Incidence, angiographic and clinical predictors, and impact of stent thrombosis: a 6-year survey of 6,545 consecutive patients

A nested case-control study was conducted at our tertiary centre. Consecutive patients with angiographically confirmed ST as defined according to Academic Research Consortium criteria [3] between March 2010 and November 2016 were included in the study. We did not incorporate cases of very late ST (i. e. ST beyond 1 year after PCI). Angiographically confirmed acute, subacute, or late ST cases were matched to controls without ST in a 1:4 ratio based on (1) clinical indication and (2) index date ± 6 weeks of the PCI, and were randomly selected from a consecutive patient cohort of 6,545 patients to comprehensively study the incidence, angiographic predictors, and impact of definite ST. We did not use any restrictions in the random selection of matched controls. The study complied with the Declaration of Helsinki, and explicit Ethics Committee approval from our institution was waived given the non-experimental design of the study.

PCI was performed according to standard techniques. Patients were treated with aspirin and a P2Y12 inhibitor prior to PCI, and 70–100 IU/kg of heparin during the procedure. Thereafter, aspirin was continued indefinitely, and the P2Y12 inhibitor was continued as defined by national and international guidelines for a period of 1–12 months: 1 month following bare-metal stents (BMS) and 6–12 months following drug-eluting stents (DES). Implantation of BMS and use of GpIIb/IIIa inhibitors were at the operators’ discretion.

Coronary angiography reassessment was conducted using Xcelera software (Philips Healthcare, Best, The Netherlands). Baseline and post-PCI angiograms of the index procedures were individually reassessed in random order by two well-experienced interventional cardiologists who were blinded to the objective of the study and to patient outcome. In case of disagreement, a consensus was reached by consulting a third interventional cardiologist. A complete list of the angiographic definitions used in our study is provided in the Supplementary Appendix. Subsequently, the level of inter-observer reliability [4] (i. e. Cohen’s kappa coefficient, κ) was calculated. Cohen’s kappa coefficient ranges from −1.0 to +1.0, and was considered: poor (κ < 0), slight (κ = 0–0.20), fair (κ = 0.21–0.40), moderate (κ = 0.41–0.60), or substantial (κ = 0.61–0.80). Lesion complexity was defined according to the modified American College of Cardiology/American Heart Association criteria [5] as lesion type B2 or C.

Detailed clinical, procedural, and angiographic data of the index PCI were obtained for both cases and controls. After hospital discharge, the duration and compliance of antiplatelet and antithrombotic agents were verified from prescription data linked to pharmacy records. To study the use and impact of antiplatelet therapy, we compared dual antiplatelet therapy (DAPT) use at the time of index PCI to angiographically confirmed ST or 30 days prior to ST for cases. Then, we compared this to the same time period for the matched controls. Follow-up data were collected during hospital admissions, routine visits to the outpatient clinic, by a medical questionnaire, and by telephone assessment at 12 months after PCI. Information regarding the patients’ vital status was obtained by the Dutch National Civil Registration.

Continuous variables are reported as means ± standard deviations (SD), and categorical variables as numbers (n) and percentages (%) compared using chi-square or Fisher exact test, Student’s t-test or Mann-Whitney U test, as appropriate. The cumulative incidence of angiographically confirmed ST with a 95% confidence interval (CI) was calculated using the Wilson method as appropriate for events with low incidence rates. After visually checking the randomness of missing data, multiple imputation chained equation was performed using 5 datasets with 10 iterations [6]. A multivariable conditional logistic regression model was built to identify independent predictors of ST with odd ratios (OR) with 95% CI as a summary statistic. Kaplan-Meier estimates for the risk of death were analysed using a log-rank test, and reported according to good clinical practice statement [7]. An adjusted Cox proportional-hazard regression for the risk of death was performed with adjustment for: age, sex, body-mass index, hypertension, dyslipidaemia, diabetes mellitus, smoking status, left ventricular ejection fraction, estimated glomerular filtration rate, prior acute coronary syndrome, stroke, peripheral artery disease, malignancy, and three-vessel disease. All statistical analyses were performed using SPSS version 25 (SPSS Inc., Chicago, IL, USA), and figures were generated using GraphPad Prism software version 7 (GraphPad, Inc., San Diego, CA, USA).

The inter-observer reliability for the overall angiographic reassessment was substantial (κ = 0.63, Supplementary Appendix). Reassessment of index angiograms showed a significant difference in angiographic stent underexpansion (43.8% vs 15.5%, p < 0.001), angiographic stent edge dissection (18.8% vs 7.2%, p = 0.005), residual coronary artery disease (CAD) (17.2% vs 8.0%, p = 0.027), and no-reflow phenomenon (9.1% vs 2.3%, p < 0.030) in ST as compared to the matched controls. Patients who suffered ST had a suboptimal angiographic result more often than the matched controls (21.9% vs 7.6%, p < 0.001) (Fig. 3).

Independent clinical, angiographic, and procedural predictors of angiographically confirmed ST were identified using a multivariable logistic regression model (Tab. 3). The strongest independent predictor of angiographically confirmed ST was DAPT non-use (OR 10.9, 95% CI 2.47–48.5, p < 0.001), followed by: stent underexpansion (OR 5.70, 95% CI 2.39–13.6, p < 0.001), lesion complexity B2/C (OR 4.32, 95% CI 1.43–13.1, p = 0.010), uncovered stent edge dissection (OR 4.16, 95% CI 1.47–11.8, p = 0.007), diabetes mellitus (OR 3.23, 95% CI 1.25–8.36, p = 0.016), and residual CAD 5 mm proximal or distal to the stent (OR 3.02, 95% CI 1.02–8.92, p = 0.045).

Kaplan-Meier estimates of the cumulative risk of death were obtained for cases and matched controls with right censoring at 5 years’ follow-up (Fig. 4). Vital status was available for 273 patients (99.3%) with a median follow-up of 4.0 ± 2.1 years. Angiographically confirmed ST was associated with an increased risk of death as compared to the matched controls (27.3% vs 11.3%, p_log-rank < 0.001). No differences were found in mortality with respect to cases that were treated for acute PCI as compared to planned PCI (30.8% vs 24.1%, p = 0.31). After adjustments, ST patients had a two-fold increased risk of death as compared to matched controls (adjusted hazard ratio 2.29, 95% CI 1.03–5.10, p = 0.042).

This study has some limitations that should be acknowledged. First, by definition we did not incorporate cases of very late ST that may have underestimated the cumulative incidence by approximately 15%. Moreover, we did not consider ST from: (1) sudden cardiac death or (2) silent stent occlusion. Second, although no independent angiographic core laboratory was consulted, angiograms were reassessed and quantified in a structured, blinded, and reliable manner. We believe that the lack of an independent core laboratory did not affect the main results of this study. Third, even though we consider verified DAPT usage from linking prescription data to pharmacy records as a strength of our analysis, we obtained complete data in roughly three-quarters of our control patients that may have introduced a certain degree of bias that we tried to minimise by using data imputation. Finally, our study reflects an analysis of BMS and DES implantations, given the time period during which it was conducted. However, since BMS are obsolete in current practice, it would have been interesting to study DES exclusively.