Objective:

The aim of this study was to analyze risk factors and avoidance techniques for proximal junctional kyphosis (PJK) and proximal junctional failure (PJF) in the upper thoracic spine with an emphasis on bone mineral density (BMD) as estimated by Hounsfield units (HU).

Methods:

A retrospective chart review identified patients at least 50 years of age who underwent instrumented fusion extending from the pelvis to an upper instrumented vertebra (UIV) between T1 and T6 and had a preoperative CT, pre- and postoperative radiographs, and a minimum follow-up of 12 months. HU were measured in the UIV, the vertebral body cephalad to the UIV (UIV+1), and the L3 and L4 vertebral bodies. Numerous perioperative variables were collected, including basic demographics, smoking and steroid use, preoperative osteoporosis treatment, multiple frailty indices, use of a proximal junctional tether, UIV soft landing, preoperative dual-energy x-ray absorptiometry, spinopelvic parameters, UIV screw tip distance to the superior endplate, UIV pedicle screw/pedicle diameter ratio, lumbar lordosis distribution, and postoperative spinopelvic parameters compared with age-adjusted normal values.

Results:

Eighty-one patients were included in the study (21 men and 60 women) with a mean (SD) age of 66 years (6.9 years), BMI of 29 (5.5), and follow-up of 38 months (25 months). Spinal fusion constructs at the time of surgery extended from the pelvis to a UIV of T1 (5%), T2 (15%), T3 (25%), T4 (33%), T5 (21%), and T6 (1%). Twenty-seven patients (33%) developed PJK and/or PJF; 21 (26%) had PJK and 15 (19%) had PJF. Variables associated with PJK/PJF with p < 0.05 were included in the multivariable analysis, including HU at the UIV/UIV+1, HU at L3/L4, DXA femoral neck T-score, UIV screw tip distance to the superior endplate, UIV pedicle screw/pedicle diameter ratio, and postoperative lumbar lordosis distribution. Multivariable analysis (area under the curve = 0.77) demonstrated HU at the UIV/UIV+1 to be the only independent predictor of PJK and PJF with an OR of 0.96 (p = 0.005). Patients with < 147 HU (n = 27), 147-195 HU (n = 27), and > 195 HU (n = 27) at the UIV/UIV+1 had PJK/PJF rates of 59%, 33%, and 7%, respectively.

Conclusions:

In patients with upper thoracic-to-pelvis spinal reconstruction, lower HU at the UIV and UIV+1 were independently associated with PJK and PJF, with an optimal cutoff of 159 HU that maximizes sensitivity and specificity.