NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

Bruening W, Uhl S, Fontanarosa J, et al. Noninvasive Diagnostic Tests for Breast Abnormalities: Update of a 2006 Review [Internet]. Rockville (MD): Agency for Healthcare Research and Quality (US); 2012 Feb. (Comparative Effectiveness Reviews, No. 47.)

Cover of Noninvasive Diagnostic Tests for Breast Abnormalities

Noninvasive Diagnostic Tests for Breast Abnormalities: Update of a 2006 Review [Internet].

Show details

Introduction

Background

Breast Cancer

Breast cancer is the second most common malignancy of women.1 The American Cancer Society estimates that in the United States in 2010, 54,010 women were diagnosed with new cases of in situ cancer, 207,090 women were newly diagnosed as having invasive breast cancer, and there were 39,840 deaths due to this disease.1 In the general population, the cumulative risk of being diagnosed with breast cancer by age 70 is estimated to be 6 percent (lifetime risk of 13%).93,94

The most common type of breast cancer, accounting for over 85 percent of cases diagnosed, is ductal carcinoma.95 Ductal carcinoma arises within the ducts of the breast from the cells lining the ducts. Early-stage breast cancer confined to the inside of the duct is referred to as ductal carcinoma in situ (DCIS). Later stages of ductal carcinoma that have invaded or broken through the walls of the ducts into nearby tissues may be referred to as invasive or infiltrating ductal carcinoma. Cases of invasive ductal carcinoma that are found to be well-differentiated specific subtypes (such as mucinous, medullary, tubular, or papillary) are much rarer than the common “otherwise not specified” type of invasive ductal carcinoma.

Another type of invasive carcinoma is lobular carcinoma. Lobular carcinoma is similar to ductal carcinoma, first arising in the terminal ducts of the lobules and then invading through the walls of the ducts and invading nearby tissues. Other rare types of potentially life-threatening breast tumors include papillary carcinoma, inflammatory breast cancer, and sarcomas, among others.95

A number of different breast lesions have been described that, while not malignant, are believed to predispose to the development of invasive breast carcinomas. These lesions include atypical ductal hyperplasia (ADH), papillary lesions, radial scars, atypical lobular hyperplasia (ALH), and lobular carcinoma in situ (LCIS).96 However, the most commonly reported breast abnormalities diagnosed after screening are benign: benign fibrocystic changes, cysts, and benign fibroadenomas.

Breast Cancer Diagnosis

Breast cancer is usually first detected by feeling a lump on physical examination (either self-examination or an exam conducted by a health practitioner) or by observing an abnormality during x-ray screening mammography. Survival rates depend on the stage of disease at diagnosis. At stage 0 (carcinoma in situ) the 5-year survival rate is close to 100 percent. The five-year survival rate for women with stage IV (cancer that has spread beyond the breast) is only 23 percent.1 Because early breast cancer is asymptomatic, the only way to detect it is through screening of asymptomatic women. Mammography is a widely accepted and used method for breast cancer screening.24 Meta-analyses of large clinical trials have demonstrated that mammography screening reduces breast cancer mortality.97,98

Mammography uses x-rays to examine the breast for clusters of microcalcifications, circumscribed and dense masses, masses with indistinct margins, architectural distortion compared with the contralateral breast, or other abnormal structures. The United States Preventive Services Task Force (USPSTF) has recently recommended routine screening mammography every two years for women aged 50 to 74, with decisions to screen women under the age of 50 made on an individual basis.4 After identification of a possible abnormality on screening mammography or physical examination, women typically undergo additional imaging studies (diagnostic mammography and/or ultrasound) and a physical examination. If these studies suggest the abnormality may be malignant, a biopsy of the suspicious area may be recommended.

The American College of Radiology has created a standardized system for reporting the results of mammography, the Breast Imaging-Reporting and Data System (BI-RADS®).99101 There are seven categories of assessment, each with an accompanying clinical management recommendation:

0.

Need additional imaging evaluation and/or prior mammograms for comparison

1.

Negative

2.

Benign finding

3.

Probably benign finding. Initial short interval followup suggested.

4.

Suspicious abnormality. Biopsy should be considered.

5.

Highly suggestive of malignancy. Appropriate action should be taken.

6.

Known biopsy-proven malignancy. Appropriate action should be taken.

Noninvasive breast imaging tests have multiple uses, including image-guidance of biopsy procedures, searching for multifocal lesions in a woman diagnosed with or at high risk of breast cancer, and screening women at high risk of breast cancer. This evidence review specifically focuses only on the use of noninvasive imaging studies that can be conducted after the discovery of a possible abnormality on screening mammography or physical examination-studies intended to guide patient management decisions. In other words, these studies are not intended to provide a final diagnosis as to the nature of the breast lesion; rather, they are intended to provide additional information about the nature of the lesion such that women can be appropriately triaged into “biopsy/watchful waiting/return to normal screening intervals” care pathways.

It is important to accurately triage women into the correct care pathway. Women with readily treatable breast cancers who get incorrectly triaged into “return to normal screening care pathways” may experience a significant delay in diagnosis and treatment of the cancer. However, the majority of women who are recalled for further assessment after a screening mammogram do not have cancer. Elmore et al. estimated that the cumulative risk for a woman having a false-positive finding on screening mammography is close to 50 percent after 10 years of yearly screenings.5 In addition, diagnostic mammography performed after a mammographic screening recall often leads to identification of a “probably benign” (BI-RADS 3) lesion. Women with “probably benign” lesions are usually referred for short-interval repeat mammography examinations, meaning that they wait for three to six months before being re-tested. Many women experience considerable emotional distress and anxiety during this waiting period.102 If an available noninvasive diagnostic test could assist clinicians in evaluating women recalled for further investigation after mammographic screening, namely, in assisting in accurately distinguishing between “benign,” “probably benign,” and “probably not benign” lesions, then some women could avoid having to spend several months wondering if they have cancer or not.

The majority of women who traditionally have been referred for biopsy also do not have cancer. Studies in the U.S. generally find that only 20 to 30 percent of women who undergo biopsy are diagnosed with breast cancer.6,103 Exposing large numbers of women who do not have cancer to invasive procedures may be considered an undesirable medical practice. In conclusion, current workup after recall results in a large number of false-positives. If additional tests could reduce the false-positive rate without increasing the false-negative rate then it is possible that women could benefit from adding these tests to standard workup.

Because there are no available studies that directly evaluate whether women benefit from additional noninvasive imaging, we addressed this important question indirectly. First we evaluated the accuracy of the imaging tests in distinguishing between “benign” and “malignant” breast lesions. Inaccurate tests will lead to sub-optimal management decisions and less than desirable patient outcomes. The accuracy of the noninvasive imaging tests was primarily measured in terms of sensitivity and specificity. Sensitivity is a measure of how accurately the test can identify women with cancer; specificity is a measure of how accurately the test can identify women who do not have cancer. A test with high sensitivity will rarely misclassify women with cancer as not having cancer, and a test with high specificity will rarely misclassify women without cancer as having cancer.

The accuracy of a test can also be expressed in a more clinically useful measure, namely, likelihood ratios. When making medical decisions a clinician can use likelihood ratios and test results to estimate the probability of an individual woman having breast cancer. Clinicians use individual patient characteristics (such as age and family history) and features seen on the diagnostic mammogram (such as microcalcifications or distortions) to estimate a woman’s risk of malignancy. This estimate is known as a “pre-test” or “prior” probability. The clinician can then use the likelihood ratios (that express the accuracy of the test) and Bayes’ theorem to decide if an additional imaging test will be helpful in guiding management decisions.

After establishing the accuracy of the various imaging tests we used the summary likelihood ratios to prepare simple models of various clinical scenarios to attempt to indirectly address the implicit question of whether women benefit from the addition of noninvasive imaging tests to standard work-up after recall for evaluation of a possible breast abnormality detected by screening mammography or physical examination. This information may be useful to clinicians in deciding when, or if, it is clinically appropriate to use various types of noninvasive technologies to evaluate breast abnormalities.

Because women with a previous history of breast cancer and women known to be at high risk of breast cancer (due to carrying BRCA1 and BRCA2 mutations or having a very strong family history of breast cancer) have a very different risk profile than the rest of the population, we did not evaluate the use of noninvasive technologies for such women in this review. Instead, we focused on the use of noninvasive imaging technology for women from the general population who present with an abnormal finding by screening mammography or physical examination. We also (as the evidence permitted) examined the influence of age; the size and morphological characteristics of the lesion; and other key clinical risk factors on the accuracy of the noninvasive imaging methods.

Noninvasive Imaging

Noninvasive imaging technologies generally fall into two primary groups: technologies that examine the anatomy, or physical structure, of the breast; and technologies that detect abnormal metabolic patterns. Some noninvasive imaging technologies are slightly invasive in that they require the infusion or injection of a tracer or contrast agent; and some technologies expose patients to radiation. Each of the noninvasive technologies considered in this review is briefly introduced in the Results section of this report.

Conceptual Framework

The analytical framework (Figure 1) demonstrates the links between patients, tests, interventions, and outcomes. The numbers on the diagram refer to the Key Questions (see next section) and their placement in Figure 1 illustrates the many links separating the Key Questions from the patient-oriented outcomes. Fryback and Thornbury have proposed a six-level model of assessing diagnostic efficacy.104 Level 1 is analytic validity; Level 2 is diagnostic accuracy; Level 3 is diagnostic thinking; Level 4 is impact on choice of treatment; Level 5 is patient-oriented outcomes; and Level 6 is societal impact. Demonstration of efficacy at each lower level is logically necessary, but not sufficient, to assure efficacy at higher levels. Patients and health-care providers are generally most interested in studies that evaluate the impact of diagnostic tests on Level 5, patient-oriented outcomes, and on Level 4, impact on choice of treatment. However, studies that directly link diagnostic tests to patient-oriented outcomes are expensive, require very long followup, and are difficult to conduct. In the absence of direct evidence, the effect of diagnostic tests on patient-oriented outcomes can sometimes be estimated by creating indirect chains of evidence by evaluating other levels. Our literature searches did not identify any relevant studies that directly reported the impact of the diagnostic tests on patient-oriented outcomes.

Figure 1 depicts the analytical framework that defines the scope and approach used in this systematic review. The figure demonstrates the links between patients, tests, interventions, and outcomes. The figure is essentially a flow chart- the patient population of interest enters the figure at the left, undergoes diagnostic mammography, and then moves to be evaluated by one of the non-invasive technologies of interest. Changes in patient management in response to the results of the imaging are depicted next, followed by diagnostic test characteristics, and then at the far right of the figure patient oriented outcomes affected by the changes in patient management and the diagnostic accuracy of the imaging tests are depicted.

Figure 1

Analytical framework. CT = computed tomography; MRI = magnetic resonance imaging; PET = positron emission tomography; SC = scintimammography Note: Figure 1 depicts the Key Questions within the context of the patient population, diagnostic tests, subsequent (more...)

Therefore, we chose to approach this project by conducting a systematic review of the diagnostic accuracy of various noninvasive methods of evaluating breast abnormalities (Level 2). After establishing the accuracy of the tests, we constructed an indirect chain of evidence in an attempt to address Level 4 (impact on choice of treatment or use of additional diagnostic tests), and where possible Level 5 (impact on patient-oriented outcomes). We used the estimates of accuracy and the usual clinical scenario to address the implicit, very important question of whether women benefit from the additional use of these noninvasive imaging tests.

Diagnostic Test Characteristics

No diagnostic test is perfect. Studies of test performance compare test results on a group of individuals, some of whom have the disease and some of whom do not. Each individual undergoes the experimental test as well as a second reference test to determine “true” disease status. The relationship between the diagnostic test results and disease status is described using diagnostic test characteristics. It is important that the reference test is very accurate in measuring “true” disease status, or else the performance of the experimental diagnostic test will be poorly estimated.

Sensitivity and Specificity

The results of the experimental and reference standard test and their relationship are commonly presented as two-by-two (2 × 2) tables (see Table 1). From the 2 × 2 table, sensitivity and specificity are readily calculated:

Table 1. Example of a 2 × 2 table.

Table 1

Example of a 2 × 2 table.

Sensitivity and specificity are test properties that are useful when deciding whether to use the test. Sensitivity is the proportion of people with the disease who have a positive test for the disease. A test with high sensitivity will rarely misclassify people with the disease as not having the disease (the test rarely has false-negative errors). Specificity is the proportion of people without the disease who have a negative test. A test with high specificity will rarely misclassify people without the disease as diseased (the test rarely has false-positive errors).

Predictive Values and Likelihood Ratios

To make sense of a diagnostic investigation, a clinician needs to be able to make an inference regarding the probability that a patient has the disease in question according to the result obtained from the test. Sensitivity and specificity do not directly provide this information. The predictive values and likelihood ratios can also be directly calculated from a 2 × 2 table:

The positive predictive value of a test is the probability of a patient having the disease following a positive test result. The negative predictive value is the probability of a patient not having the disease following a negative test result. Predictive values describe the probabilities that positive or negative results are correct for an individual patient. However, predictive values depend on the prevalence of disease in the population. A study that enrolled a patient population with a disease prevalence of 70 percent may report a positive predictive value of 80 percent. If a clinician tests a patient from a population with a disease prevalence of 70 percent, and the test comes back positive, the clinician knows the patient has an 80 percent chance of having the disease in question. However, if the patient comes from a population with a disease prevalence of 20 percent, the clinician cannot apply the results of the study directly to this patient.

Because sensitivity and specificity are difficult to directly apply to clinical situations, and predictive values vary markedly as a function of disease prevalence (i.e., may be different for each patient subpopulation) a combined measure of diagnostic performance, the likelihood ratio, is a more clinically useful diagnostic test performance measure. Negative likelihood ratios measure the ability of the test to accurately “rule out” disease, and positive likelihood ratios measure the ability of the test to accurately detect disease.

Likelihood ratios are independent of prevalence and therefore can be directly applied in the clinic to update an individual’s estimated chances of disease according to their test result. Likelihood ratios can be used in Bayes’ theorem to calculate post-test odds of having a disease from the pre-test suspicion of the patient’s odds of having that disease. Clinicians may be familiar with simple nomograms that allow a direct visualization of post-test chances of disease given a positive or negative test result, without the need to go through the tedious calculations of Bayes’ theorem; see, for example, the interactive form of the nomogram provided by the Center for Evidence-based Medicine at http://www.cebm.net.

When making medical decisions a clinician can use likelihood ratios and the test results to estimate the probability of an individual woman having breast cancer. Clinicians use individual patient characteristics such as age, family history, and personal history; and features seen on the diagnostic mammogram, such as microcalcifications or distortions, to estimate a woman’s risk of malignancy. This estimate is known as a “pre-test” or “prior” probability. The clinician can then use the likelihood ratios (that express the accuracy of the test) to help decide if an additional imaging test will be helpful in guiding management decisions. For example, if a clinician estimates a woman’s risk of malignancy as “very high >50 percent” or “very low <1 percent” most likely the use of any additional imaging test will not change the clinician’s management recommendations, and therefore additional imaging will not be beneficial to the woman. However, if a clinician estimates a woman’s risk of malignancy as being uncertain or in an intermediate area, the likelihood ratios can be used to estimate whether an additional test is likely to change management decisions.

Scope and Key Questions

This systematic review was commissioned by the Agency for Healthcare Research and Quality (AHRQ) to address the following Key Questions:

Key Question 1. What is the accuracy (expressed as sensitivity, specificity, predictive values, and likelihood ratios) of noninvasive tests for diagnosis of breast cancer in women referred for further evaluation after identification of a possible breast abnormality on routine screening (mammography and/or clinical or self-detection of a palpable lesion)? The noninvasive tests to be evaluated are:

  • Ultrasound (conventional B-mode, color Doppler, power Doppler, tissue harmonics, and tomography)
  • Magnetic resonance imaging (MRI) with breast-specific coils and gadolinium-based contrast agents, with or without computer-aided diagnosis (CADx)
  • Positron emission tomography (PET) with 18-fluorodeoxyglucose (FDG) as the tracer, with or without concurrent computed tomography (CT) scans
  • Scintimammography (SMM) with technetium-99m sestamibi (MIBI) as the tracer, including Breast Specific Gamma Imaging (BSGI)

Key Question 2. Are there demographic (e.g., age) and clinical risk factors (e.g., morphologic characteristics of the lesion) that affect the accuracy of the tests considered in Key Question 1?

Key Question 3. Are there other factors and considerations (e.g., safety, care setting, patient preferences, ease of access to care) that may affect the accuracy or acceptability of the tests considered in Key Questions 1 and 2?

This report is an update of a Comparative Effectiveness Review (CER) of the same title originally published in 2006. The Key Questions have been revised and additional diagnostic tests have been added to the list of tests to be evaluated. The 2006 version of the CER only evaluated B-mode ultrasound, MRI (without CADx), PET (without CT), and full-body scintimammography.

PubReader format: click here to try

Views

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...