European Heart Journal Advance Access originally published online on January 11, 2006
European Heart Journal 2006 27(4):419-426; doi:10.1093/eurheartj/ehi700
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Forecasting mortality: dynamic assessment of risk in ST-segment elevation acute myocardial infarction
1Department of Medicine, University of Alberta, 2-51 Medical Sciences Building, Edmonton, Alberta, Canada T6G 2H7
2Duke Clinical Research Institute, Durham, NC, USA
3Department of Cardiology, University Hospital of Uppsala, Uppsala, Sweden
4Center for Thrombosis and Vascular Research, Katholieke Universiteit, Leuven, Belgium
Received 25 July 2005; revised 31 October 2005; accepted 1 December 2005; online publish-ahead-of-print 11 January 2006.
* Corresponding author. Tel: +1 780 492 0951; fax: +1 780 492 9486. E-mail address: paul.armstrong{at}ualberta.ca
This paper was guest edited by Peter L. Thompson, University of Western Australia, Sir Charles Gairdner Hospital, Perth - Nedlands, Australia
 |
Abstract |
|---|
 Top Abstract IntroductionMethodsResultsDiscussionSupplementary materialAcknowledgementsReferences |
|---|
Aims To demonstrate the feasibility and clinical utility of developing dynamic risk assessment models for ST-segment elevation myocardial infarction (STEMI) patients.
Methods and results In 6066 STEMI patients enrolled in the Assessment of the Safety and Efficacy of a New Thrombolytic-3 (ASSENT-3) trial with complete electrocardiographic data, we assessed the probability of 30-day mortality over the following forecasting periods beginning at day 0 (baseline), 3 h, day 2, and day 5 using multiple-logistic regression. These models were validated and simplified in independent samples of 1622 similar fibrinolytic-treated patients from the ASSENT-3 PLUS trial and in 814 STEMI patients undergoing primary percutaneous coronary intervention in the COMplement inhibition in Myocardial infarction treated with Angioplasty (COMMA) trial. The discriminatory power of these predictive models, from baseline to day 5, was excellent (c-statistics 0.80 to 0.87); and their predictive ability was supported by strong gradients in mortality outcomes as the risk score increased. Dynamic modelling also provided information on the change in prognosis over time which may be used to advise more appropriate therapeutic decisions, e.g. the identification of high-risk patients for possible co-interventions.
Conclusion Dynamic modelling for STEMI patients enhances the risk assessment and stratification and should provide valuable ongoing guidance for their management.
Key Words: Prognosis • Dynamic modelling • Simplified risk scores • ST-segment elevation myocardial infarction
|
Introduction |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionSupplementary materialAcknowledgementsReferences |
|---|
By virtue of its nature and the demands of the health care system, medical decision-making in acute coronary syndromes (ACS) is a dynamic process. Patients are continually assessed from the time of entry into the health care system until discharge and throughout the follow-up care. Critical decisions, based in part on the expected outcomes, must be promptly made on admission and over the next several hours and days, as well as at the time of hospital discharge. To help guide patient management and reflect the rapid transition of patient status during hospital stay, we previously introduced the concept of dynamic risk modelling in non-ST-segment elevation ACS patients enrolled in the Global Utilization of Strategies to Open Occluded Arteries-IIb (GUSTO-IIb) trial.1 Because the short-term morbidity and mortality for ST-segment elevation myocardial infarction (STEMI) patients exceed that of patients without ST-segment elevation, we now extend this concept to provide relevant and timely prognostic information at key decision points for this important population.
As prognostic indices derived from complex models are rarely used in clinical practice,2 our aim is to further develop simplified and valid risk scores for this dynamic process that can be used at the bedside for risk assessment and clinical management. More specifically, we undertook to: (1) develop a series of prognostic models (dynamic risk models) incorporating clinically relevant information unfolding during the hospital stay; (2) validate these models in independent STEMI patient populations; and (3) generate simplified risk scores and explore their clinical utility.
|
Methods |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionSupplementary materialAcknowledgementsReferences |
|---|
Data sources
Data from the Assessment of the Safety and Efficacy of a New Thrombolytic-3 (ASSENT-3) trial were used to develop a series of dynamic models, which were validated in two independent samples: one with fibrinolytic therapy and the other with primary percutaneous coronary intervention (PCI). The details of these trials have been published previously.3–5 Briefly, in the ASSENT-3 trial, 6095 STEMI patients presenting
6 h of symptom onset were randomly assigned to one of the three treatment groups: full-dose tenecteplase and enoxaparin for a maximum of 7 days, full-dose tenecteplase with weight-adjusted unfractionated heparin for 48 h, or half-dose tenecteplase with weight-adjusted low-dose unfractionated heparin and a 12-h infusion of abciximab. Our study sample consisted of 6066 patients after excluding 29 patients with missing baseline electrocardiographic data.
Our first validation sample consisted of 1622 out of 1639 patients from the ASSENT-3 PLUS trial (after excluding 13 patients who died prior to hospital admission and four patients with missing 30-day mortality data), which enrolled STEMI patients presenting 6 h of symptom onset in the pre-hospital setting. Patients were randomly assigned to treatment with tenecteplase and either with enoxaparin or with weight-adjusted unfractionated heparin for 48 h as in the two arms of the ASSENT-3 trial common to ASSENT-3 PLUS.
To broaden the applicability of this approach to patients with primary PCI, we used a second validation sample consisted of 814 STEMI patients enrolled in the COMplement inhibition in Myocardial infarction treated with Angioplasty (COMMA) trial between January 2000 and April 2002, who arrived <6 h of symptom onset with ST-segment elevation 
2 mm in two contiguous leads or new left-bundle branch block (LBBB). The patients, who were to be treated with primary PCI, were randomly assigned to receive placebo bolus and placebo infusion, 2.0 mg/kg bolus of pexelizumab and placebo infusion for 20 h administered 4 h after the bolus, or 2.0 mg/kg bolus and 0.05 mg/kg/h of infusion of pexelizumab for 20 h.
Categorization of data
We categorized the continuous variables according to the conventions (e.g. age <65, 65–74, 75+; and ST-segment resolution of <30, 30–70, 70+) or the quartiles of these variables (heart rate, systolic BP, total ST-segment deviation, and QRS scores) as explained subsequently. Given the prognostic importance of the baseline electrocardiogram (ECG) and subsequent ST-segment resolution data in ACS patients,1,6,7 we incorporated serial ECG data collected after admission to assess the effect of the response to therapy. ECGs were collected in all patients at baseline and at 1 and 3 h after treatment in ASSENT-3 and ASSENT-3 PLUS6 and at baseline, 30 min post-PCI, and 24 h after enrolment in COMMA.5 ECG data were incorporated using three measures: total baseline ST-segment deviation categorized into <12, 12–17, >17 mm; baseline Selvester QRS score8 categorized into 0–1, 2–4, >4; and ST-segment resolution at 1 and 3 h categorized according to Schroeder's method9 as complete resolution (70%), partial resolution (30–70%), and no ST-segment resolution (<30%). In all trials, however, there were missing or incomplete ECG data/confounders (i.e. without the protocol-defined amount of ST-segment elevation, had LBBB, paced rhythm, ventricular rhythm, or poor quality ECGs) that did not allow for an accurate assessment of ST-segment resolution and QRS scores. Also excluded from the baseline QRS-score assessment were right-bundle branch block (RBBB), Wolff Parkinson White pattern (WPW), left anterior fascicular block (LAHB), left posterior fascicular block (LPHB), left ventricular hypertrophy (LVH), and right ventricular hypertrophy (RVH). The proportion of technically suitable ECGs varied between the time of collection and among trials, as shown in Table 1.
|
Model development
Our primary endpoint was 30-day all-cause mortality. Four models were developed sequentially in our study (Figure 1), incorporating the information collected at baseline (day 0 model), at 3 h (3 h model), at day 2 (day 2 model), and at day 5 of the hospital stay (day 5 model). These time points were selected to coincide with conventional clinical practice: at baseline to devise an initial assessment/management plan; at 3 h to incorporate the ST-resolution status relating to the success of reperfusion therapy; at the end of day 1 to reassess the early intervention strategies; and at day 4 to further adjust patient management strategies including possible early discharge. The following variables were included in the 30-day mortality models. Day 0 model: Baseline patient data such as demographics (e.g. age, sex, and race), medical histories (e.g. prior MI, hypertension), and presenting characteristics (e.g. Killip class, systolic blood pressure, total ST-deviation, and QRS score). 3 h model: The baseline data plus ST-segment resolution status at 1 and 3 h. Day 2 model: The baseline and 3 h data plus data on revascularization procedures and adverse events that occurred during day 0–1. Day 5 model: The baseline and 3 h data plus data on revascularization procedures and adverse events that occurred prior to day 5.
|
Multiple-logistic regression procedures for patients who survived to the start of the forecasting period, based on the stepwise, backward variable selection method, were used to develop these models. We assessed the relative contribution of a prognostic factor in the logistic regression model as the proportion of the
2 value associated with that factor out of the sum of all 2 values associated with all significant factors. In developing these models, we monitored the use of revascularization [i.e. PCI and coronary artery bypass grafting (CABG)] procedures and the occurrence of serious in-hospital adverse events, defining a broad clinical category of heart failure as a composite of cardiogenic shock, pulmonary edema, or right heart failure and an electrical disorder as a composite of asystole, electro-mechanical dissociation, or ventricular fibrillation after day 1. Missing data were treated either as separate categories or imputed as non-events when they were shown to be associated with the lower 30-day mortality.
Model validation
We evaluated our models based on the discriminatory capacity (i.e. c-statistic) and the model calibration (i.e. concordance between the predicted and the observed outcomes).10,11 Bootstrapping was performed to estimate the degree of overoptimism associated with c-statistics.12 The validation was performed internally on ASSENT-3 and externally on ASSENT-3 PLUS and COMMA, as noted earlier. In validating these models, we checked the correlation matrices, performed backward stepwise regression, and examined the standard errors of coefficients and interactions among these variables to assess any collinearity and confounding factors.
Simplified risk scores and clinical applications
To illustrate the potential of dynamic risk modelling for bedside use, we further developed and validated simplified risk scores from the day 0, 3 h, day 2, and day 5 models. This was done by assigning points to the coefficients of the most significant (i.e. 210 in Supplementary material online, Table S1) predictors in the models in a 0.5 increment as follows: one point for the coefficient ß=0.5±0.25, two points for ß=1.0±0.25, three points for ß=1.5±0.25, and so on.13 We then calculated the total risk score for each patient at each period as the sum of the points assigned to the patient's applicable risk factors in the model. These simplified risk scores were validated as for the full models, in terms of both model discrimination and calibration.
To compare our approach with a conventional risk assessment tool, we also calculated for each patient a simple Thrombolysis In Myocardial Infarction (TIMI) risk index as follows: (heart ratex[age/10]2)/(systolic blood pressure).14 This composite index was then used in lieu of the three separate factors of age, heart rate, and systolic blood pressure in our four full models in ASSENT-3 to evaluate the relative contributions of this index to mortality prediction over time. All analyses were performed using SPSS version 13.0 (Chicago, IL, USA) except for bootstrapping, which was performed using STATA version 7 (College Station, TX, USA).
|
Results |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionSupplementary materialAcknowledgementsReferences |
|---|
Our ASSENT-3 study sample consisted of 6066 patients whose median age was 61, 24% were women, and 85% were Caucasians. There were some differences in baseline patient characteristics between ASSENT-3 and the validation datasets: patients in ASSENT-3 PLUS were slightly older with fewer Caucasians in the sample, and more patients in COMMA had hypertension and anterior MI as well as higher heart rates (Table 1). Serious in-hospital adverse events among patients in ASSENT-3 were as follows: 3.0% for recurrent ischaemia and also for re-infarction, 3.6% for electrical disorders, 5.5% for heart failure, 1.5% for stroke, 3.1% for major bleeding, and 5.1% for death. Most of these complications occurred within the first few days. In-hospital PCI and CABG procedures were performed on 28.6 and 4.5% of the patients, respectively, with a median of 3 and 6 days to these procedures (Table 2).
|
Model development
A series of predictive models forecasting 30-day mortality from baseline, 3 h, day 2, and day 5 were developed on the basis of patient characteristics, complications, and procedures acquired up until the beginning of each forecasting period (Tables 1 and 2) and are hereafter referred to as the ‘full ASSENT-3 models’ (Supplementary material online). As shown in Figure 2A, the three most influential factors in the baseline model were age, systolic blood pressure, and heart rate, which accounted for 70% of the relative contribution to the mortality prediction. At 3 h, the relative contribution of these factors declined to 63%, and at both day 2 and day 5, it further dropped to 26%. The ECG measures, including the total ST-segment deviation and QRS score at baseline and the ST-segment resolution status at 3 h, accounted for 13% of the relative contribution at baseline, 23% at 3 h, but attenuated to 11 and 15% at day 2 and day 5, respectively. The in-hospital events of heart failure and stroke were particularly influential and accounted for 45 and 32% of the relative contribution at day 2 and day 5, respectively. Figure 2B depicts the trends in prognosis over time according to low (
1%), medium (1–5%), and high risk (>5%) of 30-day mortality. Interestingly, the proportion of low-risk patients increased steadily from 18.7% at baseline to 52.2% at day 5, whereas that of high-risk patients declined steadily from 32.1 to 8.4% between baseline and day 5, with an additional 3.7% expiring during that period.
|
Model validation
As detailed in Supplementary material online, the c-statistics associated with the four full ASSENT-3 models ranged from 0.82 to 0.87, and increased from 0.76 at baseline to 0.84 at day 5 when restricted only to those patients who survived the first four days of hospitalization. The amount of overoptimism was minimal, ranging from 0.004 to 0.005 for the baseline to day 2 models and 0.010 for the day 5 model. None of the Hosmer–Lemeshow statistics for these models was significant, and the Pearson correlation coefficients ranged from 0.997 to 0.999 between the predicted and the observed 30-day mortality probabilities for all models based on the deciles of these probabilities.
The full ASSENT-3 models were then externally validated in ASSENT-3 PLUS and COMMA (Supplementary material online). The c-statistics ranged from 0.78 to 0.82 when the full ASSENT-3 models were applied to the ASSENT-3 PLUS data, and none of the Hosmer–Lemeshow statistics for these four models was significant. The Pearson correlation coefficients ranged from 0.87 to 0.93 between the predicted and the observed 30-day mortality probabilities for all models based on the deciles of these probabilities. When applied to the COMMA data, the c-statistics ranged from 0.83 to 0.86; and the Pearson correlation coefficients ranged from 0.93 to 0.97 between the predicted and the observed 30-day mortality probabilities, with non-significant Hosmer–Lemeshow statistics for all four models; thus showing the robustness of our ASSENT-3 models. Overoptimism was negligible (<0.001) for all these (full and their simplified) models, because only one predictor variable derived from the ASSENT-3 models was involved in each of these validation models.
Simplified risk scores and clinical applications
The simplified risk scores from the full ASSENT-3 models are detailed in Table 3, and the strong gradients of mortality rates associated with them are depicted in Figure 3. Their discriminatory capacity was also excellent, with the c-statistics ranging from 0.80 to 0.86 (Table 3). A more detailed account of the validation of the simplified models together with their applications to individual patients is provided in Supplementary material online.
|
|
Dynamic modelling can help identify high-risk patients who should be treated more aggressively. For instance, while the simplified risk score remained unchanged between baseline and 3 h for patients with either complete or partial ST-segment resolution, 97.5% of patients without ST-segment resolution or with an ECG confounder had a higher risk score at 3 h. The rate of in-hospital revascularization in ASSENT-3 was 33.2 and 30.8%, respectively, for those with no ST-segment resolution or an ECG confounder. This was virtually identical to the observed 31.3 and 34.2% for those with partial or complete ST-segment resolution, respectively. The 30-day mortality for those with and without in-hospital revascularization was 5.2 and 10.0% (P=0.041) among patients without ST-segment resolution and 4.6 and 12.8% (P<0.001) among those with an ECG confounder. The previous results were unchanged after adjusting for the propensity for in-hospital revascularization. Hence, it is possible that a proportion of the 177 patients who died without complete or partial ST-segment resolution by 3 h but were not revascularized may have been better served by more aggressive treatment.
Our approach is flexible and may also be used to incorporate any traditional prognostic index for predicting clinical outcomes. For instance, when forecasting 30-day mortality in ASSENT-3 based on the simple TIMI risk index in place of individual baseline covariates of age, heart rate, and systolic blood pressure, a similar decline in its influence was also demonstrated (Figure 2C): while this index accounted for 72% of the relative contribution to mortality prediction at baseline, that percentage was reduced to 63% at 3 h, and then to 23 and 26% at day 2 and day 5, respectively. Conversely, other factors, such as the ECG measures and in-hospital events, assumed an increasingly greater role and became predominant in mortality prediction and risk stratification in day 2 and day 5 models.
|
Discussion |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionSupplementary materialAcknowledgementsReferences |
|---|
Dynamic modelling attempts to capture the texture of clinical risk assessment from the time of first medical contact with acute coronary patients and thereafter, especially during the critical early phase of patient management. As the care of these patients requires prompt and timely adjustments, having implications for both risk-benefit and cost-effectiveness, developing a risk assessment process to reflect this evolutionary process is a worthwhile goal1,15 and distinguishes our approach from the traditional baseline and discharge models.16,17
Several key messages have emerged from our study. First, the dynamic modelling framework enables us to risk-stratify patients at any clinically important time point (e.g. at day 5 for early discharge) as well as over several time points. A unique strength of dynamic modelling, therefore, is that it captures change in prognosis over time. An advantage of monitoring changes in prognosis is that it is feasible to devise decision rules for patient management, e.g. identifying candidates for early discharge based on the criterion of patients being stable and remaining at a low risk during the first few days of hospitalization.1 In this paper, we further demonstrated the feasibility of applying simplified risk scores to monitor change in prognosis between baseline and 3 h for identifying opportunities for more aggressive treatment. As factors not available at baseline, such as the ECG tracking of ST-segment evolution and in-hospital events, convey significant prognostic information in addition to baseline values, the performance of these models improved over the traditional baseline model as a result of their inclusion. To illustrate this point, we have used vital signs (i.e. systolic blood pressure, heart rate) collected at both baseline and 24 h in the COMMA trial to recalculate the simple TIMI risk index14 for each patient at these time points. Logistic regression based on this index alone was performed to predict day 0–30 and day 2–30 mortality, resulting in the c-statistics of 0.85 and 0.92, respectively, when the sample was restricted in both models to those who survived to the start of day 2. A corollary, therefore, is that it is essential to test the optimal methods of collecting significant baseline and in-hospital predictors that can change over time, e.g. heart rate, blood pressure, ECG measures, and biomarkers.1
Secondly, a spectrum of models, from very simple to very sophisticated, should be developed and utilized using the dynamic risk modelling framework. For scientific research at tertiary care institutions and academic centres, sophisticated full models should be further developed, tested, and used to provide most valid and reliable answers to clinical questions arising from various health care settings and systems. On the other hand, it is imperative to also develop simplified risk scores, because prognostic indices derived from complex models are rarely used in clinical practice.2 For rapid, user-friendly bedside use, we developed in this study highly reliable, simplified risk scores (c-statistics 0.80–0.86 after excluding the QRS score) to provide quick risk assessment. The implementation of dynamic risk modelling may be further facilitated through the use of centralized electronic medical records to automatically calculate these probabilities and reduce the burden of data entry to clinicians. Integrating computerized ECG analyses, which could generate standardized ECG measures, into this health information system should also be considered.18
Thirdly, our dynamic models are applicable not only to fibrinolytic-treated STEMI patients but also to those undergoing primary PCI. This is important, because an estimated 25–50% of AMI patients currently undergo primary-PCI worldwide.19 As shown in Supplementary material online, our dynamic models performed extremely well in both fibrinolytic-treated and primary-PCI patients: the c-statistics for the simplified risk scores ranged from 0.79 to 0.86 in ASSENT-3 PLUS and 0.81 to 0.86 in COMMA; and these scores were strongly and positively associated with mortality rates in all models in both validation datasets. In comparison, the c-statistics for other baseline-risk models were 0.84 for GUSTO-I full model,20 0.79 for the (simplified) TIMI risk score,11 and 0.78 for a simple TIMI risk index (based only on age, heart rate, and systolic blood pressure).14
Some limitations of our study should be noted. First, our risk models were developed and validated using specific-clinical trials data, which may differ from the general STEMI patient population. It should be noted, however, that over 14% of them were 75 years or more. Although we have validated our models also on an independent sample of primary-PCI patients, further research is warranted to test the generalizability and reproducibility of our results in other settings. Secondly, unlike the probability measures derived from the full models, simplified risk scores are not consistently defined and standardized, and hence may not be strictly comparable, across the forecasting periods. For example, a score of 7 in our study was associated with an observed 30-day mortality rate that changed from 19.8%, through 14.0, 2.9, and 5.9% as it was ascertained at baseline, 3 h, day 2, and day 5, respectively (Figure 3). For tracking change in prognosis over time using simplified risk scores, it is crucial to always refer to the levels of risk associated with these scores (Figure 3) rather than comparing them directly across the periods. Thus, for comparing change in prognosis over time, a better approach may be to use programmable calculators or handheld computers, which can readily compute mortality probabilities for individuals or groups of patients directly from the full models.1,15
In conclusion, we have extended the dynamic modelling methodology to STEMI patients receiving either fibrinolytic therapy or primary PCI and demonstrated that this strategy of continuous risk assessment and stratification is feasible and may sharpen evidence-based decision-making in the management of STEMI patients within the critical days after hospital admission.
|
Supplementary material |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionSupplementary materialAcknowledgementsReferences |
|---|
Supplementary material is available at European Heart Journal online.
|
Acknowledgements |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionSupplementary materialAcknowledgementsReferences |
|---|
The ASSENT-3 study was supported by Boehringer Ingelheim, Germany; Genentech, South San Francisco, CA, USA, and Aventis, Bridgewater, NJ, USA. The ASSENT-3 PLUS study was supported by Boehringer Ingelheim, Germany and Aventis, Bridgewater, NJ, USA. The COMMA study was funded by Procter & Gamble Pharmaceuticals and Alexion Pharmaceuticals, Inc.
Conflict of interest: none declared.
|
References |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionSupplementary materialAcknowledgementsReferences |
|---|
- Chang W-C, Boersma E, Granger CB, Harrington RA, Califf RM, Simoons ML, Kleinman NS, Armstrong PW, for the GUSTO IIb and PURSUIT Investigators. Dynamic prognostication in non-ST elevation acute coronary syndromes: insights from GUSTO IIb and PURSUIT. Am Heart J 2004;148:62–71.[CrossRef][Medline]
- Redelmeier DA, Lustig AJ. Prognostic indices in clinical practice. JAMA 2001;285:3024–3025.
[Free Full Text] - Assessment of the Safety and Efficacy of a New Thrombolytic Regimen (ASSENT)-3 Investigators. Efficacy and safety of tenecteplase in combination with enoxaparin, abciximab, or unfractionated heparin: the ASSENT-3 randomised trial in acute myocardial infarction. Lancet 2001;358:605–613.[CrossRef][ISI][Medline]
- Wallentin L, Goldstein P, Armstrong PW, Granger CB, Adgey AA, Arntz HR, Bogaerts K, Danays T, Lindahl B, Makijarvi M, Verheugt F, Van de Werf F. Efficacy and safety of tenecteplase in combination with the low-molecular-weight heparin enoxaparin or unfractionated heparin in the prehospital setting: the Assessment of the Safety and Efficacy of a New Thrombolytic Regimen (ASSENT)-3 PLUS randomized trial in acute myocardial infarction. Circulation 2003;108:135–142.[CrossRef][ISI][Medline]
- Granger CB, Mahaffey KW, Weaver WD, Theroux P, Hochmann JS, Filloon TG, Rollins S, Todaro TG, Nicolau JC, Ruzyllo W, Armstrong PW, for the COMMA Investigators. Pexelizumab, an anti-C5 complement antibody, as adjunctive therapy to primary percutaneous coronary intervention in acute myocardial infarction: the COMplement inhibition in Myocardial infarction treated with Angioplasty (COMMA) trial. Circulation 2003;108:1184–1190.[CrossRef][ISI][Medline]
- Armstrong PW, Wagner G, Goodman SG, Van de Werf F, Granger C, Wallentin L, Fu Y, for the ASSENT 3 Investigators. ST segment resolution in ASSENT 3: insights into the role of three different treatment strategies for acute myocardial infarction. Eur Heart J 2003;24:1515–1522.
[Abstract/Free Full Text] - Fu Y, Goodman S, Chang W-C, Van de Werf F, Granger CB, Armstrong PW, for the Assessment of the Safety and Efficacy of a New Thrombolytic (ASSENT-2) Trial Investigators. Time to treatment influences the impact of ST-segment resolution on one-year prognosis. Insight from the ASSENT-2 Trial. Circulation 2001;104:2653–2659.
[Abstract/Free Full Text] - Hindman NB, Schocken DD, Widmann M, Anderson WD, White RD, Leggett S, Ideker RE, Hinohara T, Selvester RH, Wagner GS. Evaluation of a QRS scoring system for estimating myocardial infarct size. V. Specificity and method of application of the complete system. Am J Cardiol 1985;55:1485–1490.[CrossRef][ISI][Medline]
- Schroder R, Wegscheider K, Schroder K, Dissmann R, Meyer-Sabellek W, for the INJECT trial group. Extent of early ST segment elevation resolution: a strong predictor of outcome in patients with acute myocardial infarction and a sensitive measure to compare thrombolytic regimens. J Am Coll Cardiol 1995;26:1657–1664.[Abstract]
- Morrow DA. New insight into clinical risk scores for patients with acute coronary syndromes. Am Heart J 2003;146:754–756.[Medline]
- Morrow DA, Antman EM, Charlesworth A, Cairns R, Murphy SA, de Lemos JA, Giugliano RP, McCabe CH, Braunwald E. TIMI Risk Score for ST-Elevation Myocardial Infarction: A Convenient, Bedside, Clinical Score for Risk Assessment at Presentation: An Intravenous nPA for Treatment of Infarcting Myocardium Early II Trial Substudy. Circulation 2000;102:2031–2037.
[Abstract/Free Full Text] - Steyerberg EW, Harrell FE Jr, Borsboom GJ, Eijkemans MJ, Vergouwe Y, Habbema JD. Internal validation of predictive models: efficiency of some procedures for logistic regression analysis. J Clin Epidemiol 2001;54:774–781.[CrossRef][ISI][Medline]
- Moons KGM, Harrell FE. Letter to the Editor: Should scoring rules be based on odds ratios or regression coefficients? J Clin Epidemiol 2002;55:1054–1055.[CrossRef][ISI][Medline]
- Morrow DA, Antman EM, Giugliano RP, Cairns R, Charlesworth A, Murphy SA, de Lemos JA, McCabe CH, Braunwald E. A simple risk index for rapid initial triage of patients with ST-elevation myocardial infarction: an InTIME II Substudy. Lancet 2001;358:1571–1575.[CrossRef][ISI][Medline]
- Granger CB, Goldberg RJ, Dabbous O, Pieper KS, Eagle KA, Cannon CP, Van de Werf F, Avezum A, Goodman SG, Flather MD, Fox KA, for the Global Registry of Acute Coronary Events (GRACE) Investigators. Predictors of hospital mortality in the global registry of acute coronary events. Arch Intern Med 2003;163:2345–2353.
[Abstract/Free Full Text] - Newby LK, Hasselblad V, Armstrong PW, Van de Werf F, Mark DB, White HD, Topol EJ, Califf RM. Time-based risk assessment after myocardial infarction. Implications for timing of discharge and applications to medical decision-making. Eur Heart J 2003;24:182–189.
[Abstract/Free Full Text] - Walter LC, Brand RJ, Counsell SR, Palmer RM, Landefeld CS, Fortinsky RH, Covinsky KE. Development and validation of a prognostic index for 1-year mortality in older adults after hospitalization. JAMA 2001;285:2987–2994.
[Abstract/Free Full Text] - Pelter MM, Adams MG, Drew BJ. Computer versus manual measurement of ST-segment deviation. J Electrocardiol 1997;30:151–156.[Medline]
- Ui S, Chino M, Isshiki T. Rates of primary percutaneous coronary intervention worldwide. Circ J 2005;69:95–100.[Medline]
- Lee KL, Woodlief LH, Topol EJ, Weaver WD, Betriu A, Col J, Simoons M, Aylward P, Van de Werf F, Califf RM, for the GUSTO-I Investigators. Predictors of 30-day mortality in the era of reperfusion for acute myocardial infarction. Results from an international trial of 41021 patients. Circulation 1995;91:1659–1668.[ISI][Medline]
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  J. L. Anderson, C. D. Adams, E. M. Antman, C. R. Bridges, R. M. Califf, D. E. Casey Jr, W. E. Chavey II, F. M. Fesmire, J. S. Hochman, T. N. Levin, et al. ACC/AHA 2007 Guidelines for the Management of Patients With Unstable Angina/Non-ST-Elevation Myocardial Infarction: A Report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines (Writing Committee to Revise the 2002 Guidelines for the Management of Patients With Unstable Angina/Non-ST-Elevation Myocardial Infarction) Developed in Collaboration with the American College of Emergency Physicians, the Society for Cardiovascular Angiography and Interventions, and the Society of Thoracic Surgeons Endorsed by the American Association of Cardiovascular and Pulmonary Rehabilitation and the Society for Academic Emergency Medicine J. Am. Coll. Cardiol., August 14, 2007; 50(7): e1 - e157. [Full Text] [PDF]
|
|
|
|
|
|
J. L. Anderson, C. D. Adams, E. M. Antman, C. R. Bridges, R. M. Califf, D. E. Casey Jr, W. E. Chavey II, F. M. Fesmire, J. S. Hochman, T. N. Levin, et al. ACC/AHA 2007 Guidelines for the Management of Patients With Unstable Angina/Non ST-Elevation Myocardial Infarction Executive Summary: A Report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines (Writing Committee to Revise the 2002 Guidelines for the Management of Patients With Unstable Angina/Non ST-Elevation Myocardial Infarction) Developed in Collaboration with the American College of Emergency Physicians, the Society for Cardiovascular Angiography and Interventions, and the Society of Thoracic Surgeons Endorsed by the American Association of Cardiovascular and Pulmonary Rehabilitation and the Society for Academic Emergency Medicine J. Am. Coll. Cardiol., August 14, 2007; 50(7): 652 - 726. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||