Evidence and medicine do not always run together

Evidence-based medicine (EBM) means basing medical practice on scientific evidence. Its ascendency is disturbingly recent. 

Yogi Berra — the baseball player, not the cartoon bear — identified the commonest complaint about evidence-based medicine many decades ago. “In theory, there is no difference between theory and practice. In practice, there is.”

I practice evidence-based medicine. At least, I recently attempted to. Twice.

The first time was my final encounter of a tiring 50-patient day. I wasn’t his usual GP. He complained about widespread aches and pains, seeking reassurance he wasn’t having a heart attack. 

I eased his concerns but recommended he keep an existing appointment with a heart specialist the next day. I wondered aloud if his aches might be caused by his cholesterol-lowering medication, known as statins. Muscle aches are a very common, and still under-recognised, side effect of statins. 

I wrote the words, “number needed to treat”, on a piece of paper and handed it to him. I explained the number needed to treat was how many people in his situation needed to take a statin to prevent one heart attack or stroke over, say, the next five years. 

If acronyms and statistics make your eyes glaze then the next two paragraphs might be challenging. But you’ll be okay. Trust me, I’m a doctor.

This is the key concept: The number needed to treat (NNT) is the inverse of the absolute risk reduction (ARR), which is of course totally different to the relative risk reduction (RRR).  

If, for instance, a drug reduces the incidence of stroke from 2% to 1%, then the RRR is 50% (because 1% is half of 2%), the ARR is 1% (because 2% minus 1% equals 1%) and the NNT is 100 (because the inverse of 1% or 1/100 is 100).

ARR is critically important. In contrast, RRR is deceptive drivel found in advertisements, media releases, sloppy medical articles, and the occasional National Prescribing Service handout. You generally hear about RRRs.

But I digress. I also briefly explained the concept of number needed to harm (NNH). Assuming the magnitudes of harm and benefit are comparable, then the NNT of any treatment must be less than its NNH. If the NNT is greater than the NNH, then the treatment will harm more people than it helps.

I suggested my patient ask the cardiologist for his statin’s NNT. He was taking it for primary prevention, having had no previous heart attacks or strokes. In his circumstances, the NNT was very likely to be over 50. In other words, there was less than one chance in 50 the tablet would do him any good within the next few years and it might well be harming him.