Meta-analyses of fracture risk assessment tools according to outcome specifications
Tool | Outcome specifications (BMD/site/sex) | Number of studies | Number of participants | Meta-analysis: random effect model AUC (95% CI) | Heterogeneity, I2 |
---|---|---|---|---|---|
FRAX (10-year prediction) | Y/MOP/W | n=539 41 43 44 60 | 14 224 | 0.67 (0.64 to 0.71)* | 80.2%* |
N/MOP/W | n=739 41 42 44 47 48 76 | 24 726 | 0.65 (0.63 to 0.68)* | 67.6%* | |
N/Hip/W | n=939 41–44 47 48 53 57 | 131 244 | 0.74 (0.68 to 0.80)* | 94.3%* | |
Y/Hip/W | n=539 41 44 53 57 | 115 611 | 0.79 (0.73 to 0.85)* | 93.3%* | |
N/MOP/M | n=245 47 | 11 199 | 0.63 (0.60 to 0.66)* | 0.0% | |
N/Hip/M | n=245 47 | 11 199 | 0.71 (0.65 to 0.77)* | 40.8% | |
Y/MOP/B | n=346 51 | 276 786 | 0.63 (0.60 to 0.66)* | 97.1%* | |
Y/Hip/B | n=346 51 | 276 786 | 0.77 (0.73 to 0.81)* | 69.8%* | |
N/MOP/B | n=346 51 | 276 786 | 0.61 (0.57 to 0.64)* | 96.3%* | |
N/Hip/B | n=346 51 | 276 786 | 0.67 (0.61 to 0.73)* | 94.7%* | |
GARVAN-GRX (10-year prediction) | Y/Hip/W | n=268 39 | 5574 | 0.74 (0.61 to 0.87)* | 88.2%* |
Y/MOP/W | n=339 68 69 | 6932 | 0.70 (0.64 to 0.75)* | 93.8%* | |
Y/MOP/M | n=268 69 | 5010 | 0.73 (0.68 to 0.78)* | 59.0% | |
QFracture (10-years prediction) | N/MOP/W | n=333 70 | 1 778 570 | 0.81 (0.78 to 0.834)* | 97.8%* |
N/MOP/M | n=233 70 | 1 741 983 | 0.72 (0.67 to 0.76)* | 99.2%* | |
N/Hip/W | n=333 42 70 | 1 779 154 | 0.89 (0.88 to 0.89)* | 96.3%* | |
N/Hip/M | n=233 70 | 1 741 983 | 0.87 (0.86 to 0.88)* | 71.0% |
Moderate heterogeneity: Higgins I² ∼50%; high heterogeneity, Higgins I² ∼75%.
*p<0.05.
AUC, area under the curve; B, both sexes; BMD, bone mass density; Hip, hip fractures; M, men; MOP, major osteoporotic fractures (MOPs are defined differently for the different instruments); N, without BMD; W, women; Y, with BMD.