A systematic literature search utilizing a predefined strategy identified 904 articles. After removing duplicates, 632 articles underwent screening based on title and abstract, resulting in the exclusion of 621. The full text of the remaining 11 articles underwent meticulous review. Seven articles were excluded after thorough examination because they did not report at least one diagnostic accuracy measure for CEUS in evaluating pediatric solid organ injuries. Ultimately, four articles meeting the inclusion criteria were identified and included. These studies provided sufficient data for constructing 2 × 2 tables, facilitating the DTA meta-analysis. The screening process and eligibility criteria adhered to PRISMA guidelines, with a flow diagram presented in Fig. 1.

PRISMA flow diagram showing the review process. PRISMA, Preferred Reporting Items for Systematic Reviews and Meta-Analyses

This review incorporates findings from four studies assessing the diagnostic efficacy of CEUS in identifying solid organ injuries among 364 pediatric patients. The studies included in this review were geographically diverse: two were conducted in Italy, one in the USA, and one in Saudi Arabia [20, 31,32,33]. All studies utilized a cohort approach, two prospective and two retrospective.

The diagnostic performance of CEUS was specifically reported for various solid organs. Injuries to the spleen and liver were examined in all studies, while kidney injuries were assessed in three studies [20, 32, 33]. Pancreatic injuries were reported in two studies, and adrenal gland injury was reported in only one study, preventing us from performing subgroup analysis for these lesions [32, 33].

All studies used CT scans as the reference standard, specifically mentioning contrast-enhanced CT imaging in all but Armstrong et al.’s study [33]. Importantly, CEUS was performed by radiologists at the point of care (POC) in all included studies. However, the studies demonstrated discrepancies in sample inclusion criteria, utilized ultrasound equipment and techniques, and the definitions of solid organ injuries. Detailed information on these aspects is provided in Table 1. Noteworthy is that the study by Menichini et al. only included patients with abnormal findings on unenhanced diagnostic ultrasound [20].

The methodological quality of the included studies, as evaluated using the QUADAS-2 tool, is detailed in Table 2. This assessment indicates that the studies generally maintain acceptable methodological quality across all evaluated domains, with no major concerns identified. However, ambiguities were noted in the patient selection methods of three studies [20, 31, 33]. Additionally, the blinding of interpreters of the reference test (CT) to the results of CEUS, as well as the blinding of CEUS interpreters to the results of the reference test, was not explicitly addressed in some of the studies [31, 33].

Figure 2 displays the paired funnel plots associated with the overall meta-analysis to assess the potential effects of publication bias or small study effects. Although no statistical test was conducted to specifically evaluate publication bias, a visual examination of these plots reveals evidence of asymmetry despite the limited number of studies.

Funnel plots for the assessment of the effect of potential publication bias/small study effect in the reported diagnostic performance indices in overall meta-analysis. FPR, false positive rate; Sen, sensitivity

An overall meta-analysis was initially conducted, pooling data regardless of the specific organs involved. Figure 3 displays the paired forest plots for this comprehensive analysis, revealing a pooled sensitivity of 88.5% (95%CI 82.5–92.6%) and a pooled specificity of 98.5% (95%CI 94.9–99.6%). Additionally, Fig. 4 illustrates the SROC curve for the entire meta-analysis, indicating an AUC of 96% (95%CI 88–99%).

Paired forest plots of the bivariate model random effect meta-analysis of diagnostic performance of contrast-enhanced sonography (CEUS) in diagnosing various solid organ injuries in pediatric patients (overall analysis). CI, confidence interval

Summary receiver operating curve (SROC) of the bivariate model random effect meta-analysis of diagnostic performance of contrast-enhanced sonography (CEUS) in diagnosing various solid organ injuries in pediatric patients (overall analysis). AUC, area under the curve; SROC, summary receiver operating curve

Significant heterogeneity (I2 = 95%, 95%CI 38.6–81.5%) was observed across the four studies, prompting sensitivity analysis. This analysis identified the study by Menichini et al. as an outlier. The pooled sensitivity and specificity post-exclusion of this study were 84.8% (95%CI 80.4–88.4%) and 96.9% (95%CI 93.6–98.6%), respectively, with an AUC of 89% (95%CI 83–99%).

Figure 5 presents the likelihood ratio scattergram for this general meta-analysis. The observations in likelihood ratio values are supported by Fagan’s nomogram, as presented in Fig. 6, indicating that the post-test probability of a solid organ injury remains above 95%, even with a low pre-test probability of 25%. Conversely, in scenarios with a high pre-test probability of 75%, CEUS may still miss 26% of solid organ injuries. Notably, these findings are largely consistent even after the exclusion of the outlier study by Menichini et al., and the post-test probability remained higher than 90% (90.9%) even with a pre-test probability as low as 25%, whereas the post-test probability after a negative test with a pre-test probability of 75% was calculated as 29.8%.

Likelihood scattergram of the bivariate model random effect meta-analysis of diagnostic performance of contrast-enhanced sonography (CEUS) in diagnosing various solid organ injuries in pediatric patients (overall analysis). The scattergram is divided into four quadrants based on likelihood ratios, with data points having a positive likelihood ratio (LRP) greater than 10 classified as optimal for confirmation, and those with a negative likelihood ratio (LRN) less than 0.1 classified as optimal for exclusion. LLQ, left lower quadrant; LRN, negative likelihood ratio; LRP, positive likelihood ratio; LUQ, left upper quadrant; RLQ, right lower quadrant; RUQ, right upper quadrant

Fagan’s nomogram plots with three different pre-test probability assumptions and pooled diagnostic performance indices based on the bivariate model random effect meta-analysis of diagnostic performance of contrast-enhanced sonography (CEUS) in diagnosing various solid organ injuries in pediatric patients (overall analysis). Neg, negative; NLR, negative likelihood ratio; PLR, positive likelihood ratio; Pos, positive

The analysis included an organ-specific subgroup assessment focusing on kidney, liver, and spleen injuries, each represented by at least three studies, as illustrated in Fig. 7. The pooled sensitivities for these subgroups were 88.1% (95%CI 65.8–96.6%) for the kidney, 75.9% (95%CI 67.1–83%) for the liver, and 89.5% (95%CI 84.6–92.9%) for the spleen. The corresponding pooled specificities were 99.3% (95%CI 96.8–99.9%) for the kidney, 96% (95%CI 92.5–97.9%) for the liver, and 95.4% (95%CI 90.2–97.9%) for the spleen. Figure 8 presents the summary receiver operating characteristic (SROC) curves for these subgroups, with pooled AUCs of 99% (95%CI 85–99%) for the kidney, 80% (95%CI 75–97%) for the liver, and 90% (95%CI 87–98%) for splenic injuries. Significant differences were observed between subgroups (P-value = 0.03), with post hoc analysis indicating higher specificity in the kidney compared to other organs (P-value < 0.05) and lower sensitivity in liver injuries compared to splenic injuries (P-value < 0.01).

Paired forest plots of the bivariate model random effect meta-analysis of diagnostic performance of contrast-enhanced sonography (CEUS) in diagnosing solid organ injuries of the kidney, liver, and spleen in pediatric patients (subgroup analysis). CI, confidence interval

Summary receiver operating curves (SROCs) of the bivariate model random effect meta-analysis of diagnostic performance of contrast-enhanced sonography (CEUS) in diagnosing solid organ injuries of the kidney, liver, and spleen in pediatric patients (subgroup analysis). AUC, area under the curve; SROC, summary receiver operating curve

The liver subgroup demonstrated significant heterogeneity (I2 95%CI 22.6–64.3%), leading to a sensitivity analysis that identified Menichini et al.’s study as an outlier. Importantly, subgroup analysis after removing this outlier showed that the higher sensitivity in the spleen compared to the liver remained significant (P-value < 0.01), whereas the higher specificity in renal injuries was no longer significant (P-value = 0.09).

Figure 9 illustrates the pre-exclusion likelihood ratio scattergrams for this analysis. Supplementary Figs. 1, 2, and 3 present the pre-exclusion Fagan’s nomogram plots for the kidney, liver, and spleen, respectively, with a post-test probability exceeding 40% in cases of high suspicion (pre-test 75%) but negative CEUS results.

Likelihood scattergram of the bivariate model random effect meta-analysis of diagnostic performance of contrast-enhanced sonography (CEUS) in diagnosing solid organ injuries of the kidney, liver, and spleen in pediatric patients (subgroup analysis). The scattergram is divided into four quadrants based on likelihood ratios, with data points having a positive likelihood ratio (LRP) greater than 10 classified as optimal for confirmation, and those with a negative likelihood ratio (LRN) less than 0.1 classified as optimal for exclusion. CI, confidence interval; LLQ, left lower quadrant; LRN, negative likelihood ratio; LRP, positive likelihood ratio; LUQ, left upper quadrant; RLQ, right lower quadrant; RUQ, right upper quadrant