Baseline Characteristics
A total of 180 patients were included in our study. Our patients had an average age of 29.8±9.8 years, a higher proportion (83.3%) of females, and an average BMI of 20.8±3.2 kg/m2. A majority of these patients’ conditions was a direct progression from adolescent idiopathic scoliosis (146 patients, 81.1%). The rest were affected with congenital scoliosis (24 patients, 13.3%), neuromuscular scoliosis (9 patients, 5.0%), or marfanoid scoliosis (1 patient, 0.6%). The onset of symptoms lasted for an average of 15.7 months before being clinically diagnosed. Clinical presentation varied, including back pain (120 patients, 66.7%), neurological symptoms (10 patients, 5.6%) and dyspnea (18 patients, 10.0%). In addition to scoliosis, other observed deformities included spinal stenosis (6 patients, 3.3%), vertebral rotation (19 patients, 10.6%), herniated disk (19 patients, 10.6%), syringomyelia (22 patients, 12.2%) and kyphosis (19 patients, 10.6%). In terms of patients’ medical histories, incidence of smoking (7 patients, 3.9%), heart disease (6 patients, 3.3%), respiratory disease (23 patients, 12.8%), hypertension (3 patients, 1.7%), and anemia (11 patients, 6.1%) were all present. A summary of baseline characteristics is shown in Table 1.
Surgical characteristics
The preoperative assessment showed that 39.6% of the patients in our sample had mild or severe systemic disease (ASA 2-3). The average levels of plasma Hgb and albumin were 131.6±14.4 and 43.1±3.5 g/L, respectively. For correction of deformity, distal fusions of most cases were at L4 or higher (144 patients, 81.4%), with some extending to L5 (27 patients, 15.3%), S1 (3 patients, 1.7%) or S2 (3 patients, 1.7%), with an average of 11.6±2.8 fused vertebrae. Fusion levels, ranging from 3 to 16, could be subdivided into three groups: 3 to 7 (19 patients, 10.7%), 8 to 12 (79 patients, 44.6%) and 13 to 16 (79 patients, 44.6%). Decompression and osteotomy were performed in 6.1% (11 of 180) and 20.6% (37 of 180) of our patients, respectively. The mean operative time was 259.7±81.1 minutes, and the mean length of stay (LOS) was 14.6±4.6 days. With an average estimated blood loss of 755.8±483.1 ml, 162 cases (90.0 %) used an autologous blood transfusion system, 73 cases (40.6%) accepted allogeneic RBC transfusion with 3.0 ± 1.4 U, and 70 cases (38.9%) received pure plasma transfusion, averaging out to 414.1± 134.1 ml. The radiographic parameters were based on anterior-posterior and lateral whole-spine X-rays. Prior to surgery, the average Cobb angle, clavicle angle (CA), sacral obliquity (SO), T1 tilt angle (T1TA), L5 tilt angle (L5TA), T1 pelvic angle (T1PA), coronal vertical axis (CVA), sagittal vertical axis (SVA), thoracic kyphosis (TK), pelvic tilt (PT), sacral slope (SS), lumbar lordosis (LL), pelvic incidence (PI), and pelvic incidence minus lumbar lordosis (PI -LL) were as follows: 66.6±24.1°, 2.3±2.9°, 2.3±2.6°, 7.2±7.4°, 12.4±8.9°, 8.2±6.7°, 15.1±12.4 mm, 22.8±17.0 mm, 33.9±18.2°, 9.1±8.6°, 33.2±10.9°, 49.3±16.1°, 42.3±13.1°, 15.9±14.5°; the respective postoperative data were the following: 30.0±21.9°, 3.2±2.6°, 2.0±2.1°, 6.5±7.3°, 7.0±6.2°, 9.0±6.9°, 15.2±11.7 mm, 20.5±16.7 mm, 33.5±13.3°, 10.5±8.8°, 32.7±7.6°, 47.3±11.6°, 43.2±10.5°, 10.5±9.3°. A summary of surgical characteristics is shown in Table 2.
Perioperative Complications
Twenty-five (13.9%) patients had at least one perioperative complication, while a total of 31 perioperative complications were observed. The majority (21 patients, 84%) experienced a perioperative complication only once, 11 of which were cardiopulmonary-related (35.5%), including seven pleural effusions, one arrhythmia, one congestive heart failure, one pneumothorax, and one atelectasis. Outside of systemic complications, there were also six incision-related surgical complications. Four patients (14%) suffered two or more complications: one case with pleural effusion and ileus; one case with pleural effusion and cerebrospinal fluid leakage; one with pleural effusion, pneumothorax, and arrhythmia; and the last one with pleural effusion, atelectasis, and pulmonary infection. Two patients were readmitted into the hospital during the perioperative period: one for severe pulmonary infection, one was suffering from radiating pain as a result of malposition of a pedicle screw (Table 3).
Different strategies were employed to deal with perioperative complications (Table 4). Closed thoracic drainage proved beneficial for patients with pleural effusion, pneumothorax and atelectasis. To treat congestive heart failure, the patient’s fluid status was closely monitored, and drugs were given to promote diuresis. Infection was treated with antibiotics effective against the strain of bacteria cultured from the site of infection. When presenting with symptoms suggestive of septic shock, patients were sent to the intensive care unit. Acute neurological problems could be rapidly corrected by dehydration and steroid treatment, while chronic neurological symptoms required patience and a nerve-nurturing treatment. If a patient suffered from unbearable pain due to malposition of the implants, a revision surgery was deployed as soon as possible. Generally, all symptoms tended to improve in the follow-up period.
Univariate analysis
Patients were divided into two groups based on whether they had any perioperative complications: a complication-free group (155) and a group with complications (25). The results of univariate analysis investigating the relationships between baseline/surgical characteristics and perioperative complications are shown in Tables 1 and 2. Factors that were found to carry a statistically significant weight in risk prediction were ASA classification (P=0.043), RBC transfusion (P=0.028), total length of stay (LOS) (P=0.004), total EBL (P=0.020), levels of fusion (P=0.038), osteotomy (P=0.003), anatomical grades of resection (P=0.045), preoperative Cobb angle (P=0.015), change in Cobb angle (P=0.003) and CVA change (P=0.005). Predictors with P values < 0.2 were also considered eligible to be factored into risk calculations. Therefore, factors that can affect the incidence of perioperative complications further encompass the duration of symptoms (P=0.071), preoperative L5TA (P=0.088), change in L5TA (P=0.103), postoperative T1PA (P=0.160), change in T1PA (P=0.056), preoperative CVA (P=0.106), postoperative CVA (P=0.143), preoperative TK (P=0.136), change in TK (P=0.113), postoperative PT (P=0.117), change in PT (P=0.171), postoperative SS (P=0.144), postoperative LL (P=0.052) and change in LL (P=0.128).
Multivariate analysis
Factors whose P value < 0.2 in the univariate analysis were selected for multivariate analysis. Observations from clinical experience and previously published research were considered while trying to find the suitable predictors for complications. A binary logistic regression model was used to eliminate influences of confounding factors, and determine the independent predictors of perioperative complications. Finally, we proposed a triple risk factor model: change in Cobb angle (P=0.015, OR=1.058, 95% CI=1.011~1.108), change in CVA (P=0.006, OR=1.066, 95% CI=1.019-1.116) and RBC transfusion (P=0.005, OR=5.631, 95% CI=1.676~18.924) were the three independent risk factors for perioperative complications (Table 5).
Nomogram construction and validation
A prognostic nomogram that integrated the two independent risk factors from the multivariate analysis in the cohort was constructed (Figure 2A). The predictive accuracy for perioperative complication as evaluated by C-index was 0.746, with a 95% CI of 0.639-0.853 (P<0.001) (Figure 2B). Calibration curves for the probability of perioperative complications showed a good correlation between the nomogram-predicted and observed values (Figure 2C).