Thanks so much. Thank you to AOSSM for this opportunity, and in particular, thanks to the entire ROC Group for making this possible. Here are disclosures of note Allosaurus and Varacil have in the past given unrestricted research grants to the ROC Group to support this research. So who is the ROC Group? Well, we are a group of national and now international centers designed to look at, study, and hopefully improve the outcomes of OCD of the knee. We are truly multidisciplinary. We include surgeons, basic scientists, radiologists, PTs, ATCs, and veterinarians. So what is OCD? Well, as you know, it's a challenging problem to treat clinically. It's been defined as a focal idiopathic alteration of subchondral bone with risk for instability and disruption of adjacent articular cartilage that may result in premature osteoarthritis. The ROC Group has previously validated a arthroscopy classification to better define the diagnosis and direct care. We've also established a prospective cohort, which is now the largest collection of OCD lesions in patients, now numbering over 1,000 patients. So what can a predictive model do? Well, we're trying to help improve preoperative planning, counseling, and discussions, and we wanted to know whether we could utilize prospective cohort preoperative data to predict the ROC arthroscopy classification at the time of surgery, and we wanted to develop a clinical algorithm to distinguish between mobile and immobile lesions. Again, the cohort has over 1,000 patients, not all of which got surgery, but we have close to 800 surgical patients. When we call this down to patients with just medial formal condyle and lateral formal condyle lesions, single lesions within those knees and had complete data, we had about 319 patients. For statistical methods, we divided our group into a 75% partition that was used for our model training. We then held back 25% to test our model. We used multivariable logistic regression on the training data to determine factors associated with the likelihood of a mobile lesion versus immobile lesion adjusted for age, sex, and weight. We then used stepwise model selection on the training data. Again, we used CART and ROC analyses on the training data to dichotomize relevant variables for a final predictive model, and we used binary clinical variables such as age, sex, and weight cutoffs. We then used quantitative model fit statistics on the validation data to get our sensitivity, and area under the curve. So what did we find? Our best model was chronologic age equal to or greater than 14 at the time of surgery, a fusion on physical exam when they presented, and any loss of range of motion. We did include male sex and weight as these had been shown in previous studies to potentially be valuable. Again, our final model included the increased likelihood of a mobile lesion being associated with a chronologic age equal to or greater than 14, any effusion on exam when presenting compared to the contralateral side, any loss of range of motion on physical exam, and we controlled for male sex and weight. When we looked at our 25% holdout sample, this model yielded a sensitivity of 83%, specificity of 75%, and area under the curve of 0.86. So what does this tell us? So if a patient presents and is greater than or equal to 14 years of age, a fusion on exam, and has some loss of range of motion, they're about 90% to 95% likely to have a mobile lesion at the time of surgery. If that patient is less than 14, has no effusion, and no loss of range of motion, they're only about 8% to 16% likely to have a mobile lesion, and this has a high predictive probability. Again, in conclusion, this predictive model for knee OCD lesions indicates age, evidence of knee effusion, and reduced range of motion are 90% to 95% likely to predict a mobile lesion. And again, the ability to predict the mobility of a knee OCD lesion before surgical intervention can facilitate improved surgical planning, as well as communication with families and patients regarding these procedures. Special thanks to the entire ROC team, and particularly Trish Miller, who's our biostatistician who did a lot of this work. Thank you.

ROC Group

osteochondritis dissecans

OCD

predictive model

knee effusion