Androgen deprivation therapy is the standard treatment for advanced prostate cancer and it reliably produces the thing these drugs treat: weight gain, insulin resistance and cardiovascular risk. That makes men on it a mechanistically sensible group to study, and a retrospective cohort compared them against two other diabetes drug classes [1]. People arriving at these drugs while already carrying another diagnosis are the rule rather than the exception, as what else weight-drug users were already carrying sets out.
Against DPP-4 inhibitors the results look striking. All-cause mortality came in at a hazard ratio of 0.60 (95% CI 0.46–0.79), major kidney events at 0.63 and thrombotic events at 0.53.
That selection problem is acute in a cancer population. Nobody starts a new weekly injectable in a man whose cancer is advancing and whose prognosis is short, so a comparison between people who were and were not started on one is partly a comparison between people who were and were not expected to live. A mortality hazard ratio of 0.60 is roughly what that would produce by itself.
The comparison worth weighing is the second one, because SGLT2 inhibitors are an active and genuinely effective comparator rather than a near-neutral one. Against them, only mortality reached significance, at 0.76 with an upper bound of 0.99, and cardiac events, kidney events and thrombotic events all showed no difference. The advantage largely disappears when the comparison is against something that also works — the same pattern this desk found in the digital clinic analysis, where a crude comparison and a controlled one disagreed.
The authors say plainly that their neutral findings should be read cautiously because limited events may have reduced power, which is the correct reading of 659 and 1,009 patients per group. What survives is a reasonable hypothesis about a population with a real metabolic problem, not a demonstrated survival benefit. The broader question of what these drugs do during and after cancer treatment is covered in after cancer treatment, and the general difficulty of reading observational mortality figures in the cohort past eighty.