Screen-detected breast cancer: does presence of minimal signs on previous mammograms predict staging/grading of cancer?

AIM
To investigate whether the presence of minimal signs on prior mammograms predict staging or grading of cancer.


MATERIALS AND METHODS
The previous mammograms of 148 consecutive patients with screen-detected breast cancer were examined. Women with an abnormality visible (minimal signs) on both current and prior mammograms formed the study group; the remaining patients formed the control group. Age, average size of tumour, tumour characteristic, histopathology, grade, and lymph node status were compared between the two groups, using Fisher's exact test. Cases in which earlier diagnosis would have made a significant prognostic difference were also evaluated.


RESULTS
Eighteen percent of patients showed an abnormality at the site of the tumour on previous mammograms. There was no statistically significant difference between the two groups with respect to age, average size of tumour, histopathology, grade or lymph node status with p-values being 0.609, 0.781, 0.938, and 0.444, respectively. The only statistically significant difference between the two groups was tumour characteristics with more microcalcifications associated with either mass or asymmetrical density seen in the study group (p=0.003). Five patients in the study group showed lymph node positivity and were grade 3, and therefore, may have had possible gain from earlier diagnosis.


CONCLUSION
The present study did not demonstrate a statistical difference in grading or staging between the group that showed "minimal signs" on prior mammograms versus normal prior mammograms. Microcalcification seems to be the most common characteristic seen in the missed cancer and a more aggressive management approach is suggested for breast microcalcifications.

Introduction Seventeen percent of women undergoing surgery for screen-detected breast cancer undergo more than one operation (BASO audit 2008). The aim of this study is to identify pre-operative data items that predict which cases will require more than one application. Methods Over an 11 year period (April 1996 to April 2007) 1,193 women presented with screen-detected breast cancer through the South East London Breast Screening Programme. For each case, imaging and biopsy data were reviewed in order to predict which women undergoing wide local excision (WLE) would require further operations. Age, invasive cancer size, ductal carcinoma in situ (DCIS) size, histology, cancer grade, nodal status, mammographic sign and level of suspicion of malignancy (RCR Breast Group 1-5 scale) were compared. Results Of the 1,193 women, 916 (77%) had a WLE or excision biopsy. These were split into two groups: group 1 (520 women (57%)) had a single operation; and group 2 (396 women (43%)) went on to have further operations, including repeat WLE to clear margins 238 (60%), mastectomy 133 (34%) and axillary dissection 25 (6%). Twenty-eight women went on to have a third operation. There were significant associations with increasing size of DCIS (P < 0.001), increasing tumour grade (P = 0.005), radiologists suspicion of 3/4 (P = 0.002) and mammographic sign (spiculated mass plus microcalcifications and asymmetric density (ASD), P = 0.003) for repeat operations. There was no statistical difference in patient's age, invasive cancer size, histology or nodal status. Conclusion The multidisciplinary team should pay particular attention to DCIS extent and mammographic features of the tumour when planning surgical management in order to minimise the number of women requiring repeat operations.

O5
Multistatic radar: first trials of a new breast imaging modality M Shere 1 , A Preece 2 , I Craddock 3 , J Leendertz 2 , M Klemm 3 1 Frenchay Hospital, Bristol, UK, 2 Medical Physics Department, University of Bristol, Bristol, UK, 3 Engineering Department, University of Bristol, Bristol, UK Breast Cancer Research 2009, 11(Suppl 2):O5 (doi: 10.1186/bcr2369) None of the current breast imaging modalities are perfect. Mammography has quite a low sensitivity, uses ionising radiation and is uncomfortable. Ultrasound has better sensitivity, is cheap, but is user dependent and not very good for screening. Magnetic resonance imaging has good sensitivity but is expensive. Positron emission tomography is extremely expensive. Other modalities have been tested, such as thermal imaging, red light imaging and elastography. We present a new modality -microwave RADAR imaging. This uses electromagnetic radiation in the gigahertz range (similar to a mobile phone, but much lower power). Research has been going on for many years using single emitters and antennae, which has proven the concept but was unable to produce high definition images. The breakthrough came with using technology developed for land mine radar detection -multistatic arrays. The third generation imaging apparatus uses 31 microwave emitters arranged in a hemisphere in which the breast sits in a prone table system. This gives 435 different pathways through the breast, which go into a network analyser and then are translated into a three-dimensional image for viewing and manipulation. We have now done the initial trials with first use in humans; 160 women with lesions have been scanned and are being compared with mammography and ultrasound. The results are encouraging; most lesions are visible and it has a good ability to distinguish between benign and malignant lesions with a modality that is quicker, cheaper, safer and more comfortable. Pictures of the apparatus and images produced will be shown. Interval cancer review is an important part of quality assurance within the NHSBSP. No equivalent assessment is performed within the symptomatic service. We attempted to apply a similar review to the symptomatic work performed at our centre. Methods All patients referred to our centre with a breast symptom are seen in 'one-stop' triple assessment clinics. We identified, and reviewed the imaging of, patients diagnosed with in situ or invasive carcinoma at our unit during 2007 who had undergone breast imaging via the symptomatic service within 3 years prior to diagnosis. Results Between January 2005 and December 2007, 11,796 ultrasounds and 17,530 mammograms were performed within the symptomatic service. Of the total 681 cancers diagnosed in 2007, 34 of these were 'interval cancers' to the symptomatic service. The interval since discharge was ≤12 months in 9 patients, 13 to 24 months in 18 and 25 to 36 months in 7.
Fourteen cancers were felt to be true intervals, six mammographically occult, and five suspicious. Nine were unclassifiable. Of the unclassifiable cases, one had missing imaging and eight did not have mammography at prior attendance. Only four of these were symptomatic in the same area as their cancer was later diagnosed. Conclusion 'Interval cancers' within the symptomatic population are a small but heterogeneous group. We discuss the details and present a pictorial review of the interesting cases. Introduction Fibroadenomas (FAs) present as common breast lesions in young women, often necessitating core biopsy/fine needle aspiration. Our unit protocol has been to biopsy suspected FAs in women aged 20 years and over. Literature suggests there is a case for safe non-biopsy in the under 25s. We wanted to establish whether it would be safe practice to stop biopsying FAs in women <30 years of age. Methods A theoretical incidence of a benign presentation of breast cancer in our unit was established using national statistics and Stavros criteria [1]. Using this, an imaging criteria-based protocol for nonbiopsy of FAs was devised, which we retrospectively tested against our departmental practice over the period 2000 to 2008 in women <30 years of age. Results Between 2000 and 2008, 9 cancers were diagnosed in this age group and 490 fibroepithelial lesions were core biopsied, resulting in 479 histologically B2 FAs and 11 B3 lesions. Of the 11 B3 lesions, subsequent surgical histology downgraded 9 to FAs, the remaining 2 being phyllodes tumours. Case review of both the phyllodes and the cancers, applying retrospective non-biopsy criteria, demonstrated that none of these lesions (B3 and above) would have been missed. Conclusion Our new protocol provides a sound evidence base for non-biopsy of FAs in women <30 years of age. This is safe practice provided there is rigid adherence to the protocol and this is currently being audited. Methods A retrospective 6-month audit was carried out reviewing all cases coded as R3. Further investigation and outcome of these lesions was then assessed using radiology, pathology and multi-disciplinary meeting records. All cases that proved to be malignant on histology had their imaging reviewed by three independent radiologists to see if they had been correctly coded as R3.

Reference
Results This study included 140 lesions. Of these, 28 (20%) were downgraded to benign on further imaging and 111 (79%) were biopsied. One patient refused biopsy. Nineteen (14%) lesions were malignant. On imaging review of malignant cases, the consensus was R3 in all.
Conclusion Review of our practice shows that we are adequately managing R3 breast lesions in 100% of cases. It is important that all R3 lesions are biopsied as a significant proportion will be malignant. In our study 14% were malignant.

P7
How important is mammographic image manipulation when examining digital screening cases? Introduction A group of screeners was presented with recent digital screening cases on a mammographic workstation and asked to examine these images either with or without using any image manipulation functions. Their performance and visual search behaviour was measured to determine how using these functions affected their case reading behaviour and performance.
Methods Two sets of 20 cases were matched for abnormality presence and mammographic appearances as closely as possible. Seven radiologists and advanced practitioners then examined these cases on a GE digital mammography workstation whilst their eye movements were recorded using a head-mounted eye tracker. For 20 cases they were not allowed to manipulate the images and for the other 20 they could manipulate the images (that is, pan, zoom and adjust contrast and window level) if they wanted to. Case viewing order was randomised. For each case they rated their confidence in abnormality presence, its location, case density and their screening decisions. Their performance and search behaviour were also compared to those of an experienced radiologist who was very familiar with the case set.
Results The data demonstrated that participants were as able to identify abnormalities without the need of using image enhancement manipulations as they were with them (P > 0.5). However, using these tools increased their rated confidence in their case decisions as well as resulted in overall slower examination times compared to the experienced radiologist. Conclusion Whilst image post-processing manipulations are not necessary for reporting screening cases appropriately, they do affect reporting confidence and mammographic case visual examination. Results Twenty-six percent (216 of 844) of incident round cancers were potentially detectable on the previous screening mammograms (group 1). Of these, 69% were interpreted as subtle/uncertain and 29% as suspicious with the majority in the 'milky way' (55%). Seventy-four percent (628 of 844) were not visible previously (group 2). The most frequent mammographic sign at diagnosis was a spiculated mass (group 1, 56%; group 2, 48%; P = 0.0025). If present previously (group 1), the most likely signs were a mass (57%) (P = 0.001), microcalcification (13%) or an asymmetric density (10%). There was a significant difference in mammographic size between the cancers at diagnosis (mean 17 mm) and on the previous round (mean 10 mm) (P = 0.01). Most tumours were grade 2 at diagnosis (group 1, 46%; group 2, 45%). In group 1 there were significantly more grade 1 tumours (43%) and nearly half the amount of grade 3 tumours (11%) (P = 0.0001).

P8
Conclusion If visible on previous mammograms, incident round cancers are likely to be small, of low grade and appear as a mass, asymmetric density or focus of microcalcification. and clinician of speed, cost, convenience and cosmesis, it is not commonly used in the UK. As fibroadenomata are diagnosed by triple assessment and do not usually need to be excised for clinical reasons, it becomes a cosmetic procedure. We therefore did a survey of patients to determine their satisfaction with the procedure.

Role of large volume 'mammotome' biopsy in the management of screen-detected radial scars S Rajan, K Mankad, A-M Wason, P Carder
Methods We looked at the case records of 198 patients who had had 211 fibroadenomata removed in our unit between 1999 and 2006. We sent these patients a questionnaire about satisfaction and cosmetic outcome.

Results
The mean age of patients was 30.1 years (range 15 to 64). The mean size of fibroadenoma excised was 17.9 mm (range 6.6 to 36.5). 102 questionnaires were returned (51.5%). The mean score (5 point scale) for anxiety surrounding the procedure was 3.2, the mean pain score was 2.5. Only 21.6% could see a visible scar and of these only two said that it bothered them; 16.7% said that they could still feel a lump; and 83.9% of those who had previously had a surgical excision said that they preferred the mammotome excision. Conclusion This is a well tolerated procedure and has advantages for the patient, the surgeon and the health care provider. As it has become a cosmetic procedure, the measure of outcome is patient satisfaction.
We feel that this procedure should be the standard of care.

P11
In All BSUs on the UKBSP annually take part in the PERFORMS scheme as way of self-assessing their film-reading skills. We looked at the performance of all film-readers who had completed the last PERFORMS round (SA08) by BSU size in order to explore any group differences mediated by unit capacity. Methods Each BSU's size was approximated by ranking each unit by number of readers who had completed the last PERFORMS round. Subsequently, these BSUs were allocated into three main groups approximating their unit's size: small = 1 to 4 readers, 30% (n = 157); medium = 5 to 7 readers, 34.5% (n = 181); and large = 8 or more readers), 35% (n = 186). Several performance measures were compared, including 'percentage correct recall' and 'malignancies detected' (measures of sensitivity), 'percentage correct return to screen' (a measure of specificity), and negative and positive predictive value scores. Results Analysis of variance (one-way) did not produce any significant findings (P = not significant) for any of the measures, indicating equivocal performance. Descriptive statistics showed smaller units scored less than 1% below medium/large BSUs for malignancies detected, correct recall and negative predictive value only. In the transition to digital mammography the prior mammograms are in film format. There are difficulties making comparisons between digital current and film prior mammograms due to differences in image appearance and display brightness. This study investigates cancer detection performance in digital mammography with and without film prior mammograms. Methods Two radiologists and two radiography advanced practitioners read a set of 160 (41% malignant) difficult digital mammography cases twice, once with film prior mammograms and once without. Participants noted whether they would recall each case in the NHS breast screening programme.

Results
The number of false negative cases (that is, missed cancers) did not differ between conditions. There was a trend towards a larger number of false positive cases (that is, normal cases that were recalled) when prior mammograms were not available (24% increase, P = 0.058). If this trend continued throughout the NHS Breast Screening Programme, then discarding prior mammograms would correspond to an increase in recall rate from 4.6% to 5.3%, and 12,414 extra women recalled annually for assessment.

Conclusion
Initial results indicate that film prior mammograms may be beneficial to cancer detection performance in digital mammography, and therefore should be displayed. Completion of the study with four extra participants will allow firmer conclusions to be drawn. This study was performed to compare the ipsilateral post-operative mammography findings, frequency of ultrasound and image-guided biopsy post-TM with a group of women who had undergone wide local excision (WLE). Methods Ninety-one women post-TM were compared with 86 women post-WLE. All women had intact breast irradiation and were of similar age (mean age 57 years and 56 years, respectively), had had at least one post-operative mammogram and the same surgeon was present at all operations (DM).

P16
The presence and type of mammographic calcification, and focal and generalised reaction were noted as were ultrasound (US) examination and image guided biopsies. The chi-square test and Fisher's exact test were used for statistical analysis.

Results
The average length of follow-up in the WLE group was longer than in the TM group (5.3 years versus 2.9 years). Rates of postoperative site mammographic calcification and calcification requiring image-guided biopsy were similar in both groups (31 of 86 (36%) versus 31 of 91 (34%) and 2 (2%) versus 2 (2%), respectively. Rates of US and image-guided biopsy were also similar: 26 of 86 (30%) versus 23 of 91 (25%) and 10 of 98 (12%) versus 9 of 91 (10%), respectively. Having corrected for length of follow-up, no statistically significant difference in the frequency of features analysed was observed. Conclusion TM is not associated with greater post-operative mammographic calcification, US examinations or image-guided biopsy.

P20
Negative axillary ultrasound in primary breast cancer: how reassured should we really be? Methods Screening and symptomatic patients were identified from pathology records and information collated from pathology and imaging records.

Results
Of the 155 normal ultrasounds, 45 (29%) were positive at axillary surgery. True and false negative groups were compared in terms of the following: tumour size, pathological type and grade, lymphovascular invasion and oestrogen receptor (ER) status. Breast tumour size was significantly different, with the average size in the true negative group 21 mm and in the false negative group 30 mm (P <0.002).
There was no significant difference in tumour grade or ER status. However, the histological type varied significantly between the groups, with excess lobular carcinomas in the false negative group (6 of 110 versus 6 of 45, P < 0.001). The false negative group was more likely to show lymphovascular invasion in the breast (31% versus 5%, P < 0.001). Conclusion There are significant differences in tumour characteristics between women with true negative and those with false negative axillary ultrasound in terms of size, primary tumour histological type and presence of lymphovascular invasion. In particular, axillary assessment in primary lobular carcinoma may be more difficult and a negative result should be interpreted with caution. Methods We collected imaging and pathological data on 385 women with invasive breast cancer from before and after the introduction of axillary ultrasound. Results Without axillary ultrasound, 136 of 208 (65%) women underwent axillary node sampling (ANS) based on mammographic and clinical findings. The remaining 72 (35%) women had axillary clearance surgery (ANC). Following ANS, 29 (21%) were histologically positive, 27 having subsequent ANC. Of those who had initial ANC, 37 of 72 (51%) were positive. In our study 177 axillary ultrasounds were performed, and 112 were normal. Of these, 32 (29%) had ANC, 14 (44%) being histologically positive. Following abnormal axillary ultrasound, 51 of 65 (78%) underwent ANC with 45 (88%) being positive. There was no significant difference in primary tumour size or mean patient age between the groups before and after the introduction of ultrasound. Statistical analysis shows significant reduction (P < 0.02) in the proportion of ANC in the ultrasound group and, importantly, a reduction in histologically negative ANCs (P < 0.01). In our study, axillary ultrasound had a sensitivity of 54%, a specificity of 83%, a positive predictive value of 71% and a negative predictive value of 71%. Conclusion Introduction of axillary ultrasound was associated with less frequent ANC. Importantly, there has been significant reduction in histologically negative clearances. Our study supports the use of axillary ultrasound in guiding axillary surgical management.

P22
Screen-detected ductal carcinoma in situ -fine needle aspiration versus core biopsy  2008/2009and 88% in 2007/2008versus 63% in 2006/2007-and fewer repeat cores -11% in 2008/2009and 16% in 2007/2008versus 33% in 2006/2007 The continued use of FNA in the investigation of microcalcification will be reviewed in terms of its role in giving a same-day diagnosis in clinics, particularly when VACB is used for first-line diagnosis. Consenting women returned the questionnaire, allowing further data collection, including demographics, menopausal status and hormone replacement therapy use. Data were correlated against breast density measurements to determine the degree of association. Mammograms were assessed on a Hologic™ workstation and breast density calculated using Quantra™. Quantra™ is an automated algorithm for the volumetric assessment of breast tissue composition from digital mammograms. We invited 683 women to participate (those with implants or mastectomy were excluded) and 321 completed returned questionnaires were assessed. The mean age of participants was 59 years (range 49 to 81). Mean density was 19.4% (range 8.5 to 49.0%).

Results
There was a decrease in density with age (Spearman rank correlation coefficient -0.21). Correlation between density and hormone replacement therapy use showed a significant positive result. Quantra™ has shown to be an accurate, reproducible tool for quantifying breast density, demonstrated by its correlation with lifestyle and demographic data. Given its ease of acquisition, this may be the future of breast density quantification in the digital age.

P29
The feasibility of vision-supported computer-based training in digital mammography Y Chen 1 , AG Gale 1 , A Evans 2 1 Loughborough University, Loughborough, UK, 2 Nottingham Breast Institute, Nottingham, UK Breast Cancer Research 2009, 11(Suppl 2):P29 (doi: 10.1186/bcr2399) Introduction Full-field digital mammography (FFDM) screening necessitates training more users to interpret digital images, as well as facilitating computer-based training employing a range of display devices. The feasibility of training naïve observers to examine mammographic images using different forms of vision-supported training on a PC monitor was examined.
Methods A set of recent screening cases were first examined by an experienced radiologist and both his visual search behaviour and verbal commentary recorded. Twenty naive observers were then familiarised with abnormal mammographic appearance, concentrating on masses and calcifications. They were then split into four different training groups (examining mammographic images with: overlay of the radiologist's visual search; playback of the radiologist's commentary; mammographic regions of interest highlighted; or only regions of interest presented) using 20 two-view cases and a control group. Before and after training, each participant was tested on a set of 21 cases and required to identify whether an abnormality was present. Participants' eye-movements were recorded and a 21" LCD monitor was used throughout to view the images. Results Examination of visual search and performance data, pre-and post-training, indicated that only 14% of responses identified the correct features and their locations. Errors were due to search (>60%), detection (<20%) and interpretation (<18%) factors. Approaches that emphasised the region of interest around an abnormality caused observers to fixate these areas for longer periods and produced fewer errors.
Conclusion The introduction of FFDM allows a variety of displays and computer-based approaches to be used for training purposes. Results To date, 60 responses have been received, with 59 completed questionnaires. The BSU size varies considerably, with 3 BSUs inviting less than 10,000 women and 6 inviting over 50,000 annually. The percentage of invited SAOW also varies, from <5% to 25 to 30%. Only one BSU sends the first invitation/reminder in South Asian languages. Sixteen BSUs (27%) record the patient's language and 20 (34%) give leaflets in their language when they attend for mammography. Four BSUs give them normal recall letters in their language. Three BSUs send their assessment recall letter and nine give them biopsy leaflets in their language. Three BSUs send them a normal assessment results letter in their language. Seventy-six percent of the BSU think it would be useful to record a patient's language for improving services. No BSU has separate funding for targeting SAOW for breast screening; only one has a dedicated team to do so. Only 17 (29%) of BSUs have Link Officers/Promotion Officers.