Patient-controlled analgesia (PCA) is widely used as a mode of primary postoperative analgesia in children.1 This delivery method uses a computer-controlled pump activated by pressing a button or squeezing a “trigger” to initiate delivery of a predetermined bolus of IV opioid analgesic. The advantage of this type of analgesic delivery compared with continuous IV infusion or intermittent nurse-administered injections is that the patient can titrate the opioid delivery to achieve satisfactory pain control. Randomized controlled trials, mainly in adults, have demonstrated that PCA pain control is superior, with greater patient satisfaction and a similar incidence of opioid-related adverse events.2
In addition to the PCA bolus dose, a “background” infusion of opioid was traditionally administered to minimize fluctuations in plasma opioid levels, which could result in periods of inadequate pain relief—for example, when patients fall asleep and awake as a consequence of pain. However, several retrospective observational studies involving almost 10,000 adult patients showed an increased incidence of not only mild or moderate adverse events, such as nausea and sedation, but also severe adverse events, such respiratory depression with hypoxemia, with no increase in efficacy, with the addition of a background infusion.3–6 The recently updated guidelines for perioperative pain management from the American Society of Anesthesiologists do not recommend the use of a background infusion, because there appears to be no additional analgesic benefit.7 In contrast, PCA bolus dosing with background infusion is commonly used in the pediatric setting. A survey of pediatric pain management in the United States found that 29% of institutions routinely used background infusions when administering IV PCA; 63% offered it on a case-by-case basis, and 8% never used a continuous infusion.1 Hospitals with a pediatric pain service were more likely to routinely use a basal infusion (41% vs 12%), whereas those without a pediatric pain service were more likely never to use a basal infusion (17% vs 2%).
We conducted a systematic review and meta-analysis of randomized trials to assess whether the addition of a background infusion to bolus administration of an opioid analgesic via a PCA device is more effective, in terms of patient-reported pain scores, without a significant increase in the risk of adverse events than PCA bolus-alone in the postoperative pediatric population.
We registered our study protocol with the international prospective register of systematic reviews with health-related outcomes (PROSPERO, CRD42013005797).16
We conducted a comprehensive search of Medline, Embase, and the Cochrane Central Register of Controlled Trials (CENTRAL) for relevant studies reported from database inception to January 2015. Search terms included extensive controlled vocabulary and keyword searches for (randomized controlled trials) AND (patient-controlled analgesia) AND (children) AND (pain) OR (opioid consumption). The search strategies are available upon request. We searched the gray literature for registered and ongoing trials included in the meta-Register of Controlled Trials and ClinicalTrials.gov. In addition, we searched the reference lists of review articles and included articles for additional trials not identified by our electronic search of the primary databases. There were no language restrictions.
We included randomized controlled studies comparing PCA bolus-alone with PCA bolus plus background infusion for postoperative analgesia in children (0–18 years) and adolescents (13–21 years), undergoing any form of surgery that used patient-reported pain scores or opioid consumption as a primary or secondary outcome measure. If potentially eligible trials did not report either pain scores or opioid consumption, the trial was ineligible. Two reviewers independently screened the titles and abstracts of articles. The full text of any title or abstract deemed potentially eligible by either reviewer was retrieved. Two reviewers then independently assessed the eligibility of each full-text article, and any disagreements were resolved by consensus and intervention of a third author if required.
Data Extraction and Quality Assessment
Two reviewers independently extracted data from the list of included studies using standardized, pretested, data extraction forms with accompanying instructions. We extracted data on patient demographics, interventions, comparators, study results, and study methodology. Data were extracted from graphs or figures when necessary. Discrepancies were resolved by discussion. A third reviewer, not otherwise involved in the data abstraction process, made the final decision if reviewers failed to reach consensus.
Using the Cochrane Risk of Bias instrument, 2 reviewers independently assessed the risk of bias, including the likelihood of selection, performance, detection, attrition, and reporting bias17 for each outcome. For each category, the risk was classified as “low,” “unclear,” or “high.” Published guidelines for categorizing each of the domains as low, unclear, or high risk of bias were used.18 An overall risk of bias (low or unclear/high) was then assigned to each outcome.
We also independently rated the overall quality of evidence (certainty in effect estimates) for each outcome using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach, in which randomized trials begin as high-quality evidence but may be rated down by one or more for each of 5 categories of limitations: risk of bias, inconsistency, indirectness, imprecision, and reporting bias.19,20
Data Synthesis and Analysis
Data were analyzed using the RevMan Analyses statistical package in Review Manager, version 5.3 (Cochrane Collaboration, Copenhagen, Denmark). All outcomes were pooled using a random-effects model because there may be heterogeneity within the eligible trials related to variation among populations, interventions, and outcomes.
Data for pain scores were reported using different pain instruments, thus precluding pooled estimates as a weighted mean difference (WMD). Therefore, we pooled the data as a standardized mean difference (SMD) with corresponding 95% confidence intervals (CIs). To interpret the results in the context of the minimal clinically important difference (MCID), we used methods endorsed by OMERACT and the Cochrane Collaboration for reporting pain in meta-analyses for converting continuous data on different instruments to the mean difference in natural units (Visual Analogue Scale 0–10).21,22 To obtain the standard deviation values for each of the scales, we determined the median standard deviation of the control groups from the included studies.
For opioid consumption, data from each study were converted to the same units (µg/kg/h of IV morphine) to allow for results to be presented as a WMD with corresponding 95% CIs. Doses of opioids were converted to IV morphine equivalents according to the Canadian Compendium of Pharmaceuticals and Specialties 2013.23
Adverse events reported as dichotomous outcomes (nausea and vomiting, incidence of sedation) were pooled as a relative risk with corresponding 95% CI. For postoperative nausea and vomiting, the most recent Consensus Guidelines recommend antiemetic prophylaxis if the risk increases from approximately 10% in the low-risk group to 30% in the moderate-risk group.24 Therefore, we considered the MCID for the incidence of postoperative nausea and vomiting (PONV) to be 20%. To our knowledge, the MCID for the incidence of excessive sedation in children has not been reported. Therefore, we considered the same difference in incidence of 20% to be clinically important. Data for the remaining adverse event outcomes (oxygen desaturation, nighttime sleep, pruritus, patient/parent satisfaction) were not amenable to pooling for various reasons, including the use of different measurement instruments between studies or a lack of reported variance data.
Statistical heterogeneity was investigated using the χ2 test and the I 2 statistic, where I 2 was defined as the percentage of total variability from study to study accounted for by heterogeneity rather than chance. We used previously published guidelines for quantifying heterogeneity (low I 2 = 25%–49%, moderate = 50%–74%, high >75%).25
Analysis of Subgroups
Subgroup analyses included dose of background infusion and risk of bias. For background dose, based on clinical experience of the investigators, we classified the reported infusion as either “low dose” (≤15 μg/kg/h) or “high dose” (>15 μg/kg/h). A priori, we hypothesized that (1) high-dose background infusions would result in reduced pain scores, greater total opioid consumption, and a higher incidence of opioid-related adverse events compared with no background infusion, and (2) low-dose background infusions would also result in reduced pain scores, greater total opioid consumption, but no difference in opioid-related adverse events, compared with no background infusion.16 For our risk of bias subgroup analysis, a priori we hypothesized that studies at high or unclear risk were more likely to favor PCA bolus plus background infusion for the reduction in pain.16
Two additional confounding factors that were identified during the peer-review process as potential sources of heterogeneity included the type of surgery and the pain measurement scale (4-point versus >10-point) used. Therefore, we performed a post hoc subgroup analysis to explore the effect of each factor on the reduction of pain with the addition of a background infusion.
For each subgroup, we tested for interaction using a χ2 significance test and, if a subgroup proved significant, we planned to apply published criteria to evaluate the credibility of the subgroup effects.26
Our search yielded 810 citations. After removing 305 duplicates, we were left with 505 citations. After independent screening of the titles and abstracts, 32 studies were selected for full-text review. Of these, independent review yielded 7 eligible randomized trials (338 patients) for inclusion (Table 1).8,9,11–15 Figure 1 illustrates the results of our literature search. Patients’ ages ranged from 5 to 20 years, and all underwent abdominal (general or urologic) or orthopedic (limb or spine) surgical procedures. None of the included studies described the use of acetaminophen or nonsteroidal antiinflammatory drugs (NSAIDs). In Doyle et al,9 2 of the 3 groups received a low-dose (4 and 10 μg/kg/h, respectively) background infusion but were otherwise similar (Table 1). In accordance with the Cochrane guidelines, and to avoid a unit of analysis error, we combined these 2 groups and compared them with the bolus-alone group.18
Patient-Reported Pain Scores
Pain scores were assessed as a continuous variable in 6 of 7 included trials. Our analysis for pain was based on the most frequently reported time points, 4, 12, and 24 hours after surgery (Table 2). Because 4-hour pain scores may be confounded by anesthetics or opioids given in the postanesthesia care unit, we chose pain scores at 12 (±4) and 24 (±4) hours for analysis.
Pain scores after 12 hours included 5 trials8,9,12,13,15 (203 patients) and were not significantly different for the PCA bolus plus background infusion group compared with the PCA bolus-alone group (SMD, 0.06; 95% CI, −1.02 to 1.15; Figure 2A). Subgroup analysis did not show any statistically significant difference in pain scores between the PCA bolus plus low-dose background infusion group (n = 121; SMD, 0.40; 95% CI, −1.65 to 2.45) and the PCA bolus plus high-dose background infusion group (n = 82; SMD, −0.40; 95% CI, −0.84 to 0.04) compared with the PCA bolus-alone group.
Pain scores at 24 hours included 5 trials8,9,12,13,15 (n = 203) and did not demonstrate a statistically significant difference with the addition of a background infusion (SMD, −0.22; 95% CI, −0.86 to 0.41; Figure 2B). Subgroup analysis did not show any statistically significant difference in pain scores between the PCA bolus plus low-dose background infusion group (n = 121; SMD, −0.33; 95% CI, −1.47 to 0.80) and the PCA bolus plus high-dose background infusion group (n = 82; SMD, −0.08; 95% CI, −0.66 to 0.51) compared with the PCA bolus-alone group.
We compared the pain scores after conversion to a VAS 0 to 10 scale to the reported MCID of ±1.27 At 12 hours, the mean (95% CI) difference was −0.27 (−0.96 to 0.43), where a negative value favors the control/no background group (Figure 3). This result is not statistically significant but may possibly represent harm with treatment (background infusion) because the lower limit of the 95% CI approaches −1. At 24 hours, the mean (95% CI) difference was 0.26 (−0.38 to 0.90), which is not statistically significant and approaches, but does not reach, clinical significance favoring the background infusion group (Figure 3).
The high degree of heterogeneity for pain scores at 12 and 24 hours (I 2 values of 92% and 80%, respectively) was not explained by our a priori subgroup analysis of high-dose versus low-dose background infusion rate (P = 0.46 and P = 0.70, respectively). Therefore, additional post hoc subgroup analyses based on trials using different pain scales (4-point instruments versus >10-point instruments) and types of operation (open appendectomy versus lower limb orthopedic versus spinal surgery) were performed. Based on the test of interaction, there was no significant difference between the subgroups at 12 and 24 hours according to different pain scales (SMD, 0.06; 95% CI, −1.02 to 1.15; P = 0.2; Figure 4, A and B). With respect to types of operation, pain scores favored a background infusion at 12 and 24 hours for lower limb orthopedic surgery. However, only 1 study12 with 36 patients populated this subgroup, and the overall test of interaction for subgroup effect was significant only at 24 hours (P = 0.002; Figure 5, A and B).
Two trials reported pain outcomes as categorical data on a 3-point scale and were not included in our pooled pain analysis. The first study reported the average pain score over the first 24 hours, as well as the percentage of patients with mild, moderate, or severe pain aggregated over the first 48 hours.11 Although there was no difference in average pain scores among those with PCA plus low-dose background infusion versus no infusion over the first 24 hours, the incidence of moderate and severe pain scores over 48 hours was significantly lower in the background infusion group (15 μg/kg/h of morphine). The second study reported no difference in the proportion of patients who experienced none, mild, or moderate/severe pain with the addition of a high-dose background infusion of morphine at 20 μg/kg/h.14
Seven studies recorded opioid consumption and were included in our pooled estimate (Table 3).8,9,11–15 Overall, there was no difference in opioid consumption with PCA bolus plus background infusion (WMD, 2.31; 95% CI, −3.77 to 8.39; Figure 6). When subgroup analysis was performed, no differences in opioid consumption were shown between the PCA bolus plus high-dose (WMD, 5.35; 95% CI, −0.41 to 11.11) and the low-dose (WMD, −1.12; 95% CI, −14.87 to 12.64) background infusion groups compared with the PCA bolus-alone group (Figure 6).
The high degree of heterogeneity for opioid consumption (I 2 value of 88%) was not explained by our a priori subgroup analysis of high-dose versus low-dose background infusion rate (P = 0.40).
Nausea and Vomiting
Seven studies reported data for nausea and/or vomiting, but the incidence (%) or number of patients who experienced nausea and/or vomiting during the study was reported in only 5 trials (239 patients randomly allocated; Table 3).8,9,12,14,15 The incidence of PONV in the background infusion group was 28% (35/127) versus 19% (21/112) in the no background infusion group, the difference being less than the MCID of 20% (Figure 7). The pooled analysis showed that there was no difference in the risk of nausea and/or vomiting between PCA bolus-alone versus PCA bolus plus background infusion (risk ratio [RR], 1.20; 95% CI, 0.78 to 1.84; Figure 7). The risks of nausea and/or vomiting in 2 studies with PCA bolus plus low-dose background infusion, and in the 3 studies with PCA bolus plus high-dose background infusion, were not significantly different. The 2 studies not included in the pooled analysis found no difference in the occurrence of nausea and/or vomiting with the addition of a background infusion.11,13
There was a low degree of statistical heterogeneity for nausea and/or vomiting (I 2 value of 0%), with no significant difference between our a priori subgroups of high-dose versus low-dose background infusion rate (P = 0.50).
The incidence of excessive sedation was reported in 4 trials (161 patients),8,9,12,13 and the pooled analysis demonstrated no significant increase among patients with the addition of a low-dose background infusion (RR, 3.48; 95% CI, 0.41 to 29.29; Figure 8). Two studies8,13 reported results but had no events in either group and thus were not estimable. The difference in the incidence of excessive sedation between the background and the no background infusion groups was less than the MCID of 20% (4.6% vs 0%, respectively; Figure 8).
Arterial Oxygen Desaturation
Two studies reported the total number of desaturation events (SpO2 < 94%) per group.8,9 In both studies, the use of a background infusion (10 and 20 μg/kg/h) was associated with significantly more episodes of desaturation than PCA bolus-alone or PCA bolus plus background infusion at 4 μg/kg/h. We were not able to pool these studies because the authors reported the number of events rather than patients. In both studies, there were more events than patients, and after multiple attempts to contact the authors, we were unable to clarify the total number of patients with desaturation. McNeely and Trentadue12 noted that the cumulative percentage time with SpO2 < 91% was significantly greater in the background infusion group than the bolus-only group (6% vs 3.3%; P < 0.05). Weldon et al15 reported no significant difference in the incidence of SpO2 < 90% during 5 separate time periods.
Four studies8,9,14,15 (203 patients) reported on sleep. Three studies suggested significantly greater nighttime sleep in the PCA bolus plus background groups,8,9,14 whereas 1 study, Weldon et al,15 found no difference.
Two studies (82 patients) presented results for pruritus (Table 3)13,15: Weldon et al15 found no difference in the incidence of pruritus over 5 time periods (0–8, 8–16, 16–24, 24–48, and 48–72 hours) and Yildiz et al13 described no occurrence of pruritus in either group.
Berde et al11 reported higher patient satisfaction scores with PCA bolus plus background infusion than PCA bolus-alone (Table 3). Two studies reported qualitative results that described equivalent satisfaction with either technique.12,15
Risk of Bias and Quality of the Evidence
All 7 included studies were deemed to be at an overall high or unclear risk of bias (Table 4). Therefore, analysis of outcomes according to low versus unclear/high risk of bias subgroups could not be performed. Using GRADE methodology, the certainty in the estimates of effect for pain scores at 12 and 24 postoperative hours was deemed to be of very low quality (ie, we are very uncertain about the estimate; Table 5). Reasons for downgrading the score include the unclear or high risk of bias present in all the studies, the limited number of patients observed among included studies (small sample size), and inconsistency (high degree of heterogeneity). The evidence for nausea and/or vomiting was deemed to be of low quality (ie, further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate) because of small sample sizes and the unclear or high risk of bias present in all the studies. As for the incidence of excessive sedation, the evidence was deemed to be of very low quality because of the unclear risk of bias, wide 95% CI surrounding the RR, and small sample size.
Summary of Main Results
The results of this systematic review and meta-analysis suggest that the addition of an opioid background infusion to PCA for postoperative analgesia does not result in a significant reduction in patient-reported pain scores at 12 and 24 hours after surgery or produce an increased risk of patient-important adverse events, such as nausea and/or vomiting and sedation. Not only were the results not statistically significant, none reached the threshold of an MCID. Using GRADE methods, we independently rated the quality of evidence as low to very low for pain reduction and adverse events. The quality of evidence was poor, generally owing to issues of imprecision (too few patients randomly assigned and a sparse number of events observed) and risk of bias (unclear reporting of allocation concealment and blinding). We also found no significant difference in opioid consumption among those randomly assigned to background infusion versus no infusion.
Two studies reported results that were in contrast to the others. The combined background infusion group (4 and 10 μg/kg/h) from Doyle et al9 showed higher pain scores compared with the bolus-only group at 12 and 24 hours. However, within the study itself, there was no significant difference in pain scores between either of the 2 background infusion groups and the bolus group at any of the 4-hour time intervals. The increased morphine consumption in the 10 μg/kg/h group likely accounted for the significantly greater number of hypoxemic episodes and incidence of nausea and vomiting. In contrast, the 4 μg/kg/h group had fewer hypoxemic episodes than either of the other 2 groups, and a similar number of emetic episodes but more nighttime sleep than the bolus-only group. In the second study, Yildiz et al13 reported more opioid consumption in the bolus-only group. This result is not surprising because the bolus dose in the background infusion group was 50% less than that in the bolus-only group, and the background infusion was relatively low (equivalent to 2 μg/kg/h of morphine).
Overall Completeness and Applicability of Evidence
Although most studies included in this meta-analysis addressed our critically important outcomes (pain scores, adverse events), significant gaps in knowledge still remain because of the limited number of children randomized, the large degree of heterogeneity that cannot be accounted for through a priori subgroup analyses, and the risk of bias of the included studies. In particular, gaps in outcome data that need to be addressed include subcategorization of pain scores into “rest” and “movement” to better characterize the quality of pain management under different “real-life” conditions, quantitative data for the incidence of pruritus, and patient/parent satisfaction with pain management.
Although the results of this meta-analysis are relevant to current clinical practice, their applicability may be limited for numerous reasons. First, many of the surgical techniques, such as open appendectomy, are used much less commonly now. Second, because morphine was used in 6 of 7 studies, it is unknown whether the results are applicable to other opioids, such as hydromorphone. Studies comparing the analgesic efficacy and adverse event profiles of PCA with different opioids have shown minimal or no significant difference, which suggests that the results of this study may be applicable to other opioids.28–32 However, larger randomized controlled studies comparing PCA bolus-alone with PCA bolus plus background infusion for each opioid would be required to adequately address this issue. Third, none of the included studies used a multimodal analgesic approach, which may include the addition of agents such as acetaminophen and NSAIDs, for postoperative pain management. Fourth, the number of patients observed was small (n = 338) and heterogeneous (many different types of abdominal and orthopedic surgeries, wide range of ages), thus limiting the strength of evidence and conclusions.
Potential Biases and Limitations in the Review Process
Our review has limitations. First, although we attempted to obtain unpublished or raw data, which would have allowed for a more in-depth analysis, only 1 of 7 corresponding authors of the included studies responded and could not provide any additional information. Second, the doses we chose for the low and high background infusion rates were based on clinical practice standards at our institution and previous studies9,33 but may not reflect practice at other institutions. Last, the number of studies and patients included in our analysis were very small and variable, thereby increasing the risk of type 1 errors, in particular. We could have chosen to increase the CI to 99%; however, all our results, with the exception of the 1 study in the Lower Limb Orthopedic subgroup for pain scores at 12 and 24 hours, were nonsignificant at the 95% CI. Thus, increasing the CI to 99% would not have changed the clinical implications of any of the meta-analyses.
Agreements and Disagreements with Other Studies or Reviews
To our knowledge, there are no systematic reviews or meta-analyses examining the use of PCA with a background infusion in children. The majority of the individual studies included in the analysis found an increase in opioid consumption, minimal or no reduction in pain scores, and an increase in opioid-related adverse events with the addition of a background infusion to PCA bolus doses.
Although pooling of the data does not demonstrate a reduction in pain scores with the addition of a background infusion, benefit may apply only to operations associated with significant postoperative pain (eg, thoracotomy, laparotomy, and osteotomies). The types of operations in the studies analyzed were quite variable and included open appendectomy, various urologic procedures, lower extremity osteotomies and/or soft tissue procedures, and spinal fusions. Some of these operations may not have resulted in significant amounts of postoperative pain, which may reduce the potential benefit of the addition of a background infusion to bolus doses. Based on suggestions during the peer-review process, we performed a post hoc subgroup analysis to explore the benefits of a background infusion according to the type of operation. Although pain scores favored a background infusion for the lower limb orthopedic surgery, this subgroup included only one study (36 patients), and the effect was significant only at 24 hours postoperatively. Therefore, we consider these findings to be hypothesis generating, and future trials may wish to stratify study groups according to the type of surgery. The post hoc subgroup analysis we performed for the type of pain scale found no effect, although the 11-point numerical rating scale has been shown to be more sensitive in detecting clinically important differences.34
The benefits of a background infusion may also be “time-limited” (ie, helpful only for the first 24 hours when the pain is most severe). The effect of the addition of a background infusion on the occurrence of adverse events is variable and would depend on many factors, such as dose, duration, type of drug, and the use of prophylactic medications to prevent adverse events (antiemetic, antipruritic).
According to GRADE precision guidelines, a sample size of at least 300 is required to be certain in the pooled estimates. Based on all available data, 203 patents have been randomly assigned and followed for pain intensity across all studies, suggesting that at least 97 patients need to be further randomly assigned. Therefore, further rigorous trials are needed to adequately address these issues. Such trials should include children and adolescents undergoing an operation associated with significant postoperative pain who are randomly assigned to a PCA bolus plus background infusion (low-dose and high-dose) group(s) versus PCA bolus-alone group, with all groups receiving multimodal analgesic medications (acetaminophen and NSAIDs, if not contraindicated). Methodologic features should include an adequate sample size (eg, 130 patients based on α = 0.05, β = 0.20, δ = 1 for the minimal important difference for pain reduction,27 and SD = 2 from the results from the included trials) with central randomization and blinding of patients and outcome assessors. Critically important outcome measurements should include patient-reported pain scores at multiple time points over the first 48 hours using a validated, age-appropriate, pain scale; opioid consumption standardized to IV morphine equivalents; adverse events such as nausea and vomiting, excessive sedation, and pruritus using a validated instrument; and, if possible, patient/parent satisfaction scores. The trial should be registered a priori (eg, ClinicalTrials.gov) to avoid selective reporting bias.
We did not show a significant difference in patient-reported pain scores, opioid consumption, or increased risk of adverse events with the addition of an opioid background infusion to PCA bolus doses of opioid. However, the quality of the evidence and certainty in the estimates of effect from the included studies are low to very low, and further high-quality studies are necessary.
Name: Jason Hayes, MD.
Contribution: This author helped designed the study, collect and analyze data, and prepare the manuscript.
Name: Jeremiah J. Dowling, MD.
Contribution: This author helped collect and analyze data and prepare the manuscript.
Name: Arie Peliowski, MD.
Contribution: This author helped collect and analyze the data.
Name: Mark W. Crawford, MBBS.
Contribution: This author helped prepare the manuscript.
Name: Bradley Johnston, PhD.
Contribution: This author helped design the study, analyze the data, and prepare the manuscript.
This manuscript was handled by: James DiNardo, MD.
1. Nelson KL, Yaster M, Kost-Byerly S, Monitto CL. A national survey of American Pediatric Anesthesiologists: patient-controlled analgesia and other intravenous opioid therapies in pediatric acute pain management. Anesth Analg 2010;110:754–60.
2. Hudcova J, McNicol E, Quah C, Lau J, Carr DB. Patient controlled opioid analgesia versus conventional opioid analgesia for postoperative pain. Cochrane Database Syst Rev 2006:CD003348.
3. Schug SA, Torrie JJ. Safety assessment of postoperative pain management by an acute pain service. Pain 1993;55:387–91.
4. Fleming BM, Coombs DW. A survey of complications documented in a quality-control analysis of patient-controlled analgesia in the postoperative patient. J Pain Symptom Manage 1992;7:463–9.
5. Sidebotham D, Dijkhuizen MR, Schug SA. The safety and utilization of patient-controlled analgesia. J Pain Symptom Manage 1997;14:202–9.
6. Notcutt WG, Morgan RJ. Introducing patient-controlled analgesia for postoperative pain control into a district general hospital. Anaesthesia 1990;45:401–6.
7. American Society of Anesthesiologists Task Force on Acute Pain Management. Practice guidelines for acute pain management in the perioperative setting: an updated report by the American Society of Anesthesiologists Task Force on Acute Pain Management. Anesthesiology 2012;116:248–73.
8. Doyle E, Robinson D, Morton NS. Comparison of patient-controlled analgesia with and without a background infusion after lower abdominal surgery in children. Br J Anaesth 1993;71:670–3.
9. Doyle E, Harper I, Morton NS. Patient-controlled analgesia with low dose background infusions after lower abdominal surgery in children. Br J Anaesth 1993;71:818–22.
10. Doyle E, Mottart KJ, Marshall C, Morton NS. Comparison of different bolus doses of morphine for patient-controlled analgesia in children. Br J Anaesth 1994;72:160–3.
11. Berde CB, Lehn BM, Yee JD, Sethna NF, Russo D. Patient-controlled analgesia in children and adolescents: a randomized, prospective comparison with intramuscular administration of morphine for postoperative analgesia. J Pediatr 1991;118:460–6.
12. McNeely JK, Trentadue NC. Comparison of patient-controlled analgesia with and without nighttime morphine infusion following lower extremity surgery in children. J Pain Symptom Manage 1997;13:268–73.
13. Yildiz K, Tercan E, Dogru K, Ozkan U, Boyaci A. Comparison of patient-controlled analgesia with and without a background infusion after appendicectomy in children. Paediatr Anaesth 2003;13:427–31.
14. Skues MA, Watson DM, O’Meara M, Goddard JM. Patient-controlled analgesia in children. A comparison of two infusion techniques. Paediatr Anaesth 1993;3:223–8.
15. Weldon BC, Connor M, White PF. Pediatric PCA: the role of concurrent opioid infusions and nurse-controlled analgesia. Clin J Pain 1993;9:26–33.
17. Akl EA, Sun X, Busse JW, Johnston BC, Briel M, Mulla S, You JJ, Bassler D, Lamontagne F, Vera C, Alshurafa M, Katsios CM, Heels-Ansdell D, Zhou Q, Mills E, Guyatt GH. Specific instructions for estimating unclearly reported blinding status in randomized trials were reliable and valid. J Clin Epidemiol 2012;65:262–7.
18. Higgins JPT, Green S.Cochrane Handbook for Systematic Reviews of Interventions. 2011Chichester, West Sussex, England: Wiley.
19. GRADEpro GDT. GRADEpro Guideline Development Tool. 2015. Available at: gradepro.org
. Accessed April 1, 2015.
20. Guyatt GH, Oxman AD, Kunz R, Vist GE, Falck-Ytter Y, Schünemann HJ; GRADE Working Group. What is “quality of evidence” and why is it important to clinicians? BMJ 2008;336:995–8.
21. Busse JW, Bartlett SJ, Dougados M, Johnston BC, Guyatt GH, Kirwan JR, Kwoh K, Maxwell LJ, Moore A, Singh JA, Stevens R, Strand V, Suarez-Almazor ME, Tugwell P, Wells GA. Optimal strategies for reporting pain in clinical trials and systematic reviews: recommendations from an OMERACT 12 Workshop. J Rheumatol 2015;42:1962–70.
22. Johnston BC, Patrick DL, Thorlund K, Busse JW, da Costa BR, Schünemann HJ, Guyatt GH. Patient-reported outcomes in meta-analyses-part 2: methods for improving interpretability for decision-makers. Health Qual Life Outcomes 2013;11:211.
23. CPS. CPS 2013: Compendium of Pharmaceuticals and Specialties. 1st ed.2013; 1593Ottawa, Ontario, Canada: Canadian Pharmacists Association.
24. Gan TJ, Diemunsch P, Habib AS, Kovac A, Kranke P, Meyer TA, Watcha M, Chung F, Angus S, Apfel CC, Bergese SD, Candiotti KA, Chan MT, Davis PJ, Hooper VD, Lagoo-Deenadayalan S, Myles P, Nezat G, Philip BK, Tramèr MR; Society for Ambulatory Anesthesia. Consensus guidelines for the management of postoperative nausea and vomiting. Anesth Analg 2014;118:85–113.
25. Peters JL, Sutton AJ, Jones DR, Abrams KR, Rushton L. Comparison of two methods to detect publication bias in meta-analysis. JAMA 2006;295:676–80.
26. Sun X, Briel M, Walter SD, Guyatt GH. Is a subgroup effect believable? Updating criteria to evaluate the credibility of subgroup analyses. BMJ 2010;340:c117.
27. Voepel-Lewis T, Burke CN, Jeffreys N, Malviya S, Tait AR. Do 0-10 numeric rating scores translate into clinically meaningful pain measures for children? Anesth Analg 2011;112:415–21.
28. Vetter TR. Pediatric patient-controlled analgesia with morphine versus meperidine. J Pain Symptom Manage 1992;7:204–8.
29. Hong D, Flood P, Diaz G. The side effects of morphine and hydromorphone patient-controlled analgesia. Anesth Analg 2008;107:1384–9.
30. Felden L, Walter C, Harder S, Treede RD, Kayser H, Drover D, Geisslinger G, Lötsch J. Comparative clinical effects of hydromorphone and morphine: a meta-analysis. Br J Anaesth 2011;107:319–28.
31. Collins JJ, Geake J, Grier HE, Houck CS, Thaler HT, Weinstein HJ, Twum-Danso NY, Berde CB. Patient-controlled analgesia for mucositis pain in children: a three-period crossover study comparing morphine and hydromorphone. J Pediatr 1996;129:722–8.
32. Woodhouse A, Hobbes AF, Mather LE, Gibson M. A comparison of morphine, pethidine and fentanyl in the postsurgical patient-controlled analgesia environment. Pain 1996;64:115–21.
33. White I, Ghinea R, Avital S, Chazan S, Dolkart O, Weinbroum AA. Morphine at “sub-analgesic” background infusion rate plus low-dose PCA bolus control pain better and is as safe as twice a bolus-only PCA regimen: a randomized, double blind study. Pharmacol Res 2012;66:185–91.
© 2016 International Anesthesia Research Society
34. Kwong WJ, Pathak DS. Validation of the eleven-point pain scale in the measurement of migraine headache pain. Cephalalgia 2007;27:336–42.