When I was a medical student, I did a lot of rotations at the Boston VA in JP. I loved my patients there -- they were patient and kind and stoic. One of the best rotations I did was Hematology, where Lou Fiore was my preceptor. Lou was not only an excellent teacher, but also a terrific doctor and a good human being all around. He used to start our days together by saying, "I'm gonna teach you one thing today." And teach us he did, at least one thing per day. Now I teach. And on occasion I have used the Lou Fiore "I'm gonna teach you one thing today" promise. Well, today is one of those days: I'm gonna teach you one thing.
And here is that thing. I am sure I am not the first one to notice this, but I still think of it as the "Zilberberg rule of thirds." The gist of it is that, for clinical research purposes, one can think of patient populations crudely in thirds: there is one third who are too sick to benefit from any of our interventions, there is one third who are too healthy, so that no matter how we try to tweak, their outcomes will not change, and the middle third, which comprises the "sweet spot" for intervention. So it is a fool's errand to pursue proof of concept studies in either of the bracketing thirds, since it is only the middle third that is likely to show a signal.
Pharmaceutical manufacturers do not always appreciate this trichotomy. Look at Vioxx, for example: when used in patients who were essentially healthy, an unacceptable safety signal arose that drove the drug off the market. Same for SSRIs, where the ill-conceived enthusiasm for treating marginal depression cases seems to be debunking the entire serotonin hypothesis. The flip side is sepsis research: septic shock patients are so far gone that it is difficult for any single therapy to alter their outcomes. Just look at the Xigris story, as well as myriad other therapies that tried and failed. This is the rule of thirds at its most pronounced.
In HEOR the rule of thirds holds as well. To prove cost effectiveness the following questions need to be asked:
1. Is the disease in question prevalent?
2. Is the economic impact of the disease known and substantial?
3. Does the diagnostic/therapy in question alter the course of the disease in such a way as to be significant?
If the answer to any of the questions above is "no," you really need to think carefully about the value proposition.
Some of you will bring up the inter-individual differences, the heterogeneous treatment effect, etc. And yes, these are supremely important. However, though the framework I propose here is simplistic, we have to start somewhere. To be sure, there is a more nuanced approach to this beast, but generally, one will not go wrong by asking these questions before committing huge resources to a project, particularly if the answer to question 2 or 3 is a resounding "no." So, even in health economics it behooves one to know the Zilberberg rule of thirds: choose the right population where the diagnostic/therapeutic advance and its costs can be justified by a substantial gain in the outcomes.
And that is your one thing for today.