User login
Our knowledge base in cardiovascular diseases has progressed in the last 20 years largely on the strength of randomized clinical trials. We pay special attention to "megatrials" because of their statistical heft, and often apply the lessons learned to the patients we treat.
But critics abound, and with good reason. A personal favorite: "Randomised controlled trials are primarily about medical interventions and not patients," that is, randomization permits comparisons of interventions not confounded by the individuality of patients." The paradox of the clinical trial is that it is the best way to assess whether an intervention works, but is arguably the worst way to assess who will benefit from it (Lancet 1999;353:743-6).
So therein lies the core of the debate about RCTs: How representative are they? If simvastatin worked in the 4S Study (Lancet 1994;344:1383-9), would the drug work in American, non-Scandinavian patients? If implantable cardioverter defibrillators reduced mortality in a primary prevention cohort with a mean age in the lower 60s, would they work in octogenarians (Arch. Intern. Med. 2012: 2012;172:67-8)? A combination of hydralazine and nitrates worked in black patients with heart failure treated in A-HeFT (N. Engl. J. Med. 2004;351:2049-57), but do the findings apply to Caucasians?
Krumholz and colleagues pointed out 10 years ago that women and the elderly were underrepresented in clinical trials of heart failure interventions (Arch. Intern. Med. 2002; 1682-8), andthese findings have remained germane (Arch. Intern. Med. 2011;171:550-6). Recently, Dhruva and Redberg proposed with ample justification (JAMA 2012;307:1145-6) that device trials increase representation of women and that the Food and Drug Administration should be focused on enforcing this directive.
The problem is that there are limits to how completely we can "cover the waterfront," since patient factors that likely influence response (ethnicity, race, sex and other related inherent genetic, biological and social parameters) can be so varied. Shall we set a minimum enrollment for elderly Asian females without diabetes? What about young Hispanic males who smoke more than one pack of cigarettes a day? And how do we handle in-group differences? To some degree, this line of questioning is pure nonsense. We have learned something from the science of clinical trials; namely, subgroups should be prespecified, and stratification at time of randomization is necessary. However, these requirements represent an impossible statistical barrier if multiple potential subgroups are determined to be of interest.
There is no easy answer to the proposal of Dhruva and Redberg. To simply state "Registry" or "Pharmacogenomics" is not sufficient. On the other hand, the potential "Balkanization" of clinical trials in which each defined subgroup is given its own trial is not tenable. Selection of one favored subgroup for greater representation will not solve the problem either. The thorny representation issues of randomized trials must be addressed. The future of RCTs lies in the balance.
Our knowledge base in cardiovascular diseases has progressed in the last 20 years largely on the strength of randomized clinical trials. We pay special attention to "megatrials" because of their statistical heft, and often apply the lessons learned to the patients we treat.
But critics abound, and with good reason. A personal favorite: "Randomised controlled trials are primarily about medical interventions and not patients," that is, randomization permits comparisons of interventions not confounded by the individuality of patients." The paradox of the clinical trial is that it is the best way to assess whether an intervention works, but is arguably the worst way to assess who will benefit from it (Lancet 1999;353:743-6).
So therein lies the core of the debate about RCTs: How representative are they? If simvastatin worked in the 4S Study (Lancet 1994;344:1383-9), would the drug work in American, non-Scandinavian patients? If implantable cardioverter defibrillators reduced mortality in a primary prevention cohort with a mean age in the lower 60s, would they work in octogenarians (Arch. Intern. Med. 2012: 2012;172:67-8)? A combination of hydralazine and nitrates worked in black patients with heart failure treated in A-HeFT (N. Engl. J. Med. 2004;351:2049-57), but do the findings apply to Caucasians?
Krumholz and colleagues pointed out 10 years ago that women and the elderly were underrepresented in clinical trials of heart failure interventions (Arch. Intern. Med. 2002; 1682-8), andthese findings have remained germane (Arch. Intern. Med. 2011;171:550-6). Recently, Dhruva and Redberg proposed with ample justification (JAMA 2012;307:1145-6) that device trials increase representation of women and that the Food and Drug Administration should be focused on enforcing this directive.
The problem is that there are limits to how completely we can "cover the waterfront," since patient factors that likely influence response (ethnicity, race, sex and other related inherent genetic, biological and social parameters) can be so varied. Shall we set a minimum enrollment for elderly Asian females without diabetes? What about young Hispanic males who smoke more than one pack of cigarettes a day? And how do we handle in-group differences? To some degree, this line of questioning is pure nonsense. We have learned something from the science of clinical trials; namely, subgroups should be prespecified, and stratification at time of randomization is necessary. However, these requirements represent an impossible statistical barrier if multiple potential subgroups are determined to be of interest.
There is no easy answer to the proposal of Dhruva and Redberg. To simply state "Registry" or "Pharmacogenomics" is not sufficient. On the other hand, the potential "Balkanization" of clinical trials in which each defined subgroup is given its own trial is not tenable. Selection of one favored subgroup for greater representation will not solve the problem either. The thorny representation issues of randomized trials must be addressed. The future of RCTs lies in the balance.
Our knowledge base in cardiovascular diseases has progressed in the last 20 years largely on the strength of randomized clinical trials. We pay special attention to "megatrials" because of their statistical heft, and often apply the lessons learned to the patients we treat.
But critics abound, and with good reason. A personal favorite: "Randomised controlled trials are primarily about medical interventions and not patients," that is, randomization permits comparisons of interventions not confounded by the individuality of patients." The paradox of the clinical trial is that it is the best way to assess whether an intervention works, but is arguably the worst way to assess who will benefit from it (Lancet 1999;353:743-6).
So therein lies the core of the debate about RCTs: How representative are they? If simvastatin worked in the 4S Study (Lancet 1994;344:1383-9), would the drug work in American, non-Scandinavian patients? If implantable cardioverter defibrillators reduced mortality in a primary prevention cohort with a mean age in the lower 60s, would they work in octogenarians (Arch. Intern. Med. 2012: 2012;172:67-8)? A combination of hydralazine and nitrates worked in black patients with heart failure treated in A-HeFT (N. Engl. J. Med. 2004;351:2049-57), but do the findings apply to Caucasians?
Krumholz and colleagues pointed out 10 years ago that women and the elderly were underrepresented in clinical trials of heart failure interventions (Arch. Intern. Med. 2002; 1682-8), andthese findings have remained germane (Arch. Intern. Med. 2011;171:550-6). Recently, Dhruva and Redberg proposed with ample justification (JAMA 2012;307:1145-6) that device trials increase representation of women and that the Food and Drug Administration should be focused on enforcing this directive.
The problem is that there are limits to how completely we can "cover the waterfront," since patient factors that likely influence response (ethnicity, race, sex and other related inherent genetic, biological and social parameters) can be so varied. Shall we set a minimum enrollment for elderly Asian females without diabetes? What about young Hispanic males who smoke more than one pack of cigarettes a day? And how do we handle in-group differences? To some degree, this line of questioning is pure nonsense. We have learned something from the science of clinical trials; namely, subgroups should be prespecified, and stratification at time of randomization is necessary. However, these requirements represent an impossible statistical barrier if multiple potential subgroups are determined to be of interest.
There is no easy answer to the proposal of Dhruva and Redberg. To simply state "Registry" or "Pharmacogenomics" is not sufficient. On the other hand, the potential "Balkanization" of clinical trials in which each defined subgroup is given its own trial is not tenable. Selection of one favored subgroup for greater representation will not solve the problem either. The thorny representation issues of randomized trials must be addressed. The future of RCTs lies in the balance.