| | Determinants of coronary events in patients with stable angina: Results from the Impact of Nicorandil in Angina StudyReceived 8 December 2004; accepted 27 March 2005. BackgroundIONA was a randomized trial of nicorandil (20 mg twice daily) versus placebo in 5126 patients with stable angina of effort. We present the characteristics as measured at baseline that were found to be associated with increased risk of major outcomes. MethodsThe primary end point of the study was coronary heart disease death, nonfatal myocardial infarction (MI), or unplanned hospitalization for cardiac chest pain. The secondary outcome excluded the softer outcome of unplanned hospitalization for chest pain. Potential prognostic factors included nicorandil treatment, sex, smoking, severity of angina (Canadian Cardiovascular Society [CCS] grading), previous MI, coronary artery bypass graft/percutaneous transluminal coronary angioplasty (CABG/PTCA), left ventricular dysfunction, left ventricular hypertrophy (LVH) on electrocardiography (ECG), histories of diabetes or hypertension, baseline cardiovascular drugs, age, blood pressures, heart rate, height, weight, and body mass index. ResultsA total of 5047 patients had complete baseline information. There were 724 primary outcomes and 240 secondary outcomes. The strongest determinants of the hard secondary end point were CCS score (hazards ratio [HR] 2.25, 95% confidence intervals [CI] 1.50-3.38) for grades III and IV versus grade I, age (HR 1.23, 95% CI 1.13-1.33) for 5-year risk, and previous MI (HR 2.05, 95% CI 1.48-2.85). The results for the primary outcome were similar, with stronger associations with LVH on ECG and smoking, and a weaker association with age. ConclusionsWe present models of risk factors for coronary outcomes in a population with angina of effort. The strongest risk factors were the CCS score and previous MI. The IONA study was sponsored by Merck Pharmaceuticals, Aventis Pharma, and the Chugai Pharmaceutical Company. Stable angina is a common manifestation of coronary heart disease and is a major cause of disability in Europe and in North America.1, 2 Moreover, the real burden of angina in the population may be underestimated, as it is commonly undiagnosed and has consequently been described as a “submerged clinical iceberg.3” In addition to morbidity, angina consumes a substantial proportion of health-care resources and, as a consequence, represents a substantial cost to health-care systems.4 Although the risk of future major coronary events associated with angina is high, it is also highly variable. The objective of the current study is to determine the baseline factors that were prognostic of outcome, to create explicit multivariate models that can be used for risk prediction in the way that the Framingham models are used for subjects without a history of coronary disease, and to determine the distribution of risk across the study population. Prognostic models may be used to identify high-risk subjects, to help guide the cost-effectiveness of treatment strategies in populations and to help in the selection of high-risk patients for clinical trials. Aspirin, ACE inhibitors, and statins have been shown to reduce the risk of cardiovascular events in subgroups of subjects with coronary disease.5, 6, 7, 8, 9 Recently, the IONA trial became the first study to demonstrate a similar effect with a specific antianginal agent.10 This study provides a database from which prognostic models can be developed. Methods  Study design The design and conduct of the IONA study has been reported in detail elsewhere.11 Patients with a history of angina were recruited from hospital and primary care centers in the United Kingdom. Standard background antianginal therapy (one or more oral antianginal medications including β-blockers, calcium-channel blockers, or long-acting nitrates) was not specified but was the optimal therapy as judged by the investigator for the individual patient. Patients were randomized to double-blind treatment with nicorandil (in a dose of 10 mg twice daily for 2 weeks and 20 mg twice daily thereafter) or placebo. A strategy of recruiting subjects with clearly established coronary heart disease or a positive exercise test with additional risk factors was adopted to ensure that the study would recruit a group of subjects who had an increased risk of experiencing a primary end point than would otherwise be the case. The recruits were men >45 years or women >55 years. They were required to have a history of myocardial infarction (MI) or coronary artery bypass graft (CABG), or a definite diagnosis of coronary heart disease by angiography or a documented positive exercise test. In the category that qualified only by a definite diagnosis of coronary heart disease, there had to be one of the following high-risk features: left ventricular hypertrophy (LVH) on electrocardiography (ECG); left ventricular ejection fraction of ≤45% or an echocardiographic end diastolic dimension of >55 mm; aged 65 years or older; type I or II diabetes; hypertension or documented evidence of other vascular disease (peripheral or central). The most important exclusion factor apart from cardiovascular instability was concomitant treatment with a sulphonylurea because these may inhibit K+ channel opening. Patients were followed up for at least 1 year and up to 3 years, with an average of 1.6 years. All patients provided written informed consent, and the study was approved by a multicenter research ethics committee and local research ethics committees in the United Kingdom. End points The primary end point of the study was the combined outcome of coronary heart disease death, nonfatal MI, or unplanned hospitalization for cardiac chest pain. The secondary end point was the combined outcome of coronary heart disease death or nonfatal MI. Formal definitions of the study end points have been reported previously.11 Prognostic variables The following (categorical) baseline prognostic factors were included in the statistical analysis: randomized treatment (nicorandil or placebo), sex, current smoking status, severity of angina as defined by the Canadian Cardiovascular Society (CCS) score12, 13, 14; previous MI, CABG, and/or percutaneous transluminal coronary angioplasty (PTCA), left ventricular dysfunction, or evidence of LVH on ECG; histories of diabetes or hypertension. Note that the CCS score has 4 grades, which can be summarized as follows: (1) no pain during ordinary activity, (2) slight limitation of ordinary activity, (3) marked limitation of ordinary activity, and (4) inability to do any physical activity. The following continuous factors were also considered: age; blood pressures (systolic, diastolic, and pulse pressures); heart rate; height, weight, and body mass index. In addition to these “traditional” prognostic factors, we also considered the concomitant use of antianginal drugs (long-acting nitrates, calcium antagonists, β-blockers, and the number of these drugs taken). Other cardiovascular drugs that were taken at baseline were also examined, namely, ACE inhibitors, antiplatelets, statins, and diuretics. Statistical methods All analyses were reported on the subset of 5047 of the 5126 randomized subjects who had complete baseline data. Summary statistics were reported as counts and percentages for categorical data and as means and standard deviations for continuous variables. All formal analyses were based on survival analysis methodology. Clinical outcomes were sought on all subjects until death, the end of the study, or withdrawal of informed consent for follow-up (whichever came first). Univariate and multivariate models were fitted using Cox proportional hazards models, with the associated estimation of hazards ratios (HR) and 95% confidence intervals. P values were calculated using the Wald statistic. The continuous variables were presented with the units changed to make the HRs more meaningful (eg, the risk for an increase of 5 years in age is more suitable than that for only 1 year) and also in fifths using quintiles as the cut-points (the lowest fifth was always the referent level). Grades III and IV of the CCS score were combined for the models. Multivariate models were derived using a forward stepwise fitting procedure (using a strict significance level of .05). Previous MI and previous CABG together constitute 2 of the main inclusion criteria of the IONA study. To protect against a major impact of this design feature, the chosen multivariate models were refitted in the subgroups defined by whether the patients had a previous MI and/or a previous CABG. The models were also used to estimate the 18-month risk for each subject in the study, and hence the distribution of risk in the population. All analyses were carried out using SAS version 8.02. Results The average age was 66 years, and 76% were men (Table I). Most of the patients had a prior MI (66%). The distribution of patients in the CCS severity-of-angina grades was 27% (grade I), 62% (grade II), 11% (grade III), and 1% (grade IV). Most of the factors studied were associated with the primary end point in univariate models (724 subjects with events) and the secondary end point (240 subjects with events) (see Table II, Table III). Nondrug variables having the strongest associations with the primary end point were, in decreasing order of statistical significance: CCS angina score (HR 2.46, 95% CI 1.95-3.09) for grades III and IV relative to grade I), previous MI (HR 1.62, 95% CI 1.37-1.91), LVH (HR 1.50, 95% CI 1.21-1.86), and current smoking (HR 1.37, 95% CI 1.14-1.64). In addition, left ventricular dysfunction, age, and heart rate were significant and associated with a higher risk of an event. Treatment with nicorandil, higher weight, and body mass index were significantly associated with a lower risk. The following variables were not significant in univariate analysis: sex, history of CABG/PTCA, diabetes, or hypertension, blood pressures, and height. | ⁎ Primary end point = coronary heart disease death, nonfatal myocardial infarction, or unplanned hospitalization for cardiac chest pain. †Secondary end point = coronary heart disease death or nonfatal myocardial infarction. ‡n = number of classes of antianginal drugs that were used before the study. |
| ⁎ Primary end point = coronary heart disease death, nonfatal myocardial infarction or unplanned hospitalization for cardiac chest pain. †Secondary end point = coronary heart disease death or nonfatal myocardial infarction. |
Most of the drugs taken at baseline were associated with the subsequent risk of a primary end point (with the exception of antiplatelet agents). The strongest drug association with risk was the use of long-acting nitrates (HR 1.66, 95% CI 1.42-1.93). Baseline use of ACE inhibitors, diuretics, and calcium antagonists was associated with an increased risk of an event, whereas baseline use of β-blockers and statins was associated with a reduced risk. The univariate associations with the secondary end point were roughly similar, with perhaps a general trend for associations to become stronger, because of the harder and more specific nature of the end point, but less statistically significant because of the smaller number of events. In particular, the effect of previous MI (perhaps expectedly) rose substantially to an HR of 2.34 (95% CI 1.69-3.24). Previous CABG/PTCA became significant (HR 0.56, 95% CI 0.41-0.76). Antiplatelet agents also became significantly associated with a lower risk (HR 0.71, 95% CI 0.51-0.99), and the use of calcium-channel blockers became nonsignificant (Table II). The effects of age, heart rate, weight, and body mass index all strengthened, and finally, raised pulse pressure became significantly associated with a higher risk (Table III). All of the significant variables were entered into forward stepwise selection procedures, for possible inclusion in multivariate models. This was done separately for the primary and secondary end points. The additional prognostic value of the baseline medications was then considered (also using a forward stepwise approach), using the previous models as a starting point. The resulting models are shown in Table IV for the primary end point and Table V for the secondary end point. The following variables were chosen in the model for the primary end point: CCS angina score, nicorandil treatment, current smoking, previous MI, LVH, and age. As with the univariate results, the strongest determinants of the primary end point were CCS scores (adjusted HR for grades III and IV vs grade I = 2.30, 95% CI 1.82-2.90) and previous MI (adjusted HR 1.55, 95% CI 1.31-1.84). When the baseline medications were considered, long-acting nitrates and diuretics were also entered into the model (and appeared to be associated with a higher risk). When these 2 classes of drugs were added, the HRs of the other factors were slightly attenuated. The equivalent models for the secondary end point were slightly different. As in the primary analysis of the trial,10 treatment with nicorandil did not achieve statistical significance. Previous CABG/PTCA and heart rate were included in the model. In addition, long-acting nitrates were not significant, but ACE inhibitors and statins did get included in this analysis (Table V). A total of 3799 patients (75%) had a previous MI or undergone a previous CABG. Of the patients who had either of these prior events, 15.6% had a subsequent primary end point and 5.4% had a secondary end point. The equivalent figures for those without these prior events are 10.4% and 2.72%, respectively. The models in these subgroups (not shown) were similar for patients with and without prior MIs or CABGs (with some results becoming nonsignificant in the smaller group of patients without these prior events). The mean 18-month risk for the primary end point was estimated as 13.7% (SD 5.3%). Only 5% of the population had a risk of an event by 18 months, which was >1 in 4. Alternatively, 25% of the population had a chance of an event by 18 months that was estimated at ≤10% (inclusive), 70% had an 18-month risk of up to 15%, 89% had a risk of up to 20%, and 95% of the population had a risk of up to 25%. The average 18-month risk was 12.6% for patients with nicorandil and 14.8% for patients receiving placebo. The distributions of 18-month risk for the primary and secondary end points are illustrated in Figure 1. Similarly, the mean 18-month risk for the secondary end point was 4.6% (SD 3.3%), with 18% of the population having a risk of up to 2% (inclusive), 54% having a risk of up to 4%, 75% having a risk up to 6%, and 87% of the population having a risk of up to 8%. When treatment with the study drug was forced into the model (on top of the chosen factors), the average risk was 4.1% for patients with nicorandil and 5.1% for patients receiving placebo. Discussion IONA was the first study of a specific antianginal agent to demonstrate a statistically significant benefit in reducing major coronary events. These results were achieved in addition to standard treatments for coronary heart disease and, in particular, on top of what the investigator considered to be optimal antianginal therapy. In this study, we used the study cohort as a resource to identify which factors were useful for identifying high-risk patients, in a population with angina of effort. A key finding is that the CCS angina score was a significant independent factor for both end points. Other variables that were significant for both end points in the multivariate models were previous MI, LVH on the ECG, smoking, age, and current smoking. The CCS angina score is the standard measurement of the severity of the symptoms of angina of effort, and its origins have been described by Cox and Naylor.13 There have been some criticisms of the CCS score for being vague and not being suitable for grading unstable angina.13 Also, Cox and Naylor noted that the CCS score has previously been found to be only weakly predictive of survival (compared with other variables), although “patients have tended not to differ by more than one grade within any given trial.”13 The presence of angina is itself predictive of coronary events,15, 16, 17, 18 but demonstrating the predictive power of severity grades within an angina population has been more problematic. Other types of severity scale have been successfully used with unstable angina patients.19, 20 However, one small study of unstable angina patients did find that the most important predictors of outcome were the CCS score and left ventricular ejection fraction.21 In a study that included patients both with and without angina, the CCS score was also found to be strongly (and independently) associated with the occurrence of stroke.22 The results were slightly different for the secondary end point, consisting of the hard end points of coronary heart disease death and nonfatal MI. The CCS scores did not dominate the model as much, with age and previous MI being strongly associated with this outcome. This is perhaps not surprising, because severity of angina is more likely to be associated with hospitalization for chest pain, and prior MI is strongly associated with recurrent MI. However, there was still a very strong association between the CCS score and the secondary end point, with HRs of 1.31 and 2.25 in the multivariate model for grade II and grade III and IV combined, respectively (Table V). It is interesting that age is more associated with the secondary outcome (univariate HR associated with being 5 years older = 1.26, 95% CI 1.17-1.37) than the primary end point (HR 1.06, 95% CI 1.01-1.10). We also examined the additional effects of various baseline drugs, with long-acting nitrates and diuretics being independently associated with the primary end point, and ACE inhibitors, statins, and diuretics being independently associated with the secondary outcome. These drugs appeared to be associated with a higher risk of an outcome, with the exception of statins (multivariate HR 0.71, 95% CI 0.55-0.94). It should be noted that drugs may appear to be either protective or to imply increased risk, depending upon prescribing policies and/or which patients they are “channeled” toward by individual physicians.23, 24 For example, a prescribing pattern could imply worse disease, coexistent heart failure, or coexistent hypertension. One previous article has found that long-term nitrate use was associated with increased mortality, and that the effect withstood statistical devices for control of confounding such as propensity scores and sensitivity analyses.25 Despite these attempts to control confounding, it is still possible that the use of nitrates merely serves as a surrogate for unknown factors that are associated with a poor prognosis.24, 26 The use of ACE inhibitors or diuretics may also be markers for congestive heart failure. Both β-blockers and antiplatelet agents were associated with a low risk in the univariate analyses, but these drugs did not contribute to the multivariate models. We were able to create parsimonious multivariate models that included a range of 6 to 9 variables of a list of 26 possible prognostic factors. Although individual factors in the models had high statistical significance in their relation to outcome, this does not mean that the models are likely to be effective in accurately predicting the time until an event will occur in an individual patient.27 A previous study in patients with unstable angina, but who had no previous MI or CABG, found that electrocardiographic changes had a high sensitivity (84%) but a low specificity (30%) for detecting cardiac mortality and nonfatal MI.28 In our study, if we combine the CCS angina grades of III and IV as a potential “test,” we have a sensitivity of 20% and a specificity of 90% for the primary end point, or 21% and 89%, respectively, for the secondary end point. The primary role of risk factor models is likely to be in guiding and optimizing the cost-effectiveness of treatment strategies at a population level rather than at an individual level. In addition, such models can play an important role in selecting high-risk patients for inclusion in future clinical trials. In many contexts, identification and treatment of patients who are at highest risk may represent the most cost-effective use of limited health-care resources. In fact, such approaches are now widespread in the national and international guidelines for the initiation of treatments for cardiovascular disease. Many previous risk factor models have focused on subjects without a history of disease. The results from IONA provide tools for the calculation of risk in the group of high-risk subjects with existing stable coronary heart disease who experience angina of effort. Appendix A. Calculating the 18-Month Risk for Every Subject The survival function for a particular time, ti is given by: For example, the model for the primary end point included the following variables:  | X1 = indicator variable for CCS class 2 | β1 = 0.22892 | | | X2 = indicator variable for CCS classes 3 and 4 | β2 = 0.83212 | | | X3 = treatment with nicorandil | β3 = −0.18636 | | | X4 = current smoking | β4 = 0.29802 | | | X5 = MI | β5 = 0.44050 | | | X6 = LVH | β6 = 0.35794 | | | X7 = centered age (age divided by 5 y and 13.32 subtracted) | β7 = 0.05392 | | | | |
The baseline survival function at 18 months (ie, Ŝ0 = 547) for a subject with all of the x variables set to 0 was estimated to be 0.92099. Note that when the variables are set to 0, the exp( xTβ) part of the equation cancels to equal 1. Therefore, the survival function for other subjects can be calculated as follows: (a)For a subject with a CCS score of 2, treated with nicorandil, who smokes, with previous MI but no previous LVH, with a centered age of 0.37309, has an xTβ value equal to (b)survival = exp(exp(0.80120) × loge(0.92099)) = 0.83244 (c)risk = 1 − survival = 0.16756, approximately 17%. For information, the equivalent baseline survival function for the secondary end point at 18 months (Ŝ0 = 547) for a subject with all of the x variables set to 0 was estimated to be 0.98109. The equivalent parameter estimates are the following: | X1 = indicator variable for CCS class 2 | β1 = 0.26830 | | | X2 = indicator variable for CCS classes 3 and 4 | β2 = 0.81273 | | | X3 = current smoking | β3= 0.38341 | | | X4 = MI | β4= 0.71886 | | | X5 = CABG/PTCA | β5= −0.40971 | | | X6 = LVH | β6= 0.39716 | | | X7 = centered age (age divided by 5 y and 13.32 subtracted) | β7= 0.20565 | | | X8 = heart rate (divided by 10 and 6.656 subtracted) | β8= 0.13941 | | | | |
Appendix B. The IONA Study Group Scientific Steering CommitteeVoting member: Prof HJ Dargie (chairman, Glasgow Western Infirmary, Glasgow), Prof I Ford (Glasgow University, Glasgow), Prof KM Fox (Royal Brompton Hospital, London), and Prof WS Hillis (Glasgow University, Glasgow) Sponsor representatives (nonvoting): Dr M Morris (Merck Pharmaceuticals) and Dr M Ford (Aventis Pharma Ltd) Critical Events Committee Prof WS Hillis (Glasgow University, Glasgow), Prof JJV McMurray (Glasgow University, Glasgow), and Dr AL Clark (University of Hull, Kingston upon Hull) Data and Safety Monitoring Committee Prof J Hampton (chairman, University Hospital, Nottingham), Dr A Skene (Nottingham Clinical Research Ltd), and Dr J Birkhead (Northampton General Hospital, Northampton) Statistical and Data Center Robertson Centre for Biostatistics, University of Glasgow, Glasgow: Prof I Ford (director), Dr AD McMahon (study statistician), Ms A Trainer, Ms H Christie (database managers), Ms A Nears (nurse/coder), Ms C Ferrell (study administrator), and Dr B Shaw (clinical coordinator). Study Monitoring Ingenix Pharmaceutical Services: Ms V Diment (project manager) Investigators References 1. 1Tunstall-Pedoe H. Angina pectoris: epidemiology and risk factors. Eur Heart J. 1985;6:1–5. 2. 2Gandhi MM. Clinical epidemiology of coronary heart disease in the UK. Br J Hosp Med. 1997;58:23–27. MEDLINE 3. 3Hemingway H, Shipley M, Britton A, et al. Prognosis of angina with and without a diagnosis: 11 year follow up in the Whitehall II prospective cohort study. BMJ. 2003;327:895. 4. 4Stewart S, Murphy N, Walker A, et al. The current cost of angina pectoris to the National Health Service in the UK. Heart. 2003;89:848–853. 5. 5Antiplatelet Trialists' Collaboration. Collaborative overview of randomised trials of antiplatelet therapy prevention of death, myocardial infarction, and stroke by prolonged antiplatelet therapy in various categories of patients. BMJ. 1995;308:81–106. 6. 6Yusuf S, Sleight P, Pogue J, et al. Effects of an angiotensin-converting-enzyme inhibitor, Ramipril, on cardiovascular events in high-risk patients. N Engl J Med. 2000;342:145–153. MEDLINE |
CrossRef
7. 7Scandinavian Simvastatin Survival Study Group. Randomised trial of cholesterol lowering in 4444 patients with coronary heart disease: the Scandinavian Simvastatin Survival Study (4S). Lancet. 1994;344:1383–1389. Abstract 8. 8Long-term Intervention with Pravastatin in Ischaemic Disease (LIPID) Study Group. Prevention of cardiovascular events and death with pravastatin in patients with coronary heart disease and a broad range of initial cholesterol levels. N Engl J Med. 1998;339:1349–1357. MEDLINE |
CrossRef
9. 9Rubins HB, Sander JR, Collins D, et al. Gemfibrozil for the secondary prevention of coronary heart disease in men with low levels of high density lipoprotein cholesterol. N Engl J Med. 1999;341:410–418. MEDLINE |
CrossRef
10. 10The IONA Study Group. Effect of nicorandil on coronary events in patients with stable angina: the Impact Of Nicorandil in Angina (IONA) randomised trial. Lancet. 2002;359:1269–1275. Abstract | Full Text |
Full-Text PDF (111 KB)
|
CrossRef
11. 11The IONA Study Group. Impact Of Nicorandil in Angina (IONA):design, methodology and management. Heart. 2003;85:E9. 12. 12Campeau L. The Canadian Cardiovascular Society grading of angina pectoris revisited 30 years later. Can J Cardiol. 2002;18:371–379. 13. 13Cox J, Naylor CD. The Canadian Cardiovascular Society grading scale for angina pectoris: is it time for refinements?. Ann Intern Med. 1992;117:677–683. MEDLINE 14. 14Dagenais GR, Armstrong PW, Theroux P, et al. Revisiting the Canadian Cardiovascular Society grading of stable angina pectoris after a quarter of a century of use. Can J Cardiol. 2002;18:941–944. 15. 15Lampe FC, Whincup PH, Shaper AG, et al. Variability of angina symptoms and the risk of major ischemic heart disease events. Am J Epidemiol. 2001;153:1173–1182. MEDLINE |
CrossRef
16. 16Brand FN, Larson M, Friedman LM, et al. Epidemiologic assessment of angina before and after myocardial infarction: the Framingham study. Am Heart J. 1996;132:174–178. Abstract |
Full-Text PDF (558 KB)
|
CrossRef
17. 17Herlitz J, Karlson BW, Richter A, et al. Occurrence of angina pectoris prior to acute myocardial infarction and its relation to prognosis. Eur Heart J. 1993;14:484–491. 18. 18Lampe FC, Whincup PH, Wannamethee SG, et al. Chest pain on questionnaire and prediction of major ischaemic heart disease events in men. Eur Heart J. 1998;19:63–73.
CrossRef
19. 19Rizik DG, Healy S, Margulis A, et al. A new clinical classification for hospital prognosis of unstable angina pectoris. Am J Cardiol. 1995;75:993–997. Abstract |
Full-Text PDF (697 KB)
|
CrossRef
20. 20Van Miltenburg A, Van Zijl AJ, Simoons ML, et al. Incidence and follow-up of Braunwald subgroups in unstable angina pectoris. J Am Coll Cardiol. 1995;25:1286–1292. Abstract |
Full-Text PDF (668 KB)
|
CrossRef
21. 21Drozdz J, Krzeminzska-Pakula M, Chrzanowski L, et al. Predictors of long term outcome in medically treated patients with unstable angina. Can J Cardiol. 2003;19:135–139. 22. 22Tanne D, Shotan A, Goldbourt U, et al. Severity of angina pectoris and risk of ischemic stroke. Stroke. 2003;33:245–250.
CrossRef
23. 23McMahon AD. Observation and experiment with the efficacy of drugs: a warning example using a cohort of NSAID and ulcer healing drug users. Am J Epidemiol. 2001;154:557–562. MEDLINE |
CrossRef
24. 24McMahon AD. Approaches to combat with confounding by indication in observational studies of intended drug effects. Pharmacoepidemiol Drug Saf. 2003;12:551–558. MEDLINE |
CrossRef
25. 25Nakamura Y, Moss AJ, Brown MW, et al. Long-term nitrate use may be deleterious in ischemic heart disease: a study using databases from two large-scale postinfarction studies. Am Heart J. 1999;138:577–585. Abstract | Full Text |
Full-Text PDF (168 KB)
|
CrossRef
26. 26Teo KK, Catellier DJ. Long-term nitrate use in chronic coronary artery disease: need for a randomized controlled trial. Am Heart J. 1999;138:400–402. Full Text |
Full-Text PDF (104 KB)
|
CrossRef
27. 27Wald NJ, Hackshaw AK, Frost CD. When can a risk factor be used as a worthwhile screening test?. BMJ. 1999;319:1562–1565. 28. 28Seres L, Valle V, Marrugat J, et al. Usefulness of hospital admission risk stratification for predicting nonfatal acute myocardial infarction or death six months later in unstable angina pectoris. Am J Cardiol. 1999;84:963–969. Abstract | Full Text |
Full-Text PDF (137 KB)
|
CrossRef
 Reprint requests: Alex D. McMahon, Robertson Centre for Biostatistics, University of Glasgow, G12 8QQ Glasgow, Scotland, UK. E-mail: alexm@stats.gla.ac.uk
PII: S0002-8703(05)00339-X doi:10.1016/j.ahj.2005.03.040 © 2005 Mosby, Inc. All rights reserved. | |
|
FULL TEXT
