Thinking about treatment effects, price signals and F. A. Hayek

I was recently interviewed on The External Medicine Podcast (here and here) by brothers and medical students Daniel and Mitch Belkin. They asked a bunch of great questions and we explored a range of topics important to me; the closest being treatment effects and more specifically, what they tell us and what they don’t.

They caught me by surprise, when early in the interview, they asked me to talk about a blog post I wrote many years ago entitled “A Hayekian Defense of Evidence-Based Medicine.” I wanted to explore that here.

Evidence-based medicine or EBM, is a systematic way to appraise medical research to arrive at objective judgements about its quality, clinical applicability, and knowledge gaps. It holds that randomized controlled trials provide the highest level of evidence for determining a treatment’s effectiveness.

First, what does “treatment effect” in a randomized trial mean? It is the ratio of events or outcomes that are observed for one group, who receive an intervention (e.g., drug, surgery, device, etc.), compared to another group that either does not receive the intervention or receives a different intervention. It can be expressed in relative or absolute terms.

Imagine 20,000 patients are randomly assigned to receive either Drug X or a placebo pill so that there are 10,000 patients in each group. In the group receiving Drug X, 100 patients die (1% of patients) and in the group receiving the placebo, 200 patients die (2% of patients). The treatment effect of Drug X in this case can be expressed as a relative risk reduction of 50% (1% ÷ 2% = 50%) or as an absolute risk reduction or risk difference of 1% (2% - 1% = 1%).

As you might imagine, how a treatment effect is expressed (relative versus absolute terms) makes a big difference on peoples’ perceptions and significantly influences decision making but is beyond the scope of this essay (will talk about this in a future post).

A final way to express a treatment effect is in terms of number of patients needed to treat to prevent one patient from experiencing the outcome of interest, which is also known as number needed to treat. It is determined by dividing the number 1 by the absolute risk difference (i.e., 1 ÷ 0.01 = 100).

For the case above, strictly as it pertains to the patient population enrolled in the trial, the treatment effect can be expressed in 3 ways:

Drug X reduces the relative risk of dying by 50% compared to placebo

Drug X reduces the absolute risk of dying by 1% compared to placebo

If you treat 100 patients with Drug X instead of a placebo, you will prevent 1 person from dying

This is powerful information that tells us something very real about Drug X that we could not have predicted without performing the clinical trial.

Biologic reductionism, while important for facilitating the discovery of new interventions, is completely insufficient to tell us anything about their effectiveness in a population of human beings. This is why I associate treatment effects from clinical trials to price signals in a free market economy.

The analogy is simple, treatment effects like prices are impossible to predict. Thus, my reference to Friedrich Hayek and the price system.

Here is what I wrote in my original post:

“In The Use of Knowledge in Society (1945) F.A. Hayek explained how prices communicate information. To state his theory in an over-simplistic way: there is a lot of knowledge out in the economy (this is sometimes referred to as the knowledge problem) and there is no way to collate all of it into a single place or central body and thus, economic calculation or central planning will usually fail to achieve its desired ends. Better economic order arises based on information from price signals [in a free market]. Only prices can communicate what is ultimately valuable to consumers and entrepreneurs. Value roughly equates to truth and in the Austrian economic sense it is an unequivocally subjective property. Thus, prices serve to share and synchronize local and personal knowledge and values, allowing society’s members to achieve diverse and complex ends through a principle of spontaneous organization.”

In the same sense that economic planners lack sufficient knowledge to best coordinate economic activity and arrive at prices, physicians and medical scientists lack sufficient knowledge to predict the success of biomedical interventions in human beings. Also, from the original post:

“Like the economy, the human body is a complex system composed of 100 trillion individual cells – the number of chemical reactions occurring at any one time is too numerous to count and furthermore, each human being is different. Its complexity multiplied hundreds of billions of times. Like the knowledge problem that exists in economics; the same could be applied to the complexity of managing disease and preventing illness [predicting the success of treatments].”

Medicine is full of interventions that have been instituted on the basis of biomedical reasoning only to be disproven in clinical trials (i.e., medical reversals). There is a series of classic articles as well as a book by physicians Vinay Prasad and Adam Cifu dedicated to the study of medical reversals in contemporary practice.

The recent pandemic provided a plethora of cases including: 1) early intubation, which probably increased death; 2) therapeutic anticoagulation; 3) hydroxychloroquine; 4) withdrawal of renin angiotensin aldosterone system inhibitors; 5) lopinavir; 6) colchicine; 7) azithromycin; and 8) ivermectin.

My personal bias, as a medically conservative physician, is generally to avoid interventions that lack data from randomized trials. Why? Because I lack confidence in their efficacy regardless of the biologic rationale they are based on and I would rather err on the side of omission than commission. Hopefully, that much has been made obvious. But that is just the tip of the iceberg. In future pieces I intend to explore a range of topics related to inferences drawn from clinical trials and how I judge the validity of such inferences along with how I incorporate data from clinical trials into my own practice as a general cardiologist.