Sensitivity and specificity of mammographic screening as practised in Vermont and Norway

Objective

The aim of this study was to examine the sensitivity and specificity of screening mammography as performed in Vermont, USA, and Norway.

Methods

Incident screening data from 1997 to 2003 for female patients aged 50–69 years from the Vermont Breast Cancer Surveillance System (116 996 subsequent screening examinations) and the Norwegian Breast Cancer Screening Program (360 872 subsequent screening examinations) were compared. Sensitivity and specificity estimates for the initial (based on screening mammogram only) and final (screening mammogram plus any further diagnostic imaging) interpretations were directly adjusted for age using 5-year age intervals for the combined Vermont and Norway population, and computed for 1 and 2 years of follow-up, which ended at the time of the next screening mammogram.

Results

For the 1-year follow-up, sensitivities for initial assessments were 82.0%, 88.2% and 92.5% for 1-, 2- and >2-year screening intervals, respectively, in Vermont (p=0.022). For final assessments, the values were 73.6%, 83.3% and 81.2% (p=0.047), respectively. For Norway, sensitivities for initial assessments were 91.0% and 91.3% (p=0.529) for 2- and >2-year intervals, and 90.7% and 91.3%, respectively, for final assessments (p=0.630). Specificity was lower in Vermont than in Norway for each screening interval and for all screening intervals combined, for both initial (90.6% vs 97.8% for all intervals; p<0.001) and final (98.8% vs 99.5% for all intervals; p<0.001) assessments.

Conclusion

Our study showed higher sensitivity and specificity in a biennial screening programme with an independent double reading than in a predominantly annual screening program with a single reading.

Advances in knowledge

This study demonstrates that higher recall rates and lower specificity are not always associated with higher sensitivity of screening mammography. Differences in the screening processes in Norway and Vermont suggest potential areas for improvement in the latter.In a previous study in which selected early outcome measures of mammographic screening in Vermont, USA, and Norway were compared, higher recall and interval cancer rates were shown for Vermont than for Norway. The rate of screen-detected cancers did not differ [1]. The findings were consistent with other international studies [2-4]. Different radiological reading procedures have been suggested as a possible reason for the findings [1,2,4].Breast cancer screening involves a series of events that begins with the screening examination (bilateral two-view mammography), and may continue with a recall for diagnostic work-up. The diagnostic work-up may lead to a recommendation for a biopsy, which determines whether the suspect lesion is benign or malignant. In both Vermont and Norway, the decision to recall a female patient is based on the assessment of her initial screening mammogram. In the USA, single reading is the usual practice, while in Norway an independent double reading with consensus is performed, in accordance with the European guidelines [5]. In a single reading, a radiologist decides whether the female patient should be recalled for diagnostic work-up, while in an independent double reading with consensus, two radiologists discuss the findings and a consensus is reached as to whether to recall the patient. In both processes, a final assessment is reached after additional breast imaging (including ultrasound) to determine whether to recommend a biopsy.We surmise that the different procedures for initial assessment will affect the sensitivity and specificity of both the initial and the final assessments. However, this can be difficult to ascertain when comparing countries that also have differing screening intervals. To better understand how differences in the interpretation procedures of screening mammography may influence cancer detection, we have taken a detailed look at the sensitivity and specificity of initial and final assessments in our previously studied cohort of female patients aged 50–69 years who underwent screening mammography in Vermont or Norway during 1997–2003. The aim of this study was to determine and compare the sensitivity and specificity of the initial and final assessments of mammographic screening as practised in Vermont and Norway.