1. Pragmatic Integrated Randomised Controlled Trials in Screening: Experience from a trial in 1.2million women attending breast screening
10min 5 questions. Insert mammo after radar
Dr Sian Taylor-Phillips
Associate Professor of Screening and Test Evaluation

2. NIHR Postdoctoral Fellowship
Advisory Group
Janet Dunn
Aileen Clarke
Matthew Wallis
Alison Duncan
Olive Kearins
David Jenkinson
Nigel Stallard
Helen Parsons
Ala Szczepura
Simon Gates
Sue Hudson

3. Contents Pragmatic Integrated Trials
The CO-OPS reverse reading trial (Changing case Order to Optimise patterns of Performance in Screening)
The vigilance decrement
Methods
Results
Discussion

4. Pragmatic Integrated Trials
Explanatory trials measure efficacy
- the benefit a treatment produces under ideal conditions
Pragmatic trials measure effectiveness
- the benefit the treatment produces in routine clinical practice
Integrated into the current systems to reduce workload and cost

5. CO-OPS RCT
Changing case Order to Optimise patterns of Performance in Breast Screening (CO-OPS)

6. The vigilance decrement

7. NHS Breast Screening

8. NHS Breast Screening s.taylor-phillips@warwick.ac.uk
Malignant spiculated mass

9. The CO-OPS Trial
Changing Case Order to Optimise Patterns of Performance in Screening
Is there a vigilance decrement?
Does reverse reading help?

10. The CO-OPS Trial
Pragmatic randomised controlled trial of a software intervention to change case order so that any vigilance decrement will occur for the first and second readers when examining different cases
Intervention ↓↑ or ↑↓
Control ↓↓ or ↑↑

11. The CO-OPS Trial Integration
Centre level recruitment and informed consent
Selection of women: all taking part in breast screening during the study period, automated
Randomisation: Automated using breast screening programme software
Intervention: Applied automatically through the software
Data collection: Routine data systems modified to collect trial results

12. The CO-OPS Trial Primary Outcome Cancer detection rate
Secondary Outcome
Recall rate
Patterns of performance with time since a break

13. The CO-OPS Trial Participation:
46 English Breast Screening Centres have completed the trial
One year of data collection

14. The CO-OPS Trial Participation: 22157 (9227) 20356 (10765) 73.5 (3.6)
Centres taking part (n=46)
Centres not taking part (n=34)
Mean number of women screened in one year (sd)
22157 (9227)
20356 (10765)
Mean percentage uptake (sd)
73.5 (3.6)
73.1 (5.7)
Mean SDR prevalent round (sd)
1.46 (0.4)
1.61 (0.5)
Mean SDR incident round (sd)
1.51 (0.20)
1.50 (0.23)

15. s.taylor-phillips@warwick.ac.uk s.taylor-phillips@warwick.ac.uk
Randomized (n=1,207,633)
in 37,724 batches
Allocation
Intervention Arm
(n=603,528) in 18,797 batches
Received allocated intervention (n=524,971, 87%)
Did not receive allocated intervention but included in analysis: Not read in intended order (n=51,599, 8.5%) Reader trainee (n=26,110, 4.3%) Reader 1 or 2 non-film reader (n=1, 0.0%) Only one reader (n=895, 0.1%) Read using bar code not ordered list (n=16,952, 2.8%) No readers (n=1, 0.0%)
Control Arm
(n=604,105) in 18,927 batches
Received allocated intervention (n=560,760, 93%)
Did not receive allocated intervention but included in analysis: Not read in intended order (n=40,528, 6.7%) Reader 1 or 2 non-film reader (n=1, 0.0%) Only one reader (n=625, 0.1%) Read using bar code not ordered list (n=17,176, 2.8%) No readers (n=0)

16. Follow-Up s.taylor-phillips@warwick.ac.uk
Intervention Arm
Lost to follow-up:
NBSS records not updated
(n=172, 0.03%)
Inconclusive biopsy and did not attend any further test
(n=115, 0.02%)
Recalled for assessment but did not attend appointment
(n=118, 0.02%)
Control Arm
Lost to follow-up:
NBSS records not updated
(n=126, 0.02%)
Inconclusive biopsy and did not attend any further test
(n=118, 0.02%)
Recalled for assessment but did not attend appointment
(n=140, 0.02%)

17. Analysis Analysed (n=596,642, 98.9%)
Excluded from analysis: Technical Recall
(n=6,339, 1.05%)
Subsequent screen of same woman (n=142, 0.02%)
Analysed
(n=597,505, 98.9%)
Excluded from analysis: Technical Recall
(n=6,087, 1.01%)
Subsequent screen of same woman (n=129, 0.02%)

18. Primary and Secondary Outcomes
Intervention
Control
Cancer Detection Rate (95%CI)
0.89%
(0.86% to 0.91%)
0.88%
(0.85% to 0.90%)
Recall Rate
(95%CI)
4.14%
(4.09% to 4.19%)
4.17%
(4.12% to 4.22%)
JAMA. 2016;315(18):

19. Patterns of recall: Single reader

20. Discussion Pragmatic integrated RCTs Requirements
Must fit easily into current systems
Same IT systems across centres (e.g. screening)
Time and willingness to adapt clinical IT systems
Two versions of standard practice if no individual consent

21. Discussion Pragmatic integrated RCTs Advantages Large sample size
Increase involvement of non-research centres
Cost ~10% of other trials
Low loss to follow-up using routine data (particularly if wait for annual reporting)

22. Thank you to the participating centres

23. Control arm Intervention arm Vigilance Vigilance Reader 1: ForwardsB C D E F A B C D E F
Woman
Woman
Vigilance
Vigilance
Reader 2: Forwards
Reader 2: Backwards A B C D E F A B C D E F
Woman
Woman
Vigilance
Vigilance
Both readers together
Both readers together A B C D E F A B C D E F
Woman
Woman
Number of disagreements = Low
Recall Rate = Low
Cancer Detection Rate=Low
Number of disagreements = High
Recall Rate = High
Cancer Detection Rate = High