Accepted answer
Take it from the SURPASS-3 adverse-event table by arm, and check the unit before you use it. An incidence can be the proportion of participants who reported the event at least once, or the count of events divided by exposure time, and the two differ by however many people had it repeatedly. Then subtract the placebo arm, because the untreated rate is not zero. And read the discontinuation column beside it: an event that made people leave the trial is under-counted at every later visit, so a low late-timepoint incidence can mean the event was severe rather than rare.
Worth being precise here: a trial establishes what happened to a defined group under a defined protocol. Extending it beyond that group is inference, and inference is allowed as long as it is labelled.
Confidence intervals matter more than point estimates when two trials disagree. Two studies reporting fifteen and twenty per cent whose intervals overlap heavily have not disagreed about anything.
Relative to absolute, worked
| Quantity | Value | Derivation |
|---|
| Control-arm event rate | 8.0 % | From the trial table, not the abstract |
| Hazard ratio | 0.80 | Reported |
| Treated event rate | 6.4 % | 8.0 × 0.80 |
| Absolute risk reduction | 1.6 pp | 8.0 − 6.4 |
| Number needed to treat | 63 | 1 ÷ 0.016 |
| Relative risk reduction | 20 % | 1 − 0.80 |
The last two rows describe the same finding. Only one of them is used in headlines.
Non-inferiority and superiority designs are not interchangeable. A non-inferiority result says the new agent is not meaningfully worse against a pre-specified margin — it does not say it is as good, and it certainly does not say it is better.
Registry entries at ClinicalTrials.gov carry the pre-specified primary endpoint with a timestamp, which is the cheapest available check on whether an endpoint was changed after the data were seen.
When two sources disagree, the answer is almost always in the methods section of the one you have not read.