Supine versus Prone 3D Abus Accuracy in Breast Tumor Size Evaluation

Round 1

Reviewer 1 Report

In this manuscript, D’Angelo and colleagues compare the efficacy of various methods of determining the tumor size (specifically, the largest diameter of the tumor) in breast cancer patients in two different centers of breast cancer care. The main measurement that the authors focus on is the tumor size as determined by 3D automated breast ultrasound (ABUS). One of the centers of this two-center study, ABUS was carried out with the patients in supine position. In the second center, ABUS was carried out on a different set of patients in the prone position. The authors compare the efficacy of the two ABUS methods using the tumor size with the histopathologically determined tumor size as the standard reference.

Using interclass correlation coefficient (ICC), the authors find that ABUS outperforms many other measures, and prone ABUS correlates slightly better with the standard (and also with a few other measures) than the supine ABUS method.

The major flaw, which in this reviewer’s view is fatal in and of itself, is that ICC is not a reliable measure for evaluating agreement between various metrics. In general, samples with large enough variance can produce spuriously high correlations. This after all, was the rationale for Bland and Altman to develop their Bland-Altman difference plots.

Moreover, the authors make little effort to measure the statistical significance of the observed differences in ICCs. This is unfortunate, because such an exercise would have made the aforementioned flaw to the authors.

The authors do show Bland-Altman plots in Figure 3, but the two plots they show are largely beside the point. Besides, one of the plots in Figure 3 (ABVS vs HISTOLOGY) essentially confirms the result previously by a much more rigorous study by Girometti et al, 2018 (which the authors duly cite).

Another major concern is that there are several unresolved confounds in the study, all of which have to do with the fact that the two sets of ABUS measurements were made at two different facilities by two different sets of providers on two different cohorts of patients. The data provided in Table 1 (especially the results of the statistical tests provided therein) alleviate this concern a little bit, but not fully. In other words, the concern about the potential confounds remains unresolved.

 

MINOR TYPOGRAPHICAL ISSUES:

            The manuscript is best with them. But in fairness to the authors, these are very minor issues that can be resolved with the help of a good copy editor. I stopped taking notes on the typos after the first two pages or so, but the typos I did note are listed below.

(1) Abstract (line 14): “Breast cancer size is one of the main determining of TNM staging.”

            Define TNM.

(2) Abstract (throughout): Inconsistent use of abbreviations. Sometimes the abbreviations are defined prior to first use, sometimes not.

(3) Abstract, line 16: In this prospective two center study, […]

            Should be hyphenated  as “two-center”. The authors spell correctly in Materials and Methods (line 72).

(4) Introduction (lines 38-40). “It was approved by the Food and Drug Administration (FDA) in 2012 as a supplemental screening tool for women with heterogeneously 39 and extremely dense breasts [4].”

            Please spell out what “It” stand for.

__________________________________________

 

Author Response

Please see the attachment

Author Response File: Author Response.docx

Reviewer 2 Report

This article presents research in the field of breast tumor size assessment.

 

The article is an interesting approach in the aspect of examining the accuracy of SUPINE 3D ABUS test compared to PRONE 3D ABUS in assessing tumor size in breast cancer patients. The study was performed on patients with percutaneous biopsy confirmed early stage breast cancer

 

Minor remarks:

1) I would suggest describing in more detail how the selection of patients' medical conditions flows out in terms of their impact on the outcome of the study.

2) The authors could refer to other methods in terms of what is the influence of the method and what type of data studied on their effectiveness using particular types of tests.

3) The article needs minor language corrections.

 

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 3 Report

This paper presents the results of the accuracy evaluation of breast tumor size by comparing Supine 3D ABUS and Prone 3D ABUS. This is what is said in the abstract, but along the text and in the conclusions seems that the focus was to evaluate Supine and Prone ABUS against MRI.

 

If the evaluation is focused on the positions, it is not clear the biggest benefit of one or another for the patient. Is it really necessary? 

Besides this, other concerns are related to the text that needs an extensive review.

 

Below I provide some more detailed comments:

 

• Starting with the abstract, there are several acronyms used that are shown in the middle of the text far from there. 
• There are several inconsistencies. For example, why "PRONE" is in capital letters? Moreover, the authors change it along with the text to "Prone" and keep going back and forth. 
• Also, some sentences seem to be out of context in the middle of the text for example: "It was approved by the Food and Drug Administration (FDA) in 2012 as a supplemental..." What was approved? Another one: "Although the evidence for long-term benefits..." Evidence of what?
• In lines 48-50 is stated that "Breast-Conserving Surgery (BCS) is the standard treatment for early breast cancers". Is that right? How about chemotherapy? It seems that the authors have some point here but is not clear.
• There is some lack of organization along with the paper, for example, in the Materials and Methods section. We will know the number of patients only in the results section.
• The conclusions don't make sense compared to the abstract. What was the evaluation proposed ABUS vs. MRI or Prone ABUS vs. Supine ABUS?

  

Thank you.

Regards.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

First of all, I made many typos of my own in my earlier review. Ironically, I made typos even when complaining about the authors’ typos (e.g., I said the manuscript was “best” with typos when I meant to say “beset”). My sincere apologies. I am heartened to see that the overall import of the comments nonetheless did adequately get through to the authors.

Having said that, I must note that the revision does not fully address the two major concerns I had expressed about the original manuscript. In fairness to the authors, the revised version does (seem to) try to address these concerns, but it clearly falls short.

My first major concern was about the authors’ use of ICCs. In the revised version, the authors have done away with the ICCs and replaced them with the Bland-Altman difference plot analyses. This is a step in the right direction. But the truly disquieting part is that no data points are shown in four of the difference plots (Figures 4 through 7), even as the authors refer to individual ‘dots’ in the relevant part of the text, and include a nominal dot in the key for each figure. I duly note that the authors do show the mean difference in each of these figures, but that is not the same as showing the individual data points.

My second major concern was that the authors did not adequately take note of the possible confounds in their study. The authors have done nothing to address this concern. In the relevant section of the Discussion, they do allude to the fact that “… the sample study differs from the two hospitals, which could be biased.” But this does not even adequately describe, much less address, the issue of confounds. (As the authors no doubt know, from a statistical viewpoint, ‘bias’ is not necessarily the same thing as ‘confound’. Bias is just one of the many problems that confounds can cause.)

A reasonable reader reading the aforementioned sentence of the authors would not even realize that they are referring to potential confounds; I certainly did not. The authors’ sentence makes it sounds like they are referring to some minor sampling issue. But the problem is potentially more serious than that: As I noted in my original review, the confounds in the present case arise from the fact that the authors make one set of measurements in one hospital and another set of measurements in another hospital. And then they make many comparisons between the two sets of measurements, and find some differences. Are the differences attributable to the actual efficacies of the measurement techniques per se, or to any number of other differences between the two hospitals, or both? There is no way to tell. The authors owe it to the reader to clearly and forthrightly explain this issue.

Author Response

Please see the attachment

Author Response File: Author Response.docx

Reviewer 3 Report

The authors have addressed most of my concerns.

I suggest again going over the text and fixing some possible typos and small errors. For example, in Table 1, the p-value column shows the comma "," instead of the dot ".".

Author Response

Please see the attachment

Author Response File: Author Response.docx