Total hip replacement but not clinical osteoarthritis can be predicted by shape variations of the hip: a prospective cohort study (check)

Abstract

Purpose: Osteoarthritis (OA) of the hip can be defined by both radiographic and clinical criteria. Previously, it has been shown that radiographic criteria associate to the shape of the hip. However, it is unknown whether shape variations of the hip also associate with hip OA defined by clinical criteria. We aimed to determine the contribution of hip shape in the prediction of both end-stage OA (THR) and general clinical OA (ACR criteria) in a prospective cohort. Methods: Individuals were extracted from the CHECK (cohort hip and cohort knee) study. CHECK is a prospective cohort study of individuals with assumed early symptomatic OA of knee or hip. Of the 1002 individuals, a random subset of 865 had standardized anteroposterior radiographs (AP) taken at both baseline and 5 years follow-up. AP-pelvic radiographs of sufficient quality were obtained in 723 subjects at baseline and in 770 subjects at 5 years follow-up. The shape of the proximal femur and acetabulum on the AP radiographs was assessed using statistical shape modelling (SSM). The shape was defined by a set of landmark points that are positioned along the contour of the bone in the image (Figure 1). An SSM describes all variation in shape of the complete cohort by several independent shape aspects called modes. Hips that had received Total Hip Replacement (THR) within 5 years were classified as end-stage OA and hips fulfilling the ACR criteria after 5 years based on history, physical examination, and laboratory findings were classified as clinical OA. The association between each mode and both outcome measures was calculated by Generalized Estimating Equations (GEE), corrected for sex and age. To correct for multiple testing, a p-value of 0.0017 was used to determine statistical significance (0.05 was divided by the 30 modes tested). The predictive value of the GEE model including all univariate significant modes was then tested by the area under the ROC curve (AUC). Results: For this study, 682 females (mean age 55.8) and 183 males (mean age 56.4) were included. At baseline, 75% of the hips had no signs of OA (K&L=0) whereas 25% had doubtful OA (K&L=1). When corrected for age and sex, nine modes at baseline associated significantly with the development of end-stage OA within 5 years (Table 1). For example, mode 12 which resulted in the largest odds ratio of 2.34, illustrates superior joint space narrowing and overcoverage of the acetabulum. When combining these modes, only the four modes which associated most with end-stage OA (mode 4, 11, 12, 15) were selected for calculating the AUC, to prevent overfitting of the model. The predictive value of those four modes in the GEE model resulted in an AUC of 0.81. At baseline, no modes associated significantly with the presence of clinical OA at 5 years. Conclusions: These results show that the shape of the hip as quantified by an SSM is able to predict the risk of THR, whereas variation in shape can not predict the development of clinical OA. The latter might be a result of the poor reliability of the ACR criteria. We observed that 87% of the individuals who had clinical OA at baseline, did not have clinical OA anymore after 5 years. In conclusion, minor shape variations at baseline have a good predictive value for the development of end-stage OA and may be used as a biomarker to predict the future risk for THR.View Large Image Figure ViewerDownload Hi-res image Download (PPT)