2017-18 HSC Section 3 Green Book

outcome measurement; high: significant variation in outcome measurement. 6. Selective reporting. Low: well-described inclusion/exclusion criteria; mid: unclear inclusion/exclusion criteria; high: no inclusion/exclusion criteria. 7. Complete data. Low: all data present for all subjects; mid: 0% to 10% data missing; high: > 10% data missing. Data were compiled from included studies by direct extrac- tion of individual patient-level data and subjected to meta- analysis. Data collected were as follows: patient age (or age range), study design, perioperative ketorolac administration (yes/no), and postoperative hemorrhage (yes/no). The summary relative risk (RR) of post-tonsillectomy hemorrhage for ketorolac compared to control was calculated using a random effects model (metan function, Stata 12.1; StataCorp, College Station, TX). All RR values are presented with 95% confidence intervals (CIs) and P values assessing statistical significance relative to the null hypothesis of no difference in risk. Subgroup analyses were performed on specific subsets of studies based on their design and patient population. Performance, reporting, and analysis of this systematic review and meta-analysis has been done in accordance with PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines. 7 Of the 21 abstracts reviewed, 12 studies fulfilled the inclusion criteria of being primary studies describing the use of ketorolac in association with tonsillectomy. Nine studies were excluded on title/abstract review for the following reasons: six were review articles, two did not involve ketorolac, and one did not involve tonsillectomy. Two of the remaining 12 studies were subsequently excluded after full-length review for not reporting postoper- ative hemorrhage rates. Of the remaining 10 studies 1,4,8–15 (Fig. 1), seven exclusively included children under the age of 18, though the exact age range varied from one study to another. Two studies 4,8 included both adults and children; data specific to children, as defined by the study, were extracted from each study and included for further analy- sis. One study 1 exclusively included adults over the age of 18 years. Except for one study, 8 all patients > 18 years old were considered adults, and all patients 18 years old were considered children. In one study, 8 the reporting cutoff was at age 13 years, therefore some patients from age 13 to 18 years would have been reported together with the over- 18 adults. Definition of postoperative bleeding varied between studies, ranging from any bleeding noticed by the patient to a return to operating room for control. Analysis of risk of bias (Table I) revealed significant risk of bias across many domains for all reports, including both pro- spective and retrospective studies. Meta-analysis When all 10 studies are considered together, a sum- mary RR of 2.04 (95% CI: 1.32–3.15; P < .002; Fig. 2) was calculated. This implies a significantly increased risk of post-tonsillectomy hemorrhage associated with ketorolac use ( P < .001). However, when pediatric and adult patients were assessed independently, there was a greatly increased RESULTS Systematic Review

Fig. 1. Flowchart of study inclusion. [Color figure can be viewed in the online issue, which is available at www.laryngoscope.com.]

MATERIALS AND METHODS Systematic Review

PubMed search from 1970 to 2013 with the terms “[ketoro- lac OR toradol] AND tonsillectomy” was performed by two inde- pendent authors ( D . K . C ., S . R . P .). Inclusion criteria were as follows: 1) original study reporting unique individual patient- level data; 2) patients underwent tonsillectomy by any tech- nique; 3) two groups assessed, those who received ketorolac in the perioperative period and those who did not. Studies in any language were considered. All patients, including children and adults, were included. There was no mini- mum or maximum length of follow-up required after interven- tion. Abstracts were reviewed. Studies that obviously did not fulfill inclusion criteria, including review articles, basic-science studies, case reports, as well as reports either not involving ketor- olac or tonsillectomy were excluded. Conflicts between the two reviewers would be resolved after discussion; however, this did not apply for any of the studies considered. The remaining articles were submitted to full-length review and individual-level data extracted for meta-analysis (Fig. 1). In addition, the refer- ence lists for each included article were reviewed, and no other articles fulfilling inclusion criteria were identified. Each study was assessed for risk of bias according to guidelines put forth by the Cochrane Collaboration. 6 This was evaluated on seven domains as follows: 1. Randomization. Low risk: adequate description of appropri- ate randomization technique; mid: randomization performed, but no technique specified; high: no randomization. 2. Treatment allocation. Low: allocation procedure unbiased and described fully; mid: allocation procedure described, but unclear bias; high: no allocation procedure described 3. Blinding. Low: all parties blinded; mid: some parties blinded; high: no parties blinded. 4. Standardization (treatment). Low: all subjects undergoing consistent and well-defined treatment plans; mid: some vari- ation in treatment; high: significant variation in treatment. 5. Standardization (outcome). Low: outcome measures deter- mined identically for all subjects; mid: some variation in

Laryngoscope 124: August 2014

Chan and Parikh: Ketorolac and Tonsillectomy

253