We retrospectively reviewed the records of all consecutive patients receiving free-hand PSI in the thoracic or lumbar spine from December 2010 through July 2011. Thirty-six patients (17 males and 19 females) were included in this study. Free-hand PSI was performed by a young academic spinal surgeon. The index surgeon had performed spinal surgeries for two years by himself, excluding the period of his fellowship program. The index surgeon had practiced fluoroscopy-guided PSI during his first year. He then learned the free-hand technique from the senior author and practiced this skill on cadavers [14]. Three times he implemented the free-hand technique under the guidance of the experienced surgeon. He has used the free-hand technique continuously since then.

Postoperative computed tomography (CT) scans were obtained the day after surgery to evaluate the position of the pedicle screws by 3 mm thickness. The accuracy of each pedicle screw placement was objectively evaluated in three-dimensional CT scans. A prior investigation divided the pedicle violations into three grades as follows: grade I, screws within the pedicular cortical margin; grade II, screws that extended <2 mm beyond the pedicular cortex; and grade III, screws that extended ≥2 mm beyond the pedicular cortex [5]. In another study, the cortical breach was classified as <2 mm, 2-4 mm or >4 mm [18]. Other studies categorized the pedicle breach in 2-mm increments [4,15]. Similar to previous studies of the accuracy of PSI, we defined a breach as an extension of more than 2 mm outside the pedicular margin (Fig. 1). The direction of the cortical breach, such as medial, lateral, superior, or inferior, was also noted in each case. The malposition of the pedicle screws was determined by two independent spine surgeons who reviewed the CT scans.

The index surgeon performed free-hand PSI as described by Kim et al. [14,19]. The technique is described in detail in our previous cadaveric study [20]. The first important component of a successful thoracolumbar screw placement is the meticulous exposure of the posterior elements along the subperiosteal plane. The spine was bilaterally exposed from lateral 1 cm from the transverse processes and to the lateral border of the superior articular process. The entry point of the vertebra was at the junction of the bisected transverse process and the lamina at the lateral border of the pars. Detailed entry points by each vertebra were followed based on the method by Kim et al. [14]. The curved pedicle probe was initially pointed laterally as a safety measure to avoid medial wall perforation (Fig. 2). After inserting the pedicle probe tip approximately 15-20 mm (to beyond the isthmic part of the pedicle), the pedicle probe was removed, and five bony walls were confirmed: a bottom and four walls (medial, lateral, superior, and inferior) through ball-tipped sounder. Then, the tip of the pedicle probe was turned to face medially. The pedicle probe was advanced to a maximum depth of 25-40 mm in the thoracic spine. A ball-tipped pedicle sounding device was utilized to repalpate five distinct bony borders. The screw was slowly placed down the pedicle into the body in the same alignment to confirm that it was threaded properly and to allow for viscoelastic expansion. Intraoperative anteroposterior and lateral radiographs were obtained in each patient to assess the presence of pedicle breaches and to confirm the harmonious position of the screws (Fig. 3). Intraoperative spinal cord monitoring including somatosensory and motor evoke potential was used in cases of deformities, tumor, and thoracic spinal cord-compressing lesion. However, direct screw stimulation using triggered electromyogram was not implemented.