In the 1970s and early 1980s, as a consequence of the exaggerated concern over mammography and the overestimated radiation risk to the breast, investigators looked to ultrasound as a replacement for mammographic screening. To facilitate ultrasound screening, whole-breast ultrasound units were devised that permitted scanning of the entire breast to try to detect early-stage breast cancer. Several large studies showed that ultrasound, at that time, was incapable of detecting cancers that were not also evident by either physical examination or mammography (1,4). As noted above, however, some investigators, using more modern equipment, have reported cancers detected only by ultrasound. Although encouraging, these studies have not been properly performed (blinded interpretation), and the scan techniques and criteria for intervention were not clearly defined. Although ultrasound is expected to be used in conjunction with mammography screening, the fact that it is being used to try to find unsuspected cancer makes it a true screening test, and it must be properly validated. To truly understand the ability of a secondary screening test to find cancer, it must first be studied in a blinded fashion so that the interpreter of the ultrasound examination does not know the results of the mammogram or clinical breast examination. Once the blinded evaluation is completed, then it can be reinterpreted with the information from the other studies. Without blinding, the ultrasound study can be biased by the information from the other studies. Without blinding, a divining rod could be shown to point to breast
cancer (see Ultrasound Screening for Breast Cancer). None of the studies to date has been blinded. The ACRIN/Avon trial 6666 is a blinded study and will be the first to determine the true ability of ultrasound to find mammographically occult cancers.

The biggest hurdle for a breast cancer screening test, however, is to prove that finding cancers using the new test actually results in a decrease in cancer deaths. Without this proof, a screening test may not only not be efficacious, but, because of false-positive studies, its use could actually be detrimental.

Although we must wait for the proper studies and scientific proof of benefit, I think that the mounting evidence suggests that ultrasound may be able to detect some early cancers that are occult on mammography and clinical breast examination. If this proves to be the case, systems for whole-breast ultrasound evaluation will need to be developed to remove the current operator dependence of the procedure. High-quality mammography remains the only proven imaging technique that has efficacy for the detection of clinically occult early-stage cancer, but as technology improves other tests may begin to show efficacy.

In our experience with hand-held, real-time contact breast ultrasound, as well as our original experience with whole-breast scanning with a dedicated water-path system, we frequently have had difficulty separating the hypoechoic, ill-defined appearance of breast cancer from normal, irregularly marginated, hypoechoic mammary tissue (Fig. 17-3). The normal structures of the breast produce findings on ultrasound that are similar to those seen with some breast cancers (Fig. 17-4). Fat is hypoechoic, and “fibrocystic” tissue can mimic the hypoechogenicity of breast cancer. Cooper’s ligaments and other fibrous structures in the breast can produce acoustic shadowing that can also
be indistinguishable from breast cancer (Fig. 17-5). If these problems can be overcome so that the false-positive rate for ultrasound becomes reasonable, and if ultrasound can be shown to have screening efficacy (decreases deaths) in properly performed trials, it could be a second-level screen, in addition to mammography, for women with dense breast tissue.

The fact that ultrasound finds lesions that mimic breast cancers but prove to be benign (false-positives) was originally demonstrated in our study of more than 1,000 women. We undertook long-term follow-up of 94 women who had suspicious lesions found only by ultrasound (they all had normal clinical breast examinations and mammograms). Over a 3- to 4-year period, none of these women developed a clinically evident cancer. This study demonstrated that scanning women with normal mammograms and negative physical examinations raises concern for lesions that are invariably benign. There is as yet no support for surveying the entire breast by ultrasound, and with a high false-positive rate, scanning the breast is undesirable unless it is part of a research protocol. Our study used old, whole-breast ultrasound imaging, but it provides a model for how the ability of ultrasound to find cancers should be tested. It was prospective and the readers were blinded. The radiologist who interpreted the mammograms did not know the results of the ultrasound or the clinical examination. The radiologist who interpreted the ultrasound study (performed in a standardized, automated fashion) was blinded to the findings on the clinical examination and the mammogram. This is the only way to determine whether there is a separate, efficacious role for a test. Despite anecdotal claims, there has not yet been a prospective, blinded study of lesions detected only by ultrasound that validates its use for detecting clinically and mammographically occult breast cancer in asymptomatic women. These observations do not negate the usefulness of ultrasound in diagnosis for evaluating lesions whose presence has been established by physical examination or mammography.

Investigators have described the detection of cancer by ultrasound alone, but these reports remain anecdotal (1,5,6,7,8,9,10,11,12). The selection criteria in many have not been clearly defined, and the scan techniques as well as criteria for intervention have not been clearly documented. One of the biggest problem with interpreting the results from these studies lies in the fact that the investigators using ultrasound to “detect” breast cancers have not been blinded to the clinical examination and the mammogram (13). What all have failed to understand is that the only way to prevent bias would be to conduct blinded interpretation of the ultrasound without access to the mammogram and clinical examination. The reason that this is so important is that blinded analysis is the only way to determine the independent contribution of a new technology for screening. If the individual evaluating the new technology (in this case ultrasound) has access to other information, that information can bias the evaluation. Consciously or otherwise, the analyst may use the information from the mammogram or the clinical examination to “detect” an abnormality using ultrasound that would not have been detected without that information. It has been argued that ultrasound would never be used without mammography for screening, but this is irrelevant. Without knowing the independent contribution of ultrasound, its use in conjunction with mammography is indeterminate. Once the blinded comparision is made, then the analyst can have access to all of the data, and the analysis can determine how this changes from the independent review (14). This is the only way to determine how ultrasound functions as a screening test.

Even after the sensitivity of ultrasound for detecting occult cancer is determined, it remains to be seen whether finding cancers by ultrasound alone actually saves lives. The same issues involved in demonstrating whether mammography screening is efficacious must be addressed for ultrasound screening.

The reason validation is so important can be summarized in the following manner. Suppose that ultrasound screening found only the cancers that were nonlethal (most women who develop breast cancer do not die from it). On the other hand, what if ultrasound found only cancers that had already metastasized and were incurable even though they were found earlier? If either or both of these situations is the case, then finding cancer using ultrasound would be of no benefit. Since ultrasound has a very high false-positive rate and false-positive results lead to biopsies, and since every cancer that is detected will be treated, a screening test that finds only incurable or nonlethal cancers will cause
only harm, with no benefit. It is for this reason that rigorous testing is needed to prove that a screening test has efficacy (15). It is clear that there are palpable cancers that are not visible by mammography but are visible by ultrasound (16). What is not clear is how many cancers that are not palpable and not visible by mammography are detectable by ultrasound.

Although I have little doubt that there are cancers that are detectable only by ultrasound, this remains to be proven. If the ACRIN trial is performed properly it will provide important information on the sensitivity and specificity of ultrasound in a triple-blind evaluation of clinical breast examination, mammography, and ultrasound, with each investigator initially blinded to the findings of the others and then with knowledge of all the available information (this is the only way to avoid the significant potential for bias). The technique of scanning needs to be defined and reproducible. If ultrasound screening is to be feasible, it will likely require dedicated screening devices. It is not likely to be cost-effective to use hand-held systems for screening. The criteria for intervention must be carefully defined, because there are many ultrasound findings in any normal breast that could be mistaken for cancer and thus the false-positive rate could be extremely high.

Until the true ability of ultrasound to detect occult cancer is established, it should not be used outside of scientific trials for screening. Once it is established that ultrasound does have the ability to detect occult cancers, then it will need to be determined whether finding small cancers is sufficient to justify screening using ultrasound, or whether mortality reduction must first be established. Given the improvements in the technology that have taken place over the past years, ultrasound should be re-evaluated as a screening technique, but until its efficacy has been demonstrated, ultrasound should not be considered as a screening test.

It has been said that “if your tool is a hammer, then everything is a nail.” There are enthusiasts in every field, and ultrasound is no exception. Many radiologists are highly proficient in the use of ultrasound in the evaluation of the breast. Numerous papers have been written about the various ultrasound criteria whose presence increases the likelihood that a lesion is benign or that suggest it is malignant. Most are not recent observations. Although Stavros is often credited with establishing criteria that might be used to differentiate benign from malignant lesions, most of the characteristics he used were developed decades ago. There is little doubt that oval or slightly lobulated, sharply marginated, well-circumscribed masses with homogeneous echotexture and enhanced through-transmission of sound that are longer than they are wide (the long access parallel to the skin) have a very high probability of being benign. The problem lies in the fact that cancer occasionally has all of these characteristics.

What has not been determined is how reliable a noninvasive test must be to avoid the need for tissue sampling (needle or excisional biopsy). As a consequence of the costs associated with breast biopsy (i.e., economic, psychological, and physical), there has been great emphasis placed on using imaging to avoid a biopsy. This is perfectly reasonable as long as the woman is informed of the options and has the opportunity to decide how much certainty she requires. Because the stakes may be as high as life and death, it should not be sufficient that a test such as ultrasound be able to predict the nature of a lesion, say, 70% of the time; this leaves too much room for error. Fortunately, a breast biopsy, even surgical excision, has an extremely low morbidity and virtually no mortality. The only biopsy that is safer is a skin biopsy. The cost of a breast biopsy can be reduced. If it is performed skillfully, there should be little trauma, and there need be little or no residual evidence of the needle biopsy, or even surgical biopsy. If ultrasound, or any other noninvasive test, is promoted as a method to differentiate benign from malignant lesions, the false-negative rate becomes critical, and women should be properly informed of the probabilities that the test result will be incorrect; they should also be provided with their options.

The real question lies in how accurately ultrasound can differentiate the borderline lesions, not the cancers that are obvious clinically or by mammography. There is no reason to use ultrasound to try to determine whether a spiculated lesion by mammography is benign or malignant. Nevertheless, the studies that have been used to suggest that ultrasound is good for separating benign lesions from malignant have included obvious cancers. Including these lesions only dilutes the data and makes ultrasound appear more accurate than it actually is. This is compounded by the fact that, once again, the ultrasound analyst has not been blinded to the mammographic findings and those from clinical breast examination. Knowing that a lesion is a spiculated mass by mammography, it would be unlikely for an investigator to suggest that it was a benign lesion by ultrasound.

The uses of imaging in breast evaluation are unique. The only proven efficacy is from using mammography to screen asymptomatic women to reduce the mortality from breast cancer through earlier detection. Using ultrasound to differentiate cysts from solid masses is another efficacious use for sonography. Given the statistical probabilities and the fact that breast cancer is fairly rare relative to the number of women at risk, the vast majority of abnormalities found by mammography or clinical breast examination prove to be benign. Investigators should constantly be reminded that in a population of women being screened repetitively, there will only be two to four new cancers each year. When the accuracy of a test is considered, it should be remembered
that if 1,000 screening mammograms were reported as negative, without the radiologist even reviewing them, the accuracy would be 99.96% to 99.98%. This is a greater accuracy than virtually any other medical test. The problem, of course, is that this approach would have failed to detect the two to four cancers among the 1,000 mammograms. Although the test was highly “accurate,” it would actually have been totally without value and might have even had a negative impact if the reports were used to falsely reassure the women being screened.

What is usually overlooked when a new test is promulgated is the critical question, “How does use of the new test actually alter the care of the patient, and is this alteration beneficial for the patient?” Many papers and books describe how features seen on ultrasound can be used to predict whether a lesion is likely to be benign or malignant, but these are irrelevant if the lesion must still be biopsied or removed to provide certainty. If a test cannot safely alter the care of an individual, then its efficacy must be questioned.

In a highly quoted and referenced article, Stavros et al (17) claimed that ultrasound can accurately classify “some” solid lesions as being benign such that biopsy can be avoided. They presented criteria that they claimed permitted them to accurately differentiate benign lesions from malignant lesions (Table 17-1). Unfortunately, this publication was an unstructured report of a conglomerate of retrospective findings. The selection criteria for women who were evaluated were poorly specified and produced an aggregate of lesions that were inappropriately analyzed as a whole. As noted above, they included masses that were spiculated by mammography and for which ultrasound was not indicated diagnostically. They included lesions that were also typically benign, and there was no blinding as to the clinical and mammographic features of the lesions. Without the proper study design, the results are impossible to interpret and the conclusions are unsupported; however, many who use sonography rely on this classification.

The Stavros et al report was essentially a compilation of anecdotal observations. The authors speculated on the origin of ultrasound findings as if they were fact when in reality they have never been substantiated. For example, they stated that an echogenic border was indicative of infiltration. This is mere speculation. They suggested that “punctate calcifications that are sonographically visible within solid nodules are more likely to be associated with malignant than benign lesions.” This relationship has never been substantiated. They stated as fact that “markedly hyperechoic tissue that is well circumscribed and of uniform echogenicity represents fibrous tissue.” This has never been documented, and hyperechoic cancers, although rare, have been seen (personal communication from Alan Semine, MD).

The report also mixed palpable and nonpalpable lesions indiscriminately, which reflects the failure to appreciate the efficacious use of an imaging technique.

The majority of the criteria used by Stavros to differentiate benign lesions from malignant were nothing new. Most had been described over many years by other authors (18,19,20) and had been found to be insufficiently accurate to be used reliably. The authors added one new criterion: they described a thin echogenic rim that they felt was associated with benign masses (Fig. 17-6). The authors stated, “Sonography helped to correctly classify 100 nodules as malignant in comparison to 38 classified as malignant with mammography.” What they overlooked was the fact that comparing sonography to mammography as a diagnostic test is a meaningless endeavor. It is well established that
mammography is not a very accurate “diagnostic” technique; it is a screening test. The attempt to differentiate borderline lesions into benign versus malignant by imaging is meaningless, since all of these lesions require a tissue diagnosis, regardless of their ultrasound classification. Even though the analyst might suggest that these would prove to be malignant, there was no proof that a biopsy could be avoided. Once it is clear that a tissue diagnosis is needed, there is no value to additional imaging unless the additional study can safely eliminate some of that intervention. The real reason for using ultrasound to evaluate a solid mass is to determine whether it can be seen by ultrasound to decide if it can be biopsied under ultrasound guidance.

Another major problem with the Stavros paper is the fact that the age distribution of the patients with the various lesions was not provided. If a high proportion of the women with benign lesions were under age 30, then this would shift a large number of lesions to benign categories based on the age of the women alone without even looking at the images. If the radiologist who interpreted the images knew that the patient was under 30, then the image interpretation would likely be biased by this knowledge (cancer is extremely rare among women under age 30) and the results would be influenced. Including a large number of these patients would make the overall statistics seem better then they actually are by diluting the percentage of misses.

The real question is not whether ultrasound is helpful in evaluating spiculated lesions (virtually always malignant) or masses among women under age 30 (you can “bet the farm” that these latter cases are fibroadenomas, and imaging these women adds little to their care). The real question is whether ultrasound has any value in differentiating the borderline lesions (sharply marginated round, oval, or smoothly lobulated masses) or the lesions with ill-defined/obscured margins that raise concern.

In addition to having not been validated scientifically, the Stavros criteria do not add much to the sonographic evaluation of lesions. The authors required that to be classified as a benign lesion, the mass could not have any of the indeterminate or malignant features, despite the fact that these could all be seen with lesions that proved to be benign. The “echogenic capsule” can not only be seen with malignant lesions, but it is very uncommon even among benign lesions, especially if the fibroadenomas found among women under age 30 are removed from the analysis. In a study of 403 solid lesions that were biopsy-proven, 262 benign masses were included. Among these, only 66 (25%) had a thin echogenic pseudocapsule, and 3 of these (5%) were malignant lesions (21). Other criteria that the authors evaluated also showed the overlap between benign and malignant lesions using ultrasound criteria. There were 193 lesions with well-circumscribed margins. Among these, 20 (10%) were malignant. Among the 107 lesions that were “taller than wide,” 76 (71%) were malignant, but 31 (29%) were benign. Among the 91 lesions that had enhanced through transmission, 58 (64%) were benign, but 33 (36%) were malignant. This paper concluded that many of the descriptors used to try to differentiate benign from malignant lesions are, in total, “statistically significantly” different for differentiating all benign lesions from all malignant lesions. What the authors failed to explain is that the percentage of lesions with a given criterion might be significantly different when comparing all benign lesions with all malignant lesions, but this ignores the critical question, “Can the criterion be used to accurately determine whether a specific individual’s lesion is benign or malignant?” Since all criteria overlap, it is somewhat risky to rely on them to differentiate specific individual lesions.

Because we all would like to find imaging tests that differentiate benign from malignant lesions, and because many ultrasound enthusiasts do not appear to understand or wish to participate in scientific validation studies, there is the perception in the breast imaging community that ultrasound can be used to differentiate solid breast masses. It is important to realize that there is no scientifically clear support for this perception. The reader is cautioned that this capability has not been shown in any scientifically valid study.