Weeding Out the Problem: The Impact of Preoperative Cannabinoid Use on Pain in the Perioperative Period

Christopher W. Liu, BSc, MBBS, MMED; Anuj Bhatia, MBBS, MD, FRCA, FRCPC, FIPP, FFPMRCA, EDRA; Arlene Buzon-Tan, RN, MN, NP; Susan Walker, RN, MN, NP; Dharini Ilangomaran, BSc; Jamal Kara, BSc, PG, Diploma; Lakshmikumar Venkatraghavan, MBBS, MD, FRCA, FRCPC; Atul J. Prabhu, MBBS, FRCA, FRCPC


Anesth Analg. 2019;129(3):874-881. 

In This Article


Study Design

This was a retrospective propensity score-matched cohort study conducted on data collected between April 1, 2015 and June 30, 2017 at Toronto Western Hospital, Toronto, Ontario, Canada. The study was approved by the University Health Network Institutional Research Ethics Board, and waiver for consent was approved in view of the anonymized nature of the data. Strengthening of Reporting of Observational Studies in Epidemiology guidelines were followed during conduct and reporting of this retrospective study.[19]


We identified all adult (18 years and older) patients who underwent hip or knee arthroplasty or major spine surgery (defined as decompression and fusion at 2 or more levels) at our hospital between April 1, 2015 and June 30, 2017 and required Acute Pain Service follow-up with data entry into our proprietary Networked Online Processing of Acute Pain Information database.

Primary and Secondary Outcomes

The primary outcome of this study was numerical rating scores for pain with movement in the early postoperative period (up to 36 hours after surgery). The secondary outcomes included pain at rest, opioid consumption, incidence of pruritus, nausea and vomiting, sedation, delirium, constipation, impairment of sleep and physical activity, patient satisfaction with analgesia in the early postoperative period, and the length of Acute Pain Service follow-up.

Intraoperative and Postoperative Pain Management

Majority of patients (approximately 90%) undergoing hip and knee arthroplasties at our hospital receive a spinal anesthetic (2–3 mL of 0.5% isobaric bupivacaine) with 100 mg of intrathecal morphine (unless contraindicated) with the remainder receiving a general anesthetic. Patients undergoing total knee arthroplasty also receive an adductor canal block with 15 mL of 0.2% ropivacaine for intra- and postoperative analgesia. Patients undergoing lumbar spine fusion surgery (involving more than 2 levels) receive general anesthesia. All patients undergoing hip and knee arthroplasties and spine fusion surgery are followed up by the Acute Pain Service team at our hospital because they receive patient-controlled oral analgesia or IV patient-controlled analgesia.

Postoperatively, all patients are given regular doses of acetaminophen (650–1000 mg every 6 hours for 5 days) and nonsteroidal anti-inflammatory drugs (celecoxib 100–200 mg twice daily or meloxicam 7.5–15 mg daily for 5 days) unless these are contraindicated. Breakthrough pain is managed with patient-controlled oral analgesia or IV patient-controlled analgesia. For patient-controlled oral analgesia, a set of short-acting oral opioid (typically a single dose of oxycodone or morphine 5 or 10 mg or hydromorphone 1 or 2 mg) is provided at the patient's bedside, and the patient is allowed to have a maximum of 1 dose every 2 hours. The use of patient-controlled oral analgesia is monitored by the Acute Pain Service and ward nurses, and the medications are replenished by the nurses on consumption. As part of the protocol, patient-controlled oral analgesia is replaced by IV patient-controlled analgesia if the patient's pain is not controlled by patient-controlled oral analgesia (numerical rating score >3/10) or if the patient is unable to tolerate oral intake. IV hydromorphone is commonly used for IV patient-controlled analgesia at our hospital with a bolus dose of 0.2–0.4 mg allowed every 5 minutes and a dose limit of 40 mg in 4 hours. Patients are assessed daily by the Acute Pain Service until satisfactory pain control (mild pain or return to preoperative baseline numerical rating score) is achieved with oral analgesics.

Data Source and Extraction

The study was performed using data obtained from the proprietary Networked Online Processing of Acute Pain Information database of our department that was developed and is maintained in-house. The Networked Online Processing of Acute Pain Information database is an online secure data platform that was created using Filemaker Pro v13 software (Filemaker Inc, Santa Clara, CA). On enrollment into Acute Pain Service, details about the baseline demographics (age, gender), medical characteristics (comorbidities, medication use), and intraoperative anesthetic techniques (general anesthesia, type of regional anesthesia) are entered into the Networked Online Processing of Acute Pain Information database by the anesthesiologist enrolling the patient into the Acute Pain Service. All patients who undergo elective surgery are seen by a pharmacist preoperatively for medication reconciliation. During that visit, the patients are routinely asked about their current use of cannabinoids. This information is also entered into the preoperative case record as either their medical history (if cannabis or cannabis oil was used for recreational or medical purposes) or a medication that the patient is taking (if a synthetic cannabinoid was used). In the postoperative period, each patient is typically assessed once or twice a day by the Acute Pain Service team. At each follow-up visit, data on pain-related variables are entered into the database. Intensity of pain and analgesic requirements are entered as continuous variables, while physical activity and quality of sleep are assessed using ordinal categories (no disruption, some disruption, severe disruption). Opioid consumption is converted into oral morphine equivalents per day using a standardized conversion protocol available online (http://nationalpaincentre.mcmaster.ca/opioid/cgop_b_app_b08.html). Adverse effects (nausea and vomiting, sedation, pruritus, urinary retention, and delirium) are also assessed using ordinal scales (none, mild, moderate, severe).

Data on the variables of interest abstracted for this study included age, sex, coexisting morbidities, use of regional analgesia techniques for hip and knee arthroplasties, preoperative medications, and pain scores (numerical rating score) at rest and with movement in the postoperative period. Pain- and analgesia-related adverse effects including pruritus, nausea and vomiting, sedation, constipation, delirium, impairment of physical activity, and/or sleep due to pain were also abstracted as was duration of Acute Pain Service care in days. Pain scores, medication use, and adverse effects of medications were evaluated for the early postoperative period (up to 36 hours after surgery) because the median duration of Acute Pain Service care at our hospital is 1–2 days following major orthopedic surgery. Regional analgesia was defined as patients who received "single shot" nerve blocks including subarachnoid blocks or perineural catheters or intrathecal morphine or any combination of these interventions. Data were extracted into a Microsoft Excel (Microsoft Inc, Redmond, WA) spreadsheet. Patients with missing data for the primary outcome of interest (pain scores with movement) were excluded from the analysis. No imputation techniques were used for missing data for secondary outcomes. Categories for impairment of physical activity, quality of sleep, and adverse effects were collapsed into binary outcomes—"none" or "present" (mild, moderate, and severe were considered as "present"). To identify cannabinoid users, we searched for patients who reported cannabis use as part of their medical history as well as patients who reported use of any synthetic cannabinoids (such as Nabilone [Apotex Inc, North York, Ontario, Canada] and Sativex [Bayer Inc, Mississauga, Ontario, Canada]). In the Networked Online Processing of Acute Pain Information database, the pain scores were recorded using the 11-point (0–10) numerical rating scale. For the purpose of analysis for factors impacting severity of pain, we defined mild pain as pain scores between 1 and 3, moderate pain as 4–6, and severe pain as 7 or greater. Some patients had more than 1 Acute Pain Service follow-up each day. In such instances, the mean pain score on that day was used for analysis.

Statistical Analyses

Propensity Score Matching. The propensity score for a patient having pain after major orthopedic surgery was defined as the probability of being on cannabinoids (or not) in the preoperative period conditional on pretreatment factors (ie, potential confounders). These pretreatment factors were identified based on clinical experience and from the existing literature on pain following major orthopedic surgery. We then selected covariates from our database that approximated these factors. The variables used for calculation of the propensity score included elderly age (age 65 years or greater), type of surgery (hip arthroplasty, knee arthroplasty, or spine fusion), sex, history of preoperative opioid use, history of depression or anxiety, and administration of a regional analgesia technique in the perioperative period. These variables were included in a multiple logistic regression model with preoperative cannabinoid use or not as the dependent variable. We matched patients who were on cannabinoids to those who were not in a 1:1 ratio. A standardized mean difference of less than 0.1 and nearest neighbor matching without replacement were used for propensity score matching. Balance between the cannabinoid and noncannabinoid cohorts was assessed by standardized mean difference to ensure that the factors likely to impact pain and its treatment had a similar distribution in both study cohorts.[20]

The data collected from the study contained continuous and categorical variables. Continuous data were examined for normality of distribution using the Kolmogorov–Smirnov test. Continuous data with normal distributions were summarized as means and SDs, and data with non-normal distributions were summarized as medians and interquartile ranges. Categorical data were summarized as numbers and percentages. The 2 cohorts generated by propensity score matching were considered to be paired. The Wilcoxon signed-rank test was used for comparing variables that were nonnormally distributed, and the paired t test was used for normally distributed variables. Categorical variables were compared using the McNemar test. We planned 2 sensitivity analysis of outcomes. The first analysis was for patients in major joint (hip and knee) surgery and spine surgery cohorts because patients undergoing spine surgery tend to have higher levels of pain. We also planned to analyze data for patients with moderate-to-severe levels of pain (numerical rating score pain 4–10) in the early postoperative period. All analyses were conducted using IBM SPSS Statistics for Macintosh (Version 24.0; IBM Corp, Armonk, NY).