Introduction

Spinal metastasis affects up to 14% of cancer patients and potentially leading to disabilities and poor quality of life [1]. The goal of spinal metastasis surgery is to relieve pain, maintain or improve neurological or function status, and quality of life of the patients [2, 3]. The benefit of these procedures is inevitably accompanied by possible complications and mortality which should be cautiously weighed during the surgical decision-making process [4,5,6,7,8,9].

Perioperative complication rate in spinal metastasis surgery is relatively high—reported to range from 5.3% to 51% [4,5,6, 10]. Pereira et al. demonstrated that 30-day complications after surgery for spinal metastasis were associated with worsened survival [6]. Postoperative respiratory failure (PRF) is one of the most common major medical complications that lead to increased mortality, length of hospital stays, and overall cost of treatment [4, 5, 11,12,13]. Arrigo et al. [5] reported that PRF is the most common complication within 30 days of surgery. Currently, the data regarding PRF in spinal metastasis population are still lacking. To the best of our knowledge, the predictive factors of PRF after surgery for spinal metastasis have not been investigated. Given the high impact of PRF on the outcomes following spinal metastasis surgery, it is essential to understand the risk factors that are predictive of PRF to identify high-risk patients more accurately.

This study aims (1) to investigate the prevalence of and predictive factors for PRF following spinal metastasis surgery and (2) to identify the mortality rate of and risk factor for inhospital death in this patient population. We hypothesized that preoperative patient’s comorbidity and surgical parameters would be risk factors for PRF and in-hospital death following spinal metastasis surgery. The identification of these risk factors may guide the surgeon in treatment planning and avoiding this complication and help in perioperative risk counseling. If PRF is anticipated, the treating physician can prepare necessary resources, such as scheduling ICU care, and collaborate the team care to monitor and treat this complication.

Materials and methods

Study design and patient characteristics

This is a retrospective review of medical records of consecutive patients diagnosed with spinal metastasis at a single institution between October 2008 and September 2018. The institutional review board approval was obtained at our center (MURA 2018/1009). The inclusion criteria were patients aged 18 years and older who underwent operative treatment for metastatic disease in the cervical, thoracic, or lumbar spine. Patients with metastases to the spine from the hematologic malignancies—i.e., multiple myeloma and lymphoma—were also included. In patients who underwent multiple surgical procedures for spinal metastatic disease, we only included the first procedure. We excluded patients with preoperative mechanical ventilation dependence and those who underwent only spinal biopsy, stereotactic radiosurgery, and vertebroplasty (Fig. 1).

Fig. 1
figure 1

Flow diagram of patient selection

Full size image

Data collection

Data collected were demographics, American Society of Anesthesiologists (ASA) Physical Status classifications, comorbidities, Charlson comorbidity index, functional status, American Spinal Injury Association (ASIA) Impairment scale [14], length of hospital stay, history of venous thromboembolism, and preoperative blood transfusion requirement. Cancer-related data included type of primary cancer, visceral organ metastasis, central nervous system (CNS) metastasis, vertebral and non-vertebral skeletal metastasis, number of spinal levels involved, index level of spinal cord compression, Epidural Spinal Cord Compression (ESCC) scale, history of chemotherapy/radiotherapy within 30 days prior to the surgery, and cancer prognosis according to Katagiri et al. [15]. Chest radiograph/computed tomography was reviewed for evidence of lung metastasis, pneumonia, pleural effusion, and atelectasis. Laboratory data included hematocrit levels, platelet count, white blood cell count, blood urea nitrogen (BUN), creatinine level, aspartate aminotransferase (AST) level, alanine aminotransferase (ALT) level, total bilirubin (TB), albumin level, prothrombin time (PT), and partial thromboplastin time (PTT). Surgery-related data collected were surgical procedure type, surgical approach, number spinal level operated, cement usage, urgency of surgery, operative time, measured intraoperative blood loss, and intraoperative and postoperative blood transfusion. The amount of intraoperative blood loss for each case was quantified by the attending anesthesiologist. The recorded blood loss was represented by the following formula: measured blood loss = (suction fluid [mL] – irrigation fluid [mL]) + blood loss in surgical gauze measured by gravimetric method [16]. Postoperative complications, included pulmonary complications—i.e., pulmonary edema, pulmonary embolism, pleural effusion, or pneumonia—urinary tract infection, sepsis, and acute renal failure, were also recorded.

The incidence of PRF and in-hospital mortality were recorded. For the purposes of this analysis, the occurrence of PRF was defined as unplanned postoperative intubation (UPI) at any time point during admission or mechanical ventilation dependence for more than 48 h (MVD) postoperatively [11]. In-hospital mortality was defined as any fatality during admission which occur after spinal metastasis surgery.

Statistical analysis

Statistical analysis was performed with STATA version 12.0, StataCorp, College Station, Texas, USA. Frequency (percentage) was used for categorical data. Mean (SD) and median (range) were used to describe continuous data. Pearson Chi-squared or Fisher's exact tests were used for categorical data, and independent t-test or rank-sum Wilcoxon test was used for continuous variables. For significant continuous variables, receiver operating characteristic (ROC) curve was used to determine cutoff values which were then used for logistic regression analysis.

Stepwise logistic regression was used to determine predictive factors for PRFs and in-hospital mortalities. Factors with a p value ≤ 0.1 in univariate logistic regression were then advanced to multivariate logistic regression studies. Due to the different timing of occurrences of UPI and MVD, we created separate logistic regression models for each specific event. Statistical significance was determined at p value < 0.05.

Results

Demographic data

During the study period, 240 patients underwent spinal metastasis surgery at our institution. Four patients who received only spinal biopsy were excluded, leaving 236 patients enrolled into the study. The mean age was 56.6 ± 13.7 years. A total of 126 (53.4%) patients were male. The median length of hospital stay was 12 days. The most common primary cancers were lung (51; 21.6%), breast (29; 12.3%), and prostate (23; 9.8%) (Table 1). Cancer prognosis according to Katagiri classification was favorable in 144 (61.0%) patients. Visceral organ metastasis occurred in 233 patients (98.7%). Lung and CNS metastasis were found in 135 (57.2%) patients and 12 (5.1%) patients, respectively. Preoperative pleural effusion and atelectasis were found in 50 (22.2%) patients and 41 (17.4%) patients, respectively. The patient physical status was classified as ASA grades 1–3 in 204 (86.4%) patients. There were 12 (5.1%) patients with Charlson comorbidity index score ≥ 2. Spinal metastasis occurred at multiple vertebrae in 128 (54.2%) patients. The region of spinal cord compression was thoracic (167; 70.8%), lumbar (46; 19.5%), and cervical (23; 9.8%), respectively. High-grade epidural spinal cord compression occurred in 178 (75.4%) subjects. According to ASIA impairment scale, the preoperative neurologic status was categorized based on functional status as grades A, B, or C (functionally dependent) in 99 (41.9%) patients, and grade D or E (functionally independent) in 137 (58.1%) patients.

Table 1 Patient demographics
Full size table

Surgery-related data

Posterior decompression with instrumentation was performed in 153 (64.8%) patients and posterior decompression alone in 37 (15.7%) patients (Table 2). Posterior approach was performed in 212 (90%) patients, and anterior approach was performed in 17 (7.2%) patients. Spinal instrumentation was performed in 185 (78.4%) patients. The median instrumented spinal level was 4. Emergency surgery within 24 h of admission was performed in 51 (21.6%) patients. The median measured blood loss was 1000 mL. Intraoperatively, the median transfusion requirement was 2 units of red blood. Red blood transfusion was required postoperatively in 76 (32.2%) patients.

Table 2 Surgery-related data
Full size table

Incidence of PRFs and in-hospital mortality and associated patients’ characteristics

PRFs occurred in 26 (11.0%) patients. Among these, 13 (5.5%) patients required MVD postoperatively. Thirteen (5.5%) patients received UPI (Fig. 2). Patients who developed PRFs had a significantly higher in-hospital mortality rate than those who did not develop PRFs (50% [13/26 patients] vs. 0.5% [1/210 patients], p < 0.0001). There were 14 patients with in-hospital mortality, 10 patients had received UPI, and three patients had received MVD prior to their decease. Another patient developed postoperative respiratory failure but denied intubation and was treated with palliative care (Fig. 2).

Fig. 2
figure 2

The occurrence of postoperative respiratory failure and in-hospital mortality among 236 patients. There were 14 patients with in-hospital mortality, 10 patients had received UPI, and three patients had received MVD prior to their decease. Another patient developed postoperative respiratory failure but denied intubation and was treated with palliative care

Full size image

Among those who required mechanical ventilation > 48 h, there were significantly higher percentage of patients classified as ASA grade ≥ 4 (46.1% vs. 13.2%, p < 0.01) and having index spinal compression at cervical spine region (30.7% vs. 8.5%, p = 0.01). Patients with UPI had longer hospital stay (41 days vs. 14 days, p < 0.01), higher rate of preoperative pleural effusion (46.2% vs. 19.7%, p = 0.04), proportion of patients classified as ASIA A, B, or C (76.9% vs. 39.9%, p < 0.01), and lower preoperative hematocrit (33% vs. 36%, p = 0.03).

In-hospital mortality was observed in 14 (5.9%) patients (Fig. 3). Fatality occurred in cases with relatively longer length of hospital stay (37 days vs 11 days, p < 0.01), and higher proportion of patients with ASA classification ≥ 4 (38.5% vs 12.1%, p = 0.03), and ASIA grade A, B or C (76.9% vs 39.9%, p = 0.02). The patients who died during admission had higher creatinine level (0.93 mg/dL vs 0.80 mg/dL, p = 0.04), PT (13.0 secs vs 12.0 secs, p = 0.01) and lower serum albumin than those who survived (30.3 g/L vs 34.1 g/L, p = 0.04).

Fig. 3
figure 3

Incidence and risk factors of postoperative respiratory failure and in-hospital mortality

Full size image

Predictive factors for postoperative requirement for mechanical ventilation > 48 h

Univariate analysis of possible factors predicting the requirement for MVD is shown in Table 3. Significant risk factors observed in univariate logistic regression include ASA classification ≥ 4 (p = 0.02), surgery involving cervical spine (p = 0.04), and measured blood loss > 2000 mL (p = 0.01).

Table 3 Univariate logistic regression analysis
Full size table

Stepwise multivariate logistic regression reveals the independent predictors of MVD as the followings: intraoperative blood loss > 2000 mL (odds ratio [OR] 12.28, 95% confidence interval [CI] 2.88–52.36), surgery involving cervical spine (OR 9.58, 95% CI 1.94–47.25), and ASA classification ≥ 4 (OR 6.59, 95% CI 1.85–23.42) (Table 4).

Table 4 Multivariate logistic regression analysis
Full size table

Predictive factors for unplanned postoperative intubation

Using univariate analyses, possible factors predicting UPI were determined from baseline patient demographics, surgical factors, and postoperative patients’ status and are shown in Table 3. Sequential multivariate logistic regression reveals that the risk factors UPI are postoperative sepsis (OR = 20.48, 95% CI 3.47–120.86), CNS metastasis (OR = 10.21, 95% CI 1.42–73.18), lung metastasis (OR = 7.18, 95% CI 1.09–47.40), and postoperative pulmonary complications (OR = 6.85, 95% CI 1.44–32.52) (Table 4).

Predictive factors for in-hospital mortality

Using univariate analyses, possible factors predicting for in-hospital mortality were determined from baseline patient demographics, surgical factors, and postoperative patients’ status and are shown in Table 3. These factors were advanced to multivariate analysis, the independent risk factors for in-hospital mortality were found as postoperative sepsis (OR = 13.15, 95% CI 2.92–59.26), CNS metastasis (OR = 10.55, 95% CI 1.54–72.05), and postoperative pulmonary complications (OR = 9.87, 95% CI 2.35–41.15) (Table 4).

Discussion

Although spinal metastasis surgery has undergone numerous advances to improve patient outcomes, the surgery in these patients carries a risk of major medical complications and death [4,5,6,7,8,9]. PRF, defined as MVD and UPI, is a life-threatening complication that a treating physician may encounter postoperatively [11]. To date, it remains unclear what risk factors attribute to PRF following spinal metastasis surgery. In the present study, we identified [1] independent risk factors for MVD including ASA grade ≥ 4, measured intraoperative blood loss > 2000 mL, and surgery involving cervical spine; and [2] independent risk factors for UPI including preoperative lung metastasis, preoperative CNS metastasis, postoperative sepsis, and postoperative pulmonary complications. Furthermore, we found that patients who developed PRF had a significantly higher in-hospital mortality (50% vs. 0.5%, p < 0.0001).

ASA classification has been proved to be helpful for preoperative evaluation in patients undergoing spine surgery [17, 18]. We found that patients with preoperative ASA grade ≥ 4 had a higher risk for MVD which is consistent with prior studies [17, 19]. Fu et al. [17] reported that respiratory complication rates were significantly higher with ASA higher grade patients. In addition, there was eightfold increase in the respiratory complication rate in ASA grade 4 versus ASA grade 1 patients. Schuss et al. reported that preoperative ASA grade ≥ 3 is an independent predictor for early postoperative complications in patients who undergo surgery for spinal metastasis [19].

Blood loss in spinal metastasis surgery has been reported to range from 500 to 2300 ml [20,21,22]. We found that intraoperative blood loss greater than 2000 ml is an independent surgical factor for MVD. Our findings were supported by the previous studies [23, 24]. Raw et al. indicated that blood loss greater than 30 ml/kg is a surgical factor that predicts the need for postoperative ventilation for spinal surgery in adults [23]. In adult spinal deformity surgeries, Baron et al. revealed that blood loss greater than 20 ml/kg also necessitates postoperative mechanical ventilation [24]. Spinal surgery in highly vascular metastatic cancer associates with a risk for major perioperative blood loss [25]. Therefore, preoperative embolization is recommended to reduce this event especially in debulking procedure or even in separation surgery [26]. Minimizing intraoperative blood loss may reduce postoperative mortality and other uneventful complications [27,28,29].

Regarding location of spinal metastasis surgeries, our results demonstrated that cervical spine surgeries are a predictor of MVD. We believe that the need for MVD following cervical spinal metastasis surgeries might be explained by 1) for cervical tumors, an anterior approach is often performed. The cervical spine is a complex area with the nearby trachea and neurovascular structures. Cervical metastases often invade anterior vertebral column making surgical approach difficult and increase risk of airway complication, and 2) spinal cord compression from metastatic tumors in the cervical spine can result in respiratory failure and may contribute to the risk of requiring prolonged ventilation [30, 31]. Recently, Hussain et al. revealed that cervical spine tumor was associated with the high rate of pulmonary complication and mortality [32]. Yang et al. revealed that prolonged postoperative endotracheal intubation was the most common major early postoperative complication following cervical spine metastasis surgeries [33].

Our results demonstrated that postoperative sepsis was associated with postoperative UPI and in-hospital mortality. Sepsis is considered as an important cause of reintubation following any surgical procedure [34]. Ramos et al. revealed that patients who developed sepsis postoperatively were 6.9 times more likely to receive reintubation following adult spinal deformity surgery [35].

Postoperative pulmonary complication—including pneumonia, pleural effusion, and pulmonary edema—was also an independent risk factor for UPI and in-hospital mortality in this study with an OR of 6.85 and 9.87, respectively. Incidence of pulmonary complication following spine surgery for metastases ranges from 1 to 18% [8, 9, 33, 36, 37]. Serious adverse event that may occur in these patients includes the acute respiratory distress syndrome (ARDS) in which the associated mortality is over 30% [38]. Management of ARDS consists of mechanical ventilation and oxygen therapy, adequate sedation, bronchodilators, lung recruitment maneuver, and prone positioning [39].

Recognition of the presence of visceral metastases is clinically important for decision-making especially for determining the aggressiveness of operation. Our results revealed CNS metastasis and lung metastasis as predictors for UPI. Moreover, CNS metastasis was also the predictive factor for in-hospital mortality in this study. Brain metastasis and lung metastasis have been considered as poor prognostic factors in many previous studies [40,41,42,43]. Recently, Schoenfeld et al. [7] indicated that the New England Spinal Metastasis Score (NESMS) can be used as a means of prognosticating the risk of major perioperative complications, including UPI, and mortality within 30 days of surgery. One parameter of the NESMS is visceral metastases. We believe that CNS or lung metastasis may affect respiratory function of the patients. To avoid complications and deaths, special attention should be paid to those with advanced cancer especially for patients with CNS or lung metastasis.

Our findings highlight the need for adequate recognition of spinal metastasis surgical candidate who may develop PRF and urges for adequate prevention strategies. However, there are limitations to our study. First, it was a single-center retrospective analysis, future prospective studies with a larger sample size are needed to corroborate our findings. Second, our study population comprises different cancer types resulting in heterogeneity. However, in univariate analysis, no significant difference of cancer types was associated with PRFs. Third, the measurement of blood loss by gravimetric method tended to provide lower blood loss volume [16, 44]. We acknowledge that the accuracy and reliability of blood loss measurement is of utmost importance. Currently, there is no gold standard for quantifying intraoperative blood loss [16, 44]. In our study, we eagerly used the same technique systematically evaluated intraoperative blood loss for the whole cohort, representing the consistency of our clinical practice. However, the optimization to improve methodology for blood loss evaluation may still be required in the future study [45]. Finally, our center is the academic center; thus, our results may not reflect outcome pattern in the community setting.

In conclusion, in this study, PRFs and in-hospital death occur at a relatively high rate following spinal metastasis surgery. Predictive factors for MVD included ASA classification ≥ 4, surgeries involving cervical spine, and intraoperative blood loss > 2000 ml. On the other hand, CNS metastasis, lung metastasis, postoperative sepsis, and postoperative pulmonary complications were predictive factors for UPI. Identification of risk factors may help for surgical decision-making and preoperative patient counseling with regard to risks of morbidity and mortality.