Objective To reassess the efficacy of cangrelor efficacy using the universal definition of myocardial infarction (MI).
Design We adopted a novel approach to systematically implement the universal definition of MI. Two physicians blinded to treatment allocation reviewed plots of CK-MB and troponin values in relation to time of randomisation and percutaneous coronary intervention (PCI) to identify patients with stable or falling biomarkers pre-PCI (ie, primary cohort), and those with post-PCI CK-MB elevations.
Setting The CHAMPION PLATFORM trial.
Patients Non-ST-elevation acute coronary syndromes (95%) and stable angina patients (5%).
Interventions Cangrelor versus placebo.
Main outcome measures The efficacy of cangrelor compared with placebo using the reclassified events (type 4a MI) and the original clinical events committee-adjudicated (CEC PCI-MI) results was investigated.
Results Of 5295 patients, 3406 (64.4%) were in the primary cohort. Type 4a MI occurred in 4.3% (226 events/5295 patients) while original CEC PCI-MI occurred in 6.5% (344 events/5295 patients), a significant difference (p<0.0001). Using the reclassified MI events, the primary composite endpoint of death, MI, or ischaemia-driven revascularisation through 48 h occurred in 5.4% of patients (4.9% cangrelor, 6.0% placebo; OR 0.80; 95% CI 0.63 to 1.02) as opposed to 7.5% of the primary analyses (7.0% cangrelor, 8.0% placebo; OR 0.87; 95% CI 0.71 to 1.07).
Conclusions Systematic, strict implementation of the universal MI definition with emphasis on baseline assessment may enhance discrimination in detecting PCI-MI and may allow for more rigorous assessment of interventions in patients undergoing early PCI.
Statistics from Altmetric.com
In patients with acute coronary syndromes (ACS), creatine kinase-MB (CK-MB) elevation after percutaneous coronary intervention (PCI) has been extensively associated with an increased risk of death, characterised by a direct continuous relationship between the amount of CK-MB elevation and the degree of risk.1–5 This has supported the inclusion of myocardial infarction (MI) related to PCI as an endpoint (EP) in clinical trials. However, the assessment of PCI-related MI is becoming increasingly complex.
Data from registries and clinical trials document decreasing times from hospital admission to cardiac catheterisation in non-ST-segment elevation (NSTE) ACS.6–9 Recently, in the Cangrelor versus Standard Therapy to Achieve Optimal Management of Platelet Inhibition (CHAMPION) PLATFORM trial, the median time from admission to PCI was only 7.9 h.10 The CHAMPION PLATFORM trial evaluated cangrelor, a short-acting intravenous P2Y12 inhibitor, in patients with NSTE-ACS or stable angina undergoing PCI. The trial results failed to show a significant reduction in the primary EP of death, MI or ischaemia-driven revascularisation (IDR) at 48 h in cangrelor-treated patients. However, a benefit on pre-specified secondary EPs not dependent on biomarkers, including death, Q-wave MI and stent thrombosis, was observed. It has been hypothesised that challenges related to the definition of recurrent MI in the setting of very early PCI may have confounded the MI EP adjudication and potentially obscured the detection of treatment effects.
Identification of early recurrence of MI in patients rapidly treated with PCI limits the ability to discriminate new EP MI events related to PCI from the index MI event because cardiac biomarkers at the time of PCI are potentially still rising and the two curves of biomarker release tend to overlap. The need to comprehensively identify all PCI-related MI (sensitivity) competes with the necessity to differentiate this EP from the index MI (specificity). Novel approaches that maximise discrimination between PCI-related MI and index MI in the setting of very early PCI may be helpful in clinical practice and clinical trials.
In 2007 and 2012, the task force on the universal definition of MI defined recurrent MI as occurring only if cardiac biomarkers are stable or falling, and suggested a 20% cut-off to define reinfarction.11 ,12 In the present study, we adopted a methodology to systematically implement the definition of PCI-related MI using the criteria proposed by the Universal MI Working Group, and explored the efficacy of cangrelor against placebo using this definition in the CHAMPION PLATFORM study.
Patients enrolled in the CHAMPION PLATFORM trial were considered for this analysis. This study has been previously described and reported.10 In brief, the trial compared cangrelor, a short-acting, potent and reversible intravenous P2Y12 inhibitor, administered as a bolus (30 μg/kg body weight) followed by an infusion (4 μg/kg/min) with a placebo bolus and infusion in 5362 patents with NSTE-ACS managed with PCI who were not treated with clopidogrel prior to randomisation. The infusion was continued for at least 2 h with a maximum of 4 h. At the end of the infusion, patients in both groups received a loading dose of 600 mg of clopidogrel. A small group of patients with stable angina (N=287; 5% of the total patients) were initially eligible in the study before a protocol amendment.
Conducted concurrently with PLATFORM, the CHAMPION PCI trial compared cangrelor with clopidogrel 600 mg rather than placebo in a population that also included ST-segment elevation MI.13 These analyses were focused on the PLATFORM database because no ST-segment elevation MI patients were enrolled, and using a single trial would likely maximise internal validity of the findings.14
Protocol requirements for biomarkers
The PLATFORM protocol required collection of at least one CK or CK-MB sample and one troponin sample at the time of randomisation before PCI and at least three CK-MB samples after PCI (first sample to be collected ‘immediately after PCI’ (within 2 h), with additional samples collected every 8 h (±1 h) for the next 24 h). Troponin collection was not required after PCI. If available, CK-MB and troponin values collected as part of clinical care pre-PCI were requested, but no verification was performed to confirm complete reporting.
Plots of cardiac biomarkers and key clinical events were created. The plots were based on a semi-logarithmic scale that displayed biomarker values as coloured dots in two dimensions, where the time from randomisation (in hours) is provided on the x-axis (linear scale) and biomarkers indexed by the upper limit of normal (ULN) are provided on the y-axis (logarithmic scale). Vertical and horizontal lines help orient the reviewer. The horizontal dotted lines are fixed and mark the lines corresponding to 1, 3 and 5×ULN for a given biomarker. The vertical lines on the x-axis indicate specific time of interest including the time of randomisation (time 0; R); the time of PCI start; the time of hospital discharge; and the 72-h marker after PCI (72 h). Red dots correspond to local CK-MB values, green dots to local troponin I values, and blue dots to local troponin T values. A progressive number labelled each CK-MB dot for identification purposes. A typical biomarker plot is shown in figure 1.
Adjudication process: type 4a MI assessment
Biomarker plots were generated for all patients undergoing PCI. Two reviewers (SL and AT), blinded to treatment allocation and prior MI adjudication results, independently reviewed each plot and made the following determinations: (1) baseline biomarker status, defined as whether temporal trends of cardiac markers at the time of PCI were either visually normal, stable, or falling versus rising or unknown; (2) presence or absence of post-PCI CK-MB (re)-elevation; and (3) peak and nadir CK-MB values in patients with CK-MB (re-)elevations post-PCI.
Baseline biomarker status was assessed using cardiac biomarkers (CK-MB or troponin) prior to PCI or using the sample collected immediately after PCI. CK-MB was favoured for status definition but troponin was allowed, at the reviewer's discretion, if it was the only biomarker available. A discordant adjudication on any of the three adjudicated fields defined a disagreement. These were resolved by consensus or with a third expert reviewer (PT), if the conflict remained unresolved.
Universal MI definition for PCI-related MI (type 4a MI)
In the current analyses we applied the universal MI definition for PCI-related MI—that is, type 4a MI. This definition requires an elevation of >3×ULN in patients with normal baseline biomarkers and a re-elevation of at least 20% between peak and nadir with a peak >3×ULN in patients with abnormal but stable/falling biomarkers. According to the universal MI definition, type 4a MI cannot be assessed when biomarkers are rising or unknown at the time of PCI.
Original CHAMPION MI definition (CEC PCI-related MI)
The definition of PCI-related MI used by the original CHAMPION PLATFORM clinical events committee (CEC) is shown in the online supplementary appendix. This definition, which is similar to definitions used prior to the publication of the universal MI definition, relied on clinical judgement rather than a strict assessment of baseline biomarker status to assess relationships between the timing of cardiac ischaemia-related symptoms, biomarker trends and PCI. Thus, stable or falling biomarkers at the time of PCI were not required to assess PCI-related MI. In addition, the relative increase required with the original CEC PCI-related MI definition was at least 50%, while it is at least 20% in the universal MI definition.
The definition of spontaneous (non-procedural) MI in CHAMPION PLATFORM required clinical manifestations of myocardial ischaemia based on signs, symptoms and/or electrocardiographic changes plus an elevation of at least 2×ULN of CK-MB, which was favoured over troponin.
Patients adjudicated as having normal or stable/falling biomarker status defined the primary cohort. If the biomarker status was unknown, typically due to insufficient or no biomarker data before PCI, the patient was classified by convention as rising and excluded from the primary cohort. After the reconciliation between reviewers was finalised, the statistician linked the peak and nadir identification to the actual CK-MB value and calculated the relative increase. If the peak was >20% higher than the nadir and >3×ULN, a type 4a MI was assigned. Inter-observer agreement rate was calculated with κ statistics. Only CK-MB (re-)elevations in patients in the primary cohort were included in the final analysis.
Categorical data are presented as proportions and continuous data are presented as medians (25th, 75th percentiles). Baseline differences between the primary and non-primary cohort patients and between rates of type 4a MI and CEC-adjudicated PCI-MI were compared using the χ2 test or the Mann–Whitney test, as appropriate.
The universal MI definition encourages a full reporting of biomarker elevations after PCI and specifically thresholds of 5 and 10×ULN.11 Thus, additional type 4a MI EPs were defined using these thresholds for the CK-MB peak.
Differences in the incidence of type 4a MI between cangrelor and placebo were evaluated by fitting a logistic regression model with type 4a MI as outcome and randomised treatment as predictor. Similar analyses were performed with type 4a MI defined by thresholds of 5 and 10×ULN. We then used type 4a MI definitions in lieu of the original CEC PCI-related MI definition and explored randomised treatment effect on the primary composite EP of death, MI and IDR at 48 h. For the latter analysis, the MI component (total MI) included the newly defined PCI-related MI (type 4a MI) instead of the original PCI-related MI definition (CEC PCI-related MI), while MI unrelated to PCI was left as originally defined by the CEC. Q-wave MIs (related or unrelated to PCI) were also included.10 According to the original primary analysis in PLATFORM, all analyses were performed in the modified intention-to-treat population, defined as randomised patients who received at least one dose of study drug and underwent the index PCI.
Due to the exploratory nature of the study, no formal adjustment for multiple comparisons was performed. A p value <0.05 was considered statistically significant. All statistical analyses were performed with SAS V.9.2.
We favoured CK-MB to define biomarker status due to the faster release kinetic of CK-MB over troponin in both the rising and falling phase,15 and due to less missing data. However, as a sensitivity analysis, we also used troponin to define baseline biomarker status. In this analysis, patients were categorised as ‘normal’ if all pre-PCI troponin samples were normal, ‘abnormal’ if at least one troponin sample was above the ULN, and ‘unknown’ if no troponin data were available pre-PCI. Patients with only one abnormal troponin pre-PCI were excluded from the troponin-defined cohort. Patients with two or more troponin samples before PCI were categorised as rising versus stable/falling. Only patients categorised as normal or abnormal and stable/falling were included in the troponin-defined cohort.
Of the 5295 patients in the modified intention-to-treat population, 140 (2.6%) had missing CK-MB values in the first 24 h after PCI and biomarker plots were not assessed. In the 5155 patients included (87.4%), the median (25th, 75th) number of CK-MB samples was 5 (4, 5) and only 12% of patients (N=624) had >1 CK-MB value before PCI. In the remaining patients, the first post-PCI sample was used to assess baseline status. The first post-PCI CK-MB sample was collected a median of 2.1 h (2, 2.3) after PCI. Overall, 3406 (66.3%) had cardiac biomarkers adjudicated as normal or stable/falling at the time of PCI and defined the primary cohort. For 29 patients (0.6%), biomarker status was assessed with troponin data only. The κ statistic for inter-observer agreement (SL and AT) was 0.86, with the most common reason for disagreement being the misidentification of the peak or the nadir CK-MB value (57% of the disagreements).
Baseline characteristics for the primary cohort and excluded patients are presented in table 1. Patients in the primary cohort tended to have more conventional risk factors for coronary artery disease and prior PCI. The median time from hospital admission to angiography was significantly longer in the primary cohort compared with the excluded patients (8.6 h (2.5, 23.6) vs 4.7 h (2.0, 17.7); p<0.001).
In total, 226 (4.3%) type 4a MI events were identified in the primary cohort compared with 344 (6.5%) PCI-related MI events, which were identified by the original CEC adjudication, a difference that was highly significant (p<0.0001). Baseline characteristics of patients with type 4a MI and CEC PCI-MI are detailed in the appendix (see online supplementary table S1). There were 175 patients with both type 4a MI and CEC-adjudicated PCI MIs (77.4% of type 4a PCI-MIs and 50.9% of CEC-adjudicated PCI-MIs). There were 51 patients classified as having a type 4a MI but not a CEC PCI-MI. There were 169 patients classified as having a CEC PCI-MI but not a type 4a PCI-MI. Of the 226 type 4a MIs, 118 (52%) had a peak CK-MB >5×ULN and 46 (20%) had a CK-MB peak >10×ULN. The incidence of type 4a MIs and CEC PCI-related MIs by randomised treatment assignment are presented in table 2 and figure 2A, while randomised treatment effect on the total MI component (MI related and unrelated to PCI) and the composite primary EP are listed in table 3 and figure 2B. Using the reclassified MI events, the primary composite EP of death, MI and IDR through 48 h occurred in 5.4% of patients (4.9% cangrelor, 6.0% placebo; OR 0.80; 95% CI 0.63 to 1.02) as opposed to 7.5% of the primary analyses (7.0% cangrelor, 8.0% placebo; OR 0.87; 95% CI 0.71 to 1.07). Cangrelor was associated with a trend towards an increased effect using a higher threshold to define type 4a MI.
Of the 5295 modified intention-to-treat patients included, 515 (9.7%) had no troponin data pre-PCI (unknown). Of the 4780 patients with troponin data, 1591 were normal (33%) and 3189 (67%) were abnormal. Of those categorised as abnormal, 1889 (59%) had only one sample and were excluded from the primary cohort and 1300 (41%) had two or more troponin samples. In these 1300 patients, troponin was rising in 723 patients (who were excluded) and stable/falling in 577. Therefore, a total of 2168 patients (45% of those with troponin data) were included in the troponin-defined cohort.
In this cohort, 116 type 4a MI events were observed. Of these, 69 (59%) had peak CK-MB >5×ULN and 25 (22%) had a CK-MB peak >10×ULN. Treatment effects on type 4a MI, total MI component and primary composite EP in the troponin-defined primary cohort are shown in table 4. The online supplementary table S2 in the appendix reports treatment effects using a threshold of 5 and 10×ULN to define type 4a MI. By restricting type 4a MI assessment to the troponin-defined cohort, lower event rates but higher treatment effects were observed.
In a contemporary NSTE-ACS population treated with a very early invasive strategy, we documented that over one-third of patients had baseline cardiac markers that precluded an assessment of PCI-related MI using the universal MI definition. A systematic implementation of this definition was associated with lower rates of PCI-related MI. These findings suggest that the universal MI definition may enhance discrimination in detecting PCI-related MI and may allow a more rigorous assessment of interventions.
According to the universal MI definition,11 if cardiac biomarkers are not stable or falling at the time of PCI, assessment cannot be made as to whether PCI-related MI has occurred or not. In CHAMPION PLATFORM, this group of patients represented more than one-third of the overall population, a proportion that was not anticipated.16 Although patients with rising cardiac biomarkers may develop new post-PCI CK-MB elevation, the prognostic implications of these re-elevations as well as their relationship with PCI-related myocardial damage are unclear. Novel approaches to accurately measure the procedural outcome in the setting of very early PCI and to maximise discrimination between index and EP MI are needed.
To maximise discrimination between the index and PCI-related MI, biomarkers—especially CK-MB with its fast release kinetics15—may provide a more robust comparison of interventions that target peri-PCI MI. Q-waves typically require days to develop and thus may have limited value in distinguishing two MI events that are temporally close; furthermore, Q-waves are rare after PCI.10 ,13 On the other hand, symptoms are hard to standardise and are common even after a successful PCI. While other approaches, such as advanced imaging modalities, might be available in the future, logistical considerations and cost are potential limitations in large clinical trials.
By considering only patients in whom type 4a MI could be assessed (ie, primary cohort patients), the incidence of PCI-related MI using the universal MI definition was similar to the original CEC adjudication (ie, 6.6% (226 type 4a MI/3406 primary cohort patients) vs 6.5% (344 CEC PCI-related MI/5295 total patients)). This highlights the importance of a careful assessment of the baseline biomarker status during event adjudication. Biomarker plots provided a simple and reproducible tool to assess baseline biomarker status at the time of PCI, and post-PCI CK-MB elevations, and to identify specific CK-MB measurements of interest. This novel methodology could therefore be useful to streamline CEC operations and improve the efficiency of large clinical investigations.
The short time from randomisation to PCI in CHAMPION PLATFORM resulted in diagnostic challenges for PCI-related MIs, so the use of the universal MI definition to guide the definition of PCI-related MI was considered. White et al performed a pooled analysis of the two CHAMPION trials and first applied this definition to reclassify PCI-related MI.17 In this analysis, the original CEC MI definition was maintained for patients with baseline normal or unknown troponin values, while only CEC-adjudicated Q-wave MIs were used if baseline troponin levels were abnormal. The overall MI incidence was 2.8% (365 events in 13 049 patients) as opposed to 6.9% (902 events in 13 049 patients) using the original CEC adjudication MI results. This analysis showed a significant 18% OR reduction in MI favouring cangrelor (OR 0.82; 95% CI 0.68 to 0.99).
Similar to White et al, we used the universal MI definition to define PCI-related MI and showed similar reductions in PCI-related MI with cangrelor. The two analyses have important differences. First, each patient was reviewed individually in the present analysis. This allowed the identification of MI events excluded by White et al, like CK-MB re-elevations in patients with abnormal but falling troponin before PCI not associated with Q-waves (figure 1). Second, in agreement with the universal MI definition, we used a 20% cut-off rather than the original 50% criterion, which is preferred from an analytical perspective.18
As a sensitivity analysis, baseline biomarker status was also defined using troponin data only instead of CK-MB. By restricting the assessment of PCI-related MI only to patients with normal or stable/falling troponins, we observed lower event rates, suggesting that troponin may provide enhanced specificity to detect recurrent MI in the setting of early PCI and also higher treatment effects. The optimal combination of biomarkers to best assess the baseline status before PCI and the EP MI after PCI is unknown. While the latest universal MI definition12 recommended troponin for both assessments, there is uncertainty on the use of troponin to assess EP MI after PCI. Specifically, it is debated whether the same degree of post-PCI elevation should be used for troponin (especially with high-sensitivity assays) as compared with CK-MB.19 ,20 The third universal MI definition is appropriately cautious on the use of high-sensitivity troponin assays,12 which are expected to increase substantially the number of potential MI due their very high analytical sensitivity. Overall, the present analysis and the White analysis suggest that the universal MI definition may provide a stricter and more specific assessment of PCI-related MI and a better quantification of treatment effects.
Based in part on these analyses, CHAMPION PHOENIX, a large phase III trial of cangrelor, has been initiated.21 This trial will compare cangrelor with clopidogrel standard of care in patients who require PCI and are clopidogrel naïve; it is expected to enrol 10 900 patients. In this trial, the universal MI definition and biomarker plots will be implemented to identify and adjudicate suspected MI EPs with an emphasis on the assessment of baseline status, which will be performed by integrating at least two samples of troponin, and the presence of new ischaemic symptoms and of new ischaemic ECG changes.21
This analysis has several limitations. First, the biomarker status at the time of PCI was mostly assessed using a CK-MB value drawn immediately after PCI (rather than immediately before PCI) on the basis of the general assumption that a PCI-related CK-MB elevation required >2 h. Although the validity of this assumption is untested, the observation that this early post-PCI sample appeared unaffected, even in patients with important CK-MB elevations, is reassuring. Also, troponin data before PCI were limited, with most patients having <2 troponins prior to PCI. Therefore, treatment effects in the sensitivity analyses should be interpreted with caution.
Second, the precision of data on PCI-related MI may be limited by the heterogeneity of the assay used. In CHAMPION PLATFORM, virtually all patients had local CK-MB (activity or mass) assays rather than core lab CK-MB. Also, CK-MB was favoured over troponin for MI adjudication. However, recent data suggest CK-MB may be favoured for diagnostic purposes in PCI settings.22
Finally, the definition of spontaneous or type 1 MI was different than the universal MI definition, as CK-MB was favoured over troponin and 2×ULN rather than 1×ULN was required. However, the primary goal of the present post-hoc analysis was exploratory and we recognise that the findings of a possible effect of cangrelor on MI and other EPs should be interpreted with caution and need to be tested prospectively.
In CHAMPION PLATFORM, over one-third of patients had cardiac biomarkers that were not stable or falling at time of PCI. This complicates the diagnosis of PCI-related MI and the interpretation of treatment effect on the procedural outcome. We adopted a novel approach to systematically implement the universal MI definition on PCI-related MI. This approach may enhance specificity for PCI-related MI diagnosis and may allow a more rigorous assessment of treatment effects in ACS patients undergoing early PCI. This methodology is being tested prospectively in the ongoing CHAMPION PHOENIX study, a phase III clinical trial comparing cangrelor with clopidogrel.
We thank Brian Tinga of the Duke Clinical Research Institute for his tremendous help in the creation of biomarker plots. We also thank Elizabeth E S Cook of the Duke Clinical Research Institute and Jayne Prats of the The Medicines Company for excellent editorial support.
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Files in this Data Supplement:
- Data supplement 1 - Online appendix
Contributors All authors contributed to the conception and design, acquisition of data or analysis and interpretation of data; and drafting the article or revising it critically for important intellectual content; and gave final approval of the version submitted.
Funding The CHAMPION PLATFORM trial was funded by The Medicines Company. The current analyses were funded by the Duke Clinical Research Institute. The sponsors had no role in the conception and design of the present study. The analyses were performed by the Duke Clinical Research Institute statistical team and independently confirmed by the Sponsor.
Competing interests PT: Advisory board and research grant: Merck, Inc. HDW: Research grant: Bristol-Myers Squibb, Sanofi Aventis, Daiichi Sankyo, GSK, Astra Zeneca, Merck Sharpe & Dohme, Schering Plough, Johnson & Johnson, Roche, Pfizer, NIH, Medicines Company, Eli Lilly (all significant); Consultant/advisory board: Regado (modest). CMG: Research grant: The Medicines Company (significant); Honoraria: The Medicines Company (significant); Consultant/advisory board: The Medicines Company (significant). GWS: Consultant/Advisory Board: The Medicines Company (modest). RAH: Research grants: The Medicines Company, BMS, Merck, Sanofi-Aventis, Portola, Novartis (all significant); Consultant/advisory board: Merck, Astra Zeneca (all modest); Consultant/advisory board: BMS, Sanofi-Aventis (all significant). DLB: Research grants: Amarin, AstraZeneca, Bristol-Myers Squibb, Eisai, Ethicon, Medtronic, Sanofi Aventis, The Medicines Company. KWM: Research grants: Bayer, Sanofi-Aventis, Boehringer Ingelheim, Bristol-Myers Squibb, Daiichi Sankyo, Eli Lilly, Glaxo Smith Kline, Johnson and Johnson, Merck, Novartis, Portola Pharmaceuticals, Pozen, Regado (all significant); Consultant/advisory board: Bayer, Merck, Johnson and Johnson, Boehringer Ingelheim, Glaxo Smith Kline, Eli Lilly, Bristol-Myers Squibb, Ortho/McNeill, Sanofi-Aventis (all modest); Consultant/advisory board: Astra Zeneca (significant).
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement Sharing of additional data will be assessed on a case-by-case basis by the authors and CHAMPION PLATFORM Publications Committee.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.