Problems, solutions, and perspectives in the evaluation of interval cancers in Italian mammography screening programmes: a position paper from the Italian group for mammography screening (GISMa).

In this position paper, a self-convened team of experts from the Italian Group for Mammography Screening (Gruppo italiano screening mammografico, GISMa) pointed out the problems that increasingly hamper the feasibility and validity of the estimate of the proportional incidence of interval breast cancer (IBC) in Italy, suggested potential solutions and an agenda for research, and proposed that the question of the sensitivity of mammography be viewed in a larger perspective, with a greater attention to radiological review activities and breast radiology quality assurance programmes. The main problems are as follows: the coverage of cancer registration is incomplete; the robustness of using the pre-screening incidence rates as underlying rates decreases with time since the start of screening; the intermediate mammograms performed for early detection purposes may cause an overrepresentation of IBCs; the classification of many borderline screening histories is prone to subjectivity; and, finally, the composition of cohorts of women with negative screening results is uncertain, because several mammography reports are neither clearly negative nor clearly positive, and because of the limitations and instability of the electronic mammography records. Several possibilities can be considered to cope with these issues: standard methods for using the hospital discharge records in the identification of IBCs should be established; for the calculation of regional estimates of the underlying incidence, a suitable mathematical model should be identified; the definition of IBC according to the 2008 GISMa guidelines needs to be updated, especially with respect to in situ cancers and to invasive cancers with borderline screening histories; a closer adherence to standard screening protocols, with a simplified patient management, would make it easier to objectively identify IBCs; alternative methods for estimating the sensitivity of mammography should be taken into consideration; and, finally, analysis could be restricted to the absolute incidence rate of IBC, which would make comparison of the risk between neighbouring populations possible. Epidemiologists must extend their attention to the prevention of the risk of IBC and the implementation of breast radiology quality assurance practices. Epidemiologists and radiologists can share common objectives: it is necessary to promote the idea that the availability of a registry-based series of IBCs is not a prerequisite for their radiological review; radiological review of breast cancers greater than 20mm in size detected at second and subsequent screens, that are potential substitutes for IBCs, needs radiological and epidemiological validation studies; the advent of digital mammography brings about the possibility to create libraries of mammograms accessible online, which enables the conduct of large studies of the diagnostic variability of radiologists; and, finally, epidemiologists and radiologists have the responsibility to monitor the effects that a loss of cumulative professional experience in screening centres, due to the imminent retirement of a substantial proportion of healthcare workforce, could cause on their performance.

[Interval cancers as indicators of performance in screening programmes]. / I carcinomi di intervallo quali indicatori di performance di un programma di screening.

GISMa analyses the interval cancer (IS) topic providing guidelines and reference standards in addition to CE recommendations. IC identification is based on Cancer Registries (CR), if existing, or on hospital discharge records, in alternative. The optimal measure of IC frequency (inversely correlated with sensitivity) is the IC proportional incidence (observed IC/carcinomas expected in absence of screening). Other formulas (IC/IC + screen detected cancers; IC rate per 1000 negative screens) look less reliable. IC stage at diagnosis (if available through CR) must be compared with screen detected cancer and cancer detected in non-attenders. Review of mammograms preceding the IC (coded as screening error, minimal signs, or occult) should be done mainly with a blind procedure (IC mixed with negative controls), as this procedure is more representative of the original scenario and more respectful of radiologist's rights.

Screening mammography interpretation test: more frequent mistakes.

PURPOSE: To present the mammographic cases most commonly misinterpreted by the participants in the mammography self-test proposed by the Italian Society of Medical Radiology (SIRM) National Congress in Rimini, Italy, 2002, by analysing the findings responsible for errors, suggesting reasons for the errors, and assessing possible inadequacies in the format of the test. MATERIALS AND METHODS: The self-test was performed on the mammograms of 160 cases (32 positive and 128 negative for cancer as confirmed by histology). The mammograms had been taken in the four standard projections and placed on four multi-panel diaphanoscopes, each displaying a set of 40 cases comprising benign and malignant cases in equal proportions. The participants were given pre-printed forms on which to note down their diagnostic judgement. We evaluated a total of 134 fully-completed forms. Among these, we identified the 23 cases most frequently misread by over 15 participants in percentages varying between 40-90%. Of these cases, 10 were malignancies and 13 were negative mammograms. On review, we also assessed the diagnostic contribution of complementary investigations (not available the participants). RESULTS: The 134 fully-completed forms (all of the 40 cases) yielded a total of 5360 responses, 1180 of which (22.01%) were incorrect. Of these, 823 out of the 4288 cases expected to be negative (19.2%) were false positive, and 357 out of the 1072 cases expected to be positive (33.3%) were false negative. As regards the 23 most frequently misread cases, these were 10/32 (31.25%) mammograms positive for malignancy and 13/128 (10.15%) negative mammograms or mammograms showing benign disease. The 10 malignancies included 7 infiltrating ductal carcinomas, 1 infiltrating cribriform carcinoma, 1 infiltrating tubular carcinoma, and 1 carcinoma in situ. The 13 cases of benign disease--as established by histology or long-term follow-up--mistaken for malignancies by the test participants were fibrocystic breast disease in 5 cases, surgical scar in 1 case, ABBI scar in 1 case, radial scar in 2 cases, microcalcifications that had remained stable for years in 2 cases, focal sclero-adenosis in 1 case and sclero-elastosis in 1 case. CONCLUSIONS: The errors were due to microcalcifications, benign disease simulating a neoplasm, overlapping tissue, visibility of a lesion in one projection only, lesion site in relation to the corpus mammae, missed areas of asymmetry. Attention must be paid to these signs of focal breast disease since, if correctly evaluated, they enable the early diagnosis of low-grade carcinomas that frequently carry a favourable prognosis.