Recovery and prognostic value of myocardial strain in ST-segment elevation myocardial infarction patients with a concurrent chronic total occlusion

All CMRs were performed on a 1.5-Tesla scanner using a dedicated phased array cardiac receiver coil. ECG-gated steady-state free-precession cine images for LV function imaging were obtained, during repeated breath holds, in long-axis orientation (2-, 3-, and 4-chamber views) and in short-axis orientation covering the left ventricle from base to apex. Late gadolinium-enhanced (LGE) images were acquired using an inversion recovery gradient-echo pulse sequence with slice locations identical to the cine images to identify the size and extent of infarction. Images were acquired at least 10 min after administration of a gadolinium-based contrast agent in a dosage of 0.2 mmol/kg of body weight. Transmurality of scar tissue of the myocardium was assessed in patients who underwent baseline CMR of sufficient quality. To assess viability, transmural extent of infarction (TEI) was used, TEI of 0–50% per segment was considered viable [13].

The analysis of segmental wall thickening has been described before [11]. An independent core laboratory, blinded for randomization outcome, analyzed all CMR images (ClinFact Corelab using QMass MR analytical software version 7.6, Medis BV). A 16-segment model, excluding the apex, was used to analyze wall thickening. Endo- and epicardial borders on the end-diastolic and end-systolic images were manually outlined on all short-axis cine slices. Segmental wall thickening was defined as a percentage increase of LV wall thickness during systole compared with diastole. Myocardial segments were considered dysfunctional if wall thickening was less than 45% [14]. Individual segments for each patient were assigned to one of the major coronary arteries using the American Heart Association standardized myocardial segmentation and nomenclature statement [15]. Myocardial segments were assigned to the CTO, infarct, or remote territory using this standard model (in relation to the coronary anatomy scored by the angiographic corelab).

Strain measurements were performed offline using the FT-CMR software method of Medis QStrain Software (Medis Medical Imaging Systems, version 2.0.12.2.) (example of the analysis is in the Supplementary File). All three longitudinal-axis views (2-, 3-, and 4-chamber) were used to determine peak GLS. Endocardial contours were manually drawn during end-diastole and end-systole with subsequent automatic tracking during the cardiac cycle. As an example, a cine CMR movie file of the endocardial tracking is available as additional files 1 and 2. For the assessment of GCS and segmental circumferential strain, the corelab contours for the short-axis images were used. Peak GCS was calculated from 3 short-axis views (basal, mid, and apical). For peak segmental strain, short-axis images were used to define the segments according to the 16-segment model after manual insertion of a reference point (delineated at the anterior insertion of the right ventricle). All studies were loaded into the software and analyzed in a random order by one investigator blinded for randomization outcome under supervision of a CMR cardiologist with > 15-year experience (JE, supervisor: AH). The reproducibility of GLS measurements was assessed in 30 CMR scans (15 patients with baseline and follow-up CMR). The intraclass correlation coefficient for interobserver agreement was 0.97 (95% CI 0.89 to 0.99; p < 0.001).

At 4 months and 1, 2, 3, 4, and 5 years, clinical follow-up was collected to assess survival status. Survival data were censored at 5 years or known date of last contact [16].

Data are presented as mean ± standard deviation for continuous variables. Discrete variables are presented as frequencies and percentages. Baseline characteristics were compared using the independent-samples t test, or Fisher’s exact probability test in case of binary endpoints. Analyses were performed on intention-to-treat analysis. Changes in GLS and GCS within each group were tested with paired Student’s t test. Pearson’s correlation coefficient analysis was used to assess the relationship between the global strain and LV function parameters. Multivariable linear regression was used for testing the contribution of baseline, angiographic, and CMR characteristics in relation to LVEF at follow-up. Stepwise forward selection of variables was used; all variables with a p value < 0.05 were included and variables with a p value > 0.10 were removed from the model. Cumulative event rates of long-term mortality were estimated using Kaplan-Meier curves, and the Log rank statistic was used for comparing the survival curves. Hazard ratios for long-term mortality were calculated, after verification of the proportional hazard assumption, using Cox proportional hazard regression analyses. We evaluated the recovery of segmental circumferential strain in dysfunctional segments at baseline. Because within 1 patient the regional strain in the different segments is strongly related and not an independent outcome, multilevel analysis was used (linear regression) [17]. The following fixed effects were included: randomization outcome and baseline segmental strain. Multilevel analysis was also used to look for predictors of regional wall thickening recovery, and the following fixed effects were included: baseline SWT, baseline transmural extent of infarction, presence of microvascular obstruction, randomization outcome, and baseline segmental strain. All tests were two-sided, and a p value < 0.05 was considered to indicate statistical significance. Statistical analysis was performed with the Statistical Package for Social Sciences software (SPSS version 23.0 for Windows).

The baseline characteristics of the 200 patients included in this study are shown in Supplement Table 1. Patients had a mean age of 60 ± 10 years and 88% was male. Mean baseline LVEF was 41 ± 12%, LVEDV 103 ± 25 ml/m², and infarct size 12 ± 11 g. In the patients randomized to CTO-PCI, the PCI was performed on day 5 ± 2 after primary PCI. Baseline CMR was performed 4 ± 2 days after primary PCI. Baseline GLS was available in 184 patients and GCS in 176 patients. Serial GLS measurements were available in 166 and serial GCS in 160 patients (Fig. 1). GLS and GCS by FT-CMR were significantly correlated with LVEF, LVEDV, and infarct size (p < 0.0001 for all). Changes in strain parameters were also related to change in LVEF; however, change in LVEDV and infarct size showed poor correlation with change in global strain (Supplement Table 2).

GLS and GCS significantly improved from baseline to 4 months of follow-up (ΔGLS − 1.8 ± 4.3%, p < 0.001; ΔGCS − 1.7 ± 4.7%, p < 0.001) (Fig. 2). However, there was no treatment effect of CTO-PCI on the recovery of global strain parameters (∆GLS − 2.4 ± 4.2% versus − 1.4 ± 4.3%, p = 0.14; ∆GCS − 1.4 ± 4.5% versus − 1.9 ± 4.9%, p = 0.55) (Table 1).

Both GLS and GCS were univariate predictors for LVEF at 4-month follow-up. In multivariate analysis, GLS remained a significant predictor for functional outcome (4-month LVEF) (ß − 0.40, 95% CI − 0.68 to − 0.12, p = 0.006), together with baseline LVEF, LVEDV, and presence of microvascular obstruction (MVO) (Table 2). By the Kaplan-Meier analysis, the patients in the lowest quartile GCS (GCS > − 14%) had a significantly worse survival compared with patients with GCS < − 14% (Fig. 3). During a median follow-up of 4.0 (2.2–5.0) years, 13 patients died (6.5%). Univariate predictors for long-term mortality were LVEF, LVEDV, and GCS. In stepwise forward multivariate analysis, GCS remained the strongest predictor for mortality (HR 1.17, 95% CI 1.04–1.31, p = 0.009) (Supplement Table 3). In 74% of the patients, infarct data was available; when including infarct size and MVO in the model, infarct size was the only significant CMR parameter predicting mortality (HR 1.07, 95% CI 1.03–1.11, p < 0.001).

There were 2560 segments available for serial segmental circumferential strain analysis (from n = 160 patients). Segmental strain was significantly recovered from baseline to 4-month follow-up in the dysfunctional segments (wall thickening < 45% at baseline; ∆ segmental strain − 2.4 (8.9), p < 0.001; Table 3). However, no significant difference was found between CTO-PCI and no-CTO PCI. Table 3 shows the recovery of segmental strain in the dysfunctional segments in the CTO. In the CTO territory, no significant difference was found on segmental strain between CTO-PCI and no-CTO PCI (∆ segmental strain − 2.5 ± 9.5% versus − 2.5% ± 8.5%, p = 0.50).

At baseline, 59% of the myocardial segments were dysfunctional (wall thickening < 45%). Of the dysfunctional segments, 74% was viable (TEI < 50%) and MVO was present in 13%. In a multivariate analysis, baseline wall thickening, segmental strain, and infarct (TEI) were all statistically significantly related; however, based on the t statistic, wall thickening and segmental strain were stronger predictors for wall thickening recovery compared with infarct or MVO. In the CTO territory, 59% of the segments were dysfunctional and 96% of the dysfunctional segments were viable (TEI < 50%). In multivariate analysis, segmental strain was an independent predictor for regional wall thickening recovery, incremental to baseline wall thickening and infarct (coefficient − 0.44, SE 0.14, t − 3.10, p = 0.002; Table 4).

Although strain significantly recovered at follow-up, we did not find a beneficial effect of CTO-PCI compared with no-CTO PCI on this recovery, which is consistent with the main EXPLORE trial results. FT-CMR strain has substantial clinical potential to be of diagnostic and prognostic value. Global strain measurements are highly reproducible with good to excellent intra- and inter-reproducibility and analysis appears not to be influenced by the level of training [7, 18]. However, only limited data is available on reference values of strain in healthy subjects. LVEF and LVEDV are frequently used in predicting patient prognosis; nonetheless, they are limited as contractility is not measured and they are affected by patient heart rate, loading conditions, and heart valve function [19]. We found that GLS is an independent predictor for global LV function, incremental to LVEF and LVEDV. In a beating heart, there are two longitudinal movements, shortening and lengthening (mainly reflecting GLS), and two transversal movements, narrowing and widening (mainly reflecting GCS). The cardiac cycle consists of different phases as described in more detail by Torrent-Guasp et al [20]. In short, (1) a decrease in the transversal diameter of the base caused by basal loop contraction (narrowing movement), (2) a decrease in the longitudinal axis (shortening movement), (3) an increase in the longitudinal axis (lengthening movement), and (4) an increase in the transversal diameter of the base are conditioned by the relaxation of the ventricular walls and widening movement [20]. GLS is mainly determined by subendocardial myofibers (the basal loop), which are more prone to early myocardial damage before global LV parameters are affected. Therefore, diminished GLS is an early marker of LV dysfunction. This could explain why GLS was the best predictor of LV preservation at follow-up. GCS is largely based on contraction of circumferential myofibers, which stay mostly preserved during early LV deterioration and serve as a restriction to prevent expansion of the LV [21].

For mortality, infarct size remained the strongest predictor. Previous reports regarding the prognostic value of global strain have been conflicting. In a study with 470 ischemic and non-ischemic cardiomyopathy patients, GLS was a strong predictor for mortality incremental to LVEF and infarct size [22]. In 74 STEMI patients, GCS was able to predict preservation of global function (LVEF > 50%) at follow-up similar to infarct size [23]. However, in a study of 65 STEMI patients, GLS was not able to predict adverse LV remodeling [5]. In another study, GLS did not improve risk stratification compared with baseline characteristics and CMR indices in 323 STEMI patients [24]. In the largest strain study thus far (1,235 MI patients), GLS did have incremental prognostic value over LVEF and infarct size to predict mortality [25]. Nonetheless, most studies performed are of small sample sizes and with relative short follow-up.

Segmental strain significantly recovered at follow-up. Yet, there was no effect of CTO-PCI, compared with no-CTO PCI, on global nor on segmental strain recovery. However, we have previously shown that wall thickening significantly improves after CTO-PCI compared with no-CTO PCI in the dysfunctional CTO territory [11]. Although strain has the potential to detect more subtle regional differences, we could not reproduce this finding with segmental strain, although the number of segments was relatively low. Furthermore, it is important to mention that previous studies reported high degrees of measurement variability and relatively poor segmental strain reproducibility [26, 27]. Therefore, segmental strain data may be less reliable and its use in clinical practice should be done with caution.

We did find that segmental strain was a strong predictor for wall thickening recovery, incremental to infarct, especially in the dysfunctional CTO territory. This finding is consistent with previous data: in 45 STEMI patients, segmental strain was incremental to infarct and MVO in predicting recovery of wall thickening [28]. However, another study in STEMI patients showed that segmental strain was only a mild predictor of wall thickening recovery and inferior compared with infarct [29]. Furthermore, segmental strain and infarct transmurality are also related, as segmental strain is a predictor for infarcted segments [30]. Therefore, the incremental predictive value and exact relation of segmental strain and infarct needs further examination as, different from infarct (TEI), segmental strain can be measured without the use of a contrast agent, making it a possible alternative in patients with contrast allergy or renal failure.

In our high-risk patient population, new risk stratification parameters may be relevant to adequately select patients for CTO revascularization. None of the randomized CTO trails conducted showed a beneficial effect of CTO-PCI on clinical outcome nor on LVF, although ischemia and viability testing prior to inclusion were not mandated. It remains to be determined what the optimal patient selection threshold for CTO revascularization is and whether optimal selection will indeed lead to improved outcomes. Further studies are needed to investigate the true value of (stress) FT-CMR strain in clinical evaluation and whether it can be used as a diagnostic and prognostic tool to select patients that might benefit from PCI.

There are several limitations applicable to this study. Unfortunately, not all CMRs were suitable for strain analysis and not all patients underwent baseline CMR. Furthermore, infarct data was not available in all patients and only 4% of the dysfunctional segments in the CTO territory were non-viable (TEI > 50%), making sample size relatively small and underpowered. Infarct size was probably decreased by the presence of collaterals, which were present in > 90% of the patients. As with all randomized trials, patient selection has occurred. The mortality rate was relatively low in our study cohort with the risk of overfitting the model and EXPLORE was not powered for clinical outcomes.