It is important to estimate the likelihood that a pediatric fracture is caused by osteogenesis imperfecta (OI), especially the least severe type of OI (type 1).
We reviewed records of 29,101 pediatric patients with fractures from 2003 through 2015. We included patients with closed fractures not resulting from motor vehicle accidents, gunshot wounds, nonaccidental trauma, or bone lesions. Patients with OI of any type were identified through International Classification of Diseases-9 code. We randomly sampled 500 pediatric patients in whom OI was not diagnosed to obtain a control (non-OI) group. We reviewed age at time of fracture, sex, fracture type, laterality, and bone and bone region fractured. Bisphosphonate use and OI type were documented for OI patients. Subanalysis of patients with type-1 OI was performed. The Fisher exact and χ tests were used to compare fracture rates between groups. P< 0.05 was considered significant. Positive likelihood ratios for OI were calculated by fracture pattern.
The non-OI group consisted of 500 patients with 652 fractures. The OI group consisted of 52 patients with 209 fractures. Non-OI patients were older at the time of fracture (mean, 9.0±5.0 y) than OI patients (mean, 5.5±4.4 y) (P< 0.001). OI patients had more oblique, transverse, diaphyseal, and bilateral long-bone fractures than non-OI patients (all P< 0.001). Non-OI patients had more buckle (P=0.013), metaphyseal (P< 0.001), and physeal (P< 0.001) fractures than OI patients. For patients with type-1 OI and long-bone fractures (n=18), rates of transverse and buckle fractures were similar compared with controls. Transverse humerus (15.2), olecranon (13.8), and diaphyseal humerus (13.0) fractures had the highest positive likelihood ratios for OI, and physeal (0.09) and supracondylar humerus (0.1) fractures had the lowest.
Transverse and diaphyseal humerus and olecranon fractures were most likely to indicate OI. Physeal and supracondylar humerus fractures were least likely to indicate OI. Radiographic fracture pattern is useful for estimating likelihood of OI.
LEVEL OF EVIDENCE: