The Geriatric G8 Score Is Associated with Survival Outcomes in Older Patients with Advanced Prostate Cancer in the ADHERE Prospective Study of the Meet-URO Network

Round 1

Reviewer 1 Report

It was my pleasure to review this manuscript entitled “The geriatric G8 score predicts survival outcomes in older patients with advanced prostate cancer in the ADHERE prospective study of the Meet-URO network”. This subject seems very interesting. Overall article is well written English is fluently and adequate, title very informative about paper.

The authors should improve the introduction with prostate cancer epidemiology

Please provide the strength of this study

Author Response

Reviewer #1

It was my pleasure to review this manuscript entitled “The geriatric G8 score predicts survival outcomes in older patients with advanced prostate cancer in the ADHERE prospective study of the Meet-URO network”. This subject seems very interesting. Overall article is well written English is fluently and adequate, title very informative about paper.

The authors should improve the introduction with prostate cancer epidemiology.

Please provide the strength of this study.

Re: thank you very much for appreciating the manuscript and your valuable comments.

Regarding the prostate cancer epidemiology we have added the following to the introduction paragraph: “Prostate cancer is the fourth most common cancer worldwide(1). Age-standardized incidence and mortality are 68 and 10 per 100,000, respectively, in more developed regions(2). The average age at which prostate cancer (PC) is diagnosed in western populations is 66 years old. However, at the time of diagnosis, 60% of patients are 65 years or older, and by 2040, this percentage will rise to 70%(3). Nearly 70% of PC deaths occur in men aged ≥ 75(1). While the overall mortality rate is predicted to remain steady, the number of men aged 70 and older who die from prostate cancer will nearly double by 2040(1). Furthermore, the median age of men who develop the meta-static disease is considerably older, and the median age of those who die due to pros-tate cancer is eighty years(4). Treatment costs for older men with early and late pros-tate cancer are already high and projected to rise in the following decades(3).”

We agree with the Reviewer that the strength of the evidence provided by the study and the ultimate message was not evident in the original manuscript form. Guidelines from scientific societies recommend geriatric screening by G8 for patients with prostate cancer. However, supporting evidence for patients with metastatic prostate cancer is scanty, particularly for those approaching an ARPI that is generally considered a low-toxicity anticancer agent for every patient. By the present study, which focused on older patients with mCRPC treated with ARPIs, we could confirm the relevance of geriatric assessment and G8 as a reliable screening tool. In this setting, G8 can select patients who deserve comprehensive geriatric assessment (CGA) to identify frail or vulnerable patients. As suggested by the International Society of Geriatric Oncology (SIOG), best supportive care (BSC) remains the preferred option for the formers. Conversely, treatment optimisation for vulnerable patients carrying reversible clinical conditions should be considered, including the early introduction of palliative care. In the revised version, we added those considerations with appropriate reference in the Discussion paragraph. We also amended the abstract conclusion: “Baseline G8 screening is recommended for mCRPC patients aged ≥70 to optimise ARPIs in vulnerable individuals, including early introduction of palliative care.”  

Reviewer 2 Report

Thank you for the opportunity to review this study the association between geriatric G8 score and survival outcomes in older patients with advanced prostate cancer in the ADHERE prospective study.

General comments:

In the title, abstract and manuscript, could you please avoid using the words "predictive" or "predictor" and prefer the terms "prognosis" or "prognostic factor" or "associated factors". The term "predictor" should only be used in situations where a predictive model is performed, which is not the case here.

Methods:

1.               The choice of the median to create 2 groups seems to me questionable. I would advise you to choose cut-points based on validated thresholds in the literature, for example for age or Charlson index.

2.               The method should specify which guidelines are followed to assess progression (I assume Prostate Cancer Clinical Trials Working Group 3 [PCWG3]?) and cite the reference. Could you specify in the article each time that it is the radiographic PFS (rPFS)?

3.               You should specify that you are using a Cox model for the multivariate analysis and the variable selection method.

4.               There are too many variables for the number of patients and events in the multivariate analyses, I would advise you to use a variable selection method.

Results and figures:

1.             Table 1: I would advise you to also do Cox models for the univariate analysis and only present the median OS and PFS for the G8 score groups before the univariate and multivariate analyses.

2.               Table 2: Please specify for each variable the reference category. For variables with 3 categories, there must be 2 HR presented, this is not the case.

3.               Figure 1: Could you please indicate the censored observations on the survival curves (as recommended)?

Author Response

Reviewer #2

Thank you for the opportunity to review this study the association between geriatric G8 score and survival outcomes in older patients with advanced prostate cancer in the ADHERE prospective study.

General comments:

In the title, abstract and manuscript, could you please avoid using the words "predictive" or "predictor" and prefer the terms "prognosis" or "prognostic factor" or "associated factors". The term "predictor" should only be used in situations where a predictive model is performed, which is not the case here.

Re: thank you for the comment. We have substituted throughout the manuscript the “predictive” or “predictor” terms for “prognostic” or “prognostic factor” or “associated factor” as suggested. We rewrite the title (and running title) using ”is associated” instead of “predicts”.

Methods:

1. The choice of the median to create 2 groups seems to me questionable. I would advise you to choose cut-points based on validated thresholds in the literature, for example for age or Charlson index.

Re: thank you for the comment. We appreciate your standpoint well, as we had been puzzled by that choice that was ultimately taken up as no strong evidence support cut-offs for most of the numerical variables studied. A cut-off of 9 for Charlson comorbidity index (ACCI) has been used to screen patients over the age of 70 (Charlson et al: Validation of a combined comorbidity index. J Clin Epidemiol 47:1245-51, 1994) and of <12 months for time to castration resistance (Wenzel M et al, Front Oncol. 2021 Apr 23;11:659135. doi: 10.3389/fonc.2021.659135. eCollection 2021.). The cut-offs of ≥ 8 for Gleason and 6 for the six basic ADLs were adequate, while we did not find prognostic thresholds for the number of concomitant medications and age over 70. However, we assumed that 75 years could be realistic, as sometimes used in the subgroup analyses of clinical trials. Given the low evidence level for the above cut-offs, the peculiar old population of our study and your advice, we thought to analyse and report either the unsupervised median values and the literature-based thresholds for ACCI, time to castration-resistance, alongside age ≥75 years. We clarified it in the methods paragraph and reported the results in Table 1. At the univariable analysis, age ≥75 years and time to castration resistance <12 months were associated with worse OS and PFS, respectively, similar to the cut-offs of ≥78 and <31 months with non-superior p-values. The Charlson score by the cut-off of ≥9 was associated with worse PFS but not OS, whilst it was neither with the median value of ≥10. This factor was therefore included in the multivariable analysis.   

2. The method should specify which guidelines are followed to assess progression (I assume Prostate Cancer Clinical Trials Working Group 3 [PCWG3]?) and cite the reference. Could you specify in the article each time that it is the radiographic PFS (rPFS)?

Re: Thank you for the comment. Actually, in the ADHERE study, which has now been published (PMID: 35920559 DOI: 10.1093/oncolo/oyac147), the RECIST 1.1 were used to define the radiographic PFS (rPFS). We added to the methods paragraph that the rPFS was defined as the time from ARPI start to the date of disease progression on imaging as per RECIST 1.1, or death from any cause, whichever occurred first, and OS was calculated from the ARPI start date until death or the last follow-up. As suggested, we have specified throughout the manuscript, including tables and the figure, that the PFS was rPFS. Moreover, it was reported as rPFS also in the original above-mentioned ADHERE paper.

3. You should specify that you are using a Cox model for the multivariate analysis and the variable selection method.

Re: thank you very much for the comment. We apologise for that oversight. We have specified it in the Methods paragraph as it follows and in table 2. Cox-regression multivariable analysis (MVA) was performed for OS and PFS of clinical variables with a p-value < 0.05 at the UVA. When both median and literature-reported cut-off values of quantitative variables were significant, the one with the lowest p-value was carried on in the MVA.

4. There are too many variables for the number of patients and events in the multivariate analyses, I would advise you to use a variable selection method.

Re: thank you for the comment. As in reply to your previous comment, we specified that only clinical variables with a p-value < 0.05 at the univariable analysis were carried over to the multivariable analysis. Those were 4 for OS and 10 for rPFS, that perhaps can be considered acceptable given the number of patients (n=234) and the median follow-up of 15.4 months in metastatic castrate-resistant prostate cancer. However, we agree that a higher number of events would have improved the accuracy of the estimates of either the UVA or MVA, and we have added it to the study limitations.  

Results and figures:

1. Table 1: I would advise you to also do Cox models for the univariate analysis and only present the median OS and PFS for the G8 score groups before the univariate and multivariate analyses.

Re: thank you for the comment. In the Methods paragraph and Abstract, we specified that UVA and MVA were performed by Cox regression analysis. We recalculated the p-values and added relative HRs and 95% CIs by Cox in the UVA as reported in the new version table 1. As expected, little changed for the non-quantitative variables, by the Cox compared to the log-rank at UVA. Regarding the quantitative variables, the site of metastases lost its significance for OS but not for PFS, whilst we found significance for PFS according to the therapy setting (i.e., a worse outcome with using Enza or Abi post-ABI/ENZA was worse). The MVA was updated in table 2, results and abstract. Regarding the G8 score groups, curves, data and p-values by log-rank were reported in the revised Figure 1.

2. Table 2: Please specify for each variable the reference category. For variables with 3 categories, there must be 2 HR presented, this is not the case.

Re: thank you very much for the comment. As for the above comments, we analysed and presented data as you suggested in both the UVA and MVA in revised Tables 1 and 2 and manuscript results.

3. Figure 1: Could you please indicate the censored observations on the survival curves (as recommended)?

Re: thank you for the comment. We have indicated the censored observation on the survival curves in Figure 1.

Reviewer 3 Report

This is an interesting topic. However there needs to be a major revision

First of all,  G8 is abnormal if ≤ 14  and not the other way.  In the UVA analysis,  it seems that patients withs a  low G8  (≤ 14) have a longer survival. Is that the case or is there a mistake ?

PFS and OS must be defined  ( start=beginning of ARPI?).

The population characteristics should be described. The numbers in the UVA table, the text and figure 1 of supplementary data concerning G8>14 are not consistent.

Make sure abbreviations are all defined ( eg CR that seems to be castration resistance)

Author Response

Reviewer #3

This is an interesting topic. However there needs to be a major revision.

Re: thank you for appreciating the manuscript topic and your valuable comments.

1. First of all, G8 is abnormal if ≤ 14 and not the other way. In the UVA analysis, it seems that patients with a low G8 (≤ 14) have a longer survival. Is that the case or is there a mistake?

Re: thank you for the comment. We are sorry for the oversight regarding the direction of the of G8 score. Based on the other reviewers’ comments, alongside thoroughly double-checking the G8 and the other thresholds for consistency, the UVA and MVA were performed by Cox regression analysis based on median and literature-reported (where available) cut-off values of quantitative variables. We recalculated the p-values and added relative HRs and 95% CIs by Cox in the UVA as reported in the new version table 1. MVA was performed for OS and PFS of clinical variables with a p-value < 0.05 at the UVA. When both median and literature-reported cut-off values of quantitative variables were significant, the one with the lowest p-value was carried on in the MVA. For variables with 3 categories, two HR were presented. There were minimal changes in the p-values of non-quantitative variables by the Cox compared to the log rank in the UVA. Regarding the quantitative variables, the site of metastases lost its significance for OS but not for PFS, whilst we found significance for PFS according to the therapy setting (i.e., a worse outcome with using Enza or Abi post-ABI/ENZA was worse). The MVA was updated in table 2, results and abstract. Regarding the G8 score groups, curves, data and p-values by log-rank were reported in the updated Figure 1.

2. PFS and OS must be defined (start=beginning of ARPI?).

Re: thank you for the comment. By the original ADHERE study, which has now been published (PMID: 35920559 DOI: 10.1093/oncolo/oyac147), the RECIST 1.1 were used to define the radiographic PFS (rPFS). We added to the methods paragraph that the rPFS was defined as the time from treatment start to date of disease progression on imaging as per RECIST 1.1, or death from any cause, whichever occurred first, and OS was calculated from the ARPI start date until death or the last follow-up. Throughout the manuscript, including tables and the figure, we specified that the PFS was rPFS. Moreover, it was reported as rPFS also in the original above-mentioned ADHERE paper.

3. The population characteristics should be described. The numbers in the UVA table, the text and figure 1 of supplementary data concerning G8>14 are not consistent.

Re: thank you for the comment. We added Supplementary Table 1 with the characteristics of patients in the overall cohort and the ARPI-relative ones. We quoted that table at the beginning of the results paragraph. Moreover, we updated references with the full article of the ADHERE study to which the current analysis refers (PMID: 35920559 DOI: 10.1093/oncolo/oyac147). As mentioned, we have double-checked the G8-related and the other thresholds for consistency while the UVA and MVA were performed by Cox regression analysis.

4. Make sure abbreviations are all defined (eg CR that seems to be castration resistance)

Re: thank you for the comment. We double-checked alla the abbreviations, and CR for castration resistance seemed the only one was missing. We added it to the manuscript text and Tables.

Round 2

Reviewer 2 Report

Thank you for the opportunity to review this study the association between geriatric G8 score and survival outcomes in older patients with advanced prostate cancer in the ADHERE prospective study and for the reply.

1.                    Could you record only 1 cut-off per variable in the method, results and discussion, taking the new cut-offs?

2.                    Tables 1 and 2: For variables with 3 categories, it would seem better to indicate the p-value of the global variable (type III p-value) rather than p-value of each category?

Author Response

Thank you for the opportunity to review this study the association between geriatric G8 score and survival outcomes in older patients with advanced prostate cancer in the ADHERE prospective study and for the reply.

Re: thank you very much for your constructive and expert suggestions we very much appreciate it.

1. Could you record only 1 cut-off per variable in the method, results and discussion, taking the new cut-offs?

Re: thank you for this comment. In our original analysis, we considered cut-off values for quantitative variables the unsupervised median values as not much literature evidence on cut-offs is available. However, as you suggested, for three of those variables, we also tested the literature-reported thresholds, or the following: 9 for Charlson comorbidity score, <12 months for time to castration resistance (CR) and ≥ 75 for age. We specified in the methods paragraph that when both median and literature-reported cut-off values of those quantitative variables were significant, the one with the lowest p-value was carried over in the multivariable analysis. The results indicated that the literature-reported thresholds had lower p-values than the median ones for age and time to castration resistance, but the Charlson 9 score was significant on rPFS (contrarily to the median value of 10) and was therefore carried over in the multivariable Cox model. Leaving only one cut-off per those three variables would risk methodological inconsistency. Consequently, we are keen to leave the two cut-off values only for the above mentioned three variables. We hope you can agree with us.  

2. Tables 1 and 2: For variables with 3 categories, it would seem better to indicate the p-value of the global variable (type III p-value) rather than p-value of each category?

Re: thank you for this comment. In your previous comments, you suggested using the Cox model for the univariable analysis and specifying for each variable the reference category and, for variables with three categories, the two HRs. We followed your advice, and reported p-values in the univariable analysis for the two categories and relative p-values when three were present with the reference one. The site of metastases lost its significance for OS but not for rPFS in the univariable analysis, whilst we found significance for rPFS according to the therapy setting (i.e., a worse outcome with using Enza or Abi post-ABI/ENZA was worse). That granularity could not be detected by choosing the p-value of the global variable. For instance, the site of metastases would result again in a significant variable for OS that should be carried over in the multivariable analysis, which we think is not optimal as we cannot say which metastatic site is linked to the prognosis. Thus, we thought to leave it as it currently is. We hope you can agree with us.  

Reviewer 3 Report

Thank you for the corrections.

The rPFS and the % of patients withs G8 >14 and ≤ 14 are still inverted in the text p6=5 lines 120-121. They are correct in the table of figure 1

Author Response

Reviewer #3

Thank you for the corrections.

Re: thank you very much for your helpful and expert comments, very much appreciated.

1. The rPFS and the % of patients withs G8 >14 and ≤ 14 are still inverted in the text p6=5 lines 120-121. They are correct in the table of figure 1

Re: thank you for noting it. We are sorry for the reiterated mistake that we have now corrected. By checking again for G8 consistency, we have amended the numbers in Figure 1 OS.

2. English language and style are fine/minor spell check required

Re: thank you. We had already checked and amended it where needed in the previous revised version.