Professional Documents
Culture Documents
and not just ketorolac. Sources included in A number of the investigations included in the 2013
appraised SRs were not separately appraised to Cochrane Review9 were also included in four other
take advantage of higher level evidence. SRs with meta-analysis.10-13 The SR by Chan and
Parikh10 included six investigations, another11
included 14 investigations, a third12 included nine
Critical Appraisal of the Literature investigations, and a fourth13 included five investiga-
Evidence Sources tions. All but one SR10 examined multiple NSAIDs,
including ibuprofen,9,11-13 ketoprofen,9,11,12
9,11,12 9,11,12
The search yielded 88 potential sources (Figure 1). diclofenac, tenoxicam, rofecoxib,9,11
11,12
Twenty-one met inclusion criteria based on exam- indomethacin, celecoxib, flurbiprofen,11 lor-
11
ining the title, 14 based on abstract, and 5 based on noxicam, naproxen,12 and nimesulide.12
11
88 sources identified by
systematic search
67 sources excluded based on
lack of title relevancy
Figure 1. Flowchart summary of evidence examining postoperative hemorrhage in patients undergoing tonsillec-
tomy who received perioperative nonsteroidal anti-inflammatory drugs.
4
Table 1. Summary of Evidence Examining Postoperative Hemorrhage in Patients Undergoing Tonsillectomy Receiving Perioperative
Nonsteroidal Anti-inflammatory Drugs
Evidence Type
N
Evidence Source Level of Evidence* Outcome Comments
Lewis et al (2013)9 SR of 15 RCTs Outcome 1: reoperation for bleeding—14 Cochrane Review
1,101 subjects studies, 1,044 subjects—Peto OR, 1.69 (95% Large CIs suggesting inadequate
Level I CI, 0.71 to 4.01) sample size
Outcome 2: bleeding, no reoperation—10 Large degree heterogeneity (c2 and I2
studies, 745 subjects—Peto OR, 0.99 (95% statistics)
CI, 0.41 to 2.40) Fixed-effects model used
Conclusion: insufficient evidence to
Ketorolac subgroup:
exclude an increased risk of bleeding
Outcome 1: reoperation for bleeding—Peto
when NSAIDs are used in pediatric
OR, 3.82 (95% CI, 1.03 to 14.10)
tonsillectomy
Five trials, 359 children; total events: nine
ketorolac, one control
Heterogeneity: c22 5 0.56, (P 5 .76);
I2 5 0.0%
Overall effect: Z 5 2.01 (P 5 .044)
Outcome 2: bleeding, no
reoperation—Peto OR, 1.19 (95% CI, 0.45 to
5
6
Table 1. Continued
Evidence Type
N
Evidence Source Level of Evidence* Outcome Comments
reporting bias, and absences from other sources of independently reviewed and sorted studies based
bias. on NSAID type, children versus adults, type and
dose of NSAID, timing and route of administration,
Two authors independently assessed trial quality control, surgical technique, and bleeding-related
and extracted the data for the Cochrane Review.9 outcomes. A meta-analysis was performed using a
There were no studies with a low risk of bias in statistical meta-analysis program in studies in
all domains. Outcome measures included periop- which the number of individuals with each
erative bleeding requiring surgical intervention, outcome was reported. Subgroup analysis was
perioperative bleeding requiring nonsurgical conducted for subjects receiving ketorolac. To
intervention, and vomiting. Nonsurgical interven- combine studies measuring different outcomes,
tions included intravenous fluid therapy, pro- the most severe outcome was selected from each
longed observation and nonsurgical hemostasis. study. The studies were grouped by specific out-
Subgroup analyses included type of NSAID, timing comes: bleeding requiring reoperation, bleeding
of administration, and type of control. Six of the 15 requiring readmission, secondary bleeds, and
studies examined ketorolac as the intervention. bleeding that could be managed with conservative
Subgroup comparisons were achievable for ketor- or no measures. There was statistical homogeneity
olac. Outcome 1, perioperative bleeding that among the 36 studies, nine of which had no severe
required surgical intervention, was examined in outcomes.
five studies and involved 359 children. Outcome
2, perioperative bleeding that required nonsur- In the SR by Møiniche et al,12 data from 25 RCTs
gical intervention, was examined in five studies including 33 comparisons were included following
and involved 365 children. recommendations by the Quality of Reporting of
Meta-analyses statement.17 Adult and pediatric
The SR by Chan and Parikh10 included retrospec- data included 970 subjects receiving NSAIDs and
tive and prospective studies involving children 883 receiving a non-NSAID treatment or placebo.
and adults. Two authors independently identified All trials were of satisfactory methodological qual-
studies involving tonsillectomy and ketorolac ity, producing a median Oxford score of 4. After
administration during the perioperative period. an ante hoc decision, the incidence of reoperation
Ten studies were included, and individual-level because of bleeding was the primary outcome mea-
data were extracted for meta-analysis. Risk of bias sure. The presence of 0-event trials was identified
was evaluated based on randomization, treatment as a methodological problem. Zero-event trials
allocation, blinding, treatment and outcome stan- occur with very rare events or when the trials are
dardization, selective reporting, and complete of limited size. To overcome this issue, two statisti-
data. Analysis revealed a significant risk of bias cal models were used—the Peto method18 and the
for all reports across many domains, especially method described by Shadish and Haddock15—to
treatment allocation and blinding. A total relative calculate odds ratios (ORs) with 95% confidence
risk (RR) was calculated from the 10 studies, as intervals (CIs) for rare events.
well as independently for pediatric and adult
data. The adult data included three studies, two The SR with meta-analysis of RCTs by Marret et al13
retrospective and one prospective. Nine pediatric used a very rigid methodology. Included were
studies included three retrospective and six pro- seven randomized double-blind pediatric studies.
spective. Overall, postoperative hemorrhage rates Each study was subjected to the Jadad Composite
from individual studies and pooled subgroups Scale19 in an effort to decrease bias because of
were reported. the inclusion of low-quality trials defined as scores
less than 2. All studies were of at least moderate
In the SR with meta-analysis by Riggin et al,11 36 quality with a Jadad Assessment Scale score of 3
studies met inclusion criteria, which included or more. The Mantel-Haenszel20 procedure was
1,747 children and 1,446 adults. All were used, which pooled ORs that were assigned
RCTs involving NSAIDs administered to patients weights proportional to the inverse of the
undergoing tonsillectomy; a control group treated within-study OR variance. Heterogeneity statistics
with paracetamol, placebo, or opioids; and such as a c2 (k 2 1 degree of freedom) were
bleeding measured as an outcome. Two authors described. For primary evaluation criterion, the
8 PHILLIPS-REED, AUSTIN, AND RODRIGUEZ
number needed to harm—the number of NSAID (95% CI, 1.32 to 3.15; P , .002) for post-
subjects required before one more NSAID subject tonsillectomy hemorrhage with perioperative ke-
would experience a harmful outcome—was torolac administration was calculated. This finding
computed as the inverse of the difference of the implied a significantly increased risk of post-
proportion of patients who required reoperation tonsillectomy hemorrhage associated with ketoro-
in the NSAID and control groups. All tests were lac use (P , .001). When pediatric and adult data
two sided, and P , .05 was considered statistically were assessed independently, there was a greatly
significant. The primary outcome measure was the increased risk in adults (RR, 5.64; 95% CI, 2.08 to
need for surgical electrocautery to stop bleeding; 15.27; P , .001) compared with children (RR,
the secondary evaluation criterion was postopera- 1.39; 95% CI, 0.84 to 2.30; P 5 .20). From these re-
tive bleeding requiring a change in postoperative sults, adults receiving ketorolac had over five times
management. increased risk of bleeding compared with the
controls. The three adult studies included two
Findings From the Evidence retrospective RCTs and one prospective RCT. All
three studies were consistent in reporting an
Major characteristics of the evidence sources9-13 increased risk of bleeding with ketorolac.
are presented in the table. Collectively, these
accounted for a total of 46 studies, which In the nine pediatric trials included in the SR by
included 12 studies examining ketorolac. All Chan and Parikh,10 the retrospective and prospec-
five SRs9-13 included a number of investigations tive results were consistent. The three retrospec-
examining subjects who underwent tive studies produced an RR of 1.31 (95% CI,
tonsillectomy or adenotonsillectomy and who 0.66 to 2.59; P 5 .45), and the six prospective
received ketorolac. Two SRs9,13 focused only studies yielded an RR of 1.49 (95% CI, 0.71 to
on children and excluded all patients older 3.13; P 5 .30). The timing of ketorolac administra-
than 16 years. Two SRs9,12 used the Peto OR tion did not affect the RR and showed an equiva-
method for rare and nonoccurring events. The lent risk of postoperative bleeding: preoperative
disadvantage to the Peto method is that (RR, 1.43; 95% CI, 0.6 to 3.42; P 5 .43) and intra-
relevant data might be censored, thus allowing operative or postoperative (RR, 1.37; 95% CI,
bias to be introduced. 0.74 to 2.54; P 5 .32). These results suggest that
there was no significantly increased risk of
The Cochrane Review9 reported a subgroup anal- bleeding in children.
ysis for ketorolac. Five trials involving 359 subjects
compared ketorolac with another analgesic or pla- The SR by Riggin et al11 found no significant
cebo and assessed cases of perioperative bleeding increased risk of post-tonsillectomy bleeding in
that required further surgical intervention. The adults and children receiving ketorolac. Outcome
intervention group had an increased risk (Peto data from eight statistically homogenous ketorolac
OR, 3.82; 95% CI, 1.03 to 14.10), and the statistical studies (heterogeneity P 5 .717) were subana-
test for differences between subgroups was not lyzed, and five reported no adverse outcomes in
significant (P 5 .1). Five studies with 365 subjects either treatment arm. The one significant result
assessed ketorolac and perioperative bleeding that was in the secondary subanalysis of studies in
required nonsurgical intervention. The difference which subjects were given NSAIDs postopera-
in the effect estimates was smaller (Peto OR, tively only, which revealed an increased risk of
1.19; 95% CI, 0.45 to 3.14) and again showed no bleeding (OR, 2.02; 95% CI, 1.25 to 3.27;
statistical evidence of a difference between sub- P 5 .0042). There was no significant bleeding
groups (P 5 .36). Based on results consistent risk in the preoperative or perioperative analysis.
with an increased and decreased risk of bleeding,
it was concluded that insufficient evidence exists The SR by Møiniche et al12 examined intraopera-
to exclude an increased risk of bleeding when tive blood loss, postoperative bleeding, and admis-
NSAIDs are used in pediatric tonsillectomy. sion as measures of bleeding along with
reoperation because of bleeding as endpoints
Ten studies examining ketorolac were reviewed by and found that the necessity for reoperation was
Chan and Parikh,10 and a summary RR of 2.04 0% to 12.5% in the NSAID group compared with
TONSILLECTOMY AND KETOROLAC 9
0% to 10% in the control group. The OR was statis- (95% CI, 17 to 144), which suggests that using
tically significant (Peto OR, 2.33; 95% CI, 1.12 to NSAIDs in 29 subjects after tonsillectomy would
4.83) and borderline significant with the method be accompanied by a hemorrhage severe enough
described by Shadish and Haddock15 (OR, 1.92; to require reoperation of one subject. Six of the
95% CI, 1.00 to 3.71). The number needed to treat seven trials yielded an OR greater than one for re-
was 60 (95% CI, 34 to 277). operation for hemostasis, although the differences
were not statistically significant. One of the
When the NSAID was given postoperatively, included studies suggested an increase in immedi-
Møiniche et al12 found that 3.2% of the subjects ate postoperative bleeding.
required a reoperation compared with 0.7% of
the control subjects. This difference was statisti- Summary
cally significant (Peto OR, 4.36; 95% CI, 1.69 to
11.26; OR, 3.27; 95% CI, 1.39 to 7.68), and the The results differed for bleeding associated with
number needed to treat was 40 (95% CI, 23 to the administration of an NSAID after tonsillectomy
123), meaning that for every 40 subjects who even between the SRs.9-13 This may be because of
receive an NSAID dose postoperatively, one sub- the differences in outcomes measured or to the
ject will require a reoperation. Of the four bleeding heterogeneity of the control groups. Bleeding
endpoints used, only reoperation showed a statis- requiring further surgical intervention is an
tically significant result in favor of the non-NSAID uncommon event after tonsillectomy. The
treatment and was dependent on the selection of interpretations from SRs9-13 available to date
the statistical model. Event rates of admission yielded incongruent conclusions on the
because of bleeding were very small, 2.5% without increased risk of bleeding after tonsillectomy
and 5.1% with NSAIDs, and calculations were when NSAIDs were given. The 2013 Cochrane
based on 506 subjects. A post hoc power analysis Review9 concluded that there is insufficient evi-
using these numbers and a significance level of dence to exclude an increased risk of bleeding
5% revealed a statistical power of approximately when NSAIDs are used in pediatric tonsillectomy.
50% only. Both Riggin et al11 and Møiniche et al12 The subgroup of children receiving ketorolac had
found the timing of NSAID administration to be sig- an increased risk of bleeding. One SR10 concluded
nificant. that ketorolac can be used safely in children but is
associated with a fivefold increased bleeding risk
Marret et al13 reported that of the 243 subjects in adults. Another SR11 concluded that NSAIDs
who did not receive NSAIDs, 13 subjects had pri- can be considered safe in pediatric tonsillectomy.
mary or secondary bleeding (5.3%; range, 0% to The fourth SR12 concluded that NSAIDs should
25%), and of the 262 subjects who received be used cautiously in tonsillectomy. The final
NSAIDs postoperatively, 24 subjects (9.2%; range, SR13 concluded that NSAIDs should not be used af-
0% to 25%) had postoperative bleeding (OR, 1.8; ter tonsillectomy.
95% CI, 0.9 to 3.4). The overall incidences of post-
operative bleeding treated medically or surgically The evidence9-13 offered conflicting conclusions.
and of postoperative bleeding requiring reopera- Given the danger of bleeding after tonsillectomy,
tion were 7.3% and 2.6%, respectively. A significant the perianesthesia team should weigh carefully
difference in the rate of reoperation was reported the risks and complications of NSAIDs and
with 0.8% for the controls and 4.2% for the NSAID- consider using alternative analgesics. Future
treated subjects (OR, 3.8; 95% CI, 1.3 to 11.5; studies need to be carefully designed and
P 5 .02) representing a very large increase in the controlled to avoid the risks of bias identified in
OR. The number needed to harm was 29 prior efforts.
References
1. Baugh RF, Archer SM, Mitchell RB, et al. Clinical practice 2. Gillis JC, Brogden RN. Ketorolac. A reappraisal of its phar-
guideline: Tonsillectomy in children. Otolaryngol Head Neck macodynamic and pharmacokinetic properties and therapeutic
Surg. 2011;144(1 suppl):S1-S30. use in pain management. Drugs. 1997;53:139-188.
10 PHILLIPS-REED, AUSTIN, AND RODRIGUEZ
3. Judkins JH, Dray TG, Hubbell RN. Intraoperative ketorolac 13. Marret E, Flahault A, Samama CM, Bonnet F. Effects of
and posttonsillectomy bleeding. Arch Otolaryngol Head Neck postoperative, nonsteroidal, antiinflammatory drugs on
Surg. 1996;122:937-940. bleeding risk after tonsillectomy: Meta-analysis of randomized,
4. McNeill RA. A history of tonsillectomy: Two millenia of controlled trials. Anesthesiology. 2003;98:1497-1502.
trauma, haemorrhage and controversy. Ulster Med J. 1960;29: 14. Melnyk BM, Fineout-Overholt E. Making the case for
59-63. evidenced-based practice and cultivating a spirit of inquiry. In:
5. Isaacson G. Tonsillectomy healing. Ann Otol Rhinol Melnyk BM, Fineout-Overholt E, eds. Evidence-Based Practice
Laryngol. 2012;121:645-649. in Nursing and Health Care: A Guide to Best Practice, 2nd
6. Sinha VR, Kumar RV, Singh G. Ketorolac tromethamine for- ed. Philadelphia, PA: Lippincott Williams & Wilkins; 2011:3-24.
mulations: An overview. Expert Opin Drug Deliv. 2009;6: 15. Shadish WR, Haddock CK. Combining estimates of effect
961-975. size. In: Cooper H, Hedges LV, eds. The Handbook of Research
7. Forrest JB, Heitlinger EL, Revell S. Ketorolac for postop- Synthesis, 1st ed. New York, NY: The Russell Sage Foundation;
erative pain management in children. Drug Saf. 1997;16: 1994:261-280.
309-329. 16. Guyatt G, Furukawa TA. Advanced topics in the validity
8. Richardson WS, Wilson MC, Nishikawa J, Hayward RS. The of therapy trials: An illustration of bias and random error. In:
well-built clinical question: A key to evidence-based decisions. Guyatt G, Rennie D, Meade M, Cook D, eds. Users’ Guide to
ACP J Club. 1995;123:A12-A13. the Medical Literature, 2nd ed. New York, NY: McGraw-Hill;
9. Lewis SR, Nicholson A, Cardwell ME, Siviter G, Smith AF. 2008:109-112.
Nonsteroidal anti-inflammatory drugs and perioperative 17. Moher D, Cook DJ, Eastwood S, Olkin I, Rennie D,
bleeding in paediatric tonsillectomy. Cochrane Database Syst Stroup DF. Improving the quality of reports of meta-analyses
Rev 2013;CD003591. of randomised controlled trials: The QUOROM statement. Lan-
10. Chan DK, Parikh SR. Perioperative ketorolac increases cet. 1999;354:1896-1900.
post-tonsillectomy hemorrhage in adults, but not children. 18. Yusuf S, Peto R, Lewis J, Collins R, Sleight P. Beta
Laryngoscope. 2014;124:1789-1793. blockade during and after myocardial infarction: An overview
11. Riggin L, Ramakrishna J, Sommer DD, Koren G. A 2013 of the randomized trials. Prog Cardiovasc Dis. 1985;27:
updated systematic review & meta-analysis of 36 randomized 335-371.
controlled trials; no apparent effects of non steroidal anti- 19. Jadad AR, Moore RA, Carroll D, et al. Assessing the quality
inflammatory agents on the risk of bleeding after tonsillectomy. of reports of randomized clinical trials: Is blinding necessary?
Clin Otolaryngol. 2013;38:115-129. Control Clin Trials. 1996;17:1-12.
12. Møiniche S, Rømsing J, Dahl JB, Tramer MR. Nonsteroidal 20. Lachin JM. Power and sample size evaluation for the
antiinflammatory drugs and the risk of operative site bleeding Cochran-Mantel-Haenszel mean score (Wilcoxon rank sum)
after tonsillectomy: A quantitative systematic review. Anesth test and the Cochran-Armitage test for trend. Stat Med. 2011;
Analg. 2003;96:68-77. 30:3057-3066.