Study characteristics

Figure 1 illustrates the systematic screening process conducted in this study. Initially, we searched 139 potentially relevant studies (PubMed, 21; Web of Science, 31; Embase, 27; Cochrane Library database, 46; Clinical Trials gov.14). After removing 67 duplicate studies, the remaining documents were subjected to comprehensive title and abstract screening. Subsequently, 48 studies that were deemed irrelevant, such as those that were non-randomized controlled trials, reviews, animal experiments, and those with inconsistent research purposes, were excluded. The articles were further assessed for eligibility, and 24 full texts were shortlisted. Ultimately, only 9 studies satisfied the inclusion criteria and were considered for meta-analysis. The reasons for excluding the remaining 15 studies were as follows: 4 were clinical registration trials, 2 were cardiac surgery, 5 did not report the results of postoperative atelectasis, 1 reported atelectasis was diagnosed by chest radiograph instead of lung ultrasound, and 3 performed ultrasound-guided or ultrasound-assessed LRM on both the experimental group and the control group. Finally, 9 randomized controlled trials were included in 443 patients (Acosta et al. 2018; Park et al. 2021; Lee et al. 2020; Song et al. 2017; Liu et al. 2022; Jang et al. 2020; Yang et al. 2021; Acosta et al. 2021; Acosta et al. 2020). The detailed information of the included study is shown in Table 2.

Fig. 1

Flow chart of study screening

Table 2 Study characteristics

Table 2 outlines the fundamental characteristics of the included study. We included RCTs with ultrasound-guided LRM. These investigations need to use lung ultrasonography to evaluate atelectasis, count the number of cases in the experimental and control groups, and determine whether the difference is statistically significant. Our study defined the experimental group as those receiving ultrasound-guided LRM.

Figure 2 presents the results of the quality evaluation conducted using Review Manager 5.3. We assessed the integrity of outcome data and the risk of selective reporting of research results as low in all trials, indicating low risk in random sequence generation. However, Three trials (Acosta et al. 2018; Yang et al. 2021; Acosta et al. 2020) did not provide information about random assignment concealment. Additionally, two studies (Liu et al. 2022; Acosta et al. 2020) did not apply blinding measures to the researcher or the subject, while four studies (Acosta et al. 2018; Song et al. 2017; Yang et al. 2021; Acosta et al. 2021) omitted descriptions of whether blinding procedures were implemented for the researcher or the subject. Furthermore, five studies (Acosta et al. 2018; Song et al. 2017; Yang et al. 2021; Acosta et al. 2021; Acosta et al. 2020) failed to report whether the outcome evaluator was blinded. Nonetheless, nine studies (Acosta et al. 2018; Park et al. 2021; Lee et al. 2020; Song et al. 2017; Liu et al. 2022; Jang et al. 2020; Yang et al. 2021; Acosta et al. 2021; Acosta et al. 2020) pose a low risk of other biases.

Fig. 2

Evaluation of risk bias for included RCTs: a percentage plot of seven types of bias for the included studies; b summary of bias for each study

Grading evidence quality

The evaluation of the quality of evidence using GRADEpro is presented in Table 3. The assessment is based on several key parameters such as risks of bias, inconsistency, indirection, imprecision, and publication bias, and the evidence is subsequently categorized into four categories: high, medium, low, and extremely low. The risk of bias was evaluated by considering a total of 18 indicators, all of which were assessed as not serious. Due to I2, the inconsistency of LUS and LUS of each part were rated as serious > 50%, indicating unacceptable heterogeneity.

Table 3 Quality of evidence by GRADE

In terms of indirectness and imprecision, since all studies directly compared ultrasound-guided LRM with a certain sample size and a control group, the indicators were classified as non-serious. Given the aforementioned assessment, we can confidently state high confidence in all the results.

Primary outcomesIncidence of postoperative atelectasis

The incidence of postoperative atelectasis was reported in nine studies with 443 patients (Acosta et al. 2018; Park et al. 2021; Lee et al. 2020; Song et al. 2017; Liu et al. 2022; Jang et al. 2020; Yang et al. 2021; Acosta et al. 2021; Acosta et al. 2020). Among them, there were 222 cases in the ultrasound-guided LRM group and 221 cases in the control group (Fig. 3). The incidence of postoperative atelectasis in the ultrasound-guided LRM group was lower than in the control group. Low heterogeneity was observed in the results (RR 0.31,95% CI 0.25 to 0.40,p < 0.05,heterogeneity P > 0.10,I2 = 37%).

Fig. 3

Forest plot for the incidence of postoperative atelectasis between the ultrasound-guided and control groups. CI = confidence interval, RR = risk ratio, M-H = methods of merging dichotomous variables

Subgroup analysis of postoperative atelectasis by LRM or non-LRM used in the control group

Whether the control group used LRM or not was reported in the nine studies with 443 patients (Acosta et al. 2018; Park et al. 2021; Lee et al. 2020; Song et al. 2017; Liu et al. 2022; Jang et al. 2020; Yang et al. 2021; Acosta et al. 2021; Acosta et al. 2020), of which 126 patients in the control group used LRM in two studies (Park et al. 2021; Lee et al. 2020), and 317 patients in five studies did not use LRM (Acosta et al. 2018; Song et al. 2017; Liu et al. 2022; Jang et al. 2020; Yang et al. 2021; Acosta et al. 2021; Acosta et al. 2020). The findings demonstrated that whether the control group did not employ LRM or did so with non-ultrasound-guided LRM, the incidence of postoperative atelectasis was reduced in patients with ultrasound-guided LRM (compared to non-LRM in the control group: RR = 0.33,95% CI 0.25–0.43,P < 0.05,heterogeneity p > 0.10,I2 = 30%,compared to LRM in the control group: RR = 0.26,95% CI 0.15–0.46,P < 0.05,heterogeneity P < 0.10,I2 = 73%) (Fig. 4).

Fig. 4

Forest plot for subgroup analysis of the incidence of postoperative atelectasis between the ultrasound-guided and control groups. Grouped by LRM or non-LRM used in the control group: compared to LRM in the control group, compared to non-LRM in the control group. CI = confidence interval, RR = risk ratio, M-H = methods of merging dichotomous variables

Subgroup analysis of the effect of adults and children on postoperative atelectasis

The incidence of postoperative atelectasis in adults and children was reported in the nine studies with 443 patients (Acosta et al. 2018; Park et al. 2021; Lee et al. 2020; Song et al. 2017; Liu et al. 2022; Jang et al. 2020; Yang et al. 2021; Acosta et al. 2021; Acosta et al. 2020). The results showed that ultrasound-guided LRM reduced the incidence of postoperative atelectasis in adults (RR = 0.49,95% CI 0.36 to 0.67,p < 0.05,heterogeneity p > 0.10,I2 = 0%) (Fig. 5). Using ultrasound-guided LRM also reduces the incidence of postoperative atelectasis in children (RR = 0.23,95% CI 0.17 to 0.33,p < 0.05,heterogeneity p > 0.10,I2 = 0%). It may be more effective in children than adults (heterogeneity p < 0.05,I2 = 89.6%; P for subgroup differences < 0.01).

Fig. 5

Forest plot for subgroup analysis of the incidence of postoperative atelectasis between the ultrasound-guided and control groups. Grouped by age: age ≥ 18 years (adult), age < 18 years (children). CI = confidence interval, RR = risk ratio, M-H = methods of merging dichotomous variables

Subgroup analysis was performed in the present study to investigate the effect of PEEP after LRM on postoperative atelectasis

In 9 studies, 443 patients (Acosta et al. 2018; Park et al. 2021; Lee et al. 2020; Song et al. 2017; Liu et al. 2022; Jang et al. 2020; Yang et al. 2021; Acosta et al. 2021; Acosta et al. 2020) reported the incidence of postoperative atelectasis with PEEP after LRM. Among 7 studies, 360 patients maintained the same PEEP before and after LRM (Park et al. 2021; Lee et al.