Consecutive patients admitted to the coronary care unit of Anzhen Hospital with an initial diagnosis of ASTEMI were screened for inclusion. The study was conducted in accordance with the Declaration of Helsinki and approved by the Ethics Committee of Beijing Anzhen Hospital. Written informed consent was obtained from all participants. Between June 2016 and December 2022, a total of 296 patients who underwent gadolinium-enhanced CMR following primary PCI were evaluated. Inclusion criteria: (1) diagnosis of ASTEMI according to the Fourth Universal Definition of Myocardial Infarction [16] and (2) completion of CMR within 3–7 days of hospital admission. Exclusion criteria: (1) known cardiomyopathy, congenital or valvular heart disease, pericardial disease, or severe arrhythmia; (2) incomplete or missing essential laboratory data including brain natriuretic peptide (BNP) and creatine kinase-myocardial band (CK-MB); (3) insufficient follow-up information; and (4) poor image quality.

All examinations were performed on 3.0 T cardiovascular MR systems, including the Achieva (Philips Healthcare, Best, Netherlands) and the Discovery MR750w (GE Healthcare, Milwaukee, WI, USA), both equipped with a 32-channel phased-array cardiac coil, electrocardiographic (ECG) gating, and respiratory navigation. The standard imaging protocol featured steady-state free-precession (SSFP) cine sequences during breath-holding, T2-weighted short-axis imaging, and late gadolinium enhancement (LGE). A cine CMR study was conducted using an SSFP sequence, which captured contiguous short-axis images from the mitral annulus to the apex, covering both ventricles. Long-axis views in two-, three-, and four-chamber planes were also acquired, with 25–30 reconstructed cardiac phases per cycle. T2-weighted short-axis imaging used a short tau inversion recovery technique. Myocardial perfusion data were collected during the intravenous administration of 0.1 mmol/kg gadolinium-based contrast at a rate of 4 mL/s. For LGE, prospectively ECG-gated gradient-echo images in short- and long-axis planes were obtained 10–15 min after injecting 0.2 mmol/kg of contrast (TR/TE = 4.1/1.6 ms; flip angle = 20°; matrix size = 256 $\times$ 130).

All CMR datasets were transferred to a dedicated workstation and analyzed using commercial software CVI42 (version 5.2.0, Circle Cardiovascular Imaging Inc., Calgary, Canada). Ventricular function was evaluated using the Short 3D module, which semiautomatically traced endocardial and epicardial borders at end-diastolic and end-systolic frames on cine short-axis stacks, including papillary muscles. Contours were visually inspected and manually adjusted by two experienced cardiovascular radiologists ($>$10 years of experience). Quantitative indexes, including left-ventricular ejection fraction (LVEF), end-diastolic volume (LVEDV), end-systolic volume (LVESV), stroke volume (LVSV), and left-ventricular mass (LVM), were automatically derived by the software. Left-atrial volume (LAV) was obtained by manual endocardial contouring in both four- and two-chamber views, and the maximal LAV was calculated at LV end-systole (just before mitral valve opening) using the biplane area-length method.

LGE was identified as regions with signal intensity of $>$5 standard deviations above normal myocardium on short-axis images. The infarct size was expressed as LGE mass normalized to the total LV mass. Hypointense cores within areas of hyperenhancement on LGE, indicating microvascular obstruction (MVO), were classified and reported as a percentage of LV mass. Zones of hypointensity on T2-weighted short tau inversion recovery sequences that match infarct regions were considered to represent intramyocardial hemorrhage (IMH).

A CMR-based model estimating LVFP from LAV and LVM has been recently proposed. The equation was developed using data from a cohort of 835 individuals evaluated for suspected heart failure who underwent CMR, echocardiography, and invasive hemodynamic assessment [15]. In this study, the formula was applied to calculate LVFP values using measured LAV and LVM:

LVFP_cmr = 6.1352 $+$ (0.07204 $\times$ LAV) $+$ (0.02256 $\times$ LVM)

Composite endpoint events were identified by reviewing Anzhen Hospital’s electronic medical records, with additional follow-up via telephone for events that occurred after discharge. MACE was defined as a composite endpoint including all-cause death, recurrent myocardial infarction, or heart failure requiring rehospitalization following AMI. If a patient experienced multiple events, only the most severe one was recorded, based on the following hierarchy: death, reinfarction, then heart failure. Each patient was counted once for a MACE event in the analysis.

Continuous data were expressed as the mean $\pm$ standard deviation when normally distributed and as the median with interquartile range (IQR) when non-normally distributed. Categorical variables were presented as counts and percentages. Baseline characteristics between patients with and without MACE were compared using the independent t-test or Wilcoxon rank-sum test for continuous variables and the chi-square test for categorical variables. Associations between continuous parameters were evaluated using Spearman’s rank correlation. The discriminative performance of LVFP_cmr in predicting MACE was determined by performing receiver operating characteristic (ROC) analysis, where the area under the curve (AUC) and Youden index were used to calculate the optimal cutoff. Event-free survival was analyzed using the Kaplan-Meier method, and intergroup differences were assessed with the log-rank test. Univariate Cox proportional hazards models were used to estimate hazard ratios (HRs) for MACE and mortality. Variables with p 0.05 in univariate testing were subsequently entered into multivariable Cox regression using a stepwise selection approach, and those with p 0.05 were retained. In addition, a hierarchical composite outcome analysis was performed using the win ratio (WR) method (death $>$ reinfarction $>$ heart failure hospitalization). Patients were compared in all possible unmatched pairs. A pair was declared a “win” for the patient when a higher-priority event occurred later, and ties moved on to the next level. The WR was calculated as total wins divided by total losses, with 95% confidence intervals (CIs). All statistical analyses were performed using SPSS (version 29, Statistical Package for the Social Sciences, International Business Machines, Armonk, NY, USA) and R software (version 4.4.2; R Foundation for Statistical Computing, Vienna, Austria).

A sex-specific CMR-derived PCWP (sex-specific LVFP_cmr) was additionally computed from LAV and LVM, with sex coded as female = 0 and male = 1, as proposed by Garg et al. [17]: Sex-specific LVFP_cmr = 5.7591 + 0.07505 $\times$ LAV + 0.05289 $\times$ LVM – 1.9927 $\times$ sex. The equation was validated using PCWP measurements obtained via RHC in heart failure cohorts. Compared with the generic equation, it demonstrated a better reduction of sex-related bias and enhanced the prognostic accuracy [17]. Using identical censoring rules, endpoints, and covariate adjustment as the main analysis, univariable and multivariable Cox models were fit, replacing the generic LVFP_cmr with the sex-specific LVFP_cmr. Subsequently, the optimal cut-point was derived via the Youden index for Kaplan-Meier curves. To avoid multicollinearity, entering the generic and sex-specific metrics in the same model was avoided.

In summary, this single-center retrospective study included 296 patients with ASTEMI who underwent CMR after primary PCI. Quantitative CMR parameters such as LAV, LVM, and derived LVFP_cmr were examined in relation to future MACE. The prognostic value of LVFP_cmr was evaluated using Kaplan-Meier survival curves and Cox regression, adjusting for key clinical and imaging factors. These analyses aimed to determine if LVFP_cmr provides independent prognostic insights beyond conventional measures, such as LVEF and infarct size.

A total of 309 patients initially met the inclusion criteria (Fig. 1). Of these, 7 lacked 6-month follow-up data, and 6 had poor CMR image quality. The final cohort comprised 296 patients, with a median age of 58 years (IQR, 49–66); 248 (83.78%) were male, and 48 (16.22%) were female. The last follow-up was in December 2024, with a median duration of 1563 days (IQR, 1442–1714). Of these, 38 patients (12.84%) experienced MACE, including 8 deaths from all causes, 12 reinfarctions, and 18 hospitalizations for heart failure. The primary baseline clinical and CMR characteristics are summarized in Tables 1,2.

Fig. 1.

Flowchart of patient inclusion. MACE, major adverse cardiovascular events.

Compared with the group that did not experience MACE, those with MACE displayed significantly reduced LVEF (38.49% $\pm$ 14.85 vs. 49.47% $\pm$ 11.53; p 0.001). Moreover, these patients exhibited higher LVEDV (148.73 mL [121.72, 174.10] vs. 123.00 mL [97.35, 145.64]; p 0.001), LVFP_cmr (14.57 mmHg [13.17, 15.99] vs. 13.30 mmHg [12.05, 14.51]; p 0.001), and infarct size (% LV mass) (35.73% [27.56, 48.04] vs. 27.68% [18.48, 36.58]; p 0.001). In addition, B-type natriuretic peptide (BNP) levels were substantially higher in the MACE cohort (283.00 pg/mL [117.00, 486.00] vs. 166.50 pg/mL [81.00, 294.00]; p 0.01).

The cross-sectional associations between LVFP_cmr and chamber structure/function were explored using Spearman’s rank correlation ($\rho$) (Table 3). LVFP_cmr demonstrated the strongest correlations with LV mass ($\rho$ = 0.62; p 0.001) and LVEDV($\rho$ = 0.59; p 0.001), followed by LVESV ($\rho$ = 0.51; p 0.001). Moderate positive relationships were noted with stroke volume and cardiac output ($\rho$= 0.31 and 0.25, respectively; both p 0.001). As expected for a filling-pressure marker, LVFP_cmr was inversely associated with LAEF and LVEF ($\rho$ = –0.27 and –0.24; both p 0.001). Correlations with infarct size were small but significant ($\rho$ = 0.15; p = 0.008). Measures of microvascular injury showed weak yet significant associations: extent of MVO ($\rho$ = 0.19; p 0.001), MVO present ($\rho$ = 0.14; p = 0.020), and IMH present ($\rho$ = 0.15; p = 0.012). Scatter plots depicting the association between LVFP_cmr and LA/LV structural indexes are presented in Fig. 2. Consistent with the Spearman correlation coefficients, the visual patterns suggested small to moderate monotonic relationships for most parameters. A correlation matrix plot of LVFP_cmr and major CMR-derived volumetric and functional parameters is shown in Fig. 3.

Fig. 2.

Scatter plots showing the correlations between LVFP_𝐜𝐦𝐫 and cardiac structural/functional parameters. (A) LAEF, (B) LVEF, (C) LVEDV, (D) LVESV, (E) SV, and (F) CO. Spearman’s correlation coefficients ($\rho$) and p values are displayed. Shaded areas represent 95% confidence intervals of the fitted regression lines.

Fig. 3.

Correlation matrix plot of LVFPcmr and key CMR-derived parameters. The correlation matrix displays pairwise Spearman correlation coefficients among LVFP_cmr, LVEF, LVEDV, LVESV, SV, and CO. Circle size and color intensity represent the magnitude and direction of the correlation (blue = positive; red = negative).

The LVFP_cmr demonstrated considerable variability within the cohort; the median LVFP_cmr value in the overall cohort was 13.37 mmHg (IQR, 12.16–14.72 mmHg). The AUC of LVFP_cmr for MACE was 0.67 (95% CI, 0.58–0.76). Based on ROC curve analysis, an optimal cutoff of 14.30 mmHg (sensitivity 0.58, specificity 0.71) was identified to stratify patients into a low-LVFP_cmr group (n = 198) and a high-LVFP_cmr group (n = 98). The high-LVFP_cmrgroup exhibited a significantly higher incidence of MACE than the low-LVFP_cmr group (22.45% vs. 8.08%, p 0.001). Fig. 4 presents the CMR findings and corresponding clinical data from patients with and without MACE. Kaplan-Meier curves showed significantly lower MACE-free survival in patients with higher LVFP_cmr, lower LVEF, and larger infarct size (all log-rank p 0.001; Fig. 5). Optimal cutoffs for variable dichotomization—specifically for LVEF and infarct size (% of LV mass)—were derived from ROC analysis by maximizing the Youden index in the present cohort, followed by application to Kaplan-Meier and Cox analyses.

Fig. 4.

Representative CMR findings and their relationship with MACE. (A,B) Patient 1, who died during follow-up (age 40 years). (A) Short-axis end-diastolic cine image with LV endocardial (red) and epicardial (green) contours used to calculate LV mass (LVM 325.15 g). (B) Four-chamber cine image with left-atrial contour (pink) used to calculate left-atrial volume (LAV 135.08 mL). The derived LVFP_cmr was 23.20 mmHg. (C,D) Patient 2, who survived (age 53 years). (C) Short-axis cine image with LV mass measurement (LVM 86.06 g). (D) Four-chamber cine image with left-atrial contour for volume assessment (LAV 30.36 mL). The derived LVFP_cmr was 10.26 mmHg. LVM, left-ventricular mass.

Fig. 5.

Kaplan-Meier curves of MACE-free survival. (A) Stratified by LVFP_cmr (low vs. high using the study cutoff of 14.30 mmHg). (B) Stratified by LVEF (36.14% vs. $\ge$36.14%). (C) Stratified by infarct size (30.83% vs. $\ge$30.83% of LV mass). (D) Stratified by GRACE risk category (low, intermediate, high). Shaded bands denote 95% confidence intervals. GRACE, Global Registry of Acute Coronary Events.

In univariable Cox regression, higher LVFP_cmr was significantly linked to MACE (HR, 1.31; 95% CI, 1.14–1.51; p 0.001). In multivariable analysis, LVFP_cmr remained an independent predictor of MACE (HR, 1.25; 95% CI, 1.07–1.46; p 0.01). Furthermore, LVEF (per 1% increase: HR, 0.96; 95% CI, 0.93–0.99; p = 0.017), infarct size (% LV mass; per 1% increase: HR, 1.03; 95% CI, 1.01–1.05; p = 0.017), and GRACE risk category (high vs. low risk: HR, 4.72; 95% CI, 1.77–12.19; p 0.01) were independently associated with MACE (Table 4). The Harrell C-index of the model incorporating LVFP_cmr, along with clinical and imaging covariates, was 0.77 (95% CI, 0.68–0.88), indicating robust discriminatory ability for predicting MACE.

In the hierarchical composite analysis, the high-LVFP_cmr group displayed significantly worse outcomes than the low-LVFP_cmr group. The WR was 0.35 (95% CI, 0.33–0.37), signifying that patients with elevated LVFP_cmr experienced more severe adverse events earlier and more frequently. These results highlight the prognostic significance of LVFP_cmr beyond time-to-first-event analysis.

The sex-specific LVFP_cmr equation showed that higher sex-specific LVFP_cmr was associated with increased risk of MACE in univariable analysis (HR, 1.18; 95% CI, 1.09–1.29; p 0.001) (Table 4). In multivariable analysis, sex-specific LVFP_cmr remained an independent predictor of MACE (HR, 1.17; 95% CI, 1.07–1.28; p 0.001) (Table 5). The AUC of sex-specific LVFP_cmr for MACE was 0.67 (95% CI, 0.57–0.76). The optimal cut-point (14.45 mmHg) separated Kaplan-Meier curves with significantly lower event-free survival in the high sex-specific LVFP_cmr group (log-rank p 0.001) (Fig. 6). Therefore, the sex-specific equation yields comparable prognostic power in our ASTEMI cohort.

Fig. 6.

Kaplan-Meier survival curves stratified by sex-specific LVFP_cmr.

However, our study has certain limitations that should be acknowledged. Our findings, from a retrospective single-center study, are subject to inherent selection and information biases. Patient inclusion was limited to those who underwent clinically indicated CMR after PCI, which might have introduced referral bias toward more stable or evaluable cases. Additionally, the small sample size and limited number of outcome events could have led to model overfitting and decreased statistical robustness. Consequently, the discriminative ability and robustness of our multivariable models should be validated in larger, multicenter prospective studies. Furthermore, this study did not incorporate emerging CMR parameters such as myocardial strain, native T1/T2 mapping, or extracellular volume. Therefore, whether these advanced tissue characterization metrics add incremental prognostic value beyond LVFP_cmr could not be assessed, and their mechanistic relevance could not be explored. Future research should incorporate multicenter recruitment with standardized CMR acquisition and postprocessing protocols to minimize institutional bias and enable robust external validation of the prognostic value of LVFP_cmr.

In this study, a detailed prognostic assessment of post-ASTEMI patients who underwent CMR was conducted, combining conventional clinical parameters with CMR-derived measures, including LVFP_cmr. The findings established that elevated LVFP_cmr was associated with unfavorable outcomes and served as an independent marker for risk stratification. The proposed LVFP_cmr metric could be incorporated into post-PCI risk assessment protocols, particularly in patients undergoing early CMR. Its noninvasive estimation allows for routine integration without additional scan time, potentially guiding follow-up intensity and the selection of adjunctive therapy. In clinical practice, an elevated LVFP_cmr value may help identify patients who require intensified follow-up, closer monitoring for heart failure progression, or optimization of guideline-directed medical therapy. These findings provide a preliminary framework to identify high-risk post-ASTEMI patients and may inform future studies on surveillance strategies and targeted interventions.