A Preliminary Study Regarding Construction Of Clinical Predictive Models Based On The Outcome Of MOWHTO And MUKA In The Treatment Of Isolated Medial Knee Osteoarthritis

ObjectiveThe study aimed to detect: 1)The efficacy of MOWTHO(Medial Opening-Wedge High Tibial Osteotomy)and MUKA(Medial Unicompartmental Knee Arthroplasty)in the treatment of isolated medial compartment knee osteoarthritis and whether there was difference or not between those two surgical techniques,2)To investigate the risks factors that determined the satisfied outcome in the treatment of isolated medial compartment knee osteoarthritis through univariate and multivariate regression analysis,3)To construct clinical predictive models based on the demographic,clinical and radiographic data in patients underwent MOWHTO and MUKA through machine learning methods,and aimed to aiming to lay the foundation for subsequent perspective clinical trial.MethodsPatients who had underwent MOWHTO and MUKA for primary knee osteoarthritis were reviewed from January 2016 to December 2020.Patients aged 55 years to 70 years for men and 50 years to 70 years for women were included,and the follow-up more than or equal to12 months.Patients diagnosed with multiple compartment osteoarthritis,inflammatory arthritis,septic arthritis and follow-up less than 12 months were excluded.Patients demographic data contained gender,age,Body Mass Index(BMI),TCM symptom types,Anesthesia risks of America(ASA)grades,anesthesia type,operating sides and follow-up time.Knee ROM(range of motion),KSSK(Knee Society Knee Score),KSSF(Knee Society Functional Score)and knee flexion contracture were collected.K-L(Kellgren-Laurence),FTA(Femoral Tibial Angle),MPTA(Medial Proximal Tibial Angle),JLCA(Joint Line Congruence Angle),PTS(Posterior Tibial Slope),Ahlback grades,Outerbridge grades were also obtained preoperatively and postoperatively.Firstly,preoperative and postoperative KSSK,KSSF,ROM,FTA and MPTA and their difference were compared intra-groups and inter-groups.Secondly,risks factors containing demographic,clinical and radiographic data in all patients which were correlated to postoperative satisfactory outcome were detected through univariate and multiple variate regression model.Thirdly,4 commonly applied clinical predictive models,such as Support Vector Machine,Random Forest,Neural Networks and Logistic Regression were constructed via machine learning methods and the ideal model was determined with ROC analysis.Fourthly,a dynamic Nomogram model was further proposed under the guidance of Logistic Regression model,which was supposed to be meaningful for providing clinical decisions.Quality data was analyzed using Fisher test,normal distributed continuous quantality was analyzed using student's t test,and non-normal distributed continuous data was analyzed using Wilcoxon test.R Software was used for data analysis in the construction of predicting clinical model.ResultsNinety-three patients(97 knees)had underwent MOWHTO,and 148 patients(155knees)had underwent MUKA.Apart for gender,the remaining demographic data,consisted of age,BMI,TCM symptom types,ASA grades,anesthesia types and operating side between the two groups had no significance(P>0.05).There were more women than men in both groups(P<0.01).Ahlback grades,Outerbridge grades,Altman grades,knee contracture angle and PTS also revealed no significance(P>0.05).Patients in MOWHTO group mainly had K-L grade 2 and grade 3,while patients in the MUKA group mainly had K-L grade 3 and grade 4,which showed significance difference(P<0.01).FTA and MPTA obviously changed between the groups(P<0.01).To be specifically,FTA decreased and MPTA increased both immediate postoperatively and at the final follow-up(P<0.01).KSSK,KSSF and KSSF difference showed no difference between groups(P>0.05),but significantly improved when compared preoperatively in each group(P<0.01).Preoperative and postoperative ROM in each group and ROM between groups showed no difference(P>0.05).Gender(OR=1.80,95% CI:1.10-3.00),Outerbridge grades(OR=1.50,95% CI :1.10-1.90),preoperative MPTA(OR=0.93,95% CI:0.86-1.00)and preoperative KSSF(OR=0.82,95% CI:0.79-0.86),and were risk factors in univariate analysis,while only Outerbridge grades(OR=1.70,95%CI:1.30-2.30)and preoperative KSSF(OR=0.82,95% CI:0.78-0.86)were risk factors in multivariate analysis.The AUC,sensitivity and specificity in the Support Vector Machine,Random Forest,Neural Networks and Logistic Regression models was 0.91(95% CI 0.84-0.98),0.84(95% CI:0.75-0.93),0.91(95% CI:0.84-0.98),0.95(95% CI:0.91-0.99),0.85(95% CI:0.76-0.94),0.87(95% CI:0.78-0.96),0.94(95% CI:0.88-1.00),0.83(95%CI:0.72-0.94),0.88(95% CI:0.79-0.97),0.70(95% CI:0.60-0.80),0.63(95% CI:0.49-0.77),0.72(95% CI:0.58-0.86),respectively.The dynamic Nomogram owed good prediction value of satisfactory outcomes for patients with isolated compartment knee osteoarthritis who underwent either MOWHTO or MUKA.ConclusionsThe study had three conclusions: Firstly,both MOWHTO and MUKA were very effective procedures in the treatment of isolated medial compartment knee osteoarthritis,but no diffecence was found between the two groups(P>0.05).Secondly,univariate and multivariate regression analysis suggested that preoperative Outerbridge and KSSF scores were the independent risk factors in determining the satisfied treatment outcomes.Thirdly,four clinical predictive models with relatively high accuracy,sensitivity and specificity were constructed.Random Forest model showed the highest accuracy,while Logistic Regression model appeared to be the least accuracy.The dynamic Nomogram also revealed well reliability.