Clinical Takeaway
| Domain | Key finding | Evidence level |
| The mystery | A gene variant carried by ~10% of people produces higher circulating GLP-1 yet worse metabolic disease — the cleanest case of true GLP-1 resistance ever documented. | High (human + murine + meta-analysis) |
| Real resistance | PAM loss-of-function carriers lose 44% of the HbA1c benefit on GLP-1RA — but respond normally to metformin, sulfonylureas, and DPP-4i. Pathway-specific, written in the genome. | High |
| Mostly mislabeled | Almost no “non-responder” has this. Most are under-titrated, non-adherent, on the wrong agent, or in the left tail of a continuous distribution (10–17% lose <5%). | High (STEP/SURMOUNT) |
| Plateau ≠ resistance | The ~60–72-week stall is a defended energy-balance setpoint, not a dying drug. Regain on stopping is the disease reasserting — proof the drug was still working. | High (regain meta-analyses) |
| Clinical move | Before writing “non-responder”: confirm dose, adherence, and time; reset expectations to the distribution; switch agents before abandoning the class. | Practice point |