Dose-Limiting Toxicity (DLT)

A dose-limiting toxicity is defined in advance, not judged afterwards. The protocol names which events, at which severity grade, in which organ systems, occurring inside which observation window, will count. Anything outside that definition, however unpleasant, does not stop escalation. The maximum tolerated dose is derived from the resulting rate: in a conventional three-plus-three design, escalation continues while cohorts stay clear, and the maximum tolerated dose is the highest level at which the observed rate stays below target, conventionally about one in three.

The concept comes from cytotoxic oncology, where more drug meant more tumour kill and toxicity was the only ceiling. Metabolic and hormonal peptides behave differently. In incretin escalation studies the ceiling has repeatedly been tolerability rather than organ injury, with nausea and vomiting limiting how fast and how far dose can rise, which is why approved products escalate in steps over weeks rather than starting at target. A tolerability ceiling moves with the escalation schedule; a true toxicity ceiling does not.

The consequence is that a maximum tolerated dose is not a recommended dose. For chronic non-oncology treatment the useful dose usually sits well below it, on the flat part of an exposure-response curve. The FDA Project Optimus initiative exists precisely because dose selection driven by the maximum tolerated dose has repeatedly produced approved doses higher than necessary.

Where this misleads is in reading no dose-limiting toxicity observed as evidence of safety. A phase 1 study of a few dozen participants cannot detect an event occurring once in a thousand exposures, and a window of weeks cannot see toxicity that is cumulative or delayed by months. Escalation beyond studied doses outside a trial inherits none of the monitoring that made a DLT observable in the first place.