Concordance of breast imaging reporting and data system assessments and management recommendations in screening mammography

Reader characteristics and mammogram features associated with breast imaging reporting scores

Objectives:

This study aims to explore the reading performances of radiologists in detecting cancers on mammograms using Tabar Breast Imaging Reporting and Data System (BIRADS) classification and identify factors related to breast imaging reporting scores.

Methods:

117 readings of five different mammogram test sets with each set containing 20 cancer and 40 normal cases were performed by Australian radiologists. Each radiologist evaluated the mammograms using the BIRADS lexicon with category 1 - negative, category 2 - benign findings, category 3 - equivocal findings (Recall), category 4 - suspicious findings (Recall), and category 5 - highly suggestive of malignant findings (Recall). Performance metrics (true positive, false positive, true negative, and false negative) were calculated for each radiologist and the distribution of reporting categories was analyzed in reader-based and case-based groups. The association of reader characteristics and case features among categories was examined using Mann-Whitney U and Kruskal-Wallis tests.

Results:

38% of cancer-containing mammograms were reported with category 3 which decreased to 32.3% with category 4 and 16.2% with category 5 while 16.6 and 10.3% of cancer cases were marked with categories 1 and 2. Female readers had less false-negative rates when using categories 1 and 2 for cancer cases than male readers (p < 0.01). A similar pattern as gender category was also found in Breast Screen readers and readers completed breast reading fellowships compared with non-Breast Screen and non-fellowship readers (p < 0.05). Radiologists with low number of cases read per week were more likely to record the cancer cases with category 4 while the ones with high number of cases were with category 3 (p < 0.01). Discrete mass and asymmetric density were the two types of abnormalities reported mostly as equivocal findings with category 3 (47–50%; p = 0.005) while spiculated mass or stellate lesions were mostly selected as highly suggestive of malignancy with category 5 (26%, p = 0.001).

Conclusions:

Most radiologists used category 3 when reporting cancer mammograms. Gender, working for BreastScreen, fellowship completion, and number of cases read per week were factors associated with scoring selection. Radiologists reported higher Tabar BIRADS category for specific types of abnormalities on mammograms than others.

Advances in knowledge:

The study identified factors associated with the decision of radiologists in assigning a BIRADS Tabar score for mammograms with abnormality. These findings will be useful for individual training programs to improve the confidence of radiologists in recognizing abnormal lesions on screening mammograms.

New mammography screening performance metrics based on the entire screening episode

BACKGROUND

Established mammography screening performance metrics use the initial screening mammography assessment because they were developed for radiologist performance auditing, yet these metrics are frequently used to inform health policy and screening decision making. The authors have developed new performance metrics based on the final assessment that consider the entire screening episode, including diagnostic workup.

METHODS

The authors used data from 2,512,577 screening episodes during 2005-2017 at 146 facilities in the United States participating in the Breast Cancer Surveillance Consortium. Screening performance metrics based on the final assessment of the screening episode were compared with conventional metrics defined with the initial assessment. Results were also stratified by breast density and breast cancer risk.

RESULTS

The cancer detection rates were similar for the final assessment (4.1 per 1000; 95% confidence interval [CI], 3.8-4.3 per 1000) and the initial assessment (4.1 per 1000; 95% CI, 3.9-4.3 per 1000). The interval cancer rate was 12% higher when it was based on the final assessment (0.77 per 1000; 95% CI, 0.71-0.83 per 1000) versus the initial assessment (0.69 per 1000; 95% CI, 0.64-0.74 per 1000), and this resulted in a modest difference in sensitivity (84.1% [95% CI, 83.0%-85.1%] vs 85.7% [95% CI, 84.8%-86.6%], respectively). Absolute differences in the interval cancer rate between final and initial assessments increased with breast density and breast cancer risk (eg, a difference of 0.29 per 1000 for women with extremely dense breasts and a 5-year risk >2.49%).

CONCLUSIONS

Established screening performance metrics underestimate the interval cancer rate of a mammography screening episode, particularly for women with dense breasts or an elevated breast cancer risk. Women, clinicians, policymakers, and researchers should use final-assessment performance metrics to support informed screening decisions.

Frequency and Determinants of a Short-Interval Follow-up Recommendation After an Abnormal Screening Mammogram

PURPOSE

After imaging assessment of an abnormal screening mammogram, a follow-up examination 6 months later is recommended to some women. Our aim was to identify which characteristics of lesions, women, and physicians are associated to such short-interval follow-up recommendation in the Quebec Breast Cancer Screening Program.

METHODS

Between 1998 and 2008, 1,839,396 screening mammograms were performed and a total of 114,781 abnormal screens were assessed by imaging only. Multivariate analysis was done with multilevel Poisson regression models with robust variance and generalized linear mixed models.

RESULTS

A short-interval follow-up was recommended in 26.7% of assessments with imaging only, representing 2.3% of all screens. Case-mix adjusted proportion of short-interval follow-up recommendations varied substantially across physicians (range: 4%-64%). Radiologists with high recall rates (≥15%) had a high proportion of short-interval follow-up recommendation (risk ratio: 1.82; 95% confidence interval: 1.35-2.45) compared to radiologists with low recall rates (<5%). The adjusted proportion of short-interval follow-up was high (22.8%) even when a previous mammogram was usually available.

CONCLUSIONS

Short-interval follow-up recommendation at assessment is frequent in this Canadian screening program, even when a previous mammogram is available. Characteristics related to radiologists appear to be key determinants of short-interval follow-up recommendation, rather than characteristics of lesions or patient mix. Given that it can cause anxiety to women and adds pressure on the health system, it appears important to record and report short-interval follow-up and to identify ways to reduce its frequency. Short-interval follow-up recommendations should be considered when assessing the burden of mammography screening.

Outcomes of unconventional utilization of BI-RADS category 3 assessment at opportunistic screening

Background An important difficulty regarding the Breast Imaging Reporting and Data System (BI-RADS) category 3 assessment is the need for extensive diagnostic workup and an additional 6-month follow-up study. Purpose To evaluate the feasibility of the BI-RADS category 3 assessments at opportunistic screening. Material and Methods Mammography charts of 9062 screening patients in a major teaching hospital situated in an urban setting of a developing country were evaluated retrospectively (1997-2010). BI-RADS category 3 patients, called for a 6-month follow-up, which comprised a single-view spot or magnification mammogram. The length of follow-up period, compliance to periodic mammographic surveillance, cancer detection rate, and negative predictive values of category 3 assessments were calculated. Results Of the screened population, 9.2% were assigned BI-RADS category 3, and 31.2% of these cases were lost to follow-up. The mean follow-up period for 606 patients was 36.9 months. The negative predictive values for 6-month, 12-month, and final control studies were 90.9%, 87.5%, and 100%, respectively. Patient compliance for 6 months, 12 months, and any control evaluations beyond 12 months was low (50.0%, 29.8%, and 47.5%, respectively). Cancer detection rate was 0.8%. Conclusion Results of the study supports the feasibility of the BI-RADS category 3 assessments at opportunistic screening without any additional diagnostic workup. The practice of category 3 assessment following screening mammograms may be a more cost-effective method for developing countries with high recall rates and low resources in eliminating the maximum risk with minimum cost within the limits of available resources.

Annexin A3 is a mammary marker and a potential neoplastic breast cell therapeutic target

Breast cancers are the most common cancer-affecting women; critically the identification of novel biomarkers for improving early detection, stratification and differentiation from benign tumours is important for the reduction of morbidity and mortality.

To identify and functionally characterise potential biomarkers, we used mass spectrometry (MS) to analyse serum samples representing control, benign breast disease (BBD) and invasive breast cancer (IDC) patients. Complementary and multidimensional proteomic approaches were used to identify and validate novel serum markers.

Annexin A3 (ANX A3) was found to be differentially expressed amongst different breast pathologies. The diagnostic value of serum ANX A3 was subsequently validated by ELISA in an independent serum set representing the three groups. Here, ANX A3 was significantly upregulated in the benign disease group sera compared with other groups (P < 0.0005).

In addition, paired breast tissue immunostaining confirmed that ANX A3 was abundantly expressed in benign and to a lesser extent malignant neoplastic epithelium. Finally, we illustrated ANX A3 expression in cell culture lysates and conditioned media from neoplastic breast cell lines, and its role in neoplastic breast cell migration in vitro.

This study confirms the novel role of ANX A3 as a mammary biomarker, regulator and therapeutic target.

Concordance of BI-RADS Assessments and Management Recommendations for Breast MRI in Community Practice

OBJECTIVE

The purpose of this study was to evaluate concordance between BI-RADS assessments and management recommendations for breast MRI in community practice.

MATERIALS AND METHODS

Breast MRI data were collected from four regional Breast Cancer Surveillance Consortium registries from 2005 to 2011 for women who were 18-79 years old. Assessments and recommendations were compared to determine concordance according to BI-RADS guidelines. Concordance was compared by assessment category as well as by year of examination and clinical indication.

RESULTS

In all, 8283 MRI examinations were included in the analysis. Concordance was highest (93% [2475/2657]) in examinations with a BI-RADS category 2 (benign) assessment. Concordance was also high in examinations with category 1 (negative) (87% [1669/1909]), category 0 (incomplete) (83% [348/417]), category 5 (highly suggestive of malignancy) (83% [208/252]), and category 4 (suspicious) (74% [734/993]) assessments. Examinations with categories 3 (probably benign) and 6 (known biopsy-proven malignancy) assessments had the lowest concordance rates (36% [302/837] and 56% [676/1218], respectively). The most frequent discordant recommendation for a category 3 assessment was routine follow-up. The most frequent discordant recommendation for a category 6 assessment was biopsy. Concordance of assessments and management recommendations differed across clinical indications (p < 0.0001), with the lowest concordance in examinations to assess disease extent.

CONCLUSION

Breast MRI BI-RADS management recommendations were most concordant for assessments of negative, incomplete, suspicious, and highly suggestive of malignancy. Lower concordance for assessments of probably benign and known biopsy-proven malignancy and for examinations performed to assess disease extent highlight areas for interventions to improve breast MRI reporting.

Breast Cancer Characteristics Associated With Digital Versus Film-Screen Mammography for Screen-Detected and Interval Cancers

OBJECTIVE

The purpose of this study was to determine whether pathologic findings of screen-detected and interval cancers differ for digital versus film mammography.

MATERIALS AND METHODS

Breast Cancer Surveillance Consortium data from 2003-2011 on 3,021,515 screening mammograms (40.3% digital, 59.7% film) of women 40-89 years old were reviewed. Cancers were considered screen detected if diagnosed within 12 months of an examination with positive findings and interval if diagnosed within 12 months of an examination with negative findings. Tumor characteristics for screen-detected and interval cancers were compared for digital versus film mammography by use of logistic regression models to estimate the odds ratio and 95% CI with adjustment for age, race and ethnicity, hormone therapy use, screening interval, examination year, and registry. Generalized estimating equations were used to account for correlation within facilities.

RESULTS

Among 15,729 breast cancers, 85.3% were screen detected and 14.7% were interval. Digital and film mammography had similar rates of screen-detected (4.47 vs 4.42 per 1000 examinations) and interval (0.73 vs 0.79 per 1000 examinations) cancers for digital versus film. In adjusted analyses, interval cancers diagnosed after digital examinations with negative findings were less likely to be American Joint Committee on Cancer stage IIB or higher (odds ratio, 0.69; 95% CI, 0.52-0.93), have positive nodal status (odds ratio, 0.78; 95% CI, 0.64-0.95), or be estrogen receptor negative (odds ratio, 0.71; 95% CI, 0.56-0.91) than were interval cancers diagnosed after a film examination with negative findings.

CONCLUSION

Screen-detected cancers diagnosed after digital and film mammography had similar rates of unfavorable tumor characteristics. Interval-detected cancers diagnosed after a digital examination were less likely to have unfavorable tumor features than those diagnosed after film mammography, but the absolute differences were small.

A history of breast cancer and older age allow risk stratification of mammographic BI-RADS 3 ratings in the diagnostic setting

OBJECTIVE

The objective was to investigate whether risk stratification of mammographic Breast Imaging: Reporting and Data System (BI-RADS) 3 can be accomplished in the diagnostic setting.

METHODS

We analyzed 4941 BI-RADS-3-rated patients (23 malignant outcomes) and built logistic-regression models with age, personal and family history of breast cancer, fibroglandular density, and additional mammographic findings as predictive variables.

RESULTS

A personal history of breast cancer (odds ratio: 5.53) and older age (odds ratio: 12.44/10.93 for age 50-64/>64) are independent risk factors. Patients with both risk factors have a risk >2%.

CONCLUSION

Biopsy may be warranted in older patients with a history of breast cancer who would be otherwise assigned BI-RADS 3.

The influence of mammographic technologists on radiologists' ability to interpret screening mammograms in community practice

RATIONALE AND OBJECTIVES

To determine whether the mammographic technologist has an effect on the radiologists' interpretative performance of screening mammography in community practice.

MATERIALS AND METHODS

In this institutional review board-approved retrospective cohort study, we included Carolina Mammography Registry data from 372 radiologists and 356 mammographic technologists from 1994 to 2009 who performed 1,003,276 screening mammograms. Measures of interpretative performance (recall rate, sensitivity, specificity, positive predictive value [PPV1], and cancer detection rate [CDR]) were ascertained prospectively with cancer outcomes collected from the state cancer registry and pathology reports. To determine if the mammographic technologist influenced the radiologists' performance, we used mixed effects logistic regression models, including a radiologist-specific random effect and taking into account the clustering of examinations across women, separately for screen-film mammography (SFM) and full-field digital mammography (FFDM).

RESULTS

Of the 356 mammographic technologists included, 343 performed 889,347 SFM examinations, 51 performed 113,929 FFDM examinations, and 38 performed both SFM and FFDM examinations. A total of 4328 cancers were reported for SFM and 564 cancers for FFDM. The technologists had a statistically significant effect on the radiologists' recall rate, sensitivity, specificity, and CDR for both SFM and FFDM (P values <.01). For PPV1, variability by technologist was observed for SFM (P value <.0001) but not for FFDM (P value = .088).

CONCLUSIONS

The interpretative performance of radiologists in screening mammography varies substantially by the technologist performing the examination. Additional studies should aim to identify technologist characteristics that may explain this variation.

Breast MRI BI-RADS assessments and abnormal interpretation rates by clinical indication in US community practices

RATIONALE AND OBJECTIVES

As breast magnetic resonance imaging (MRI) use grows, benchmark performance parameters are needed for auditing and quality assurance purposes. We describe the variation in breast MRI abnormal interpretation rates (AIRs) by clinical indication among a large sample of US community practices.

MATERIALS AND METHODS

We analyzed data from 41 facilities across five Breast Cancer Surveillance Consortium imaging registries. Each registry obtained institutional review board approval for this Health Insurance Portability and Accountability Act compliant analysis. We included 11,654 breast MRI examinations conducted in 2005-2010 among women aged 18-79 years. We categorized clinical indications as 1) screening, 2) extent of disease, 3) diagnostic (eg, breast symptoms), and 4) other (eg, short-interval follow-up). We characterized assessments as positive (ie, Breast Imaging Reporting and Data System [BI-RADS] 0, 4, and 5) or negative (ie, BI-RADS 1, 2, and 6) and provide results with BI-RADS 3 categorized as positive and negative. We tested for differences in AIRs across clinical indications both unadjusted and adjusted for patient characteristics and registry and assessed for changes in AIRs by year within each clinical indication.

RESULTS

When categorizing BI-RADS 3 as positive, AIRs were 21.0% (95% confidence interval [CI], 19.8-22.3) for screening, 31.7% (95% CI, 29.6-33.8) for extent of disease, 29.7% (95% CI, 28.3-31.1) for diagnostic, and 27.4% (95% CI, 25.0-29.8) for other indications (P < .0001). When categorizing BI-RADS 3 as negative, AIRs were 10.5% (95% CI, 9.5-11.4) for screening, 21.8% (95% CI, 19.9-23.6) for extent of disease, 17.7% (95% CI, 16.5-18.8) for diagnostic, and 13.3% (95% CI, 11.6-15.2) for other indications (P < .0001). The significant differences in AIRs by indication persisted even after adjusting for patient characteristics and registry (P < .0001). In addition, for most indications, there were no significant changes in AIRs over time.

CONCLUSIONS

Breast MRI AIRs differ significantly by clinical indication. Practices should stratify breast MRI examinations by indication for quality assurance and auditing purposes.

Design and implementation of handheld and desktop software for the structured reporting of hepatic masses using the LI-RADS schema

RATIONALE AND OBJECTIVES

The Liver Imaging Reporting and Data System (LI-RADS) can enhance communication between radiologists and clinicians if applied consistently. We identified an institutional need to improve liver imaging report standardization and developed handheld and desktop software to serve this purpose.

MATERIALS AND METHODS

We developed two complementary applications that implement the LI-RADS schema. A mobile application for iOS devices written in the Objective-C language allows for rapid characterization of hepatic observations under a variety of circumstances. A desktop application written in the Java language allows for comprehensive observation characterization and standardized report text generation. We chose the applications' languages and feature sets based on the computing resources of target platforms, anticipated usage scenarios, and ease of application installation, deployment, and updating.

RESULTS

Our primary results are the publication of the core source code implementing the LI-RADS algorithm and the availability of the applications for use worldwide via our website, http://www.liradsapp.com/. The Java application is free open-source software that can be integrated into nearly any vendor's reporting system. The iOS application is distributed through Apple's iTunes App Store. Observation categorizations of both programs have been manually validated to be correct. The iOS application has been used to characterize liver tumors during multidisciplinary conferences of our institution, and several faculty members, fellows, and residents have adopted the generated text of Java application into their diagnostic reports.

CONCLUSIONS

Although these two applications were developed for the specific reporting requirements of our liver tumor service, we intend to apply this development model to other diseases as well. Through semiautomated structured report generation and observation characterization, we aim to improve patient care while increasing radiologist efficiency.

Probably benign breast MRI lesions: frequency, lesion type, and rate of malignancy

PURPOSE

To determine the frequency, malignancy rate, and imaging characteristics of BI-RADS (Breast Imaging Reporting and Data System) 3 MRI detected lesions.

MATERIALS AND METHODS

A total of 4370 consecutive contrast-enhanced breast MRI exams from March 1, 2004, to March 1, 2009, were retrospectively reviewed. The study was HIPAA (Health Insurance Portability Accountability Act) compliant and Institutional Review Board approved. The frequency of BI-RADS 3 MRI interpretations and lesion type, as well as malignancy rate at follow-up were calculated.

RESULTS

There were 348/4370 (8%) BI-RADS 3 exams performed on 345 patients. There were 125 foci (35.9%), 52 non-masslike enhancements (14.9%), and 171 masses (49.1%). There were 207 lesions with biopsy or imaging follow-up of at least 24 months (mean, 32 months; range, 6-77 months) of the BI-RADS 3 lesion. Five of these 207 (2.4%) lesions were found to be malignant.

CONCLUSION

Rate of malignancy among BI-RADS 3 MRI exams with follow-up was 2.4%. Types of lesions assigned as probably benign were 35.9% foci, 14.9% non-masslike enhancements, and 49.1% masses.

Breast imaging reports for malignant lesions: are we maintaining recommended BI-RADS(®) lexicon standards?

Background

The purpose of this study was to evaluate mammography reports for diagnosed breast cancer cases in major government and private centers in Karachi, Pakistan, with respect to concordance with the Breast Imaging Reports And Data System (BI-RADS^®) lexicon.

Methods

A prospective, descriptive, multicenter study was conducted in the radiology sections of the Aga Khan University Hospital, Pakistan Naval Station Shifa Hospital, Advanced Radiology Clinic, Karachi Institute of Radiotherapy and Nuclear Medicine, and Civil Hospital Karachi between May and October 2010 after approval from the ethical review committee of Aga Khan University. Mammograms reported as BI-RADS category 4 and 5 were included in the study. Mammograms reported as BI-RADS category 0, 1, 2 and 3 were excluded. Fifty reports were collected from each center. Data were collected about the clinical indication, breast density, location and description of the lesion, calcification, and comments on axillary lymph nodes. This description was compared with the BI-RADS lexicon.

Results

The mean age of the patients was 50 ± 12 years. The clinical indication, breast parenchymal density, lesion location, and presence of calcification were better described by the private centers, while description of lymph node status was better stated by the government centers. This difference was statistically significant, except for lesion description. The description of masses by the two reporting groups was comparable.

Conclusion

Mammographic reporting of malignant breast lesions in the private sector is more in line with the BI-RADS lexicon, as compared with government sector hospitals in Karachi, Pakistan. Lymph node documentation was better in government sector reports.

Second Opinion Assessment in Diagnostic Mammography at a Breast Cancer Centre

Purpose: The aim of this retrospective study was to evaluate the importance of second opinion assessment for diagnostic mammography and sonography in a breast cancer centre. Material and Method: We analysed a total of 374 diagnostic mammographies and sonographies. All patients had previously undergone mammography and sonography examination in different external clinics, and the findings had been classified according to the BI-RADS system. All patients underwent additional sonography investigation in the outpatient department of our university clinic with additional mammography where necessary. The final diagnosis (histological clarification in 316 cases, follow-up in 58 cases) was compared with the BI-RADS classification made by the external clinics and by the university clinic, and the correlation between their findings and the final diagnosis was analysed. Results: The final diagnosis yielded 146 benign lesions and 228 cancers. In 74 % of cases (277/374), the BI-RADS classification of the first assessment corresponded to that of the second assessment. 26/55 lesions (47 %) were upgraded at the second assessment from BI-RADS 3 to BI-RADS 4, and 71/186 findings (38 %) were downgraded at the second assessment from BI-RADS 4 to BI-RADS 3. The correlation between the initial diagnosis made in the external facilities and the final diagnosis was low (kappa: 0.263), but the correlation between the second opinion assessment and the final diagnosis was significantly (p < 0.001) higher (kappa: 0.765). The second assessment increased the sensitivity from 91 % (208/228) to 99 % (225/228) and the specificity from 32 % (46/146) to 74 % (108/146). 20 additional malignant lesions were only detected at the second assessment; however the second assessment also resulted in 3 additional false-negative findings. Surgical biopsy was prevented in 49 women after the second assessment. Conclusion: An independent second diagnostic evaluation can significantly improve the correlation between BI-RADS classification and the final diagnosis, resulting in a benefit for the patient.

Risk of developing breast cancer by utilizing Gail model

The Gail model has been widely used to quantify an individual woman's risk of developing breast cancer by using important clinical parameters, usually for clinical counselling purposes or to determine eligibility for mammography and genetic tests. The aim of the present study was to estimate the five-year and lifetime breast cancer risk among women in Rasht, Iran. In this cross-sectional study, 314 women were evaluated at Alzahra Women Hospital in 2007. Participants were ≥35 years of age without a history of breast cancer. Risk estimation was performed using the computerized Gail model. A five-year risk >1.66% was considered high-risk; 5.1% of women were high-risk. The mean five-year breast cancer risk was 0.8% (SD±1). Mean breast cancer risk up to the age of 90 years (lifetime risk) was 9.0% (SD±3.9%); 16.2% of the participants had a five-year risk higher than the average woman of the same age, and 18.2% had the same risk. Also for the lifetime risk, 11.1% of the women had higher risk and 1.6% had the same risk as the average woman. Routine use of the Gail model is recommended for identifying women at high average risk for increasing the survival of women from breast cancer.

Follow-up frequency and compliance in women with probably benign findings on breast magnetic resonance imaging

RATIONALE AND OBJECTIVES

Six-month short-interval follow-up is recommended for probably benign findings on breast magnetic resonance imaging (MRI). We wanted to examine patient adherence to follow-up recommendation for Breast Imaging-Reporting and Data System (BI-RADS) category 3 lesions at a tertiary care medical center.

MATERIALS AND METHODS

We performed a retrospective review of frequency and adherence rates to follow-up recommendation for women with an initial BI-RADS 3 breast MRI between 2005 and 2007.

RESULTS

A total of 132 women with BI-RADS 3 breast MRI recommendations were included. Ninety-six of 132 (72.7%) women adhered to the first follow-up recommendation or elected to have tissue diagnosis; 78/132 (59.1%) had follow-up MRI and 18/132 (13.6%) had tissue diagnosis. Thirty-six of 132 (27.3%) women did not return for follow-up. Nine of nine (100%) of BRCA carriers returned for follow-up or had tissue diagnosis, compared to 87/123 (70.7%) of non-BRCA carriers. A total of 35/41 (85.4%) of patients with a prior history of breast cancer returned for follow-up or had tissue diagnosis, compared to 61/91 (67%) of patients without a history of breast cancer. Only 5/15 (33%) of patients undergoing MRI for symptom alone adhered to follow-up recommendations.

CONCLUSION

Adherence to BI-RADS category 3 follow-up recommendation is often low. Women with a history of breast cancer or who were BRCA carriers were significantly more likely to adhere to follow-up recommendation than women without a history of breast cancer or women undergoing MRI for symptoms alone. Strategies to improve adherence should be developed.

The Effect of Budgetary Restrictions on Breast Cancer Diagnostic Decisions

We develop a finite-horizon discrete-time constrained Markov decision process (MDP) to model diagnostic decisions after mammography where we maximize the total expected quality-adjusted life years (QALYs) of a patient under resource constraints. We use clinical data to estimate the parameters of the MDP model and solve it as a mixed-integer program. By repeating optimization for a sequence of budget levels, we calculate incremental cost-effectiveness ratios attributable to consecutive levels of funding and compare actual clinical practice with optimal decisions. We prove that the optimal value function is concave in the allocated budget. Comparing to actual clinical practice, using optimal thresholds for decision making may result in approximately 22% cost savings without sacrificing QALYs. Our analysis indicates short-term follow-ups are the immediate target for elimination when budget becomes a concern. Policy change is more drastic in the older age group with the increasing budget, yet the gains in total expected QALYs related to larger budgets are predominantly seen in younger women along with modest gains for older women.

The Breast Imaging Reporting and Data System (BI-RADS) in the Dutch breast cancer screening programme: its role as an assessment and stratification tool

Objectives

To assess the suitability of the Breast Imaging Reporting and Data System (BI-RADS) as a quality assessment tool in the Dutch breast cancer screening programme.

Methods

The data of 93,793 screened women in the Amsterdam screening region (November 2005–July 2006) were reviewed. BI-RADS categories, work-up, age, final diagnosis and final TNM classification were available from the screening registry. Interval cancers were obtained through linkage with the cancer registry. BI-RADS was introduced as a pilot in the Amsterdam region before the nationwide introduction of digital mammography (2009–2010).

Results

A total of 1,559 women were referred to hospital (referral rate 1.7 %). Breast cancer was diagnosed in 485 women (detection rate 0.52 %); 253 interval cancers were reported, yielding a programme sensitivity of 66 % and specificity of 99 %. BI-RADS 0 had a lower positive predictive value (PPV, 14.1 %) than BI-RADS 4 (39.1 %) and BI-RADS 5 (92.9 %; P < 0.0001). The number of invasive procedures and tumour size also differed significantly between BI-RADS categories (P < 0.0001).

Conclusion

The significant differences in PPV, invasive procedures and tumour size match with stratification into BI-RADS categories. It revealed inter-observer variability between screening radiologists and can thus be used as a quality assessment tool in screening and as a stratification tool in diagnostic work-up.

Key Points

• The BI-RADS atlas is widely used in breast cancer screening programmes.

• There were significant differences in results amongst different BI-RADS categories.

• Those differences represented the radiologists’ degree of suspicion for malignancy, thus enabling stratification of referrals.

• BI-RADS can be used as a quality assessment tool in screening.

• Training should create more uniformity in applying the BI-RADS lexicon.

Timeliness of follow-up after abnormal screening mammogram: variability of facilities

PURPOSE

To describe the timeliness of follow-up care in community-based settings among women who receive a recommendation for immediate follow-up during the screening mammography process and how follow-up timeliness varies according to facility and facility-level characteristics.

MATERIALS AND METHODS

This was an institutional review board-approved and HIPAA-compliant study. Screening mammograms obtained from 1996 to 2007 in women 40-80 years old in the Breast Cancer Surveillance Consortium were examined. Inclusion criteria were a recommendation for immediate follow-up at screening, or subsequent imaging, and observed follow-up within 180 days of the recommendation. Recommendations for additional imaging (AI) and biopsy or surgical consultation (BSC) were analyzed separately. The distribution of time to follow-up care was estimated by using the Kaplan-Meier estimator.

RESULTS

Data were available on 214,897 AI recommendations from 118 facilities and 35,622 BSC recommendations from 101 facilities. The median time to subsequent follow-up care after recommendation was 14 days for AI and 16 days for BSC. Approximately 90% of AI follow-up and 81% of BSC follow-up occurred within 30 days. Facilities with higher recall rates tended to have longer AI follow-up times (P < .001). Over the study period, BSC follow-up rates at 15 and 30 days improved (P < .001). Follow-up times varied substantially across facilities. Timely follow-up was associated with larger volumes of the recommended procedures but not notably associated with facility type nor observed facility-level characteristics.

CONCLUSION

Most patients with follow-up returned within 3 weeks of the recommendation.

Effectiveness of computer-aided detection in community mammography practice

BACKGROUND

Computer-aided detection (CAD) is applied during screening mammography for millions of US women annually, although it is uncertain whether CAD improves breast cancer detection when used by community radiologists.

METHODS

We investigated the association between CAD use during film-screen screening mammography and specificity, sensitivity, positive predictive value, cancer detection rates, and prognostic characteristics of breast cancers (stage, size, and node involvement). Records from 684 956 women who received more than 1.6 million film-screen mammograms at Breast Cancer Surveillance Consortium facilities in seven states in the United States from 1998 to 2006 were analyzed. We used random-effects logistic regression to estimate associations between CAD and specificity (true-negative examinations among women without breast cancer), sensitivity (true-positive examinations among women with breast cancer diagnosed within 1 year of mammography), and positive predictive value (breast cancer diagnosed after positive mammograms) while adjusting for mammography registry, patient age, time since previous mammography, breast density, use of hormone replacement therapy, and year of examination (1998-2002 vs 2003-2006). All statistical tests were two-sided.

RESULTS

Of 90 total facilities, 25 (27.8%) adopted CAD and used it for an average of 27.5 study months. In adjusted analyses, CAD use was associated with statistically significantly lower specificity (OR = 0.87, 95% confidence interval [CI] = 0.85 to 0.89, P < .001) and positive predictive value (OR = 0.89, 95% CI = 0.80 to 0.99, P = .03). A non-statistically significant increase in overall sensitivity with CAD (OR = 1.06, 95% CI = 0.84 to 1.33, P = .62) was attributed to increased sensitivity for ductal carcinoma in situ (OR = 1.55, 95% CI = 0.83 to 2.91; P = .17), although sensitivity for invasive cancer was similar with or without CAD (OR = 0.96, 95% CI = 0.75 to 1.24; P = .77). CAD was not associated with higher breast cancer detection rates or more favorable stage, size, or lymph node status of invasive breast cancer.

CONCLUSION

CAD use during film-screen screening mammography in the United States is associated with decreased specificity but not with improvement in the detection rate or prognostic characteristics of invasive breast cancer.

The mammographic density of a mass is a significant predictor of breast cancer

PURPOSE

To determine whether the mammographic density of noncalcified solid breast masses is associated with malignancy and to measure the agreement between prospective and retrospective assessment.

MATERIALS AND METHODS

The institutional review board approved this study and waived informed consent. Three hundred forty-eight consecutive breast masses in 328 women who underwent image-guided or surgical biopsy between October 2005 and December 2007 were included. All 348 biopsy-proved masses were randomized and assigned to a radiologist who was blinded to biopsy results for retrospective assessment by using the Breast Imaging Reporting and Data System (retrospectively assessed data set). Clinical radiologists prospectively assessed the density of 180 of these masses (prospectively assessed data set). Pathologic result at biopsy was the reference standard. Benign masses were followed for at least 1 year by linking each patient to a cancer registry. Univariate analyses were performed on the retrospectively assessed data set. The association of mass density and malignancy was examined by creating a logistic model for the prospectively assessed data set. Agreement between prospective and retrospective assessments was calculated by using the κ statistic.

RESULTS

In the retrospectively assessed data set, 70.2% of high-density masses were malignant, and 22.3% of the isodense or low-density masses were malignant (P < .0001). In the prospective logistic model, high density (odds ratio, 6.6), irregular shape (odds ratio, 9.9), spiculated margin (odds ratio, 20.3), and age (β = 0.09, P < .0001) were significantly associated with the probability of malignancy. The κ value for prospective-retrospective agreement of mass density was 0.53.

CONCLUSION

High mass density is significantly associated with malignancy in both retrospectively and prospectively assessed data sets, with moderate prospective-retrospective agreement. Radiologists should consider mass density as a valuable descriptor that can stratify risk.

SUPPLEMENTAL MATERIAL

http://radiology.rsna.org/lookup/suppl/doi:10.1148/radiol.10100328/-/DC1.

Positive predictive value of mammography: comparison of interpretations of screening and diagnostic images by the same radiologist and by different radiologists

OBJECTIVE

The purpose of this study was to evaluate whether the positive predictive value (PPV) after a recommendation for biopsy differs when one as opposed to more than one radiologist performs the workup after abnormal findings are discovered at screening mammography.

MATERIALS AND METHODS

Using data in a mammography registry for the years 1996-2005, we identified 6,391 diagnostic examinations with a recommendation for biopsy that were performed on a day other than the day of the screening examination. The PPV after a recommendation for biopsy was calculated for two scenarios. In the first scenario, the radiologist interpreting the diagnostic images had interpreted the screening images. In the second scenario, the radiologist read diagnostic images after another radiologist had read the screening images. We used conditional logistic regression analysis to perform within-radiologist comparisons, controlling for covariates known to be associated with PPV after a recommendation for biopsy.

RESULTS

Of the screening examinations with positive findings, 2,335 (36.5%) were scenario 1, and 4,056 (63.5%) were scenario 2. We found no difference between the two scenarios with respect to PPV after a recommendation for biopsy when we controlled for age, breast density, family history of breast cancer, history of breast procedures, time since last mammogram, use of ultrasound at any point in the workup after abnormal results of screening mammography, and interval in days between the screening and diagnostic studies.

CONCLUSION

Who interprets the follow-up images after screening mammograms show abnormal findings does not appear to be an important factor influencing the wide variability in PPV among radiologists.

Effect of previous benign breast biopsy on the interpretive performance of subsequent screening mammography

BACKGROUND

Most breast biopsies will be negative for cancer. Benign breast biopsy can cause changes in the breast tissue, but whether such changes affect the interpretive performance of future screening mammography is not known.

METHODS

We prospectively evaluated whether self-reported benign breast biopsy was associated with reduced subsequent screening mammography performance using examination data from the mammography registries of the Breast Cancer Surveillance Consortium from January 2, 1996, through December 31, 2005. A positive interpretation was defined as a recommendation for any additional evaluation. Cancer was defined as any invasive breast cancer or ductal carcinoma in situ diagnosed within 1 year of mammography screening. Measures of mammography performance (sensitivity, specificity, and positive predictive value 1 [PPV1]) were compared both at woman level and breast level in the presence and absence of self-reported benign biopsy history. Referral to biopsy was considered a positive interpretation to calculate positive predictive value 2 (PPV2). Multivariable analysis of a correct interpretation on each performance measure was conducted after adjusting for registry, year of examination, patient characteristics, months since last mammogram, and availability of comparison film. Accuracy of the mammogram interpretation was measured using area under the receiver operating characteristic curve (AUC). All statistical tests were two-sided.

RESULTS

A total of 2,007,381 screening mammograms were identified among 799,613 women, of which 14.6% mammograms were associated with self-reported previous breast biopsy. Multivariable adjusted models for mammography performance showed reduced specificity (odds ratio [OR] = 0.74, 95% confidence interval [CI] = 0.73 to 0.75, P < .001), PPV2 (OR = 0.85, 95% CI = 0.79 to 0.92, P < .001), and AUC (AUC 0.892 vs 0.925, P < .001) among women with self-reported benign biopsy. There was no difference in sensitivity or PPV1 in the same adjusted models, although unadjusted differences in both were found. Specificity was lowest among women with documented fine needle aspiration-the least invasive biopsy technique (OR = 0.58, 95% CI = 0.55 to 0.61, P < .001). Repeating the analysis among women with documented biopsy history, unilateral biopsy history, or restricted to invasive cancers did not change the results.

CONCLUSIONS

Self-reported benign breast biopsy history was associated with statistically significantly reduced mammography performance. The difference in performance was likely because of tissue characteristics rather than the biopsy itself.

Frequency of malignancy seen in probably benign lesions at contrast-enhanced breast MR imaging: findings from ACRIN 6667

PURPOSE

To determine the frequency of malignancy in probably benign lesions seen at magnetic resonance (MR) screening of the contralateral breast in patients with known breast cancer enrolled in American College of Radiology Imaging Network (ACRIN) protocol 6667.

MATERIALS AND METHODS

ACRIN conducted a prospective multi-institutional MR imaging screening trial of the contralateral breast in women in whom breast cancer had been diagnosed recently. Each participating institution obtained institutional review board approval before patient accrual and was compliant with HIPAA. Informed consent was obtained from the patients. At enrollment, all women had negative clinical breast examination results and negative mammograms of the study breast. At image interpretation, radiologists scored lesions by using the Breast Imaging and Reporting and Data System (BI-RADS) lexicon. Of the 969 women who comprised the final study group, 106 were classified as having a BI-RADS category 3 lesion as their highest scoring lesion at MR imaging. There were 145 BI-RADS category 3 lesions in 106 patients.

RESULTS

In the 106 patients with at least one BI-RADS category 3 lesion, there were 37 masses (25.5%), 59 areas of nonmass enhancement (40.7%), and 47 foci of enhancement (32.4%). In two (1.4%) of these patients, no findings were reported. Eighty-three (78.3%) of the 106 patients had no evidence of malignancy in the study breast after 2 years of follow-up; the remaining 23 (21.7%) received a tissue diagnosis. Seventeen (16.0%) of the 106 patients elected to undergo biopsy. Biopsy was recommended in the remaining six patients (5.7%) on the basis of follow-up imaging findings. The biopsy results were benign in 18 (78%) of the 23 patients, whereas they showed atypical hyperplasia in two (9%). One (4%) of the 23 patients had ductal carcinoma in situ. Overall, malignancy was diagnosed in one (0.9%) of the 106 patients.

CONCLUSION

In a multi-institutional study, the frequency of malignancy in MR-detected BI-RADS category 3 lesions was 0.9% (95% confidence interval: 0.02%, 5.14%).

Time trends in radiologists' interpretive performance at screening mammography from the community-based Breast Cancer Surveillance Consortium, 1996-2004

PURPOSE

To examine time trends in radiologists' interpretive performance at screening mammography between 1996 and 2004.

MATERIALS AND METHODS

All study procedures were institutional review board approved and HIPAA compliant. Data were collected on subsequent screening mammograms obtained from 1996 to 2004 in women aged 40-79 years who were followed up for 1 year for breast cancer. Recall rate, sensitivity, and specificity were examined annually. Generalized estimating equation (GEE) and random-effects models were used to test for linear trend. The area under the receiver operating characteristic curve (AUC), tumor histologic findings, and size of the largest dimension or diameter of the tumor were also examined.

RESULTS

Data on 2,542,049 subsequent screening mammograms and 12,498 cancers diagnosed in the follow-up period were included in this study. Recall rate increased from 6.7% to 8.6%, sensitivity increased from 71.4% to 83.8%, and specificity decreased from 93.6% to 91.7%. In GEE models, adjusted odds ratios per calendar year were 1.04 (95% confidence interval [CI]: 1.02, 1.05) for recall rate, 1.09 (95% CI: 1.07. 1.12) for sensitivity, and 0.96 (95% CI: 0.95, 0.98) for specificity (P < .001 for all). Random-effects model results were similar. The AUC increased over time: 0.869 (95% CI: 0.861, 0.877) for 1996-1998, 0.884 (95% CI: 0.879, 0.890) for 1999-2001, and 0.891 (95% CI: 0.885, 0.896) for 2002-2004 (P < .001). Tumor histologic findings and size remained constant.

CONCLUSION

Recall rate and sensitivity for screening mammograms increased, whereas specificity decreased from 1996 to 2004 among women with a prior mammogram. This trend remained after accounting for risk factors. The net effect was an improvement in overall discrimination, a measure of the probability that a mammogram with cancer in the follow-up period has a higher Breast Imaging Reporting and Data System assessment category than does a mammogram without cancer in the follow-up period.

Recommendation for short-interval follow-up examinations after a probably benign assessment: is clinical practice consistent with BI-RADS guidance?

OBJECTIVE

American College of Radiology BI-RADS guidance suggests that women with a probably benign finding on mammography receive a management recommendation for short-interval follow-up; historically, radiologists in community practice have not consistently linked this assessment with short-interval follow-up. We evaluated predictors of discordance between probably benign assessments and short-interval follow-up recommendations.

MATERIALS AND METHODS

We linked data on 196 radiologists who completed a survey on demographic and practice patterns to 15,515 diagnostic mammograms they interpreted with probably benign assessments between 2001 and 2006. Patient characteristics were collected at the time of the mammography. Using logistic regression, we examined whether patient and radiologist characteristics were associated with the odds of short-interval follow-up recommendations (relative to a recommendation for normal follow-up, additional imaging evaluation, or biopsy or surgical consultation).

RESULTS

Overall, 90.9% of mammograms with probably benign findings were recommended for short-interval follow-up; 4.3% were recommended for normal follow-up, 3.0% for additional imaging, and 1.8% for biopsy or surgical consultation. Women with probably benign findings were less likely to receive a short-interval follow-up recommendation if they had extremely dense breasts versus almost entirely fatty breasts (odds ratio [OR], 0.61; 95% CI, 0.39-0.96) or had a breast lump versus no symptoms (OR, 0.55; 95% CI, 0.38-78). Radiologists were less likely to recommend short-interval follow-up if they had >/= 20 years of experience versus < 10 years of experience (OR, 0.57; 95% CI, 0.36-0.90) but more likely if they practiced primarily at an academic medical center versus other institutions (OR, 2.66; 95% CI, 1.14-6.21).

CONCLUSION

In contrast to older studies, the majority of probably benign assessments are now recommended for short-interval follow-up, but the probability of short-interval follow-up recommendations varies by patient and radiologist characteristics.

Modelling the cumulative risk of a false-positive screening test

The goal of a screening test is to reduce morbidity and mortality through the early detection of disease; but the benefits of screening must be weighed against potential harms, such as false-positive (FP) results, which may lead to increased healthcare costs, patient anxiety, and other adverse outcomes associated with diagnostic follow-up procedures. Accurate estimation of the cumulative risk of an FP test after multiple screening rounds is important for program evaluation and goal setting, as well as informing individuals undergoing screening what they should expect from testing over time. Estimation of the cumulative FP risk is complicated by the existence of censoring and possible dependence of the censoring time on the event history. Current statistical methods for estimating the cumulative FP risk from censored data follow two distinct approaches, either conditioning on the number of screening tests observed or marginalizing over this random variable. We review these current methods, identify their limitations and possibly unrealistic assumptions, and propose simple extensions to address some of these limitations. We discuss areas where additional extensions may be useful. We illustrate methods for estimating the cumulative FP recall risk of screening mammography and investigate the appropriateness of modelling assumptions using 13 years of data collected by the Breast Cancer Surveillance Consortium (BCSC). In the BCSC data we found evidence of violations of modelling assumptions of both classes of statistical methods. The estimated risk of an FP recall after 10 screening mammograms varied between 58% and 77% depending on the approach used, with an estimate of 63% based on what we feel are the most reasonable modelling assumptions.

The ACR BI-RADS experience: learning from history

The Breast Imaging Reporting and Data System (BI-RADS) initiative, instituted by the ACR, was begun in the late 1980s to address a lack of standardization and uniformity in mammography practice reporting. An important component of the BI-RADS initiative is the lexicon, a dictionary of descriptors of specific imaging features. The BI-RADS lexicon has always been data driven, using descriptors that previously had been shown in the literature to be predictive of benign and malignant disease. Once established, the BI-RADS lexicon provided new opportunities for quality assurance, communication, research, and improved patient care. The history of this lexicon illustrates a series of challenges and instructive successes that provide a valuable guide for other groups that aspire to develop similar lexicons in the future.

Predictors of interobserver agreement in breast imaging using the Breast Imaging Reporting and Data System

The Breast Imaging Reporting and Data System (BI-RADS) was introduced in 1993 to standardize the interpretation of mammograms. Though many studies have assessed the validity of the system, fewer have examined its reliability. Our objective is to identify predictors of reliability as measured by the kappa statistic. We identified studies conducted between 1993 and 2009 which reported kappa values for interpreting mammograms using any edition of BI-RADS. Bivariate and multivariate multilevel analyses were used to examine associations between potential predictors and kappa values. We identified ten eligible studies, which yielded 88 kappa values for the analysis. Potential predictors of kappa included: whether or not the study included negative cases, whether single- or two-view mammograms were used, whether or not mammograms were digital versus screen-film, whether or not the fourth edition of BI-RADS was utilized, the BI-RADS category being evaluated, whether or not readers were trained, whether or not there was an overlap in readers' professional activities, the number of cases in the study and the country in which the study was conducted. Our best multivariate model identified training, use of two-view mammograms and BI-RADS categories (masses, calcifications, and final assessments) as predictors of kappa. Training, use of two-view mammograms and focusing on mass description may be useful in increasing reliability in mammogram interpretation. Calcification and final assessment descriptors are areas for potential improvement. These findings are important for implementing policies in BI-RADS use before introducing the system in different settings and improving current implementations.

Observer Variability of the Breast Imaging Reporting and Data System (BI-RADS) Lexicon for Mammography

AIM: We aimed to determine the inter- and intra-observer variabilities between breast radiologists and a general radiologist in categorizing mammographic lesions using the Breast Imaging Reporting and Data System (BI-RADS), and to evaluate the effects of the histopathologic results on the variability. METHODS: Mammograms from 142 women who underwent biopsy were evaluated. 3 breast radiologists (2 with >10 years experience and 1 with 1 year experience) and 1 general radiologist retrospectively reviewed mammograms twice within an 8-week interval. Inter- and intra-observer variabilities were assessed with Cohen's kappa statistic, and the positive predictive value for final assessments was calculated. RESULTS: The intra-observer variability for mass and calcification assessments was moderate to almost perfect (kappa values: 0.41-1) for breast imagers and was fair to substantial for the general radiologist (kappa values: 0.21-0.8). Inter-observer agreement between the breast imagers was higher than between the breast and general radiologists. There was no apparent difference in agreement between observers for malignant and benign subgroups. CONCLUSIONS: The differences in intra- and inter-observer agreement between the breast imagers and the general radiologist affirm the utility of the BI-RADS lexicon. The histopathologic results of the lesions do not affect the agreement. BI-RADS is a simple and adequate tool for assessing mammograms, even after only 1 year of training.

Delay in diagnostic testing after abnormal mammography in low-income women

PURPOSE/OBJECTIVES

To identify factors associated with diagnostic delay after an incomplete or abnormal mammogram among women participating in a state mammography screening program.

RESEARCH APPROACH

Retrospective case-control design using bivariate and multivariate logistic regression analyses to explore the associations between age, race, ethnicity, marital status, breast cancer history, and self-reported breast symptoms and delay.

SETTING

A statewide program of free screening mammography for women who are under- or uninsured.

PARTICIPANTS

11,460 women enrolled in a free, statewide screening program from 2002-2006.

METHODOLOGIC APPROACH

Using the Tennessee Breast and Cervical Cancer Screening Program database, further analyses were conducted.

MAIN RESEARCH VARIABLES

The outcome measure was delay in completion of all diagnostic tests and was defined as women who did not complete testing within 60 days.

FINDINGS

Thirty-seven percent of women required follow-up, and of a subset used in the analysis, 30% experienced delay of more than 60 days. Controlling for marital status, age, and breast cancer history, women who experienced delay were more likely to be African American versus Caucasian (odds ratio [OR] = 1.45, 95% confidence interval [CI] = 1.13, 1.85) or Hispanic (OR = 0.72, 95% CI = 0.55, 0.93) and to have self-reported breast symptoms (OR = 1.50, 95% CI = 1.27, 1.77).

CONCLUSIONS

In a sample of women with low income needing mammography follow-up, delay was associated with three intrapersonal variables, potentially reducing the effectiveness of mammography screening for women who were African American, or Hispanic, or had self-reported breast symptoms.

INTERPRETATION

Nurses providing cancer screening examinations are uniquely positioned to assess the knowledge, beliefs, and resources of women using the program and to navigate women through barriers to completion. Knowledge of factors associated with delay is valuable for planning interventions and allocating program resources.

Variability in interpretive performance at screening mammography and radiologists' characteristics associated with accuracy

PURPOSE

To identify radiologists' characteristics associated with interpretive performance in screening mammography.

MATERIALS AND METHODS

The study was approved by institutional review boards of University of Washington (Seattle, Wash) and institutions at seven Breast Cancer Surveillance Consortium sites, informed consent was obtained, and procedures were HIPAA compliant. Radiologists who interpreted mammograms in seven U.S. regions completed a self-administered mailed survey; information on demographics, practice type, and experience in and perceptions of general radiology and breast imaging was collected. Survey data were linked to data on screening mammograms the radiologists interpreted between January 1, 1998, and December 31, 2005, and included patient risk factors, Breast Imaging Reporting and Data System assessment, and follow-up breast cancer data. The survey was returned by 71% (257 of 364) of radiologists; in 56% (205 of 364) of the eligible radiologists, complete data on screening mammograms during the study period were provided; these data were used in the final analysis. An evaluation of whether the radiologists' characteristics were associated with recall rate, false-positive rate, sensitivity, or positive predictive value of recall (PPV(1)) of the screening examinations was performed with logistic regression models that were adjusted for patients' characteristics and radiologist-specific random effects.

RESULTS

Study radiologists interpreted 1 036 155 screening mammograms; 4961 breast cancers were detected. Median percentages and interquartile ranges, respectively, were as follows: recall rate, 9.3% and 6.3%-13.2%; false-positive rate, 8.9% and 5.9%-12.8%; sensitivity, 83.8% and 74.5%-92.3%; and PPV(1), 4.0% and 2.6%-5.9%. Wide variability in sensitivity was noted, even among radiologists with similar false-positive rates. In adjusted regression models, female radiologists or fellowship-trained radiologists had significantly higher recall and false-positive rates (P < .05, all). Fellowship training in breast imaging was the only characteristic significantly associated with improved sensitivity (odds ratio, 2.32; 95% confidence interval: 1.42, 3.80; P < .001) and the overall accuracy parameter (odds ratio, 1.61; 95% confidence interval: 1.05, 2.45; P = .028).

CONCLUSION

Fellowship training in breast imaging may lead to improved cancer detection, but it is associated with higher false-positive rates.

Probabilistic computer model developed from clinical data in national mammography database format to classify mammographic findings

PURPOSE

To determine whether a Bayesian network trained on a large database of patient demographic risk factors and radiologist-observed findings from consecutive clinical mammography examinations can exceed radiologist performance in the classification of mammographic findings as benign or malignant.

MATERIALS AND METHODS

The institutional review board exempted this HIPAA-compliant retrospective study from requiring informed consent. Structured reports from 48 744 consecutive pooled screening and diagnostic mammography examinations in 18 269 patients from April 5, 1999 to February 9, 2004 were collected. Mammographic findings were matched with a state cancer registry, which served as the reference standard. By using 10-fold cross validation, the Bayesian network was tested and trained to estimate breast cancer risk by using demographic risk factors (age, family and personal history of breast cancer, and use of hormone replacement therapy) and mammographic findings recorded in the Breast Imaging Reporting and Data System lexicon. The performance of radiologists compared with the Bayesian network was evaluated by using area under the receiver operating characteristic curve (AUC), sensitivity, and specificity.

RESULTS

The Bayesian network significantly exceeded the performance of interpreting radiologists in terms of AUC (0.960 vs 0.939, P = .002), sensitivity (90.0% vs 85.3%, P < .001), and specificity (93.0% vs 88.1%, P < .001).

CONCLUSION

On the basis of prospectively collected variables, the evaluated Bayesian network can predict the probability of breast cancer and exceed interpreting radiologist performance. Bayesian networks may help radiologists improve mammographic interpretation.

Clinical application of the BI-RADS final assessment to breast sonography in conjunction with mammography

OBJECTIVE

The objective of our study was to report the results of classification of sonographic findings according to BI-RADS and to calculate the positive predictive value (PPV) for each BI-RADS assessment category.

SUBJECTS AND METHODS

We prospectively classified 4,668 breast sonograms according to BI-RADS final assessment category. Suspicious sonographic findings were divided into major and minor suspicious findings. Category 1 was normal and category 2 was a benign finding such as cyst or nodule with uniform and intense hyperechogenicity. A nodule neither category 2 nor category 4 or 5 was defined as category 3. A nodule with one or more suspicious findings, not category 5, was defined as category 4. A nodule with two or more major suspicious findings was defined as category 5.

RESULTS

Of the 4,668 cases, 321 cases failed to undergo follow-up of at least 1 year. The PPV was 0.1% in category 1 (3/2,191), 0% in category 2 (0/773), 0.8% in category 3 (6/737), 31.1% in category 4 (161/519), and 96.9% in category 5 (123/127). In palpable lesions (n = 751), the PPV was 2.2% in category 1 (2/93), 0.9% in category 3 (2/217), 54% in category 4 (107/198), and 98% in category 5 (98/100). In nonpalpable lesions (n = 3,596), the PPV was 0.05% in category 1 (1/2,098), 0.8% in category 3 (4/520), 16.8% in category 4 (54/321), and 92.6% in category 5 (25/27).

CONCLUSION

As with mammography, placing sonographic lesions into BI-RADS categories is useful for predicting the presence of malignancy.

Accuracy of short-interval follow-up mammograms by patient and radiologist characteristics

OBJECTIVE

The purpose of our study was to examine the accuracy of short-interval follow-up mammograms and evaluate patient and radiologist characteristics associated with accuracy.

MATERIALS AND METHODS

We evaluated 45,007 initial short-interval follow-up mammograms from the Breast Cancer Surveillance Consortium interpreted 3-9 months after a probably benign assessment on a screening or diagnostic examination between 1994 and 2004. We linked these mammograms with patient characteristics and breast cancer diagnoses within 12 months. A subset of short-interval follow-up examinations (n = 13,907) was merged with radiologist characteristics collected from survey data from 130 interpreting radiologists. Using logistic regression, we fit generalized estimating equations to model sensitivity and specificity of short-interval follow-up mammograms by patient and radiologist characteristics.

RESULTS

For every 1,000 women, 8.0 women (0.8%) were diagnosed with breast cancer within 6 months and 11.3 (1.1%) within 12 months. Sensitivity was 83.3% (95% CI, 79.4-87.3%) for cancers diagnosed within 6 months and 60.5% (56.2-64.7%) for those diagnosed within 12 months. Specificity was 97.2% (96.9-97.6%) at 6 months and 97.3% (96.9-97.6%) at 12 months. Sensitivity at 12 months increased among women with unilateral short-interval follow-up mammograms (odds ratio, 1.56 [95% CI, 1.06-2.29]) and when the interpreting radiologist spent more than 10 hours a week in breast imaging (odds ratio, 3.25 [1.00-10.52]).

CONCLUSION

Initial short-interval follow-up mammography examinations had a lower sensitivity for detecting breast cancer within 12 months than other diagnostic mammograms (61% for short-interval follow-up vs 80% for diagnostic mammograms reported in the literature). However, sensitivity within the 6-month interval that is usually recommended for subsequent follow-up was 83%. Accuracy of short-interval follow-up mammograms was influenced by few patient and radiologist characteristics.

Radiologist characteristics associated with interpretive performance of diagnostic mammography

BACKGROUND

Extensive variability has been noted in the interpretive performance of screening mammography; however, less is known about variability in diagnostic mammography performance.

METHODS

We examined the performance of 123 radiologists who interpreted 35895 diagnostic mammography examinations that were obtained to evaluate a breast problem from January 1, 1996, through December 31, 2003, at 72 facilities that contribute data to the Breast Cancer Surveillance Consortium. We modeled the influence of radiologist characteristics on the sensitivity and false-positive rate of diagnostic mammography, adjusting for patient characteristics by use of a Bayesian hierarchical logistic regression model.

RESULTS

The median sensitivity was 79% (range = 27%-100%) and the median false-positive rate was 4.3% (range = 0%-16%). Radiologists in academic medical centers, compared with other radiologists, had higher sensitivity (88%, 95% confidence interval [CI] = 77% to 94%, versus 76%, 95% CI = 72% to 79%; odds ratio [OR] = 5.41, 95% Bayesian posterior credible interval [BPCI] = 1.55 to 21.51) with a smaller increase in their false-positive rates (7.8%, 95% CI = 4.8% to 12.7%, versus 4.2%, 95% CI = 3.8% to 4.7%; OR = 1.73, 95% BPCI = 1.05 to 2.67) and a borderline statistically significant improvement in accuracy (OR = 3.01, 95% BPCI = 0.97 to 12.15). Radiologists spending 20% or more of their time on breast imaging had statistically significantly higher sensitivity than those spending less time on breast imaging (80%, 95% CI = 76% to 83%, versus 70%, 95% CI = 64% to 75%; OR = 1.60, 95% BPCI = 1.05 to 2.44) with non-statistically significant increased false-positive rates (4.6%, 95% CI = 4.0% to 5.3%, versus 3.9%, 95% CI = 3.3% to 4.6%; OR = 1.17, 95% BPCI = 0.92 to 1.51). More recent training in mammography and more experience performing breast biopsy examinations were associated with a decreased threshold for recalling patients, resulting in similar statistically significant increases in both sensitivity and false-positive rates.

CONCLUSIONS

We found considerable variation in the interpretive performance of diagnostic mammography across radiologists that was not explained by the characteristics of the patients whose mammograms were interpreted. This variability is concerning and likely affects many women with and without breast cancer.

The Probably Benign Assessment

The probably benign assessment (category 3 in the Breast Imaging Reporting and Data System) is associated with a less than 2% probability of malignancy. Its use in mammography is well supported by robust data from various large-scale prospective studies. Use of the probably benign assessment for lesions visible only at ultrasound or MR imaging is much less well established. This article examines in depth the use of the probably benign assessment: which lesions should be assessed as probably benign, the published evidence supporting such use, pitfalls in misuse, and areas of potentially expanded use that currently are under investigation.

Influence of computer-aided detection on performance of screening mammography

BACKGROUND

Computer-aided detection identifies suspicious findings on mammograms to assist radiologists. Since the Food and Drug Administration approved the technology in 1998, it has been disseminated into practice, but its effect on the accuracy of interpretation is unclear.

METHODS

We determined the association between the use of computer-aided detection at mammography facilities and the performance of screening mammography from 1998 through 2002 at 43 facilities in three states. We had complete data for 222,135 women (a total of 429,345 mammograms), including 2351 women who received a diagnosis of breast cancer within 1 year after screening. We calculated the specificity, sensitivity, and positive predictive value of screening mammography with and without computer-aided detection, as well as the rates of biopsy and breast-cancer detection and the overall accuracy, measured as the area under the receiver-operating-characteristic (ROC) curve.

RESULTS

Seven facilities (16%) implemented computer-aided detection during the study period. Diagnostic specificity decreased from 90.2% before implementation to 87.2% after implementation (P<0.001), the positive predictive value decreased from 4.1% to 3.2% (P=0.01), and the rate of biopsy increased by 19.7% (P<0.001). The increase in sensitivity from 80.4% before implementation of computer-aided detection to 84.0% after implementation was not significant (P=0.32). The change in the cancer-detection rate (including invasive breast cancers and ductal carcinomas in situ) was not significant (4.15 cases per 1000 screening mammograms before implementation and 4.20 cases after implementation, P=0.90). Analyses of data from all 43 facilities showed that the use of computer-aided detection was associated with significantly lower overall accuracy than was nonuse (area under the ROC curve, 0.871 vs. 0.919; P=0.005).

CONCLUSIONS

The use of computer-aided detection is associated with reduced accuracy of interpretation of screening mammograms. The increased rate of biopsy with the use of computer-aided detection is not clearly associated with improved detection of invasive breast cancer.