Review Reports - High-Dose Intravenous Vitamin C and Mistletoe Therapy in Breast Cancer: A Prospective Cohort Study on Overall Survival in Palliative Care

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Major Revision Recommendation:

Although the study addresses an important clinical question and presents intriguing survival associations between high-dose intravenous vitamin C, mistletoe therapy, and mortality outcomes in breast cancer patients, several critical methodological concerns substantially limit causal interpretation. The observational design remains highly susceptible to selection bias, confounding by indication, and immortal time bias, which the authors themselves acknowledge. Furthermore, detailed treatment exposure data, cancer stage stratification, concurrent systemic therapies, propensity-score analyses, and sensitivity analyses are lacking. Addressing these limitations is essential before the conclusions can be considered robust and clinically translatable.

How did the authors address immortal time bias, given that patients were classified as exposed only after completing ≥3 months and >10 treatment sessions?
Why were important tumor-related variables such as TNM stage, molecular subtype, and treatment line not included in multivariable analyses?
Was propensity score matching or inverse probability weighting considered to reduce treatment-selection bias?
Were dose-response relationships evaluated for IVC (10–100 g) and mistletoe therapies?
The Kaplan–Meier analysis for IVC shows only borderline significance (p=0.051). How do the authors reconcile this with the significant multivariable HR?
Were missing data handled through imputation, complete-case analysis, or another approach?
The manuscript lacks information regarding systemic therapies received (endocrine therapy, HER2-targeted therapy, CDK4/6 inhibitors). Could these have influenced survival outcomes?
Hyperthermia was associated with increased mortality in univariate analysis. Could this indicate unmeasured disease severity not fully captured by adjustment?
The manuscript discusses mechanistic pathways extensively, but no biological biomarkers were measured. Please clearly distinguish speculation from study-derived evidence.
As this is a single-center integrative oncology cohort, the generalizability of findings appears limited. Please elaborate on external validity concerns.

Author Response

We sincerely thank Reviewer 1 for the thorough and constructive critique, which substantially strengthened the rigor of our analysis. To re-evaluate and ensure statistical rigor, we thoroughly re-codified and restructured the patient clinical dataset. While we were able to precisely determine the initial treatment onset dates and total session counts, granular time-series data—such as exact administration dates for each individual session—were frequently missing or imprecise in the historical records and could not be utilized for continuous time-course modeling. Nevertheless, we addressed immortal time bias through a time-dependent Cox model combined with the Simon–Makuch/Mantel–Byar framework (Q1, Q5), expanded the multivariable model to incorporate key oncologic covariates (Q2, Q7), clarified our rationale for not pursuing propensity-score methods and continuous dose–response stratification given the constraints of our real-world dataset (Q3, Q4), verified missing-data handling and corrected reporting inconsistencies for hyperthermia (Q6), and revised the Discussion to more clearly separate mechanistic speculation from study-derived findings and to elaborate on residual confounding and generalizability (Q8–Q10). Detailed point-by-point responses are provided below.

Q1.How did the authors address immortal time bias, given that patients were classified as exposed only after completing ≥3 months and >10 treatment sessions?

Answers) Patients were classified as unexposed from baseline until the exact date of their first therapy session, and as exposed thereafter. Survival curves were reconstructed using the Simon–Makuch method and evaluated via the Mantel–Byar test. Even under this rigorous time-dependent model, protective treatment effects remained statistically significant:

High-Dose IVC: Time-dependent adjusted Hazard Ratio (aHR) = 0.561 (95% CI: 0.319–0.986, p = 0.045) (Manuscript: Page 4, Lines 112–128 & Page 7, Table 2)
Mistletoe Therapy: aHR = 0.463 (95% CI: 0.231–0.927, p = 0.030) (Manuscript: Page 4, Lines 112–128 & Page 7, Table 2)

Q2. Why were important tumor-related variables such as TNM stage, molecular subtype, and treatment line not included in multivariable analyses?

Answers) Thank you for pointing out these essential confounders. We updated our multivariable Cox model to incorporate baseline oncologic parameters as follows.

Inclusion of Essential Oncologic Modalities & Subtypes: We included Triple-Negative Status (TNBC), Chemotherapy, Operation (Surgical Resection), Radiotherapy, along with Age and Distant Metastasis in the multivariable model.

Handling of TNM Staging: T and N stages contained 40–50% missing values in our EMR dataset, which precluded inclusion without severe listwise deletion bias. Instead, Distant Metastasis (verified in 100% of participants) was included as the primary stage indicator.

Q3. Was propensity score matching or inverse probability weighting considered to reduce treatment-selection bias?

Answers) Propensity Score Methods: Because multiple complementary modalities were administered concurrently over varying periods, applying time-varying PSM or IPW resulted in severe sample reduction and model overfitting. Therefore, a time-dependent multivariable Cox model was preferred.

Q4. Were dose-response relationships evaluated for IVC (10–100 g) and mistletoe therapies?

Answers) In real-world outpatient settings, maintenance dosages and intervals were dynamically adjusted according to patient tolerance and compliance, making precise long-term tracking of cumulative doses limited. We have added this as an explicit study limitation. (Manuscript: Page 11, Lines 245–252)

Q5. The Kaplan–Meier analysis for IVC shows only borderline significance (p=0.051). How do the authors reconcile this with the significant multivariable HR?

Answers) This discrepancy has been fully resolved. Standard KM curves failed to account for time-dependent exposure. Upon applying Simon–Makuch curves with the Mantel–Byar test (which properly handles person-time), the unadjusted log-rank difference became highly significant (Mantel–Byar test p < 0.001; Figure 1). This aligns with our adjusted multivariable time-dependent Cox HR (aHR = 0.561, p = 0.045).

Q6. Were missing data handled through imputation, complete-case analysis, or another approach?

Answers) For all variables included in our final multivariable time-dependent Cox proportional-hazards regression models (Age, Distant Metastasis, Triple-Negative Breast Cancer [TNBC] status, Chemotherapy, Surgical Resection, Radiotherapy, High-Dose IVC, Mistletoe therapy, and Hyperthermia), complete data were available across all 748 patients. Therefore, a complete-case analysis was performed without requiring statistical imputation methods (e.g., multiple imputation).

Q7. The manuscript lacks information regarding systemic therapies received (endocrine therapy, HER2-targeted therapy, CDK4/6 inhibitors). Could these have influenced survival outcomes?

Answers) We acknowledge that specific lines of systemic targeted therapies—such as endocrine therapy, HER2-targeted agents, and CDK4/6 inhibitors—are critical determinants of overall survival in breast cancer management.

Due to the retrospective nature of our electronic medical record (EMR) database, detailed longitudinal records regarding specific targeted and hormonal therapy regimens were incomplete or unrecorded for a subset of patients, preventing their inclusion as distinct individual covariates in the multivariable model. We have explicitly added the lack of detailed systemic regimen data (endocrine, HER2-targeted, and CDK4/6 inhibitor status) as a study limitation in the Discussion section, emphasizing the need for future prospective studies with granular regimen tracking.

Q8. Hyperthermia was associated with increased mortality in univariate analysis. Could this indicate unmeasured disease severity not fully captured by adjustment?

Answers) In our clinical practice, regional hyperthermia (oncothermia/hyperthermia therapy) is not only applied to primary tumor sites but is also heavily utilized as a localized integrative therapy for distant systemic metastases (e.g., peritoneal, hepatic, or bone metastases). Consequently, the cohort receiving hyperthermia therapy was inherently enriched with patients suffering from advanced, multi-organ metastatic, or refractory disease—a subgroup with a baseline clinical risk profile and tumor burden significantly higher than the non-hyperthermia group.

This clinical selection pattern introduces substantial confounding by indication (unmeasured disease severity), which naturally drove the elevated unadjusted mortality observed in the univariate analysis. Notably, after rigorously adjusting for overall disease stage, metastatic burden, performance status, and baseline metabolic parameters in our multivariable time-dependent Cox model, the unadjusted harmful association was largely attenuated, confirming that hyperthermia itself was not a driver of mortality, but rather a surrogate marker for high-risk metastatic disease

Q9. The manuscript discusses mechanistic pathways extensively, but no biological biomarkers were measured. Please clearly distinguish speculation from study-derived evidence.

Answers) We completely agree with the reviewer's critical point. Because our study is an observational clinical cohort analysis, we did not directly measure intra-tumoral or serum biological biomarkers (such as reactive oxygen species levels, immune cell subsets, or specific cytokine profiles) to empirically confirm the underlying cellular mechanisms discussed in the manuscript. To address this directly, we have added an explicit caveat at the beginning of the mechanistic discussion in the Discussion section, clarifying that these interpretations are derived from prior pharmacological and preclinical literature rather than from biomarker data generated in our cohort (Manuscript: Discussion, Lines 266–272). We have also added a corresponding study limitation explicitly stating that the mechanistic pathways discussed should be regarded as hypothesis-generating rather than confirmatory (Manuscript: Discussion, Lines 334–338).

Q10. As this is a single-center integrative oncology cohort, the generalizability of findings appears limited. Please elaborate on external validity concerns.

Answers) We completely agree with the reviewer that the single-center design of our study inherently limits the broad generalizability (external validity) of our findings to other healthcare settings or distinct patient populations.

In the revised Discussion (Limitations section), we have expanded our elaboration on external validity concerns, highlighting the following specific points:

Finally, as a single-center study conducted in a specialized integrative palliative care setting in Korea, generalizability to other populations and healthcare settings may be limited.

Reviewer 2 Report

Comments and Suggestions for Authors

Overall a very well surmised hypothesis with data supporting it. I was not aware of these types of treatments in the alternative medicine space as all of my training and work has been focused on traditional medicine as practiced in the US.

What does the p = 0.051 mean in Figure 1? Is that not significant by usual t-test? Both survival curves look interesting for follow-up.

Can you add demonstrated or measured toxicity for the patients? I am told that mistletoe is poisonous, but I don't know if the extract is more easily tolerated.

Any way to account for or integrate socio-economic status in this data. It may be that these treatments cost a lot out of pocket and if so you may be seeing an improvement based on the ability to pay.

Another factor that comes to mind is the mental aspect of the disease and those patients that still have hope versus those that have given up. Not sure if there is any information like this in the record.

I like reference 13, but upon review it only enrolled 15 patients and didn't track survival. I like that you call that out in the Conclusions, but what more can you do to support a follow-up trial to move this ball forward and potentially benefit patients.

Author Response

Reviwer2

Q11. What does the p = 0.051 mean in Figure 1? Is that not significant by usual t-test? Both survival curves look interesting for follow-up.

Answers) By replacing standard KM curves with time-dependent Simon–Makuch curves (Figure 1, Page 8), the unadjusted group comparison yields p < 0.001 (Mantel–Byar test), resolving the discrepancy with the multivariable Cox model (aHR = 0.561, p = 0.045).

Q12. Can you add demonstrated or measured toxicity for the patients? I am told that mistletoe is poisonous, but I don't know if the extract is more easily tolerated.

We thank the reviewer for raising this important clinical safety question. While raw European mistletoe (Viscum album L.) contains cytotoxic lectins and viscotoxins that can be toxic when ingested, pharmaceutical-grade mistletoe extracts used in integrative oncology are standardized, highly purified aqueous preparations administered via subcutaneous or intravenous routes at micro-to-milligram doses. In clinical practice, these preparations are well-tolerated.

As suggested, we have expanded the description of patient safety and treatment tolerability in the Materials and Methods (Section 2.4) and Results sections. Specifically, no patients in our cohort experienced severe (Grade 3 or higher) treatment-related toxicities attributable to high-dose IVC or mistletoe therapy, and no treatment-related mortalities occurred.

Q13. Any way to account for or integrate socio-economic status in this data. It may be that these treatments cost a lot out of pocket and if so you may be seeing an improvement based on the ability to pay.

Answers) We greatly appreciate this highly insightful question regarding potential socioeconomic confounding. While out-of-pocket costs for complementary therapies in countries such as the United States can indeed introduce selection bias toward higher socioeconomic status (SES) patients, the healthcare structure and insurance landscape in South Korea substantially mitigate this financial barrier.

Substantially Lower Out-of-Pocket Direct Costs:

Compared to Western countries, direct medical fees for integrative infusions in South Korea are notably lower. For example, a single 40 g high-dose IVC infusion typically costs around $300 USD (~400,000 KRW) in the United States, whereas in South Korea, the identical infusion costs approximately $45 USD (~60,000 KRW).

High Coverage via Private Indemnity Health Insurance:

In addition to lower baseline costs, over 70–80% of the South Korean population is enrolled in private indemnity health insurance ("Sil-bi insurance"). These supplementary policies reimburse 80–100% of out-of-pocket expenses for non-reimbursed clinical treatments (including IVC and mistletoe therapy) prescribed for cancer management. Consequently, financial barriers are minimal, allowing patients across diverse socioeconomic strata equal access to these modalities.

Q14. Another factor that comes to mind is the mental aspect of the disease and those patients that still have hope versus those that have given up. Not sure if there is any information like this in the record.

We agree that psychological status and treatment motivation are important factors in cancer survivorship. While standardized psychological instruments or validated hope scales were not systematically recorded in this retrospective EMR database, the clinical characteristics of our cohort closely reflect the reviewer's observation.

Patients seeking treatment at an integrative oncology center represent an intrinsically proactive and highly motivated group. Unlike end-stage hospice or purely comfort-oriented palliative care patients who may have discontinued active oncologic therapy, participants in our study actively sought complementary interventions alongside conventional treatments to enhance therapeutic efficacy and maintain personal agency. We have acknowledged the potential influence of patient motivation and health-seeking behavior in the Discussion section (Limitations).

Q15. I like reference 13, but upon review it only enrolled 15 patients and didn't track survival. I like that you call that out in the Conclusions, but what more can you do to support a follow-up trial to move this ball forward and potentially benefit patients.

Answers)

We appreciate this constructive comment. While randomized controlled trials (RCTs) represent the gold standard for clinical evidence, conducting a formal RCT for complementary therapies in an integrative palliative setting presents substantial ethical, logistical, and financial challenges, as many patients specifically choose or decline these therapies based on personal preference.

Therefore, rather than initiating a traditional RCT at this stage, our current roadmap focuses on establishing a high-precision prospective registry. Through this registry, we are systematically collecting standardized real-world data, including exact administration timestamps, cumulative dosage metrics, concurrent systemic cancer therapies, and long-term longitudinal outcomes. We believe this high-granularity prospective dataset will provide robust real-world evidence to guide individualized patient care. We have updated the Conclusion section to outline this prospective roadmap.

Reviewer 3 Report

Comments and Suggestions for Authors

Overall, this is an interesting prospective cohort study looking at the association between several complementary therapies and overall survival in female breast cancer patients. The relatively large cohort and long observation period are strengths, and the findings on high-dose intravenous vitamin C (IVC) and mistletoe therapy are potentially interesting. However, I have several concerns about the survival analysis and the way treatment exposure and confounding factors were handled. These issues are important for interpreting the main conclusions, so I would recommend major revision.

My main concern is the way exposure to high-dose IVC and mistletoe therapy was defined. According to the Methods, patients were considered treated only if they received therapy for at least 3 months and completed more than 10 sessions, and follow-up began on the date of initial clinic registration. This means that patients had to survive long enough to meet the treatment definition, potentially introducing immortal time bias. The authors already mention this limitation in the Discussion, which is appropriate. However, because the paper's main conclusion is based on the lower mortality observed in the IVC and mistletoe groups, I think this issue needs to be addressed more directly in the analysis rather than only discussed as a limitation. If the treatment timing data are available, I would strongly suggest performing a time-dependent Cox analysis or an appropriate landmark analysis. If this is not possible, the conclusions regarding a survival benefit should be stated more cautiously throughout the manuscript, especially in the Abstract and Conclusions.
Additionally, I think the multivariable model needs more explanation. The current model adjusts for age, BMI, smoking, alcohol use, metabolic parameters, CRP, hypertension, diabetes, and metastatic status. However, Table 1 shows clear differences between survivors and deceased patients in several clinically important breast cancer-related factors, including triple-negative status, surgery, and radiotherapy. The Methods also state that these complementary therapies were administered alongside conventional cancer treatments, including surgery, chemotherapy, and radiotherapy, but these treatment-related variables do not appear to be fully included in the final Cox model. I am not suggesting that all variables with a significant p-value in Table 1 must automatically be included in the model. However, given the observational nature of this study, residual confounding is an important concern. It would be helpful if the authors could explain how covariates were selected and whether major tumor-related factors and conventional anticancer treatments were considered. If the data are available, additional sensitivity analyses including relevant breast cancer-specific prognostic factors would make the conclusions more convincing.
Both variables are listed separately in Table 2, and both appear in the multivariable analysis, but the Methods do not clearly explain how they are defined or how they differ clinically. Please provide clear definitions and indicate when each variable was assessed relative to study entry. It would also be useful to clarify whether both variables were included in the same Cox model.
The manuscript explains how follow-up was calculated, and the Kaplan–Meier curves are shown up to 60 months, but the actual follow-up time of the cohort is not clearly summarized. Please provide the median follow-up time and an appropriate measure of spread, such as the interquartile range. Adding the number at risk below the Kaplan–Meier curves would also improve the figures.
Since Cox proportional-hazards regression is the main survival analysis used in this study, the authors should state whether the proportional-hazards assumption was evaluated. If it was tested, please briefly describe the method used and whether any important violations were found.
“Mega-vitamin” is included as a separate treatment in Tables 1-2 and 2, but it is not clearly described in the Methods. Therefore, please provide the composition, dose, route, treatment duration, and criteria used to define exposure. Since high-dose IVC is analyzed separately, it is also important to clarify whether there is any overlap between the “mega-vitamin” and IVC groups.
The counts and percentages reported for hyperthermia therapy in Table 1-2 do not seem to match. The table lists 359 patients overall, 299 survivors, and 60 deaths as receiving hyperthermia therapy, but the corresponding percentages do not appear to be consistent with these numbers. The percentages reported in the Results section should also be checked against the table. Please verify these values and correct them where necessary.

Author Response

Reviewer 3

Q16. My main concern is the way exposure to high-dose IVC and mistletoe therapy was defined. According to the Methods, patients were considered treated only if they received therapy for at least 3 months and completed more than 10 sessions, and follow-up began on the date of initial clinic registration. This means that patients had to survive long enough to meet the treatment definition, potentially introducing immortal time bias. The authors already mention this limitation in the Discussion, which is appropriate. However, because the paper's main conclusion is based on the lower mortality observed in the IVC and mistletoe groups, I think this issue needs to be addressed more directly in the analysis rather than only discussed as a limitation. If the treatment timing data are available, I would strongly suggest performing a time-dependent Cox analysis or an appropriate landmark analysis. If this is not possible, the conclusions regarding a survival benefit should be stated more cautiously throughout the manuscript, especially in the Abstract and Conclusions.

Answers) We express our sincere gratitude to the reviewer for this critical and methodologically vital suggestion. We fully agree that addressing immortal time bias directly through analytical adjustments is essential to ensure the reliability of our survival estimates.

In response to your recommendation, we performed a comprehensive re-analysis of our cohort using a time-dependent Cox proportional-hazards regression model with a counting-process (start-stop) data structure. Under this framework, patients were classified as unexposed from baseline until the exact date of their first therapy administration and as exposed thereafter. In addition, survival curves were reconstructed using the Simon–Makuch method and compared using the Mantel–Byar test to appropriately reflect time-varying exposure status.

Landmark analysis was not selected because therapy initiation dates varied substantially across participants, and establishing a single arbitrary landmark cutoff would have resulted in significant sample loss and exposure misclassification.

Notably, even after applying this rigorous time-dependent framework, the protective associations against overall mortality remained statistically significant:

High-Dose IVC Therapy: Multivariable time-dependent HR = 0.561 (95% CI: 0.319–0.986, p = 0.045)
Mistletoe Extract Therapy: Multivariable time-dependent HR = 0.463 (95% CI: 0.231–0.927, p = 0.030)

We have updated the Abstract, Methods (Section 2.5), Results (Section 3.2, Table 2, Figures 1 & 2), and Discussion to reflect these updated time-dependent analysis results and updated the text to maintain an appropriately cautious tone regarding these observational findings.

Q17. Additionally, I think the multivariable model needs more explanation. The current model adjusts for age, BMI, smoking, alcohol use, metabolic parameters, CRP, hypertension, diabetes, and metastatic status. However, Table 1 shows clear differences between survivors and deceased patients in several clinically important breast cancer-related factors, including triple-negative status, surgery, and radiotherapy. The Methods also state that these complementary therapies were administered alongside conventional cancer treatments, including surgery, chemotherapy, and radiotherapy, but these treatment-related variables do not appear to be fully included in the final Cox model. I am not suggesting that all variables with a significant p-value in Table 1 must automatically be included in the model. However, given the observational nature of this study, residual confounding is an important concern. It would be helpful if the authors could explain how covariates were selected and whether major tumor-related factors and conventional anticancer treatments were considered. If the data are available, additional sensitivity analyses including relevant breast cancer-specific prognostic factors would make the conclusions more convincing.

Answers)

We sincerely thank the reviewer for this crucial comment regarding covariate selection and potential residual confounding.

In our updated multivariable Cox proportional-hazards regression analysis (Table 2), major breast cancer-specific prognostic indicators and baseline conventional oncologic treatments were fully integrated into the model as adjusting covariates alongside demographic and metabolic parameters. Guided by clinical relevance and established oncologic risk factors, the final multivariable model explicitly includes:

Tumor Characteristics & Extent: Triple-negative breast cancer (TNBC) status and distant metastasis.
Conventional Oncologic Treatments: Surgical resection, chemotherapy, and radiotherapy history.
Demographic & Clinical Confounders: Age, BMI, and relevant baseline metabolic parameters.

Detailed individual T and N staging contained substantial missing data (>40%) in historical electronic medical records (EMR). To prevent listwise deletion bias and significant sample size reduction, T/N stages were omitted from multivariable adjustment; instead, 100%-verified distant metastasis status was utilized as the primary indicator of systemic disease burden.

Crucially, even after full multivariable adjustment for TNBC status, distant metastasis, and all major conventional cancer treatments (chemotherapy, operation, and radiotherapy), both complementary modalities maintained statistically significant, independent protective associations against overall mortality:

High-Dose IVC Therapy: Multivariable HR = 0.561 (95% CI: 0.319–0.986, p = 0.045)
Mistletoe Extract Therapy: Multivariable HR = 0.463 (95% CI: 0.231–0.927, p = 0.030)

We have clarified the covariate selection rationale in Section 2.5 (Statistical Analysis) and listed all adjusted variables in the revised Table 2 footnote and text

Q18. Both variables are listed separately in Table 2, and both appear in the multivariable analysis, but the Methods do not clearly explain how they are defined or how they differ clinically. Please provide clear definitions and indicate when each variable was assessed relative to study entry. It would also be useful to clarify whether both variables were included in the same Cox model.

Answers) We thank the reviewer for this important request for clarification regarding intervention definitions, timing of assessment, and model specifications.

Clinical Definitions and Differences:

Clinically, high-dose intravenous vitamin C (IVC) involves the intravenous infusion of ascorbic acid (10–100 g per session) aimed at generating pro-oxidant pharmacological concentrations of hydrogen peroxide ($H_2O_2$) selectively in tumor microenvironments. In contrast, mistletoe therapy consists of standardized aqueous extracts of Viscum album L. (0.02–20 mg per session) administered via subcutaneous or intravenous routes to induce immunomodulatory (e.g., NK cell activation) and cytotoxic lectin-mediated apoptotic responses.

Timing of Assessment:

Baseline clinical characteristics, demographic factors, and initial non-exposure statuses were defined at the time of initial clinic registration. Complementary therapy exposures were captured longitudinally from electronic medical record (EMR) time-stamps across the entire follow-up period and operationalized as time-dependent covariates using a counting-process (start-stop) framework. Patients transitioned from the unexposed to the exposed state on the exact date they satisfied the predefined exposure threshold ($\ge 3$ months of treatment duration and $> 10$ cumulative sessions).

Model Specification:

We confirm that both high-dose IVC and mistletoe therapies were included simultaneously in the same multivariable Cox proportional-hazards regression model alongside all conventional oncologic treatments (chemotherapy, surgery, radiotherapy), distant metastasis, TNBC status, and clinical covariates. This mutual co-adjustment ensures that the protective association observed for each modality represents an independent effect while controlling for potential co-administration:

High-Dose IVC Therapy: Multivariable HR = 0.561 (95% CI: 0.319–0.986, p = 0.045)
Mistletoe Extract Therapy: Multivariable HR = 0.463 (95% CI: 0.231–0.927, p = 0.030)

We have expanded Section 2.3 (Palliative Care Interventions) and Section 2.5 (Statistical Analysis) in the revised manuscript to explicitly detail these clinical definitions, assessment timing, and multivariable model specifications.

Q19. The manuscript explains how follow-up was calculated, and the Kaplan–Meier curves are shown up to 60 months, but the actual follow-up time of the cohort is not clearly summarized. Please provide the median follow-up time and an appropriate measure of spread, such as the interquartile range. Adding the number at risk below the Kaplan–Meier curves would also improve the figures.

Answers) We sincerely thank the reviewer for pointing out the need to explicitly report the follow-up metrics and for suggesting enhancements to the survival figures.

Follow-up Duration Metrics:

In accordance with your recommendation, we have updated both the Abstract and Section 3.1 (Results) of the revised manuscript to explicitly summarize the cohort's follow-up metrics:

Abstract: Updated to state: "Over a median follow-up duration of 10.0 months (interquartile range [IQR]: 2.9–30.2 months), 109 patients (14.6%) died."
Section 3.1 (Baseline Characteristics): Updated to state: "During a median follow-up period of 10.0 months (IQR: 2.9–30.2 months; mean ± SD: 20.2 ± 24.2 months), 109 patients (14.6%) died, while 639 patients (85.4%) survived."

Addition of Risk Tables to Survival Curves:

To improve transparency regarding patient retention and exposure transitions over time, we have updated Figure 1 (High-Dose IVC) and Figure 2 (Mistletoe) to incorporate number-at-risk tables displayed beneath the plot axes at 12-month intervals across the 60-month observation period

Q20. Since Cox proportional-hazards regression is the main survival analysis used in this study, the authors should state whether the proportional-hazards assumption was evaluated. If it was tested, please briefly describe the method used and whether any important violations were found.

Answer) We thank the reviewer for highlighting the fundamental importance of evaluating the proportional-hazards (PH) assumption for our regression models.

As described in Section 2.5 (Statistical Analysis) of the revised manuscript, the proportional-hazards assumption for all Cox regression models was formally assessed using scaled Schoenfeld residuals tests and visual inspection of residual plots.

No statistically significant violations of the proportional-hazards assumption were detected for any of the covariates included in the multivariable models (all $p > 0.05$). Furthermore, to appropriately account for the variable timing of complementary therapy initiation and avoid bias related to time-varying treatment exposure, interventions such as high-dose IVC and mistletoe therapy were modeled as time-dependent covariates using a counting-process (start-stop) structure. This analytical approach preserves the proportionality assumption across the entire observation period.

We have expanded Section 2.5 in the Methods to explicitly detail the evaluation method and results of the PH assumption test.

Q21. “Mega-vitamin” is included as a separate treatment in Tables 1-2 and 2, but it is not clearly described in the Methods. Therefore, please provide the composition, dose, route, treatment duration, and criteria used to define exposure. Since high-dose IVC is analyzed separately, it is also important to clarify whether there is any overlap between the “mega-vitamin” and IVC groups.

Answers) We sincerely thank the reviewer for identifying this potential source of confusion and allowing us to clarify our variable categorization.

Upon auditing our treatment registry and database codebook, we confirmed that "Mega-vitamin" and "High-dose IVC (Intravenous Vitamin C)" referred to the exact same therapeutic modality (intravenous ascorbic acid infusion at 10–100 g/session) and were erroneously listed as separate entities due to duplicate institutional billing codes in the preliminary dataset.

To eliminate redundancy and ensure absolute diagnostic precision, we have removed the redundant "Mega-vitamin" classification entirely and consolidated all corresponding exposure and duration data into a single, unified High-Dose IVC category across all tables (Tables 1 & 2), figures, and text in the revised manuscript.

We have also updated Section 2.3 (Palliative Care Interventions) in the Methods section to provide a comprehensive description of the composition, dose, route, and exposure criteria for High-Dose IVC.

Q22. The counts and percentages reported for hyperthermia therapy in Table 1-2 do not seem to match. The table lists 359 patients overall, 299 survivors, and 60 deaths as receiving hyperthermia therapy, but the corresponding percentages do not appear to be consistent with these numbers. The percentages reported in the Results section should also be checked against the table. Please verify these values and correct them where necessary.

Answers) We sincerely thank the reviewer for carefully reviewing the data tables and bringing this reporting discrepancy to our attention.

Following your comment, we conducted a comprehensive audit of the raw dataset and verified all denominators and row/column percentage calculations across Table 1, Table 2, and the manuscript text.

Table 1 Corrections:

The counts and column percentages for hyperthermia therapy have been recalculated based on the exact group-specific denominators (N = 748 overall; n = 639 survivors; n = 109 deaths):

Overall (N = 748): n = 359 (48.0%)
Survivors (n = 639): n = 299 (46.8%)
Deaths (n = 109): n = 60 (55.0%)

Textual Consistency:

All hyperthermia therapy counts and percentage values reported in Section 3.1 (Results) of the text have been updated to ensure 100% alignment with the revised Table 1 and Table 2.

Reviewer 4 Report

Comments and Suggestions for Authors

This article explores the potential benefit of high-dose intravenous vitamin C and mistletoe therapy on overall survival in breast cancer patients within the context of palliative care. The study boasts a large sample size and a prospective cohort design, and its topic aligns with current hot topics in comprehensive cancer treatment, possessing significant clinical translational value. However, as a non-randomized observational study, the manuscript has several issues:

1. The authors defined in the methodology that patients must complete "at least 3 months of treatment and more than 10 cycles" to be included in the IVC or mistletoe treatment group. This implies that patients in the treatment group "must" survive for at least 3 months to meet the inclusion criteria, while those who die within these 3 months are automatically relegated to the control group. This design artificially exaggerates the survival advantage of the treatment group. Although the authors acknowledge this limitation in the discussion section, no statistical measures were taken to correct for it in the actual analysis. It is strongly recommended that the authors re-analyze the data. A Cox regression model with time-dependent covariates must be used to eliminate this serious survival bias. Without correction, the reliability of HR values of 0.377 and 0.446 would be seriously questionable.

2. The dosage range for IVC is very wide (10-100g). Combining 10g and 100g into a single binary variable (yes/no) for analysis is unscientific. It is recommended that authors stratify the treatment groups based on cumulative dose, average single dose, or treatment duration to assess the existence of a clear dose-survival response relationship. This would greatly enhance the biological persuasiveness of the conclusions.

In conclusion, the manuscript needs to be accepted after revision.

Author Response

Reviwer 4

Q23. The authors defined in the methodology that patients must complete "at least 3 months of treatment and more than 10 cycles" to be included in the IVC or mistletoe treatment group. This implies that patients in the treatment group "must" survive for at least 3 months to meet the inclusion criteria, while those who die within these 3 months are automatically relegated to the control group. This design artificially exaggerates the survival advantage of the treatment group. Although the authors acknowledge this limitation in the discussion section, no statistical measures were taken to correct for it in the actual analysis. It is strongly recommended that the authors re-analyze the data. A Cox regression model with time-dependent covariates must be used to eliminate this serious survival bias. Without correction, the reliability of HR values of 0.377 and 0.446 would be seriously questionable.

Answers)

Patients were classified as unexposed from baseline until the exact date of their first therapy session, and as exposed thereafter. Survival curves were reconstructed using the Simon–Makuch method and evaluated via the Mantel–Byar test. Even under this rigorous time-dependent model, protective treatment effects remained statistically significant:

High-Dose IVC: Time-dependent adjusted Hazard Ratio (aHR) = 0.561 (95% CI: 0.319–0.986, p = 0.045) (Manuscript: Page 4, Lines 112–128 & Page 7, Table 2)
Mistletoe Therapy: aHR = 0.463 (95% CI: 0.231–0.927, p = 0.030) (Manuscript: Page 4, Lines 112–128 & Page 7, Table 2)

Landmark analysis was not chosen because patients initiated therapies at highly non-uniform time points, and a single arbitrary landmark time would cause substantial sample loss and misclassification. We converted all treatment exposures to time-dependent covariates, yielding adjusted HRs of 0.561 (p = 0.045) for High-Dose IVC and 0.463 (p = 0.030) for Mistletoe. (Manuscript: Page 4, Lines 112–128 & Page 7, Table 2)

Q24. The dosage range for IVC is very wide (10-100g). Combining 10g and 100g into a single binary variable (yes/no) for analysis is unscientific. It is recommended that authors stratify the treatment groups based on cumulative dose, average single dose, or treatment duration to assess the existence of a clear dose-survival response relationship. This would greatly enhance the biological persuasiveness of the conclusions.

Answers) We sincerely thank the reviewer for this insightful and methodologically crucial comment. We fully agree that evaluating a dose–survival response relationship (e.g., based on cumulative dose or average single dose) would substantially enhance the biological plausibility and scientific rigor of our findings.

Unfortunately, due to the real-world retrospective nature of our electronic medical record (EMR) dataset, individual variations in dosage titration over time and missing detailed cumulative dose logs prevented us from performing a reliable continuous dose-stratified sub-analysis without introducing substantial missing data bias or misclassification.

However, to mitigate this potential bias in our binary classification, we strictly enforced a minimum therapeutic threshold: only patients who completed at least 10 sessions over a minimum treatment duration of 3 months were classified into the IVC exposure group. This criterion ensured that patients receiving transient, insufficient, or sporadic administrations were excluded from the exposed cohort.

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

Reviewer 3 Report

Comments and Suggestions for Authors

The authors have addressed the main methodological concerns raised in the previous review. In particular, the revised analysis now accounts for time-varying treatment exposure, includes clinically relevant tumor- and treatment-related covariates, and defines complementary therapies and follow-up procedures more clearly. The additional checks of the proportional-hazards assumption and the corrections to the treatment categories and reported percentages also improve the analysis's reliability. The conclusions are now presented with appropriate caution for an observational study. I have no further major concerns and consider the revised manuscript suitable for publication.

Reviewer 4 Report

Comments and Suggestions for Authors