261
EDITORIALS
Surrogate measures
in clinical
trials Management of acute myocardial infarction was radically changed when thrombolysis was shown to improve case-fatality.l-4 Like most advances in medicine, the solution of one problem has raised new and important questions. Among the most pressing is whether different agents have a similar impact on the outcome of infarction; differences in the cost of these agents are also relevant. In this issue (p 254) Dr Hogg and his colleagues in Glasgow report a comparison of two leading thrombolytic compounds-intravenous streptokinase and anistreplase. Anistreplase is a complex of streptokinase and acylated plasminogen, and has the theoretical advantage of local streptokinase release at the site of clot formation and the practical advantage of being administered over 5 minutes rather than 1 hour. There was no difference in coronary artery patency rates measured at angiography 90 minutes and 24 hours after administration of the drugs in 128 patients who had been randomised to treatment after sustaining an acute infarction. The ultimate aim of thrombolysis is not to secure coronary artery patency-no patient will be grateful to his doctor for improvement in angiographic
appearances-but to improve post-infarction morbidity and especially mortality. The difficulty here is that the mortality rate for myocardial infarction is so low that very large numbers are needed to achieve the necessary statistical power. Trials of streptokinase have recruited 118061 and 17 1872 patients while 13 318 were mustered to test alteplase The AIMS study of anistreplase was exceptionally small by these standards since it was halted after 1004 patients had been recruited.3All these studies assessed the benefits of thrombolytic therapy by comparison with a control group who did not receive such treatment. Since each agent was shown to be beneficial, the differential benefit achieved when different agents are compared would be smaller and so even larger numbers would be required to achieve satisfactory statistical power. Apart from the cost and difficulty of setting up such massive trials, the time taken to achieve a definitive result is correspondingly lengthened. What should one do? One possibility is to pool the results of existing mortality trials. The Glasgow workers do this and
conclude that there is no difference in outcome between streptokinase and anistreplase. Such analyses are of limited value: selection of patients for clinical trials in myocardial infarction is notoriously variable, as the mortality rates in control groups show, and spurious differences can be created and real differences overlooked when trials are pooled. Another approach is to use a surrogate measure of success which will yield enough end-points for adequate analysis. Coronary artery patency provides a ready surrogate for coronary mortality but raises scientific and ethical issues that are inextricably linked. The scientific issue is simple since there is nothing intrinsically unacceptable in the use of surrogate measures in clinical trials. A veritable industry has been built on clinical trials devoted to one surrogate measure for cardiovascular disease-blood pressure. Evidence of efficacy of two widely used classes of drugs in prevention of cardiovascular disease-angiotensin converting enzyme inhibitors and calcium antagonists-is based entirely on trials in which blood pressure reduction was accepted as a surrogate measure of cardiovascular events. The criteria for acceptability are pragmatic: how far is the use of a surrogate measure necessary and how appropriate is the particular measure selected? The first question has already been answered since adequate mortality data will not be available for a considerable time. The second question is more difficult. Under normal circumstances prognosis is clearly better with a patent coronary artery than with an occluded artery, but this assumption cannot be made about a coronary artery that is patent as a result of thrombolysis. However, the trial datal-4 clearly indicate persistence of benefit after discharge of patients from hospital, and there is nothing to suggest that coronary artery patency will carry a different subsequent prognosis as a result of different pharmacological properties of the agents used. In GISSI-2 (streptokinase versus alteplase) the pursuit of relevant end-points has taken a different direction. On p 289 the Italian trial organisers justify their use of a combined index-mortality plus survival to discharge with poor left-ventricular function. What about the ethical issue? The method of demonstrating coronary artery patency used by Dr
262
Hogg and colleagues was invasive, necessitating two coronary angiograms within 24 hours. That coronary angiography has been used extensively in the management of acute myocardial infarction as a prelude to angioplasty or surgery does not automatically justify its use in this study because the procedure did not lead to any change in treatment of individual patients. Moreover, the most common clinical complication-bleeding from the catheter site-was directly attributable to coronary angiography. Ethical acceptability of this trial requires that additional serious risks are very slight and that its scientific value is commensurate with the extra discomfort and risks involved. We have previously expressed disquiet about experiments at the Hammersmith Hospital in which multiple for a physiological investigation was used angiography of the action of vasoactive peptides in the coronary tree.6 By contrast, the clinical value of the Glasgow trial seems undeniable, but would of course not justify exposing patients to a significant risk. Thus acceptability stands or falls by the definition of acceptable risk in this case. In a fully equipped cardiac unit with supporting facilities, the risks of serious complications are low6 and would seem justifiable in the Glasgow study. These results are a valuable step in the evaluation of the main therapeutic advances of recent years; open discussion of the issues involved is nevertheless essential to preserve standards of scientific clinical investigation. Italiano Per Lo Studio Della Streptochinasi Nell’Infarto Miocardico (GISSI). Effectiveness of intravenous thrombolytic treatment in acute myocardial infarction. Lancet 1986; i: 397-401. 2. ISIS-2 Collaborative Group. Randomised trial of intravenous streptokinase, oral aspirin, both, or neither among 17 187 cases of suspected acute myocardial infarction: ISIS-2. Lancet 1988; ii: 349-60. 3. AIMS Trial Study Group. Effect of intravenous APSAC on mortality after acute myocardial infarction: preliminary report of a placebocontrolled clinical trial. Lancet 1988; i: 545-49. 4. Wilcox RG, Olsson CG, Skene AM, et al. Trial of tissue plasminogen activator for mortality reduction in acute myocardial infarction. Anglo-Scandinavian Study of Early Thrombolysis (ASSET). Lancet 1988; ii: 525-30. 5. Editorial. Peptides into the coronary arteries. Lancet 1989; ii: 958-59. 6. Cumberland DC. Coronary angiography and angioplasty. Medicine Int 1989; 68: 2834-36.
1. Gruppo
Are you
taking the medicine?
Thomas Willisl observed in the 17th century that severe asthma could be life-threatening, the low case-fatality rate from the disorder among patients of an individual practitioner subsequently encouraged the view of Osler two hundred years later that asthma is seldom fatal. New Zealand has lately experienced a second "epidemic" of asthma deaths,33 equal in magnitude to that of the 1960s, while the death rate in the UK has probably been rising steadily over the past ten years, especially among males aged 15-34 years.4 The increase in asthma mortality has occurred against a backdrop of increased prescribing of anti-asthma therapy. There are many reasons why
Although
influenced case-fatality rates, most notably poor assessment of the severity of asthma by patients and doctors,6 and a reluctance to prescribe adequate therapy, especially steroids.7 Another important factor is that patients often do not adhere to the prescribed drug regimen. Studies of patient compliance are remarkably consistent in revealing full compliance rates of only 30-50% (in whatever condition is studied8), and patients with asthma are no different, despite the symptomatic nature of their disease.9-11 treatment
has
not
Compliance may be best defined as the extent to which a person’s behaviour (whether in terms of taking drugs, following diets, or executing changes in personal lifestyle) coincides with medical advice. Such definition assumes that the recommendations of the patients’ medical advisors are always correct and should therefore be followed. Full and complete compliance is the state in which the patient takes all the prescribed drug for the entire period intended by the prescribing physician. The term partial compliance implies that the patient has taken only some of the total intended number of doses, either taking the drug fewer times each day, though for the correct number of days, or possibly the full dose but for a shorter period than prescribed. The biological effects of taking all doses for only the first half of a course may, of course, be profoundly different from those of taking only half the prescribed number of tablets every day. Asthmatic patients unfortunately may require both long-term treatment taken 2-4 times a day and short courses of oral drugs such as steroids. Systematic study of the extent and determinants of patients’ compliance and non-compliance with antiasthma therapy has been hampered by the difficulties of assessment. Simply asking patients whether they are taking their drugs is easy and universally applicable but is notoriously unreliable, overestimating true compliance by roughly 30%.12 Assessment by weighing returned pressurised aerosol cannisters is confounded both by the widespread practice of conducting "test-firings" before actual inhalation of drug and by the concurrent use of multiple different inhalers of the same drug (especially of rescue beta agonists). Counting returned dry powder capsules may merely drive poorly compliant patients away from follow-up, increasing noncompliance further. Mechanical devices have been devised to monitor use of pressurised aerosols automatically (eg, the ’Nebuliser Chronolog’ manufactured by ATP, Denver, USA13), but until they are an integral part of the inhaler the influence of the device itself on compliance will remain uncertain. Reiser and Warner14 argued that the mere inclusion of patients in a prospective study may lead to a
unrepresentatively high compliance rates simply because these patients receive more attention. However, there are no objective prospective studies of the influence of such trials on compliance for good or bad. Only unannounced assay of drugs in biological
EDITORIALS
Surrogate measures
in clinical
trials Management of acute myocardial infarction was radically changed when thrombolysis was shown to improve case-fatality.l-4 Like most advances in medicine, the solution of one problem has raised new and important questions. Among the most pressing is whether different agents have a similar impact on the outcome of infarction; differences in the cost of these agents are also relevant. In this issue (p 254) Dr Hogg and his colleagues in Glasgow report a comparison of two leading thrombolytic compounds-intravenous streptokinase and anistreplase. Anistreplase is a complex of streptokinase and acylated plasminogen, and has the theoretical advantage of local streptokinase release at the site of clot formation and the practical advantage of being administered over 5 minutes rather than 1 hour. There was no difference in coronary artery patency rates measured at angiography 90 minutes and 24 hours after administration of the drugs in 128 patients who had been randomised to treatment after sustaining an acute infarction. The ultimate aim of thrombolysis is not to secure coronary artery patency-no patient will be grateful to his doctor for improvement in angiographic
appearances-but to improve post-infarction morbidity and especially mortality. The difficulty here is that the mortality rate for myocardial infarction is so low that very large numbers are needed to achieve the necessary statistical power. Trials of streptokinase have recruited 118061 and 17 1872 patients while 13 318 were mustered to test alteplase The AIMS study of anistreplase was exceptionally small by these standards since it was halted after 1004 patients had been recruited.3All these studies assessed the benefits of thrombolytic therapy by comparison with a control group who did not receive such treatment. Since each agent was shown to be beneficial, the differential benefit achieved when different agents are compared would be smaller and so even larger numbers would be required to achieve satisfactory statistical power. Apart from the cost and difficulty of setting up such massive trials, the time taken to achieve a definitive result is correspondingly lengthened. What should one do? One possibility is to pool the results of existing mortality trials. The Glasgow workers do this and
conclude that there is no difference in outcome between streptokinase and anistreplase. Such analyses are of limited value: selection of patients for clinical trials in myocardial infarction is notoriously variable, as the mortality rates in control groups show, and spurious differences can be created and real differences overlooked when trials are pooled. Another approach is to use a surrogate measure of success which will yield enough end-points for adequate analysis. Coronary artery patency provides a ready surrogate for coronary mortality but raises scientific and ethical issues that are inextricably linked. The scientific issue is simple since there is nothing intrinsically unacceptable in the use of surrogate measures in clinical trials. A veritable industry has been built on clinical trials devoted to one surrogate measure for cardiovascular disease-blood pressure. Evidence of efficacy of two widely used classes of drugs in prevention of cardiovascular disease-angiotensin converting enzyme inhibitors and calcium antagonists-is based entirely on trials in which blood pressure reduction was accepted as a surrogate measure of cardiovascular events. The criteria for acceptability are pragmatic: how far is the use of a surrogate measure necessary and how appropriate is the particular measure selected? The first question has already been answered since adequate mortality data will not be available for a considerable time. The second question is more difficult. Under normal circumstances prognosis is clearly better with a patent coronary artery than with an occluded artery, but this assumption cannot be made about a coronary artery that is patent as a result of thrombolysis. However, the trial datal-4 clearly indicate persistence of benefit after discharge of patients from hospital, and there is nothing to suggest that coronary artery patency will carry a different subsequent prognosis as a result of different pharmacological properties of the agents used. In GISSI-2 (streptokinase versus alteplase) the pursuit of relevant end-points has taken a different direction. On p 289 the Italian trial organisers justify their use of a combined index-mortality plus survival to discharge with poor left-ventricular function. What about the ethical issue? The method of demonstrating coronary artery patency used by Dr
262
Hogg and colleagues was invasive, necessitating two coronary angiograms within 24 hours. That coronary angiography has been used extensively in the management of acute myocardial infarction as a prelude to angioplasty or surgery does not automatically justify its use in this study because the procedure did not lead to any change in treatment of individual patients. Moreover, the most common clinical complication-bleeding from the catheter site-was directly attributable to coronary angiography. Ethical acceptability of this trial requires that additional serious risks are very slight and that its scientific value is commensurate with the extra discomfort and risks involved. We have previously expressed disquiet about experiments at the Hammersmith Hospital in which multiple for a physiological investigation was used angiography of the action of vasoactive peptides in the coronary tree.6 By contrast, the clinical value of the Glasgow trial seems undeniable, but would of course not justify exposing patients to a significant risk. Thus acceptability stands or falls by the definition of acceptable risk in this case. In a fully equipped cardiac unit with supporting facilities, the risks of serious complications are low6 and would seem justifiable in the Glasgow study. These results are a valuable step in the evaluation of the main therapeutic advances of recent years; open discussion of the issues involved is nevertheless essential to preserve standards of scientific clinical investigation. Italiano Per Lo Studio Della Streptochinasi Nell’Infarto Miocardico (GISSI). Effectiveness of intravenous thrombolytic treatment in acute myocardial infarction. Lancet 1986; i: 397-401. 2. ISIS-2 Collaborative Group. Randomised trial of intravenous streptokinase, oral aspirin, both, or neither among 17 187 cases of suspected acute myocardial infarction: ISIS-2. Lancet 1988; ii: 349-60. 3. AIMS Trial Study Group. Effect of intravenous APSAC on mortality after acute myocardial infarction: preliminary report of a placebocontrolled clinical trial. Lancet 1988; i: 545-49. 4. Wilcox RG, Olsson CG, Skene AM, et al. Trial of tissue plasminogen activator for mortality reduction in acute myocardial infarction. Anglo-Scandinavian Study of Early Thrombolysis (ASSET). Lancet 1988; ii: 525-30. 5. Editorial. Peptides into the coronary arteries. Lancet 1989; ii: 958-59. 6. Cumberland DC. Coronary angiography and angioplasty. Medicine Int 1989; 68: 2834-36.
1. Gruppo
Are you
taking the medicine?
Thomas Willisl observed in the 17th century that severe asthma could be life-threatening, the low case-fatality rate from the disorder among patients of an individual practitioner subsequently encouraged the view of Osler two hundred years later that asthma is seldom fatal. New Zealand has lately experienced a second "epidemic" of asthma deaths,33 equal in magnitude to that of the 1960s, while the death rate in the UK has probably been rising steadily over the past ten years, especially among males aged 15-34 years.4 The increase in asthma mortality has occurred against a backdrop of increased prescribing of anti-asthma therapy. There are many reasons why
Although
influenced case-fatality rates, most notably poor assessment of the severity of asthma by patients and doctors,6 and a reluctance to prescribe adequate therapy, especially steroids.7 Another important factor is that patients often do not adhere to the prescribed drug regimen. Studies of patient compliance are remarkably consistent in revealing full compliance rates of only 30-50% (in whatever condition is studied8), and patients with asthma are no different, despite the symptomatic nature of their disease.9-11 treatment
has
not
Compliance may be best defined as the extent to which a person’s behaviour (whether in terms of taking drugs, following diets, or executing changes in personal lifestyle) coincides with medical advice. Such definition assumes that the recommendations of the patients’ medical advisors are always correct and should therefore be followed. Full and complete compliance is the state in which the patient takes all the prescribed drug for the entire period intended by the prescribing physician. The term partial compliance implies that the patient has taken only some of the total intended number of doses, either taking the drug fewer times each day, though for the correct number of days, or possibly the full dose but for a shorter period than prescribed. The biological effects of taking all doses for only the first half of a course may, of course, be profoundly different from those of taking only half the prescribed number of tablets every day. Asthmatic patients unfortunately may require both long-term treatment taken 2-4 times a day and short courses of oral drugs such as steroids. Systematic study of the extent and determinants of patients’ compliance and non-compliance with antiasthma therapy has been hampered by the difficulties of assessment. Simply asking patients whether they are taking their drugs is easy and universally applicable but is notoriously unreliable, overestimating true compliance by roughly 30%.12 Assessment by weighing returned pressurised aerosol cannisters is confounded both by the widespread practice of conducting "test-firings" before actual inhalation of drug and by the concurrent use of multiple different inhalers of the same drug (especially of rescue beta agonists). Counting returned dry powder capsules may merely drive poorly compliant patients away from follow-up, increasing noncompliance further. Mechanical devices have been devised to monitor use of pressurised aerosols automatically (eg, the ’Nebuliser Chronolog’ manufactured by ATP, Denver, USA13), but until they are an integral part of the inhaler the influence of the device itself on compliance will remain uncertain. Reiser and Warner14 argued that the mere inclusion of patients in a prospective study may lead to a
unrepresentatively high compliance rates simply because these patients receive more attention. However, there are no objective prospective studies of the influence of such trials on compliance for good or bad. Only unannounced assay of drugs in biological
