Perioperative Epidural Analgesia and Outcome After Major Abdominal Surgery in High-Risk Patients : Anesthesia & Analgesia

Secondary Logo

Journal Logo


Perioperative Epidural Analgesia and Outcome After Major Abdominal Surgery in High-Risk Patients

Peyton, Philip J. MBBS, FANZCA; Myles, Paul S. MBBS, MPH, MD, FANZCA; Silbert, Brendan S. MBBS, FANZCA; Rigg, John A. MBBS, FANZCA; Jamrozik, Konrad MBBS, Dphil, FAFPHM, MFPHM; Parsons, Richard MSc, PhD

Author Information
Anesthesia & Analgesia 96(2):p 548-554, February 2003. | DOI: 10.1213/00000539-200302000-00046
  • Free


The influence of perioperative epidural analgesia on outcome after major abdominal surgery has been a source of controversy. A meta-analysis (1) of randomized, controlled trials (RCTs) found that, compared with other analgesic techniques, neuraxial block was associated with significantly decreased perioperative morbidity and mortality. However, there was no significant difference between groups when the analysis was restricted to non-orthopedic surgery, nor when neuraxial block was combined with general anesthesia.

We have recently published the findings of the Multicentre Australian Study of Epidural Anesthesia (the MASTER Anesthesia Trial) (2). This multicenter RCT investigated the influence of perioperative epidural analgesia on outcome in 888 patients undergoing major abdominal surgery who were deemed at high risk because of the presence of one or more important co-morbidities. In comparison with a control group who received IV opioid analgesia, we found no difference in mortality or in the incidence of major morbidity (as defined by our protocol; see Appendix 1), with the exception of the incidence of respiratory failure. However, postoperative analgesia was found to be clinically superior on the basis of pain visual analog scores (VAS) in patients randomized to the epidural group. In the epidural group, mean pain VAS with coughing was 30% less than in the control group in the first 24 h after surgery and 20% less for the remaining 48 h.

Whereas primary analysis of data from the MASTER Trial indicated that case fatality in abdominal surgery is not improved by epidural analgesia, the findings of recent studies have suggested the possibility of a beneficial effect in some categories of postoperative morbidity or where specific preoperative co-morbidities or types of surgery are involved. Another recent large (n = 1021) randomized trial (3) of epidural opioid analgesia demonstrated no difference in mortality or major complications in patients having major abdominal surgery except for a subgroup undergoing aortic surgery in which a reduction in cardiac morbidity, stroke, and respiratory failure was observed. A systematic meta-analysis of RCTs by Ballantyne et al. (4) found a significant reduction in the incidence of pulmonary infection where epidural local analgesia was used. Beattie et al. (5) conducted a meta-analysis of RCTs looking at cardiac outcome and showed a significant reduction in the rate of postoperative myocardial infarction with epidural analgesia. Other authors have found that mortality and cardiac morbidity may be worsened by inadequate or failed regional anesthesia, suggesting that epidural efficacy or expertise with the technique may contribute to postsurgical outcome (6).

Positive findings from meta-analyses should be confirmed by a large RCT (7,8). We therefore examined our data to determine if certain categories of patients within the MASTER Trial showed a benefit in surgical outcome from epidural analgesia suggested by these other reports.


The design and protocol of the MASTER Trial have been described in more detail elsewhere (2,9). The trial involved patients undergoing elective, open, and major abdominal surgery (including esophagectomy). To maximize statistical power, we prospectively selected those at high risk of an adverse outcome through having one or more important preexisting co-morbidities (Appendix 1). Comparison of treatment groups showed no statistically significant differences between them in the distribution of any inclusion criterion (2).

The trial protocol provided guidelines for premedication, intraoperative monitoring, induction, and maintenance of relaxant general anesthesia and reversal of neuromuscular block, blood, fluid and temperature homeostasis, optimization of respiratory and cardiac function, and immediate postoperative medical management. Criteria for extubation were to extubate as early as possible where vital capacity was more than 10 mL/kg and temperature was more than 35°C with stable hemodynamics. By necessity in a large multicenter study, sufficient flexibility was present to allow participating centers to optimize clinical management.

Patients were randomized to receive general anesthesia with intra- and postoperative epidural analgesia or general anesthesia with IV opioid analgesia. The epidural management protocol called for catheter placement before surgery at a level that would provide epidural block two spinal segments above the upper end of the patient’s wound, intraoperative loading with local anesthetic (bupivacaine or ropivacaine), and continuation by infusion of local anesthetic and opioid (pethidine or fentanyl) after surgery for 72 h or more. Postoperative analgesia in the control group was primarily achieved with patient- or physician-controlled IV opioid infusions initially, supplemented by rectal and oral nonsteroidal antiinflammatory drugs, oral narcotics, and paracetamol. The primary end-points studied were death within 30 days of surgery and a broad range of precisely defined perioperative complications (Appendix 2). Relevant data were collected by study nurses at each participating institution, but whether particular end-points had occurred was defined by a computer algorithm at the time of data entry by staff of the Trial Secretariat who were blinded to treatment allocation. The study was approved by the human investigation committee at each of the 25 participating centers, and written informed consent was obtained from all patients randomized.

The frequency of primary end-points was compared in patients allocated to control and epidural groups within the subsets of participants in the trial described below. In addition, we compared the proportions requiring prolonged mechanical ventilation (for a total of more than 24 h, either continuously or in discrete episodes).

A subgroup at high risk of postoperative respiratory complications was identified as all patients who were recruited because of the presence of one or more of the following (Appendix 1): 1) respiratory insufficiency; 2) age over 75 yr having required hospital admission in the previous 2 yr for an acute exacerbation of respiratory disease; and 3) morbid obesity (body weight over twice ideal).

The subgroup of patients at risk for cardiac complications had a documented history of one or more of the following (see Appendix 1): 1) cardiac failure; 2) acute myocardial infarction; 3) myocardial ischemia; 4) age over 75 yr and myocardial infarction at any time; and 5) diabetes mellitus.

All patients who underwent open abdominal aortic surgery were included in a separate subgroup.

We compared the incidence of primary end-points between our control group and those patients randomized to the epidural group who had early epidural failure requiring abandonment of their epidural analgesia and use of an alternative analgesic regimen. Early epidural failure was defined as those patients in the epidural group who reported a VAS pain score of 6 or more and had their epidural infusion terminated within the first 24 h after surgery or in whom insertion of the epidural had failed.

We looked for evidence that experience and presumed expertise with the use of epidural analgesia was related to outcome. This intention-to-treat analysis compared three major hospitals (identified here as Centers A, B, and C) that each recruited in excess of 100 subjects and collectively recruited over half of the total for the trial. Each center audited its total frequency of epidural administration for the calendar year 2000, as well as the total number of abdominal surgical procedures performed, as an index of its experience with epidural analgesia. A treatment-center interaction was sought in the incidence of death and the combined end-point (death or at least one morbid complication).

We compared the mean and median postoperative intubation times, length of stay in intensive care, and length of hospital stay in patients randomized to control and epidural groups.

Comparison of incidence of end-points between treatment groups was performed using χ2 analysis. Odds ratios and corresponding 95% confidence limits were calculated using a logistic regression model. Mean intubation times and lengths of stay were compared using the t-test, and median values were compared using the Wilcoxon’s ranked sum test. Because the distribution of all these variables was skewed by the presence of small but important numbers of patients with very prolonged times, logarithmic transformation of the data was performed to allow more reliable parametric testing. Logistic regression analysis was used to determine whether a treatment-center interaction was present. A P value <0.05 was considered significant.


Table 1 compares the incidence of primary end-points for at-risk respiratory, cardiac, and aortic subgroups and for the trial overall. The only significant difference between control and epidural groups was frequency of respiratory failure. However, there were no significant differences in the frequency of prolonged mechanical ventilation in the trial or any subgroup.

Table 1:
Comparison of incidence of primary end-points for the high risk subgroups

In comparison with controls (n = 441), there was no significant difference in the incidence of the combined end-point (death or at least one morbid complication) in the subgroup with early epidural failure (n = 32), as defined by our criteria (control group, 61%, early epidural failure subgroup, 53%;P = 0.39).

Figure 1 shows the comparison between the three major recruitment centers and their audit data of use of epidural techniques in calendar year 2000. Logistic regression analysis showed no significant relationship between treatment effect and frequency of use in routine practice (P = 0.24 for the combined end-point;P = 0.39 for mortality).

Figure 1:
Comparison of treatment effect (odds ratio [OR], 95% confidence interval [CI]) on the incidence of death and the combined end-point (death or at least one morbid end-point) among the three major recruitment centers. Frequency of epidural use is the number of epidurals performed in routine clinical practice (outside the trial) as a percentage of the total number of elective open abdominal procedures performed in calendar year 2000.

Mean and median postoperative intubation times, length of stay in intensive care, and length of hospital stay are listed in Table 2. There was only a small and not significant difference in duration of intubation between treatment groups. There was no significant difference between them in length of stay in intensive care or in length of hospital stay.

Table 2:
Mean and Median Postoperative Intubation Times, Length of Stay in Intensive Care, and Length of Hospital Stay


We found no difference in the incidence of any major complication or death between control and epidural groups in any of the subgroups we studied, with the exception of respiratory failure. This result is similar to that found for the trial overall (2). However, the reduction in the duration of postoperative intubation with epidural analgesia, although consistent with this, was clinically insignificant. This apparent contradiction may be attributable to the use of commonly accepted definitions of respiratory failure based on arterial blood gases (Appendix 2). An increased arterial carbon dioxide tension is commonplace in postoperative patients, particularly those receiving opioid analgesia, and is probably a less helpful marker of respiratory failure and the need for ventilatory support than in other clinical settings.

Furthermore, no statistically significant difference in need for prolonged ventilation was present either overall or in any subgroup. This probably explains the similar duration of stay in intensive care. The incidence of pneumonia was not significantly different in the trial overall or in our subgroup defined by high respiratory risk. This result from our 888 randomized patients conflicts with the findings of the meta-analysis of Ballantyne et al. (4), which showed an effect in favor of epidural analgesia with local anesthetic among 5 RCTs (n = 215 patients).

We were also unable to demonstrate any difference in the incidence of death or cardiac complications in our subgroups. Again, this is consistent with the findings of the MASTER Trial overall but different from the results of the meta-analysis of Beattie et al. (5). Their pooled data showed a significant reduction in perioperative myocardial infarction with thoracic epidural catheters in comparison with control groups among 11 randomized studies involving 1173 patients. Their inclusion criteria demanded that epidural analgesia continue for at least 24 hours after surgery, but their article does not state how they accounted for mortality among those patients randomized to the epidural group who may have died within the first 24 hours. In our study, close to one-third of postoperative deaths occurred within the first 48 hours (2). A strict intention-to-treat analysis of our data reduced the possibility of survivor bias in our results.

There are a number of examples of large RCTs reaching different conclusions from those of previous meta-analyses of smaller trials addressing the same clinical question. LeLorier et al. (7) found that this occurred 35% of the time among 40 outcomes examined in 19 meta-analyses and 12 subsequent large RCTs. Reasons suggested for these discrepancies include publication bias and inclusion of older, outdated studies (8). Meta-analyses must also attempt to avoid including studies with duplicated data and consider the extent of heterogeneity among studies included. Assimilation of incidental data from studies primarily designed to examine other end-points may increase the possibility of false positive findings. Pogue and Yusuf (10) point out that meta-analyses that include mainly small trials are most vulnerable to positive publication bias and may overestimate effect of treatment because only large between-group differences in a small trial can reach statistical significance.

Our finding that morbidity and mortality were unaffected by epidural analgesia in the aortic surgery subgroup fails to confirm the findings of Park et al. (3). In a secondary analysis, they found a significant reduction in cardiovascular complications, myocardial infarction, stroke, and respiratory failure associated with the use of epidural analgesia. The contrast with our results is interesting given that the overall incidence of complications was approximately half that in our study. The reason for this difference is that the MASTER Trial deliberately selected a high-risk series of patients to maximize the power of the study. For example, in contrast to this study, Park et al. (3) actively excluded patients with recent myocardial infarction, whereas this was one of our inclusion criteria. However, we were unable to show any difference in cardiac or other complications between our treatment groups.

A possible explanation for the difference in findings between these studies and the MASTER Trial is the influence of unblinded assessment of outcomes. Studies of the outcome of epidural anesthesia are open unless their protocol includes a sham epidural catheter, an option we rejected on ethical grounds, as have most previous investigators in this field. However, we largely eliminated observer bias from our study by moving decisions about end-points away from our clinical sites to assessors at the trial headquarters who used strict assessment algorithms and were blinded to treatment group allocation.

Another criticism of the design of many previous randomized trials examining the influence of epidural analgesia on perioperative outcome has been inadequate sample size, resulting in underpowered studies. For instance, the average size of the 141 trials included in the meta-analysis of Rodgers et al. (1) was 67 subjects. Furthermore, incremental advances in many aspects of perioperative care over the last 30 years may make it more difficult to now demonstrate a major improvement in outcome from any one treatment modality.

The design of the MASTER Trial followed the recommendations of Yusuf et al. (11) on the requirements for reliable RCTs in clinical medicine. It measured clinically important end-points (major morbidity and mortality) and was designed to achieve sufficient statistical power to be able to detect a modest difference in the occurrence of these end-points between control and epidural groups. This was achieved by selecting patients at high risk of perioperative complications. The sample size of 888 patients provided more than 80% power to detect a clinically important absolute difference of 10% in the incidence of end-points where the baseline frequency was 50%. The frequent incidence of major complications in our sample confirmed that this power was exceeded for the combined end-point.

Recruitment of such large numbers of selected high-risk patients requires a multicenter approach, which also provides other important advantages. Large broad-based trials are also less likely to suffer type I error, the chance rejection of the null hypothesis, or false positive result. This is illustrated by our analysis to determine whether experience and presumed expertise with epidural analgesia at a participating institution is related to treatment effect. Figure 1 shows that the three major recruitment centers (A, B, and C) had very different frequencies of epidural use in their routine practice, which reflect not only different clinical case loads, but also differences in propensity to use of the technique within similar surgical populations. The center with the smallest proportion of epidurals (Center A) found a borderline improvement with epidural analgesia in the incidence of one or more major complications or death within the MASTER Trial, but this benefit was not confirmed by the other two centers. This finding suggests that including a broad range of centers is an important part of a clinical trial such as this. The series of 177 patients recruited at Center A itself constitutes a relatively large randomized trial of epidural analgesia compared with the bulk of studies conducted on this topic in the past. However, in the overall context of the MASTER Trial, the positive result at this center represents a random finding and indicates how a type I error can arise from a relatively small sample.

In conclusion, the MASTER Trial reveals no important differences in major postsurgical morbidity or mortality between control and epidural groups in the trial or in selected subgroups at increased risk of respiratory or cardiac complications or undergoing aortic surgery. No relationship could be demonstrated between typical frequency of use of epidural analgesia in the three major recruitment centers and treatment effect. We were unable to find any significant improvement in major morbidity or mortality after major abdominal surgery from perioperative epidural analgesia.

Appendix 1.


Table. Appendix 1. Inclusion Criteria for the MASTER Anesthesia Trial

Appendix 2.


Table. Appendix 2. Brief Description of End-point Definitions in the MASTER Anesthesia Trial


1. Rodgers A, Walker N, Schug S, et al. Reduction of post-operative mortality and morbidity with epidural or spinal anesthesia: results from overview of randomized trials. BMJ 2000; 321: 1493–7.
2. Rigg JRA, Jamrozik K, Myles PS, et al. Epidural anesthesia and analgesia and outcome of major surgery: a randomized trial. Lancet 2002; 359: 1276–82.
3. Park W, Thompson J, Lee K. Effect of epidural anesthesia and analgesia on perioperative outcome: a randomized, controlled Veterans Affairs cooperative study. Ann Surg 2001; 234: 560–71.
4. Ballantyne J, Carr D, deFerranti S, et al. The comparative effects of postoperative analgesic therapies on pulmonary outcome: cumulative meta-analyses of randomized, controlled trials. Anesth Analg 1998; 86: 598–612.
5. Beattie W, Badner N, Choi P. Epidural analgesia reduces postoperative myocardial infarction: a meta-analysis. Anesth Analg 2001; 93: 853–8.
6. Bode RH Jr, Lewis KP, Zarich SW, et al. Cardiac outcome after peripheral vascular surgery: comparison of general and regional anesthesia. Anesthesiology 1996; 84: 3–13.
7. LeLorier J, Grégoire G, Benhaddad A, et al. Discrepancies between meta-analyses and subsequent large randomized, controlled trials. N Engl J Med 1997; 337: 536–42.
8. Myles P. Why we need large randomized studies in anesthesia. Br J Anaesth 1999; 83: 833–4.
9. Rigg JRA, Jamrozik K, Myles PS, et al. Design of the multicenter Australian study of epidural anesthesia and analgesia in major surgery: the MASTER Trial. Control Clin Trials 2000; 21: 244–56.
10. Pogue J, Yusuf S. Overcoming the limitations of current meta-analysis of randomized controlled trials. Lancet 1998; 351: 47–52.
11. Yusuf S, Collins R, Peto R. Why do we need some large simple randomized trials? Stat Med 1984; 3: 409–20.
© 2003 International Anesthesia Research Society