3% female). We used these data to explore the effect of adjusting for different body mass compartments on the HBM–OA relationship (Supplementary Table 7). Using our age and gender adjusted model, adjustment

for height and then either weight, or fat and lean mass, produced similar degrees of attenuation compared with BMI adjustment. When each parameter was added individually to the regression model, fat mass resulted in the greatest attenuation of the HBM–OA association (similar to that for BMI) whereas lean mass, despite representing a greater proportion of overall mass, appeared less important. If anything, adjustment for individual fat compartments (trunk, Lenvatinib peripheral [arms and legs], android and gynoid) led to less attenuation than adjustment for total fat mass, suggesting that overall weight and fat mass are more important than fat distribution. Patterns of knee compartment involvement were examined first in all knees with KL grade ≥ 2, and then in knees with KL ≥ 3 (definite osteophyte see more plus narrowing) only (Table 4), excluding those HBM cases with a self-reported history of inflammatory arthritis. Predominant

medial compartment disease was the most prevalent pattern in both HBM cases and controls, in whom OA patterns were similar. If anything, amongst narrowed knees, the proportion of medial compartment disease was slightly greater in HBM cases compared with the control group (p = 0.037); however, this association did not persist after age and gender adjustment. 315 X-rays (15 HBM case knees, 300 control knees) were considered to be poor quality in terms Farnesyltransferase of resolution/penetration/artefact etc. A further 210 knees (58 case knees, 152 control knees) had significant rotation or tilt. Excluding all of these knees from the analysis did not materially affect the HBM–knee OA association observed (OR 2.45 [1.82,3.30], p < 0.001 for KL ≥ 2, adjusted for age and gender). Findings for JSN (most likely to be affected by tilt) were also essentially unchanged (data not shown).

A person-level analysis, in which the worst knee per participant was analysed, also gave similar results (Supplementary Table 8). Radiographic knee replacements were excluded from the main analysis; including these knees (n = 32) and grading them as KL = 4 resulted in marginally increased odds ratios for knee OA in HBM (Supplementary Table 9). A small number of HBM cases and family controls reported a history of inflammatory arthritis: excluding these knees from the overall combined analysis (n = 35 HBM case knees, 4 family control knees) again did not materially change our findings (OR 2.33 [1.76,3.09], p < 0.001 for KL ≥ 2, adjusted for age and gender). Data on inflammatory arthritis were not available for the population controls. Restricting the analysis to those HBM cases meeting our index definition on the basis of their hip BMD alone (total hip Z-score ≥ + 3.