OUP user menu

Implementation of standardized assessment and reporting of myocardial infarction in contemporary randomized controlled trials: a systematic review

Sergio Leonardi, Paul W. Armstrong, Phillip J. Schulte, E. Magnus Ohman, L. Kristin Newby
DOI: http://dx.doi.org/10.1093/eurheartj/eht003 894-902 First published online: 25 January 2013

Abstract

Myocardial infarction (MI) is a key endpoint in randomized controlled trials (RCTs), but heterogeneous definitions limit comparisons across RCTs or meta-analyses. The 2000 European Society of Cardiology/American College of Cardiology MI redefinition and the 2007 universal MI definition consensus documents made recommendations to address this issue. In cardiovascular randomized trials, we evaluated the impact of implementation of three key recommendations from these reports—troponin use to define MI; separate reporting of spontaneous and procedure-related MI; and infarct size reporting. We searched ClinicalTrials.gov and MEDLINE databases for cardiovascular RCTs with more than 500 patients in which enrolment began between September 2000 and July 2012 and that listed MI in the primary endpoint. We searched English-language publications with primary results or design papers. Of 3222 studies screened, 96 (3.0%) met our criteria. We extracted enrolment start date, number of patients and MI events, follow-up duration, and coronary revascularization rate. Data extraction quality was assessed by duplicated extractions. Of 96 RCTs, 80 had a primary results publication, comprising 608 091 patients and 43 621 endpoint MIs. Myocardial infarction represented 45.3% (95% confidence interval, 40.2–50.4) of events in the primary composite endpoint. Troponin defined MI in 57% (53/93) of trials with an MI definition available. Of these RCTs, three used troponin only if creatine kinase-MB was unavailable, six used troponin to define peri-procedural MI, seven specified the 99th percentile as the MI decision limit, and three reported spontaneous and procedure-related MI separately. None reported biomarker-based infarct size, but five reported MI as multiples of the assay upper limit of normal. Although MI is a major component of cardiovascular RCT primary endpoints, standardized MI reporting and implementation of consensus document recommendations for MI definition are limited. Developing appropriate strategies for uniform implementation is required.

  • Myocardial infarction
  • Clinical trials
  • Systematic reviews

Introduction

Myocardial infarction (MI) is a widely accepted non-fatal cardiovascular endpoint employed to assess efficacy and/or safety of new treatments in clinical trials. However, failure to use a standard MI definition has emerged as a major challenge. In September 2000, recognizing the superior sensitivity and prognostic utility of troponin compared with creatine kinase (CK)-MB, an expert consensus document from the European Society of Cardiology (ESC) and American College of Cardiology (ACC) provided guidance to the scientific and clinical communities on redefining MI and proposed troponin as the diagnostic ‘gold standard’.1 To improve consistency across randomized controlled trials (RCTs), this document specified that MI endpoints in RCTs be classified as spontaneous vs. related to coronary revascularization procedures and that the quantity of myonecrosis be determined. A 2007 and, most recently, a 2012 update reasserted troponin as the preferred biomarker for myonecrosis and also recommended that clinicians and investigators categorize MI type according to a five-category classification scheme, including whether the MI was spontaneous or revascularization-related.2,3 Prior to the release of the 2012 Universal Definition of MI document, we undertook the current study to evaluate the extent to which the 2000 and 2007 consensus recommendations had been implemented in contemporary cardiovascular RCTs.

Methods

We performed a systematic review of cardiovascular RCTs with more than 500 patients in which MI was part of the primary endpoint. The 500-patient threshold was selected to identify RCTs with enough MI endpoints to adequately assess all recommended components of MI reporting and that would be most likely to affect clinical practice. For the same reason, we limited our search to published trials. Our systematic review was conducted in accordance with the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) statement.4

Search strategy

We separately searched two databases: ClinicalTrials.gov, the official source of the US National Institutes of Health (NIH) for clinical trial registration, and the National Library of Medicine's MEDLINE database. Although RCTs could be registered on ClinicalTrials.gov for the entire period explored (1 September 2000 through 4 July 2012), the NIH mandated registration of all RCTs only after 27 September 2007.5 Because some trials that started before September 2007 may not have been captured, we used MEDLINE to complement the ClinicalTrials.gov search. Finally, we reviewed online materials using the Google search engine. We restricted our search to RCTs in cardiovascular disease, but did not restrict the type of intervention to which patients were randomized. From this pool of RCTs, those with an enrolment start date before September 2000 were excluded from further analysis. Figure 1 outlines the results of our database searches, culminating in a final set of 96 RCTs for our study; 80 had primary results published and 16 had only a design paper. Additional details of our search methodology are explained in Supplementary material online, Appendix 1.

Figure 1

Systematic review and selection of clinical trials.

Data abstraction

We abstracted study name; enrolment start date; publication year; numbers of patients, MI events, and components in the primary endpoint composite; reference within the publication to the 2000 ESC/ACC consensus document and the 2007 Universal Definition of MI; use of a clinical events classification committee; follow-up duration; and text of the endpoint MI definition used. Accuracy of the MI definition abstracted from the design paper and/or primary manuscript was assessed by contacting trial investigators and obtaining the endpoint MI definition for their respective trials. To ensure the quality of the data extraction process, a 30% random sample of included RCTs was re-reviewed, and the data were collected in a second abstraction sheet. In addition, a random 30% of records from each database was re-reviewed after initial screening was finalized as a quality check on RCT selection. No errors in data abstraction were identified by these quality control measures.

Metrics of guideline recommendations adherence

We determined the proportion of RCTs in which the ESC/ACC 2000 document and the 2007 Universal Definition of MI were referenced by manually reviewing the reference list of the design paper, primary results manuscript, or both. We also determined the proportion of RCTs referencing other consensus documents endorsed by the ESC, ACC, or American Heart Association (AHA). Then, we examined each trial for the use of the following key recommendations in the 2000 ESC/ACC consensus document for MI redefinition and in the 2007 Universal Definition of MI:

  • Recommendation 1: Use of troponin for MI diagnosis and MI decision limit provided. We evaluated this recommendation in RCTs for which the MI definition (including biomarkers employed) used in the trial was published. In five trials, the biomarker employed was unavailable in the design paper or primary manuscript but was obtained by contacting the investigators.610 We also explored adherence to this recommendation according to coronary revascularization group and tertiles of duration of time between the publication of the ESC/ACC 2000 MI redefinition document and the start of enrolment. To explore the level of adherence to Recommendation 1 as a function of coronary revascularization rate, we created three RCT categories:

    • Interventional RCTs: RCTs in which percutaneous coronary intervention (PCI) or coronary artery bypass grafting (CABG) was required by protocol either as part of the randomized intervention or as an inclusion criterion. Only protocol violators did not undergo coronary revascularization, and the rate of revascularization was assumed to be near 100%.

    • Acute coronary syndrome (ACS) RCTs: RCTs in which a coronary revascularization could be performed as part of treatment for the index ACS event but was not required.

    • Other RCTs: RCTs in which coronary revascularization was possible but not expected (e.g. secondary prevention trials enrolling patients with risk factors but no prior MI or patients with chronic stable coronary artery disease).

    • Recommendation 2: Separate reporting of spontaneous MI and procedure-related (PCI or CABG) MI.

    • Recommendation 3: Report of infarct size using area under the troponin or CK-MB curve or peak biomarker values (ESC/ACC 2000 recommendation) or as a multiple of the upper limit of normal (ULN) of the applied cardiac biomarker (2007 Universal Definition of MI recommendation).

Finally, for the subgroup of trials that started after the publication of the 2007 Universal Definition of MI, we also assessed adherence to the five-type classification scheme proposed.

Survey

To assess underpinnings of gaps in implementing consensus document recommendations, we surveyed principal investigators of all 96 RCTs via an online questionnaire. Because some investigators (n = 17) led more than 1 RCT, 66 individual investigators were surveyed. The full text of question and answer options is provided in Supplementary material online, Appendix 2. To encourage full participation, we sent e-mail reminders and scheduled telephone calls or in-person visits as needed. Responses were received from 61 of 66 investigators (92.4%), corresponding to responses for 91 of 96 RCTs. Every investigator who responded answered all survey questions completely.

Statistical analysis

The contribution of MI to the primary composite endpoint was examined using a random effects meta-analysis to estimate the mean proportion. This approach assigns weights to trials accounting for both (i) sampling error due to finite sample size within trials and (ii) random variability due to heterogeneous populations among trials. Results are presented using mean point estimates [95% confidence intervals (CIs)] of the proportion. This proportion was also evaluated considering RCTs by coronary revascularization group and groupings of follow-up duration (timing of endpoint assessment). The same method was used to identify the revascularization proportion in ACS RCTs and Other RCTs. To address the primary hypotheses, for each recommendation, the level of adherence was calculated as the proportion of RCTs fulfilling that recommendation. In addition, for Recommendation 1, we also explored adherence by coronary revascularization group and by time from the publication of the 2000 consensus document to the start of enrolment. In a sensitivity analysis, we examined the use of the three consensus document recommendations in RCTs with more than 100 but 500 or less patients (Supplementary material online, Appendix 3).

Results

Table 1 shows RCTs with a primary results publication available. Of these 80 RCTs, all but 4 (5%)26,27,30,78 reported the use of a clinical events classification committee blinded to treatment assignment. These MI definitions are detailed in Supplementary material online, Table S1A–C, and grouping of trials by coronary revascularization rate is provided in Supplementary material online, Table S2. The rate of coronary revascularization was 50.6% (95% CI 40.0–61.1) in ACS RCTs and 4.4% (95% CI 2.1–9.0) in Other RCTs. One published trial reported MI event rates only graphically.26

View this table:
Table 1

Characteristics of 80 cardiovascular published randomized clinical trials ordered by the start date of enrolment

Study nameStart datePubYearPatientsMI, nMI, %Follow-up duration
REPLACE 111November 2000200410565075.848 h
PROVE IT TIMI 2212November 20002004416229228.82 yearsa
MATCH13December 2000200475991219.818 months
ESTEEM14January 2001200318838332.76 months
ISAR SWEET15January 200120047013254.21 year
SIRIUS16February 2001200310583220.59 months
PROactive6May 20012005523822320.534.5 monthsa
ADVANCE17June 2001200711 140915.14.3 yearsa
ICTUS18July 20012005120014956.71 year
SYNERGY19August 2001200410 027120785.130 days
FIRE20October 2001200365112488.630 days
REPLACE 221October 20012003601039268.330 days
OnTARGET22November 2001200825 620129130.656 monthsb
PRIMO CABG23January 20022004309925285.430 days
JIKEI HEART24January 2002200730813614.93.1 yearsb
CHOIR25March 2002200614323817.116 monthsb
ALMICAD26April 200220071233NAcNAc24 months
Lee et al.27April 200220056891381.330 days
RACS28April 2002200710042064.5180 days
ExTRACT TIMI 2529October 2002200620 47976734.230 days
HEART 2D30October 20022009111513635.532 monthsa
CHARISMA31October 2002200615 60330127.228 monthsb
Nussmeier et al.32January 20032005167122.040 days
CLARITY TIMI 2833February 20032005349110816.9Index H
ARTS II34February 200320076073844.230 days, 1 year
PASSION35March 200320066191116.71 year
OASIS 536March 2003200620 07852745.79 days
Windecker et al.37April 2003200510123237.29 months
ARISE38June 20032008614424322.924 monthsa
ACTIVE W39June 2003200667065914.81.28 yearsb
PROXIMAL40July 20032007594712.330 days
ACUITY41August 2003200613 81970445.630 days
OASIS 642August 2003200612 09231725.130 days
EASY43October 20032006100510751.230 days
TYPHOON44October 200320062019911.71 year
PRIMO CABG II45July 20042011425454982.730 days
SORT OUT II46August 2004200820989843.418 months
MULTI STRATEGY47October 200420087441548.430 days
MERLIN TIMI 3648October 20042007656047732.930 days
TRITON TIMI 3849November 2004200713 608109576.915 months
EARLY ACS50November 20042009949269076.096 h
BEAUTIFUL51December 2004200810 91742525.419 monthsb
DOORS52January 200520129006266.030 days
SYNTAX53March 2005200918007126.91 year
HORIZONS-AMI54March 2005200836026516.930 days
PRECOMBAT55March 20052011600813.31 year
I CARE56April 2005200814342449.018 monthsd
COSTAR II57May 2005200817005033.38 months
ISAR LEFT MAIN58July 200520096072932.61 year
AIM HIGH10September 20052011341417230.93 yearsa
ISAR REACT 359November 20052008457023861.230 days
CRESCENDO60December 2005201018 69528238.213.8a
CHAMPION-PCI61April 20062009887753494.348 h
CURRENT62June 2006201025 08651447.630 days
RIVAL63June 20062011702112546.830 days
ISAR REACT 464July 20062011172111761.930 days
SPIRIT IV65August 2006201036878245.11 year
CH. PLATFORM66September 20062009502236893.248 h
PLATO67October 2006200918 093109758.412 months
MEND CABG II68November 20062008302324389.030 days
LEADERS69November 2006200817078852.49 months
ZEST70November 20062010264516456.91 year
CORONARY71November 20062012475232868.630 days
PRODIGY72December 2006201220138040.424 months
COMPARE73February 2007201018004371.71 year
REAL LATE74July 2007201027011753.12 years
ECLAT STEMI75July 200720127867477.930 days
ISAR TEST 476September 2007200926039927.61 year
TRA 2°P TIMI 5077September 2007201226 449123756.130 monthsb
MI FREEE78November 20072011585542340.113.1 months
ISAR CABG79November 200720116103027.31 year
TRACER80December 2007201112 944131961.8502 daysb
ISAR TEST 581February 2008201130027719.71 year
RESOLUTE AC82May 2008201022929350.01 year
AIDA STEMI83July 2008201220653425.090 days
ISAR REACT 3A84August 20082010250520992.530 days
ATLAS ACS85November 2008201215 52661361.213 months
PLATINUM86January 2009201115302128.81 year
APPRAISE 287March 20092011739237665.7241 days
Litt et al.88July 20092012137015100.030 days
  • PubYear is the year of publication; MI, n is the absolute number of myocardial infarction within the primary endpoint; MI, % is the proportion of myocardial infarction endpoint events within the primary composite.

  • aMean follow-up duration.

  • bMedian follow-up duration.

  • cThe ALMICAD trial did report the actual number of MI events.

  • dThe I CARE trial was terminated early at 18 months for efficacy.

Contribution of myocardial infarction to the primary endpoint

Using 79 RCTs with primary results provided as MI counts, we estimated that MI contributed 45.3% (95% CI 40.2–50.4) of events in the primary composite outcomes of large cardiovascular RCTs. The proportion that MI contributed to the primary composite endpoint decreased with increasing number of endpoint components (Supplementary material online, Figure S1) and was highest among Interventional RCTs and trials with a follow-up duration of ≤1 month (Table 2).

View this table:
Table 2

Myocardial infarction contribution to the primary endpoint by revascularization group and by follow-up duration

CategoryGroupnMI % (95% CI)
Revascularization rateInterventional4653.2 (45.7–60.5)
ACS1646.8 (37.1–56.8)
Other1724.3 (18.1–31.8)
Follow-up duration≤1 month3164.3 (55.5–72.1)
Between 1 month and 1 year2535.6 (29.0–42.9)
>1 year2332.2 (24.8–40.6)
  • Proportion of MIs in the primary composite endpoint was calculated using random effects estimates with 95% CIs.

  • CI, confidence interval; MI, myocardial infarction.

Proportion of RCTs referencing published myocardial infarction definition documents

Among 96 RCTs with a primary results and/or design paper, 7 (7.3%) referenced the ESC/ACC 2000 document for endpoint MI definition and 18 (18.8%) referenced the Universal Definition of MI 2007 publication. In addition, 34 (35.4%) cited another consensus document (27 referenced the Academic Research Consortium definition89,90; 7 referenced the ACC Key Standards document,91 both of which referred to the ESC/ACC 2000 document). Fourteen (14.6%) RCTs published no MI definition.16,17,22,24,31,39,46,51,58,74,9295 After contacting investigators, we received and examined 93 of 96 endpoint MI definitions (96.9%). Table 3 presents an overview of key features of these definitions in the 10 largest RCTs, which accounted for 199 448 patients (32.8% of all patients in RCTs we evaluated). In these trials, both the threshold for spontaneous MI (1× or 2× ULN) and the biomarker used (CK-MB or troponin) varied. Definitions of re-infarction in RCTs that enrolled patients with acute MI also varied (Supplementary material online, Appendix 4 and Table S3).

View this table:
Table 3

Key features of the myocardial infarction definition in the 10 largest randomized clinical trials

TrialSpontaneous MIPCI-related MICABG-related MI
Preferred biomarkerFold elevation above ULNPreferred biomarkerFold elevation above ULNPreferred biomarkerFold elevation above ULN
TRA 2°P TIMI 5077acTn1CK-MB3CK-MB/cTn5
ONTARGET22bNANANANANANA
CURRENT62CK-MB/cTn2CK-MB3CK-MB5 (w/ Qw) or 10 (w/o Qw)
ExTRACT TIMI 2529CK-MB/cTn1CK-MB3CK-MB5 (w/ Qw) or 10 (w/o Qw)
OASIS 536CK-MB2CK-MB3CK-MB5 or Qw
CRESCENDO60CK-MB/cTn/CK2/1/2CK-MB3CK-MB5 (w/ Qw) or 10 (w/o Qw)
PLATO67CK-MB/cTn1CK-MB3CK-MB5 (w/ Qw) or 10 (w/o Qw)
CHARISMA31cCK-MB/cTn2/1
ATLAS85dCK-MB/cTn1CK-MB3CK-MB5 (w/ com) or 10 (w/o com)
ACUITY96CK-MB/cTn1CK-MB3CK-MB5 (w/ Qw) or 10 (w/o Qw)
  • aTRA 2°P TIMI 50 definition of CABG-related MI required additional complications beyond biomarker criterion (new Q-waves or new left-bundle branch block; angiographically documented graft or native coronary artery occlusion; or imaging evidence of new loss or viable myocardium).

  • bNo MI definition published for ONTARGET.

  • cThe CHARISMA trial did not use a separate MI definition for procedure-related MI.

  • dATLAS definition of CABG-related MI required additional complications beyond biomarker criterion (new Q-waves or new left-bundle branch block; angiographically documented graft or native coronary artery occlusion; or imaging evidence of new loss or viable myocardium) for CK-MB elevation between 5× and 10× the ULN.

  • CABG, coronary artery bypass grafting; CK-MB, creatine-kinase-MB; com, additional complications; cTn, cardiac troponin; MI, myocardial infarction; NA, not available; PCI, percutaneous coronary intervention; Qw, Q-waves; ULN, upper limit of normal; w/, with; w/o, without.

Recommendation 1: use of troponin for myocardial infarction diagnosis and myocardial infarction decision limit provided

Among 93 RCTs with an MI definition provided, troponin was used to define the MI endpoint in 53 (57%). This proportion was 66.7% (n = 34) among 51 RCTs that referenced any consensus document for MI definition. Three of 93 RCTs used troponin to define endpoint MI only if CK-MB was unavailable,36,62,88 6 used troponin to define procedural MI, and 7 specified the 99th percentile as the MI decision limit. All other RCTs (n = 40) used CK-MB or total CK to define endpoint MI. The use of troponin to define endpoint MI by revascularization group and by time between ESC/ACC 2000 publication and the start of RCTs enrolment is shown in Figure 2A and B, respectively. Among trial types, adherence to Recommendation 1 was lowest among Interventional RCTs, and by time from publication to enrolment start, the highest use of troponin occurred among RCTs that began >74 months after the publication of the ESC/ACC 2000 document.

Figure 2

(A) Troponin use for endpoint myocardial infarction definition by coronary revascularization group. (B) Troponin use for endpoint myocardial infarction definition by time from the publication of 2000 ESC/ACC MI document to the start of trial enrolment.

Recommendation 2: separate reporting of spontaneous myocardial infarction and myocardial infarction related to surgical or percutaneous coronary revascularization procedures

Three trials (3.7%) reported spontaneous and procedural MI separately.18,77,80 The ICTUS trial, which adhered to the 2000 ESC/ACC MI redefinition, compared an early invasive strategy with a selective invasive strategy in patients with non-ST-segment elevation MI.18 Overall, 149 MIs occurred; 32.9% were spontaneous and 67.1% were procedure-related. The TRACER80 and TRA 2°P TIMI 5077 trials reported MI using the five-type classification scheme proposed by the 2007 Universal Definition of MI.

Recommendation 3: report of comparison of size of myocardial infarction between treatment and control groups in addition to the presence/absence of myocardial infarction

No trials reported a comparison of infarct size between treatment and control groups, using area under the biomarker curve. Five trials compared MI rates between treatment and control groups, using size thresholds (e.g. 3×–5× ULN, 5×–10× ULN, or >10× ULN), but did not provide actual peak levels.11,20,21,40,68

Similar results for the use of the three recommendations explored were observed in sensitivity analyses of RCTs with more than 100 but 500 or less patients (Supplementary material online, Table S4).

2007 Universal Definition of Myocardial Infarction adherence

Of the 96 trials included, 25 (26.0%) started enrolment after the publication of the 2007 Universal Definition of MI. Of these, 11 (44.0%) referenced this definition in the design or primary results paper. Troponin defined endpoint MI in 19 (76.0%). Nine (36.0%) used the five-type classification scheme to define endpoint MI, but seven of these nine adapted this classification, typically for PCI-related (type 4a) MI definition, by using CK-MB instead of troponin.

Survey results

A survey was administered to the investigators between April 2011 and July 2012. Of 61 unique investigators responding to the survey, 54 (88.5%) responded that the use of a standard MI definition in RCTs was important, but 40 of those 54 (74.1%) also indicated that the ESC/ACC/AHA/WHF (World Heart Federation) task force definition created challenges related to assay variability and definition of re-MI and PCI-related MI. Overall, 54 investigators (88.5%) said troponin should be used to define MI in RCTs, but 18 added that they would not use high-sensitivity troponin assays. Of these 54, 38 (70.4%) specified their preference for troponin to define spontaneous but not procedural MI. The most common reasons for not favouring troponin in the procedural setting were as follows: (i) the recommended diagnostic threshold for troponin was too low (n = 23, 60.5%); (ii) a belief that there was a lack of clinical relevance of asymptomatic troponin elevations after procedures (n = 10, 26.3%); and (iii) these elevations are a marker of atherosclerosis burden but have no independent relationship with mortality (n = 10, 26.3%).

For each of the 96 trials in our analysis, we asked the principal investigator to specify whether a standardized MI definition was used; we obtained responses for 91 trials. A standard MI definition was reported in 49 trials (53.8%); 22 reported using the universal MI definition.

Discussion

In this large and broad cross-section of contemporary RCTs in cardiovascular disease, we estimated that MI contributes an average of 45.3% of events in primary endpoint composites. Despite the frequency and importance of MI as an endpoint component in RCTs that establish the evidence for the use of therapies in cardiovascular practice, MI definition was heterogeneous and implementation of recent consensus recommendations for defining MI was low. Overall, only 57.0% of RCTs used troponin to define endpoint MI, whereas 43.0% of trials used the less-specific CK-MB or total CK. This trend was especially evident among RCTs studying revascularization, of which only six trials specified troponin to define procedure-related MI. One in seven trials failed to publish any criteria for MI definition. Although the use of troponin was more frequent among RCTs in which enrolment started >74 months after the publication of the 2000 ESC/ACC consensus MI definition document, overall these findings suggest a lack of standardized implementation of one of the most important endpoint definitions in cardiovascular RCTs.

Use of troponin to define endpoint myocardial infarction

Owing to nearly absolute myocardial specificity, the use of troponins I and T was emphasized in the ESC/ACC 2000 redefinition of MI consensus statement1 and reinforced in the 2007 Universal Definition of MI document.2 Thus, it is noteworthy that slightly more than half of cardiovascular RCTs we examined used troponin to define endpoint MI. Even among RCTs that referenced a consensus document, only 66.7% used troponin to define endpoint MI and only 7 specified the recommended 99th percentile as the MI decision limit. These findings are in sharp contrast to the rapid uptake of troponins to define MI in clinical practice.97

Because failure to implement a standard approach to define MI in RCTs impairs systematic comparison of results across trials, our findings underscore the need to (i) increase awareness of consensus documents; (ii) overcome impediments to implementation of consensus recommendations in RCTs; and (iii) develop uniform standards for determining and reporting endpoint MI in RCTs and defining the role of troponin.

Spontaneous and procedural myocardial infarction reporting and infarct size

Remarkably, only three of 80 RCTs separately reported spontaneous and procedural MI, whereas five reported MI rates between treatment arms, using threshold biomarker size, and no RCT provided actual peak biomarker levels. Although there is general agreement that procedure-related myocardial necrosis has some prognostic implication, it remains controversial as to how best to weigh its impact relative to spontaneous MI.96 Indeed, separate reporting was a key consideration driving the five-group classification in the 2007 and 2012 MI definition documents.2,3 Finally, it may be useful to compare the effect of treatments using a more sensitive, continuous measure of myonecrosis, such as the peak biomarker value, in addition to the presence/absence of MI. Without consistently defining and reporting MI, it will remain challenging to compare studied populations and reported results across trials.

Understanding gaps in the implementation of consensus document recommendations

Implementation of consensus recommendations for MI definition and reporting is likely time-dependent. The use of troponin to define endpoint MI was higher in RCTs that began >74 months after ESC/ACC MI 2000 redefinition, compared with earlier trials. This may reflect increased acceptance over time of troponin as the preferred biomarker to define MI. It may also reflect that establishing endpoint definitions in the design phase of clinical trials often precedes the start of enrolment. However, even at 34–74 months (about 3–6 years after the 2000 redefinition document was published), only 48.5% of trials used troponin to define endpoint MI: this suggests it was unlikely the only factor in our findings.

Troponin was used least often to define MI in Interventional RCTs. Although the association of CK-MB elevation (mainly defined as >3× ULN) with death after coronary revascularization has been demonstrated, whether this relationship exists for troponin and at what level of elevation above the 99th percentile is less certain.98100 This has prompted debate about the use of troponin to define procedure-related MI.101 Our investigator survey revealed a reluctance to use troponin to define procedure-related MI; only 16 of 61 investigators favoured troponin to define procedure-related MI. The most common reason cited (60.5%) was the low recommended diagnostic threshold (i.e. 3× ULN for PCI-related MI). It is reassuring that the most recent version of the Universal Definition of MI has addressed this concern by raising the recommended threshold for peri-PCI MI from 3× ULN to 5× ULN, although the task force recognizes that these thresholds are arbitrary.3

About one-fourth of respondents were concerned about clinical relevance, including that asymptomatic troponin elevation post-procedure is not clinically important or that it is a marker of atherosclerosis burden but has no independent relationship with mortality. Concerns have also been raised that interpreting peri-procedural troponin elevations is challenging and may be less relevant among patients with pre-procedural elevation.102,103 Increasingly sensitive troponin assays will add more complexity to this debate by increasing the proportion of patients with elevated baseline troponin levels. However, a recent analysis of pooled data from more than 10 000 clinical trial patients with non-ST-segment elevation ACS that assessed troponin trends before and at the time of PCI found that (i) in using pre-procedural troponin trends to identify patients with stable or falling troponin levels, 57% of patients could be assessed for peri-PCI MI, and (ii) if troponin levels were stable or falling before PCI, then a new troponin re-elevation post-PCI was associated with worse outcomes, even after accounting for pre-procedural troponin levels.104 The importance of assessing cardiac marker trends before PCI in adjudicating peri-PCI MI was also emphasized by the recent 2012 Universal Definition of MI.2,3 Thus, enhanced use of a standard MI definition that incorporates troponin will likely be most successful by iteratively addressing investigator concerns. Only reassessment of the degree of implementation of the 2012 Universal Definition of MI in several years will confirm whether the changes (which are consistent with our survey findings of reasons for failure of significant implementation to date) were successful in increasing standardized MI definition in clinical trials. We believe our survey not only has provided important insights into investigator concerns about implementing standard MI definitions in clinical trials, but it can also serve as the framework for the future reassessments of implementation.

Concern about the relevance of peri-procedural troponin elevations and MI definition thresholds does not entirely explain the low implementation of the Universal Definition of MI. Even ACS RCTs and Other RCTs incompletely incorporated ESC/ACC consensus recommendations for troponin use to define MI, particularly at the 99th percentile. This may reflect ongoing concerns about precision of some assays at the 99th percentile. Even if assays are very precise at the 99th percentile, it is also unknown whether smaller MIs that may be detected and that generally are associated with lower risk for adverse outcomes would affect the ability to detect overall treatment differences in RCTs. New generation troponin assays with even higher sensitivity will likely add more uncertainty, as recently acknowledged by the Third Universal Definition of MI.3 It is noteworthy that more than one-quarter of investigators surveyed favoured troponin to define MI, but not the use of high-sensitivity assays due to potential ‘noise’ or non-clinically relevant MI event detection. Although these concerns could lead to troponin use at a cutoff above the 99th percentile, they would not solely explain overall low rates of incorporation of troponin testing into MI endpoint definitions. To this point, 11.5% of investigators felt troponin should not be used at all to define MI in RCTs.

Future directions

The low use of consensus document recommendations we observed identifies a pressing need to implement standard definition and reporting of MI endpoints in cardiovascular RCTs. We believe that the reporting criteria developed by the ACC/AHA/ESC/WHF Writing Group for Redefinition of Myocardial Infarction should be required as part of the standardized CONSORT (Consolidated Standards of Reporting Trials) reporting system for clinical trials now adopted by major medical journals.105 Using this type of standardized reporting system, such as that recently developed for bleeding,106 would make it easier for both clinicians and clinical trialists to understand findings of RCTs. Moreover, we suggest that the redefinition of MI writing group consider modifications that address concerns and perspectives among investigators that were identified in our survey. Other specific efforts and measures that may facilitate uniform definition and reporting include the following: (i) requirements from regulatory authorities for standard MI definition in RCTs in which MI is a key component of the primary endpoint, including appropriate distinction of MI types; and (ii) requirements by journal editors that actual data for necrosis markers used to determine MI endpoints in RCTs be submitted as a supplementary file with each primary results manuscript so that they are publically available. This would provide an important opportunity for investigators, regulators, and critics to examine varying definitions of MI (e.g. varying troponin thresholds for MI) and how they affect trial results.

To successfully implement standard definition and reporting of MI in RCTs, logistical issues, including pros and cons of local laboratory vs. core laboratory troponin measurement, must be addressed. In particular, if local measurements are used, how best to incorporate and account for the wide variety and lack of standardization of available troponin assays and their varying performance characteristics must be resolved.

Limitations and strengths

We acknowledge some limitations of our systematic review but also would point to its strengths. Whereas one author (S.L.) screened and abstracted all the data, we implemented rigorous quality checks to ensure complete and consistent data abstraction. Also, the inclusion/exclusion criteria and parameters assessed were stringent, reducing the likelihood of bias that could substantially alter the results. These contentions are supported by our sensitivity analysis that examined smaller RCTs, revealing similar findings. Furthermore, we limited our search to primary results and/or design papers. RCTs may report the incidence of spontaneous and procedural MI in a secondary publication.107 However, because the consensus recommendations were designed for use in primary MI endpoint determination, we felt that a stringent approach to addressing their implementation for primary results presentation was important. More importantly, reporting results according to these key recommendations in secondary publications, which are more difficult to identify and often occur well after the primary results are published, creates unnecessary obstacles to understanding the primary results. Finally, the quality of data management and monitoring, which is usually not publicly or readily available, was not analysed in our systematic review.

Conclusions

Although MI contributes significantly to primary outcome measures in contemporary RCTs, low implementation of the 2000 ESC/ACC MI redefinition and 2007 Universal Definition of MI consensus recommendations was evident. This was particularly true regarding the use of troponin to define MI. Given the importance of MI as a metric in clinical practice, clinical trials, and epidemiological studies, it is necessary to better understand this failure and to create appropriate strategies for uniform implementation of recommendations. Our survey, from a broad cross-section of contemporary RCTs, underscores the need for further investigation to establish the most appropriate and clinically meaningful definition of peri-procedural MI.

Funding

This work was supported by the investigators and by the National Institutes of Health (T32HL079896 to P.J.S.). No other commercial, foundation, philanthropic, or government funding sources were used.

Conflict of interest: S.L. and P.J.S. report no conflicts of interest with the submitted work. P.W.A.: Consultancy—F. Hoffmann-LaRoche Ltd., Axio/Orexigen, Eli Lilly & Co. Merck & Co., Inc. in conjunction with DCRI. Grant/grant pending—Boehringer Ingelheim, F. Hoffman-LaRoche Ltd. & Sanofi-Aventis Canada, Inc.; Sanofi-Aventis Canada Inc.; Scios, Inc., Ortho Biotech, Johnson & Johnson, Jansen Ortho in conjunction with DCRI; GlaxoSmithKline; Amylin Pharmaceuticals, Inc. in conjunction with DCRI; Merck & Co., Inc. in conjunction with DCRI. Payment for development of educational presentations—AstraZeneca and Eli Lilly & Co. L.K.N.: Board membership—Society of Chest Pain Centers. Consultancy—Amgen, AstraZeneca, Daiichi Sankyo, Eli Lilly & Co., Genentech, Johnson & Johnson, Novartis. Grants/grants pending—Amylin, AstraZeneca, Bristol-Myers Squibb, diaDexus, GlaxoSmith Kline, Merck & Co., Inc. MURDOCK Study, NHLBI, Regado Biosciences, Roche. Payment for lectures including service on speakers bureaus—Johnson & Johnson, American Diabetes Association. Travel/accommodations/meeting expenses unrelated to activities listed—AHA, Society of Chest Pain Centers. E.M.O.: Consultancy—AstraZeneca, Boehringer Ingelheim, Bristol-Myers Squibb, Gilead Sciences Inc., Janssen Pharmaceuticals Inc., LipoScience, Merck & Co., Inc. Pozen, Roche, Sanofi-Aventis, The Medicines Company, WebMD. Grants/grants pending—Daiichi Sankyo, Daiichi Sankyo, Maquet.

Acknowledgements

The authors wish to thank Morgan deBlecourt for editorial assistance in the preparation of the manuscript for submission. Finally, the authors are grateful to the investigators who participated in the survey for their important feedback and contributions.

References

View Abstract