This systematic review and meta-analysis was initiated by manual search of the studies and references from two articles.23,37 Searches were performed irrespective of where the studies were conducted, without language restrictions, and were limited to humans. The study plan obeyed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines for meta-analyses of interventional studies.38

The authors completed searches for the key evidence in ten scientific databases from March 1 to 31, 2018; trials were published between January 1, 2001, and December 31, 2017: MEDLINE (OvidSP), Embase, PubMed, CENTRAL, LILACS, ISRCTN Registry, CINAHL (EBSCOHost), Web of Science (CPCI-S) for RCTs, and for ongoing trials the authors searched Clinicaltrials.gov and WHO-ICTRP. For each database, a comprehensive search strategy was developed by using Boolean operators, and controlled through MeSH terms. Some of the search terms, as detailed in Table 1, included: "anemia, sickle cell," "pain," "acute pain," "randomized controlled trial," "placebo," "clinical trial," and "double blinded." Additionally, for further searches Google Scholar was used, and the authors manually searched the other bibliographies from included studies, which were considered for references supplied to other studies or were missed in the primary electronic search. Only the studies published under peer-review were considered in the search.

Two review authors (MIS and SS) independently searched the databases. Also, they independently identified the pertinence of the retrieved studies and examined the full-text papers to decide upon eligibility according the criteria. Any discrepancies about the inclusion of full-text studies were elucidated and resolved by consensus of all investigators.

To be considered, clinical trials needed to have at least ten participants in each study arm, in any clinical setting, and were required to be randomized, double-blinded, placebo-controlled studies, with parallel or single group and/or pilot study designs, and whose interventions had an analgesic purpose. The study population was pediatric patients that were aged between 3 and 21 years old, with three SCD genotypes – limited to HbSS, HbSC, and HbS-β thalassemia – experiencing an uncomplicated acute pain episode from VOCs. Non-RCTs and RCTs that reported the eligibility combination of both pediatric and adult participants, studies in animals, and studies that outlined outcomes in participants with complicated severe sickle-cell pain crisis (e.g., acute chest syndrome, acute splenic sequestration, and cerebrovascular accident) and chronic sickle-cell pain, studies published with only abstract, and studies with double-blinded active comparator trials of drugs were excluded. Therefore, seven studies were excluded: three non-RCTs,39–41 one study with a paucity of available participants,42 one including merely two genotypes,43 one with chronic pain,44 and one with an active comparator.45 The included studies had fully outlined two of these following outcome measures: assessing pain score, length-of-stay, cumulative amount of narcotic use, and adverse events (the definitions of outcome measures are reported in Table 2). Effectively, the studies that reported on disease-modifying therapy (e.g., hydroxyurea) for SCD were excluded, which is beneficial for VOCs. The studies addressed interventions particularly in regular blood transfusion, erythro-exchange regimens, colloids, fluid replacement, medications that drive to dwarf the less-deformable red blood cells, and anticoagulants which are thought to preclude the SCD-crisis and to minimize the frequency of pain by impeding the sickling process were also excluded. Others studies designed interventions featuring aggressive analgesic therapy with systemic treatment (oral, subcutaneous, epidural, and intramuscular administration) involving strong opioids, weak opioids (codeine and dextropropoxyphene), and non-opioids agents (paracetamol and NSAIDs) were not included in the review.

One reviewer (MIS) performed the data extraction from the included trials, which was reviewed by all investigators (SS and DZ), and all discrepancies were also resolved with all authors by agreement. One investigator (MIS) extracted the general details from each study according to the Cochrane Handbook for Systematic Reviews of Interventions,46 which was subsequently crosschecked by all reviewers. The form included the first author and year of publication, study location, setting, design, characteristics of participants, description of interventions, and study results (pre-specified outcomes).

Risk of bias for methodological quality of the eligible trials was ascertained and assessed by one review author (MIS) and followed with a review by all investigators using each domain of relevant bias in accordance with the Cochrane risk assessment tools for randomized study designs.47 Each trial was judged as low, high, or unclear risk of selection bias (random sequence generation and allocation concealment), performance bias (blinding of participants and personnel), detection bias (blinding of outcome assessment), attrition bias (incomplete outcome data), reporting bias (selective reporting), and other bias (Table 3, more details in Supplementary material). Any discrepancies were resolved by agreement of all investigators. As advocated in the Cochrane Collaboration, the Grading of Recommendations, Assessment, Development, and Evaluation (GRADE) Profile48 was used to rate the quality of the evidence for individual outcome by using the five leading domains, including the following: limitation of the study design or execution (risk of bias); inconsistency, indirectness, and imprecision of results; and, publication of bias. Hence, the quality of evidence was judged as very low, low, moderate, or high for each outcome.

From each outcome measure, the authors retrieved and directly extracted the mean (standard deviation [SD]) or calculated it from the median and interquartile ranges (IQRs) by using the method of Hozo et al.49 Continuous outcomes were recorded as mean difference (MD) along with 95% confidence intervals (CIs) in outcomes measured using the same scale among studies (length-of-stay, primarily in-hospital), and as standardized mean difference (SMD) among studies that had used scales or units in different ways for measuring identical outcomes (change in ladder of pain score and amount of narcotics used during the study). Also, the number of events from each study in the two arms (split by the same investigator into minor and serious) and the summary statistic, presented by dichotomous outcomes as risk ratios (RR) and 95% CIs, were extracted. Contact was established with the principal authors of the identified reviews and trials to ask for any unpublished or missing data. The present authors derived pooled estimates of SMD, MD, and 95% CIs from the mean (SD) with inverse variance (IV)-weighted random-effects models where heterogeneity was moderated. In contrast, fixed-effects models were used where heterogeneity was considered as unimportant or nil. Overall summary estimates of RR and 95% CIs were computed with Mantel–Haenszel weighted fixed-effects models for events during the study. To determine the heterogeneity assessment between the effects of intervention, the authors used Cochran’s chi-squared (chi2) test along with p-value less than 0.10 to assess for subgroup analyses. The authors also used an alternative measurement of the statistical heterogeneity with the inconsistency (I2) method, which expresses a range from 0% to 100%, and it was calculated as unimportant (I2<50%), moderate (I2≥50%), or substantial (I2<80%).50,51

Analyses were performed to explore the impact of the pharmaceutical analgesic interventions versus control on the SMD of the change in the ladder of pain score for uncomplicated acute sickle-cell pain crisis, on the MD of hospital length-of-stay, and on the SMD of the amount of narcotics used during the study. Subgroup analyses were performed to consider the diverse adverse events that occurred in the two arms throughout each study.

Assessment of publication bias in order to generate funnel plots and sensitivity analyses could not be performed, since only a few studies were retrieved (less than 10).50,52 Furthermore, it would not be possible to achieve a meta-analysis review if the heterogeneity identified was 80% or more. All analyses were performed with Review Manager (Review Manager (RevMan) [Computer program]. version 5.3. Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration, 2014).53

The summary of the study characteristics are described in Table 4 and in Supplementary Material. Moreover, these studies were published between the year 2003 and 2016. A total of four trials involved 227 participants.54–57 Among these, 111 patients were incorporated into intervention arms and 116 into placebo arms.

In the studies selected for inclusion in meta-analysis, three RCTs were found with interventional drugs that have crucial interventional role in the pathological mechanism of VOCs, including intravenous magnesium, arginine, and nitric oxide, and one RCT of intranasal fentanyl, which is an opioid derivative. All of the RCTs were compared with a non-treatment arm as placebo control, and those drugs were subsequently investigated. There were two studies with parallel design54,55; one single group design56 and one pilot design.57 All trials included were of patients admitted to the emergency department (ED) for an acute episode of pain crisis caused by VOC that used scale measurement of pain; and that were of shorter duration of treatment, and were achieved in a normal period. Three of the included studies were done in United States54,56,57 and one was conducted in Canada.55

The total number of participants in the included trials was not different regarding the intervention (111) and placebo (116) arms, and total number of participants within each trial was between 20 and 104. All studies focused on unpredictable and uncomplicated acute episodes of sickle-cell pain crisis with phenotypes identified as HbSS, HbSC, and HbSβ-thalassemia. Of the 227 participants involved, 149 (65.6%) had HbSS, 61 (26.9%) had HbSC, and only 17 (7.5%) had HbSβ-thalassemia; 115 (50.66%) were male and 112 (49.34%) were female. Forty-nine participants were eligible and were included for analysis in the intranasal fentanyl trial, 104 in the intravenous magnesium trial, 54 in the arginine trial, and only 20 in the nitric oxide trial. Across two groups (treatment and control) in all studies, the lowest mean (SD) age of participants ranged from 10.6 (5.3) to the highest mean (SD) of 17.6 (2.4) years. The splits of participants in each arm (intervention and placebo) through all trials were almost equal (24–25) in the intranasal fentanyl trial, 51–53 in the intravenous magnesium trial, 26–28 in the arginine trial; and ten-ten in the nitric oxide trial.54–57

The Cochrane risk of bias assessment tool for individual trials is shown in Table 3 and in Supplementary Material. Additionally, the quality of the evidence is shown for each outcome, which in accordance with the GRADE approach was classified from “low” to “moderate” (Table 5). The three randomized controlled trials were judged to be at low risk of selection bias given that randomization method was reported by using the tables in block, and the sequence was engendered through the pharmacy research.54–56 One trial was judged to have an unclear risk of selection bias, as the study did not suitably describe the method of randomization and had insufficient information regarding allocation concealment (Supplementary material).57 All four studies were graded to be at low risk of performance, attrition, and selective reporting bias because all reported that their trials were double-blinded and that the placebo treatments were not discernible from the study intervention. All analyzed participants had been included in the final data. Reports of all intended pre-specified outcome measures were cited in the online trial registry.54–57 Three trials had an unclear risk of detection bias because in the trial registry there were insufficient clear details to judge, and no supplementary statements of information were given for the outcome assessors that were incorporated into the blinding.55–57 One trial was considered as having a low risk of detection bias because the blinding incorporated outcome assessors that were stated in the trial registry.54 Furthermore, all studies had a high risk of other potential sources of bias; the reason was that the target sample size was not attained (Supplementary material).

Owing to the poor number of included trials into the meta-analysis review (less than ten), implementation of a funnel plot to explore the risk of publication bias and sensitivity analyses were not feasible.

In the analysis of the change in the ladder of pain score, pooling the four studies (pain intensity rating was assessed either by visual-analogue scale (VAS) or other tools with age-appropriateness, i.e., the Faces Pain Scale) in an IV-weighted random-effects meta-analysis (Fig. 2) indicated that the SMD estimate did not show a significant difference between analgesic drug and control arms (SMD: −0.08, 95% CI: −0.53, 0.37); p=0.72. There was moderate heterogeneity (I2>50%) among studies that estimated the outcome of change in pain rating (chi2=7.66, p=0.05, I2=61%). The quality of evidence was low (Table 5).

Figure 2.

Change in the ladder of pain.

(0.16MB).

For the three trials that reported hospital length-of-stay for uncomplicated acute episodes of sickle-cell pain crisis,55–57 the pooled mean difference estimate calculated with IV-weighted fixed-effects meta-analysis (Fig. 3) indicated no statistically significant difference between the analgesic drug and control groups (MD: −5.05h, 95% CI: −26.64, 16.55); p=0.65. The quality of evidence was low (Table 5), with no evidence of heterogeneity for the hospital length-of-stay outcome among those three trials (χ2=1.94, p=0.38, I2=0%).

Figure 3.

Length-of-stay primarily in hospital.

(0.17MB).

The same three trials55–57 reported the estimated pooled standardized mean difference, calculated with IV-weighted fixed-effects meta-analysis (SMD: −0.25, 95% CI: −0.55, 0.05); p=0.10 (Fig. 4). A non-statistically significant difference between the analgesic drug and control arms for the amount of narcotics used throughout the study was found, along with no important heterogeneity (χ2=3.06, p=0.22, I2=35%). The quality of evidence was moderate (Table 5). This outcome analysis was reported by one study (intranasal fentanyl) as an event.54 Five (20%) participants in placebo group received parenteral morphine compared to one (4%) participant in intervention group. In another study (arginine),56 two participants in placebo arms were ruled out from the analysis of the amount of opioid consumption on account of the narcotic records being incomplete. Thus, the number of participants in placebo arm for the total opioid use outcome in this study was decreased from 28 to 26. However, these participants were involved in all other outcomes analyses.

Figure 4.

Narcotics used during the study.

(0.17MB).

The subgroup analyses results are shown in Figs. 5 and 6. Each trial had their own manner to describe and tabulate the different numbers of adverse events. Therefore, the present authors planned to exhibit adverse events data in proportion to the minor (transient hypotension, pruritus, hives, pain at the drug delivery site, drowsiness, nausea or vomiting, hypoxia, headache, and tachycardia) and serious adverse events (acute chest syndrome; liver function enzymes increased; hospitalization for pain management; requiring blood transfusion; return to ED within 24h, 72h, or one week; and readmission for pain management within one month) during the study. For the proportion experiencing minor adverse events presented by pain at the drug delivery site, there was a statistically significant difference (p=0.03) in the combination of data from two studies (intranasal fentanyl citrate and intravenous magnesium)54,55 along with no important heterogeneity (χ2=0.17, p=0.68, I2=0%). The quality of evidence was low (Table 5).

Figure 5.

Proportion experiencing minor events.

(0.78MB).

Figure 6.

Proportion experiencing serious events.

(0.64MB).