Altogether, 1012 abstracts were retrieved from the literature search (Fig. 1). Of these, articles were excluded based on abstract or title. Most of the excluded studies were overlaps between both databases, animals studies, no original articles or were articles investigated other pathologies or included cervical or lumbar factures, or exclusively evaluated non-operative treatment or anterior approaches. Altogether, 78 articles were analyzed completely. Of these articles 26 were additionally excluded, not focusing specifically on the thoracic spine, including geriatric patients or insufficiently describing the method of posterior stabilization. A total of 16 articles analyzed the timing of surgery in polytraumatized patients, reported of patients suffering of neurologic deficits after midthoracic fractures, and evaluated the impact of concomitant thoracic injuries and were excluded. Altogether, 976 articles were excluded (Fig. 1). All 36 remaining original articles, which covered the period from 1971 to 2018 are summarized in Tables 1, 2 and 3. Levels of evidence were defined as described by Bassler and Antes .
A total of five studies dealt with mainly biomechanical aspects of posterior stabilization of thoracic fractures (Table 1). Generally, the thoracic spine is biomechanically stiffer than the other regions of the spine, because of two anatomical characteristics: the first one is the articulation of the head of the ribs with the articular facets of the adjacent vertebral bodies in combination with the radiate ligaments, which attach to the head of the ribs and both adjacent vertebral bodies, and the costotransverse ligaments. The second is the structure of the thoracic cage itself, which increases the resistance to all directions of motion . Watkins et al.  evaluated the amount of stability provided by the rib cage and the sternum. In their study, the intact rib cage provided 40% of the stability of the thoracic spine in flexion–extension, 35% in lateral bending and 31% in axial rotation. A sternal fracture decreased the stability of the thoracic spine significantly by 42% in flexion–extension, 22% in lateral bending and 15% in axial rotation. Berg et al.  observed two clinical cases of combined sternal and thoracic spine fractures, which developed significant kyphotic deformities after nonoperative treatment and postulated the sternal-rib complex as the fourth column of the spine based on the three column theory of the spine by Denis. In general, the structural instability of thoracic fractures is treated with posterior instrumentation 2 levels above and below the fracture site, but in case of intact rib cage a short segment fixation with 1 level above and below the fractured vertebra could be an alternative. Therefore, Perry et al.  created a burst fracture at T9 in eight human thoracic spines (C7–L1) with intact rib cages and tested a long segment instrumentation (3 above, 2 below), a short segment instrumentation (1 above/1 below) with and without vertebral augmentation and vertebral augmentation without instrumentation. In their study, the long segment instrumentation showed a significant reduction of ROM during flexion–extension (− 90%), whereas the other instrumentations only tended to reduce motion. However, Perry et al.  suggested that in case of intact rib cages short segment instrumentation might adequately stabilize the spine. A common strategy to increase stability of short segment fixation is the addition of cross-links or screws at fracture site (index screws). Lazaro et al.  evaluated seven human thoracic spine segments after creating a wedge fracture in five conditions: long segment fixation (2 above/2 below) with cross-link, short segment fixation (1 above/1 below), short segment fixation with cross links, short segment fixation with index screws and short segment fixation with index screws and cross-link. The long segment fixation was significantly stiffer than short segment fixation, but adding index screws to the short segment construct significantly improved stability by 25%. Adding the cross-link increased stability only during axial rotation. Alternative fixation devices or techniques for the thoracic spine have been described, but pedicle screw systems still provide superior biomechanical properties [18, 26]. All relevant articles are summarized in Table 1.
Placement of thoracic pedicle screws
Transpedicular pedicle screws
Seventeen studies evaluated screw positioning, screw implantation and intra-operative control of screw placement (Table 2). The Placement of pedicle screws in the straight ahead technique promoted by Roy-Camille et al.  is associated with a penetrating rate of 41% . Generally, structures at risk were the intercostal vessel (T4–5), esophagus (T5–9), diaphragm, azygos vein (T5–11), inferior vena cava (T11–12) on the right side as well as aorta (T5–12) and esophagus (T4–9) on the left side.
Dwahan et al.  analyzed the effective pedicle diameter and the mean insertion angle comparing three types of insertion techniques (straight ahead, straight forward with angulation in the axial plane, and anatomic with angulation in the axial and sagittal planes) and found the largest effective diameter using the anatomic way.
Additionally, the funnel technique, opening the insertion point to visualize the medial cortex of the lamina, was analyzed in a cadaver study . The authors had a low perforating rate of 10% grade 1 and 0.4% grade 2. Another way to improve the accuracy could be the use of a special drill guide as tool for thoracic pedicle screw placement . Only 5 of 66 screws had a perforation, all less than 2 mm.
Lehmann et al.  analyzed 229 pedicles in an anatomic study. The authors found an ideal starting point 2–3 mm lateral to the midline of the superior articular facet (line between the lateral and the middle third of the superior facet). The cephalocaudal point is more level depending. T7–9 at the cranial border of the transverse process, T6 and T10 between the cranial border and proximal one-third of transverse process, T4–5 and T11 proximal one third of transverse process and T1–3 and T12 bisected transverse process.
In a clinical setting, Bransfold et al.  reported of 1.2% revision surgery rate due to screw misplacement in 245 patients treated because of thoracic fractures in an open technique under fluoroscopic control. Thereby, there seems to be a correlation between pedicle diameter and penetration rate with a 33% misplacement rate in pedicle diameters of less than 5 mm to 11% in diameters between 5 and 7 mm and no misplacement in diameters above 7 mm . Altogether, there seems to be no difference in accuracy of screw placement in thoracic spine between the open and percutaneous technique .
Two studies investigated the benefit of navigation for pedicle screw placement in the thoracic spine [35, 45]. Both studies found significant higher accuracy in the navigated pedicle screw placement technique. Alternatively, intraoperative 3D-Imaging using a cone-beam device can be used to improve accuracy . A total of 3.8% of the pedicle screws were re-implanted due to the findings in the 3D-Scan. No penetration of more than 2 mm was seen postoperatively. Additionally, Fischer et al.  used preoperative CT-guided transpedicular guide wires and reported a high accuracy and a low complication rate.
In contrast, intraoperative electromyographic monitoring could not improve the accuracy of transpedicular screw placement in thoracic spine .
Parapedicular pedicle screws
The insertion of pedicle screws parapedicularly is an alternative to the transpedicular screws placement . The accuracy of this technique under computer-assisted navigation was good and reliable .
Husted et al.  performed a cadaver study for the parapedicular approach. In this technique, the screws were inserted cephalad to the tip of the transverse process and advanced between the transverse process and the rib. The direction of insertion was caudad in an oblique direction following the course of the rib medially to its articulation with the vertebral body under fluoroscopic control. All screws had an extraspinal position and were positioned in the pedicle rib unit.
Alternatively, translaminar screw fixation has been evaluated, which can be inserted with high accuracy under clinically control .
Hu et al.  analyzed the laminar of the high thoracic spine (Th 1–3) in an Asian population. The authors found larger lamina in males than females but sufficient corridors in all cases.
Outcome after posterior stabilization
There were 14 clinical studies evaluating the outcome in patients with unstable thoracic fractures (Table 3). Most studies on non-operative treatment of unstable fractures are historical. They report treatment algorithms that include bedrest for 2–6 weeks [5, 17, 48]. Hospitalisation times with non-operative treatment range from 3 weeks to 3 months [5, 17] and can be reduced to between less than 2 weeks and 3 weeks by operative treatment in patients with isolated thoracic vertebral fractures [13, 14].
After surgical treatment of thoracic vertebral fractures by posterior stabilization in general, non-surgery-related complications occur in 49%–60%, especially in patients with complete or incomplete paraplegia [28, 46]. Deep venous thrombosis is observed in 4%–9% [13, 36]—compared to 24.5% in conservatively treated patients . This results in an in-hospital mortality rate of 6%–8% [46, 47]. Neurologic worsening during follow-up can be reduced from around 2% [5, 48] with non-operative treatment to less than 1% with posterior stabilization [28, 29, 45, 46]. The only complication directly linked to the non-operative treatment is brace-related skin complications in 16% . For surgical posterior stabilization, the need for revision surgery is reported in 0 to 22% [13, 15, 29, 36, 41, 46]. Some authors differentiate between early revision surgery in 4%–19% and late revision surgery in 16%–22%—with late revisions being mainly those for pain and low-grade-infection [13, 14, 28]. One study identified lamina hooks as source of potential neurological complications requiring revision surgery .
Radiological secondary deformity was seen in 94% with non-operative-management, particularly in fractures with associated injuries to the adjacent disc and/or posterior ligaments [17, 48]. With posterior stabilization, a loss of reduction of 1° to 4° was observed after 1 year and of 2° to 4° after 2 years and more [13, 14, 29, 41]. The only study reporting data on bone healing stated a fusion rate of 95% at 12 months on conventional radiographs .
Pain at follow-up after non-operative treatment was reported by 21% to 48% of the patients at 5 years and longer, while this was the case in 35% with a mean of VAS 3 at 15 months after operative treatment with posterior stabilization [5, 8, 17, 48]. Harkönen et al.  reported that 13% of the patients had “poor” mobility at 5 years with non-operative treatment, while Yue et al.  reported “very good to excellent levels of satisfaction with regards to pain, mobility, posture, and activity” 22 months after posterior stabilization.
Only few studies report patient-reported outcomes for operatively treated patients with values for the SF 36—PCS of 36–40 and the SF 36-MCS of 43–56 [36, 41]. No difference in functional outcome was seen between short and long segmental construct. Generally, return to work was less likely in patients with concomitant spinal cord injuries (25% of ASIA A/B/C and 88% of ASIA D/E) . One year after posterior stabilization of a thoracic vertbral fracture, 7.8% received compensation payments for chronic back pain .
There is also some evidence that percutaneous posterior stabilization of thoracic spine fractures is associated with a reduced inflammatory response, less bleeding, shorter hospitalisation time and earlier return to activities of daily living [20, 42].