Optimal Choice as First-Line Therapy for Patients with Triple-Negative Breast Cancer: A Bayesian Network Meta-Analysis
Pier Adelchi Ruffini
Round 1
Reviewer 1 Report
The authors reviewed randomized clinical trials in advanced or metastatic TNBC to identify the optimal first line choice. They applied a Bayesian framework to this analysis to facilitate indirect comparisons
They acknowledge limitations of their work (page 7) but there are more
MAJOR OBSERVATIONS
In fact, this reviewer’s main concern is that heterogeneity inherent with TNBC was overlooked in this work, leading to clinically counterintuitive conclusions
As an example, chemotherapy plus AKT inh emerged as the optimal first choice. This is based upon 3 clinical trials where ipatasertib or capivasertib, both AKT inhibitors, in combination with paclitaxel were compared with paclitaxel alone. Trials with ipatasertib (Kim S-B et al., Lancet Oncology 2017 and Dent R et al., SABCS 2020) overall failed to demonstrate an advantage over paclitaxel alone in TNBC. On the other hand, a relatively small study with capivasertib (Schmid P et al., Journal of Clinical Oncology 2020) showed promise of superior efficacy over paclitaxel alone and this hypothesis is being formally tested in a phase 3 trial (NCT03997123). Thus, the conclusion that chemotherapy + AKTinh. Is likely the most effective therapy is challenged by negative results of ipatasertib and can barely rely on positive results of capivasertib. Furthermore, the authors should specify that these agents may be more effective in TNBC with alterations in PIK3CA/AKT1/PTEN pathway
That being said, the findings are interesting and could prompt additional clinically relevant investigations
MINOR OBSERVATIONS
Introduction:
2nd line, “proportion” should be “of all new diagnoses of breast cancer”
10th line, “this population” should be “in this population”
Materials and Methods:
2.1 Study selection and outcome measures
Why were only 4 TRAEs considerd for the analysis?
line 16 “by the edition” lacks the number of the edition
Results:
1st line, “literature” should be “articles”
Figure 1 legend, “society” should be “Society”
Figure 2 legend, “undergone” should be “included in”
Figure 3 legend: “enrolled” should be “selected” or “evaluated”
Discussion:
1st line “decent” should be removed
Author Response
Response to Reviewer 1 Comments
Point 1: In fact, this reviewer’s main concern is that heterogeneity inherent with TNBC was overlooked in this work, leading to clinically counterintuitive conclusionAs an example, chemotherapy plus AKT inh emerged as the optimal first choice. This is based upon 3 clinical trials where ipatasertib or capivasertib, both AKT inhibitors, in combination with paclitaxel were compared with paclitaxel alone. Trials with ipatasertib (Kim S-B et al., Lancet Oncology 2017 and Dent R et al., SABCS 2020) overall failed to demonstrate an advantage over paclitaxel alone in TNBC. On the other hand, a relatively small study with capivasertib (Schmid P et al., Journal of Clinical Oncology 2020) showed promise of superior efficacy over paclitaxel alone and this hypothesis is being formally tested in a phase 3 trial (NCT03997123). Thus, the conclusion that chemotherapy + AKTinh. Is likely the most effective therapy is challenged by negative results of ipatasertib and can barely rely on positive results of capivasertib.
Response 1: Thank you for your critical comment! We have concluded the respective data included in our study as Table S1. As the table shows, the survival outcomes varied among different clinical trials, to which several confounding factors could lead. Actually, the heterogenous outcomes in these clinical trials were exactly the main reason for us to perform this pooled analysis and obtain the synthesized conclusion. We understand the limitations potentially affect the reliability of the conclusion in meta-analysis and have elaborated on them in the Discussion section. At present, we are carrying out a retrospective population-based study to explore the optimal therapeutic option for patients with metastatic TNBC. We think the findings at that time will provide more practical evidence for physicians. We hope this kind of revision could meet your approval. We sincerely hope this kind of revision could meet your approval. Thank you!
Point 2: Furthermore, the authors should specify that these agents may be more effective in TNBC with alterations in PIK3CA/AKT1/PTEN pathway
Response 2: Thank you for your rigorous consideration. We have added this specification in the Discussion section (as tracked change suggested). Thank you!
Point 3: That being said, the findings are interesting and could prompt additional clinically relevant investigations
Response 3: Really appreciate your comment.
Point 4: 2nd line, “proportion” should be “of all new diagnoses of breast cancer”
Response 4: Thank you very much. We have revised this phase in accordance with your opinion.
Point 5: 10th line, “this population” should be “in this population”
Response 5: Thank you very much. We have revised this phase in accordance to your opinion.
Point 6: Why were only 4 TRAEs considerd for the analysis?
Response 6: Thank you very much for your critical comment. For the safety profiles, based on the publicly available data (from published articles or ClinicalTrials.gov), we first selected the most common Top 10 adverse events (AEs) recorded in clinical trials (N=10). Next, considering that we should take into account all the enrolled treatments, we selected the overlapped AEs to represent the general side effects tending to occur in patients with TNBC. In the meantime, the supervisors (B.X. and J.W.) reviewed and decided to pick up the AEs most relevant to the clinic to proceed. Finally, we adopted the most common Top 4 as well as the clinical-associated AEs for this meta-analysis. We considered this processing will avoid bias to the utmost extent and provide the most significant reference to practitioners. We hope this interpretation could meet your approval. Thank you!
Point 7: line 16 “by the edition” lacks the number of the edition
Response 7: Sorry about this confusion. We have revised this phase in accordance with your suggestion.
Point 8: 1st line, “literature” should be “articles”
Response 8: Thank you very much. We have revised this phase in accordance to your suggestion.
Point 9: Figure 1 legend, “society” should be “Society”
Response 9: Sorry about this mistake. We have revised this phase in accordance with your suggestion.
Point 10: Figure 2 legend, “undergone” should be “included in”
Response 10: Thank you very much. We have revised this phase in accordance with your suggestion.
Point 11: Figure 3 legend: “enrolled” should be “selected” or “evaluated”
Response 11: Thank you very much. We have revised this phase in accordance with your suggestion.
Point 12: 1st line “decent” should be removed
Response 12: Thank you very much. We have revised this phase in accordance with your suggestion.
Reviewer 2 Report
The authors provide a good meta-analysis identifying a potentially effective therapy for TNBC. The overall study looks into treatment for significant number of patients, thus possibly important for their findings. Some of the comments
1. they authors could describe R packages used for the analysis. It is not very clear how the SUCRA analysis was performed. How does it differ from standard ROC curve analysis.
2. is it possible to add any km plot for the PFS? the fire 3 and 4 are confusing. it is not immediately clear how the hazard ratios are correlated with treatment and how different treatment differ. The authors could clarify.
3. Were these analysis for HR conducted using univariate/multivariate model. Was there any other factor affecting treatment other than TNBC.?
Author Response
Response to Reviewer 2 Comments
Point 1: The authors could describe R packages used for the analysis. It is not very clear how the SUCRA analysis was performed.
Response 1: Sorry about this confusion. All the meta-analyses in this study were performed using the R package gemtc (version 1.0-1). We have added this detail in the Statistical analysis of Materials and Methods section.
Regarding the SUCRA analysis, I post the R codes as follows:
rank.probability(result.fe,preferredDirection=-1)
rankings<-rank.probability(result.fe,preferredDirection=-1)
sucra(rankings)
sucra<-sucra(rankings)
plot(sucra)
Thank you!
Point 2: How does it differ from standard ROC curve analysis.
Response 2: Thank you for your comment. The plot of the ROC curve is based on the individual data and our study, in contrast, was carried out by virtue of a published study (the pooled analysis of individual data). Therefore, we could not accomplish the ROC curve right now. However, really appreciate your consideration, and we will adopt this option for statistical analysis in the following research. Thank you!
Point 3: Is it possible to add any km plot for the PFS? the fire 3 and 4 are confusing. it is not immediately clear how the hazard ratios are correlated with treatment and how different treatment differ. The authors could clarify.
Response 3: Thanks for your consideration and sorry about this kind of confusion. Figures 3-4 presented the comparative HR and 95% CrI for the PFS and OS. Data were presented as hazard ratio and 95% credible interval. The upper right values displayed the top treatment compared to the bottom treatment, while the lower left values demonstrated the converse comparisons. For the network meta-analysis, this type of illustration could be the collective data and reveal the most significant findings. Regarding the KM curve, it is plotted based on the individual survival data which is different from our data type, thus we cannot complete it at this moment. We are designing further studies to explore the better therapeutic choice for TNBC according to the population-based data. Thank you!
Point 4: Were these analysis for HR conducted using univariate/multivariate model. Was there any other factor affecting treatment other than TNBC.?
Response 4: Thank you for your comment. The pooled analysis for HR and 95% CrI was based on the Bayesian model for network meta-analysis. The main reason for this choice is the lack of direct head-to-head comparisons among different therapeutic agents and exploring the optimal choice for mTNBC is a big challenge for current clinical practice. The manipulation of univariate/multivariate models requires individual-based data with complete clinicopathological characteristics. We are sorry that we cannot facilitate this statistical model right now. However, as you mentioned, we will manipulate these models to rule out confounding and identify the most significant variables affecting the therapeutic options in the following population-based study. We hope this kind of revision could meet your approval. Thank you!
Round 2
Reviewer 1 Report
The authors have considerably improved the manuscript
This reviewer has only a few comments/suggestions at this stage
§ 2.1 lines 71-72: TRAEs of rash, pruritus, neuropathy, neutropenia are to be recorded and analyzed but in
§ 3.2 lines 156-157: TRAEs of rash, diarrhoea, neuropathy, neutropenia are listed
it should be the same TRAEs in both sections, this reviewer believes
§ 2.1 line 61: society should be Society
§ 2.1 line 62: “were also undergone assessment” could be “were also subjected to assessment” or some other rephrasing
Table S1 publication by Dent R et al is most likely referred to cohort A of the Ipatunity trial, as one can infer from NCT number. The correct reference is Dent R et al., Cancer Res (2021) 81 (4_Supplement): GS3-04.
DOI: https://doi.org/10.1158/1538-7445.SABCS20-GS3-04
it is an abstract from SABCS 2020
The authors may want to include in the reference list and possibly comment on also final results with OS data from the LOTUS trial with ipatasertib. The publication is Dent R et al., Breast Cancer Res Treat 189:377-386, 2021
Author Response
- 2.1 lines 71-72: TRAEs of rash, pruritus, neuropathy, neutropenia are to be recorded and analyzed but in
- 3.2 lines 156-157: TRAEs of rash, diarrhoea, neuropathy, neutropenia are listed
it should be the same TRAEs in both sections, this reviewer believes
Response: Sorry about this confusion. Have revised the content according to your comment.
- 2.1 line 61: society should be Society
Response: Thank you for your critical comment. We already revised the word correctly.
- 2.1 line 62: “were also undergone assessment” could be “were also subjected to assessment” or some other rephrasing
Response: Thank you for your critical comment. We already revised the phrase correctly.
Table S1 publication by Dent R et al is most likely referred to cohort A of the Ipatunity trial, as one can infer from NCT number. The correct reference is Dent R et al., Cancer Res (2021) 81 (4_Supplement): GS3-04.
DOI: https://doi.org/10.1158/1538-7445.SABCS20-GS3-04
it is an abstract from SABCS 2020
The authors may want to include in the reference list and possibly comment on also final results with OS data from the LOTUS trial with ipatasertib. The publication is Dent R et al., Breast Cancer Res Treat 189:377-386, 2021
Response: Thank you for your rigorous comment. We have modified the references according to your information.
Thank you again for your efforts in our manuscript!
