OBJECTIVES: In a phase 2 trial, abemaciclib demonstrated tumour responses in refractory HR+ HER2- mBC. The most common investigator-reported treatment emergent adverse events (TEAEs) were diarrhoea, fatigue, nausea, decreased appetite, and abdominal pain. Fatigue is known to play a central role in patient experience. We present an analysis of patient reports of these symptoms over the course of the trial, with fatigue as the central concept. We explicitly consider missing data, the relationships between symptoms, and the relationships with survival.

METHODS: Data came from a single-arm, open-label study of 132 previously-treated patients with mBC. The EORTC QLQ-C30 v3 was administered at baseline and every 28 days thereafter. Domains for fatigue, pain, nausea and vomiting, appetite loss, and diarrhoea were analysed from baseline to visit 10 using extended pattern mixture modelling (ePMM). ePMM is a latent-variable longitudinal mixture model approach used to examine individual variability across all time points and identify subgroups; the subgroup identifier is regressed on missing data indicators at each time point. The final models assessed 1) the prediction of fatigue by the remaining symptoms, and 2) the prediction of overall survival (OS), and progression-free survival (PFS) by fatigue.

RESULTS: Three fatigue subgroups were identified: ‘no change’ (55%; n=73), ‘improvement’ (9%; n=12), ‘worsening then improvement’ (36%; n=47). Pain significantly predicted fatigue (Ρ< 0.01) but the remaining TEAEs did not. Fatigue predicted OS and PFS: Patients in the ‘no change’ fatigue subgroup had significantly longer OS and PFS than those in the ‘worsening then improvement’ subgroup (Ρ=0.01 and Ρ = 0.001, respectively).

CONCLUSIONS: Most patients reported no change in fatigue. Fatigue was predicted by changes in pain, and was predictive of changes in OS and PFS. Patient reports of symptom experience are informative and when modelled appropriately they can inform understanding of differential survival.