Immediate Rescue Designs in Pediatric Analgesic Trials: A Systematic Review and Meta-analysis

CLINICAL trials to evaluate efficacy or effectiveness of analgesics pose ethical and scientific challenges for all ages, but especially for children. For sound scientific reasons, the standard approach to adult acute pain trials involves enrollment of patients with moderate to severe pain, randomization between active drug and placebo, and comparison of pain scores over time between active and placebo subjects as the primary measure of analgesic efficacy (fig. 1A).¹  The ethical basis of this approach rests on adults making informed decisions to bear the risk of assignment to a placebo group and potentially to experience continued pain during the study period. Comparative effectiveness trials involving no placebo group are relevant for guiding clinical decision-making, but they pose statistical problems for establishing drug efficacy.² 

For children, as vulnerable subjects, there is greater ethical concern about a significant risk of unrelieved pain in clinical trials, particularly where there are existing effective treatments. For ethical and practical reasons, pediatric analgesic trials have proceeded slowly. The U.S. Congress, the Food and Drug Administration, and the European Union have generated incentives for pharmaceutical companies to test analgesics in children, and some funding mechanisms via the Food and Drug Administration and the National Institutes of Health have supported a small number of investigator-initiated pediatric trials for off-patent analgesics that lack commercial incentives.³  Despite these initiatives, enrollment rates in pediatric analgesic trials have been very low.⁴ 

In November 2010, an expert consensus group was convened by the Anesthetic, Analgesic and Addiction Drugs section of the Food and Drug Administration to address pediatric analgesic trial design issues. Recommendations from this group were published in Pediatrics in 2012.¹  One prominent recommendation was to regard rescue-analgesic sparing as a pragmatic surrogate primary endpoint in pediatric analgesic trials. In order to maintain the scientific advantages of the traditional adult design, subjects could still be blinded and randomized between study drug and placebo. By providing immediate access to incremental rescue analgesia, especially via patient-controlled analgesia (PCA) or nurse-controlled analgesia (NCA) pumps, we expected that these designs would reduce the odds that subjects, especially those randomized to placebo, would experience unrelieved severe pain (fig. 1, B and C). This recommendation was based on committee members’ opinions and informal review of a small number of successful trials, but not on any systematic review or quantitative analysis.

With this background, we now attempt a systemic review and quantitative analysis of rescue analgesic designs in pediatric trials of four commonly used classes of analgesics for acute pain: opioids, nonsteroidal antiinflammatory drugs (NSAIDs), acetaminophen, and local anesthetics. Our hypotheses were (1) immediate opioid rescue designs provide reasonable assay sensitivity as pragmatic surrogate measures of analgesia efficacy, (2) surrogate efficacy effect sizes vary with drug class and study design, choice of opioid-sparing endpoints, method of rescue analgesic administration, and type of surgery, and (3) subjects randomized to either study drug or control/placebo groups with these designs had acceptably low mean pain scores (additional technical aspects and details relevant to researchers in the field can be found in Supplemental Digital Content 1, https://links.lww.com/ALN/B91).

The search strategy is described in detail in Section A1 of Supplemental Digital Content 1, https://links.lww.com/ALN/B91.

Studies were included if they met the following criteria: (1) randomized or controlled clinical trial; (2) children and adolescents aged ≤18 yr; (3) use of immediate rescue paradigms; and (4) assessed rescue medication and/or pain scores in postoperative pain setting.

For the purpose of this study, we included articles only if they (1) included placebo or control groups; (2) used IV opioids as rescue medication; and (3) used opioids, NSAIDs, acetaminophen, or local anesthetics as the “study drug,” which was tested for efficacy against a placebo or other control.

We chose to evaluate the following three analgesic sparing outcomes: (1) rescue opioid usage (milligram per kilogram per hour), (2) percentage of subjects requiring rescue medication, or (3) time to first rescue medication (minutes). No language restrictions were applied. In calculating rescue opioid usage, opioids were converted to morphine equivalents.^* In assessing the hourly usage, we chose the shortest reported timeframe over an hour and divided the morphine equivalent by this timeframe. For articles reporting pain scores, we standardized all the pain scores to a 0 to 10 scale.^5,6 

Data were extracted independently by two reviewers (J.K. and C.D.). Inconsistencies were resolved in consensus meetings. Additional study variables were extracted and tabulated, including age range of patients; sample sizes; type(s) of surgery; dose regimes for study drugs and rescue opioid, method of pain assessment; criteria for opioid administration; and duration of follow-up. Each included study was graded for quality and scored using the Jadad criteria.⁷ 

Two methods of assessing opioid sparing were used. First, we used a meta-analytic approach. We estimated Hedges’ g and the 95% CI between the study drug and control groups.⁸  Data management and calculations were performed using Comprehensive Meta-Analysis, version 2.0.^† Since considerable heterogeneity was expected, all analyses were performed with a random-effects model.⁹  We assessed the presence of publication bias using the fail-safe N method.¹⁰  To assess heterogeneity between studies, Q-statistics were calculated. A statistically significant Q indicates a heterogeneous distribution of odds ratios between studies, meaning that systematic differences, possibly influencing the results, are present.¹¹  In addition, the degree of inconsistency was quantified by the I ² statistic, which measures the percentage of variation across studies that is due to heterogeneity rather than chance.¹²  A 0% value means no heterogeneity, and higher values represent an increase in heterogeneity. Generally, heterogeneity is categorized at 25% (low), 50% (moderate), and 75% (high). Given the large number of possible factors influencing the standardized effect size, we chose not to conduct any sensitivity analyses to explain the heterogeneity, except for subgroup analyses with regard to type of surgery to assess its influence on assay sensitivity and choice of outcome.

Second, mean opioid sparing by the study drug was expressed as percent of maximum possible effect (placebo – study/placebo),¹³  as in pharmacologic studies.¹⁴ 

A third approach to data synthesis explored the distribution of pain scores in subjects randomized to study drug or control conditions. A “perfect rescue paradigm” should in theory result in low and nearly equal pain scores in both active and placebo groups (fig. 1C).¹  We wanted to evaluate the degree to which this aim is achieved in practice and to depict this balance between assay sensitivity and potential for unrelieved pain in control subjects. To do this we plotted each of these opioid-sparing endpoints against the mean pain scores for subjects in the control/placebo group. We refer to the relationship between these opioid-sparing endpoints and the pain in the controls as the “efficacy–burden relationship” for each study. This relationship is meant to assess both how well a study performed in terms of effect size (opioid sparing between test drug and control groups) and how severe the pain was in the group that was randomized not to receive the test drug.

The study selection procedure is summarized in figure 2. The summary of characteristics of the studies is shown in table 1.^15–98  All included studies were controlled blinded trials with quality scores of 3 to 5 on the Jadad scale.⁷  Selected results of publication bias and tests of heterogeneity, as well as efficacy measures of the individual studies and the graphical representation of the efficacy–burden relationship are presented in tables 2 and 3 in Section A3 of Supplemental Digital Content 1, https://links.lww.com/ALN/B91.

Details of the 33 included articles (48 active study drug arms) can be seen in Section A2 of Supplemental Digital Content 1, https://links.lww.com/ALN/B91. Two special cases are noteworthy. Trials using intrathecal or epidural opioids as the study drug gave the largest effect sizes and highest efficacy to burden relationships. Intrathecal and epidural opioids, especially morphine and hydromorphone, have markedly higher potency and duration of action compared to the same opioids given systemically. Conversely, when the ultra–short-acting opioid remifentanil³¹  was given intraoperatively as study drug, it produced a significant negative postoperative opioid-sparing effect, that is, greater opioid use in patients receiving study drug versus placebo, consistent with previous animal and adult human studies showing remifentanil’s potential to induce hyperalgesia and acute tolerance.⁹⁹ 

This endpoint was recorded for 32 study drug arms among 22 studies including a wide range of surgeries.

Significant standardized mean differences (reductions in opioid use) were found for 16 of 28 active study drug arms (Hedges’ g = −0.84; 95% CI, −1.22 to −0.47, P < 0.001) (fig. 3). Subgroup analyses found a highest effect in scoliosis surgeries (Hedges’ g = −3.23; 95% CI, −5.01 to −1.45, P < 0.001), intermediate effect sizes in thoracic and cardiac surgeries, (Hedges’ g = −0.90; 95% CI, −1.25 to −0.55, P < 0.001), and very modest or insignificant effects in adenotonsillectomies (Hedges’ g = −0.47; 95% CI, −1.16 to 0.22, P = 0.18), and urological/abdominal procedures (Hedges’ g = −0.17; 95% CI −0.43 to 0.09, P = 0.20). Significant percent reductions in opioid use were found in 17 of 32 study drug arms. Intrathecal morphine²⁵  after spinal fusion showed the highest sparing effect and efficacy–burden relationship.

This endpoint was recorded for 21 study drug arms among 16 studies. It was used commonly for ambulatory/short-stay surgeries, including urologic surgeries and adenotonsillectomies, not at all for scoliosis surgery or thoracic surgery.

Percentage requiring rescue medication showed significant standardized mean difference from control in 11 of 21 treatment arms (Hedges’ g = −0.83; 95% CI, −1.15 to −0.52; P < 0.001). Subgroup analyses showed that the effect magnitude is higher for the adenotonsillectomies (Hedges’ g = −0.99; 95% CI, −1.53 to −0.44; P < 0.001) (driven by one article)¹⁵  and smaller for urological procedures (Hedges’ g = −0.42; 95% CI, −0.89 to −0.05; P = 0.08). Ten of 21 study drug arms demonstrated significant percent opioid-sparing effect.

This endpoint was recorded for 20 study drug arms among 15 studies. It was used commonly for ambulatory/short-stay surgeries, as well as in one open-heart surgery and one spinal fusion surgery.

Time to first rescue medication shows a strong mean difference in favor of the study drug versus control in 11 of 19 treatment arms (Hedges’ g = −1.64; 95% CI, −2.43 to −0.84; P < 0.001). Subanalyses by the type of surgery found small differences in the magnitude of the effect for adenotonsillectomies (Hedges’ g = −1.44; 95% CI, −2.80 to −0.09; P = 0.04), and urological procedures (Hedges’ g = −1.86; 95% CI, −3.96 to −0.25; P = 0.08). Thirteen out of 20 treatment arms demonstrated significant percentage opioid-sparing effect. Overall, time to first rescue seems to be an outcome with high assay sensitivity in single-dose opioid studies.

Details of the 29 included articles (38 active study drug arms) can be seen in Section A2 of Supplemental Digital Content 1, https://links.lww.com/ALN/B91. The largest number of studies involved adenotonsillectomies (13 articles). Overall the degree of opioid-sparing effect of NSAIDs varied considerably.

This endpoint was recorded for 21 study drug arms among 20 studies. It was used commonly for adenotonsillectomies but also for one study involving idiopathic scoliosis surgery and one involving other types of orthopedic surgery. Significant standardized mean differences (reductions in opioid use) were found for 10 of 15 active study drug arms (Hedges’ g = −0.92; 95% CI, −1.32 to −0.52; P < 0.001) (fig. 4). The effect magnitude was found to be slightly higher in tonsillectomy surgeries than the general mean (Hedges’ g = −1.15; 95% CI, −1.92 to −0.38; P = 0.003). No other type of surgery could be evaluated separately due to the small number of studies. Significant percent reductions in opioid use were found in 11 of 21 study drug arms.

This endpoint was recorded for 26 study drug arms among 18 studies. It was used commonly for adenotonsillectomies. Percentage requiring rescue medication showed significant standardized mean difference from control for 8 of 18 treatment arms (Hedges’ g = −0.52; 95% CI, −0.66 to −0.38; P < 0.001). Ten out of 17 articles demonstrated significant percent opioid sparing effect.

This endpoint was recorded for 15 study drug arms among 11 studies. It was used commonly for ambulatory/short-stay surgeries, mainly for adenotonsillectomies. Time to first rescue medication shows a moderate mean difference in favor of the study drug versus control in one of four treatment arms (Hedges’ g = −0.32; 95% CI, −0.53 to −0.10; P = 0.004). Since there were only four studies that could be included, no subgroup analyses could be conducted. Only 2 of 11 treatment arms demonstrated significant percent opioid-sparing effect and one article had a slightly negative, yet nonsignificant sparing effect.⁵³ 

Overall, total opioid dose in milligram per kilogram per hour seems to be the opioid-sparing outcome with the highest assay sensitivity for NSAID trials.

Details of the 11 included articles (17 active study drug arms) can be seen in Section A2 of Supplemental Digital Content 1, https://links.lww.com/ALN/B91. General conclusions were (1) rectal acetaminophen showed greatest effect size in all three measures at doses of 40 or 60 mg/kg, and variable effects at lower doses; and (2) only one oral and two IV acetaminophen studies met full inclusion criteria for this review; effect size using IV acetaminophen was strongly influenced by study methodology.

This endpoint was recorded for 13 study drug arms among 9 studies. It was used commonly for ambulatory/short-stay surgeries. Significant standardized mean differences (reductions in opioid use) were found for four of eight active study drug arms (Hedges’ g = −2.12; 95% CI, −3.50 to −0.75; P = 0.002) (fig. 5), though this was heavily influenced by the effectiveness of 40 and 60 mg/kg in the study by Korpela et al.⁷⁹  Significant percent reductions in opioid use were found in 4 of 13 study drug arms.

This endpoint was recorded for 10 study drug arms among 6 studies. It was used commonly to evaluate rectal acetaminophen in ambulatory/short-stay surgeries. Percentage requiring rescue medication showed significant standardized mean difference from control in 5 of 10 treatment arms (Hedges’ g = −0.82; 95% CI, −1.20 to −0.44; P < 0.001). Noteworthy, the effects sizes found in the study reported by Korpela et al.⁷⁹  were much smaller compared to the previous outcome and similar to the group mean. Four of six study drug arms demonstrated significant percent opioid-sparing effect.

This endpoint was recorded for five study drug arms among four studies. Time to first rescue medication shows no effect in favor of the study drug versus control in any arms (Hedges’ g = −0.07; 95% CI, −0.32 to 0.19; P = 0.60). Since there were only four studies that could be included, no subgroup analyses could be conducted. No study found any percent differences between the study drug and control, including even those finding significant differences for 40 mg/kg rectal acetaminophen when assessing total rescue opioid usage.^59,73 

Overall, for acetaminophen trials, total opioid dose in milligram per kilogram per hour seems to be the outcome with the highest assay sensitivity, yet due to low number of studies, these results should be considered preliminary.

Details of the general characteristic of the 20 included articles (27 active study drug arms) can be seen in Section A2 of Supplemental Digital Content 1, https://links.lww.com/ALN/B91.

This endpoint was recorded for 18 study drug arms among 13 studies. Half of the articles administered the local anesthetics via infiltration and the other half via peripheral block or epidural. Significant standardized mean differences (reductions in opioid use) were found for 6 of 14 active study drug arms (Hedges’ g = −0.72; 95% CI, −1.18 to −0.27; P = 0.002) (fig. 6). Abdominal surgeries (including inguinal hernia procedures) showed a nonsignificant effect (Hedges’ g = −0.40; 95% CI, −1.01 to 0.20; P = 0.19). No other type of surgery could be evaluated due to the low number of studies. Single-dose studies showed significant standardized mean differences (Hedges’ g = −1.16 95% CI, −1.85 to −0.47; P = 0.001), whereas studies evaluating continuous infusion showed none (Hedges’ g = −0.16; 95% CI, −0.55 to 0.22; P = 0.41). Significant percent reductions in opioid use were found in 10 of 18 study drug arms.

This endpoint was recorded for 15 study drug arms among 10 studies. It was used most commonly for abdominal/urological surgeries. Percentage requiring rescue medication showed significant standardized mean difference from control for 11 of 15 treatment arms (Hedges’ g = −1.19; 95% CI, −1.56 to −0.82; P < 0.001). Abdominal surgeries (including inguinal hernia procedures) were very close to the mean value (Hedges’ g = −1.21; 95% CI, −1.72 to −0.70; P < 0.001). No other type of surgery could be evaluated due to the small number of studies. Eight of 15 study drug arms demonstrated significant percent opioid-sparing effect.

This endpoint was recorded for 12 study drug arms among 10 studies. It was used most commonly for abdominal/urological surgeries. Time to first rescue medication shows a strong difference in favor of the study drug versus control in 9 of 10 treatment arms (Hedges’ g = −1.55; 95% CI, −2.11 to −0.99; P < 0.001). In subgroup analyses, the urological and abdominal procedures show a smaller magnitude of the effect (Hedges’ g = −1.05; 95% CI, −1.45 to −0.65; P < 0.001). Nine of 12 treatment arms demonstrated significant percent opioid-sparing effect.

Local anesthetic trials were highly variable. They involved wound infiltration (9 of 20 articles, two of those with continuous subcutaneous infusion), peripheral nerve blocks (5 of 20 articles), and epidural blocks (4 of 20 articles, only one of those with continuous infusion). Some involved single injection, others involved continuous infusions.

Among the entire group of 85 clinical trials, 62 reported pain scores. Mean pain scores were 2.3 ± 1.5 in the study drug arms and 3.4 ± 1.2 in the control arms (P < 0.001). Eighteen trials used PCA or NCA, whereas 44 administered rescue analgesia by nurse-administered boluses. No statistical differences were found between PCA/NCA and nurse-administered rescue trials on pain scores in the study drug arms (2.5 ± 1.3 vs. 2.2 ± 1.6; P = 0.43) and in the control arms (3.4 ± 1.5 vs. 3.3 ± 2.1; P = 0.84). Studies using observational measures did not differ from those using self-report measures (3.6 ± 1.7 vs. 3.1 ± 2.1; P = 0.25). No trial using PCA/NCA had a mean pain score greater than five in either study arm or control arm. Similar results for the pain score measures were found when using parametric and nonparametric methods.

A linear regression found no associations between pain scores in the control group and type of surgery, type of pain scale (objective vs. self-report), and PCA/NCA versus nurse-administered boluses.

When opioids were used as the study drug, a positive relationship was found between pain in the control group and time to first rescue medication (β = 0.37, R² = 0.44), a small relationship between pain in the control group and total opioid use (β = 0.16, R² = 0.16), but no relationship with the percentage requiring recue medications (β = 0.06, R² = 0.01) (figs. 1 and 2 in Supplemental Digital Content 1, https://links.lww.com/ALN/B91, which depict the efficacy–burden relationship for opioids as the study drug). Although time to rescue showed good assay sensitivity for opioid sparing, trials using this endpoint had higher pain scores in control groups compared with trials using the other primary opioid-sparing endpoints.

When NSAIDs were used as the study drug, a positive strong relationship was found between pain in the control group and time to first rescue medication (β = 0.70, R² = 0.25), a small relationship between pain in the control group and percentage requiring recue medications (β = 0.18, R² = 0.025), but no relationship with the total opioid use (β = 0.12, R² = 0.01) (figs. 7 and 8).

When acetaminophen was used as the study drug, no relationship was found between pain in the control group and percentage requiring recue medications (β = 0.007, R² = 0), or total opioid use (β = 0.06, R² = 0.0001). Since only one study provided pain scores when looking at time to first rescue medication, no association could be assessed (figs. 3 and 4 in Supplemental Digital Content 1, https://links.lww.com/ALN/B91, which depict the efficacy–vurden relationship for acetaminophen as the study drug).

When local anesthetics were used as the study drug, a positive relationship was found between pain in the control group and time to first rescue medication (β = 0.53, R² = 0.53), percentage requiring recue medications (β = 0.29, R² = 0.40), and total opioid use (β = 0.29, R² = 0.40) (figs. 9 and 10).

Our systematic review and quantitative analysis examined rescue analgesic designs in pediatric trials of four commonly used classes of analgesics for acute pain. We considered these designs from the standpoint of usefulness as a surrogate measure of analgesic efficacy and from the standpoint of burden of unrelieved pain in the subjects, particularly in control groups.

Although opioid sparing could be demonstrated in a high percentage of trials in this systematic review, the magnitude of rescue opioid sparing varied greatly. Some sources of variability in these trials appears due to: (1) the test drug (dose, bioavailability, intrinsic efficacy, time course of action relative to the timing of measurements), (2) type of opioid-sparing endpoint (milligram per kilogram per hour, time to rescue, percent of subjects needing rescue), (3) method of analysis (Hedges’ g vs. percent sparing), (4) type of surgery, and (5) a range of additional demographic variables. Despite an initial survey of almost 6,000 abstracts from pediatric analgesic trials using nested search terms, ultimately only 85 trials fit our inclusion criteria for quantitative analysis. Analysis indicated great heterogeneity in each of the five sources of variability. Based on this, recommendations can be only somewhat provisional.

We evaluated the three most common reported rescue-analgesic–sparing outcomes: (1) total dose (milligram per kilogram per hour), (2) percentage of children requiring rescue medication, and (3) time to first rescue medication (minutes). Variations in the design methodologies of the analgesic clinical trial influence the sensitivity for detecting differences in each of these outcomes, making it difficult to designate one of these outcomes as an accepted standard for all studies. Total dose seems to be the outcome most often chosen. This parameter can be used in single-dose or multiple-doses trials. However, this outcome is susceptible to the nonlinear interactions between the study drug and the rescue medication. In addition, for a single-dose study drug with strong efficacy but short duration of action, effect size magnitudes will depend on the time period chosen for recording between-group differences, and use of a long-acting rescue opioid might wash out between-group differences. One recent trial published after we completed the systematic review, used PCA via sufentanil as the rescue analgesic to evaluate thoracic paravertebral blockade for the Nuss operation for pectus excavatum. This study showed excellent opioid sparing, but also had high pain scores in the control group.¹⁰⁰  In future studies evaluating rescue opioids with relatively short context-sensitive half-times, we believe that it is important to permit escalation of dosing parameters in the setting of unrelieved pain. Time to first rescue medication appears useful in single-dose studies (e.g., nerve blocks, infiltrations) and allows us to have a more clear view of the primary effect of the study drug. However, this parameter appears less useful for very short duration study drugs or multiple dose trials. Finally, percentage requiring rescue medication can be useful in small procedures, when the pain scores and the time and amount of rescue medication needed are low.

From an ethical standpoint, it was noteworthy that, among the 83 studies included in this review, the mean pain scores in the control arm were mild or moderate (averaging 3.4 out of 10 in a standardized scale), and only very rarely severe. PCA/NCA paradigms seem particularly effective for preventing high pain scores among children randomized to control groups.

Previous research indicates that, in routine clinical use, patients and nurses do not dose PCA or NCA, respectively, based solely on pain intensity, but dosing is based also on a range of other factors, including anticipation of future pain and on side-effects. In PCA/NCA trials, dosing was not titrated to give equal scores in the study drug and control groups, but rather mean pain scores were lower in the study drug arms (P = 0.013). Based on this, there is a potential for opioid sparing alone to underestimate the analgesic effect of a study drug. Future studies may consider the potential utility of composite efficacy measures based on both pain scores and opioid sparing.

As with all quantitative systematic reviews, meta-analyses are only as good as the data that are reported and the description of methods in each study. Data of many studies could not be included due to lack of necessary information. In our meta-analysis, we included multiple treatment arms from a single study. We are aware that duplicating the number of patients in a control group between two comparisons may generate a unit-of-analysis error. This could have been avoided by either splitting the shared group resulting in a smaller sample size and including two or more comparisons, by combining groups to create pairwise comparisons, or by undertaking a multiple treatment analysis. However, the goal of our study was not to test a specific drug’s efficacy in either the prevention or treatment of pain but to see if immediate rescue designs are feasible and usable outcomes in both these kinds of studies, over various types of surgeries and intervention drugs, and to consider the utility of different opioid-sparing endpoints. It is clear that variability in study design, type of surgery, method of opioid delivery, duration of study drug administration, and reported outcome measures all impact on the likelihood and degree of positive findings and thus our quantitative results should be interpreted as exploratory.

Immediate rescue analgesic trials show reasonable assay sensitivity and tolerably low burden (low-moderate pain scores) for children after surgery. From a clinician’s standpoint, opioid sparing matters only if it is associated with meaningful improvements in clinical outcome measures, including pain scores, reductions in side-effects such as nausea, vomiting, itching, bowel dysfunction, etc., as well as in the time course of recovery, rehabilitation, and postoperative behavior. Acetaminophen, NSAIDs, spinal morphine, and a range of types of regional anesthesia are indeed effective at reducing systemic opioid use, but the degree of opioid sparing varies both with drug, dose, technique, and type of surgery.

Patient-controlled analgesia/nurse-controlled analgesia paradigms seem ideal for more extensive surgeries, repeated dosing of the study drug, or trials involving long-acting study drugs. Surrogate efficacy measures are important for guiding analgesic prescribing in infants and children, but they are not the only outcome measures that are essential components of pediatric analgesic trials. As outlined by previous consensus groups, including the Initiative on Methods, Measurement, and Pain Assessment in Clinical Trials,¹⁰¹  measures of safety, side-effects, hospital stay, complications, behavioral measures, and functional recovery parameters are essential as well. The combination of these elements will allow for ethical and feasible study designs in future pediatric analgesic trials.

The authors thank Paul A. Bain, Ph.D. (Countway Library of Medicine, Harvard University, Boston, Massachusetts), and Alison Clapp, M.L.I.S. (Boston Children’s Hospital, Harvard Medical School, Boston, Massachusetts), for their assistance in searching the literature, and Ted Kaptchuk, B.A. (Beth Israel Deaconess Medical Center, Harvard Medical School, Boston, Massachusetts), for providing mentoring, which was supported by National Institutes of Health grant no. 2K24AT004095 (Bethesda, Maryland).

This study was supported by a grant project (P2BSP1_148628) awarded to Dr. Kossowsky by the Swiss National Science Foundation (Bern, Switzerland) and by the Sara Page Mayo Endowment for Pediatric Pain Research, Education and Treatment (Boston, Massachusetts) to Dr. Berde.

The authors declare no competing interests.