How to read a longevity study

1. Who — or what — was studied?

The single most important line in any paper is the one saying whether the subjects were humans, mice, worms or cells in a dish. Most longevity headlines trace back to animal or cell work. That research is valuable — it generates hypotheses — but it is not evidence that anything works in people. If the answer is not humans, you can usually stop reading the coverage there.

2. What was actually measured?

A study can only support claims about what it measured. If it measured NAD+ concentration, it supports a claim about NAD+ concentration — not energy, not ageing, not lifespan. The gap between surrogate markers and outcomes that matter is where most supplement marketing lives. Lifespan itself is almost never measured in human trials, because you would need decades; when someone claims human lifespan evidence, ask what was really counted.

3. How was it designed?

Randomised controlled trials can establish cause and effect; observational studies can establish association. The Finnish sauna cohorts are excellent science, followed for decades — and still cannot prove sauna causes longer life, because frequent sauna users differ from rare users in fitness, wealth and habits that statistics only partly capture. Meanwhile, a randomised trial answers only the exact question it asked: TREAT tested one fasting schedule for twelve weeks; that is what it can speak to.

The sauna evidence →The fasting evidence →

4. How big, and how long?

Small trials produce noisy results in both directions; short trials cannot detect slow effects or slow harms. There is no magic threshold, but be increasingly sceptical below a few dozen participants per group, and treat any trial under three months as preliminary for anything ageing-related.

5. Has anyone else found the same thing?

Single striking results frequently fail to replicate — not through fraud but through chance, publication bias and the file-drawer problem. The rapamycin immune-function story is instructive: promising randomised results in 2014 and 2018 were followed by a phase 3 programme that failed to confirm clinical benefit. Any grading system that had rewarded the early results with top marks would have been wrong.

The rapamycin evidence →

6. Who paid, and who benefits?

Industry funding does not make a study wrong — REDUCE-IT was manufacturer-funded and remains a major trial — but funded studies are known, on average, to reach sponsor-favourable conclusions more often. Look for the funding statement and the authors' declarations, and weight replication by independent groups accordingly.