Saturday, November 16, 2013

Welcome to the Harm/Hope Ratio: An Alternative to Industry Brainwashing

I have been mulling over several new posts, including one on the recently issued cholesterol treatment guidelines, and have come to the realization that underlying all of the messages I most want to convey is one basic point.

In keeping with the recent post:
--I have become increasingly impressed with the overall epidemiological picture painted by recent data on the harm caused by prescription drugs. We have reviewed Donald Light's data:
--showing that people taking their drugs as prescribed (not errors; not overdoses) is currently the 4th leading cause of death in the U.S. Just to up the ante a bit, we more recently encountered Peter G√łtzsche's statement that this should be elevated to the 3rd leading cause of death:

So the worrisome bottom line is that all this time, as we have been sold a bill of goods from the drug industry about how if we do one single thing to get in the way of their excessive profiteering, we'll suffer a screeching halt to medical progress, we have ignored the fact that the real "progress" in recent years is how many prescription drugs the average American is already taking and the harm done to them in terms of both mortality and morbidity from this over-drugging.

My friends in the evidence-based medicine movement have tried to chip away at this industry brainwashing particularly by attacking the commonly heard phrase "risk-benefit ratio." They point out quite reasonably that this term, which seems totally objective on the surface, actually hides a serious bias. We talk about "benefits" as if they were assured (no "risk" of benefit) but we only allude to harm as something that might or might not happen. The EBM gurus now say that we should always be precise and talk about the "harm-benefit ratio," unless we want to get really wordy and talk about the "probability of harm-probability of benefit ratio."

This leads me to one of my periodic modest proposals. I suggest, in light of the recent epidemiologic evidence, that we should now begin to talk about the "harm-hope ratio." I propose this term because:
  • As above we have solid epidemiologic evidence that the American public is killing ourselves by ingesting way too many prescription drugs.
  • When we do the usual studies, we are generally shocked by the high number-needed-to-treat (NNT) attached to commonly used drugs. The really great drugs like metformin for Type 2 diabetes usually run an NNT of around 10-30--that is, 10 to 30 patients have to take the drug for a certain length of time to achieve the hoped-for therapeutic benefit. The crappy drugs like statins for primary prevention of heart disease run NNT's typically in the several hundreds.
  • Therefore, the usual situation when we take prescription drugs--we can be almost certain that at least on the population level, harm is being caused; and we accept the harm given our hope that we might be the lucky soul who's the 1 out of 30, or the 1 out of 400, who will actually derive some significant therapeutic benefit, like not dying of a heart attack in the next few years.

Maybe if we were more honest and started talking widely about the harm-hope ratio, instead of the highly misleading risk-benefit ratio, we'd start to turn the proverbial ocean liner around.


Bernard Carroll said...

Howard, I think you have painted with a very broad brush here about Number Needed to Treat (NNT). This metric varies dramatically according to context of treatment.

In acute medical management, NNTs typically are low – signifying high benefit. Sumatriptan for migraine has a NNT of 2, while the better antidepressant drugs have a NNT of 5-6 for remission.

When we move to the context of prevention, NNTs are higher. Carotid endarterectomy has a NNT of 15 for preventing stroke or death in 6 years, while the NNT for preventing death over 10 years in treating hypertension ranges from 23 to 81. For prevention of heart attacks, the NNT for statins in 16-23 in patients with existing disease, and 70-250 in patients who only have risk factors.

Finally, when we move to the context of screening for disease, the NNTs go much higher. To prevent one death over 5 years the number needed to screen is 1374 for stool blood testing and it is 2451 for mammography.

I agree completely with your points about harm, and about the misleading term risk-benefit ratio. The harms and the benefits are seldom expressed in equivalent terms. How do we compare the harm of a drug-induced tardive dyskinesia, for instance, with the benefit of a modest reduction in depressive symptoms when antipsychotic drugs are used to “augment” antidepressant drugs in nonpsychotic patients?

Alice Dreger said...

Brilliant, Howard, just brilliant. Why aren't you our Surgeon General?