This study analyzed a prospective cohort of 460 patients who presented with STEMI (PHRC, N° 2006/0070). Patients were consecutively included from January 2006 to August 2018. Inclusion criteria: STEMI managed with coronary angioplasty within 12 hours of symptom onset; successful angioplasty with stenting; written informed consent. Exclusion criteria: prior myocardial infarction or coronary artery bypass surgery; age <18 years; clinical signs of cardiogenic shock; major comorbidities limiting life expectancy; contraindications for CMR (pacemaker, ferromagnetic body, claustrophobia, or severe renal insufficiency). The study adhered to the Helsinki Declaration principles. All patients provided informed consent before undergoing CMR, and the study protocol was approved by the hospital’s ethics committee.

The procedure was performed by an interventional cardiologist. TA was initiated after the lesion crossed by the guidewire. A minimum of two 40 ml syringes was recommended. To avoid systemic embolic risk, selective catheterization beyond the coronary ostia was performed before removing the TA catheter. The guide catheter was also aspirated.

Thrombus was classified as follows for the purpose of quantitative evaluation: TTS as follows:

The TTS classifies thrombus burden from 1 to 5. Scores “1 to 4” reflect increasing thrombus size without full occlusion, ranging from suspected thrombus to large thrombus >2 vessel diameters. Score “5” indicates complete vessel occlusion. For analysis, scores “1 to 4” were grouped as nonocclusive thrombus, and score “5” as occlusive.

Longitudinal follow-up with CMR was conducted. The initial exam was performed on day 6 after admission [IQR: 4; 9] and repeated at 98 days [IQR: 94; 106]. CMR machines of 1.5 or 3 Tesla power (Avanto and Skyra, Siemens Medical Solutions, Forchheim, Germany) were used. Long-axis and short-axis views were used to assess segmental and global kinetics (30-phase count). Contiguous short-axis slices (8 mm thickness) covered the entire left ventricle. Late enhancement sequences were acquired 8 to 10 minutes after the injection of 0.2 mmol/kg gadoterate meglumine (Dotarem, Guerbet Laboratories, Roissy-Charles de Gaulle, France). Inversion time was adjusted to suppress the signal from presumed healthy myocardium. Standard parameters included: repetition time = 4.9 ms, excitation time = 1.9 ms, flip angle = 15°, slice thickness = 8 mm, spatial resolution = 1.35 × 1.35 × 8 mm³. T2-weighted inversion recovery prepared fast-spin echo sequence was performed on a stack of contiguous short-axis views covering the entire LV in mid-diastole; slice thickness/gap 7/0.8 mm; TR (4,000 ms); TE 60 to 70 ms; FOV 264 × 385; matrix 176 × 320; flip angle 180°.

CMR images were transferred to a dedicated workstation for blinded analysis (Medis Suite, Leiden, The Netherlands). Endocardial and epicardial contours were manually traced on all short-axis views, excluding trabeculations and papillary muscles. End-diastolic and end-systolic volumes, and left ventricular mass were indexed to body surface area. Infarct size was calculated using the “full width at half maximum” method on late enhancement images. Late enhancement zones were summed to determine the percentage of left ventricular mass affected.

MVO was defined as a hypointense area within the hyperintense infarct zone on late enhancement sequences. MVO extent was manually measured and expressed in grams. Reproducibility of this measurement has been published.

IMH was identified on T2-weighted images as a region of hypointense signal within the hyperintense area of myocardial edema.

Variables will be described as a mean ± a standard deviation if the variable follows a normal distribution and a median [interquartile range] otherwise. Normality has been validated graphically through appropriate plots. For qualitative variables, they will be presented as numbers (proportions). Groups will be compared according to standard tests: Student’s test ( t test) or Wilcoxon’s nonparametric test for quantitative variables based on the distribution of variables and chi-square test ( χ ²) or Fischer’s test ( F -test) for qualitative variables. These tests comparing the two groups were performed descriptively.

To estimate the effect of thrombus aspiration (TA) on coronary microcirculation in STEMI patients, we used a propensity score-based inverse probability weighting approach with the average treatment effect (ATE) estimand. This allowed us to balance baseline covariates between patients who received TA and those managed with other reperfusion strategies, thereby emulating a randomized treatment assignment ( Figure 1 and Supplementary Figure 3 ). The ATE estimand aims to quantify the average effect of TA in the overall study population, independent of actual treatment allocation. In other words, it addresses the counterfactual question: “How would outcomes, on average, have differed if all patients had received TA versus none?”

Impact of thromboaspiration (TA) on microvascular obstruction (MVO) assessed by cardiac magnetic resonance (CMR): in 460 patients with ST elevation myocardial infarction (STEMI). Using an average treatment effect (ATE) propensity score, TA emerged as an independent predictor of MVO occurrence (OR 1.52; 95% CI: 1.16 to 1.98; p = 0.0024) and its extent (standardized mean difference [SMD] 0.42; 95% CI: 0.02 to 0.72; p = 0.041). Subgroup analyses showed that TA significantly increased MVO risk when performed more than 6 hours after symptom onset (OR 3.46; 95% CI: 1.92 to 6.23; p < 0.0001) and in patients with nonocclusive thrombus (TTS “1 to 4”) (OR 2.23; 95% CI: 1.29 to 3.85; p = 0.004). In sex-based analysis, TA was associated with a higher MVO risk (OR 1.52; 95% CI: 1.14 to 2.05; p = 0.005) and extent (SMD 0.44; 95% CI: 0.006 to 0.747; p = 0.049) in men, but not in women (OR 0.74; 95% CI: 0.38 to 1.46; p = 0.386; SMD 0.93; 95% CI:–0.166 to 2.019; p = 0.12).

Once covariate balance was achieved, weighted logistic regression models were used for binary outcomes, yielding odds ratios (OR) and 95% confidence intervals (CI), while weighted linear regression models were used for continuous outcomes, yielding standardized mean differences and 95% CI. Stratified analyses were also performed according to sex, symptom-to-treatment time (TTS), and treatment delay to explore potential effect modifications. All results were summarized using forest plots for visual comparison of treatment effects across outcomes. Statistical analyses were performed using R software (version 4.4.1; R Foundation for Statistical Computing, Vienna, Austria).