Hard to Say Whether Medicine Does More Good Than Harm

A draft article by Spyros Makridakis about blood pressure and iatrogenics takes issue with the statement that “The treatment of hypertension has been one of medicine’s major successes of the past half-century.” Over the last half-century, the article says, the death rate for people with high blood pressure decreased by almost exactly the same amount as the death rate for people without high blood pressure. Apparently “one of medicine’s major successes” is a case where the health benefit no more than equaled the health cost — leaving aside what the treatment cost in time and money.

Because very high blood pressure (systolic > 180 mm Hg) is quite dangerous and blood pressure drugs really work, this is a surprising outcome. Makridakis points out that doctors start treating high blood pressure when it rises above  systolic = 140 mm Hg, a point when there is little or no increase in death rate. This article tells doctors to immediately prescribe drugs when systolic blood pressure is above 160. Yet death rate clearly increases only when systolic blood pressure is above 180. Makridakis concludes (as do I) that blood pressure drugs have significant health costs as well as benefits. The drugs are so often prescribed when they do no good and the costs are so high that the overall health costs of blood pressure treatment have managed to be as high as the overall benefits. Even when handed a relatively easy-to-measure problem (high blood pressure) and a relatively simple solution (blood pressure drugs), our health care system managed to achieve no clear gain. If this is “one of medicine’s major successes”, medicine is in bad shape.

The last paragraph of Makridakis’s article makes a surprising statement: “We strongly believe that medicine is extremely useful.” It does not explain this belief, which is contradicted by the rest of the article. I was puzzled. I wrote to the author:

I recently read your paper on “High blood pressure and iatrogenics”. The main part makes good sense. Then it ends with something quite puzzling: “We strongly believe that medicine is extremely useful.” No doubt a few areas of medicine are extremely useful. For large chunks of medicine, it is hard to tell whether they do more good than harm, because so many drugs and other treatments have undisclosed or unnoticed bad effects.

For example, tonsillectomies — for a long time the most common operation — is associated with a 50% increase in mortality in one study. The notion that cutting off part of the immune system is a good idea makes as much sense as the idea that cutting out part of the brain is a good idea. Another example is sleeping pills. They are associated with a three-fold increase in death rate soon after they begin to be taken. I am not saying that medicine overall does more harm than good. I am saying that a strong belief about the outcome of such an assessment (does medicine overall do more good than harm?) doesn’t make sense.

Makridakis replied:

Thank you for your email. The paper you mention is a draft posted for comments. I agree with you that my statement is wrong. It should have read: : “We strongly believe that medicine can be extremely useful”. For instance, this could be the case in treating heart attacks, strokes, traumas from car accidents or bullet shots. But in most other cases the harm from treatment can be greater than the benefits. In addition, the harm from preventive medicine can exceed its value. Thank you for pointing out this mistake to me.

Puzzle resolved.

8 Responses to “Hard to Say Whether Medicine Does More Good Than Harm”

  1. Caleb C Says:

    Reminds me of Kurt Harris, an MDs, view of medical nihilism:

    “I won’t expand much on the influence of medical school hazing and 23 years of reading xrays, CTs, MRIs, et cetera, but this has been a huge influence, especially in making me a medical nihilist.

    A medical nihilist posits that in a world where the entire medical system (alternative and complementary not exempted) disappeared in some selective rapture, that the net effect would be positive for the economy, and no worse than neutral for the aggregate level of health and wellness”

    I’m a healthcare provider as well, and I firmly believe Kurt’s assessment of our profession is correct. Of course, I more or less believe it could be extended to most of the top elite fields; education, politics, finance, government, lawyers, medicine; if they were to disappear in a selective rapture the net effect would be positive for the economy, and often times result in a huge aggregate increase in human satisfaction and enjoyment of life.

    I guess it shouldn’t be a surprise that elites will, if allowed, rig the social system so that they capture more wealth than they contribute in utility. At least the old aristocratic elite would risk their lives to protect you from the Viking raiders every now and then, and on average had pretty good tastes in what arts and science they promoted. Comparing state sponsored art these days to the cathedrals, I can’t help but think our societies elites are far inferior in quality:)

  2. Robbo Says:

    Wow !

    “Thank you for pointing out this mistake to me.”

    Both of you deserve great credit over this exchange.

  3. Greg Says:

    Makridakis borrowed a graph from an article by statistician Sidney Port. Here’s another one by him about blood pressure and mortality, showing why researchers that may not be experts in statistics can find linear relationships where the true effects are nonlinear.

    http://www.math.ucla.edu/~scp/publications/EHJ.pdf

    This doesn’t appear to be a popular reading of the data. Framingham’s senior investigator responded and claimed to refute Port’s analysis “without resorting to any type of statistical modeling” (or even having a statistician listed as a co-author). Not sure what to think, except that I’d sure like to see more statisticians writing/co-writing medical papers.

    http://hyper.ahajournals.org/content/42/4/453.long

  4. Kim Øyhus, viking raider Says:

    They assume high blood pressure is a defect.
    They do not ask why the body raises the blood pressure.
    If the body is trying to achieve something by raising blood pressure,
    such as trying to avoid something worse, then pressure lowering medicine
    will make something worse happen.

  5. dearieme Says:

    “Thank you for pointing out this mistake to me.” I’m with Robbo.

  6. dearieme Says:

    In Greg’s second link: “However, the more relevant outcome is CVD mortality and the CVD events promoted by hypertension unconfounded by non-CVD mortality, which could be associated with low BP.”

    Against which I observe:
    (i) Relevant to what? Says who?
    (ii) All cause mortality has the advantages of (a) not being potentially polluted by errors re cause of death, and (b) being useful to anyone who’s fairly relaxed about how he dies but less so about when he dies.
    (iii) It begs the question to assume that CVD events are promoted by hypertension, probably in general and certainly when the debate concerns what constitutes hypertension.
    (iv) “non-CVD mortality, which could be associated with low BP”: I’m trying to decide whether this remark is (a) mere windy assertion, (b) question-begging, (c) evidence of not having grasped the point of the paper they are criticising. These categories are not mutually exclusive.

    The paper in the first link is altogether on a higher intellectual plane than the one in the second. That doesn’t guarantee that it gets closer to the truth, but it stacks the odds in its favour.

    Seth: I wouldn’t say that CVD mortality is “more” relevant, but it is relevant. I think the authors were trying to say this: Let’s assume 1. High blood pressure causes CVD mortality. 2. Low blood pressure does NOT cause mortality but is associated with many diseases that do (these diseases cause both death and low blood pressure). If these assumptions are true, by looking at the CVD mortality vs. blood pressure function, you get a better idea of what will happen if you lower blood pressure than if you look at the all-cause mortality vs blood pressure function.

  7. dearieme Says:

    Again concerning Greg’s second link:

    (i) “Without resorting to any type of statistical modeling” – ha bloody ha: even a primitive bit of statistical analysis is statistical modelling. It reminds me of Monsieur Jourdain’s discovery that he has “been speaking prose all my life, and didn’t even know it!” But at least M. Jourdain realised!

    (ii) A second remark with an anti-intellectual flavour: “the value that the logistic splines analysis of Port et al …”. I’ve used splines myself (admittedly decades ago) but I have to say that the use of splines doesn’t seem to me to be central to the Port et al analysis – they are just a nifty tool. The remark is intended, I suppose, to imply to a readership of medics that Port et al are using some fancy-dan trick to pull the wool over their eyes. The key parts of that analysis are instead (to quote myself) three bits of common sense, namely (a) using all-causes death as their measure of [negative] merit, (b) categorising the data by age and sex, and (c) exploiting percentiles as a ‘natural scale’ for bp. (I say “common sense” in the reasonably precise sense of “that’s what an engineer or physicist would do.”)

    Is it too rude to suggest that engineers and physicists tend to be better with numbers than medics?

  8. Kim Øyhus, physicist Says:

    A feedback system failing, are what the curves in the article look like.

    I am thinking about negative feedback, where something tries to homeostase something else, and succeeds, but eventually fail, and blood pressure rises, perhaps as a consequence, or symptom, or direct cause.

    The curve is flat, with a slight increase, like a feedback system with low amplification and no integration, and the suddenly increases, like from a saturated negative feedback with too big input.

    It should be possible to map feedback systems like that from their effects.