Table 3. Estimated Performance Measures and Results From Log Binomial Regression Analysis.
| Variable | Mammography Plus Ultrasonography | Mammography Alone (Matched) | Relative Risk (95% CI)a |
|---|---|---|---|
| BI-RADS end-of-day assessment, No. (%) | |||
| 0 (Needs additional imaging) | 21 (0.3) | 5159 (17.2) | NA |
| 1 (Negative) | 2689 (44.2) | 23 909 (79.5) | NA |
| 2 (Benign) | 2793 (45.9) | 878 (2.9) | NA |
| 3 (Probably benign) | 234 (3.8) | 64 (0.2) | NA |
| 4 (Suspicious) | 342 (5.6) | 49 (0.2) | NA |
| 5 (Highly suspicious) | 2 (0.0) | 3 (0.0) | NA |
| Total | 6081 (100) | 30 062 (100) | NA |
| Performance based on end-of-day assessment | |||
| Recall rate for additional imaging or biopsy, % (95% CI) | 9.9 (9.1-10.6) | 17.6 (17.1-18.0) | 0.52 (0.48-0.57) |
| End-of-day assessment of 0, 3, 4, 5, No. | 599 | 5275 | NA |
| Total examinations, No. | 6081 | 30 062 | NA |
| Final assessment, No. (%) | |||
| 0 (Needs additional imaging) | 5 (0.1) | 104 (0.3) | NA |
| 1 (Negative) | 2694 (44.3) | 26 848 (89.3) | NA |
| 2 (Benign) | 2798 (46.0) | 1936 (6.4) | NA |
| 3 (Probably benign) | 235 (3.9) | 341 (1.1) | NA |
| 4 (Suspicious) | 347 (5.7) | 792 (2.6) | NA |
| 5 (Highly suspicious) | 2 (0.0) | 41 (0.1) | NA |
| Total | 6081 (100) | 30 062 (100) | NA |
| Performance based on final assessment | |||
| Biopsy recommendation rate per 1000 screens (95% CI) | 57.4 (51.9-63.5) | 27.7 (25.9-29.7) | 2.05 (1.79-2.34) |
| Final assessment of 4, 5, No.b | 349 | 833 | NA |
| Total examinations, No. | 6081 | 30 062 | NA |
| Short-interval imaging follow-up rate (95% CI) | 3.9 (3.4-4.4) | 1.1 (1.0-1.3) | 3.10 (2.60-3.70) |
| Final assessment of 3, No. | 235 | 341 | NA |
| Total examinations, No. | 6081 | 30 062 | NA |
| Sensitivity (95% CI) | 78.6 (67.1-92.0) | 73.8 (68.1-80.0) | 1.08 (0.92-1.27) |
| Final assessment of 4, 5 and cancer, No. | 33 | 155 | NA |
| Total cancers, No. | 42 | 210 | NA |
| Specificity (95% CI) | 94.8 (94.2-95.3) | 97.7 (97.6-97.9) | 0.97 (0.97-0.98) |
| Final assessment of 0, 1, 2, 3 and no cancer, No. | 5719 | 29 169 | NA |
| Noncancers, No. | 6035 | 29 843 | NA |
| PPV2 (95% CI) | 9.5 (6.8-13.1) | 21.4 (19.6-23.5) | 0.50 (0.35-0.71) |
| Final assessment of 4, 5 and cancer | 33 | 367 | NA |
| Biopsy recommended (final assessment of 4, 5) | 349 | 1713 | NA |
| Cancer detection rate per 1000 screens (95% CI) | 5.4 (3.9-7.6) | 5.5 (4.7-6.4) | 1.14 (0.76-1.68) |
| Final assessment of 4, 5 and cancer | 33 | 165 | NA |
| Total examinations | 6081 | 30 062 | NA |
| False-negative rate per 1000 screens (95% CI)c | 1.5 (0.8-2.8) | 1.9 (1.4-2.4) | 0.67 (0.33-1.37) |
| Final assessment of 0, 1, 2, 3 and cancer, No. | 9 | 56 | NA |
| Total examinations, No. | 6081 | 30 062 | |
| Cancer rate per 1000 screens (95% CI) | 6.9 (5.1-9.3) | 7.4 (6.4-8.4) | 0.99 (0.70-1.42) |
| All cancers, No. | 42 | 221 | NA |
| Total examinations, No. | 6081 | 30 062 | NA |
| False-positive biopsy recommendation rate per 1000 (95% CI) | 52.0 (46.7-57.8) | 22.2 (20.6-24.0) | 2.23 (1.93-2.58) |
| Final assessment of 4, 5 and no cancer, No. | 316 | 668 | NA |
| Total examinations, No. | 6081 | 30 062 | NA |
Abbreviations: BI-RADS, Breast Imaging Reporting and Data System; DCIS, ductal carcinoma in situ; NA, not applicable; PPV2, positive predictive value of biopsy recommendation.
a
Relative risk is from log binomial model adjusted for site, age, menopausal status, first-degree family history of breast cancer, year of examination, prior benign biopsy result, and correlation among women within the same matched set using generalized estimated equations. Sensitivity was adjusted for site and first-degree family history of breast biopsy.
b
Final assessment of 4, 5 considered a positive examination result and was used to calculate performance measures.
c
False-negative rate includes both invasive and DCIS.