Short and Long-Term Outcomes of Epidural or Intravenous Analgesia after Esophagectomy: A Propensity-Matched Cohort Study

Background and Objectives As a well-established technique for postoperative pain relief, the benefits of epidural analgesia (EDA) have been under debate recently. This study aimed to determine whether EDA could improve perioperative outcomes and survival in patients undergoing esophagectomy. Methods From January 2010 to December 2012, 587 consecutive cases undergoing McKeown-type esohpageactomy were retrospectively identified from a prospectively maintained database. Results After propensity-matching, incorporating baseline characteristics, 178 cases were included in each group, and patients characteristics distributions were well-balanced between two groups. Compared with intravenous analgesia, the use of EDA significantly decreased the incidence of pneumonia from 32% to 19.7% (P = 0.008), and anastomotic leakage from 23.0% to 14.0% (P = 0.029). The change in CRP level of EDA group was significantly decreased (preoperative, 6.2 vs. 6.2; POD 1, 108.1 vs. 121.3; POD 3, 131.5 vs. 137.8; POD 7, 69.3 vs. 82.1 mg/L; P = 0.044). EDA patients had a significantly longer duration of indwelling urinary catheter (P<0.001), and lower levels in both systolic (P = 0.001) and diastolic blood pressure (P<0.001). There weren't significant differences in overall survival (log-rank P = 0.47) and recurrence (Gray-test P = 0.46) between two groups. Conclusions These findings revealed that EDA could attenuate inflammatory response and reduce the incidence of pneumonia and anastomotic leakage after esophagectomy, at the price of delayed urinary catheter removal and lower blood pressure. EDA remains an important component of multimodal perioperative management after esophagectomy.


Introduction
Epidural analgesia (EDA) is a well-established technique for postoperative pain management that has been widely used for decades. The advantages of EDA, such as better postoperative pain relief and improved perioperative outcomes, once led EDA to be considered as the gold standard for pain management after major surgery. It has been suggested that the use of EDA ameliorates perioperative immune suppression, and reduced risk of recurrence and extended survival have been demonstrated subsequently among patients with breast [1] or prostate cancer [2]. With the huge evolution in perioperative management, however, some new evidence suggests that the benefits of EDA are not as significant as previously thought, especially in less invasive operations [3,4]. The protective effects of EDA have been under great debate recently, and the use of EDA has shown a tendency of continuous decrease [5].
Mckeown-type esophagectomy, one of the most common procedures for nonmetastatic esophageal cancer, is a cervicothoracoabdominal procedure and is also one of the most invasive operations with high postoperative morbidity and mortality [6,7]. Pulmonary complications and anastomotic leakage are the most common serious morbidities, and are great challenges to surgeons, even in experienced centers [8]. Major complications contribute to substantial perioperative mortality, and dramatically deteriorate quality of life [9,10]. The importance of minimizing the risk of surgical complications can never be overstated. The investigations on the effects of EDA on postoperative outcomes for patients undergoing esophagectomy have been limited so far.
To test the hypothesis that EDA could inhibit postoperative inflammatory response, and improve perioperative outcomes and survival when compared with intravenous analgesia (IVA) in patients undergoing major surgery (esophagectomy), we performed the present study, by employing a prospectively maintained esophageal cancer database and conducting propensity matching to compensate for the differences in baseline characteristics.

Ethics statement
All study protocols were approved by the Institutional Review Board of Sun Yat-Sen University Cancer Center. Written informed consent was obtained from each patient. All patient data were anonymized and de-identified in a confidential manner.

Study population
Retrospectively screening a prospectively maintained esophageal cancer database which was constructed in 2014, a total of 587 consecutive cases undergoing elective McKeown-type esohpageactomy from January 2010 to December 2012 in Sun Yat-Sen University Cancer Center (Guangzhou, China) were identified. Among them, 543 cases receiving epidural or intravenous analgesia were eligible for this study. The database recorded information regarding sociodemographic data, treatment administration, perioperative parameters and follow-up status. Tumors were staged according to the 7th edition of the AJCC staging system [11]. The Mckeown-type procedure was described previously [10,12]: first involved esophageal mobilization and radical mediastinal lymphadectomy through right thoracotomy or thoracoscopy, and then laparotomy or laparoscopic gastric mobilization. A gastric tube is fashioned with a width of approximately 4-5 cm, and pulled through retrosternal or posterior mediastinal routes to the left neck. A mechanical circular stapler was used to complete the esophagogastric anastomosis.
For patients in the EDA group, epidural catheter was inserted at thoracic 6-10 level before induction of general anesthesia. A test dose of of 3-5 mL of 1% lidocaine was administered as soon as the epidural catheter was in place, a mixture of 0.2% ropivacaine and mophine 0.02 mg/ml was injected epidurally as a loading dose before the end of the operation, then a continuous perfusion of 0.125% ropivacaine with morphine 0.06-0.1 mg/ml were administered through the epidural catheter at a rate of 2 mL/h until 48 h after surgical procedure. For the IVA group, the methods varied with the preference of anesthesiologists. The most common solution was a mixture of sufentanil (1-1.5 ug/ml) plus flurbiprofen (1-1.5 mg/mL), 10-15 ug sufentainl plus 50 mg flurbiprofen as a loading dose before the end of the operation; less frequently used solutions were mixtures of fentanyl (0.008-0.012 mg/ml) and flurbiprofen (1-1.5 mg/ml), 0.1-0.2 mg fentanyl and 50 mg fluriprofen as a loading dose; both were administered by intravenous at a rate of 2 mL/h until 48 h after surgical procedure. Patient-controlled technique wasn't routinely used.

Perioperative management and follow-up
The preoperative workup regularly included physical examination, complete blood cell count, serum biochemistry tests, computed tomography, endoscopic ultrasonography, pulmonary function test and histological diagnosis. Surgical techniques have been described previously [12]. Normally, blood cell count, serum biochemistry tests were performed in postoperative day (POD) 1, 3 and 7, and temperature, heart rate and blood pressure were measured at 7 a.m. everyday.
Pneumonia was defined as clinical manifestation of pneumonia or bronchopneumonia confirmed by a new or progressive infiltrate on chest radiography and a positive sputum culture during postoperative hospital stay [9]. Anastomotic leakage was defined liberally by any extravasation of water-soluble contrast during swallow study, visualization of anastomotic dehiscence or fistulae during endoscopy or visible loss of saliva through the cervical wound [10].
The follow-up protocol was in accordance with our previous studies [13], and performed by the oncologic outpatient clinic or official contact with patients or their relatives by telephone. The last follow-up was April 30 th , 2015, and the median follow-up was 34.9 (interquartile range, 17.7-43.5) months.

Statistical analysis
Categorical variables were expressed as percentage, and compared using χ 2 test. All imbalanced variables with a significance level of P<0.10 on χ 2 test between two groups were included in the logistic regression to calculate the propensity score, modelling the probability of a patient receiving EDA. A one-to-one match without replacement was performed by using nearestneighbor matching method with a caliper of 0.02 (0.2 of standard deviation) as recommended by Dr. Austin [14]. Normal continuous variables were expressed as mean ± standard deviation and compared by t test, and nonnormal ones as expressed as median (range) and compared by Mann-Whitney U test. Two-way repeated-measures analysis of variance (ANOVA) was applied to compare changes in the parameters of two groups.
Overall survival was analyzed with Kaplan-Meier curves and log-rank test. The risk of recurrence (defined as the cumulative incidence of recurrence [CIR]) was estimated using a cumulative incidence function, which accounted for death without recurrence as a competing event [15]. Differences in CIR between groups were assessed using the methods by Gray [16]. Statistical analyses were undertaken using SPSS 22.0 software for windows (SPSS Inc., Chicago, IL, USA) and R version 3.1.0 (http://www.r-project.org/). Statistical significance was set at p<0.05 and all tests were two-sided.

Baseline patient characteristics
The selection and matching process of participants was showed in Fig 1. Among 543 eligible patients, 183 patients received EDA successfully, and the other 360 ones received IVA. Patients in EDA group were more likely to be male (P = 0.012) and undergo open esophagectomy (P<0.001). There was a trend that EDA group had higher proportions of normal preoperative percentage of predicted forced expiratory volume in 1 second (FEV1%) (P = 0.062) and C-reactive protein (CRP) level (P = 0.053). Propensity score were estimated by the above mentioned parameters, and matching based on similar scores produced 178 patients in each group. The distributions of patients' characteristics were well-balanced between EDA and IVA groups after matching. The baseline characteristics of two groups, before and after matching, were summarized in Table 1. Table 2 summarized the perioperative outcomes between the two groups. Among 178 included patients in each group after matching, 35 patients (19.7%) in EDA group and 57 (32%) in IVA group developed pneumonia (P = 0.008). EDA also reduced the incidence of acute respiratory distress syndrome (ARDS), although it didn't reach statistical significance (P = 0.082). The incidence of anastomotic leakage in EDA group was 14.0%, and was significantly lower than that of IVA group (23.0%, P = 0.029).

Short-term outcomes
The median ICU stay was 1.5 days (range, 0-31 days) for EDA patients, and 1 day (range, 0-65 days) for IVA patients (P = 0.063). EDA patients had a significantly longer duration of  Fig 2A). There were marginally significantly lower levels of white blood cell (P = 0.086, Fig 2B) and heart rate (P = 0.055, Fig 2C) in EDA group. Comparisons of changes in mean systolic blood pressure (SBP, P = 0.001, Fig 3A) and diastolic blood pressure (DBP, P<0.001, Fig 3B) revealed significantly lower levels in EDA patients.

Discussion
Esophageal cancer is one of the most common malignancies worldwide, and the incidence of adenocarcinoma has been dramatically increasing in western countries [17]. It is of huge importance to determine the optimal postoperative pain management for esophagectomy, a highly invasive procedure. Although it has been established that EDA provides better acute pain control and perioperative pathophysiology, whether these benefits after esophagectomy could be translated to improvements in short and long-term clinical outcomes has been scantly investigated. Previous studies yielded conflicting results, and were habitually limited by small sample sizes [18][19][20]. The present large-scale propensity-matched analysis reveals that the benefits EDA include attenuated inflammation, reduced risk of pulmonary complications and anastomotic leakage, although it also delays urinary catheter removal and lowers postoperative blood pressure transiently.
The current anesthetic practice in esophagectomy is one lung ventilation (OLV) to facilitate surgical exposure. The ischemia/reperfusion of the ipsilateral lung and high oxygen concentrations of the ventilated lung could trigger an inflammatory response [21]. Systemic inflammatory response plays a crucial role on the postoperative pathophysiological changes, and may result in acute lung injury [22]. Besides, the discharge of sputum after esophagectomy is usually insufficient due to severe wound pain. Collectively, major respiratory complications can reach above 30%, even in experienced centers [8]. The use of EDA provides excellent solutions to these problems. First of all, the lower levels of CRP, white blood cell and heart rate in EDA patients indicate that EDA could attenuate the inflammatory response after esophagectomy, in accordance with the results of various operations [23,24]. In addition, the outstanding control of acute pain after esophagectomy [25] help patients to cough vigorously and discharge sputum timely and sufficiently, preventing pulmonary complications.
Another important finding of current study is the protective effects of EDA on anastomotic leakage, and it could be easily explained. Ischemia of gastric conduit [26] and impairment in oxygen supply [27] are the most important predisposing factors for leakage, besides anastomotic   techniques. Michelet et al [28] used a laser Doppler flowmeter to measure the gastric mucosal blood flow of patient undergoing esophagectomy, and found that EDA improved the microcirculation of the gastric tube. In addition, EDA reduces the incidence of pulmonary complications, maintains adequate oxygen delivery postoperatively, and consequently promotes the tissue healing of anastomosis.
These findings have significant research implications as well. Current clinical investigations and trials on the perioperative outcomes after esophagectomy may neglect the effects of analgesia methods sometimes. Future such studies should obtain detailed data on analgesia method, consider it as a potential confounding factor, and adjust for or stratify by it during analysis.
The favorable prognostic effects of EDA weren't observed on neither overall survival nor disease recurrence among patients with esophageal cancer. We evaluated the contribution of EDA on recurrence in competing risks regression models. This modeling technique accounts for the effect of death without recurrence, which precludes the occurrence of a recurrence event. Competing risks regression provides a more conservative estimate of the effect relative to Cox regression or Kaplan-Meier models [29]. In competing risk regression analysis, EDA didn't reduce the CIR, compared with IVA.
There are some limitations that should be taken into consideration in interpretation of our results. As with all retrospective studies, this study was exposed to selection bias. Although we used propensity-score matching to compensate for some differences in baseline characteristics that may influence the outcomes after esophagectomy, intrinsic biases may still remain. Additionally, although in the minimally invasive esophagectomy (MIE) subgroup, EDA was also associated with lower incidence of pneumonia and anastomotic leakage, but neither reached statistical significance for the small sample size (data not shown). Further studies are still warranted to address this issue in the setting of MIE. Lastly, CRP, white blood cell and heart rate are non-specific markers of systematic inflammation, and the current study failed to measure specific inflammatory markers, such as interleukin-6, due to its retrospective nature.
In conclusion, the present propensity-matched analysis reveals that the use of EDA among patients undergoing esophagectomy, could attenuate the postoperative inflammatory response, decrease the incidence of pneumonia and anastomotic leakage. The main drawbacks of EDA were delayed urinary catheter removal and transient hemodynamic instability. Therefore, EDA remains to be considered as an important component of multimodal perioperative recovery after esophagectomy.
Supporting Information S1 File. Patient data for statistical analyses. (XLS)