This prospective, randomized, double-blind, controlled trial was conducted at the Department of Orthopaedics, Mulago National Referral Hospital. The Institutional Review Board of Makerere University Medical School and Mulago National Referral Hospital approved it as part of completion of Master of Medicine in Anaesthesia and Critical Care. The Makerere University School of Medicine Research and Ethics Committee (SOMREC) provided ethical approval for this trial. This study was conducted for 8months between June 2016 and February 2017.
The trial was conducted in accordance with the Declaration of Helsinki and Good Clinical Practice and registered at ClinicalTrials.gov. (NCT03056521). A written informed consent was obtained from all patients. The authors have followed the applicable CONSORT guidelines.
For this single centre, we consecutively recruited all patients during the study period who were 18 years and above, scheduled for elective orthopaedic surgery, that consented to participate in the study. We excluded those with surgically unrelated chronic pain, bone cancer and those that were previously recruited but came back for re-surgery for the same/different condition.
Randomization and concealment: Block randomization was used for this study with block size of 5 to ensure equal distribution of patients in each arm. A computer program was used to generate the randomization sequence of patient allocation to either the intervention or control arms by an independent statistician. Arm1 received specific information about pain preoperatively (intervention) while Arm 2 did not receive any information (control). The study numbers were allocated sequentially following the aforementioned codes. The randomization code was sent to the principal investigator in an opaque carrier envelope by an independent statistician who was not part of the study. They were similar to computer generated number sequence, becoming the patients study number.
Blinding: The study was double-blind with the research assistants, anaesthesia providers and the statistician blinded to the assignment and the arms to which the patients belonged.
All patients admitted for elective orthopaedic surgery, were expected to have a pre-operative visit by the anaesthetic provider on the day before surgery. The principal investigator recruited all the patients to ensure consistency. Randomization was based on computer-generated codes.
To conceal allocation, sealed opaque envelopes were opened only shortly before administering the intervention. After consent bedside, they were randomized. Demonstration of the Verbal Numerical Rating Scale (VNRS) and how to indicate the level of pain they experienced was done. The intervention arm privately received specific preoperative information about pain from the principal investigator in addition to the preoperative assessment. A trained assistant assessed the pain scores postoperatively at 0, 12, 24, 48hours or till discharge if less than 48hours.
We used a pretested interviewer administered questionnaire for both the intervention and the control arm. The questionnaire included the patient demographics, vital baseline clinical characteristics, intra operative and postoperative parameters at 0, 12, 24 and 48hours. The patient satisfaction with pain management were also recorded. This was adapted from an unpublished study by Kimenye et al in Mulago National Referral Hospital and adjusted to suit our study .
When a patient was found with postoperative pain, the doctor/nurse on duty was informed so as to manage it according to their discretion. All filled data collection tools were checked for completion on a daily basis. Data was entered into EPIDATA-Entry software.
The intervention: Specific preoperative information about pain was given verbally following an order on the leaflet. This information involved the following: emphasis on patient’s own role in pain management, benefits of well-treated postoperative pain, role of physiotherapy, use of basic pain medication, disadvantages of poorly controlled pain, practical physical methods of pain management and other facts about postoperative pain. These are based on studies by Gammon et al and Louw et al. methods [22, 23].
Socio-demographic data was recorded as well as type of surgery, duration of surgery, type of anaesthesia, pain scores, pulse rate, analgesics given, patient satisfaction and anaesthesia provider.
Primary outcomes were postoperative pain experience and patient satisfaction. Postoperative pain experience was measured as pain score using the VNRS. The assessed pain using the VNRS was classified as no pain (1-3), pain (4-10) but also as no pain (0), mild (1-3), moderate (4-6) and severe pain (7-10).
We also assessed patient satisfaction with pain management which was a yes or no answer. Those that were not satisfied with the management were requested to give reasons.
Sample size calculation and statistical analysis
Sample size calculations for our trial were based on the study of Sjöling et al . Power analysis determined that a sample size of at least 165 subjects per arm would achieve 80% power to detect a difference in VNRS scores between the two arms, with a significance level α of 0.05.
Arms were primarily compared for balance in patients’ demographic data, intra-operative characteristics and postoperative variables.
Data was analyzed using STATA version 12.0. Continuous variables were summarized using means, standard deviations, medians, and ranges. Categorical variables were summarized using frequencies, proportions, and percentages. Box and Whisker plots were drawn to also present the continuous variables. The differences between arms were regarded as being statistically significant if the P-value was less than 0.05.
To assess the effect of preoperative information about pain, the outcome was dichotomized as yes (VNRS 4-10) and no (VNRS0-3). Bivariate analysis was performed for each of the independent variables to determine association with outcome. The association was assessed using logistic regression and its strength was summarized using odds ratios and 95% confidence intervals. Variables with a P-value of ≤ 0.2 at bivariate analysis were considered for multivariate model. Multivariate logistic regression was performed to determine how preoperative pain information as a main predictor is associated with the outcome. Interactions between the variables which remain in the model were assessed using the Chunk test followed by assessing for confounding using a difference of ≥ 10% between the crude and adjusted measure of effect (OR) for the variables that would have gone out at each step.