The Percival Effect
Ben Wolozin did it again. Whenever he and his colleagues turn to statins and epidemiology something unexpected happens. The data by themselves are stunning. Stunning because of the enormous sample size, stunning because of the extreme risk reduction in the simvastatin population, stunning because of the unexpected PD effect (maybe not as much unexpected to those who listened to Eliezer Masliah's talk in Salzburg). Moreover, the work shows how urgently more studies are needed to explore and validate these effects within prospective clinical trials—studies aimed to explore and validate the modes of action of statins in dementia, and studies that compare the effectiveness of different statins.

But the vast sample size came at a cost. Predominantly, the absence of a good control group, the increased likelihood of bias (because the database used does not reflect an average U.S.-American population), and missing resolution when it comes to the details. One example that highlights this is the cardiovascular risk group (CV). Statins are given to...  Read more

The Percival Effect
Ben Wolozin did it again. Whenever he and his colleagues turn to statins and epidemiology something unexpected happens. The data by themselves are stunning. Stunning because of the enormous sample size, stunning because of the extreme risk reduction in the simvastatin population, stunning because of the unexpected PD effect (maybe not as much unexpected to those who listened to Eliezer Masliah's talk in Salzburg). Moreover, the work shows how urgently more studies are needed to explore and validate these effects within prospective clinical trials—studies aimed to explore and validate the modes of action of statins in dementia, and studies that compare the effectiveness of different statins.

But the vast sample size came at a cost. Predominantly, the absence of a good control group, the increased likelihood of bias (because the database used does not reflect an average U.S.-American population), and missing resolution when it comes to the details. One example that highlights this is the cardiovascular risk group (CV). Statins are given to patients with what is considered to represent too high blood LDL-cholesterol levels. This factor alone introduces a major bias and even the comparison between the CV and statin users does not remove this bias. Obviously CV patients who aren't given a statin differ from those who are. This example could be extended endlessly, e.g., different dosages for different statins, different population profiles for different statins, etc.

At the end of the day the value of this work is not that it would provide us with new answers, rather that the value of this work is the new and important questions it raises. Sometimes a question can be more relevant than an answer. The tragedy of Percival on his quest for the holy grail was that he failed to ask the relevant question.

View all comments by Tobias Hartmann