Must Bilaterality Be Considered In Statistical Analyses of Total Knee Arthroplasty?
In studies of TKA, treating each knee as an independent case in patients with bilateral TKAs can lead to errors, because patients with bilateral and unilateral TKAs may achieve different levels of function and because the assumption of statistical independence is violated.
We determined whether (1) patients undergoing bilateral TKAs have different demographics and preoperative and postoperative function from those undergoing unilateral TKAs, and (2) means, SEs, and p values change substantially based on how the two cases of bilateral TKAs are treated in statistical analyses.
We retrospectively compared 513 patients undergoing unilateral TKAs and 602 patients undergoing bilateral TKAs regarding demographics (age, gender, BMI), preoperative and 1-year postoperative function (motion arc, American Knee Society score, patellofemoral score, WOMACTM, SF-36). Each case of a bilateral TKA was treated as independent in the reference analyses. Then changes of mean, SEs, and p values were compared with those of the reference method when three other methods of analysis were used: using mean values of both knees, randomly selecting one side, and using a generalized estimating equation.
Patients undergoing bilateral TKAs had higher female predominance (95% versus 90%) and BMI (27.4 versus 26.1 kg/m2) than those undergoing unilateral TKAs. Although patients undergoing bilateral TKAs had worse preoperative function, they achieved comparable function at 1 year. The mean values did not differ considerably depending on the four methods. The SEs and p values of the reference method were generally the smallest, but the differences appeared negligible.
Our analyses suggest bilaterality can be ignored in the analysis of TKA, since postoperative function is comparable between unilateral and bilateral TKAs and the theoretical errors in statistical analysis had little effect on the results.