An introduction to evidence-based medicine

I'm Gordon Guyatt. I'm a professor in the Department of Clinical Epidemiology and Biostatistics and Medicine, McMaster University. I practice as a specialist in internal medicine doing hospital-based care. And for many years, I've been practicing an evidence-based approach to clinical care, and in this talk, I would like to give you an introduction to evidence-based medicine.

So, in the talk, I hope to persuade you about why you might bother with evidence-based medicine, and to do so, I'm going to talk about life before evidence-based medicine. Evidence-based medicine represents a solution. What was the problem? I'm then going to describe the three principles of evidence-based medicine as I see it and how to recognize an evidence-based medicine practitioner. I'm going to give you an example to how to judge the size of treatment effects and then a little story about helping the patient decide.

If it wasn't on the basis of evidence that we use to make clinical decisions, what was it? And I think to a large extent, it was expert recommendations.

The next slide gives some insight into possible limitations of expert recommendations. It is an old story of thrombolytic or clot busting therapy for patients with myocardial infarction. It's something called a cumulative meta-analysis. Down the center of the figure is 1.0 which means that thrombolytic therapy would neither increase nor reduce the likelihood of death after myocardial infarction. 0.5 would represent thrombolytic therapy cutting the rate of death after MI in half. 2.0 on the right side of that figure would mean a doubling of the death rate with thrombolytic therapy after myocardial infarction. The dots represent the best estimate of treatment effect as the data accumulated, and the lines around those dots represent a 95 percent confidence intervals, the range of plausible truth as the data were accumulating. The first trial of thrombolytic therapy enrolled only 23 patients, and as a result, had very wide confidence intervals. It was conducted in the late 1950s. The second trial enrolled not 65 patients but 42. This is a cumulative meta-analysis, each of those numbers represents the cumulative number of patients enrolled in trials up to that point. Up to seven trials and less than 2,000 patients, you still see that the confidence interval around the point estimate overlaps no effect. We're still uncertain at this point whether thrombolytic therapy is beneficial or not. At 10 trials and 2,500 patients, we see that the confidence interval no longer overlaps no effect. It's starting to look like thrombolytic therapy is beneficial. We could have a discussion about when the answer was in but most of us would probably say by 30 trials and over 6,000 patients. The lower boundary of the confidence interval is now quite away from no effect. We're getting very low p-values and it looks like thrombolytic therapy reduces the death rate after myocardial infarction by about 25 percent. Did this stop people from doing randomized trials of thrombolytic therapy? No, it did not. And indeed, there were another 40,000 patients enrolled after the answer was in, half of whom did not receive the life-prolonging benefits of thrombolytic therapy. We end up with a very narrow confidence interval, unnecessarily narrow, and half the patients in those trials had paid the price by not receiving the benefits of thrombolytic therapy. Why then did we have to enroll another 40,000 patients after the answer is in? I think a large part of the answer to that question is on the right side of this slide which presents textbook and current review recommendations which were being made while these data were accumulated. And the categories are recommending that thrombolytic therapy given routinely, specific indications, calling it experimental treatment, or not even mentioning thrombolytic therapy. And two things I'd like to point out here, one is the divergence in expert recommendations as these data were accumulated. So, you see, in the latter part of the 1980s, some experts are saying routine, some specific indications, some calling it experimental, some not even mentioning it. And secondly, it's a decade after the answer is in that you finally are approaching a consensus among the experts and that is a very large part why people had to keep doing the trials to finally convince the experts.