3.1. Study process
After a preliminary search found 2287 publications, only 27 of them were ultimately included. The flow chart for the literature screening procedure is shown in Fig. 1.
3.2. Study characteristics
There were 3 case-control studies, 23 retrospective cohort studies, and 1 prospective cohort research among the included studies. 12 of these studies were published in English and 15 in Chinese. With a cumulative case group of 1,435 and a case-control group of 19,405, the study's total sample size was 20,840. Pneumonia ranged in incidence or prevalence from 2.1% to 26.38%. The research, most of which were from China (n = 19) and Japan (n = 8), were published between 2008 and 2023. The main methods for diagnosing pneumonia include imaging, lab work, and clinical presentation. The main features of the 27 assessed studies are shown in Table 1.
3.3. Quality evaluation of included studies
Of the included studies, twenty-four were found to be of high quality by NOS, while the other three were found to be of moderate quality (Table 2).
3.4. Risk factors
Among the 27 studies, multivariate analysis identified 35 risk factors in total. A risk factor was selected and combined for the meta-analysis if it was reported in two or more studies. In the end, 16 risk factors were found for pneumonia after radical gastrectomy for gastric cancer were identified.
3.4.1. Smoking history
Eight studies included 2,615 patients undergoing radical gastrectomy; the data included information on a history of smoking (Guo et al., 2016; Han X et al., 2017; Liu et al., 2017; Tan et al., 2023; Xiao et al., 2020, 2023; Xie et al., 2020; Yu et al, 2022). In Figure 2A, the forest plot is shown. For this meta-analysis, a fixed effects model was used because there was no discernible heterogeneity between the trials (I2 = 26%, P = 0.22). Across the eight investigations, the overall OR was 2.71, with a 95% CI ranging from 2.09 to 3.50. The combined data showed that in patients undergoing radical gastrectomy for gastric cancer, a history of smoking was an independent predictor of pneumonia (P < 0.001).
3.4.2. Prolonged postoperative nasogastric tube retention
Figure 2B illustrates the inclusion of prolonged postoperative nasogastric tube retention in six eligible studies (n = 1691) (Bai et al., 2020; Han et al., 2017; Lu et al., 2023; Tan et al., 2023; Xiao et al., 2023; Xiao H et al., 2014). The random-effects model revealed a significant association between prolonged postoperative nasogastric tube retention and the risk of pneumonia after radical gastrectomy for gastric cancer, as indicated by the overall pooled OR of 2.25 (95% CI: 1.36 to 3.72, P = 0.002). The studies exhibited significant heterogeneity (I2 = 63%, P = 0.02). A sensitivity analysis encompassing five studies was conducted. Pooled analysis indicated statistical significance (OR:2.88, 95% CI:1.93–4.28, P<0.001; I2 = 24%, P=0.26). Shown in Figure 3A.
3.4.3. Intraoperative bleeding ≥ 200 ml
Data on intraoperative bleeding ≥ 200 ml were reported by five studies (n = 6,789) (Han et al., 2017; Tan et al., 2023; Tu Ru Hong et al., 2017; Xiao H et al., 2014; Yu and Gao, 2022). Testing revealed statistical heterogeneity amongst the studies (I2 = 79%, P = 0.0009). An independent association between intraoperative bleeding ≥ 200 ml and pneumonia after radical gastrectomy for gastric cancer is shown in Figure 2C, which presents the results of the random-effects model. The OR was 2.21, and the 95% CI ranges from 1.15 to 4.24 (P = 0.02). Following the exclusion of Tu's trial (Tu Ru Hong et al., 2017), sensitivity analyses revealed a decrease in heterogeneity (I2 = 36%, p = 0.19) (see Fig 3B), while the combined odds ratio remained statistically significant (OR:2.51, 95% CI: 1.64–3.83, p < 0.0001).
3.4.4. Diabetes mellitus
Nine studies (n = 3,055) provided data on diabetes mellitus (Bai et al., 2020; Lu et al., 2023; Miki Yuichiro et al., 2016; Tan et al., 2023; Wang et al., 2020; Xiao et al., 2023; Xiao H et al., 2014; Yu and Gao, 2022; Zhao et al., 2019). According to the significant heterogeneity in the number of trials (I2 = 96%, P < 0.00001), the summary effect was estimated using a random-effects model. With a pooled OR of 4.58 and a 95% CI ranging from 1.84 to 11.38, diabetes mellitus was found to significantly increase the incidence of pneumonia after radical gastrectomy for gastric cancer (P = 0.001) (see Fig. 2D). Sensitivity analyses were conducted on the included studies, and upon exclusion of Zhao's study (Zhao et al., 2019), the findings revealed a decrease in heterogeneity (I2 = 24%, p = 0.24) while maintaining a statistically significant combined OR: 2.67, 95% CI: 2.13–3.36 (p < 0.00001) (Fig. 3C).
3.4.5. Excessive duration of surgery
Eight studies totaling 10,136 people were included in this meta-analysis and systematic review to provide information on the prolonged duration of surgical (Bai et al., 2020; Guo et al., 2016; Lu et al., 2023; Tan et al., 2023; Tu Ru Hong et al., 2017; Wang et al., 2020; Xiao Haifan et al., 2019; Yu and Gao, 2022). As a result of the notable heterogeneity between the trials (I2 = 90%, P < 0.0001), this independent meta-analysis used a random-effects model. Referring to Figure 2E, the pooled OR was determined to be 1.51, with a 95% CI ranging from 1.25 to 1.83 (P < 0.0001). The results showed a strong correlation between the incidence of pneumonia after a radical gastrectomy for gastric cancer and the duration of the surgical procedure.
3.4.6. Perioperative blood transfusion
Data on perioperative blood transfusion were obtained from six studies with a total of 8,543 participants (Bai et al., 2020; Liu et al., 2017; Miki Yuichiro et al., 2016; Xiao Haifan et al., 2019; Yasuhiko et al., 2008; Zhao et al., 2019). Statistical testing revealed significant heterogeneity among the studies (I2 = 90%, P < 0.00001). As shown in Figure 2F, the random-effects model demonstrated an independent association between perioperative blood transfusion and pneumonia following radical gastrectomy for gastric cancer, with an OR of 4.21 and a 95% CI ranging from 2.51 to 7.06 (P < 0.00001).
3.4.7. Male gender
Two studies (n = 779) examining the association between pneumonia and male sex found statistical significance without heterogeneity, as validated by a meta-analysis (Satoshi et al., 2017; Yasuhiko et al., 2008) (I2 = 0, P = 0.55). The overall OR was 3.56, with a 95% CI ranging from 1.50 to 8.42. The findings demonstrated that, after a radical gastrectomy for gastric cancer, male sex was an independent prognostic factor for pneumonia (Figure 2G).
3.4.8. Advanced age
Age-related data were extracted from eight studies involving a total of 10,760 individuals (Guo et al., 2016; Miki Yuichiro et al., 2016; Redondo et al., 2016; Ryuichiro et al., 2021; Tu Ru Hong et al., 2017; Xiao et al., 2020; Xie et al., 2020; Yu and Gao, 2022). This independent meta-analysis utilized a random-effects model due to the substantial heterogeneity observed among the trials (I2 = 94%, P < 0.00001). The pooled OR was estimated at 1.91 (Fig. 2H), with a 95% CI ranging from 1.42 to 2.58. The findings revealed a significant association between the advanced age of subjects and the incidence of pneumonia following radical gastrectomy for gastric cancer (P < 0.0001).
3.4.9. Nutritious status
Nutritional status was evaluated in four eligible studies (n = 3,134), as depicted in Figure 2I(Bai et al., 2020; Guo et al., 2016; Jun et al., 2016; Zhao et al., 2019). The random-effects model demonstrated a significant association between nutritional status and the risk of postoperative pneumonia following radical gastrectomy for gastric cancer, with a pooled OR of 2.62 and a 95% CI ranging from 1.55 to 4.44 (P = 0.0003). Substantial heterogeneity was observed among the studies (I2 = 71%, P = 0.02). Sensitivity analyses, excluding Zhao's trial (Zhao et al., 2019), showed a reduction in heterogeneity (I2 = 0%, P = 0.39) (Figure 3D), while the combined odds ratio remained statistically significant (OR: 1.89, 95% CI: 1.32–2.71, P = 0.0005).
3.4.10. Total gastrectomy
Relevant data were provided by three trials, totaling 4,033 individuals who had total gastrectomy (Jun et al., 2016; Michita et al., 2020; Redondo et al., 2016). The results are shown in a forest plot in Figure 2J. For this meta-analysis, a fixed-effects model was used because there was no discernible heterogeneity between the trials (I2 = 0%, P = 0.91). Across the six trials, the overall OR was 2.59, with a 95% CI ranging from 1.83 to 3.66. Based on the combined data, total gastrectomy was found to be a significant independent predictor of postoperative pneumonia after radical gastrectomy for gastric cancer (P < 0.00001).
3.4.11. COPD
Three studies, involving 4,402 patients who underwent radical gastrectomy for gastric cancer, provided data on patients with chronic obstructive pulmonary disease (COPD) (Redondo et al., 2016; Ryuichiro et al., 2021; Xiao Haifan et al., 2019). The pooled analysis revealed that COPD was a significant risk factor for postoperative pneumonia following radical gastrectomy for gastric cancer, with an OR of 4.72 and a 95% CI ranging from 3.80 to 5.86 (P < 0.00001). The heterogeneity test did not reveal any statistically significant heterogeneity (I2 = 0%, P = 0.39) (Figure 2K).
3.4.12. Impaired pulmonary function
There were three studies (Liu et al., 2017; Miki Yuichiro et al., 2016; Xie et al., 2020) (n = 1,756) that provided information on pulmonary function. Significant heterogeneity amongst the trials was found by statistical analysis (I2 = 92%, P < 0.00001). With an OR of 1.29 and a 95% CI spanning from 0.82 to 2.03, the random-effects model did not reveal a correlation between impaired pulmonary function and postoperative pneumonia after radical gastrectomy for gastric cancer (P =0.83), as illustrated in Figure 2L. Sensitivity analyses that excluded Liu's data(Liu et al., 2017) revealed a statistically significant combined odds ratio with reduced heterogeneity(I2 = 0%, P = 1.00) (OR: 2.72, 95% CI: 1.58–4.69, P = 0.0003) (Figure 3E).
3.4.13. History of pulmonary disease
Three separate trials totaling 2,972 participants who provided information on a history of pulmonary disease were included in this analysis (Guo et al., 2016; Michita et al., 2020; Wang et al., 2020). The meta-analysis utilized a random-effects model because of the significant heterogeneity among the studies (I2 = 79%, P = 0.008). With a 95% CI spanning from 1.17 to 2.21 (P = 0.003), the calculated pooled OR was 1.61 (Figure 2M). The results showed a significant correlation between a history of lung illness and pneumonia after radical gastrectomy for gastric cancer. After Wang's trial (Wang et al., 2020) was excluded from the analysis, sensitivity analyses showed a decrease in heterogeneity (I2 = 42%, P = 0.19) (Figure 3F), the combined OR: 1.52, 95%CI: 1.10–2.09 (P = 0.01) was still statistically significant.
3.4.14. D2 lymphadenectomy
Two studies (n = 1,030) (Ryuichiro et al., 2021; Satoshi et al., 2017) investigating the relationship between pneumonia and D2 lymphadenectomy showed statistically significant results without heterogeneity, as confirmed by a meta-analysis (I2 = 0, P = 0.73). With a 95% CI spanning from 2.29 to 7.49, the combined OR was 4.14. These results indicate that, after radical gastrectomy for gastric cancer, D2 lymphadenectomy is an independent predictor of pneumonia (Figure 2N).
3.4.15. Hypertensive
Results from two trials (n = 1,739) examining the association between pneumonia and hypertension proved statistically significant and showed no heterogeneity. For this meta-analysis, a fixed-effects model (I2 = 0, P = 1.0) was used. The combined OR was estimated to be 2.21, with a 95% CI ranging from 1.29 to 3.79. These results imply that after a radical gastrectomy for gastric cancer, hypertension is an independent predictor of pneumonia (P =0.004) (Figure 2O).
3.4.16. Open operative procedure
Data on the practice of open operative procedures were reported by two researchers (n = 7,796) (Xiao Haifan et al., 2019; Tu Ru Hong et al., 2017). A statistical analysis of the studies showed that there was a high amount of heterogeneity (I2 = 98%, P < 0.00001). The random effects model results are shown in Figure 2P. They show that there is no association between an open surgical procedure and pneumonia after a radical gastrectomy for gastric cancer (OR = 2.81, 95% CI: 0.78 to 10.19, P = 0.11).
3.4.17. Other relevant risk factors
Preoperative comorbidities (Yao Z et al., 2019), wound pain (Yao Z et al., 2019), tumor located at middle third (Xiao Haifan et al., 2019), preoperative sarcopenia(Kamiya A et al., 2022), modified frailty index(Yongsheng et al., 2021), ASA classification≥3(Tu Ru Hong et al., 2017), tumor diameter(Tu Ru Hong et al., 2017), advanced stage(Jun et al., 2016), time to first diet(Redondo et al., 2016), postoperative hospital stay(Redondo et al., 2016), poor performance status(Haruhiko et al., 2021), cardia-non-preserving gastrectomy(Haruhiko et al., 2021), systemic Inflammation score(Michita et al., 2020), mechanical ventilation time(Guo et al., 2016), preoperative use of antimicrobials(Xiao et al., 2020), excessive postoperative bed rest(Xiao et al., 2020), pathologic MMP-2 in gastric cancer(Xiao et al., 2020), pathology of gastric cancer TIMP1(Xiao et al., 2020) and preoperative chemotherapy(Xie et al., 2020) were the other 19 individual risk factors identified via multiple regression analysis. Our meta-analysis did not include these 19 risk factors.
3.5. Publication bias
Since none of the studies included in this meta-analysis met the minimum threshold of ten articles, a test for publication bias was not conducted.