Nonpharmacological Interventions to Reduce Sedation and General Anesthesia in Pediatric MRI: A Meta-analysis

This systematic review and meta-analysis were structured and reported according to The Preferred Reporting Items for Systematic Reviews and Meta-analysis (PRISMA) reporting guidelines (Supplemental Table 9). The study protocol has been published²²  and registered in PROSPERO (CRD42021243236).

We systematically searched Ovid Medline, CINAHL, Embase, and the Cochrane Central Register of Controlled Trials from database inception until May 11, 2021 and updated on October 10, 2022 to identify available peer-reviewed articles. Keywords and controlled vocabulary were linked by Boolean operators using a combination of synonyms for “MRI” and “nonpharmacological interventions” and “pediatrics patients.” Supplemental Table 4 contains additional details and presents the search terms for each database. We also screened reference lists of included articles, prior systematic reviews, first-tier gray literature,²³  and ongoing trials to identify additional evidence.

Study inclusion was based on the population, interventions, comparators, and outcomes framework.²⁴  Articles were eligible if participants were 0 to 18 years with any pre-existing condition or disability, and if the study compared a nonpharmacological intervention with standard care alone. The primary outcome was use of sedation or general anesthesia, and secondary outcomes included scan time, image quality, and child and parental anxiety. RCTs and quasi-experimental designs with comparator groups were eligible, and studies were included regardless of language, publication status, and year of publication. For the inclusion of conference abstracts, we contacted the authors to obtain unreported data or to clarify information. We excluded studies with an observational or a qualitative design, studies without a comparison group, or if the comparison group was healthy controls, as well as studies not reporting on at least 1 of the predefined primary or secondary outcomes.

We used the systematic review management software Covidence²⁵  to screen titles and abstracts, as well as for full-text reviews. Three independent reviewers (J.H., T.W.M., and N.E.B.) screened articles for eligibility, and disagreements about inclusion were discussed with a fourth reviewer (L.K.G.) until consensus was reached. Potentially relevant articles not written in English were screened using Google Translate. Using a predesigned data extraction form, 2 reviewers (J.H. and J.T.) independently extracted relevant data in Covidence from the included studies. The data extracted included information about study identification, methodology, population, intervention, control groups, outcomes, and conclusions. Any discrepancies between authors in data extraction were resolved through consensus discussion.

Two reviewers (J.H. and J.T.) independently assessed the methodological quality of RCTs using the Cochrane Collaboration tool.²⁶  Risk of bias concerns were rated as low, unclear, or high, and any disagreements were resolved through discussion to achieve a final decision. One reviewer (J.T.) performed the risk of bias assessment for nonrandomized studies using the Scottish Intercollegiate Guidelines Network checklists.²⁷  Study quality for nonrandomized studies was rated as high, acceptable, or low. The certainty of evidence for primary outcome was assessed using the Grading of Recommendations, Assessment, Development, and Evaluations (GRADE) framework, and tables were generated using the Cochrane web application GRADEpro.²⁸ 

Effects of nonpharmacological interventions on the different outcomes were in general investigated in infants (<2 years) and children and adolescents (2–18 years), as most of the studies did not allow further subdivision of age groups. We performed a random effects meta-analysis of all studies reporting use of sedation or general anesthesia (dichotomous) in Review Manager version 5.4.²⁹  Meta-analysis was performed if 2 or more studies provided data on use of sedation or general anesthesia that could meaningfully be pooled. If studies included multiple intervention groups, data from all intervention groups was pooled. We separately performed meta-analysis for randomized studies and nonrandomized studies. Subgroup analyses were conducted for children aged 3 to 10 years and 11 to 18 years to examine differences in effect estimates according to age groups. Effect sizes were computed by reporting the number of participants with an event of sedation or general anesthesia relative to the number of participants in the group. If data were only available in a figure, we used the WebPlotDigitizer version 4.5³⁰  to extract data. The proportion of study participants receiving sedation or general anesthesia in MRI was pooled and forest plots were generated to present risk ratios (RRs) and 95% confidence intervals (CIs). Heterogeneity of effect sizes was assessed by l² statistics, and if heterogeneity exceeded 75%, substantial heterogeneity was indicated. Sensitivity analyses were performed to identify potential differences between findings from unpublished data and fully published peer-reviewed data. For secondary outcomes (continuous), summary statistics were performed in RStudio version 1.4³¹  to calculate means and SD.

Of the 10 424 records identified from the searches, 226 articles were considered for full-text review, 180 of which were excluded for various reasons (Fig 1). The 46 studies that were ultimately included in the systematic review were published from 1997 to 2022. Twelve were RCTs and 34 were nonrandomized. They included 1720 infants (<2 years) and 35 163 children and adolescents (2–18 years) undergoing MRI. Table 1 provides an overview of the characteristics of the included studies, whereas a detailed summary is available in Supplemental Table 8.

In studies reporting infant sex and age, 55.3% were male and the age range was 0 to 1 year, with half of the studies^32–36  targeting infants 0 to 3 months. In those reporting on children and adolescents, 52.8% were males and the age range was 2 to 18 years, with almost 40% of the studies^37–45  targeting children <10 years old. Only a few studies specified patient diagnoses, which spanned a broad spectrum of somatic and psychiatric diseases. The brain was the most frequently scanned body part, followed by the spine, heart, neck, and abdomen. Only a few studies provided information on sociodemographic characteristics (Table 1).

According to the GRADE framework, there was moderate certainty in evidence for RCTs and very low certainty for nonrandomized studies (Supplemental Table 6 and 7). The RCTs were rated as either high or unclear risk of bias, the latter mainly because of lack of blinding and unclear allocation concealment, selective reporting, and random sequence generation. The nonrandomized studies were rated as either acceptable or low quality. The most common reasons for downgrading quality in nonrandomized studies were that study groups were not comparable, lack of confounding considerations, and no confidence intervals were provided. Six conference abstracts were included but were not assessed for methodological quality because of limited information on study design. Risk of bias of the individual studies are provided in Supplemental Fig 3 and Supplemental Table 5.

The 12 studies that addressed infants were aimed at calming them and making them fall asleep before the scan using feeding (Table 2),^{32–34,36,46–48}  immobilization,^{33–36,46,49–52}  and/or noise reduction,^{32,47,48,50,51}  either alone or in combination. The 34 studies addressing children and adolescents were designed to prepare, distract, or acknowledge the patients (Table 2). Primary components of the interventions included mock scanners, toy models,^{37,40,42,44,45,53–60}  preparational videos, and reading material about MRIs,^{40,41,44,45,56–58,61–65}  creative art (eg, play or music) therapies,^37,66–69  and various types of visual and auditive distraction during the scan.^{38,39,45,47,48,53,60,69–71} 

The meta-analysis, which was performed exclusively on children and adolescents because only 1 infant study⁵¹  examined the effect of nonpharmacological interventions on the use of sedation or general anesthesia, included 33 873 children and adolescents from 6 RCTs^{37,43,56,57,62,68}  and 14 nonrandomized studies^{38,39,41,42,54,58,63,64,70–73}  comparing the need for sedation (n = 9) or general anesthesia (n = 11) between a nonpharmacological intervention and standard care. The RCTs included 1115 children and adolescents and the nonrandomized studies included 32 758. For studies that provided data on control groups, standard care was either described as pertinent information about the scanning^{37,41,42,57,58,62}  or as an alternative nonpharmacological intervention (eg, involvement of child life specialists).^38,56,70 

For the 6 RCTs, nonpharmacological interventions were associated with lower use of sedation or general anesthesia when compared with standard care, with 95 out of 677 (14%) children receiving sedation or general anesthesia in the intervention groups compared with 106 out of 438 (24.2%) in control groups (Fig 2). The pooled RR for sedation or general anesthesia in RCTs was 0.68 (95% CI, 0.48–0.95) with moderate heterogeneity (l² = 35%).

For the 14 nonrandomized studies, nonpharmacological interventions were associated with lower use of sedation or general anesthesia compared with standard care with 5653 out of 21 866 (25.9%) children receiving sedation or general anesthesia in the intervention group compared with 4471 out of 10 892 (41.0%) in the control groups. The pooled RR in nonrandomized studies was 0.58 (95% CI, 0.51–0.66) with considerable heterogeneity (l² = 91%). After removing a conference abstract³⁸  with limited data available, there was no change found in the significance of the overall effect estimates or statistical heterogeneity.

Subgroup analysis on children aged 3 to 10 years, including 6 studies,^{39,53,64,70,72,73}  showed that 1724 out of 4510 (38.2%) children received sedation or general anesthesia in the intervention group compared with 1943 out of 3496 (55.6%) in the standard care group (Supplemental Fig 4). The pooled RR for sedation or general anesthesia was 0.65 (95% CI, 0.56–0.76) with considerable heterogeneity (l² = 89%). In older children and adolescents, aged 11 to 18 years, including 4 studies,^53,64,70,73  159 out of 3170 (5.0%) children received sedation or general anesthesia in the intervention group compared with 183 out of 2624 (7.0%) in the standard care group. The pooled RR for sedation or general anesthesia was 0.69 (95% CI, 0.47–1.01) with moderate heterogeneity (l² = 48%).

Thirteen studies^{32,35,42,46–48,51,52,59,63,66,71,74}  assessed the effect of nonpharmacological interventions on scan duration, 9 of which reported the mean scan time between groups. Sedation or general anesthesia was used as standard care in all but 1 of the infant studies, and in 1 of the 3 studies on children and adolescents (Supplemental Table 8).

The pooled estimates showed that the total scan time was lower in the intervention group among infants with a mean (SD) of 41.7 minutes (19.5) in the interventions group compared with a mean (SD) of 49 minutes (19.7) in the control group (Table 3) which is supported by meta-analysis (Supplemental Fig 5). For children and adolescents, only 6 of 34 studies reported scan time. Thus, meta-analysis was not performed for this age group because of large heterogeneity in outcome measures.

Twenty-one studies evaluated image quality.^{32–36,40,42,44–52,55,62,66,67,75}  Radiologists determined whether images were diagnostic or not in 11 studies on infants and 4 studies on children and adolescents. Sedation and general anesthesia were used as standard care in the control group in almost all the studies on infants and in 2 out of 4 studies on children and adolescents (Supplemental Table 8). For infants, the pooled estimates between groups were almost similar, with 1202 of 1310 (92%) images evaluated as diagnostic in the intervention group compared with 487 of 533 (91%) in the control group (Table 3, Supplemental Fig 6). The pooled estimates for children and adolescents seemed to be in favor of the intervention group with 481 out of 632 (76%) images evaluated as diagnostic compared with 247 out of 361 (68%) in the control group (Table 3) but was not supported by estimates from meta-analysis that showed no difference between groups (Supplemental Fig 7).

Seven studies^{56,57,59,61,65,76,77}  investigated psychological outcomes in children, adolescents, and parents. In studies providing data on the control group, pertinent information about the scanning^57,59,61,77  or an alternative nonpharmacological intervention (booklet or Web site)^44,56  was employed as standard care.

Anxiety and distress level was measured using State-Trait Anxiety Inventory (STAI) for adults and for children, Revised Children’s Manifest Anxiety Scale, Visual Analog Scales, or a relaxation score.

Three studies^59,61,76  favored the intervention group and 2^56,65  the control group, whereas 2 studies^57,77  were inconclusive (Table 3). None of the studies reported a significant difference between groups, and 1 study⁷⁶  did not test for statistical significance. No meta-analysis was performed on psychological outcomes because of large heterogeneity in outcome measures.

In this systematic review and meta-analysis, we found that nonpharmacological interventions appear to reduce the need for sedation and general anesthesia in children and adolescents (2–18 years) undergoing MRI. The success rate among children and adolescents undergoing MRI without sedation or general anesthesia was as high as 86% in the RCTs and 74.1% in the nonrandomized studies for nonpharmacological interventions, eg, preparation, distraction, and acknowledgment, compared with 75.8% and 59%, respectively, in the standard care groups, which most often received pertinent information only. Moreover, nonpharmacological interventions appeared to be more effective in children aged 3 to 10 years compared with older children and adolescents (11–18 years). As only 1 study estimated the effect of nonpharmacological interventions on the use of sedation or general anesthesia in infants (<2 years), eg, feeding and immobilization, this group was not included in the meta-analysis. Further analysis indicated that nonpharmacological interventions reduce total scan time and retain image quality for all age groups. The effect on psychological parameters were inconclusive.

Compared with previous systematic reviews,^9–12  our review now provides considerable evidence on the effectiveness of nonpharmacological interventions in reducing the need for sedation and general anesthesia in children and adolescents undergoing MRI. We were able to include 20 studies in meta-analysis compared with 2^9,12  and 5¹¹  studies, respectively, in previous meta-analysis on children and adolescents and we were also able to add subgroup analysis comparing the effect across different age groups. Our findings emphasize that nonpharmacological interventions are beneficial and cannot be replaced solely by providing pertinent knowledge in the effort to reduce the need for sedation or general anesthesia in children and adolescents undergoing MRI. The review also demonstrates a broad range of opportunities for hospitals to employ nonpharmacological interventions. Whereas previous reviews with meta-analysis^9,11  on this topic have only included a limited number of studies, the findings of our meta-analysis provide a robust and consistent association based on a larger body of evidence. The positive association between nonpharmacological interventions and reduced need for sedation or general anesthesia offer support for including this in future guidelines for children and adolescents undergoing MRI.

For infants, it is even more evident that nonpharmacological interventions are needed to reduce sedation and general anesthesia, which is likely why only 1 study was identified that did not use sedation or general anesthesia as standard care in the control group. The pooled estimates of scan time imply that total scan time may be reduced in infants, by implementing nonpharmacological interventions, along with obtaining diagnostic images. The high rate of successful images (92%) is consistent with a previous meta-analysis¹⁰  performed on 53 infant studies showing a pooled success rate of 87% in interventions using nonpharmacological strategies. However, their findings were generated by single summary estimates (proportional meta-analysis) without any comparison with control conditions. The positive effects of calming infants and encouraging sleep using feeding and immobilization are promising and well recognized.^10,16,17,19  For children and adolescents, a reduction in scan time increases the likelihood for obtaining successful images without movement.¹⁵  Even though it is unclear whether scan time can be reduced by nonpharmacological interventions in this age group, the pooled estimates for image quality indicates that nonpharmacological interventions can provide at least as many diagnostic-quality scans as standard care, which in half of the included studies was sedation or general anesthesia. In some hospitals, nonpharmacological strategies are already considered standard care for children and adolescents, which is now supported by the evidence in this review.

Unlike a previous review²¹  on various methods for preparing for radiologic procedures, our review did not draw any conclusions on whether preparatory interventions may positively affect anxiety when MRIs are conducted. However, Leroy et al⁷⁸  suggest that age-appropriate preparation and distraction are essential components in reducing anxiety and distress in children undergoing medical procedures and that a child-friendly environment and family involvement may increase procedural comfort.

The pooled estimates from our subanalysis suggest that nonpharmacological interventions are most effective in children aged 3 to 10 years, possibly since this age group often needs support to understand and handle the procedure but still has the cognitive capacity to be coached successfully. Although an effect was also found in older children and adolescents (11–18 years), this group was generally more likely to complete the scan without sedation or general anesthesia by only receiving pertinent information, as only 7.0% of older children and adolescents in the control group needed sedation or general anesthesia compared with 55.6% of children aged 3 to 10 years. Even though only a modest effect was seen in older children and adolescents, some in this age group may still need more thorough preparation, and probably most in this age group would appreciate a sort of entertainment during the scan. This need could be addressed in early stages, eg, when prescribing the scan, by asking the child or adolescent as well as the parents to consider the value of using a nonpharmacological intervention. However, the degree of development and understanding may vary largely in children and adolescents, whereas other factors such as previous experience with imaging may as well influence a patient’s ability to complete a scan without sedation or general anesthesia. Furthermore, nonpharmacological interventions are not always possible, eg, in unstable patients, which is a group that was excluded in some studies.

Reducing the need for sedation and general anesthesia may have more than short-term benefits for children and their parents. The opportunity to improve children’s capacity to understand health information and cope with medical procedures may improve the general hospital experience and be important for children’s future encounters with health services. Moreover, minimizing the use of sedation and general anesthesia, in conjunction with retaining the quality of images and decreasing scan time, may reduce total MRI expenditures significantly.⁷  Considering the urgent need for initiatives to reduce healthcare spending worldwide,^79,80  along with the increasing focus on patient-centered care,^81,82  nonpharmacological interventions appear to be a cost-effective strategy for performing MRI in infants, children, and adolescents, with potential short- and long-term benefits. Future research should aim to compare nonpharmacological interventions to determine the minimum approach needed to avoid sedation and general anesthesia within different age groups, clinical contexts, and cultures.

This review has some limitations. The primary reason for downgrading evidence concerned the risk of bias. However, most studies were not expected to meet the criteria for high quality, as blinding of participants in RCTs is obviously difficult in these types of studies and most nonrandomized studies employed a retrospective design. Evidence for nonrandomized studies was further downgraded because of substantial heterogeneity between the studies. Even though various concerns were related to the methodological quality of nonrandomized studies, the pooled estimates from these studies confirmed the data gathered from RCTs, which adds to the general conclusion.

We included image quality, scan time, and anxiety as secondary outcomes, but large variability was found in the evaluation of these outcomes across studies. Moreover, a variation was found in the analytical approach, with most studies reporting the total number of children scanned, whereas a few studies only reported the total number of scans. Furthermore, the standard care applied in control groups was often vaguely described and may have varied between hospitals and demographic settings. Also, the protocols for when to use sedation or general anesthesia might differ between MRI sites and different healthcare professionals. Additionally, the complexity of each scanning session may differ considerably according to the type of MRI examination and image quality requirements. Lastly, inconsistent and brief reporting of child and parental characteristics (eg, factors impacting health literacy) limited the possibility of reporting and analyzing the impact of these factors.

Nonpharmacological interventions, such as feeding and immobilization in infants, and strategies to prepare, distract, and acknowledge children and adolescents, should be considered as standard procedures to reduce the use of sedation and general anesthesia in pediatric patients undergoing MRI.

We wish to acknowledge information specialist Janne Vendt, Medical library at Rigshospitalet, for assisting in generating the strategy for the literature search.