The optimal analgesic regimen for patients undergoing lumbar decompression and fusion is unclear. A recent systematic review of the literature was unable to show superiority of either epidural analgesia or systemic analgesia after lumbar decompression and fusion.1 When compared with systemic analgesia, some studies have determined that epidural analgesia using local anesthetic alone or in combination with epidural opioids can provide excellent pain control and reduce opioid consumption following conventional open spinal surgery.2-6 On the other hand, other studies have failed to detect any clinically meaningful advantages of epidural analgesia in similar settings.7-11

Presumably the absence of conclusive evidence stems from several factors, most notably heterogeneous samples of surgical patients, variable indications for surgery and surgical techniques, and inconsistent approaches to postoperative analgesia (e.g., limited use of multimodal regimens).12 The objective of this trial was to evaluate the analgesic effects of adding a continuous local anesthetic-based epidural infusion to standard systemic multimodal analgesia and to compare the results with use of systemic multimodal analgesia alone in a patient population undergoing lumbar decompression and fusion. Specifically, our primary objective was to determine if the addition of epidural analgesia to systemic multimodal analgesia would reduce opioid consumption over the first 48 hr postoperatively when compared with systemic multimodal analgesia alone. Our secondary objective was to assess and compare the effects of each regimen on pain scores, opioid-related side effects, and discharge time from hospital.


The study was conducted at Toronto Western Hospital, part of University Health Network, a tertiary care academic health sciences centre. The University Health Network Research Ethics Board approved the study (January 2008), and all patients provided written and informed consent. Adults (18-80 years of age and American Society of Anesthesiologists [ASA] status I-III) who were scheduled to undergo primary one- or two-level lumbar decompression and fusion for pain due to spondylolisthesis were eligible for the study. Exclusion criteria included the inability to provide informed consent, revision fusion, body mass index > 40, language barrier, allergy to local anesthetics, bleeding diathesis, pregnancy, high-dose opioid use (defined as preoperative opioid consumption > 250 mg oral morphine equivalent per day), preexisting neurologic deficits (defined as sensory or motor deficit corresponding to the spinal level of surgery), and history of drug addiction. All patients were recruited from the practice of a single spine surgeon (Y.R.R.). Patients were recruited in the surgeon’s preoperative clinic by a research coordinator.

A randomization sequence was created by an independent statistician using The randomization sequence was stratified by number of surgical levels (one or two) and sex. Permuted blocks of four were used within each stratum. The randomization list was kept in the possession of the independent research pharmacy at Toronto Western Hospital. The pharmacy created sequentially numbered identical kits containing all study materials. When the postoperative analgesic regimen was initiated, the next kit in the sequence was used. The research coordinator and all investigators, participants, care providers, outcome assessors, and data analysts were blinded to treatment group.

All patients were pre-medicated with oral acetaminophen (1,000 mg age < 70 yr, 650 mg age > 70 yr) and oral gabapentin (600 mg age < 70 yr, 300 mg age > 70 yr) as per our institutional preventive analgesic strategy. Subsequently, all patients underwent a standardized induction technique using midazolam 0.03 mg·kg−1 fentanyl 2 μg·kg−1, and propofol 2-3 mg·kg−1 iv. Endotracheal intubation was facilitated with rocuronium 0.6 mg·kg−1 iv; ventilation was maintained with an inspired oxygen concentration of 0.5, and general anesthesia was maintained with desflurane at 1.0 minimum alveolar concentration. Patients were treated with forced-air warming, and intraoperative analgesia was provided with titrated intravenous hydromorphone boluses. Muscle relaxation was reversed using neostigmine 50 μg·kg−1 iv and glycopyrrolate 10 μg·kg−1 iv, while granisetron 1 mg iv was used for antiemetic prophylaxis.

With the exception of the number of levels (stratified to one- or two-level fusion), the surgical procedure was identical for all patients in order to reduce the potential confounding effect of surgical morbidity on acute postoperative pain and narcotic requirements. All patients underwent decompression and fusion with port access and instrumentation (METRx™ fixed 22-mm tubular retractor and the SEXTANT™ percutaneous pedicle screw-rod system, Medtronic, Memphis, TN, USA). Surgery involved a paramedian muscle-splitting approach with decompression, transforaminal lumbar interbody fusion, placement of a single interbody device (CAPSTONE™, Medtronic, Memphis, TN, USA), and stabilization using bilateral percutaneous segmental pedicle screw and rod fixation. The single surgeon (Y.R.R.) had extensive experience with this technique, which provides consistent exposure to morbidity, surgical time, and blood loss.

After closure of the surgical wound, the surgeon (Y.R.R.) introduced a single-orifice 19G epidural catheter (Arrow, Teleflex Medical, Markham, ON, Canada) in a sterile fashion using a loss of resistance to air technique at one vertebral level above the operative site. The surgeon then threaded the catheter three centimeters into the epidural space with fluoroscopic guidance. Proper placement and catheter tip position inside the epidural space was confirmed using 3 mL of Omnipaque® (GE Healthcare, Mississauga, ON, Canada) dye visualized by real-time fluoroscopy.

The postoperative analgesic regimen was initiated shortly after patient arrival in the postanesthesia care unit (PACU) following motor function assessment by the surgeon. Patients randomized to the LA group received an initial epidural bolus of hydromorphone 0.6 mg diluted in 0.9% saline 3 mL followed by an infusion of 0.1% bupivacaine with hydromorphone 15 μg·mL−1 at 6 mL·hr−1 for 48 hr. Patients in the NS group received an initial bolus of 0.9% saline 3 mL followed by an infusion of 0.9% saline at 6 mL·hr−1. All epidural solutions were prepared in a sterile fashion by our independent research pharmacy according to patient group allocation. The solutions were delivered to the PACU in 5 mL syringes containing the initial study bolus solution and 250 mL container bags filled with the study infusion solution. To maintain blinding and allocation concealment, the epidural solutions were labelled with the patients’ initials, study number (1-39), and a line printed with “0.6 mg hydromorphone in 3 mL of saline or 3 mL saline placebo” on the syringes and “0.1% bupivacaine with 15 μg of hydromorphone per mL or saline placebo” on the container bags.

All patients received an intravenous patient-controlled analgesia (PCA) device programmed to deliver a bolus dose of 0.2-0.4 mg iv hydromorphone, lockout of five minutes, and a maximum dose of 10 mg over four hours. Additionally, all patients received oral acetaminophen 1,000 mg every six hours (650 mg for age > 70 yr) and oral gabapentin 200 mg every eight hours (100 mg for age > 70 yr) as part of a multimodal analgesic regimen.12 Nonsteroidal anti-inflammatory medications, however, were not used due to concerns of non-union and bleeding in the setting of spinal fusion.13 After initiation of the epidural infusion, inadequate pain control (verbal numeric rating scale [VNRS] ≥ 5) was addressed in a standardized fashion regardless of group allocation, i.e., the epidural infusion rate was increased by 2 mL·hr−1 and the intravenous PCA bolus dose was increased by 0.2 mg. This intervention was repeated again if deemed necessary by the attending anesthesiologist on the Acute Pain Service (APS) who was blinded to patient group allocation. Spinal sensory block levels were not routinely assessed in an effort to preserve patient and assessor blinding. The epidural catheter was removed 48 hr after initiation of the infusion.

All patients were treated on the same postoperative ward and received corresponding nursing care, mobilization protocol (starting postoperative day 1), and physiotherapy services.

The APS staff continued to visit all patients twice daily. As per our institutional practice, patients no longer requiring intravenous PCA were offered controlled-release and/or immediate-release oxycodone or hydromorphone at the discretion of the attending anesthesiologist on the APS in order to achieve a VNRS pain score of ≤ 4. In the event of inadequate pain control (VNRS pain score > 7), intravenous PCA was maintained. Upon discharge from hospital, patients received one of three oral analgesic preparations to be taken up to every four hours as needed: Tylenol #3® (acetaminophen 300 mg/codeine, 30 mg/caffeine, 15 mg per tablet), Percocet® (acetaminophen 325 mg, oxycodone HCl, 5 mg per tablet), or acetaminophen with immediate-release hydromorphone if intolerant to codeine. To facilitate data analysis, equianalgesic conversion ratios were employed according to the general monograph for opioids in the Canadian Pharmacists’ Association Compendium of Pharmaceuticals and Specialties (hydromorphone:morphine = 1:5, oxycodone:morphine = 2:3, intravenous:oral = 1:3).Footnote 1 As described previously, oral acetaminophen and gabapentin were continued until discharge from hospital.

Outcome measures and assessment

Patients were assessed at the following time points: (i) 15 min after arrival in the PACU (immediately prior to the epidural bolus); (ii) twice daily on postoperative days 1 and 2 (0800 hours and 1600 hours); and (iii) once daily (0800 hours) until discharge. The primary outcome measure was defined as cumulative opioid consumption at 48 hr. Secondary outcomes collected at each assessment included: pain at rest and with mobilization (VNRS from 0 = no pain to 10 = worst pain imaginable); heart rate (bradycardia defined as heart rate < 55 beats·min−1) at any time since the last assessment; blood pressure (hypotension defined as systolic blood pressure < 100 mmHg) at any time since the last assessment; presence of nausea and vomiting at any time since the last assessment; sedation (four-point scale: 0 = awake and alert; 1 = mild - easy to arouse; 2 = moderate - frequently drowsy and easy to arouse, 3 = severe - somnolence and difficult to arouse) at any time since the last assessment; presence of pruritus at any time since the last assessment; urinary retention (inability to void > 24 hr after epidural removal); time to assisted and unassisted ambulation; patient satisfaction with analgesic modality (Quality of Recovery Score administered on postoperative day 2 after removal of the epidural catheter);14 and time to discharge readiness (date of discharge order from patient’s chart).

All study patients completed baseline questionnaires with respect to medical history, pain, and functional disability (Oswestry Disability Index [ODI]), as well as surgical expectation as per the surgeon’s (Y.R.R) standard clinical practice.15 The surgical expectation score is not a validated measure but standard information that is routinely collected by our surgical department (Appendix). There were no changes to the study protocol or outcome assessments during the trial period.

Statistical analysis

In a retrospective review of 20 consecutive patients undergoing port access lumbar decompression and fusion with intravenous PCA at Toronto Western Hospital prior to the current study, results revealed a mean cumulative opioid consumption of 562.8 mg oral morphine equivalent up to and including postoperative day 2. Based on a priori discussions between the investigators and past experience with epidural analgesia, we aimed to detect a 50% reduction in cumulative opioid consumption at 48 hr. Therefore, assuming a standard deviation of 200.72 mg (based on our retrospective data), we estimated that a minimum of 16 patients would be required for randomization. We set type I error α = 0.05 (two-sided) and type II error β = 0.2. Due to the small sample size, there was no planned interim analysis or early stopping rule.

Computerized statistical analysis was performed using SPSS® 18 (IBM, Armonk, NY, USA). Differences between groups for continuous variables were analyzed using the Student’s two-sample t test, while categorical variables were analyzed using the Chi square or Fisher’s exact test if any expected cell size was less than five. Statistical tests were two-sided.


Thirty-nine of the 79 eligible patients approached consented to participate in the study during January 2008 to April 2011 (Fig. 1). Each patient was followed in hospital until they were discharged home. The trial was stopped once the planned number of patients were recruited. Twenty participants were randomized to the LA group and 19 were randomized to the NS group. One patient randomized to the NS group was withdrawn from the study due to illicit drug use that was not disclosed by the patient preoperatively. One patient randomized to the LA group requested discontinuation of the intervention and removal of the epidural catheter because of anxiety related to the sensation of sensory/motor blockade. One patient randomized to the LA group suffered intractable nausea; this patient was unblinded and the epidural study solution was changed to 0.2% ropivacaine without hydromorphone. One patient randomized to the NS group experienced an accidental dislodgement of the epidural catheter during transfer from the operating room table to the bed; epidural solution was not administered to this patient. These latter three patients consented to continued data collection and analysis. In total, 20 patients were analyzed in the LA group and 18 in the NS group.

Fig. 1
figure 1

CONSORT diagram detailing screened, randomized, and analyzed patients, n = number of patients

Patient characteristics were similar between groups with respect to age, sex, body mass index, ASA physical status, number of levels decompressed, preoperative baseline pain and opioid use, expectations from surgery, ODI, and duration of surgery (Table 1).

Table 1 Patient characteristics

At 48 hr, the mean (SD) cumulative opioid consumption was 249.3 (143.3) mg in the NS group and 184.7 (208.1) mg in the LA group (mean difference 64.6 mg; 95% confidence interval (CI) −54.3 to 183.5; P = 0.27) (Fig. 2). There were no differences in intravenous PCA hydromorphone consumption in isolation (Table 2). There were no differences in the self-reported Quality of Recovery Score or the number of days to assisted ambulation, unassisted ambulation, and time to discharge from hospital (Table 2). Postoperative pain scores from the PACU onwards (VNRS at rest) and postoperative day 1 onwards (VNRS at movement) are presented in Fig. 3. Epidural and opioid-related side effects were similar between groups (Table 3).

Fig. 2
figure 2

Cumulative 48 hr oral morphine consumption. Data presented as mean with bars denoting standard deviation

Table 2 Secondary outcomes
Fig. 3
figure 3

Line graph of verbal numeric rating scale (VNRS) pain scores. Panel A) VNRS at rest, Panel B) VNRS with movement. Data presented as median [interquartile range]

Table 3 Epidural/opioid side effects


This study was designed to evaluate the effects of continuous epidural analgesia with a local anesthetic and opioid solution in a homogeneous surgical population with concomitant systemic multimodal analgesia following lumbar decompression and fusion. We observed a mean reduction in average cumulative opioid consumption of approximately 64.6 mg at 48 hr; however, this estimate was imprecise (95% CI: −54.3 to 183.5), thus no definitive conclusion could be made about the efficacy of adjuvant epidural anesthesia. Most similar studies published previously did not employ opioid-sparing adjuncts.2-5,8 One study included standing order acetaminophen;9 one study included a nonsteroidal anti-inflammatory drug (NSAID) on an as needed basis,7 and one study included acetaminophen plus a NSAID on an as needed basis10 (Table 4). Inconsistent systemic analgesic protocols may partially explain why the optimal postoperative regimen following lumbar fusion and decompression has not yet been determined.

Table 4 Comparative studies assessing epidural analgesia with local anesthetic vs systemic opioids after spine surgery

The strength of the present trial stems from the homogeneity of both diagnosis and procedure with the presence of a standardized multimodal analgesic protocol. Previous studies investigating the role of epidural analgesia after spine surgery have been limited by methodological shortcomings, including heterogeneity in diagnosis and procedure,5,8,9 non-standardized analgesic co-intervention,9,10 either extremely high4 or extremely low5,8 doses of epidural local anesthetic, and significant loss to follow-up (patients removed from study due to protocol violations)4,8,9 (Table 4). Predictably, studies reporting issues with transient sensorimotor deficits were those utilizing high infusion rates.2-4 None of the previous studies reported issues with hemodynamic instability.

Several procedural factors related to the epidural may have affected our results. First, while our epidural infusion rate was thoughtfully selected in order to strike a balance between analgesia and undesired motor blockade, it is possible that this rate was too low. Indeed, previous studies determining the efficacy of epidural analgesia for open spine surgery used continuous infusion rates up to 14 mL·hr−1 along with a patient-controlled epidural bolus feature.3,4 We have had a similar experience using a catheter placed directly through the decompression space; however, in our view, such high rates may have been required to compensate for the loss of local anesthetic solution through the traditional surgical wound. Moreover, while infusion rates higher than 6 mL·hr−1 provide effective analgesia, our institutional experience is that these rates produce undesirable side effects and motor block in the setting of lumbar fusion. Consequently, we switched to a loss of resistance technique at a cephalad non-decompressed level. The use of a continuous infusion technique may not have provided as effective analgesia as a continuous infusion combined with a patient-controlled epidural analgesia (PCEA) bolus feature; however, our institutional experience with PCEA (infusion of 8 mL·hr−1 with a bolus of 2-4 mL every 30 min as required) culminated in a disproportionate number of patients with clinically significant hypotension. Finally, under the present study conditions, the use of PCEA may have led to patient confusion with two separate patient-controlled devices.

We recognize that our study design was based on assumptions that ultimately did not reflect the conditions of the trial. The retrospective data used to calculate our sample size were collected in the absence of a standard multimodal analgesia regimen. The trial was designed to detect a 50% reduction in oral morphine consumption from a historical level of approximately 560 mg; the morphine consumption in the control group in this trial was approximately 250 mg. The confidence interval for the opioid sparing effects of epidural analgesia added to multimodal analgesia was very wide, from 54.3 mg in favour of the control group to 183.5 mg in favour of epidural analgesia. Thus, we cannot rule out the possibility of a 50% reduction (~125 mg) in morphine consumption with the addition of an epidural anesthetic. Furthermore, we recognize that the decision to use the estimated opioid use and variance from our retrospective data in our sample size calculation may have contributed to inflating the imprecision in our estimate of effect.16

In summary, we observed that continuous epidural analgesia combined with systemic multimodal analgesia following one- or two-level lumbar decompression and fusion resulted in a reduction in opioid consumption of 64.8 mg oral morphine equivalent, but this was an imprecise estimate (95% CI: −54.3 to 183.5). Based on this study, the routine use of continuous epidural analgesia to reduce postoperative opioid consumption is not yet warranted. Our findings suggest that the optimal regimen for postoperative pain following one- or two-level lumbar decompression and fusion has yet to be determined.