Breast microcalcifications: the UK RCR 5-point breast imaging system or BI-RADS; which is the better predictor of malignancy?

Abstract

Objective:

In the UK RCR 5-point breast imaging system (UKS), radiologists grade mammograms from 1 to 5 according to suspicion for malignancy, however unlike BI-RADS, no lexicon of descriptors is published. The aim of this study was to determine whether strict categorisation of microcalcifications (MCC) according to BI-RADS was a better predictor of malignancy than the UKS and whether these descriptors could be used within the UKS.

Methods:

A retrospective review of 241 cases, with MCC on mammography, who underwent biopsy was performed. Morphology, distribution, extent, UKS score, BI-RADS category and pathology were recorded. The positive predictive value (PPV) of each classification system for malignancy was calculated.

Results:

28.6% were diagnosed with DCIS/IDC. The PPV for malignancy using the UKS was 18.9%, 69.4%, 100% for M3-5 respectively (p < 0.001) and using ΒI-RADS morphology was amorphous: 7.1%, coarse heterogeneous: 33.3%, fine pleomorphic: 48.1% and fine linear/fine linear branching: 85.2% (p < 0.001). The PPV based on distribution was grouped: 14.2%, regional: 32.3%, diffuse: 33.3% and linear/segmental: 77.8% (p < 0.001). Combining all cases of benign-appearing, amorphous and grouped coarse heterogenous and grouped fine pleomorphic MCC gave a PPV of 12.8%. Combining regional, linear or segmental coarse heterogenous and fine pleomorphic and all fine linear/branching MCC resulted in a PPV of 83.3% for malignancy.

Conclusion:

Combining morphology and distribution of MCC is accurate in malignancy prediction. Use of BI-RADS descriptors could help standardise reporting within the UKS and an algorithm using these within the UKS is proposed. Better prediction would enable more appropriate counselling and help to identify discrepancies.

Advances in knowledge:

No guidance exists on scoring of suspicious MCC in the UK breast imaging system. Use of BI-RADS morphologic/distribution descriptors can aid malignancy prediction. Findings other than morphology of MCC are important in malignancy prediction. An algorithm for use by the UK radiologist when evaluating MCC is provided.

Introduction

Breast cancer is a leading cause of mortality in females resulting in high breast imaging volumes in radiology departments. To improve the radiological management of breast cancer and communication among clinicians and other health care professionals, standardised and synoptic reporting systems have been developed worldwide and are advocated as the reporting method of choice.^1,2 This style of reporting and, more specifically, the use of checklists, have been shown to be the most effective methods of improving the content and completeness of reports with potential to improve care for females. Standardised reporting may also reduce the risk of misinformation, improve clinician satisfaction, adapt to new technologies and increase the ability to compare research or develop guidelines.^3,4 These reporting systems guide further management and in some cases grade imaging findings as to the likelihood of malignancy. Although the Breast Imaging and Reporting Data System (BI-RADS) is the commonest breast reporting system worldwide, it is not suitable for all healthcare settings. Therefore other systems have been developed and are currently in use in countries such as Australia, New Zealand, and the United Kingdom.^5,6

The American College of Radiology (ACR),^7,8 has devised a detailed classification system which combines the morphological features of lesions with the possibility of malignancy. In the Breast Imaging Reporting and Data System (BI-RADS) fifth edition (Figure 1), four morphological descriptors for suspicious MCC have been described: amorphous, coarse heterogeneous, fine pleomorphic, and fine linear/fine linear branching (Figure 2). These are further subdivided into two categories, with the first three (amorphous, coarse heterogeneous, fine pleomorphic) placed in Category 4b (10–50% possibility of malignancy) and fine linear/ fine linear branching placed in Category 4c (50–95% possibility of malignancy).⁸ BI-RADS Category 3 is reserved for an isolated group of round, punctate calcifications, that is new, increasing, linear, or segmental in distribution, or adjacent to a known cancer. The recommendation for this category is a follow-up mammogram at 6 months with incidence of malignancy less than 2%.⁸ All typically benign calcifications are placed in Category 2.

Figure 1. .

Breast Imaging Reporting and Data System (BI-RADS) fifth edition guidelines and the Royal College of Radiology (RCR) 5-point breast imaging system (UKS) concerning microcalcifications (MCC). m, months; y, years; mm, millimetres.

Figure 2. .

Magnification views of mammograms showing the BI-RADS morphological descriptors of microcalcifications A: amorphous B: coarse heterogeneous C: fine pleomorphic D: fine linear/branching.

A UK Royal College of Radiology (RCR) 5-point breast imaging system (UKS) for mammography, ultrasound and cytology was described in the UK in 1998 and was formalised by Maxwell et al on behalf of the RCR Breast Group in 2011.^6,9 Unlike BI-RADS, it does not give the likelihood of malignancy in each category and a standardised lexicon of terminology is not published. The UKS is 1 = normal; there is no significant imaging abnormality, 2 = benign findings; the imaging findings are benign, and further investigation purely on the basis of the imaging findings is not indicated, 3 = indeterminate/probably benign findings; there is a small risk of malignancy, and further investigation is indicated, 4 = findings suspicious of malignancy; there is a moderate risk of malignancy and further investigation is indicated, 5 = findings highly suspicious of malignancy; there is a high risk of malignancy and further investigation is indicated.⁶ Mammograms and ultrasounds are scored M:1–5 and U:1–5 respectively. Regarding calcifications; bilateral benign powdery microcalcification, benign secretory microcalcification and ‘‘popcorn’’ calcification in fibroadenomas are included in score M2 and there is no reference to scoring the remaining types of calcification.

The primary purpose of this study was to determine whether strict categorisation of MCC according to BI-RADS morphological descriptors is a better predictor of malignancy than the UKS, as the latter allows a radiologist to also take the distribution and size of MCC into consideration when allocating a score. Secondly, we aimed to determine whether further alterations or recommendations should be made to improve the UKS as it continues to evolve. Additionally, further data on the UKS could aid with international publication of UK breast research and as a guide to medical professionals coming to work in the UK and Ireland from abroad as we identify a gap in the literature.

Methods and materials

Data was collected in a breast unit, which provides mammography screening to over 20,000 females per year and symptomatic services to a population of over 2 million. A retrospective review was performed of all patients with suspicious MCC detected on mammography over a 1-year period (1 October 2014 to 30 September 2015) who underwent stereotactic vacuum biopsy.

All females attended screening assessment clinics (n = 182) or symptomatic clinics (n = 72) and underwent triple assessment (clinical assessment, imaging and tissue sampling) based on national guidelines.^10,11 Females had standard 2D bilateral digital mammography (mediolateral oblique and craniocaudal projections) performed. Following identification of MCC, a true lateral projection of the side-of concern, magnifications views of the area(s) of interest in two planes (lateral and craniocaudal) and targeted breast ultrasound (US) were undertaken. One of 7 dedicated consultant breast radiologists (2–20 years experience) evaluated the imaging, allocated a UKS score to the mammograms and following written consent, performed stereotactic vacuum biopsy of the suspicious MCC. Only patients who consented to vacuum biopsy and had a specimen X-ray confirming inclusion of MCC in the biopsy sample were included for analysis. Patients who had more than one focus of suspicious MCC, and underwent more than one vacuum biopsy, had each biopsy site considered separately. For each site the histopathology and surgical data were recorded. In the case of B3 biopsy result, residual MCC (if present) was removed with a vacuum excision biopsy if technically feasible. If not technically feasible females underwent follow up imaging or surgical excision according to international guidance. For B4 and B5 vacuum biopsy results, patients had surgery following image-guided wire localisation. For those patients returned to routine recall following a benign (B2) biopsy, the rate of false negative assessment and interval breast cancer was recorded.

For the purposes of the study, two experienced dedicated breast radiologists independently reviewed the pre-biopsy digital magnification views, blinded to histopathology results, and graded the morphological features of the MCC according to BI-RADS fifth edition and allocated a UKS score. In addition, the distribution and size of MCC were recorded. In cases where there was a discrepancy (n = 5) between the two readers, a third consensus reading was taken. Benign appearing MCC (BI-RADS 2–3) included round or rim, rod-like, curvilinear or pop-corn like MCC. Suspicious MCC were classified according to the specific BI-RADS fifth edition descriptors: amorphous, coarse heterogeneous, fine pleomorphic, and fine linear/fine linear branching (Figure 2). Readers recorded the distribution of MCC according to BI-RADS: grouped, regional, linear/segmental, and diffuse (Figure 3). The extent was determined by TNM staging: 0.1–0.5 cm, 0.5–1 cm, 1–2 cm, 2–5 cm, >5 cm. Patients with a mass, distortion or sonographic finding in association with MCC were excluded from the study in order to focus on the classification of pure MCC. If different morphological types of MCC were present, imaging was classified according to the most suspicious focus (Figure 4). Data also recorded included whether previous mammograms were available for comparison, whether there was a history of breast cancer, high risk family history or other increased risk lesion available (e.g., ADH) to the reporting radiologist.

Figure 3. .

Distribution descriptors of microcalcifications in the breast. A: grouped B: regional C: linear/segmental D: diffuse.

Figure 4. .

An 81-year-old female. Magnification views demonstrate several clusters of microcalcifications with different morphologies including coarse heterogenous, fine pleomorphic and linear like microcalcifications. Stereo-guided vacuum biopsies diagnosed High-Grade DCIS.

Histological results were classified based on the five point system: B1 = normal tissue, B2 = benign lesion/s, B3 = uncertain malignant potential, B4 = suspicious, B5a = malignant in situ, B5b = malignant invasive.¹² For benign lesions, the vacuum biopsy result was recorded, as this was the final histological sample and for malignant lesions, which underwent surgical excision, the data recorded was histology of the surgical specimen.

Ethics committee approval is not required for retrospective review, but permission was granted for publication by the director of education and research.

Statistical analysis

The positive predictive value (PPV) was determined based on the number of females who had histopathological diagnosis of malignancy (B5a/B5b on final histology) in each category of the two scoring systems, with each morphological descriptor, and with reference to extent and distribution. Statistical analysis was performed in Excel (Office 2016), and values were calculated with chi-square (χ²) test.

Results

265 vacuum biopsies of MCC were performed from October 2014 to September 2015, in 254 females, aged 42–82 years (mean 58 years) of which 241 were eligible for inclusion (Table 1). 130 (53.9%) biopsies were benign (B2) including sclerosing adenosis, duct ectasia, fibroadenomatoid change and stromal fibrosis. 42 (17.4%) were indeterminate risk lesions (B3) including MCC in association with flat epithelial atypia, radial scar, atypical ductal proliferation, atypical ductal or lobular hyperplasia, and papillary lesions with luminal calcifications. If the MCC were not completely removed, further diagnostic or excision vacuum biopsy was performed and none were upgraded to malignancy. Two were classified initially as B4 but following surgical biopsy, one was downgraded to B3 (LCIS) and one upgraded to B5a (DCIS). 69 (28.6%) were diagnosed as malignant, including ductal carcinoma in situ (DCIS) and invasive ductal carcinoma (IDC), 81.2 and 18.8% respectively. No patient, from this study cohort, who was returned to routine recall following a benign (B2) biopsy, was known to develop breast cancer/deemed to have a false negative assessment in the follow-up period after biopsy (mean 35 months). Of the patients with a B3 biopsy, two developed breast carcinoma in the follow-up period (mean 38 months). Both of these patients were on annual surveillance.

Table 1. .

Positive predictive value (PPV) according to Royal College of Radiology (RCR) 5-point breast imaging system (UKS) with reference to the equivalent Breast Imaging Reporting and Data System (BI-RADS category) (p value <0.001)

UKS	B2	B3	B5a	B5b	Total	PPV%	BI-RADS category
M2	1	1	0	0	2	0.0	1–2
M3	121	38	30	7	196	18.9	4b
M4	8	3	20	5	36	69.4	4c
M5	0	0	6	1	7	100.0	5
Total	130	42	56	13	241

159/241 (66%) did not have previous mammograms available for comparison to the reporting radiologist. Of the 82 with previous imaging, 54 (66%) demonstrated new microcalcifications and 28 showed progression in number and change in morphology of microcalcifications resulting in recall for biopsy. 24/54 (44.4%) of new microcalcifications and 8/28 (29%) of altered appearing microcalcifications were malignant. The large majority of mammograms were reported in a screening environment with no clinical history available. Of those reported from a symptomatic clinic (72/254), 16 had factors in the clinical history provided that increased the risk of breast cancer: six family history of breast cancer (5 high and one moderate risk), one previous diagnosis of ADH in the ipsilateral breast, seven personal history of cancer in the ipsilateral breast and two in the contralateral breast.

PPV of the UK score

The positive predictive value (PPV) for malignancy using the UKS allocated in the clinical setting was; M2: 0%, M3: 18.9%, M4: 69.4%, M5: 100% (p value < 0.0001) (Table 1). 37/69 (53.6%) of patients who were subsequently diagnosed with malignancy were given a probably benign score on imaging (M3).

Comparison of UKS with BI-RADS

The rates of malignancy within each UKS were mapped to a BI-RADS category (Table 1). 18.9% of M3 (indeterminate-probably benign findings) scores resulted in malignancy, which corresponds to BI-RADS Category 4b (likelihood of malignancy 10–50%). The rate of malignancy for an M4 score (findings suspicious of malignancy) was 69.4%, which corresponds to BI-RADS Category 4c (likelihood of malignancy 50–95%). The M5 category was equivalent to BI-RADS Category 5 with a 100% incidence of malignancy.

PPV of BI-RADS morphological descriptors

PPVs for malignancy based on the morphological descriptors in the BI-RADS fifth edition were; benign appearing (BI-RADS 2–3): 0%, amorphous: 7.1%, coarse heterogeneous: 33.3%, fine pleomorphic: 48.1%, fine linear/ fine linear branching: 85.2% (Table 2). There was a statistically significant difference in the rate of malignancy with the different morphological descriptors (p value < 0.0001).

Table 2. .

Positive predictive value for malignancy based on the BI-RADS morphology and distribution

		Histopathology Grading System
		B2	B3	B5a	B5b	Total	PPV%
BI-RADSa	Morphology
	Benign appearing	32	2	0	0	34	0.0
	Amorphous	57	21	5	1	84	7.1
	Coarse heterogenous	23	5	10	4	42	33.3
	Fine pleomorphic	16	12	21	5	54	48.1
	Fine linear/fine linear branching	2	2	20	3	27	85.2
	Distribution
	Grouped	109	30	17	6	162	14.2
	Regional	16	5	7	3	31	32.3
	Diffuse	0	2	1	0	3	33.3
	Linear/segmental	5	5	31	4	45	77.8
	Total	130	42	56	13	241

^aBI-RADS: The Breast Imaging Reporting and Data System (p value <0.001).

PPV of BI-RADS distribution descriptors

The PPVs for malignancy based on the distribution of the MCC were: grouped MCC: 14.2%, regional MCC: 32.3%, diffuse MCC: 33.3%, linear/linear segmental MCC: 77.8% (Table 2). There was statistically significant difference in the likelihood of malignancy between the different distribution types (p value < 0.0001).

PPV of TNM size categories

The PPV for malignancy increased with increasing size:<0.5 cm PPV = 0% (N = 0/32), >0.5 but less than <1 cm PPV = 13.0% (N = 10/77), >1 but less than 2 cm PPV = 30.4% (N = 17/56), >2 but<5 cm PPV = 47.8% (N = 22/46) and >5 cm PPV = 66.7% (N = 20/30). There was a statistically significant difference for malignancy with increasing size (p value < 0.0001).

Malignancy rates comparing BI-RADS morphological and distribution descriptors to UKS

The incidence of malignancy when the BI-RADS morphological descriptors and distributions are combined are described in Table 3. The differing rates of malignancy within each of the morphological descriptor groups depending on radiologist UKS is shown in Table 4. This demonstrates an increasing incidence of malignancy within each descriptor group with an increasing score.

Table 3. .

Positive predictive value for malignancy based on distribution and morphology

	Benign appearing (%)	Amorphous (%)	Coarse heterogeneous (%)	Fine pleomorphic (%)	Fine linear/ Fine linear branching (%)	Total (%)
Diffuse	0	1/3 (33.3)	0	0	0	1/3 (33.3)
Grouped	0/31 (0.0)	4/60 (6.7)	7/34 (20.6)	11/35 (31.4)	1/2 (50.0)	23/162 (14.2)
Regional	0/2 (0.0)	1/18 (5.6)	1/1 (100.0)	5/7 (71.4)	3/3 (100)	10/31 (32.3)
Linear/ segmental	0/1 (0.0)	0/3 (0.0)	6/7 (85.7)	10/12 (83.3)	19/22 (86.4)	35/45 (77.8)
Total	0/34 (0.0)	6/84 (7.1)	14/42 (33.3)	25/54 (46.3)	23/27 (85.2)	241

Table 4. .

Rates of malignancy within the descriptor groups according to Royal College of Radiology (RCR) 5-point breast imaging system (UKS) (34 cases of benign-appearing microcalcifications excluded)

		Histopathology Grading
	Morphology	B2	B3	B5a	B5b	Total	PPV %a
BI-RADSb DESCRIPTORS /UKSc	Amorphous
	M3	55	21	3	1	80	5.0
	M4	2	0	2	0	4	50.0
	Coarse heterogenous
	M3	22	5	7	2	36	25.0
	M4	1	0	3	2	6	83.3
	Fine pleomorphic
	M3	13	10	13	3	39	41.0
	M4	3	2	8	1	14	64.3
	M5	0	0	0	1	1	100.0
	Fine linear/fine linear branching
	M3	1	1	7	1	10	80.0
	M4	1	1	7	2	11	81.8
	M5	0	0	6	0	6	100
	Total	98	40	56	13	207

^aPPV: positive predictive value.

^bBI-RADS: The Breast Imaging Reporting and Data System.

^cUKS: Royal College of Radiology (RCR) 5-point breast imaging system.

Inter-reader variability of the UK score

The UK score allocated to the mammogram in the clinical setting correlated to the score allocated in the study setting with an inter reader reliability of 88%. In 13 cases, the score was upgraded by the study radiologists and in 16 cases it was downgraded.

Discussion

The Breast Imaging-Reporting and Data System (BI-RADS), although the commonest breast reporting system worldwide, is not universally used, with countries such as the UK, Ireland and Australia using an alternative system.⁸ The reasons relate predominantly to differences in practice. Lesions that are considered as indeterminate (BI-RADS 3) are recommended to have follow-up every 6 months for 2 years and in the US there is widespread practice of annual screening mammography.⁶ In the UK, females having a screening mammogram either undergo biopsy if required or are discharged until their next screening mammogram in 3 years.¹⁰ An annual follow-up mammogram is performed infrequently following routine screening, usually if biopsy is not feasible.¹⁰ An annual mammogram is only standard practice for a small group of females at significantly increased risk of developing breast cancer, including gene mutation carriers, personal history of breast cancer and biopsy proven B3 lesions.

The paper by Maxwell et al formalizing the UK RCR 5-point breast imaging system, did not provide a likelihood of malignancy within each category or provide detailed guidance on the scoring of MCC with the exception of describing what constitutes typically benign calcification.⁶ 54% of patients in the current study diagnosed with malignancy were initially given an M3 score, probably benign, suggesting an overuse of this category. RCR guidance suggests that 90% of patients subsequently diagnosed with malignancy are given an M4 or M5 score.¹³ Taylor et al, provided a quantification of the likelihood of malignancy for each category of the UKS.¹⁴ In their study, the incidence of malignancy for M1 and M2 scores was less than 2%, M3 41%, M4 95% and M5 98%. However, it did not focus on the scoring of MCC and was undertaken in a symptomatic setting rather than the screening and symptomatic setting of this study.

The categorisation of MCC is a challenging task for radiologists as benign MCC can often mimic malignant MCC and vice versa (Figures 5–7). Studies have also shown moderate to significant inter-observer variability in the interpretation of MCC by radiologists; in this study seven different radiologists scored the mammograms in the clinical setting with an inter-reader reliability of 88% between them and the study radiologists.¹⁵ Both of these variables may account for the differing rates of malignancy in our study compared to Taylor et al,¹⁴ for M3 (19 vs 41% respectively) and M4 (69 vs 95% respectively). Similar to their findings, the UKS M3 category correlated to BI-RADS 4b and UKS M4 to BI-RADS 4c in our study.

Figure 5. .

Challenging categorisation. A 68-year-old female. Magnification view of microcalcifications classified as M4 (UKS system). Histopathology demonstrated fibrosis associated with benign-type calcifications, graded B2.

Figure 6. .

Challenging categorisation. A 64-year-old female. Mammogram including magnification view demonstrating a cluster of coarse heterogenous calcifications that could easily be mistaken as benign. However, the presence of fine pleomorphic calcifications posteriorly raised suspicion and biopsy diagnosed Grade 3 Invasive Ductal Carcinoma.

Figure 7. .

Challenging categorisation. A 50-year-old female. Magnification view of a group of microcalcifications that was graded as M4 (UKS system). It has a combination of amorphous, coarse heterogenous and fine pleomorphic microcalcifications. Biopsy diagnosed a radial scar associated with microcalcifications: B3.

Several previous studies have evaluated the rates of malignancy associated with each of the BI-RADS morphological descriptors for suspicious MCC. Recent studies have found PPVs for malignancy for amorphous MCC of 6.5 and 7.9% which are concordant with our results of 7.1%^16,17 however differ considerably from the 20–25% incidence quoted in other studies.^18,19 Our study, however, has larger numbers within each morphological grouping. The PPVs for malignancy for coarse heterogeneous (33.3%) and fine pleomorphic MCC (48.1%), in this study, are concordant with the BI-RADS fifth edition and 4b category (10–50%) but again vary between studies. Incidences of malignancy for coarse heterogenous of 3–20% and fine pleomorphic of 28–63% are quoted in the literature.^16,18–20 The rates for fine linear/linear branching are between 50–100% and are concordant with our incidence of 85.2%.^16,18–20 The wide variation in results may be due to the difficulty that can exist in distinguishing between different types of calcification in some patients; large, dystrophic calcifications that can occur in different pathological processes and the early changes of popcorn calcification in a fibroadenoma can mimic the stromal reaction of a malignant tumour.²¹ In addition, over the last 15 years different morphological descriptors for MCC have been used in various BI-RADS editions and therefore terminology in the literature varies.⁸ For fine linear/ fine linear branching MCC, all studies agree with the BI-RADS categorisation of 4c and show incidences of malignancy ranging from 66.7 to 100% which is concordant with our 85% incidence (9-12).

In this study, no patient that was graded as BI-RADS Category 2/3 had a diagnosis of malignancy (14% of cases). The challenge for the radiologist in the UK is that if biopsy is not performed during the screening services, the next screening mammogram is in 3 years, which could result in an interval or delayed presentation of malignancy. The UK scoring system is in essence a binary system where the radiologist decides between no-biopsy (scores 1–2) and biopsy (scores 3–5). Breast radiologists are keenly aware of the balance between maintaining a low recall rate (not subjecting females to unnecessary vacuum biopsies of benign calcifications) and not missing a cancer diagnosis if the microcalcifications are not easily categorised. Whether MCCs are scored M3 or 4–5 prior to initial biopsy may affect the counselling of patients in assessment clinics and guide the need for repeat biopsy if a discrepancy between the radiology score and the histology result arises. However, in the vast majority of the cases, whether the initial score is M3 or M4 does not affect the primary work-up and management of the patient particularly given the MDM discussion of all cases post-biopsy. BI-RADS is not a binary system and provides a multiple choice to the reporting radiologist: discharge, follow-up or biopsy. Although short-term recall is not advised by the NHS Breast Screening Program, this may have avoided an unnecessary biopsy in the cases in our study that were graded BI-RADS Category 2/3, particularly where the radiologist is almost certain that the MCC are benign. However, a study that included indeterminate and suspicious clusters of MCC, has shown that stability (at 6 months and a year) cannot be relied on as a reassuring indicator that findings are benign.²² A recent study, by Michaels et al, highlighted that 41 of 542 (7.5%) cases of microcalcifications were upgraded from BI-RADS 3 to 4 or 5 during imaging surveillance of at least 24 months.²³ Of these 8 (1.5%) were deemed to be malignant on biopsy. The lack of a ‘follow-up’ category in the UK scoring system potentially avoids this possibility of a delayed diagnosis of malignancy in females with relatively stable but indeterminate microcalcifications.

We aimed to determine whether the use of the BI-RADS descriptors could aid radiologists in allocating a UKS score. A more formal and standardised approach could improve uniformity between radiologists and centers in the UK and act as an aid for radiologists trained abroad. Based on our results, if morphological features were considered in isolation, all round/punctate and amorphous MCC could be graded as M3, resulting in a 5.1% incidence of malignancy, corresponding to the ‘probably benign’ UKS description. If coarse heterogeneous, fine pleomorphic and fine linear/fine linear branching MCC were graded as M4 it would have resulted in an incidence of 51.2% in this category, corresponding to the ‘findings suspicious for malignancy’ description.

However, when evaluating the presence of MCC on a mammogram, the radiologist looks at both the appearance of the individual pieces of calcium and their distribution including extent. In addition, the radiologist may have previous mammograms for comparison to assess for interval change.^24,25 They may also have knowledge of patient history including gene mutation carrier status, personal or family history of malignancy, or history of B3 lesion, all of which increase the likelihood of malignancy.^26–31 It is a combination of these features that guides the UKS score allocated by the radiologist. The ability to combine these features has the potential to improve the PPV for malignancy. This is already demonstrated by the 100% PPV for malignancy in the UKS M5 group unlike BI-RADS where MCC alone is not scored BI-RADS 5. In our study, the presence of new microcalcifications since a previous mammogram, without regard to morphology, had an incidence of malignancy of 44.4% compared to 28.2% in the study population. In contrast, the BI-RADS lexicon does not provide guidance on how the distribution, changing appearance and size of MCC affects the possibility of malignancy and studies have not focused on their influence on incidence of malignancy.^{17,20,32–34} The evaluation of these features and clinical history is almost certainly the reason for our finding that the PPV of malignancy for coarse heterogenous MCC if scored M3 was 25% but if scored M4 was 83%. Similarly, the results for fine pleomorphic were 41% if scored M3 and 64% if scored M4 (Table 4). Our results show that distribution of MCC is an independent predictive factor of malignancy, with a linear/segmental distribution being the most suspicious with a PPV of 77.8% (p value <0.001) and increasing extent of MCC increases the rate of malignancy.

Combining the morphology and the distribution of MCC might be a more accurate tool in malignancy prediction. Guidance within the UKS, in this respect could improve radiology performance in predicting biopsy results and increase standarisation and uniformity of reporting among radiologists. Better prediction would enable more appropriate counselling of patients. However, increased uniformity of radiological reporting in the UKS 3 and 4 categories would not change clinical practice in assessment clinics or affect recall rates as all of these patients undergo biopsy unlike in the BI-RADS system. We have not included size criteria within the proposed categorisation as this could increase complexity and reduce accuracy. In our study, if morphology and distribution were combined: punctate/round and amorphous MCC in all distributions and grouped heterogenous and fine pleomorphic MCC could be categorised as M3, giving a PPV of malignancy of 12.8% (24/187 patients). Regional, linear or segmental coarse heterogenous and fine pleomorphic MCC and all forms of fine linear/fine linear branching MCC could be categorised as M4, with a PPV of malignancy of 83.3% (45/54 patients). Radiologists may also choose to categorise segmental fine linear branching MCC as M5 given the classical appearance for malignancy (Figure 8). This categorisation would help radiologists in determining the radiologic and pathologic concordance of biopsy results, in particular highlight the need to consider re-biopsy for patients in the M4 category given the high PPV for malignancy. This could potentially reduce the risk of a false negative assessment and interval cancer.³⁵

Figure 8. .

Proposed subcategorisation of microcalcifications according to morphology and distribution.

Limitations

This study contains higher incidences of amorphous and grouped MCC and less linear/linear branching and diffuse MCC. Our study numbers are larger than previous studies, however some of the numbers in the subgroups are low and a larger study may be of benefit.^17,18 All screening mammograms were reported by two radiologists which reduces the risk of a false negative.³⁶ However, no data is available on the incidence of interval breast cancer in females who had microcalcifications present on mammography, but were not recalled during the study period, if the microcalcifications were interpreted as benign by both radiologists. The exact impact of the comparison of the current with previous mammograms and the patients’ history of breast disease on the UKS score was not determined. A larger intra-observer variability in the UKS scoring of MCC may have occurred as seven different radiologists scored the mammograms in the clinical setting at the time of biopsy compared to two radiologists in the retrospective determination of BI-RADS morphology, size and distribution.

Conclusion

The UK Royal College of Radiology 5-point breast imaging system (UKS) was formalised only relatively recently compared to BI-RADS. The scoring of MCC, in particular, provides a diagnostic challenge for breast radiologists and BI-RADS publishes extensive information on morphological and distribution descriptors to assist radiologists. We propose guidance on the scoring of MCC within the UKS to improve and standardise mammography reports and to enable appropriate counselling of females and MDM discussion. This would result in a PPV for malignancy of 12.8% for M3 and 83.3% for M4 corresponding to the ‘probably benign’ and ‘suspicious for malignancy’ categories described in the UKS.