Patterns of Change Over Time in Knee Bone Shape Are Associated with Sex.

BACKGROUND: Knee osteoarthritis (OA) is more common in females than in males; however, the biological mechanisms for the difference in sex in patients with knee OA are not well understood. Knee shape is associated with OA and with sex, but the patterns of change in the bone's shape over time and their relation to sex and OA are unknown and may help inform how sex is associated with shape and OA and whether the effect is exerted early or later in life.Questions/purposes (1) Does knee shape segregate stably into different groups of trajectories of change (groups of knees that share similar patterns of changes in bone shape over time)? (2) Do females and males have different trajectories of bone shape changes? (3) Is radiographic OA at baseline associated with trajectories of bone shape changes? METHODS: We used data collected from the NIH-funded Osteoarthritis Initiative (OAI) to evaluate a cohort of people aged 45 to 79 years at baseline who had either symptomatic knee OA or were at high risk of having it. The OAI cohort included 4796 participants (58% females; n = 2804) at baseline who either had symptomatic knee OA (defined as having radiographic tibiofemoral knee OA and answering positively to the question "have you had pain, aching or stiffness around the knee on most days for at least one month during the past 12 months") or were at high risk of symptomatic knee OA (defined as having knee symptoms during the prior 12 months along with any of the following: overweight; knee injury; knee surgery other than replacement; family history of total knee replacement for OA; presence of Heberden's nodes; daily knee bending activity) or were part of a small nonexposed subcohort. From these participants, we limited the eligible group to those with radiographs available and read at baseline, 2 years, and 4 years, and randomly selected participants from each OAI subcohort in a manner to enrich representation in the study of the progression and nonexposed subcohorts, which were smaller in number than the OA incidence subcohort. From these patients, we randomly sampled 473 knees with radiographs available at baseline, 2 years, and 4 years. We outlined the shape of the distal femur and proximal tibia on radiographs at all three timepoints using statistical shape modelling. Five modes (each mode represents a particular type of knee bone shape variation) were derived for the proximal tibia and distal femur's shape, accounting for 78% of the total variance in shape. Group-based trajectory modelling (a statistical approach to identify the clusters of participants following a similar progression of change of bone shape over time, that is, trajectory group) was used to identify distinctive patterns of change in the bone shape for each mode. We examined the association of sex and radiographic OA at baseline with the trajectories of each bone shape mode using a multivariable polytomous regression model while adjusting for age, BMI, and race. RESULTS: Knee bone shape change trajectories segregated stably into different groups. In all modes, three distinct trajectory groups were derived, with the mean posterior probabilities (a measure of an individual's probability of being in a particular group and often used to characterize how well the trajectory model is working to describe the population) ranging from 84% to 99%, indicating excellent model fitting. For most of the modes of both the femur and tibia, the intercepts for the three trajectory groups were different; however, the rates of change were generally similar in each mode. Females and males had different trajectories of bone shape change. For Mode 1 in the femur, females were more likely to be in trajectory Groups 3 (odds ratio 30.2 [95% CI 12.2 to 75.0]; p < 0.001) and 2 than males (OR 4.1 [95% CI 2.3 to 7.1]; p < 0.001); thus, females had increased depth of the intercondylar fossa and broader shaft width relative to epicondylar width compared with males. For Mode 1 in the tibia, females were less likely to be in trajectory Group 2 (OR 0.5 [95% CI 0.3 to 0.9]; p = 0.01) than males (that is, knees of females were less likely to display superior elevation of tibial plateau or decreased shaft width relative to head width). Radiographic OA at baseline was associated with specific shape-change trajectory groups. For Mode 1 in the femur, knees with OA were less likely to be in trajectory Groups 3 (OR 0.4 [95% CI 0.2 to 0.8]; p = 0.008) and 2 (OR 0.6 [95% CI 0.3 to 1.0]; p = 0.03) than knees without OA; thus, knees with OA had decreased depth of the intercondylar fossa and narrower shaft width relative to epicondylar width compared with knees without OA. For Mode 1 in the tibia, knees with OA were not associated with trajectory. CONCLUSIONS: The shapes of the distal femur and proximal tibia did not change much over time. Sex and baseline knee radiographic OA status are associated with the trajectory of change in the bone's shape, suggesting that both may contribute earlier in life to the associations among trajectories observed in older individuals. Future studies might explore sex-related bone shape change earlier in life to help determine when the sex-specific shapes arise and also the degree to which these sex-related shapes are alterable by injury or other events. LEVEL OF EVIDENCE: Level III, prognostic study.

Development and Validation of a Multitask Deep Learning Model for Severity Grading of Hip Osteoarthritis Features on Radiographs.

Background A multitask deep learning model might be useful in large epidemiologic studies wherein detailed structural assessment of osteoarthritis still relies on expert radiologists' readings. The potential of such a model in clinical routine should be investigated. Purpose To develop a multitask deep learning model for grading radiographic hip osteoarthritis features on radiographs and compare its performance to that of attending-level radiologists. Materials and Methods This retrospective study analyzed hip joints seen on weight-bearing anterior-posterior pelvic radiographs from participants in the Osteoarthritis Initiative (OAI). Participants were recruited from February 2004 to May 2006 for baseline measurements, and follow-up was performed 48 months later. Femoral osteophytes (FOs), acetabular osteophytes (AOs), and joint-space narrowing (JSN) were graded as absent, mild, moderate, or severe according to the Osteoarthritis Research Society International atlas. Subchondral sclerosis and subchondral cysts were graded as present or absent. The participants were split at 80% (n = 3494), 10% (n = 437), and 10% (n = 437) by using split-sample validation into training, validation, and testing sets, respectively. The multitask neural network was based on DenseNet-161, a shared convolutional features extractor trained with multitask loss function. Model performance was evaluated in the internal test set from the OAI and in an external test set by using temporal and geographic validation consisting of routine clinical radiographs. Results A total of 4368 participants (mean age, 61.0 years ± 9.2 [standard deviation]; 2538 women) were evaluated (15 364 hip joints on 7738 weight-bearing anterior-posterior pelvic radiographs). The accuracy of the model for assessing these five features was 86.7% (1333 of 1538) for FOs, 69.9% (1075 of 1538) for AOs, 81.7% (1257 of 1538) for JSN, 95.8% (1473 of 1538) for subchondral sclerosis, and 97.6% (1501 of 1538) for subchondral cysts in the internal test set, and 82.7% (86 of 104) for FOS, 65.4% (68 of 104) for AOs, 80.8% (84 of 104) for JSN, 88.5% (92 of 104) for subchondral sclerosis, and 91.3% (95 of 104) for subchondral cysts in the external test set. Conclusion A multitask deep learning model is a feasible approach to reliably assess radiographic features of hip osteoarthritis. © RSNA, 2020 Online supplemental material is available for this article.

Associations between alcohol, smoking, and cartilage composition and knee joint morphology: Data from the Osteoarthritis Initiative.

Objective: To determine the cross-sectional associations of alcohol consumption and smoking history with magnetic resonance imaging (MRI) measures of cartilage composition (T2) and joint structure using data from the Osteoarthritis Initiative (OAI). Design: Subjects with radiographic Kellgren Lawrence right knee grades 0-2 were selected from the OAI database, and those with previously analyzed MRI cartilage T2 and semi-quantitative joint morphology gradings (WORMS) were included (n â€‹= â€‹2061). Alcohol consumption was categorized as: no drinks to <1 drink/week, 1-7 drinks/week, >7 drinks/week. Smoking history was categorized as none, current, or former. Linear regression was used to assess the relationships of alcohol consumption and smoking history with both WORMS scores and cartilage T2. Results: Subjects who consumed >7 drinks/week had significantly higher cartilage T2 than subjects who consumed <1 drink/week in the average of all cartilage regions and in three of five individual regions (coefficient range: 0.45-0.90, p â€‹< â€‹0.05). Subjects with moderate alcohol consumption (1-7 drinks per week) had higher cartilage and meniscus WORMS scores than subjects who consumed <1 drink/week (p â€‹< â€‹0.05). Current smokers had significantly higher cartilage T2 compared to non-smokers in the average of all cartilage regions and in three of five individual regions (coefficient range: 0.47-0.74, p â€‹< â€‹0.05). Conclusions: Alcohol consumption (1-7 drinks/week) was associated with worse cartilage and meniscus joint morphology, and >7 drinks/week was associated with elevated cartilage T2. Compared to non-smokers, current smokers had a more degenerated cartilage matrix as evidenced by greater cartilage T2.

Lumbosacral transitional vertebrae: association with low back pain.

PURPOSE: To assess the prevalence and degree of lumbosacral transitional vertebrae (LSTV) in the Osteoarthritis Initiative (OAI) cohort, to assess whether LSTV correlates with low back pain (LBP) and buttock pain, and to assess the reproducibility of grading LSTV. MATERIALS & METHODS: Institutional review board approval was obtained, and informed consent documentation was approved for the study protocol. Standard standing pelvic radiographs that included the transverse processes of L5 were graded according to Castellvi classification of LSTV in 4636 participants (1992 men and 2804 women; aged 45-80 years) from the OAI cohort. These data were correlated with prevalence and severity of LBP and buttock pain. RESULTS: Prevalence of LSTV was 18.1% (841 of 4636), with a higher rate in men than in women (28.1% vs 11.1%, respectively; P<.001). Of the 841 individuals with LSTV, 41.72% were type I (dysplastic enlarged transverse process), 41.4% were type II (pseudoarticulation), 11.5% were type III (fusion), and 5.2% were type IV (one transverse process fused and one with pseudoarticulation). Of the participants without LSTV, 53.9% reported LBP, while the prevalence of LBP for types I, II, III, and IV was 46%, 73%, 40%, and 66%, respectively (P<.05, χ2 test). Types II and IV had higher prevalence and severity of LBP and buttock pain (P<.001). CONCLUSION: LSTV types II and IV positively correlate with prevalence and severity of LBP and buttock pain.