Methods
Population, intervention and comparator (PIC) questions were generated a priori to the systematic literature review. Pertinent outcomes (O) were identified by the working group and then independently each member voted on outcomes using a scale of 1 to 9. Outcomes that received a rounded average score of 7 to 9 were deemed critical outcomes, those receiving a score of 4 to 6 were considered important but not critical and those receiving a score of 1 to 3 were considered of limited importance. Only critically important outcomes (7 to 9 rating) were considered in decision making for generating the final recommendations.
PICO 1: should NSAIDs be used in analgesic regimens for adult patients (≥18 years old) with traumatic fracture versus routine analgesic regimens that do not include NSAIDs to improve analgesia and reduce oral morphine equivalents (OMEs), without increases in non-union and acute kidney injury rates?
PICO 2: should ketorolac be used in analgesic regimens for adult patients (≥18 years old) with traumatic fracture versus routine analgesic regimens that do not include ketorolac to improve analgesia and reduce OMEs, without increasing non-union rates?
PICO 3: should selective NSAIDs (COX-2 inhibitors) be used in analgesic regimens for adult patients (≥18 years old) with traumatic fracture versus routine analgesic regimens that include non-selective NSAIDs to improve analgesia and reduce OMEs, without increasing non-union rates?
Acute kidney injury was deemed critical as an outcome for PICO 1 but not for PICO 2 and PICO 3 (score of 6).
COX-2 inhibitors included the coxib medications and meloxicam.
Identification of references
Our project was registered with the PROSPERO registry of systematic reviews and meta-analyses (CRD42020167575). Published literature was searched through MEDLINE (via Ovid), Embase (via Elsevier), Cochrane Central Register of Controlled Trials (via Wiley) and Web of Science (via Clarivate) databases by a professional librarian (SC) on March 18, 2020 and updated on February 25, 2021. The search used a combination of database-specific subject headings and keywords for the following concepts: NSAIDs, Opioids, Orthopedic Procedures, Fracture in various iterations and combinations. Results were limited to the English language. The full search strategy is available in online supplemental file 1.
Studies that included adult (≥18 years old) trauma patients with any fracture were eligible for inclusion. Case reports, case series, commentaries, reviews, editorials and animal studies were excluded. For a study to be included in our final analysis, a clear comparison between patients receiving NSAIDs and control patients had to be present, as well as at least one of the critical outcomes reported. Studies in which NSAIDs were part of a multimodal approach to pain control compared with no multimodal pain control were excluded as the impact of NSAIDs was unable to be determined. Specifically, studies that did not report the specific impact of NSAIDs independent of other medications were not included.
Titles and abstracts were screened independently by two team members for inclusion in our meta-analysis. Conflicts were blindly adjudicated by a third member. Full-text review was also performed by two team members working independently, with conflicts adjudicated by a blinded third member. Included articles had their reference lists reviewed by two team members for identification of potential additional articles. The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) flow diagram for our systematic review is depicted in figure 1.
Figure 1PRISMA flow diagram. PRISMA, preferred reporting items for systematic reviews and meta-analyses.
Data extraction and methodology
Data extraction from each included study was performed using a standardized data collection sheet and was performed in duplicate. Discrepancies were adjudicated by a third author. Data extracted included authors, journal, publication year, study design, number of patients, type of fracture(s), indication for NSAIDs (heterotopic ossification prevention vs pain control), type of NSAID(s) used, dose and duration (if available), number of patients in each experimental and control arms, as well as the critical outcomes previously listed. The definition of non-union in the literature is inconsistent and varied widely between studies. Due to the variability, the study definition was used and captured.
Meta-analysis was performed in Review Manager (RevMan V.5.3, Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration, 2014) with random-effects modeling to generate forest plots. For dichotomous outcomes, the Mantel-Haenszel random effects model was used to calculate a pooled event rate and OR. For continuous outcomes, inverse variance was used to calculate a mean difference between interventions. Treatment effects were calculated with each study weight being proportional to the number of subjects it contributed to each outcome. Heterogeneity was calculated and quantified with I2. High heterogeneity was considered present for I2 values >75%, moderate for I2 values of 50% to 74% and low if I2 <50%.
Publication bias was evaluated using the Egger test, and the GRADE framework was applied to all quantified outcomes for assessment of bias, publication bias, inconsistency, imprecision and indirectness. Evidence profiles were created for each PICO using GRADEpro GDT software (GRADEpro Guideline Development Tool, McMaster University, 2015).
All committee members voted initially independently taking into consideration the quality of evidence, relationship of benefits and harms, perceived patient values and preferences, and resource utilization. Our PICO questions and analysis results (forest plots, GradePRO table, risk of bias assessment and summary of study types) were submitted to two external GRADE experts for blind review. No Institutional ethics was not necessary as we did not examine any individual patient data.