Study Design and Ethics
The primary objective of this phase 3, multicenter, double-blind, three-arm, randomized, placebo-controlled, multiple-dose, parallel-group trial was to evaluate the safety, tolerability, and efficacy of two different doses of IV tramadol versus placebo in the management of postoperative pain over 48 h in patients undergoing a unilateral primary first metatarsal bunionectomy surgery. Secondary efficacy objectives included assessment of analgesic efficacy over 24 h, rescue medicine (ibuprofen 400 mg) use, and patient global assessment (PGA). An important aim of the study was dose finding, and thus two doses of tramadol were studied: 50 mg and 25 mg (each given with the same dosing intervals). Safety assessments included treatment emergent adverse events (TEAEs), clinical laboratory tests, vital signs, and electrocardiograms (ECGs).
The study was performed in accordance with ethical principles that have their origin in the Declaration of Helsinki and are consistent with International Conference on Harmonization (ICH)/Good Clinical Practice (GCP), applicable regulatory requirements, and the sponsor or its delegate’s policy on bioethics. Aspire Institutional Review Board (https://aspire-irb.com/, Santee, CA), the central IRB that approved the study for Avenue Therapeutics and the study sites, reviewed study documents according to current US regulations. The patient’s signed and dated informed consent form (ICF) was obtained before initiation of any study procedures. The study was registered at clinicaltrials.gov (NCT03290378).
Study Treatment and Eligibility
The study was conducted at five sites in the USA from August 30, 2017 to April 23, 2018. Eligible patients were male or female, age 18–75 years of age, undergoing unilateral first metatarsal bunionectomy surgery, and who met the American Society of Anesthesiologists (ASA) physical class 1 or 2, and were willing and able to understand the study procedures and use of pain scales. Important exclusion criterial included other concomitant opioids during the treatment within 30 days prior to surgery, other prohibited medications (analgesics, medications that lower seizure threshold), allergy or physical dependence on opioids, and not reaching a minimal postsurgical pain level of moderate to severe on a categorical pain scale and a numeric pain rating scale (NPRS) rating of at least 5 on a scale of 0–10.
Patients underwent the screening visit, preoperative assessment (within 24 h prior to surgery start time), the surgical/treatment visit (including the surgery), the primary treatment period of double-blind study medication for 48 h, and a follow-up visit (day 14). The treatment groups were:
-
IV tramadol 50 mg, given at hours 0, 2, 4, 8, 12, 16, 20, 24, 28, 32, 36, 40, and 44
-
IV tramadol 25 mg, given at hours 0, 2, 4, 8, 12, 16, 20, 24, 28, 32, 36, 40, and 44
Patients were confined at the healthcare facility during study drug administration and were discharged only if clinically stable. All patients in the study followed an anesthesia protocol and were not randomized into the trial if they deviated from the protocol as listed below.
Pre- and intraoperative anesthesia per study protocol included IV midazolam, a popliteal sciatic nerve block (0.5% ropivacaine; loading volume at the discretion of the investigator), a local Mayo block using short-acting 2% lidocaine without epinephrine, propofol induction followed by a continuous infusion for intraoperative sedation, intraoperative IV fentanyl, and no nausea prophylaxis is allowed. Postoperative anesthesia included regional anesthesia via a popliteal sciatic nerve block. After the completion of surgery, a continuous infusion of ropivacaine, mepivicaine, or bupivacaine was provided with additional boluses of local anesthetic at the discretion of the investigator. Additional pain medicine allowed were ketorolac 30 mg IV q6h until 6 h before removal of the popliteal block and 2–4 mg IV morphine sulfate every 2 h as needed until 1.5 h before removal of the popliteal block. Popliteal block was permanently discontinued between approximately 4:00 a.m. and 5:00 a.m. on the day after surgery.
Following removal of the popliteal block, patients were assessed for the dosing eligibility criteria; patients must have been awake and alert and had a pain intensity of 5 or greater on the 0 (no pain) to 10 (worst pain possible) point numerical pain rating scale (NPRS) and reported a score of moderate or severe on a four-point categorical rating scale (with categories of none, mild, moderate, or severe) just before the first dose of study drug. Patients who did not report pain at these levels within 8 h of the removal of the block were discontinued from the study (and thus were not treated with double-blind study medication). Bunionectomy, a common procedure used in clinical trials studying pain management, is a very painful orthopedic surgical model. The surgical anesthetic administration was the same for all patients and thus any effects on initial NPRS scores would be the same across all treatment groups. Further, postsurgical dosing criteria were included in the protocol to ensure that patients had sufficient pain levels when entering the study and therefore sufficient time for the ropivacaine or bupivacaine to wear off. The 8-h window for eligibility after removal of the block was sufficient as most patients met the criteria in that timeframe. Blinded study drug infusions were given over approximately 15 ± 2 min via an infusion pump.
We note that this study was designed using one of several well-established acute pain surgical models [12, 13]. The bunionectomy surgical pain model was not intended to “mimic” real-life treatment but rather to eliminate as many potential biases and confounding factors (e.g., from intrinsic and extrinsic factors) and to allow for the same procedures to be performed in a blinded setting, thus ensuring rigorous and reliable conclusions regarding the treatment benefit of each dose level of IV tramadol.
Endpoints
Pain intensity assessments were recorded immediately prior to the first dose (baseline, T0) and at 0.5, 1, 2, 3, 4, 5, 6, 8, 10, 12, 14, 16, 18, 20, 22, 24, 26, 28, 30, 32, 34, 36, 38, 40, 42, 44, 46, and 48 h after first treatment (i.e., post T0). The sum of pain intensity differences (SPID) through 48 h post first dose (SPID48) measured at rest was used as the primary measure of efficacy.
Secondary efficacy endpoints include SPID24, total consumption (mg) of rescue (supplemental) analgesia (this is the total amount of rescue analgesia given to the patient after first dose of study medication through 48 h post first dose), and PGA of efficacy at 24 and 48 h post first dose using a five-point scale [the question posed was “How would you rate the study medication in terms of its effectiveness in controlling your pain?” (0 = poor; 1 = fair; 2 = good; 3 = very good; 4 = excellent)]. Additional analgesia endpoints included the time-specific pain intensity profile over time, time (in minutes) to first rescue analgesia from the time of first dose of study medication, and number (percentage) of patients who required no rescue analgesia from T0 to T48.
TEAEs, classified by the Medical Dictionary for Regulatory Activities (MedDRA) system organ class (SOC) and preferred term, including assessment of AEs related to substance abuse (based on the FDA’s guidance “Assessment of Abuse Potential of Drugs, Guidance for Industry, January 2017”). Clinical laboratory (including full chemistry, hematology, and urinalysis panels), vital sign, physical examination, and ECG changes were assessed. Clinically meaningful changes in these safety outcomes were reported as TEAEs.
Statistical Methods
The sample size was based on an assumed effect size between tramadol and placebo of at least 40%. A sample size of 135 patients per arm (405 patients in total) provided over 90% power to detect a SPID48 effect size of 40% between each pairwise tramadol and placebo comparison, assuming an alpha of 0.05 and using a two-sided test of mean differences.
The full analysis set (FAS) population was defined as all randomized patients who received at least one dose of study medication. Patients were analyzed according to the treatment group they were randomized to. The safety population was defined as all patients who received at least one dose of study medication. Patients were analyzed according to the actual treatment they received. There was one patient randomized to placebo who received a single dose of IV tramadol 50 mg, and thus was included in the tramadol 50 mg arm for purposes of safety assessment.
All SPID calculations were performed using the standard trapezoidal rule:
$$SPID=\sum_{i=0}^{x}\left(\frac{{PID}_{i}+{PID}_{i+1}}{2}\right)* \left({T}_{i+1}- {T}_{i}\right),$$
where PIDi is the PID at time i, and (Ti + 1 – Ti) is the time difference in hours between time i and time i + 1.
Multiple imputation provides a useful strategy for analyzing data sets with missing values. Instead of filling in a single value for each missing value, Rubin’s [14] multiple imputation strategy replaces each missing value with a set of plausible values that represent the uncertainty about the correct value to impute. This methodology was used to address missing data. Specifically, for the primary endpoint (SPID48), 100 imputed data sets were created, with data imputation for missing values due to missingness at random as well as due to discontinuation due to AE (missing, not at random) and to account for use of rescue medication (the last NPRS prior to the use of any rescue medication was used to impute subsequent NPRS scores for the subsequent protocol-specified time points for measurement of pain intensity through 4 h after the time of the dosing of the rescue medication).
The 100 imputed data sets were analyzed using an analysis of covariance (ANCOVA) model with contrasts to test the primary efficacy endpoint. The model used treatment as the main effect and investigational center and baseline pain intensity (NPRS scores of 0–10) as covariates. Data from all three treatment groups were included in the same ANCOVA model for purposes of the testing procedures. The 100 resulting treatment effect parameters and standard errors (SEs) from these ANCOVA were presented individually and then combined to provide a distribution of parameters (and SEs) upon which the primary hypothesis of treatment effect was determined.
The results of the analyses were presented in summary tables with standard summary statistics as well as least squares (LS) means, active versus placebo LS mean differences (treated group – placebo), SEs, confidence intervals, and paired-comparison p values. The mean SPID values and standard errors of the mean (SEM) for each treatment group were also presented graphically. This analysis was referred to as the multiple imputation analysis.
Average patient global assessment at scheduled time points was assessed for treatment group differences using an ANCOVA with pooled study center and the baseline pain score as covariates. Total consumption of rescue medication was calculated as the total amount of rescue analgesia (mg) captured in the rescue medications electronic case report form and recorded as given to the patient between the first dose of study medication through 48 h post first dose (4 h after the start of the last dose of study medication). The total consumption of rescue analgesia was analyzed using the nonparametric Wilcoxon rank sum test to test each active versus placebo comparison separately. The incidence of TEAEs was summarized for each treatment group by SOC and preferred term sorted in descending frequency by SOC, and then by preferred term within SOC.
There were two pairwise comparisons of interest:
All inferential assessments were two-sided tests performed at the 0.05 alpha level. A hierarchical alpha testing strategy was utilized to control for the overall experiment-wise alpha. As there were multiple tests being performed (the single primary efficacy variable pairwise test and the key secondary efficacy tests), the following strategy was applied.
The high-dose tramadol arm versus placebo comparison was assessed for the primary endpoint, summary of pain intensity differences through 48 h (SPID48) at the 0.05 alpha level. If and only if the p value is 0.05 or less for this pairwise comparison was the lower dose tramadol arm versus placebo comparison to be assessed. If the p value for the high-dose arm was not significant, all testing was to cease, and it was to be concluded that neither tramadol treatment group provide better pain relief than placebo.
If the primary endpoint was significant for the tramadol versus placebo comparison (in favor of the tramadol arm) for the high dose comparison, then statistical testing was to proceed to the high-dose tramadol key secondary endpoints, to be tested in the following order:
If a statistical test within each pairwise comparison was significant for the high dose comparison at the nominal 0.05 level, two-sided (in favor of the tramadol arm) for the first endpoint, then testing was to proceed to the next endpoint in the list, and so on. Once a non-significant test occurred, endpoints lower in the list were to be considered not statistically significant. Testing for the low-dose tramadol vs placebo comparison proceeded in a similar fashion, but only if the original high-dose tramadol vs placebo SPID48 primary comparison was statistically significant and in favor of tramadol. Dose–response was assessed via testing for a linear trend across the doses (where placebo was set to zero, and the tramadol 25 mg and 50 mg groups were analyzed accordingly).
The time-specific pain intensity profile over time was assessed using a mixed model repeated measures (MMRM) with pain intensity difference as the dependent variable and fixed effects for treatment, time, the treatment by time interaction, and random patient effect, and pooled study center and baseline pain score as covariates. Pairwise comparisons at each time point were determined from the MMRM, and model-based least square means (LS means) and standard errors (SEs) were plotted for visual comparison of the treatment group outcomes. Only scheduled visits were included in the MMRM after appropriate imputation for rescue. All analyses were performed using the SAS System® version 9.4.
A computer-generated randomization scheme was prepared prior to the study start (Axiom Real Time Metrics, Toronto, Ontario). Patients who met the enrollment criteria were randomly allocated to one of three treatment arms (tramadol 50 mg, tramadol 25 mg, or placebo) in a 1:1:1 ratio. Upon randomization, a unique four-digit study-specific randomization number was assigned. Patients, study staff, the sponsor, and contract research organizations involved with the study were blinded to treatment assignments, and the study remained blinded until all patients completed all study procedures and the clinical database had been locked.