UICC World Cancer Congress 2006

Overview. A good screening test meets the following criteria. The disease has a high population prevalence; there is an accurate diagnostic test, one with good sensitivity and specificity; the test is easy to perform and acceptable to the population; the screening test is efficacious (work in the ideal) and effective (work in usual conditions); the disease being screened for is detectable in an asymptomatic phase; and treatment in the pre-symptomatic phase prevents symptomatic disease, ultimately seen in reduced mortality rates. Breast cancer meets most of these criteria. With the recommendation for screening affecting millions of women world-wide, it is essential to know if screening tests are efficacious, effective, and lead to reduced mortality. At this time we can accept that screening reduces mortality for all women screened. For appropriate evaluation of screening programs and screening includes clear definitions of the population screened, and measurement of the accuracy of screening in the environment where screening takes place, as well as audit data for radiologists who perform mammography.

Screening Programs, and screening mammography. The evaluation of screening programs requires measurement of the proportion of the population screened, the recall rate, the proportion following recommendations for recall, and the cancer detection rates. It is also very useful to evaluate the interval cancer and false positive rates.

Evaluation of Screening Performance. Audit information for radiologists and mammography facilities should include recall rate, sensitivity, specificity, positive predictive values for the screening mammogram, the recommendation for biopsy and biopsy performed. Cancer detection rates and completed work-up rate are also important. ecall rate is the proportion of screening examinations that lead to further diagnostic work-up. Sensitivity (proportion of cancers that are detected by screening) is based on having access to population-based cancer data, and this is essential for evaluation of screening outcomes. The completion of registration of cancers will affect the quality of the evaluation performed. Specificity (proportion of screening where no cancer is present that are negative on screening) is dependent of the ability to follow women with negative mammograms; women move and die of other causes. As the majority of all screening examinations are negative, errors in calculating specificity are negligible, unless there is a large loss to follow-up. Predictive values are most valuable as they can be calculated most easily, as long as women recalled are followed to the end of the work-up or pathologic outcome. Predictive value of the screening mammogram is the proportion of positive screening mammograms that have a cancer diagnosis. Predictive values for recommendation for biopsy is the proportion of recalled women recommended for biopsy, who have a cancer diagnosis; and Predictive values of biopsy is the proportion of biopsies that are cancer. Interval cancer rate is the proportion of negative screening examinations that are followed in some time period by a cancer diagnosis. Cancer detection rate is the proportion of all screening mammograms that detect cancer. For all the accuracy indices listed above, there are decision about definition of cancer, operational definitions for recall and time used for follow-up that can lead to variations in measurement. Characteristics of the cancers detected and missed at screening are also important in evaluating screening. Screen detected cancers should be associated with a high proportion of cancers being early (DCIS and cancers <10mm), and invasive cancers should be early stage, low grade, with a large proportion of node negative tumors. Factors that are associated with variation in screening mammography occur in three main areas: factors related to the women, the screening practice and the radiologist. In addition there are clearly differences in performance related to geographical location and national policies. Characteristics of the women that are important to understand when interpreting statistics about screening performance include: age, breast density, family history of breast cancer, history of breast biopsy, use of hormone therapy, and time from the previous screening visit. Characteristics of the practices that have an effect on screening performance include: use of CAD or double reading, systems in place for recall or follow-up, the practice type (whether a mammography practice or full breast center) and imaging procedures available at the practice in addition to mammography (clinical breast exam, Ultrasound, digital mammography or MRI). Finally, the experience and characteristics of the radiologists themselves may have an important role in understanding variability of screening performance, including age, training, volume read and percent effort for interpreting mammography exams, among other characteristics. Quality of data and performance measures require standardized data collection, training and monitoring data completion. Understanding definitions used in preparation of evaluation data improves the ability to compare performance within and among countries.