femur, radius/ulna, vertebrae, humerus, pelvis or ribs) in TZD users. Time-dependent adjustments were made for age, gender, comorbid- ity and drug use. Results: Of all current TZD users, 215 experienced a fracture during follow-up. Compared with non-users of antidiabetic drugs, current TZD users had a 1.6-fold increased risk of osteoporotic fracture: fully adjusted (adj.) HR 1.6 (95% CI 1.4–1.9). The adj. HR of hip fracture was 1.5 (95% CI 1.2–1.8). The increases in risks of fracture were larger in women than in men: adj. HR of osteoporotic fracture was 1.7 (95% CI 1.4–2.0) in women and 1.3 (95% CI 1.0–1.9) in men. When we compared current TZD use with current use of other antidiabetic medication, we found that the risks were lower: adj. HR 1.4 (95% CI 1.2–1.6) for osteoporotic fracture, and adj. HR 1.2 (95% CI 1.0–1.5) for hip fracture. Conclusions: Users of TZDs were at increased risk of osteoporotic fracture. However, our results suggest that the findings were partially confounded by underlying disease severity. When observational studies assess risk of fracture in patients with TZDs, the severity of type 2 diabetes mellitus should be taken into account. doi:10.1016/j.bone.2010.09.124 OP36 Prediction of bone loss by MR perfusion imaging and spectroscopy James Griffith 1 , David K.W. Yeung 1 , Jason C.S. Leung 2 , Timothy C.Y. Kwok 3 , Ping C. Leung 2 1 Department of Imaging and Interventional Radiology, The Chinese University of Hong Kong, Hong Kong SAR, China 2 Jockey Club Centre for Osteoporosis Care and Control, The Chinese University of Hong Kong, Hong Kong SAR, China 3 Department of Medicine and Therapeutics, The Chinese University of Hong Kong, Hong Kong SAR, China Background: Our ability to select those people most prone to bone loss is limited, with clinical risk factors such as age, weight, weight loss over two years and baseline bone mineral density (BMDa) being the best-recognised markers of future bone loss. Several studies have indicated a link between reduced bone perfusion, increased marrow fat and osteoporosis. Objective: The aim of this study was to determine if marrow fat content or perfusion indices assessed on MR imaging at baseline are predictive of future bone loss. Methods: From an initial cohort of 120 subjects who underwent dual X-ray absorptiometry (DXA) as well as MR perfusion imaging and spectroscopy of the right hip at baseline repeat DXA of the hip at two years (52 subjects) and four years (45 subjects) was performed. Results: Percentage reduction in femoral neck BMDa at two and four years post-baseline was greater in subjects with below median acetabulum E slope (r = 0.517, p =0.0003) or muscle E max (r = 0.306, p = 0.043) after adjusting for baseline BMDa, age, weight and weight change. Baseline MR parameters correlated with reduc- tion in BMDa at four years as well as traditionally applied clinical risk factors. Acetabulum enhancement slope (E slope ), femoral neck maximum enhancement (E max ), and marrow fat content at baseline had sensitivities of 89%, 81% and 72% , respectively, at distinguishing between fast (>1% /annum) and slow bone loss. Conclusion: Elderly female subjects with below median perfusion indices at baseline had increased femoral neck bone loss at four years. Selected perfusion indices and marrow fat content have a high sensitivity in discrimi- nating between fast and slow bone losers. doi:10.1016/j.bone.2010.09.125 OP37 Analysis of COL1A1, COL1A2, CRTAP and LEPRE1 mutations in Chinese patients with osteogenesis imperfecta Hao Zhang, Zhen-lin Zhang Department of Osteoporosis & Bone Diseases, Metabolic Bone Disease & Genetic Research Unit, Shanghai JiaoTong University Affiliated Sixth People's Hospital, Shanghai, China Objective: Osteogenesis imperfecta (OI) with dominant inheri- tance is caused by mutations in COL1A1 or COL1A2, the genes that encode type I collagen. More recently, CRTAP and LEPRE1 mutations were detected in patients with OI with recessive inheritance. However, these findings have been mostly restricted to Western populations, and CRTAP and LEPRE1 mutations have never been reported in an Asian population. The aim of this study is to investigate mutations of OI patients in China. Methods: In the present study, mutations analyses were performed for COL1A1, COL1A2, CRTAP and LEPRE1 in a cohort of 55 unrelated Chinese patients with OI type I–IV. Results: A total of 53 mutations were identified in COL1A1 and COL1A2, including 42 mutations in COL1A1 and 11 mutations in COL1A2. Of the mutations identified, 24 mutations are novel. Most mutations resulted in substitutions for glycine (14 in COL1A1 and 7 in COL1A2). The c.2299 G>A (p.Gly767Ser) mutations in COL1A1 were found in four unrelated OI probands (11%), which indicates that the Gly767Ser mutation is a mutation hot-spot for Chinese patients with OI. In two independent samples without collagen mutations, we found two novel compound heterozygous mutations in LEPRE1. Interestingly, one patient who carried the LEPRE1 mutation was 24 years old and is the oldest reported carrier of this mutation so far. Conclusions: Our findings indicate that COL1A1 and COL1A2 mutations are common in Chinese patients with OI, and LEPRE1 mutations are relatively common in Chinese autosomal recessive OI. The correlation between genotype and phenotype is still unclear in Chinese patients with OI, and this correlation will be determined in further studies. doi:10.1016/j.bone.2010.09.126 OP38 Genome-wide bivariate association study suggested the DDX6 gene affecting both obesity and osteoporosis Xi Li 1 , Jian Li 2 , Shu-Feng Lei 1 , Xiang-Ding Chen 1 , Hui Shen 2 , Yong-Jun Liu 2 , Yao-Zhong Liu 2 , Qing Tian 2 , Hong-Wen Deng 1,3 1 Laboratory of Molecular & Statistical Genetics; and the Key Laboratory of Protein Chemistry & Developmental Biology of Ministry of Education, College of Life Sciences, Hunan Normal University, Changsha, Hunan, China 2 School of Medicine, University of Missouri-Kansas City, Kansas City, MO, USA 3 Center of Systematic Biomedical Research, Shanghai University of Science & Technology, Shanghai, China Objective: Obesity and osteoporosis are closely related diseases. As a potential mechanism underlying the close relationship between obesity and osteoporosis, pleiotropic genes may exist to influence the risks of both obesity and osteoporosis. Currently, univariate genome- wide association studies (GWAS) are normally performed in analyz- ing obesity and osteoporosis, which ignores the genetic correlations and co-predisposition between these two diseases. A newly devel- oped bivariate GWAS strategy can account for co-variation of two correlated disease phenotypes and usually is more powerful than regular individual univariate association strategy in identifying pleiotropic genes underlying diseases of shared genetic susceptibility. Here, we performed the genome-wide bivariate analysis to identify Abstracts / Bone 47 (2010) S361–S375 S373