| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Breast Imaging |
1 From the Department of Radiology, University of New Mexico Health Sciences Center, MSC10 5530, 1 University of New Mexico, Albuquerque, NM 87131 (R.D.R.); Department of Radiology, University of North Carolina, Chapel Hill, NC (B.C.Y.); Group Health Center for Health Studies, Seattle, Wash (L.A.A., D.S.M.B., W.E.B.) and Cancer Research and Biostatistics (W.E.B.), Seattle, Wash; Department of Radiology, University of California San Francisco School of Medicine, San Francisco, Calif (E.A.S.); Department of Radiology, University of Washington Medical Center, Seattle, Wash (C.D.L.); Office of Health Promotion Research (B.M.G., D.L.W.), Department of Pathology (D.L.W.), and Vermont Cancer Center (B.M.G., D.L.W.), University of Vermont, Burlington, Vt; Department of Community and Family Medicine, Dartmouth Medical School, Hanover, NH (P.A.C.); General Internal Medicine Section, Departments of Veterans Affairs and Epidemiology and Biostatistics, University of California, San Francisco, Calif (K.K.); and Applied Research Program, Division of Cancer Control and Population Sciences, National Cancer Institute, Bethesda, Md (R.B.). Received September 7, 2005; revision requested November 10; revision received January 7, 2006; final version accepted February 1. Supported by cooperative grants U0169976 (R.D.R.), U01CA70040 (B.C.Y.), U01CA63740 (E.A.S., K.K.), U01CA70013 (B.M.G., D.L.W.), U01CA86076 (L.A.A., W.E.B.), U01CA63731 (D.S.M.B.), and U01CA86082-01 (P.A.C.) from the National Cancer Institute. Address correspondence to R.D.R. (e-mail: rrosenb{at}unm.edu).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONADVANCES IN KNOWLEDGEReferences |
|---|
Materials and Methods: Institutional review board approval was obtained, and study was HIPAA compliant. Informed consent was or was not obtained according to institutional review board guidelines. Data from 188 mammographic facilities and 807 radiologists obtained between 1996 and 2002 were analyzed from six registries from the Breast Cancer Surveillance Consortium (BCSC). Contributed data included demographic information, clinical findings, mammographic interpretation, and biopsy results. Measurements calculated were positive predictive values (PPVs) from screening mammography (PPV1), biopsy recommendation (PPV2), biopsy performed (PPV3), recall rate, cancer detection rate, mean cancer size, and cancer stage. Radiologist performance data are presented as 50th (median), 10th, 25th, 75th, and 90th percentiles and as graphic presentations by using smoothed curves.
Results: There were 2 580 151 screening mammographic studies from 1 117 390 women (age range, <30 to 
80 years). The respective means and ranges of performance outcomes for the middle 50% of radiologists were as follows: recall rate, 9.8% and 6.4%–13.3%; PPV1, 4.8% and 3.4%–6.2%; and PPV2, 24.6% and 18.8%–32.0%. Mean cancer detection rate was 4.7 per 1000, and the mean size of invasive cancers was 13 mm. The range of performance outcomes for the middle 80% of radiologists also was presented.
Conclusion: Community screening mammographic performance measurements of cancer outcomes for the majority of radiologists in the BCSC surpass performance recommendations. Recall rate for almost half of radiologists, however, is higher than the recommended rate.
Supplemental material: radiology.rsnajnls.org/cgi/content/full/241/1/55/DC1
© RSNA, 2006
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONADVANCES IN KNOWLEDGEReferences |
|---|
A medical audit is a compilation of specific important patient outcomes over a defined period of at least a year. This allows a radiologist and a facility to recognize areas of strength, as well as those areas that may need improvement. The medical audit is recognized as one of the best quality assurance tools (2,3). There are limited data available, however, with which the practicing radiologist and facility may compare their results. There is a lack of generalizable literature concerning the actual performance of radiologists in the United States and, thus, a limited knowledge of optimal performance targets that are achievable by general radiologists. The opinion of experienced radiologists (4, p 83) and guideline targets for performance for some parameters have been set in some countries (5, pp 4–5;6, pp 147–148). There are problems, however, with using these data for U.S. radiologists. These targets have not been quantified within the U.S. health care environment because of a lack of appropriate population-based screening data. Therefore, the value of existing guidelines is limited.
The Breast Cancer Surveillance Consortium (BCSC) is a National Cancer Institute–funded research initiative of seven population-based research sites with a Statistical Coordinating Center that collects and analyzes mammographic and pathologic data in defined populations (7). The BCSC has published data on its methods (8), confidentiality issues (9), and overall community performance (10–13). These prior results were based on a population of patients that has characteristics that are similar to the national demographic characteristics in terms of age, ethnicity, and urban or rural residence (11). For comparison purposes, use of BCSC data presents two barriers for the average radiologist: (a) The methods used in prior publications cannot be applied by the average community radiologist. (b) BCSC results represent averages rather than the distribution of the range of performance outcomes. The key methodologic limitation is that most community radiology groups do not have the ability to link their mammographic data to regional cancer registries. In addition, the large variations in measurements of performance of mammographers that have been extensively documented (14–19) underscore the importance of understanding where individual performance lies within the distribution of performance of other radiologists.
The American College of Radiology has created medical audit methods intended for use by community radiologists. The 4th edition of the American College of Radiology Breast Imaging Reporting and Data System (BI-RADS) manual (20) provides standardized terminology and assessments used in breast imaging for mammography, ultrasonography, and magnetic resonance imaging. In addition, the BI-RADS manual contains instructions on the use of these assessments to compute outcome measurements from mammographic data that are possible for many community practices. The measurements proposed by the American College of Radiology are more extensive than the minimal measurements required by the Mammography Quality Standards Act for mammographic accreditation. The purpose of our study was to retrospectively evaluate the range of individual radiologist performance outcomes in an audit of screening mammography by using a representative group of U.S. radiologists to allow development of performance benchmarks for screening mammography.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONADVANCES IN KNOWLEDGEReferences |
|---|
Data Sources
Data were collected from six BCSC registries: Carolina Mammography Registry (Chapel Hill, NC), Group Health Cooperative (Seattle, Wash), New Hampshire Mammography Network (Lebanon, NH), New Mexico Mammography Project (Albuquerque, NM), Vermont Breast Cancer Surveillance System (Burlington, Vt), and San Francisco Mammography Registry (San Francisco, Calif). To determine cancer outcomes, each registry links its data to a state tumor registry or to a Surveillance Epidemiology and End Result, also known as SEER, program. Six of these registries also collect some benign pathologic results (7).
Each registry and the Statistical Coordinating Center of the BCSC have received a Federal Certificate of Confidentiality and approval from each institutional review board for the protection of human subjects to collect and send data to the Statistical Coordinating Center and to conduct research with these data. Three of seven sites were granted a waiver of informed consent. At three of the other sites, women had the option to exclude their data from research. At one site, the patient's signature was required to allow inclusion of data for research. Our study was Health Insurance Portability and Accountability Act compliant. All registries have strict procedures for deidentification of patient information and protection of confidentiality (9). Linkage procedures follow protocols specifically designed to preserve patient confidentiality.
Data Collected
Approximately 188 mammographic facilities contributed to the pooled data. This number of facilities represents about 2% of the approximately 10 000 Food and Drug Association–certified mammographic facilities in the United States in 2000. We compared the demographic makeup of the population living in the catchment areas of the six BCSC registries included in our study to that of the entire U.S. population by using 2000 census data. To describe the BCSC population, we (L.A.A., W.E.B.) included census data from all counties in which there was a participating mammographic facility.
Study Group
The study included women who had undergone at least one screening mammographic examination during the years 1996–2002. Screening mammographic examinations performed after December 2002 were excluded to ensure that there was at least 12 months following the screening examination during which cancer could be diagnosed and there was adequate time for cancer reporting. A screening mammographic examination was defined as one characterized by the interpreting radiologist as having an indication of screening.
The pooled data contain screening mammographic interpretations determined by 807 identified radiologists. A radiologist identifier was not available from some facilities but was present for 84.0% (2 166 970 of 2 580 151) of the studies in this report. Some radiologists contributed data from multiple facilities. Many radiologists also interpreted some mammograms at facilities outside of the consortium, and therefore only a subset of their interpretations would have been captured. This inclusion of only a subset of their interpretations also occurs because radiologists move between facilities or serve as temporary radiologists in a facility.
Mammographic Data Collection Procedures and Definitions
Across all BCSC registries, patients undergoing mammography complete a questionnaire at each imaging visit that requests medical history and demographic data, including date of most recent mammographic examination, family history of breast cancer, previous breast biopsy, personal history of breast cancer, and description of recent breast symptoms. Women were considered to have a family history of breast cancer if they reported having at least one female first-degree relative (mother, sister, or daughter) with breast cancer. Women were considered to have a personal history of breast cancer if they had self-reported previous breast cancer or had evidence of previous breast cancer in the cancer registry or pathology database. Each woman was considered to have a previous mammographic examination if she had a self-reported prior mammographic examination or there was indication of information about a prior mammographic examination in the BCSC database.
Generally, screening mammography is performed for women without breast symptoms, but some women with symptoms are included in all screening populations (10,21,22). In this analysis, we used mammograms identified as screening mammograms by the interpreting radiologist independent of whether or not symptoms were present at the time of the examination. We included mammograms that are variably considered screening mammograms that had been obtained because of other special cases, and these mammograms included those obtained in patients with breast implants and in patients with prior breast cancer if the mammograms were designated as from a screening study.
The mammographic registry also captures data about image interpretation, including management (imaging, biopsy, and clinical evaluations) recommendations and the BI-RADS assessment categories assigned by the interpreting radiologist for each mammographic examination (9,20). A separate assessment often is recorded for each breast. For the purpose of this study, we created an overall assessment for the entire examination by using the more serious abnormal BI-RADS assessment category according to the following hierarchy: negative (category 1), benign (category 2), probably benign (category 3), needs additional evaluation (category 0), suspicious (category 4), and highly suggestive of malignancy (category 5). A positive result was defined as one classified with BI-RADS assessment categories 0, 4, or 5, and a negative result was defined as one classified with BI-RADS assessment categories 1, 2, or 3. Results in a previously published investigation (23) showed only very small nonsignificant differences between woman-specific and breast-specific outcome data, and these results indicated that woman-specific data are sufficiently accurate measurements of interpretive performance.
A report about screening mammography from the BCSC (24) indicated that 10%–15% of examinations with positive (abnormal) results (BI-RADS categories 0, 4, or 5) were discordant between the BI-RADS assessment category assigned and subsequent management recommendations provided by the interpreting radiologist compared with the recommendations that the BI-RADS assessment category should inherently suggest. These nonstandard approaches tend to undercount a sizable proportion of positive mammograms (25). This undercount is caused by the common use of BI-RADS category 3 (probably benign finding) with additional imaging recommended instead of BI-RADS 0 (needs additional imaging) (24). Thus, a negative assessment (BI-RADS category 3) is used instead of a positive assessment (BI-RADS category 0). Because of the differences in how practicing radiologists implement BI-RADS, an important percentage of women with similarly abnormal mammographic findings may appear to be classified in different categories of assessment. To create comparable performance benchmarks across facilities, we made two modifications to the collected data: (a) If additional imaging was performed at the time of the screening, the screening mammogram was considered positive. (b) If a recommendation for immediate work-up was given along with an assessment that indicated a probably benign assessment (BI-RADS category 3), then the assessment was considered positive and classified as BI-RADS category 0. According to BI-RADS audit rules, any mammogram with a BI-RADS category 6 assessment (known breast cancer) was excluded from the analysis.
Patients undergoing mammography were considered to have breast cancer if a state tumor registry, Surveillance Epidemiology and End Result program registry, or pathology database indicated the diagnosis of invasive carcinoma or ductal carcinoma in situ within 12 months after a screening mammographic examination.
Outcome Measurements and Statistical Analysis
A true-positive mammogram was defined as a screening mammographic examination with a positive interpretation that was followed by the diagnosis of invasive breast cancer or ductal carcinoma in situ within 12 months. Cancer detection rate was defined as the number of cancers following a positive mammogram divided by the total number of screening mammographic examinations. Conversely, a false-positive mammogram was defined as a screening mammographic examination with a positive interpretation and no breast cancer diagnosed within the next 12 months.
We calculated the positive predictive value (PPV) by dividing the number of true-positive examinations by the sum of true-positive and false-positive examinations. Three separate PPV calculations were performed by using BI-RADS methods: PPV1 (probability of cancer following a positive mammographic interpretation), PPV2 (probability of cancer following a BI-RADS assessment of 4 or 5), and PPV3 (probability of cancer among patients actually undergoing biopsy after a BI-RADS assessment of 4 or 5). For screening examinations with an initial BI-RADS assessment of category of 0, the final assessment was determined by looking ahead 180 days to determine whether additional imaging had been performed. Final assessment was used when PPV2 and PPV3 were computed. A final BI-RADS assessment of category 4 or 5 was assumed to be a biopsy recommendation. PPV3 included the performance of any type of biopsy (fine-needle aspiration, cyst aspiration, core, or surgical biopsy). PPV2 and PPV3 are both important, as they are measurements of different aspects of the process; PPV2 is a measurement of PPV for biopsy recommendations, whereas PPV3 is a measurement for biopsies actually performed.
Because few mammographic facilities have adequate resources to estimate sensitivity or specificity, we report those calculations only on the BCSC Web site at http://breastscreening.cancer.gov.
Simple descriptive statistics (frequency, percentile, mean, and median values) were chosen to provide clinically relevant screening performance benchmarks. We illustrated the variability found among radiologists by using percentile values to indicate ranges that describe where the middle 50% and 80% of performance outcomes was found for specific outcome measurements. For example, the combination of 25th and 75th percentile values defines the range within which the middle 50% of performance outcomes was found, and the combination of 10th and 90th percentile values defines the range within which the middle 80% of performance outcomes was found. To reduce the amount of random statistical variation in these data, we reported outcomes from only those radiologists who contributed at least a designated, subjectively determined minimum number of mammographic examinations or cancers for each outcome displayed, as follows: recall rate and cancer detection rate, 1000 examinations; PPV1, 100 abnormal interpretations; PPV2, 30 biopsy recommendations; PPV3, 30 biopsies performed; and for cancer measurements, 15 cancers with complete information on the outcome criteria. We used graphic presentations (frequency distributions overlaid with percentile values) to display these data in an easily understandable format and present the tabular data in the BCSC Web site. More complex analytic methods, such as those designed to elucidate statistically significant interactions among the data variables collected, are beyond the scope of our study.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONADVANCES IN KNOWLEDGEReferences |
|---|
|
|
Mammographic Performance Measurements
The recall rate was 9.8% (253 169 of 2 580 151). PPV1 (percentage of cancers determined after a positive screening examination) was 4.8% (12 068 of 253 169), PPV2 (percentage of cancers determined after a BI-RADS assessment category of 4 or 5 was assigned) was 24.6% (9342 of 37 987), and PPV3 (percentage of cancers determined after a BI-RADS assessment category of 4 or 5 was assigned and a biopsy was performed) was 33.8% (8901 of 26 340) (Table 3). Prior to reclassification of some otherwise negative or benign examinations as BI-RADS category 0, such as when additional imaging was performed at the time of the screening examination (to increase consistency of BI-RADS assessments between radiologists), the recall rate was 7.5% (193 265 of 2 580 151) and the PPV1 was 6.0% (11 560 of 193 265).
|
The percentage of all cancers diagnosed as ductal carcinoma in situ was 21.6% (2603 of 12 068) (Table 4). Of the invasive cancers with known size, 37.1% (3215 of 8656) were 10 mm or smaller, and 21.2% (1839 of 8656) were larger than 2 cm. The median size was 13 mm, and the mean size was 16.4 mm. The percentage of cancers considered minimal (cases of ductal carcinoma in situ or invasive cancer of 10 mm or smaller) was 51.7% (5818 of 11 259 [known size]) (Table 4).
|
Performance Benchmarks
The range of recall rate of the middle 50% of radiologists was 6.4%–13.3%, and that of 80% of radiologists was 4.4%–16.8% (Table 5, Fig 1). The range of PPV1 of the middle 50% of radiologists was 3.4%–6.2%, and that of 80% of radiologists was 2.6%–8.6%. The range of PPV2 (BI-RADS assessment 4 and 5) was 18.8%–32.0% for the middle 50% of radiologists, and the range for 80% of radiologists was 14.1%–38.8%. Most of the smoothed curves for performance outcomes (Figs 1, 2) have clearly defined peaks, except recall rate, where there is a flattening around the average of performance outcomes.
|
|
|
|
|
|
|
|
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONADVANCES IN KNOWLEDGEReferences |
|---|
Publication of our results follows that of the recent Institute of Medicine report of 2005 (36, p 5). That report recommends adding many of the measurements reported in our study to the mammographic facility audit requirements.
Unlike European screening programs, performance targets published in the United States have not been used to enforce performance outcomes. Just recently, estimates of performance outcomes of diagnostic mammography in a large group of radiologists in the United States provided empiric evidence on which to base performance targets (10,11).
Variation of mammographic outcome and accuracy measurements is well known, but the literature is generally limited to one practice or geographic area (19; 33; 34;36, p 5; 37), research methods not applicable to the community (38), or review of selected mammograms (39,40). Our study addresses these limitations by using graphic representations of outcome measurements from routine practice and from a broad base of representative radiologists and by using methods accessible to most radiologists. Thus, results of our study allow a radiologist to compare his or her outcome measurements with those of a group of radiologists who are representative of those in U.S. practice (11).
Findings in our report extend results of prior research because the range of performance outcomes in community practice for screening mammography is documented. Because only 10% of our mammograms were initial mammograms, results are best compared with results with subsequent mammograms. The average values of our results were similar to those in prior U.S. reports (8;10;29;36, p 5). In general, the results for most radiologists are within the desirable ranges recommended for highly skilled radiologists (4, p 83; 6, pp 147–148) (Tables 6, 7), with the exception of the recall rate (median, 9.7%) and the PPV1 (median, 4.5%).
|
|
The majority of radiologists appear to practice within the Agency for Health Care Policy and Research and European minimal guidelines for cancer detection rate and ductal carcinoma in situ detection, and the percentage of cancers that were classified as early stage appeared to substantially exceed the guidelines. These favorable outcomes are perhaps the result of the higher recall rates and/or shorter screening intervals (median of 18 months in the United States vs 36 months in the United Kingdom) (42), which are commonly observed in practice in the United States.
Study Limitations
Although we believe a major strength of our study is its large sample of clinical practices drawn from a diverse geographic area, features that allow it to be a mirror of the typical U.S. community practice, limitations in this analysis should be noted. We did not evaluate how characteristics, such as physician volume or the practice of double reading, influenced observed outcomes, nor did we evaluate how the outcome parameters interact (eg, an examination of the relationship of recall rate to sensitivity or cancer detection rate). These types of issues are crucial to the creation of new guidelines and are being evaluated in other research efforts in which BCSC data are used.
Use of BCSC Performance Benchmarks Data by Radiologists
For these data to be valuable to community radiologists, these radiologists need to collect the necessary data about their own practice. Efficient audit systems of screening mammographic practices are needed for collection of prospective, long-term, standardized, and high-quality data as recommended by the recent Institute of Medicine report (36). In many practices, maybe most, radiologists do not collect much of these data and cannot therefore evaluate their performance outcomes relative to the BCSC data about performance benchmarks for mammographic screening. Evaluation of the quality of care delivery is not an easy process and is best performed with specialized software, knowledgeable personnel, and access to pathologic data. There is no reimbursement for the substantial costs involved.
The collection of cancer data is likely to be incomplete in community radiology, and acquisition of complete follow-up information for outcomes for all biopsies is a greater challenge when no centralized registry exists.
Another important limitation is the small number of cancers present for each radiologist on screening mammograms performed during 1 year. A difference of even one or two cancers has a major effect on the resulting performance measurement. This imprecision may be offset by aggregation of several years of data and/or data from an entire group of radiologists (Appendix E1 [radiology.rsnajnls.org/cgi/content/full/241/1/55/DC1]).
Patient population differences also will alter these performance measurements. Patients' demographic characteristics such as age, race and ethnicity, family history of breast cancer, and prior mammography will affect all outcomes, especially cancer detection rate and recall rate (10–12,29,30,40).
The benchmark data displayed in our study represent the current range of community performance outcomes. It should not be inferred that the average community performance outcome is necessarily the recommended target for performance. Desirable goals, targets, or guidelines are created by a panel of experienced radiologists by using an evidence-based process that includes a thorough review of peer-reviewed literature and their own experiences. Examples of evidence-based review processes are those commissioned by the Agency for Healthcare Research and Quality–managed U.S. Preventive Services Task Force or a panel of experienced radiologists provided by a committee of the American College of Radiology. Benchmark data are useful to create these goals or guidelines, but they are useful to the practicing radiologist only if he or she has data from a similar population that are calculated by using the same methods. Our study findings, therefore, should be helpful in updating any future guidelines, as they demonstrate the variation present in community practice.
The wide variation in recall rate likely represents a lack of consensus among practicing radiologists concerning performance targets. A wide variation is of concern because the literature suggests that wide variation in the processes of care delivery may be associated with lower quality or worse outcomes from care delivery (43, p 88). These benchmark data demonstrate the need for improvement in screening mammographic performance in the United States if the guidelines recommended by the Agency for Healthcare Research and Quality are to be met.
A major caution in the use of screening mammographic guidelines is that a single performance measurement in isolation may not be meaningful. In particular, the acceptability of a value for recall rate also depends on the parallel acceptability of values for PPV1, cancer detection rate, and size of invasive cancer detected. For example, a radiologist with a higher than average recall rate may need those additional recalled patients to achieve a sufficiently high cancer detection rate and a sufficiently high detection rate for invasive cancers of small size. Investigators in another BCSC study are examining how the parameters of recall rate, cancer detection rate, and sensitivity interact in clinical practice.
Fortunately, efforts have already begun to evaluate screening mammographic performance through an integrated assessment of recall rate, cancer detection rate, and PPV. One such innovative approach has been developed in the United Kingdom where facilities receive feedback in a single graphic display that indicates whether they are operating within an acceptable range of three measurements: recall rate, cancer detection rate, and PPV1 (44). This feedback allows identification of facilities or radiologists who are operating outside of the acceptable range for these parameters. With on-site review, experienced radiologists can then confirm problems and assist in processes necessary to improve performance outcome (5, pp 4–5; 45).
|
CONCLUSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONADVANCES IN KNOWLEDGEReferences |
|---|
Although most measurements of the performance of radiologists for screening examinations are similar to published recommendations, for many radiologists, recall rate is higher and PPV1 is lower than the recommendations. Variability in all of these measurements for any practice may relate in part to methods used but primarily reflects actual differences. Additional research and involvement by panels of experienced radiologists will be required to better define the optimal performance targets appropriate to the U.S. health care environment.
|
ADVANCES IN KNOWLEDGE |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONADVANCES IN KNOWLEDGEReferences |
|---|
|
FOOTNOTES |
|---|
Abbreviations: BCSC = Breast Cancer Surveillance Consortium • BI-RADS = Breast Imaging Reporting and Data System • PPV = positive predictive value
Authors stated no financial relationship to disclose.
Author contributions: Guarantor of integrity of entire study, R.D.R.; study concepts/study design or data acquisition or data analysis/interpretation, all authors; manuscript drafting or manuscript revision for important intellectual content, all authors; manuscript final version approval, all authors; literature research, R.D.R., B.C.Y., E.A.S., P.A.C., K.K.; clinical studies, K.K.; statistical analysis, L.A.A., K.K., W.E.B.; and manuscript editing, B.C.Y., L.A.A., E.A.S., C.D.L., B.M.G., P.A.C., K.K., D.S.M.B., D.L.W., W.E.B., R.B.
|
References |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONADVANCES IN KNOWLEDGEReferences |
|---|
This article has been cited by other articles:
|
|
 |
|
  E. J. A. Bowles and B. M. Geller Best Ways to Provide Feedback to Radiologists on Mammography Performance Am. J. Roentgenol., July 1, 2009; 193(1): 157 - 164. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S. L. Jackson, S. H. Taplin, E. A. Sickles, L. Abraham, W. E. Barlow, P. A. Carney, B. Geller, E. A. Berns, G. R. Cutter, and J. G. Elmore Variability of Interpretive Accuracy Among Diagnostic Mammography Facilities J Natl Cancer Inst, June 2, 2009; 101(11): 814 - 827. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. M. Lee, D. Georgian-Smith, G. S. Gazelle, E. F. Halpern, E. A. Rafferty, R. H. Moore, E. D. Yeh, H. A. D'Alessandro, R. A. Hitt, and D. B. Kopans Detecting Nonpalpable Recurrent Breast Cancer: The Role of Routine Mammographic Screening of Transverse Rectus Abdominis Myocutaneous Flap Reconstructions Radiology, August 1, 2008; 248(2): 398 - 405. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 W. A. Berg, J. D. Blume, J. B. Cormack, E. B. Mendelson, D. Lehrer, M. Bohm-Velez, E. D. Pisano, R. A. Jong, W. P. Evans, M. J. Morton, et al. Combined Screening With Ultrasound and Mammography vs Mammography Alone in Women at Elevated Risk of Breast Cancer JAMA, May 14, 2008; 299(18): 2151 - 2163. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. G. Elmore and R. J. Brenner The More Eyes, the Better to See? From Double to Quadruple Reading of Screening Mammograms J Natl Cancer Inst, August 1, 2007; 99(15): 1141 - 1143. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
