Abstract
<title>Abstract</title> <p>Many chronic conditions involve interacting symptom, functional, and physiological domains measured repeatedly over time. Randomized trials frequently summarize efficacy using baseline-to-endpoint contrasts in a designated primary domain. When the clinical question concerns sustained multidomain stabilization rather than single-domain magnitude change, it is uncertain whether scalar endpoint summaries uniquely determine the estimand of interest. We represent trial outcomes as multidomain longitudinal trajectories and define prespecifiable trajectory-based estimands, including coherent multidomain response, sustained stabilization, and a fragmentation index describing oscillatory temporal organization. We establish structurally that scalar baseline-to-endpoint contrasts are generally non-injective mappings when at least two domains and three measurement occasions are present. We then illustrate the implications using published summary statistics from three chronic randomized trials (anxiety, chronic low back pain, and COPD), and conduct a Monte Carlo study to estimate how often endpoint equivalence can coexist with divergence in trajectory-based estimands under plausible longitudinal dependence structures. All illustrations use published summary statistics only; no individual participant data were accessed. Distinct multidomain longitudinal organizations can produce similar scalar endpoint contrasts while differing in sustained stabilization and risk of short-term deterioration. Across three disease areas, endpoint similarity in a primary domain coexisted with divergence in multidomain response patterns. In simulations reflecting realistic cross-domain correlation and temporal autocorrelation, endpoint equivalence frequently co-occurred with material differences in sustained stabilization or fragmentation. Scalar endpoint analyses remain valid for endpoint-specific estimands. However, when the clinical objective involves sustained multidomain stabilization, endpoint-only contrasts may not uniquely determine longitudinal organization. Trajectory-based estimands can be prespecified within the ICH E9(R1) framework and computed from standard repeated-measure data without altering randomization or intention-to-treat principles. Augmenting endpoint analyses with multidomain longitudinal summaries may improve alignment between the estimand of interest and chronic regulatory disease dynamics.</p>