Author + information
- Received October 14, 2010
- Revision received December 23, 2010
- Accepted January 20, 2011
- Published online May 1, 2011.
- Donald E. Cutlip, MD⁎,†,⁎ (, )
- Gaku Nakazawa, MD§,
- Mitchell W. Krucoff, MD∥,
- Marc Vorpahl, MD‡,
- Roxana Mehran, MD¶,
- Aloke V. Finn, MD§,
- Pascal Vranckx, MD#,
- Carey Kimmelstiel, MD†,‡,
- Clifford Berger, MD†,
- John L. Petersen, MD⁎⁎,
- Theresa Palabrica, MD† and
- Renu Virmani, MD§,††
- ↵⁎Reprint requests and correspondence:
Dr. Donald E. Cutlip, Cardiovascular Division, Beth Israel Deaconess Medical Center, 330 Brookline Avenue, Boston, Massachusetts 02215
Objectives This study sought to validate the sensitivity and specificity of the Academic Research Consortium's (ARC) classification of stent thrombosis.
Background Classification of stent thrombosis according to ARC criteria has become widely accepted. The criteria have not been validated against an autopsy standard.
Methods An autopsy registry of 139 subjects with prior coronary stenting underwent detailed histopathological analysis to assess for stent thrombosis. Based on clinical data only, cases were adjudicated according to ARC stent thrombosis criteria, including a proposed modification of the possible classification to include death beyond 30 days due only to sudden death or acute ischemia.
Results Autopsy results confirmed 51 cases as positive and 88 as negative for stent thrombosis. Clinical adjudication classified 105 cases as definite (10), probable (31), or possible (64) ARC stent thrombosis. Specificity was high for definite (99%) and definite plus probable (83%) criteria, but sensitivity was poor at 18% and 51%, respectively. Including the possible cases improved sensitivity to 92% but reduced specificity to 34% (58 false positives). The modified possible criteria eliminated 13 false positive cases (specificity = 49%) and was the best approximation of a hypothetical gold standard in a sensitivity analysis if late death represented at least 20% of all stent thrombosis cases.
Conclusions In a selected autopsy sample, restricting ARC stent thrombosis to definite or definite plus probable criteria results in substantial under-reporting of confirmed cases. Inclusion of a modified possible classification may provide the best estimate of late and very late stent thrombosis rates.
Stent thrombosis is a catastrophic complication of coronary stenting, presenting as sudden death or nonfatal myocardial infarction (MI) in almost all cases (1,2). After a controversy surrounding possible inconsistent and incomplete reporting of stent thrombosis in drug-eluting stent (DES) trials, especially those events occurring more than 1 year after stenting (3), the Academic Research Consortium (ARC) proposed a standardized classification (4). This classification standardized the inconsistencies of previous trial protocol definitions, by including uniform criteria for angiographic confirmation, allowing for inclusion of probable events presenting as new acute myocardial infarction without angiographic confirmation, and maintaining the principle of intention-to-treat by not excluding patients with interval target lesion revascularization. Concern remained, however, surrounding those events occurring beyond 1 year that may present as unexplained death. Although these events were considered as possible stent thrombosis by ARC, it became recognized that an increasing proportion of deaths remain unexplained over time, leading to a potential exaggeration of the rates of very late stent thrombosis as well as possible dilution of a signal for differences between groups if events other than stent thrombosis were included in a substantial number (5,6). For these reasons, most studies have restricted the definition of cases to those meeting definite or probable criteria, and thereby have not accounted for the vast majority of cases presenting as late death (7,8).
Although the ARC classification has been accepted and recommended as a standard for determining stent thrombosis, it has not been validated for the adjudication of death events. We sought to apply the ARC classification for adjudication of stent thrombosis based on symptomatology immediately before death and other clinical data among subjects who had died after prior stent implantation and had been selected for detailed autopsy analysis of the stented coronary segments.
Patients and lesions
From the CVPath registry of autopsies with coronary stents, 139 consecutive patients referred between June 2002 and March 2008 and who received at least 1 DES were examined. The study cohort included only sirolimus-eluting or paclitaxel-eluting DES. Cases were routinely referred to CVPath Institute, Inc. (Gaithersburg, Maryland) from the local medical examiners when the patients had received coronary stents. Cases from other institutions were referred to CVPath Institute, Inc. when the pathologist was unable to section the coronary artery with the stent intact. Causes of death were reported as stent-related cardiac deaths, nonstent-related cardiac deaths, and non-cardiac deaths as previously defined (9). Briefly, cardiac death was determined after a complete autopsy, including examination of the myocardium. The presence of an occlusive luminal thrombus or nonocclusive thrombus with distal embolization was considered a stent-related cardiac death. The presence of acute thrombus was defined as a platelet-rich thrombus that occupied >30% of the cross-sectional area of the lumen. Nonstent-related cardiac deaths were defined if the stent was patent without evidence of thrombus or restenosis (luminal stenosis <75% cross-sectional area) but other cardiac causes were likely, including other coronary artery disease with severe narrowing (>75% cross-sectional area stenosis) of 1 or more major coronary arteries, presence of prior myocardial infarction and history consistent with arrhythmia, or other myocardial or valvular heart disease that was deemed responsible for the death.
Available clinical data including any details related to the previous stent procedure and relevant cardiac and other medical history leading up to the time of death were compiled for review by a clinical events committee (CEC). These data included pre-morbid electrocardiography, hospital discharge summary, and report of coronary angiography if performed, but the CEC did not have access to actual angiograms. The CEC was not aware of any information from the autopsy study.
The clinical data summary and a standard adjudication case report form were sent to each of 2 reviewers. The reviewers were asked to determine cause of death as cardiac versus noncardiac, and in cases of cardiac death, they were asked if the clinical data were consistent with ARC definite, probable, or possible stent thrombosis. If there was concordance for all aspects of the adjudication, this result was entered into the event database. If there was discordance, the event was reviewed by the full CEC panel for consensus and this decision was then entered into the database. All adjudications were patient-based. If >1 stent had been implanted, then stent thrombosis for any stent was considered as a confirmed case. If >1 stent thrombosis was confirmed for a patient, then the case with the highest ARC level of certainty was adjudicated.
Adjudicated endpoint definitions
The CEC determined cardiac death as any death related to cardiac diagnosis or occurring because of a complication from a cardiac procedure. Any other death for which a clear noncardiac cause was not identified was also considered as cardiac. MI was defined as any event with acute ischemic signs or symptoms and associated with any elevation of creatine kinase-myocardial band or troponin. MI was attributed to the target vessel unless a non-target vessel could be clearly implicated by electrocardiography or angiography. Stent thrombosis was defined according to the ARC classification except that cases adjudicated as possible were also further subclassified as either: 1) sudden cardiac death or acute ischemia likely; or 2) unexplained death, acute ischemia unlikely (Table 1). A new category termed “ARC modified possible” included only those cases classified as sudden death or acute ischemia likely. ARC definite stent thrombosis required angiographic report of thrombus or occlusion and clinical report consistent with acute coronary syndrome.
Stent thrombosis was classified according to CEC adjudication as definite, definite or probable, any ARC (definite, probable, or possible), or any ARC modified (definite, probable, or modified possible). The autopsy confirmation of stent thrombosis was considered as a true positive diagnosis and the absence of stent thrombosis by autopsy was considered as a true negative. Sensitivity was calculated for each category of stent thrombosis as the number of autopsy-confirmed cases that were adjudicated as confirmed cases divided by total number of autopsy-confirmed cases. Specificity was calculated for each category of stent thrombosis as the number of cases that were not confirmed as stent thrombosis by autopsy and were adjudicated as not being stent thrombosis divided by the total number of cases not confirmed as stent thrombosis by autopsy.
The time-to-death was calculated based on time from most recent stent procedure. The time-to-stent-thrombosis was calculated as time from most recent stent procedure unless prior specific stent other than most recent was identified as thrombotic. Time-to-stent-thrombosis could not be determined specifically for 3 cases that were presumed to be beyond 30 days and within 1 year, that is, late stent thrombosis. Results for time-to-event are presented as median and interquartile range.
We performed a simulated sensitivity analysis based on a hypothetical sample of 10,000 patients with variable proportion of stent thrombosis events presenting as late death, using the calculated sensitivity and specificity for possible and modified possible criteria from this study when stent thrombosis presented as late death and assuming 90% sensitivity and 99% specificity for definite and probable criteria when stent thrombosis events did not present as late death.
There were 139 cases referred for autopsy. Of these, cardiac death was confirmed in 111 (80%) cases. The median time-to-cardiac-death was 115 days (interquartile range 6 to 366). Among the cardiac deaths, stent thrombosis was confirmed in 51 (45%) cases, including 37 occlusive and 13 nonocclusive thrombi. Other causes of death are shown in Table 2. Of 12 subjects with analyzable stents in >1 vessel, thrombosis was present in >1 location in 8 subjects. There were 3 stent thrombosis cases that had DES and bare-metal stents in different locations with thrombosis of DES only in 1 subject, of bare-metal stent only in 1 subject, and of both DES and bare-metal stent in the other. The median time-to-stent-thrombosis was 51 days (interquartile range: 12 to 260 days). The timing of confirmed stent thrombosis was early in 21 cases, late in 22 cases, and very late in 8 cases. Stent thrombosis involved the left main in 3 cases, left anterior descending artery in 28 cases, circumflex artery in 13 cases, right coronary artery in 12 cases, left internal mammary artery bypass graft in 1 case, and a saphenous vein bypass graft in 1 case.
Clinical event committee adjudication
The CEC adjudicated any ARC stent thrombosis in 105 (76%) patients from this autopsy population. Sensitivity and specificity for each ARC level of classification according to timing of suspected stent thrombosis are shown in Table 3. ARC definite or probable stent thrombosis identified stent thrombosis associated with death slightly over 50% of the time. This was substantially improved (92%) when the possible classification was also included, but at a cost of reducing specificity from nearly 100% for ARC definite or 83% for definite or probable stent thrombosis to 34% for any ARC stent thrombosis. The 4 cases that were not identified by any ARC classification are described in Table 4.
Application of the modified possible criteria resulted in improved specificity over the current ARC possible criteria, with 13 fewer false positive cases, but did fail to detect stent thrombosis in an additional 2 cases. This included 1 subject who died 353 days after stenting with reported cause of sepsis in the clinical record and 1 subject who died 979 days after stenting with clinical report of being found after an alcohol-related fall and apparent head trauma. Despite improved specificity, ARC modified possible criteria continued to overdiagnose stent thrombosis in 45 of 88 cases (Table 5).
Sensitivity analysis based on prevalence of stent thrombosis presenting as late death
The impact of the sensitivity and specificity of ARC-adjudicated late stent thrombosis depends on the prevalence of late death as a presentation for stent thrombosis. Table 6 depicts the results of ARC definite or probable stent thrombosis, any ARC stent thrombosis, and any ARC-modified stent thrombosis compared with a hypothetical gold standard as the proportion of stent thrombosis cases presenting as late death varies. When the frequency of stent thrombosis presenting as late death exceeds 20% of all stent thrombosis events, the ARC-modified possible criteria provide a better estimate of the gold standard rate.
This is the first report to test the validity of the ARC stent thrombosis definitions in a relatively large number of autopsy cases. The results demonstrate that the ARC definition can be applied by experienced investigators to accurately identify stent thrombosis in this setting, but that there is large variability in sensitivity and specificity depending on ARC criteria being applied. Furthermore, the accuracy of the adjudicated rates depends on the prevalence of late death as the presenting manifestation of stent thrombosis.
At the height of the DES stent thrombosis controversy, the ARC definitions were welcomed by regulatory agencies and clinical trialists as a standardized method for defining stent thrombosis across clinical trials and various stent types. Such standardization has value for assessment of potential safety differences between new devices and other therapies compared with current treatments or alternative strategies. Despite such standardization, however, defining an event can only add value if there is adequate sensitivity to detect true and likely very weak safety signals while avoiding the introduction of noise that might dilute the signal.
The evaluation of any clinical definition of stent thrombosis is hampered by the unavailability of a gold standard—pathological examination of the stent segment—in most cases. In our study, detailed pathological evaluation was performed by a single and experienced pathology group and provides a gold standard against which to test each of the ARC criteria. Notably, the highly specific criteria for ARC definite stent thrombosis detected only 1 case that was not confirmed by autopsy, but was only present in 18% of autopsy-confirmed cases. Likewise, the most widely used ARC classification of definite or probable was falsely positive in only 17% of unconfirmed cases but present in only 51% of confirmed cases. The lack of sensitivity for definite and definite or probable criteria was related to the failure to diagnose stent thrombosis presenting as or resulting in deaths beyond 30 days. These late deaths represented 30 of the 51 (59%) confirmed stent thrombosis cases. Concern for failure to account for late and very late stent thrombosis presenting as death was the basis for including the possible category in the ARC definition.
Indeed, when the cases of late death that were adjudicated as ARC possible stent thrombosis were included, the sensitivity improved to >90% in our study, but unfortunately resulted in a false positive diagnosis in 58 of 88 (66%) subjects that were not confirmed by autopsy. Such a false positive rate more than doubles the estimate of stent thrombosis and may mask any signal that might be detectable by more specific measures.
We assessed whether the signal-to-noise ratio might be improved by inclusion of only those late deaths that suggested a sudden ischemic event. Limiting the ARC possible criteria to sudden cardiac death or death associated with probable acute ischemia substantially improved specificity with 13 fewer false positives and only reduced sensitivity by 2 cases. Nevertheless, there remained 45 of 88 (51%) false positive diagnoses among this sample.
There are several questions raised by this analysis and limitations to be considered. There is obvious selection bias inherent in the autopsy sample. Thus, our results only apply to those cases with fatal presentation or death early after suspected stent thrombosis. Despite high fatality rates, clinical studies suggest that most cases of stent thrombosis present with nonfatal acute ischemia (1,2). It is also possible that more complete access to clinical data, including angiography and electrocardiography, in these cases would improve positive and negative diagnostic accuracy. As noted in Tables 4 and 5, inadequate clinical data inherent in the referral nature of the autopsy cases may have also reduced both sensitivity and specificity of the ARC definitions. This points to the importance of careful clinical follow-up and complete ascertainment of clinical data for the duration of any clinical study to minimize the uncertainty in determining cause of death.
Despite these limitations, the results demonstrate that late or very late stent thrombosis presenting as death is not uncommon and that ignoring these events by only reporting ARC definite or probable stent thrombosis will clearly underestimate the true frequency of stent thrombosis. Because most cases likely present with acute ischemia and have some level of clinical evaluation rather than unexplained late death, this underestimation is clearly less than suggested by the 51% sensitivity in our study. Nevertheless, as demonstrated by sensitivity analysis, as the proportion of stent thrombosis presenting as late deaths increases, as might be expected during longer-term follow-up, the accuracy of the definite or probable classification decreases. It appears that a modification of the ARC possible category as was used in our study offers improved diagnostic accuracy in this setting.
Other modifications to the application of the ARC criteria may also improve accuracy. As was suggested previously, removing the inclusion of unexplained death within 30 days after stenting for acute MI may improve diagnostic accuracy of the ARC probable category (4). In our study, 4 of 13 false positive probable ARC stent thromboses were related to deaths within 30 deaths after MI.
Finally, our results point to the critical importance of obtaining autopsy in clinical studies in which accurate definition of stent thrombosis is a major determinant of safety. Furthermore, protocol definitions must include the potential to classify those events presenting as death beyond 30 days. Until there are more accurate clinical criteria available, we believe the preferred reporting of the ARC classification should include the modified possible category.
Dr. Cutlip is a principal investigator for Medtronic. Dr. Krucoff reports receiving consulting fees and grants via DCRI. Dr. Mehran is on the advisory board for Abbott Vascular, AstraZeneca, Ortho-McNeil, Regado Biosciences, and she received institutional research grants from Bristol-Myers Squibb/sanofi-aventis. Dr. Finn reports receiving research grants from Medtronic and St. Jude Medical, and is a consultant for Abbott and Cordis. Dr. Petersen reports research relationships for Abbott Vascular, Boston Scientific, and InfraReDx Inc. Dr. Virmani is a consultant for Medtronic, Abbott Vascular, Terumo Corporation, and Atrium Medical Corporation. All other authors report that they have no relationships to disclose.
- Abbreviations and Acronyms
- Academic Research Consortium
- clinical events committee
- drug-eluting stent(s)
- myocardial infarction
- Received October 14, 2010.
- Revision received December 23, 2010.
- Accepted January 20, 2011.
- American College of Cardiology Foundation
- Cutlip D.E.,
- Baim D.S.,
- Ho K.K.,
- et al.
- Camenzind E.,
- Steg P.G.,
- Wijns W.
- Cutlip D.E.,
- Windecker S.,
- Mehran R.,
- et al.,
- Academic Research Consortium
- Vranckx P.,
- Kint P.P.,
- Morel M.A.,
- Van Es G.A.,
- Serruys P.W.,
- Cutlip D.E.
- Weisz G.,
- Leon M.B.,
- Holmes D.R. Jr..,
- et al.
- Ellis S.G.,
- Stone G.W.,
- Cox D.A.,
- et al.,
- TAXU IV Investigators
- Joner M.,
- Finn A.V.,
- Farb A.,
- et al.