Experimental and numerical investigation of secondary screw perforation in the human proximal humerus

Surgical treatment of proximal humerus fractures remains challenging, with a reported failure rate ranging from 15% to 35%. The dominant failure mode is secondary, i.e. post-operative screw perforation through the glenohumeral joint. A better understanding and the ability to predict this complication could lead to improved fracture fixation and decreased failure rate. The aims of this study were (1) to develop an experimental model for single screw perforation in the human humeral head and (2) to evaluate the ability of densitometric measures and micro finite element (microFE) analyses to predict the experimental failure event. Screw perforation was investigated experimentally in twenty cuboidal specimens cut from four pairs of fresh-frozen human cadaveric proximal humeral heads. A centrally inserted 3.5 mm screw was pushed quasi-statically at a constant displacement rate until perforation of the articular cartilage in each specimen. Force and displacement were recorded and evaluated at both initial screw loosening and perforation events. Bone volume was calculated around and in front of the screw and tip-to-joint distance was measured on the combined pre- and post-instrumentation micro computed tomography (microCT) scans. Implicit linear and explicit non-linear microFE models were created based on the microCT scans. The strength of these densitometric, geometrical and microFE methods to predict the experimental results was evaluated via correlation analysis. The bone volume measures were optimized in a parametric analysis to maximize correlation coefficients. The strongest and quantitatively correct predictions of perforation force (R² = 0.93) and displacement (R² = 0.77) were achieved using the explicit, non-linear microFE models. Linear microFE simulations provided the strongest predictions of loosening force (R² = 0.90). Correlation strengths reached by optimized bone volume measures for predicting experimental force and by tip-to-joint distance for predicting displacement were only slightly inferior compared to the results of microFE models. The strong correlations achieved with densitometric and geometric measures indicate that monotonic perforation of single screws through the articular surface of the humeral head can be well predicted with these easily accessible measures. However, non-linear microFE models delivered even stronger correlations and quantitatively correct predictions of perforation force and displacement. This indicates that if computational resources are available, non-linear simulations may have a high potential to investigate more complex fixations and loading scenarios.