Rectus Femoris Transfer In Children With Cerebral Palsy: Comparing A Propensity Score-Matched Observational Study To A Randomized Controlled Trial

Aim To test whether an observational study employing propensity score matching could accurately estimate the causal treatment effects of rectus femoris transfer (RFT) as part of single-event multilevel surgery (SEMLS) in ambulatory children with cerebral palsy. Method We used a large clinical database to derive a propensity score for treatment assignment (SEMLS +/- RFT) and used this score to generate a matched patient cohort. We compared the causal treatment effects estimated from this matched cohort with a previously published randomized controlled trial (RCT). Results The treated arms of the observational study and RCT were well matched. There were 129 limbs (81 males) with a mean age of 10 years 7 months (4y 7mo) in the treated arm of the observational study, and 129 limbs (68 males) with a mean age of 10 years 2 months (3y 9mo) in the control arm of the observational study. Differences between the observational study and RCT cohorts were clinically meaningless for knee flexion kinematics (1-4 degrees), timing of knee angle extrema (<3% gait cycle), and speed (<5mm/s). Postoperative changes in the observational study matched those from the RCT. All but one of the observational study confidence intervals were completely contained within the corresponding RCT confidence interval; there were no meaningful differences in magnitude or sign of key outcomes related to stiff knee gait. Interpretation Propensity score matching is an accurate method for estimating the causal treatment effects of RFT as part of an SEMLS. It seems reasonable to extend this approach to other common components of SEMLS treatment in this population.