Exploratory analysis of predictors of revision surgery for proximal junctional kyphosis or additional postoperative vertebral fracture following adult spinal deformity surgery in elderly patients: a retrospective cohort study

The rates of revision surgery due to proximal junctional kyphosis/failure (PJK/PJF) following correction and fusion surgery for adult spinal deformity (ASD) is high, and its risk assessment and countermeasures are necessary but still insufficient [1‐7]. The potentially modifiable risk factors are greater curvature correction, combined anterior-posterior spinal fusion, hybrid instrumentation (proximal hooks and distal pedicle screws), fusion to the sacro-pelvis, thoracoplasty procedure, and residual sagittal imbalance [4, 5, 7]. Non-modifiable risk factors include older age (> 55 years) and severe preoperative sagittal imbalance [4, 5, 7]. Other less well-established but likely risk factors of PJK/PJF following ASD surgery are low bone density, high body mass index, and presence of a comorbidity [4, 5], and the risk of revision surgery due to complications associated with not only PJK/PJF due to adjacent segment disease but also additional remote level vertebral fracture following posterior instrumentation fusion surgery has been a concern for elderly patients with low vertebral bone density [8, 9]. Bone density has been normally evaluated in an anteroposterior projection of the lumbar vertebra using dual-energy X-ray absorptiometry (DXA), but there are some problems with lumbar DXA in ASD patients. It is unclear whether it reflects the bone density of the lower thoracic vertebrae, in which upper instrumented vertebra (UIV) of ASD surgery is frequently present. Besides, no study in ASD patients on the bone density in the vertebrae around the UIV has been reported. The use of spinal CT substituting for lumbar DXA to complement this disadvantage has been reported, and a significant positive correlation between the T-score on DXA and Hounsfield units (HU) on spinal CT was reported [10]. Bone density on spinal CT can be measured at any vertebral level [10‐13], and it can be investigated separately in the vertebral body and pedicle. In addition, accurate measurement is possible even in the presence of spinal deformity. Given the risk for possible neurological damage as well as severe back pain or impaired quality of life, PJK/PJF or additional remote level vertebral fracture is particularly serious complications for ASD patients. The etiology of PJK/PJF is multifactorial as no study has evaluated a single factor that strongly and consistently predicts their development. Furthermore, there remains conflicting evidence with regard to whether the number of levels fused, the UIV implant types, or the location of the UIV influence the risk of PJK/PJF development. Thus, we aimed to identify factors associated with revision surgery following correction and fusion surgery for ASD in elderly patients. Demographics, clinical data, and radiographic variables were analyzed.

Mean age at the time of surgery was 73.0 and 74.2 years old, respectively (P = 0.55) (Table 1). Groups A and B included 14 (100%) and 36 (90%) female patients, respectively (P = 0.28). Group B had a greater mean BMI (P = 0.01). Indication for the index surgical procedure included degenerative scoliosis in 4 and 12 patients, degenerative kyphosis in 4 and 8, degenerative kyphoscoliosis in 2 and 8, and posttraumatic kyphosis in 2 and 10, in groups A and B respectively, and posttraumatic kyphoscoliosis in 2 in group B, and iatrogenic kyphosis in 2 in groups A (P = 0.18). The mean YAM value on DXA was 92.9% and 96.2%, respectively (P = 0.62). The number of patients with osteoporosis or osteopenia evaluated on lumbar DXA were 6 (42.9%) and 18 (45%), respectively (P = 1.0). The mean bone densities of T8 (P = 0.002) and T9 (P = 0.01) on spinal CT were significantly lower in group A (Fig. 2), whereas those of L4 (P = 0.002) and L5 (P = 0.01) were significantly higher in group A. The mean bone densities of the T8 (P = 0.03) and T9 (P = 0.02) pedicles were significantly lower in group A (Fig. 3), whereas that of the L3 pedicle (P = 0.04) was significantly higher in group A. Preoperative treatment of osteoporosis was performed in 6 (42.8%) and 22 patients (55%), respectively (P = 0.54). Regarding the radiographic parameters before surgery, no significant difference was noted in SVA, PT, TK, LL, PI, PI-LL, or UIV + 2 angle between the groups. The proportions of the SRS-Schwab ASD classification were not also significantly different between the groups (Table 2).

Regarding the perioperative factors, mean operation time were 282 and 304 min in groups A and B, respectively (P = 0.37), mean blood losses were 698 and 1128 mL, respectively (P = 0.02), mean number of levels fused were 6.9 and 7.6, respectively (P = 0.24), and UIV implant type was PS in more than half of the patients (P = 0.30) (Table 3). The level of uppermost PS was not significantly different between the groups (P = 0.32) (Fig. 4). PSO was applied to 14% and 30% in groups A and B, respectively (P = 0.21), and sacral fusion was applied in 57% and 70%, respectively (P = 0.28). No significant difference was noted between the groups in any radiographic parameter immediately after surgery.

Regarding the postoperative factors, the mean durations of follow-up after surgery were 40 and 37 months in groups A and B, respectively (P = 0.61) (Table 4), additional vertebral fracture developed in 85% and 25%, respectively (P < 0.001), PJK developed in 85% and 60%, respectively (P = 0.07), and PS loosening occurred in 100% and 95%, respectively (P = 0.54). As for the PJK levels in revision surgeries, there were 1 case at T7, 2 cases at T8, 5 cases at T9, 2 cases at T10, and 2 cases at T12, respectively. Loosening of the uppermost PS occurred in 100% and 75%, respectively (P = 0.04), and it occurred within 3 months after surgery in 71% and 40%, respectively (P = 0.04). Postoperative treatment of osteoporosis was performed in 100% and 77%, respectively (P = 0.04), and 14% and 40% of them were treated with teriparatide, respectively (P = 0.07).

In this study, patients who required revision surgery due to complications associated with PJK or additional remote level vertebral fracture after ASD surgery had significantly lower mean bone densities of T8 and T9 vertebra and significantly higher mean bone densities of L4 and L5 vertebra. The rates of loosening of the uppermost PS within 3 months after surgery and additional vertebral fracture were significantly higher in patients who required revision surgery.

Since the sagittal alignment of the spine is often corrected largely in ASD surgery and this loads a large physical stress on the adjacent intervertebral segments, PJK is likely to develop [5, 7, 16]. PJK caused by bone failure and implant/bone interface failure are strongly influenced by the bone strength of UIV and nearby vertebrae. Bone strength is generally evaluated based on the bone density. Bredow et al. measured the mean bone density of vertebrae on spinal CT in 365 patients aged 59 years on average treated with PS fixation of one or two levels and observed that PS loosening occurred in the vertebrae with a mean bone density of 116 HU, but it did not occur in those with a mean bone density of 132 HU [12]. Kumano et al. reported that lower BMD assessed using HU values from preoperative CT is associated with adjacent segment fracture after spinal fusion surgery [8]. The present study clarified that in patients who required revision surgery due to complications associated with PJK or additional remote level vertebral fracture after ASD surgery, PS loosening occurred early and the mean bone density was low in the T8 and T9 vertebral bodies and high in the L4 and L5 vertebral bodies. For these patients, certain countermeasures may be necessary, such as strengthening of osteoporosis treatment and augmentation of PS.

The incidence of uppermost PS loosening was significantly higher in group A, and it occurred within 3 months after surgery in 71% and 40% in groups A and B, respectively, being significantly higher in group A. PS loosening does not necessarily cause PJK or additional vertebral fracture, but it was clarified that the probability of revision surgery is high when uppermost PS loosening occurs within 3 months after surgery. To prevent uppermost PS loosening early after surgery, risk assessment and certain countermeasures are necessary. It has been reported that the threshold of the mean bone density of the vertebra for applying augmentation to prevent PS loosening is about 120 HU [12]. In our study, the most frequent level of the uppermost PS was T10 and the mean bone densities of the vertebral body were 143 and 129 HU in groups A and B, respectively, being not significantly different between the groups, and both values were higher than 120 HU. However, the mean bone densities of the T8 and T9 vertebral bodies were about 100 HU in group A whereas these were about 140 HU in group B, being significantly different. Besides, there were more than half of the patients whose PJK levels were at T8 (2cases) and T9 (5cases) out of the 12 cases in group A. These results clarified that PJK should be strongly influenced by the bone density of nearby UIV. Also, mean bone densities of the T8 and T9 vertebral bodies whose uppermost PS levels except for T9 and T10 were 107 and 109 HU in group A, respectively, being significantly low. It was suggested that the risk of uppermost PS loosening early after ASD surgery is high in elderly patients regardless of the level of the uppermost PS when the mean bone densities of the T8 and T9 vertebral bodies are lower than about 140 HU.

Ohtori et al. reported the effect of drugs for treatment of osteoporosis on PS loosening [17]: PS loosening occurred in 26% and 25% of patients on CT in the oral risedronate treatment and control groups at 12 months after surgery, respectively, whereas the incidence was 13% in the teriparatide injection group, being significantly lower. In our study, no significant difference was noted in the incidence of PS loosening between patients with and without preoperative treatment of osteoporosis, and on the contrary, fewer patients received postoperative treatment of osteoporosis in group B. However, limited to teriparatide, the drug was more frequently used in group B, although the difference was not significant. To prevent uppermost PS loosening early after surgery, pre- and postoperative treatment with teriparatide or certain PS augmentation such as polymethylmethacrylate cement injection and/or expandable PS may be necessary [1, 6, 18].

The limitations of this study were the retrospective design, multiple factors associated with revision surgery to be tested, and small number of patients who were included in the study. In the baseline characteristics of SRS-Schwab ASD classification, there were more kyphosis patients in group A and more scoliosis patients in group B. Kyphosis patients tend to have highly degenerated lumbar vertebra, and sclerosed changes could occur in curved vertebra in scoliosis patients. Patients in group A could have more sclerosed lumbar vertebra, and patients in group B could have more sclerosed thoracic vertebra. The higher mean bone density of L4 and L5 in group A and T8 and T9 in group B could be caused by the sclerosed vertebra. Therefore, if we performed a matched pair analysis in each curve type, results could be different from the current study. However, baseline characteristics excluding BMI, preoperative parameters, and postoperative parameters excluding blood loss were not statistically different between the groups. Second, there were some biases in deciding to perform additional surgery. There were many patients who developed PJK in both group A (85%) and group B (60%) in the current study. Clinical symptoms of patients with PJK/PJF were variable, and patients who needed revision surgery could have not decided to have additional surgery. However, patients who received revision surgery had severe symptoms such as severe pain, neurological deficits, or progressive sagittal deformity. Thus, there were no patients with mild symptoms included in group A who needed not have revision surgery. Another limitation was the diversity of UIV implant types, such as bilateral PS, bilateral hooks, and unilateral PS with unilateral hook, and this made the risk assessment of PS loosening based on the mean bone density of UIV difficult. Therefore, we compared the mean bone density of the extensive region from T8 to the ilium. Another limitation was the slightly broad range of the UIV level from lower thoracic to upper lumbar vertebrae. We considered that the inclusion of patients with UIV at an upper lumbar level is more practical because no significant bias was noted in the basic background between the groups.

We focused on the factors associated with revision surgery for proximal junctional kyphosis or additional postoperative vertebral fracture following correction and fusion surgery for ASD in elderly patients. In patients who required revision surgery, the mean bone densities of vertebral bodies and pedicles at T8 and T9 were significantly lower. The mean bone density represented in HU on spinal CT may be useful for risk assessment of and countermeasures against revision surgery after ASD surgery in elderly patients.