The Prognostic Value of Serum Biomarkers for Survival of Children with Osteosarcoma of the Extremities
, Marco Manfrini 1, Shinji Tsukamoto 3, Andreas F. Mavrogenis 4, Barbara Bordini 2, Davide Maria Donati 1 and Costantino Errani 1,*
Round 1
Reviewer 1 Report
The authors conducted a retrospective study to investigate prognostic factors in 210 children who underwent treatment for appendicular osteosarcoma. The study examined several factors, including the patient's age and sex, tumor size and location, presence of metastasis, response to chemotherapy, and various blood biomarkers. According to a multivariate Cox regression model, a high level of alkaline phosphatase (AP), poor response to chemotherapy, and the presence of metastatic disease were found to be significantly associated with an increased risk of death within 5 years. Based on these findings, the authors recommend measuring levels of AP and ESR prior to treatment for osteosarcoma. This can help identify children who may have an increased risk of death or local recurrence, respectively.
However, the authors acknowledged several limitations to the study, including its retrospective nature, the fact that it was conducted at a single institution, and the small sample size. These limitations may reduce the statistical power of the study and should be taken into account when interpreting the results.
In addition, this manuscript can be improved in the following respects:
1. In all the Kapplan-Meier figures, the meaning of red and blue needs to be clearly indicated. Meanwhile, The y-axis legend should be more specific, rather than just stating "Kaplan-Meier". In addition, it’s better to show the exact p-values in the figures.
2. Grammatical errors should be carefully checked throughout the manuscript. For instance, in Figure 1, "Patients who did not underwent surgery" should be corrected to "Patients who did not undergo surgery."
The English language used in this manuscript is good
Author Response
Dear reviewer,
We changed the manuscript following your suggestions.
Thank you,
Costantino Errani
Author Response File: 
Author Response.pdf
Reviewer 2 Report
From a biostats and clinical epidemiology point of view, I have some questions for the Authors:
- table 1, number and % should be renamed as absolute/relative frequencies
- table 1 and everywhere, age and other continuous covariates should be reported as median/IQR and not mean/sd
- line 104, for the same reasons, you have to estimate and report median f-fup for the only surviving pts and not mean one!
- line 130, OS should be defined as "... death for any cause". It would be interesting to estimate the CSS (Cancer Specific Survival) too, even I do imagine that the vast majority of deaths were cancer-related
- line 140 and everywhere, p-values always with 3-sign digits
- line 142, death causes are lacking, all due to primary cancer!? please, specify it
- table 2, to be totally redone. It seems you are reporting absolute frequencies (since % is lacking). But, as for the title, you have to report OS at 5 yrs, ok!
- line 164, this KM curve and the relative estimations have to be redone by a landmark point! (i.e. Analysis of Survival by Tumor Response, JCO 1983). The current estimation is biostatistically wrong!
- table 3, to be totally redone. You have to report, side to side, the numerical results for the whole series of univariate models and the multivariate one, without forest plots. Moreover, clinical response must be treated as a time-varying covariate in the Cox models (i.e. Analysis of Survival by Tumor Response, JCO 1983), the current estimation is biostatistically wrong!
- figure 5, not to be shown, due to the poor number of events
- table 3 and 5, are they OS estimations? please, state it clearly
Minor editing
Author Response
Dear Reviewer,
We changed the manuscript following your suggestions.
1) table 1, number and % should be renamed as absolute/relative frequencies:
we changed table 1
2) table 1 and everywhere, age and other continuous covariates should be reported as median/IQR and not mean/sd
we changed everywhere
3) line 104, for the same reasons, you have to estimate and report median f-fup for the only surviving pts and not mean one!
We disagree on the utility of median follow-up limited to survivors. Given that unfortunately in our study 64 of the 210 patients die within 5 years, what information should the follow-up calculated only on the survivors? Considering then the fact that, as shown in figure 1, we excluded from the analysis the alive patients with a follow up of less than 5 years
4) line 130, OS should be defined as "... death for any cause". It would be interesting to estimate the CSS (Cancer Specific Survival) too, even I do imagine that the vast majority of deaths were cancer-related
We analyzed children with osteosarcoma and sadly they all died of the disease
5) line 140 and everywhere, p-values always with 3-sign digits
We changed following your suggestions.
6) line 142, death causes are lacking, all due to primary cancer!? please, specify it
Yes all patients died of disease
7) table 2, to be totally redone. It seems you are reporting absolute frequencies (since % is lacking). But, as for the title, you have to report OS at 5 yrs, ok!
We changed table 2 following your suggestion.
8) line 164, this KM curve and the relative estimations have to be redone by a landmark point! (i.e. Analysis of Survival by Tumor Response, JCO 1983). The current estimation is biostatistically wrong!
table 3, to be totally redone. You have to report, side to side, the numerical results for the whole series of univariate models and the multivariate one, without forest plots. Moreover, clinical response must be treated as a time-varying covariate in the Cox models (i.e. Analysis of Survival by Tumor Response, JCO 1983), the current estimation is biostatistically wrong!
The two articles cited are very interesting and useful in the case of statistical analyzes in which the interest is on the treatment, in fact they consider the different times in which a treatment can take effect. But in our article the scope of the work is different and the analysis performed is appropriate to try to understand if there are prognostic factors at the onset of the disease without analyzing the treatment which is the same for all patients. See our previous articles ( The Prognostic Value of the Serum Level of C-Reactive Protein for Survival of Children with Ewing's Sarcoma Cancers (Basel). 2023 Mar 3;15(5):1573. doi: 0.3390/cancers15051573 and C-reactive protein and tumour diagnosis predict survival in patients treated surgically for long bone metastases Int Orthop. 2021 May;45(5):1337-1346)
9) figure 5, not to be shown, due to the poor number of events
We deleted fig 5 following your suggestion.
10) table 3 and 5, are they OS estimations? please, state it clearly
No you are right.. they are HR! We changed it following your suggestion.
Author Response File: 
Author Response.docx
Round 2
Reviewer 2 Report
Some critical points do remain to be solved, i.e. landmark analysis for KM curves stratified by clinical response (actually they are wrong) and median follow-up estimation technique
Author Response
I think that the misunderstanding regarding your comments and our answers is related to different study population. Our patients are children with osteosarcoma in which one of the the main prognostic factor known so far was not the clinical response to chemotherapy but the chemotherapy-induced tumor necrosis. Anderson's article (Analysis of Survival by Tumor Response James R. Anderson, Kevin C. Cain, and Richard D. Gelber) suggests a statistical analysis of elderly cancer patients in which clinical response clearly stratifies patients into those who will survive and those who will not survive. In children with sarcoma, on the other hand, poor tumor necrosis does not necessarily stratify patients who will survive from those who will not survive but into patients who are more likely to relapse. Following your suggestions, we made a statistical analysis only of the good responders, ie children with necrosis more than 90%, and even in these group of patients the AP predicts the prognosis (see the following figure).
To avoid misunderstandings, we decided to change the terminology in the article, referring to the article by Lozano et al. (Complete tumor necrosis after neoadjuvant chemotherapy defines good responders in patients with Ewing sarcoma, Cancer. 2023 Jan 1;129(1):60-70. doi: 10.1002/cncr.34506. Epub 2022 Oct 28), which divides the patients in good responders and poor responders based on the term of “chemotherapy-induced tumor necrosis” and not “response to chemotherapy”.
Author Response File: Author Response.pdf
Round 3
Reviewer 2 Report
The Authors decided not to follow several reviewer' suggestions
minor
