Pharmacological control of opioid-induced pruritus: a quantitative systematic review of randomized trials

Transcription

1 European Journal of Anaesthesiology 2001, 18, REVIEW Pharmacological control of opioid-induced pruritus: a quantitative systematic review of randomized trials F. Kjellberg and M. R. Tramèr Division of Anaesthesiology, Department APSIC, Geneva University Hospitals, Geneva, Switzerland. Summary Background and objective Numerous drugs have been used to prevent or to treat opioid-induced pruritus in the surgical setting. Their relative efficacy is not well understood. Methods The methods employed involved the systematic search (MEDLINE, EMBASE, Cochrane library, bibliographies, without language restriction, up to June 2000) for full reports of randomized comparisons of any intervention which is thought to be anti-pruritic (active) compared with placebo or no treatment (control) in surgical (including labour) patients receiving opioids. The number of patients who had no pruritus were analysed using relative risk and number-neededto-treat with 95% confidence interval. Results Twenty-two trials (1477 patients) were analysed. Two trials (66 patients), both with low-dose propofol, were on treatment of established pruritus; propofol had no anti-pruritic effect compared with Intralipid. In prophylaxis trials, the average incidence of pruritus with control was 59% (range, 10% to 100%). Most mu-receptor antagonists were efficacious: intravenous naloxone mgkg 1 h 1, relative risk 2.31 (95% confidence interval, 1.5 to 3.54), numberneeded-to-treat to prevent pruritus compared with control 3.5; oral naltrexone 9 mg, relative risk 2.80 ( ), number-needed-to-treat 1.7; naltrexone 6 mg was less effective and 3 mg did not work; different intravenous and epidural nalbuphine regimens, relative risk 1.71 ( ), number-needed-to-treat 4.2. Intravenous nalmefene 0.5 or 1 mg was not antipruritic. Intravenous (but not epidural) droperidol 2.5 mg was efficacious, relative risk, 1.71 ( ), number-needed-to-treat 4.9. There was a lack of evidence for any anti-pruritic efficacy with prophylactic propofol, epidural or intrathecal epinephrine, epidural clonidine, epidural prednisone, intravenous ondansetron, or intramuscular hydroxyzine. Conclusion Naloxone, naltrexone, nalbuphine and droperidol are efficacious in the prevention of opioidinduced pruritus; minimal effective doses remain unknown. There is a lack of valid data on the efficacy of interventions for the treatment of established pruritus. Keywords: PHARMACOLOGY, naloxone, naltrexone, droperidol; PAIN, postoperative; COMPLICATIONS, pruritus, vomiting; META-ANALYSIS. Introduction Opioids are an integral part of the treatment of pain due to surgery and labour. Opioids, however, may induce adverse effects. In a cohort study on admissions of 4031 adults to medical and surgical units, morphine, demerol (pethidine) and oxycodone accounted for 18% of all adverse drug events [1]. Pruritus is a Accepted October 2000 Correspondence: M. R. Tramèr. common adverse effect of opioids; it is thought to occur most frequently after epidural and intrathecal administration [2]. The mechanism of opioid-related pruritus is not fully understood. The role of histamine is unclear; pruritus may be induced by opioids which do not release histamine. The fact that mu-receptor antagonists (naloxone, nalbuphine) can reverse opioid-related pruritus suggests that there may be a mu-receptor mediated central mechanism [3]. Many different drugs with various potential mechanisms of action have been used to prevent or 346 # 2001 European Academy of Anaesthesiology

2 Pharmacological control of opioid-induced pruritus 347 to treat established pruritus. Their relative efficacy, optimal doses and likelihood for adverse effects are not well understood. Naloxone, for instance, although widely used to control opioid-related pruritus, may reverse the analgesic effect of opioids. The purpose of this study was to identify all pharmacological therapies that have been used to control opioid-related pruritus in the surgical setting (including labour), and to establish their relative efficacy. Methods Data search We performed a systematic search for full reports of randomized comparisons of any intervention which is thought to be anti-pruritic (active) compared with placebo or no treatment (control) in surgical patients (including labour) receiving an opioid. Comparisons of an active drug with another active drug but without a placebo or no treatment arm (i.e. head-to-head comparisons) were not considered. We searched in different electronic databases (MEDLINE from 1966, EMBASE from 1974, Cochrane library issue IV 1999) using the free text keywords pruritus, itching, antipruritic, opioid, postoperative, postsurgical, anaesthesia, anesthesia, randomized, randomised, controlled, human, and their combinations. We also checked reference lists of retrieved reports. We did not contact manufacturers or authors. The search was without language restriction, and included reports up to June We did not include data from letters, abstracts, case reports, experimental studies and review articles. Studies with group sizes <10 patients were excluded. Critical appraisal and data extraction Both authors read all included trials and scored them independently for methodological validity using the three-item, five-point Oxford scale which takes into account randomization, blinding and description of withdrawals [4]. In a trial with a no treatment control, the method of blinding was a priori regarded as inadequate (i.e. 0 points for blinding in the Oxford scale). Both authors extracted data independently. The primary end-point of efficacy was the number of patients who were completely free of pruritus with active and with control. Itching was considered as pruritus. Data on decreases in pruritus scores or visual analogue scales were not further analysed, as these data were inconsistently reported. Secondary efficacy end-points were the incidences of other opioidinduced adverse effects, for instance, nausea and vomiting. Data on adverse drug reactions were extracted when they were reported in a dichotomous form. Information on patients, clinical settings, doses and routes of administration of opioids and anti-pruritic drugs, and duration of observation periods were recorded. Discrepancies in validity scores and extracted data were resolved by discussion. Quantitative analyses Patients receiving no treatment were considered as placebo controls. In propofol trials, Intralipid was regarded as an inactive control. Statistical significance of any treatment effect was calculated as relative risk with 95% confidence interval [5]. For combined data, we used a fixed effect model throughout because we combined data only when they were clinically homogenous, and because heterogeneity tests lack sensitivity [6]. As an estimate of the clinical relevance of a treatment effect, the number-needed-to-treat was calculated [7] with 95% confidence interval [8]. A positive number-needed-to-treat indicated the number of patients who had to be exposed to an active intervention to show a particular end-point in one of them (for instance, prevention of pruritus) who would not have shown that end-point had they all received a control treatment. Upper and lower limits of the 95% confidence interval around the number-neededto-treat, regardless of whether they were positive or negative, are reported [9]. When the 95% confidence interval of the relative risk excluded one, we assumed that the result was not statistically significant. In this case, we would expect the 95% confidence interval around the number-needed-to-treat to range from a positive limit to a negative limit, indicating that the confidence interval includes zero and thus infinity. Results Retrieved trials We identified 70 reports addressing the issue of opioid-induced pruritus (Fig. 1) [10]. Twenty were

3 348 F. Kjellberg and M. R. Tramèr Table 1. Pharmocological control of opioid-induced pruritus. Analysed randomized controlled trials Ref. Quality score (random blinding withdrawals) Surgical setting Opioid (PCEA ¼ patient controlled epidural analgesia; i.t. ¼ intrathecal; epid ¼ epidural) Comparisons ( ) Number of patients [ ] Extrapolated regimens [35] 2/1/0 Caesarean section Morphine 4 mg epid 1. Naltrexone 6 mg oral (15) 2. Naltrexone 9 mg oral (15) 3. Placebo oral (15) [36] 2/1/0 Caesarean section Morphine 250 mg i.t. 1. Naltrexone 3 mg oral (10) 2. Naltrexone 6 mg oral (12) 3. Placebo oral (13) [33] 2/1/0 Caesarean section Morphine 250 mg i.t. 1. Propofol 10 mg 1-2 x (17) 2. Intralipid (12) [37] 1/0/0 Labour Sufentanil 10 mg kg 1 i.t. 1. Epinephrine 200 mg i.t. (20) 2. No treatment (20) [38] 1/1/0 Caesarean section Morphine 200 mg þ fentanyl 10 mg i.t. 1. Naloxone 48 mg h -1 [0.6 mg kg 1 h 1 ] i.v. (20) 2. Nalmefene 0.25 mg kg 1 i.v. x 2 (12 h) (20) 3. Nalmefene 0.5 mg kg 1 i.v. x 2 (12 h) (20) 4. Placebo i.v. (20) [39] 1/0/0 Caesarean section Morphine 2 mg x 2 (18 ) epid 1. Prednisone 50 mg epidurally x 2 (18 ) (37) 2. No treatment (58) [40] 2/2/0 Abdominal Morphine-PCA 1. Naloxone 0.25 mg kg 1 h 1 i.v. (20) hysterectomy 2. Naloxone 1 mg kg 1 h 1 i.v. (20) 3. Saline i.v. (20) [41] 1/1/1 Orthopaedic Morphine 250 mg i.t. 1. Propofol 30 mg h 1 i.v. for up to 20 h (40) 2. Intralipid (41) [42] 1/0/1 Caesarean section Morphine 2 mg epid 1. Droperidol 2.5 mg i.v. (54) 2. No treatment (53) Time of administration Observation period 5 min after morphine 60 min after morphine Treatment of pruritus 6 h With intrathecal sufentanil 3h At the time of cord clamp and 12 h later (groups 2 and 3) End of surgery 18 Postanaesthetic Care Unit End of surgery After delivery

4 Pharmacological control of opioid-induced pruritus 349 [43] 2/0/0 Caesarean section Morphine 2 mg epid Morphine 4 mg epid 1. Droperidol 2.5 mg i.v. (35) 2. Droperidol 5 mg i.v. (35) 3. No treatment (35) 4. Droperidol 2.5 mg i.v. (35) 5. Droperidol 5 mg i.v. (35) 6. No treatment (35) [44] 1/1/0 Caesarean section Morphine 5 mg epid 1. Hydroxyzine 50 mg i.m. (20) 2. Saline i.m. (20) [45] 1/1/0 Any surgery Alfentanil 10 mg kg 1 i.v. 1. Ondansetron 4 mg iv (40) 2. Saline i.v. (40) [46] 1/1/0 Caesarean section Morphine 5 mg epid 1. Nalbuphine 20 þ10 þ10 mg i.v. (28) 2. Saline i.v. (32) [47] 1/2/0 Hip replacement Morphine 4 mg epid 1. Droperidol 2.5 mg epid (20) 2. Saline epid (20) [48] 1/1/1 Caesarean section Hydromorphone PCEA 1. Nalbuphine 20 mg ml 1 epid (19) 2. Nalbuphine 40 mg ml 1 epid (16) 3. Nalbuphine 80 mg ml 1 epid (20) 4. Saline epid (20) [49] 2/0/1 Major gynaecologic Morphine 250 mg i.t. 1. Epinephrine 200 mg i.t. (25) 2. No treatment (22) 3. Saline (no morphine) (23) [50] 1/0/0 Caesarean section Morphine 5 mg epid 1. Droperidol 2.5 mg epid (32) 2. Droperidol 2.5 mg i.v. (32) 3. No treatment (33) [51] 1/1/0 Caesarean section Morphine 4 mg epid 1. Naloxone 400 mg i.v. þ 1600 mg 12h 1 [2.4 mg kg 1 h 1 ] (15) 2. Saline i.v. (15) 3. Morphine i.m. þ saline epid (15) [52] 1/1/0 Arthroplasty Morphine 300 mg it 1. Propofol 10 mg i.v. þ 30 mg 1 inf (20) 2. Intralipid (20) After delivery 10 min after epidural morphine 30 min before alfentanil 5 min At skin closure, and at 6 h, and 12 h postoperatively 18 h On admission to recovery room Postanaesthetic Care Unit 32 h With intrathecal morphine With epidural morphine 1 h after epidural morphine With intrathecal morphine

5 350 F. Kjellberg and M. R. Tramèr Table 1. (Continued ) Observation period Time of administration Comparisons ( ) Number of patients [ ] Extrapolated regimens Opioid (PCEA ¼ patient controlled epidural analgesia; i.t. ¼ intrathecal; epid ¼ epidural) Quality score (random blinding withdrawals) Surgical setting Ref. End of surgery 1. Epinephrine 2.5 mg ml 1 (1:400,000) epid (20) 2. Clonidine 3 mg ml 1 [53] 1/0/0 Caesarean section Sufentanil 4 mg i.t. þ sufentanil PCEA epid (20) 3. No treatment (20) End of surgery [54] 2/1/1 Abdominal Morphine 3 mg x 2 (12 h) epid 1. Nalbuphine 60 mg kg 1 h 1 hysterectomy i.v. (23) 2. Naloxone 2 mg kg 1 h 1 i.v. (22) 3. Saline i.v. (23) Prevention and treatment 8 h [34] 1/2/1 Caesarean section Morphine 200 mg i.t. 1. Propofol 10 mg i.v. (29) 2. Intralipid (29) not considered as they were not randomized trials, and six because they were not relevant for the purpose of this study. Of the remaining 44 reports, 22 were subsequently excluded. In 10, the number of patients who were completely free from pruritus after the administration of the study drugs was not reported [11 20]. Five had no placebo or no treatment control group [21 25]. Three were not properly controlled [26 28]. In two, group sizes were <10 patients [29,30]. Finally, two were in children [31,32]; these were excluded as all other trials were in adults. Of the 22 analysed trials, one was on treatment of established pruritus [33], and one on both, prevention and treatment (in two separate randomized designs) [34]. The others were on prevention only [35 54] (Table 1). Data on 1477 patients were analysed. Fourteen trials (1072 patients, 73% of all patients) were in obstetrics (labour or Caesarean section). Average trial size was 67 patients (range, ). Median validity score was 2 (range, 1 4); three scored 4, six 3, nine 2, and four scored 1. In most trials, the observation period was (minimum 5 min [45], maximum 48 h [50]). Underlying risk In trials on prevention, analgesia was with epidural morphine in 10 trials [35,39,42 44,46,47,50,51,54], with intrathecal morphine in five [34,36,41,49,52], with intrathecal sufentanil in two [37,53], with a combination of intrathecal morphine and fentanyl in one trial Fig. 1. Control of opioid-induced pruritus. Flow chart of retrieved and analysed reports.

6 Pharmacological control of opioid-induced pruritus 351 that very few patients obtained any relief from their pruritus. Pruritus was relieved in 2 of 17 (11.8%) patients who received propofol compared with 1 of 12 (8.3%) of those receiving Intralipid (not significant). All failures were successfully treated with intravenous naloxone mg. Three women receiving propofol felt dizzy compared with one who received Intralipid [33]. Prevention of pruritus Fig. 2. Incidence of pruritus in control patients (i.e. control event rate) with different opioids and routes of administrations. i.t. ¼ intrathecal; i.v. ¼ intravenous; PCA ¼ intravenous patient-controlled analgesia; PCEA ¼ patient-controlled epidural analgesia. [38], with i.v. morphine in one [40], with i.v. alfentanil in one [45], and with epidural hydromorphone in one [48]. In placebo and no treatment groups, the incidence of pruritus (i.e. the control event rate) was on average 58% with intrathecal morphine, 60% with epidural morphine, and 55% with all other opioids and routes of administration. There was graphically no evidence of a relationship between the dose of morphine and the incidence of pruritus with both intrathecal and epidural routes (Fig. 2). The incidence of pruritus in parturients was similar to those in the other settings. Data on prevention came from 21 trials. Opioid antagonists. There were four trials with i.v. naloxone 48 mg h 1 [38], 0.25 and 1.0 mgkg 1 h 1 [40], 400 mg i.v. plus 1600 mg 12h 1 [51] and 2 mg kg 1 h 1 [54] (Table 1). To test for dose-responsiveness, we extrapolated fixed doses [38,51] to variable doses (i.e. mgkg 1 h 1 ) using patients average body weight as reported in the original trials. For the resulting dose range, mg kg 1 h 1, there was graphically no evidence of dose-responsiveness (Fig. 3). Combined data suggested that naloxone was efficacious; compared with control, the number-needed-to-treat to prevent pruritus was 3.5 (Table 2a). With naxolene 0.6 mg kg 1 h 1, there was no significant difference in pain intensity compared with control [38]. With 1 mgkg 1 h 1, cumulative morphine consumption was increased compared with 0.25 mgkg 1 h 1, but there was no difference in the verbal pain rating scores [40]. With 2 mgkg 1 h 1, the number of patients requiring Treatment of established pruritus Two trials reported on treatment of established pruritus; both were with low-dose propofol [33,34]. In one [34], parturients received intrathecal morphine 200 mg, and were randomized to prophylactic propofol or Intralipid (control). Twenty-seven control patients developed pruritus, and were further randomized to treatment with propofol 10 mg or Intralipid. Within 8 h, one parturient in each group was completely free of pruritus (not significant). In the other trial [33], patients received 250 mg morphine intrathecally. Those who developed pruritus were treated either with propofol 10 mg once or twice, or Intralipid. The trial was terminated prematurely because the investigators noted Fig. 3. Antipruritic efficacy with different doses of naloxone. Vertical bars are 95% confidence intervals. Grey diamond: combine estimate.

7 Table 2. Prevention of opioid-induced pruritus. Efficacy data Incidence of pruritus Number of patients without pruritus/total number of patients Relative risk Number needed to treat Active Regimen Active Control Active Control (95% confidence interval) References A. mu-receptor antagonists Naloxone 0.25 to 2.4 mg kg 1 h 1 i.v 46.2% 80.3% 52/97 24/ (1.51 to 3.54) 3.5 (2.4 to 6.3) [38,40,51,54] Naltrexone 3 mg oral 70.0% 92.3% 3/10 1/ (0.47 to 32.1) 4.5 (1.8 to 10) [36] 6 mg oral 25.9% 78.6% 20/27 6/ (1.69 to 6.80) 1.9 (1.3 to 3.3) [35,36] 9 mg oral 6.7% 66.7% 14/15 5/ (1.35 to 5.80) 1.7 (1.1 to 3.0) [35] Nalmefene 0.5 or 1 mg i.v. 92.5% 95.0% 3/40 1/ (0.17 to 13.5) 40 (6.6 to 10) [38] Nalbuphine 40 mg 12 h 1 or 60 mg kg 1 h 1 i.v. or mg ml 1 epid 53.8% 77.3% 49/106 17/ (1.12 to 2.62) 4.2 (2.7 to 9.9) [46,48,54] B. D2-receptor antagonist Droperidol 2.5 mg i.v. 50.6% 71.2% 77/156 45/ (1.28 to 2.29) 4.9 (3.2 to 10) [42,43,50] 5 mg i.v. 60.0% 70.0% 28/70 21/ (0.84 to 2.11) 10 (3.9 to 17) [43] 2.5 mg epid 50.0% 64.2% 26/52 19/ (0.90 to 2.15) 7.1 (3.0 to 22) [47,50] 2.5 or 5 mg i.v or epid 47.1% 30.5% 131/278 75/ (1.27 to 1.98) 6.0 (4.0 to 12) [42,43,47,50] C. Intravenous anaesthetic Propofol 10 mg or 10 mg þ 30 mg 1 D. Catecholamine Epinephrine 1:400,000 epid or 200 mg i.t. 56.9% 59.7% 28/65 25/ (0.72 to 1.64) 36 (5.0 to 6.9) [45] E. H 1 -receptor antagonist Hydroxyzine 50 mg i.m. 25.0% 10.0% 15/20 18/ (0.62 to 1.12) 6.7 ( 2.6 to 12) [44] F. Alpha2-agonist Clonidine 3 mg ml 1 epid 25.0% 15.0% 15/20 17/ (0.65 to 1.21) 10 ( 2.9 to 6.9) [53] G. Corticosteroid Prednisone 50 mg epid 8.1% 13.8% 34/37 50/ (0.93 to 1.23) 18 (5.5 to 15) [39] H. 5-HT3 receptor antagonist Ondansetron 4 mg i.v. 30.0% 42.5% 28/40 23/ (0.87 to 1.70) 8.0 (3.0 to 12) [37,49,53] epid ¼ epidural, i.m. ¼ intramuscular, i.t. ¼ intrathecal, i.v. ¼ intravenous. or 30 mg h % 47.8% 52/89 47/ (0.87 to 1.45) 16 (4.8 to 12) [34,41,52] 352 F. Kjellberg and M. R. Tramèr

8 Pharmacological control of opioid-induced pruritus 353 rescue analgesia was significantly increased compared with control [54], and with 2.4 mg kg 1 h 1, there was less pain relief compared with control [51]. Oral naltrexone 3, 6 and 9 mg was tested in two trials [35,36]. There was evidence of dose-responsiveness (Table 2a). Naltrexone 3 mg was not different from control. The 6 and 9 mg doses, however, were significantlymoreefficaciousthancontrol;numbers-neededto-treat were below 2. With both 6 mg and 9 mg, the average duration of analgesia was shortened compared with control; with 9 mg, more patients reported unsatisfactory analgesia compared with 6 mg. Intravenous nalmefene 0.25 or 0.5 mg kg 1, tested in one trial [38], was not anti-pruritic (Table 2a). Five different regimens of i.v. nalbuphine were tested in three trials: 60 mgkg 1 h 1 [54], 40 mg [46], and 20, 40 and 80 mgml 1 epidurally [48]. With this variety of regimens it was impossible to test for dose-responsiveness. Combined data suggested anti-pruritic efficacy with nalbuphine; the numberneeded-to-treat to prevent pruritus compared with control was 4.2 (Table 2a). With 40 mg i.v., there was no difference in pain scores compared with control, but one patient who received nalbuphine required oxygen therapy [46]. With mgml 1 epidurally, less patients reported moderate or severe pain compared with control (33% to 47% with nalbuphine vs. 56% with control), but more were drowsy (23% to 39% with nalbuphine vs. 15% with control) [48]. When all trials on opioid antagonists which reported emesis data were combined, the anti-nausea effect was significant [average incidence of nausea with all opioid antagonists 44.4% vs. 54.4% with control, relative risk 1.30 (1.03 to 1.63)]. There was no effect on vomiting [34.7% with opioid antagonists vs. 37.7% with control, relative risk 1.07 ( )]. Droperidol Intravenous droperidol 2.5 or 5 mg was tested in three trials [42,43,50], and epidural droperidol 2.5 mg in two [47,50]. Intravenous droperidol 2.5 mg was efficacious, whereas both i.v. 5 mg and epidural 2.5 mg were not significantly different from control (Table 2b). When all droperidol data were combined, the number-needed-to-treat to prevent pruritus compared with control was 6. Droperidol significantly prevented nausea and vomiting [15% with droperidol vs. 24.8% with control, relative risk 1.15 (1.25 to 10.3)], but induced somnolence [11.9% vs. 2.6%, relative risk 4.57 (1.77 to 10.8)]. Other interventions Three different low-dose propofol regimens were tested in three trials (Table 1c). With 30 mg h 1 [41], there was more pruritus compared with Intralipid; 10 mg [34] was no different from Intralipid; and with 10 mg bolus plus 30 mg 24- h 1 [52], there was significantly less pruritus compared with Intralipid. To test for dose-responsiveness was not possible. When all data were pooled, there was no significant difference in anti-pruritic efficacy with propofol compared with Intralipid; the numberneeded-to-treat to prevent pruritus was 16. With the 30 mg h 1 infusion, one of 40 patients receiving propofol needed naloxone for respiratory depression, and eight of 40 needed dose reduction due to oversedation [41]. Intrathecal or epidural epinephrine, tested in two trials [37,49], had no significant effect on pruritus (Table 2d). Epidural clonidine 3 mgml 1 [53], intramuscular hydroxyzine 50 mg [44] and epidural prednisone 50 mg [39] were tested in one trial each; none was significantly different from control (Table 2e g). Prophylactic i.v. ondansetron 4 mg [45], was also tested in only one valid trial (Table 2h). During the administration of i.v. alfentanil, there was significantly less perinasal scratching [42.5% with ondansetron vs. 70% with saline (0.9% NaCl)]. However, 5 min after the administration of i.v. alfentanil, when patients were asked about the presence of itching, there was no significant difference (30% vs. 42.5%). Discussion We retrieved 70 reports addressing the issue of the pharmacological control of opioid-induced pruritus. This suggests that many investigators perceive pruritus as an important issue that is worthwhile studying. However, there were several problems with these reports. Numerous potentially relevant randomized trials had to be excluded as they were invalid for the purpose of this analysis. Many trialists reported on self-invented and non-validated measurements of anti-pruritic efficacy. The decrease of 4 points on a 10-point visual analogue pruritus scale may indeed suggest efficacy. However, such an end-point is likely to be chosen arbitrarily, or post hoc. These data cannot be put into a wider context; they cannot be compared or combined with data from other trials. Our dichotomous hurdle of complete absence of pruritus after

9 354 F. Kjellberg and M. R. Tramèr the application of the study drugs may be unnecessarily high. An active intervention which does not completely prevent pruritus may alleviate symptoms to a great extent. Such an intervention may be, of course, useful. However, in order to extract homogeneous data and to minimize the risk of observational bias due to different definitions of end-points, we decided to concentrate on complete absence of pruritus. Data from 11 randomized trials could thus not be used for the purpose of this analysis. The quality of the analysed trials, according to the validated Oxford scale [54] was often unsatisfactory. More than half scored 1 or 2, and none scored 5. Poor quality trials are likely to report biased estimates of efficacy [55]. This effect cannot be ruled out here, although it was impossible to test this hypothesis because of the large variety of different drugs and regimens tested. There was also the problem of active controlled trials without a placebo or no treatment control. It has been shown in other settings that head-to-head comparisons are unlikely to further our understanding on the efficacy of interventions unless there is a gold standard treatment against which all other active treatments may be compared [56]. This also applies to anti-pruritic drugs. The message that can be derived from this large body of data is of limited value only to clinical practice. Most trials were on prevention of pruritus. It may be interesting from a pharmacological point of view to know that very different molecules (for instance, antagonists of the mu or the dopamine receptors) may possess some anti-pruritic properties. The clinical relevance of this, however, is doubtful. It is unlikely that anaesthetists will change practice based on these data, and give, for instance, prophylactic droperidol or naloxone to their patients in an attempt to prevent pruritus. Only a very limited number of valid data on the treatment of established pruritus could be found. Thus, we still do not know what treatment a patient should receive who is complaining about opioidrelated pruritus. What then is the clinical message? Antagonists of the mu receptor are effective anti-pruritic drugs. Intravenous naloxone was most frequently tested. Of four patients receiving prophylactic naloxone, one will not have pruritus who would have had pruritus had they all received a placebo or no treatment. The failure to show dose-responsiveness most likely indicated that the doses tested in these trials were too high. Pain outcomes suggested that the dose of naloxone should probably not exceed 2 mgkg 1 h 1. Naltrexone did show dose-responsiveness; the lowest dose tested, 3 mg orally, did not work, and higher doses, 6 mg and 9 mg, showed significant anti-pruritic efficacy. Nine milligrams seemed to be too high a dose, as pain scores were increased. Nalbuphine, an agonist/antagonist at the mu-receptor, was also efficacious, although the variety of regimens tested makes it difficult to provide clear treatment recommendations. There was some evidence for less pain intensity but increased drowsiness with nalbuphine. Nalmefene, a pure opioid antagonist, tested in one trial with 40 treated patients only, did not show any efficacy. Finally, opioid-related nausea (but not vomiting) was decreased with mu-receptor antagonists. An alternative for prophylaxis may be i.v. (but not epidural) droperidol. The optimal dose remains unknown (the dose-response with 2.5 and 5 mg i.v. was inverse), and there is less efficacy too compared with the mu antagonists (number-neededto-treat about 6). Again not unexpectedly, patients are less likely to suffer from nausea or vomiting, but, with these relatively high doses tested, they are more likely to be somnolent [57]. None of the other tested drugs, epinephrine, propofol, clonidine, hydroxyzine or prednisone, showed any worthwhile benefit. This may be partly due to the limited number of tested patients. Ondansetron showed some efficacy on perinasal scratching but not on pruritus (itching). If the absolute risk difference of pruritus was about right (i.e. 30% with ondansetron vs. 42.5% with placebo), then the hypothetical number-needed-totreat to prevent pruritus with prophylactic ondansetron compared with placebo was about 8 (i.e. the reciprocal of 12.5%). The clinical relevance of this result is not obvious, and the fact the authors terminated the study 5 min after the administration of alfentanil does not make it easier to put the result into a clinical context. A further message relates to the risk of pruritus with opioids. In placebo and no treatment, controlled randomized trials, we may assume that there is some relationship between the control event rate (i.e. the incidence of pruritus in patients receiving a placebo or no treatment ) and the true underlying risk for pruritus. In these trials, on average 60% of patients

10 Pharmacological control of opioid-induced pruritus 355 receiving an opioid without an anti-pruritic drug had some pruritus. Thus, we may assume that these trials represent a high-risk population for opioid-induced pruritus. With epidural and intrathecal morphine, there was no evidence that the dose, in the range of doses tested in these trials, made any difference. Other opioids represented the same range of risk although the data were limited (Fig. 2). Knowing that pruritus due to opioid analgesia happens frequently, independent on the opioid used, the route of administration, and the dose, begs the question as to whether this adverse effect is clinically important. Pruritus never becomes chronic and it does not kill. An expert panel of anaesthesiologists ranked opioid-induced pruritus almost last in a hierarchy of the clinical relevance of different anaesthesia outcomes [58]. Even postoperative nausea, pain, or shivering ranked much higher. We do not know how patients themselves would rank the importance of pruritus. Research agenda A large variety of different drugs have been tested in these trials. Often, it was unclear if the trials were designed to provide a new and clinically useful treatment against pruritus, or if the authors primary interest was to study the origin of opioid-induced pruritus. As with other systematically searched trials, there was not much evidence for any scientific basis for most of these studies [59]. This systematic review reveals more what we do not know than what we know about the control of opioid-induced pruritus. Almost 1500 patients have been randomized in the 22 analysed trials, and hundreds in the trials which had to be excluded from our analyses. Many questions, however, remain unanswered. More data are needed on the treatment of established pruritus, on the efficacy of these drugs in children, on dose-responsiveness, and on optimal doses (i.e. minimal effective doses with an acceptable risk of adverse effects). Valid trials are needed, properly randomized, and double-blind, with a reasonable number of patients, and an adequate length of follow-up (ideally, ). End-points should be standardized, and should include the number of patients who are completely free of pruritus after administration of the study drugs. In the absence of a gold-standard treatment, trials should be placebo controlled. Acknowledgments We gratefully acknowledge Prosper Grant No from the Swiss National Research Foundation (MRT). We thank Daniel Haake from the Documentation Service of the Swiss Academy of Medical Sciences (DOKDI) for his help in searching electronic databases. References 1 Bates DW, Cullen D, Laird N et al. Incidence of adverse drug events and potential adverse drug events. Implications for prevention. ADE Prevention Study Group. JAMA 1995; 274: Kam PC, Tan KH. Pruritus-itching for a cause and relief? Anaesthesia 1996; 51: Benedetti C, Butler SH. Systemic analgesics. In: Bonica JJ, ed. The Management of Pain, Vol. II, 2nd edn. Philadelphia: Lea & Febiger, 1990: Jadad AR, Moore RA, Carroll D et al. Assessing the quality of reports of randomised clinical trials: is blinding necessary. Control Clin Trials 1996; 17: Morris JA, Gardner MJ. Calculating confidence intervals for relative risk, odds ratio and standardised ratios and rates. In: Altman DG, Gardner MJ, (eds) Statistics with Confidence Confidence Intervals and Statistical Guidelines. London: BMJ Books, 1995: Gavaghan DJ, Moore RA, McQuay HJ. An evaluation of homogeneity tests in meta-analyses in pain using simulations of individual patient data. Pain 2000; 85: McQuay HJ, Moore RA. Using numerical results from systematic reviews in clinical practice. Ann Intern Med 1997; 126: Cook RJ, Sackett DL. The number needed to treat: a clinically useful measure of treatment effect. Br Med J 1995; 310: Altman DG. Confidence intervals for the number needed to treat. Br Med J 1998; 317: Moher D, Cook DJ, Eastwood S, Olkin I, Rennie D, Stroup DF. Improving the quality of reports of meta analyses of randomized controlled trials: the QUOROM statement. Lancet 1999; 354: Borgeat A, Wilder-Smith OH, Saiah M, Rifat K. Subhypnotic doses of propofol possess direct antiemetic properties. Anesth Analg 1992; 74: Borgeat A, Stirnemann HR. Ondansetron is effective to treat spinal or epidural morphine-induced pruritus. Anesthesiology 1999; 90: Colbert S, O Hanlon DM, Galvin S, Chambers F, Moriarty DC. The effect of rectal diclofenac on pruritus in patients receiving intrathecal morphine. Anaesthesia 1999; 54:

11 356 F. Kjellberg and M. R. Tramèr 14 Davies GG, From R. A blinded study using nalbuphine for prevention of pruritus induced by epidural fentanyl. Anesthesiology 1988; 69: Douglas MJ, Kim JH, McMorland GH. The effect of epinephrine in local anaesthetic on epidural morphineinduced pruritus. Can Anaesth Soc J 1986; 33: Gal TJ, DiFazio CA. Prolonged antagonism of opioid action with intravenous nalmefene in man. Anesthesiology 1986; 64: Gambling DR, Howell P, Huber C, Kozak S. Epidural butorphanol does not reduce side effects from epidural morphine after cesarean birth. Anesth Analg 1994; 78: Joshi GP, Duffy L, Chehade J, Wesevich J, Gajraj N, Johnson ER. Effects of prophylactic nalmefene on the incidence of morphine-related side effects in patients receiving intravenous patient-controlled analgesia. Anesthesiology 1999; 90: Lawhorn CD, McNitt JD, Fibuch EE, Joyce JT, Leadley RJ. Epidural morphine with butorphanol for postoperative analgesia after cesarean delivery. Anesth Analg 1991; 72: Raisch DW, Barreuther AD, Osborne RC. Evaluation of a non-food drug administration-approved use of cimitidine: Treatment of pruritus resulting from epidural morphine analgesia. Ann Pharmacotherapy 1991; 25: Alhashemi JA, Crosby ET, Grodecki W, Duffy PJ, Hull KA. Treatment of intrathecalmorphine-induced pruritus following cesarean section. Can J Anaesth 1997; 44: Cohen SE, Ratner EF, Kreitzman TR, Archer JH, Mignano LR. Nalbuphine is better than naloxone for treatment of side effects after epidural morphine. Anesth Analg 1992; 75: Kendrick WD, Woods AM, Daly MY, Birch RF, DiFazio C. Naloxone versus nalbuphine infusion for prophylaxis of epidural morphine-induced pruritus. Anesth Analg 1996; 82: Saiah M, Borgeat A, Wilder-Smith OH, Rifat K, Suter PM. Epidural-morphine-induced pruritus: propofol versus naloxone. Anesth Analg 1994; 78: Somrat C, Oranuch K, Ketchada U, Siriprapa S, Thipawan R. Optimal dose of nalbuphine for treatment of intrathecal-morphine induced pruritus after cesarean section. J Obstet Gynaecol Res 1999; 25: Colbert S, O Hanlon DM, Chambers F, Moriarty DC. The effect of intravenous tenoxicam on pruritus in patients receiving epidural fentanyl. Anaesthesia 1999; 54: Dailey PA, Brookshire GL, Shnider SM, Abboud TK, Kotelko DM, Noueihid R. The effects of naloxone associated with the intrathecal use of morphine in labor. Anesth Analg 1985; 64: Rawal N, Schött U, Dahlström B et al. Influence of naloxone infusion on analgesia and respiratory depression following epidural morphine. Anesthesiology 1986; 64: Gowan JD, Hurtig JB, Fraser RA, Torbicki E, Kitts J. Naloxone infusion after prophylactic epidural morphine: effect on incidence of postoperative side-effects and quality of analgesia. Can J Anaesth 1988; 35: Penning JP, Samson B, Baxter AD. Reversal of epidural morphine-induced respiratory depression and pruritus with nalbuphine. Can J Anaesth 1988; 35: Lawhorn CD, Brown Jr RE. Epidural morphine with butorphanol in pediatric patients. JClinAnesth1994; 6: Bailey AG, Valley RD, Freid EB, Alhoun P. Epidural morphine combined with epidural or intravenous butorphanol for postoperative analgesia in pediatric patients. Anesth Analg 1994; 79: Beilin Y, Bernstein HH, Zucker Pinchoff B, Zahn J, Zenzen WJ. Subhypnotic doses of propofol do not relieve pruritus induced by intrathecal morphine after cesarean section. Anesth Analg 1998; 86: Warwick JP, Kearns CF, Scott WE. The effect of subhypnotic doses of propofol on the incidence of pruritus after intrathecal morphine for caesarean section. Anaesthesia 1997; 52: Abboud TK, Afrasiabi A, Davidson J et al. Prophylactic oral naltrexone with epidural morphine: effect on adverse reactions and ventilatory responses to carbon dioxide. Anesthesiology 1990; 72: Abboud TK, Lee K, Zhu J et al. Prophylactic oral naltrexone with intrathecal morphine for cesarean section: effects on adverse reactions and analgesia. Anesth Analg 1990; 71: Camann WR, Minzter BH, Denney RA, Datta S. Intrathecal sufentanil for labor analgesia. Effects of added epinephrine. Anesthesiology 1993; 78: Connelly NR, Rahimi A, Parker RK. Nalmefene or naloxone for preventing intrathecal opioid mediated side effects in cesarean delivery patients. Int J Obstet Anesth 1997; 6: Etchin A, Perl A, Bider D, Ben Rafael Z. Prevention of a side effect of epidural morphine by epidural steroid administration in cesarean section. Gynecol Obstet Invest 1990; 29: Gan TJ, Ginsberg B, Glass PS, Fortney J, Jhaveri R, Perno R. Opioid-sparing effects of a low-dose infusion of naloxone in patient-administered morphine sulfate. Anesthesiology 1997; 87: Grattidge P. Nausea and vomiting after major arthroplasty with spinal anaesthesia including morphine: a randomised trial of subhypnotic propofol infusion as prophylaxis. Acta Anaesthesiol Scand 1998; 42: Horta ML, Horta BL. Inhibition of epidural morphineinduced pruritus by intravenous droperidol. Reg Anesth 1993; 18: Horta ML, Ramos L, da Goncalves ZR, et al. Inhibition of epidural morphine-induced pruritus by intravenous

12 Pharmacological control of opioid-induced pruritus 357 droperidol. The effect of increasing the doses of morphine and of droperidol. Reg Anesth 1996; 21: Juneja MM, Ackerman WE, Bellinger K. Epidural morphine pruritus reduction with hydroxyzine in parturients. J Ky Med Assoc 1991; 89: Kyriakides K, Hussain SK, Hobbs GJ. Management of opioid-induced pruritus: a role for 5-HT3 antagonists? Br J Anaesth 1999; 82: Morgan PJ, Mehta S, Kapala DM. Nalbuphine pretreatment in cesarean section patients receiving epidural morphine. Reg Anesth 1991; 16: Naji P, Farschtschian M, Wilder Smith OH, Wilder Smith CH. Epidural droperidol and morphine for postoperative pain. Anesth Analg 1990; 70: Parker RK, Holtmann B, White PF. Patient-controlled epidural analgesia: Interactions between nalbuphine and hydromorphone. Anesth Analg 1997; 84: Paech MJ. The influence of adrenaline on postoperative analgesia after subarachnoid morphine. Anaesth Intensive Care 1993; 21: Sanansilp V, Areewatana S, Tonsukchai N. Droperidol and the side effects of epidural morphine after cesarean section. Anesth Analg 1998; 86: Thind GS, Wells JC, Wilkes RG. The effects of continuous intravenous naloxone on epidural morphine analgesia. Anaesthesia 1986; 41: Törn K, Tuominen M, Tarkkila P et al. Effects of subhypnotic doses of propofol on the side effects of intrathecal morphine. Br J Anaesth 1994; 73: Vercauteren MP, Vandeput DM, Meert TF, Adriaensen HA. Patient-controlled epidural analgesia with sufentanil following caesarean section: The effect of adrenaline and clonidine admixture. Anaesthesia 1994; 49: Wang JJ, Ho ST, Tzeng JI. Comparison of intravenous nalbuphine infusion versus naloxone in the prevention of epidural morphine-related side effects. Reg Anesth Pain Med 1998; 23: Ernst E, Pittler MH. Re-analysis of previous meta-analysis of clinical trials of homeopathy. J Clin Epidemiol 2000; 53: Tramèr MR, Reynolds DJ, Moore RA, McQuay HJ. When placebo controlled trials are essential and equivalence trials are inadequate. Br Med J 1998; 317: Henzi I, Sonderegger J, Tramèr MR. Efficacy, doseresponse, and adverse effects of droperidol for prevention of postoperative nausea and vomiting. Can J Anaesth 2000; 47: Macario A, Weinger M, Carney S, Kim A. Which clinical anesthesia outcomes are important to avoid? The perspective of patients. Anesth Analg 1999; 89: Picard P, Tramèr MR. Prevention of pain on injection with propofol: quantitative systematic review. Anesth Analg 2000; 90: