Author + information
- Received November 18, 2016
- Revision received February 22, 2017
- Accepted March 8, 2017
- Published online June 5, 2017.
- Stephen G. Ellis, MDa,∗ (, )
- M. Nicholas Burke, MDb,
- M. Bilal Murad, MDc,
- John J. Graham, MDd,
- Ramy Badawi, MDe,
- Catelin Toma, MDf,
- Henry Meltser, MDg,
- Ravi Nair, MDa,
- Chris Buller, MDh,
- Patrick L. Whitlow, MDa,∗,
- CAPS Group
- aDepartment of Cardiovascular Medicine, Cleveland Clinic, Cleveland, Ohio
- bMinneapolis Heart Institute, Minneapolis, Minnesota
- cUnited Heart and Vascular Center, St. Paul, Minnesota
- dToronto Heart Center, Toronto, Ontario, Canada
- eQueens Heart, Honolulu, Hawaii
- fUniversity of Pittsburgh Medical Center Heart and Vascular Institute, Pittsburgh, Pennsylvania
- gKenmore Mercy Hospital, Buffalo, New York
- hSt. Michaels Hospital, Toronto, Ontario, Canada
- ↵∗Address for correspondence:
Dr. Stephen G. Ellis, Cleveland Clinic, 9500 Euclid Avenue, J2-3 Cleveland, Ohio 44195.
Objectives The aim of this study was to develop a hybrid approach–specific model to predict chronic total coronary artery occlusion (CTO) percutaneous coronary intervention success, useful for experienced but not ultra-high-volume operators.
Background CTO percutaneous coronary intervention success rates vary widely and have improved with the “hybrid approach,” but current predictive models for success have major limitations.
Methods Data were obtained from consecutively attempted patients from 7 clinical sites (9 operators, mean annual CTO volume 61 ± 17 cases). Angiographic analysis of 21 lesion variables was performed centrally. Statistical modeling was performed on a randomly designated training group and tested in a separate validation cohort. The primary outcome of interest was technical success.
Results A total of 436 patients (456 lesions) met entry criteria. Twenty-five percent of lesions had prior failed percutaneous coronary interventions at the site. The right coronary artery was the most common location (56.4%), and mean occlusion length was 24 ± 20 mm. The initial approach was most often antegrade wire escalation (70%), followed by retrograde (22%). Success was achieved in 79.4%. Failure was most closely correlated with presence of an ambiguous proximal cap, and in the presence of an ambiguous proximal cap, specifically defined collateral score (combination of Werner and tortuosity scores) and retrograde tortuosity. Without an ambiguous proximal cap, poor distal target, occlusion length >10 mm, ostial location, and 1 operator variable contributed. Prior failure, and Werner and tortuosity scores alone, were only weakly correlated with outcomes. The basic 7-item model predicted success, with C statistics of 0.753 in the training cohort and 0.738 in the validation cohort, the later superior (p < 0.05) to that of the J-CTO (Multicenter CTO Registry of Japan) (0.55) and PROGRESS CTO (Prospective Global Registry for the Study of Chronic Total Occlusion Intervention) (0.61) scores.
Conclusions Success can be reasonably well predicted, but that prediction requires modification and combination of angiographic variables. Differences in operator skill sets may make it challenging to create a powerful, generalizable, predictive tool.
Chronic total coronary artery occlusions (CTOs) are found in 15% to 30% of diagnostic catheterization procedures, are inconsistently treated with percutaneous coronary intervention (PCI), and are often the reason for referral for coronary artery bypass graft surgery (1–4). Improvements in angioplasty equipment, technique, and strategy (especially the “hybrid approach,” which is now the preferred strategy for many if not most experienced CTO operators ) have led to increased success rates for some operators in treating CTOs, but reported success rates continue to vary between 60% and 90% depending on the anatomy, the operator’s experience, and willingness to take on difficult procedures (6–8). Current models to predict angiographic success in this situation are limited by only modest predictive value (C statistics ranging from 0.65 to 0.75) and are generally not specific to the hybrid approach (9–13). Given the risk and cost of the procedure, as well as varied physician expertise, better predictive models with contemporary strategies are needed.
Clinical sites were chosen on the recommendation of leadership of the OPEN CTO (Outcomes, Patient Health Status, and Efficiency in Chronic Total Occlusion) study (14). Sites were asked to provide consecutive series of 40 to 80 patients from their recent experience, treated by 1 or more experienced hybrid operators, regardless of angiographic and clinical outcome. Patients were deidentified. Participation in this study was approved by the relevant local Institutional Review Board.
Clinical data were obtained directly from the clinical sites using dedicated case report forms.
All angiographic analyses were performed by a single hybrid trained CTO operator, the Director of the Angiographic Core Laboratory at the Cleveland Clinic, using Siemens software and guide catheter calibration for quantitative measurements (15). All pre-intervention analyses were performed blinded to angiographic and clinical outcomes.
To assess reproducibility, 20 randomly selected angiograms were reviewed independently by 2 operators (S.G.E., R.N.) to assess concordance of selected morphological parameters, as reflected by Cohen’s kappa statistic (>0.75 indicates strong agreement, 0.40 to 0.75 fair to good agreement, and <0.40 poor agreement).
Ambiguous proximal cap (APC) was defined as a flush occlusion with uncertainty as to the vessel’s initial course and likely unwillingness of an experienced operator to use an aggressive wire to “stick” the cap. CTO was defined as a 100% stenosis with TIMI (Thrombolysis In Myocardial Infarction) flow grade ≤1 of presumably >3 months’ duration.
Excessive proximal or retrograde tortuosity was defined as a >90° bend in diastole in a nonforeshortened projection in need of being traversed to access the CTO.
Interventional collateral vessels
Septal collateral vessels were graded for: 1) Werner classification (16), with grade 1 subdivided into 1A (definite but spindly, connecting with and filling all or much of the retrograde tree) and 1B (small but definitely present and filling a small portion of the coronary tree); 2) number of septal collateral vessels fulfilling at least Werner criterion 1B; 3) tortuosity score classification for the best interventional collateral vessel (0 = generally straight, 1 = moderate corkscrew without kinks, 2 = tight corkscrew and/or kinked); and 4) the presence or absence of a 90° turn into or out of the best intervention collateral vessel (17).
Other potential interventional collateral vessels were scored similarly, with the exception of number. On the basis of observations from the training cohort, a summary collateral risk score was defined as 0 = Werner 2 septal vessels, or Werner 2, tortuosity score <1 epicardial vessels, or bypass graft <90° access bend; 2 = Werner 0 septal vessels, or Werner 0 or tortuosity score 2 epicardial vessels, or bypass graft >120° access bend; and 1 = other.
Landing zones for dissection re-entry were graded on a scale of 0 to 2, with grade 2 indicating angiographically assessed >2 mm in diameter (not presumed on the basis of vessel size), >15 mm in length, no moderate to severe calcium, and no major side branches; grade 1 indicating 1 to 2 mm in diameter, >10 mm in length, and no severe calcium or major side branches; and grade 0 indicating anatomy less favorable than for a score of 1. Landing zones were also scored 0 or 1 for the presence of a side branch emanating from the landing zone supplying the left ventricle and with diameter >1.5 mm.
Lesion calcification was graded on a scale of 0 to 3: 0 = none visible, 1 = visible only with motion, 2 = visible without motion (moderate), and 3 = extends across the diameter of the coronary artery (severe).
A poor target vessel was defined as a landing zone score of 0 or reference vessel size <2.0 mm.
Preferred technical approach was coded as per the hybrid algorithm (5), with occasional modification on the basis of local anatomy (e.g., heavy calcification, excessive tortuosity).
All other angiographic definitions are consistent with those used previously in published research.
Intraobserver variability testing revealed kappa values as follows: APC (κ = 0.50), epicardial collateral vessels (Werner κ = 0.46, Kato κ = 0.29), septal collateral vessels (Werner κ = 0.91, Kato κ = 0.20), target vessel (κ = 0.59), and summary collateral vessel score (κ = 0.64).
Operators were identified as A through I.
Technical and procedure success
Angiographic or technical success was defined as present if there was restoration of TIMI flow grade 3 into branches constituting at least one-half of the segments distal to the occlusion and no stenosis >50% within the site of occlusion.
Angiographic success, when present, was attributed to the dominant technical approach leading to success (antegrade wire escalation, antegrade dissection re-entry, retrograde) or in rare cases in which there was not a clear dominant effective approach to the 2 approaches leading to success. Technical failure was attributed to all approaches used not leading to angiographic success.
Procedural success was defined as angiographic success in the absence of tamponade, myocardial infarction, and emergency bypass surgery or periprocedural death.
Continuous data are summarized as mean ± SD (normally distributed data) or as median and interquartile range (skewed data) and were compared using the Student t test or Wilcoxon rank sum test as appropriate. Categorical data are expressed as frequencies and were compared using chi-square or Fisher exact tests, depending on the expected cell frequency.
Patients were randomly divided 2:1 into the training and validation cohorts.
Five outcomes of interest were assessed: overall technical success, and procedural success and technical success with the antegrade wire escalation, dissection re-entry, and retrograde approaches alone. Technical success was deemed the primary outcome.
Given the ratio of lesions to patients, the final analysis was performed on a per lesion basis.
Within the training cohort, continuous variables were assessed for possible dichotomization or transformation using cubic spline analysis. Pre-procedural variables (Table 1) with p values <0.20 were eligible for backward elimination regression analysis. For the primary outcome analysis, hierarchical logistic regression analysis on the basis of the strategy of the hybrid approach and regression results was performed to identify sequential covariates. For example, the potential effect of collateral vessels was assessed separately for patients with and without APCs. Septal and epicardial collateral vessel patterns were assessed separately. Logically related variables correlated with outcomes in the training set were tested separately and in combination. A basic model was constructed using only data from this cohort. An extended model was constructed considering a limited number of variables shown in multiple other studies to correlate with success (calcium, left circumflex coronary artery occlusion, prior PCI failure, prior coronary artery bypass grafting), vetted on the basis of their correlation with success after consideration of the basic score and their effect on the C statistic in the training set.
The resultant models were evaluated using receiver-operating characteristic curves and C statistics, and the overall model was compared with the J-CTO (Multicenter CTO Registry of Japan) and PROGRESS CTO (Prospective Global Registry for the Study of Chronic Total Occlusion Intervention) scores in the validation cohort.
Data from 461 patients (481 lesions) treated in 2014 and 2015 (except for those from the senior author) were submitted. Twenty-one patients (21 lesions) were rejected as not having target CTOs (all with TIMI flow grade 2 not due to bridging collateral vessels), and 4 patients (4 lesions) were rejected as having inadequate pre-PCI images for analysis.
Notably, 24.5% of all CTOs had 1 or more previous failed attempts, the mean J-CTO score was 1.4 ± 0.9, the first technical approach was most commonly antegrade wire escalation (70%), and technical success and procedural success were achieved in 79.4% and 77.9%, respectively.
Correlates of success
Procedural success by operator ranged from 61.1% to 90.9% (p = 0.023).
Individual correlates of outcomes are presented in Table 3 and Figures 1, 2, 3, and 4. The presence of an APC or a poor landing zone (score = 0) was the most statistically powerful individual correlate in the training set.
Importantly, many of the previously found correlates of failure (e.g., prior failed attempt [odds ratio (OR): 1.80; p = 0.19], moderate to severe lesion calcium [OR: 1.80; p = 0.29], occlusion length >20 mm [OR: 1.10; p = 0.82], left circumflex coronary artery occlusion [OR: 1.40; p = 0.45]) were not well correlated with success in this cohort.
A 7-item model described in Figure 5A predicted success, with C statistics of 0.753 in the training cohort and 0.738 in the validation cohort, the later superior (p < 0.05) to that of the J-CTO (0.55) and PROGRESS CTO scores (0.61).
When considered in addition to the basic score, the OR for moderate to severe calcium was 1.90 (p = 0.098), for prior failure was 1.50 (p = 0.30), for left circumflex lesion was 1.50 (p = 0.40), and for prior bypass surgery was 1.10 (p = 0.75). When calcium was added to the basic score, the C statistics for the training and validation cohorts were 0.746 and 0.714, respectively. When other potential covariates were added in addition to calcium, the C statistics fell below 0.70 in the validation cohort.
Correlates of success by technical approach
Global (refitted) correlates of the antegrade wire escalation approach (n = 375)
Univariate correlates of failure included occlusion length >10 mm (p < 0.001), APC (p < 0.001), ostial location (p = 0.004), and TIMI flow grade 1 (p = 0.043). Independent correlates were occlusion length >10 mm (OR: 2.77; p < 0.001), APC (OR: 2.64; p < 0.001), and ostial location (OR: 1.90; p = 0.049).
Global (refitted) correlates of antegrade dissection retry approach (n = 83)
No significant correlates of success were found, including measures of lesion calcium, vessel size, occlusion length and operator.
Global (refitted) correlates of retrograde approach (n = 166)
The collateral vessel score was the only parameter approaching statistical significance (p = 0.055 overall, p = 0.029 for septal collateral vessel–based attempts). Other parameters trending toward correlation included prior coronary artery bypass grafting (p = 0.098), retrograde tortuosity (p = 0.10), Werner score <2 (p = 0.14), and tortuosity score >0 (p = 0.21).
Correlates of procedural complications (death, myocardial infarction, emergency coronary artery bypass grafting, or tamponade)
Independent correlates of complications included moderate to severe lesion calcium (OR: 4.85; p = 0.023) and low left ventricular ejection fraction (OR: 0.96 per unit; p = 0.029.) Age >70 years trended toward correlating with complications (OR: 3.04; p = 0.086). Technical success was unrelated to complications: 2.2% in technically successful procedures and 1.1% in failed cases (p = 0.48).
Although there have been several recent studies of predictors of successful CTO PCI, this differs from most in that: 1) the essential technical approach to the procedures followed the “hybrid algorithm”; 2) the focus was on results of well-trained moderate- to high-volume, but not ultra-high-volume, CTO operators’ results; and 3) the angiographic analysis was highly detailed and granular.
The main findings of this study relate to the prognostic importance of an APC, and in its presence, the nature and access to retrograde collateral vessels, and in its absence, the important of factors related to antegrade success, namely, occlusion length, the distal target vessel and ostial site of occlusion, and finally, in particular, the prognostic capability of the derived score relative to other scores.
These findings have some similarities with, but are also different from, other studies assessing correlates of procedural success with the hybrid CTO approach. For example, Sapontis et al. (7) found occlusion length, tortuosity, and proximal cap ambiguity to be related to outcomes, as in this study, but also J-CTO score, which had little impact in this study. A summary score was not derived in their analysis. Other studies have also questioned the predictive utility of the J-CTO score with hybrid-approach CTO PCI (10).
The importance of the operator skill set, even considering that all operators’ annual CTO volumes were relatively high, also carried independent prognostic value. This corroborates similar findings with complex CTO PCI (18).
A score considering all of these factors performed quite well in the independent validation cohort, derived from the same operators’ cases, with a C statistic of 0.738.
This predictive capability was considerably better than that of the J-CTO (C statistic = 0.55) and PROGRESS CTO (C statistic = 0.61) scores. Interestingly, several factors used in those and other scores (e.g., prior failure, left circumflex coronary artery occlusion, simple Werner collateral vessel score) had very limited or no predictive value in this analysis (19–21).
The statistical analysis plan was a priori slightly modified from the usual multivariate logistic regression approach used in previous studies of this topic, for reasons that should be intuitive to the interventional operator. These have to do principally with technique choice. Consider an example in which 1 technical approach, perhaps antegrade wire escalation, is clearly favored (e.g., in the absence of interventional collateral vessels). If the anatomic complexity for the antegrade approach is simple, then the absence of collateral vessels is largely meaningless, whereas if the antegrade related anatomy is challenging, the fact that there are no interventional collateral vessels is likely to make an important impact on the likelihood of success. Traditional models would “count” the impact of the absence of interventional collateral vessels equally in the 2 examples.
The extended model, also including evaluation of lesion calcium, considers the broader issue of modeling in this scenario from a more Bayesian perspective on the basis of results of other studies and yielded similar predictive power to the basic model in our validation cohort.
Although these scores need to be validated in fully independent datasets, they are likely to have the most utility for operators similar to those who contributed to this study, not rare ultra-high-volume operators with success rates in excess of 90%, or infrequent or non–hybrid strategy–trained CTO operators. Given the highly varying success rates reported in the published research, the score may have merit in identifying relatively “easy” CTOs for which it may be reasonable to have lower volume operators attempt, while steering those away for more complex lesions, but the score would have to be tested among that cohort of operators.
A number of limitations should be considered when reviewing these results. First, a relatively limited number of operators and procedures were analyzed. Some rare but powerful correlates of outcome may have been missed. Furthermore, because interventionalists will avoid approaches with apparent low likelihood of success, fewer data will be generated in such scenarios. Consider, if there are no evident interventional collateral vessels, the retrograde approach is unlikely to be undertaken, hence the impact of the absence of interventional collateral vessels will be underestimated when one evaluates correlates of success using the retrograde approach. A hierarchical modeling approach can obviate only some of the impact of this problem. Second, the operators whose results were analyzed have advanced skills and perform large numbers of CTO procedures annually; hence the predictive tool may well have less value for less, or even more, skilled operators. Even within this group, varied skills sets could be discerned.
Third, attributing success to a single approach when they are often used in tandem (the retrograde approach usually requires reverse CART (controlled antegrade and retrograde tracking), i.e., the ability to cross into the proximal cap architecture) is problematic. Last, interobserver agreement on assessing key anatomic variables ranged from excellent (septal Werner score) to poor (septal Kato score), similar to that reported for other angiographic morphologic variables (22). More uniform agreement on precise definitions might improve the reproducibility and usefulness of this and similar scoring systems.
WHAT IS KNOWN? CTO PCI success rates vary considerably by operator and anatomic complexity. Current models predicting technical success with the hybrid approach have important limitations, often applying only to truly expert operators and/or having poor predictive power.
WHAT IS NEW? This study provides a model, tested and validated in high-CTO-volume operators with realistic approximate 80% success rates, that outperforms previously available models.
WHAT IS NEXT? The generalizability of this model needs to be tested with other operators and patients, with the understanding that it is likely that the factors that influence success may vary among low-volume, high-volume, and truly elite CTO operators.
The authors acknowledge the financial support of Boston Scientific and the technical expertise of Ms. Janet Doak to allow us to complete this project and report.
The authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- ambiguous proximal cap
- chronic total coronary artery occlusion
- odds ratio
- percutaneous coronary intervention
- Received November 18, 2016.
- Revision received February 22, 2017.
- Accepted March 8, 2017.
- 2017 American College of Cardiology Foundation
- Fefer P.,
- Knudtson M.L.,
- Cheema A.N.,
- et al.
- Brilakis E.S.,
- Grantham J.A.,
- Rinfret S.,
- et al.
- Brilakis E.S.,
- Banerjee S.,
- Karmpaliotis D.,
- Lombardi W.L.,
- et al.
- Galassi A.R.,
- Boukhris M.,
- Azzarelli S.,
- et al.
- Luo C.,
- Huang M.,
- Li J.,
- et al.
- Franco-Nombela L.,
- Urena M.,
- Valero-Jerez M.,
- et al.
- Alessandrino G.,
- Chevalier B.,
- Lefvre T.,
- et al.
- Christopoulos G.,
- Wyman R.M.,
- Alaswad K.,
- et al.
- Christopoulos G.,
- Kandzari D.E.,
- Yeh R.W.,
- et al.
- ↵Outcomes, Patient Health Status, and Efficiency in Chronic Total Occlusion (OPEN-CTO). Available at: https://clinicaltrials.gov/ct2/show/NCT02026466. Accessed March 22, 2017.
- ↵Siemens Healthineers. syngo QCA. Available at: https://usa.healthcare.siemens.com/angio/options-and-upgrades/clinical-software-applications/syngo-qca. Accessed March 22, 2017.
- Werner G.S.,
- Ferrari M.,
- Heinke S.,
- et al.
- Galassi A.R.,
- Sianos G.,
- Werner G.S.,
- et al.
- Rathore S.,
- Katoh O.,
- Matsuo H.,
- et al.
- Popma J.J.,
- Lansky A.J.,
- Yeh W.,
- et al.