Professional Documents
Culture Documents
meta分析 NSAIDs节省阿片
meta分析 NSAIDs节省阿片
doi: 10.1093/bja/aew391
Review Article
Abstract
Background. Morphine, and analgesics other than morphine (AOM), are commonly used to treat postoperative pain after major
surgery. However, which AOM provides the best efficacy-safety profile remains unclear.
Methods. Randomized trials of any AOM alone or any combination of AOM compared with placebo or another AOM in adults
undergoing major surgery and receiving morphine patient-controlled analgesia were included in a network meta-analysis.
The outcomes were morphine consumption, pain, incidence of nausea, vomiting at 24 h and severe adverse effects.
Results. 135 trials (13,287 patients) assessing 14 AOM alone or in combination were included. For all outcomes, comparisons with pla-
cebo were over-represented. Few trials assessed combinations of two AOM and none the combination of three or more. Network meta-
analysis found morphine consumption reduction was greatest with the combination of two AOM (acetaminophen þ nefopam,
acetaminophen þ NSAID, and tramadol þ metamizol): -23.9 (95% CI -40;-7.7), -22.8 (-31.5;-14) and -19.8 (35.4;-4.2) mg per 24 h, respec-
tively. For AOM used alone, morphine consumption reduction was greatest with a-2 agonists, NSAIDs, and COX-2 inhibitors. When con-
sidering the risk of nausea, NSAIDs, corticosteroids and a-2 agonists used alone were the most efficacious (OR 0.7 [95% CI: 0.6-0.8], 0.36
[0.18-0.79], 0.41 [0.15-.64], respectively). The paucity of severe adverse effects data did not allow assessment of efficacy-safety balance.
Conclusions. A combination of aetaminophen with either an NSAID or nefopam was superior to most AOM used alone, in
reducing morphine consumption. Efficacy was best with three AOM used alone (a-2 agonists, NSAIDs and COX-2 inhibitors)
and least with tramadol and acetaminophen. There is insufficient trial data reporting adverse events.
Clinical trial registration. PROSPERO: CRD42013003912.
Key words: analgesics; balanced analgesia; postoperative pain; surgery; systematic review
C The Author 2016. Published by Oxford University Press on behalf of the British Journal of Anaesthesia. All rights reserved.
V
For Permissions, please email: journals.permissions@oup.com
22
Non-opioid analgesics in adults after major surgery | 23
Apfel classification.24 26 27 Secondary outcomes also included eFig. 1). Tables 1 and 2 present the characteristics of the selected
severe adverse events (SAEs) (as defined in each trial). trials (detailed description in Supplementary data Table 2).
Intention-to-treat analysis was used whenever available. About half of the trials concerned abdominal surgery and about
one quarter gynaecologic surgery. All included trials were pub-
Review of network geometry lished between 1986 and 2014.
The overall risk of bias was low for 36 (27%) trials, high for 13
We examined the pattern of comparisons among the different (10%) and unclear for 86 (63%). More specifically, the risk of
AOM within the network of trials (i.e. we assessed the geometry selection bias (sequence generation and allocation conceal-
of the networks for each outcome separately) and produced net- ment) was low for 43% of trials and was unclear for 57%. The
work graphs with nodes representing the competing AOM and risk of performance bias (blinding of patients and personnel)
two nodes linked together by an edge, if at least one trial com- and detection bias (blinding of outcome assessors) was low for
pared the two corresponding AOM. We examined the connec- 62% and 57% of trials, respectively. Finally, 70% of trials featured
tions between AOM (i.e. which of the considered treatments low risk of bias regarding incomplete outcome data
were compared head-to-head in trials and which were con- (Supplementary data eFig. 2). Detailed assessments of the risk of
nected only indirectly by one or more “common comparators” bias for each trial are in Supplementary data eTable 3.
and the amount of evidence informing each comparison.
Table 2 Trial characteristics by analgesic monotherapy. Data are number (percentage) unless stated otherwise. Data for number of age,
weight and gender were not available for 15, 27, and 19 trials, respectively. DDD, defined daily dose; NSAIDs, nonsteroidal anti-inflamma-
tory drugs; COX-2, cyclooxygenase.
NSAIDs 72 (84) 32 (38) 28 (33) 2007 (1987-2014) 19 (31) 50 [18.1-66.4] 71.6 [53-96.6] 73% [3-100]
COX-2 inhibitor 37 (97) 26 (67) 19 (49) 2007 (2001-2013) 10 (40) 52.3 [32.3-68.5] 69.5 [54.7-93] 51% [0-100]
Acetaminophen 19 (70) 3 (12) 8 (31) 2008 (1995-2013) 7 (29) 60.5 [42.3-55] 55.8 [51.4-65] 70% [55-93]
Corticosteroids 6 (55) 10 (91) 11 (100) 2004 (1999-2008) 1 (20) 47.3 [25.5-62.6] 73.3 [65-85.7] 61% [30-100]
A-2 agonist 8 (80) 7 (70) 3 (30) 2005 (1992-2008) 1 (10) 48.8 [27-73.3] 76.8 [57.8-88.7] 63% [40-80]
Tramadol 4 (33) 5 (42) 5 (42) 2001 (1992-2008) 2 (16) 47.5 [42-53] 69.1 [58.4-74] 73% [36-100]
Nefopam 5 (56) 2 (22) 1 (11) 2003 (1990-2010) 1 (11) 57.8 [46-73] 70.9 [62.5-77] 71% [63-77]
Metamizol 3 (100) 0(0) 0 (0) 2009 (1996-2009) 1 (33) 54.3 [52-58.5] 75.8 [75.8-75.8] 50% [18-100]
COX-2 inhibitor
10
(n=2161)
10 Corticosteroid
(n=377)
25
5
Nefopam
(n=260)
6
2 Placebo
(n=4843)
Nefopam+NSAID 2
(n=28) 60 Acetaminophen+NSAID
2 (n=85)
3
20
3
NSAID
(n=3259) 8
Acetaminophen+metamizol
(n=23)
2
Acetaminophen
(n=699) Metamizol
(n=97)
Acetaminophen+nefopam
(n=38)
Fig 1 Network geometry for trials reporting treatment effect for morphine consumption.
26 | Martinez et al.
frequently evaluated; 15 trials assessed six different combina- sparing effect was significantly greater for NSAIDs than cortico-
tions of two AOM. No trial assessed the combination of three or steroids. Some combinations were associated with a signifi-
more AOM (Fig. 1). The difference in the reporting of outcomes cantly superior effect: acetaminophen þ NSAIDs was superior to
varied by the AOM: the ratio between the number of trials tramadol, nefopam, corticosteroids, and metamizol, and acet-
reporting nausea and that reporting morphine consumption aminophen þ nefopam was superior to corticosteroids
ranged from 0% to 100% (Supplementary data eFig. 3). (Supplementary data eTable 4).
Pain
Synthesis of results
Pain was significantly decreased with two AOM used alone
Morphine consumption and pain at 24 h (NSAIDs and COX-2 inhibitors) and two associations of AOM
Pooled analysis revealed that morphine consumption was sig- (acetaminophen þ NSAID and tramadol) as compared with pla-
nificantly lower with six AOM administered alone (tramadol, cebo, with mean reductions ranging from 5.2 to 23 mm/100 at
nefopam, acetaminophen, NSAIDs, COX-2 inhibitors, and a2- 24 h (Fig. 3). The treatment effects did not reach a clinically
adrenergic agonists) as compared with placebo, with mean meaningful level for any AOM for pain (Supplementary data
reductions ranging from 7.4 to 14.6 mg per 24 h; mean reduction eTable 4).
was 20 mg per 24 h with three associations of AOM
(acetaminophen þ NSAIDs, acetaminophen þ nefopam, and tra- Postoperative nausea and vomiting
madol þ metamizol) (Fig. 3). In the absence of direct compari- The risk of PONV was significantly decreased with three AOM
sons, network meta-analysis revealed that the morphine- used alone (NSAIDs, corticosteroids, a-2 agonists) as compared
Non-opioid analgesics in adults after major surgery | 27
Morphine Pain
Treatment Mean difference [95%CI] Mean difference [95%CI]
Nausea Vomiting
Treatment Odds ratio [95%CI] Odds ratio [95%CI]
Acetaminophen+metamizol
Acetaminophen+nefopam
Tramadol+NSAID
0.10 0.20 0.25 0.33 0.50 1.0 2.0 0.10 0.20 0.25 0.33 0.50 1.0 2.0
Fig 3 Forest plots of network meta-analysis estimates of analgesics other than morphine vs placebo for morphine consumption, pain, nausea, and vomiting.
with placebo. Across all drugs, the ORs ranged from 0.36 to 0.89 vomiting) (Fig. 3). Use of a-2 agonists and corticosteroids did not
for nausea and 0.26 to 0.98 for vomiting. The largest reduction differ in reducing the risk of PONV. a-2 agonists were signifi-
was obtained with corticosteroids (OR 0.36 for nausea and 0.26 cantly superior to NSAIDs for nausea, and corticosteroids were
for vomiting) and a-2 agonists (OR 0.41 for nausea and 0.44 for significantly superior to NSAIDs for vomiting. In the absence of
28 | Martinez et al.
direct comparisons, network meta-analysis revealed that corti- for SAE. Furthermore, the differential reporting of SAEs varied
costeroids were significantly superior to COX-2 inhibitors in pre- greatly by the AOM tested: the ratio between the number of tri-
venting nausea and to COX-2 inhibitors, nefopam, NSAIDs, and als reporting SAEs and reporting morphine consumption ranged
acetaminophen in preventing vomiting (Supplementary data from 0% for a2 agonists, tramadol, nefopam, metamizol and all
eTable 4). Evidence was not sufficient to show reduced risk of combinations, to 36% for COX-2 inhibitors. Third, the trials of
PONV with any association of AOM. combinations of AOM were few. Only 1.5% of all trials evaluated
the efficacy of the combination of two AOM and no trial eval-
Serious adverse events uated the combination of three AOM. All combinations deserve
The network of trials reporting treatment effects on serious more investigation, especially those involving the most effective
adverse events was sparse, which precluded synthesizing data. analgesics, such as NSAIDs and a-2 agonists, which have never
In conventional random-effects meta-analysis, evidence was been assessed. This lack of evidence contrasts greatly with clini-
insufficient for increased risk of SAEs with NSAIDs (nine trials, cal practice. National and European surveys reported the use of
OR 1.28 [95% CI 0.65;2.53], with low between-trial heterogeneity, more than two analgesics for 30% to 75% of patients.32 33
I2 ¼ 10%), acetaminophen (two trials, OR 3.09 [95% CI 0.33;28.60], Altogether, 30 yr of clinical research on the subject has not met
with moderate heterogeneity, I2 ¼ 18%), and COX-2 inhibitors clinicians’ needs, who still need an answer to which AOM or
(five trials, OR 2.85 [95% CI 0.75;10.83], with substantial heteroge- AOM combination has the best efficacy/safety balance?
neity, I2 ¼ 79%) (Supplementary data eFig. 4).
Relative effectiveness
with the most recent conventional MAs of clonidine and dexme- substantial between-trial heterogeneity, particularly for the
detomidine, which reported a reduction in morphine consump- morphine consumption outcome, which was reported previ-
tion, pain, nausea and vomiting as compared with placebo.36 37 ously in conventional meta-analyses.13 17 36–38 43 There is grow-
A limitation of the currently available evidence is the lack of ing evidence that morphine consumption depends more on
information on a-2 agonists regarding SAEs, which needs to be individual pain vulnerability rather than on surgical trauma.
addressed before recommending such prescription in practice. Indeed, several factors are considered to explain pain vulner-
Finally, we found that dexamethasone produced the largest ability such as genetics, previous morphine consumption, pre-
reduction in PONV incidence, with no morphine-sparing effect. operative chronic pain, and psychological factors.44–49 None of
These results confirm that dexamethasone at commonly used these factors was monitored and evaluated in our dataset.
dosages should be considered more as an antiemetic than as a Finally, our work focused only on analgesics and did not
real analgesic in the postoperative setting.38 Other analgesics account for other drug classes or methods often used in associa-
including acetaminophen, tramadol, metamizol and nefopam tion with the AOM, such as antihyperalgesic drugs and/or
showed smaller benefit. As was previously shown, acetamino- regional analgesia. Future analyses may cover larger networks
phen and nefopam should not be recommended alone with of evidence including other classes. An additional limitation of
morphine.7 11 14 16 The combination of tramadol, a weak opioid, our syntheses is the frequent lack of direct evidence from head-
with a strong opioid did not confer a significant benefit, which to-head trials. As a consequence, the network meta-analysis
agrees with a recent meta-analysis.17 Metamizol is not widely estimates rely on indirect information. However, our explora-
used because it has been banned in many countries because of tion of inconsistency did not show evidence of discrepancy
Supplementary material 10. Elia N, Lysakowski C, Tramer MR. Does multimodal analgesia
with acetaminophen, nonsteroidal antiinflammatory drugs,
Supplementary material is available at British Journal of or selective cyclooxygenase-2 inhibitors and patient-
Anaesthesia online. controlled analgesia morphine offer advantages over mor-
phine alone? Meta-analyses of randomized trials.
Anesthesiology 2005; 103: 1296–304
Declaration of interest 11. Evans MS, Lysakowski C, Tramer MR. Nefopam for the pre-
V.M. has received payments and travel funding for lectures vention of postoperative pain: quantitative systematic
from Jansen, Pfizer, Astellas, D.F. has received payments and review. Br J Anaesth 2008; 101: 610–7
travel funding for lectures from Grunenthal, Biocodex, 12. Marret E, Kurdi O, Zufferey P, Bonnet F. Effects of nonsteroi-
Mundipharma. H.B., E.M., P.R. and L.T. have no interest dal antiinflammatory drugs on patient-controlled analgesia
declared. morphine side effects: meta-analysis of randomized con-
trolled trials. Anesthesiology 2005; 102: 1249–60
13. Maund E, McDaid C, Rice S, Wright K, Jenkins B, Woolacott N.
Funding Paracetamol and selective and non-selective non-steroidal
anti-inflammatory drugs for the reduction in morphine-
This study was funded by the institute UPSA de la douleur (IUD). related side-effects after major surgery: a systematic review.
IUD had no role in the design and conduct of the study; the col- Br J Anaesth 2011; 106: 292–7
lection, management, analysis, and interpretation of the data;
29. Rücker G. Network meta-analysis, electrical networks and 39. McQuay HJ, Poon KH, Derry S, Moore RA. Acute pain: combi-
graph theory. Res Synth Methods 2012; 3: 312–24 nation treatments and how we measure their efficacy. Br J
30. Dias S, Welton NJ, Sutton AJ, Caldwell DM, Lu G, Ades AE. Anaesth 2008; 101: 69–76
Evidence synthesis for decision making 4: inconsistency in 40. De Oliveira GS Jr, Agarwal D, Benzon HT. Perioperative single
networks of evidence based on randomized controlled trials. dose ketorolac to prevent postoperative pain: a meta-
Med Decis Making 2013; 33: 641–56 analysis of randomized trials. Anesth Analg 2012; 114: 424–33
31. Dias S, Welton NJ, Caldwell DM, Ades AE. Checking consis- 41. Mills EJ, Ghement I, O’regan C, Thorlund K. Estimating the
tency in mixed treatment comparison meta-analysis. Stat power of indirect comparisons: a simulation study. PLoS One
Med 2010; 29: 932–44 2011; 6: e16237
32. Benhamou D, Viel E, Berti M, et al. [PATHOS study on postop- 42. Thorlund K, Mills EJ. Sample size and power considerations
erative pain management in Europe: French data]. Ann Fr in network meta-analysis. Syst Rev 2012; 1: 41.
Anesth Reanim 2008; 27: 664–78 43. Trinquart L, Ioannidis JP, Chatellier G, Ravaud P. A test for
33. Fletcher D, Fermanian C, Mardaye A, Aegerter PA. Patient- reporting bias in trial networks: simulation and case studies.
based national survey on postoperative pain management BMC Med Res Methodol 2014; 14: 112
in France reveals significant achievements and persistent 44. Gerbershagen HJ, Pogatzki-Zahn E, Aduckathil S, et al.
challenges. Pain 2008; 137: 441–51 Procedure-specific risk factor analysis for the development
34. Chen LC, Elliott RA, Ashcroft DM. Systematic review of the anal- of severe postoperative pain. Anesthesiology 2014; 120:
gesic efficacy and tolerability of COX-2 inhibitors in post- 1237–45