
In breast cancer, preoperative evaluation of the tumor size is important for clinical stage determination and treatment planning. As the number of young patients with breast cancer increases, the importance of oncoplastic surgery, which obtains negative resection margins and does not excessively remove surrounding normal tissues, is also increasing.(1) Ultrasonography (USG) is mainly used for preoperative evaluation; however, the use of magnetic resonance imaging (MRI) has recently increased. USG is noninvasive, simple to perform, and can evaluate the size of a tumor relatively accurately.(2,3) However, this depends on the radiologist, and the size of the probe and posterior acoustic shadowing may limit the measurement of tumor size.(4,5) MRI is a sensitive imaging modality that shows multiplanar imaging and is useful for diagnosing multifocal or occult lesions.(6-8) However, it overestimates tumor size.(9,10) Although several studies have compared the accuracy of preoperative imaging modalities, each study has different results.(9-13) In this study, tumor size measured using USG and MRI was compared with the pathologic tumor size. We further analyzed the accuracy of the imaging modalities according to the menopausal status, MRI findings, pathological subtypes, and intrinsic subtypes.
Between March 2016 and February 2019, breast cancer patients with a single mass lesion on USG who underwent surgery at the National Cancer Center were reviewed. Patients with non-mass lesions, multiple lesions, microcalcifications except for intratumoral, positive surgical margins, or had undergone neoadjuvant chemotherapy (NAC) were excluded. Ultimately, 996 female patients were analyzed retrospectively.
The original official interpretations were made by four breast radiologists with 8 to 15 years of experience in breast imaging on a dedicated workstation. The largest tumor diameter was measured for each test. Fig. 1 and 2 display USG, MRI, and pathological images utilized in this study. These figures provide examples of patients with high and poor concordance, respectively.
1) USGUSG (iU22, Philips Medical Systems; Aixplorer, Super-sonic Imagine; Aplio i800, Canon Medical Systems) was handheld using a 12 MHz linear transducer. Both breasts were scanned with the patient in the supine position. Both transverse and longitudinal images were obtained from the suspicious lesions. For patients undergoing breast conserving surgery, skin marking was performed on the tumor location with USG before surgery.
2) MRIThe following three MRI scanners were used: Signa HDxt 3.0 T (GE Healthcare, Milwaukee, WI, USA), Achieva 3.0 T TX (Philips N.V., Eindhoven, The Netherlands), and Ingenia 3.0 T (Philips N.V.) with dedicated breast surface coils, and the patient in a prone position. The MRI protocol consisted of the following: an axial fat-suppressed T2-weighted sequence and a dynamic axial fat-suppressed T1-weighted sequence before and 90, 180, 270, and 360 s after the intravenous injection of gadoteric acid (0.2 ml/kg body weight, Dotarem; Guerbet, Aulnay-sous-Bois, Fran-ce). The T1-weighted images were sectioned at a thickness of 2 mm. Subtraction and maximum intensity projection images were generated using a dynamic series. A single mass lesion on USG was classified as a mass or non-mass enhancement (NME) feature on MRI and an official interpretation was used. NME was defined as an enhancing lesion exhibiting distinct features of a mass or background parenchymal enhancement around the index tumor, situated less than 1 cm apart. MRI were interpreted by four breast imaging radiologists. As the data were collected retrospectively, there were no instances of consensus or discordant reports among the radiologists.
The largest tumor diameter was obtained according to an official interpretation by a breast pathologist. In the analysis according to pathologic subtype, only patients with ductal carcinoma
The unit of tumor size was cm. The Mann-Whitney U test or Student's t-test was used for continuous variables after verifying the normality of variables using the Kolmogorov–Smirnov test, and the means with standard deviations were described, as appropriate. Chi-square test or Fisher's exact test was used for categorical variables. The intraclass correlation coefficient (ICC) was used to confirm the concordance of tumor size between imaging and pathology. The scale of Landis and Koch was used according to the ICC values. Using the scale, the interpretation of reliability was as follows: 0.00-0.20; slight, 0.21-0.40; fair, 0.41-0.60; moderate, 0.61-0.80; substantial, 0.81-1.00; almost per-fect.(14) All analyses were performed using STATA version 18 (StataCorp LP, College Station, TX, USA) and R version 4.2.1 (R Foundation for Statistical Computing, Vienna, Austria).
A total of 996 patients were analyzed and their clinicopathological characteristics are presented in Table 1. The average age of all the patients was 53.53 years old, and there were 449 (45.68%) premenopausal and 534 (54.32%) postmenopausal women. According to the MRI features and NME, there were 915 (91.87%) and 81 (8.13%) masses, respectively. For the distribution of pathological subtypes, 70 (7.03%) had DCIS, 808 (81.12) had IDC, 2 (0.20%) had LCIS, 35 (3.51%) had ILC, and the others included mucinous carcinoma and papillary cancer. Patients with IDC and ILC were classified according to the intrinsic subtype: 597 (72.36%) were luminal A, 72 (8.73%) were luminal B, 47 (5.09%) were HER-2 positive, and 114 (13.82%) were triple-negative breast cancer (TNBC). In the clinical stage, T0 was 77 (7.73%), 1 was 606 (60.84%), 2 was 312 (31.33%), and 3 was 1 (0.10%); however, the distribution changed to 77 (7.73%), 684 (68.67%), 232 (23.29%), and 3 (0.30%), respectively in the pathologic stage.
The average tumor size was 1.70 cm on USG, 2.02 cm on MRI, and 1.79 cm upon pathological examination (Table 2). The ICC was obtained to compare the concordance of tumor size between pathology and imaging. The ICC was 0.6019 (0.5600-0.6406) for USG and 0.7208 (0.6558-0.7711) for MRI compared to pathology.
All patients were divided according to menopausal status, MRI features, and pathological subtypes, and the concordance was compared (Table 3). According to the menopausal status, the concordance rate was higher for both USG and MRI in postmenopausal women (ICC, 0.6671; 95% CI, 0.617-0.7118 for USG, and ICC 0.7298; 95% CI, 0.6381-0.794 for MRI) than in premenopausal women (ICC, 0.513; 95% CI, 0.4413-0.5782 for USG, and ICC 0.7041; 95% CI, 0.6406-0.7562 for MRI). In the analysis according to the MRI features, it was confirmed that there was a higher concordance in masses (ICC, 0.7471; 95% CI, 0.7171-0.7744 for USG, and ICC 0.7313; 95% CI, 0.665-0.7818 for MRI) than in NME (ICC, 0.1409; 95% CI, -0.0456-0.3277 for USG, and ICC 0.6035; 95% CI, 0.3914-0.744 for MRI). There was a large difference between the ICCs of USG and MRI for the premenopausal, NME, and lobular types.
Analysis was according to the intrinsic type of invasive cancer, as shown in Table 4. In luminal A and HER-2 type breast cancer, MRI showed higher ICC than USG. In TNBC, USG showed a higher ICC; however, the difference between USG and MRI was small.
The imaging methods for diagnosing breast cancer include mammography, USG, and MRI. USG is the primary imaging test used to evaluate breast cancer before surgery, and when used as a supplement to mammography, it can increase the sensitivity and specificity of breast cancer diagnosis, especially in young women.(15) MRI is a more sensitive examination and helps in staging and surgical planning, as it can identify the extent of invasion and multiple lesions in the ipsilateral or contralateral breast.(16) This retrospective study used medical records to compare tumor sizes on preoperative USG, MRI, and postoperative pathology in 996 patients with breast cancer. The average tumor size in all patients was 1.70 cm on USG, 2.02 cm on MRI, and 1.79 cm on pathology, and was larger on MRI than on USG. Previous studies have also confirmed that MRI overestimates tumor size.(9,10) However, MRI showed a higher concordance with pathology in the ICC value in this study.
We also analyzed concordance according to menopausal status, MRI features, and pathological subtypes. There was a large difference between the ICC of USG and MRI for the premenopausal, NME, and lobular types. Therefore, MRI may be more useful for patients with premenopausal, NME, or lobular type. Upon close examination of the results of the postmenopausal women, both USG and MRI concordance were higher than those of premenopausal women. This may be due to decreased fibroglandular tissue (FGT) and background parenchymal enhancement (BPE) in postmenopausal women.(17) In young women, the sensitivity and specificity of MRI may be limited owing to the large number of dense breasts and BPE.(17,18)
Additionally, high concordance was observed in the case of the mass. In NME, an area of enhancement without definite features of a mass is defined by the Breast Imaging Reporting and Data System lexicon of the American College of Radiology, slight concordance with USG, and fair concordance with MRI.(19) The reason is that this study was designed primarily on patients with a single tumor on USG. Another reason is that NME is difficult to evaluate using MRI alone. The evaluation of NME on MRI is complicated and difficult to measure the size. Therefore, patients with NME may exhibit positive resection margins.(20,21) NME is associated with DCIS and has low accuracy in MRI owing to NME and microcalcifications.(13,20,22) However, it was difficult to find a correlation because the number of patients with DCIS in this study was only 70 (7.03%).
Lobular cancer is difficult to detect by imaging because of the clinical characteristics of diffuse spreading.(23) MRI has a higher accuracy than USG or mammography for the lobular type.(10,24,25) In the lobular type, concordance was higher with MRI than with USG. This is consistent with a previous study showing that MRI is a useful method for diagnosing lobular cancer. However, the small number of patients with lobular cancer included in this study may pose a constraint in interpreting the results.
Comparisons of image accuracy according to tumor biology have mostly been conducted in patients who have undergone neoadjuvant chemotherapy.(26-28) Moreover, the response to NAC is an important factor in determining the type and range of breast surgery. Among imaging modalities, MRI is a sensitive method for confirming the response after NAC, it is more accurate in TNBC and HER2-positive breast cancer than in hormone receptor-positive breast cancer.(27,28) Contrary to previous studies, we compared concordance according to tumor biology in patients who did not receive NAC. MRI showed higher ICCs than USG for luminal A and HER-2 type breast cancer. USG showed a higher ICC for TNBC; however the difference with MRI was small. Therefore, MRI may be more useful in patients with luminal A or HER-2 positive types. MRI may be omitted in patients with TNBC for whom surgery is the first-line treatment.
This study had several limitations. First, because our study targeted patients with single mass lesions, there are limitations in its application to patients with non-mass lesions, or multifocal or multicentric diseases. Second, there were limitations to breast density evaluation. There may have been inaccuracies because there were no data on the FGT and BPE grades.
However, the strengths of our study are its large sample size compared to previous studies and comparison according to various clinicopathological characteristics, including tumor biology. This study provided information on patients with a single mass lesion who did not undergo NAC.
MRI may be more helpful in predicting tumor size for patients that are premenopausal, with NME on MRI, and lobular cancer. However, in luminal A and HER-2, USG is considered more helpful. Concordance differs depending on the clinicopathological characteristics and should be considered in the surgical plan.
An abstract of this paper was presented in a poster session at the Global Breast Cancer Conference 2023.
There are no conflicts of interest.
This study was approved by the Institutional Review Board of the NCC, Korea (IRB No. NCC2022-0318) and was performed in accordance with the principles of the Declaration of Helsinki. The need for informed consent was waived owing to the retrospective nature of the study.
Clinicopathologic Features of Patients
Clinicopathologic features | n (%) |
---|---|
Age (years) | 53.53 ± 10.83 |
Menopausal state (missing = 13) | |
Premenopause | 449 (45.68) |
Postmenopause | 534 (54.32) |
MRI feature | |
Mass | 915 (91.87) |
NME | 81 (8.13) |
Pathology | |
DCIS | 70 (7.03) |
IDC | 808 (81.12) |
LCIS | 2 (0.20) |
ILC | 35 (3.51) |
Others | 81 (8.13) |
Clinical T stage | |
0 | 77 (7.73) |
1 | 606 (60.84) |
2 | 312 (31.33) |
3 | 1 (0.10) |
Pathological T stage | |
0 | 77 (7.73) |
1 | 684 (68.67) |
2 | 232 (23.29) |
3 | 3 (0.30) |
Intrinsic subtype (missing = 153) | |
Luminal A | 597 (72.36) |
Luminal B | 72 (8.73) |
HER2 | 47 (5.09) |
TNBC | 114 (13.82) |
MRI = magnetic resonance imaging; NME = non-mass enhance-ment; DCIS = ductal carcinoma in situ; IDC = invasive ductal carcinoma; LCIS = lobular carcinoma in situ; ILC = invasive lobular carcinoma; HER-2 = human epidermal growth factor receptor 2; TNBC = triple negative breast cancer.
Tumor Size and Intraclass Correlation Coefficient Values between Imaging and Pathology
Measured method | Tumor size (cm), mean ± SD | ICC (95% CI) |
---|---|---|
USG | 1.70 ± 0.80 | 0.6019 (0.5600-0.6406) |
MRI | 2.02 ± 1.12 | 0.6019 (0.5600-0.6406) |
Pathology | 1.79 ± 0.93 | reference |
USG = ultrasonography; MRI = magnetic resonance imaging; ICC = intra-class correlation coefficient.
Intraclass Correlation Coefficient Values in Subgroup
Menopausal state | MRI features | Pathologic subtype | ||||||
---|---|---|---|---|---|---|---|---|
Premenopause (n = 449) |
Postmenopause (n = 534) |
Mass (n = 915) | NME (n = 81) | Ductal (n = 878) | Lobular (n = 37) | |||
USG MRI |
0.513 (0.4413-0.5782) 0.7041 (0.6406-0.7562) |
0.6671 (0.617-0.7118) 0.7298 (0.6381-0.794) |
0.7471 (0.7171-0.7744) 0.7313 (0.665-0.7818) |
0.1409 (-0.0456-0.3277) 0.6035 (0.3914-0.744) |
0.592 (0.547-0.6336) 0.7419 (0.6718-0.7939) |
0.4535 (0.1655-0.6727) 0.7738 (0.6025-0.8768) |
USG = ultrasonography; MRI = magnetic resonance imaging; NME = non-mass enhancement; ICC = intraclass correlation coefficient.
Tumor Size and Intraclass Correlation Coefficient Value in Intrinsic Subtype
Measured method | Luminal A (n = 597) | Luminal B (n = 72) | HER2 (n = 47) | TNBC (n = 114) | |||||||
---|---|---|---|---|---|---|---|---|---|---|---|
Tumor size (cm) | ICC (95% CI) | Tumor size (cm) | ICC (95% CI) | Tumor size (cm) | ICC (95% CI) | Tumor size (cm) | ICC (95% CI) | ||||
USG MRI Pathology |
1.63 ± 0.74 1.92 ± 1.07 1.74 ± 0.92 |
0.5519 (0.4928-0.6058) 0.7447 (0.6903-0.7887) reference |
1.89 ± 0.74 2.12 ± 0.89 1.98 ± 0.72 |
0.6096 (0.4426-0.7361) 0.7207 (0.5873-0.8159) reference |
2.15 ± 0.95 2.82 ± 1.60 2.38 ± 1.32 |
0.3262 (0.0497-0.5574) 0.7591 (0.5616-0.8673) reference |
2.03 ± 0.83 2.31 ± 0.89 1.947 ± 0.781 |
0.7077 (0.603-0.7885) 0.6270 (0.3973-0.7636) reference |
USG = ultrasonography; MRI = magnetic resonance imaging; HER-2 = human epidermal growth factor receptor 2; TNBC = triple negative breast cancer; ICC = intraclass correlation coefficient.