Baseline characteristics predicting lower return rates of missing patient-reported quality of life data over 5 years: evidence from the IMPORT HIGH and IMPORT LOW breast cancer radiotherapy trials.

Abstract

Long-term patient reported outcomes (PROs) from questionnaires are important to capture late adverse effects in breast cancer treatment. Declining return rates over time may introduce selection bias and reduce robustness of results. Using data from two trials which had collected data prospectively, we investigated factors associated with non-return of PROs at 5 years. We also investigated which questions were more likely missed in PROs. IMPORT HIGH (ISRCTN47437448) and IMPORT LOW (ISRCTN12852634) investigated different radiotherapy treatments for breast cancer patients with high and low ipsilateral breast tumour relapse risk respectively. Both trials had PRO sub-studies with similar design, using questionnaires such as EORTC QLQ-C30, QLQ-BR23 and Body Image Scale. We used univariable and multivariable logistic regressions for the analysis. The return rate for PROs was 99.5 % (1034/1039) at baseline and 81.4 % (766/941) at year 5, 100 % (1257/1257) and 84 % (974/1165) for IMPORT HIGH/LOW respectively. Participants with moderate/severe depression/anxiety at baseline were more likely to miss year-5 response: IMPORT HIGH 16 % with depression/anxiety, OR = 1.60, CI = [1.03-2.49]; IMPORT LOW 9 % with depression/anxiety, OR = 2.62, CI = [1.66-4.14]. In multivariable analysis, age, smoking status, depression/anxiety, and ethnicity were significant predictors of missing PRO returns in IMPORT HIGH while depression/anxiety, relationship status and deprivation in IMPORT LOW. Participants who missed sexual functioning questions initially were more likely to continue missing them (OR = 1.02, CI = [1.00-1.03] IMPORT HIGH; OR = 1.02, CI = [1.01-1.02] IMPORT LOW). Participants who were younger, more deprived, smoked at baseline, had depression/anxiety may require targeted support to return PRO during follow-up. Sensitive questions need 'Prefer not to say' or 'Not applicable' options available and require further work to optimise wording.