Investigating the presence of surgical learning in the Timing of Primary Surgery for cleft palate randomised trial

Background/Aims
When conducting a randomised controlled trial (RCT) in surgery, it is important to consider surgical learning, where surgeons familiarity with one, or both, of the interventions increases during the trial. If present, learning may compromise trial validity. We demonstrate a statistical investigation into surgical learning within a trial of cleft palate repair.

Methods
TOPS compared primary surgery, using the Sommerlad technique, for cleft palate repair delivered at six- or twelve-months of age. Participating surgeons had varying levels of experience with the intervention and in repair across the age groups. . Trial design aimed to reduce the surgical learning via pre-trial surgical technique training and balancing the randomisation process by surgeon. We explore residual learning effects by applying visual methods and statistical models to a surgical outcome (fistula formation) and a process indicator (operation time).

Results
26 surgeons operated on 521 infants. As the trial progressed, operation time reduced for surgeons with no pre-trial Sommerlad experience (n=2), before plateauing at 30 operations, whereas it remained stable for those with prior experience. Fistula rates remained stable regardless of technique experience. Pre-trial age for primary surgery experience had no impact on either measures.

Conclusions
Managing learning effects through design was not fully achieved but balanced between trial arms and residual effects were minimal. This investigation explores the presence of learning, within an RCT that may be valuable for future trials. We recommend such investigations are undertaken to aid trial interpretation and generalisability and determine success of trial design measures.