The aim of this study was to evaluate the independent prognostic value of qualitative and quantitative admission electrocardiographic (ECG) analysis regarding long-term outcomes after non–ST-segment elevation acute coronary syndromes (NSTE-ACS). From the Fragmin and Fast Revascularization During Instability in Coronary Artery Disease (FRISC II), Invasive Versus Conservative Treatment in Unstable Coronary Syndromes (ICTUS), and Randomized Intervention Trial of Unstable Angina 3 (RITA-3) patient-pooled database, 5,420 patients with NSTE-ACS with qualitative ECG data, of whom 2,901 had quantitative data, were included in this analysis. The main outcome was 5-year cardiovascular death or myocardial infarction. Hazard ratios (HRs) were calculated with Cox regression models, and adjustments were made for established outcome predictors. The additional discriminative value was assessed with the category-less net reclassification improvement and integrated discrimination improvement indexes. In the 5,420 patients, the presence of ST-segment depression (≥1 mm; adjusted HR 1.43, 95% confidence interval [CI] 1.25 to 1.63) and left bundle branch block (adjusted HR 1.64, 95% CI 1.18 to 2.28) were independently associated with long-term cardiovascular death or myocardial infarction. Risk increases were short and long term. On quantitative ECG analysis, cumulative ST-segment depression (≥5 mm; adjusted HR 1.34, 95% CI 1.05 to 1.70), the presence of left bundle branch block (adjusted HR 2.15, 95% CI 1.36 to 3.40) or ≥6 leads with inverse T waves (adjusted HR 1.22, 95% CI 0.97 to 1.55) was independently associated with long-term outcomes. No interaction was observed with treatment strategy. No improvements in net reclassification improvement and integrated discrimination improvement were observed after the addition of quantitative characteristics to a model including qualitative characteristics. In conclusion, in the FRISC II, ICTUS, and RITA-3 NSTE-ACS patient-pooled data set, admission ECG characteristics provided long-term prognostic value for cardiovascular death or myocardial infarction. Quantitative ECG characteristics provided no incremental discrimination compared to qualitative data.
The prognostic value of the admission 12-lead electrocardiogram (ECG), 1 of the most important diagnostic tools in acute coronary syndromes (ACS), has been shown previously. The presence of isolated T-wave inversion or ST-segment deviation was associated with worse short-term outcomes. Moreover, quantitative analysis of the ST segment and T waves predicted mortality or myocardial infarction (MI) up to 1 year. The long-term independent predictive value of the admission ECG and the eventual incremental value of quantitative compared to qualitative evaluation in patients managed with a routine or selective invasive strategy have not yet been investigated. The Fragmin and Fast Revascularization During Instability in Coronary Artery Disease (FRIST II), Invasive Versus Conservative Treatment in Unstable Coronary Syndromes (ICTUS), and Randomized Intervention Trial of Unstable Angina 3 (RITA-3) (FIR) collaboration, a non–ST-segment elevation ACS (NSTE-ACS) patient-pooled database including 5-year follow-up, allowed us to investigate the relation between admission electrocardiographic characteristics and 5-year cardiovascular (CV) death or MI. Moreover, we investigated whether ECG variables contributed with additional predictive information to the established predictors from the FIR risk score.
Methods
The principal investigators of the FRISC II, ICTUS, and RITA-3 trials (L.W., R.J.d.W., and K.A.A.F.) initiated the FIR collaboration, and the main prespecified analyses and a common set of variables were summarized in a protocol. The setting and data collection have been described previously.
For the present analysis, we included all patients with available admission electrocardiograms. The details of the designs of the FRISC II, ICTUS, and RITA-3 trials have been published previously. These trials compared a routine invasive strategy with a conservative or “selective invasive” strategy in patients with NSTE-ACS. Patients who were originally randomized to the routine invasive group were scheduled to undergo early coronary angiography, with subsequent revascularization when appropriate. Coronary artery bypass grafting was recommended with severe left main stem or 3-vessel disease. The selective invasive strategy consisted of initial medical treatment with coronary angiography and revascularization only in the case of refractory angina despite optimal medical treatment. In the FRISC II and ICTUS trials, a predischarge ischemia detection test was routinely performed.
Qualitative data regarding the presence of ST-segment depression (≥0.1 mV cumulative), transient ST-segment elevation (≥0.1 mV cumulative), ST-segment deviation (≥0.1 mV cumulative), and left bundle branch block (LBBB) were available in the FRISC II, ICTUS, and RITA-3 trials.
Regarding quantitative analysis of the baseline electrocardiogram, ST-segment and T-wave changes were assessed in ECG core laboratories in FRISC II and ICTUS. In FRISC II, 12-lead admission ECGs were analyzed at the ECG core laboratory in Copenhagen University Hospital (Copenhagen, Denmark), without information on clinical outcomes. ST-segment deviation was measured 60 ms after the J point in intervals of 0.05 mV. In FRISC II, 719 patients were excluded because of missing admission ECGs, ECGs of very poor technical quality, or the presence of confounding factors excluding ST-segment analysis other then LBBB on electrocardiography. In ICTUS, a standard 12-lead ECG was obtained on admission and evaluated at the University of Amsterdam Academic Medical Center core laboratory (Amsterdam, The Netherlands), without information on clinical outcomes. The presence and magnitude of ST-segment deviation were measured 80 ms after the J point in intervals of 0.05 mV in each lead. In ICTUS, 37 patients were excluded because of missing admission ECGs or ECGs of very poor technical quality. For the present analysis, cumulative ST-segment depression, elevation, and deviation was defined as the summed deviation over all leads, excluding lead aVR. A negative T wave was defined as an amplitude of ≥0.05 mV.
The main outcomes for the present analysis were the composite of CV death or MI and its individual components. Death was considered CV unless an unequivocal non-CV cause could be established. The original definition of MI from FRISC II and RITA-3 and the MI definition from the 5-year follow-up ICTUS report were used.
Normally distributed continuous variables are presented as mean ± SD and variables with skewed distributions as medians with interquartile ranges. Categorical variables are presented as frequencies with percentages. Cumulative event rates according to ECG characteristics were estimated using the Kaplan-Meier method and compared using a log-rank test. Follow-up was censored at the actual date of last contact or at 5 years, whichever came first. The prognostic value of ECG characteristics was assessed by investigating the relation between CV mortality and MI in 3 sets of Cox proportional-hazards analyses: univariate analyses, multivariate analyses including independent ECG characteristics only, and multivariable analysis including independent ECG predictors and established predictors for CV death or MI. The established risk factors were derived from the FIR risk score and included age, previous MI, diabetes, hypertension, and body mass index. The ECG predictors were identified by stepwise backward selection of all ECG variables in a Cox proportional-hazards model. To be less stringent, variables with p values >0.10 indicated by the likelihood ratio test were excluded from the model. For the multivariate models, we assumed that ST-segment and T-wave abnormalities were absent in the 14 patients with LBBB in ICTUS. Interaction between treatment strategy and ECG predictors was assessed by including interaction terms in the multivariate-adjusted Cox models and presenting hazard ratios (HRs) according to treatment strategy. Predefined analyses included (1) the prognostic value of the qualitative ECG characteristics regarding short-term and long-term outcomes in Cox proportional-hazards models with a landmark set at 1-year, (2) the outcomes of a routine invasive strategy compared to a selective invasive strategy according to qualitative ECG characteristics, and (3) the incremental value of quantitative versus qualitative ECG characteristics to established risk factors and treatment strategy. For this analysis, we constructed receiver-operating characteristic curves and compared C statistics. Moreover, the additional discriminative value of quantitative ECG characteristics was assessed using the category-less net reclassification improvement and integrated discrimination improvement indexes. The proportional-hazards assumption for all analyses was verified graphically by checking the parallelism of log-log survival curves and by testing the significance of the interaction with time. No relevant violations were observed.
Results
After excluding 47 patients because of missing qualitative ECG data, 5,420 patients with NSTE-ACS were included in the present analysis. The baseline characteristics of the study patients are listed in Table 1 .
| Characteristic | Patients for Qualitative Analysis | Patients for Quantitative Analysis |
|---|---|---|
| (n = 5,420) | (n = 2,901) | |
| Age (years) | 64 (56–72) | 63 (56–72) |
| Body mass index (kg/m 2 ) | 27.2 ± 4.1 | 27.1 ± 3.8 |
| Men | 3,677 (67.8%) | 2011 (69.3%) |
| MI | 1314 (24.2%) | 649 (22.4%) |
| Percutaneous coronary intervention | 307 (5.7%) | 193 (6.7%) |
| Coronary artery bypass grafting | 108 (2.0%) | 104 (3.6%) |
| Current cigarette smoking | 1803 (33.2%) | 1,022 (35.2%) |
| Hypertension (history requiring medication) | 1820 (33.6%) | 952 (32.8%) |
| Hypercholesterolemia (history requiring medication) | 1242 (22.9%) | 594 (20.5%) |
| Diabetes mellitus | 701 (12.9%) | 369 (12.7%) |
| ST-segment depression (≥0.1 mV) | 2,325 (42.9%) | — |
| ST-segment elevation (transient) (≥0.1 mV) | 478 (8.8%) | — |
| ST-segment deviation (≥0.1 mV) | 2,556 (47.2%) | — |
| LBBB | 121 (2.2%) | — |
| FIR risk score ⁎ | 4 (2–7) | 4 (2–6) |
| FRISC II | 2,447 (45.1%) | 1738 (59.9%) |
| ICTUS | 1,163 (21.5%) | 1,163 (40.1%) |
| RITA-3 | 1,810 (33.4%) | Excluded |
⁎ Age: <60 years = 0, 60 to 64 years = 1, 65 to 69 years = 2, 70 to 74 years = 3, and ≥75 years = 5; diabetes = 4; hypertension = 1; ST-segment depression = 2; body mass index: <25 kg/m 2 = 1, 25 to <35 kg/m 2 = 0; and ≥35 kg/m 2 = 2.
Regarding qualitative ECG characteristics, Figure 1 presents Kaplan-Meier estimates of long-term CV death or MI and the individual components according to the presence of different ECG characteristics. Overall, in the 5,420 patients, ST-segment depression (HR 1.61, 95% confidence interval [CI] 1.41 to 1.83, p < 0.001) and LBBB (HR 2.06, 95% CI 1.46 to 2.92, p < 0.001) were independently associated with increased long-term CV death or MI after backward selection. These HRs were, respectively, 1.43 (95% CI 1.25 to 1.63, p < 0.001) and 1.64 (95% CI 1.18 to 2.28, p = 0.003) after adjustment for established predictors. The adjusted HRs for LBBB and ST-segment depression were, respectively, 1.66 (95% CI 1.07 to 2.57) and 1.46 (95% CI 1.21 to 1.75) in the selective invasive group and 1.59 (95% CI 0.96 to 2.62) and 1.39 (95% CI 1.14 to 1.70) in the routine invasive group. We separately assessed the prognostic value of isolated transient ST-segment elevation that was not associated with increased CV death or MI rates in univariate Cox regression (HR 0.85, 95% CI 0.60 to 1.21, p = 0.38).
Regarding the composite outcome, no interaction was observed between ST-segment depression and treatment strategy (p = 0.53) or LBBB and treatment strategy (p = 0.74). This indicated that these variables could not identify patients who benefited from a routine invasive strategy compared to a selective invasive strategy. In univariate Cox regression taking the 1-year landmark into account, generally, the presence of all ECG abnormalities was associated with increased rates of the composite outcome and its individual components. The weakest relation was observed with transient ST-segment elevation. After backward selection in multivariate Cox regression, the presence of ST-segment depression or LBBB was an independent predictor for 1-year and long-term (1- to 5-year follow-up) outcomes. Although ST-segment depression was mainly predictive of short-term outcomes, the presence of LBBB was mainly predictive of long-term outcomes. Adjustment for established predictors of CV death or MI did not materially change the HR coefficients. However, LBBB lost its significance regarding short-term outcomes.
Admission electrocardiograms for quantitative analysis were available in 2,901 patients. Baseline quantitative ECG characteristics, including the ST segment and T wave, are listed in Table 2 . Several ECG characteristics were associated with increased 5-year CV death or MI hazard in univariate Cox regression. In backward selection of these univariate ECG characteristics, all remained independently associated with long-term CV death or MI. After backward selection in multivariate Cox regression including age, body mass index, hypertension, diabetes mellitus, and history of MI in addition to the ECG characteristics, only cumulative ST-segment depression of 5 mm, ≥6 leads with inverse T waves, and LBBB were independent quantitative ECG predictors of 5-year CV death or MI. LBBB was the most powerful predictor (p < 0.001). The multivariate Cox regression models are listed in Table 3 .
| Characteristic | Frequency (%) | HR (95% CI) | p Value |
|---|---|---|---|
| General | |||
| Rhythm | 0.45 | ||
| Sinus rhythm | 2,819/2,897 (97.3%) | Reference | |
| Atrial fibrillation/flutter | 55/2,897 (1.9%) | 1.36 (0.78–2.37) | |
| Other | 23/2,897 (0.8%) | 1.33 (0.55–3.21) | |
| QRS duration ≥0.12 seconds | 97/2,898 (3.3%) | 2.22 (1.56–3.16) | <0.001 |
| LBBB | 56/2,901 (1.9%) | 2.56 (1.65–3.96) | <0.001 |
| Pathologic Q wave | 782/2,853 (27.4%) | 1.31 (1.08–1.58) | 0.006 |
| ST segment | |||
| Cumulative ST-segment depression (mm) | <0.001 | ||
| <1.0 | 1,437/2,844 (50.5%) | Reference | |
| 1–2.5 | 574/2,844 (20.2%) | 1.32 (1.04–1.67) | |
| 3–4.5 | 349/2,844 (12.3%) | 1.36 (1.03–1.80) | |
| ≥5 | 484/2,844 (17.0%) | 1.65 (1.31–2.09) | |
| Cumulative transient ST-segment elevation (mm) | 0.10 | ||
| <1.0 | 1,433/2,844 (50.4%) | Reference | |
| 1–2.5 | 714/2,844 (25.1%) | 1.15 (0.93–1.42) | |
| 3–4.5 | 410/2,844 (14.4%) | 0.87 (0.65–1.15) | |
| ≥5 | 287/2,844 (10.1%) | 0.77 (0.55–1.09) | |
| Cumulative ST-segment deviation (mm) | 0.06 | ||
| <1.0 | 715/2,844 (25.1%) | Reference | |
| 1–2.5 | 589/2,844 (20.7%) | 1.17 (0.89–1.54) | |
| 3–4.5 | 564/2,844 (19.8%) | 1.08 (0.81–1.43) | |
| ≥5 | 976/2,844 (34.3%) | 1.37 (1.37–1.74) | |
| T wave | |||
| Number of leads with inverse T waves ≥6 | 439/2,842 (15.4%) | 1.38 (1.10–1.73) | 0.006 |
| Cumulative inverse T-wave amplitude (mm) | 0.26 | ||
| <1.0 | 1,255/2,841 (44.2%) | Reference | |
| 1–2.5 | 470/2,841 (16.5%) | 1.13 (0.87–1.45) | |
| 3–4.5 | 293/2,841 (10.3%) | 0.97 (0.70–1.34) | |
| ≥5 | 823/2,841 (29.0%) | 1.22 (0.99–1.51) |
| Characteristic | HR (95% CI) | p Value |
|---|---|---|
| Backward selection of ECG characteristics | ||
| QRS duration ≥0.12 seconds | 1.92 (1.13–3.26) | 0.02 |
| Pathologic Q wave | 1.29 (1.07–1.57) | 0.009 |
| Cumulative ST-segment depression (mm) | 0.001 | |
| <1.0 | Reference | |
| 1–2.5 | 1.31 (1.03–1.66) | |
| 3–4.5 | 1.33 (1.00–1.77) | |
| ≥5 | 1.59 (1.25–2.02) | |
| Number of leads with inverse T waves ≥6 | 1.27 (1.00–1.60) | 0.05 |
| LBBB | 1.91 (0.99–3.69) | 0.05 |
| Backward selection of ECG characteristics and outcome predictors | ||
| Cumulative ST-segment depression (mm) | 0.09 | |
| <1.0 | Reference | |
| 1–2.5 | 1.25 (0.99–1.58) | |
| 3–4.5 | 1.16 (0.87–1.54) | |
| ≥5 | 1.34 (1.05–1.70) | |
| Number of leads with inverse T waves ≥6 | 1.22 (0.97–1.55) | 0.09 |
| LBBB | 2.15 (1.36–3.40) | 0.001 |
| Age (years) | <0.001 | |
| <60 | Reference | |
| 60–64 | 1.14 (0.85–1.54) | |
| 65–69 | 1.37 (1.04–1.79) | |
| 70–74 | 1.49 (1.12–2.00) | |
| ≥75 | 2.15 (1.65–2.79) | |
| Body mass index (kg/m 2 ) | 0.008 | |
| <25 | 1.33 (0.93–1.38) | |
| 25–34 | Reference | |
| ≥35 | 1.79 (1.22–2.63) | |
| Hypertension | 1.20 (0.99–1.44) | 0.06 |
| Diabetes mellitus | 1.90 (1.53–2.36) | <0.001 |
| Previous MI | 2.00 (1.66–2.41) | <0.001 |
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
