The STREAM-1 and STREAM-2 findings have been previously reported. ^, The trials were performed in accordance with the Declaration of Helsinki principles. The protocols were approved by national regulatory authorities and local ethics committees at each of the participating centers, and patients gave informed consent.

In the current report, we examined all STEMI patients undergoing PPCI: they received standard of care, including antiplatelet and anticoagulant therapies according to guideline-based local practice in both studies. Patients with STEMI in STREAM-1 of any age ( n = 1,892) presenting within 3 hours of symptoms, were randomly assigned to receive either conventional full-dose tenecteplase (with rescue intervention, if needed) and scheduled coronary angiography 6 to 24 hours after randomization (pharmacoinvasive strategy), or to PPCI. In STREAM-2 ( n = 604), older patients with STEMI ≥60 years presenting within 3 hours of symptoms were randomly assigned (2:1) to receive half-dose tenecteplase (with rescue intervention, if needed) and scheduled coronary angiography 6 to 24 hours after randomization (pharmacoinvasive strategy) or to PPCI. In both studies, patients were randomized either prehospital in ambulances ( n = 832) or in community hospitals ( n = 285). In STREAM-2, a higher proportion of patients were randomized at community hospitals (52.4 vs 19.2%) in STREAM-1.

In the current study, only those patients assigned to PPCI in both STREAM-1 and STREAM-2 studies were included, provided time from randomization-to-sheath insertion (employed as a proxy for first medical contact-to-device time in both STREAM trials) was available (97%). Notwithstanding the studies’ intent to exclude patients who were able to undergo PPCI within 1 hour ( Figure 1 ), 29% underwent PPCI within 1 hour of randomization (277 patients in STREAM-1 and 49 patients in STREAM-2).

Consort diagram. PPCI, primary percutaneous coronary intervention; STREAM, STrategic Reperfusion Early After Myocardial Infarction; S1, STREAM-1; S2, STREAM-2.

Both trials also underwent similar core laboratory ECG measurements of reperfusion and the endpoints of interest were: the proportion of patients with worst lead ≥50% ST segment elevation resolution, as well as resolution of sum ST deviations (binary ≥50%) and residual ST deviation (continuous millimeters) following angiography. Baseline ECG, postreperfusion, and discharge measurements were interpreted at a core laboratory blinded to 30-day clinical outcomes. The presence of Q waves was defined using the Selvester QRS screening criteria : a Q wave of ≥30 ms in lead aVF (inferior); ≥40 ms in leads I and aVL (lateral); or ≥40 ms in ≥2 leads V4, V5 or V6 (apical); or any Q wave ≥20 ms or QS complex in leads V2 and V3 (anterior). Q wave equivalents were defined as follows: R wave ≥40 ms in V1 (posterior), or R wave ≤0.1 mV and ≤10 ms in V2 (anterior). ST segment deviation was measured at the J point with magnified calipers. The total ST deviation was calculated by adding the sum of ST-elevations and ST depressions. Worst lead ST-elevation resolution was also reported and dichotomized as >50% or <50% based on guideline recommendations for reperfusion assessment. Selvester 54-criteria/32-point QRS scoring system, a validated ECG estimate of infarct size, was measured at hospital discharge. The Cardio Calipers version 3.3 program (Iconico Inc., New York, NY) performed electronic measurements. When there was uncertainty regarding an ECG measurement, a consensus process between readers was used for final ECG reporting.

Angiographic findings of interest were thrombolysis in myocardial infarction (TIMI) flow grade of the culprit vessel at first and last coronary angiography, as reported by the site investigators.

Patient baseline characteristics, ECG and angiographic measurements were reported according to time from randomization-to-sheath insertion and categorized by whether this was <60 minutes (early PPCI) versus greater or ≥60 minutes. Categorical variables were summarized as numbers (percentages), whereas continuous variables were presented as medians with 25th and 75th percentiles. Comparisons of baseline variables between trials and within trials were conducted using Chi-square tests for categorical variables and Wilcoxon rank sum tests for continuous variables.

Clinical efficacy and safety outcomes at 30 days and mortality at 1-year are reported as observed rates. Mortality within 1 year is displayed using Kaplan–Meier curves according to early PPCI with logrank test. The relative association of early PPCI with the clinical efficacy outcomes are summarized as a relative risk (RR) with 95% confidence interval (CI), calculated from a Poisson regression model with robust variance. In the pooled cohort, the studies (STREAM-1 and-2) were included as covariates. Unadjusted and adjusted estimates are reported, with the latter adjusted for the simplified TIMI Risk Score and radial artery access. Inverse probability weighting (IPW) was used to account for selection bias and imbalance of baseline characteristics between early PPCI and late PPCI groups. To establish the weights, a logistic regression model for early vs late PPCI was first developed by including relevant patient characteristics available at randomization (ie, age, sex, history of diabetes, hypertension, prior MI and study). Subsequently, the inverse of predicted probabilities was used as a weight in the Poisson regression model for the clinical outcomes (ie, clinical efficacy endpoints at 30 days and 1-year mortality). Assessment of heterogeneity in early/late PPCI on clinical outcomes between trials was also undertaken.

The relationship of early PPCI to cardiogenic shock and Q waves at hospital discharge was also assessed using time from randomization-to-sheath insertion as a continuous variable. To investigate the relationship visually, predicted events rates with 95% confidence interval (CI) limits were plotted according to time to sheath insertion (44 patients with times greater than 180 minutes were truncated to 180 minutes) using the restricted cubic splines with 4 knots in a logistic regression model.

There was no imputation for missing data. All statistical analyses were conducted using SAS (version Base 9.4; Cary, NC). P -values for two-sided tests are provided for descriptive purposes.