Our Institutional Review Board approved this study, and the requirement for informed consent was waived. This study was a retrospective review of 1446 consecutive patients who underwent CT of the spine, including the thoracolumbar and lumbosacral junctions, for back pain between March 2012 and July 2016. We excluded 1346 patients who did not have spinal variants (n = 1269) or additional images (n = 77) (i.e., whole-spine radiographs, magnetic resonance [MR] imaging scans or computed tomography [CT] imaging scans) that would allow us to count the number of presacral mobile vertebra (Fig. 1A). Posteroanterior chest and supine abdominal radiographs were regarded as wholespine radiographs because these images could be used to count the number of presacral mobile vertebra (Fig. 2A, B). Although it was not possible to count down from the first cervical vertebra, the number of cervical vertebra is known to be fixed at seven (1234567). Therefore, on the posteroanterior chest and supine abdominal radiographs, the number of presacral mobile vertebra could be determined by counting down from the first rib-bearing thoracic vertebra and adding the seven cervical vertebra. In addition, posteroanterior chest radiographs included the cervicothoracic junctions. Thus, the first rib could be differentiated from a cervical rib because the former arises from a superiorly directed transverse process, while the latter articulates from an inferiorly directed transverse process (18). For the same reason, chest CT and abdominal CT images were also regarded as whole-spine CT scans. The existence of spinal variants in each case was evaluated by a radiologist, based on axial spine CT with curved planar reformatting (CPR) images and images to determine the number of presacral mobile vertebra (i.e., both posteroanterior chest and supine abdominal radiographs, both chest CT and abdominal CT images, whole-spine radiographs, MR imaging, or CT imaging). Spinal variants included the following: 1) alteration of the typical distribution of 12 thoracic and five lumbar vertebra (Fig. 3), 2) anomalous total number of presacral mobile vertebra (e.g., 22, 23, or 25) (Fig. 1), and 3) TLTV or LSTV (Fig. 2). Park et al. (19) suggested differentiating vertebral segments at the thoracolumbar junction as TLTV or non-TLTV (thoracic or lumbar) segments on the basis of definitive criteria for differentiating most caudal ribs in consideration of several variants of the thoracolumbar junction. In accordance with the previous literature (1319), a thoracic vertebra is defined by the presence of paired ribs that are 3.8 cm or greater in length and originate from the facet at the pedicle (or the body) of the vertebra (Figs. 1B, C, 2C, 3C). A lumbar vertebra was identified as a vertebra without articulating ribs, without facets for ribs on the pedicle (or the body of the vertebra), and with fused transverse processes (Figs. 1D, 2F, 3D) (13). TLTV were defined as any segments that did not meet the criteria for thoracic or lumbar vertebra in the thoracolumbar junction (Fig. 2D) (19). CPR images were used for the accurate measurement of rib length. Rib length was measured by drawing a line at the midpoint of the rib width from the proximal head of a rib to the distal body (Fig. 2E) (19). Only type I or II LSTV, as defined by the Castellvi classification (20), were included in this study, because we focused our analysis on the number of presacral mobile vertebra. Finally, 100 consecutive patients with spinal variants were included in the study.

All CT images were acquired with a dual-source CT scanner (Somatom Definition Flash; Siemens Healthcare, Erlangen, Germany). Patients were scanned with a detector configuration of 128 × 0.6 mm with combined modulation (CARE Dose 4D; Siemens). Standard-of-care spinal CT imaging was performed at 275 quality reference mAs. All CT imaging was performed using the following parameters: pitch, 0.7; gantry rotation time, 0.5 seconds; section width, 1 mm; and reconstruction increment, 0.7 mm.

The first lumbar vertebra, as established through morphologic analysis of the thoracolumbar junction, was labeled by each reader as the 21st segment in 65.0% of the patients, as the 20th segment in 31.0% of the patients, and as the 19th segment in 4.0% of the patients (Fig. 4). Inter-observer agreement between the two readers in determining the first lumbar vertebra on the basis of morphologic analysis was nearly perfect (κ value: 1.00). Although there was a discrepancy in the distribution of vertebra in one case, it did not affect establishment of the first lumbar vertebra. In that case, 11 thoracic vertebra, one TLTV, and five lumbar vertebra were identified by reader 1, while 12 thoracic vertebra and five lumbar vertebra were identified by reader 2. This disagreement in values was because of a discrepancy in the measurement of the rib length by the two readers. Reader 1 considered the 19th vertebra as a TLTV, because the vertebra had paired separated bones, one of which was less than 3.8 cm in length, although it articulated with the facet at the pedicle. In contrast, reader 2 considered the 19th vertebra as a thoracic vertebra, because reader 2 regarded the paired ribs as being 3.8 cm or greater in length.

The 20th vertebra, which was identified by counting in a cranial-to-caudal manner by reader 1, was morphologically classified as a TLTV in 60.0% of the patients, as the first lumbar segment in 31.0% of the patients, as the second lumbar segment in 4.0% of the patients, and as a thoracic segment in 5.0% of the patients (Fig. 5).