Literature Search Findings
Two researchers (X.H.Z Y.W.H) searched PubMed, Embase, Web of science, Scopus, Cochrane, Clinicalkey, and Microsoft Academic databases, and a total of 2533 studies were obtained. X.H.Z used EndNote X9 to remove 1268 duplicate studies. We excluded 12 records marked as ineligible by automation tools and 2 studies due to incomplete information. And two investigators independently reviewed the initially included studies and excluded 542 non-clinical studies (Review: 349; Meta-analysis: 26; Case report: 42; Letter: 28; Animal experiments: 26; Laboratory studies: 26; Guidelines or Conference Abstract: 45). Based on the number of "stars" marked in EndNote X9, we screened clinical studies in the second stage, and 126 studies can be reviewed in full text. After excluding 99 studies, 27 included articles were identified. We evaluated the quality of NRCT among them, three studies with a score of <12 were excluded[18–22]. This review finally included 22 articles. The literature search findings are represented in PRISMA_2020_flow_diagram (Fig. 1).
Study and Patient Characteristics
The characteristics of the included studies are summarized in Table 2. A total of 19 studies (RCT:12 NRCT:10) with 2097 patients were included in this review. Among them, 9 studies were published after 2015. Most of the included studies are from Asia, including 8 from mainland China[23–29] and Taiwan ROC[30], 10 from Japan[31–39] and Korea[40]. The other three studies are from Belarus[41], Italy[42], Ukraine[43], and Brazil[44]. Matched factors mainly include the following: age, gender, histology, stage, lymphadenectomy, type of gastrectomy. If the above content is reported in the research, the investigators will use the corresponding number of the representative to express it. The included studies all reported the patient’s age, gender, and stage of gastric cancer. For the staging of gastric cancer, 8 studies[25, 27, 30, 34, 35, 37, 40, 41] reported the TMN staging situation, and 15 studies[23, 24, 26, 28, 29, 31–33, 36, 38–40, 42–44] listed the number of patients in each stage. The studies of Liu[26] et al., Cui[28] et al., and Reutovich[41] et al. included only stage III patients.
Intervention characteristics
Two investigators summarized the intervention characteristics of the included studies in Table 3. In the included studies, the choice of chemotherapeutics for HIPEC was mainly MMC or Cisplatin or a combination of the two. The HIPEC protocol chosen by cui and his colleagues[28] is as follows: Cisplatin (60 mg/m2) and 0.9% sodium chloride solution (3000 mL) at 1,4days; fluorouracil (0.75 g) and 0.9% sodium chloride solution (3000 mL) at 2 ,3days; Dexamethasone (10 mg) and 2% lidocaine (10 mL) 1-4days; temperature: 41‑43˚C. Xie[25] et al. used Cisplatin (50 mg/L) and 0.9% sodium chloride solution for 60min, temperature: 42-43˚C. Reutovich[41] et al. chose Ringer's solution (5-6L), cisplatin (50 mg/m2) and doxorubicin (50 mg/m2), Temperature: 42˚C. Zhu[23] et al.'s solution is: Cisplatin (75 mg/m2) and 2000ml 0.9% sodium chloride solution for 60min, temperature: 41.5-42.5˚C. Beeharry[29] et al. and his colleagues used cisplatin (50mg/L) for 60 min, temperature: 41-43˚C. The researchers used MMC chemotherapy in 8 studies. The method of Hamazoe[33] et al. is: MMC (10 µg/ml) in 0.9% saline solution for 50-60min, temperature: 44-45˚C. Koga[32] et al. chose MMC (8-10 mg/L) in 2000ml saline solution for 50-60min, temperature: 44‑45˚C. Diniz[44] et al. used MMC (38 mg/m2) in saline solution for Time: 90min, temperature: 41-42˚C. Fujimoto and his colleagues[35] chose MMC (10 mg/ml), and 0.9% sodium chloride solution (3-4L) for 120min, temperature: 44.5‑45˚C. Kim[40] et al. chose MMC (40 mg) and dialysis solution (4000cc) for 60 min, temperature: 44.5-45.7˚C in RCT. Ikeguchi[37] et al. chose MMC (80-100 mg/m2) for 50-60 min, temperature: 44-45˚C. Takahashi[38] et al method is: MMC (50 mg), Activated carbon (375mg), Saline (100ml). In addition, 7 studies chose the HIPEC scheme of MMC+ cisplatin. Fujimura[34] et al. used MMC (30 mg) and cisplatinum (300 mg) in 10 liters saline solution for 60min, temperature: 41‑42˚C. Rosa[42] et al. used Cisplatin (75 mg/m2), MMC (15 mg/m2), and 0.9% sodium chloride solution (2L/m2) for 90min, temperature: 41‑42˚C. Yonemura[31] et al. and his colleagues used Cisplatin (300mg), MMC (30mg), and 0.9% sodium chloride solution (8L) for 60min, temperature: 41.5-43.5˚C. Hirose[36] et al. reported cisplatin (100mg), MMC (20mg) and etoposide (100mg), temperature: 41‑44.5˚C. Kunisaki[39] et al. chose cisplatin (150mg), MMC (15mg), and etoposide (150mg) in 5 to 6 L physiologic saline for 40min, temperature: 42-43˚C. Kang[30] et al. chose cisplatin (30 mg/L), MMC (10 mg/L), and etoposide (20 mg/L) in 3-4 L of lactated Ringer solution for 60min, temperature: 41-43˚C. Yarema[43] et al. used MMC (12.5 mg/m2), cisplatin (75 mg/m2) for 90min, temperature: 41-43.6˚C. In addition, the study by Liu[26] et al. used Oxaliplatin (100 mg/m2) and 0.9% sodium chloride solution (3000 mL) at 1,4days; fluorouracil (0.75 g) and 0.9% sodium chloride solution (3000 mL) at 2 ,3days; Dexamethasone (10 mg) and 2% lidocaine (10 mL) 1-4days, temperature: 42‑44˚C. Zhong[24] et al. used Lobaplatin (50 mg/m2) and 3000mL 5% intravenous glucose solution for 60min, Temperature: 43˚C.
Risk of Bias Assessment and Study Quality
Two investigators used RevMan 5.3 to assess the risk of bias for 12 RCTs. The evaluation result is shown in Fig. 2, Fig. 3. Among all the included RCTs, 3 studies[27, 29, 38] reported the method of random sequence generation, Beeharry[29] et al. and Takahashi[38] et al. reported on the concealment of random sequences. Hirose[36] et al. and Kim[40] et al. pointed out in the study that random sampling cannot be achieved due to the particularity of interventions. Although we evaluate this as high risk, this cannot be the basis for excluding these two studies. Only Beeharry et al. reported blinding the researchers responsible for data statistics, and none of the other studies mentioned blinding. In addition, the simple size is small in two studies[32, 36], and there may be a risk of reporting bias.
Meta-analysis and Synthesis
Overall 1-year survival (Fig. 4.1A)
Four studies (2RCTs, 3NRCTs), 415 patients were reported overall 1-year survival[25, 26, 28, 34, 43]. Analyzing under the random effects model, the overall heterogeneity (I2=62%) is acceptable. The Overall 1-year survival rate was significantly favorable to the HIPEC (OR=5.10, 95%CI = 1.41-18.45).
Overall 3-year survival (Fig. 4.1B)
Ten studies (7RCTs, 4NRCTs), 1049 patients were reported overall 3-year survival[24–28, 32, 34, 36, 38, 39, 41]. Analyzing under the random effects model, the overall heterogeneity (I2=57%) is acceptable. The Overall 3-year survival rate was significantly favorable to the HIPEC (OR=2.47, 95%CI = 1.58-3.85).
Overall 5-year survival (Fig. 4.2)
Seven studies (4RCTs, 5NRCTs), 997 patients were reported overall 5-year survival[26, 30, 31, 33, 36, 37, 39, 40, 42]. Analyzing under the random effects model, the overall heterogeneity (I2=56%) is acceptable. The Overall 5-year survival rate was significantly favorable to the HIPEC (OR=1.96, 95%CI = 1.27-3.04).
Overall 3-year survival in different HIPEC ways (Fig. 5)
Eight studies (3Cisplatin, 2MMC, 3Cisplatin + MMC), 696 patients were reported overall 3-year survival in different HIPEC ways. Analyzing under the random effects model, the overall heterogeneity (I2=64%) is acceptable. The overall 3-year survival in different HIPEC ways was significantly favorable to the HIPEC (OR= 2.41, 95%CI = 1.32-4.40).
Overall recurrence rate (Fig. 6A)
Seven studies, 807 patients were reported overall recurrence rate[23, 25, 26, 28, 36, 40, 44]. Using random effects model analysis, the heterogeneity is significant. Therefore, we conducted a sensitivity analysis and finally excluded studies[44] that caused significant heterogeneity. Six studies with 538 patients were evaluated. Using fixed effects model analysis, the heterogeneity is no longer significant (I2=27%). The overall recurrence rate was significantly favorable to the HIPEC (OR=0.43, 95%CI = 0.29-0.65).
Peritoneal recurrence rate (Fig. 6B)
Eleven studies, 1119 patients were reported peritoneal recurrence rate[23–25, 29, 36, 37, 39, 40, 42, 43]. Using random effects model analysis, the heterogeneity is significant. The previous sensitivity analysis has excluded study with significant heterogeneity[44]. Using random effects model analysis, the heterogeneity is no longer significant (I2=53%). The peritoneal recurrence rate was significantly favorable to the HIPEC (OR=0.26, 95%CI = 0.15-0.48).
Complication: myelosuppression (Fig. 7A).
In six studies, 646 patients reported the incidence of postoperative myelosuppression[25, 28, 29, 35, 38, 40]. Analyzing under the fixed effects model, the overall heterogeneity (I2=0%) is not significant. The overall effect is not significantly different (OR=1.37, 95%CI = 0.77-2.42).
Complication: leakage (Fig. 7B).
In twelve studies, 1107 patients reported the incidence of postoperative leakage[24, 25, 27, 29, 32, 33, 35, 36, 38–40, 42]. Analyzing under the fixed effects model, the overall heterogeneity (I2=0%) is not significant. The overall effect is not significantly different (OR=0.90, 95%CI = 0.50-1.63).
Complication: intestinal obstruction (Fig. 7C).
In ten studies, 889 patients reported the incidence of postoperative intestinal obstruction[24, 25, 27, 29, 32, 36, 38–40, 42]. Analyzing under the fixed effects model, the overall heterogeneity (I2=0%) is not significant. The overall effect is not significantly different (OR=0.81, 95%CI = 0.38-1.70).
Complication: liver dysfunction (Fig. 8A).
In five studies, 426 patients reported the incidence of postoperative liver dysfunction[23, 24, 27, 29, 39]. Analyzing under the fixed effects model, the overall heterogeneity (I2=18%) is not significant. The overall effect is not significantly different (OR=1.05, 95%CI = 0.49-2.24).
Complication: renal dysfunction (Fig. 8B).
In seven studies, 647 patients reported the incidence of postoperative renal dysfunction[23–25, 29, 36, 39, 40]. Analyzing under the fixed effects model, the overall heterogeneity (I2=0%) is not significant. The occurrence of renal dysfunction after surgery was significantly favorable to the control (OR=2.44, 95%CI = 1.12-5.28).
Complication: pulmonary dysfunction (Fig. 8C).
In five studies, 524 patients reported the incidence of postoperative pulmonary dysfunction[24, 25, 36, 39, 40]. Analyzing under the fixed effects model, the overall heterogeneity (I2=37%) is not significant. The occurrence of pulmonary dysfunction after surgery was significantly favorable to the control (OR=6.03, 95%CI = 3.08-11.80).
Overall complications (Fig. 8D).
In five studies, 384 patients reported the incidence of overall complications[29, 36, 40, 42, 44]. Analyzing under the fixed effects model, the overall heterogeneity (I2=43%) is not significant. The overall effect is not significantly different (OR=0.95, 95%CI = 0.61-1.48).
Deaths due to recurrence after surgery: liver recurrence (Fig. 9A).
In two studies, 181 patients reported the deaths due to liver recurrence after surgery[34, 35]. Analyzing under the fixed effects model, the overall heterogeneity (I2=30%) is not significant. The overall effect is not significantly different (OR=0.81, 95%CI = 0.28-2.31).
Deaths due to recurrence after surgery: lymph node and local recurrence (Fig. 9B).
In three studies, 221 patients reported the deaths due to lymph node and local recurrence after surgery[33–35]. Analyzing under the fixed effects model, the overall heterogeneity (I2=9%) is not significant. The overall effect is not significantly different (OR=1.19, 95%CI = 0.54-2.62).
Deaths due to recurrence after surgery: peritoneum recurrence (Fig. 9C).
In four studies, 395 patients reported the deaths due to peritoneum recurrence after surgery[33–35, 37]. Analyzing under the random effects model, the overall heterogeneity (I2=61%) is not significant. The overall effect is not significantly different (OR=0.37, 95%CI = 0.13-1.06).