THE ASSOCIATION BETWEEN STATISTICAL SHAPE VARIATIONS OF THE HIP AND THE DEVELOPMENT OF RADIOGRAPHIC HIP OSTEOARTHRITIS WITHIN 8 YEARS OF FOLLOW-UP: DATA FROM 16,112 HIPS IN THE WORLD COACH CONSORTIUM
M.M.A. van Buuren, F.D.E.M. Boel, N.S. Riedstra,H. Ahedi, N. Arden, S.M.A. Bierma-Zeinstra, C.G. Boer, F.M. Cicuttini, T.F.Cootes, D.T. Felson, W.P. Gielis, G. Jones, S. Kluzek, N.E. Lane, C. Lindner,J. Lynch, J. van Meurs, A.E. Nelson, M.C. Nevitt, E.H. Oei, J. Runhaar, T.D. Spector,J. Tang, H. Weinans, R. Agricola.
Hip osteoarthritis (OA) is a highly prevalent joint disease, and a leading cause of physical disability. Partly due to a lack of knowledge on its etiology, very limited effective preventive strategies currently exist. Hip morphology has been marked as an important risk factor for the development of hip OA, with variants such as cam morphology and acetabular dysplasia being the most established morphological risk factors. However, there may be other unknown morphological variants or a combination of morphological variants that also pose a risk. Statistical shape modeling (SSM) allows to quantify the whole projected shape of the proximal femur and hemipelvis, and maybe used to identify these variants. The purpose of this study was to determine the association between baseline hip shape and the development of definite radiographic hip OA within 8 years of follow-up.
We used data from the Worldwide Collaboration on OsteoArthritis prediCtion for the Hip (World COACH) consortium. The WorldCOACH consortium was established to collect and harmonize all available individual participant data from 9 prospective cohort studies with sequential pelvic or hip imaging available. For the current study,we included the 6 cohorts with baseline pelvic radiographs and radiographic OA scores available within a maximum of 8 years. Standardized anteroposterior (AP) pelvic and/or hip radiographs were taken at baseline and at a follow-up visit between 4-8 years in each included cohort. Scores for radiographic hip OA were already available for each cohort, either by Kellgren & Lawrence grade, Croft grade, or a modification of either. For this analysis, we harmonized available scores into “definitely no OA” (any score 0), “doubtful OA” (any score 1), or“definite OA” (any score ≥2 or total hip replacement). We only included hips without OA (any score 0) at baseline to avoid capturing any variations in hip shape resulting from OA. An automatic BoneFinder® search model was used to annotate all baseline radiographs, outlining the bony shape with 80 points per hip. The same software was used tocreate an SSM, providing the population mean shape and arange of independent shape variations called shape modes. The SSM was built from all hips without OA at baseline that hadsufficient quality radiographs for reliable annotation. We analyzed the first 22 shape modes that together explained 90% ofthe total shape variance, and applied Bonferroni correction to account formultiple testing. The associations between each baselinehip shape mode and the development of definite radiographic hip OA were estimatedusing a generalized linear mixed-effects model with three levels: hip side,person, and cohort. The results are expressed as odds ratios (OR) for havingincident OA at follow-up, adjusted for baseline age, sex and BMI.
From the 6 included cohorts (Table 1), we had data on 67,214 hips. We excluded 22,186 hips with doubtful or definite OA and 17,484 hips with missing OA scores at baseline, leaving 27,544 hips with definitely no radiographic OA at baseline. After exclusion of hips with incomplete baseline data (n=2,022), doubtful OA or missing OA scores at follow-up (n=10,090), we included 16,112 hips from 10,151 participants (72% female). The mean age was 63 (±8.3) years and mean BMI was 26.8 (±4.2) kg/m2 at baseline. After a mean of 6.9 years of follow-up, 777 (4.8%) of the included hips had developed incident definite radiographic hip OA. From the first 22 shape modes, 6 shape modes (modes 4, 8, 12, 13, 15, 21) showed statistically significant associations with the development of incident radiographic hip OA (Table 2), even after Bonferroni correction with a p-level of <0.0022 (Figure 1).
In this unique, large individual participant data analysis within the World COACH consortium, baseline hip shape, as quantified by SSM, was associated with the development of radiographic hip OA. Identified shape variants included previously identified risk factors such as cam and pincer morphology. Interestingly, other shape variants were also associated with increased radiographic hip OA incidence, suggesting we need to look further than the established morphological risk factors.