Presentation (CTI)

OBJECTIVES: Traditional univariate or domain-by-domain analyses, particularly those relying on binary significance thresholds, treat PRO domains independently and therefore overlook their multivariate, conceptually related structure. This discards information about how symptoms and functioning co-occur, limiting clarity and stakeholder relevance of PRO findings. Profile analysis offers an intuitive multivariate framework to evaluate patterns and group differences across multiple correlated PRO domains. Here, we illustrate its application to assess whether baseline ECOG performance status (ECOG-PS) differentiates multidomain PRO profiles using data from the RATIONALE-305 trial.
METHODS: Profile analysis was conducted using one-way multivariate analysis of variance (MANOVA) with Wilk’s lambda to compare PRO profiles between ECOG-PS 0 and ECOG-PS 1 at baseline. Seven PRO outcomes were assessed: GHS/QoL, physical and role functioning, fatigue, pain, constipation, diarrhea, and EQ-5D VAS. Three hierarchical hypotheses were tested: (1) parallelism, whether groups exhibit the same pattern of means across domains; (2) levels, whether overall PRO means differ between groups; and (3) flatness, whether domain means are similar overall. Where relevant, post-hoc sensitivity analyses explored domain-specific differences.
RESULTS: Parallelism was nonsignificant, suggesting that ECOG-PS groups exhibited a similar pattern of PRO responses across domains. Equal levels test was significant, demonstrating that patients with ECOG-PS 1 had consistently worse outcomes than those with ECOG-PS 0 across the PRO profile. Given the hierarchical testing structure, flatness was not evaluated further.
CONCLUSIONS: Patients with poorer PS experienced greater symptom burden and reduced functioning, consistent with clinical expectations. Profile analysis revealed a clear separation between ECOG-PS groups across multiple PRO domains, illustrating how this approach can capture patterns of patient experience that are difficult to summarize using domain-specific or dichotomized endpoints alone. This application demonstrates the value of profile analysis for generating more intuitive presentations of PRO data for patients, clinicians, and health technology assessment bodies, as datasets become more complex.