Magnetic Resonance Imaging in the Preoperative Evaluation of Invasive Lobular Carcinoma; A Useful Investigation?

Abstract

Aim: Quantify the added benefit of preoperative magnetic resonance imaging (MRI) staging of invasive lobular carcinoma of the breast (ILC) in addition to conventional imaging with ultrasound and mammography (US/MMG). Methods: Retrospective study ILC patients at our center to investigate effect of MRI in detecting additional disease, as well as effects on subsequent surgical management, reoperation rates and disease recurrence. Results: Preoperative MRI detected additional disease in 30.4% and altered surgical management in 39.1%. MRI better predicted histopathological lesion size but did not significantly alter reoperation or recurrence rates. Conclusion: MRI is superior in ILC detection and upstages a significant proportion of patients when used in preoperative staging. The impact on surgical outcomes warrants further investigation in larger prospective studies.

Keywords: :

• breast-conserving surgery
• breast cancer
• ILC
• invasive lobular carcinoma
• magnetic resonance imaging
• MRI
• mastectomy
• ultrasound and mammography
• US/MMG

Invasive lobular carcinoma (ILC) is the second most common histologic subtype of breast cancer, accounting for 5–15% of all breast cancers [1]. Histologically, ILC arises from the lobular epithelium and infiltrates the stroma in single cell file strands (Figure 1). Typically, it does not destroy the underlying anatomical tissues or cause a desmoplastic stromal reaction. This insidious growth pattern leads to low-density lesions and a lack of hemorrhage, calcification or necrosis, making it difficult to identify on mammography and ultrasonography [2]. Additionally, there is a characteristic tendency for ILC to present with multifocality, multicentricity or occult involvement of the contralateral breast [1]. These characteristics may explain the higher failure rate of breast-conserving therapy in patients with ILC [3], as additional ipsilateral or contralateral disease is often missed on conventional imaging. Indeed, the rates of positive margins following breast-conserving surgery (BCS) for ILC generally range from 30–60% [4–6].

Figure 1. Histopathological findings of metastatic invasive lobular carcinoma from a peritoneal sample.

The single file pattern of small cells is typical of invasive lobular carcinoma (hematoxylin and eosin, ×400 magnification).

Reproduced from Osaku et al. (2015) [7], licensed under CC BY 4.0.

Mammography (MMG) is considered the ‘gold standard’ imaging modality in the early detection of breast cancer, with a sensitivity of 77–87% [8,9]. It is much less sensitive for the detection of ILC, with a reported sensitivity of 57–79% [10]. Ultrasound (US) is often used in conjunction with MMG, providing up to a 40% increase in the detection of asymptomatic cancers, including ILC [11]. Magnetic resonance imaging (MRI) has a higher sensitivity of 90% for breast cancers and 93% for ILC [12]. The increase in sensitivity is attributed to neovascularization which allows a relative increase in gadolinium-based contrast uptake to the breast cancer stroma (Figure 2) [13]. Effectiveness of MRI in preoperative staging must be weighed up against its cost, increased delay time to treatment and increased anxiety to patients. The use of MRI as a screening tool for breast cancer does not seem to be cost effective for the general population, even with thresholds above $100,000 per quality-adjusted life-year [14]. Several authors have however demonstrated quantifiable benefits in preoperative staging with MRI in patients with known breast cancers, including ILC [15,16]. These include reduction in rate of conversion mastectomy, reduction in recurrence and reduction in re-excision rates [15].

Figure 2. Imaging findings of invasive lobular carcinoma.

(A) Craniocaudal and (B) mediolateral oblique views of the right breast demonstrate a spiculated lesion (ovals). (C) Ultrasound image demonstrates a hypoechoic lesion with echogenic rim. (D) Postcontrast MRI demonstrates rim-enhancement of the mass with extension of enhancement within adjacent tissue (oval).

Adapted with permission from Dhillon et al. [17].

Australian guidelines suggest that MRI should be considered if conventional imaging is insufficient to guide treatment [18]. MRI has also been suggested to be beneficial to patients with a significant hereditary history of cancers, past breast or ovarian cancers, including ductal carcinoma in situ (DCIS) or lobular carcinoma in situ (LCIS) or carriers of high-risk breast cancer gene mutations, although no firm recommendations have been made for the use of MRI in the preoperative workup of ILC [19].

Due to the relatively low incidence of ILC, few studies have investigated the role of MRI in preoperative staging and the subsequent impact on surgical and long-term outcomes. In this study, we aimed to assess the impact of preoperative MRI on tumor grading and proposed surgical management, as well as reoperation and cancer recurrence rates at our center.

Methods

The current study was a retrospective analysis of all patients diagnosed with ILC between January 2015 and December 2020 at the Royal Brisbane and Women's Hospital (RBWH).

A total of 133 patients with ILC were identified from the RBWH breast cancer database. Patients with concurrent histological types of invasive cancer other than ILC, as well as those with ductal carcinoma in situ, were excluded. Corresponding breast imaging findings, pathology reports and clinical notes were retrieved from the integrated electronic medical records (ieMR). The influence of preoperative findings on surgical management was ascertained from electronic copies of clinical notes and multidisciplinary team meeting recommendations. The position of the ILC lesions were recorded for each patient on each imaging modality (MMG, US and MRI) based on RBWH radiology reports. The size of lesions (measured as the greatest linear dimension of each lesion) on MRI and US was also recorded from radiology reports as the combined size of all lesions in each breast quadrant, as well as the size of the largest lesion in each breast quadrant (denoted as the index lesion for comparison). Electronic histopathology records were used to document corresponding position and size data for each patient. Outcomes, including the need for reoperation or regional or distal cancer recurrence, were also recorded from electronic copies of outpatient clinic notes and operation records.

Conventional ultrasound and mammography were performed by BreastScreen Australia, a national breast-screening program. All mammograms were interpreted using the Breast Imaging Reporting and Data System (BI-RADS).

All patients underwent MRI in the prone position using a dedicated bilateral breast coil at the same hospital (Siemens MAGNETOM Vida 3T, Siemens MAGNETOM Skyra 3T). All protocols included a series of T1-weighted sequences that were repeated at least four-times, first prior to the administration of a gadolinium contrast agent and then several times after intravenous contrast administration. tumor size was measured in coronal, axial and sagittal planes.

The diagnostic sensitivity of ILC detection between MRI and US compared with histological findings were reported as the sensitivity (%) and the corresponding 95% CI estimated using exact methods. Comparisons of the diagnostic sensitivity of ILC detection between MRI and US, the detection of multifocal and multicentric disease, and the detection of contralateral disease between methods were assessed using McNemar's test with exact probabilities. Comparisons of total and index lesion sizes between the methods were assessed using an exact sign test. Statistical significance was set at a p-value < 0.05 (two-sided). Comparison between disease recurrence between patients having MRI compared with those having conventional imaging only was performed using two-tailed Student's t-test. Stata version 15 (StataCorp, TX, USA) was used for all analyses.

Preoperative histopathological characterization and staging was performed by the RBWH pathology department. Patel et al. (2003) detail some recent techniques for preoperative diagnosis from breast cancer from histopathological images [20].

This project was classified as a quality assurance project in the opinion of the RBWH Human Research Ethics Committee (HREC), and as such was exempt from full ethical review. A letter of exemption was provided by the HREC. A waiver of consent was approved for collection and use of de-identified patient data, given the retrospective nature of this study.

Results

Patient & tumor characteristics

A total of 133 women diagnosed with ILC met the inclusion criteria. All 133 patients underwent surgery and 23 (17.2%) underwent preoperative MRI. The mean age of patients in the MRI- group was 65.4 years (SD: 12.28), and 53.5 years for patients in the MRI+ group (SD: 12.1, p < 0.001), although age was not found to correlate with additional MRI findings (X ² = 0.02, df = 1, p = 0.87). When assessing breast density on MRI, 9 patients had a Bi-RADS score of B and 14 patients had a Bi-RADS score of C or D. Increased breast density on MRI did not correlate with additional findings on MRI (X ² = 0.18, df = 1, p = 0.68).

Five women had known metastatic disease prior to surgery, and one woman was known to have a BRCA1 mutation. There were no other underlying health conditions in the 133 women that were thought to affect breast imaging.

Most patients who underwent preoperative MRI did so to assess the feasibility of BCS when they were on the borderline between being suitable for BCS or requiring a mastectomy, either due to borderline lesion size (∼5 cm) or poor characterization of lesions on ultrasound and mammography.

Imaging findings

The diagnostic sensitivity of ILC detection for the 133 patients imaged by US/MMG was 93.3% (95% CI: 87.6–96.9%), and the diagnostic sensitivity of ILC detection for the 23 patients imaged with MRI was 91.7% (95% CI: 73.0–99.0%). There were 23 patients with known lesions imaged by both MRI and MMG/US; for these cases there was no evidence of a difference in the diagnostic sensitivity of detection of ILC between MRI and MMG/US (p = 0.69), with respective diagnostic sensitivities of 91.3 and 82.6% (Table 1). Of these 23 patients, 17 had lesions detected by both MRI and MMG/US. Two patients had lesions detected only by MMG/US and four patients had lesions detected only by MRI.

Table 1. Diagnostic sensitivity of MRI and US.

Imaging modality	Overall		Matched cases
	N	Diagnostic sensitivity (%) (95% CI)	N	Diagnostic sensitivity (%) (95% CI)	McNemar's test p-value
MRI	24	91.7 (73.0–99.0)	23	91.3 (72.0–98.9)	0.69
US/MMG	134	93.3 (87.6–96.9)	23	82.6 (61.2–95.0)

MMG: Mammography; US: Ultrasound.

MRI detected additional ipsilateral and contralateral disease in 30.4% of patients, 17% of whom had no disease detected on conventional imaging. Presurgical MRI staging led to a change in surgical management in 9 of 23 patients (39.1%). Table 2 summarizes additional MRI findings. Of the patients in the MRI+ group, seven (30.4%) underwent a switch from BCS to mastectomy, while two (8.7%) required a switch from BCS to bilateral mastectomy.

Table 2. Additional findings on MRI.

Findings	Number	Percentage
Occult disease on ultrasound seen on MRI	4	17.4%
Contralateral disease	2	8.7%
Significantly larger disease (>5 cm)	2	8.7%
Additional ipsilateral disease	1	4.35%
No new findings	14	60.9%

There was no evidence of a difference in the total lesion size between MRI and histology (p = 0.30), with median total lesion sizes of 25 mm (IQR: 18–50.5) for MRI and 40 mm (IQR: 18–55) for histology. There was strong evidence of a difference in the total lesion size between US and histology (p < 0.001), with a smaller median total lesion size of 19 mm (IQR: 10–30) for US compared with 25 mm (IQR: 14–48) for histology. Note that the relatively small sample size of patients in the MRI+ group reduced the statistical power to detect a significant difference. There was also some evidence of a difference in the total lesion size between MRI and US (p = 0.039), with a smaller reported median total lesion size for US compared with MRI (25 (IQR: 14.5–35.5), 34 (IQR:21.5–55.3), respectively) (Table 3).

Table 3. Total lesion size comparison between imaging methods.

Imaging modality	N	Total size median (IQR)	Exact sign test p-value
MRI	17	25 (18–50.5)	0.30
Histology		40 (18–55)
US	113	19 (10–30)	<0.001
Histology		25 (14–48)
MRI	12	34 (21.5–55.3)	0.039
US		25 (14.5–35.5)

IQR: Interquartile range; US: Ultrasound.

Where possible, the largest lesion was categorized as an index lesion and compared between the imaging modalities and histology (Table 4). There was no evidence of a difference in the size of the index lesion between MRI and histology (p = 0.63), with median lesion sizes of 21 mm (IQR: 18–50) and 35 mm (IQR: 14–55), respectively. Some evidence of a difference in the size of the index lesion was found for US compared with both histology and MRI (p = 0.021 in both cases), with US reporting smaller median index lesion sizes in both comparisons. On average, US underreported lesion size compared with histology by 23% and MRI by 7% (Figure 3).

Table 4. Index lesion size comparison between imaging methods and histology.

Imaging modality	N	Largest size median (IQR)	Exact sign test p-value
MRI	19	21 (18–50)	0.63
Histology		35 (14–55)
US	109	16 (10–26)	0.021
Histology		22 (13–45)
MRI	16	25 (18–40)	0.021
US		15 (12–25.5)

IQR: Interquartile range; US: Ultrasound.

Figure 3. Size measured on histopathology for each individual lesion compared with the size measured prior to surgery on MRI and US/MMG.

Size measurements on MRI correlated better with lesion size on histopathology than conventional imaging. Dotted line represents a 1:1.

MMG: Mammography; US: Ultrasound.

Surgical management & long-term outcomes

Of the patients in the MRI+ group, 21.7% underwent BCS, and 78.2% underwent primary mastectomy. In the MRI- group, 45.9% underwent BCS and 54% primary mastectomy (OR 3.0, 95% CI: 1.0–8.8). The overall reoperation rate for patients who underwent BCS was 40% in the MRI+ subgroup and 37% in the MRI- subgroup (Figure 4 & Table 5), although the relatively small sample size of patients in the MRI+ group who received BCS (n = 5) makes it difficult to determine the statistical significance of these findings. There was no disease recurrence in the MRI+ subgroup, compared with one case of breast recurrence and two cases of distal disease recurrence in the remaining patients (p = 0.78), with a median time since commencement of therapy of 4.6 years (SD: 1.68 years).

Table 5. MRI outcomes by patient.

Age	Additional findings on MRI	Change in management	Reoperation following BCS	Breast density (BI-RADS)
29	Nil	Nil	Nil	D
38	Significantly larger disease (>5 cm)	BCS to mastectomy	N/A	D
41	Nil	Nil	Nil	D
41	Occult disease on US/MGG, seen on MRI	Mastectomy	N/A	C
44	Nil	Nil, opted for mastectomy	N/A	B
44	Additional discrete lesions	BCS to mastectomy	N/A	B
45	Nil lesion seen on MRI	Nil	Nil	B
45	Nil lesion seen in MRI	Nil	Nil	C
47	Nil	Nil	Nil	C
48	Nil	Nil	Nil	B
49	Occult disease on US/MMG, seen on MRI	Bilateral mastectomy	Nil	C
50	Occult disease on US/MMG, seen on MRI	Mastectomy	Nil	B
52	Nil	Nil	Yes	C
53	Nil	Nil	Nil	D
62	Nil	Nil	Nil	C
62	Nil	Nil	Yes	C
63	Multifocal disease on MRI	Nil	Yes	B
64	Occult disease on US/MMG, seen on MRI	Mastectomy	Nil	B
69	Contralateral disease	Bilateral mastectomy	Nil	C
70	Nil	Nil	Nil	B
70	Nil	Nil	Nil	C
72	Contralateral disease	Bilateral mastectomy	Nil	B
72	Multifocal disease	Mastectomy	Nil	C

BCS: Breast-conserving surgery; MMG: Mammography; N/A: Not applicable; US: Ultrasound.

Figure 4. Flowchart demonstrating initial surgical management and reoperation rates for ILC patients with and without preoperative MRI (percentage of patients, number of patients).

BCS: Breast-conserving surgery; ILC: Lobular carcinoma of the breast.

Discussion

MRI is known to be superior to conventional imaging in characterizing ILC [21], though there is currently little evidence of reduced rates of recurrence or disease-free survival with the use of preoperative MRI. The addition of preoperative MRI detected additional disease in 30.4% of our patients and led to a change in management in 39.1% (Table 2). MRI also provided closer estimates of true lesion size compared with conventional imaging, underreporting lesion size by a median of 7% compared with 23% on ultrasound. These findings are in line with studies by other authors, who have also noted the detection of additional cancer as well as a change in planned therapy of approximately a third of patients due to the addition of MRI.

In our cohort, the superiority of MRI in imaging ILC did not translate to improved clinical outcomes. The overall proportion of patients requiring re-excision of margins or completion mastectomy after BCS for ILC in our center (37.5%) was similar to that in other studies [5,22,23], and there was no significant difference between the MRI+ and MRI- groups. This is despite the fact that almost three-fourths of patients in the MRI+ group underwent MRI to assess the feasibility of BCS. Likewise, other authors have not noted a significant difference in reoperation rates with the addition of MRI for both lobular and ductal invasive carcinomas [24,25]. In fact, in a recent meta-analysis of 3374 patients assessing outcomes of ILC with preoperative MRI (the largest comparative analysis in this subgroup of patients in the existing literature), Houssami et al. [26] found that the addition of MRI does not improve surgical outcomes in women with ILC, suggesting that it is an unnecessary cost when used routinely. We also found no significant difference in locoregional or distal disease recurrence between the MRI- and MRI+ BCS subgroups.

It should be acknowledged that there is a significant resource burden associated with not only the use of MRI itself but also the subsequent need for repeat imaging and biopsies, time taken to analyze results, and multidisciplinary team discussions. The psychological burden on patients waiting for completion of preoperative workup should also be considered; anxiety and psychological distress in patients with a new diagnosis of breast cancer are exacerbated by lengthening the duration of the workup prior to definitive management [27].

This study has several limitations. First, our relatively modest sample of patients who underwent preoperative MRI may have had limited statistical power. Additionally, a significant proportion of patients who underwent preoperative MRI ended up undergoing mastectomy as 39% of these patients were upstaged. This left only a small subgroup of patients who underwent both MRI and BCS, making it difficult to compare reoperation rates between the subgroups with adequate statistical power. Second, the mean age of patients who underwent MRI was almost 11 years younger than that of patients who underwent conventional imaging only, which may suggest a selection bias of younger patients favoring BCS over mastectomy, and therefore, opting for MRI to assess the feasibility of BCS. It is unclear whether this bias is due to the preference of the surgeons or patients, but it may be relevant, as breast density is typically inversely correlated with age. However, our analysis found that breast density did not predict the presence of additional findings on MRI. Lastly, this was a non-randomized and retrospective study from a single center, which should be taken into account. To assess the impact of MRI on disease recurrence, we would recommend a larger sample size of patients with ILC. Irrespective, we hope our data will contribute to a growing body of evidence in the literature on this topic.

Conclusion

In conclusion, MRI is superior to conventional imaging in the preoperative workup of ILC, detecting additional disease in 30.4% of cases and changing management in 39.1% of cases in our center over a 6-year period. Our analysis however, in line with much of the current literature suggests that these superior imaging findings do not seem to translate to improved clinical outcomes. Further research, ideally in the form of prospective studies with larger sample sizes or meta-analyses, are required to investigate this issue.

Summary points

• Invasive lobular carcinoma (ILC) of the breast is challenging to diagnose and manage due to its occult nature and multifocality.

• MRI shows superiority in imaging ILC compared with mammography and ultrasound.

• Our retrospective single center study found that the addition of preoperative MRI detected additional disease in 30.4% of cases.

• MRI led to a change in surgical management, with 39.1% of patients switching to mastectomy from breast conserving surgery.

• MRI was significantly more accurate in predicting histological lesion size compared with conventional imaging methods.

• Despite these benefits, preoperative MRI did not significantly affect reoperation rates or disease recurrence, though larger populations are required to make guiding statements about these findings.

• While MRI enhances ILC detection, its impact on surgical outcomes is still unclear.

• Larger prospective studies are needed to further investigate the relationship between MRI imaging, surgical decisions, and clinical outcomes in ILC patients.

Author contributions

All authors contributed substantially to data collection, analysis and authorship of the manuscript. All authors have read and approved the final manuscript.

Financial disclosure

The authors have no financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.

Writing disclosure

No writing assistance was utilized in the production of this manuscript.

Ethical conduct of research

This project was classified as a quality assurance project in the opinion of the Royal Brisbane and Women's Hospital Human Research Ethics Committee (HREC), and as such was exempt from full ethical review. A letter of exemption, including an informed consent waiver, has been provided by the HREC. All methods were performed in accordance with relevant guidelines and regulations.

Acknowledgments

The authors would like to thank S Llewellyn for assisting with statistical analysis.

Competing interests disclosure

The authors have no competing interests or relevant affiliations with any organization or entity with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.