Gastric cancer (GC) is one of the most common malignant tumors and ranks fourth in cancer incidence and second in cancer mortality around the world [1, 2]. For patients with primary advanced gastric cancer (AGC), surgical resection with D2 lymphadenectomy remains the current standard [3]. In recent decades, minimally invasive surgery (MIS), including robotic and laparoscopic surgery, has been rapidly developed and improved, providing a new alternative surgical method for the treatment of AGC and gradually becoming the mainstream surgical treatment for GC. To date, many studies have evaluated the safety and efficacy of robot-assisted gastrectomy (RAG) and laparoscopic-assisted gastrectomy (LAG) in the treatment of AGC and have demonstrated that RAG has many advantages over LAG [4,5,6,7]. However, only a few reports have assessed and compared the incidence of complications following RAG with that following LAG using their own subjective criteria, which can easily lead to bias and misjudgment, and it is difficult to compare the incidence of complications between these studies.

In 1992, Clavien et al. [8] established a novel approach for grading the severity and incidence of postoperative complications based on the management of the complications. Dindo et al. [9] revised and improved this grading system in 2004, introducing a new five-scale classification with the aim of presenting an objective, simple, reliable, reproducible, flexible, and applicable way of evaluating postoperative complications irrespective of cultural background. This grading system was more widely applicable after being revised and improved, and it was eventually named the Clavien–Dindo (C–D) classification [8,9,10]. The C–D classification was subsequently validated and was widely used to systematically evaluate the severity and incidence of complications in many surgical disciplines [11,12,13,14]. Although several studies [15,16,17,18,19] evaluated complications following LAG using the C–D classification, complications assessed by the C–D classification in RAG have rarely been well described. Therefore, it is necessary to systematically evaluate the applicability of the C–D classification in RAG.

The aim of the present study was to grade and compare the severity and incidence of complications following RAG with those following LAG using the C–D classification in a propensity score matching (PSM) cohort and to identify the risk factors associated with complications after MIS for AGC.

Materials and methods

Patients and data collection

The study was approved by the institutional review board of Lanzhou General Hospital of Chinese People’s Liberation Army, and written informed consent was obtained from all the patients in the study.

We retrospectively collected data from a prospective database of Western China Gastric Cancer Collaboration (WCGCC) group participated by the Department of General Surgery, Lanzhou General Hospital of Chinese People’s Liberation Army. Inclusion criteria for the study were: (i) All patients were diagnosed with GC by biopsy; (ii) All patients with pT2-4aN0-3M0 GC were eligible; (iii) All patients were conducted gastrectomy with D2 lymphadenectomy; (iv) The physical status of all patients was evaluated as American Society of Anesthesiology (ASA) Classes I, II, or III; (v) All patients did not undergo radiotherapy or chemotherapy before surgery. (vi) All patients did not undergo splenectomy during surgery. However, patients with benign gastric tumors, gastrointestinal stromal tumors, or combined with other organ malignancies, and residual GC were excluded from this study (Fig. 1). In this study, the pathological stages were classified according to the eighth edition of the American Joint Committee on Cancer TNM Staging System for GC [20, 21]. The division of tumor sites was based on the guidelines of Japanese classification for GC [22]. The work of this study has been reported in accordance with the Strengthening the Reporting of Cohort Studies in Surgery (STROCSS) statement [23].

Fig. 1
figure 1

CONSORT diagram for case enrollment in the present study. RAG robot-assisted gastrectomy, LAG laparoscopy-assisted gastrectomy, ASA American Society of Anesthesiologist

Full size image

Operative procedures

The extent of gastrectomy and lymph node (LN) dissection was carried out based on the Japanese gastric cancer treatment guidelines [3]. The extent of resection was decided according to the tumor location and extent. When the marginal state is suspicious, we performed an intraoperative frozen section to confirm the negative proximal resection margin to ensure R0 resection. The type of reconstruction was determined by the surgeon’s experience. Roux-en-Y oesophago-jejunal anastomosis was performed to reconstruct the digestive tract for total gastrectomy, and Billroth II or Roux-en-Y gastro-jejunal anastomosis was applied to distal gastrectomy. The Da Vinci Surgical System (Intuitive Surgical Inc., Sunnyvale, CA) was used for RAG. The surgical procedures for RAG and LAG have been described as before in detail elsewhere [24, 25].

Perioperative management

All patients underwent preoperative evaluation through chest X-rays, electrocardiogram, lung function tests and laboratory tests, etc. Prophylactic use of antibiotics was routinely performed 1 h before surgery and every 3 h during surgery. Blood biochemical tests were routinely performed every 2–3 days after surgery. Patients were given sipping water after the first gastrointestinal ventilation after surgery, and were given liquid diet on the 3rd or 4th day after surgery. Once the patient tolerated a liquid diet for 2 days, a soft diet was given next. When the patient’s soft diet is satisfactory and there are no complications, the patient is advised to be discharged. Patients without chemotherapy contraindications were given first adjuvant chemotherapy at the 4th week or 5th week after surgery.

Definition and assessment of complications

The C–D classification was adopted to grade the severity of the postoperative complications for each patient [9, 10, 26]. According to the C–D classification system, the severity of complications was assessed from Grade I to V. The complications assessed in our study mainly included wound problem, bleeding, anastomotic leakage or stenosis, remnant gastric infarction, postoperative ileus, and other complications within 2 months of the initial operation. These complications were divided into overall complications (Grade I–V), serious complications (Grade ≥ IIIa) and mortality (Grade V). If the patient had multiple complications, the most serious complication was used for grading. Three independent attending surgeons assessed C–D grade for each patient, and any divergences were solved by discussion. Table 1 shows the details of the complications recorded in our center according to the C–D classification.

Table 1 Classification of the common complications following gastrectomy for AGC according to C–D classification system
Full size table

Propensity score matching analysis

PSM analysis was used to limit confounders and overcome possible patient selection bias. Propensity scores for all patients were calculated using a logistic regression model based on the following variables: age, gender, BMI, ASA, tumor size, and stage of primary tumor, etc. Variables with p values less than 0.05 between groups were included in the PSM analysis. Calculating the propensity scores also require that any potential clinical confounders be included as covariates regardless of their statistical range [27]. A 1:1 nearest neighbour matching algorithm with an optimal calliper width of 0.2 without replacement was used to match the propensity scores [28,29,30]. Standardized mean differences (SMD) were assessed to determine whether a sufficient balance was achieved after matching (SMD < 0.1) [31]. This procedure was conducted using the SPSS-R plugin developed by felixthoemmes, wliao229 (https://sourceforge.net/projects/psmspss/files/).

Statistical analysis

Statistical Package for Social Science version 22.0 (SPSS, Chicago, IL, USA) was used for the calculation of the correlation coefficients. Continuous variables were assessed for normality of distribution using the one-sample Kolmogorov–Smirnov test. Continuous variables with a normal distribution are reported as mean (standard deviations, SD) and compared using Student’s t test. Continuous variables that were not normally distributed are expressed as median (inter quartile range, IQR) and analyzed using Mann–Whitney U test. Categorical variables are expressed as absolute numbers and percentages and compared using Pearson’s χ2 test or Fisher’s exact test as appropriate. Multivariable binary logistic regression analysis was performed to identify independent risk factors for complications and odds ratios (ORs) with 95% confidence intervals (CIs) were calculated. A two-sided P value < 0.05 was considered to be statistically significant.

Results

Demographics and surgical outcomes

Among 527 patients with AGC, 251 (47.6%) received RAG and 276 (52.4%) LAG (Fig. 1). The mean age was 57.7 years in the RAG group and 56.8 years in the LAG group (P = 0.374). Patient characteristics and surgical outcomes of the two groups are provided in Table 2. RAG patients presented with a smaller tumor size compared to LAG patients (4.5 ± 2.3 vs. 5.1 ± 2.1, P = 0.004). In the RAG group, there were significantly less patients with previous abdominal operation compared to the LAG group (8.8% vs. 15.9%, P = 0.013). No significant differences between the two groups were noted in gender, body mass index (BMI), ASA, tumor location, extent of resection, histologic type, or pathological stage (Table 2).

Table 2 Demographics and surgical outcomes of the entire cohort and PSM cohort
Full size table

The mean number of retrieved LNs for the RAG group was significantly larger than that for the LAG group (41.2 ± 15.1 vs. 36.3 ± 14.8, P < 0.001). The time to ambulation (2.1 ± 0.6 vs. 2.2 ± 0.4 days, P = 0.024) and time to first flatus (2.6 ± 0.7 vs. 2.9 ± 0.6 days, P < 0.001) were significantly shorter in the RAG group. The postoperative hospital stay for patients who underwent RAG was significantly shorter than who underwent LAG (10.9 ± 2.8 vs. 12.3 ± 3.2 days, P < 0.001). The difference between two groups in the operating time, estimated blood loss, and time to first liquid and soft intake was not significant (all P > 0.05, Table 2).

Propensity score matching analysis

After PSM analysis, 446 patients (223 patients in each group) were enrolled into further analysis. Standardized difference analysis demonstrated a successful match between all baseline characteristics (SMD = 0.006, Fig. 2). The baseline characteristics of the matched cohort were similar between the two groups (Table 2). Regarding surgical outcomes in the matched cohort, the time to ambulation, time to first flatus, and postoperative hospital stay were still significantly shorter in the RAG group than those in the LAG group (all P < 0.05; Table 2).

Fig. 2
figure 2

Density distributions of standardized differences before and after matching in the present study

Full size image

Postoperative morbidity and mortality

Morbidities and mortalities for all patients in the matched cohort are summarized in Table 3. After PSM, we analyzed overall and severe complications classified by C–D classification for the patients. In all, 42 patients (24.5%) incurred complications after RAG, compared with 78 patients (34.9%) presented complications after LAG (OR 0.43, 95% CI 0.28–0.67, P < 0.001). There were fewer major complications experienced following RAG than the following LAG (8.9% vs. 17.5%; OR 0.39, 95% CI 0.21–0.71, P = 0.002).

Table 3 Comparison of complications following robotic and laparoscopic gastrectomy for AGC using the Clavien–Dindo classification in the PSM cohort
Full size table

RAG patients had fewer grade I complications compared to LAG and there was no statistically significant difference (P = 0.848, Fig. 3). The most frequent grade I complications in both groups were fever followed by wound problem, delayed recovery of bowel movement, and atelectasis. We found a significant difference between the two groups in the rate of grade II complications, such as wound infection, fluid collection, pneumonia, postoperative ileus, and anastomosis leakage (4.9% vs. 10.8%, P = 0.022, Fig. 3). The incidence of grade IIIa complications was significantly lower in the RAG group (3.1% vs. 8.1%, P = 0.024, Fig. 3). Among grade IIIa complications, anastomosis stenosis and leakage, pleural effusion, leakage of lymphatics, and fluid collection were relatively common. Grade IIIa complications occurred in five patients (anastomosis leakage in two patients, intra-abdominal bleeding in one, small-bowel perforation in one, and postoperative ileus in one) in the RAG group and 14 patients (anastomosis leakage in five patients, intra-abdominal bleeding in three, postoperative ileus in three, and other complications in three) in the LAG group (2.2% vs. 6.3%, P = 0.035, Fig. 3). Moreover, no significant differences were noted in grade IVa and grade IVb complications rates between the two groups (grade IVa: 0.9% vs. 1.8%, P = 0.407; grade IVb: 0.4% vs. 0.9%, P = 0.559, Fig. 3). There were three deaths (2 in RAG and one in LAG) due to anastomosis leakage during hospitalization. Five patients experienced more than two complications before and after, and the most severe complication was used as a reference for final grading in the present study.

Fig. 3
figure 3

Overall distributions of complications following robotic and laparoscopic gastrectomy for AGC using the Clavien–Dindo classification in PSM cohort. RAG robot-assisted gastrectomy, LAG laparoscopy-assisted gastrectomy, AGC advanced gastric cancer, PSM propensity score matched

Full size image

Subgroup analysis

Subgroup analyses were further performed for overall and severe complications in the matched cohort, respectively. In this study, we converted the clinicopathological parameters into dichotomous variables and selected the median of these parameters as the threshold.

Within the overall complications alone, there were significant differences between the two groups in all stratified parameters except age < 65 years (P = 0.064), female (P = 0.082), and previous abdominal operation (P = 0.132), as shown in Fig. 4. Regarding the severe complications alone, subgroup analyses revealed that the statistical differences between the RAG and LAG groups were present in these stratifications, such as BMI, tumor size, histologic type, operating time, and estimated blood loss (all P < 0.05, Fig. 5), whereas no differences were observed in these stratified parameters including age ≥ 65 years (P = 0.147), female (P = 0.272), ASA class III (P = 0.121), upper third tumor (P = 0.231), total gastrectomy (P = 0.062), stage I (P = 0.183), and retrieved LNs < 30 (P = 0.129) (Fig. 5).

Fig. 4
figure 4

Subgroup analyses of overall complications following robotic and laparoscopic gastrectomy for AGC using the Clavien–Dindo classification in PSM cohort. RAG robot-assisted gastrectomy, LAG laparoscopy-assisted gastrectomy, OR odds ratio, BMI body mass index, PAS previous abdominal operation, LNs lymph nodes, AGC advanced gastric cancer, PSM propensity score matched

Full size image
Fig. 5
figure 5

Subgroup analyses of severe complications following robotic and laparoscopic gastrectomy for AGC using the Clavien–Dindo classification in PSM cohort. RAG robot-assisted gastrectomy, LAG laparoscopy-assisted gastrectomy, OR odds ratio, BMI body mass index, PAS previous abdominal operation, LNs lymph nodes, AGC advanced gastric cancer, PSM propensity score matched

Full size image

Risk factors for overall complications

Logistic regression was performed to determine factors associated with overall complications following RAG and LAG for AGC in the matched cohort. Univariate analysis revealed that age ≥ 65 years, BMI ≥ 30 kg/m2, total gastrectomy, stage T3–T4a, stage N1–N3b, stage II–III, laparoscopic operation, and operation time ≥ 250 min were significantly associated with higher risk of overall complications (all P < 0.05, Table 4). In multivariate analysis, age ≥ 65 years, total gastrectomy, stage T3–T4a, stage II–III, and operation time ≥ 250 min were significant independent risk factors for overall complications (all P < 0.05, Table 4).

Table 4 Risk factors for overall and severe complications following robotic and laparoscopic gastrectomy for AGC in the PSM cohort
Full size table

Risk factors for severe complications

Regarding severe complications, higher complication rates correlated with age ≥ 65 years, BMI ≥ 30 kg/m2, ASA class III, previous abdominal operation, total gastrectomy, stage T3–T4a, stage N1–N3b, stage II–III, laparoscopic operation, operation time ≥ 250 min and retrieved LNs ≥ 30 in univariate analysis (all P < 0.05, Table 4). Furthermore, the multivariate analysis showed that age ≥ 65 years, stage II–III, and operation time ≥ 250 min were independent risk factors for severe complications following RAG and LAG (all P < 0.05, Table 4).

Discussion

Accumulating evidence has demonstrated that RAG has been increasingly used in the treatment of AGC, and many advantages over LAG have been reported [5, 6, 24]. However, there is still no consensus on the definition and classification criteria of complications following RAG and LAG, which greatly hampers the assessment and comparison of different surgical procedures. Therefore, we have adopted the C–D classification system, which has proven to be a standard for surgical safety assessment in many fields [11,12,13,14, 32,33,34]. This study aimed to compare the severity and incidence of complications following RAG vs. those following LAG using C–D classification and to identify risk factors related to complications.

To date, complication reports assessed complications following RAG and LAG using the C–D classification are quite limited and the results are still controversial [4, 35,36,37]. Obama et al. [4] and Kim et al. [35] reported no differences in the incidence of overall and major complications between RAG and LAG. However, a retrospective cohort study from a Japanese institution reported a higher incidence of overall complications (C–D grade ≥ III), local complications, and pancreatic fistula in the LAG group than in the RAG group [36]. Seo et al. [37] also reported that the frequency of pancreatic fistula was higher in the LAG group. Using C–D classification to grade the severity of complications, we observed significant reduction in the overall and major complication rates of RAG compared with LAG, respectively. However, the incidence of complications appears to be slightly higher than those reported in previous studies, which may be due to the fact that our study included all complications from grade I to V, while most of the previous studies did not assess grade I complications.

Besides, we evaluated specific subgroups of the matched cohort to further test the independent relationship between the clinicopathological parameters and overall and severe complications, respectively. In the present study, we focused on the advantages of RAG in terms of the complications. Interestingly, subsequent subgroup analyses of overall complications supported RAG except age < 65 years, female, and previous abdominal operation. We also performed subgroup analyses testing the effect of the stratified parameters on severe complications and again obtained similar results as in the overall complications, whereas no differences were observed in these stratified parameters including age ≥ 65 years, female, ASA class III, upper third tumor, total gastrectomy, stage I, and retrieved LNs < 30. This difference may be caused by a limited number of cases in the above subgroup, so the above results also require greater data validation.

Understanding the relevant risk factors is important to minimize the incidence of complications. We then performed univariate and multivariate analyses in the mathed cohort to identify risk factors associated with overall and severe complications following RAG and LAG. The results showed that age, extent of resection, pTNM stage, and operation time were independent risk factors for overall complications, while age, pTNM stage, and operation time were associated with severe complications.

Age was identified as an independent risk factor for overall and severe complications following RAG and LAG in our study, and this result was consistent with many previous reports [38, 39]. This may be due to the poor tolerance of elderly patients to surgical stress leading to an increased incidence of complications. However, there are still studies reporting no correlation between patient age and complications [40, 41]. Therefore, more prospective randomized clinical trials are needed to confirm the effect of age on surgical outcomes.

Extent of resection was also identified as an independent risk factor related to overall complications. The incidence of the overall complications following total gastrectomy was significantly higher than distal gastrectomy, and the major complications following total gastrectomy was an anastomosis complication. These results were consistent with Lee et al. [39] and Ji et al. [42]. This could be explained by the difference in blood supply around the anastomosis. The other possible cause is that the number of retrieved LNs following total gastrectomy is greater, which may compromise the vascular supply around the duodenal stump. Therefore, surgeons should always be more careful when removing LNs around the blood vessels, especially in total gastrectomy.

As expected, our study showed that pTNM stage to be an independent risk factor for overall and severe complications. Zhou et al. reported [38] that patients with higher stage often have anemia, weight loss and hypoproteinemia, which may affect the vulnerability of surgical stress and the occurrence of complications. Some studies [43,44,45] also suggested that surgical resection for higher-stage AGC patients requires more complicated surgical procedures and longer operative time, resulting in excessive surgical stress and a high incidence of complications.

The current study also demonstrated a close link between the operation time and a higher incidence of overall and severe complications. This result is inconsistent with that reported by Watanabe et al. [46] and Lin et al. [47]. This may be because the surgical procedure with a long operation time is more technically complicated. For example, total gastrectomy needs to resect more areas of LNs, and the oesophago-jejunal anastomosis is more complicated and requires more time than the gastro-jejunal anastomosis.

Nevertheless, our study has certain limitations to be acknowledged. First, this study is limited by its retrospective nature, which places the study at risk for possible selection and information bias and limits the evidence level of our conclusions. Although PSM was used to mitigate these biases, some residual confounders may persist. Second, this is a single-center study regarding Chinese populations and this demographic condition may limit the applicability of our results to other populations. Third, our study does not assess long-term complications owing to the relatively short follow-up period. Fourth, postoperative complications in this research did not include pancreatic fistula, which may affect the assessment of the overall and severe complications. Next, we will continue to follow-up these patients to further investigate these issues. Finally, some factors such as patients’ preoperative hematologic parameters and surgeon’s proficiency which may affect postoperative morbidity and mortality are not fully incorporated into our study.

Conclusions

In conclusion, the present study demonstrated RAG with D2 lymphadenectomy is feasible and safe for the treatment of AGC in terms of the lower incidence and severity of complications. In our opinion, the C–D classification facilitates to objectively assess the incidence and severity of specific complications and is an important vehicle for comprehensive comparison and evaluation of the safety of different types of gastrectomy. However, more large-scaled, multicenter, and prospective randomized control studies using the C–D classification are still warranted to fully evaluate the complications following robotic and laparoscopic gastrectomy for AGC.