In the past, many standard and accepted practices for clinical problems were simpler and more straightforward than those that today's clinicians face—and these practices seem to have worked, despite the paucity of good research evidence. Physicians simply made subjective, intuitive decisions about what worked based on what they observed. The problem today is that the growing complexity of medicine bombards clinicians with a chaotic array of clinical choices, ambiguities and uncertainties that exceeds the inherent limitations of the unaided human mind. As a result, many of today's standard clinical practices bear no relation to any evidence of effectiveness.
Instead, physicians frequently base their decisions on shortcuts, such as the actions of the average practitioner ("if everyone is doing it, the intervention must be appropriate"); the commonness of the disease ("if the disease is common, we have no choice but to use whatever treatment is available"); the seriousness of the outcome ("if the outcome without treatment is very bad, we have to assume the treatment will work"); the need to do something ("this intervention is all we have"); and the novelty or technical appeal of the intervention ("if the machine takes a pretty picture, it must have some use").
Drug prescribing is another blatant example of medical practice that is often evidence-free. Drugs that are known to be effective may work well for only 60 percent of people who take them. But about 21 percent of drug prescriptions in the United States are for "off-label" use, that is, to treat conditions for which they have not been approved by the U.S. Food and Drug Administration. That's more than 150 million prescriptions per year. Off-label use is most common among cardiac medications (46 percent) and anticonvulsants (46 percent). Here's the real punch line: in 73 percent of the cases where drugs are used in unapproved ways, there is little or no evidence that they work. Physicians prescribe drugs well over a million times a year with little or no scientific support.
These are fighting words, saying that such a big chunk of medical practice is not based on science. To illustrate just how provocative this topic is, look at what happened in the 1990s when the Federal Agency for Health Care Policy and Research (now the Agency for Healthcare Research and Quality) released findings from a five-year investigation of the effectiveness of various treatments for low back pain—one of the leading reasons that Americans see physicians.
Between 1989 and 1994, an interdisciplinary Back Pain Patient Outcomes Assessment Team (BOAT) at the University of Washington Medical School in Seattle set out to determine what treatment strategies work best and for whom. Led by back expert Richard A. Deyo, MD, MPH, the team included orthopedic surgeons, primary-care physicians, physical therapists, epidemiologists and economists. Together, they examined the relative value of various diagnostic tests and surgical procedures.
They conducted a comprehensive review of clinical literature on back pain. They exhaustively examined variations in the rates at which different procedures were being used to diagnose and treat back pain. Their chief finding was deeply disturbing: what physicians thought worked well for treating low back pain doesn't. The implication was that a great many standard interventions for low back pain may not be justified. And that was immensely threatening to physicians, especially surgeons who perform back operations for a living.
Among the researchers' specific findings: no evidence shows that spinal-fusion surgery is superior to other surgical procedures for common spine problems, and such surgery leads to more complications, longer hospital stays and higher hospital charges than other types of back surgery.
Disgruntled orthopedic surgeons and neurosurgeons reacted vigorously to the researchers' conclusion that not enough scientific evidence exists to support commonly performed back operations. The surgeons joined with Congressional critics of the Clinton health plan to attack federal funding for such research and for the agency that sponsored it. Consequently, the Agency for Healthcare Policy and Research had its budget for evaluative research slashed drastically.
The back panel's guidelines were published in 1994. Since then, even though there are still no rigorous, independently funded clinical trials showing that back surgery is superior to less invasive treatments, surgeons continue to perform a great many spinal fusions. The number increased from about100,000 in 1997 to 303,000 in 2006.
What are physicians to do? They need a great deal more reliable information than they have, especially when offering patients life-changing treatment options. Before recommending surgery or radiation treatment for prostate cancer, for example, physicians and their patients must compare the benefits, harms and costs of the two treatments and decide which is the more desirable.
One treatment might deliver a higher probability of survival but also have bad side effects and high costs, while the alternative treatment might deliver a lower probability of survival but have no side effects and lower costs. Without valid scientific evidence about those factors, the patient may receive unnecessary and ineffective care, or fail to receive effective care, because neither he nor his physician can reliably weigh the benefits, potential harm and costs of the decision.
Recognizing that the quality and reliability of clinical-research information vary greatly, entities like the U.S. Preventive Services Task Force (USPSTF) have devised rating systems to rank the strength of available evidence for certain treatments. The strongest evidence is the scarcest and comes from systematic review of studies (randomized, controlled trials) that are rigorously designed to factor out biases and extraneous influences on results. Weaker evidence comes from less rigorously designed studies that may let bias creep into the results (for example, trials without randomization or cohort or case-control analytic studies). The weakest evidence comes from anecdotal case reports or expert opinion that is not grounded in careful testing.
Raymond Gibbons, MD, a professor of medicine at the Mayo Clinic and past president of the American Heart Association, puts it well: "In simple terms, Class I recommendations are the 'do's'; Class III recommendations are the 'don'ts'; and Class II recommendations are the 'maybes.'" The point is this: even physicians who follow guidelines must deal with scientific uncertainty. There are a lot more "maybes" than "do's."
Even the "do's" require value judgments, and it is important to be clear about what evidence-based practice guidelines can and cannot do, regardless of the strength of their scientific evidence. Guidelines are not rigid mandates or "cookie-cutter" recommendations that tell physicians what to do. They are intended to be flexible tools to help physicians and their patients make informed decisions about their care.
Even guidelines that are rooted in randomized, controlled trial research do not make clinical decisions for physicians; rather, they must be applied to individual patients and clinical situations based on value judgments, both by physicians and their patients. Clinical decision-making must entail value judgments about the costs and benefits of available treatments. What strong guidelines do is to change the anchor point for the decision from beliefs about what works to evidence of what works. Actual value-based treatment decisions are a necessary second step.
For example, should a physician recommend an implantable cardioverter-defibrillator (ICD) to his or her patient when a randomized-control trial shows that it works? The device is a small, battery-powered electrical-impulse generator implanted in patients at risk of sudden cardiac death due to ventricular fibrillation (uncoordinated contraction of heart chamber muscle) and ventricular tachycardia (fast heart rhythm). A published randomized trial compared ICDs to management with drugs for heart-attack patients and found that ICDs reduced patients' probability of death at 20 months by about one-third.
Armed with such a guideline, the physician and patient must still make a value judgment: whether the estimated decrease in chance of death is worth the uncertainty, risk and cost of the procedure. The ultimate decision is not in the guideline, but it is better informed than a decision made without the evidence to help guide it. The guideline has lessened uncertainty but not removed it.