Randomised controlled trials represent the gold standard in the evaluation of outcome of treatment. They are needed because differences between treatment effects have been minimised and observational studies may give a biased estimation of the outcome. However, conducting this kind of trial is challenging. Several methodological issues, including patient or surgeon preference, blinding, surgical standardisation, as well as external validity, have to be addressed in order to lower the risk of bias. Specific tools have been developed in order to take into account the specificity of evaluation of the literature on non-pharmacological intervention. A better knowledge of methodological issues will allow the orthopaedic surgeon to conduct more appropriate studies and to better appraise the limits of his intervention.

Purpose: Surgical techniques must be evaluated before proceeding with widespread use. The evaluation system is usually copied after that used for drugs, relying on randomised trials. This system fails however to assess one dimension, i.e. quality control. The purpose of the present study was to demonstrate the usefulness of an evaluation technique taking into account the capacity to control the quality of a surgical procedure and determine the reliability, reproducibility, and controllability of the technique. This method of evaluation was applied to computer-assisted surgery for total knee arthroplasty.

Material and methods: Computer-assisted surgery based on 3D bone reconstruction was used for 78 knees. The main outcome assessment criterion was the mechanical axis from the centre of the femoral head to the centre of the ankle. The desired alignment was between 3° varus and 3° valgus. Alignments were measured continuously by an independent operator. A Cusum curve was drawn over time and tested after each operation to determine whether the procedure under evaluation could be continued. The same method was applied to the position of the individual femoral and tibial implants setting the limits at two degrees around the perpendicular mechanical axis (AP view).

Results: The mechanical axis was between 3° varus and 3° valgus in 91% of the knees. The continuous evaluation curve showed that the procedure was under control throughout the trial. There was a learning curve that plateaued at 27 knees. Evaluation of the position of the individual implants showed a trend towards femur valgus over time which was compensated by a trend towards tibial varus.

Discussion: Industrial quality control procedures are well established and widely used. The goal of this work was to apply the same quality control methodology to a surgical procedure. Before undertaking a randomised trial this type of evaluation can affirm that the procedure is well controlled and that using an innovating technique under these conditions does not expose patients to undue risks.