Raising a biomarker does not necessarily cause disease

A change in a biomarker is not sufficient, on its own, to establish that disease has been caused or prevented. This is a settled principle of clinical epidemiology, resting on the distinction between surrogate endpoints (measurable intermediates such as LDL cholesterol, HDL cholesterol, blood glucose, or heart rhythm) and hard clinical outcomes (heart attacks, deaths). Two well-understood facts support it. First, a statistical association between a biomarker and a disease can arise from confounding or reverse causation, so an elevated marker need not be a cause. Second, and more decisively, there are documented cases in which an intervention moved a biomarker in the "right" direction yet failed to help, or even harmed, patients: antiarrhythmic drugs suppressed abnormal heartbeats but increased mortality in the CAST trial; drugs that raised HDL cholesterol (such as torcetrapib and, in outcome trials, niacin) did not reduce cardiovascular events. Because the claim is framed with "not necessarily," a single such counterexample is enough to establish it, and many exist.

The practical upshot, and the context in which the claim was raised, is that a randomized trial showing an intervention changes a biomarker does not by itself settle whether that intervention changes disease risk; a marker is a valid surrogate only when trials show its changes reliably track hard outcomes. The claim says nothing about any particular biomarker being non-causal: some markers are excellent surrogates. It asserts only that biomarker change and disease causation are not the same thing, which no informed party disputes.