In-hospital measurement of left ventricular ejection fraction and one-year outcomes in acute coronary syndromes: results from the IMMEDIATE Trial

This study analyzed a subset of data on participants enrolled in the IMMEDIATE Trial [8, 10], a randomized, placebo-controlled, double-blind clinical effectiveness trial of GIK conducted from December 2006 through July 2011, in which paramedics, aided by electrocardiograph-based decision support, enrolled 871 patients aged ≥ 30 years with high probabilities of ACS. Participants were given either GIK (30 % glucose, 50 U/L of regular insulin, and 80 mEq of KCl/L) intravenously at 1.5 mL/kg/h for 12 h, or identical-appearing placebo. This investigation included the subset of participants who had their LVEF measured during their index hospitalization by left ventriculography or echocardiography [8, 10].

Screened patients included all those transported by emergency medical services (EMS) in response to a 9-1-1 call for symptoms suggestive of ACS who were ≥ 30 years of age and had an out-of-hospital electrocardiogram (ECG) performed. Inclusion was based on paramedics’ clinical assessment of a patient likely having ACS, supplemented by decision support by the electrocardiograph-based Acute Cardiac Ischemia Time-Insensitive Predictive Instrument (ACI-TIPI) and Thrombolytic Predictive Instrument (TPI). Patients were candidates for enrollment if the ACI-TIPI prediction of ACS was 75 % or higher, if STEMI was detected by the TPI, and/or the patient met local standards for EMS identification of STEMI [10]. Patients were excluded if they had a language barrier or impaired reasoning, were prisoners or pregnant, or had clinically significant rales (Killip Class 3 or 4 HF) [8, 10].

In addition, a National Institutes of Health appointed Data Safety Monitoring Board oversaw enrollment to ensure safe and ethical study conduct throughout the trial. Informed consent was obtained from all patients in accordance with the Exception from Informed Consent Requirements for Emergency Research per the Code of Federal Regulations 21 CFR 50.24 and included community consultation, institutional review board approval from all sites, assent from patients prior to randomization, and written consent once stabilized at the hospital [11].

The Coordinating Center’s IRB at Tufts Medical Center provided approval for the overall trial along with all participating sites (University of Texas Southwestern Medical School, Dallas; Cambridge Health Alliance, Cambridge, Massachusetts; Alaska Regional Hospital, Anchorage; University of New Mexico School of Medicine, Albuquerque; Medical College of Wisconsin, Milwaukee; Medical Center of Central Georgia, Macon; Regions Hospital EMS, St. Paul, Minnesota; Avera Medical Group, Sioux Falls, South Dakota; Penn State Hershey Medical Center, Hershey, Pennsylvania; Emerson Hospital, Concord, Massachusetts; St Joseph Medical Center, Bellingham, Washington; Texas Tech University Health Sciences Center, El Paso; Yale New Haven Hospital, Connecticut)

Data were collected by trained study staff. They were instructed to record LVEF as measured by cardiac catheterization; if the catheterization was not done, and LVEF was available by echocardiogram, this was recorded. The dates and times of echocardiograms were not routinely recorded, so medical records were reviewed to obtain them where possible. Based on emergency department (ED) presentation date and dates of LVEF measurement by left ventriculography or echocardiography, we classified a participant’s LVEF as “early” (within 24 h of ED presentation), or “not early” if measured more than 24 h after ED presentation. We could not find the date and time of LVEF measurement for 94 out of 445 participants; for this study we assigned them to the not early group, as the majority (92) had echocardiograms, and our search showed that the majority of the echocardiograms were done more than 24 h after ED presentation.

Statistical analyses were performed using R, version 2.15.2. Tests were two-sided, using alpha ≤ 0.05 as statistical significance. Comparison of means of in-hospital LVEF between the GIK and placebo groups, between early and not early LVEF measurement, and between catheterization and echocardiogram measurement all used two-sample student t-tests. Just for the Kaplan–Meier survival curves, they were plotted for participants with LVEFs in normal (55–70 %), mildly abnormal (40–54 %), moderately abnormal (25–39 %) and severely abnormal (<25 %) categories. Cox proportional hazards models were used to estimate univariate hazard ratios (HRs) for LVEF associated with the composite outcome of death or hospitalization for HF. Based on clinical and statistical significance in those analyses, we chose candidate variables for possible inclusion in a multivariable model to estimate adjusted HRs for the composite outcome of death or hospitalization for HF at 1 year. Using the final adjusted model, we also tested for dichotomous variables of timing of LVEF (early vs. not early) and the measurement method (catheterization vs. echocardiogram). Two-way interactions between LVEF and timing (early vs. not early) were tested in the Cox proportional hazard model of the composite outcome, adjusting for age and coronary artery disease (CAD). A similar analysis was done to test the two-way interaction between LVEF and test method (catheterization vs. echocardiogram). We checked assumptions needed for proportional hazards analyses by plotting and testing Schoenfeld residuals as related to time.

Table 1 shows demographic and clinical features of participants having the composite outcome of death or hospitalization for HF at 1 year (n = 52) and those who did not (n = 393). Participants with the composite outcomes were older and more frequently had medical histories of CAD, HF, diabetes, hypertension, stroke, and hyperlipidemia. Rates of acute myocardial infarction were similar in both groups, but those participants that had the composite outcome presented with a significantly higher Killip Class. Please see Appendices below.

Data on LVEF measured by left ventriculography or echocardiography during the index hospitalization were available for 445 of 871 study participants. There were significant differences in baseline demographics and clinical characteristics of participants with and without in-hospital LVEFs recorded. Of participants with in-hospital LVEF measured, 92.6 % had confirmed diagnoses of ACS, compared with 33.1 % among those without measured LVEF. There were no significant differences between characteristics of participants who had in-hospital LVEF recorded by left ventriculography vs. echocardiography. For 248 participants, LVEF was measured early (219 [80.1 %] by catheterization), defined as measurement within 24 h of presentation, and for 197 participants, the LVEF measurement was performed not early (54 [19.7 %] by left ventriculography).

Participants with the lowest LVEF had the lowest survival and the greatest incidence of composite outcomes (Fig. 1). Univariate Cox proportional hazard models were used to identify factors associated with the composite outcome (Table 2). A reduction of LVEF by 5 % decrements was significantly associated with higher hazards of death or hospitalization for HF over 1 year (HR = 1.31, 95 % CI: 1.20–1.45, P < 0.001). Also significant was age, expressed in 10-year increments (HR = 1.79, 95 % CI: 1.44–2.24, P < 0.001), prior CAD (HR = 4.46, 95 % CI: 2.47–8.04, P < 0.001), HF (HR = 3.56, 95 % CI: 1.86–6.79, P < 0.001), diabetes (HR = 1.88, 95 % CI: 1.06–3.34, P = 0.030), hypertension (HR = 2.17, 95 % CI: 1.12–4.23, P = 0.022), and stroke (HR = 2.56, 95 % CI: 1.20–5.43, P = 0.015).

Multivariate Cox models were used to better characterize the association of LVEF with the composite outcome. In stepwise regression, candidate parameters for inclusion included LVEF expressed by 5 % decrements; age, sex, history of CAD, HF and diabetes. The model that had the lowest akaike information criteria (best fit) following the selection process included only LVEF, age, and CAD. The C statistic (which is equivalent to the receiver-operating characteristic (ROC) curve area) for that model is 0.805 (SE = 0.04).

The effect of 5 % lower LVEF on 1-year death or hospitalization for HF remained statistically significant after adjusting for age and history of CAD (HR = 1.26; 95 % CI: 1.15–1.38, P < 0.001) when LVEF was used as a continuous variable. When LVEF was tested as a binary variable and dichotomized, participants with LVEF < 40 % had higher hazard of 1-year mortality and hospitalization for HF compared to participants with LVEF ≥ 40 %, after adjusting for age and history of CAD (HR = 3.59; 95 % CI: 2.05–6.27, P < 0.001). The proportional hazard assumption was not violated. Separating the composite outcome into its components, reductions of LVEF by 5 % decrements were also significantly associated with higher hazard of death and hospitalization for HF when adjusted for age and history of CAD (HR = 1.18; 95 % CI: 1.06–1.32, P = 0.002; and HR = 1.37; 95 % CI: 1.18–1.58, P < 0.001, respectively) (Table 2).

The HRs for the association of LVEF with the study outcomes were similar whether measured by left ventriculography or by echocardiography, (respectively, HR = 1.32; 95 % CI 1.15, 1.51 and 1.21; 95 % CI 1.106, 1.35) and whether done within 24 h or not within 24 h (respectively, HR = 1.28; 95 % CI 1.10, 1.50 and 1.23; 95 % CI 1.10, 1.38). Tests for interactions of LVEF and measurement method and LVEF and timing did not reach significance (Table 3). Controlling for method of LVEF measurement (left ventriculography vs. echocardiography) or timing of measurement (early vs. not early) did not alter the association of LVEF with the outcome (Table 4). Additionally, while having an echocardiogram (vs. left ventriculography) and being not early (vs. early) were associated with the 1-year outcome, neither was statistically significant.

Sensitivity analyses were performed in which we re-estimated HRs for LVEF, adjusting not only for already specified covariates age and CAD history, but also for sex, HF history, and diabetes. Reductions of LVEF by 5 % were significantly associated with higher hazard of the 1-year composite outcome (HR = 1.25; 95 % CI: 1.14–1.38, P < 0.001). Also, the HRs for the association of LVEF with the study outcome when measured by left ventriculography vs. by echocardiography, and for when measured early vs. not early, were similar. Finally, we studied the association of the IMMEDIATE Trial treatment, GIK vs. placebo, on in-hospital LVEF, stratified by method (catheterization vs. echocardiogram) and timing (early vs. not early. There was a trend toward higher in-hospital LVEF among those receiving GIK. When measured by echocardiography alone, median LVEF with GIK 50 % vs 45 % with placebo (P = 0.08). When measured by either echocardiography or catheterization, the median LVEF with GIK was 50 % vs 45 % with placebo (P–0.07) (Table 5).