Secondary Logo

Journal Logo

Nonsteroidal Antiinflammatory Drugs and the Risk of Operative Site Bleeding After Tonsillectomy: A Quantitative Systematic Review

Møiniche, Steen, MD*,; Rømsing, Janne, MS, Pharm, PhD†,; Dahl, Jørgen B., MD, DMSc*,; Tramèr, Martin R., MD, DPhil

doi: 10.1213/00000539-200301000-00015
PEDIATRIC ANESTHESIA: Review Article
Free

The use of nonsteroidal antiinflammatory drugs (NSAIDs) for analgesia after tonsillectomy is controversial because NSAIDS, through platelet inhibition, may increase the risk of perioperative bleeding. We systematically searched for randomized, controlled trials that reported on the incidence of perioperative bleeding attributable to the use of NSAIDs in patients undergoing tonsillectomy. As secondary outcome measures, we analyzed the quality of pain relief and the incidence of postoperative nausea and vomiting. Twenty-five studies with data from 970 patients receiving a NSAID and 883 receiving a non-NSAID treatment or a placebo were analyzed. Data were combined using a fixed-effect model. Of four bleeding end points (intraoperative blood loss, postoperative bleeding, hospital admission, and reoperation because of bleeding), only reoperation happened significantly more often with NSAIDs: Peto-odds ratio, 2.33 (95% confidence interval [CI], 1.12–4.83) and number-needed-to-treat, 60 (95% CI, 34–277). Compared with opioids, NSAIDs were equianalgesic, and the risk of emesis was significantly decreased (relative risk, 0.73; 95% CI, 0.63–0.85; numbers-needed-to-treat, 9; 95% CI, 5–19).

*Department of Anesthesiology and Intensive Care Medicine, Herlev University Hospital; †Department of Pharmaceutics, The Royal Danish School of Pharmacy, Copenhagen, Denmark; and ‡Division of Anesthesiology, Geneva University Hospitals, Switzerland

September 23, 2002.

Dr Dahl was supported by a grant (No. 22000947) from the Danish Medical Research Council, Copenhagen, Denmark, and Dr Tramèr is a beneficiary of a PROSPER grant (No. 3233–051939.97/2) from the Swiss National Science Foundation.

Address correspondence to Steen Møiniche, MD, Department of Anesthesiology and Intensive Care Medicine, Herlev University Hospital, DK-2730, Copenhagen, Denmark. Address e-mail to moiniche@dadlnet.dk.

Nonsteroidal antiinflammatory drugs (NSAIDs) are an attractive alternative to opioids for the control of postoperative pain. They are effective analgesics (1), and they lack opioid-related adverse effects such as sedation, emesis, pruritus, or respiratory depression. However, NSAIDs inhibit platelet function, and this feature may limit their usefulness in patients who are prone to postoperative bleeding. Prospectively collected data on the risk of operative site bleeding associated with the perioperative use of NSAIDs are sparse. The evidence for a causative association is weak (2,3), and the magnitude of any deleterious effect remains unknown.

The aim of this systematic review was to critically appraise the existing data on the incidence of perioperative bleeding complications caused by NSAIDs and to quantify the potential impact of NSAIDs on bleeding. As a clinical setting, we have chosen tonsillectomy for two reasons: first, tonsillectomy is a very common and well-defined surgical procedure. Second, bleeding is a serious complication after tonsillectomy, which may contribute to postoperative morbidity (4,5). Thus, for this setting, we found it important to evaluate the risk-benefit ratio of NSAIDs.

Back to Top | Article Outline

Methods

To perform this systematic review, we followed the recommendations of the QUOROM statement (6). A systematic search of the literature was performed without language restriction. We searched in Medline (from 1966), Embase (from 1989), and the Cochrane Controlled Trials Register (2001) using the words NSAID, nonsteroidal antiinflammatory drug, tonsillectomy, and individual drug names (diclofenac, ketorolac, fenbufen, flurbiprofen, ibuprofen, indomethacin, ketoprofen, naproxen, piroxicam, meloxicam, tenoxicam, nimesulide, rofecoxib, celecoxib, tolmetin, fenoprofen, sulindac, etodolac, nabumetone, tiaprofenic acid, meclofenamic acid, acemetacin, mefenamic acid, aceclofenac, tolfenamic acid, lornoxicam, parecoxib, fentiazac, suprofen, benzydamine, and niflumic acid). The last electronic search was done in December 2001. Additional studies were identified from reference lists of retrieved reports and of relevant review articles. Abstracts or unpublished data were not considered.

Reports were considered if they were randomized, controlled trials of a NSAID (experimental intervention) compared with a non-NSAID treatment (control intervention) in adults or children undergoing tonsillectomy with or without adenoidectomy. Patients in the control group could receive a non-NSAID analgesic (for instance, an opioid), a placebo, or no treatment. Only trials in which the NSAID was administered systemically (i.e., orally, rectally, IV, or IM) and that reported data on one of three end points (intra- or postoperative bleeding, postoperative pain, or postoperative nausea and vomiting [PONV]) were considered.

Retrieved reports were screened by two authors (SM and JR). Each report that met the inclusion criteria was read independently by all authors to assess adequacy of randomization and blinding and to assess description of withdrawals (7). Discrepancies were resolved by discussion. Because there was an a priori agreement that reports without randomization would be excluded, the minimum score of an included trial was 1, and the maximum score was 5.

Information about NSAID regimens (drug, dose, and route of administration), comparators, number of patients enrolled and analyzed, study design, observation periods, surgical technique, and primary and secondary outcome measures were entered on standard collection sheets. This was performed by one investigator (JR) and checked by another (SM). Our primary outcome measures were bleeding complications. These included the volume of intraoperative blood loss, the incidence of postoperative bleeding, the rate of readmission because of bleeding, and the incidence of reoperation because of bleeding. Definitions of intraoperative blood loss were taken as reported in the original trials. Reoperation because of bleeding was regarded as the most relevant end point in this context. Secondary outcomes of interest were quality of pain relief and the incidence of PONV.

To estimate the frequency of rare events (postoperative bleeding, reoperation, and admission), we calculated odds ratios with 95% confidence intervals (CI) using two different models. First, we used the Peto-modified Mantel-Haenszel method (Peto-OR) that ignores trials without events (8). Second, if both cells of a sample were 0, we added 0.5 to all cells of that sample; we shall refer to this as the method described by Shadish and Haddock (9) (OR). Results were challenged with further sensitivity analyses, for instance, by comparing different NSAID regimens. For binary data of frequent events (PONV), we calculated relative risks (RR) with 95% CI. When the OR, Peto-OR, or RR indicated a statistically significant result (i.e., the 95% CI did not include 1), we calculated numbers-needed-to-treat (NNT) with 95% CI using the weighted means of the pooled experimental and control event rates (10). For continuous data, we calculated weighted mean differences (WMD), taking into account study size and sd as reported in the individual trials (ReviewManager software, version 4.0, Cochrane Collaboration, The Nordic Cochrane Center, Copenhagen). We used a fixed effect model throughout because we combined data only when they were clinically homogenous and because heterogeneity tests lack sensitivity (11).

Data from individual trials were graphically displayed using both forest plots to show relative efficacy of treatment effects and event rate scatters (12) to explore variability in event rates.

Quantitative analysis of data on analgesic efficacy was considered inappropriate because a large variety of outcome measures were reported in these trials. Therefore, for each trial, we qualitatively evaluated if the analgesic efficacy of the NSAID treatment was inferior, superior, or equal to the comparator treatment. This assessment was based on reported statistical significance (P < 0.05) of standardized pain mea-sures (scores of pain intensity and pain relief or consumption of supplementary analgesics) in the original investigations.

Back to Top | Article Outline

Results

We identified 63 potentially relevant trials of which 38 were subsequently excluded (13–50) (Fig. 1). Data from 25 randomized trials, including 33 comparisons, were included (51–75) (Table 1). In those, 1853 patients (970 received a NSAID) underwent tonsillectomy. The median number of patients included in the studies was 77 (range, 27–183 patients). The median quality score was 4 (range, 1–5; one trial scored 1, five scored 2, six scored 3, seven scored 4, and six scored 5). Twenty-one trials were double-blinded, of which 18 described an adequate method of blinding (for instance, identical ampoules). Eight NSAIDs were tested, all in analgesic doses (diclofenac, ketorolac, ibuprofen, indomethacin, ketoprofen, tenoxicam, naproxen, and nimesulide). Controls received a placebo or no treatment in 11 studies, an opioid (morphine, pethidine, papaveretum, tramadol, tilidine, and codeine) in 9 studies, and acetaminophen or acetaminophen plus codeine in 3 studies each (Table 1). The surgical technique was reported in 10 studies; tonsillectomy was performed by electrocautery dissection in 8 (52,54–56,61,62,64,67) and ad modum Boyle-Davis in 2 (53,59). Nine studies were in adults (51,56,57,59,60,63–65,70), 14 in children (54,55,58,61, 62,66–69,71–75), and 2 in both adults and children (52,53).

Figure 1

Figure 1

Table 1

Table 1

Table 1

Table 1

In 9 studies with 12 treatment arms, a NSAID was administered before tonsillectomy, and intraoperative blood loss was an end point (56,60–62,64–66,70,71). Blood loss was estimated by measuring the volume of blood in the suction bottles and in the swabs. Mean measured blood loss was 2.1 ± 0.9 mL/kg with a NSAID compared with 1.8 ± 0.9 mL/kg with the control. The WMD was not significantly different (WMD, 0.38 mL/kg; 95% CI, −0.06 to –0.81).

Dichotomous data on postoperative bleeding were reported in 16 studies with 19 treatment arms (51–59,62–64,66,67,69,71) (Fig. 2). With both NSAIDs and controls, there was a wide variability in event rates ranging from 0% to almost 30%. In 2 trials, no bleedings were reported (57,58). With NSAIDs, 73 of 685 patients (10.7%) had postoperative bleeding compared with 58 of 619 (9.4%) control patients, a difference that was not statistically significant and independent of the model used (Fig. 2).

Figure 2

Figure 2

Readmission or unanticipated admission because of bleeding was reported in 8 studies with 9 treatment arms (51,52,54–57,60,66) (Fig. 3). With NSAIDs, the range of admissions was 0% to approximately 15%, and with controls, the range was between 0% and <10%. In 2 trials, no admissions were reported (57,60). With NSAIDs, 14 of 275 (5.1%) patients required admission or readmission compared with six of 231 (2.6%) control patients. This difference was not statistically significant and independent of the model used (Fig. 3).

Figure 3

Figure 3

Presence or absence of reoperation was reported in 21 studies with 27 treatment arms (51–71) (Fig. 4). With NSAIDs, the range of reoperations was 0% to approximately 13%, and with controls, the range was 0%–10%. In 7 trials, no patients were reoperated on because of bleeding (52,56–58,60,61,69). With NSAIDs, 24 of 877 (2.7%) patients had a reoperation compared with eight of 749 (1.1%) control patients. The OR was statistically significant with the Peto method (2.33; 95% CI, 1.12–4.83) and was borderline significant with the method described by Shadish and Haddock (9) (OR, 1.92; 95% CI, 1.00–3.71); the NNT was 60 (95% CI, 34–277).

Figure 4

Figure 4

The quantity of data and the quality of data reporting allowed us to perform four sensitivity analyses. Primary bleeding (within 24 h of tonsillectomy) was reported in 11 studies with 14 treatment arms (52–57,62–64,69,71). With NSAIDs, 44 of 494 (8.9%) patients had a primary bleed compared with 35 of 425 (8.2%) control patients. This difference was not statistically significant and independent of the model used (Peto-OR, 1.31; 95% CI, 0.80–2.13; OR, 1.27; 95% CI, 0.79–2.04). Secondary bleeding (beyond the initial 24-h period) was reported in 7 studies with 8 treatment arms (52–57,63). With NSAIDs, 14 of 237 (5.9%) patients had a secondary bleed compared with 15 of 246 (6.1%) control patients. This difference was not statistically significant and independent of the model used (Peto-OR, 0.86; 95% CI, 0.39–1.89; OR, 0.88; 95% CI, 0.41–1.89).

In five trials, NSAID treatment was started before surgery, and postoperative bleeding was an end point (56,62,64,66,71). With preoperative NSAIDs, 24 of 214 (11.2%) patients had a postoperative bleed compared with 22 of 189 (11.6%) control patients. This difference was not statistically significant and independent of the model used (Peto-OR and OR, 1.06; 95% CI, 0.57–1.98). In 13 trials, NSAID treatment was started after surgery, and postoperative bleeding was an end point (51–55,57–59,62–64,67,69). With postoperative NSAIDs, 49 of 471 (10.4%) patients had a bleed compared with 41 of 475 (8.6%) control patients. This difference was not statistically significant and independent of the model used (Peto-OR, 1.42; 95% CI, 0.90–2.24; OR, 1.41; 95% CI, 0.90–2.20).

In 10 trials, NSAID treatment was started before surgery, and reoperation was an end point (56,60–62,64–66,68,70,71). With preoperative NSAIDs, nine of 406 (2.2%) patients had a reoperation compared with five of 319 (1.6%) control patients. This difference was not statistically significant and independent of the model used (Peto-OR, 1.33; 95% CI, 0.45–3.94; OR, 1.20; 95% CI, 0.45–3.20). In 13 trials, NSAID treatment was started after surgery (51–55,57–59,62–64,67,69), and reoperation was an end point. With postoperative NSAIDs, 15 of 471 (3.2%) patients had a reoperation compared with three of 475 (0.7%) control patients. This difference was statistically significant (Peto-OR, 4.36, 95% CI, 1.69–11.26; OR, 3.27; 95% CI, 1.39–7.68), and the NNT was 40 (95% CI, 23–123).

In 11 trials, a single-dose NSAID regimen was given, and reoperation was an end point (54,56,60–63,65,66,68,69,71). With the single-dose NSAIDs, 13 of 448 (2.9%) patients had a reoperation compared with five of 421 (1.2%) control patients. This difference was not statistically significant and independent of the model used (Peto-OR, 2.42; 95% CI, 0.93–6.29; OR, 1.96; 95% CI, 0.83–4.64). In 10 trials, a multiple-dose NSAID regimen was given, and reoperation was an end point (51–53,55,57–59,64,67,70). With the multiple-dose NSAID regimen, 11 of 429 (2.6%) patients had a reoperation compared with three of 328 (1%) control patients. This difference was not statistically significant and independent of the model used (Peto-OR, 2.21; 95% CI, 0.72–6.84; OR, 1.87; 95% CI, 0.68–5.14).

There was a lack of data to evaluate the risk with individual NSAIDs, of different routes of administration, or to differentiate between adults and children.

Eleven studies compared a NSAID with placebo or no treatment and reported on pain outcomes (53,57,60,62,64,69,70,72–75) (Table 1). All studies except one (60) showed significantly improved pain relief with NSAIDs compared with control.

Eight studies compared a NSAID with an opioid and reported on pain outcomes (52,54,56,63,68,71, 72,75) (Table 1). Opioid regimens were IM or IV morphine 0.1–0.2 mg/kg (54,68), papaveretum 0.2–0.3 mg/kg (56,71,75), pethidine 1 mg/kg (63,72), tilidine 2.5 mg/kg (63), or oral tramadol approximately 1 mg/kg (52). In two studies (56,75), pain relief was significantly improved with a NSAID, in five studies, there was equivalence (52,54,63,71,72), and in one study, pain relief was significantly better with the opioid (68).

Three studies compared a NSAID with rectal acetaminophen 13–35 mg/kg (61,65) or oral acetaminophen approximately 10–15 mg/kg (58) and reported on pain outcomes (Table 1). All three studies showed equivalence in terms of analgesic efficacy.

Three studies compared a NSAID with acetaminophen- codeine administered in doses of approximately 10 mg/kg of acetaminophen and 0.4–1 mg/kg of codeine and reported on pain outcomes (55,59,67) (Table 1). Results were inconsistent with one trial each showing significantly improved (59), equivalent (67), or inferior (55) pain relief with the NSAID.

Most studies that reported on PONV did not distinguish between nausea and vomiting, and they only rarely reported on the length of the observation period (Table 1). With placebo or no treatment (data from 10 trials), 77 of 279 patients (27.6%) had PONV symptoms, with NSAIDs (19 trials), there were 251 of 794 (31.6%), and with codeine, pethidine, morphine, papaverine, or tramadol (11 trials), 217 of 445 (48.8%) patients vomited or were nauseous (Table 1). When all opioid-controlled trials were analyzed, the risk with opioids was significantly increased compared with NSAIDs (RR 0.73; 95% CI, 0.63–0.85); the NNT was 9 (95% CI, 5–19).

Back to Top | Article Outline

Discussion

These trials were of satisfactory methodological quality; the median Oxford score was 4. All studies were randomized; thus, selection bias should not have been a factor. At least 18 trials (72%) were properly blinded, minimizing the risk of observer bias. However, the results from this systematic review are ambiguous and should be interpreted carefully. Several issues contribute to this uncertainty.

First, of four bleeding end points (i.e., intraoperative blood loss, postoperative bleeding, admission because of bleeding, and reoperation because of bleeding), only one, reoperation, showed a statistically significant result in favor of the non-NSAID treatment. Furthermore, this statistical significance was dependent on the statistical model used (the Peto method). This, perhaps, reflects the difficulty in choosing the most relevant end point in this context. Posttonsillectomy bleeding, for instance, is difficult to determine. Numbers from the literature range from 0.28%–20%(5), and this large variability is most likely because of the uncertainty of investigators on how to define significant bleeding. Not surprisingly, we were unable to differentiate between minor and major bleeding episodes. Our ante hoc decision was that the incidence of reoperation because of bleeding was the most relevant primary outcome measure, both from a clinical and an economic point of view. Also, this end point is clearly dichotomous, and thus, there is a minor risk only of observation and interpretation bias. According to the literature, reoperation because of bleeding after tonsillectomy without NSAID exposure happens in approximately 0.3%–1.9%(4,76,77). In these trials, the reported incidence of reoperation without NSAID exposure (i.e., the control event rate) was in agreement with these figures.

Second, the risk-benefit ratio was not straightforward. If the NSAID-related absolute increase in the incidence of reoperation because of bleeding is about right, then we have to assume that of 100 patients undergoing tonsillectomy, less than 2 will require a reoperation (NNT, approximately 60) who would not have required one had they not received a NSAID. However, of the same 100 patients, 11 will not have PONV symptoms (NNT, 9) who would have had these symptoms had they received an opioid. This is a beneficial effect, although its magnitude is small compared with the efficacy of truly antiemetic interventions (78). Also, with NSAIDs, pain relief was significantly improved in 9 of 10 placebos or no treatment controlled trials and was of comparable efficacy in 12 of 14 comparisons with a non-NSAID analgesic. Thus, NSAIDs may be regarded as a rational choice for analgesia after tonsillectomy. However, the question is whether avoiding PONV in 11 patients is worth the potential risk and expense of bringing two patients back to the operation room.

Third, we were unable to identify any predictive factor for bleeding. The impact of single-dose versus multiple-dose regimens, for instance, could not be demonstrated. Data on children compared with adults, or on the individual NSAIDs, were too sparse to allow for subgroup analysis. One sensitivity analysis only provided a hint of information on an increased risk of reoperation (but not of bleeding per se) when NSAIDs were given after surgery compared with preoperative administration. The biological basis for this differential effect remains obscure. The lack of information on predictive factors for bleeding complications may have been due to the lack of relevant data or to poor reporting of relevant data. The surgical technique was not expected to be of major importance here because all these trials were strictly controlled. If a surgical technique had an impact on intra- or postoperative bleeding, then this would concern all patients, those who received a NSAID and those who did not, and thus, the RR would stay the same. Also, the reported surgical techniques represent widely used and accepted gold standard interventions for tonsillectomy.

Fourth, the presence of zero-event trials was a methodological problem. Zero-event trials occur when the event is very rare or when the trials are of limited size. Indeed, perioperative bleeding complications are relatively rare, and these trials were small; group sizes rarely exceeded 40 patients. There is no general agreement on how to deal with zero cells in meta-analysis. In zero-event trials, it cannot be excluded if a particular event actually happened but was not reported. It may also be assumed that these trials were unable to identify an (rare) event (i.e., the trials were too small); there is then an argument to regard such trials as invalid essays and, therefore, to exclude them from a combined analysis. To overcome this methodological problem, we applied two statistical models to calculate the combined ORs for rare (bleeding) events. The Peto-method ignores trials with zero events (8). The disadvantage of this method is that potentially relevant data are censored. As a result, combined event rates increase artificially because the denominator decreases; both the underlying risk and the risk with drug exposure may then be overestimated. The advantage of the method described by Shadish and Haddock (9) is that all trials can be included in the analysis independent of whether they report or not on a particular end point. The disadvantage of that method is that the analysis may be flawed if the sum of the manipulated zero cells outweighs the number of true events. Finally, we calculated the NNT based on combined event rates that were weighted by trial size and not manipulated. The results of all analyses using these models were surprisingly consistent.

Finally, it is important to note that these meta-analyses had low power. For example, event rates of admission because of bleeding were very small (2.6% without and 5.1% with NSAIDs), and calculations were based on 506 patients. A post hoc power analysis using these numbers and a significance level of 5% reveals a statistical power of approximately 50% only. Thus, it cannot be excluded that some of the nonsignificant findings in the meta-analysis were caused by Type II errors.

Consequently, the agenda is one of further research rather than of clinical recommendations. The recently introduced cyclo-oxygenase type 2 (COX-2) inhibitors may have advantages compared with classic NSAIDs. They seem to be effective, safe, and well tolerated for treating postoperative pain (79), and their COX-2 selectivity produces minimal effect on platelet aggregation. Two studies have found that COX-2 inhibitors have analgesic efficacy equivalent to that of NSAIDs for dental pain (80,81). In contrast, a study of tonsillectomy in children has found that the addition of ibuprofen to acetaminophen significantly reduced the need for early analgesia, whereas the addition of a COX-2 inhibitor to acetaminophen did not (50). There were no differences between groups in perioperative bleeding episodes. Further studies are required to carefully evaluate the risk-benefit ratio of COX-2 inhibitors in tonsillectomy. However, if the risk of, for example, reoperation because of bleeding with classic NSAIDs was approximately 2.7% and with COX-2 inhibitors was as with placebo (1.1%), then 2400 patients would need to be randomized to ensure a statistically significant result in favor of the COX-2 antagonist. Finally, it remains unknown if these data from patients undergoing tonsillectomy can be extrapolated to other surgical settings. Further trials are required to evaluate the potential additional bleeding risk with NSAIDs for different surgeries. In future studies, the recording of the incidence of perioperative bleeding requires standard definitions.

In conclusion, the evidence for NSAIDs to increase the incidence of bleeding after tonsillectomy remains ambiguous. The cautious clinical message must be that there is some evidence from randomized, controlled trials that NSAIDs may increase the likelihood of reoperation because of bleeding, particularly when NSAIDs are given in the postoperative period. However, there is a lack of evidence for NSAIDs to increase intraoperative blood loss or to increase the incidence of postoperative bleeding or readmission because of bleeding. Likewise, there is a lack of evidence that the regimen per se (single- versus multiple-dose) has any impact on bleeding and that there is any difference in the risk profile between different NSAIDs or between adults and children. Compared with opioids, NSAIDs seem to be equianalgesic for this type of surgery, and they decrease the risk of PONV to some extent. Whether these data can be extrapolated to other surgical settings needs to be shown. Overall, these results may indicate that NSAIDs should be used cautiously in tonsillectomy until further data are available.

Back to Top | Article Outline

References

1. McQuay HJ. Acute pain. In: Tramèr MR, ed. Evidence-based resource in anaesthesia and analgesia. London: BMJ Books, 2000: 87–106.
2. Strom BL, Berlin JA, Kinman JL, et al. Parenteral ketorolac and risk of gastrointestinal and operative site bleeding: a postmarketing surveillance study. JAMA 1996; 275: 376–82.
3. Forrest JB, Camu F, Greer IA, et al. Ketorolac, diclofenac, and ketoprofen are equally safe for pain relief after major surgery. Br J Anaesth 2002; 88: 227–33.
4. Handler SD, Miller L, Richmond KH, Baranak CC. Post-tonsillectomy hemorrhage: incidence, prevention and management. Laryngoscope 1986; 96: 1243–7.
5. Liu JH, Anderson KE, Willging JP, et al. Posttonsillectomy hemorrhage: what is it and what should be recorded? Arch Otolaryngol Head Neck Surg 2001; 127: 1271–5.
6. Moher D, Cook DJ, Eastwood S, et al. Improving the quality of reports of meta-analyses of randomised controlled trials: the QUOROM statement. Lancet 1999; 354: 1896–1900.
7. Jadad AR, Moore RA, Carroll D, et al. Assessing the quality of reports of randomized clinical trials: is blinding necessary? Control Clin Trials 1996; 17: 1–12.
8. Yusuf S, Peto R, Lewis J, et al. Beta blockade during and after myocardial infarction: an overview of the randomized trials. Prog Cardiovasc Dis 1985; 27: 335–71.
9. Shadish WR, Haddock CK. Combining estimates of effect size. In: Cooper H, Hedges LV, eds. The handbook of research synthesis. 1st ed. New York: The Russel Saga Foundation, 1994: 261–80.
10. Laupacis A, Sackett DL, Roberts RS. An assessment of clinically useful measures of the consequences of treatment. N Engl J Med 1988; 318: 1728–33.
11. Gavaghan DJ, Moore RA, McQuay JH. An evaluation of homogeneity tests in meta-analyses in pain using simulation of individual patient data. Pain 2000; 85: 415–24.
12. L’Abbé KA, Detsky AS, O’Rourke K. Meta-analysis in clinical research. Ann Intern Med 1987; 107: 224–33.
13. Tawalbeh MI, Nawasreh OO, Husban AM. Comparative study of diclofenac sodium and paracetamol for treatment of pain after adenotonsillectomy in children. Saudi Med J 2001; 22: 121–3.
14. Rømsing J, Østergaard D, Drozdziewicz D, et al. Diclofenac or acetaminophen for analgesia in paediatric tonsillectomy outpatients. Acta Anaesthesiol Scand 2000; 44: 291–5.
15. Virtaniemi J, Kokki H, Nikanne E, Aho M. Ketoprofen and fentanyl for pain after uvulopalatopharyngoplasty and tonsillectomy. Laryngoscope 1999; 109: 1950–4.
16. Nakayama K. Analgesic effect of voltaren suppository following tonsillectomy in children. Pract Otol 1982; 75: 1455–60.
17. Balli R. Results and observations of the action of the combination of indomethacin and d-propoxyphene in tonsillectomized subjects. Otorinolaringol Ital 1973; 39: 171–84.
18. Mansur ES, de Mesquita CS. Use of naproxen sodium in the postoperative period after tonsillectomy. Folha Med 1980; 81: 487–9.
19. Stamm AC, Menon AD. The use of piroxicam suppository in the postoperative period of patients submitted to tonsillectomy. Revista Brasileira de Clinica e Therapeutica 1985; 14: 40–3.
20. Knudsen KE, Brofeldt S, Mikkelsen S, et al. Peritonsillar infiltration with low-dose tenoxicam after tonsillectomy. Br J Anaesth 1995; 75: 286–8.
21. Valijan A. Pain relief after tonsillectomy: effect of benzydamine hydrochloride spray on postoperative pain relief after tonsillectomy. Anaesthesia 1989; 44: 990–1.
22. Young JR. A comparative study of benzydamine hydrochloride (“Difflam” pump spray) and placebo as analgesics following tonsillectomy. Int J Tissue React 1987; 9: 131–3.
23. Raj TB, Wickham MH. The effect of benzydamine hydrochloride (difflam) spray on post-tonsillectomy symptoms: a double-blind study. J Laryngol Otol 1986; 100: 303–6.
24. Young JR. A comparative study of benzydamine hydrochloride 0.15% w. v. (”Difflam“ pump spray) and placebo as analgesics following tonsillectomy. J Int Med Res 1985; 13: 245–7.
25. Irwin BC, Acharya KB. A comparative study of benzydamine hydrochloride 0.15% w/v (”Difflam“ Oral Rinse) and acetyl salicylic acid as analgesics following tonsillectomy. J Int Med Res 1984; 12: 216–8.
26. Eigel T. Analgesic and antiphlogistic effect of a benzydamine gargle solution in a double-blind experiment. Med Welt 1981; 32: 746–8.
27. Wetterwald Value of primary anti-inflammatory agents, benzydamine, in the sequelae of cold loop tonsillectomy. Lille Med 1969; 14: 681–2.
28. Togoni L. Analgesic and anti-inflammatory action of benzydamine in otorhinolaryngo-logical surgery. Valsalva 1968; 44: 46–53.
29. Kristensen S, Tveteraas K, Hein P, et al. Relief of pain and trismus in patients treated with naproxen or acetylsalicylic acid after tonsillectomy. J Laryngol Otol 1988; 102: 39–42.
30. Kristensen S, Tveteraas K, Outzen KE, Poulsen HB. Treatment of pain after tonsillectomy: comparison between naproxen (Naprosyn) and acetylsalicylic acid. Ugeskr Læger 1986;148:2832–5.
31. Azuma F, Itoh J, Takeda T, Matsuoka I. A double-blind clinical trial of diclofenac sodium suppositories for pain and inflammation after tonsillectomy. Jibiinkoka-Rinsho 1979; 72: 63–75.
32. Metternich FU, Brusis T, Parandeh-Shab F. Post-tonsillectomy treatment of pain in adults. HNO 1998; 46: 50–5.
33. Thorneman G, Åkervall J. Pain treatment after tonsillectomy: advantages of analgesics regularly given compared with analgesics on demand. Acta Otolaryngol 2000; 120: 986–9.
34. Engelhardt T, Crawford M. Sublingual morphine may be a suitable alternative for pain control in children in the postoperative period. Paediatr Anaesth 2001; 11: 81–3.
35. Mendham JE, Mather SJ. Comparison of diclofenac and tenoxicam for postoperative analgesia with and without fentanyl in children undergoing adenotonsillectomy or tonsillectomy. Paediatr Anaesth 1996; 6: 467–73.
36. Nordbladh I, Ohlander B, Björkman R. Analgesia in tonsillectomy: a double-blind study on pre and post-operative treatment with diclofenac. Clin Otolaryngol 1991; 16: 554–8.
37. Swanepoel A, Semple P. Oral versus rectal diclofenac for postoperative tonsillectomy pain in children. Anaesthesia 1999; 54: 298–9.
38. D’Apuzzo V, Zambelli D. Clinical experience with fentiazac suspension in inflammatory process of childhood. Curr Ther Res 1983; 34: 818–24.
39. Ejnell H, Björkman R, Wåhlander L, Hedner J. Treatment of postoperative pain with diclofenac in uvulopalatopharyngoplasty. Br J Anaesth 1992; 68: 76–80.
40. Watcha MF, Jones MB, Lagueruela RG, et al. Comparison of ketorolac and morphine as adjuvants during pediatric surgery. Anesthesiology 1992; 76: 368–72.
41. Angot A. Action of oropivalone-bacitracin compared to a NSAID (niflumic acid) in the after-care of adult amygdalectomy. J Fr Oto-Rhino-Laryngol 1987; 36: 497–500.
42. Pradillo JA, Takane J, Méndez EE. Effects of sodium ketoprofen on the postoperative evolution of patients with tonsillectomy. Invest Med Int 1984; 11: 152–6.
43. Tanner K, Novoselac M, Eggenschwiler E. The analgesic efficacy of suprofen in hospitalized patients after tonsillectomy: a double blind study after rectal application. Schweiz Rundsch Med Prax 1985; 74: 600–2.
44. Fassolt A. The control of pain after tonsillectomy with mefenamic acid. Schweiz Rundsch Med Prax 1974; 63: 1040–3.
45. Francois M, Consten L, Rezvani Y, et al. The efficacy and tolerance of tiaprofenic acid vs paracetamol after tonsillectomy in children. J Fr Oto-Rhino-Laryngol 1994; 43: 437–40.
46. Lacomme Y, Pessey J-J. Effects of a niflumic acid derivative in tonsil surgery in children. Rev Laryngol Otol Rhinol 1978; 99: 159–63.
47. Metta H, Haidar N, Mendoza JJ, et al. Lysine clonixinate vs ibuprofen in otorhinolaryngology. Prensa Med Argent 2000; 87: 906–18.
48. Azuma F. The clinical efficacy and plasma concentration of a 25 mg suppository of diclofenac sodium in children with post-tonsillectomy pain and inflammation. Pract Otol 1982; 75: 1445–53.
49. Tamura M, Yamamura M, Yoshioka A, et al. A double-blind study of a new non-steroidal anti-inflammatory analgesic GP 45 840 on the inflammatory symptoms in post-tonsillectomized patients. Jibiinkoka-Rinsho 1972; 65: 583–9.
50. Pickering AE, Bridge HS, Nolan J, Stoddart PA. Double-blind, placebo-controlled analgesic study of ibuprofen or rofecoxib in combination with paracetamol for tonsillectomy in children. Br J Anaesth 2002; 88: 72–7.
51. Bailey R, Sinha C, Burgess LPA. Ketorolac tromethamine and hemorrhage in tonsillectomy: a prospective, randomized, double-blind study. Laryngoscope 1997; 107: 166–9.
52. Courtney MJ, Cabraal D. Tramadol vs diclofenac for posttonsillectomy analgesia. Arch Otolaryngol Head Neck Surg 2001; 127: 365–88.
53. Dommerby H, Rasmussen OR. Diclofenac (Voltaren®): pain-relieving effect after tonsillectomy. Acta Otolaryngol 1984; 98: 185–92.
54. Gunter JB, Varughese AM, Harrington JF, et al. Recovery and complications after tonsillectomy in children: a comparison of ketorolac and morphine. Anesth Analg 1995; 81: 1136–41.
55. Harley EH, Dattolo RA. Ibuprofen for tonsillectomy pain in children: efficacy and complications. Otolaryngol Head Neck Surg 1998; 119: 492–6.
56. Kotecha B, O’Leary G, Bradburn J, et al. Pain relief after tonsillectomy in adults: intramuscular diclofenac and papaveretum compared. Clin Otolaryngol 1991; 16: 345–9.
57. Parker DA, Gibbin KP, Noyelle RM. Syrup formulations for post-tonsillectomy analgesia: a double-blind study comparing ibuprofen, aspirin and placebo. J Laryngol Otol 1986; 100: 1055–60.
58. Pasquale G, Scaricabarozzi I, D’Agostino R, et al. An assessment of the efficacy and tolerability of nimesulide vs paracetamol in children after adenotonsillectomy. Drugs 1993; 46: S234–7.
59. Petruson B. The analgesic effect of diclofenac (Voltaren®) after tonsillectomy. OPMEAR 1991; 36: 65–7.
60. Rorarius MGF, Baer GA, Siirtola M, et al. Effect of intravenous diclofenac or indomethacin on the emergence from anaesthesia for tonsillectomy. Acta Anaesthesiol Scand 1993; 37: 616–21.
61. Rusy LM, Houck CS, Sullivan LJ, et al. A double-blind evaluation of ketorolac tromethamine versus acetaminophen in pediatric tonsillectomy: analgesia and bleeding. Anesth Analg 1995; 80: 226–9.
62. Rømsing J, Østergaard D, Walther-Larsen S, Valentin N. Analgesic efficacy and safety of preoperative versus postoperative ketorolac in paediatric tonsillectomy. Acta Anaesthesiol Scand 1998; 42: 770–5.
63. Saarnivaara L, Metsä-Ketelä T, Männistö P, Vapaatalo H. Pain relief and sputum prostaglandins in adults treated with pethidine, tilidine and indomethacin after tonsillectomy: a double-blind study. Acta Anaesthesiol Scand 1980; 24: 79–85.
64. Salonen A, Kokki H, Tuovinen K. I.v. ketoprofen for analgesia after tonsillectomy: comparison of pre- and post-operative administration. Br J Anaesth 2001; 86: 377–81.
65. Schmidt A, Björkman S, Åkerson J. Preoperative rectal diclofenac versus paracetamol for tonsillectomy: effects on pain and blood loss. Acta Anaesthesiol Scand 2001; 45: 48–52.
66. Splinter WM, Rhine EJ, Roberts DW, et al. Preoperative ketorolac increases bleeding after tonsillectomy in children. Can J Anaesth 1996; 43: 560–3.
67. St Charles CS, Matt BH, Hamilton MM, Katz BP. A comparison of ibuprofen versus acetaminophen with codeine in the young tonsillectomy patient. Otolaryngol Head Neck Surg 1997; 117: 76–82.
68. Sutherland CJ, Montgomery JE, Kestin IG. A comparison of intramuscular tenoxicam with intramuscular morphine for pain relief following tonsillectomy in children. Paediatr Anaesth 1998; 8: 321–4.
69. Sutters KA, Levine JD, Dibble S, et al. Analgesic efficacy and safety of single-dose intramuscular ketorolac for postoperative pain management in children following tonsillectomy. Pain 1995; 61: 145–53.
70. Tarkkila P, Saarnivaara L. Ketoprofen, diclofenac or ketorolac for pain after tonsillectomy in adults? Br J Anaesth 1999; 82: 56–60.
71. Thiagarajan J, Bates S, Hitchcock M, Morgan-Hughes J. Blood loss following tonsillectomy in children: a blind comparison of diclofenac and papaveretum. Anaesthesia 1993; 48: 132–5.
72. Watters CH, Patterson CC, Mathews HML, Campbell W. Diclofenac sodium for post-tonsillectomy pain in children. Anaesthesia 1988; 43: 641–3.
73. Mather SJ, Peutrell JM. Postoperative morphine requirements, nausea and vomiting following anaesthesia for tonsillectomy: comparison of intravenous morphine and non-opioid analgesic techniques. Paediatr Anaesth 1995; 5: 185–8.
74. De Lima Pontes PA, Gregório LC. Double-blind clinical trial: naproxen sodium suspension vs placebo in postoperative adeno-tonsillectomy in children. Rev Bras Med 1985; 42: 57–61.
75. Bone ME, Fell D. A comparison of rectal diclofenac with intramuscular papaveretum or placebo for pain relief following tonsillectomy. Anaesthesia 1988; 43: 277–80.
76. Hellier WPL, Knight J, Hern J, Waddell T. Day case paediatric tonsillectomy: a review of three years experience in a dedicated day case unit. Clin Otolaryngol 1999; 24: 208–12.
77. Schroeder WA. Post tonsillectomy hemorrhage: a ten-year retrospective study. Mo Med 1995; 92: 592–5.
78. Tramèr MR. A rational approach to the control of postoperative nausea and vomiting: evidence from systematic reviews. I. Efficacy and harm of antiemetic interventions, and methodological issues. Acta Anaesthesiol Scand 2001; 45: 4–13.
79. Desjardins PJ, Grossman EH, Kuss ME, et al. The injectable cyclooxygenase-2-specific inhibitor parecoxib sodium has analgesic efficacy when administered preoperatively. Anesth Analg 2001; 93: 721–7.
80. Morrison BW, Christensen S, Yuan W, et al. Analgesic efficacy of the cyclooxygenase-2-specific inhibitor rofecoxib in post-dental surgery pain: a randomized, controlled trial. Clin Ther 1999; 21: 943–53.
81. Ehrich EW, Dallob A, De Lepeleire I, et al. Characterization of rofecoxib as a cyclooxygenase-2 isoform inhibitor and demonstration of analgesia in the dental pain model. Clin Pharmacol Ther 1999; 65: 336–47.
© 2003 International Anesthesia Research Society