Impact of response shift in longitudinal postoperative quality of life (QoL) analysis among breast cancer (BC) patients: A randomized multicenter cohort study.

Abstract

6121 Background: Response shift (RS) could bias QoL analyses when comparing treatments. One objective of our study was to characterize the change in internal standard, change in value, reconceptualization components of RS and to study its impact on longitudinal QoL in patients with primary BC. Methods: A French prospective multicentric cohort included women hospitalized or consulting for a primitive BC or a suspicion. QLQ-C30, BR23 and EuroQOL EQ-5D was assessed at the beginning (before treatment) and at the end of the first hospitalization as well as 3 and 6 months after. We used then-test, ideal scale and successive comparison approaches to characterise the 3 components of RS. Change in internal standard at the end of the first hospitalization was identified calculating mean difference (MD) between pretest and then-test (retrospective measure) baseline QoL assessment and using matched pair signed rank tests. A − (or +) MD reflects retrospectively a higher QoL level at baseline for functional (or symptoms) dimensions. Results: From February 2006 to February 2008, 381 women were included in 4 centers. Mean age was 56 years, 36% had a suspicion of BC while only 10% were not confirmed, 28% had mastectomy and 43% sentinel node biopsy, 35% adjuvant chemotherapy. At the end of the first hospitalization, retrospective assessment of QoL at baseline (then-test) highlighted that women had initially underestimated QoL level (p < 0.0001) for emotional (MD = -5.3 points) and cognitive (MD = -2.9 points) functions resulting respectively in a true QoL change of +3.3 points (p = 0.0008) and -3.3 points (p =0.0002) instead of observed change of +8.6 and -0.4 points respectively. In the same way fatigue (MD = -1.3), dyspnea (MD = -2.2), insomnia (MD = -5.1), appetite (MD = - 3.4) and diarrhea (MD = -3.1) symptoms as well as sexual enjoyment (MD = -2.9) and future perspective of BR23 (MD = +7.9) were initially significantly better at inclusion when retrospectively evaluated (p < 0.01). Conclusions: Occurrence of RS after first hospitalization for 7/15 dimensions of QLQ-C30 and 2/8 of BR-23 suggests to take RS into account to interpret QoL change in BC. Other components of RS are under investigation. No significant financial relationships to disclose.