Minimal important difference in opioid consumption based on adverse event reduction—A study protocol

Abstract

Background

The patient‐relevant minimal important difference for opioid consumption remains undetermined, despite its frequent use as primary outcome in trials on postoperative pain management. A minimal important difference is necessary to evaluate whether significant trial results are clinically relevant. Further, it can be used as effect size to ensure that trials are powered to find clinically relevant effects. By exploring the dose–response relationship between postoperative opioid consumption and opioid‐related adverse effects, we aim to approximate the minimal important difference in opioid consumption anchored to opioid‐related adverse effects.

Methods

This is a post‐hoc analysis of aggregated data from two clinical trials (PANSAID NCT02571361 and DEX2TKA NCT03506789) and one observational cohort study (Pain Map NCT02340052) on pain management after total hip and knee arthroplasty. The primary outcome is the Hodges–Lehmann median difference in opioid consumption between patients with no opioid‐related adverse effects and patients experiencing the mildest degree of one or more opioid‐related adverse effects (i.e., mild nausea, sedation and/or dizziness or vomiting). Secondary outcomes include the Hodges–Lehmann median difference in opioid consumption that corresponds to one point on a cumulated opioid‐related adverse event 0–10 scale. Further, we will explore the proportion of patients that experience opioid‐related adverse effects for consecutive opioid dose intervals of 2 mg iv morphine equivalents. Quantile regression will be used to assess any significant interactions with patient baseline characteristics.

Conclusions

This study will hopefully bring us one step closer to determining relevant opioid reductions and thereby improve our understanding of intervention effects and planning of future trials.

1. INTRODUCTION

Opioids still hold a major role in treating acute pain, ¹ , ² despite risks of opioid‐related adverse effects. ³ , ⁴ , ⁵ Thus, clinicians have to balance their use to provide sufficient pain relief while avoiding opioid‐related adverse effects. ⁶ , ⁷ A variety of non‐opioid analgesics have demonstrated statistically significant opioid‐reducing effects in clinical trials. ⁸ , ⁹ , ¹⁰ However, to ensure clinical significance, researchers should designate a relevant minimal important difference. ¹¹ Opioid consumption has been used as the primary outcome in a major part of pain trials. ¹² As effect size or minimal important difference, trialists often use an absolute reduction of around 10 mg iv morphine equivalents the first 24 hours after surgery. ¹³ This value has been chosen arbitrarily and may not reflect what patients find clinically relevant. The problem with determining a patient‐relevant minimal important difference may be that opioid reduction is not directly relevant to patients but functions as a surrogate marker for other outcomes such as pain relief and avoidance of opioid‐related adverse effects.

We believe that a patient‐rated opioid minimal important difference should optimally be determined with anchor‐based methods to these outcomes. Further, an anchor‐based method could also help illuminate a potential acceptable opioid cut‐off dose based on ‘no opioid‐related adverse effects’ or ‘no worse than mild opioid‐related adverse effects’, corresponding to ‘no worse than mild pain’ for pain relief. ¹⁴ , ¹⁵ , ¹⁶ This has not been explored in the current literature and generally, the dose–response relationship between opioid consumption and both opioid‐related adverse effects and pain is poorly described. ¹⁷ , ¹⁸

In this post hoc analysis of aggregated data from two clinical trials and one observational cohort study on pain management after total hip and knee arthroplasty, we aim to explore the dose–response relationship between postoperative opioid consumption and opioid‐related adverse effects. We hope to approximate an ‘opioid‐related adverse effects’‐anchored minimal important difference for 0–24 h postoperative opioid consumption. Secondly, we will explore the relationship between opioid doses and the proportion of patients that experience opioid‐related adverse effects.

2. METHODS

2.1. Study type

Post hoc analysis of the results of three studies on postoperative pain management after total hip or knee arthroplasty conducted between 2014 and 2021.

2.2. Study aim

Our aim is to explore the dose–response relationship between opioid consumption and opioid‐related adverse effects and to provide estimates of a minimal important difference in opioid consumption anchored to opioid‐related adverse effects.

2.3. Eligibility criteria and data extraction

Two randomised trials, PANSAID ¹⁹ and DEX2TKA, ²⁰ and one observational cohort study, Pain Map, ²¹ constitute our study population (Table 1). Two trials investigated postoperative pain management after total hip arthroplasty ¹⁹ , ²¹ and one after total knee arthroplasty. ²⁰ All were multicentre and pre‐registered at clinicalTrials.gov (NCT02571361, NCT03506789 and NCT02340052) and have been published in peer reviewed journals. Data will be sufficiently anonymized with the extraction of only age, sex and perioperative data on medication use and side effects. We will exclude patients with incomplete data on 0–24 h opioid consumption; severity of nausea, dizziness and sedation at 6 and 24 h postoperatively; and incidence of vomiting.

TABLE 1.

All three trials registered nausea, dizziness and sedation as none, mild, moderate or severe graded by patients at 6 and 24 h postoperatively and vomiting as yes or no

Trial/ID	Study type/participants	Basic analgesic regimen	Antiemetic treatment
PANSAID NCT02571361	RCT, parallel 4‐group in 556 THA patients (523 eligible)	Group A: 1 g PCM + 400 mg Ibuprofen q6h Group B: 1 g PCM + placebo q6h Group C: 400 mg ibuprofen + placebo q6h Group D: 0.5 g PCM + 200 mg ibuprofen q6	Rescue: Ondansetron up to 8 mg, DHB 0.625–1.25 mg
DEX2TKA NCT03506789	RCT, parallel 3‐group in 485 TKA patients (444 eligible)	All: 1 g PCM + 400 mg ibuprofen q6h Group DX1: 24 mg iv dexamethasone x 2 (perioperative and 24 h postoperatively) Group DX2: 24 mg iv dexamethasone perioperatively and placebo at 24 h Group Placebo: Placebo x 2	Group DX1/DX3: Dexamethasone 24/48 mg All: Prophylactic: ondansetron 4 mg Rescue: 2. dose ondansetron or DHB 0.625–1.25 mg
Pain Map DK NCT02340052	Prospective cohort of 501 THA patients (449 eligible)	All: 1 g PCM 217 patients: 400 mg ibuprofen q6h 93 patients: gabapentin 29 patients: LFCN	Rescue: dexamethasone 4 mg, ondansetron 4 mg and/or DHB 0.625–1.25 mg

Abbreviations: DHB: dehydrobenzperidol; LCFN, lateral femoral cutaneous nerve block; PCM, paracetamol; q6h, every 6 h; THA: total hip arthroplasty; TKA, total knee arthroplasty.

Vomiting was registered dichotomously. Nausea, dizziness and sedation were graded by patients on a 4‐point Likert scale as none, mild, moderate or severe. All opioid‐related adverse effects were assessed twice (6 and 24 h postoperatively), and we will use only the worst experienced severity degree of the two time points.

2.4. Outcome measures

2.4.1. Primary outcome

The Hodges–Lehmann median difference in opioid consumption between patients with no opioid‐related adverse effects and patients experiencing the mildest degree of one or more opioid‐related adverse effects (i.e., mild nausea, sedation and/or dizziness or vomiting). Data will be presented with Hodges–Lehmann confidence intervals and box plots.

2.4.2. Secondary outcome

1. The Hodges–Lehmann median difference in opioid consumption between patients with mild versus moderate and moderate versus severe opioid‐related adverse effects (nausea, dizziness and/or sedation). Vomiting was registered dichotomously and will therefore not be included in this analysis.

2. The Hodges–Lehmann median difference in opioid consumption that corresponds to one point on a cumulated opioid‐related adverse event 0–10 scale (0–3 for nausea, dizziness and sedation and 0–1 for vomiting). Ordinal none‐mild–moderate–severe gradings will be converted to 0–3 numeric scales for the purpose of data handling and statistical analyses. To avoid misuse of ordinal values in statistical analyses, we will not perform a regression analysis. Instead, the Hodges–Lehmann median difference in opioid consumption for each step on the 0–10 point scale will be reported separately. We will primarily focus on the median value of these 10 Hodges–Lehmann median differences (Table 2). ²²

TABLE 2.

Each patient are given 0–3 points (none‐mild–moderate–severe) for nausea, dizziness and sedation and 0–1 point for vomiting, cumulated to a 0–10 scale

Note: The overall median difference is the median value of the ten Hodges–Lehmann median differences corresponding to each step on the scale.

With an emphasis on the primary outcome, the results of the three outcomes will be discussed and aggregated into one suggested minimal important difference for 0–24 h opioid consumption.

2.4.3. Subgroup analyses

Quantile regression will be used to assess any significant interactions between adverse events (none/mind) and the following baseline variables: procedure, age, ASA‐score, sex, perioperative use of systemic glucocorticoids and perioperative use of ondansetron.

2.4.4. Exploratory outcomes

We will assess the proportion of patients that experience opioid‐related adverse effects for consecutive opioid dose intervals to explore and better understand this relationship. We know from the dataset that around 90% of participants received 0–60 mg morphine equivalents and the remaining 10% received 61–195 mg. We will only assess the population receiving 0–60 mg and divide these patients into opioid consumption intervals of 2 mg morphine to ensure interval blocks with enough patients to minimise inaccuracy. Data will be reported in a histogram with opioid consumption on the X‐axis and proportion of patients experiencing any opioid‐related adverse effect on the Y‐axis (Figure 1).

FIGURE 1.

Visualisation of the proportion of patients experiencing opioid‐related adverse events in relation to cumulated opioid consumption 0–24 h

2.4.5. Statistical considerations and data handling

Hodges–Lehmann median differences with 95% confidence intervals will be calculated in R and quantile regression will be performed in STATA. ²³ , ²⁴

We will remove extreme outliers defined as data points that are more extreme than (first quartile − 3 × interquartile range) or (third quartile +3 × interquartile range). ²⁵

3. DISCUSSION

We will use three approaches to estimate the minimal important difference in opioid consumption based on (1) the difference between participants with overall none versus mild opioid‐related adverse effects, (2) differences between overall mild versus moderate and moderate versus severe opioid‐related adverse effects and (3) differences corresponding to one severity degree on a cumulated opioid‐related adverse effect 0–10 scale. We believe this study could be important for the planning of future trials in postoperative pain management, despite its post hoc design and the inclusion of only four adverse effects. Based on our findings, we will plan a subsequent prospective study to test and elaborate on the results.

To the best of our knowledge, only a single study has investigated the relationship between opioid dose and opioid‐related adverse effects using a similar approach. ¹⁷ In 2004, Zhao et al. published a post hoc analysis on 193 patients undergoing laparoscopic cholecystectomy and found that every 3–4 mg iv morphine equivalent resulted in a ‘clinically meaningful event’ in one of 12 different opioid‐related effects. ¹⁷ The two most frequent events were ‘Feeling of general fatigue or weakness’ and ‘Dry mouth’. We will assess the four opioid‐related adverse effects we consider most important to patient satisfaction and the postoperative courses: nausea, vomiting, dizziness and sedation. ²⁶ Other opioid‐related adverse effects were not systematically assessed in the included studies and will therefore not be included in this analysis. Respiratory depression is rarely seen with normal postoperative intravenous morphine doses. ²⁶ Zhao et al. administered hydrocodone and fentanyl and converted them to morphine equivalents, which could skew opioid‐related adverse effect profiles. ²⁶ , ²⁷ Of the approximately 1500 patients we expect to include in our analysis, the majority received patient‐controlled intravenous morphine alone. Further, to increase the clinical relevance of our analysis, we will provide three different estimates for minimal important differences in opioid consumption. It is obvious that the duration of opioid‐related adverse effects is important as well. This aspect will not be optimally assessed in the current analysis but will be incorporated in the following prospective study. In anchor‐based methods, the anchor should optimally be a validated scale or known patient‐relevant value. ²⁸ As there are no current reports on patient‐relevant opioid‐related adverse effects severity, we chose a conservative approach with the lowest detectable change in the overall opioid‐related adverse effect burden as the primary outcome.

Overall, this study will provide clinical data on the relationship between opioid dose and opioid‐related adverse effects. We expect this to be useful for defining a minimal important difference in opioid consumption, which is necessary in order to power trials to detect clinical important differences and to determine whether significant differences are actually clinically important.

AUTHOR CONTRIBUTIONS

Anders Karlsen, Casper Pedersen, Jens Laigaard and Ole Mathiesen conceived the idea for this protocol. Anders Karlsen, Janus Christian Jakobsen, Casper Pedersen and Ole Mathiesen drafted the first version of this protocol. Anders Karlsen and Janus Christian Jakobsen planned the statistical analyses. As primary investigators of the three trials included in this post hoc analysis, Anja Geisler, Kasper Thybo and Kasper Gasbjerg facilitated a better understanding of data. All authors critically revised the manuscript and take responsibility for the content.

FUNDING INFORMATION

There is no funding.

CONFLICT OF INTEREST

The authors declare no conflicts of interest.

Karlsen APH, Pedersen C, Laigaard J, et al. Minimal important difference in opioid consumption based on adverse event reduction—A study protocol. Acta Anaesthesiol Scand. 2023;67(2):248‐253. doi: 10.1111/aas.14175