Evaluating health-related quality-of-life therapeutic effectiveness in a clinical trial with extensive nonignorable missing data and heterogeneous response: results from a phase III randomized trial of gemcitabine plus paclitaxel versus paclitaxel monotherapy in patients with metastatic breast cancer

Abstract

This manuscript presents health-related quality of life (HRQL) results from a phase III trial of gemcitabine-paclitaxel (GT) versus paclitaxel (T) in metastatic breast cancer patients. Patients completed the Rotterdam Symptom Checklist (RSCL) and Brief Pain Inventory (BPI) at baseline and at the end of each cycle. Sensitivity analyses for six longitudinal pattern mixture models (PMMs) assessed potential bias due to informative dropout. Cumulative probability plots with 50% confidence intervals indicated the proportion of patients whose HRQL was likely to improve, decline, or stay the same. Sensitivity analyses addressing nonignorable missing RSCL data included 351 patients. The mean RSCL global HRQL score for GT was significantly and consistently better than that for T (all PMMs P<0.040). The slope estimate of 1.5 points (100-point scale) per cycle from one PMM translated to a clinically significant 9-point improvement over six cycles with GT versus T. For GT, ~25% of patients were more likely than not to have improved HRQL, whereas that proportion for T was ~5%. PMMs showed no consistent treatment arm differences for BPI or other RSCL outcomes. Adding gemcitabine to paclitaxel for the treatment of metastatic breast cancer is more likely to improve global HRQL over time compared to monotherapy treatment.