Abstract
<title>Abstract</title> <p>Background Regulatory precedent research—surveying prior approvals, labels, review documents, and trials to inform a Target Product Profile (TPP), Target Product Label (TPL), or value dossier—has no defined endpoint. The base grows and decays while work proceeds, so "comprehensive" is an aspiration, not a decision, and the regulatory intelligence (RI) literature defines no stopping rule. Methods We mapped the RI and TPP literature against four mature stopping frameworks—Value of Information (VoI), data/meaning saturation, capture–recapture horizon estimation, and living-review update triggers—and synthesized them into a decision-responsive rule organized by concentric rings. As a proof of concept we applied the rule to two glucagon-like peptide-1 (GLP-1) receptor agonist decisions—a cardiovascular (CV) and a chronic kidney disease (CKD) claim—checking its inputs against live regulatory data. Results The rule stopped at 60% of the CV corpus, where the mechanism class was self-sufficient, but ran to 100% of the CKD corpus, where the standard had been set by neighbouring classes. A third example, from a real anonymised patient-reported-outcome engagement, reproduced the periphery pattern. A volume-only rule and a lower-bound coverage estimator stopped prematurely in every case, reading approximately 95%, 92% and 97% against true values of 80%, 76% and 41%. Live label histories corroborated the chronology: semaglutide gained CV (2020) then CKD (2025) indications five years apart, the CKD paradigm set earlier by sodium-glucose cotransporter-2 (SGLT2) inhibitors. Conclusions A defensible stopping rule must respond to the decision, not a fixed budget: VoI-ranked rings, code and meaning saturation tracked separately, capture–recapture, and update triggers.</p>