Contrast-Enhanced CT Texture Analysis in Colon Cancer: Correlation with Genetic Markers
Round 1
Reviewer 1 Report
This retrospective study correlated texture analysis features from portal venous phase CTs performed within 1 month before locally advanced colon cancer resection with cancer genomic features of interest, specifically KRAS mutation, NRAS mutation, BRAF mutation, and MSI. Of the 56 textural features assessed, the authors found one feature that was a significant predictor of NRAS mutation and four features that were significant predictors of MSI. Although the study's findings will need to be validated on other patient cohorts / scanners, the results are promising.
My main criticisms are the following:
1. The authors assessed whether 56 textural features could predict 4 genomic features. This amounts to 224 statistical tests. At an alpha of 0.05, one would expect more than 10 textural features to be associated with at least one genomic feature due to chance alone. Thus, the manuscript would be strengthened if the associations persist after correcting for multiple comparisons.
2. The justification for identifying textural features that predict these genomic features is that tumors can be heterogeneous and therefore subject to sampling error. However, all tumors in this study were resected, thereby providing the entire tumor for assessment of mutation status. Do the authors think that predicting mutation status via textural analysis would be useful for making management decisions prior to resection of the primary tumor? If so, how? The clinical value of the manuscript would be strengthened by clarifying this point.
Author Response
First of all, we want to thank the reviewer for the suggestions that have helped us to improve substantially the manuscript.
This retrospective study correlated texture analysis features from portal venous phase CTs performed within 1 month before locally advanced colon cancer resection with cancer genomic features of interest, specifically KRAS mutation, NRAS mutation, BRAF mutation, and MSI. Of the 56 textural features assessed, the authors found one feature that was a significant predictor of NRAS mutation and four features that were significant predictors of MSI. Although the study's findings will need to be validated on other patient cohorts / scanners, the results are promising.
RE: Thank you for the positive comment
My main criticisms are the following:
- The authors assessed whether 56 textural features could predict 4 genomic features. This amounts to 224 statistical tests. At an alpha of 0.05, one would expect more than 10 textural features to be associated with at least one genomic feature due to chance alone. Thus, the manuscript would be strengthened if the associations persist after correcting for multiple comparisons.
RE: Thanks for raising the point. We did not write it in the text but we used a correction for the Mann-Whitney test results with Bonferroni and we selected only those feature that showed significant p value in both. We have added this point in the materials and methods section.
“Texture parameters were compared between the two groups using the Mann-Whitney test and Bonferroni correction. A p-value of <0.05 was considered statistically significant.”
- The justification for identifying textural features that predict these genomic features is that tumors can be heterogeneous and therefore subject to sampling error. However, all tumors in this study were resected, thereby providing the entire tumor for assessment of mutation status. Do the authors think that predicting mutation status via textural analysis would be useful for making management decisions prior to resection of the primary tumor? If so, how? The clinical value of the manuscript would be strengthened by clarifying this point.
RE: We thank the reviewer for underlying this important matter. Indeed, chemotherapy is moving toward mutation-selective treatments and so the knowledge of which mutations are present inside a colon cancer before surgery could be of huge importance for the oncologists.
In order to better underline this point, we have added a sentence in the discussion:
“Indeed, it has been demonstrated that for certain types of mutations a targeted therapy is feasible (e.g. KRAS G12C inhibitor), hence, the knowledge of which mutations are present inside a colon cancer before surgery could be of paramount importance for the oncologists in order to set up a correct and targeted chemotherapy [29,30].”
Author Response File: Author Response.docx
Reviewer 2 Report
The authors aimed to correlate histogram patterns of venous-enhanced CT examinations of patients with CRC with the presence or absence of mutations. The primary tumour was examined in the case of advanced stages AJCC III/IV, in connection with BRAF, KRAS, NRAS, and MSI analyses. The results are in part somewhat sobering, in part promising. For BRAF, KRAS, NRAS rather sobering, for MSI better, but overall within expectations.
The methodology of the study is fine as far as it goes, execution and results including the statistical methods without any recognisable formal deficiencies.
The discussion is detailed and contains essential points. In particular, the limitations were elaborated. These include in particular the small number of patients in the overall rather extended time interval, which reflects the general problem of such complex evaluations, which is probably difficult to avoid.
Some minor formal errors in the manuscript should be mentioned here:
The abbreviation MSS is not explained
The abbreviation WT is not explained
In Figure 2, "Stage 1-2" should be written in Roman numerals, i.e. "Stage I/II".
On page 4, "Houlsfield" should be replaced by "Hounsfield".
Author Response
First of all, we want to thank the reviewer for the suggestions that have helped us to improve substantially the manuscript.
The authors aimed to correlate histogram patterns of venous-enhanced CT examinations of patients with CRC with the presence or absence of mutations. The primary tumour was examined in the case of advanced stages AJCC III/IV, in connection with BRAF, KRAS, NRAS, and MSI analyses. The results are in part somewhat sobering, in part promising. For BRAF, KRAS, NRAS rather sobering, for MSI better, but overall within expectations.
The methodology of the study is fine as far as it goes, execution and results including the statistical methods without any recognisable formal deficiencies.
The discussion is detailed and contains essential points. In particular, the limitations were elaborated. These include in particular the small number of patients in the overall rather extended time interval, which reflects the general problem of such complex evaluations, which is probably difficult to avoid.
RE: We thank the reviewer for the positive comments.
Some minor formal errors in the manuscript should be mentioned here:
The abbreviation MSS is not explained
The abbreviation WT is not explained
RE: Thank you for raising the point. We have added the acronyms written in extenso in the materials and methods section both for MSS and WT.
In Figure 2, "Stage 1-2" should be written in Roman numerals, i.e. "Stage I/II".
RE: Thank you for the suggestion, we have modified the Figure accordingly.
On page 4, "Houlsfield" should be replaced by "Hounsfield".
RE: Thank you, we have corrected the typo.
Author Response File: Author Response.docx