Introduction
Materials and methods
Study design
Data collection
Statistical analysis
Results
Descriptive statistics
Study characteristic | Reader 2 Blinded | % | Reader 2 Not blinded | % | |
---|---|---|---|---|---|
Age of woman (mean) | Mean age | 59.2 | - | 59.3 | - |
Group 1 (52 or less) | 90,505 | 23.66 | 167,217 | 22.70 | |
Group 2 (53-59) | 111,642 | 29.19 | 214,996 | 29.18 | |
Group 3 (60 or more) | 180,343 | 47.15 | 354,488 | 48.12 | |
Total | 382,490 | 100.00 | 736,701 | 100.00 | |
First screen/subsequent screen | Subsequent screen | 300,820 | 78.65 | 581,080 | 78.88 |
First screen | 81,670 | 21.35 | 155,621 | 21.12 | |
Total | 382,490 | 100.00 | 736,701 | 100.00 | |
Reader 1 Recall (pre arbitration) | Not recalled | 363,034 | 94.91 | 697,294 | 94.65 |
Recalled | 19,456 | 5.09 | 39,407 | 5.35 | |
Total | 382,490 | 100.00 | 736,701 | 100.00 | |
Reader 2 Recall (pre arbitration) | Not recalled | 364,387 | 95.27 | 698,512 | 94.82 |
Recalled | 18,103 | 4.73 | 38,189 | 5.18 | |
Total | 382,490 | 100.00 | 736,701 | 100.00 | |
Recall (after arbitration) | Not recalled | 367,341 | 96.04 | 706,082 | 95.84 |
Recalled | 15,149 | 3.96 | 30,619 | 4.16 | |
Total | 382,490 | 100.00 | 736,701 | 100.00 | |
Cancers detected by reader 1 | Cancer detected | 3066 | 0.80 | 5717 | 0.78 |
No Cancer detected | 379,424 | 99.20 | 730,984 | 99.22 | |
Total | 382,490 | 100.00 | 736,701 | 100.00 | |
Cancers detected by reader 2 | Cancer detected | 3226 | 0.84 | 6117 | 0.83 |
No cancer detected | 379,264 | 99.16 | 730,584 | 99.17 | |
Total | 382,490 | 100.00 | 736,701 | 100.00 | |
Cancers detected overall (after arbitration) | Cancer detected | 3355 | 0.88 | 6301 | 0.86 |
No Cancer detected | 379,135 | 99.12 | 730,400 | 99.14 | |
Total | 382,490 | 100.00 | 736,701 | 100.00 |
Alliterative bias
Recall rate, cancer detection rate, and PPV
Recall rate
Recall rate overall (after arbitration) multilevel model | |||
---|---|---|---|
Fixed Effects | Odds ratioa | 95% credible interval for the odds ratiob | Pr(>|z|)c |
Blinding Yes (versus no as the reference category) | 0.869 | 0.826, 0.913 | < 0.001 |
Age (centred) | 1.007 | 1.005, 1.009 | < 0.001 |
First screen (versus subsequent screen as the reference category) | 2.823 | 2.728, 2.922 | < 0.001 |
Blinded yes * age (interaction term) | 1.005 | 1.001, 1.008 | 0.015 |
Blinding yes * first screen (interaction term) | 1.060 | 0.999, 1.125 | 0.057 |
Cancer detection rate
Cancer detection overall (after arbitration) multilevel model | |||
---|---|---|---|
Fixed effects | Odds ratioa | 95% credible interval for the odds ratiob | Pr(>|z|)c |
Blinding yes (versus no as the reference category) | 1.029 | 0.970, 1.089 | 0.341 |
Age (centred) | 1.052 | 1.049, 1.056 | < 0.001 |
First screen (versus subsequent screen as the reference category) | 1.696 | 1.591, 1.807 | < 0.001 |
Tumour characteristics
Positive predictive value
Reader 2 Blinded | Reader 2 Not blinded | Total | |
---|---|---|---|
Cancer detected (TP) after arbitration | 3355 | 6301 | 9656 |
Recall (after arbitration), no cancer detected (FP) | 11,794 | 24,318 | 36,112 |
Total recalled (TP + FP) | 15,149 | 30,619 | 45,768 |
PPV (TP/total recalled) % | 22.1% | 20.6% | - |
Cancer detected (TP) (reader 2) | 3226 | 6117 | 9343 |
Recall (reader 2), no cancer detected (FP) | 14,877 | 32,072 | 46,949 |
Total recalled (TP + FP) | 18,103 | 38,189 | 56,292 |
PPV (TP/total recalled)% | 17.8% | 16.0% | - |
Further test accuracy estimates based on interval cancer data
Reader 2 Blinded | Reader 2 Not blinded | Equality of proportions tests | |||||||||
---|---|---|---|---|---|---|---|---|---|---|---|
Cancer | Cancer | ||||||||||
Detected | Interval | None | Total | Detected | Interval | None | Total | ||||
Recall (after arbitration) | Yes | 3355 | 44 | 11,750 | 15,149 | 6301 | 90 | 24,228 | 30,619 | - | |
No | 0 | 724 | 366,617 | 367,341 | 0 | 1346 | 704,736 | 706,082 | - | ||
Total | 3355 | 768 | 378,367 | 382,490 | 6301 | 1436 | 728,964 | 736,701 | - | ||
Test accuracy statistics | 95% confidence interval | 95% confidence interval | χ2(1) | p | |||||||
Sensitivity | 82.44% | (81.28%,83.60%) | 82.60% | (81.76%,83.45%) | 0.0390 | 0.843 | |||||
Specificity | 96.89% | (96.84%,96.95%) | 96.68% | (96.64%,96.72%) | 37.6 | < 0.001 | |||||
PPV | 22.44% | (21.77%,23.10%) | 20.87% | (20.42%,21.33%) | 14.7 | < 0.001 | |||||
NPV | 99.80% | (99.79%,99.82%) | 99.81% | (99.80%,99.82%) | 0.491 | 0.483 |