The beauty of the nudge principle is that nobody’s choice is being taken away. In 2015, you can choose to participate in a trial, or not. In 2020, the choice would still be yours. All we propose is to change the default, and the usual patient’s response goes from No thanks to Sure, why not. This change would have to be accompanied by the creation of a robust clinical-trials enterprise able to design studies and conduct them, at low cost, with fewer barriers across diverse settings. For most clinical questions, those for which we really do not know the answer, there should be no barriers to additional trials. If a physician has a patient with a problem for which there are no treatments supported by a robust evidence base, that physician should be able to begin a trial, quickly and easily. Obviously, the busy practicing clinician cannot do this alone. The infrastructure needs to be created.
There will be reasonable opposition to using the nudge principle to increase trial enrollment. Groups that have historically been marginalized may feel that being enrolled in trials is exploitative. Although we would argue, as Thaler and Sunstein do, that the choice to not participate is still present and that the default option is merely switched, patients’ ability to opt out of trials would have to be made very clear to them so there is no hint of coercion. Great care would have to be taken to make sure the therapies being offered in a trial really are truly considered equivalent. The reality is that few of us are likely to feel strongly about the decision; most of us would be indifferent about which of two putatively equivalent treatments we would receive.
Another real concern is, Who designs these studies? We know that the pharmaceutical and device industries tend to design trials with their interests in mind. We endorse the nudge principle only for trials designed and conducted by teams who have no conflicts of interest. The nudge principle is a tool. If used correctly, it could do a lot of good; if used improperly, it could accelerate harm.
Finally, the nudge principle should be instituted alongside a broader commitment to conduct clinical trials for less money. Right now, researchers estimate that it costs $5,000 to $20,000 per patient per trial just to handle enrollment and data management. Like most costs in health care, this is absurd. Researchers know that being in a trial does not increase the cost of your medical care and that the cost of handling the data from that trial must be decreased. There are many ways of achieving this. One potential solution is the development of registry-based randomized trials. These are trials conducted within existing observational studies that would therefore decrease costs of patient recruitment. In one such study, the cost to randomize each patient fell to $50.
MEDICAL CARE IN AN ERA OF A HIGH BURDEN OF PROOF AND STREAMLINED CLINICAL TRIALS
Instituting a clear requirement of the burden of proof in medical innovation and using the nudge principle to increase enrollment in clinical trials are closely related and interdependent recommendations. Setting the evidentiary bar higher for the adoption of new therapies risks slowing down medical innovation. Reforming how we recruit people into trials would not only avoid this negative consequence but actually improve the present situation. If innovations are tested quickly, we can assure that an innovation that is adopted is really a beneficial one.
Patient care would look different in a world with a high burden of proof and a streamlined system of trial recruitment. There might actually be fewer options for treatment, but this would not be a bad thing. The options that would no longer exist would be the ones that do not work (or actually cause harm). Currently, patients who exhaust proven remedies are offered a menu of unproven treatments. In the world we envision, these patients would not be treated with guesses but would have a range of clinical trials to choose from. They would be given novel treatments, but in a setting that protects their health. The safest way to receive a new drug is in a trial with a control arm. The randomized-controlled-trial design provides a builtin safeguard—trials are stopped if the treatment turns out to be harmful (remember the CAST study from chapter 1). When patients are treated with unproven medicines outside trials, there is no safety net.
::
There are more unanswered questions in medicine than we could ever count. Each day patients and doctors must make choices. The byproduct of this situation is that many decisions turn out to be wrong and many accepted medical practices turn out to be major missteps. We need a way forward that assures that we get it right more often. We propose reforms that we believe would go a long way toward curbing reversal: medical innovations should be subject to higher regulatory and professional standards that are based on adoption of the concept of the burden of proof and a commitment to answering relevant clinical questions, facilitated through use of the nudge principle. The specific solutions we suggest should be evaluated. Clever trials can assess whether the nudge principle improves patient outcomes or whether the trials that get done are trivial and accomplish little. We believe these proposals are logical and compelling and that it is time to put them to the test.
17 HOW NOT TO BECOME A VICTIM OF REVERSAL
THE STRUCTURE AND CULTURE of the medical field makes it likely that you will be offered therapies that are unproven. Some of these therapies will eventually be reversed. In this chapter we propose a way to approach your medical care that minimizes the risk that you will receive a therapy today that will be found to be ineffective or harmful tomorrow. By taking a smart approach to your care and forming a productive alliance with your doctor, you can (mostly) assure yourself excellent, evidence-based, care. We say “mostly” because, as we have outlined, there are strong forces behind the epidemic of reversal. Financial incentives, a confidence in mechanistic science, and a medical literature filled with studies designed and run by the companies that stand to profit from the outcome—all of these promote faulty therapies. At times, even we, as physicians, have been tricked. During our careers we have written prescriptions for estrogen-replacement therapy, fenofibrate, and other interventions that were based on flawed evidence but good stories.
Your basic goal as a patient is easily stated: you need to be confident that every medical intervention you accept has been shown, in robust studies, to improve the outcome in which you are most interested. If no such treatment exists (not an uncommon scenario, even in 21st-century medicine), you should be aware that the effectiveness of any therapy you consider is based more on conjecture and hope than on science. Sometimes accepting such therapies is reasonable, but that acceptance should be a highly informed decision.
There is much working against you, coming from both sides of the doctor-patient relationship, in achieving this goal. From the patient side of the equation, there is what has been called “white-coat silence.” This term is analogous to white-coat hypertension, a phenomenon in which people who have normal blood pressures become hypertensive when they come to the doctor. White-coat silence occurs when a usually informed, confident, and empowered person walks into a doctor’s office and fails to ask the important questions. The causes of this phenomenon are likely legion and, to a great extent, unknown, but it is real. But simply asking questions is not everything. It takes skill to ask the right questions about decisions that you do not know much about. How is a patient just diagnosed with angina supposed to know what medications should be prescribed for him?
On the doctor’s side, the two challenges are time and knowledge. Patients returning to Adam’s practice are scheduled in 20-minute intervals to allow (or force) him to see 12 patients in a four-hour session. This, of course, ignores the one or two people who are added during each session with urgent issues. Vinayak often finds himself having to see two patients at once—one admitted in the hospital and another arriving at the clinic. In accordance with Murphy’s Law, this usually happens right when a third patient calls. For these reasons, it is hard to get a harried doctor to sit back and reconsider decisions that he has made. Even more of a challenge is that many physicians do not have the knowledge at hand to answer the questions you should ask. A busy doctor often gets information about thera
peutics from guidelines and pharmaceutical detailers rather than the weekly medical journals. Even for those who do religiously consume the New England Journal of Medicine and the Journal of the American Medical Association, it has become harder and harder to identify the interventions that really make a difference. The design and reporting of trials is frequently (and in some cases purposely) complicated in an effort to make drugs look more effective than they are.
So, where do we start? We start by asking the right questions. Patients often gravitate toward the nuts and bolts of a recommendation. How does it work? What are the side effects? Does my insurance cover it? These are great questions—but are for a later time. Vertebroplasty, discussed in chapter 1, came with a great story. It also had few side effects and was covered by most insurance plans. The real question should have been, Does it work?
STEP 1. WHAT END POINTS MATTER?
When presented with a medical recommendation, start with a bit of soul-searching. What end points matter to you? We discussed many of these issues in chapter 3 when we talked about surrogate end points. Anything measurable can be an end point, but do not be fooled; not all end points are equal. If we are perfectly honest, only two end points matter: morbidity and mortality. Morbidity: will this treatment relieve my symptoms, or free me from or prevent future disability? Mortality: will this treatment make me live longer?
When a doctor suggests a treatment, take a second to think about what end points matter to you. This really gets back to surrogate end points, those end points that are easy to measure, which generally track with more important clinical end points but are, in themselves, meaningless. We covered these in detail in chapter 3. Let us imagine that you are meeting with your doctor to initiate treatment for recently diagnosed diabetes. You feel well; in fact you did not even suspect there was anything wrong with you until a blood test showed that you had elevated blood sugars. Your doctor suggests a medicine and says that this medication will effectively lower your blood sugar and hemoglobin A1c (HbA1c), the number that we use to monitor the average blood sugar. The questions to ask at this point are these: Why do we treat diabetes? What are the end points we are trying to avoid? The responses to these questions serve as the launching point for the rest of the discussion.
We treat diabetes not to get better blood-sugar numbers but because lowering blood sugars leads to fewer symptoms (frequent urination, insatiable thirst) and, ideally, avoids the late complications of the disease: blindness, nerve injury, kidney failure, heart attacks, strokes, and death. The specifics are important because not all treatments that lower blood sugar lessen the real burden of the disease—recall Thomas Galbraith back in chapter 3.
Sometimes the disconnect between the real, important outcomes and the stand-in, the surrogate outcome, can be truly disconcerting. A patient with bladder cancer that has spread to his liver and lung might be offered a treatment that has been proved to shrink tumors. This is not an important outcome. The patient may not feel the tumors. What matters is whether the drug helps the person to live as long as possible with the fewest symptoms. There are plenty of examples of treatments that shrink tumors but do not improve the end points you care about.
Now you could try to take this a step further and research the relationship between the surrogate and the end point you care about. How reliable is bone density as a marker of future hip fracture? How good is carotid-artery intimal thickness at predicting heart attacks? Be careful. Surrogates mislead frequently, and no one has perfectly figured out the rhyme or reason to it.
STEP 2. WHAT TYPE OF STUDIES SHOWS THAT THE TREATMENT IMPROVES THE END POINTS THAT MATTER?
If you are lucky enough to find that there is evidence that the treatment you have been offered affects the end point you are interested in, you next must determine whether this is reliable evidence. We hope our earlier discussions have convinced you that faulty evidence is not really any better than no evidence at all. If the evidence is one doctor’s experience or an observational study—you should doubt it. You would like the evidence to come from a randomized trial done at hospitals across the country.
But then you should also ask, How good is that randomized trial? This is a question that attentive doctors spend a lot of time thinking about but one that will be very hard for you, as a patient, to answer on your own. Not all randomized controlled trials are created equal and not every trial applies to every patient. Are the patients in the trial like you? Same age? Same problems? Same fitness? Same country? Who paid for this trial? Was it a government-sponsored trial that was trying to discover the truth or an industry-sponsored one that was trying to sell a drug?
Before going on to Step 3, it is time to pause and consider what you have been offered. If the treatment suggested has not been proved, in a well-done randomized trial, to benefit the outcome in which you are interested, you need to ask about alternative treatments. We are fortunate to live at a time when there are usually alternatives. Many of the alternatives are older, less expensive, well-proven treatments. They may not be the hottest new drug with an impressive advertising campaign, but they may be more effective. Note the “may be.” It goes without saying that you should ask the same questions about the alternative treatments.
It may be hard to get these alternatives. Doctors will sometimes gloss over them. They have reasons for their recommendations, and there are reasons why the alternatives are just that, alternatives—but this is your decision. Get a couple of options, write them down, and ask the right questions. A key question at this point is, What happens if we do nothing? In doctorspeak, this means, What is the natural history of the disease? Lots of doctors do not know the answer to this question for diseases they treat all the time. Why? Well, because it is so rare that we do nothing. For some conditions, the natural history is just to get better (or not to get worse). That is the case for most musculoskeletal injuries and, as the medical profession has learned recently, for some early-stage, caught-only-by-screening, cancers. For some conditions, the natural history is death. Natural history can be based on old data, data for patients with severe versions of illness, or data from a few cases. Do these data apply today, to you?
STEP 3. HOW MUCH WILL THIS TREATMENT HELP?
If you are lucky and have been offered a treatment that addresses the end points you care about and has been proved in a well-done, randomized controlled trial, the answer to this question will help you decide whether the therapy is worth it. It will help you weigh risks and benefits. You need to ask your doctor the “number needed to treat.” How many people like me need to get this therapy, for what period of time, for one person to benefit from it? The size of this number will always surprise you. Our best therapies that are meant to prolong life (outside the realm of cancer care) require us to treat 20 to 50 patients to provide benefit to one. Treating patients with cholesterol-lowering statin drugs after an MI may be gospel in medicine, but you need to treat 20 people for five years to save one life. The numbers are even greater for preventive interventions. Few people question colon-cancer screening, but it takes doing endoscopy on 191 people and following them for 11 years to prevent one case of colon cancer. You will need to consider this likelihood of benefit compared to side effects, costs, and the hassle of the therapy to decide whether this treatment is right for you.
STEP 4. MAKING THE DECISION
If you have been offered a treatment that is proved to affect end points you care about, you can decide whether you want it. You have considered the “number needed to treat.” Now ask those questions that you wanted to ask from the start. How does it work? What are the side effects? Does my insurance cover it?
If the only therapy for a problem is an unproven one, your decision whether to accept it depends on your state of health. If you are lucky enough not to have a medical problem and you feel well, the decision is easy. In this case, the unproven treatment that you are being offered is to prevent you from becoming ill. To quote Nancy Reagan, “Just say no.” The likelihood that such a treatme
nt will improve the things that matter is very low. The list of medical practices that improve outcomes among healthy individuals is a very short one.* If you are sick or suffering, the decision is harder. The treatment you are being offered may help, but it also may not. This is true no matter how convincing the surrogate end points are, how convincing the explanation of how the treatment should work is, or how well produced the advertisements are. It is in your best interest (as well as your doctor’s) to understand that you are accepting the treatment knowing that there is a fairly good possibility it will not work. Remember that the majority of people do not even benefit from the treatments that every doctor agrees are indicated.†
Consider three patients. The first is a patient with elevated bad cholesterol and low good cholesterol. In 2006 she was 55 years old and had had diabetes for five years. Her doctor prescribed niacin. There were trials showing that it raised good cholesterol and lowered bad cholesterol. The mechanism of action is elegant. The drug acts in several parts of the cholesterol pathway. Moreover, the patient tolerated the drug well, without flushing, the most common side effect. Finally, the drug was FDA-approved and was covered by the patient’s insurance. However, there were no studies that proved the drug improved clinical end points—cardiovascular outcomes and death. The patient took the medication for years before it became clear that niacin does not improve these outcomes.
The second patient is a 44-year-old man, an avid runner. On a morning jog he tripped and fell, rupturing his hamstring. Traditionally this injury was treated with rest and physical therapy. Some patients did well; others had persistent pain and disability. His doctor, knowing how important running was to this man, referred him to an orthopedic surgeon who recommended surgical repair. The studies that support this intervention were poorly designed: they compared the outcomes of patients who were offered (and chose to have) surgery with those who were not (or did not). The rationale behind the procedure was logical; the surgeon would reattach the hamstring to where it had been for the previous 44 years. The patient, wanting to know he had done everything to improve his chances, chose to have the surgery. After six weeks of immobility and six weeks of therapy, he is back to his baseline performance. Of course, we do not know whether the surgery made the difference, but the runner does not care.
Ending Medical Reversal Page 22