05 Warren-cpp 5/11/04 12:03 PM Page 180
180
ORIGINAL PAPER
An audit of assessment procedures in women who develop breast cancer after a negative result R Warren, P Allgood, G Hunnam, S Godward, S Duffy on behalf of the East Anglian Breast Screening Programme ................................................................................................... J Med Screen 2004;11:180–186
Objectives: A case audit was undertaken to determine the extent to which the early diagnosis of
See end of article for authors’ affiliations
...............
Correspondence to: Ruth Warren, Cambridge Breast Unit, Addenbrooke’s Hospital, Cambridge CB2 2QQ; Email:
[email protected] Accepted for publication 30 June 2004
...............
cancer could be improved by better adherence to screening guidelines, and to estimate the effect that this might have on breast cancer survival. Although affecting only a small proportion of the cancers of the screening programme, this exercise had an educational function for screening radiologists. Setting: The East Anglian breast screening programme, a group of seven centres offering screening to a total population of 2.2 million inhabitants. Women were screened every three years between the ages of 50 and 64. Methods: Adherence to the guidelines of the UK National Breast Screening Programme (as published in 2001) was tested in women assessed between the start of screening on 1 April 1989 and 31 December 1999, in cases where the screen was negative but who were subsequently diagnosed with breast cancer. Results: In this period the programme screened 503,493 women, recalled 25,346 and diagnosed 3689 with cancer. 194 cancers in 193 women were reviewed, comprising those cancers that arose at the site of the lesion previously assessed. 96 women (49.5%) had calcifications, 48 (24.7%) had opacities. 139 of 194 cases were judged to have been inadequately assessed. A recurring theme showed that biopsies not undertaken or with false negative findings led to failure to diagnose lesions which were subsequently shown to be cancer. Microcalcifications and opacities were more likely to have been inadequately assessed than spiculate masses, parenchymal deformities, or asymmetric densities. In the earliest time period (1989–1993), there were a larger proportion of inadequately assessed cases than in the period 1994–1999. Conclusion: Scrupulous adherence to good guidelines will result in a greater proportion of cancers being diagnosed. Failure to perform effective percutaneous biopsy was the usual cause of missed diagnoses. Although an infrequent occurrence this may have an effect on subsequent survival from breast cancer.
INTRODUCTION
A
udit is an important tool for improvement of a clinical service when standards of performance are available against which to compare. Improvement of performance driven by audit has been a tool in use in the UK National Health Service Breast Screening Programme (NHSBSP) since its inception, with great evidence of benefit. An example of its use to improve the national performance of fine needle aspiration cytology is given by Wells and is relevant to the present study.1 The present audit was undertaken by the Quality Assurance team and the radiologists and radiographers of the East Anglian NHSBSP for reflective practice and education with a view to improving screening effectiveness. The purpose was to investigate the diagnostic shortfalls and effects of breast cancer arising in women following a negative assessment at breast screening. Categorising these tumours allows us to ascertain the frequency and characteristics of missed tumours, the reasons why they were missed and what can be done to minimise the numbers of tumours missed in the future. Audit requires a standard against which to benchmark performance and the only available standard was the NHSBSP guideline on radiological assessment, which was published in 2001.2 It should be noted that the cases were assessed some years before this publication, since the last date under consideration was the Journal of Medical Screening
2004
Volume 11
Number 3
end of 1999. By examining the characteristics of the cancers subsequently diagnosed, and the time course of their presentation, it may be possible to predict the benefit foregone by missing the first opportunity of diagnosis. Assessment in this study is defined as diagnostic confirmation of a positive screening mammography.
BACKGROUND Previous studies have shown that many of the cancers that arise during the breast screening interval were in fact visible on previous mammograms when viewed in retrospect; that is, they were false negatives.3–5 Similarly, other studies have shown that this is also true of some cancers that were diagnosed at incident screens.6–8 These studies provide valuable information. Some of the cases (a small subset) in these publications may have been assessed but the studies do not specifically focus on screening assessment. There has only been one study to date that specifically looked at this issue. A useful study from Nottingham examined breast cancers diagnosed in screening-aged women following a negative screening assessment. This study was small (n=28) and limited to one centre9 but it offers some useful recurring themes that are associated with false negative events. Notable among these themes was the frequency of cancers
www.jmedscreen.com
05 Warren-cpp 5/11/04 12:03 PM Page 181
Screening assessment audit in women subsequently developing breast cancer
arising after unbiopsied calcification. It should be noted that these cancers are very few by comparison with the cancers successfully detected by the screening programme, and so the gain to be achieved by the present scrutiny will be small compared with the more usual audit of the screening and detection process. However, the most effective screening will be achieved when every link in the diagnostic chain functions to the best possible standards. When UK breast screening was introduced in 1987 it was recommended that assessment should be carried out by multidisciplinary teams. Guidance was subsequently published for the organisation of assessment clinics and some procedural standards have been included in the guidelines for each of the involved disciplines. However, there were no specific guidelines published on breast screening assessment until 2001.2 The current audit looks at assessment performance before the publication of these guidelines and compares them with the guidelines as they stand today. Clinical practice evolves slowly, and different centres respond to these changes at different rates. This will reflect the effects of education, attendance at meetings, and access to new forms of equipment or new techniques. In running a programme for a large region there will be opinion leaders and those who follow, and it is the duty of the quality assurance leaders to encourage the most effective practice across the entire region of responsibility. The NHSBSP publication forms a standard for the use of individual radiologists and teams and has required an early revision, which is now due, because of changes in practice that have occurred since the 2001 publication.
PATIENTS AND METHODS The audit covered the seven breast screening units of East Anglia, now part of Eastern Region. This sub-region covers a population of 2.2 million people and comprises the following services: Cambridge and Huntingdon, Kings Lynn, Peterborough, East Suffolk, West Suffolk, Norfolk and Norwich, and Great Yarmouth. Data were primarily obtained from the individual units and from the East Anglian Cancer Registry. The screened women were aged 50–64 years and were offered screening every three years. The study ran from the start of screening in the old East Anglian region, 1 April 1989, to a last assessment date of 31 December 1999. The population comprised all women who were given a negative result from breast screening assessment and who later developed breast cancer at the same site as the abnormality that had been assessed. This included women who subsequently presented with an interval cancer, a cancer at a following incident screen or at early recall, or women who were lost to the screening programme either because they were beyond the screening age or because they failed to attend a following breast screening appointment. These women presented through the symptomatic service, and their cancers were in due course recorded at the cancer registry. For this audit the categories are defined below:
Screen detected tumours Incident: Women who were diagnosed with breast cancer at a subsequent routine screening examination after three years. This also included women over the upper screening age who referred themselves at the three year anniversary of their last screening event. Early recalls: Women who were diagnosed at an early recall screen. This was usually one year after the original assessment.
www.jmedscreen.com
181
Symptomatic tumours Interval: Women of screening age who subsequently presented with breast cancer in the interval between screening rounds. 65+: Women who had routinely referred themselves since they reached the upper screening age limit (>65 years until 2003) and subsequently presented symptomatically with breast cancer in the interval between screening rounds; they would be classified as interval cancers if they had been within the screening age at diagnosis. From 2003 screening was offered to women up to age 70 (seven screening rounds), but this is likely to have had little impact on the present study. Lapsed attendees: Women who had not attended at least the last routine screen, either by invitation (for women of screening age) or by referring themselves (women above screening age), before they were diagnosed. Lapsed attendees had more than 3 years (screening round length) between the negative assessment and date of presentation.
The reviewing panel The reviewing panel of the assessment procedures consisted of the then Quality Assurance Director (RW, a radiologist), the QA Radiologist (GH) and the Screening Unit Director, working on a consensus basis with the radiologists and radiographers of the unit under review. No individual reviewer saw all the films. These professionals were supported in the work by a member of the QA reference centre staff (PA). In order to ascertain whether the lesion was the same as the abnormality that had been assessed, the mammograms taken at screening assessment were compared with the diagnostic mammograms. Women whose lesion was in the other breast to that which had been assessed and women whose lesion was at a different site in the same breast were excluded. Women who did not have both screening and diagnostic mammograms to compare were also excluded. All procedures carried out at the original assessment were recorded. These included additional mammography, clinical examination, ultrasound, fine needle aspiration (FNA) and core biopsy (CB). A decision was then made as to whether the assessment procedures had been carried out correctly according to the 2001 guidelines – the publication has good flow charts for the different screening queries, which were useful for comparison. The study ran from 1 April 1989 to a last assessment date of 31 December 1999. Surveillance for subsequent breast cancer diagnosis continued until 31 December 2000, both in order that a full year had elapsed after the last audited case and because the East Anglian Cancer Registry has a time lag between diagnosis and registration of 18 months. Death details were obtained from the Cancer Registry and followed-up until the beginning of December 2003. There is a time lag of approximately two weeks for deaths to be recorded on the cancer registry for the majority of breast cancers. Deaths were also checked at the individual screening units.
Data analysis and statistical methods The primary aims of data analysis were to identify the abnormalities which most frequently gave rise to the original assessments and their relationship to subsequent screening or clinical presentation, the factors associated with assessment being retrospectively judged as inadequate, and to describe the prognosis of the subsequent tumours in relation
Journal of Medical Screening
2004
Volume 11
Number 3
05 Warren-cpp 5/11/04 12:03 PM Page 182
182
Warren, Allgood, Hunnam, et al.
to original mammographic features and subsequent presentation. For the latter, the Nottingham Prognostic Index (NPI),10 a score based on the pathological size, lymph node status and histological grade of the tumour, was used. Statistical analysis consisted mainly of simple tabulations and cross-tabulations. Associations between factors were assessed for significance by chi-squared tests, and where ordinal data permitted, by Cuzick’s non-parametric test for trend.11
RESULTS In the period studied by this audit (1 April 1989 to 31 December 1999), 600,333 women were invited to be screened (plus 33,464 self referrals), 474,280 were actually screened (plus 29,213 self referrals), 23,832 were recalled to assessment12 (plus 1514 from self referrals), and 3270 were diagnosed with cancer before 31 December 2000 (plus 419 from self referrals). These figures take account of the cases arising from the last screening examination and the subsequent complete year for interval cancers to be diagnosed. Deaths reported until 1 December 2003 are included. The screening units of the region met performance targets laid down by the UK NHSBSP13 and the mortality benefit has been predicted for the second round of screening.14 We identified 556 breast tumours in 555 women who had been assessed at breast screening, given a negative outcome, and later developed breast cancer. One woman was diagnosed with bilateral tumours, both of which had been previously assessed. For 130 women either one set or both of their assessment and diagnostic mammograms were missing. In 232 women the tumour was located in the contra-lateral breast or at a different site in the same breast to that which had been previously assessed. Thus the study population was made up of 194 tumours diagnosed in 193 women who had been given a negative outcome at screening assessment for a lesion that later developed into a tumour. Thus of those with all the required information, about 45% (194/[555–130]) had tumours at the anatomical sub site that was the target of the original assessment. Assuming that the same proportion applies to the 130 with missing information, the total expected number of such tumours is approximately 253, 1.1% of the total number of assessments in the programme to date (23,832). There were 3689 cancers detected in the screening programme over the period, suggesting a sensitivity of the assessment process (episode sensitivity)12 of about 94% (3689/[3689+254]). 107 (55.2%) of the tumours were in the left breast, 87 (44.9%) were in the right. Almost 64% (124/194) were located in the upper outer quadrant of the breast. Over half the women (106/194) were assessed at their prevalent screen and most were recalled due to a mammographic abnormality alone (188/194). Four women were
Table 1
recalled because of both a clinical and mammographic abnormality, and two women were recalled because of a clinical abnormality alone; one of these had suspected Paget’s disease of the nipple. This woman was referred to a surgical clinic but no further action was taken at that time. Just over seven years later she presented with Paget’s disease of the nipple as a lapsed attendee. Women were most commonly assessed for microcalcifications (96/194) followed by opacities (48/194), parenchymal deformities (31/194), asymmetric densities (28/194), and spiculated masses (27/194; Table 1). 36 of the lesions had multiple features, such as microcalcification plus opacity. At assessment, clinical examination was performed in over 80% of women (159/194), almost three quarters (144/194) had additional mammographic views taken, and 127/194 were examined by ultrasound. However, in only 30/194 was FNA performed and in 14/194 a core biopsy. The mean (median) age of the women at assessment was 57.7 (57.5) years. The majority (121/194) were returned to routine screening, but a substantial minority (66/194) were recalled early. Seven women had an open biopsy; six of them with benign result (i.e. lesional tissue) and one had normal tissue excised. The mean (median) age at diagnosis was 60.5 (60.2) years, giving a mean (median) time between negative assessment and diagnosis of breast cancer of 2.8 (2.6) years. The range of time between assessment and diagnosis was very wide (0.21–10.00 years). There were eight women who were diagnosed with breast cancer less than 6 months after their initial assessment; five of these presented as symptomatic cancers, and two referred themselves back to the screening unit early due to breast pain. In total, more than half the women (113/193) were subsequently diagnosed with breast cancer by the screening programme, 77 at a routine screen and 36 at an early recall screen. Of 81/194 (41.8%) women who presented symptomatically, 55 presented as interval cancers; 13 were women over 65 years who routinely self-referred and also presented with an interval cancer, and 13 were lapsed attendees. Screening-aged women with interval cancers had a mean (median) time to presentation of 2.1 (1.9) years; women over 65 years who presented with an interval cancer had a mean (median) time to presentation of 4.1 (3.2) years; lapsed attendees had a mean (median) time to presentation of 6.7 (6.9) years. Table 2 compares the time from the original assessment to diagnosis according to the mammographic abnormality assessed for screen-detected and symptomatic tumours. Although there was slightly more variation in time to diagnosis according to mammographic abnormality for symptomatic tumours, there was little difference between the average times to diagnosis. Our panel of assessors judged procedures to be inadequate
Original assessment procedures according to mammographic abnormality Procedures carried out at original assessment n (%)
Mammographic abnormality n=194 (100%)*
Additional mammograms
Clinical examination
Ultrasound
FNA biopsy
CB biopsy
Open biopsy
Microcalcification 96 (49.5) Opacity 48 (24.7) Parenchymal deformity 31 (16.0) Asymmetrical density 28 (14.4) Spiculated mass 27 (13.9)
81 31 23 17 21
73 40 29 24 25
41 43 28 19 26
19 (20.0) 7 (14.9) 5 (16.1) 4 (14.3) 2 (7.4)
9 2 1 2 1
4 (4.2) 0 0 1 (3.6) 2 (7.4)
(84.4) (64.6) (74.2) (60.7) (77.8)
(76.8) (85.1) (93.6) (85.7) (92.6)
(43.2) (91.5) (90.3) (67.9) (96.3)
(9.5) (4.3) (3.2) (7.1) (3.7)
* Totals add up to more than study population because the mammographic abnormalities are not mutually exclusive.
Journal of Medical Screening
2004
Volume 11
Number 3
www.jmedscreen.com
05 Warren-cpp 5/11/04 12:03 PM Page 183
Screening assessment audit in women subsequently developing breast cancer
Table 2
assessment
183
Time to diagnosis and mode of detection according to mammographic abnormality at original Mode of detection and mean (median) time (years) from assessment to diagnosis
Mammographic abnormality n=194 (100%)
Screen n=113 (100%)
Time to diagnosis mean (median) years
Symptomatic n=81 (100%)
Time to diagnosis mean (median) years
Microcalcification 96 (49.5) Opacity 48 (24.7) Parenchymal deformity 31 (16.0) Asymmetrical density 28 (14.4) Spiculated mass 27 (13.9) Total n* and mean time to diagnosis
59 23 18 15 17 132*
2.5 (2.4) 2.5 (2.9) 2.7 (3.0 3.5 (2.9) 2.6 (3.0) 28 (2.6)
37 25 13 13 10 98*
3.5 3.4 2.7 3.1 2.0 2.9
(3.1) (2.4) (2.2) (2.4) (1.5) (3.1)
*Totals add up to more than study population because the mammographic abnormalities are not mutually exclusive. Time to diagnosis given above is taken from figures in this table (n>100%). Mean (median) for all screen detected=2.6 (2.9) years and for all symptomatic cases=3.2 (2.3).
in 139/194 (71.7%) according to the guidelines.2 This varied significantly between breast screening units, ranging from 37.0%–91.7% (Pearson χ2[6]=22.73, p=0.001). The reasons given as to why these assessment procedures were judged inappropriate in light of the 2001 guidelines are shown in Table 3a according to principal mammographic abnormality. In all mammographic abnormalities, the most common cause of false negative findings related to biopsy procedures – not done, yielding inadequate tissue or a benign finding accepted as a true negative when it was in fact a targeting error. Of those assessments judged to be adequate, percutaneous biopsy (CB or FNA) was performed in 13/55 (23.6%) of cases, and of those judged to be inadequate, it was performed in 27/139 (19.4%) of cases. The difference is not significant (p=0.5). Of the 112/139 inadequate assessments, 98/139 were due to failure of the CB or FNA. This comprises 71/112 (63.4%) cases where no percutaneous biopsy was undertaken and 27/139 where it was undertaken but gave a spurious negative result due to inadequate tissue or targeting error. The association between lesions judged to be correctly assessed and the mammographic abnormality at the original
screening was significant for microcalcifications and asymmetric density (both χ2 [1] p
