Table 2.
Radiological surveillance of interval breast cancers: methods and results of mammographic review and classification
| Study (first author) | Methods | Distribution of radiological classificationa of interval breast cancers | Additional findings |
|---|---|---|---|
| Weber13 | Review of 800 interval BCs from southern screening region of Dutch breast screening program (2000–11 spanning transition from FSM to DM) by two radiologists based on prior screen and diagnostic mammogram. | Year 1 of inter-screening interval True interval: 43.1% Missed/FN: 31.3% Minimal signs: 25.7% Year 2 of inter-screening interval True interval: 60.2% Missed/FN: 19.1% Minimal signs: 20.7% | Majority of missed or minimal-signs cases were masses at prior FSM or DM. No differences in mammography features (for FSM vs. DM) for cases emerging year 1 of inter-screen interval; however, cases in year 2 of inter-screen interval for DM were more frequently true (than missed) interval BCs compared to those for FSM (p = 0.03). |
| Blanch10 | Review of 1012 interval cancers (Spanish screening program 2000–06) by panels of three experienced radiologists using semi-informedb review of screening and diagnostic mammograms, independent double-reading and arbitration for discordant classification. | True interval: 48.2% Missed/FN: 23.2% Minimal signs: 17.2% Occult cancer: 11.3% | Factors associated with interval BC differed by radiological category, for example, family history of BC was mainly associated with true interval BC, whereas density was more strongly associated with occult BCs followed by true interval BCs. |
| Nederend38 | Review of 224 interval cases from FSM or DM screening (prior screen and those taken at diagnosis) by two experienced radiologists: readers aware they were reviewing interval BCs but blinded to each other's review; consensus for discordant classification | True intervalc: FSM 47.1%; DM 65.3% Missed/FN: FSM 30.8%; DM 20.2% Minimal signs: FSM 22.1%; DM 14.5% | Majority of missed interval BCs were masses at prior FSM or DM, followed by asymmetry or architectural distortion. |
| Domingo14 | Study of 2245 invasive BCs (948 were interval cases) diagnosed 2000–09 in women participating in biennial population screening in Spain; interval BCs were classified by semi-informedb review of the screening and diagnostic mammograms by panels of three radiologists. | True interval: 48.0% Missed /FN: 23.6% Minimal signs: 17.5% Occult cancer: 10.9% | True interval BCs were associated with HER2 and triple-negative tumour phenotypes and with extremely dense (>75% density) breasts; extreme breast density was most strongly associated with occult interval BCs |
| Renart-Vicens26 | Review of 22 interval cases (Girona Health Region screening program 2000–06) by panel of expert radiologists, using semi-informedb independent double-reads of screening and diagnostic mammograms, with arbitration for discordant classification. | True interval: 54.5% Missed /FN: 13.6% Minimal signs: 13.6% Occult cancer: 18.2% | Distribution of pathological features differed between interval and screen-detected BCs (see Table 3) |
| Fong18 | Review of 692 interval BCs, with comparison to screen-detected BC (Breast Test Wales 1998–2001): blindedd review of screening and 'symptomatic' mammograms by two readers, with consensus for discordant classification. | True interval: 57.8% Missed /FN: 17.7% Occult cancer: 10.0% Unclassified: 2.2% | 10-year all-cause survival rate for screen-detected BC (81.6%) was higher than that for interval BC (72.4%) [p < 0.001]: this differed by radiological category, true interval BC (77.5%), FN interval BC (55%), occult (54.4%) with latter two types having lower survival rates than screen-detected. |
| Carbonaro15 | Review of 130 interval BCs in population screening program, Italy 2001–06: three expert radiologists blindly reviewed mammograms, mixed with negative screens: cases not recalled classified as true interval BC, those recalled by only one reviewer as minimal signs, and those recalled by >2 reviewers as missed interval BCs | True interval: 55.0% Missed /FN: 22.0% Minimal signs: 24.0% | A higher rate of larger (T3-T4) tumours was evident for missed interval BCs (18%) than minimal signs (6%) or true interval BCs (8%); and the rate of node metastases (N2-N3) for minimal signs (19%) or missed cancers (25%) was higher than that for true interval BCs (10%). |
| Payne39 | Review of 332 interval BCs (Nova Scotia screening program 1991–2004): blindedd and independent review by three experienced radiologists; classified as true interval BC if >2 radiologists reported index screen as normal (otherwise classified as missed interval BC if >2 reported abnormal screen). | Classified into two categories: True interval: 74.1% Missed/FN: 25.9% | Breast density distribution varied between the two types of interval BC and differed across age-group; rate of true interval BCs was higher for longer screening interval but this was not the case for FN cases. |
| Pellegrini40 | Review of 103 interval BCs in population screening program Trento, Italy 2001–08: external (three radiologists) and internal (five radiologists) panel with varying screening experience blindly reviewed pre-diagnosis screening mammograms, mixed with negative controls. Classification based on majority report ('missed' if recalled by most reviewers). | External review True intervalc: 67.0% Missed/FN: 18.4% Minimal signs: 14.6% Internal review True intervalc: 62.1% Missed/FN: 17.4% Minimal signs: 20.4% | No significant difference between external and internal radiological review. |
| Caumo32 | Review of 100 interval BCs in Verona, Italy, screening program 2000–06: three expert radiologists blindly reviewed pre-diagnosis mammograms, mixed with negative controls. Classification according to majority report. | True intervalc: 71.0% Missed/FN: 15.0% Minimal signs: 14.0% | Interval BC proportional incidence 10.8% in year 1 and 40.0% in year 2 of inter-screening interval. Interval BCs associated with denser breasts compared with negative controls (p = 0.02). |
| Pirola34 | Review of pre-diagnosis screening mammograms of 30 interval BCs from Milan, Italy, screening program (2005) performed by an expert radiologist who had read >300,000 mammograms, blindedd to interval BCs by case-mix with negative screens. | True intervalc: 76.7% Missed /FN: 16.6% Minimal signs: 6.7% | Interval BC proportional incidence estimated as 17.4% for 2-year inter-screening interval. |
| Hofvind37 | Review of 231 interval BCs in Norwegian population screening program 1995–98: six experienced radiologists reviewed and classified cases in a consensus meeting, using screening and diagnostic mammograms; classified as missed if all radiologists agreed tumour was visible at screening mammogram | True interval: 35% Missed /FN: 35% Minimal signs: 23% Occult cancer: 7% | Of the combined missed and minimal signs interval BCs, 50% were poorly defined masses or asymmetric densities, 26% were MC with/without associated density or mass, at the baseline screen. |
| Bare31 | Review of 57 interval BCs in population screening program in Northeast Spain 1995–2001: 'informed consensus review' by three experienced radiologists using screening and diagnostic mammograms. | Excludes 19 'unclassifiable' cases: True interval: 39.5% Missed/FN: 21.1% Minimal signs: 26.3% Occult cancer: 13.2% | No major differences in the prognostic features of interval BCs when examined by radiological type or time elapsed since last screening mammogram. |
| Ciatto33 | Independent review of 100 screening mammograms (20 interval BCs, 80 negative screens) by six radiologists, using three sequenced review methods (separated by 2 weeks) with increasing information: (1) blindedd (no IC information, case-mix) (2) partially informedb (aware IC) (3) fully informed (with diagnostic mammograms) | Method 1 average (range): Missed /FN: 24% (10–40) Minimal signs: 6% (5–15) Method 2 average (range) Missed/FN: 33% (20–55) Minimal signs: 10% (10–20) Method 3 average (range) Missed/FN: 42% (35–50) Minimal signs: 20% (15–30) | A classification of 'missed' or minimal-signs interval BC was more likely using method 2 (odds ratio (OR) = 1.78, p = 0.033 or method 3 (OR = 3.91, p = 0.000) relative to method 1, but no reader effect was evident. |
| Evans35 | Review of 208 interval BCs from a multi-centre RCT of screening from age 40–41 years: review by two radiologists with arbitration by a third, using semi-informedb review of screening mammograms followed by diagnostic mammograms. Abnormalities further classified as malignant, subtle (features difficult to detect), or non-specific (features only seen in retrospect after reviewing diagnostic films). | True interval: 42% Missed/ FN: 26% Occult cancer: 32% | Features frequently misinterpreted were granular MC (38%), asymmetric density (27%), distortion (22%). Of abnormal previous screens, 37% were classified malignant, 39% subtle change and 21% non-specific. MC more common on diagnostic mammograms of FNs than those of true interval BCs (28 vs. 14%). Cases with true interval or FN findings had similar background parenchymal patterns, but those with occult interval BC had higher proportion of dense patterns, p < 0.05 |
| Gao36 | Review of 59 interval BCs (Singapore screening program 1994–97) by three radiologists using index screens; semi-informed (aware reviewing interval cases but unaware of tumour location). | Missed/FN: 17% (based on 'worst diagnoses' from five screen-readers, two from initial reads and three from re-review). | In 3 years of successive follow-up from index screen, interval BC rates per 10,000 women-years were 2.1, 10.6 and 10.8 each year. |
BC breast cancer, DM digital mammography, FSM film-screen mammography, FN false-negative, MC micro-calcifications, IC interval cancer, RCT randomised controlled trial
a Classification of interval BCs: true interval (cancer is not visible at the index mammographic screen but becomes visible at the diagnostic mammogram); missed/FN (cancer is visible on the index mammogram but is not recalled or is misinterpreted); minimal-signs (subtle abnormality is visible on the index mammogram but one that is unlikely to warrant recall); occult (cancer that is not visible on the index screen and not visible on the diagnostic mammogram)
b Semi-informed radiological review generally involved screen-readers knowing that interval BC cases were being reviewed, without information on the side and location of the interval cancer
c In some studies 'true interval' BCs are also referred to as 'occult' at the index or pre-diagnosis screen; this should not be confused with the conventional 'occult cancer' classification of interval cases, which usually refers to a BC that is not seen on the index mammography screen and also occult on the diagnostic mammogram
d Blinding or blinded methods of review: this generally refers to (a) interval cases being interspersed with screen-reading as part of the routine screening workflow; or (b) interval cases being mixed with normal screening mammograms but not integrated into routine screen-reading workflow (study-specific methods described in table)