M E D I C A L A U D I T O F M A M M O G R A P H Y S C R E E N I N G E X A M I N A T I O N S
A l i T ü r k , M . D . * / E r k in A n b a l , M . D . *
* D e p a r t m e n t o f R a d io lo g y , S c h o o l o f M e d ic in e , M a r m a r a U n iv e rs ity , Is ta n b u l, T u rk e y .
A B S T R A C T
O b je c t iv e : To audit our mammography
screening practice and evaluate the usefulness of BI-RADS (Breast Imaging Reporting and Data System) lexicon in the classification of mammography findings.
M ethod: In this study we performed a medical
audit of our mammography screening practice between 1 st December 1999- 1st December
2 0 0 0 .
R e s u lt s : The number of baseline screening
examinations were 2007 (67.2%) while the remaining 1041 (32.8%) were performed on women already screened at least once. Recall rate is 3.5%, PPV1 9.9%, PPV2 20%, and PPV3 25%. Overall cancer detection rate is 0.69% for 21 diagnosed cancers. Minimal cancer ratio is 14.3% and axillary lymph node involvement ratio is 33.3%.
C o n c lu s io n : Standardisation of mammographie
reporting and auditing the data are important in assessing an optimal screening facility. Every mammography screening center should audit its data to improve the outcome and to monitor the effectiveness.
K e y W o r d s : Medical audit, BI-RADS categorisation, Mammography
IN T R O D U C T IO N
Quality control for a mammography unit, in addition to assessment of equipment, imaging and image processing, includes the evaluation of the accuracy of image interpretation.
At the individual practice level, the record keeping of mammography examinations and the results of follow up procedures are necessary to evaluate and improve the validity and reliability of mammography practice. This will ensure appropriate medical follow up and therapy in women with suspicious mammography findings (1-4).
Mammography auditing is the collection of data that allows the routine and periodic evaluation of screening mammography performance and the record keeping and information retrieval system required to support this practice (5). The physicians have to find a way to control and improve their diagnostic skills in countries where there are no state regulations to control mammography units according to international standards. The medical audit is the only way to
evaluate and improve the subjective
interpretation of the mammographic images because it includes not only technical but also interpretive performance of the mammography practice. It demonstrates one’s success and failure in detection of breast cancer which is the ultimate indicator of mammography performance (6,7).
( A c c e p t e d 7 M a r c h , 2 0 0 2 ) M a r m a r a M e d ic a l J o u r n a l 2 0 0 2 ; 1 5 ( 1 ) : 2 7 - 3 2 Correspondance to: Ali Türk, M.D, ■ Department of Radiology, Marmara University Hospital, 81190 Altunizade, Istanbul, Turkey.
In this study we performed a medical audit of our screening mammography practice between 1st December 1999- 1st December 2000 and evaluated the usefulness of BI-RADS (Breast Imaging Reporting and Data System) lexicon in the classification of mammography findings.
M E T H O D S
For auditing, a variety of data was collected from both the mammography and any breast biopsy reports. The mammography report consists of
demographic information, results and
recommendations The reports must be
constructed to allow easy collection of useful audit data . Results and recommendations must be categorized using standardized codes such as BI-RADS lexicon (8). Integration of the mammography and pathology results are mandatory and Important items of audit study. Breast biopsy results are coded, necessary data are collected using standard pathology nomenclature (4,7,9). After mammography and biopsy reports are appropriately coded , using computer software programs or manually necessary data are collected.
Raw data includes demografic information, mammograpy interpretation results, any subsequent recommendations, and pathology findings collected from the mammography and pathology reports . Essential raw data include audit period dates, number of mammographic examinations, number of recalls, number of recommended biopsy, biopsy results and tumor staging. Derived data refer to calculated measures of various mammographic and pathologic parameters based on the collected raw data. Essential raw data includes number of true positives, number of false-positives, positive predictivity values, cancer detection rate, percentage of minimal cancers, percentage of node positive cancers and recall rate (10). Three different PPV were calculated.
1-PPV1: (Abnormal findings of screening) The percentage of all screening examinations with abnormal findings that results in cancer diagnosis (1,9,11,12).
PPV1= True posItlves/Number of screening examinations with abnormal findings.
2- PPV2: (Biopsy recommended): The
percentage of all biopsy recommended cases after abnormal findings on mammography examinations (13,14).
PPV2 = True positives/number of biopsy recommended cases
3- PPV3: ( Biopsy performed): Percentage of all performed which resulted In the diagnosis of cancer (1,11,12,14).
Cancer detection rate, prevalent and incident cancer rates were calculated. Cancer detection rate is the number of cancers detected per 1000 cases examined by mammography (1,7,9,10). Prevalent cancer is the cancer detected in first screening examinations and incident cancer is the cancer detected in follow up examinations (1,
10,12,13).
The percentage of screened women who are recommended for further imaging evaluation such as coned compression and magnification views were assessed as recall rate (1,10,12,13). Breast carcinomas measuring 1 cm or less in greatest dimension have more favorable diagnosis than larger ones (15,16). Therefore stage of the detected disease is an important indicator of quality of mammography practices (17) .
All mammographically detected lesions have been classified in 6 categories as defined in BI RADS lexicon of American College of Radiology. (18)
RESULTS
In our mammography department, we performed 3048 screening mammography examinations between 1st December 1999-1 st December 2000. The Number of baseline screening examinations was 2007 (67.2%) while the remaining 1041 (32.8%) were performed on women already screened at least once ( Table I). The number of women 50 years or older was 1256 (41.2%). The remaining 1789 (58.8%) women were under 50 years old (Table II). According to BI-RADS assessment categories, 1660 women (54.4%) were placed in category 1,
M e d i c a l a u d i t o f m a m m o g r a p h y s c r e e n i n g e x a m i n a t i o n s
Table I: Number of Screenees
Women screened once 2007 (67.2%) Women screened at least once 1041 (32.8%)
Total 3048
Table II: Age Distribution of Screenees
l< 40 74 ( 2.4%)
40-50 1182 (38.8%)
50-70 1690 (55.4%)
>70 99 (3.4%)
925 women (30%) in category 2, 359 women (11%) in category 3, 94 women (3%) in category 4 and 11 women (0.3%) in category 5 (Table 111). Ultrasonography was performed In 1157 (50.4%) women.
In category 1 and 2, we performed 37 coned compression, 9 magnification, 4 coned compression and magnification views, summing up in total 50 (%1) additional mammographie examinations. In category 3, we performed 39 coned compression, 1 magnification, 2 coned compression and magnification views. In category 3, the additional mammographie examination rate is 11 %. In category 4, of the 15 (16%) women who were imagined with additional
mammographie views 7 were coned
compression, 5 were magnification and 3 were
both coned compression and magnification views (Table IV). The recall rate is 3.5%.
In 25 women tru-cut biopsy, in 17 women mammotome vacuum assisted biopsy ,in 5 women fine needle aspiration biopsy (FNAB), in 23 women surgical biopsy were performed (Table V).
The ratios are summarized in table VI.
Overall cancer detection rate is 0.69% for 21 diagnosed cancers.
The cancer detection rate for women with prevalent cancer is 6.5 per 1000, that for incident cancer is 3 per 1000.
Three of 21 cancers (14.3%) measured less than 10 mm and one of these were in situ. 27.8 % were T1, 66.7% were T2, and 5.6% were T3 according to tumor size (Table VII). Axillary lymph node involvement ratio is 33.3%.
Cancers were staged according to the current criteria of the American Joint Commitee on cancer. 20% were stage 1, 66% stage 2, 7% stage 3, and 7% stage 0. In pathologic examination 20% were Gradel, 66% were Grade2, 7% Grade 3, and 7% carcinoma in-situ.
Table III: BI-RADS Categorization
Category 1 1660 (54.5%) Negative Category 2 924 (30.3%) Benign lesions
Category 3 359 (11.8%) Probably benign and for which 6 month follow up was recommended Category 4 94 (3%) Suspicious or indeterminate lesions
Category 5 11 (0.4%) Highly suggestive of malignancy
Category 6 Additional imaging evaluation recommended
Table IV: Additional Imaging Evaluation
Coned Compression Magnification Compression+ Magnification Category 1+2 37 9 4 50(1.9%) Category 3 39 1 2 42 (11.6%) Category 4 7 5 3 15 (16%)
Table V: Biopsy Types
Tru-cut Biopsy 25
Vacuum Aspiration Biopsy 17 Fine Needle Aspiration Biopsy 5
Surgical Biopsy 23
Total 70
Table VI: Ratios
Numbers of cancer detected 21 Screening cancer detection rate 0.69% Prevalent cancer detection rate 0.65% Incident cancer detection rate 0.03%
PPV1 9.9%
PPV2 20%
PPV3 25%
Table VII: Tumor Size
Tis 1
T1 (2 cm or < 2 cm) 5 (27.8%) T2 (> 2cm, < 5 cm) 12 (66.7%)
T3 (> 5 cm) 1
Minimal cancer ratio 14.3% Lymph node involvement ratio 33.3%
D IS C U S S IO N
The audit as a teaching tool has many significant functions. Audit statistics providing comparisons
of performance between different
mammographic practices determines existant problems and indicates the corrective actions to be taken to improve future outcome (4,11,19,20). If’ the audit outcome is within expected range, providing positive reinforcement on individual performance leads to confidence in the accuracy of the interpretations. If the outcome is below expected values, it will indicate a need for additional education through continuing medical education to correct the problem. By increasing the confidence in the accuracy of interpretation, audit increases compliance of both referring physicians and patients (1,10).
Due to the lack of a national mammographic registry system in Turkey we could not determine the exact numbers of false negatives and the ratio of interval cancers correctly. In spite of this limitation , in known cases, determining their causes allows technical and interpretative shortcomings to be corrected, improving future performance.
Review and correlation of pathology reports with
their corresponding mammographic
examinations in both benign and malign cases is a very valuable teaching tool enhancing the learning process of both pathologist and radiologist. We could not track all the women with follow up or biopsy recommendations for abnormal mammographic findings due to the lack of a national mammography registry system. We tried to get malign or benign pathology reports from referring clinicians or patients themselves but could not obtain all of them.
Because no official breast carcinoma screening program is present in Turkey and women’s
compliance varies greatly in routine
mammographic practices, we could not get reliable data for sensitivity, specifity and negative predictive value.
Our PPV1 is 9.9% (Recommended range 5-10%) and PPV3 is 25% (20-40%). Both are in accordance with recommended ranges. But PPV2 is 20% (25-40%) and lower than the recommended threshold value. All three definitions of PPV can be used for practices in both screening and diagnostic examinations. But in screening mammography, PPV1 should be used to evaluate the data (10). Our mammography practice also performs mostly screening examinations and our PPV1 value is within the recommended range. A high PPV is desirable to decrease morbidity but not by the practise of underreading. PPV2 value is lower than the recommended which shows low true positive rate when compared to biopsy recommendation. We reached 70% of biopsy recommended cases, and we do not know the results of the rest . We believe that this has caused our low rate of PPV2 other than higher biopsy recommendation.
Our cancer detection rate is 0.69% (0.02-1%) and it is in accordance with recommended values. However when comparing local to international data, characteristics of screened population must be considered, because of variations.
Our recall rate for US was high (50%). We performed additional US examination on all women who had heteregenous or dense breast parenchyma. This high rate of additional US examination did not increase the rate of detected
malignancies. Now, we performed US
examinations only for patients who have equivocal on mammography findings.
In many studies it has been shown that , breast carcinoma tumours measuring 1 cm or less in size has a more favorable prognosis than larger ones. Therefore the stage of the detected disease is a quality indicator of breast screening programs (16,17,21). We believe the high number of advanced cancers seen during our audit period were due to the fact that poor socioeconomic conditions of the women lead to a very low attendance in routine mammographic examinations and they tend to get medical
M e d i c a l a u d i t o f m a m m o g r a p h y s c r e e n i n g e x a m i n a t i o n s
treatment when they are symptomatic or a mass is palpable. Though the radiologist’s interpretive ability seems to be the primary factor in detecting small cancers, future audit studies are needed to show the expected increase in minimal cancer detection rate.
A standardised reporting system is necessary for audit studies to be effective. In order to develop a national mammography registry to improve the interpretive quality and improve follow up of patients with mammographie abnormalities, all the mammographie reports must be uniform.
Mammography terminology must be
standardised for assessment of abnormal findings and recommendations. The absence of uniformity results in ambigious reports so that patients and referring clinicians will not have a clear understanding regarding mammography results and recommended action (further imaging, biopsy or follow up). Also, absence of uniform results make the audit studies very difficult. Ambigious reports make it difficult to determine which films were interpreted as positive and which negative.
In our department, we have been using BI-RADS lexicon for two years and we observe that various BI-RADS assessment categories appear to be useful predictors of the malignancy of a lesion. We placed 924 cases in category 2 and it was 100% predictive of a benign lesion while the 12 cases were placed in category 5 with a PPV of 100%.
To our knowledge this is the first systematic audit study using BI-RADS lexicon in Turkey.
We believe that standardisation of
mammographie reporting and auditing the data is important in assessing an optimal screening facility. Every mammography screening center should audit its data to improve the outcome and to monitor its effectiveness.
REFERENCES
1. Sickles EA. Q u ality assurance : how to a u dit y o u r own m a m m o g ra p h y practice. Radiol Clin
n orth Am 19 9 2 ; 3 0 :2 6 5 -2 7 5 .
2. N o n tic c io lo DL, Sickles EA. C om pu terized fo llo w -u p o f a b n o rm a litie s d e te c te d a t
m a m m o g ra p h y s creen in g . AJR 1 9 9 0 ; 1 5 5 :7 5 1 -7 5 3
3. flaug FJ, Tocino IM, Clayton FD. A utom ated m a n a g e m e n t o f screen in g an d diagnostic m am m ography. Radiology 1 9 8 7 ; 1 6 4 :7 4 7 - 7 5 2
4. Sickles EA, O m insky Sfl, Sollitto RA. M edical audit o f a rapid-throughput m am m ography screening practice: m ethotology and results o f 2 7 1 1 4 exam inations. Radiology 1 9 9 0 ; 175: 3 2 3 -3 2 7
5. Brown M, flo u n E. Q uality assurance audits o f c o m m u n ity s c re e n in g m a m m o g ra p h y practices. AJR 1 9 9 4 ; 16 3 :8 2 5 -8 2 9
6. Murphy WA Jr, D estouet JM, Monsees BS. P ro fessio n al q u a lity a s s u ran ce fo r m a m m o g ra p h y s c re e n in g program s. Radiology 1 9 9 0 ; 1 7 5 :3 1 9 -3 2 3
7. Spring DB, Rim brell-W ilcot R. Evaluating the success o f m am m ography at the local level: how to conduct an a u dit o f y o u r practice. Radiol Clin north Am 1 9 8 7 ; 2 5 :9 8 3 -9 9 2 8. R opans DB, D O rsi CJ. B reast im aging
reporting a n d data system. 2 nd ed. Reston, VA: Am erican College o f Radiology, 1 9 9 3 ;2 5 - 5 0 .
9. R obertson CL. A p riv a te b reast im aging practice: m ed ical au d it o f 2 5 7 8 8 screening and 1 0 7 7 diagnostic exams. Radiology 1 993; 3 0 :2 6 5 -2 7 5
10. L in v e r M, O such J, B re n n e r J. The m a m m o g ra p h y 'a u d it: a p rim e r fo r the m am m ography quality standards act (MQSA). AJR 1 9 9 5 ; ¡6 5 :1 9 -2 5
1 1. Moseson D. A udit o f m am m ography in a c o m m u n ity settin g. Am J Surg 1 9 9 2 ;
1 6 3 :5 4 4 -5 4 6
12. B urhenne LJW, I ¡¡slop TG, Burhenne MJ. The British C olum bia m am m o graph y screening program : evaluation o f the first 15 months. AJR 1 9 9 2 ; 1 5 8 :4 5 -4 9
13. Bird RE. Low-cost screening m am m ography : report on finances an d review o f 2 1 7 1 6 cases. Radiology 1 9 8 9 ; 171: 8 7 -9 0
14. L in v e r MIT, P ater SB, R osenberg RD. Im p ro v e m e n t in m am m o graph y interpretation skills in a co m m u n ity radiology practice after d edicated teacning courses: 2-year m edical a u d it o f 3 8 6 3 3 cases. R adiology 1 9 9 2 ;
1 8 4 :3 9 -4 3
15. Tabar L, Fegenberg G. Update o f the Swedish tw o-cou n ty p ro g ram o f m a m m o g ra p h ic screening fo r breast cancer. Radiol Clin ITorth Am 1 9 9 2 ; 3 0 :1 8 7 -2 1 0
16. K inne DW. Staging a n d follow -up o f b reast cancer patients. C ancer 1 9 9 1 ; 6 7 :1 1 9 6 -1 1 9 8 17. S chim idt T, tla rtw a g n e r li. M edical au dit a fte r
2 6 7 1 1 breast im aging studies. C ancer 1 9 9 8 ; 8 3 :2 5 1 6 -2 5 2 0 .
18. R opans DB, D O r s i CJ. B reast im agin g reporting an d data system (BI-RADS). 2 nd ed. Reston, VA: A m e ric a n C o lleg e o f Radiology, 1 9 9 5 :2 7 -3 8 .
19. Bird RE, Wallace TW. Analysis o f cancers misses a t screening m am m ography. Radiology
1 9 9 2 ; 1 8 4 :6 1 3 -6 1 7
2 0 . B u rh e n n e tIJ , B u rh e n n e LW, G o lb e rg E. In te rv a l b re a s t c a n c e rs in th e s c re e n in g m am m o graph y program s o f British C olom bia: analysis a n d c la s s ific a tio n . AJR 1 9 9 4 ;
1 6 2 :1 0 7 2 -1 0 7 5
