3.1 Search results
A total of 13918 literatures were searched and 6431 articles were excluded because of duplicate results. Among the remaining 7487 records, 7362 were excluded through title or abstract screening, leaving 125 records for further full-text screening. Finally, 81 literatures were included in our meta-analysis (Figure S1). Three studies assessed two outcomes resulting in a total of 84 reports [7, 20, 21]. Of these, 7 studies reported the incidence risk for all sites combined or site-specific cancer [7-9, 20-23], 50 studies examined the impact of splenetomy on the progression of gastric carcinoma [24-73], 4 studies investigated the role of splenetomy in the progression of colorectal cancer [74-77], 6 studies examined the impact of splenetomy on the progression of esophago‑gastric junction cancer [78-83], and 14 studies examined the impact of splenetomy on the progression of hepatocellular carcinoma [84-97]. Among 50 studies for gastric carcinoma contained 7 RCTs [30, 32, 42, 48, 50, 52, 68]. The detailed study characteristics and the quality of studies are shown in Table S1-S6.
3.2 Associations between splenectomy and the risk of cancer incidence
The meta-analysis results of 8 reports showed that splenectomy was associated with a significantly increased risk of overall cancer incidence from random-effects model (RR: 1.476, 95% CI: 1.182–1.843; Table 1, Figure S2), with a high heterogeneity (I2= 89.6%, P < 0.001). Similarly, splenectomy also increased the incidence risk of Non Hodgkin lymphoma (NHL, RR: 3.867, 95% CI: 2.055–7.277; I2= 71.4%; Table 1, Figure S3) and Hodgkin lymphoma (HL, RR: 6.714, 95% CI: 1.502–30.003; I2= 65.9%; Table 1, Figure S4) with high heterogeneity. For other specific cancers, significant associations were observed in lung cancer (RR: 1.28, 95% CI: 1.145–1.440; I2= 15.0%; Table 1, Figure S5), leukemia (RR: 5.376, 95% CI: 4.431–6.523; I2= 34.1%; Table 1, Figure S6), breast cancer (RR: 1.574, 95% CI: 1.105–2.242; I2= 0%; Table 1, Figure S7), oral and pharyngeal cancer (RR: 1.285, 95% CI: 1.034–1.598; I2= 0%; Table 1, Figure S8), skin cancer (RR: 1.760, 95% CI: 1.004–13.088; I2= 0%; Table 1, Figure S9), esophagus cancer (RR: 2.058, 95% CI: 1.529–2.770; I2= 41.7%; Table 1, Figure S10), prostate cancer (RR: 1.253, 95% CI: 1.070–1.467; I2= 6.0%; Table 1, Figure S11), bladder cancer (RR: 1.300, 95% CI: 1.005–1.683; I2= 0%; Table 1, Figure S12), gastric cancer (RR: 1.920, 95% CI: 1.394–2.643; I2= 30.2%; Table 1, Figure S13), pancreas cancer (RR: 1.779, 95% CI: 1.228–2.577; I2= 0%; Table 1, Figure S14), and brain and nervous system cancer (RR: 3.941, 95% CI: 1.080–14.380; I2= 0%; Table 1, Figure S15) with low heterogeneity. However, splenectomy had no effect on the incidence risk of liver cancer (RR: 1.436, 95% CI: 0.469–4.397; I2= 97.7%; Table 1, Figure S16), colorectal cancer (RR: 1.152, 95% CI: 0.919–1.444; I2= 14.9%; Table 1, Figure S17), and kidney cancer (RR: 1.121, 95% CI: 0.724–1.736; I2= 15.2%; Table 1, Figure S18). Furthermore, there was no publication bias among included studies as demonstrated by Egger´s test (Table 1).
The reasons of splenectomy included trauma and no trauma, such as hypersplenism, benign hematologic disorders, and other no trauma reasons. Subgroups analyses were performed where there were three or more studies based on reasons of splenectomy (trauma, protal hypertension, and other no trauma). Interestingly, subgroup analyses based on reasons of splenectomy showed that splenectomy for trauma increased the incidence risk of liver cancer (RR: 2.068, 95% CI: 1.525–2.805; I2= 0%; Table 2, Figure S16) while splenectomy for protal hypertension decreased the incidence risk of liver cancer (RR: 0.524, 95% CI: 0.423–0.649; I2= 0%; Table 2, Figure S16) with low heterogeneity. Our results from subgroup analyses also found only splenectomy for no trauma reasons increased the incidence risk of breast cancer (RR: 1.639, 95% CI: 1.046–2.568; I2= 0%; Table 2, Figure S7) and skin cancer (RR: 1.915, 95% CI: 1.020–3.596; I2= 0%; Table 2, Figure S9) while only splenectomy for trauma increased the incidence risk of oral and pharyngeal cancer (RR: 1.265, 95% CI: 1.011–1.583; I2= 0%; Table 2, Figure S8) and prostate cancer (RR: 1.236, 95% CI: 1.052–1.460; I2= 0%; Table 2, Figure S11). The results of incidence risk of bladder cancer changed from significant to non-significant in subgroup analyses (Table 2, Figure S12). The results of subgroup analyses on reasons of splenectomy for the incidence of all cancer, NHL, lung cancer, leukemia, esophagus cancer, gastric cancer, pancreas cancer, and colorectal cancer were consistent with the main effect and subgroup analyses partially or completely resolved heterogeneity (Table 2, Figure S2, S3, S5, S6, S10, S13, S15, S17).
3.3 Associations between splenectomy and the 5-year OS of cancer patients
We next investigated the associations between splenectomy and the 5-year OS of gastric cancer, liver cancer, colorectal cancer, and esophageal and esophago‑gastric junction cancer. The meta-analysis results of 39 reports showed that splenectomy was associated with a significantly lower 5-year OS of gastric cancer based on a random-effects model (RR: 0.787, 95% CI: 0.734–0.844; Table 3, Figure S19), with a high heterogeneity (I2= 60.1%, P < 0.001). Similarly, splenectomy was also associated with a significantly lower 5-year OS of colorectal cancer (RR: 0.769, 95% CI: 0.555–0.859; I2= 0%; Table 3, Figure S20), esophageal and esophago‑gastric junction cancer (RR: 0.541, 95% CI: 0.397–0.737; I2= 34.8%; Table 3, Figure S21) based on a fixed-effects model. Interestingly, the meta-analysis results of 10 reports showed that splenectomy was associated with a significantly higher 5-year OS of liver cancer based on a fixed-effects model (RR: 1.173, 95% CI: 1.076–1.279; Table 3, Figure S22), with a low heterogeneity (I2= 39.7%, P = 0.093). Similarly, there was also no publication bias among included studies as demonstrated by Egger´s test (Table 3).
Subgroups analyses were performed for gastric cancer based on study design (RCT and nRCT), surgical method of gastrectomy (total gastrectomy), surgical method of splenectomy (splenectomy and pancreaticosplenectomy), and tumor stage (Ⅰ, Ⅱ, Ⅲ, Ⅳ). Our results indicated that the patients underwent splenectomy had lower 5-year OS only in nRCT (RR: 0.760, 95% CI: 0.712–0.810; I2= 43.2%; Table 4, Figure S19) but RCT (RR: 0.988, 95% CI: 0.882–1.108; I2=23.5%; Table 4, Figure S19). Surgical methods of gastrectomy and splenectomy did not affect the 5-year OS of gastric cancer (Table 4). Interestingly, we found only stage Ⅱ (RR: 0.821, 95% CI: 0.707–0.953; I2= 0%; Table 4, Figure S23) and stage Ⅲ (RR: 0.723, 95% CI: 0.618–0.846; I2= 0%; Table 4, Figure S23) gastric cancer patients underwent splenectomy had lower 5-year OS but stage Ⅰ (RR: 0.936, 95% CI: 0.839–1.044; I2= 0%; Table 4, Figure S23) and stage Ⅳ (RR: 1.048, 95% CI: 0.545–2.014; I2= 29.4%; Table 4, Figure S23).
3.4 Associations between splenectomy and the hospital mortality of cancer patients
There were 26 studies reported the hospital mortality of gastric cancer patients after splenectomy, 9 studies for liver cancer, and 5 studies for esophageal and esophago‑gastric junction cancer. As shown in the results of meta-analysis, gastric cancer patients in splenectomy group showed a significantly higher hospital mortality based on a fixed-effects model, with a low heterogeneity (RR: 1.461, 95% CI: 1.229–1.736; I2= 47.2%; Table 5, Figure S24). There were no significant differences in hospital mortality for liver cancer (RR: 1.572, 95% CI: 0.757–3.080; I2= 0%; Table 5, Figure S25) and esophageal and esophago‑gastric junction cancer (RR: 1.658, 95% CI: 0.948–2.900; I2= 47.7%; Table 5, Figure S26). There was also no publication bias among included studies as demonstrated by Egger´s test (Table 5).
Similarly, subgroups analyses were performed for gastric cancer based on study design (RCT and nRCT), surgical method of gastrectomy (total gastrectomy), and surgical method of splenectomy (splenectomy and pancreaticosplenectomy). Our results indicated that there were no significant differences in hospital mortality for gastric cancer who underwent total gastrectomy with a low heterogeneity (RR: 0.906, 95% CI: 0.544–1.508; I2= 0%, Table 6, Figure S27). Study design and surgical methods of splenectomy did not affect the hospital mortality of gastric cancer (Table S7).
3.5 Associations between splenectomy and the complications after surgery
There were 35 studies yielded a pooled RR of 1.818 (95% CI: 1.523–2.171; Table 6) for the association between the total postoperative complications of gastric cancer patients and splenectomy based on random-effects model analyses. High heterogeneity was observed with I2 = 80.7%. Regarding to specific postoperative complication, our result showed that splenectomy group had higher incidence of anastomotic leakage (RR: 1.853, 95% CI: 1.501–2.288; I2= 0%, Table 6), pulmonary complication (RR: 1.584, 95% CI: 1.257–1.996; I2= 0%, Table 6), intraabdominal abscess (RR: 2.914, 95% CI: 1.672–5.081; I2= 67.2%, Table 6), pancreatic fistula (RR: 5.390, 95% CI: 2.571–11.298; I2= 73.3%, Table 6), and wound infection (RR: 1.461, 95% CI: 1.023–2.088; I2= 0%, Table 6). There was no significant difference between two groups in the incidence of ileus, postoperative bleeding, cardiac complication, liver dysfunction, pancreatits, anastomotic stenosis, renal dysfunction, and sepsis (Table 6). The results of Egger´s test showed there was publication bias in the analysis of total postoperative complications and liver dysfunction.
Our results showed that the splenectomy for liver cancer did not increase the total postoperative complication rate (RR: 1.124, 95% CI: 0.949–1.331; Table 7) based on fixed-effects model analyses, with a low heterogeneity between studies (I2= 15.7%). Regarding to specific postoperative complication, we found splenectomy for liver cancer increased the incidence of postoperative bleeding (RR: 2.230, 95% CI: 1.077–4.617; I2= 0%, Table 7) and portal vein or splenic vein thrombosis (PSVT; RR: 23.912, 95% CI: 6.782–80.306; I2= 21.6%, Table 7) with a low heterogeneity based on fixed-effects model analyses. There was no significant difference between two groups in the incidence of pleural effusion, bile leakage, ascites, wound infection, upper gastrointestinal haemorrhage, pulmonary infection, and liver failure (Table 7). As showed in Table 8, splenectomy for esophageal and esophago‑gastric junction cancer increased the incidence of leakage (RR: 1.935, 95% CI: 1.116–3.355; I2= 72.3%, Table 8) with a high heterogeneity based on random-effects model analyses. There was no significant difference between two groups in the incidence of pulmonary complications, cardiac complications, wound infection, and postoperative bleeding (Table 8).