Detection of Bilateral Breast Cancer at Biennial Screening Mammography in the Netherlands: A Population-based Study

Introduction

Mammography is widely used to screen asymptomatic women for breast cancer. Several studies have shown that screening mammography reduces breast cancer mortality by helping detect breast malignancies at an early stage (1,2). It is estimated that about 20%–30% of breast cancers emerge as interval cancers (cancers diagnosed after a negative screening examination [defined as no recommendation for referral] and before the next scheduled biennial screening round). Of these interval cancers, 18%–29% are considered to be missed at screening mammography (3–5).

Breast cancer may be diagnosed bilaterally. The overall frequency of bilateral breast cancer has been shown to range from 4% to 10%. The frequency of bilateral synchronous breast cancer is more uncommon and has been reported to range from 1% to 3% (6–9). Radiologic improvements in breast cancer detection, such as the use of magnetic resonance (MR) imaging, have raised interest in bilateral breast cancer because of an increase in the detection of contralateral breast cancer. In women with newly diagnosed breast cancer, MR imaging has been reported to enable detection of contralateral cancer in 3%–5% of patients (10–12).

Studies about the sensitivity of breast radiology in the detection of bilateral breast cancers have mainly focused on symptomatic patients (12,13). Although screening mammography programs are increasingly implemented throughout the world, to our knowledge no data have been published with regard to bilateral breast cancer detection at screening mammography. Therefore, the aim of the current study was to determine the incidence of bilateral breast cancer at biennial screening mammography and to assess the sensitivity of screening in the detection of bilateral breast cancers.

Materials and Methods

Screening Procedure and Imaging Technique

We included all 302 196 screening mammograms obtained in 80 466 women aged 50–75 years at one fixed center and one mobile unit in the southern breast cancer screening region of the Netherlands (Bevolkings Onderzoek Borstkanker Zuid) between May 1, 1998, and July 1, 2008. Women participating in the nationwide Dutch screening program are asked to give written informed consent for the use of their data for scientific purposes, and all but three women in the study period had given this informed consent. According to the Dutch Central Committee on Research Involving Human Subjects, approval by our local institutional review board was not required for this study.

Two-view mammography (mediolateral oblique and craniocaudal views) of each breast was performed at the initial examination (ie, the first time the women were screened). Subsequent screening examinations primarily consisted of only one-view mammography (mediolateral oblique view). An additional craniocaudal view was obtained in 43.7% of the subsequent screening examinations (115 399 of 264 043 subsequent screening mammograms). Indications for this second view included any changes in mammographic findings at screening, a complicated judgment regarding interpretation owing to dense fibroglandular tissue, a history of breast surgery, and an interval of more than 2 years since the previous screening examination.

All mammographic examinations were performed by specialized screening mammography technicians. Mammograms were obtained by using commercially available units (at fixed center: Performa, Oldelft, Tuusula, Finland; at mobile unit: Alfa RT, Oldelft, and, from 2004 on, Performa). Dedicated mammography screens were used (at fixed center: Mamoray MR-R, Agfa, Schroenhausen, Germany; at mobile unit: Mamoray MR-S, Agfa, and, from 2004 on, Mamoray MR-R). Both dedicated film (both centers: Mamoray HDR; Agfa, Mortsel, Belgium) and extended-cycle dedicated processing were used.

Image Interpretation and Referral

During the study period, 11 screening radiologists participated in the screening program. Each radiologist read more than 5000 mammograms annually. All screening mammograms were assessed independently by two certified screening radiologists. From May 1998 until January 2001, discrepant readings for which the two screening radiologists did not reach consensus about referral were presented to a panel of three radiologists. Women were referred for further analysis when at least one of the three panel radiologists recommended referral. Arbitration panel reading was abandoned in January 2001 because this reading strategy did not enable detection of all lesions that subsequently proved to be malignant (14). From 2001 to 2002, discrepant readings for which the two screening radiologists did not reach consensus were routinely referred for further analysis. From 2003 on, the mammograms were independently read by two mammography screening technicians and independent double reading was performed by the radiologists (3,15). A woman was referred for additional work-up if both screening radiologists considered the mammogram to be positive or, in the case of discrepant readings, if at least one radiologist considered referral necessary after consensus meeting. In addition, mammograms that the technologists had considered to be positive but for which the radiologists did not think referral was necessary were reviewed by two screening radiologists. A woman was referred if, on review, at least one of the radiologists considered work-up to be necessary. A woman was not referred for further diagnostic work-up in case of normal findings, benign mammographic findings (eg, lymph nodes, calcified fibroadenoma, lipoma, and vascular calcifications), or nonspecific minimal signs (eg, vague area of density with an incomplete sharp border and a diameter of 5–30 mm [density comparable to that of glandular tissue], fewer than six clustered nonspecific microcalcifications, or subtle architectural distortions that include asymmetric glandular tissue) (16). Lesions determined to be suspicious or malignant at screening mammography were classified by radiologists into one of the following categories: (a) suspicious high-density mass (spiculated mass or a mass with indistinct borders); (b) suspicious microcalcifications (pleomorphic, branching, or amorphous and/or indistinct microcalcifications); (c) high-density mass in combination with microcalcifications; (d) architectural distortion; or (e) asymmetry.

Diagnostic Work-up

If a suspicious or malignant lesion was seen at screening mammography, the woman was referred by her general practitioner to a surgical oncologist. After the surgeon performed a physical examination, clinical two-view mammography of each breast was performed; local compression or magnification mammograms were obtained at the radiologist’s discretion. The radiologist then decided whether breast ultrasonography (US), MR imaging, and/or biopsy should be performed. At diagnostic work-up, radiologists classified the radiologic findings according to the American College of Radiology Breast Imaging Reporting and Data System (BI-RADS) (17,18). Patients with probably benign breast imaging results (BI-RADS category 3) or benign biopsy results at work-up usually underwent their first follow-up mammographic examination at 6 months. Depending on the findings at follow-up mammography, a repeat mammogram may be obtained at a later stage to exclude malignancy.

Screening Follow-up Procedure and Review of Bilateral Breast Cancers

During a follow-up period of about 2 years (until the next biennial screening examination), we collected screening mammography findings, clinical data, additional breast imaging reports, biopsy results, and breast surgery reports from all women with a positive screening result (ie, those who required additional evaluation). Procedures for the detection of interval cancers have been described previously (14,19); most interval cancers were identified by linking the screening records to the regional cancer registry (Eindhoven Cancer Registry) and regional pathology laboratories.

Two screening radiologists (L.E.M.D., F.H.J., with 13 and 15 years of screening experience, respectively) reviewed the screening and clinical mammograms from all women in whom bilateral breast cancer was diagnosed within 2 years after the latest screening examination. They retrospectively determined whether tumors that were not detected with screening mammography (first and/or second examinations) had been missed, had shown a nonspecific minimal sign, or had been mammographically occult at screening. At review, the radiologists were informed about the bilateral nature of the malignancies but had no information about whether the cancers had been detected at screening or had emerged as interval cancers. The radiologists also assessed the breast density on the most recent screening mammograms according to BI-RADS (18) and classified the mammographic abnormality of each visible cancer into one of the five categories described previously. Finally, to determine the cause of a delay in breast cancer diagnosis after referral, the radiologists also reviewed the initial clinical mammograms of those women in whom contralateral breast cancer was diagnosed more than 3 months after referral. These clinical mammograms were also classified in accordance with BI-RADS. At review, the radiologists were initially blinded to each other’s referral opinion, and discrepant assessments were followed by consensus reading.

Synchronous and metachronous bilateral breast cancers were defined as bilateral cancers in which the cancer in the contralateral breast was diagnosed within 3 months or more than 3 months, respectively, after the diagnosis of the index cancer for which the woman had been referred.

Statistical Analysis

The primary outcome measures were the incidence of bilateral breast cancer at biennial screening mammography and the sensitivity of screening mammography in the detection of these cancers. Descriptive statistics were obtained by using software (SPSS, version 17.0; SPSS, Chicago, Ill). The χ² and Fisher exact tests were used to test the differences in breast density and histologic tumor characteristics between referred women whose bilateral cancer had or had not been detected at screening. The significance level was set at P = .05.

Results

Overall Screening Outcome

A total of 302 196 screening mammograms were obtained in 80 466 women between May 1, 1998, and July 1, 2008. Altogether, 3801 (1.3%) screening mammograms were referred for further diagnostic assessment. Assessment of women with positive screening mammograms was performed at 18 hospitals; at least 500 women were evaluated at four of the hospitals. At follow-up, breast cancer was diagnosed in 1555 women, yielding an overall cancer detection rate of 5.1 per 1000 women screened and a true-positive referral rate of 40.9%. At 2-year follow-up, interval cancers had been diagnosed in 585 women. The overall screening sensitivity for breast cancer detection, irrespective of the uni- or bilateral presence of breast cancer, was 72.7% (1555 of 2140 women; 95% confidence interval [CI]: 70.9%, 74.6%).

Bilateral breast cancer was diagnosed in 41 of the 1555 women with screening-detected cancers and 11 of the 585 with interval cancers (ie, 2.4% [52 of 2140 women] of all cases), resulting in a bilateral breast cancer detection rate of 0.17 per 1000 screening mammograms.

Screening Sensitivity for Bilateral Breast Cancer Detection

Ten of the 52 women with bilateral breast cancer (including the women with bilateral interval cancers) had been referred for the assessment of a bilateral abnormality noted at screening mammography, resulting in a screening sensitivity for bilateral breast cancer of 19% (10 of 52 women; 95% CI: 8.5%, 29.9%). Another 31 women with bilateral breast cancer had been referred for a unilateral abnormality alone; the contralateral tumor was detected within 3 months after referral in 16 women (synchronous bilateral cancer) and more than 3 months after referral in 15 (metachronous bilateral cancer). The remaining 11 bilateral breast cancers were interval cancers (Figure).

Download as PowerPointOpen in Image Viewer

In the 52 women with bilateral breast cancer (104 cancers), 51 tumors were detected at screening mammography and 53 were not. At blinded review, 18 of the 53 (34%) tumors not detected at screening were considered to have been missed at screening mammography, 11 (21%) had shown a minimal sign, and 24 (45%) had been mammographically occult.

Characteristics of Bilateral Cancers in Referred Women

At review of the screening mammograms of the 16 synchronous contralateral cancers not detected at screening, the tumor was considered mammographically occult in one case (6.3%), to have shown a minimal sign in four cases (25%), and to have been missed in 11 cases (69%). Review of the screening mammograms and initial clinical mammograms (ie, the first mammogram obtained after referral) of the 15 metachronous contralateral cancers not detected at screening showed a suspicious lesion in four cases, a lesion with minimal signs in one case, and no mammographic abnormalities in 10 cases. The diagnostic delay of the five cancers that had been visible at screening was due to misclassification of BI-RADS categories (BI-RADS category 3 instead of BI-RADS categories 4 or 5) with the initial clinical mammograms in three cases and to a false-negative core biopsy of a BI-RADS category 4 lesion in one case. In one patient, a cluster of microcalcifications, located at the periphery of the screening mammogram and defined as showing minimal signs of abnormality at review, was not depicted at initial clinical mammography. Biopsy was performed when the microcalcifications were properly visualized at follow-up mammography 6 months after breast-conserving surgery. The diagnostic delay of these five contralateral cancers ranged from 6 to 17 months.

The 10 metachronous contralateral cancers not detected at screening, which were considered mammographically occult on both the latest screening mammogram and the initial clinical mammogram, were diagnosed 6–23 months after the first cancer (mean, 14 months). These cancers were diagnosed after they manifested as a palpable mass (three cases) or as a radiologic abnormality at follow-up of the first cancer (seven cases).

A BI-RADS breast density category of less than 2 was assigned in 38 of the 51 women (75%) with screening-detected cancer (51 tumors, of which 41 were index tumors and 10 were screening-detected contralateral tumors) and 20 of the 31 (65%) with interval cancer (31 contralateral cancers); the difference was not significant (P = .23).

The 41 bilateral breast cancers (82 cancers) in referred women comprised 55 invasive ductal cancers (67%), 19 invasive lobular cancers (23%), and eight in situ ductal cancers (9.7%). There was no significant difference in detection of invasive ductal cancers compared with invasive lobular cancers (71% [39 of 55 women] vs 47% [nine of 19 women], respectively; P = .06).

Characteristics of Bilateral Interval Cancers

All but one of the 11 bilateral interval cancers were detected synchronously shortly after clinical presentation. The predominant index lesion manifested as a palpable abnormality (eight cases) or nipple retraction (three cases). The mean time between the last screening mammogram and the clinical presentation of the interval cancers was 11 months (range, 5–9 months). Nine index lesions were classified as BI-RADS category 4 or 5 at clinical mammography, whereas two index lesions with normal findings at clinical mammography were found at breast US of a palpable mass. The 10 synchronously diagnosed contralateral cancers were detected at clinical mammography (seven cancers), breast US of a palpable mass (two cancers), or MR imaging (one cancer). The single metachronously diagnosed contralateral cancer showed a BI-RADS category 3 density on the initial clinical mammogram, and the malignant nature of this lesion was confirmed with percutaneous biopsy at 12-month follow-up.

Review showed that nine of the 22 interval cancers (41%) had either been missed at the latest screening examination (three cases, of which two were densities and one was architectural distortion) or demonstrated a minimal sign (six cases, of which four were densities and two were clusters of nonspecific microcalcifications). The 22 interval cancers comprised 16 invasive ductal cancers (73%) and six invasive lobular cancers (27%).

The total group of 52 bilateral breast cancers (104 cancers) comprised 71 invasive ductal cancers (68.3%), 25 invasive lobular cancers (24.0%), and eight ductal cancers in situ (7.7%). There was no significant difference in the detection of invasive ductal cancers compared with that of invasive lobular cancers (55% [39 of 71 cases] vs 36% [nine of 25 cases], respectively; P = .08).

Discussion

To our knowledge, this is the first study to address the detection of bilateral breast cancers at screening mammography. In our population, bilateral breast cancer comprised 2.4% of all screening-detected and interval cancers. Variations in cancer detection rates at screening programs may partly be explained by differences in characteristics of screened women and differences in screening protocols. Nevertheless, the overall 72.7% screening sensitivity in our study is in line with that reported in other series (20,21).

Although controversy exists as to the exact effect bilateral breast cancer has on survival, several studies have shown that the prognosis of women with bilateral breast cancer tends to be worse than that of women with unilateral cancer (8,9,22); a delay in the diagnosis of breast cancer will further worsen the prognosis (23). It may therefore be even more important to timely detect bilateral breast cancer at screening. Unfortunately, the screening sensitivity of 19% for the detection of bilateral breast cancer in our study was frankly disappointing. Several factors may have contributed to this low sensitivity. First, some of the contralateral cancers could simply not be detected at screening because they were mammographically occult. However, more often they were missed as a result of perception error (mammographic abnormality not detected at screening) or interpretation error (mammographic abnormality misinterpreted at screening). Perception errors and interpretation errors both account for approximately half of missed breast cancers visible at screening (21). Review showed that most of the contralateral breast cancers in women who had been referred for a unilateral lesion only (and in whom the contralateral tumor was detected within 3 months) had been missed at screening; only one contralateral cancer had been mammographically occult. An important cause for missing contralateral cancers may be the happy eye syndrome, or satisfaction of search (24). The observation of a suspicious lesion may mislead a screening radiologist into not looking carefully for other, contralateral lesions. The sensitivity of mammography in breast cancer detection also depends on the histologic characteristics of the tumor. Invasive lobular breast cancer accounts for approximately 8%–10% of breast cancers in the general population but is found more frequently in bilateral breast cancers (25). Compared with invasive ductal cancers, invasive lobular cancers are more difficult to detect at mammography because these tumors more commonly manifest as subtle architectural distortions or focal asymmetric densities resembling those of normal breast parenchyma or show no mammographic abnormalities at all (26). We also observed a relatively high percentage of invasive lobular cancers (24%) in women with bilateral interval cancers or bilateral cancers diagnosed after referral owing to a unilateral or bilateral screening abnormality. The detection of invasive lobular cancer in our series, however, was not worse than that of invasive ductal cancer, which may be explained by the relatively small number of bilateral breast cancers in our study.

Five of the referred women experienced a 6–17-month delay in the diagnosis of the contralateral breast cancer despite the fact that the mammographic abnormality was visible on the latest screening mammogram and/or the initial clinical mammogram. Previous studies have shown that the work-up of referred patients should be improved to prevent an unnecessary delay in cancer diagnosis (19,27,28). These diagnostic delays may be due, especially, to misinterpretation of mammographic lesions as benign or probably benign at clinical mammography.

In women with bilateral interval cancers, one of the malignancies was found at contrast material–enhanced MR imaging. Contrast-enhanced MR imaging has emerged as a highly sensitive imaging modality for the detection of synchronous contralateral cancers or high-risk lesions in patients with newly diagnosed breast cancer (29). Compared with conventional mammography, contrast-enhanced MR imaging has a higher sensitivity for detecting invasive lobular cancers and may be used for a more accurate determination of tumor size, tumor multifocality, and pectoral muscle invasion by tumor growth (30,31). Although randomized controlled trials are needed to establish the long-term effects of contrast-enhanced MR imaging in women newly affected by breast cancer (32), the use of this diagnostic modality could probably have led to an earlier diagnosis of the metachronous cancers in our study.

Our study has certain limitations. The referral rate in the Dutch nationwide screening program is much lower than that in other screening programs (33). We are not able to quantify the effect of this lower referral rate on the sensitivity of detecting breast cancer, regardless of whether it is unilateral or bilateral. Patients with nonspecific minimal signs, which are present on approximately 11% of screening mammograms, were not referred for further diagnostic assessment because these lesions have a low cancer risk of 0.5% and a favorable tumor stage if malignant (16). After a critical reconsideration of nonspecific minimal signs and the implementation of screening BI-RADS, an increased referral rate is currently observed in the Dutch breast screening program (34). Screening outcome parameters will be influenced by the screening interval used in screening programs. Many European programs, including the Dutch one, offer biennial screening for women aged 50–75 years. In contrast, women in the United Kingdom are screened every 3 years and those in the United States are usually offered annual screening (33). In contrast to programs where two-view mammography is routinely performed, the Dutch screening program offers single-view mammography (mediolateral oblique) at subsequent screening examinations. An additional craniocaudal view is obtained only if indicated, and the detection of breast cancer, whether unilateral or bilateral, may have been hampered by this limited use of two-view mammography at subsequent screening mammography in our study (35). The applicability of our study may be somewhat limited by the fact that only mammograms obtained with analog screening units were included; most mammography units are now digital. In the Netherlands, the conversion from analog to digital screening has recently been completed for the nationwide breast screening program. Despite the similar or even higher cancer detection rate found at digital screening when compared with analog screening (36,37), we currently cannot predict the effect of digital screening on the detection of bilateral breast cancers. Finally, the knowledge of a high probability of bilateral breast cancer at review of the screening and clinical mammograms in our population is not a true reflection of daily screening practice and may have introduced detection bias.

In summary, we found that screening mammography has a low sensitivity for the detection of bilateral breast cancer. Both screening radiologists and clinical radiologists should pay utmost attention to the contralateral breast to detect bilateral malignancies without diagnostic delay. Although bilateral breast cancers comprise a small proportion of all screening-detected and interval cancers, a timely diagnosis of the bilateral disease is important to prevent a worsening in survival prognosis.

Disclosures of Potential Conflicts of Interest: W.S.P. Financial activities related to the present article: none to disclose. Financial activities not related to the present article: none to disclose. Other relationships: none to disclose. L.E.M.D. Financial activities related to the present article: none to disclose. Financial activities not related to the present article: none to disclose. Other relationships: none to disclose. J.H.G. Financial activities related to the present article: none to disclose. Financial activities not related to the present article: none to disclose. Other relationships: none to disclose. A.C.V. Financial activities related to the present article: none to disclose. Financial activities not related to the present article: none to disclose. Other relationships: none to disclose. F.H.J. Financial activities related to the present article: none to disclose. Financial activities not related to the present article: none to disclose. Other relationships: none to disclose. M.J.H.H.H. Financial activities related to the present article: none to disclose. Financial activities not related to the present article: none to disclose. Other relationships: none to disclose. M.W.J.L. Financial activities related to the present article: none to disclose. Financial activities not related to the present article: none to disclose. Other relationships: none to disclose.

Author Contributions

Author contributions: Guarantors of integrity of entire study, W.S.P., F.H.J., M.J.H.H.H.; 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, W.S.P., L.E.M.D., M.J.H.H.H.; clinical studies, W.S.P., L.E.M.D., F.H.J., M.J.H.H.H.; statistical analysis, W.S.P., L.E.M.D., J.H.G., M.J.H.H.H., M.W.J.L.; and manuscript editing, all authors