(we are making an exception to our ‘paywall’ policy due to the wide dissemination of the WaPo story discussed below)
The Enbrel Imbroglio: Finally, Some Genuinely ‘Fake News’
“Pfizer had clues its blockbuster drug could prevent Alzheimer’s. Why didn’t it tell the world?” –Rowland, C. Washington Post 6/4/19
The term ‘fake news’ has been irresponsibly and numbingly hurled around, generally in the service of denial and obfuscation. But while the exception by no means proves the rule, this piece by the usually-highly-responsible Washington Post does fall into the category. Many readers of this Neurogram will already be familiar with the premise of the story, which boils down to this: A post hoc epidemiological study of insurance data found that the proportion of Enbrel users (110 out of 127,000) in an Alzheimer’s population was lower than that proportion of Enbrel users (302 out of 127,000) in a non-Alzheimer’s population. Which was taken as evidence that Enbrel can ‘prevent’ Alzheimer’s, but that Pfizer chose not to pursue it as a treatment because of its nearterm loss of patent exclusivity, choosing profits over the greater good.
It is a story that is compatible with the common vilification of the Pharma industry as an engine for greed, but is not congruent with some basic principles of data analysis, starting with this one:
Correlation does not equal causality.
We are vaguely curious as to how many hundreds of variables were screened in the comparison of these two populations, populations defined by virtue of the diagnosis codes filed with insurance carriers, a criterion that surely did not include anything approaching the current standard of practice, which defines Alzheimer’s via PET or CSF evidence of amyloid pathology. This kind of fishing expedition is sure to catch
something that looks like an association, we would also be curious as to how many other variables showed a numerical contrast of this seeming magnitude. But the leap to presume a causal relationship between Enbrel use and an Alzheimer’s outcome, based on this correlation, is not only unjustified, but is incorrect.
One could in theory choose to use such an association as the basis for a prospective, double-blinded, placebo-controlled trial, but that is not a decision to be taken lightly. The notion that one can run a 3000-4000pt Alzheimer’s study for “$80 million”, as was suggested in the article, most likely was a shock to every company that has run a major Phase III Alzheimer’s program, where the tab tends to be in the half-billion dollar range, and up. The fact that the patent life of the drug in question was dwindling is not irrelevant to the calculation of whether or not to proceed, but hypothesizing that this was the only factor driving Pfizer’s (and Amgen’s as well) decision is to overlook the likelihood that this ‘finding’ was due to statistical noise, and would not have been replicated in a $500 million pivotal program.
We do believe that this population data should be made available for review by responsible academic and NIH entities (Pfizer apparently did make it available to at least one external researcher) in keeping with the principle of data transparency that we believe should be applied to human drug trial data, both positive and negative (acknowledging that this was not a drug trial per se).
NIR has probably been more critical of Pfizer in recent years than has been any other appraiser of the biopharm scene. The Company engaged in a cowardly gutting of its once-formidable neuroscience R&D group, and grotesquely over rewarded Ian Read’s mediocre performance as CEO. But they are getting an unfair rap here: The idea that this correlation in and of itself would mandate a major prospective trial is an underinformed oversimplification, perpetuating an equally oversimplified caricature of the pharma industry.