Saturday 5 January 2013

Evidence based medicine vs the Golden Rule

I am a big fan of evidence based medicine (EBM). Not the cookbook type (“do it this way or else”), but the idea that medicine is a science and therefore should be approached scientifically. We should use the principles of logic and rational thinking to reduce the errors that result from our often irrational, subjective “human” way of making decisions. Sometimes, however, we try to use EBM to justify something that doesn’t need scientific support – something that should be the default, and only changed if there is evidence against it. Something like the Golden Rule.

The Golden Rule states that we should do unto others as we would have them do unto us. Some claim that this originates from the bible, but Richard Dawkins and others tell us that it is innate and has been demonstrated in humans that have never heard of a god or the bible, and it has also been shown in other primates. This wiki link covers it pretty well, but basically it is pretty hard to find anybody that disagrees with the concept.

But medicine is full of rules that ‘sound’ good that end up being bad (like screening for diseases, tightly controlling blood sugar in diabetics, resuscitating newborn babies with oxygen instead of air, and unblocking narrowed blood vessels) and my blog is full of examples of why we should not accept things without good evidence just because they sound self-evident. But if I was to accept only one truth as being self-evident, it would have to be the Golden Rule. You would have to show me a reason not to apply it.

The example
This is an example of where the Golden Rule should be applied as the default, and yet we are running around in circles trying to find the evidence to support it instead of just accepting it and making sure there is no evidence against it.

For decades, those treating elderly patients with hip fractures have been demanding that patients have their surgery as soon as possible, using the association between later surgery and increased mortality as justification. Stating that the rate of mortality increases with delays to surgery may be an effective strategy to demand the extra resources to treat hip fracture sooner, but it is not true. It is an example of interpreting the data in a biased manner in order to fit our aims.

The literature on this comes from observational studies – there are no randomised trails on surgical timing for hip fractures. This prospective series showed no independent association between timing of surgery and mortality. Neither did this retrospective study. And neither did this large study, mainly because the authors of these studies allowed for the fact that sicker patients tend to be the ones that get their surgery performed later. This one didn’t either, but it did show that short-term mortality for this fracture is decreasing over time. Yet, this systematic review, and this massive ‘meta-analysis and meta-regression of over 190,000 patients’ concluded that early surgery (within 24-48 hours) was associated with a lower mortality.

So there is some evidence for an association between early surgery and better survival. Remember though, that an association does not mean cause-and-effect. For example, how do we explain the association when there is no increase in mortality from never performing the surgery in the first place? (Cochrane review here, go to page 26 for RCT evidence of operative vs non-operative treatment). And how do we explain the association when the in-hospital deaths usually occur peri-operatively and are often avoidable. These patients often die within 24 hours after their surgery, and from their surgery, rather than from waiting for surgery.

My take on the evidence
The evidence that surgery, at any time, saves lives in this group of patients is weak, and given that the mortality is not higher if they never have surgery, and that most patients die around the time of surgery, the association even lacks biological plausibility. As my regular readers know, biological plausibility is easy to find, and does not imply proof of cause and effect with any association, but if you can’t even get biological plausibility, then you are really struggling.

So why do surgery?
Surgery does not save lives, but it does result in shorter length of stay, better mobility and better independence, so when the doctor tells you that surgery for hip fractures is a life-saving procedure, they are overestimating the benefit of surgery.

Use the Golden Rule
Due to my opinion on the evidence above, some accuse me of suggesting that surgery for hip fractures should not be done early. That is not the case: I think that elderly patients with hip fractures should be treated as soon as practicable (when they have been medically cleared and theatre time and an experienced surgical team is available), preferably within 24 hours. Why? Not because it will influence mortality, but because that is the compassionate thing to do; it is what you would expect for yourself, for your elderly mother or grandmother, or any other person. Shouldn’t that be justification enough?

The bottom line
Trying to manipulate the evidence to support your position will inevitably weaken your position. Some things should not need such support when they can be aligned with simpler but more universal rules, like the Golden Rule. Unless there is evidence to the contrary, the Golden Rule should be the default. Policy makers should realise this logic instead of waiting for a systematic review that convinces them otherwise.


  1. Hi Dr Skeptic,
    I have recently been introduced to your blog, and I have thoroughly enjoyed reading your views on the subjects tackled. Knowing of your keen interest in "honest" research I thought you might be interested in an article I found in "Ethics and Health Law News" ( It seems many researchers do not follow the Golden Rule, and in many clinical situations the patient as a person, seems to have been forgotten.

    I look forward to your next article.

    Cheers, Arthur (retired surgeon)

    PS: I have just completed the Master of Health Law degree (Syd Uni)

    1. Thanks Arthur,

      The link is a good one, and publication of ALL results is currently the big push. We thought that trial registries were the answer, but that didn't work as planned. The overestimation of effect from under-reporting is well covered in Ben Goldacre's book Big Pharma (see my review: and for anti-depressants, see my blog here:

      Stay tuned.


Note: only a member of this blog may post a comment.