Accepted answer
SURMOUNT-2 would have written it into the protocol, and the wording is the part that matters. Escalation protocols in this class generally permit a delay at the current level, sometimes a single step down with a later re-attempt, and count a participant as remaining in the arm throughout. That is analytically important: an intention-to-treat analysis keeps them at their randomised assignment regardless of the dose they were actually taking, so the "top dose" arm contains people who never reached the top dose. Find the protocol amendment history as well as the paper, because tolerability rules are among the things most often revised mid-programme, and dose-escalation decisions are made under supervision — nothing here is medical advice.
Start with the half-life, because the interval between steps should be at least four half-lives or you are escalating before the previous step has expressed itself.
Liraglutide titrates weekly rather than four-weekly because its half-life is about thirteen hours, so steady state is reached within days. The interval follows the pharmacokinetics in both cases.
Holding a step for eight weeks instead of four is a legitimate approach with no efficacy cost at the destination, and it is what the trials effectively did for anyone who tolerated poorly.
Trial data show gastrointestinal adverse events concentrated in the escalation phase and declining at stable doses.
Hold rather than escalate while symptoms are active. Always.