Prevention of pain on injection of propofol: systematic review and meta-analysisBMJ 2011; 342 doi: https://doi.org/10.1136/bmj.d1110 (Published 15 March 2011) Cite this as: BMJ 2011;342:d1110
- Leena Jalota, visiting research fellow1,
- Vicki Kalira, medical student12,
- Elizabeth George, staff research associate1,
- Yung-Ying Shi, staff research associate1,
- Cyrill Hornuss, research fellow1,
- Oliver Radke, attending physician13,
- Nathan L Pace, professor14,
- Christian C Apfel, associate professor1
- On behalf of the Perioperative Clinical Research Core
- 1Department of Anesthesia and Perioperative Care, University of California at San Francisco, San Francisco 94115, CA, USA
- 2West Virginia School of Medicine, Morgantown, WV, USA
- 3Department of Anaesthesia and Critical Care Medicine, University Hospital Carl Gustav Dresden at the TU Dresden, Dresden, Germany
- 4Department of Anaesthesiology, University of Utah, Salt Lake City, UT, USA
- Correspondence to: C C Apfel or
- Accepted 17 January 2011
Objective To systematically determine the most efficacious approach for preventing pain on injection of propofol.
Design Systematic review and meta-analysis.
Data sources PubMed, Embase, Cochrane Library, www.clinicaltrials.gov, and hand searching from the reference lists of identified papers.
Study selection Randomised controlled trials comparing drug and non-drug interventions with placebo or another intervention to alleviate pain on injection of propofol in adults.
Results Data were analysed from 177 randomised controlled trials totalling 25 260 adults. The overall risk of pain from propofol injection alone was about 60%. Using an antecubital vein instead of a hand vein was the most effective single intervention (relative risk 0.14, 95% confidence interval 0.07 to 0.30). Pretreatment using lidocaine (lignocaine) in conjunction with venous occlusion was similarly effective (0.29, 0.22 to 0.38). Other effective interventions were a lidocaine-propofol admixture (0.40, 0.33 to 0.48); pretreatment with lidocaine (0.47, 0.40 to 0.56), opioids (0.49, 0.41 to 0.59), ketamine (0.52, 0.46 to 0.57), or non-steroidal anti-inflammatory drugs (0.67, 0.49 to 0.91); and propofol emulsions containing medium and long chain triglycerides (0.75, 0.67 to 0.84). Statistical testing of indirect comparisons showed that use of the antecubital vein and pretreatment using lidocaine along with venous occlusion to be more efficacious than the other interventions.
Conclusions The two most efficacious interventions to reduce pain on injection of propofol were use of the antecubital vein, or pretreatment using lidocaine in conjunction with venous occlusion when the hand vein was chosen. Under the assumption of independent efficacy a third practical alternative could be pretreatment of the hand vein with lidocaine or ketamine and use of a propofol emulsion containing medium and long chain triglycerides. Although not the most effective intervention on its own, a small dose of opioids before induction halved the risk of pain from the injection and thus can generally be recommended unless contraindicated.
Propofol is the drug of choice for induction of anaesthesia in millions of patients every year because of its rapid onset and short duration of action, easy titration, and favourable profile for side effects.1 Despite these positive attributes, about three out of five patients experience pain on injection of propofol, with one of these patients reporting severe or excruciating pain. Some patients recall the induction of anaesthesia as the most painful part of the perioperative period. As a result several interventions have been investigated to alleviate the pain associated with propofol injection. A systematic review in 2000 suggested pretreatment using lidocaine (lignocaine) in conjunction with venous occlusion as the most effective intervention.2 Despite that recommendation the technique failed to gain widespread popularity, possibly because of the time needed to apply the tourniquet. As a result the pain associated with injection of propofol remains a challenge and more than 100 new studies have explored additional and alternative strategies. These include novel propofol emulsions,3 4 modified emulsions, and microemulsion formulations,5 6 7 as well as diverse drugs and their combinations. We summarised all the available evidence from trials that compared the use of any drug or non-drug interventions (or combinations) with an active or inactive control in adults receiving intravenous propofol.
This qualitative systematic review included studies published up to December 2010. We searched PubMed, Cochrane Library, and Embase using the search terms “propofol” AND (“injection pain” OR “pain on injection”). We limited our search to clinical trials and randomised controlled trials (see web extra 1 for details of search strategy).
To identify all available evidence we identified additional relevant randomised controlled trials by hand searching the reference lists of the original papers until no further relevant references could be found. We also searched reviews on pain associated with propofol injection for similar randomised controlled trials. Although we applied no language restrictions, the only relevant studies were in English, German, and Japanese.
To minimise data duplication as a result of multiple reporting we compared papers from the same author. In addition, we searched www.clinicaltrials.gov for studies. Two authors (LJ and VK) screened and retrieved reports and excluded irrelevant studies. Relevant data were extracted by one author (VK) and checked by another (LJ). Additional investigators (CCA, OR, and NLP) participated in the review process when uncertainty about eligibility criteria arose. From each study we extracted details on patients’ characteristics (adults only), use of non-drug interventions (for example, site of venous cannulation, speed of injected propofol, temperature of injected propofol), use of analgesic interventions, and use of combinations of interventions (see web extra 2 for characteristics of included studies).
Selection of studies for review
Selected studies included all randomised controlled trials that compared the use of any drug or non-drug intervention, or a combination, with an active or inactive control, and reported the response rate and severity of pain in adults receiving intravenous propofol. All included studies had numerical data presented in the text or a table; if data were not presented as such, we extracted the information from the graphs if the scale allowed a sufficiently precise estimation. We included all studies that met the eligibility criteria, regardless of language of publication.
Assessment of risk of bias
We assessed risk of bias in each of four domains in studies meeting the inclusion criteria: adequate sequence generation, adequate concealment of allocation, adequate blinding, and completeness of reporting data on outcomes (see web extra 2). The specific domains of risk of bias were graded as “yes” for low risk, “unclear,” and “no” for high risk. As more than 95% of the primary studies were designed to search for pain reduction during anaesthesia induction with propofol, selective outcome reporting bias was considered unlikely and not assessed.
Meta-analyses were carried out by direct comparisons of intervention versus control (pairwise) and indirect comparisons between the network of interventions shown to be significant individually. The primary outcome was the number of patients reporting any pain (pain response rate); the effect size was the relative risk. We did not carry out meta-analyses of pain scores (for example, numerical or verbal rating scales) because they were reported both rarely and inconsistently. For studies with multiple intervention groups, we partitioned the count of events and patients in the control group into two or more control groups within any meta-analysis to avoid a unit of analysis error. Similarly, for the studies participating in the indirect comparisons, we partitioned the comparator group according to how many times it was used for indirect comparisons (across meta-analyses). The summary relative risks and 95% confidence intervals were estimated using a random effects Mantel-Haenszel method in RevMan 5.0 (Cochrane Collaboration). Statistical heterogeneity was assessed by the I2 value. If interventions involved 10 or more studies, we used funnel plots to visualise small study effects or reporting bias; asymmetry was tested using the arcsine transformation and method of moments linear regression implemented in the R package meta (R Foundation for Statistical Computing, Austria).10 We considered P values less than 0.05 and relative risks not crossing the identity line as statistically significant.
We analysed the network of randomised controlled trials within an indirect comparison framework using previously described models11 and implemented in frequentist mixed effects metaregression12; we selected only interventions that significantly reduced pain by a direct intervention comparison with six or more included studies. The summary statistic was the relative risk, with 95% confidence intervals. The common comparator was the placebo or control group. The moderators in the mixed effects models were the interventions entered as categorical covariates. Assumptions in this analysis included a sufficient homogeneity of the different trials, treatment effects (logRR) distributed normally around a typical value, and the same residual heterogeneity (τ2) among the moderators. This analysis was carried out using the R package metafor using restricted maximum likelihood estimation (see web extra 3 for details of the model). We adjusted the test statistics of individual estimates of moderator variables and omnibus hypotheses of all moderators by the method of Knapp and Hartung (t and F distributions).13 Residual heterogeneity was assessed by χ2 tests. As the methods of estimation are different, the relative risk values from RevMan and metafor differ slightly.
A search of PubMed, Embase, and Cochrane databases identified 674 potentially relevant papers (fig 1⇓), of which 427 were excluded: 83 of the 249 identified through PubMed (three were duplicates and 80 did not measure pain from propofol injection), 229 of the 283 identified through Embase (183 were identified in the previous search and 46 did not measure pain from propofol injection), and 105 of the 142 identified through the Cochrane databases (98 were previously identified and seven did not measure pain from propofol injection). In addition to the 257 potential studies a further 24 were identified after hand searching references of relevant papers, and two were from the US government clinical trials website (www.clinicaltrials.gov). Thus, 283 studies were retrieved as potential clinical trials for further evaluation. A further 106 studies were excluded for the following reasons: reviews (seven studies),1 3 4 14 15 16 17 not carried out in humans (n=2),18 19 not randomised controlled trials (n=15),20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 improper assessment of pain (n=7),35 36 37 38 39 40 41 duplicate publication (n=2),42 43 methodological concerns (n=15),44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 did not measure pain on injection of propofol (n=20),60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 intervention not aimed at pain reduction on injection of propofol (n=3),80 81 82 incomplete data or inability to extract data (n=17),83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 and studies in children (n=18).100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 Thus 177 studies were included in the analysis.
Overall, a low risk of bias was identified for adequate sequence generation in 40% of included studies (n=71), adequate allocation concealment in 43% (n=76), blinding in 85% (n=151), and whether incomplete outcome data were addressed in 88% (n=156).
Thus this systematic review includes data from 25 260 adults (177 randomised controlled trials). The average trial size was 142 patients (range 24 to 388). Nineteen drugs and eight different non-drug interventions and combinations were tested (see web extra figure). About 60% of patients in the control group reported pain on injection of propofol alone. Trials reported pain scores rarely and on different scales. Therefore this analysis is based exclusively on the response rate of pain.
Because of the wide variety of interventions investigated, three categories of studies were established: non-drug interventions, drug interventions and their combinations, and both drug and non-drug interventions. Each category was further divided into several subcategories. Finally, subanalyses were carried out for interventions involving more than five studies.
Efficacy according to categories
The non-drug category consisted of studies that used mechanical interventions such as different infusion rates,118 119 120 venous occlusion,119 needle sizes,121 injection sites,122 123 124 125 126 microfiltration,127 128 temperature,90 129 130 131 132 133 134 135 136 137 and bacteriostatic saline.138 The most efficacious intervention in this subcategory was selection of an antecubital vein compared with a hand vein as the injection site (relative risk 0.14, 95% confidence interval 0.07 to 0.30; table 1⇓ and fig 2⇓).119 123 124 125 126 139 Conversely, non-effective interventions were cold propofol (4ºC), propofol at room temperature, venous occlusion by itself, and modifying the speed of the intravenous carrier fluid (table 2⇓).
The drug category comprised various drugs or drug combinations (table 3⇓). The studies were divided into 19 subcategories based on drug class—for example, antiemetics, local anaesthetics, benzodiazepines, barbiturates. Most of these drugs were partially successful in reducing the risk of pain from propofol injection.
A lidocaine-propofol admixture (25 trials) was the most effective intervention in this subcategory (0.40, 0.33 to 0.48, fig 3⇓).125 126 135 138 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 The funnel plot was, however, asymmetrical (arcsine transformation regression, t=−5.3, df=39, P<0.001) suggesting a strong small study effect or reporting bias for this intervention (fig 4⇓). No other funnel plots were asymmetrical. A lidocaine-propofol admixture was of similar efficacy to pretreatment with lidocaine alone (24 studies) (0.47, 0.40 to 0.56, fig 5⇓).70 119 125 146 150 159 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178
Pretreatment with opioids showed analgesic benefit (0.49, 0.41 to 0.59, fig 6⇓). Various opioids were studied: alfentanil (six studies),179 180 181 182 183 184 remifentanil (n=5),185 186 187 188 189 sufentanil (n=1),187 fentanyl (n=3),180 185 190 tramadol (n=3),70 161 163 meperidine (pethidine) (n=3),161 173 186 and butorphanol (n=1).191 All of these opioids were successful in reducing pain from propofol injection.
Pretreatment with non-steriodal anti-inflammatory drugs was also effective in seven trials (0.67, 0.49 to 0.91, fig 8⇓). Flurbiprofen, diclofenac, and ketorolac were the primary agents explored for potential reduction of pain from propofol injection.147 177 197 198 199 200 201
Modified propofol formulations containing medium and long chain triglycerides compared with formulations containing long chain triglycerides were effective in 24 trials (0.75, 0.67 to 0.84, fig 9⇓).5 6 151 177 197 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 Combining trials that studied various combinations of standard and modified emulsion formulations with lidocaine had a similar effect (0.61, 0.44 to 0.84).149 151 203 206 220 221 222 223 224 225 226 227
Combined drug and non-drug category
The combined drug and non-drug category incorporated non-drug techniques such as site of injection,119 123 124 126 alteration of temperature of propofol,90 132 133 135 and venous occlusion (table 4⇓). The most commonly studied intervention was venous occlusion in conjunction with various drugs, such as antiemetics, non-steroidal anti-inflammatory drugs, β blockers, lidocaine, and opioids; many combinations reduced the risk of pain from propofol injection. In this category pretreatment using lidocaine in conjunction with venous occlusion was the most effective intervention at preventing the pain from propofol injection (0.29, 0.22 to 0.38, fig 10⇓).133 142 152 177 228 229 230 231 232 233 234 235 236 One trial found that lidocaine applied ionophoretically was more effective than a sham application (0.31, 0.14 to 0.69).237 Three trials found statistically significant results with modifications of propofol’s temperature in combination with drugs such as lidocaine and nafamostat.132 133 136
Risk of bias assessment
Eight interventions statistically significantly reduced the pain from propofol injection. A sensitivity analysis to assess the potential effect of four criteria for the risk of bias assessment was carried out. When the point estimates or confidence intervals of the individual domains were compared with the overall point estimates, no appreciable difference occurred that would change the interpretation of the results (table 5⇓).
To be able to rank the interventions, a network approach was used to estimate indirect comparisons among effective interventions involving more than six studies.11 Indirect treatment comparisons were estimated for the eight pairwise (intervention versus control) statistically significant interventions; the data were derived from 167 treatment arms in 101 studies. These eight interventions were included as moderators in a mixed effects metaregression (table 6⇓). An omnibus test for inclusion of the moderators was significant (F=46.58, 159, P<0.001) and each individual regression coefficient was significant (t statistics, P<0.05 for all interventions). While the residual heterogeneity (τ2=0.10) remained significant (χ2=402, df=159, P<0.001), about 44% of the residual heterogeneity had been accounted for by the inclusion of the eight moderators in the model; the Akaike information criterion was also reduced in the full model.
The relative risk of using an antecubital vein was lower than for six of the other interventions, with the indirect relative risks ranging from 0.19 (modified propofol formulation) to 0.34 (lidocaine-propofol admixture). Pretreatment using lidocaine in conjunction with venous occlusion also had a lower relative risk than six of the other interventions, with the indirect relative risks ranging from 0.39 (modified propofol formulation) to 0.69 (lidocaine-propofol admixture). Although the indirect relative risk for using an antecubital vein compared with pretreatment using lidocaine in conjunction with venous occlusion was 0.50, the 95% confidence interval extended beyond the identity line.
The risk of pain was similarly reduced for five interventions (lidocaine-propofol admixture, and pretreatment with lidocaine, opioids, ketamine, and non-steroidal anti-inflammatory drugs), with direct relative risks varying from 0.43 to 0.67. Confidence intervals for nine of the 10 indirect relative risks between the five interventions were non-significant (table 6). Six interventions had lower indirect relative risks compared with a modified propofol formulation.
About 60% of patients experience pain on injection with standard propofol alone—that is, without any preventive measures. A previous systematic review and meta-analysis identified pretreatment using lidocaine (lignocaine) in conjunction with venous occlusion using a tourniquet to be the most efficacious intervention to reduce pain from propofol injection.2 Since then more than 100 randomised controlled trials have been published on the topic. Our systematic review and meta-analysis confirms the efficacy of the previously suggested technique (relative risk 0.29). However, selecting an antecubital vein instead of a hand vein was numerically the most efficacious intervention (relative risk 0.14). In addition, we identified six other efficacious interventions that are commonly used—namely, lidocaine-propofol admixture; pretreatment using lidocaine (without venous occlusion), opioids, non-steroidal anti-inflammatory drugs, or ketamine; and a propofol emulsion containing medium and long chain triglycerides. Furthermore, we carried out indirect comparisons across the meta-analyses and found that choosing the antecubital vein and venous occlusion along with pretreatment using lidocaine were similarly efficacious and clearly superior to the other six interventions.
The results of this analysis show that injection of propofol through an antecubital vein is one of the most effective interventions to reduce associated pain. From a physiological standpoint, differences in vein diameter, flow rate, and endothelial structure might account for the reduction in pain. Presuming that propofol is injected mid-stream into the lumen of the vein, the larger diameter of and faster flow rate through the antecubital vein will minimise the extent to which a high concentration of propofol comes into contact with the sensitive endothelial wall. Alternatively, propofol may be buffered more effectively when more blood is available to dissipate and mask the “full effect” of the bolus. Additionally, the composition of nociceptors along the endothelial wall might differ between the smaller veins of the hand and the larger antecubital veins.119 139 238 239 In contrast to careful selection of veins, other non-drug interventions—for example, both cold and warm propofol, adjusting the speed of intravenous carrier fluid, and microfiltration—were disappointingly ineffective approaches for alleviating pain from propofol injection.
The other similarly effective intervention was a combination of a drug and non-drug techniques—that is, pretreatment using lidocaine in conjunction with venous occlusion before injection. Although this has been considered the most efficacious technique, it has not become standard.2 A reason for this may be the additional procedural steps involved in the intervention, leading to some delay when swift induction is expected. In addition, venous occlusion has also been paired with many other drugs (for example, antiemetics, non-steroidal anti-inflammatory drugs, opioids) and was found to have some measurable success, albeit less so than when venous occlusion was combined with pretreatment using lidocaine. Although some of these combinations of interventions reached statistical significance, they were generally only investigated in a few studies, which makes it difficult to draw meaningful conclusions.
Although pretreatment of a hand vein using lidocaine in conjunction with proximal venous occlusion seems as effective as using an antecubital vein, clinicians may prefer the antecubital vein because it is an effective route and simple to use.
Similarly, of the drug interventions, a lidocaine-propofol admixture was similarly efficacious when compared with pretreatment using lidocaine alone. Both interventions were, however, considerably less efficacious than pretreatment with lidocaine in conjunction with venous occlusion. Interestingly, in the newer studies a trend was towards using a lidocaine-propofol admixture as opposed to propofol alone as the control group, suggesting that this clinical practice has become widely spread. Additionally, the funnel plot for the lidocaine-propofol admixture showed significant asymmetry (fig 4).10 240 Although this intervention with lidocaine may be efficacious, its treatment effect may be well overestimated.
Our analysis of almost 1500 patients showed that pretreatment with opioids resulted in a relative risk of about 0.50. Thus, unless contraindicated otherwise, it seems reasonable to use opioids as standard pretreatment several minutes before induction.
Multiple trials investigated a variety of propofol formulations, such as lipid-free formulations, modified emulsion formulations, and propofol containing bismuth. Of these, the most commonly studied emulsions, those containing medium and long chain triglycerides, were compared with the conventional emulsions containing long chain triglyceride (2344 patients, 24 studies), with a relative risk of 0.75 for emulsions containing medium and long chain triglycerides.
Other promising drug interventions were pretreatment with ketamine and with non-steroidal anti-inflammatory drugs. Pretreatment with either of these drugs should not only decrease the pain from propofol injection but also reduce postoperative pain, postoperative nausea and vomiting, and the need for postoperative opioids.241 242 However, diclofenac sodium is itself associated with pain on injection, which may lead to thrombophlebitis.62 243 Although this may be avoided by using a newer formulation, dilution, or slow intravenous infusion, the pain on injection using diclofenac limits its use for reducing the pain from propofol injection.
Based on the comparisons carried out here, it seems that among the wide arrays of interventions tested, eight had sufficient evidence of benefit. When interventions seem similarly efficacious, choices for intervention can be made on factors such as cost, personal choice, and simplicity of application.
Our results of direct and indirect comparisons suggest a possible strategy that is both efficacious and easy to apply in clinical practice (fig 11⇓). Since opioids are used commonly as part of a balanced anaesthesia protocol, it seems reasonable to use them as routine premedication in preparation for induction for all three options, as they halve the risk of pain from propofol injection (relative risk 0.50). We do not recommend the use of non-steroidal anti-inflammatory drugs as the results for these agents were heterogeneous and some themselves cause pain on injection. One approach could be to use an antecubital vein, whenever practicable, with its relative risk reduction of about 85%. Based on the assumption that interventions of independent pathways work independently,244 the risk of pain on injection is likely to be only 5% when preoperative opioids are combined with the antecubital approach (60%×0.49×0.14=4.1%). In other words, further interventions are unlikely to benefit more than 1 out of 20 patients, thereby additional interventions would only provide limited additional benefit from a clinical standpoint.
As an intravenous line in the antecubital vein may be occluded when the elbow is flexed, unintentional extravasations may not be detected as quickly as when the dorsum of the hand is used. Therefore when intravenous placement into the antecubital vein is challenging, we consider a fair alternative to be the hand vein with preoperative opioids plus lidocaine in conjunction with venous occlusion as this would also bring the risk down to less than 10% (60%×0.49×0.29=8.5%). Notably, a lidocaine-propofol admixture was also statistically significantly superior to placebo and is probably the most commonly used approach to reduce the pain from propofol injection. Owing to possible publication bias, however, the “true treatment effect” is unclear so we prefer similarly efficacious methods that have no evidence of publication bias. The other practical alternative could be to use preoperative opioids in conjunction with pretreatment using lidocaine or ketamine before the injection of a propofol emulsion containing medium and long chain triglycerides, thereby also reducing the risk of pain to about 10-12% (60%×0.49×0.47×0.75=10.3%) or 12% (60%×0.49×0.56×0.75=12.3%). Nevertheless, these estimates of multiplicative treatment effects are based on the assumption of independence and strictly speaking require confirmation in randomised controlled trials.
Limitations of the study
A range of other techniques reached statistical significance in a limited number of studies (often only one or two) and some of them lacked biological plausibility, such as the efficacy reported for antiemetics, cholinesterase inhibitors, antihistamines, stimulants, and combinations of interventions. Further research is needed to verify or refute these results and, if these interventions are truly efficacious, it will be essential to uncover underlying mechanisms. Furthermore, assessment of the intensity of pain score as an additional outcome was unachievable.
Unless contraindicated we recommend the routine use of a small dose of opioids before induction of anaesthesia using propofol injection in all patients. On the basis of efficacy and convenience we also recommend using an antecubital vein instead of a hand vein. If the hand vein is the site of injection, we recommend pretreatment using lidocaine in conjunction with venous occlusion, or a combined intervention such as pretreatment with ketamine or lidocaine before injection of a propofol emulsion containing medium and long chain triglycerides.
What is already known on this topic
Pretreatment with lidocaine (lignocaine) in conjunction with venous occlusion has been suggested as the best intervention to reduce pain from propofol injection
This technique failed to gain widespread popularity and the search for alternative interventions continues
What this study adds
Using an antecubital vein instead of a hand vein is a simple and effective way to avoid the pain from propofol injection
If the hand vein is chosen, pretreatment using lidocaine in conjunction with venous occlusion is equally efficacious, although not widely used
A third option could be the combination of “less efficacious interventions,” such as using a modified propofol emulsion in conjunction with pretreatment of the hand vein using lidocaine or ketamine
Cite this as: BMJ 2011;342:d1110
We thank Wolfgang Viechtbauer (School for Public Health and Primary Care, Maastricht University, the Netherlands) for the syntax of arguments used in the R package metafor; David Moher (Ottawa Hospital Research Institute, University of Ottawa) for proving critical input for the study; and Sakura Kinjo (University of California San Francisco) for translating the Japanese manuscripts. RevMan 5.0 (Review Manager, Cochrane Collaboration, UK) statistical package was used in this systematic review for analysis. Additionally, the R packages metafor and meta were used (R Development Core Team (2010). R: A language and environment for statistical computing. R Foundation for Statistical Computing).
Contributors: VK, LJ, YYS, OR, and CCA wrote the study protocol. LJ and VK collected and collated the data. CCA and NLP provided statistical expertise. CCA, LJ, and NLP analysed and interpreted the data. LJ, YYS, EG, CH, CCA, and NLP drafted the article. LJ, EG, YYS, CH, CCA, OR, and NLP critically revised the article for important intellectual content. LJ, EG, YYS, CH, CCA, and NLP provided administrative, technical, or logistic support. All authors approved the final article, had full access to all the data (including statistical reports and tables), and can take responsibility for the integrity of the data and the accuracy of the data analysis. CCA, LJ, and NLP are the guarantors.
Funding: This study was funded by the Perioperative Clinical Research Core, University of California, San Francisco.
Competing interests: All authors have completed the Unified Competing Interest form at www.icmje.org/coi_disclosure.pdf and declare that LJ, VK, EG, YYS, CH, OR, and CCA have support from the Perioperative Clinical Research Core, Department of Anaesthesia and Perioperative Care, University of California, San Francisco, CA for the submitted work; LJ, VK, EG, YYS, CH, NLP, OR, and CCA have no relationships that might have an interest in the submitted work in the previous 3 years; their spouses, partners, or children have no financial relationships that may be relevant to the submitted work; and LJ, VK, EG, YYS, OR, and CCA have no non-financial interests; and NLP is an unpaid statistical editor for a Cochrane Collaborative Review Group.
Ethical approval: Not required.
Data sharing: The technical appendix, statistical code, and dataset are available from the corresponding author at.
This is an open-access article distributed under the terms of the Creative Commons Attribution Non-commercial License, which permits use, distribution, and reproduction in any medium, provided the original work is properly cited, the use is non commercial and is otherwise in compliance with the license. See: http://creativecommons.org/licenses/by-nc/2.0/ and http://creativecommons.org/licenses/by-nc/2.0/legalcode.