Cohort study

A cohort study follows a large group of people over years, recording what they do and what happens to them, without assigning anyone to anything. Most of what is known about how habits relate to lifespan comes from studies of this kind, because nobody can randomise thousands of people to walk a certain amount for a decade.

Its strength is scale and realism. A cohort can include tens of thousands of people living ordinary lives, and it can run long enough for all-cause mortality to become measurable — something almost no trial can afford. The 2022 meta-analysis that produced the best current figures on daily steps pooled 15 such cohorts covering 47,471 adults and 297,837 person-years of follow-up.

Its weakness is fixed and cannot be analysed away. People who walk more differ from people who walk less in many ways at once, including ones nobody recorded. Researchers adjust for what they measured — age, smoking, income, existing illness — and the differences they did not measure remain. A particular problem in mortality research is reverse causation: illness reduces activity before it kills, so the least active group contains people who were already dying.

Pooling many cohorts increases precision without changing this. A meta-analysis of observational studies produces a confident estimate of an association, and an association is what it remains. This is why a cohort sits below a randomized controlled trial on the evidence ladder even when it is a hundred times larger.