Proximal femoral geometry and low-trauma hip fracture risk beyond bone mineral density: an updated and expanded systematic review and meta-analysis

Bone mineral density (BMD) remains central to hip fracture assessment, but it does not fully capture the structural mechanics of the proximal femur. An earlier meta-analysis evaluated conventional hip geometry primarily in relation to femoral neck fracture (FNF); however, substantial additional evidence has since accumulated, including studies examining broader hip fracture outcomes and associations evaluated in models incorporating BMD. We aimed to determine whether conventional proximal femoral geometry is associated with low-trauma hip fracture and whether these associations persist beyond BMD. This PRISMA-compliant systematic review and meta-analysis included English-language studies comparing proximal femoral geometry between hip fracture cases and non-fracture controls or reporting geometry-based fracture risk estimates. Random-effects models pooled mean differences (MDs), odds ratios (ORs), and hazard ratios. Subgroup analyses were performed by sex, fracture type, imaging modality, and BMD adjustment. Forty-seven studies comprising 136,026 participants, including 7,426 hip fracture cases and 128,600 controls, were included. Compared with controls, cases had significantly greater hip axis length (HAL; MD, 0.16 cm; 95% CI, 0.09–0.24), femoral neck axis length (FNAL; MD, 0.07 cm; 95% CI, 0.00–0.14), femoral neck width (FNW; MD, 0.08 cm; 95% CI, 0.05–0.11), and neck-shaft angle (NSA; MD, 1.74°; 95% CI, 1.10–2.38). In OR analyses stratified by BMD inclusion in the regression model, HAL (OR, 1.31; 95% CI, 1.20–1.42) and FNW (OR, 1.32; 95% CI, 1.10–1.59) remained significantly associated with hip fracture in models incorporating BMD, whereas FNAL (OR, 1.06; 95% CI, 0.51–2.22) and NSA (OR, 1.37; 95% CI, 0.99–1.88) did not. Associations were more evident in women and FNF subgroups, and were broadly consistent across DXA and radiographic studies. Sensitivity analyses supported the robustness of the pooled estimates; however, heterogeneity was substantial and prediction intervals (PIs) crossed the null for all continuous geometric parameters, indicating limited generalizability across populations and clinical settings. Conventional proximal femoral geometry, particularly length- and width-based parameters, may provide fracture-related structural information beyond BMD. However, substantial heterogeneity and null-crossing PIs limit the generalizability of the pooled estimates. Evidence for fracture-risk associations persisting in models incorporating BMD was strongest for HAL and, to a lesser extent, FNW, whereas evidence for FNAL and NSA was less consistent. These measures should therefore be regarded as complementary, rather than stand-alone, markers of fracture risk.