression model and b) propensity matching

model - Comparisons are shown as x : y (y being the reference), such that a position

left of OR 1 indicates greater survival with x and a position right of OR 1 greater

survival with y.

25