Answering the narrow version, because the broad one does not have a single answer. Read four things before the headline number. The population, because trial populations are selected and supported in ways that real cohorts are not. The comparator, because "better than placebo" and "better than the current standard" are different claims and get reported identically. The primary endpoint as pre-registered, because a secondary endpoint promoted after the fact is a hypothesis rather than a finding. And the completion rate, because a large effect in the half of participants who finished is a different result from a large effect in everybody enrolled.

DataDave said:

Read four things before the headline number.

Bayesian meta-analysis perspective on the trial evidence: traditional frequentist meta-analyses report point estimates and confidence intervals. Bayesian approaches provide probability distributions that are more intuitive for clinical decision-making.

For example: "There is a 98.5% probability that semaglutide 2.4mg produces >10% weight loss vs placebo" is more actionable than "RR 3.4, 95% CI 2.8-4.1, p<0.001."

The the trial evidence evidence is strong under both frameworks, but Bayesian analysis better communicates the degree of certainty for individual patient counseling.

KarenAZ_mom said:

I keep finding that the number in the press summary and the number in the paper are not the same number, and the difference is always in the same…

Same position here, arrived at the long way round. Relative and absolute effects need reading together. A 20% relative reduction on a high baseline risk is a large absolute benefit; the same relative figure on a low baseline risk is a small one, and press summaries almost always quote the relative number because it is bigger.

Adding the clinical framing, because it changes how the question reads.

KarenAZ_mom said:

...regarding the trial evidence...

I think this is an underappreciated point. To expand on it with some data:

A recent meta-analysis of 22 RCTs (n=18,900) found that the trial evidence was associated with a consistent effect size across diverse patient populations[1].

The NNT was 20, which is comparable to statins for secondary prevention. That's a strong clinical argument for this approach.