This paper asks whether we can more plausibly extend trial results to a different population by assuming the treatment’s relative effect is stable, rather than assuming the whole outcome distribution is the same. The authors develop identification results for failure-time outcomes (time until an event, like death) under an assumption they call risk ratio transportability. They then apply the idea to lung‑cancer screening data to see how the methods work in practice.

At a high level, the method separates two pieces of information: the baseline risk in the target population and the relative effect of the intervention estimated in the trial. The trial used here is the National Lung Screening Trial (NLST), which enrolled people aged 55–74 who were current or former smokers with at least 30 pack‑years of smoking between August 2002 and April 2004. The target population comes from the 2005 National Health Interview Survey (NHIS), restricted to people who would have been eligible for the NLST, with follow-up truncated to six years to match the trial. The researchers combined individual data from the NLST and NHIS, using common baseline variables such as age, sex, smoking history, body mass index, and several comorbidities.

Typical methods for transporting trial results often assume distributional transportability: that the full distribution of outcomes is exchangeable between the trial and the target population after adjusting for measured covariates. That assumption can be hard to justify. The authors propose instead to assume exchangeability of the relative effect on the risk ratio scale. A risk ratio compares the risk under one strategy to the risk under another. If that ratio is stable across populations (after adjustment), one can apply the trial’s relative effect to the target population’s baseline risk to estimate outcomes there. The analysis used a discrete‑time setup, splitting follow-up into quarterly intervals up to six years, and focused on the intention‑to‑treat effect of three annual low‑dose computed tomography (CT) screens versus control.