Introduction

Advances in laparoscopic surgical techniques have brought about benefits of minimally invasive treatment to many patients. Since Kitano et al. [1] first reported performing laparoscopic distal gastrectomy (LDG) for a patient with gastric cancer, surgical outcomes for LDG comparable to those of open distal gastrectomy (ODG) have been reported by some clinical trials [2,3,4,5,6]. On the basis of such clinical studies, LDG has become established as standard practice [7]. However, one major problem in the field of surgical oncology that has recently been identified is that the results of clinical trials conducted by centers of excellence may not be applicable to the surgical community in general. This is, in part, because high-risk patients and elderly or frail individuals tend to be excluded from clinical trials. Furthermore, in fields in which there are rapid technological advances, such as laparoscopic surgery, there is a concern regarding considerable inter-surgeon and inter-institutional differences in terms of skills in both surgery and perioperative care.

Although the safety and oncological feasibility of various types of laparoscopic surgery have not been thoroughly elucidated in Japan as a whole, the number of laparoscopic gastrectomies has, for example, increased approximately sixfold over the past 10 years and is expected to increase further [8]. Furthermore, as mentioned above, the reported surgical outcomes from specialized hospitals might include a certain selection and publication bias. Considering these situations, a nationwide survey was needed to clarify the safety of LDG in the real world.

Therefore, it is imperative to examine whether LDG, which has started to be widely performed throughout Japan, is actually being performed safely and that it deserves to be regarded as a standard procedure.

The National Surgical Quality Improvement Program (NSQIP) of the American College of Surgeons (ACS) is known to provide practical and real-time clinical data about surgical risk for surgeons [9, 10]. In Japan, the National Clinical Database (NCD) system was introduced in 2011 [11, 12]. This system covers more than 90% of general surgeries performed in Japan, with 1,200,000 patients from 4105 institutions registered each year [11, 12]. Detailed analyses of surgical risks and postoperative complications have been conducted for various surgical procedures using the NCD in the same way as has been conducted by the NSQIP.

In the present nationwide prospective cohort study using the NCD system, we evaluated the surgical outcome of LDG. Simple comparisons of surgical outcomes between LDG and ODG revealed a significant selection bias as a matter of course. Therefore, some important variables that affect surgeons’ decisions with regard to allocating patients to laparoscopic or open gastrectomy were added to the NCD data collection system. In addition, to improve comparability between different surgical procedures, potential confounding factors were adjusted for using propensity score matching (PSM) [13]. The results of the present study will provide surgeons with useful real-world information that is relevant to their clinical practice.

Methods

Study design and cohort development

The present study was a prospective cohort study. To extract a cohort that could be considered representative of the whole of Japan, target institutions were selected by stratified random sampling from institutions registered with the Japanese Society of Endoscopic Surgery and using the NCD system. First, Japan was divided into ten regions (Fig. 1), and all municipalities were classified into three urban levels: metropolitan areas (population ≥ 1 million; black squares in Fig. 1), larger cities (≥100 thousand), and small cities (<100 thousand). Second, using hospital volume (number of surgical cases per year), hospital type (university hospital, specialized hospital, and “others”), and location (10 regions and three urban levels) as adjustment factors, random sampling was performed to select a cohort of institutions that was representative of the country. Requests to participate were sent to 179 institutions, of which 169 ultimately agreed to cooperate. During a 1-year period from August 2014 to July 2015, consecutive patients who underwent distal gastrectomy for gastric cancer were prospectively enrolled in the study. To focus on the surgical outcomes of gastrectomy, patients who underwent concurrent operations other than cholecystectomy were excluded from the analysis.

Fig. 1
figure 1

Geographical regions of Japan used for the stratified random sampling. Hospitals registered with the National Clinical Database (NCD) system were classified into ten regions and three urban levels according to their address. Stratified random sampling was performed to adjust for four hospital types and to identify the cohort to be used in the present study

Full size image

Study team and consensus meeting

To adjust for confounding factors to enable a comparison of surgical outcomes between LDG and ODG using PSM [13], consensus meetings were held by a study team consisting of surgeons and biostatisticians. Preoperative factors that served as a basis for determining allocation to either open or laparoscopic surgery were identified to estimate propensity scores. As a result, covariates for propensity score estimation included patient age at the time of surgery, body mass index, the American Society of Anesthesiologists physical status (ASA-PS), comorbidities, clinical T, N, and M factors (7th edition of the American Joint Committee on Cancer (AJCC) TNM classification), histological type of the biopsy specimen, and presurgical endoscopic resection. Some of these variables were collected explicitly for the present study in addition to the standard list of variables in the NCD gastroenterology registry system.

Propensity score estimation and matching

Logistic regression analysis was used to create a propensity score model for undergoing laparoscopic gastrectomy rather than open surgery using the clinically relevant variables presented in Table 1 (except for clinical and surgical stages, which are collinear with TNM classifications). In addition, several additional risk predictors identified in a previous study [14] were included in the model, including mode of arrival to hospital, smoking status, habitual alcohol consumption, need for assistance with daily activities, the presence of ascites, congestive heart failure, peripheral vascular disease, cerebrovascular accident, preoperative transfusions, hemoglobin <10 g/dl, platelets >350,000/µl, albumin <3 g/dl, alkaline phosphatase >500 IU/l, potassium <138 mEq/l, prothrombin time (international normalized ratio) >1.1, white blood cell count >9000/µl, and the number of procedures (procedure volume, including both LDG and open gastric resection surgery) at the treating hospital. Greedy nearest neighbor matching was performed using a caliper of 0.2 standard deviations of the logit of the estimated propensity score at a ratio of 1:1 without replacement using the macro program provided by Marcelo Coca-Perraillon [15]. The distribution of variables in the LDG and open surgery patients was compared in the pre- and post-matching cohorts using the standardized difference calculated using the macro program by Dongsheng Yang and Jarrod E. Dalton [16].

Table 1 Baseline characteristics before and after propensity score matching
Full size table

Outcomes and statistical analysis

The main outcomes were the in-hospital mortality rate and the incidence of severe postoperative complications. Secondary endpoints included the length of the postoperative stay, incidence of reoperation, readmission, operative time, blood loss, and the incidence of conversion to open surgery as a result of intra-operative accidents. For categorical variables, the proportion of patients experiencing these outcomes was compared between the LDG and ODG groups using Pearson’s chi-squared test or Fisher’s exact test. Continuous variables were compared using the Wilcoxon rank sum test. All statistical tests were two-sided, and P ≤ 0.05 was considered significant. All statistical analyses were performed using SAS version 9.4 (SAS Institute Inc., Cary, NC, USA).

Results

During the study period, 5261 patients underwent distal gastrectomy for gastric cancer in participating hospitals. After PSM, 1067 patients were ultimately selected for each of the LDG and ODG groups.

Patient characteristics

Table 1 lists patient characteristics before and after PSM. At baseline, the ODG group included many high-risk patients with an ASA-PS of ≥3 as well as many elderly patients. In addition, the ODG group had a relatively high number of patients with advanced clinical TNM staging, with more than half the patients having ≥T3 invasion and lymph node metastasis. Conversely, in the LDG group, 70.8% of patients had early stage gastric cancer, and many patients had previously undergone endoscopic resection. Emergency surgery was more common in the ODG than LDG group. After adjustment for these characteristics using PSM, significant differences in background clinicopathologic factors were no longer evident between the two groups. Propensity score distributions are shown in Fig. 2.

Fig. 2
figure 2

Distribution of propensity scores, a before and b after matching, showing the distributions of open surgery (solid line) and laparoscopic surgery (dotted line)

Full size image
Table 2 Surgical outcomes
Full size table

Surgical Outcomes

Surgical outcomes are given in Table 2. There was no significant difference in mortality between the ODG and LDG groups, with three (0.3%) and six (0.6%) postoperative in-hospital deaths, respectively (P = 0.510). With regard to postoperative complications, wound infection and dehiscence were common in the ODG group. The incidence of pancreatic fistula of grade B or above was more common in the LDG than ODG group (2.2 vs. 1.0%, respectively; P = 0.04).

In the LDG group, the mean length of the surgery was approximately 1 h longer than in the ODG group, whereas the estimated blood loss was approximately one-third of that in the ODG group.

Although D2 lymph node dissection was performed more in the ODG group compared with the LDG group (59.2 vs. 51.9%, respectively; P < 0.001), the number of harvested lymph nodes was higher in the LDG group compared with the ODG group (32 vs. 36, respectively; P < 0.001). Of all LDG patients, 36 (3.4%) were switched to ODG because of an intraoperative accident in two patients and after determining that ODG was suitable from an oncological perspective in 25 patients. Reoperation was performed in 20 and 29 patients in the ODG and LDG groups, respectively. No significant difference was observed in the incidence of readmission within 60 days of surgery between the two groups.

Discussion

The primary goal of the present study was to compare the surgical outcomes of LDG with those of ODG on a nationwide scale across Japan. To this end, to ensure the thoroughness of the data, we used the NCD system, which covers 90% of clinical institutions in Japan. To obtain a representative sample from the whole of Japan, participating hospitals were identified using the method of stratified random sampling according to hospital type, region, and urban level. In addition, detailed information specific for gastrectomy and prospectively enrolled patients was added to the NCD system. As a result, the clinical interpretability of propensity score estimation was secured in the present study, enabling comparisons of surgical outcomes between the two procedures after adjustment for potential confounding factors using PSM, which is highly accurate. In this way, the results of the present study are expected to provide relevant and pertinent real-world information in real time.

In terms of the surgery itself, there were no significant differences between ODG and LDG with regard to the reconstruction method, and it is thus inferred that ODG and LDG are comparable. Mortality rates were comparably low, being <1% in both the ODG and LDG groups, indicating that both procedures are feasible. However, operative time was shorter and blood loss was greater for ODG than LDG. Such results have been reported in previous studies [2, 4, 6, 17]. We interpret these findings to reflect that, in LDG, although the surgery itself takes time because manipulation of the forceps is highly restricted, hemostasis is more meticulous because of the magnifying effect of the laparoscopic image, thereby preventing unexpected blood loss. In the present study, the overall incidence of complications was almost the same between the LDG and ODG groups, and the length of postoperative hospital stay was significantly shorter in the LDG group; this appears to demonstrate the efficacy of minimally invasive endoscopic surgery. Wound infection and dehiscence were more common in the ODG group. Laparotomy with a large surgical wound carries a high risk of surgical site infection, which is consistent with the results of the KLASS-01 randomized control trial [18] conducted in Korea. Interestingly, our results show the incidence of grade B or higher pancreatic fistulas was significantly more common in the LDG group, even though more intensive (D2) lymph node dissection was performed in ODG compared with LDG. The following point might be given as one of the reasons.

The greatest difficulty in LDG is manipulation of forceps by the surgeon without tactile sensation [19]. Although utmost caution should be paid to prevent organ injury due to the inappropriate use of forceps in LDG, it is speculated that the reason for the high incidence of pancreatic fistulas in the LDG group could be the assistant applying greater force than is actually required to displace the pancreas to expand the operative field as well as to the surgeon inflicting thermal injury on the pancreas because of the use of energy devices. In particular, assistants should avoid manipulating the pancreas with more force than is necessary in D2 lymph node dissection procedures involving extensive dissection near the pancreas and in obese patients, and these types of surgeries should be performed with extreme care [20]. Leakage of pancreatic juice not only increases the length of hospital stay, but may require reoperation and could even result in mortality [21, 22]. In LDG, increased pancreatic juice leakage is a complication that must not be overlooked and one that requires active management. Preventing both thermal and pressure injuries to the pancreas is essential. Therefore, we recently developed a novel laparoscopic approach that avoids direct compression of the pancreas during suprapancreatic lymphadenectomy. In this procedure, the assistant pulled and controlled the connective tissues along the inferior border of the pancreas to the caudal or dorsal side instead of compressiing the pancreas itself using a gauze/sponge. This retrospective study showed the relationship of pancreas compression with the postoperative Drain-AMY and postoperative complications23. Therefore, avoidance of pancreatic compression is one of the important methods to prevent pancreatic fistulas during LDG.

In terms of the design of the present study, in general pragmatic phase III clinical trials are thought to provide the highest scientific level of evidence in examining causal relationships between intervention and outcome. Conversely, big data analyses, such as using the NCD system, have gained attention as a study method that enables the actual state of medical care to be ascertained and the causal relationship between intervention and outcome to be comprehensively demonstrated. That is, clinical trials are a means to enhance scientific verification (internal validity) that place an emphasis on pathophysiological elucidation, whereas the premise for the analysis of big data is that it provides feedback on actual clinical practice and is a practical means of obtaining data with high external validity. In the present study, to make maximal use of the advantages of both methods, we created a study design with particularly thorough confounding factor adjustment and random sampling. That is, factors serving as the basis for the selection of the surgical procedure were identified at consensus meetings held by specialists and, for any information that was lacking, variables were newly added to the database system and registered prospectively. Furthermore, subject sampling was performed to create an overall representative sample of Japan. Because the results of the present study are clinically and logically valid on the basis of its design, we were able to reach a conclusion with regard to the safety of LDG.

Nevertheless, due attention should be paid to the study’s limitations because the results of the present study are purely the outcome of research using the Japanese NCD; the extrapolation of these findings to other countries, such as the US and in Europe, remains open to discussion. In East Asia, in countries such as Korea and China, where patients have similar tumor locations and characteristics [24,25,26], we can expect similar results; however, in Western countries, there are many factors that may have a disadvantageous effect on surgical outcomes, including a higher incidence of advanced cancer and a greater number of patients with comorbidities and obesity. Furthermore, in North America, the actual prevalence of gastric cancer is low and the learning curves of surgeons may vary.

In conclusion, the present study demonstrates the safety of LDG. Although this procedure is generally being performed safely, pancreatic fistulas tend to occur more often in patients undergoing laparoscopic than open surgery; therefore, due care should be taken when performing dissections near the pancreas.