Ong, Cliff K.-S. DDS*; Lirk, Philipp MD†; Seymour, Robin A. PhD‡; Jenkins, Brian J. MD§
The concept of preemptive analgesia to reduce postoperative pain was founded on a series of successful animal experimental studies that demonstrated central nervous system plasticity and sensitization after nociception (1). Preemptive analgesia is defined as an antinociceptive treatment that prevents the establishment of altered central processing of afferent input, which amplifies postoperative pain (2). By decreasing the altered central sensory processing, preemptive analgesia is thought to consequently decrease the incidence of hyperalgesia and allodynia after surgery (3). It is important to consider this definition in clinical trials for determining the effectiveness of preemptive analgesia. The emphasis of preemptive analgesia is on the pathophysiologic phenomenon that it should prevent: altered sensory processing. Therefore, preemptive may not simply mean “before incision.” An insufficient afferent blockade cannot be preemptive, even if it is administered before the incision.
Whether preemptive analgesic interventions are more effective than conventional regimens in managing acute postoperative pain remains controversial. Several reviews have addressed this question and have drawn fundamentally different conclusions. For example, some reviews have concluded that preemptive analgesia is effective as such (3,4), but some have concluded it to be effective only for certain analgesic drugs (5,6). Some analyses have attributed no beneficial effect to any drug (7), whereas some have postulated dependence on a range of factors (2,8,9), and some reviews have been unable to draw a final conclusion regarding efficacy (10–13). Therefore, it can be stated that whereas the evidence on preemptive analgesia in animal studies is very convincing (14), results from human clinical studies remain inconsistent.
Many of the summaries of results of clinical studies of the above-mentioned reviews typically took the form of narrative reviews, which may have been subjective. Therefore, the purpose of this meta-analysis was to synthesize the data statistically from previous randomized and double-blind controlled trials to determine whether preemptive application of analgesic regimens is of superior efficacy in the treatment of acute postoperative pain as compared with the same analgesic regimens initiated after surgical incision. This was based on the original observations in experimental studies suggesting that the timing of analgesic regimens was important to obtain a reduction of postinjury pain hypersensitivity (1). In the event that the clinical studies yield comparable effects, we will then be able to report this effect more confidently.
This meta-analysis used the methods proposed by the Cochrane Collaboration (15). Full published reports of randomized controlled trials (RCTs) on preemptive analgesia for postoperative pain were sought in MEDLINE, EMBASE, CINAHL, and PubMed covering January 1987 to October 2003. A broad free-text search with restriction to publications in English was undertaken with all variants of terms. “Preemptive analgesia,” “postoperative pain,” “preoperative,” and “preincisional” were entered as major subject headings, and “randomized controlled trial” was entered as a publication type selected from the dictionaries menu. Reference lists of retrieved reports and reviews were searched for additional trials. Unpublished reports and abstracts were not considered. Authors were not contacted for original data.
The retrieved reports were stratified according to the type of analgesics or interventions (local anesthetic wound infiltration, N-methyl-d-aspartic acid [NMDA] receptor antagonists, nonsteroidal antiinflammatory drugs [NSAIDs], and opioids), mode of administration (systemic, epidural, or wound infiltration), and surgical procedure.
Articles that met the following criteria were included in this meta-analysis: 1) comparison of the same analgesic intervention before and after surgical incision by the same route and 2) randomized and double-blind study design. Exclusion criteria were 1) comparison of preoperative treatment with placebo or no treatment, 2) comparison of preoperative treatment with a combination of preoperative plus postoperative treatment, and 3) comparison of different preoperative and postoperative treatment regimens.
Where possible, the following outcome measures data were extracted from the retrieved reports in the form of mean/median data plus dispersion values or dichotomous data:
1. Pain intensity in the form of the various pain scores, e.g., visual analog scale (VAS) scores during the first 24–48 postoperative hours.
2. Supplemental postoperative analgesic requirements.
3. Time to first rescue analgesic.
In cases in which trials reported outcome as graphs, the means and standard deviations were estimated from these graphs.
The quality of the included studies was assessed regarding the extent to which the RCT design, data collection, and statistical analysis minimized or avoided bias in treatment comparisons. A modification of a validated scale (16) was used to perform the quality assessment. This scale includes five items pertaining to description of randomization, appropriate blinding, dropouts and withdrawals, and other pain outcome measures. In brief, the following rules were applied to assess study quality:
1. Randomization—if the reports were described as randomized, one point was given. An additional point was given if the method of randomization was described and adequate, e.g., computer generated or a table of random numbers. However, one point was deducted if the method of randomization was inappropriate, e.g., randomization according to age or birthday.
2. Blinding—if the reports were described as double-blind, one point was given. An additional point was given if the blinding was described and appropriate, e.g., use of double-dummy. One point was deducted if blinding was inappropriate.
3. Patients’ withdrawals—if the reports described the numbers and reasons for withdrawals, one point was given.
4. Pain intensity—to ensure that a clinically relevant effect could be detected, one point was given if the pain scores were ≥30 mm on a VAS or more than moderate on a verbal rating scale (VRS).
5. Sample size—studies that performed a power calculation to estimate the sample size required to detect the treatment difference were awarded one point. Furthermore, RCTs with a sample size of ≤10 were not considered in this meta-analysis (17).
According to these assessments, the minimum score of an included trial was 2, and the maximum was 7.
The statistical software Comprehensive Meta Analysis™ (Biosta Inc., NJ) was used for the synthesis of data from all the included studies. The meta-analysis consisted of a two-stage process. In the first stage, a summary statistic was calculated for each study. These values describe the treatment effects observed in each individual study. In the second stage, a summary (pooled) treatment effect estimate was calculated as a weighted average of the treatment effects estimated in the individual studies. A weighted average is defined as follows:
where Ti is the treatment effect estimated in study i, Wi is the weight given to study i, and the summation is across all studies. In this meta-analysis, we were interested in the difference between the pain outcome variables of the patients between the two treatment groups (pretreatment and posttreatment). To perform meta-analyses of such data, the mean difference in the outcome variable between groups for each study was converted to an effect size (ES) by entering the mean values, the standard deviations, and the number of participants on whom the outcomes were assessed in each of the two groups into the software (18). The three outcome variables measured a single theoretical construct (pain experience). First, the ES for each outcome measure for all the studies for each intervention was combined, and a separate analysis on each outcome measure was performed. Second, the ES values of the three outcome measures for each intervention were combined mathematically into one. In addition, the significance tests for each included trial were combined.
Equation (Uncited)Image Tools
The outcome variables used for combining results were the pain intensity scores during the first 24–48 h, total supplemental postoperative analgesic consumption, and time to first analgesic. Because the outcome measures data were entered as mean, standard deviation, and P value of the difference between treatment groups, the effect size index (ES) for this meta-analysis is the standardized mean difference (SMD). The ES is expressed in standardized units. A positive ES indicates that preemptive analgesia is effective, and a negative value indicates that preemptive analgesia is ineffective. Therefore, a point estimate of 0 indicates no effect, values more than 0 reflect a better outcome for the pretreated group, and values <0 reflect a better outcome of the posttreated group. If the point estimate and confidence interval (CI) were more than 0, the study would meet the criterion for statistical significance (α was set at 0.05). If the CI overlapped 0, the P value would exceed 0.05, and the study would not be statistically significant. Results of our meta-analyses are graphically displayed in Forrest plots (19).
Each of the three outcome variables was analyzed separately. In the case of preemptive analgesia being effective for that outcome variable as evidenced by the overall point estimate, it was decided that preemptive treatment should reduce that pain outcome measure by at least 10% of points to be clinically useful. These “percentage points” are units relative to the standard deviations of the outcome measure. This tells us whether the intervention is more effective for the outcome. Forrest plots were plotted for each outcome measure and for each different type of analgesic intervention.
There will always be confounding covariates (study population, age, sex, type of surgery, quality of study, and publication year) and manipulation, and these may affect the analysis. Fixed-effects models were used throughout, unless statistical heterogeneity was observed by the Cochran Q test (P < 0.05). When heterogeneity was significant, random-effects models were used. Potential publication bias was not assessed with funnel plots because these tests have been shown to be unhelpful (20). A Mann-Whitney U-test was also used to assess the relationships between positive and negative trials and quality scores.
The literature search identified 102 RCTs of preemptive analgesia for acute postoperative pain. Of these, 36 trials were excluded because they did not meet the inclusion criteria; these are summarized in Appendix 1 (21–56).
Subsequently, 66 RCTs, with a total of 3261 patients, were included in this meta-analysis (Tables 1–5). These 66 studies were stratified into 5 groups according to surveyed intervention as follows: 1) 19 trials of epidural analgesia (57–75), 2) 15 trials of peripheral local anesthetic infiltrations (76–90), 3) 7 trials of systemic NMDA receptor antagonists (91–97), 4) 17 trials of systemic NSAIDs (98–114), and 5) 8 trials of systemic opioids (115–122). Not all the 66 trials used all 3 outcome measures (pain intensity, supplemental analgesic, and time to first analgesic) that were needed for the meta-analysis. Some used only one or two of the outcome measures. A total of 50 trials used pain intensity, 44 trials used supplemental analgesic, and 28 trials used time to first analgesic as 1 of the outcome measures.
The median quality score for the studies that favored pretreatment was 4 (range, 2–7) and was 4 (range, 2–7) for studies that favored posttreatment. The percentage of trials that favored pretreatment did not differ from that which favored posttreatment (Mann-Whitney U-test; P = 0.7). There was also no significant difference between the higher-quality (range, 5–7) trials and lower-quality (range, 2–4) trials (P = 0.44).
The data from the RCTs of each of the five stratified analgesic intervention groups were synthesized separately to obtain a combined value that reflects the ES for each of the pain outcome measures (Figs. 1–3). Finally, data from all three outcome measures were combined for each analgesic intervention to give a final ES (Fig. 4). Below are the details for the analyses.
Pain Intensity Outcome Measure
In the included studies, various different types of pain scales were used for preemptive studies, including the VAS, VRS, numerical rating scale, and objective pain scale (in the case of young children). Of the 50 trials reporting pain intensity as an outcome measure, there were 13 epidural, 11 local anesthetic, 7 NMDA antagonist, 12 NSAID, and 7 opioid studies. The results are summarized in the Forrest plot in Figure 1.
Epidural Analgesic RCTs.
Thirteen RCTs with a total of 653 patients comparing preincisional versus postincisional epidural analgesia were included (Table 1). Of the 13 RCTs, 7 were statistically significant favoring pretreatment, whereas the other 6 were reported as not significant. Figure 1 shows that the 13 studies had a combined ES of +0.25, with a CI of +0.10 to +0.41. The combined P value is 0.002.
Local Anesthetic RCTs.
Eleven RCTs with a total of 535 patients comparing preincisional versus postincisional peripheral local anesthetic wound infiltration were included (Table 2). Of the 11 RCTs, 5 were statistically significant favoring pretreatment, whereas the other 6 were reported as not significant or significantly favored posttreatment. Figure 1 shows that the 11 studies have a combined ES of +0.10, with a CI of −0.07 to +0.27. The combined P value is 0.26.
NMDA Antagonist RCTs.
Seven RCTs with a total of 418 patients comparing preincisional versus postincisional systemic NMDA antagonists were included (Table 3). Of the seven RCTs, only two were statistically significant, whereas the other five were reported as not significant. Figure 1 shows that the seven studies have a combined ES of 0.0, with a CI of −0.19 to +0.20. The combined P value is 0.97. The 0.0 point estimate indicates that preemptive NMDA antagonists have no effect on postoperative pain intensity scores.
Twelve RCTs with a total of 617 patients compared preincisional versus postincisional systemic NSAIDs (Table 4). Of the 12 RCTs, 6 were statistically significant favoring pretreatment, whereas the other 6 were reported as not significant or significantly favored posttreatment. Figure 1 shows that the 12 studies have a combined ES of +0.14, with a CI of −0.02 to +0.30. The combined P value is 0.09.
Seven RCTs with a total of 324 patients comparing preincisional versus postincisional systemic opioids were included (Table 5). Of the seven RCTs, only one was statistically significant favoring pretreatment, whereas the other six had been reported as not significant or significantly favored posttreatment. Figure 1 shows that the 7 studies have a combined ES of −0.24, with a CI of −0.01 to -0.41. The combined P value is 0.39.
Supplemental Analgesic Outcome Measure
The supplemental analgesics taken included patient-controlled analgesia; IM or IV morphine, meperidine, and fentanyl; and oral paracetamol or NSAIDs. Of the 66 included trials, 44 trials used postoperative supplemental analgesic consumption as an outcome measure for assessing the efficacy of the treatment groups. There were 13 epidural, 8 local anesthetic, 7 NMDA antagonist, 12 NSAID, and 4 opioid trials. The results are summarized in the Forrest plot in Figure 2.
Epidural Analgesic RCTs.
Thirteen RCTs with a total of 640 patients comparing preincisional versus postincisional epidural analgesia were included (Table 1). Of the 13 RCTs, 10 were statistically significant favoring pretreatment, whereas the other 3 were reported as not significant. The 13 studies have a combined ES of +0.58, with a CI of +0.42 to +0.74. The schematic in Figure 2 shows that the majority of the studies are within the pretreatment effective range, and the CI for all 14 studies includes the combined ES of +0.58. The combined P value is <10−8, a highly significant difference favoring pretreatment.
Local Anesthetic RCTs.
Eight RCTs with a total of 360 patients comparing preincisional versus postincisional peripheral local anesthetic wound infiltration were included (Table 2). Of the eight RCTs, five were statistically significant favoring pretreatment, whereas the other three were reported as not significant. The eight studies have a combined ES of +0.44, with a CI of +0.23 to +0.65. The combined P value is 0.00006.
NMDA Antagonist RCTs.
Seven RCTs with a total of 418 patients comparing preincisional versus postincisional systemic NMDA antagonists were included (Table 3). Of the seven RCTs, three were statistically significant favoring pretreatment, whereas the other four were reported as not significant or significantly favored posttreatment. The seven studies had a combined ES of +0.17, with a CI of −0.24 to +0.37. The combined P value is 0.09.
Twelve RCTs with a total of 582 patients comparing preincisional versus postincisional systemic NSAIDs were included (Table 4). Of the 12 RCTs, 8 were statistically significant favoring pretreatment, whereas the other 4 were reported as not significant. The 12 studies have a combined ES of +0.48, with a CI of +0.31 to +0.65. The combined P value is 0.00000003, a highly significant difference favoring pretreatment.
Four RCTs with a total of 194 patients comparing preincisional versus postincisional systemic opioids were included (Table 6). Two of the RCTs were reported as statistically significant favoring pretreatment, and the other two were reported as not significant. The four studies have a combined ES of +0.23, with a CI of −0.06 to +0.52. The combined P value is 0.12.
Time to First Analgesic Outcome Measure
A total of 28 trials used time to first analgesic as an outcome measure for assessing the efficacy of treatment. There were nine epidural, seven local anesthetic, four NMDA antagonist, six NSAID, and two opioid studies. Time to first analgesic was defined as the time from the end of surgery to the first rescue analgesic request. The results are summarized in the Forrest plot in Figure 3.
Epidural Analgesic RCTs.
Nine RCTs with a total of 368 patients comparing preincisional versus postincisional epidural analgesia were included (Table 1). Of the nine RCTs, five were statistically significant favoring pretreatment, whereas the other four were reported as not significant or significantly favored posttreatment. The nine studies have a combined ES of +0.31, with a CI of +0.10 to +0.52. The combined P value is 0.004.
Local Anesthetic RCTs.
Seven RCTs with a total of 306 patients comparing preincisional versus postincisional peripheral local anesthetic wound infiltration were included (Table 2). Of the seven RCTs, five were statistically significant favoring pretreatment, whereas the other two were reported as not significant. The seven studies have a combined ES of +0.44, with a CI of +0.21 to +0.68. The combined P value is 0.0002.
NMDA Antagonist RCTs.
Four RCTs with a total of 258 patients comparing preincisional versus postincisional systemic NMDA antagonists were included (Table 3). Of the four RCTs, two were statistically significant favoring pretreatment, whereas the other two were reported as not significant. The four studies have a combined ES of +0.12, with a CI of −0.13 to +0.37. The combined P value is 0.34.
Six RCTs with a total of 307 patients comparing preincisional versus postincisional systemic NSAIDs were included (Table 4). Of the six RCTs, five were statistically significant favoring pretreatment, whereas the other was reported as not significant. The six studies have a combined ES of +0.68, with a CI of +0.44 to +0.91. The combined P value is <10−8, a highly significant difference favoring pretreatment.
Two RCTs with a total of 74 patients comparing preincisional versus postincisional systemic opioids were included (Table 5). Both RCTs were reported as not significant. The two studies have a combined ES of −0.34, with a CI of −0.81 to +0.13. The combined P value is 0.16.
Combined Three Outcome Measures
The three outcomes were combined to achieve an ES for each group of analgesic interventions (Fig. 4).
A total of 37 combined outcome variables in 19 trials were analyzed. They had a combined ES of +0.38, with a CI of +0.28 to +0.47. The combined P value is <10−8, a highly significant difference favoring pretreatment.
Local Anesthetic RCTs.
A total of 26 combined outcome variables in 15 trials were analyzed. They had a combined ES of +0.29, with a CI of +0.17 to +0.40. The combined P value is 0.000001.
NMDA Antagonist RCTs.
A total of 16 combined outcome variables in 7 trials were analyzed. They had a combined ES of +0.09, with a CI of −0.03 to +0.22. The combined P value is 0.12.
A total of 30 combined outcome variables in 17 trials were analyzed. They had a combined ES of +0.39, with a CI of +0.27 to +0.48. The combined P value is <10−8, a highly significant difference favoring pretreatment.
A total of 13 combined outcome variables in 8 trials were analyzed. They had a combined ES of −0.10, with a CI of −0.26 to +0.07. The combined P value is 0.25.
This meta-analysis was conducted to assess the ability of preemptive analgesic interventions to attenuate postoperative pain scores, decrease supplemental postoperative analgesic requirements, and prolong the time to first rescue analgesia. The main result is that with these outcome measures, preemptive analgesia showed an overall beneficial effect in selected analgesic regimens that was most pronounced after epidural analgesia, local wound infiltrations, and systemic NSAID administration.
Pain intensity measures rated by the patient have been described as one of the most reliable estimates of treatment efficacy (123). In this meta-analysis, preemptive epidural analgesia could reduce postoperative pain intensity by approximately 25% points (ES, +0.25; 95% CI, 0.10–0.41). Therefore, it can be concluded that, according to the currently available data, preemptive epidural analgesia is effective and clinically useful in reducing postoperative pain intensity scores. In contrast, effects of preemptive local anesthesia and NSAID administration on postoperative pain intensity did not reach levels of statistical significance sufficient to draw a positive conclusion, even though a trend toward reduced postoperative pain scores was noted in the NSAID cohort. Synthesis of preemptive systemic opioid and NMDA antagonist analgesia yielded a negative and zero ES, respectively, suggesting that these preemptive treatments are not effective in reducing postoperative pain intensity scores.
Furthermore, it has been shown that total analgesic consumption is perhaps the most adequate outcome measure for showing a true preemptive effect (124). Preemptive epidural analgesia proved effective in reducing supplemental analgesic consumption, featuring a very large ES (ES, +0.58; 95% CI, 0.42–0.74), and the entire range of the CI exceeds our a priori criterion of at least 10%. Similarly, it could be concluded that preemptive local anesthetics (ES, +0.44; 95% CI, 0.23–0.65) and NSAIDs (ES, +0.48; 95% CI, 0.31–0.64) are clinically useful in reducing supplemental analgesic consumption. The preemptive administration of systemic NMDA antagonists and opioids was not shown to elicit a significant beneficial effect. Reducing the supplemental analgesia by approximately 44% to 58% by using preemptive analgesic techniques is clinically useful. Thus, there may be an economic facet that favors the incorporation of preemptive analgesia into the clinical routine. Although it has been stated that the difference in the requirement for postoperative supplemental analgesic consumption is not an important medical issue, provided that adequate doses are available when the need arises (125), in today’s climate of financial constraints on health care expenditure, it is important to consider the economic effect of adopting preemptive analgesia with the potential of reducing overall postoperative analgesic requirements. Forty-four percent to 58% less postoperative analgesic consumption for surgical patients could mean that a large sum of money could be saved through adopting the technique of preemptive analgesia, which essentially does not add any additional cost to the existing armamentarium of pain management other than changing the timing of administration. This is significant because this effect was observed not only for more costly and time-consuming interventions, such as epidural analgesia, but also for NSAIDs and local anesthesia.
Time to the first rescue analgesic request was used in many of the included trials as an outcome measure. In this meta-analysis, we could show that preemptive epidural analgesia was clinically useful in prolonging the time to first analgesic request (ES, 0.31; 95% CI, 0.10–0.52). Similarly, local anesthesia (ES, 0.44; 95% CI, 0.21–0.68) and NSAID (ES, 0.68; 95% CI, 0.44–0.91) administration are clinically useful in prolonging the time to first analgesic intake. In contrast, the administration of NMDA antagonists and systemic opioids did not yield effects consistent enough to draw conclusions regarding their clinical utility. Although it has been stated that the time to first analgesic intake is probably less accurate for assessing the preemptive effect and that a decreased time to first analgesic request is not a treatment problem provided that timely medication is available when the need arises (125), there is potential clinical usefulness for prolonging the time to the first analgesic request. Prolonging the time to first analgesic request means that the analgesic duration has outlasted the pharmacological duration of action of the drug. In theory, this is a technique of increasing the duration of analgesia without increasing the dosage or dosing frequency. Often, this would translate into less pain, less total analgesic consumption, and better patient comfort. As an example, ketorolac administered before surgery for impacted third-molar surgery has a mean analgesic duration of 8.9 hours, compared with 6.9 hours when given after surgery (98). This is clinically significant, because pain for this type of procedure is usually most severe between six to eight hours after the surgery (126).
Considering the efficacy of the individual interventions, it can be stated that preoperative epidural analgesic treatment is more effective in managing acute postoperative pain, attenuating pain scores, decreasing the total supplemental analgesic consumption, and prolonging the time to first rescue analgesic. Other interventions, such as local anesthesia and NSAID intake, failed to elicit significant effects in all outcome measures but did affect selected variables (Table 6). As for the latter interventions, the relatively large CI regarding ES on pain intensity suggests that additional information may be required to establish the role of these analgesics in reducing pain intensity scores. The least proof of efficacy was found in the case of systemic NMDA antagonist and opioid administration.
Our findings are therefore largely at odds with a recent systematic review (7) of preemptive analgesia, which did not find a beneficial effect on postoperative pain intensity scores. We believe that Moiniche et al. (7) have done a good job in reviewing the topic, and it is probably neither us nor Moiniche et al. who are wrong, because a number of studies have been published since 2000 that may have substantially changed the overall picture. This meta-analysis included 10 recent trials from 2001 to 2003 (58,59,76,79,91,98–102) that were not included in Moiniche et al.’s review, which included only studies published up to 2000 (7). Furthermore, the methods and data used were different from our present meta-analysis. In Moiniche et al.’s review, all the scores in the different pain scales were converted into a single VAS score, such that the scores could be combined. The weighted mean difference (WMD) was calculated by using this converted pain score for each group of analgesic interventions. In contrast, our meta-analysis used the SMD in lieu of WMD as the ES. This allowed us to combine and analyze the data without having to convert. WMD is a standard statistic that measures the absolute difference between mean values in two groups in a clinical trial. It estimates the amount by which the treatment changes the outcome on average. However, it is important to note that this method assumes that all of the trials have measured the outcome on the same scale. Conversely, SMD is used as a summary statistic in meta-analysis when the trials all assess the same outcome but measure it in a variety of ways. When an outcome (such as pain) is measured in a variety of ways across studies (by using different scales), it may not be possible to directly compare or combine study results in a meta-analysis. By expressing the effects as a standardized value (SMD), the results can be combined because they have no units. Moiniche et al.’s review also included several studies that were excluded from this meta-analysis for reasons of methodological deficiencies (Appendix 2) (36,40,42,44,45). The present meta-analysis featured stricter inclusion criteria, as reflected by the fact that we evaluated only 66 (64%) of 102 potentially eligible trials, whereas Moiniche’s review included 80 (86%) of 93 potential studies (7).
One possible caveat of preemptive analgesia may be the theoretical risk of complications when certain drugs are administered before surgery. This is exemplified by the possible increased risk of intraoperative and postoperative bleeding problems from the use of preoperative NSAIDs. However, existing data from RCTs on the incidence of perioperative bleeding complications caused by NSAIDs have been conflicting (127). A recent meta-analysis concluded that the evidence that NSAIDs increase the incidence of bleeding after surgery is ambiguous (128).
The differences in the efficacy of the individual analgesic interventions for a preemptive effect may be due to the degree of sufficiency of the afferent blockade, the nature of the pain, and its inflammatory component. Preemptive analgesia cannot be effective if the analgesic intervention is not adequate. The analgesic interventions need to produce a sufficiently dense and long duration of blockade for them to block the transmission of noxious afferent information from the periphery to the spinal cord and brain (2). In this respect, it appears that systemic opioids do not provide a sufficiently dense and long duration of blockade of spinal nociception to prevent central sensitization. In contrast, epidural analgesia can provide the sufficiently dense blockade required for a positive outcome. In addition, it has been suggested that if preemptive analgesia extends its duration into the postoperative period, then prevention of pain hypersensitivity could be even more pronounced. For acute postoperative pain, inflammatory mediators and nociceptive input should be kept inhibited well into the postoperative period. Central sensitization may not be prevented if the treatment is terminated too early.
In conclusion, on the basis of the surrogate outcome measures “postoperative pain scores,” “total analgesic consumption,” and “time to first rescue analgesic,” this meta-analysis demonstrates the possible efficacy of preemptive analgesia to improve postoperative acute pain management in selected analgesic regimens. The ES was most pronounced for preemptive administration of epidural analgesia, local anesthetic wound infiltration, and NSAIDs. Although preemptive epidural analgesia resulted in consistent improvements in all three outcome variables, preemptive local anesthetic wound infiltration and NSAID administration improved analgesic consumption and time to first rescue analgesia, but not postoperative pain scores. The least proof of efficacy was found in the case of systemic NMDA antagonist and opioid administration, and the results remain equivocal.
The authors thank V. T. W. Fan, MD, Department of Oral & Maxillofacial Surgery, Faculty of Dentistry, National University of Singapore, for reviewing the manuscript.
Table. Excluded Tria...Image Tools
Table. Detailed Reas...Image Tools
1. Woolf CJ, Wall PD. Morphine-sensitive and morphine-insensitive actions of C-fiber input on the rat spinal cord. Neurosci Lett 1986;64:221–5.
2. Kissin I. Preemptive analgesia. Anesthesiology 2000;93:1138–43.
3. Wilder-Smith OH. Pre-emptive analgesia and surgical pain. Prog Brain Res 2000;129:505–24.
4. Kelly DJ, Ahmad M, Brull SJ. Preemptive analgesia. II. Recent advances and current trends. Can J Anaesth 2001;48:1091–101.
5. McQuay HJ. Do pre-emptive treatments provide better pain control? In: Gebhart GF, Hammond DL, Jesen TS, eds. Proceedings of the 7th World Congress on Pain: progress in pain research and management. Vol 2. Seattle: IASP Press, 1994:709–23.
6. Niv D, Lang E, Devor M. The effect of preemptive analgesia on acute postoperative pain. Minerva Anestesiol 1999;65:127–40.
7. Moiniche S, Kehlet H, Dahl JB. A qualitative and quantitative systematic review of preemptive analgesia for postoperative pain relief. Anesthesiology 2002;96:725–41.
8. Kissin I. Preemptive analgesia: why its effect is not always obvious. Anesthesiology 1996;84:1015–9.
9. Katz J, McCartney CJL. Current status of preemptive analgesia. Curr Opin Anaesthesiol 2002;15:435–41.
10. Dahl JB, Kehlet H. The value of pre-emptive analgesia in the treatment of postoperative pain. Br J Anaesth 1993;70:434–9.
11. Woolf CJ, Chong MS. Pre-emptive analgesia: treating postoperative pain by preventing the establishment of central sensitization. Anesth Analg 1993;77:362–79.
12. Penning JP. Pre-emptive analgesia: what does it mean to the clinical anaesthetist? Can J Anaesth 1996;43:97–101.
13. Ong KS, Seymour RA. An evidence-based medicine approach to pre-emptive analgesia. Am J Pain Manage 2003;13:150–64.
14. Wall PD. The prevention of postoperative pain. Pain 1988;33:289–90.
15. Mulrow CD, Oxman A, eds. How to conduct a Conchrane systematic review. The Cochrane Library (Issue 2). The Cochrane Collaboration Handbook (version 3.0.2). San Antonio Cochrane Collaboration, 1997.
16. Jadad A, Moore A, Carrol D, et al. Assessing the quality of randomized trials: is blinding necessary? Control Clin Trials 1996;17:1–12.
17. L’Abbe KA, Detsky AS, O’Rourke K. Meta-analysis in clinical research. Ann Intern Med 1987;107:224–33.
18. Hedges LV, Olkin I. Vote-counting methods in research synthesis. Psychol Bull 1980;88:359–69.
19. Lewis S, Clarke M. Forrest plots: trying to see the wood and trees. BMJ 2001;322:1479–80.
20. Sterne JA, Gavaghan D, Egger M. Publication and related bias in meta-analysis: power of statistical tests and prevalence in the literature. J Clin Epidemiol 2000;53:1119–29.
21. Beilin B, Bessler H, Mayburd E, et al. Effects of preemptive analgesia on pain and cytokine production in the postoperative period. Anesthesiology 2003;98:151–5.
22. Burmeister MA, Gottschalk A, Freitag M, et al. Pre- and intra-operative epidural ropivacaine has no early preemptive analgesic effect in major gynecologic surgery. Can J Anaesth 2003;50:568–73.
23. Neustein SM, Kreitzer JM, Krellenstein D. Preemptive epidural analgesia for thoracic surgery. Mt Sinai J Med 2002;69:101–4.
24. Flisberg P, Tornebrandt K, Walther B, Lundberg J. A comparison of the effects on postoperative pain relief of epidural analgesia started before or after surgery. Eur J Anaesthesiol 2000;17:627–33.
25. Kirdemir P, Ozkocak I, Demir T, Gogus N. Comparison of postoperative analgesic effects of preemptively used epidural ketamine and neostigmine. J Clin Anesth 2000;12:543–8.
26. Vaida SJ, David BB, Somri M, et al. The influence of preemptive spinal anesthesia on postoperative pain. Clin Anesth 2000;12:374–7.
27. Wu CT, Yeh CC, Yu JC, et al. Pre-incisional epidural ketamine, morphine and bupivacaine combined with epidural and general anaesthesia provides pre-emptive analgesia for upper abdominal surgery. Acta Anaesthesiol Scand 2000;44:63–8.
28. Gottschalk A, Smith DS, Jobes DR, et al. Preemptive epidural analgesia and recovery from radical prostatectomy. JAMA 1998;279:1076–82.
29. Kucuk N, Kizilkaya M, Tokdemir M, et al. Preoperative epidural ketamine does not have a postoperative opioid sparing effect. Anesth Analg 1998;87:103–6.
30. Rockemann MG, Seeling W, Pressler S, et al. Reduced postoperative analgesic demand after inhaled anesthesia in comparison to combined epidural-inhaled anesthesia in patients undergoing abdominal surgery. Anesth Analg 1997;84:600–5.
31. Williams-Russo P, Sharrock NE, Haas SB, et al. Randomized trial of epidural versus general anesthesia: outcomes after primary total knee replacement. Clin Orthop 1996;331:199–208.
32. Moiniche S, Hjortso NC, Hansen BL, et al. The effect of balanced analgesia on early convalescence after major orthopaedic surgery. Acta Anaesthesiol Scand 1994;38:328–35.
33. Cerfolio RJ, Brant AS, Bass CS, Martolucci AA. A prospective double-blinded randomized trial evaluating the use of preemptive analgesia of the skin before thoracotomy. Ann Thorac Surg 2003;76:1055–8.
34. Maestroni U, Sortini D, Devito C, et al. A new method of preemptive analgesia in laparoscopic cholecystectomy. Surg Endosc 2002;16:1336–40.
35. Gatt CJ, Parker RD, Tetzlaff JE, et al. Preemptive analgesia: its role and efficacy in anterior cruciate ligament reconstruction. Am J Sports Med 1998;26:524–9.
36. Campbell WI, Kendrick RW, Fee JPH. Balanced pre-emptive analgesia: does it work? A double-blind, controlled study in bilaterally symmetrical oral surgery. Br J Anaesth 1998;81:727–30.
37. Hoard MA, Bill TJ, Campbell RL. Reduction in morbidity after iliac crest bone harvesting: the concept of preemptive analgesia. J Craniofac Surg 1998;9:448–51.
38. Ke RW, Portera SG, Lincoln SR. A randomized blinded trial of preemptive local anesthesia in laparoscopy. Prim Care Update Ob Gyns 1998;5:197–8.
39. Ko CY, Thompson JE, Alcantara A, Hiyama D. Preemptive analgesia in patients undergoing appendectomy. Arch Surg 1997;132:874–7.
40. Campbell WI, Kendrick RW. Pre-emptive analgesia using local anaesthesia: a study in bilaterally symmetrical oral surgery. Br J Anaesth 1997;79:657–9.
41. Pedersen JL, Crawford ME, Dahl JB, et al. Effect of preemptive nerve block on inflammation and hyperalgesia after human thermal injury. Anesthesiology 1996;84:1020–6.
42. Elhakim M, Abdel HH. Comparison of preoperative with postoperative topical lidocaine spray on pain after tonsillectomy. Acta Anaesthesiol Scand 1995;39:1032–5.
43. Wu CT, Yu JC, Liu ST, et al. Preincisional dextromethorphan treatment for postoperative pain management after upper abdominal surgery. World J Surg 2000;24:512–7.
44. Fu ES, Miguel R, Scharf JE, et al. Preemptive ketamine decreases postoperative narcotic requirements in patients undergoing abdominal surgery. Anesth Analg 1997;34:1086–90.
45. Tverskoy M, Oz Y, Isakson A, et al. Preemptive effect of fentanyl and ketamine on postoperative pain and wound hyperalgesia. Anesth Analg 1994;78:205–9.
46. Oztekin S, Hepaguslar H, Kar AA, et al. Preemptive diclofenac reduces morphine use after remifentanil-based anaesthesia for tonsillectomy. Paediatr Anaesth 2002;12:694–9.
47. Hanlon DM, Thambipillai T, Colbert ST, et al. Regional anesthesia and pain: timing of pre-emptive tenoxicam is important for postoperative analgesia. Can J Anaesth 2001;48:162–6.
48. Mixter CG, Hacket TR. Preemptive analgesia in the laparoscopic patient. Surg Endosc 1997;11:351–3.
49. Espinet A, Henderson DJ, Faccenda KA, et al. Does pre-incisional thoracic extradural block combined with diclofenac reduce postoperative pain after abdominal hysterectomy? Br J Anaesth 1996;76:209–13.
50. Romej M, Vopel-Lewis T, Merkel SI, Reyneolds PI. Effect of preemptive acetaminophen on postoperative pain scores and oral fluid intake in pediatric tonsillectomy patients. AANA J 1996;64:535–40.
51. Sandin R, Sternlo JE, Stam H, et al. Diclofenac for pain relief after arthroscopy: a comparison of early and delayed treatment. Acta Anaesthesiol Scand 1993;37:747–50.
52. Vogel RI, Desjardins PJ, Major KVO. Comparison of presurgical and immediate postsurgical ibuprofen on postoperative periodontal pain. J Periodontol 1992;63:914–8.
53. Flath RK, Hicks ML, Dionne RA, et al. Pain suppression after pulpectomy with preoperative flurbiprofen. J Endod 1987;13:339–46.
54. Chiaretti A, Viola L, Pietrini D, et al. Preemptive analgesia with tramadol and fentanyl in pediatric neurosurgery. Childs Nerv Syst 2000;16:93–9.
55. Motamed C, Mazoit X, Ghanouchi K, et al. Preemptive intravenous morphine-6-glucuronide is ineffective for postoperative pain relief. Anesthesiology 2000;92:355–60.
56. Collis R, Brandner B, Bromley LM, et al. Is there any clinical advantage of increasing the pre-emptive dose of morphine or combining pre-incisional with postoperative morphine administration? Br J Anaesth 1995;74:396–9.
57. Katz J, Cohen L, Schmid R, et al. Postoperative morphine use and hyperalgesia are reduced by preoperative but not postoperative epidural analgesia: implications for preemptive analgesia and the prevention of central sensitization. Anesthesiology 2003;98:1449–60.
58. Esmaoglu A, Cuha Y, Boyaci A. Preemptive efficacy of epidural fentanyl in elective abdominal surgery. Eur J Anaesthesiol 2001;18:59–63.
59. Aida S, Fujihara H, Taga K, et al. Involvement of presurgical pain in preemptive analgesia for orthopedic surgery. Pain 2000;84:169–73.
60. Subramaniam B, Pawar DK, Kashyap L. Preemptive analgesia with epidural morphine or morphine with bupivacaine. Anaesth Intensive Care 2000;28:392–8.
61. Aida S, Baba H, Yamakura T, et al. The effectiveness of preemptive analgesia varies according to the type of surgery. Anesth Analg 1999;89:771–6.
62. Obata H, Saito S, Fujita N, et al. Epidural block with mepivacaine before surgery reduces long-term post-thoracotomy pain. Can J Anaesth 1999;46:1127–32.
63. Kundra P, Deepalakshmi K, Ravishankar M. Preemptive caudal bupivacaine and morphine for postoperative analgesia in children. Anesth Analg 1998;87:52–6.
64. Richards JT, Read JRM, Chambers WA. Epidural anaesthesia as a method of preemptive analgesia for abdominal hysterectomy. Anaesthesia 1998;53:296–307.
65. Wong CS, Lu CC, Cheng CH, Ho ST. Preemptive analgesia with ketamine, morphine and epidural lidocaine prior to total knee replacement. Can J Anaesth 1997;44:31–7.
66. Choe H, Choi YS, Kim YH, Ko SH. Epidural morphine plus ketamine for upper abdominal surgery: improved analgesia from preincisional versus postincisional administration. Anesth Analg 1997;84:560–3.
67. Kundra P, Gurnani A, Bhattacharya A. Preemptive epidural morphine for postoperative pain relief after lumbar laminectomy. Anesth Analg 1997;85:135–8.
68. Aguilar J, Domingo V, Espachs P, et al. Absence of an early preemptive effect after thoracic extradural bupivacaine in thoracic surgery. Br J Anaesth 1996;76:72–6.
69. Dhal JB, Daugaard JJ, Rasmussen B. Immediate and prolonged effects of pre- versus postoperative epidural analgesia with bupivacaine and morphine on pain at rest and during mobilization after total knee arthroplasty. Acta Anaesthesiol Scand 1994;38:557–61.
70. Holthusen H, Eichwede F, Stevens M, et al. Preemptive analgesia: comparison of preoperative with postoperative block on postoperative pain in children. Br J Anaesth 1994;73:440–2.
71. Rockemann MG, Seeling W, Bischof C, et al. Prophylactic use of epidural mepivacaine. Anesthesiology 1996;84:1027–34.
72. Katz J, Clairoux M, Kavanagh BP, et al. Preemptive lumbar epidural anaesthesia reduces postoperative pain and patient-controlled morphine consumption after lower abdominal surgery. Pain 1994;59:395–403.
73. Pryle BJ, Vanner RG, Enriquez N, Renolds F. Can preemptive lumbar epidural blockade reduce postoperative pain following lower abdominal surgery? Anaesthesia 1993;48:120–3.
74. Dahl JB, Hansen BL, Hjortso NC, et al. Influence of timing on the effect of continuous extradural analgesia with bupivacaine and morphine after major abdominal surgery. Br J Anaesth 1992;69:4–8.
75. Rice LJ, Pudimat MA, Hannallah RS. Timing of caudal block placement in relation to surgery does not affect duration of postoperative analgesia in pediatric ambulatory patients. Can J Anaesth 1990;37:429–31.
76. Mahfouz AK, Nabawi KS. Preemptive analgesia in rhegmatogenous retina detachment surgery: is it effective? Retina 2002;22:602–6.
77. Reuben SS, Sklar J, El-Mansouri M. The preemptive analgesic effect of intraarticular bupivacaine and morphine after ambulatory arthroscopic knee surgery. Anesth Analg 2001;92:923–6.
78. Kristin N, Schonfeld CL, Bechmann M, et al. Vitreoretinal surgery: preemptive analgesia. Br J Ophthalmol 2001;85:1328–31.
79. Gill P, Kiani S, Victoria BA, et al. Preemptive analgesia with local anesthetic for herniorrhaphy. Anaesthesia 2001;56:414–7.
80. Altintas F, Bozkurt P, Ipek N, Kaya G. The efficacy of pre- versus postsurgical axillary block on postoperative pain in paediatric patients. Paediatr Anaesth 2000;10:23–8.
81. Hanlon DM, Colbert ST, Keane PW, et al. Preemptive bupivacaine offers no advantages to postoperative wound infiltration in analgesia for outpatient breast biopsy. Am J Surg 2000;180:29–32.
82. Fischer S, Troidl H, Maclean AA, et al. Prospective double-blind randomized study of a new regime of preemptive analgesia for inguinal hernia repair. Eur J Surg 2000;166:545–51.
83. Ke RW, Porte SG, Bagous W, Lincoln SR. A randomized blinded trial of preemptive analgesia in laparoscopy. Obstet Gynecol 1998;92:972–5.
84. Molliex S, Haond P, Baylot D. Effect of pre- versus postoperative tonsillar infiltration with local anesthetics on postoperative pain after tonsillectomy. Acta Anaesthesiol Scand 1996;40:1210–5.
85. Pasqualucci A, Angelis VD, Contardo R, et al. Preemptive analgesia: intraperitoneal local anesthetic in laparoscopic cholecystectomy. Anesthesiology 1996;85:11–20.
86. Dahl V, Raeder JC, Erno PE, et al. Preemptive effect of preincisional versus postincisional infiltration of local anaesthetic on children undergoing hernioplasty. Acta Anaesthesiol Scand 1996;40:847–51.
87. Huffnagle HJ, Norris MC, Barbara LL, et al. Ilioinguinal iliohypogastric nerve blocks—before or after cesarean delivery under spinal anesthesia? Anesth Analg 1996;82:8–12.
88. Orntoft S, Longreen A, Moiniche S, et al. A comparison of pre- and postoperative infiltration with bupivacaine on pain after tonsillectomy. Anaesthesia 1994;94:151–4.
89. Turner GA, Chalkiadis G. Comparison of preoperative with postoperative lignocaine infiltration on postoperative analgesic requirements. Br J Anaesth 1994;72:541–3.
90. Ejlersen E, Andersen HB, Eliasen K, et al. A comparison between preincisional and postincisional lidocaine infiltration and postoperative pain. Anesth Analg 1992;74:495–8.
91. Helmy AK, Ayham B. The effect of the preemptive use of NMDA receptor antagonist dextromethorphan on postoperative analgesic requirements. Anesth Analg 2001;92:739–44.
92. Dahl V, Ernoe PE, Steen T, et al. Does ketamine have preemptive effects in women undergoing abdominal hysterectomy procedures? Anesth Analg 2000;90:1419–22.
93. Menigaux C, Fletcher D, Dupont X, et al. The benefits of intraoperative small-dose ketamine on postoperative pain after anterior cruciate ligament repair. Anesth Analg 2000;90:129–35.
94. Adam F, Libier M, Oszustowicz T, et al. Preoperative small-dose ketamine has no preemptive analgesic effect in patients undergoing total mastectomy. Anesth Analg 1999;89:444–7.
95. Chia YY, Liu K, Chow LH, Lee TY. Preoperative administration of intravenous dextromethorphan reduces postoperative morphine consumption. Anesth Analg 1999;89:748–52.
96. Wu CT, Yu JC, Yeh CC, et al. Preincisional dextromethorphan treatment decreases postoperative pain and opioid requirement after laparoscopic cholecystectomy. Anesth Analg 1999;88:1331–4.
97. Mathisen LC, Aasbo V, Raeder J. Lack of preemptive analgesic effect of ketamine in laparoscopic cholecystectomy. Acta Anaesthesiol Scand 1999;43:220–4.
98. Ong KS, Seymour RA, Chen FG, Ho VCL. Preoperative ketorolac has a preemptive effect after oral surgery. Int J Oral Maxillofac Surg 2004;33:274–8. doi:10.1016/j.ijom.2003.09.015.
99. Reuben SS, Bhopatkar S, Maciolek H, et al. The preemptive analgesic effect of rofecoxib after ambulatory arthroscopic knee surgery. Anesth Analg 2002;94:55–9.
100. Priya V, Divatia JV, Sareen R, Upadhye S. Efficacy of intravenous ketoprofen for preemptive analgesia. J Postgrad Med 2002;48:109–12.
101. Norman PH, Daley MD, Lindsey RW. Preemptive analgesic effects of ketorolac in ankle surgery. Anesthesiology 2001;94:599–603.
102. Rosaeg OP, Krepski B, Cicutti N, et al. Effect of preemptive multimodal analgesia for arthroscopic knee ligament repair. Reg Anesth Pain Med 2001;26:125–30.
103. Nagatsuka C, Ichinohe T, Kaneko Y. Preemptive effects of a combination of preoperative diclofenac, butorphanol and lidocaine on postoperative pain management following orthognathic surgery. Anesth Prog 2000;47:119–24.
104. Colbert ST, Hanlon DM, McDonnell C. Analgesia in day case breast biopsy—the value of pre-emptive tenoxicam. Can J Anaesth 1998;45:217–22.
105. Romsing J, Ostergaard D, Larsen SW, Valentin N. Analgesic efficacy and safety of preoperative versus postoperative ketorolac in paediatric tonsillectomy. Acta Anaesthesiol Scand 1998;42:770–5.
106. Hanlon JJ, Muldoon T, Lowry D, at al. Improved postoperative analgesia with preoperative piroxicam. Can J Anaesth 1996;43:102–5.
107. Vanlersberghe C, Lauwers MH, Camu F. Preoperative ketorolac administration has no preemptive analgesic effect for minor oral surgery. Acta Anaesthesiol Scand 1996;40:948–52.
108. Fletcher D, Zetlaoui P, Monin S, at al. Influence of timing on the analgesic effect of intravenous ketorolac after orthopedic surgery. Pain 1995;61:291–7.
109. Rogers JEG, Fleming BG, Macintosh KC, et al. Effect of timing of ketorolac administration on patient-controlled opioid use. Br J Anaesth 1995;75:15–8.
110. Buggy DJ, Wall C, Carton EG. Preoperative or postoperative diclofenac for laparoscopic tubal ligation. Br J Anaesth 1994;73:767–70.
111. Nelson WE, Henderson RC, Almekinders LC, et al. An evaluation of pre- and postoperative nonsteroidal antiinflammatory drugs in patients undergoing knee arthroscopy: a prospective, randomized, double-blinded study. Am J Sports Med 1993;21:510–6.
112. Murphy DF, Medley C. Preoperative indomethacin for pain relief after thoracotomy: comparison with postoperative indomethacin. Br J Anaesth 1993;70:298–300.
113. Sisk AL, Grover BJ. A comparison of preoperative and postoperative naproxen sodium for suppression of postoperative pain. J Oral Maxillofac Surg 1990;48:674–8.
114. Sisk AL, Mosley RO, Martin RP. Comparison of preoperative and postoperative diflunisal for suppression of postoperative pain. J Oral Maxillofac Surg 1989;47:464–8.
115. Doyle E, Bowler GMR. Preemptive effect of multimodal analgesia in thoracic surgery. Br J Anaesth 1998;80:147–51.
116. Millar AY, Mansfield MD, Kinsella J. Influence of timing of morphine administration on postoperative pain and analgesic consumption. Br J Anaesth 1998;81:373–6.
117. Griffin MJ, Hughes D, Knaggs A, et al. Late-onset preemptive analgesia associated with preincisional large-dose alfentanil. Anesth Analg 1997;85:1317–21.
118. Sarantopoulos C, Fassoulaki A. Sufentanil does not preempt pain after abdominal hysterectomy. Pain 1996;65:273–6.
119. Fassoulaki A, Sarantopoulos C, Zotou M, Papoulia D. Preemptive opioid analgesia does not influence pain after abdominal hysterectomy. Can J Anaesth 1995;42:109–13.
120. Wilson RJT, Leith S, Jackson JB, et al. Preemptive analgesia from intravenous administration of opioids: no effect. Anaesthesia 1994;49:591–3.
121. Mansfield M, Meikle R, Miller C. Influence of timing of preoperative alfentanil on postoperative pain and analgesic requirements. Anaesthesia 1994;49:1091–3.
122. Richmond CE, Bromley LM, Woolf CJ. Preoperative morphine preempts postoperative pain. Lancet 1993;342:73–5.
123. Moher D, Jadad AR, Nichol G, Penman M. Assessing the quality of randomized trials: an annotated bibliography of scales and checklists. Control Clin Trials 1995;16:62–73.
124. McQuay HJ. Pre-emptive analgesia. Br J Anaesth 1992;69:1–3.
125. Hogan QH. No preemptive analgesia—is that so bad? Anesthesiology 2002;96:526–7.
126. Seymour RA, Meechan JG, Blair GS. An investigation into post-operative pain after third molar surgery under local analgesia. Br J Oral Maxillofac Surg 1985;23:410–8.
127. 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.
128. Moiniche S, Romsing J, Dahl JB, Tramer MR. NSAIDs and the risk of operative site bleeding after tonsillectomy: a quantitative systematic review. Anesth Analg 2003;96:68–77.