Using a predefined search strategy, 1332 articles were initially identified. After removing duplicates, 653 papers were screened based on title and abstract, resulting in the exclusion of 625 articles. The full text of the remaining 28 papers was thoroughly reviewed, leading to the exclusion of 19 articles not aligned with the study's aim. Ultimately, 9 articles meeting the inclusion criteria were identified. A flow diagram following PRISMA guidelines (Fig. 1) summarizes the screening process and eligibility criteria.

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

The selected studies included in our analysis comprised of 7 cohort studies [2, 5,6,7,8, 19, 20] and 2 case–control studies [1, 21]. These studies were conducted in various countries, including the United States (n = 4), France, Colombia, Canada, Brazil, and Turkey (each n = 1). The age range of patients across the studies varied from 13 to 86 years, with the majority of patients being male.

Among the nine articles selected, the majority of studies utilized thoracic, abdominal, or a combination of both types of MDCT scans. Three studies used oral contrast agents [2, 8, 20], three studies employed rectal administration of contrast material [2, 7, 8], and seven studies utilized IV contrast agents [1, 2, 7, 8, 19,20,21]. It is important to note that some studies incorporated multiple routes of contrast administration. Table 1 provides a summary of the general characteristics of the examined studies.

A total of 933 patients with penetrating trauma underwent both CT scan and the reference standard, resulting in the identification of 294 patients with penetrating diaphragmatic injury. Table 2 provides an overview of the general characteristics of diaphragmatic injuries resulting from penetrating trauma across all the included studies. Among these studies, bilateral injury was reported in only two studies [1, 20], while left-sided injury was the most commonly observed. Surgical exploration served as the reference standard in all of the studies included.

Regarding the mechanism of injury, two studies reported stab wounds [5, 19], while two studies specifically focused on gunshot injuries [8, 20]. Four studies documented cases involving both stab and gunshot wounds [1, 2, 7, 21], and one study did not provide specific information on the penetrating trauma mechanism [6]. Table 3 presents the diagnostic accuracy of MDCT scan in cases of diaphragmatic injury caused by penetrating trauma in the included studies.

Among the included studies, the presence of a contiguous injury sign on both sides was found to be the most sensitive indicator of penetrating diaphragmatic injury. Common signs observed on MDCT scans indicating potential penetrating diaphragmatic injury included herniation of abdominal viscera or fat into the thorax, the collar sign, dependent viscera sign, transdiaphragmatic trajectory, diaphragmatic discontinuity, and diaphragmatic thickening. Additionally, associated injuries frequently observed with penetrating diaphragmatic injury included pleural effusion, pneumothorax, hemothorax, hemoperitoneum, and pneumoperitoneum (Table 3).

Deek's funnel plot asymmetry test indicated no significant evidence of publication bias (P = 0.81) among the analyzed studies. As a result, we did not proceed with the trim-and-fill test (Supplementary Fig. 1).

Supplementary Fig. 2 provides a visual representation of the quality assessment of the included studies. Further details regarding the quality assessment for each individual study can be found in Supplementary Table 1. Overall, the majority of studies included in this review exhibited satisfactory methodological quality, indicating a low risk of bias and minimal concerns regarding applicability.

The meta-analysis of nine studies investigating the diagnostic performance of MDCT in assessing diaphragmatic injury in penetrating trauma revealed a pooled sensitivity of 74% (95% CI: 56%-87%) and a pooled specificity of 92% (95% CI: 79%-97%) (Fig. 2). However, significant heterogeneity was observed in both sensitivity (I2 = 88.85%, 95% CI: 82.94–94.77) and specificity (I2 = 95.72%, 95% CI: 94.1–97.43) across the included studies. The SROC curve demonstrated an AUC of 0.90 (95% CI: 0.88–0.93) (Fig. 3).

Forest plot and summary statistics of diagnostic test accuracy (DTA) meta-analysis of the included studies. CI: Confidence Interval

Summary receiver operating characteristic curve (SROC) of diagnostic test accuracy (DTA) meta-analysis of the included studies. AUC: Area Under the Curve. SENS: Sensitivity. SPEC: Specificity. SROC: Summary Receiver Operating Characteristic

We conducted univariate meta-regression analyses on several covariates, including mean age, mechanism of injury, and gender, to explore potential factors contributing to the heterogeneity observed. However, none of these covariates could account for the observed heterogeneity.

The Fagan plot analysis demonstrated that with pre-test probabilities of 25%, 50%, and 75%, the corresponding positive post-test probabilities were 75%, 90%, and 96%, respectively. Conversely, the negative post-test probabilities were 9%, 22%, and 46%, respectively. These findings are visually represented in Figs. 4, 5, and 6.

Fagan plot analysis using summary sensitivity and specificity results of the meta-analysis of the included studies with a pre-test probability of 25%

Fagan plot analysis using summary sensitivity and specificity results of the meta-analysis of the included studies with a pre-test probability of 50%

Fagan plot analysis using summary sensitivity and specificity results of the meta-analysis of the included studies with a pre-test probability of 75%