The Good Doctor

by Kenneth Brigham

  There are some other reasons why a lot of what we believe about health and disease may be wrong. One reason is that we are endlessly fascinated by observational studies—studies that link two apparently unrelated events. These make headlines almost daily. Heartburn treatments cause dementia! Coffee causes pancreatic cancer! Eating breakfast cereal causes women to give birth to more boys! Those all make good headlines and may even be based on solid observational science, but none of them is proven to be true.

  Stanley Young, Assistant Director of Bioinformatics at the U.S. National Institute of Statistical Sciences, and his colleague Alan Karr identified twelve papers describing observational studies that were subsequently tested in randomized clinical trials.92 What do you think they discovered? Not a single one of the more critical studies agreed with the original observation, and in five of the studies the results of the definitive studies were in the opposite direction of the original report. There is a message here (alluded to in the earlier chapters). Don’t trust observational studies that show associations between one variable and another, regardless of what the variables are. Remember the bucket brigade and the ebbing tide. Could be true (the bucket brigade removes water from the ocean) and true (the shoreline recedes) but unrelated.

  Sometimes things happen because one thinks they should or even just because one wants them to. Of the various eponyms for these phenomena (Placebo effect, Hawthorne effect), we like Pygmalion effect.93 The reference is not to G.B. Shaw’s play but to the fictional Cypriot sculptor (in Ovid’s Metamorphoses) whose ivory carving of a woman so enraptured him that he no longer desired real women. After Pygmalion made offerings to Aphrodite wishing for a bride, “the living likeness of my ivory girl,” the sculpture came to life. So the Pygmalion effect is an outcome determined solely by what was expected (or desired) of the subjects entering the study. There are plenty of examples. There are also plenty of studies showing real physiological effects of a sugar pill (Placebo effect), and definite effects resulting solely from the fact that the subjects in the study know that they are being observed (Hawthorne effect). These are not imaginings or magic tricks. These are real effects!

  But are these effects that are related more to the conditions of the experiment than to the thing being tested legitimate medical evidence? If you believe something works that has no obvious scientific explanation and you do it and your condition improves, what would the good doctor who is comfortable with uncertainty say about that? She would probably say go for it! We don’t understand why a lot of things work in medicine. Doctors prescribed and we took megatons of aspirin before someone discovered that it inhibited production of prostanoids, biochemicals that are major players in pain and inflammation. Doctors felt a lot better using the drug once they knew that, but aspirin didn’t work any better than it ever had.

  This good doctor considers herself a medical scientist, that is, she feels most confident in decisions based on sound experimental evidence. But she is also a perpetual student and never stops asking why. She realizes that important things can be learned from human experience and so she looks hard at clinical results that may lack scientific proof—pragmatic evidence. Two specific areas that interest her are eastern traditional medicine, and empirical results in everyday medical practices vis-à-vis formal clinical trials.

  Eastern traditional medical practices have been used for centuries by millions of people who believe that they are effective even though the explanations for their effects are difficult for the western medical mind to accept. A science-based physician believes that any such practices must be tested by well-designed clinical trials before they can be incorporated into the professionally sanctioned armamentarium of the practicing doctor. This is no different than is required of any medical intervention. However, the doctor comfortable with uncertainty also knows that even good clinical trials with negative results do not rule out the possibility of a therapeutic effect in some circumstances. Suppose you are doing or taking something that is clearly outside the bounds of sanctioned medical practice and you are convinced that it makes you better. If it is unlikely to do harm and there is no reason to think that it will interfere with your sanctioned treatment, the good doctor would be very likely to tell you to go ahead with it. You begin to see the healing power of uncertainty. Sometimes what’s important to you is what your doctor knows she doesn’t know for sure.

  Pragmatic evidence also becomes an issue when there is a difference between the subjects who participate in randomized prospective clinical trials and the patients seen in actual medical practices. The very fact that design of a clinical trial requires focusing on the condition to be studied and minimizing any confounding conditions means that the people in the study may not be exactly like you or the other people who sit in doctors’ offices waiting for advice.

  Recognition of this problem has birthed the emerging field of pragmatic research trials.94 The reasoning is simple. Test the therapy in patients in real medical practices who have all of the confounding conditions that may affect outcomes. Include essentially everybody with the condition being studied, regardless of what else they have. The question is, are the results of the tightly controlled trial the same as results from studies in unselected patients being seen in an everyday medical practice? The pragmatic researcher might argue that if they aren’t, then the data from the more rigidly controlled studies is interesting but meaningless in the real medical world.

  We strongly suspect that Dr. Ioannidis would threaten apoplexy at the thought of such experiments. How does one interpret the data? What kinds of statistics can be applied to such situations? Dr. Iaonnidis would not be at all surprised if the messier experiment in the clinic failed to find the statistically significant results that the more rigidly controlled experiment claimed. The good doctor will keep abreast of these pragmatic studies, but will view them in the context of all of the other available information about your condition and will do her best to understand what the aggregated information means to you.

  Try as you might to avoid it, there will always be uncertainty. All evidence-based conclusions are subject to revision as more evidence comes to light. The good doctor knows that evidence plays a vital role in health care, but she also knows that the available evidence may not be the final word.

  Karl Smallwood, “Sherlock Holmes Never Said “Elementary, My Dear Watson,” Today I Found Out, August 27, 2013, http://www.todayifoundout.com/index.php/2013/08/sherlock-holmes-never-said-elementary-dear-watson/.

  “Times Topics: Andrew Wakefield,” The New York Times, http://www.nytimes.com/topic/person/andrew-wakefield.

  “The Anti-vaccination Movement: A Study in Propaganda, Disinformation, and Dishonesty, What Would Jack Do?, February 3, 2015, http://whatwouldjackdo.net/2015/02/the-anti-vaccination-movement-a-study-in-propaganda-disinformation-and-dishonesty.html.

  “Measles Cases and Outbreaks,” Centers for Disease Control and Prevention, http://www.cdc.gov/measles/cases-outbreaks.html.

  “Carl Sagan Quotes,” BrainyQuote, http://www.brainyquote.com/quotes/quotes/c/carlsagan589698.html.

  Nassim Nicholas Taleb, The Black Swan: The Impact of the Highly Improbable (Random House Trade Paperbacks, 2010).

  Hriday Shah and Kevin Chung, “Archie Cochrane and His Vision for Evidence-Based Medicine,” Plastic and Reconstructive Surgery 124 (2009): 982-988.

  A. L. Cochrane, Effectiveness and Efficiency: Random Reflections on Health Services (London: Nuffield Provincial Hospitals Trust, 1973).

  A. Levin, “The Cochrane Collaboration,” Annals of Internal Medicine 135 (2001): 309-312.

  David Freedman, “Lies, Damned Lies, and Medical Science,” The Atlantic, November 2010.

  “Hype Cycle,” Wikipedia, https://en.wikipedia.org/wiki/Hype_cycle.

  Stanley Young and Alan Karr, “Demi
ng, Data, and Observational Studies: A Process Out of Control and Needing Fixing,” Significance, pp. 116-120, September 2011.

  Stephen W. Draper, “The Hawthorne, Pygmalion, Placebo and Other Effects of Expectation: Some Notes,” http://www.psy.gla.ac.uk/~steve/hawth.html.

  “Explanatory and Pragmatic Research,” Open Philanthropy Project, http://www.openphilanthropy.org/explanatory-and-pragmatic-research.

  CHAPTER 9

  Information Is Not Necessarily Knowledge

  Paradoxically, the more I learn about medical problems,

  the less I seem to know . . .

  —PHILLIP K. PETERSON, M.D., in Get Inside Your Doctor’s Head95

  Don’t worry very much about whether your doctor has enough information about your condition. One doesn’t have to remember all that stuff anymore; it’s all there and available at the click of a mouse to anyone with a WiFi connection. But you should worry a lot about two things: Can your doctor critically evaluate the unfiltered deluge of information released by that mouse click, pick out the flowers from among the garbage? And does your doctor understand what the information means, does she transform information into knowledge?

  TMI (too much information), a texter’s common response to oversharing, is an appropriate comment on the state of medical information available on the internet.

  Google virtually any term you can think of that has to do with a health-related topic of interest, and a mouse click will flood your computer screen with thousands of hits that purport to contain reliable information on the topic. So the problem is not getting information. The problem is that you can access more information than you can possibly assimilate and that what you get is an olio of facts of varying credibility and pure fictions, each nicely presented with similar claims, equally attractive computer graphics, and equally convincing narratives. The odds of coming away from such a blind search with anything like an accurate view of the topic you were trying to understand aren’t very good.

  Suppose you want the latest inside scoop on how to be as sure as you can be that your new infant will not die in her sleep. You google a bunch of phrases like “Infant sleep position,” “Infant co-sleeping,” “pacifier sleeping,” etc.; Dr. Rachel Moon, professor of pediatrics at the University of Virginia, an expert on sudden infant death syndrome (SIDS), and colleagues did precisely this search using 13 different terms.96 Fewer than half of the 1,300 websites that Moon et al. analyzed had information consistent with the official American Academy of Pediatrics recommendations. And a lot of the sites were carefully disguised sales pitches or appeals from special interest groups. A Mayo Clinic study several years ago concluded that medical advice about several common ailments gleaned from the internet was more likely to be either unavailable or incomplete than correct and useful.97 You desperately need a knowledgeable doctor’s help to navigate this morass if you are to extract any accurate information. She will help you do that, but she will also warn you that getting information, even accurate information, doesn’t mean that you understand the problem. Some even argue that our propensity for conflating information and knowledge creates a paradox—more information, less understanding.

  A couple of examples may help to make the point. Suppose you are in the business of manufacturing men’s suits and your production manager comes up with a brilliant idea that will dramatically enhance the bottom line. He has discovered an extensive database with detailed physical measurements of men from each major ethnic group in the U.S. Based on such measurements in one hundred thousand randomly selected men in each category, statisticians have constructed in exquisite quantitative detail a physical model of the body of the average American male of each ethnicity. This intrepid manager suggests that you manufacture only one size suit in each ethnic category perfectly fitted to the statistical models. Think how much you’ll save in production costs!

  What are the odds of a man entering a local shop that carries your brand and finding one of your suits that fits him? Well, the odds are certainly low and depending on how good a fit the customer insists on, the odds may be zero. As soon as you start getting feedback from retailers, that production manager will be toast. His problem was not lack of accurate information; he had information in spades. His problem was that he failed to understand what the information meant and so made a devastating mistake in applying it.

  Here is another example of the difference between information and understanding that Todd Rose, director of the Mind, Brain, and Education program at Harvard, uses in his book The End of Average.98 If you wanted to know how typing speed relates to accuracy, you could gather data from all the typists you could find and make a graph of the two variables, words typed per minute versus number of mistakes per minute. You would find an inverse relationship—expert typists type faster and also make fewer mistakes than we hunt-and-peck types. However, if you want to know how to improve your typing accuracy, you would be making a big mistake to force yourself to type faster; the faster you try to type the more mistakes you will make (assuming you are not a highly skilled typist). Make a graph of your individual words per minute versus mistakes per minute and the relationship will be positive, exactly opposite the data from the expert group. Understanding the relationship between typing speed and accuracy requires that you know the circumstances. The population data do not predict an individual result.

  Your doctor should know that statistically analyzed population data don’t fit any specific person. When judged by such information, we are all misfits. The clothing industry deals with this by producing a range of sizes from which the shopper can select the one that fits best, recognizing that the fit won’t be perfect. Likewise, there is usually a range of options in a given medical situation and choices are made that take into account characteristics of the specific patient. However, the thinking doctor understands that even so, the fit may not be as good as it can be, and so she is constantly looking for anything about you that could cause you to respond differently from the subjects studied in clinical trials. Population studies rigorously done and carefully analyzed are critically important to your health care, but a doctor who knows and understands those studies will realize that their value is in demonstrating interventions’ effects or lack thereof in groups of your fellow humans and not necessarily in how precisely they predict what will happen to you.

  The good doctor knows that the same information may be crucial in one situation and useless in another. For example, she understands the difference between practicing public health and practicing medicine.

  The public health practitioner relies on data from large populations to implement policies aimed at improving overall health outcomes for a population. Success or failure can be measured by whether or not the targeted outcome improves. Count the bodies (or whatever) before and after implementing the policy, tote up the data, and do the statistics. There are many important public health successes. Fewer people get arbovirus infections when mosquito populations are controlled. Tooth decay decreases when public water supplies are fluorinated. The incidence of thyroid goiters goes down when iodine is added to table salt. The incidence and consequences of several infectious diseases and environment- or behavior-related health problems improve when the public health people aim aggressive programs at things like clean water, flush toilets, and cigarette smoking. The good doctor encourages her patients to respond to public health campaigns that promote immunizations, smoking cessation, healthy diets, etc.

  But, the public health practitioner treats large aggregates of people and addresses health issues that many individuals share—the most good for the most people. The medical doctor’s job is to deal with people one by one and each one she faces is different from everyone else in potentially important ways; every one of her patients is a one-off. Your doctor’s responsibility is to help you to be healthier and to get you back to good health as quickly as possible when you get sick. She knows that what the available information means
to your situation may be very different from what it means to the public health policy maker. You may be a real outlier when measured against the general population, or even a subpopulation of people with a problem similar to yours. You can’t make yourself type faster and expect fewer mistakes, no matter what happens in the typing world at large. You and your doctor need to figure out where you are on the spectrum of possibilities and what course of action is most likely to benefit you. The good doctor values the experts’ recommendations based on what happens in large groups of people and knows that applying that knowledge to unique human beings may require some flexibility. That’s why this doctor is open to the whole range of possibilities. She knows that the essence of understanding is knowing what she doesn’t know for sure, and she is prepared to use that ignorance to help you, unique individual N of 1 that you are, to be as healthy as you can be.