Did any trial in this class ever test a split weekly schedule, or a different dosing frequency at all?

Accepted answer

Nothing resembling a split weekly schedule was ever tested for the approved weekly agents. But dosing frequency has been compared head to head within a single incretin molecule, twice, and in both cases the less frequent schedule performed at least as well and was better tolerated. That is not proof about semaglutide or tirzepatide, but it is the only directional evidence that exists and it runs against the splitting argument.

The frequency comparison that was run

Exenatide existed in two formulations of the same molecule: an immediate-release version dosed twice daily and an extended-release version dosed once weekly. They were compared directly in a randomised trial, and the once-weekly formulation produced a larger HbA1c reduction — roughly −1.9 % against −1.5 % — with less nausea, on the order of 26 % against 35 % [1].

Read that carefully, because it is doing two things at once. The weekly formulation had a flatter concentration profile and less nausea and better efficacy. The splitting argument predicts that more frequent administration flattens the curve and improves tolerability. Here the less frequent administration flattened the curve — because the extended-release formulation removed the twice-daily absorption spikes — and improved everything. The lesson is not that frequency is irrelevant; it is that what matters is the shape of the exposure curve, and for a compound engineered to have a long half-life, less frequent dosing produces the flatter curve, not the more frequent.

That is the general principle. Splitting the dose of a compound whose half-life already exceeds the dosing interval does not flatten anything much, as the ratio arithmetic shows. Splitting the dose of a compound with rapid absorption and short elimination does — which is why twice-daily exenatide existed at all, and why it was superseded.

The frequency change that happened during development

The other relevant history is that semaglutide's obesity programme did not begin as a weekly programme. The phase 2 dose-ranging trial administered semaglutide daily subcutaneously, across doses from 0.05 to 0.4 mg per day, with liraglutide 3.0 mg and placebo as comparators, and reported weight reductions rising to roughly 13.8 % at the top dose over 52 weeks against about 2.3 % on placebo [2].

Note what 0.4 mg daily is: 0.4 × 7 = 2.8 mg per week, delivered in seven equal fractions. That is the maximally split version of a weekly schedule slightly above the eventual 2.4 mg weekly dose, and it is the closest thing to a split-schedule dataset that exists for this molecule. It produced roughly 13.8 % at 52 weeks. The weekly 2.4 mg phase 3 trial produced roughly 14.9 % at 68 weeks [3].

Those are not comparable numbers — different durations, different populations, different escalation schedules, no randomisation between them — and anyone who quotes them as a frequency comparison is over-reading badly. What they do establish is that a fully split schedule at a slightly higher weekly dose did not produce a visibly better result, and that the programme moved to weekly dosing without any efficacy or tolerability signal that argued against it. Development chose weekly for convenience and the data did not object.

Why the trial you want will never be run

Four reasons, in descending order of how decisive they are.

1. There is no device. The approved presentations in this class are single-dose autoinjectors and fixed-dose pens. A trial of half-doses twice weekly would require a bespoke presentation, manufactured, stability-tested and regulated, for a trial nobody is required to run. This alone ends it.
2. There is no commercial question. A sponsor runs frequency trials when frequency is a competitive axis. In this class the competitive direction is unambiguously less frequent — monthly and oral programmes — because that is what patients prefer and what wins prescriptions. A twice-weekly presentation is a commercial step backwards regardless of the result.
3. The predicted effect size is below the noise floor. A 15 % reduction in peak exposure, in a class with 30 %-plus between-person exposure variability, on an endpoint like nausea incidence with wide confidence intervals. Powering that comparison adequately would need a large trial to detect something the pharmacokinetics says is small.
4. The tolerability problem was solved a different way. The titration ladder is the intervention that addresses gastrointestinal symptoms in this class, and it works well enough that discontinuation for adverse events runs in the single-digit percentages in the obesity programmes. There is no unmet problem large enough to justify the device work.

So the evidence base is: no direct trial for these agents, one direct frequency comparison in a different incretin that went the other way, and one abandoned daily programme in the same molecule that showed nothing remarkable. That is thin, but it is not empty, and it points in one direction.