Author + information
- Received February 24, 2006
- Revision received March 17, 2006
- Accepted March 21, 2006
- Published online August 1, 2006.
- ↵⁎Reprint requests and correspondence:
Dr. Christopher B. Granger, Duke Clinical Research Institute, Duke University Medical Center, 2400 Pratt Street, Room 0311 Terrace Level, Durham, North Carolina 27705.
With an increasing burden of cardiovascular disease and many promising novel treatments in development, the need for efficient systems to evaluate treatments has never been greater. To understand whether a treatment should be used in practice, we need to know whether it makes patients live longer, feel better, prevents adverse events, or does these things with better tolerability or lower cost. But therapeutic development is expensive, inefficient, and is generally focused on short-term treatment effects, rather than on prevention and on long-term impact. Could measures of disease progression, combined with trends on clinical outcomes and post-marketing surveillance to assess safety, serve as the foundation for therapeutic development? Experience and principles of clinical research tell us no. Especially in the field of heart failure, numerous treatments have appeared promising based on disease markers, yet caused harm when tested in studies that assessed clinical outcomes. The intersection of complex human disease, intended and unintended targets of therapy, and overall risk and benefit make it impossible to accurately predict the effect on clinical outcomes based on impact on a disease marker. While reliable measures of disease progression are important to guide which treatments to study in trials, clinical outcome trials must remain the basis for informing clinicians on which treatments improve clinical outcomes. Improved reliability and capacity require the development of more efficient clinical trial methods, streamlined regulatory processes, rational use of privacy protection, leveraging of electronic medical records, and recruitment of a larger proportion of the clinical community to participate in clinical trials.
Square in your ship’s path are Sirens, crying beauty to bewitch men coasting by; woe to the innocent who hears that sound!Homer (1)
There has never been a greater need to enhance our ability to determine which current and new therapies are beneficial. Cardiovascular disease will grow as the number one cause of death and disability worldwide (2), in part related to the aging of the population. At the same time, especially in the U.S., the growth of health care spending is creating intense pressure to assure that additional expenditures are worthwhile and that quality care is being delivered. Key components of quality involve the consistent and efficient application of treatments that are proven to improve clinical outcomes in a safe manner (3). The medical establishment, including the pharmaceutical industry and the Food and Drug Administration (FDA), is under increased scrutiny to foster the development of drugs and devices in a manner that provides information about safety and effectiveness in general practice. Commonly used treatments like hormone replacement therapy and cyclooxygenase-2 inhibitors have served to highlight the hazards of wide use of drugs for which safety and/or efficacy is not well established in large trials that include broad patient populations and long-term follow-up.
To decide whether a treatment is worthwhile to use, clinicians, patients, and payers need to understand the impact of treatment based on multiple factors—including whether the new treatment makes patients live longer (for fatal diseases), feel better (for symptomatic diseases), prevent non-fatal clinical events (including need for hospitalization), or do these things as well as alternatives but with better tolerability or lower cost (4).
The FDA has a more focused responsibility when deciding whether to approve a drug. It must determine whether a drug is safe and effective. Effectiveness can be defined by effects on clinical outcomes, or in some cases by biologic markers, including blood pressure or lipid lowering. Even these markers, however, may only partially predict the impact of treatments on clinical outcome. Post-marketing event reporting and surveillance occasionally provide informative safety data.
In cardiovascular disease, thanks to a tradition led by clinical trials groups and countless collaborating investigators, we are fortunate to have a number of medical (5–18) (Table 1)and device treatments that have been proven to reduce mortality, most of which also improve quality of life. This list forms the foundation of indicators driving quality of care.
In a Viewpoint in this issue of the Journal, Cohn (19) has outlined limitations in our current capacity to develop therapies. We are limited in our ability to evaluate the effects of therapies on clinical outcomes early in a disease process, when the risk of events is low, because we cannot predict long-term outcomes from short-term studies and because it is difficult to conduct very large and long-term trials efficiently and cost-effectively. Moreover, trials performed in patients with most advanced disease may not be reliable guides to the balance of benefit and risk in less severe disease, whereas trials performed in very low-risk populations (that may also be at low risk for adverse effects) may be insufficiently powered to detect important treatment effects, from a public health perspective, that could be detected in higher-risk cohorts.
To address these deficiencies, Cohn (19) argues that we should be using measures of “disease progression,” when valid, as important determinants of treatment effect, in the context of effects on quality of life and of trends on effects on hard clinical outcomes. Thus, shorter-term, smaller trials could more rapidly bring promising drugs to market, and post-marketing surveillance could be used to further assess safety. While appealing and critically important to guide decisions on which treatments are worthy of study in large randomized trials, using measures of disease progression as important information to inform treatment use has substantial limitations and must be vigorously resisted.
It is ironic that Cohn (19) is suggesting that disease markers can reliably predict the impact on important clinical outcomes, because drug development for heart failure (HF), to which he has made major contributions, is rife with examples of failures of such an approach (20–30) (Table 2).In fact, beta-blockers, not long ago, were believed to be contraindicated because of a negative impact on myocardial function, while angiotensin-converting enzyme inhibitors were feared to trigger atherosclerotic events in HF due to impaired perfusion from blood pressure lowering. Plasma norepinephrine level, strongly related to outcome, was believed to be a reliable guide to impact of treatment for HF on clinical outcomes (31,32), until moxonidine, one of the most potent treatments to reduce norepinephrine, was found to paradoxically increase mortality by 50% (27).
Why is using effect on measures of disease progression inadequate, at least for the foreseeable future, as a method to determine the clinical impact of treatments?
1. The complexity of human disease does not enable any marker to be reliable enough. For most complex diseases, we simply do not understand the disease well enough to use markers of disease to predict clinical effect. For many lifesaving treatments, including aspirin, we have little understanding of the exact mechanism of benefit. Inhibition of tumor necrosis factor-alpha had supportive pre-clinical data and measurable impact on left ventricular (LV) size and function (30), yet in trials, the net effect appeared detrimental (29). Recently, calcium and vitamin D were shown to significantly improve disease progression measured by calcium content in hip bones, yet when the Women’s Health Initiative trial was completed, they had no significant effect on the clinical outcome of fracture (33). Thus, using markers poses 2 hazards: treatments that affect markers may not do so in a way that improves outcomes, and treatments that improve outcomes may not affect the markers (34).
2. Unintended targets are common. It is naive, in particular, given multiple examples to the contrary, to believe that we can predict the overall impact of a treatment based on the effect on a single measure of a disease. Even when there are measures that accurately reflect the impact of a treatment on disease progression—for example, preservation of ejection fraction after fibrinolytic therapy—these measures cannot guide treatment decisions without being evaluated in the context of other effects, such as the increase in intracranial hemorrhage. Hormone replacement therapy has an important beneficial impact on low-density lipoprotein and high-density lipoprotein cholesterol, yet increases short-term risk of myocardial infarction through poorly defined mechanisms (35). Quite simply, focus on a mechanism does not allow the essential evaluation of the balance between benefit and risk.
3. Assessing disease progression in fatal disease is confounded, including by “survival bias.” Because a basic condition of validity of the randomized clinical trial depends on assessing (nearly) all randomized patients to assure balance of unmeasured confounders, use of an outcome that depends on acquisition and survival presents a methodologic challenge. For example, use of improvement in LV ejection fraction as a measure of benefit from fibrinolytic therapy was limited by the fact that more sick patients in the control group died, so that surviving patients with lower ejection fraction in the fibrinolytic therapy group masked the beneficial effect (36).
4. Integration of safety, efficacy, and cost-effectiveness is not possible based on assessment of markers of disease progression. More and more, we need to know not only if a treatment is effective, but how effective and the net magnitude of effect given the cost. Thus, not only are large trials needed with adequate numbers of clinical end points to assess risk and benefit in an integrated fashion, but the trials should be done in populations that represent where the treatment is intended to be used.
If studying the impact of treatments on disease progression will not provide the solution to the challenges Cohn (19) outlines, what will? It is true that exclusive focus on all-cause mortality will prevent insights that can help refine effective treatments by understanding the balance of beneficial and harmful effects and best dosing. Because <1 in 10 drugs tested in humans will ever come to market, smarter choices based on better markers of disease progression are needed to guide which treatments to study in large (and expensive) trials. As we contemplate the era of genomic medicine, a refined ability to decide which treatments to test in which patients based on details of disease manifestation will be especially important. We agree with Cohn (19) that certain measures, including impact of HF treatments of ventricular volume, may both correlate with clinical outcomes and reflect a major pathway by which certain treatments translate into improved clinical outcomes, and such measures may be very important in helping to decide which treatments will be likely to improve survival and quality of life.
Most importantly, we need improved methods to efficiently test the impact of promising therapies on clinical outcomes. This requires the development of more efficient clinical trial methods, streamlined regulatory processes, rational use of privacy protection, leveraging of electronic medical records, recruitment of a larger proportion of the clinical community to participate in clinical trials, and integration of trials with registries to assess the use and safety of treatments in observational studies. Efficient conduct of trials in developing health care systems and focus of the National Institutes of Health on re-engineering the U.S. clinical research enterprise (37) offer important opportunities.
Ultimately, both improved ability to predict which treatments will be effective by better ability to measure disease progression and more efficient, long-term, real-world clinical trials are needed. For the time being, large, long-term clinical trials must serve as the standard for determining risk and benefit to establish which treatments improve cardiovascular health.
↵a Dr. Granger receives research funding from AstraZeneca, Novartis, Genentech, Boehringer Ingelheim, Glaxo Smith Kline, Procter and Gamble, Alexion, Sanofi Aventis and consulting and/or honoraria from AstraZeneca, Glaxo Smith Kline, Sanofi Aventis, Medicines Company.
↵b Dr. McMurray receives research funding/consulting fees or lecture honoraria from: Actelion, Amgen, Astra-Zeneca, Atherogenics, BMS-Sanofi, Cardiokinetics, GSK, J&J, Novartis, Pfizer, Solvay, and Takeda
- Abbreviations and Acronyms
- Food and Drug Administration
- heart failure
- left ventricular
- Received February 24, 2006.
- Revision received March 17, 2006.
- Accepted March 21, 2006.
- American College of Cardiology Foundation
- ↵Homer. The Oddyssey. Book 12, lines 41–44. Available at: http://deoxy.org/alephnull/sirens.htm. Accessed May 1, 2006.
- Murray C.J.L.,
- Lopez A.D.
- Committee on Quality of Health Care in America,
- Institute of Medicine
- DeMets D.L.,
- Califf R.M.
- Antithrombotic Trialists’ Collaboration
- Fibrinolytic Therapy Trialists’ (FTT) Collaborative Group
- ACE Inhibitor Myocardial Infarction Collaborative Group
- Pfeffer M.A.,
- McMurray J.J.,
- Velazquez E.J.,
- et al.,
- VALIANT-Valsartan in Acute Myocardial Infarction Trial Investigators
- Pitt B.,
- Remme W.,
- Zannad F.,
- et al.,
- Eplerenone Post-Acute Myocardial Infarction Heart Failure Efficacy and Survival Study Investigators
- The Heart Outcomes Prevention Evaluation Study Investigators
- Young J.B.,
- Dunlap M.E.,
- Pfeffer M.A.,
- et al.,
- Candesartan in Heart failure Assessment of Reduction in Mortality and morbidity (CHARM) Investigators and Committees
- Cohn J.N.
- Packer M.,
- Rouleau J.,
- Swedberg K.,
- et al.
- Massie B.M.,
- Berk M.R.,
- Brozena S.C.,
- et al.
- Swedberg K.,
- Bristow M.R.,
- Cohn J.N.,
- et al.,
- Moxonidine Safety and Efficacy (MOXSE) Investigators
- Mann D.L.,
- McMurray J.J.,
- Packer M.,
- et al.
- Bozkurt B.,
- Torre-Amione G.,
- Warren M.S.,
- et al.
- Swedberg K.,
- Eneroth P.,
- Kjekshus J.,
- Wilhelmsen L.,
- the CONSENSUS Study Group
- Zerhouni E.