Background/Purpose: There is a need to design models that predict the risk of total joint replacement (TJR) in patients with knee osteoarthritis (OA). Our aim is to design a model combining clinical and molecular data to predict the risk of TJR in patients from the Osteoarthritis Initiative (OAI).

Methods: Caucasian patients from the OAI were classified into two groups based on their progression to total knee replacement (TKR) during a 108-month follow-up period: TKR progressors and non-progressors. We combined a series of clinical and molecular variables to construct the predictive models. Among the clinical variables we included: sex, age, body mass index (BMI), blood pressure, adherence to the Mediterranean diet (MD) and knee alignment. The molecular variables: mitochondrial DNA (mtDNA) haplogroups, 8 single nucleotide polymorphisms (SNPs) robustly associated with the risk of knee OA occurrence in GWAS studies (rs12107036.TP63, rs8044769.FTO, rs11177.GNL3, rs4730250.DUS4L, rs143383.GDF5, rs10948172.SUPT3H, rs11842874.MCF2L, rs3771501.TGFA). Polygenic Risk Scores (PRS) was derived from a GWAS summary statistics for TKR phenotype, and it was standardized with Z transformation. A panel of 10 serum proteins (APOA4, AZ2G, A2AP, APOA1, FETUA, RET4, TSP1, COMP, YKL40 and SAMP). Appropriate statistical analyses including Cox regression models and the calculation of the area under the curve (AUC) were carried out.

Results: Due to the incomplete data set for all the OAI patients, especially in terms of serum proteins, we constructed two models. A complete model with all the variables except serum proteins (N=345 TKR progressors), and a reduced model that also included the proteins (N=149 TKR progressors). The complete model showed that older age (p< 0.001), higher BMI (p< 0.001), higher systolic blood pressure (p=0.011), a lower adherence to MD (p=0.026), varus alignment (p=0.034), TT vs CC genotype at rs11177.GNL3 (0.042) and the simultaneous presence of the haplogroup Uk and the GG genotype at rs12107036.TP63 (p=0.006) are risk factors to progress to TKR.

The reduced model showed that older age (p< 0.001), higher BMI (p< 0.001), higher systolic blood pressure (p=0.049), a lower adherence to MD (p=0.041), TT genotype at rs11177.GNL3 (p=0.027), above median levels of A2AP (p< 0.001), as well as below median levels of COMP (p< 0.001) and YKL40 (p=0.004) are risk factors to TKR progression.

The AUC for the complete model was 0.705 (0.662–0.748) and for the reduced model was 0.761 (0.722-0.799). Excluding the serum proteins, the AUC of the reduced model was 0.698 (0.655-0.741) (Figure 1).

Conclusion: We constructed a series of predictive models capable to predict the risk of TKR combing clinical and molecular data. Despite the reduced data set, the inclusion of serum proteins significantly improves the predictive capability of the model.

« Back to ACR Convergence 2024

ACR Meeting Abstracts - https://acrabstracts.org/abstract/a-model-based-on-clinical-and-molecular-data-predcits-the-risk-of-total-joint-replacement-progression-in-knee-oa-patients-from-the-oai/