Palliative Performance Scale Predicts Survival in Patients with Bone Metastasis Undergoing Radiotherapy

Reviewer comment

Authors’ response

Continuous covariates have to be reported only as median/IQR.

Thank you for this suggestion. All continuous covariates (e.g., age, BMI, Charlson Comorbidity Index) are now consistently reported as median and interquartile range (IQR) throughout the manuscript, including the baseline characteristics and Results section. Mean ± standard deviation values were removed accordingly.

Add the median follow-up for the whole cohort.

We agree that reporting follow-up time is important. Given the palliative intent of treatment and the absence of structured follow-up after radiotherapy, all patients experienced the event of interest (death). Therefore, the median follow-up time corresponds to the median overall survival and is now explicitly reported in the Results section for clarity.

Add the number of events, total and stratified by cancer subtype.

This information has now been added. The total number of overall survival events as well as event numbers and rates stratified by cancer subtype are summarized in a new table (Table 5), providing a descriptive overview of outcome heterogeneity across tumor entities.

Restricted mean survival time: explain the reader why you added this (correct!) info!

We thank the reviewer for highlighting this point. A detailed explanation has been added to the Statistical Analyses section. Restricted mean survival time (RMST) was included as an absolute and clinically interpretable measure of average survival time that does not rely on the proportional hazards assumption and is particularly suitable for palliative cohorts with short and heterogeneous survival and limited follow-up.

What about treating PPS as a time-dependent covariate in a Mantel–Byar model? In such way, you will be able to estimate the PPS contribution too, that is impossible stratifying the cohort by PPS.

We appreciate this thoughtful methodological suggestion. A Mantel–Byar or other time-dependent approach requires longitudinal PPS assessments with clearly defined change times. In our dataset, PPS was assessed only at baseline prior to radiotherapy, and no repeated measurements during follow-up were available. Consequently, modeling PPS as a time-dependent covariate would not be methodologically justified. Instead, to appropriately address the violation of the proportional hazards assumption for PPS, we performed a stratified Cox regression as a sensitivity analysis. This approach allows valid estimation of the effects of other covariates while acknowledging non-proportionality of PPS. This rationale is now explicitly stated in the Methods and Sensitivity Analysis sections.

Reviewer comment

Authors’ response

Secondary Outcome (Discharge Destination): The inclusion of logistic regression analyzing discharge to home versus other destinations is a good addition. It might be worth commenting briefly on this negative finding in the Discussion section.

We thank the reviewer for this insightful comment. We have now added a dedicated paragraph to the Discussion section addressing the negative finding of the discharge destination analysis. Specifically, we discuss that discharge disposition after palliative radiotherapy may be influenced by factors not captured in this study, such as availability of informal caregivers, social support, organizational aspects of post-acute care, and rapid changes in clinical condition after treatment. We also note that the relatively high proportion of patients discharged to other or unspecified destinations (41%) may have limited the statistical power to detect significant associations. This addition provides a more balanced interpretation of the secondary outcome beyond survival.

Inconsistency regarding the threshold selected for the PPS (70% in Kaplan–Meier analysis vs. 60% in Cox regression model).

We appreciate the reviewer for highlighting this inconsistency. To ensure methodological coherence and improve clarity, all analyses were harmonized to use a uniform PPS cutoff of 60%. This threshold is clinically established in palliative care and distinguishes patients with preserved ambulatory function from those with marked functional impairment. The Kaplan–Meier analysis was updated accordingly, and the rationale for selecting this cutoff is now explicitly stated in the Methods section, ensuring consistency across all analyses.