Personalized Medicine vs Guideline-Based Medicine FREE ONLINE FIRST
Published online May 27, 2013
Two philosophical approaches to the implementation of optimal health care are emerging—the use of evidence-based guidelines and the application of personalized (or “precision”) medicine. Even though both approaches have important merits, they both also can present conflicting priorities that must be reconciled before they can be best leveraged.
Evidence-based guidelines are generated based on the body of clinical data available for a particular question. The highest level of evidence assigned in a guideline is based on multiple randomized controlled clinical trials. In general, randomized clinical trials have specific inclusion and exclusion criteria designed to represent a population broad enough and sufficiently enriched to attain a requisite number of end points and demonstrate a statistically and clinically significant difference in outcome. Subgroup analyses (both those that are prespecified and other post hoc analyses) are often performed to identify characteristics within the study population that are associated with greater benefit from the intervention, with no benefit, or even with harm. Yet these analyses are accompanied by warnings that findings should be cautiously interpreted.1
Indeed, there is well-deserved skepticism regarding the utility and accuracy of subgroup analysis from clinical trials, and these analyses are therefore generally not used in the formulation of guidelines. Patients (including those enrolled in trials) have multiple characteristics, each of which may influence the behavior and significance of other characteristics. Analysis of a subgroup showing that a single characteristic influences outcome is of limited clinical significance unless multiple variables that may modify the importance of the single variable are considered. However, if well-conducted analyses from multiple sources demonstrate concordant findings, perhaps these subgroup analyses should be considered when guidelines are constructed and revised, given the impracticality of performing randomized clinical trials to answer the question of appropriateness for every possible subgroup.
In the end, “the guidelines” are usually established based on the inclusion criteria for the trial. The applicability of the guidelines may be questioned, or even suspect, when individual patients within the heterogeneous population to which the guidelines are applied in clinical practice differ in certain critical characteristics from those of the trial population on which the guideline recommendation is based. That is, the generalizability of trial results to clinical practice may be compromised by a number of factors involved in execution of the trial, such as where patients were recruited (eg, inpatient vs outpatient venue, tertiary referral centers vs primary care centers).
The President's Council of Advisors on Science and Technology noted that personalized medicine “refers to the tailoring of medical treatment to the individual characteristics of each patient. It does not literally mean the creation of drugs or medical devices that are unique to a patient, but rather the ability to classify individuals into subpopulations that differ in their susceptibility to a particular disease or their response to a specific treatment. Preventive or therapeutic interventions can then be concentrated on those who will benefit, sparing expense and side effects for those who will not.”2 Although the increasing attention directed to personalized medicine has largely focused on the interaction of an individual's genome with specific treatments, any individual characteristics that affect treatment outcomes may be relevant to clinical decision making. If certain subpopulations within the total cohort of a clinical trial were considered unlikely to benefit from the intervention—based, for example, on actual subgroup analysis—this hypothesis would need to be tested prospectively in a separate clinical trial to achieve a sufficient level of evidence about the value of the intervention in this patient subpopulation.
Although it is possible to make a case for equipoise in such a situation, once the guidelines include this subpopulation in the general group in which the intervention is recommended based on the results (or entry criteria) of clinical trials, it is difficult to overcome the multidimensional resistance to actually testing “not providing the intervention” when the guidelines recommend otherwise. Thus, the development of evidence-based guidelines based on a relatively broad set of enrollment criteria inhibits the subsequent development of personalized medicine within this “enrollment criteria” space.
Several examples highlight this issue. For instance, the MADIT-II (Multicenter Automatic Defibrillator Implantation Trial II) randomized clinical trial3 demonstrated the efficacy of implantable cardioverter-defibrillator (ICD) therapy in patients with prior myocardial infarction and left ventricular ejection fraction of 30% or less. The improved survival reported among the patients randomized to receive ICD therapy in the MADIT-II study, as well as in other supporting trials, formed the basis for a class IA recommendation by the 2006 American College of Cardiology/American Heart Association/Heart Rhythm Society guidelines for implanting ICDs in this patient population.
As would be expected, a wide range of comorbidities was present in various degrees in the MADIT-II population, including diabetes, hypertension, heart failure, atrial fibrillation, and renal dysfunction. In retrospective analyses of MADIT-II data, 5 clinical factors were found to be associated with survival. Patients with 3 or more of these risk factors, accounting for 19% of the total population in the trial, did not derive any survival benefit from the ICD.4 Nevertheless, ICD implantation for patients who meet the enrollment criteria and have 3 or more of these factors is still recommended by the guidelines.5 Based on these post hoc analyses, it would be reasonable to hypothesize that up to 1 of every 5 ICDs may be implanted in patients who derive no benefit. However, the medicolegal and regulatory environment, as well as the limited financial resources for studies of this kind, present major barriers6 to testing whether it is possible that better, more individualized care for these patients might not include an ICD (ie, if there is no survival benefit, but there is a risk of complications or inappropriate shocks).
The MADIT-II trial excluded patients who had experienced a myocardial infarction within the last month. In another post hoc subgroup analysis,7 all the benefit of ICD therapy was noted in patients whose myocardial infarction occurred 18 months or more prior to enrollment. Furthermore, the DINAMIT (Defibrillator in Acute Myocardial Infarction Trial)8 and IRIS (Immediate Risk Stratification Improves Survival)9 trials enrolled patients within 30 to 40 days of their myocardial infarction and found no benefit of ICD therapy over several years of follow-up. These and other supporting data10 provide a strong basis to hypothesize that ICD implantation even up to 1 to 2 years after myocardial infarction may provide no benefit. Given the small but present risks of ICD therapy (acute implant complications, device infection, inappropriate shocks), the goal of personalized medicine should be to determine whether this therapy is warranted even in the first 1 to 2 years after myocardial infarction.
Further tension between personalized medicine and guideline-based medicine could occur in the setting of well-intentioned programs that grade or report compliance rates with guideline-based treatments. This can be particularly problematic when the guidelines encompass recommendations for which evidence is extrapolated from clinical trials. For example, the MADIT-II trial excluded patients receiving dialysis. The benefit of ICD implantation in this population is unclear, but the guidelines do not address this subgroup. In systems that monitor compliance with guidelines, should the decision not to implant an ICD in a patient receiving dialysis but with prior myocardial infarction and left ventricular ejection fraction of 25% reflect a “deviation from the guidelines”?
The conflict between guideline-based medicine and personalized medicine predominantly occurs when considering withholding a therapy that is recommended or supported by the guidelines but that may not be beneficial for an individual patient. If a subpopulation may not benefit from the therapy, it is important to identify the subpopulation and verify this finding in an appropriate clinical trial. This presents a genuine opportunity to deliver better health care at lower costs by withholding the intervention. Cultivating a health care culture poised to explore these opportunities is critical. This will entail more active participation from a range of stakeholders, including physicians who will need to embrace equipoise when the data support this position; insurers (including the Centers for Medicare & Medicaid Services) who have traditionally not been involved in the design, funding, and conduct of clinical trials; regulators who will need to develop policies to enable and support this type of patient-centered research; and health care organizations and quality-measurement groups who will need to develop more nuanced approaches to assessing quality of care and processes that monitor guideline implementation.
Published Online: May 27, 2013. doi:10.1001/jama.2013.6629
Conflict of Interest Disclosures: The authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Dr Goldberger reported serving as a consultant to Red Bull and providing expert testimony. He directs the Path to Improved Risk Stratification, NFP which has received unrestricted educational grants from Boston Scientific, Medtronic, and St Jude, which have also provided grants to his institution and speaker honoraria. He serves on the DSMB for a clinical trial sponsored by Zoll. Dr Buxton reported serving on the events committee for clinical trials supported by Medtronic, Boston Scientific, and St Jude, receiving support for travel from Medtronic, and receiving royalties for editing a textbook from Blackwell Publishing. Dr Buxton also reported that his institution has received grants from Medtronic, Biosense-Webster, and Boston Scientific.
President's Council of Advisors on Science and Technology. Priorities for Personalized Medicine. White House website. http://www.whitehouse.gov/files/documents/ostp/PCAST/pcast_report_v2.pdf
Epstein AE, DiMarco JP, Ellenbogen KA, et al. ACC/AHA/HRS 2008 Guidelines for device-based therapy of cardiac rhythm abnormalities: a report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines (Writing Committee to Revise the ACC/AHA/NASPE 2002 Guideline Update for Implantation of Cardiac Pacemakers and Antiarrhythmia Devices) [published correction appears in Circulation. 2009;120(5):e34-e35]. Circulation. 2008;117(21):e350-e408
PubMed | Link to Article
PubMed | Link to Article