Validation of a protocol for pre-operative flap thickness assessment on dual-view digital mammography for the decision-making process in implant-based breast reconstruction

Introduction

Delayed healing following mastectomy and prosthetic reconstruction can often be related to poor quality mastectomy skin flaps. This may be due to a variety of factors that include surgical technique as well as variations in the thickness of the normal subcutaneous fat amongst patients. Pre-operative breast imaging is a critical component of the decision-making process both for the oncological and the reconstructive surgeons. Oncological surgeons plan their approach to optimally excise the breast parenchyma, whereas reconstructive surgeons assess the remaining thickness of the skin and subcutaneous fat layers of the breast to opt for the reconstructive technique that will lead to fewer complications and a better aesthetic outcome. Awareness of both components can play a relevant role in the assessment of specific breast characteristics that are extremely important for the decision-making process in implant-based breast reconstruction (IBBR) following conservative mastectomies (CMs).

The thickness of the non-glandular breast tissue coverage (BTC), including the skin and the subcutaneous fat layer overlying the anterior lamella of the capsula mammae, varies according to the single patient’s breast anatomy. It is this layer that correlates with an anatomical and properly performed mastectomy to optimize mastectomy flap thickness (1). It is known that the subcutaneous fat layer of the breast is not homogeneously represented in all women and that it can vary within the same breast, being a unique characteristic of the patient (2). These characteristics highlight the need for an individualized and reproducible method for the pre-operative assessment of the mastectomy flap thickness to reduce post-surgical complications and improve aesthetic outcomes for IBBR, warranting the oncological safety of the CM (3). The thickness of the mastectomy flap reflects its perfusion and vitality. Thicker mastectomy skin flaps are more likely to have less injured underlying arterial subdermal plexus, which is ultimately responsible for cutaneous perfusion. Therefore, the mastectomy flap thickness should be included when considering the hazard factors associated with postoperative ischemic complications.

The primary purpose of this study is to validate a BTC assessment modality that uses digital mammography (DM) imaging in both the cranio-caudal (CC) and the medio-lateral oblique (MLO) views. The robustness of the proposed pre-operative assessment will be determined by deriving the metrics associated with the inter- and intra-observer agreement (p-values of the matched measurements, measurement error, repeatability, limits of agreement, correlation) for each testing site and for their possible aggregations. It is our hope that this methodological study will assist with preoperative surgical planning by providing a clinical validation tool to estimate mastectomy flap thickness and its impact on pre-operative decision-making for optimizing surgical-reconstructive outcomes, both from the oncological and the aesthetic point of view. We present this article in accordance with the STROBE reporting checklist (available at https://gs.amegroups.com/article/view/10.21037/gs-2026-1-0139/rc).

Methods

Study design and dataset

We designed a retrospective, multi-reader study on a dataset of 40 unilateral DM cases obtained in the two standard views (CC and MLO), using the same commercially available manufacture (Selenia Dimensions Unit, Hologic, Bedford, MA, USA). The study set was extracted from all consecutive healthy cases recorded from January 2022 to March 2022 in the Picture Archiving and Communication Systems (PACS) of two Breast Imaging Academic Centers involved in the project (the Breast Unit of the University of Naples Federico II and the Breast Unit of the University of Turin).

In each Center, two senior radiologists selected 20 cases, uniformly distributed according to the American College of Radiology (ACR) Breast Imaging Reporting and Data System (BI-RADS) (4) breast density classification with 4 different categories: A (almost entirely fatty), B (scattered fibroglandular), C (heterogeneously dense), and D (extremely dense). The selected cases were anonymized and grouped into 20 ACR A and B (hereon referred to as non-dense) breasts and 20 ACR C and D (hereon referred to as dense) breasts. This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was based on retrospective analysis of fully anonymized pre-existing imaging data collected from institutional databases of the participating centers, so no approval was needed. All patients had previously provided informed consent for the use of their imaging data for research purposes.

BTC measurement procedure

Our pre-operative assessment of the mastectomy flap was based on 12 measurements for each breast, according to a standardized protocol, as follows:

• Each breast was divided into 4 equal sections by manually drawing three parallel lines perpendicular to the antero-posterior distance of the breast extending from the nipple to the pectoralis muscle. These lines correspond to anterior, middle, and posterior depth planes, positioned at one-quarter, one-half, and three-quarters of the total length. This partitioning method is applied to both CC and MLO views.
• The flap thickness was measured by drawing calipers from the interface between each line and the glandular profile, perpendicularly reaching the skin; 6 measurements (3 per quadrant) are performed for each view, as reported on the representative DM images in Figure 1.

Figure 1 Protocol for the pre-operative measurement of the mastectomy flap, based on 12 measurements (3 per quadrant) for each breast, applied consistently to both cranio-caudal (A) and medio-lateral oblique (B) mammographic views: the flap thickness is measured by drawing calipers from the interface between each line and the glandular profile, perpendicularly reaching the skin. AM, antero-medial; AL, antero-lateral; MM, middle-medial; ML, middle-lateral; PM, postero-medial; PL, postero-lateral; AS, antero-superior; AI, antero-inferior; MS, middle-superior; MI, middle-inferior; PS, postero-superior; PI, postero-inferior.

The 12 flap thickness measurements are labeled on CC view as antero-medial (AM), antero-lateral (AL), middle-medial (MM), middle-lateral (ML), postero-medial (PM), postero-lateral (PL); on MLO view as antero-superior (AS), antero-inferior (AI), middle-superior (MS), middle-inferior (MI), postero-superior (PS), postero-inferior (PI).

Image evaluation: readers and reading process

The study involved 7 radiologists from three different academic sites. Four were senior radiologists with extensive experience in diagnostic and/or screening mammography accrued over 6 to 16 years, interpreting more than 2,000 mammographic exams annually. The remaining radiologists were junior faculty members who dedicated over 50% of their clinical time in the last year to interpreting breast imaging studies (more than 1,000 mammographic exams). Radiologists performed their readings individually on workstations. Before starting the reading process, a remote training session was organized to familiarize the readers with the study protocol and allow them to interact with each other. Radiologists were free to use various tools for adjustments, such as window/level settings and zooming, as well as an electronic caliper. All measurements were recorded in millimeters using a predefined Excel sheet form. To prevent potential biases in their assessments, the radiologists were unaware of their colleagues’ measurements. After at least 1 month, two senior radiologists performed a second measurement session on the same cases presented in a different randomized sequence from the first reading session to minimize visual memory biases.

Statistical analysis

The N>2 sets of measurements on the same subject at the same site were analyzed with the repeated measures analysis of variance (ANOVA), with main outcome P value, within-subject standard deviation s_w, and intraclass correlation (ICC). If s_w is not significantly correlated with the magnitude of measurements (tested by the non-parametric Kendall tau coefficient), its value gives the measurement error, i.e., for 95% of observations, the difference between a measurement and its true value is expected to be within± 1.96s_w (coefficient of accuracy) and the difference between the measurements by two observers on the same subject is expected to be <√2 ×1.96s_w [coefficient of repeatability (CoR)] (5,7).

Matched measurements on n subjects at the same site performed at different times by the same observer were compared with the non-parametric Wilcoxon’s test. In this case, s_w = Ö(Sd_i²/2n) = Ö2×s_d, where d is the difference between the two matched measurements and s_d its standard deviation.

Data from N sets of measurements were represented in Bland Altman plots for N=2 (region of agreement = 2×1.96×s_d), or in limits of agreement from the mean (LoAM) plots for N>2, where LoAM = ±1.96s_w (8,9). Categorical variables were expressed as numbers and percentages and compared with the chi-square test. For all tests, significant association corresponded to P<0.05. Analyses were performed using StatPlus for Macintosh Build 8.0.4.0/Core v7.11, 2022 (AnalystSoft, Walnut, CA, USA).

Results

BTC measurement protocol validation

The proposed BTC measurement procedure was validated using the measurements conducted by the four senior radiologists on 20 non-dense breasts and 20 dense breasts in the 12 pre-defined breast sites. Measurements of the flap thickness in the PI site were incomplete for 10% of the non-dense breasts and 25% of the dense ones.

Inter-observer agreement

Table 1 presents the results of the repeated measure ANOVA applied to the measurements of the four senior radiologists: P value, within-subject standard deviation s_w, and correlation coefficient ICC for each testing site. The P values, all >0.05, excluded statistically significant differences among the measurements by the four readers.

The lack of a significant correlation between the within-standard deviation s_w and the measurements’ magnitude was tested by plotting the individual subject’s standard deviation against their mean and computing Kendall’s tau coefficient (8). The values of Kendall’s tau for all sets of measurements are reported in Table 2, with the relative P value. The average value of the Kendall tau coefficient is 0.18 for the 12 sets of measurements for non-dense breasts and 0.22 for the dense ones, with all values of P≥0.05. We can then safely consider the within-standard deviation s_w representative of the size of true measurement error and use it to derive the associated metrics (9).

The level of correlation among the measurements of the four readers was good: ICC =0.74±0.11 for the non-dense breasts and 0.82±0.13 for the dense ones (P=0.21).

The LoAM and scatter plots of the thickness measurements in the AL position are shown in Figure 2A-2D: the data relative to non-dense breasts are spread over a larger region than the data for dense ones (measurement error s_w=2.6 mm for the former vs. 1.2 mm for the latter ones) and have a lower correlation with the mean (ICC=0.54 vs. 0.93). This difference in behavior related to the density is common to most sites: the non-dense breasts have a significantly larger measurement error than the dense ones (average over 12 sites 2.5±1.1 vs. 1.9±0.86 mm, P=0.03), and this reflects on associated metrics, including repeatability (7.5±2.9 vs. 5.6±3.2 mm).

Figure 2 LOAM plots of repeated thickness measurements in the AL position for non-dense breasts (A) and dense breasts (B). Scatter plots for non-dense breasts (C) and dense breasts (D). A-B, non-dense breasts; ACR, American College of Radiology; AL, antero-lateral; C-D, dense breasts; LL, lower limit; LOAM, limits of agreement from the mean; UL, upper limit.

Intra-observer agreement

The intra-observer agreement was assessed by comparing the matched data of the first and second delayed sequence of measurements for two senior radiologists, R1 and R2. Table 3 reports as an example, the outcome for R1: the P value from the Wilcoxon test excludes significant differences between the two sequences, a conclusion supported by the correlation coefficients ICC (all values ≥0.50). As already seen for the data on the inter-observer agreement, the measurement error is larger for the non-dense breasts than for the dense ones: 2.6±0.6 mm for the former vs. 1.6±0.7 mm for the latter (P=0.001). Similar results hold for radiologist R2. Figure 3 shows the Bland-Altman plots of the measurements performed by R2 in the PM site: the region of agreement is 12.8 mm for the non-dense breasts vs. 3.5 mm for the dense ones.

Figure 3 Bland-Altman plots of the matched measurements performed by R2 in the PM site for non-dense breasts (A) and dense breasts (B). Scatter plots for non-dense breasts (C) and dense breasts (D). A-B, non-dense breasts; ACR, American College of Radiology; CD, dense breasts; LL, lower limit; PM, postero-medial; UL, upper limit.

Performance of the protocol

BTC maps

The measurements of the 20 non-dense breasts by the 4 Senior Readers in the 12 testing positions were averaged to obtain the BTC map representative of that density category and the same was done for the dense breasts. Table 4 reports the average flap thickness for the 12 testing positions in the anterior, middle, and posterior depth planes. Even if the flap thickness is generally greater for non-dense breasts than dense breasts, in more than 50% of cases with significant differences, the trend as a function of the position is similar for both density categories, increasing from the anterior depth plane to the posterior one (Figure 4). Table 4 also reports the p-value of the comparison between measurements performed in adjacent positions along the same depth plane, with the aim to identify couples of sites without significant thickness differences. For the non-dense breasts, this condition was satisfied by the set AL_AM (P=0.22), and ML_MM (P=0.37).

Figure 4 Average flap thickness values (with error bars) obtained for the 11 testing positions (postero-inferior is not displayed because of missing data) in the anterior, middle, and posterior depth planes for non-dense breasts and dense breasts. A-B, non-dense breasts; ACR, American College of Radiology; C-D, dense breasts; LL, lower limit; MI, middle-inferior; PL, postero-lateral; UL, upper limit.

For the dense breasts, the number of adjacent positions without significant differences in thickness increased to 4: AL_AM (P=0.53), AS_AI (P=0.82), ML_MM (P=0.18), and PL_PM (P=0.66). The two results in the anterior positions hinted at the possibility of the whole anterior depth plane being a homogeneous area. To test this hypothesis, the Repeated Measures ANOVA was applied to the AL_AM_AS_AI set of measurements, yielding P=0.46 and within-subject standard deviation s_w=1.7 mm. The associated LoAM plot is shown in Figure 5. The existence of adjacent sites of uniform thickness suggests the possibility of reducing the number of measurements: −2 for non-dense breasts (−17%) and −5 for dense breasts (−42%).

Figure 5 LOAM plots of repeated thickness measurements in the AL_AM_AS_AI set (dense breasts). A-B, non-dense breasts; AI, antero-inferior; AL, antero-lateral; AM, antero-medial; AS, antero-superior; C-D, dense breasts; LOAM, limits of agreement from the mean.

Length of the measuring procedure by a senior reader

One of the senior radiologists (M.D.) was timed to estimate the length of the measurement procedure over the mammographs. The time necessary to measure the flap thickness in each position in the CC and in the MLO view was about 12 s, leading to a grand total of less than 3 min for each mammograph.

Performance of protocol with less experienced radiologists

The proposed measurement procedure was further tested by having 3 junior faculty members (J) perform the same measurements as the 4 senior radiologists (excluding PI). Table 5 compares their values against the benchmark established by the SR: average thickness, LOAM and CoR. The values measured by the 3 junior members in each site satisfy the precision requirement ôJ-SRô<LOAM, (for brevity, the difference ôJ-SRô is reported only for J1). Also, the absolute differences between the measurements performed in the same position by two junior readers fall within the repeatability limits (ôJi-Jyô < COR, with i and y =1, 2, 3).

Discussion

In this paper, we present the procedure used to validate a BTC assessment based on DM imaging: our choice to adopt DM for flap assessment considers the widespread use of this modality for screening and diagnostic work-up of breast cancer. Nowadays, “oncoplastic imaging” refers to a broader use of breast imaging than the recognized one of localizing and characterizing tumors. Breast imaging can greatly assist the surgeon in the preoperative assessment of flap thickness, supporting breast reconstruction planning to potentially reduce complications and better aesthetic outcomes (3,10).

The protocol under test consists of 12 measurements performed in CC and MLO views [adapting the method proposed by Frey et al. for breast MRI (11)]. The method validation was based on the measurements on 20 non-dense breasts and 20 dense breasts performed by 4 senior radiologists with many years of experience. The main findings of our study were:

• The inter-observer agreement was highly satisfactory: no significant differences and good correlation among the four senior radiologists’ measurements in all positions and on breasts of different density;
• The intra-observer agreement, assessed separately for the two senior radiologists who performed a delayed second set of measurements, was also good: no significant differences and good correlation for either one;
• The method performed well also when used by less expert radiologists, satisfying the requirements for accuracy and repeatability set by the senior team;
• The analysis of the measured thickness consistently indicated a larger measurement error for the non-dense breast data than for the dense ones; besides being scattered over a larger region of agreement, the former were also significantly less correlated than the latter.
• For both non-dense and dense breasts, it was possible to assess a few adjacent positions in the anterior and middle depth planes with uniform thickness values.

There are several subjective and objective methods for preoperative flap thickness assessment. Rancati et al. firstly proposed DM as a breast imaging tool to assess the thickness of the non-glandular BTC (10,14). Successively, both DM and MRI assessments were introduced by Radu and Frey (11,13,15). While promising, adding MRI to DM requires computational time that may be too demanding in most clinical settings. Most recently, Pagliara et al. proposed combining intraoperative ultrasound assessment of the mastectomy skin flap with the preoperative DM and using the ratio as a reliable predictor of ischemic complications and aesthetic outcomes (16).

As reported by several authors, there is variability in the subcutaneous breast tissue identifiable on mammography. Hence, it is extremely important to identify the appropriate and more accurate method for the preoperative evaluation of the skin flap. Rancati et al. used only one view (MLO) (10,14), while Radu et al. (13) applied measurements in both mammographic views. Our dual-view protocol, applied to 20 ACR non-dense and 20 dense breasts, evidenced a very good inter- and intra-observer agreement. The outcomes of the 4 senior radiologists showed good agreement on the thickness determined for a given position, with acceptable errors of measurement and coefficients of repeatability; also, the two delayed readings by the same radiologist were highly consistent.

The topographic analysis of the measurements performed by the 4 senior radiologists was used to explore the potential to reduce the number of testing sites without compromising the robustness of the protocol. In every position, the ACR non-dense breasts had consistently larger and more scattered thickness values than the dense breasts, but both categories had thickness values that increased from the anterior to the posterior positions.

We searched in each depth plane for regions containing 2 or more measuring sites with uniform thickness values. For non-dense breasts, we identified the couples AL, AM in the anterior plane and ML, MM in the middle plane; for dense breasts, all 4 anterior sites (AL, AM, AS, AI), and the couples ML, MM in the middle plane, and PL, PM in the posterior plane. These results suggest that it may be possible to spare measurements in 2 sites when dealing with the non-dense breasts and 5 when dealing with the dense ones. Although multiple-point assessment of thickness on superior, inferior, medial, and lateral mastectomy flap can provide more information on the overall thickness, excluding a few selected positions may help establish a quicker technique for preoperative evaluation of BTC without losing reliability. For our senior radiologist, with several years of specific experience in different breast visualizing techniques, the exclusion of 5 sites would imply cutting barely 1 min, i.e., lowering the time for analyzing a mammograph from 2.5–3 to 1.5–2 min. However, we hope to extend the measurement procedure also to radiologists with less expertise: we already tested the good level of their performance, and for them to reduce the number of measurements would be a real plus.

Based on the location of the most used CMs’ incisions (“italic-s” at the crossing between upper and lower outer quadrants and at the lateral half of the inframammary fold), of the overall physiological thinner thickness of the central quadrant (where the sub-nipple-areola complex dissection becomes more superficial) and of the inferior quadrants (where the implant gravitational effect is more impacting), the more useful measurement sites for evaluating the aforementioned zones, are AL, ML and PL for the first, AM, AL, AS, AI for the second and AI, MI, PI for the latter. These 8 sites include the quadruplet (AL, AM, AS, AI) identified as a homogenous region for dense breasts and the couple (AL, AM) for non-dense breasts. Some difficulties may be encountered when trying to measure the flap thickness in the PI position. Sometimes, it may be a DM technical problem when the inframammary fold is not correctly included in the MLO view, hindering the thickness measure (i.e., in small breasts and/or extremely dense ones). Moreover, especially in ACR almost entirely fatty breasts, it may be extremely difficult to distinguish the real distance between the breast skin and Cooper’s ligaments surrounding density.

This study has some limitations. First, it is known that DM images were acquired after breast compression that could deform the actual distance between the skin and the anterior lamella of the capsula mammae; thus, despite the good reproducibility of the measurements, it cannot be conclusively demonstrated that mammographic images are always sufficient for the correct evaluation of mastectomy flap thickness. Moreover, the measurements at the pre-operative DM might not always exactly correspond to the actual mastectomy flap thickness (16). This may be particularly true when the mastectomy is not properly performed following the correct anatomical plane, that is, the anterior layer of the breast capsule (1). Furthermore, the results reported in the paper were obtained from 40 cases or, more precisely, 20 non-dense plus 20 dense breast cases, so we cannot assert that the statistical tests had the power necessary to identify very small differences. The gold standard 80% test power would however require a prohibitive increase of the sample size by at least a factor of 10.

Conclusions

The proposed approach seems to be well-standardized and reproducible, allowing good performance at different levels of experience. To the best of our knowledge, no past studies have assessed the efficiency of the measurement methods before correlating the measured average flap thickness to clinical findings.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the STROBE reporting checklist. Available at https://gs.amegroups.com/article/view/10.21037/gs-2026-1-0139/rc

Data Sharing Statement: Available at https://gs.amegroups.com/article/view/10.21037/gs-2026-1-0139/dss

Peer Review File: Available at https://gs.amegroups.com/article/view/10.21037/gs-2026-1-0139/prf

Funding: None.

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://gs.amegroups.com/article/view/10.21037/gs-2026-1-0139/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was based on retrospective analysis of fully anonymized pre-existing imaging data collected from institutional databases of the participating centers, so no approval was needed. All patients had previously provided informed consent for the use of their imaging data for research purposes.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Duncan AM, Al Youha S, Joukhadar N, et al. Anatomy of the Breast Fascial System: A Systematic Review of the Literature. Plast Reconstr Surg 2022;149:28-40. [Crossref] [PubMed]
2. Robertson SA, Rusby JE, Cutress RI. Determinants of optimal mastectomy skin flap thickness. Br J Surg 2014;101:899-911. [Crossref] [PubMed]
3. Nava MB, Catanuto G, Rocco N. Breast reconstruction with form-stable implants In: Gabriel A, Nahabedian MY, Maxwell GP, et al. editors. Spear’s Surgery of the Breast: Principles and Art. Fourth Edition. Philadelphia: Wolters Kluwer; 2020:472-87.
4. Sickles EA, D’Orsi CJ, Bassett LW, et al. ACR BI-RADS Mammography. ACR BIRADS Atlas, Breast Imaging Reporting and Data System. Reston, Virginia: American College of Radiology; 2013.
5. Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986;1:307-10.
6. Bland JM, Altman DG. Measurement error. BMJ 1996;313:744. [Crossref] [PubMed]
7. Altman DG, Bland JM. Comparing several groups using analysis of variance. BMJ 1996;312:1472-3. [Crossref] [PubMed]
8. Jones M, Dobson A, O’Brian S. A graphical method for assessing agreement with the mean between multiple observers using continuous measures. Int J Epidemiol 2011;40:1308-13. [Crossref] [PubMed]
9. Christensen HS, Borgbjerg J, Børty L, et al. On Jones et al.’s method for extending Bland-Altman plots to limits of agreement with the mean for multiple observers. BMC Med Res Methodol 2020;20:304. [Crossref] [PubMed]
10. Rancati A, Angrigiani C, Hammond D, et al. Preoperative digital mammography imaging in conservative mastectomy and immediate reconstruction. Gland Surg 2016;5:9-14. [Crossref] [PubMed]
11. Frey JD, Salibian AA, Choi M, et al. Mastectomy Flap Thickness and Complications in Nipple-Sparing Mastectomy: Objective Evaluation using Magnetic Resonance Imaging. Plast Reconstr Surg Glob Open 2017;5:e1439. [Crossref] [PubMed]
12. Pagliara D, Schiavone L, Garganese G, et al. Predicting Mastectomy Skin Flap Necrosis: A Systematic Review of Preoperative and Intraoperative Assessment Techniques. Clin Breast Cancer 2023;23:249-54. [Crossref] [PubMed]
13. Radu M, Bordea C, Noditi A, et al. Assessment of Mastectomy Skin Flaps for Immediate Implant-Based Breast Reconstruction. J Med Life 2018;11:137-45.
14. Rancati AO, Angrigiani CH, Hammond DC, et al. Direct to Implant Reconstruction in Nipple Sparing Mastectomy: Patient Selection by Preoperative Digital Mammogram. Plast Reconstr Surg Glob Open 2017;5:e1369. [Crossref] [PubMed]
15. Frey JD, Salibian AA, Choi M, et al. Optimizing Outcomes in Nipple-sparing Mastectomy: Mastectomy Flap Thickness Is Not One Size Fits All. Plast Reconstr Surg Glob Open 2019;7:e2103. [Crossref] [PubMed]
16. Pagliara D, Montella RA, Garganese G, et al. Improving Decision-making in Prepectoral Direct-to-implant Reconstruction After Nipple Sparing Mastectomy: The Key Role of Flap Thickness Ratio. Clin Breast Cancer 2023;23:e37-44. [Crossref] [PubMed]