To explore the biomechanical stability of the medial column reconstructed with the exo-cortical placement of humeral calcar screw by three-dimensional finite element analysis. A 70-year-old female volunteer was selected for CT scan of the proximal humerus, and a wedge osteotomy was performed 5 mm medially inferior to the humeral head to form a three-dimensional finite element model of a 5 mm defect in the medial cortex. Then, the proximal humeral locking plate (PHILOS) was placed. According to distribution of 2 calcar screws, the study were divided into 3 groups: group A, in which 2 calcar screws were inserted into the lower quadrant of the humeral head in the normal direction for supporting the humeral head; group B, in which 1 calcar screw was inserted outside the cortex below the humeral head, and the other was inserted into the humeral head in the normal direction; group C, in which 2 calcar screws were inserted outside the cortex below the humeral head. The models were loaded with axial, shear, and rotational loadings, and the biomechanical stability of the 3 groups was compared by evaluating the peak von mises stress (PVMS) of the proximal humerus and the internal fixator, proximal humeral displacement, neck-shaft angle changes, and the rotational stability of the proximal humerus. Seven cases of proximal humeral fractures with comminuted medial cortex were retrospectively analyzed between January 2017 and December 2020. Locking proximal humeral plate surgery was performed, and one (5 cases) or two (2 cases) calcar screws were inserted into the inferior cortex of the humeral head during the operation, and the effectiveness was observed. Under axial and shear force, the PVMS of the proximal humerus in group B and group C was greater than that in group A, the PVMS of the internal fixator in group B and group C was less than that in group A, while the PVMS of the proximal humerus and internal fixator between group B and group C were similar. The displacement of the proximal humerus and the neck-shaft angle change among the 3 groups were similar under axial and shear force, respectively. Under the rotational torque, compared with group A, the rotation angle of humerus in group B and group C increased slightly, and the rotation stability decreased slightly. All the 7 patients were followed up 6-12 months. All the fractures healed, and the healing time was 8-14 weeks, with an average of 10.9 weeks; the neck-shaft angle changes (the difference between the last follow-up and the immediate postoperative neck-shaft angle) was (1.30±0.42)°, and the Constant score of shoulder joint function was 87.4±4.2; there was no complication such as humeral head varus collapse and screw penetrating the articular surface. For proximal humeral fractures with comminuted medial cortex, exo-cortical placement of 1 or 2 humeral calcar screw of the locking plate outside the inferior cortex of the humeral head can also effectively reconstruct medial column stability, providing an alternative approach for clinical practice.