Results

Study selection

A total of 1,048 publications was retrieved and screened by two investigators (SV, JG or CV). Subsequently, 41 studies were assessed for eligibility based on their full text (SV, CV), of which ten studies fulfilled the selection criteria and were included in this review (Fig. 1 ).44-53 We approached the first author of two studies with unavailable full text articles. However, none of them responded. In total, thirty-one full text articles were excluded. The majority of the excluded studies did not report on our predefined outcomes (n=9).54-62.

Characteristics of included studies

Study characteristics and the risk of bias of the ten included studies are summarized in Table 1 . The studies included a total of 422 participants aged between 6 and 21 years old. All participants had SCD including the genotypes HbSS, HbSC, HbSß-thalassemia and HbSD. Studies were published between 1987 and 2020. All studies were performed in the USA. The type of intervention studied were psychological interventions (n=7), physical intervention (n=1) and integrative medicine (n=2). In this review, psychological interventions included CBT and biofeedback.44,45,47,48,51-53 The physical intervention included the practice of yoga51, and integrative medicine was provided in the form of massage and virtual reality46,50. In total, five different types of interventions were studied including CBT (n=5),44,48,51-53 biofeedback (n=2),45,47 yoga (n=1),49 massage (n=1)50 and virtual reality (n=1).46Eight studies were performed in the outpatient clinic (n=6)44,45,47,51-53 or at home (n=2),48,50 while two studies were performed during hospital admission.46,49
In the outpatient clinic, children received one single training session in four studies,45,46,52,53 in the remaining studies the number of sessions ranged between 2 and 13.44,47-51. The duration of the intervention period that included both the training session and the practices ranged from 30 days45,50,52 to 4.5 months.44Duration of follow-up after the non-pharmacological invention ranged from 1 to 12 months. Pain-related outcomes were assessed using self-reported daily pain diaries in seven studies,45,47,48,50-53 in one of them outcomes were electronically monitored.53 In the other three studies, pain was reported personally to researcher(s) while inpatient or by telephone assessment.
All studies primarily focused on pain. They examined either the frequency of pain (n=2)45,48, the intensity of the reported SCD-related pain on a 5- or 10-point pain scale (n=4),44,46,49,50 or both pain outcomes (n=4).47,51-53 Four studies clearly defined the type of pain they were targeting, namely acute pain in three studies,46,47,49 and chronic pain in one study.44 The remaining six studies did not specify the type of SCD-related pain they were targeting.45,48,50-53 Therefore, we decided to state this pain as daily pain. In total, eight studies investigated the secondary outcomes that we specified, with seven of them investigating health service use44,47-52 and four the use of analgesics.47,49,51,52 In total, four of the studies studied both health service and analgesic use.47,49,51,52

Quality assessment of included studies

In this review, five RCTs48-51,53 and five QED44-47,52 studies were included. The quality of the five RCTs had a mean score of 9.0 out of 13 points. The five included QED studies scored an average of 7.6 out of 10 points. Scores per study are shown in Table 1 . The detailed scoring is listed inSupplemental Figure S2 . In RCTs, the lack of follow-up48-50 and selective reporting were the main reasons for a risk of bias.50,51,53 Three studies did not report on blinding of outcome and allocation concealment, so risk of bias could not be scored on these items.48,50,51 Only one study blinded their outcome assessors for the intervention.49 In QED studies, four out of five studies did not include a control group45-47,52 and two studies did not have a follow-up causing a risk of bias.46,52 In total, five out of the ten studies achieved a follow-up rate of at least 80% of the patients.

Summary of main findings

Effects of interventions

Five studies reported significant reductions in pain-related outcomes after a study period between 1 and 12 months,44,46,47,50,52 while one study reported a significant reduction in analgesic use.47 None of the included articles reported a significant reduction of health service use. In the following paragraphs the results will be discussed according to setting (in- or outpatient) and type of intervention.
Outpatient setting

Psychological interventions

Among the five studies that investigated CBT,44,48,51-53 two showed improvements on pain-related outcomes.44,52 First, Dobson et al. showed that after guided imagery, a technique within CBT, the intensity of daily reported pain decreased from 2.4 (SD 1.2) to 0.7 (SD 1.2) on a 5-point scale (p < 0.01). In addition, the number of painful episodes per month decreased significantly, from 5.6 (SD 3.3) to 2.5 (SD 4.1; p = 0.003). However, no significant changes were reported in the number of ED visits and hospital admissions.52 Second, Sil et al. reported reduced chronic SCD pain intensities on a 10-point scale from 5.5 (SD 2.2) to 3.8 (SD 2.8. p = 0.009). The other three CBT studies did not show any effects on pain, health service or analgesic use.48,51,53 One study used electronic devices to provide CBT electronically (e-CBT) and to record outcomes digitally. In this study, there were large discrepancies between self-reported and device-reported frequencies of skill practice; children reported approximately two to four times more skill practice than what the device recorded.53The efficacy of biofeedback was examined by two studies, with one showing a significant reduction of the pain-related outcomes.45,47 This study by Cozzi et al. studied the effect of biofeedback on acute pain at home.47 Pain intensities reduced from 1.9 to 0.5 (SD not reported) on a 5-point scale (p < 0.001) and the number of self-treated crises per month decreased from 2.2 to 0.4 (p < 0.05). Furthermore, the number of days on which analgesics were taken in a month was reduced from 8.0 to 1.3 (p < 0.001), but health service use did not change compared to baseline.

Integrative medicine

Lemanek et al. investigated massage as integrative medicine at the outpatient clinic. They reported a significant decrease of pain-related outcomes after massage (p = 0.05), however no specific outcomes or effect sizes were reported.49 Therefore, this result was not classified as significant.
Inpatient settings
In the inpatient setting, the effect of yoga and virtual reality were studied for uncomplicated VOCs.46,49 Children reported a reduction in acute pain intensity on a 10-point scale after a single VR session from 7.3 (SD 2.7) to 5.8 (SD 3.2, p < 0.001). Yoga reduced pain intensity from 5.6 to 5.0 (p = 0.029) after an average of 2.5 sessions. No differences were found in LOS and opioid use after yoga.