Near-Infrared Spectroscopy Predicts Cardiovascular Outcome in Patients With Coronary Artery Disease

Study population and design

The ATHEROREMO-NIRS (The European Collaborative Project on Inflammation and Vascular Wall Remodeling in Atherosclerosis–Near-Infrared Spectroscopy) substudy (2) was a prospective, single-center, observational study assessing the prognostic value of coronary NIRS, performed at the Thoraxcenter, Erasmus Medical Center, Rotterdam, the Netherlands. All patients had an indication, as determined by their treating physician (as part of routine clinical care), for diagnostic coronary angiography and/or percutaneous coronary intervention (PCI) due to either stable angina pectoris or an ACS. Detailed inclusion and exclusion criteria are listed in the Online Table 1.

Subsequent to the standard angiography and PCI (when applicable), NIRS of a nonculprit coronary artery was performed. The NIRS target segment of the nonculprit coronary artery was required to be at least 40 mm in length and without significant luminal narrowing (<50% stenosis) as assessed by online angiography. The order of preference for selection of the nonculprit vessels was predefined in the study protocol: 1) left anterior descending artery; 2) right coronary artery; and 3) left circumflex artery.

This study was approved by the Medical Ethics Committee of the Erasmus Medical Center, and performed in accordance to the Declaration of Helsinki (2008, 6th revision). Written informed consent was obtained from all participants.

Sample size

The ATHEROREMO-IVUS (The European Collaborative Project on Inflammation and Vascular Wall Remodeling in Atherosclerosis–Intravascular Ultrasound) study had a pre-specified sample size of 800 patients, and was designed to explore multiple relations between genetic and serum biomarkers and coronary plaque characteristics (2). It was during the course of the ATHEROREMO-IVUS study that intracoronary NIRS became commercially available and accessible for our cardiac catheterization lab. The ultimate sample size of the ATHEROREMO-NIRS substudy (203 consecutively consenting patients) was not based on prior effect estimates but rather on the time point of availability and local institutional review board approval (April 2009) of the NIRS technique as ATHEROREMO evolved.

Near-infrared spectroscopy

The U.S. Food and Drug Administration–approved NIRS system, as used in this study, consists of a 3.2-F rapid exchange catheter, a pullback and rotation device, and a console (InfraReDx, Burlington, Massachusetts). Image acquisition was performed by a motorized catheter pullback at a speed of 0.5 mm/s and 240 rpm in a proximal segment of a nonculprit artery, starting distal to a side branch. The system performed 1,000 chemical measurements per 12.5 mm, in which each measurement interrogated 1 to 2 mm² of vessel wall from a depth of approximately 1 mm in the direction from the luminal surface toward the adventitia (1). Areas of the artery with spectral characteristics of a lipid core were displayed in yellow within the image map, called a chemogram. NIRS images were analyzed offline by an independent core laboratory (Cardialysis, Rotterdam, the Netherlands). Core laboratory personnel were blinded to all other patient and outcome data.

Study endpoints

The pre-specified primary endpoint was the incidence of MACCE, defined as the composite of all-cause mortality, nonfatal ACS, stroke, and unplanned coronary revascularization during 1-year follow-up, exclusive of events related to the culprit lesion at the index angiography. Secondary endpoints included: 1) the composite of all-cause mortality and nonfatal ACS; 2) the composite of all-cause mortality, nonfatal ACS, and stroke; and 3) the composite of all-cause mortality, nonfatal ACS, and unplanned coronary revascularization during follow-up. Endpoints were adjudicated by a clinical events committee on the basis of original source data. Members of the clinical events committee were blinded to other patient data and NIRS imaging characteristics. Post-discharge survival status was obtained from municipal civil registries. Nonfatal ACS included ST-segment elevation myocardial infarction (STEMI), non-STEMI, or unstable angina pectoris as defined in accordance with the guidelines of the European Society of Cardiology (3,4). Stroke was defined according to the guidelines of the European Stroke Organization (5). Unplanned coronary revascularization was defined as PCI or coronary artery bypass grafting, which initially was not planned after the index angiography and enrollment in the study.

Whenever possible, all events were further adjudicated as related or unrelated to the coronary site that was treated during the index procedure. In case of follow-up angiography, events were classified either as a definite culprit lesion–related (CLR) event or as related to a coronary site that was not treated during the index procedure (non-CLR event). In case angiographic information on the endpoint related coronary site was not available, the event was classified as indeterminate. The pathophysiological processes of definite CLR events (e.g., in-stent restenosis and stent thrombosis) differ from the pathophysiology of spontaneous plaque rupture. Therefore, data are presented both exclusive of definite CLR events (by default), as well as inclusive of definite CLR events.

Statistical analysis

Normally distributed continuous variables are presented as mean ± SD. Non-normally distributed continuous variables (e.g., the lipid core burden index [LCBI]) are presented as median (interquartile range [IQR]). Categorical variables are presented in numbers and percentages. Differences in baseline continuous variables between those with an LCBI below versus those equal to and above the median were analyzed by Mann-Whitney U tests, categorical variables by Fisher’s exact and Pearson chi-square tests (in case of more than 2 categories). Log or square root transformations were applied whenever homoscedasticity was a required assumption of the used statistical test. Linear regression was used to determine predictors of LCBI. No prior data were ever reported on LCBI distribution of the nonculprit artery. The statistical analytical plan therefore pre-specified the median LCBI value as cutoff between “low” and “high” LCBI groups, in case LCBI would appear to be non-normally distributed.

Patients lost to follow-up were considered at risk until the date of last contact, at which time point they were censored. For patients with more than 1 event, the first was counted. Cumulative event rates were estimated according to the Kaplan-Meier method and compared by the log-rank test. Backward stepwise regression analyses were used to determine the predictors of the primary endpoint. Univariable and multivariable Cox proportional hazards regression analyses were applied to evaluate the association between LCBI and 1-year outcome. Three models were used throughout the manuscript: an unadjusted model, an age- and sex-adjusted model, and a “full model.” Given the number of events available, adjustment according to a propensity score was used in order to assure parsimony of the full model (6). Variables for the propensity score of the full model were selected on the basis of clinical relevance and significance after backward stepwise regression. The propensity scores were derived from predicted probabilities in logistic regression models with LCBI above the median as dependent variable (7). The propensity score that was entered into the full model accounts for age, sex, hypercholesterolemia, diabetes, hypertension, indication for index coronary angiography (stable angina pectoris versus ACS), history of myocardial infarction, peripheral artery disease (PAD) or stroke, and prior PCI. Crude and adjusted hazard ratios (HRs) are presented with 95% confidence intervals (CIs). Discrimination of the full model with respect to event prediction was evaluated with receiver-operating characteristic curves. Heterogeneity of the effect of LCBI on MACCE was tested between patients presenting with stable angina and those presenting with ACS at the time of enrollment (8). All statistical tests were 2-sided with a type I error level of 0.05. Analyses were performed with IBM SPSS statistics version 20.0 (IBM Corp., Armonk, New York).