To the Editor:—
Marret et al. provided a meta-analysis of nonsteroidal antiinflammatory drugs (NSAIDs) and bleeding risk after tonsillectomy using return to the operating room as their primary endpoint. 1The authors carefully reviewed the literature and selected seven articles for inclusion. 2–8We are concerned that the conclusions reached are invalid and that the wording in the current Marret et al. article 1provides fodder for lawyers without adequate scientific knowledge to support these admonitions.
The authors have combined repeated single-dose intramuscular administration, continuous infusion, and chronic administration of NSAIDs. Unlike aspirin, the effects of most NSAIDs on platelet function are very short-lived. For pediatric patients, the half-lives of ketorolac and ibuprofen are approximately 2 h and 1.5 h, respectively. 9,10Because significant effects on platelet function are reversible and related in part to a blood concentration of NSAIDs, virtually no antiplatelet effect would be present after 5 to 6 drug half-lives. Therefore, any bleeding attributable to single doses of these medications would have to occur in the immediate postoperative period, and certainly within the first 24 h. 11,12Hemorrhage that occurs after 24 h is unlikely to be caused by these medications unless they are administered after hospital discharge; we do not know if this is different with infusions or intramuscular administration.
When examining the duration of treatment (table 1 of the Marret et al. review), studies by St. Charles et al. 7and Harley and Dattolo 8involved patients who received treatment for 2 weeks. Both patients who had bleeding in the Harley and Dattolo study did so 2 to 7 days after the operation, so it is possible that their bleeding could be attributed either to the continued use of the NSAIDs or to the loss of eschar at 5–10 days. The patient in the St. Charles et al. 7study returned to the operating room because of lack of cooperation.
Two studies (Salonen et al. 2and Bailey et al. 3) allowed administration of ketorolac for the first 24 h. In the Bailey et al. study (intramuscular × 24 h), there was a higher incidence of bleeding in the ketorolac group; however, on average, incidence of bleeding occurred on day 4, thus suggesting that bleeding had nothing to do with the administration of ketorolac. One adult patient in the Salonen et al. study (24 h infusion) experienced bleeding 14 h after surgery, which could be attributable to the NSAIDs.
The main studies of interest are the articles (Gunter et al. , 5Rømsing et al. , 4and Sutters et al. 6) involving single-dose administration after the end of the procedure. One patient in the Gunter et al. study returned to the operating room on postoperative day 5; that patient’s bleeding should not be attributed to the NSAIDs. In the Rømsing et al. study, 5 of the first 15 patients experienced bleeding (divided between placebo, presurgical, and postsurgical ketorolac administration), and all were operated on by the same surgeon. After eliminating that surgeon’s participation in the study, only 1 of the next 45 patients required reoperation for bleeding. The time of reoperation was not described.
When examining the timing of the bleeding that was significant enough to require reoperation (the primary endpoint of the Marret et al. analysis), it appears that nearly half of the cases presented should not attribute bleeding directly to NSAIDs. Rather, the bleeding occurred because of poor surgical technique or at a time when it was much more likely caused by dehiscence of the eschar (5–10 days after surgery). Thus, just as anesthesiologists talk about effect site half-lives for drugs such as remifentanil, this meta-analysis should have considered the effect-site antiplatelet effects of the NSAIDs before attributing the hemorrhage to the NSAID therapy.
Marret et al. suggest that the use of NSAID therapy should be abandoned both in the hospital and at home in these patients. There are insufficient data to support this strong recommendation. Only 3 of 71 patients receiving ibuprofen versus 0 of 66 controls experienced bleeding, and some of these could have been attributed to the loss of eschar (table 1); clearly, there are insufficient numbers for analysis. Regarding single-dose administration of ketorolac, in our experience, a single dose after the operation is completed and after hemostasis is obtained has not been associated with an increased incidence of bleeding. 13, 14Marret et al. mentioned our article to describe the general incidence of bleeding, but they did not comment on the fact that the incidence of bleeding in our retrospective review of more than 300 children having a tonsillectomy did not reveal a higher incidence of hemorrhage, even though they had received ketorolac. In the Marret et al. analysis, the incidence of bleeding within the first 24 h after a single dose of ketorolac was 2 of 109 patients receiving ketorolac versus 0 of 104 controls (table 1). Again, this cohort is too small for adequate analysis.
The risk of 1 in 29 patients having hemorrhage after NSAIDs therapy as described by Marret et al. is a gross overstatement. At least two bleeding events were attributed to surgical technique, whereas five others occurred at a time well after the drug had been eliminated from the body. Therefore, antiplatelet effect caused by that NSAID drug was no longer present. Using the data presented comparing 0 of 198 controls versus 4 of 257 patients receiving NSAIDs with bleeding less than 24 h after surgery, including one patient who returned to the operating room because of lack of cooperation, the rate of bleeding is statistically insignificant (P = 0.209 chi-square). A crude power analysis suggests that to demonstrate a difference with these frequencies, more than 4,000 patients would need to be studied (α= 0.05, with power of 0.8) to demonstrate a difference in treatment groups.
We suggest that Marret et al. reexamine their data, taking into consideration the effect of the drug’s half-life on platelet function. Elimination of cases in which the hemorrhage occurred beyond the period when the drug could have had antiplatelet effects, in which patients returned to the operating room because of poor surgical technique or lack of cooperation, and in which patients received the ibuprofen chronically (their bleeding could be attributable to either loss of eschar or antiplatelet effects), would allow the authors to make more accurate recommendations. We certainly acknowledge that ketorolac can cause hemorrhage if administered preoperatively or intraoperatively before hemostasis is completed, but we believe that it is a very safe analgesic when administered as a single dose after hemostasis is obtained.