Benchmarks for Sensitivity, Specificity, and False Negative Rate for Diagnostic Mamography Exams from 2007 - 2013
by Indication for Exam*
Sensitivity
| Work-Up of Abnormal Screening Examination | Short-Interval Follow-Up | Evaluation of Breast Problem / No Lump or Lump Unknown | Evaluation of Breast Problem / Palpable Lump | All Diagnostic Examinations | |
| Sensitivity** | 91.9 | 61.7 | 83.8 | 93.3 | 87.8 |
| Readers with 10+ cancers | 126 | 23 | 52 | 76 | 210 |
| True Positives† | 5,497 | 251 | 1,474 | 1,857 | 13,369 |
| Cancers‡ | 5,980 | 407 | 1,758 | 1,991 | 15,230 |
| Sensitivity (Median-50th percentile) | 92.7 | 60 | 84.6 | 93.6 | 88.2 |
| Sensitivity (25th, 75th percentiles) | 86.7, 96.3 | 53.8, 70 | 77, 90.8 | 88.3, 100 | 82.6, 92.3 |
| Sensitivity (10th, 90th percentiles) | 81.8, 100 | 47.8, 86.7 | 73.3, 94.4 | 84.2, 100 | 76.5, 95.2 |
Smoothed Plots of Sensitivity by indication for exam+
for 15,230 Cancers That Were Identified at Diagnostic Mammography (Among Radiologists Finding 10+ Cancers)
Specificity
| Work-Up of Abnormal Screening Examination | Short-Interval Follow-Up | Evaluation of Breast Problem / No Lump or Lump Unknown | Evaluation of Breast Problem / Palpable Lump | All Diagnostic Examinations† | |
|---|---|---|---|---|---|
| Specificity§ | 87 | 95.8 | 93 | 86.4 | 90.6 |
| Readers with 100+ non-cancers | 157 | 115 | 98 | 83 | 246 |
| True Negatives|| | 103,890 | 43,870 | 62,502 | 23,969 | 344,600 |
| Non-cancers# | 119,448 | 45,771 | 67,183 | 27,735 | 380,518 |
| Specificity (Median-50th percentile) | 88.7 | 96.7 | 93.5 | 89.3 | 91.6 |
| Specificity (25th, 75th percentiles) | 84.2, 92.7 | 94.3, 98.4 | 91.1, 96.1 | 85.7, 92.6 | 88.5, 94.8 |
| Specificity (10th, 90th percentiles) | 77.2, 95.1 | 91.9, 99.2 | 87.4, 98.1 | 81.4, 95.7 | 85.6, 96.6 |
Smoothed Plots of Specificity BY INDICATION FOR EXAM+
for 380,518 Non-cancers (Among Radiologists Finding 100+ Non-Cancers), 2007 - 2013
False Negative Rate
| Work-Up of Abnormal Screening Examination | Short-Interval Follow-Up | Evaluation of Breast Problem / No Lump or Lump Unknown | Evaluation of Breast Problem / Palpable Lump | All Diagnostic Examinations† | |
|---|---|---|---|---|---|
| False negative rate (per 1000)†† | 4 | 6.6 | 4.7 | 4.5 | 4.8 |
| Readers with 100+ exams | 157 | 116 | 98 | 87 | 249 |
| False negatives‡‡ | 507 | 310 | 331 | 138 | 1,902 |
| All examinations | 125,610 | 46,639 | 69,153 | 30,157 | 396,263 |
| False negative rate (Median-50th percentile) | 3.8 | 5.9 | 5 | 4.3 | 4.7 |
| False negative rate (25th, 75th percentiles) | 1.3, 6.4 | 2.4, 9.3 | 2.8, 8.3 | 0, 7.9 | 2.7, 7 |
| False negative rate (10th, 90th percentiles) | 0, 9.7 | 0, 16.3 | 0, 11.6 | 0, 13.2 | 0, 10.0 |
Smoothed Plots of False Negative Rate BY INDICATION FOR EXAM+
for 396,263 Diagnostic Mammography Exams (Among Radiologists with 100+ Exams), 2007 - 2013
Footnotes:
Numbers are frequencies unless otherwise indicated.
The first non-zero assessment within 90 days was used as the final assessment for analysis.
* Includes examinations for radiologists with minimum numbers of cases in a given column as designated; excludes examinations for which the radiologist is unknown.
** Sensitivity = the percentage of cancers that had an abnormal final interpretation (BI-RADS category 4 or 5; Note: 0 or 3 with a recommendation for biopsy was recoded as 4).
† True Positive = the number of cancers that had an abnormal final interpretation (BI-RADS category 4 or 5; Note: 0 or 3 with a recommendation for biopsy was recoded as 4).
‡ Cancers = the number of examinations with a tissue diagnosis of cancer within 1 year following the examination.
§ Specificity = the percentage of non-cancers that had a negative final interpretation (BI-RADS category 1, 2, or 3; Note: 0 with no recommendation for biopsy nor additional imaging/workup was recoded as 3, while 3 with a recommendation for biopsy was recoded as 4).
|| True Negative = the number of non-cancers that had a negative final interpretation (BI-RADS category 1, 2, or 3; Note: 0 with no recommendation for biopsy nor additional imaging/workup was recoded as 3, while 3 with a recommendation for biopsy was recoded as 4).
# Non-cancers = the number of examinations without a tissue diagnosis of cancer within 1 year following the examination.
The first non-zero assessment within 90 days was used as the final assessment for analysis.
* Includes examinations for radiologists with minimum numbers of cases in a given column as designated; excludes examinations for which the radiologist is unknown.
** Sensitivity = the percentage of cancers that had an abnormal final interpretation (BI-RADS category 4 or 5; Note: 0 or 3 with a recommendation for biopsy was recoded as 4).
† True Positive = the number of cancers that had an abnormal final interpretation (BI-RADS category 4 or 5; Note: 0 or 3 with a recommendation for biopsy was recoded as 4).
‡ Cancers = the number of examinations with a tissue diagnosis of cancer within 1 year following the examination.
§ Specificity = the percentage of non-cancers that had a negative final interpretation (BI-RADS category 1, 2, or 3; Note: 0 with no recommendation for biopsy nor additional imaging/workup was recoded as 3, while 3 with a recommendation for biopsy was recoded as 4).
|| True Negative = the number of non-cancers that had a negative final interpretation (BI-RADS category 1, 2, or 3; Note: 0 with no recommendation for biopsy nor additional imaging/workup was recoded as 3, while 3 with a recommendation for biopsy was recoded as 4).
# Non-cancers = the number of examinations without a tissue diagnosis of cancer within 1 year following the examination.
†† False negative rate = the rate (per 1000 exams) of false-negative exams.
‡‡ False negatives = the number of cancers that had a negative final interpretation.
+ For figures, An overlaid solid line indicates the 50th percentile (median), paired dashed lines indicate the 25th and 75th percentiles, and paired dotted lines indicate the 10th and 90th percentiles. A single dashed line below the median indicates the 10th and 25th percentiles are the same and a single dashed line above the median indicates the 75th and 90th percentiles are the same.
Table & Figures last updated Mar 26, 2017.
The following must be cited when reproducing this data:
"The Breast Cancer Surveillance Consortium and its data collection and sharing activities are funded by the National Cancer Institute (P01CA154292). Downloaded xx/xx/xxxx from the Breast Cancer Surveillance Consortium Web site - http://www.bcsc-research.org/."