Accepted answer
Stated carefully, start by separating chemical degradation from physical degradation, because they fail differently and they are detected differently. Chemical degradation changes the molecule and shows up as new peaks on a chromatogram. Physical degradation aggregates the molecule and often shows up as nothing at all on reverse-phase HPLC, because the aggregate never makes it onto the column.
Freeze-concentration is the mechanism people miss. As ice forms, everything that is not water is excluded into a shrinking unfrozen fraction, so the local concentration of peptide, buffer salts and preservative rises sharply. If the buffer components crystallise at different rates, local pH can shift by more than a unit. That is why a phosphate-buffered solution can behave badly on freezing while an unbuffered one is fine.
Reported and extrapolated stability by condition
| State | Condition | Usable window | Basis |
|---|
| Lyophilised solid | −20 °C, sealed, dry | 24–36 months | Supplier guidance |
| Lyophilised solid | 2–8 °C, sealed | 12–24 months | Supplier guidance |
| Lyophilised solid | 25 °C, sealed | 4–8 weeks | Extrapolated (Arrhenius) |
| Lyophilised solid | 40 °C, sealed | 1–2 weeks | Extrapolated |
| Solution, preserved | 2–8 °C | 28 days | USP microbiological convention |
| Solution, preserved | 25 °C | 3–7 days | Extrapolated |
| Solution, unpreserved | 2–8 °C | 24 hours | USP microbiological convention |
Windows for the solid state are chemical; windows for solution are microbiological and usually shorter than the chemical limit.
Worth being precise here: the 28-day figure for a reconstituted preserved vial is microbiological, not chemical. Chemically, a well-behaved peptide at 5 mg/mL at 4 °C will typically lose well under a per cent of content per month. The reason to respect the date is bioburden, and bioburden is a function of how many times you have opened it, not of the calendar.
The licensed semaglutide and tirzepatide presentations carry in-use periods of several weeks at room temperature in their labelling, which is the closest thing to real stability data in this space — and it applies to a buffered, surfactant-containing, preservative-containing formulation, not to a reconstituted research vial.
I would be careful about generalising across sequences. Stability is sequence-specific, and a rule derived from semaglutide will not transfer cleanly to a tri-agonist with different residues in different local environments.
The single highest-value change most people can make is buying a cheap logging thermometer, because it converts an assumption about their storage into a record.