Research methods · Measurement

Surrogate vs hard outcomes

You take a pill for twelve weeks. Your follow-up panel comes back and a marker has dropped — fasting insulin, an inflammation reading, whatever the bottle promised. The brand will frame that drop as the win. But the drop is a number on paper. What you actually wanted was something the number can't show you on its own: fewer heart attacks, no diabetes diagnosis, more years alive. Those are hard outcomes — things that happen to you. The number is a surrogate: a stand-in, a marker someone hopes is tracking the outcome. Sometimes it is. Often it isn't. The whole game is telling the two apart before you pay for the difference.

Why a marker can lie

A surrogate is a shortcut. Hard outcomes take years and thousands of people to measure, so researchers reach for something faster — a value in the blood, a reading on a scan — and assume it stands in for the real thing. The shortcut is only honest if the marker sits on the causal path: change it, and the outcome changes with it. The trap is that a marker can be correlated with disease without causing it, or it can be one of several roads to the outcome — so you pave one road and the disease takes another.

LDL cholesterol is the rare marker that mostly earned its place: lower it by several independent methods, and heart attacks reliably fall, so the link from marker to outcome holds. Most markers never get that test. HDL — the "good cholesterol" — predicts heart disease beautifully, yet drugs that raise it have failed to prevent heart attacks; raising the number didn't buy the outcome. Most inflammatory markers, "biological aging" clocks, and NAD+ levels are in the same unproven bucket: real associations, no validated proof that moving them on purpose changes how long or how well you live.

Work one, then finish one

Worked: the CAST trial, late 1980s. Survivors of a heart attack often have extra, irregular heartbeats on an ECG — a marker that predicts dying. The reasoning seemed airtight: give drugs that suppress those beats, smooth the ECG, save lives. The drugs worked on the marker; the irregular beats fell exactly as intended. Then the trial counted the hard outcome — deaths — and the people on the drug were dying at more than twice the rate of those on placebo. Moving the marker the "right" way was killing people. The marker predicted death but wasn't the lever; pushing it directly did harm no one saw coming until they measured the outcome itself.

Your turn: a supplement brand reports that, over twelve weeks, its product improved HOMA-IR — a marker of insulin resistance. The caption: "clinically shown to prevent diabetes." What's the gap? (HOMA-IR is a surrogate, not a hard outcome. A hard outcome would be fewer actual type 2 diabetes diagnoses over years. Moving the marker is unproven to move the disease — and as CAST showed, a moved marker can even run the wrong way. The "prevents diabetes" claim isn't supported by what was measured.)

Why this matters

Stand in a supplement aisle, or scroll a longevity podcaster's clip, and notice what the evidence actually is. Almost every claim is a moved marker: this compound lowered inflammation, that protocol improved your biological age, this powder raised NAD+. Practically none of it rests on a hard outcome — a trial showing the people who took it got sick less or lived longer than the people who didn't. The marketer's move is to show you a real, measured change in a number and let you supply the leap to "so I'll be healthier." That leap is exactly the link CAST proved can be broken. Before you spend money or rearrange your life around a marker, ask one question: has moving this number, on purpose, ever been shown to change the thing I actually care about? If the honest answer is "no one has checked," you're buying a number, not a longer life.