Statins and Bisphosphonates Failed to Replicate

Viagra and statins may slow cancer metastasis—findings drawn from laboratory work showing these drugs interfere with cellular mechanisms that allow tumors to spread.

Newspapers have run versions of this story before, and they will run it again. We mistake this pattern for scientific progress even though history teaches us otherwise.

In 1996, epidemiologists noticed something suspicious in their data. People taking statins seemed to develop fewer cancers. The effect was consistent enough to be real, spawning dozens of clinical trials and a decade of serious investigation by oncologists who had legitimate reason to believe a widely available drug might have been hiding a second life as a cancer preventive.

The trials mostly failed. Some showed modest benefits in specific populations, but nothing remotely large enough to change clinical practice. The signal that had looked so clear in observational studies dissolved once researchers controlled for confounding variables—the underlying health behaviors and socioeconomic factors that make certain people both more likely to take statins and less likely to develop aggressive cancers in the first place. Bisphosphonates followed a nearly identical trajectory, with bone density drugs showing anti-cancer associations before trials were mounted and mechanisms proposed with unwarranted confidence.

The gap between "this protein responds to this compound in a petri dish" and "this will help a patient" is not a small problem that time and money solve.

What makes this cycle repeatable is not the scientists' incompetence but the structure itself. When you observe a million people over time, you find hundreds of statistical correlations that are real but useless—the correlation is true, the causation almost never there. The gap between "this protein responds to this compound in a petri dish" and "this will help a patient" is not a small problem that time and money solve. It is an abyss, and it has been an abyss for decades.

The Viagra story will now collect the same earnest mechanism papers, the same preliminary evidence, the same careful press releases about "promising directions" that filled the statin literature. Some researchers will spend five years testing it. Most will be right to be suspicious. But the structure that rewards finding these patterns—journal editors want novelty, funders want breakthroughs, institutions want discoveries—means the next incidental correlation will look exactly the same way this one does. The real skill is learning to see the shape of failed signals before you commit your belief to them, to notice when you're pattern-matching to hope rather than to evidence.