Abstract
Introduction
Innovations in surgical robotics and navigation have significantly improved implant placement accuracy in total knee arthroplasty (TKA). However, many comparative studies have not been shown to substantially improve revision rates or other clinical outcome scores. We conducted a simulation study based on the reported distribution of patient-specific characteristics and estimated potential effect of coronal plane alignment (CPA) on risk of revision to evaluate the hypothesis that most published study designs in this area have been too underpowered to detect improvements in revision rates.
Methods
To model previously reported studies, we generated a series of simulated TKA patient populations, assigning each patient a set of patient-specific factors (age at index surgery, BMI, and sex (Fig.1a)), as well as one surgeon-controlled factor (CPA) (Fig.1b) based on registry data and published literature. We modelled the survival probability for an individual patient at time t as a Gaussian function (exp[-(t/(k∗τmax))2]), where τmax (99.5 years) is selected to ensure the mean survival probability of the patient population matched 92% at 15 years. The value of k was adjusted for simulated patients within a range of 0 to 1 as a function of their patient and surgeon-specific factors (Fig.2).
To evaluate power associated with a study design, we ran a Monte Carlo simulation generating 10,000 simulated populations of ten different cohort sizes. We divided the patient population into two groups: one group was assigned CPAs governed by the precision of a navigated/robotic approach (σ=1.5°), and the other CPAs governed by the precision of a conventional approach (σ=3°). We then simulated the time to failure for each patient, computed the corresponding Kaplan-Meier survival curves, and applied a Log-Rank test to each study to test for statistical difference. From the 10,000 simulations associated with each cohort size, we determined the percentage of simulated studies that found a statistically significant difference at each time point.
Results
Figure 3 shows a contour plot illustrating the probability that a survival analysis with a specific study design would find statistical significance between the conventional and navigated/robotic patient groups. Entries from recently published literature are overlaid for context. No studies achieved statistical significance (p<0.05).
Discussion
The effectiveness of navigated/robotic surgery is one of the most controversial debates in orthopedic surgery. The results from this simulation suggest that most revision studies aiming to settle this debate are likely significantly underpowered, falling below the normal 80% threshold.
Limitations of this analysis include using only a single surgeon-controlled variable in the survival simulation, and only a single precision for the navigated/robotic approaches. Further studies will include more implant-related risk factors and a wider range of precisions for navigated/robotic procedures.
Based on this simulation, it appears the effect size afforded by navigated/robotic surgeries on revision rates in TKA surgery is too small to recommend broad application, especially since adoption could involve added costs and unforeseen risks associated with novelty. Clinically, it may be beneficial to examine the use of robotics/navigation on high-risk patients, where studies are likely to have higher power due to larger effect sizes.
For any figures or tables, please contact the authors directly.
