The study was approved by the Ethics Review Committee of Peking University People's Hospital with an approval number of 2019PHB240. All the written informed consents were acquired from the patients.
We prospectively recruited 44 patients (male: female= 14: 30) with OVCF from the traumatic orthopedic department at a tertiary grade-A hospital. The patients with a long time between traumatic and surgery were tended to be included in this research. Patients with infection or malignancy were excluded from this study. The affected vertebrae were T5 in 1 patient, T9 in 4 patients, T10 in 1 patient, T11 in 3 patients, T12 in 9 patients, L1 in 14 patients, L2 in 8 patients, L3 in 1 patient, L4 in 3 patients, L5 in 2 patients. All the patients underwent the X-ray and MRA for the thoracic or lumbar spine. The subjects were divided into the IVC group and the non-IVC group according to the MRI.
Sample size calculation
For comparing the occlusion rate of compression vertebral artery between the non-IVC group and IVC group. Previous literature revealed that the occlusion rate of vertebral artery in patients with IVC was 57.8% . The incidence of lumbar artery occlusion was 19%-27% . Therefore, we assumed that the incidence of artery occlusion in non-IVC patient was 0.3, which is a little higher than the normal vertebral occlusion rate for the compression fracture. The incidence of artery occlusion in the IVC group was 0.6. The α was 0.05 and β was 0.2. The number of IVC group and the non-IVC group was 1:1. The calculated sample size was as follows: number of arteries in non-IVC group was 40 and IVC group was 40. 
The characteristics of the patients were age, gender, body mass index (BMI), and fracture levels. The comorbidities affecting the blood supply of the patients were recorded to reflects the blood supply condition. The vertebral avascular risk factors assessment included hypertension, diabetes mellitus, coronary heart disease, cerebral infarction, hormone usage, and smoking . Each factor was regarded as 1 point and the vertebral avascular risk factor was the sum of all scores of the patient.
All data collection including the X-ray or CT, MRI, MRA results were reviewed by the double-blind method. The IVC data and vertebral occlusion condition were separately collected by two experienced surgeons. Compression rate (CR) was calculated according to previous study . Fracture severity was graded as grade 1: mild (<25% collapse); grade 2: moderate to severe (>25% collapse) . The diagnosis of IVC was an area of low-intensity on T1-weighted and high-intensity or low-intensity on T2-weighted MRI in fracture vertebra (Fig. 1).
There were pairs of segmental arteries of each thoracic and lumbar vertebrae except the L5 . MRA covered T10 to L4 segments and corresponding arteries condition were recorded. Each vertebral artery condition was classified into patency, narrow, occlusion and was scored into 1, 0.5, 0, respectively (Fig. 2). The compression vertebral segmental artery occlusion condition was observed and recorded by an experienced surgeon twice with an interval of more than one month. The intraclass correlation efficiency in the segmental artery occlusion was 0.788 (P < 0.001). The vertebral artery condition from the T10 to the L4 of each patient were all recorded. The compression vertebral segmental arteries were labeled. The occlusion rate was divided into the lesion segmental occlusion rate (LSOR) and the total occlusion rate (TOR). The LSOR was the sum of both two sides of the fractured vertebral artery condition dividing the lesion segmental arteries number. The total occlusion rate was the sum of whole vertebrae (T10-L4) artery condition divided by the total artery number.
The basic characteristics were compared between the IVC and non-IVC groups, including age, gender, BMI, vertebral avascular risk factors, thoracolumbar levels (T11-L2), fracture severity, preoperative compression ratio, the time before MRI, LSOR, and TOR.
Univariate analysis was analyzed using SPSS Ver. 22.0 for Windows (IBM Corp. NY, USA). The continuous variables were analyzed with the chi-square tests. The categorical variables were analyzed with the Mann-Whitney U tests. The dichotomic variables were analyzed with the student's t-tests. P>0.05 was regarded as no significant difference.