Data from 1034 patients, who underwent CMR imaging after undergoing primary percutaneous coronary intervention (PCI), were analyzed in this study. All patients were enrolled within the AIDA-STEMI (Abciximab Intracoronary versus intravenously Drug Application in STEMI) and TATORT-NSTEMI (Thrombus Aspiration in Thrombus Containing Culprit Lesions in NSTEMI) trials. The aims of these previous studies were on the one hand to compare the value of intravenous versus intracoronary abciximab application in STEMI patients, which did not reveal a considerable difference between both strategies (AIDA-STEMI trial), and on the other hand to examine the effect of aspiration thrombectomy versus conventional PCI, which also did not show significant differences regarding IS, MVO or clinical outcome (TATORT-NSTEMI trial). More information including detailed study protocols and results have been previously reported [12, 13]. All patients gave written informed consent before participating. Both studies were approved by all involved local ethical committees and complied with the principles of the Helsinki Declaration.

All patients underwent an identical CMR protocol within 10 days after the index event. This protocol was applied on 1.5 or 3.0 T scanners at every study site and included balanced steady-state free precession sequences (SSFP) of long-axis 2- and 4-chamber views (CV) as well as short-axis (SAX) stacks. Typical SSFP sequence parameters were as follows: repetition time 3.2 ms, echo time 1.2 ms, flip angle 60°, 8 mm slice thickness in SAX. T2-weighted triple short-tau inversion recovery images (repetition time 2 RR intervals; echo time 80 ms; flip angle 90°) were generated. For the analysis of myocardial salvage, infarct size and microvascular obstruction (MVO), inversion recovery gradient echo sequences were acquired 10–20 min after a gadolinium bolus injection (0.15 mmol/kg bodyweight, repetition time 2.8 ms; echo time 1.1 ms; flip angle 15°; slice thickness 8 mm, with individually adjusted inversion times typically between 200 and 300 ms). More detailed information regarding study and scan protocols have been previously published [12, 13]. Typical contraindications to CMR applied to this study as previously detailed [13].

SAX segments were defined according to the American Heart Association 16-segment model [14] with segments containing LGE considered as infarct segments. RM segments were defined as unenhanced segments with one unenhanced border segment between them and infarct segments, respectively (Fig. 1). For the purposes of this paper, we defined circumferential strain (CS) in RM segments as RM CS and CS of infarct segments as infarct CS. CS of all segments is indicated as global circumferential strain (GCS). Strain analysis was performed in balanced SSFP-derived SAX stacks using dedicated evaluation software (2D CPA MR, Cardiac Performance Analysis, Version 1.1.2, TomTec Imaging Systems, Unterschleissheim, Germany). To assess SAX-derived CS, LV epi- and endocardial borders were manually tracked in basal, midventricular and apical slices to obtain global as well as regional strain values. After manual delineation of the myocardial borders at end-diastole, a semi-automated tracking algorithm was applied for tracing the contours throughout the cardiac cycle. Visual evaluations of the semi-automatically tracked contours were performed and, in case of insufficient border tracking, manual adjustments to the delineations were made with subsequent reapplication of the algorithm. All peak strain measurements are presented in percent and based on an average of three repeated and independent tracking repetitions.

For infarct area detection and quantification, dedicated post-processing software was used (QMass^®, Version 3.1.16.0, Medis Medical Imaging Systems, Leiden, Netherlands) to manually segment LV epi- and endocardial borders in SAX orientation. LGE segments containing myocardial infarction area were defined using a full width at half maximum (FWHM) approach with excellent reproducibility [15]. The slice containing the highest signal intensity infarct area enhancement was chosen and regions of interest including infarct area and normal myocardium were marked to define thresholds for enhanced and non-enhanced myocardium. After applying semi-automated infarct area detection and performing a visual accuracy review, manual adjustments were made as required. Regions with MVO were manually included, since signal intensity is not increased in these areas.

The occurrence of MACE was defined as the primary clinical endpoint of this study. All-cause mortality, reinfarction or congestive heart failure associated with rehospitalization within the first year after AMI was counted as MACE. In case of multiple occurrences of MACE within one patient, a prioritization was made (death > reinfarction > congestive heart failure) with each patient only accounting for one MACE.

Categorical parameters are presented in absolute numbers and percentages. Continuous parameters were tested for normal distribution using Shapiro–Wilk test and are reported as mean with interquartile range (IQR). For the assessment of correlations, the Spearman’s rank correlation coefficient was used. Non-parametric Mann–Whitney U test was used for comparisons of continuous data sets. A previously determined cutoff value for GCS (> − 18.3%) based on area under the curve analyses was used [10]. Similarly, an optimal dichotomization cutoff value for RM CS was determined by Youden’s Index. Moreover, high- and low-risk patient groups were categorized for LVEF, IS and MVO. For LVEF, subgroups with ≤ 35% and > 35% were analyzed according to established clinical practice. IS and MVO cutoff values were calculated using Youden’s Index. The Kaplan–Meier method was applied to analyze occurrence of clinical endpoints within predefined groups and log-rank testing for assessing differences between groups. Univariate hazard ratios (HRs) based on Cox proportional regression models were calculated in the context of MACE and mortality evaluation. Only variables with a p value < 0.05 were included in further multivariable regression calculations. Multivariable calculations comprised a stepwise approach with thresholds of 0.05 and 0.1 for p values to keep or remove variables, respectively. Due to significant correlations of CS values and LVEF, only one of these parameters was included in multivariable regression models in each case. All provided p values are two-sided with an alpha level < 0.05 considered statistically significant. IBM SPSS Statistic Software Version 24 (International Business Machines, Armonk, New York, USA) and Microsoft Excel (Microsoft, Redmond, Washington, USA) were used for all statistical calculations.

Among the overall cohort of 1235 patients enrolled in this CMR substudy (795 STEMI and 440 NSTEMI patients), 1034 patients with both SSFP and corresponding inversion recovery gradient SAX stacks covering base to apex were identified and included in the final analysis. 869 patients (648 STEMI and 221 NSTEMI) had evidence of infarct area on LGE imaging with a maximum extent allowing the definition of at least one RM segment (Fig. 2). Baseline characteristics of patients with RM are displayed in Table 1. The study cohort included mainly male patients (73.4%) with a median age of 64 years (IQR 52–72 years). Hypertension (p = 0.007) and diabetes (p = 0.03) were more frequent, whereas male patients (p = 0.008) and non-smokers (p = 0.013) were less frequent among patients with MACE during 1-year follow-up. Killip class on admission (p < 0.001) and the number of diseased vessels (p = 0.015) were significantly higher in patients with MACE.

CS was considerably higher in RM segments compared to infarct segments and GCS (− 29.7% [− 25.0% to − 34.4%] versus − 18.7% [− 14.7% to − 23.2%] for infarct segments versus − 23.8% [− 19.4% to − 28.2%] for GCS; p < 0.001 for both). Similar significant differences were present between patients with and without MACE (Table 2). All CS values were significantly reduced in patients with MACE during the 1-year follow-up (p < 0.001 for GCS and infarct CS; p = 0.007 for RM CS). IS and MVO were larger in patients with MACE (20.6% versus 14.2% for IS, p = 0.001 and 1.5% versus 0.4% for MVO, p = 0.01). Myocardial salvage index was higher in patients without MACE during follow-up (53.1 versus 41.8, p = 0.01). RM CS correlated significantly with LVEF (p < 0.001; r = − 0.3), whereas there was no significant correlation with infarct size (p = 0.257; r = − 0.039) (Fig. 3).

During the 1-year follow-up, 58 MACE were documented (death = 24, reinfarction = 16, congestive heart failure = 18). Using Youden’s Index, a cutoff value for RM CS of − 25.8% best classified the cohort into high- and low-risk groups according to RM function (Fig. 4). Further high- and low-risk grading according to infarct characteristics resulted in optimal cutoff values for a dichotomization with 19.2% (of LV mass) for IS and 1.46% (of LV mass) for MVO.

Univariable cox regression analyses in the overall cohort revealed an association of RM CS with MACE (HR 1.05, 95% CI 1.07–1.14, p = 0.003). IS (HR 1.04, 95% CI 1.02–1.06, p < 0.001), MVO (HR 1.1, 95% CI 1.03–1.17, p = 0.004) and myocardial salvage index (HR 0.99 (0.97–1.0, p = 0.013) were also associated with MACE (Table 3). Among patients considered at risk according to dichotomization cutoffs using established parameters, univariable regression-based hazard calculations identified RM CS as a strong predictor of MACE in patients with reduced LVEF ≤ 35% (HR 1.07 95% CI 1.02–1.13, p = 0.01), with reduced GCS > − 18.2% (HR 1.05 95% CI 1.02–1.09, p = 0.03) and with large MVO > 1.46% (HR 1.07 95% CI 1.02–1.1, p = 0.003). In patients with a small IS < 19.2% (HR 1.07 95% CI 1.04–1.2, p = 0.001), RM CS was also found to be a strong predictor of MACE. Applying multivariable regression calculations, Killip class (HR 1.49 95% CI 1.00–2.22, p = 0.048) and LVEF (HR 0.93, 95% CI 0.90–0.96, p < 0.001) remained significant after correction for all univariate significant parameters in the overall cohort (Table 3).

Furthermore, RM CS provided additional risk stratification among high-risk patients by identifying subgroups with higher jeopardy for the occurrence of MACE using Kaplan–Meier plots (log-rank test p = 0.015 for GCS > − 18.3%; p = 0.038 for LVEF ≤ 35% and p = 0.002 for MVO ≥ 1.46%). Regarding IS, RM CS provided additional risk stratification in small IS < 19.2% (p = 0.001). For both large IS and small MVO, a trend toward a statistical significance was documented (p = 0.085 for IS ≥ 19.2% and p = 0.074 for small MVO < 1.46%). Significant differences and trends are visualized by Kaplan–Meier plots in Fig. 5.

Several study sites performed CMR imaging using different vendors. Nevertheless, all centers followed the same study protocol and image post-processing was performed centrally in an experienced core laboratory. Regional strain reproducibility is generally lower compared to global values [29]. Nevertheless, CS has the highest reproducibility among myocardial deformation indices on a regional level [30] and was found to correlate best with LGE [26] and therefore might represent the most valid strain parameter for RM assessment. When defining RM segments, the percentage of infarct area in infarcted segments is not considered. Effects of a minimally affected infarct segment might be different compared to a completely infarcted segment. However, defining RM in accordance with the AHA 16-segment model allows an objective and standardized segment classification.