Applied Radiology

Thomas M. Kolb is in private practice in New York City, specializing in breast cancer detection in young, high-risk women.

Research protocol, introduction and significance

In our study, 27,825 consecutive asymptomatic patients were screened with a mammogram and breast physical examination. Women were divided into those with dense and nondense (fatty) breasts. Those with dense breasts (13,547) also had a complete screening breast ultrasound using Philips HDI 5000 and HDI Ultramark 9 (Philips Medical Systems, Andover, MA). All three examinations were performed during a single visit by a single examiner. Patient age, hormonal status, and risk factors were also recorded.

This study is unique because it tested each woman with three different cancer-screening methods concurrently. This method of data acquisition may more accurately determine true- and false-negative results, which are crucial for determining the accuracy/performance of a test. Specifically, if one of the three tests detects a cancer and the other two do not, then we are certain that the first is true-positive and the other two are false-negative. All previous studies that evaluated the sensitivity of mammography and physical examination have used a surrogate identifier of false- and true-negative examinations called the interval cancer method. With this method, after a mammogram is done and is interpreted as being normal, there is a defined waiting period of 12 months, during which it is observed whether or not the patient develops breast cancer. If it is determined that she has developed breast cancer, then the original, or index, mammogram is considered false-negative. If she does not, then the mammogram is considered true-negative. The interval cancer method has important limitations. One limitation is that it is completely arbitrary. One could wait 6, 12, or 24 months or any other time period and find a different sensitivity for mammography within the same set of women and mammographic results. Our method of using three contemporaneously applied tests, which is limited only if all three examinations were concurrently false-negative, may provide a more accurate assessment of how well each test performed.

Our study also evaluated, for the first time, multiple variables that may influence the performance of each screening test. We analyzed the effect of patient age, breast density, hormonal status (pre- or postmenopausal and whether taking hormonal replacement or not), and risk factors. Doing so allowed us to customize what test or combination of tests may be most effective for breast cancer detection for each particular woman. Traditionally, for mammographic screening, women have been divided by decade of life: 40 to 49 years of age, 50 to 59 years of age, and so forth. However, we evaluated whether other variables were more important than patient age in influencing the result of the mammogram. Therefore, our results have important implications as to how best to detect cancer.

Key factors

Dense breasts and mammographic sensitivity

It is considerably more difficult to mammographically detect a cancer in women with dense breasts than in those with fatty breasts. The problem is that on a mammogram the appearance of a cancerous mass is always white. It follows that it would be more difficult for a mammogram to detect a cancer in a woman with dense breasts because a white tumor would blend imperceptibly within the background of white dense breast tissue. Conversely, a tumor would be more easily detected in a woman with fatty breasts because a white tumor would be more easily seen within a background of black fatty tissue. This problem has been recognized, and the American College of Radiology has developed a four-level grading system to subjectively categorize breast density on a scale of I (fatty) to IV (most dense). However, the sensitivity of mammography for each of these density grades has never been reported prior to our study.

Physical examination sensitivity

The use of clinical breast examination (physical examination [PE]) performed by a doctor has never been well studied, yet it is widely advocated and recommended by medical associations and is routinely performed by doctors. There has never been a mortality benefit ascribed to performing PE. There are two studies that examine the use of PE in women of varying ages using the interval cancer method to determine performance. An often-quoted Canadian National Breast Screening Study that found the sensitivity of physical examination to be approximately 70% appears to have been a contaminated study, in that women who knew they had palpable breast lumps were evaluated as screening patients. ³ This scenario leads to an overly optimistic sensitivity value for screening PE, as more women who point to palpable masses in their breasts are more likely to have cancer correctly detected by a PE than women who have no known breast lumps.

Screening breast ultrasound

In 1998, we reported that screening with ultrasound can find large numbers of cancers in women with dense breasts who have normal mammograms and physical examinations. ¹ The reason is that the appearance of cancer on ultrasound is dark (the opposite of the mammographic appearance), while the background dense tissue remains white (like the mammogram). In our current study, we quantified how frequently ultrasound can find these otherwise undetected cancers and how often these cancers are small and in an early stage--an important factor in a patient's prognosis. ²

Filling the gap: Advancing ultrasound technology provides a highly effective breast cancer screening tool

Screening ultrasound is the use of ultrasound to interrogate the entirety of both breasts and is evaluated in our study as a means of detecting breast cancer in an asymptomatic woman. Diagnostic ultrasound is the use of ultrasound directed to one specific area of one breast where a question has already been raised based on the result of a mammogram and/or PE. Diagnostic ultrasound is universally recognized and is commonly used as an important adjunct to help clarify results of the mammogram and PE. Both screening and diagnostic ultrasound are performed with the same equipment (which is commonly available) and there are minor differences in technique (other than the amount of breast tissue interrogated). Screening ultrasound has not previously been routinely used nor well studied because early attempts to do so in the 1980s failed to show any added benefit. However, since the early to mid-1990s, there has been continuing improvement of ultrasound technology, which allows earlier diagnosis of breast cancer with ultrasound.

Conclusions

Our latest study demonstrated the following ² :

1) The overall sensitivity of mammography for all women viewed as a single group was 77%. It is generally stated that mammography misses 10% to 15% of cancers. Our study found this number to be higher (23%).

2) However, since women (and therefore mammographic sensitivity) differ by breast density, age, and hormonal status, it is more important to report the sensitivity for each type of woman than for all women combined. We found that the sensitivity of the mammogram is predominantly and independently determined by breast density. That is to say that, in those women with fatty breasts, mammography preformed exceedingly well, missing only 1% of cancer. However, as density increases, mammographic sensitivity decreases; and in women with the most dense breasts, mammography missed >52% (sensitivity of 47.8%) of all breast cancer. Since there are large numbers of women, predominantly young and premenopausal with dense breasts, the impact of this low sensitivity is very large.

3) Physical examination detected only 27% of cancers. Further, when a tumor was palpable (felt by the examiner), on average it measured twice as large as those tumors that were nonpalpable, and only 37% were early-stage cancers, which predicts a poorer prognosis compared with nonpalpable cancers found by either mammography and/or ultrasound. This confirms the importance of finding tumors before they become palpable. Patients' age, breast density, or hormonal status did not affect the sensitivity of PE. As detailed above, in women with fatty breasts, mammography is excellent and there appears to be little role for the performance of PE at the time of the mammogram. This is a large cost expenditure that needs re-evaluation.

4) Screening ultrasound detected an additional 42% of women with dense breasts who had invasive cancers that were not detected by either mammography or PE. This translates to a 73% increase in cancer detection. These tumors were small (~1 cm), and 90% were early-stage cancers, conferring an excellent prognosis. Women who are at high-risk for developing breast cancer benefit even more by utilizing screening ultrasound, with a higher cancer yield utilizing this technique.

5) The false-positive rate for screening ultrasound is 2.4%. This is sufficiently low enough that with the very large increased yield of cancer detected above and beyond the mammogram and PE (73%), the routine use of screening ultrasound should be considered.

6) Ultrasound is far more sensitive than PE at finding cancers in women with dense breasts and finds them at a smaller size and lower stage. Quantitatively, if we substitute ultrasound for PE in conjunction with mammography (ie, perform mammography + ultrasound instead of mammography + PE, which is the current conventional method for detection) we would increase cancer detection from 74% with the current conventional scheme to 97% with our newly proposed method. Therefore, it appears that women with dense breasts may derive significantly more benefits from screening ultrasound than from screening PE. This would be even more important for high-risk women who arbitrarily have multiple PEs each year (as in high-risk cancer
surveillance centers). Further study is necessary prior to changing any current screening guidelines.