Validation of statistical shape modelling to predict hip osteoarthritis in females: data from two prospective cohort studies (Cohort Hip and Cohort Knee and Chingford).
Agricola R., Leyland KM., Bierma-Zeinstra SMA., Thomas GE., Emans PJ., Spector TD., Weinans H., Waarsing JH., Arden NK.
OBJECTIVES: To prospectively investigate whether hip shape variants at baseline are associated with the need for future total hip replacement (THR) in women and to validate the resulting associated shape variants of the Cohort Hip and Cohort Knee (CHECK) cohort and the Chingford cohort. METHODS: Female participants from the CHECK cohort without radiographic OA (Kellgren-Lawrence score <2) at baseline were included (1100 hips); 22 hips had a THR within 5 years of follow-up. For the Chingford cohort, with only female participants, hips without radiographic OA at baseline were selected and a nested case-control design was used, with 19 THR cases within 19 years of follow-up and 95 controls matched 5 to 1 for age and BMI. Hip shape on baseline anteroposterior pelvic radiographs was assessed by statistical shape modelling (SSM) using the same model for both cohorts. RESULTS: In the CHECK and Chingford cohorts, the respective mean age was 55.8 (s.d. 5.1) and 53.6 (s.d. 5.4) and the BMI was 26.14 (s.d. 4.3) and 25.7 (s.d. 3.3), respectively. Multiple shape variants of the hip were significantly (P < 0.05) associated with future THR in both the CHECK (modes 4, 11, 15, 17 and 22) and Chingford (modes 2 and 17) cohorts. Mode 17 [odds ratio (OR) 0.51 (95% CI 0.33, 0.80) in the CHECK cohort], representing a flattened head-neck junction and flat greater trochanter, could be confirmed in the Chingford cohort [OR 0.41 (95% CI 0.23, 0.82)]. Modes 4 and 15 of the CHECK cohort also showed non-significant trends in the Chingford cohort. CONCLUSION: Several baseline shape variants are associated with the future need for THR within a cohort. Despite differences in participant characteristics, radiographic protocol and follow-up time, we could validate at least one shape variant, suggesting that SSM is reasonably transferable between cohorts.