We appreciate the three comments concerning our meta-analysis on the risk of nonsteroidal antiinflammatory drugs (NSAIDs) after tonsillectomy.
Drs. Dsida and Coté suggest excluding patients who had bleeding 24 h after tonsillectomy and who received NSAIDs for less than 24 h, or patients with bleeding induced by lack of cooperation and poor surgical technique. Under these criteria, more than half of the cases of reintervention published in the original articles would be excluded. 1–7Moreover, Disda and Coté cite one controlled, double-blind, randomized trial that studied NSAIDs given within 24 h postoperatively. 8This study reports two children in the postoperative ketoprofen group (n = 42) who required electrocautery to stop their bleeding. In contrast, none of children in the placebo group (n = 20) returned to the operating room. Including this study in the analysis proposed by Disda and Coté (table 1) leads to a statistically significant risk of primary hemorrhage (0 of 218 in the control group vs. 6 of 299 in the NSAIDs group; P = 0.04, Fisher exact test). We performed an intention-to-treat analysis based on the original data. 9Consequently, all patients included in the study were analyzed according to the groups to which they were originally randomly assigned, whatever the duration of treatment or the cooperation of the patient or surgeon. 1–7The intention-to-treat approach maintains a comparable distribution of established groups of patients that are similar apart from the randomized factor. 10Excluding randomized patients after their inclusion in the trial may lead to bias due to a modification of allocation. All of the trials included in our meta-analysis were randomized. Therefore, in our meta-analysis the distribution of patients was similar between the different groups except for the use of NSAIDs. When Drs. Dsida and Coté claim that NSAIDs have no effect, it means that the risk of bleeding should be similar whatever the NSAID dose, duration of treatment, or name of the surgeon. Moreover, they cannot deny that short-term administration of NSAIDs may have long-term consequences. NSAIDs inhibit cyclooxygenase, resulting in an inhibition of platelet aggregation. Thus, NSAIDs given immediately in the postoperative period may increase the importance of hematoma induced by the tonsil dissection. The eschar situated in the soft tissue near the tonsillar bed may be more important with the use of NSAIDs than with a control treatment and may explain secondary bleeding related to the loss of eschar of tonsillar fossa. The postoperative administration of NSAIDs, whatever the duration of treatment, may consequently increase the risk of early and late reoperation. Excluding patients who had later bleeding after a single dose of NSAIDs may underestimate the effect of NSAIDs on bleeding in the final analysis. In one of the studies, Dsida and Coté also suggested the exclusion patients who were operated on by the same surgeon and who were especially at risk for complications. 4However, the same surgeon also operated on other randomized patients. This seems to be an unacceptable handling of the data that may introduce some bias into the analysis. Excluding cases is contrary to the principle of intention-to-treat. 10,11The exclusion of patients in randomized controlled trials has been recognized as an overestimation of the benefit of treatment in comparison with the intention-to-treat analysis, based on the submitted reports. 12Similarly, the exclusion of patients who experienced adverse effects such as hemorrhage may underestimate the risk of NSAIDs use after a tonsillectomy.
Drs. Lake and Khater report a retrospective cohort of patients scheduled for tonsillectomy at Glan Clwyd Hospital between 1991 and 2002. Interestingly, their data document a clear increase in the incidence of reoperation after tonsillectomy related to the increased number of patient who used NSAIDs. When NSAIDs were administrated to less than half of the patients (from 1991 to 1998), the incidence of reoperation was less than 1%. In contrast, the incidence was greater than 1% during the past 4 yr, and as high as 2.9% in 2001 when NSAIDs were used in more than 50% of the patients. Lake and Khater also advocated that NSAIDs decrease the incidence of postoperative nausea and vomiting. Although this is a reasonable assertion, alternative treatments devoid of effect on hemostasis may be more appropriate in this setting. In fact, other treatments such as ondansetron 13or dexamethasone 14reduce postoperative nausea and vomiting and postoperative pain. 15Physicians also have cyclooxygenase-2 inhibitors to relieve postoperative pain and prevent postoperative nausea and vomiting, as suggested by Joshi et al. 16We agree that selective inhibitors of cyclooxygenase could be an interesting alternative because of their lack of effect on platelet aggregation. In contrast, anesthesiologists do not have such pharmacologic alternatives to stop bleeding after a tonsillectomy.
Dsida and Coté and Lake and Khater point out a randomized, double-blind, controlled trial that compared postoperative NSAIDs analgesia with preoperative NSAIDs or placebo, 8published in 2002 after the completion of our own research in 2001. 8In this trial, two children in the postoperative NSAIDs group returned to the operating room because of postoperative bleeding. We updated our meta-analysis to include the result of this last study. A significant increase in reoperation was still documented with the use of NSAIDs (odds ratio: 4.0; 95% confidence interval: 1.4, 11.3; P = 0.009; P = 0.98 for the heterogeneity test).
Finally, Drs. Lake and Khater concluded that urgent, randomized, double-blind trials are needed to clarify the risk of reoperation with NSAIDs use. In fact, there is now sufficient evidence to demonstrate an increased risk of bleeding after tonsillectomy with NSAIDs. Regarding this risk of reoperation, we consider that the use of NSAIDs should be abandoned.