A friend recently send me a link to a Wall Street Journal opinion piece by Norman Gleicher titled 'Expert Panels' Won't Improve Health Care. Here is part of my response: 

a) insurance companies all have panels now, but they are secretive, inconsistent, and dedicated to protecting their companies from meeting their contractual obligations 

b) government panels are more likely to be open, challengeable, and uniform–not necessarily better 

c) Dr. Gleicher is right that 'evidence-based' has much less evidence than the public believes. That would be the source of the article in my head (and likely never to see print) that is entitled "the evidence against evidence-based medicine". This article would point out that individual choice and variability, among other things, trump the evidence, especially in chronic illness. The other point is that 'evidence-based' usually is judged on a 1-5 year outcome, and may come to a completely different conclusion if the gold standard is a 10 or 20 year outcome.

A patient recently asked me about an article on People Magazine’s website about Deepak Chopra's belief that Michael Jackson had lupus and that this disease was connected to Jackson's childhood abuse.

Dr. Chopra cited ‘recent research’ in support of this contention.  

Here is some of what I wrote my patient: 

The article from the Mailman School is one of those associations that epidemiologists sometimes come up with that are not necessarily causal. What happens is that people with one medical problem virtually always have a higher frequency of a second medical problem, probably because (A.) they are much more in contact with doctors and (B.) they're much more aware of their bodies than are people who have not had to question their health. Epidemiologists (I was one once, thanks to Uncle Sam, but that's another story) are good for hypothesis generating but not for hypothesis testing. 

Other articles on this topic are more sensible. Quite a number of things about socioeconomic status have been investigated with regard to lupus. Although it is commoner in African-Americans, and African-Americans are on the whole poorer than Caucasians, there have been no bona fide associations between abuse and autoimmune illness that I know of.

It is a terrific idea to compare effectiveness of treatments for the same illness—if “effectiveness” and “same illness” mean the same to the doctors, the patients, and the payers and patients who will benefit or lose according to the rules that follow. It is a terrific idea if the comparison meets the needs of each participant and if effectiveness measured in the short term also predicts what happens after many years.

In chronic illness these ifs are mostly wrong.  For practical reasons, treatment studies mostly look at outcome after one or two years and at only one measure of success, such as, did a blood test improve? For chronically ill persons a short-term result many not predict the distant future. Some treatment choices require quick decisions: how best to repair a fracture after a fall. The immediate outcome may say the treatment worked (the fracture healed), but a later measure may give a different answer (walking is worse because the wrong repair was made). An illness relapses; the result—did permanent damage occur because the wrong treatment was used?—may take decades to know. A treatment effective in the short term (relief from pain) may be harmful later on (the chosen treatment weakens bones). The most important measure of effectiveness may be that disability does not occur, which we may not know for decades.

Treatments have side effects that some patients tolerate and others do not. Some patients choose one treatment that prevents crippling despite the fact that it increases risk of heart attacks, while another chooses a treatment that spares the heart but carries a high risk that the patient will end up in a wheel chair. Which is more effective? A third person cannot know the patient’s choice.

Effectiveness trials do not study complicated patients. Although a patient with rheumatoid arthritis who has had breast cancer cannot participate in an arthritis trial, this patient will still need treatment at some time. Patients with very mild disease and patients with severe disease are not part of trials but need to be treated as well. Exceptional patients—perhaps the majority of those with a given diagnosis—may not be able to take the “effective” drug, but, knowing the patient only by his diagnosis, the third party suggesting the most effective treatment may make the wrong call.

In a treatment trial, patients and researchers agree on the goals, relief of pain, for instance, or a joint less swollen. Outside the trial patients may have different goals. Methotrexate, the most effective first drug for rheumatoid arthritis patients, can deform or kill the baby of pregnant woman. Some women will not take it because they hope to or may accidentally become pregnant. Others will not take it for religious reasons. But even today insurers challenge the prescription of a less preferred drug while offering the physician no opportunity to explain deviation from their version of the effectiveness norm.

Personal goals can also trump the most-effective-treatment choice. To care for an ailing parent or child, to stay employed, to save a marriage, many patients will select short term gain over long-term risk. A few years ago, when Vioxx and Bextra were taken from the market, several of my patients ranked the relief provided by those drugs more important than risk of heart attack and hoarded as much of these drugs as they could find. Many patients so hate the way a drug makes them feel that they refuse a second course of treatment no matter how effective the drug is deemed.  “Effectiveness” seldom acknowledges to patients’ personal concerns.

Nonetheless, it is possible and right to compare effectiveness of medical treatments. In doing so, we should not use a one diagnosis, one time period, one outcome trial to decide the best treatment for complicated patients with chronic disease. We should instead compare treatments under real world conditions that include patients with second illnesses, with different severities of illness or different disease courses, and whose personal needs are respected when the treatment is chosen. The term “effective” should consider whether the effect is true for both the short- and for the long-term. Above all, labeling a treatment effective should weigh the patient’s opinion whether the treatment has or has not enhanced his life.

Dr. Barron Lerner in the New York Times ("A life-changing case for doctors in training," March 3, 2009) misinterprets the history of Libby Zion and the Bell Commission, and draws the wrong lessons. The concern about medical resident fatigue dates not from the 1984 case but at least as far back as a 1971 study on the effects of sleep deprivation on electrocardiogram interpretation. Widely recognized concerns about physician fatigue led to a reduction from every other night call in the 1960s to every third then every fourth night in the 1970s and 80s. The tragedy of Ms. Zion’s death was the culmination, not the stimulus, of this trend.  
The important lessons of this death, not all of which have been addressed even today, are different concerns about how hospitals run:  Is it feasible for one physician to give adequate care simultaneously to patients "boarding" in multiple areas of the hospital, so placed because bed availability is sparse? How can physicians best care for hallucinating, intoxicated patients who resist care? How can one facilitate effective communication between supervising and supervised physicians? Are medical errors due to physician fatigue less or more severe than errors due to faulty handoffs? What system changes can one make to prevent these errors?
 

Atul Gawande, the surgeon-writer, is outraged (A Lifesaving Checklist 12/30/07.) Apparently an ongoing medical study, that he endorsed, was stopped because of a bureaucratic rule. The study was trying to this question:  Are there fewer hospital infections if doctors follow a checklist, as do airline pilots on take-off (sample: Have you washed your hands? Have you draped the patient with a sterile cover) than if treat patients according to their own rules? This, Dr. Gawande fomented, is bureaucracy at its worst. Stopping the study is outrageous, because the benefit of checklists is so obvious. (He did not ask why, if the benefit is obvious, the study should be done at all.) 

The study asked the doctors for half the patients to use checklists to prepare for a procedure while the doctors for the other half would prepare as they saw fit. The reason the bureaucracy—in this case the Office for Human Research Protections (OHRP)—gave to cancel the study was that the investigators had not asked patients to consent to participate. Some might have refused. After all, if you were the patient, would you want to be under the care of the doctor who did not use a checklist?  It is, of course, possible that using a checklist makes no difference at all, that doctors do the right thing all the time, but, if you are the patient, do you really want to take that chance? Shouldn’t you be asked? 

Dr. Gawande thought that insensitive bureaucracy had stupidly ended a useful study. He missed a larger point. First, we do not know whether the checklist does or does not make a difference (preliminary studies suggests it does), so it is perfectly appropriate to do the test—improving our knowledge is clearly worthwhile. Second, an experiment on a patient is not just good or bad. Human experiments engage several value systems not all of which are easily seen. The conflict between Dr. Gawande’s wish to continue and OHRP’s command to stop was a conflict between, and a different weighting of, these different sets of values. 

In research using patients, probably the highest value is accorded to patient autonomy.  A research patient has a fundamental right: to refuse to participate. This right trumps any perceived benefit of the research. OHRP clearly focused on this value; patients should have been given the right to refuse. 

The second most highly ranked value is that of privacy. This right was emphasized in first days of AIDS, to protect against discrimination or the public release of personal information.  To protect this right, medical researchers now are burdened with cumbersome regulations; all patient-identifying information must be removed from all documents.  Research data must be kept in locked cabinets. Physicians may not transmit information about patients to anyone, including family, except another treating physician, unless a patient expressly consents (a ruling that applies to all aspects of patient care).  This is not only a matter of ethics; in many states it is a matter of law. Researchers and treating physicians find it exhausting to comply with the paperwork engendered by these rules.  Privacy rules delay research; burdensome regulations cause some potential researchers not to enter the field at all.  Since not very much that is private is a risk in the checklist research, Dr. Gawande seems to rate the privacy value low; he would dispense with the bureaucratic needs. 

A third value can be summarized by this sentence: Medical research should provide the greatest good to the greatest number.  The checklist study would do this. The investigators may have assumed that checklists and observation of doctor behavior does not count as research. They may be correct. Audits of practice, whether or not one collects additional data, are usually exempt from research rules.  Furthermore, the agencies that rule over human research exempt some research from consent—studies of population behavior, for instance—and provide expedited approvals for low risk research.    The researchers should have asked for an exemption, but they certainly know that local human research monitoring boards are inconsistent.  When the same research is done in several different sites, as the checklist study was done, one local board may consider it exempt from consent rules, while another may require full consent.  Such inconsistencies exasperate researchers and delay research if not force it to a complete halt. Had OHRP been asked before the experiment began, it might have ruled that checklists are practice audits exempt from consent, or it might not. One does not know until one asks. 

Society values innovation and rapid dissemination of research results.  These values contradict the values of autonomy and privacy. Innovation and dissemination demand flexibility of experiment rules and open and ready access by the public; autonomy and privacy restrict the rate of flow and narrow the volume of information that may be shared.  Even the creation of a widely desired a national medical database is hobbled by privacy concerns. 

The value “do no harm” has a very high priority. Pragmatically, however, the fact of the study is closer to “do less harm” or “do more good” than it is to “do no harm.” Consider the comparison of a new drug with an old; the new one is thought to be safer, but it may not be. If the new drug is better, those receiving the old will be (in one sense) harmed by the experiment; if the new drug turns out to be, in fact, more toxic, those receiving it will be harmed. Or consider the circumstance in which early surgery is thought to give a better cure rate than a medical treatment. Either the surgery or the medical-treated patient group could be harmed or helped, depending what the answer turns out to be. In some cases an experiment demonstrating good yields a very small, one might say trivial, improvement; in another case the difference might be as great as that between death and cure. The amount of potential good or potential harm can be weighed. Committees that judge the overall value of a research study on humans use common sense rules but apply no formal measurement scales to these values. Absent a metric, OHRP’s and Dr. Gawande’s opinions collide. 

So what to do?  Protect the privacy of the research subject?  Facilitate the rapid conduct of research? Protect at all costs against possible harm? Or make possible a public good? 

One cannot reconcile the conflicting demands.  One can try to make the committees more uniform, the rules more efficient, but these are minor fixes. Instead, it would be best to concede that our contradictory rules derive from different societal values, and that no one rule is absolute.  We can assign point systems—degrees of risk—that have positive (for benefit) and negative (for risk of harm) numbers instead of ruling by yes/no criteria. There are different levels of privacy and autonomy, and different degrees of good and harm. Scientific importance can be quantitated. The potential public gain can be as well.  The numbers can be summed, and the potential worth of the research scored. 

The table below gives examples. How the scoring system is actually constructed is not an important point. What is important is that the investigators should document how their proposals address each of the values; what is important is that the committee should systematically weigh each value when it decides to approve or disallow a specific project of human research. When that happ
ens, Dr. Gawande, instead of railing against a faceless bureaucracy, will see its decision from another point of view, and OHRP will understand that some quantitatively small risks can be waived.

The New York Times today had an article written by Stephanie Saul entitled THE COVERAGE GAP Avoiding Medicare’s Big Hole. The last three paragraphs discuss the ‘fact’ that many patients are reluctant to discuss the cost of their prescriptions with their doctors. The claim that patients and their doctors do not want to discuss cost of medicines is both incomplete and wrong. Several factors inform both my choice of what to write on a prescription pad and the out-of-pocket cost to the patient: * Different insurers choose as "preferred" prescriptions different medications of the same class, presumably because the insurer has purchased the chosen drug in bulk at discount. Unfortunately, when a physician works with six or more different insurers, keeping in mind which is preferred for which insurer–they are never the same–is difficult and tedious to find out. * Insurers that do have "preferred" prescriptions change their preferences at random, unpredictable intervals, so that a "preferred" prescription in October may have changed when the patient returns in April. * The same problem occurs when a patient changes insurers, as many annually do. * Because of co-payment rules, patients often ask me to write a three-month supply for a medication they need to take for 10 days, the former having no co-payment, the latter requiring out-of-pocket payment. * A patient may speak of the cost of medication at a convenient pharmacy, but the cost for the same medication will be quite different at, for instance, a Wal-Mart on the other side of town. * Some patients clearly have adverse or beneficial responses to one brand of "equivalent" drug but not to another–and are quite explicit in saying so. * New York State has a generic drug prescription law, requiring me to certify by a separate signature that a specific brand is required. If I write a prescription using the easier-to-remember proprietary name and do not sign the box certifying the brand, the pharmacist is required to supply the cheaper generic–or equivalent–instead. Most pharmacists do, so I have to ask my patients on return visit what drug they have been given in lieu of the one I prescribed. * Years of practice have given me a comfort level with some medications compared to others, an instinct, if you will, regarding good and bad things to expect. To use a usual doctor's double-negative: this experience is not irrelevant. The problem is not so much that patients and doctors are embarrassed or unwilling to discuss drug prices as it is that, when all factors are considered, the facts of cost are hard to ascertain. If there is any simple fix, it is to standardize costs and preferences across all insurers and suppliers, so that when I write a prescription I can reasonably predict what will happen next.