Breast imaging

Screening for breast cancer

 

General background

It is well known that mammography can detect breast carcinoma before it is palpable and sometimes even in a preinvasive stage (Fig. 34). Furthermore, it has been demonstrated in several controlled trials that screening with mammography can reduce breast cancer mortality. In addition, breast saving surgery can be performed in a higher proportion of cases and the need for systemic therapy can be reduced. However, several prerequisites must be fulfilled for screening to be successful. Obviously, the attendance among invited women must be high. A high quality in all steps of the imaging procedure and film reading is necessary as well as professional management of women with positive findings on the screening mammogram. A high rate of false positive findings will create undue anxiety and costs. Thus, a high specificity is important. The detection of some very slow growing carcinomas, which, in the absence of screening would not have presented during the woman's lifetime is unavoidable. Women who participate in a screening program also have to experience an excess of breast cancer years due to lead time and length bias, which will be explained later. A negative screening examination may be falsely reassuring. This, in turn may delay the diagnosis of a carcinoma appearing in the interval between screenings. Women who participate in screening programs should be instructed to perform monthly breast self-examination and to seek advice if there is a change in the breast.

Breast cancer is not a uniform disease, but rather a spectrum of diseases ranging from quite innocent to very malignant. Furthermore, in the screening situation there are mechanisms which tend to bias the sample of cancer that is detected at screening.

The length bias refers to the fact that slow growing tumors are more likely to be detected at screening than fast growing. Fast growing tumors tend to appear in the intervals between screenings, so-called interval cancers. The detection bias refers to the fact that some cancers have an appearance which is more easy to detect radiographically or to identify as malignant, while others are more difficult to detect. The self selection bias refers to the fact that those women who are invited but do not attend a screening program have, in most studies, turned out to have a greater than average risk of dying from breast cancer. This seems to be due to the fact that once the non-attenders seek advice for their breast cancer they have a more unfavourable staging.

The lead time bias refers to the period of time with which the diagnosis is advanced through screening. Thus, survival may erroneously seem prolonged. Therefore, survival cannot be used to measure the effect of screening unless the lead time is controlled for, which is difficult.

Accordingly, in all controlled trials of the effect of breast cancer screening, mortality has been used as a measure of the effect. Furthermore, all breast cancer deaths in the invited group, including cases detected in the intervals between screenings and among non-attenders, must be compared with the number of breast cancer deaths in the control group due to the biases mentioned above.

Review of trials

The effect of breast cancer screening has been estimated in several randomized trials, including one study with geographical controls (the UK Trial) and in three case control studies. The age groups invited the size of the study populations as well as screening interval and screening modalities are given in Table 2.

Table 2. Controlled trials of breast cancer screening

---

Study	Age group	No.invited/ controls	Interval (months)	Modality M= mammography P = physical examination
HIP , New York, USA	40-64	31.000/31.000	12-24	M +P
Malmoe, Sweden	45-69	21.000/21.000	21	M
Two counties, Sweden	40-74	77 .000/56.000	24-33	M
UK trial, England	45-64	46.000/127.000	12-24	M + P
Edinburgh, Scotland	45-64	23.000/23.000	12-24	M + P
Stockholm, Sweden	40-64	40.000/20.000	28	M
Gothenburg, Sweden	40-59	22.000/30.000	18	M
Canada	40-59	45.000/45.000	12	M + P *

---

 

*) only 50-59

The detection rate of breast cancer at the first screening was usually between 5 and 8 cases per 1000 women (Fig. 35) and in subsequent screening rounds between 2 and 4, depending on the screening interval. The detection rate in the first screening exceeded the control group incidence by a factor of about 2 in the younger age group (< 50 years) and by a factor of about 4 in the older age group. The lower relative detection rate among younger women is due to a lower sensitivity of mammography in this age group and possibly also a detection bias and faster growth rate. The number of advanced cancers (stage II and over) was reduced among women invited to screening, an effect that was usually seen after

Figure 35. Schematic representation of a mammographic screening program (prevalence round).

about three years of screening. The relative reduction has in most studies been in the order of20 to 30%. The positive predictive value of a re commendation of surgical biopsy has been between 40 and 80%. Currently, it is about 75% in most high quality screening studies, meaning that 3 out of 4 surgical biopsies are malignant. FNAB is used routinely in most studies and has increased the predictive value of a recommendation for surgical biopsy.

In the combined Swedish randomized trials a 24 % reduction of breast cancer mortality has been achieved. This result was statistically significant. In younger women (40 to 49 years) the reduction was only 13 % (not statistically significant), while in the age group 50 years and over the corresponding reduction was 29% (statistically significant). In absolute terms the cumulative breast cancer mortality after 12 years was 3,9 per 1000 person years in the invited group and 5, l in the control group (all ages invited). Similar results have been achieved in the UK trial.

Most experts conclude that screening women between the age of approximately 50 to 70 years is cost effective, while the issue of screening 40 to 49 year old women is still under debate.

Radiation dose

The carcinogenic risk of the radiation exposure at mammography has attracted attention, especially in the context of screening asymptomatic women. Our knowledge has increased substantially since a relationship between breast cancer and x-irradiation was first demonstrated in the sixties. The radiologist working with mammography should be aware of the state of knowledge in this field. The female breast is sensitive to the carcinogenic effects of ionizing radiation. The dose response relationship is relatively well known at high doses (one hundred to several hundred cGy (rads)). In the low dose range (up to several cGy), data are insufficient for determining the shape of the dose response curve. It is generally agreed that a linear extrapolation from the high dose range would represent the highest possible risk. Other functions, such as linear quadratic or pure quadratic which would imply a lower risk per cGy than the linear model have been discussed. Using the linear model the risk of breast cancer induction has been calculated to be in the order of 6 to 7 cases per million women irradiated per cGy and per year after a latency period of about ten years, and continuing for the rest of the women's life.

It is clear that the carcinogenic effect of ionizing radiation is age-related. In epidemiological studies of women exposed to high doses of ionizing radiation an excess number of breast cancers has been observed mainly in women who were below the age of 30 to 40 at the time of irradiation. Recent data indicate that exposure already in infancy and childhood increases the incidence of breast cancer, while earlier studies indicated that the greatest risk would be from exposure between 10 and 20 years of age with virtually no risk before that age. Furthermore, there is a latency period of 5 to 10 years between the radiation and the appearance of any excess cancer. Dose fractionation probably does not reduce the risk. Thus, the effect of low doses seems to be additive.

The mean absorbed dose in the breast gland per mammographic film is in the order of 0.1 cGy for the average breast. Comparing with the background radiation in for example Sweden, the effective dose equivalent of one mammographic exposure is only 30% of the background radiation during one year. Although the risk of inducing breast cancer with mammography is exceedingly small, it should be taken into account in two situations:
1) When screening large populations of asymptomatic women. As was pointed out above, the age of the women to be screened would be an important determinant of risk.
2) When examining symptomatic patients under the age of 30. As a general rule, every symptomatic woman aged 30 and over should have a mammogram. In the age group 25 to 30, mammography should be performed only if there is a clinical suspicion of malignancy. Under the age of 25 mammography should be performed only exceptionally.

a

Figure 36. a) and b). Two mammograms of the same breast obtained in two different institutions (and in two different countries) in a short time interval. Both examinations were performed using identical mammography film. There is a substantial difference in image contrast, image a) being clearly substandard. Imaging and processing parameters for image a) are unknown but there is reason to believe that the main reason for the low contrast was a processor problem. The patient had a 5 cm spiculated carcinoma. The consequences in terms of reduced sensitivity for the detection of early carcinoma of an image quality like the one illustrated in a) are obvious.

b

It is generally agreed that the potential risk of mammography is far outweighed by the potential benefits. However, it is important to optimize the mammographic imaging system and to monitor the imaging process by a proper quality control program in order to keep the radiation dose as low as possible.

Quality control

Several factors influence the accuracy of mammography, e.g., technical factors related to the x-ray machine and processing (Fig. 36), the examination technique and the radiologist's performance. Accordingly, there are several components of a quality assurance program.
All personnel involved should have proper training. The x-ray equipment must meet certain criteria. Medical outcome measures should be monitored. Data should be available to calculate the sensitivity, specificity and predictive values of the procedures used. Continuous correlation of radiographic findings with pathology is an essential component of a quality assurance program. In addition to its value for follow-up and training, such a correlation provides a reference base for policies on the management of different categories of lesions, e.g. calcifications and circumscribed masses.

The goal in mammography is to consistently produce high quality mammograms with minimal radiation exposure to the patient. To maintain a high image quality regular tests have to be carried out. The performance of the processor should be monitored by daily sensitometry. Usually, a 21-step sensitometer is used, producing densities ranging from gross fog to maximum density. After processing, the steps are measured in a densitometer. A minimum daily check should include speed, contrast and gross fog.

Another test that should be performed daily or at least weekly, is a phantom exposure which will provide an over-all check of the imaging system by measuring density, contrast resolution, kV and phototimer operation.

Several parameters relating to the x-ray machine such as beam quality, function of the automatic exposure device, tube current, absorbed dose, and focal spot size should be measured by a physicist semi-annually or annually. If a darkroom is used, the darkroom should be checked at regular intervals for light leaks.

In some countries an accreditation program has been implemented to guarantee a high and consistent mammographic quality.

 

Ingvar Andersson and Baldur F. Sigfússon