Missed Breast Cancers on MRI in High-Risk Patients: A Retrospective Case–Control Study

Round 1

Reviewer 1 Report

Very interesting paper, well written and quite clear

I am thankful to authors for their hard and valuable work. The manuscript is focused on a very hot topic in radiological practice and it aim to explain the lack of breast MRI to obtain the highest sensitivity and a stronger specificity, especially when used for early diagnosis in patients at high-risk of developing very aggressive cancers in young age.

There are few points I would like to point out to improve the quality of your manuscript:

Title
No comments

 

Abstract

No comments

Keywords

N/A

Introduction

No comments

 

Materials and methods:

Overall, this part is fine but probably a bit difficult to read.

• I suggest to better explain the inclusion criteria. In particular, I think it would be clearer to explain the definition of MR A and MR B in the first part of the paragraph, instead of the second part, and illustrate only after that the time between the two MRIs and the definition of missed/overlooked/misinterpreted lesions.
• I think it would be interesting to know whether there is a correlation between the number of missed lesions and the radiologist experience. Did you analyse this issue?

 

Results

I think that the results are clearly explained and supported by interesting tables.

In the first sentence the percentage of 74% is repeated twice, at the beginning and at the end of the sentence. I would fix this.

Discussion

No comments

 

Conclusions:

No comments.

 

References

No comments.

 

Tables
No comments

 

Figures
Beautiful pictures, really clear and interesting.

Author Response

 

REVIEWER 1

 

 

Comments and Suggestions for Authors

Very interesting paper, well written and quite clear

I am thankful to authors for their hard and valuable work. The manuscript is focused on a very hot topic in radiological practice and it aim to explain the lack of breast MRI to obtain the highest sensitivity and a stronger specificity, especially when used for early diagnosis in patients at high-risk of developing very aggressive cancers in young age.

Response: Thank you for your comment.

There are few points I would like to point out to improve the quality of your manuscript:

Title
No comments

 

Abstract

No comments

Keywords

N/A

Introduction

No comments

 

Materials and methods: 

Overall, this part is fine but probably a bit difficult to read.

 

(R1,1)

I suggest to better explain the inclusion criteria. In particular, I think it would be clearer to explain the definition of MR A and MR B in the first part of the paragraph, instead of the second part, and illustrate only after that the time between the two MRIs and the definition of missed/overlooked/misinterpreted lesions.

Response: Thank you for your comment. We made some changes following your kind suggestions.

 

 

(R1, 2)

I think it would be interesting to know whether there is a correlation between the number of missed lesions and the radiologist experience. Did you analyse this issue?

Response: we haven’t analyzed this factor. This an interesting topic that would warrant further publications.

 

Results

I think that the results are clearly explained and supported by interesting tables.

(R1,3)

In the first sentence the percentage of 74% is repeated twice, at the beginning and at the end of the sentence. I would fix this.

Response: This has been changed. Thank you!

 

Discussion

No comments

 

Conclusions:

No comments.

 

References

No comments.

 

Tables
No comments

 

Figures
Beautiful pictures, really clear and interesting.

Response: Thank you for your kind comments. This is very appreciated.

Author Response File: Author Response.docx

Reviewer 2 Report

This manuscript presents a retrospective case-control MRI study in high-risk breast cancer patients. The purpose is to see if MRI characteristics or molecular subtype or risk factors are associated with detectability (missed versus non missed) of breast cancer on MRI. The study selected 208 MRI exams of 104 patients for review. Each patient had a diagnosis of breast cancer (DCIS included) within 3 months of an MRI (MR B) and a most recent previous MRI (MR A). The inclusion criteria was that MR B had a recorded suspicious breast lesion (BI-RADS 4 or 5 AND MR A was either normal (BI-RADS 1), benign (BI-RADS 2) or probably benign (BI-RADS 3). The time between MR A and MR B can be between 6 to 26 months (mean: 11.1 months). In the review of 104 MR As, the radiologist categorized them into 2 groups: missed lesions and non missed lesions. Missed was defined as cancer was visible on MR A. Non missed was defined as the absence of mass or non-mass enhancement on MR A. MRI characteristics (T2 signal, size/volume, margins, distribution, enhancement kinetics) were documented. BRCA genes, molecular subtype and receptor subtype were also compared in missed or non missed. The study found 74% in 104 patients had breast cancer visible on MR A (the most recent MRI prior to their diagnosis). Their results did not find difference in MRI characteristics or receptor subtype between missed and non missed groups. They did find statistically significant association between BRCA 1 status (positive or negative) with detectability of breast cancer on MRI (odds ratio = 2.6).

These are major weaknesses of this study:

1. The manuscript was poorly written. Introduction did not show motivation or gaps in published works. Some content in Methods should go to Results. Table was numbered in a confusing way.
2. It was not clear whether one radiologist did all the review
3. Sample size was not appropriate for statistical analysis. In some categories, there were only 1 or 2 patients.
4. In Discussion, the authors claimed that this study comprises the highest number of false-negative breast MRI. It is because this is a case control study. The cohort was selected to have benign or normal MRI prior to the diagnosis. And since it’s not a blinded study, the radiologist was biased when he or she was reviewing the MR images.

Author Response

REVIEWER 2

 

Comments and Suggestions for Authors

This manuscript presents a retrospective case-control MRI study in high-risk breast cancer patients. The purpose is to see if MRI characteristics or molecular subtype or risk factors are associated with detectability (missed versus non missed) of breast cancer on MRI. The study selected 208 MRI exams of 104 patients for review. Each patient had a diagnosis of breast cancer (DCIS included) within 3 months of an MRI (MR B) and a most recent previous MRI (MR A). The inclusion criteria was that MR B had a recorded suspicious breast lesion (BI-RADS 4 or 5 AND MR A was either normal (BI-RADS 1), benign (BI-RADS 2) or probably benign (BI-RADS 3). The time between MR A and MR B can be between 6 to 26 months (mean: 11.1 months). In the review of 104 MR As, the radiologist categorized them into 2 groups: missed lesions and non missed lesions. Missed was defined as cancer was visible on MR A. Non missed was defined as the absence of mass or non-mass enhancement on MR A. MRI characteristics (T2 signal, size/volume, margins, distribution, enhancement kinetics) were documented. BRCA genes, molecular subtype and receptor subtype were also compared in missed or non missed. The study found 74% in 104 patients had breast cancer visible on MR A (the most recent MRI prior to their diagnosis). Their results did not find difference in MRI characteristics or receptor subtype between missed and non missed groups. They did find statistically significant association between BRCA 1 status (positive or negative) with detectability of breast cancer on MRI (odds ratio = 2.6).

These are major weaknesses of this study:

(R2,1)

The manuscript was poorly written.

Response: Thank you for your comment. We deeply reviewed the manuscript.

 

(R2,2)

Introduction did not show motivation or gaps in published works.

Response: We detailed the introduction following your constructive comments.

 

(R2,3)

Some content in Methods should go to Results.

Response: This was done accordingly to your comments.

 

(R2,4)

Table was numbered in a confusing way.

Response: We fixed it.

 

(R2,5)

It was not clear whether one radiologist did all the review

Response: The cases were reviewed by a fellow radiologist and the supervisor radiologist

 

(R2,6)

Sample size was not appropriate for statistical analysis. In some categories, there were only 1 or 2 patients.

 

For risk factors chest radiation and Cowden, the small numbers of observations (0, 1, or 2) in corresponding categories reflect the sparsity of the contingency tables between those risk factors and the missed/non-missed condition. The total sample sizes were not small, which can be seen from the contingency tables attached below, but the tables were quite sparse:

 

 

    We performed Fisher's Exact Test to determine the association between each of those risk factors and the missed/non-missed condition, which already accounts for the sparsity problem.

We however did not take into account the very small groups, such as the chest radiation group as statistically significant factor for missed cancers despite the below-threshold p value, given the number of individuals in this group was too small.

 

 

 

 

(R2,7)

In Discussion, the authors claimed that this study comprises the highest number of false-negative breast MRI. It is because this is a case control study. The cohort was selected to have benign or normal MRI prior to the diagnosis. And since it’s not a blinded study, the radiologist was biased when he or she was reviewing the MR images.

Response: this was added.

Author Response File: Author Response.docx

Reviewer 3 Report

Here, the authors provide an assessment of MRI characteristics and molecular subtypes in breast cancer detectability by MRI in high-risk patients. The authors utilize a small cohort of breast MRIs to assess the MRI characteristics in previously missed breast cancer MRIs. The authors conclude that neither MRI characteristics or molecular subtype influenced detectability in this study. The authors could discuss future studies in this field and how those studies may lead to the identification of MRI characteristics that may influence biopsy decision to improve breast cancer detection. While the analysis is performed well, the authors should expand some parts described below, and may consider editing their figures and tables to provide a more concise article.

Main comments

Authors should expand the introduction with the relevant literature. Here it would be beneficial to define patients at high-risk for breast cancer in this study and interval cancers, as well as including a description of the risk factors assessed in this study and why they may play a role in missing breast cancers.

Authors should describe how the MRI characteristics are measured in the methods (could provide a reference).

Authors may benefit from restructuring their figures (and ensuring they are included in the correct order and that they are referred to in the text). I suggest including a single summary figure, that could include some of the MRI images used, to show missed breast cancers at MR A and MR B for both an overlooked and misinterpreted breast cancer, and provide summary bullet points for each of these as to why they are missed.

Authors may also benefit from adding a simple flow diagram to their methods, that indicates how the cohort was selected to better explain the exclusion/inclusion criteria.

Authors should be clear on the terminology between a missed cancer and an interval cancer.

Minor comments

Authors may wish to compile some of the tables into one, such as patient cohort description with breast cancer risk factors, and pathology (hormone status and other features). Many of the other tables, particularly those with odds ratios feel unnecessary.

Authors should consider including some suggestions for future work, and how these studies would implicate clinical decision making.

Authors should double check all affiliations are provided and associated with each author’s name, as well as a corresponding author, and that names are Capitalized.

Authors should ensure they fill in the missing information for the IRB and informed consent statement since this study involves patient images, as well as author contributions and any funding.

Author should ensure proper referencing of their article.

Author Response

Here, the authors provide an assessment of MRI characteristics and molecular subtypes in breast cancer detectability by MRI in high-risk patients. The authors utilize a small cohort of breast MRIs to assess the MRI characteristics in previously missed breast cancer MRIs. The authors conclude that neither MRI characteristics or molecular subtype influenced detectability in this study. The authors could discuss future studies in this field and how those studies may lead to the identification of MRI characteristics that may influence biopsy decision to improve breast cancer detection. While the analysis is performed well, the authors should expand some parts described below, and may consider editing their figures and tables to provide a more concise article.

Main comments

(R3,1)

Authors should expand the introduction with the relevant literature. Here it would be beneficial to define patients at high-risk for breast cancer in this study and interval cancers, as well as including a description of the risk factors assessed in this study and why they may play a role in missing breast cancers.

Response: we detailed the introduction following your constructive comments.

 

(R3,2)

Authors should describe how the MRI characteristics are measured in the methods (could provide a reference).

Response: This was added. Thank you for your comment.

 

(R3,3)

Authors may benefit from restructuring their figures (and ensuring they are included in the correct order and that they are referred to in the text).

Response: This was done accordingly to your comments.

 

(R3,4)

I suggest including a single summary figure, that could include some of the MRI images used, to show missed breast cancers at MR A and MR B for both an overlooked and misinterpreted breast cancer and provide summary bullet points for each of these as to why they are missed.

Response: Thank you for your suggestion. We would like to keep the images as such following other reviewers’s comments.

(R3,4)

Authors may also benefit from adding a simple flow diagram to their methods, that indicates how the cohort was selected to better explain the exclusion/inclusion criteria.

• Response: As requested, Figure 1 has been revised using the standard PRSIMA flowchart diagram and style.
• Exclusion criteria were detailed.

(R3,5)

Authors should be clear on the terminology between a missed cancer and an interval cancer.

Response: This was added. Thank you for your comment.

 

 

Minor comments

(R3,6)

Authors may wish to compile some of the tables into one, such as patient cohort description with breast cancer risk factors, and pathology (hormone status and other features). Many of the other tables, particularly those with odds ratios feel unnecessary.

Response: this was deleted.

(R3,7)

Authors should consider including some suggestions for future work, and how these studies would implicate clinical decision making.

The time between two MRIs was between 6 and 26 months (mean 11.1 months).   Patients' ages ranged between 32 to 70 years old (mean age of 50.7 years old), at high risk of breast cancer (lifetime risk equal to or greater than 25%), including women carrying the BRCA gene mutation and other high-risk syndromes.

 

Response: this has been answered in the discussion and conclusion

(R3,8)

Authors should double check all affiliations are provided and associated with each author’s name, as well as a corresponding author, and that names are Capitalized.

Response: this was corrected

(R3,9)

Authors should ensure they fill in the missing information for the IRB and informed consent statement since this study involves patient images, as well as author contributions and any funding.

This is a retrospective review of imaging and the Sunnybrook REB waived the need for patient consent.

This work was supported by the Canadian Institute of Health Research (CIHR) [PJT-169005]

 

(R3,10)

Author should ensure proper referencing of their article.

References were listed according to the sage-vancouver guidelines.

 

 

Author Response File: Author Response.docx

Round 2

Reviewer 2 Report

I appreciate the authors' effort on revising the manuscript based on my review. I can see big improvement compared to original version. I recommend it to be published in Tomography.