Are beta-blockers effective in patients who develop heart failure soon after myocardial infarction? A meta-regression analysis of randomised trials

Ž. European Journal of Heart Failure 2 2000 333 340 Are beta-blockers effective in patients who develop heart failure soon after myocardial infarction? A meta-regression analysis of randomised trials Abstract Timothy Houghton a, Nick Freemantle b, John G.F. Cleland a, a Department of Cardiology, Castle Hill Hospital, Uni ersity of Hull, Kingston-upon-Hull HU16 5JQ, UK b Medicines E aluation Group, Centre for Health Economics, Uni ersity of York, York, UK Received 1 May 2000; received in revised form 9 June 2000; accepted 12 June 2000 Background: The great majority of post-infarction studies of beta-blockers were conducted in an era when these agents were widely held to be contra-indicated for the management of heart failure. We now know that beta-blockers are highly effective for the management of patients with chronic stable heart failure. However, there remains uncertainty about their role in the setting of post-infarction heart failure and ventricular dysfunction. Aim: the primary objective in this paper, was to investigate the extent to which heart failure or evidence of major cardiac dysfuntion influenced outcome in previous trials of beta-blockers in heart failure after myocardial infarction. Methods: We assessed the extent to which the inclusion of patients with heart failure or major cardiac dysfunction influenced outcome in randomised trials of long-term use of beta-blockade after myocardial infarction. The primary analysis was to assess the extent to which the proportion of patients included in each trial with heart failure influenced the relative odds of all-cause mortality in the trials. All randomised trials without crossover with treatment lasting more than one month and with 50 or more patients were considered. All those that provided information on the proportion of patients with heart failure or major cardiac dysfunction in the original or subsequent articles were included in the analysis. Results: Overall treatment with a beta-blocker was associated with a 22.6% reduction in the odds of death Ž 95% C1 11 32.3%.. There were very few data on the effects of beta-blockers after myocardial infarction in patients with documented left ventricular systolic dysfuntion. In the analysis that included heart failure as a factor, treatment with a beta-blocker was associated with a non-significant interaction with the presence of heart failure. However, because the group including heart failure patients were at higher risk, the absolute benefit of treatment with beta-blockers appeared greater in this group. Conclusions: This analysis suggests that the relative benefit of beta-blockers on mortality after a myocardial infarction is similar in the presence or absence of heart failure but that the absolute benefit may be greater in the former. However, as current clinical practice has changed radically from the time when the majority of these trials were conducted, further trial evidence would be desirable. 2000 Society of Cardiology. All rights reserved 1. Introduction Clinical trials show that beta-blockers are safe and effective when administered to patients with chronic, Corresponding author. Tel.: 44-1482-624084; fax: 44-1482- 624085. stable mild or moderate heart failure 1 4. Metaanalysis of published trials indicates a 35% Ž95% CI 25 45%. reduction in the odds of death. Beta-blockers have also reduced mortality after myocardial infarction in several large studies and meta-analysis has shown a 23% Ž 95% CI 15 31%. reduction in the odds of death 5. However, beta-blockers remain underutilised in both settings. Indeed, the appearance of 1388-9842 00 $20.00 2000 European Society of Cardiology. All rights reserved. Ž. PII: S 1 3 8 8-9 8 4 2 0 0 00100-8

334 ( ) T. Houghton et al. European Journal of Heart Failure 2 2000 333 340 heart failure after a myocardial infarction is often cited as a reason for withholding or even withdrawing a beta-blocker. Most of the randomised controlled trials of betablockers were conducted in the 1970s and 1980s at a time when beta-blockers were believed to be contraindicated for heart failure. Consequently, many trials of beta-blockers in this setting deliberately excluded patients with heart failure. Only two small trials of conventional beta-blockers deliberately targeted patients with heart failure or major left ventricular dysfunction, one of which showed a trend to harm 4,6. The one trial of post-infarction heart failure that did show benefit included only patients with diastolic heart failure 6. Currently, patients with evidence of major left ventricular systolic dysfunction in the aftermath of a myocardial infarction, who frequently have or are at high risk of developing heart failure, are often not treated with beta-blockers. Because of an absence of conclusive clinical trial evidence in this setting, a large randomised controlled trial ŽCAPRICORN: CArvedilol Post-infa Rct survi val COnt Rol in LV dysfunctio N., with the two primary end-points of Ž a. all-cause mortality or cardiovascular hospitalisation and Ž b. all-cause mortality, has been designed, conducted and reported elsewhere in this issue. Our objective, in this paper, is to investigate, using metaregression analysis, the extent to which heart failure or evidence of major cardiac dysfunction influenced outcome in previous trials of beta-blockers in heart failure. 2. Methods 2.1. Objecti es We assessed the extent to which the inclusion of patients with heart failure or evidence of major cardiac dysfunction influenced the outcome in randomised trials of long-term use of blockade in patients with a history of myocardial infarction. Our primary analysis was to examine the extent to which the proportion of patients included in each trial with heart failure influenced the relative odds of all cause mortality in the trials. Only four small trials measured left ventricular function objectively, one of which excluded systolic dysfunction. Only a few trials reported an alternative objective measure of cardiac dysfunction. Accordingly, this type of analysis was conducted using only heart failure criteria. In addition, we undertook a systematic review of the above trials to identify those that reported outcome in subgroups of patients with evidence of major cardiac dysfunction or heart failure to identify whether outcome differed among patients with or without such problems. 2.2. Inclusion criteria We included randomised controlled trials, without cross-over, with treatment lasting more than 1 month, and with follow up that examined the clinical effectiveness of blockers vs. placebo or control in patients who had experienced a myocardial infarction. Only trials which randomised 50 or more patients per group were included. This disqualified only three small trials, two of which excluded patients with heart failure at baseline. Treatment may have begun at any stage after myocardial infarction, and may have been commenced intravenously. Trials were only included if the proportion of patients in the trials that had signs of heart failure or evidence of major cardiac dysfunction was reported either in the original or subsequent articles. 2.3. Search strategy We conducted sensitive electronic searches of Medline, Embase, Biosis, Healthstar, Sigle, IHTA, Conference Papers Index, Derwent Drug File, Dissertation Abstracts, Pascal, International Pharmaceutical Abstracts and Science Citation Index. We reviewed the reference list of each identified study. We also examined existing bibliographies and reviews for relevant studies, including data from the Beta-Blocker Pooling Project 7. 2.4. Data abstraction and appraisal of study quality The following data from each study were abstracted; number of patients randomised to active treatment or control, the blocker used and the route and dose, duration of treatment, loss to follow up, level of blinding, concealment of allocation, specific study inclusion and exclusion criteria, duration of follow up and deaths from all causes. 2.5. Statistical analysis Standard statistical methods for meta analysis are not appropriate for meta regression analysis. Thus, we used a full random effects approach, based upon Markov chain Monte Carlo simulation, with appropriate uninformative priors using the BUGS software described by Smith et al. 8. This provides a robust treatment estimate that avoids the large sample assumptions of standard methods, can account for the

( ) T. Houghton et al. European Journal of Heart Failure 2 2000 333 340 335 Table 1a Characteristics of trials of beta-blockers where rate of heart failure is reported a Trial Follow up Drug Initiation. Double High risk Number comparison blind? patients only? randomised Multicentre Int. Study c 10 24 months Practolol Ž S. 7 28 days Yes No 3053 Taylor et al. 9 48 months Oxprenolol Ž ISA. Remote Yes No 1103 TIARA 19 24 months Timolol iv 6hrs Yes No 200 EIS c 20 1 year Oxprenolol Ž ISA. 14 36 days Yes No 1736 SSSD 4 3 years Metoprolol Ž S. 10 60 days No Yes 253 BHAT c 21 25 months Propranolol Ž NS. 5 21 days Yes No 3837 Hjalmarson et al. 22 3 months Metoprolol Ž S. iv Yes No 1395 Baber et al. c 23 3 9 months Propranolol 2 14 days Yes Anterior only 720 Julian et al. c 24 12 months Sotalol Ž NS. 5 14 days Yes No 1456 LIT 25 18 months Metoprolol Ž S. 6 16 days Yes No 2395 NMSG c 26 17 months Timolol Ž NS. 7 28 days Yes No 1884 Basu et al. 27 6 months Carvedilol Ž NS. iv Yes No 146 Hansteen et al. c 28 1 year Propranolol Ž NS. 4 days Yes No 560 APSI 29 318 days Acebutolol Ž NS. Av. 11 days Yes Yes 607 Australian Swedish Study 2 years Pindolol Ž ISA. 1 21 days Yes Yes 529 Group c 31 76% 5 days Olsson et al. 32 36 month Metoprolol Ž S. Pre-discharge Yes No 301 Aronow et al. 6 1 year Propanolol Ž NS. 2 months No Yes 158 inclusion of trials in which one or more groups experience no events without crude fixes, and is highly suited to meta regression analysis. Table 1b In the analysis the logit of the probability of death Ž pt. in the intervention group was attributed to a constant Ž., the effects of treatment Ž., the inclu- Trial Follow up Drug Initiation. Loss to % Heart Definition of comparison follow up failure heart failure Ž %. b or LVD Multicentre Int. Study c 10 24 months Practolol Ž S. 7 28 days 3.4% Excluded Not defined Taylor et al. 9 48 months Oxprenolol Ž ISA. Remote Not clear Excluded CXR TIARA 19 24 months Timolol iv 6hrs 0% Excluded Rales, S3 or CXR EIS c 20 1 year Oxprenolol Ž ISA. 14 36 days Not clear 13% Not defined SSSD 4 3 years Metoprolol Ž S. 10 60 days 1.90% 14% LVEF 20-45% All EF s 45 BHAT c 21 25 months Propranolol Ž NS. 5 21 days 0.3% 19% Not defined Hjalmarson et al. 22 3 months Metoprolol Ž S. iv 1.6% 19% Rales, i.v. diuretics Baber et al. c 23 3 9 months Propranolol 2 14 days Not clear 20% Not defined Julian et al. c 24 12 months Sotalol Ž NS. 5 14 days 0 22% Not defined LIT 25 18 months Metoprolol Ž S. 6 16 days 0.20% 25% Not defined NMSG c 26 17 months Timolol Ž NS. 7 28 days Not clear 33% Rales, S3 or CXR Basu et al. 27 6 months Carvedilol Ž NS. iv 0% 36% 32% had LVD EF EF 45% 45% HF not defined Hansteen et al. c 28 1 year Propranolol Ž NS. 4 days 0% 40% Rales and CXR APSI 29 318 days Acebutolol Ž NS. Av. 11 days 0 50% Not defined Australian Swedish Study 2 years Pindolol Ž ISA. 1 21 days Not clear 61% Rales or CXR Group c 31 76% 5 days Olsson et al. 32 36 month Metoprolol Ž S. Pre-discharge 0% 66% CXR & rales Aronow et al. 6 1 year Propanolol Ž NS. 2 months Not clear 100% Rales and CXR LVSD excluded a Patients in New York Heart Association Ž NYHA. Class II or III despite treatment with diuretics and ACE inhibitors for 2 months and with LVEF 40% Ž i.e. diastolic dysfunction.. LVEF, left ventricular ejection fraction; CTR, cardiothoracic ratio; i.v., intravenous; S3, third heart sound; CXR, chest X-ray. b Uncontrolled or NYHA class IV heart failure excluded. c Included in Beta-Blocker Pooling Project. a

336 ( ) T. Houghton et al. European Journal of Heart Failure 2 2000 333 340 sion of heart failure patients Ž I. and its effect on outcome Ž.. pt logž / I 1 pt pc logž / 1 pc 3. Results 3.1. Included trials In total, 17 trials met the inclusion criteria Žsee Table 1 for references.. These trials included 20 333 subjects, 1882 Ž 9.3%. of whom died over a mean follow-up of 22 months. The proportion of patients included with heart failure varied substantially between trials. For example, three trials contained no patients with heart failure 9,10,19. By contrast, two trials contained only patients with heart failure or major ventricular dysfunction 4,6 although one of these included only diastolic heart failure 6. The characteristics of included trials are described in Table 1. 3.2. O erall effect of blockade upon all cause mortality Overall, treatment with a blocker was associated with a 22.6% reduction in the odds of death Ž95% CI 11.0 32.3%.. This effect is very similar to the effect observed in a meta-analysis that included all postinfarction trials of beta-blockers regardless of whether or not they excluded patients with heart failure 5. The odds ratios for individual trials, exact 95% confidence intervals and full random effects estimates of the effect of blockade on all cause mortality, plotted in order of increasing proportion of heart failure patients included, are described in Fig. 1. 3.3. Meta regression analysis: the influence of the proportion of patients with heart failure included In the analysis that used the proportion of patients included with heart failure as a factor, treatment with a blocker was associated with a non-significant interaction with the presence of heart failure odds Fig. 1. Effect of blockade in trials reporting rates of heart failure. Odds ratios, pooled odds ratios and 95% confidence intervals. Trials plotted in order of increasing proportion of heart failure patients included. See Tables 1 and 2 for references.

Table 2 Characteristics of trials of beta-blockade where data from subgroups of patients with heart failure or other evidence of major ventricular dysfunction were reported. Note that patients with severe uncontrolled heart failure were excluded from these studies Trial Follow Stratified by Number of Deaths Number Randomised Ž % Mortality. up and Control Active Control Active Control Active drug overall overall low risk low risk high risk high risk EIS b 20 1 year Clinical heart 45 883 57 858 33 767 43 746 12 116 14 112 Oxprenolol failure Ž 5.1%. Ž 6.6%. Ž 4.3%. Ž 5.8%. Ž 10.4%. Ž 12.5%. BHAT b 21 25 months Clinical heart 188 1921 138 1916 121 1556 92 1571 67 365 46 345 Chadda et al. 13 Propranolol failure Ž 13.3%. Ž 10.3%. Ž 7.8%. Ž 5.9%. Ž 18.4%. Ž 13.3%. Hjalmarson et al. 22 3 months Clinical heart 62 697 40 698 37 566 27 567 25 131 13 131 Herlitz et al. 33 Metoprolol failure Ž 8.9%. Ž 5.7%. Ž 6.5%. Ž 4.8%. Ž 19.1%. Ž 9.9%. Baber et al. b 23 3 9 months Clinical heart 27 365 28 355 22 288 21 289 5 77 7 66 Propranolol failure Ž 7.4%. Ž 7.9%. Ž 7.6%. Ž 7.3%. Ž 6.5%. Ž 10.6%. a,b NMSG 26 17 months Heart 152 939 98 945 63 591 41 608 63 216 40 204 Gundersen 12 Timolol volume Ž 16.2%. Ž 10.4%. Ž 10.7%. Ž 6.7%. Ž 29.2%. Ž 19.6%. Basu et al. 27 6 months LVEF 45% 3 71 2 75 1 46 1 51 2 25 1 24 Carvedilol Ž 4.2%. Ž 2.7%. Ž 2.2%. Ž 2.0%. Ž 8.0%. Ž 4.2%. APSI a 30,31 318 days Mechanical 34 309 17 298 13 158 6 138 21 151 11 160 Acebutolol failure Ž 11.0%. Ž 5.7%. Ž 8.2%. Ž 4.3%. Ž 13.9%. Ž 6.9%. Australian Swedish 2 years Clinical heart 47 266 45 263 18 106 11 100 29 160 34 163 Study Group b 31 Pindolol failure Ž 17.7%. Ž 17.1%. Ž 17.0%. Ž 11.0%. Ž 18.1%. Ž 20.9%. Totals 558 5451 425 5408 308 4078 242 4070 224 1241 166 1205 Ž 10.2%. Ž 7.9%. Ž 7.6%. Ž 5.9%. Ž 18.0%. Ž 13.8%. Relative reduction in risk over 2 years 0.79 Ž 95% CI 0.65 0.97. 0.78 Ž 95% CI 0.061 0.99. 0.79 Ž 95% CI 0.065 0.96. Lives saved per 1000 treated over 2 years 23 of 102 deaths 17 of 76 deaths 42 of 180 deaths Aronow et al. 6 1 year Heart failure with LVDD only 60 79 Ž 75.9%. 44 79 Ž 55.7%. SSSD 4 3 years LVSD only Ž with or without heart failure. 9 123 Ž 7.3%. 17 130 Ž 13.1%. a Outcome was reported for normal, borderline and definitely enlarged hearts in the original paper. The above table reports outcome for all randomised patients and for normal sized and definitely enlarged hearts. Among 262 patients with borderline increase in heart size at baseline 19.8% of those randomised to control and 13% of those randomised to timolol died. Three patients had no heart size reported. b Included in Beta-Blocker Pooling Project. T. Houghton et al. European Journal of Heart Failure 2 ( 2000 ) 333 340 337 Downloaded from http://eurjhf.oxfordjournals.org/ at Pennsylvania State University on March 4, 2014

338 ( ) T. Houghton et al. European Journal of Heart Failure 2 2000 333 340 ratio for the interaction 0.60 Ž 95% CI 0.32 1.13%.. In other words, the proportion of patients included with heart failure had no clear effect on the benefits of beta-blockers and the results may be interpreted as a lack of sufficient evidence to show that the benefits of beta-blockers after myocardial infarction are different in patients with and without heart failure. 3.4. Subgroup analysis of patients with heart failure or entricular dysfunction after myocardial infarction A small number of trials described all cause mortality for sub-groups of patients with evidence of heart failure or major left ventricular dysfunction. These trials are shown in Table 2. These studies suggested similar relative benefits with beta-blockers compared to placebo in patients with and without heart failure or ventricular dysfunction. However, patients with heart failure were at a greater absolute risk, therefore, the absolute benefits of treatment appears greater in this group. Similarly, Table 3, which is derived from the Beta Blocker Pooling Project 7, describes the comparative benefits in patients with and without heart failure or major ventricular dysfunction, using a variety of descriptors, in the trials included in that analysis. Overall, this analysis also suggests similar relative benefits with beta-blockers in patients with and without heart failure and, therefore, that the absolute benefits of treatment may be greater in this group of patients. 4. Discussion This analysis suggests that, for patients selected for inclusion in post-infarction trials in the 1970s and 1980s, the relative benefit was similar for patients with or without heart failure or major cardiac dysfunction. As patients with heart failure are at greater absolute risk, a similar relative benefit will translate into a greater absolute benefit, in terms of the number of deaths avoided. For example, if patients without heart failure have a 5% annual mortality and patients with heart failure have 20% mortality postmyocardial infarction then, assuming a 22% relative reduction in risk with a beta-blocker, for every 1000 patients treated, 11 and 44 deaths, respectively, attributable to the blocker could be prevented. Superficially, it would now seem appropriate to recommend that most patients with post-infarction heart failure or major left ventricular dysfunction should be treated with a combination of ACE inhibitors and beta-blockers. However, care should be taken in extrapolating the above data to current practice for several reasons. The trials of beta-blockers were conducted in an era before the widespread use of imaging to assess post-infarction left ventricular function. Although it is likely that many patients with post-infarction heart failure had severe left ventricular systolic dysfunction the possibility that the patients with heart failure who benefited in the trials reported so far had well preserved left ventricular systolic dysfunction cannot be discounted. Forty percent of patients without major left ventricular function developed heart failure in the TRACE study, whereas approximately 25% of those with major left ventricular systolic dysfunction did not 11. Indeed, numerically, there were similar numbers of patients with heart failure with and without systolic dysfunction in the TRACE study. In the studies of beta-blockers post infarction there were too few data to ascertain whether similar benefits would have been observed in patients with left ventricular systolic dysfunction although patients with increased heart size did appear to benefit in one study 13. Of note, the most impressive results in this review were from a trial in which patients had heart failure due to diastolic dysfunction, many of whom also had cardiomegaly 7. The trials of beta-blockers post-infarction were conducted prior to the widespread use of ACE inhibitors and, to some extent, thrombolysis 5. It is not clear that beta-blockers remain effective after myocardial infarction when taken in combination with Table 3 Selected sub-groups reported in the Beta-Blocker Pooling Project of nine trials with 13 679 patients 7 Characteristics Numbers Placebo Mortality Odds ratio P All patients 13 679 7.2% 0.76 0.0001 History of heart failure Absent 9268 7.1% 0.78 0.004 a preceding MI Present 797 13.9% 0.87 0.52 Mechanical failure Absent 10 160 5.3% 0.78 0.007 at baseline Present 3519 12.8% 0.73 0.004 Treatment with Absent 8799 7.2% 0.78 0.004 diuretics Present 1438 11.9% 0.77 0.012 Treatment with Absent 9280 6.8% 0.81 0.01 digoxin Present 957 18.8% 0.63 0.01 a MI, myocardial infarction.

( ) T. Houghton et al. European Journal of Heart Failure 2 2000 333 340 339 ACE inhibitors. Beta-blockers are effective when added to an ACE inhibitor in patients with chronic, stable heart failure and left ventricular systolic dysfunction 1. However, the combined hypotensive effect of ACE inhibitors and beta-blockers in the setting of myocardial infarction may not be desirable. Until recently, beta-blockers were commonly believed to be contra-indicated in patients with heart failure. Consequently, it is likely that those patients with heart failure that were recruited into the trials in this retrospective analysis had rather mild heart failure. It is likely that randomised medication was frequently withdrawn if heart failure developed. These trials employed intention to treat analyses and so treatment withdrawal could have led to an underestimate of the real benefits of beta-blockade in these patients. Treatment withdrawal might also have contributed to the safety of beta-blockers in this setting. However, overall, most long-term trials of beta-blockers post-infarction have reported little overall difference in withdrawals for worsening heart failure 13,22, an early excess of withdrawals for worsening heart failure being followed, in the longer-term, by a similar number of or fewer withdrawals. The CAPRICORN study also excluded patients with severe heart failure and could include asymptomatic patients. Therefore, on current evidence, patients with severe or progressively worsening heart failure should not receive a beta-blocker immediately after myocardial infarction. However, the results of the COPERNICUS trial, ŽCarvedilOl P rosp Ective Ra Ndom Ised CUmulative Survival T rial. comparing carvedilol and placebo in patients with severe stable heart failure, suggests that delayed initiation after stabilisation is indicated even in this group of patients. All the studies conducted so far with beta-blockers in heart failure have required relatively long periods of clinical stability and have excluded patients with a recent deterioration of their heart failure. Beta-blockers have been shown to be safe in patients with chronic stable heart failure but their safety in patients with recent onset or worsening of heart failure postinfarction has not been adequately tested. In the CAPRICORN study, carvedilol was initiated 3 7 days after myocardial infarction. The protocol required clinical and haemodynamic stability for only 24 h, which is a radical departure from previous recommendations and practice. For all of the above reasons it would be desirable to confirm that beta-blockers are effective for patients with post-infarction left ventricular systolic dysfunction and or heart failure in contemporary clinical practice. A further question of key importance is whether all beta-blockers are similarly effective. An analysis of all post-infarction trials failed to show differences between beta-blockers although agents with increased intrinsic sympathomimetic activity Ž ISA. tended to be less effective 5. It is notable that in the trials of oxprenolol and pindolol reported here, patients with heart failure tended to have a higher mortality on the beta-blocker than on placebo. In addition, the data-set for atenolol, one of the more widely used beta-blockers post-infarction, proved to be weak. Classification of beta-blockers into those with, or without, important cardio selectivity, lipophilicity or ISA is not clear cut, and there is some debate in the literature on the attributes of acebutolol in particular 1 16. Increased ISA may also account for the lack of effect of bucindolol in the BEST study, a study of patients with relatively severe heart failure 17. It is clear that beta-blockers are pharmacologically different and it is uncertain whether it is appropriate to assume that there is a class benefit on morbid and fatal events. However, few direct comparisons of beta-blockers post-infarction exist. A large multi-centre, international trial comparing carvedilol and metoprolol has completed recruitment 18 Ž COMET.. The results of this study should be highly informative regarding the future choice of beta-blockers, not only for heart failure but also for patients after myocardial infarction. References 1 Cleland JGF, Freemantle N, McGowan J, Clark A. The evidence for beta-blockers equals that for ACE inhibitors in heart failure. Br Med J 1999;318:824 825. 2 CIBIS-II investigators and committees. The cardiac insufficiency bisoprolol study II Ž CIBIS-II.: a randomised trial. Lancet 1999;353:9 13. 3 Hjalmarson A, Goldstein S, Fagerberg B et al. Effects of controlled-release metoprolol on total mortality, hospitalizations, and well-being in patients with heart failure: the metoprolol CR XL randomised intervention trial in congestive heart failure Ž MERIT-HF.. J Am Med Assoc 2000;283: 1295 1302. 4 For the SSSD investigators Lopez FN, Cosin J, Marrugat J, Guindo J, Lunda ABD. Comparison of the effects of amiodarone versus metoprolol on the frequency of ventricular arrhythmias and on mortality after acute myocardial infarction. Am J Cardiol 1993;72:1234 1238. 5 Freemantle N, Cleland JGF, Young P, Mason J, Harrison J. What is the current place of beta-blockade in secondary prevention after myocardial infarction? A systematic overview and meta regression analysis. Br Med J 1999;318:1730 1737. 6 Aronow WS, Ahn C, Kronzon AI. Effect of propranolol versus no propranolol on total mortality plus non-fatal myocardial infarction in older patients with prior myocardial infarction, congestive heart failure and left ventricular ejection fraction 40% treated with diuretics plus angiotensin converting enzyme inhibitors. Am J Cardiol 1997;80:207 209. 7 Beta-Blocker Pooling Project Research Group. The Beta- Blocker Pooling Project Ž BBPP.:subgroup findings from randomized trials in postinfarction patients. Eur Heart J 1988; 9:8 16. 8 Smith TC, Spiegelhalter DJ, Thomas A. Bayesian approaches

340 ( ) T. Houghton et al. European Journal of Heart Failure 2 2000 333 340 to random-effects meta analysis: a comparative study. Stat Med 1995;14:2685 2699. 9 Taylor SH, Silke B, Ebbutt A, Sutton GC, Prout BJ, Burley DM. A long term prevention study with oxprenolol in coronary heart disease. N Engl J Med 1982;307:1293 1301. 10 Multicentre International Study. Improvement in prognosis of myocardial infarction by long term -adrenorceceptorblockade using practolol. Br Med J 1975;3:735. 11 Kober L, Torp Pedersen C, Carlsen JE et al. A clinical trial of the angiotensin-converting-enzyme inhibitor trandolapril in patients with left ventricular dysfunction after myocardial infarction. N Engl J Med 1995;333:1670 1676. 12 Gundersen T. Influence of heart size on mortality and reinfarction in patients treated with timolol after myocardial infarction. Br Heart J 1983;50Ž. 2 :135 139. 13 Chadda K, Goldstein S, Byington R, Curb JD. Effect of propranolol after acute myocardial infarction in patients with congestive heart failure. Circulation 1986;73:503 510. 14 McDevitt DJ. The assessment of -adrenoreceptor blocking drugs in man. Br J Clin Pharmacol 1977;4:413 425. 15 Feely J, Devane PJ, MacClean D. Beta blockers and sympathomimetics. Br Med J 1983;286:1043 1047. 16 Hampton JR. Choosing the right beta blocker: a guide to selection. Drugs 1994;48Ž. 4 :549 568. 17 Witte K, Thackray S, Banerjee T, Clark AL, Cleland JGF. Update of ELITE-II, BEST, CHAMP and IMPRESS clinical trials in heart failure. Eur Heart Fail 2002:107 112. 18 McGowan J, Murphy R, Cleland JGF. Carvedilol for heart failure; clinical trials in progress. Heart Fail Rev 1999;4:89 95. 19 Roque F, Amuchastegui LM, Lopez Morillos MA, Mon GA, Girotti AL et al. Beneficial effect of timolol on infarct size and late ventricular tachycardia in patients with acute myocardial infarction. Circulation 1987;76:610 617. 20 European Infarction Study Group. A secondary prevention study with slow release oxprenolol after myocardial infarction: morbidity and mortality. Eur Heart J 1984;5:189 202. 21 -Blocker Heart Attach Trial Research Group. A randomized trial of propranolol in patients with acute myocardial infarction. 1. Mortality results. J Am Med Assoc 1982; 247:1707 1714. 22 Hjalmarson A, Elmfeldt D, Herlitz J et al. Effect on mortality of metoprolol in acute myocardial infarction: A double blind randomised trial. Lancet 1981;ii:823 827. 23 Baber NS, Wainwright Evans D, Howitt G et al. Multicentre post-infarction trial of propranolol in 49 hospitals in the United Kingdom, Italy and Yuogslavia. Br Heart J 1980; 44:96 100. 24 Julian DG, Prescott RJ, Jackson FS, Szekely P. Controlled trial of sotalol for one year after myocardial infarction. Lancet 1982;i:1142 1147. 25 Lopressor Intervention Trial Research Group. The lopressor intervention trial: multicentre study of metoprolol in survivors of acute myocardial infarction. Eur Heart J 1987;8:1056 1064. 26 Norwegian Multicentre Study Group. Timolol induced reduction in mortality and reinfarction in patients surviving acute myocardial infarction. N Engl J Med 1981;304:801 807. 27 Basu S, Senior R, Raval U, van der Does R, Bruckner T, Lahiri A. Beneficial effects of intravenous and oral carvedilol treatment in acute myocardial infarction: a placebo controlled, randomized trial. Circulation 1997;96:183 191. 28 Hansteen V, Møinichen E, Lorentsen E et al. One year s treatment with propranolol after myocardial infarction: preliminary report of Norwegian multicentre trial. Br Med J 1982;284:155 160. 29 APSI Investigators Boissel JP, Leizorovicz A, Picolet H, Ducruet T. Efficacy of acebutolol after acute myocardial infarction Ž the APSI trial.. Am J Cardiol 1990;66:24c 31c. 30 For the APSI investigatorsboissel JP, Leizorovicz A, Picolet H, Peyrieux JC. Secondary prevention after high-risk acute myocardial infarction with low-dose acebutolol. Am J Cardiol 1990;66:251 260. 31 Australian and Swedish Pindolol Study Group. The effect of pindolol on the two years mortality after complicated myocardial infarction. Eur Heart J 1983;4:367 375. 32 Olsson G, Rehnqvist N, Sjogren A, Erhradt L, Lundman T. Long-term treatment with metoprolol after myocardial infarction: effect on 3 year mortality and morbidity. J Am Coll Cardiol 1985;5:1428 1437. 33 Herlitz J, Waagstein F, Lindqvist J, Swedberg K, Hjalmarson A. Effect of metoprolol on the prognosis for patients with suspected acute myocardial infarction and indirect signs of congestive heart failure Ža subgroup analysis of the Goteborg Metoprolol Trial.. Am J Cardiol 1997;80Ž 9B.:40J 44J.