ICTMC Conference November 2017 Success of randomizing trial participants to disclosure of allocation early or late: a methodological study to investigate performance bias Reeves BC, Harris RA, Rooshenas L, Ashton K, Hutton D, Rogers CA, Blencowe NS, Blazeby JM on behalf of the Bluebelle Study Group University of Bristol Clinical Trials and Evaluation Unit ICTMC Conference November 2017 The Bluebelle study is funded by the United Kingdom National Institute for Health Research (NIHR) HTA programme (ref: 12/200/04) and the Medical Research Council (MRC) ConDuCT-II Hub (MR/K025643/1). JMB is an NIHR Senior Investigator. . The views and opinions expressed are those of the authors and do not necessarily reflect those of the HTA programme, the NIHR, the MRC, the UK NHS or the Department of Health.

Aim To test the feasibility of a randomised controlled trial (RCT) to compare the effectiveness and cost- effectiveness of different dressings for primary surgical wounds: “simple” vs “complex” vs “none” Study comprised two phases: Phase A, pre-pilot feasibility research Phase B, pilot RCT

Pilot RCT: randomise to 3 groups “simple dressing” “glue-as-a- dressing” “no dressing” Complex = “glue-as-a-dressing” Wound visible with “glue-as-a-dressing” or “no dressing” Visibility of wound may influence quality of wound closure

Pilot RCT: timing of randomisation Trialists: randomise after wound closure Surgeons: not feasible!

Pilot RCT: randomise to 6 groups Disclose allocation before wound closure Disclose allocation after wound closure “simple dressing” “glue-as-a- dressing” “no dressing” “simple dressing” “glue-as-a- dressing” “no dressing”

Pilot RCT methods: key times Random’n Anaesthetic Knife-to-skin Wound closed Recovery (Rx) induction (KS) (WC) SURGERY Disclose allocation after (DAA) Disclose allocation before (DAB)

Pilot RCT methods: Information to test feasibility of disclosing allocation after wound closure Times from case report forms: “knife-to-skin” (KS) “wound closed” (WC) Times from database providing randomisation: “randomisation” (Rx) – and disclosure of allocation for the BEFORE group (DAB) “disclosure of allocation AFTER wound closed” (DAA) – when wound closed time was entered into database Supplemented by interviews about “how it worked” with surgeons and research nurses

Pilot RCT methods: key times Random’n Anaesthetic Knife-to-skin Wound closed Recovery (Rx) induction (KS) (WC) SURGERY Disclose allocation after (DAA) Disclose allocation before (DAB) AFTER group: DAA–WC (expected range <10 mins) AFTER group: DAA–Rx (expected range >30 mins) BEFORE group: WC–Rx/DAB (expected range >30 mins)

Withdrawals pre- surgery (n=0) Withdrawals pre- surgery (n=1) Results: Allocated to simple dressing (n=133; before wound closure n=66, after wound closure n=67) Withdrawals pre- surgery (n=0) Underwent surgery (n=133; before wound closure n=66, after wound closure n=67)   Number of protocol deviations: 10 when allocated timing not adhered to, 3 when allocated dressing not adhered to Allocated to glue-as-a-dressing (n=129; before wound closure n=64, after wound closure n=65) Withdrawals pre- surgery (n=1) Underwent surgery (n=128; before wound closure n=63, after wound closure n=65) Number of protocol deviations: 9 when allocated timing not adhered to, 8 when allocated dressing not adhered to Allocated to no dressing (n=132; before wound closure n=66, after wound closure n=66) Underwent surgery (n=132; before wound closure n=66, after wound closure n=66) Number of protocol deviations: 11 when allocated timing not adhered to 20 when allocated dressing not adhered to Randomised (n=394)

Results: “DAA-WC” mins [AFTER] SURGERY 20 40 60 80 100 -10 -5 5 10 50 100 150 200 Frequency -300 -200 -100 DAA – WC (minutes)

Results: “duration of operation” SURGERY AFTER: DAA – Rx BEFORE: WC – DAB 20 40 60 Frequency 200 400 600 DAA – Rx (minutes) 100 300 500 100 200 300 400 500 WC – DAB (minutes) 20 40 60 Frequency

Results: Adherence to disclosure of allocation BEFORE WC (98.9%) Adherence to disclosure of allocation AFTER WC (85.5%) Simple: BEFORE 62/63 (98.4%) before wound closure AFTER 56/65 (86.2%) after wound closure Glue: BEFORE 60/60 (100%) before wound closure AFTER 54/63 (85.7%) after wound closure No dressing: BEFORE 64/65 (98.5%) before wound closure AFTER 55/65 (84.6%) after wound closure

Comments from HCPs’ interviews Cheating the system: HCPs were aware that a ‘false’ time could be entered, to obtain the allocation before wound closure “that defeats the whole point of the study” (RN) Logistical challenges: One computer in theatre, multiple people queuing to use it - “it’s more of a traffic management control issue than any problem with the database” (RN) Performance bias: “‘Oh I’m going to make sure I do this a bit better.’ To be honest I think that wears out, I think there’s a washout phase with a few patients, so … I don’t think it’s affecting on how we close the skin at all now” (Surg) [GLUE] “And he [surgeon] said ‘Oh, I’d forgotten we were in Bluebelle, oh if I’d have thought about it I might have been a bit more careful closing the wound’” (RN)

Comments from HCPs’ interviews Timing for the AFTER group: “We’re quite strict. We’re quite anal and particularly me, because they were saying ‘what’s she having’ and I said ‘no you’re not allowed to know” (RN) “I said ‘no, not until you’ve de-scrubbed and you’re (finished)’… ” (RN) “Until you put a time on that computer and press ‘randomise’ the time of closure to the wound, so when they’re starting to close…..that’s when I’ll put it in, I’ll put the time in if I think… because when they close the skin that doesn’t take long, it only takes well a minute or so, not even that.” (RN) “cos I’m in theatre myself I can see what they’re doing, if I know they’re closing so I put it in the computer I just don’t, I don’t speak to them, I don’t speak to anybody and then I just close it down so nobody can see” (RN)

Summary / conclusions Randomisation of disclosure: surgeons “got it” and were “up for it” Feasible to disclose allocation after wound closure “I get the impression that we have been (rightfully) concerned about Bluebelle changing practices… but I don’t get a sense (from interviews) that Bluebelle has caused too much havoc out on the front line” Logging into the database twice was challenging – need a way to avoid using computer in theatre System ‘time stamps’ can be useful for auditing research team behaviour

SCHOOL OF SOCIAL AND COMMUNITY MEDICINE 07/02/14 Acknowledgements MRC Network of Hubs for Trials Methodology Research (www.methodologyhubs.mrc.ac.uk) ConDuCT-II Hub for Trials Methodology Research (www.bristol.ac.uk/social-community-medicine/ centres/conduct2) The Bluebelle Study Group and research teams (surgeons and research nurses) at the study centres All the study participants Recruitment in the pilot phase was successful. Systematic review evidence and analysis of trials did not reveal competing studies or new evidence. Practice data showed similar incidence of all three interventions supporting the need for an adaption. The revised and expanded 3-group trial with updated hypotheses was funded meaning that modification of the protocol, patient-related documents and study database was required. The randomisation system had to move from two to 3-groups and provide three balanced groups at the end of the trial. Study website and publicity materials were adapted to account for the three groups.   Implementing the adaptation, while recruitment to the trial continued was challenging and not to be underestimated. This approach allows an existing trial to maintain research of relevance to practice. SCHOOL OF SOCIAL AND COMMUNITY MEDICINE 07/02/14