2017-18 HSC Section 3 Green Book

TABLE I. Risk of Bias of Included Studies.

Treatment Allocation Blinding

Standardization (Treatment)

Standardization (Outcome)

Selective Reporting

Complete Data

Study

Year

Design Randomization

Agrawal 10

1999

R

High

High

High

Low

Mid

Low

Low

Bailey 1

1997 1995

P

Mid

High High

Low

Low

Mid Mid

Low

Low Low

Gallagher 8 Gunter 11 Judkins 4 Keidan 12 Romsing 13

R

High

High

Mid

Mid

1995

P

Low

High

Low

Low

Low

Low

Low

1996 2004

R

High

High High

High High

Mid

Mid

Mid

Low Low

P

Mid

Low

Low

Low

1998

P

Low

Mid

Low

Low

Low

Low

Low

Rusy 14

1995 1996

P P

Mid

High High

High High

Low

Low Low

Low Low

Low Low

Splinter 9 Sutter 15

Low

Mid

1995

P

High

High

Mid

Mid

Low

Low

Low

Each of the 10 included studies was assessed for risk of bias in seven domains according to Cochrane Collaboration guidelines. 6 P 5 prospective, randomized, controlled study; R 5 retrospective case-control study.

Tonsillectomy technique and indication for tonsillec- tomy are theoretical confounders for this analysis. We did not include tonsillectomy technique in the analysis because of variability in reporting; most prospective studies specified and controlled for technique in their two comparison groups, which is reflected in the low risk of bias with treatment standardization (Table I), whereas retrospective studies had incomplete reporting on technique with respect to assignment of outcomes. Tonsillectomy technique has not, however, been shown definitively to be related to differential hemorrhage risk 16 ; therefore, it likely has a limited confounding role. Indication for tonsillectomy (tonsillitis or sleep apnea) was not reported in any of the included studies at the individual patient level, so it could not be assessed as a confounder. Length of follow-up is a severe limitation in the interpretation of these studies. Most prospective studies only report 1 to 2 days of study-mandated follow- up. It is not clear from these studies what the timing of the post-tonsillectomy hemorrhage was in the individual cases. The retrospective studies relied on chart review for ascertainment of post-tonsillectomy hemorrhage, and none of them specified follow-up time in their methods. These studies are subject to significant reporting bias. In adults, two retrospective and one prospective study are highly consistent in reporting an increased risk of bleeding with ketorolac use. In children, 9 studies were analyzed; eight of these (three retrospective, five prospective) were highly consistent, with RRs ranging from 0.76 to 2.02. One of these studies 9 was an outlier, with an RR of 9.17 (95% CI: 0.53–159.14; P 5 .13). This prospective, randomized, single-blinded, placebo-con- trolled study examined preoperative ketorolac, and was halted after five post-tonsillectomy hemorrhages were noted in the ketorolac group. There were no notable study design features that would explain these aberrant results; in general, our meta-analysis did not demon- strate a difference in RR with preoperative and postop- erative ketorolac administration. It remains a possibility that an increased risk of bleeding exists in children, and that the existing studies

risk in adults (RR: 5.64; 95% CI: 2.08–15.27; P < .001) com- pared to children (RR: 1.39; 95% CI: 0.84–2.30; P 5 .20). However, adults receiving ketorolac had over five times the increased risk of bleeding, which was statistically signifi- cantly elevated compared to control, but there was no sig- nificantly increased risk of bleeding in children. The adult data included three studies, two retrospective and one pro- spective randomized, blinded, controlled; all three studies were consistent in reporting increased risk with ketorolac. Nine pediatric studies were included. The results from the retrospective and prospective studies were con- sistent; six prospective studies gave a summary RR of 1.49 (95% CI: 0.71–3.13; P 5 .30), whereas three retro- spective studies yielded a summary RR of 1.31 (95% CI: 0.66–2.59; P 5 .45). Timing of ketorolac administration also did not affect the RR: preoperative (RR: 1.43; 95% CI: 0.60–3.42; P 5 .43) and intra- or postoperative (RR: 1.37; 95% CI: 0.74–2.54; P 5 .32) ketorolac administra- tion had equivalent risk of post-tonsillectomy bleeding. Overall postoperative hemorrhage rates from individual studies and pooled subgroups are shown in Table II. DISCUSSION Perioperative administration of ketorolac is associ- ated with an increased risk of post-tonsillectomy hemor- rhage in adults but not in children. Results of seven prospective, randomized controlled studies, as well as out- comes reported in three larger retrospective case-control studies, support these findings. Analysis of validity and bias of these studies demonstrated some limitations. Among the prospective studies, there was inconsistent, and usually sparse, detail regarding the exact method of randomization and treatment allocation. Furthermore, blinding was not always sufficient to completely remove risk of bias. Though outcome assessment and reporting were consistent within each study, the definition of post- tonsillectomy hemorrhage varied from one study to another. In particular, many studies (Fig. 2) had only 24- or 48-hour follow-up, and thus would have missed all secondary hemorrhage events.

Laryngoscope 124: August 2014

Chan and Parikh: Ketorolac and Tonsillectomy

254