Perioperative epidural analgesia for major abdominal surgery for cancer and recurrence-free survival: randomised trial

Study design and participants

A detailed description of the design and 30 day outcomes of the MASTER trial has been previously published.18 19 20 MASTER was a multicentre randomised clinical trial designed to test the hypothesis that combined epidural and general anaesthesia reduces the frequency of a composite end point of mortality and major postoperative complications compared with general anaesthesia and opioid analgesia.19 It enrolled 915 patients having major abdominal surgery between July 1995 and May 2001. Patients were randomly assigned to receive intraoperative epidural anaesthesia and postoperative epidural analgesia (epidural group) or postoperative intravenous opioid based analgesia (no epidural group); all patients received general anaesthesia during surgery.

The protocol provided guidelines for premedication, intraoperative monitoring, site of the epidural (to be selected by the consultant anaesthetist to match the planned incision), use of intraoperative and postoperative epidural local anaesthetics to provide epidural block during and after surgery, induction and maintenance of general anaesthesia, replacement of blood and fluids, optimisation of core temperature and respiratory and cardiac function, criteria for tracheal extubation, and immediate postoperative medical management. With the exception of some pelvic operations, all epidural catheters were inserted in the thoracic region. Postoperative analgesia in the non-epidural group was mostly achieved with patient controlled, opioid based analgesia. In the epidural group, postoperative analgesia was managed with continuous infusions of bupivacaine or ropivacaine, supplemented with fentanyl or pethidine. Epidural analgesia was typically continued for three days after surgery.

For the clinical assessment of epidural block, we recorded minimum and maximum heart rates and systolic blood pressures during surgery in all patients as indicators of the clinical efficacy of intraoperative epidural block. After surgery, we measured intensity of pain at rest and after coughing in all patients by using a 10 cm visual analogue scale twice daily for the first three postoperative days as an indicator of effective pain control. Morphine remains the most widely used intraoperative and early postoperative opioid analgesic. Opioid usage was not recorded in the original MASTER trial database. We were none the less able to determine morphine use in the initial 72 postoperative hours from a single high recruiting site. We converted other opioids to morphine sulphate equivalents at ratios of 10:1 (pethidine) and 1:100 (fentanyl).

In the original study, among 440 patients randomised to the non-epidural group 19 (4.3%) had epidural analgesia established preoperatively or within 72 hours of surgery. Among 447 patients randomised to epidural analgesia, 29 (6.5%) did not receive an epidural, but in only 13 of these had it been unsuccessfully attempted (failure rate 4.6%). Among the remaining patients, the average duration for which the epidural remained in situ postoperatively was 74 hours. In this analysis, we analysed the height of epidural block achieved in the first three postoperative days and report maximal height as another indicator of effective postoperative epidural block. Use of nitrous oxide was not recorded in the original study, but the usual practice at that time was to include it in most cases.

An additional inclusion criterion for the follow-up study was that patients had complete surgical excision of cancer. Types of surgery included oesophagectomy, gastrectomy, hepatectomy, pancreatectomy, colectomy, nephrectomy, cystectomy, radical hysterectomy, and open prostatectomy. We excluded patients who were not having surgery for cancer (for example, aortic surgery) or in whom excision of the cancer was macroscopically incomplete, including those with distant metastasis at the time of surgery.

In this follow-up study, we collected additional data relating to type and size of cancer and extent of spread at the time of the original surgery. This required retrospective review of the medical records of all patients relating to the time of their original surgery, including pathology reports, formal reporting to cancer registries, operation reports, and discharge summaries. We used organ specific tumour-nodes-metastases staging and the American Joint Commission on Cancer tumour grading system to characterise the extent of malignancy and spread of cancer.21

Follow-up of patients

Additional ethics approvals enabled us to search existing databases for evidence of recurrence of cancer, death, and cause of death. We were sensitive to the possible distress associated with late follow-up of patients who had had surgery for cancer many years earlier, including to their families, and the patients’ rights to confidentiality of hospital and cancer registry data. We therefore developed an algorithm that would minimise direct contact with patients or families and protect the confidentiality of individual patients’ data. We sought follow-up data in the following order: the patient’s medical record from the source hospital, hospital surgery or pathology databases (or both), the patient’s general practitioner, state based cancer registry or national death index (or both), letter of introduction followed by telephone contact of the patient, and letter of introduction followed by telephone contact of the patient’s next of kin. We used an approved script for the telephone contact with the patient or next of kin. The ethics committee from one centre amended the protocol to prevent patients being contacted by researchers, as some patients may not have been made aware of their original diagnosis of cancer; another centre permitted patients to be contacted only through the patient’s general practitioner. All other centres’ ethics committees approved the follow-up procedures as outlined in the protocol.

We also searched medical records and registries with a patient identification number, date of birth, and date of surgery. The Australian national death index records multiple causes of death. These include those conditions involved in the morbid train of events leading to death that were classified as the underlying cause, the immediate cause, or any intervening causes and those conditions that contributed to death but were not related to the disease or condition that caused death. We classified the cause of death as primarily due to cancer, cardiovascular disease, sepsis, and other causes.

Study end points

The primary end point of the study was cancer-free survival after surgery; the secondary endpoint was survival (all cause mortality). We calculated recurrence of cancer and survival from the date of surgery until the event occurred or until last known contact. We censored patients who were lost to follow-up at the time of last record. Research staff blinded to exposure status collected all outcome data.

A preliminary evaluation of the original MASTER trial database suggested that about 600 patients had complete surgical excision of a cancer and would be eligible for inclusion in this follow-up study. Most of these patients had primary colorectal cancer, for which five year survival is approximately 50% and 12 year survival is less than 20%. If epidural block reduced recurrence-free survival by 25%, then 250 patients per group would provide 89% power with a two sided α of 0.05. For a reduction from 50% to 37.5% (a relative reduction of 25%) in cumulative five year survival, 247 patients per group (total 494) would be needed to provide 80% power with a two sided α of 0.05. However, a relative reduction of a third (from 50% to 33%) would require only 131 patients per group.

Statistical analysis

We based analyses on the intention to treat principle; we thus included all randomised patients who had potentially curative surgery for cancer. We constructed time to recurrence and, separately, time to death from any cause for both groups with Kaplan-Meier survival estimates and compared them by using the log-rank test. We used a Cox proportional hazards model to adjust for imbalances of potentially important covariates, with assessment of the requisite proportionality assumptions. We planned to include patients’ sex, age, and type of surgery in our adjusted analyses and tests of subgroup interaction. Also, in view of recent publications suggesting a possible effect modification of age groups under 65 years and 65 years or over,14 local spread at the time of surgery,22 and survival pattern after surgery,22 we did additional post hoc exploratory analyses to investigate such effects. Because the outcomes of cancer are usually summarised in terms of five year survival and five year disease-free survival, we also examined these end points, initially with two by two tables assessed with the χ² test and then in a multivariate logistic model with recurrence of cancer or death from any cause within five years of surgery as the end point. We express results as either hazard ratios or risk ratios with 95% confidence intervals.

Subgroup analyses assessed patients’ sex, age groups, and each major organ surgery. For these factors, we tested for interaction by using the Breslow-Day test for homogeneity of odds ratios across strata.23 Because previous studies have suggested a variable effect of epidural block over shorter (<1.5 years) and longer time frames after surgery,22 we assessed this effect in our dataset. We used SPSS for Windows, version 18 for analyses.