Table 1 Health-related quality of life endpoints, hypotheses, analysis and interpretation approach of the CARFI trial
Endpoints | HRQLdomains and/or items1 | Hypotheses | Analysis and interpretation approach |
|---|---|---|---|
Primary endpoint, primary analysis | Â | ||
Change in overall HRQL from randomization to eight months follow-up. | EORTC QLQ-C30 Summary Score | There will be no difference between the two groups2 in change from randomization to eight months follow-up. | Non-inferiority approach. Between-group difference in mean change per group estimated using linear mixed effect model for repeated measures. P-value < 0.01 defined as statistically significant. MID > 10 defined as clinically relevant, used as the agreement-limit for non-inferiority, i.e. an estimated between-group difference less than 10 points would accept the non-inferiority hypothesis. Missing data handled as missing at random. |
Supportive analysis of primary endpoint | Â | ||
Average per-patient overall HRQL from randomization to progressive disease/death/drug discontinuation/end of study (whichever came first). | EORTC QLQ-C30 Summary Score | There will be no difference between the two groups2 in per-patient mean overall HRQL averaged across all available time points from randomization to progressive disease/death/drug discontinuation/end of study (whichever came first). | Non-inferiority approach comparing the mean of the average per-patient QLQ-C30-sum score between the two groups with the rank-sum test. P-value < 0.01 defined as statistically significant. MID > 10 used as the agreement-limit for non-inferiority, i.e. an estimated between-group difference less than 10 points would accept the non-inferiority hypothesis. Missing data handled as missing not at random. |
Secondary endpoints | Â | ||
1. Change in HRQL from randomization to eight months follow-up. | All individual HRQL domains of EORTC QLQ-C30 and QLQ-MY20, and FACT/GOG-ntx instruments | There will be no difference between the two groups2 in HRQL, except for the domains potentially impacted by Kd-related side effects (nausea/ vomiting, fatigue, dyspnea, diarrhea, body image). Even for these five domains, the differences will not exceed the threshold for clinically relevant differences. | Non-inferiority approach using linear mixed effect model for repeated measures. P-value < 0.05 defined as statistical significant. Evidence-based MID threshold used as the agreement-limit for non-inferiority of between group differences for EORTC QLQ-C30 and QLQ-MY20 domains (50, 52) and 4.4 for FACT/GOG-ntx. Missing data handled as missing at random. |
2. The proportion of patients who have improved, remained stable and worsened in HRQL from randomization to eight months follow-up. | Physical, role, social and emotional functioning, GHS/QoL, body image and future perspectives | The proportion of patients who improved/remained stable/worsened will be similar between the two groups2. | Superiority approach in favor of the observation group. Patient-level analysis using the responder definition threshold of 20 points for multi item domains and 33 points for body image for improvement or worsening. Chi-square test for between group comparisons. P-value < 0.05 defined as statistical significant. Missing data handled as missing completely at random. |
3. The time to first recorded improvement in HRQL from randomization to eight months follow-up. | Physical, role, social and emotional functioning; GHS/QoL, body image and future perspectives | Patients randomized to Kd maintenance therapy will take longer time to achieve improvement in emotional functioning, role functioning and social functioning as well as in the body image domain, compared to observation, but will not differ in time to improve in physical functioning and GHS/QoL. | Superiority approach in favor of the observation group for all domains except physical functioning and GHS/QoL. Patient-level analysis to assess the first time of improvement defined as 20 points for the multi item domains and 33 points for body image compared to randomization. Proportional hazards Cox regression model was used to compare between groups. P-value < 0.05 defined as statistical significant. Missing data were handled as missing completely at random. |
4. The proportion of patients with Kd-related symptoms and the proportion of patients with moderate to severe Kd-related symptoms from randomization to eight months follow-up. | From QLQ-C30: Fatigue, nausea/vomiting, dyspnoea, insomnia, diarrhoea. From QLQ-MY20: agitation/irritation. | A larger proportion of patients randomized to Kd maintenance therapy will report more severe levels of symptoms compared to observation. | Superiority approach in favor of the observation group. Patient-level analysis using the raw score thresholds for change from randomization of ≥ 33 points for Kd-related symptoms and ≥ 66 points for severe Kd-related symptoms. Chi-square test for between group comparisons. P-value < 0.05 defined as statistical significant. Missing data handled as missing completely at random. |
5. Quality-adjusted progression-free survival from randomization to last follow-up time-point. | Two quality-adjustment metrics: EORTC QLQ-C30 Summary Score and EORTC QLU-C10D | Quality-adjusted progression-free survival will not differ substantially from the difference observed in progression free survival, since we do not expect Kd to impact significantly on overall HRQL. | Non-inferiority approach. Per-patient time in QAPFS calculation for each of the metrics and multiplied with the per-patient estimated mean time to last follow-up time point. Group-level QAPFS estimated by Kaplan-Meier method and difference between group were estimated using bootstrap methods. P-value < 0.05 defined as statistical significant. Missing data handled as missing completely at random. |