Digital Tomosynthesis for the Diagnosis of Breast Cancer

PICHON RIVIERE, A.; AUGUSTOVSKI, F.; GARCIA MARTI, S.; ALCARAZ, A.; GLUJOVSKY, D.; LOPEZ, A; REY-ARES, L.; BARDACH, A.; CIAPPONI, A; COMANDE D

Documento de Evaluación de Tecnologías Sanitarias

IECS

Año: 2013 p. 1 - 30

1668-2793

Breast cancer is a public health problem worldwide due to its high incidence and related mortality. Argentina is among the countries with higher rates of incidence and mortality due to malignant breast tumors. Every year, approximately 17,000 new cases are diagnosed and 5,400 women die, which represents an incidence and mortality rate of 74 and 20.1 every 100,000 people. Consequently, it is the first cause of mortality among women. Most breast cancers are diagnosed after an abnormal mammography, however, their diagnostic confirmation requires other imaging studies that will determine the need of biopsy. Digital Breast Tomosynthesis (DBT) or 3D mammography is an imaging diagnostic technology that might allow a better definition and location of the lesions by eliminating image overlapping and consequently a lower rate of diagnostic doubts. Technology DBT is an imaging diagnostic technique based on a modification of the digital mammography. This system consists in a rotational system fitted with an X-ray tube that rotates along an angle set between 15º and 50º. By processing the images, it is possible to get thin 1 mm slices in different planes thus allowing reconstructing the breast in three dimensions (3D). Purpose To assess the available evidence on the efficacy, safety and coverage related aspects regarding DBT use for the diagnosis of breast cancer. Methods In April 2013, a bibliographic search was carried out on the main databases: DARE, NHS EED, on Internet general search engines, in health technology evaluation agencies and health sponsors. Priority was given to the inclusion of systematic reviews; controlled, randomized clinical trials (RCTs); health technology assessments and economic evaluations; clinical practice guidelines and coverage policies of other health systems. Results Nineteen publications from 17 primary cross-sectional studies reporting data on the DBT performance were selected. Ten of these studies estimated the operating characteristics of mammography and DBT in diagnosing breast cancer in women with clinically or imaging (recalls) breast cancer suspicion; four of them described issues related with imaging reading, assessing the level of consistency between radiologists and the tumor size and five of them evaluated the addition DBT to mammography, two of them for screening and two of them in the population with suspected breast cancer. Additionally, two health technology assessments (Spain and Australia), four coverage policies and protocol of a multicenter retrospective study sponsored by the NHS (United Kingdom´s National Health System) were identified. Studies assessing the operating characteristics of mammography vs. DBT Six out of 10 studies that compared the operating characteristics of mammography versus DBT in patients with suspected breast cancer, did not find significant differences among them. Three out of four that did find statistically significant differences, observed a higher area under the ROC curve for DBT. One study conducted on 220 women only observed this difference using two projection DBT images for pathological images in general (0.86 vs. 0.80) and for masses (0.84 vs. 0.76) and calcifications (0.87 vs. 0.79). Another one, that included 185 women found an average area under ROC curve value for DBT of 0.860 and 0.766 for mammography. The third study conducted on a population of 206 women showed area under the ROC curve values of 0.87 vs. 0.83; the authors also found a significant reduction in the rate of false positives from 57% to 48% in images classified as BI-RADS 4 or 5. The fourth study that found significant differences included 144 women and observed a higher sensitivity of 88.4% vs. 73.3 and a negative predictive value (NPV) of 81.8 vs. 66.2. Studies assessing consistency in image reading One study tried to compare lesion visibility in mammography and DBT; it was conducted in 129 women and found that after assessing DBT images on deferred basis, biopsy was indicated in four women whom their mammography images did not indicate this need. In two of the women, the biopsy was positive and in two, negative. When simultaneously observing the mammography and the DBT images, radiologists considered that the visibility of malignant lesions was higher with DBT. A second study including 98 patients found that the quality of DBT images was equivalent to that of mammography in 52% of the cases, however it proved to be higher (68%) in terms of mass evaluation and lower (73%) in calcification visualization. Another study comparing mammography, DBT and ultrasound found that the three techniques underestimated the tumor size when compared with the size resulting from anatomopathology. As regards staging, 79% of the tumors were adequately staged by DBT, 52% by mammography and 73%, by ultrasound, with a statistically significant difference between mammography and DBT, while the latter behaved analogous to ultrasound. The fourth study assessing the tumor visibility in 36 patients found that 17.5% of tumors were not visualized by mammography and 12.5%, with DBT. Studies assessing the operating characteristics of mammography vs. mammography + DBT Three studies were carried out on a population of women with suspected breast cancer, one of them found that sensitivity (91.9% vs. 73.3%) and NPV (86% vs. 66.2%) were statistically higher and another one also found a significant increase in the area under the ROC (0.97 vs. 0.89). The third study did not find statistically significant differences in terms of operating characteristics but did find differences in image reading (correct image location). Finally, two studies carried out in a screened population, one on 12,631 women and the other one on 1,192 were found. The former showed that the mixed screening strategy increased the rate of cancer detection from 6.1 to 8 every 1,000 screened women and reduced the rate of false positive by 13% (61.1% to 53.1%), in the latter, two groups of radiologists assessed different sets of images, both groups found an increase in the area under the ROC curve (0.894 vs. 0.821; p