Introduction

Although its incidence is stabilizing, gastric cancer still represents a major cause of cancer deaths worldwide [1,2,3,4,5]. The foundation of gastric cancer care is surgery, which is currently performed in hospitals with different procedure volumes. Centralization of gastric cancer surgery to a small number of high-volume centers occurs in only a few countries and has not been widely accepted.

The overall estimated mortality rate of gastrectomy for cancer is surprisingly higher than the reported mortality for liver and pancreatic cancer surgeries. The mortality rates in high-volume Western centers range from 3 to 5%, which are in contrast with the rates for low-volume centers that range from 10 to 20% [6,7,8]. Postoperative morbidity rates are reported as major or minor, but these reports vary and use inconsistent descriptive terminology. As a result, studies have reported a wide range of incidence rates varying from 11 to 46% [9,10,11,12,13,14].

Understanding postoperative complications is of paramount importance, because these adverse events can have major impacts on critical outcomes. However, the wide variation in the definitions and recording of postoperative complications after gastrectomy across different institutions and countries makes establishing causal links between these complications and critical outcome measures almost impossible. A systematic review published in 2001 retrieved a total of 41 different definitions of and 13 grading scales for surgical wound infections from 82 studies. The review also found 56 separate definitions of anastomotic leaks from 107 studies reporting the outcomes of gastrointestinal surgery [15]. A recent study investigating the incidence rates of postoperative hemorrhage, respiratory failure, deep vein thrombosis, and sepsis derived from three data sources (administrative, a national clinical registry, and an institutional clinical registry) reported remarkable discordance, with Cohen’s kappa coefficients ranging from 0.02 to 0.60 [16].

These data highlight an unmet need to establish a common language to ensure consistency in the definitions of postoperative gastrectomy complications not only to create standards of care but also to facilitate outcome comparisons. Recent Asian studies have shown the benefits of standardized reporting for conducting detailed comparisons of the differences between Western and Eastern surgeons [17, 18].

The most common and universally recognized classification of postoperative complications is the Clavien–Dindo (C–D) classification, which is a treatment-related severity grading system [19,20,21]. Since a patient may develop more than one postoperative complication, C–D grading has been expanded into the Comprehensive Complications Index (CCI). The CCI is a web-based calculator that combines multiple complications and produces a final score ranging from 0 to 100 [22, 23]. Both the C–D classification and the CCI are general scores and do not provide definitions of specific postoperative complications for particular operations.

This study reports the first step of a project launched in November 2015 by a large group of gastric cancer experts who are members of the International Gastric Cancer Association (IGCA). The project aimed to define a comprehensive list of surgery-related and gastric cancer-specific complications and adverse events deemed to be essential items that should be included in multicenter studies and international databases. The list will help standardize outcome reporting after gastric resection for cancer worldwide. Recent pioneering studies by Low et al. [24, 25] have shown that standardization of data collection for complications associated with esophagectomy is the building block for achieving the ultimate goal of proposing quality-improvement projects that can benefit patient survival, quality of life, and long-term outcomes. In other words, quality improvement must start with quality measurements [26,27,28].

The road map for this study envisaged two steps. In the first step (presented here), European gastric cancer experts created and reached a consensus on a list of defined complications using Delphi surveys. This method was used by Low and coauthors in their study of esophagectomy complications [24]. The second step involves validation of the proposed list and the establishment of a large global database. The multicenter study will be fully international; many gastric cancer experts and members of the IGCA from high-volume centers in Brazil, Canada, China, Japan, Korea, and the US have all agreed to participate.

Methods

Design and assembly of the Gastrectomy Complications Consensus Group (GCCG)

The research project was developed within the portfolio of studies sponsored by European members of the IGCA. A project working group consisting of 3 gastric cancer experts was established. The project working group identified high-volume specialist gastric surgical centers in Europe from the participants of the EURECCA (EUropean REgistry of Cancer CAre) project and the members of the Italian Research Group on Gastric Cancer (GIRCG) to form the Gastrectomy Complications Consensus Group (GCCG). The GCCG was designed to include a widely heterogeneous set of gastric cancer specialists who could represent different individual institutions within a country and different health systems across Europe. Thirty-one experts from 13 European countries agreed to take part in addition to the 3 members of the project working group (Table 1).

Table 1 Members of the Gastrectomy Complications Consensus Group (GCCG)
Full size table

Delphi surveys

The project used Delphi surveys designed to be mainly web-based and coordinated by the project working group [29]. The working group thoroughly reviewed the most recent knowledge on how perioperative complications related to gastrectomy for cancer were defined and recorded in the literature. Next, the group invited the study participants to agree on the proposed methodology that would deliver an initial list of complications with their definitions. The group also proposed that (1) the project should focus on trans-abdominal gastrectomy with curative intent for gastric cancer; (2) Siewert type II and III esophago-gastric junctional cancers should be included, and (3) open and minimally invasive (laparoscopic or robotic) operative procedures should both be included in the scope of the project.

A series of questions were designed and circulated electronically to the study participants. A total of 10 groups of questions addressed the initial study design, intraoperative complications, general and surgical postoperative complications, outcome measures, and severity scores (Table 2). The project working group analyzed 660 answers. Each study participant provided a mean of 21 (range 11–36) answers.

Table 2 Results of the Delphi surveys
Full size table

For each group of questions, the study participants were asked to give their opinions on each topic. They were requested to base their views on their clinical and surgical experiences rather than referencing the literature. For select cases that were deemed particularly relevant, each expert was invited to provide remarks on a specific topic.

The responses for each group of questions were collated. Blinded discussion rounds using a modified Delphi approach were undertaken when differences of opinion were evident. When at least 80% agreement was achieved, the question group was completed, and the next question group was circulated until all of the question groups had been considered. When answers differed, the project working group summarized the various opinions and then rephrased the question in a dichotomic form.

Confirmation rounds were conducted to approve the final complications list (displayed in Table 3). A consensus was also reached for the definition of each complication (Table 4). The definitions were kept precise but simple and focused on the critical features of each complication determined by the study participants (i.e., the type and level of the clinical response to the complication itself rather than its severity grade). A preliminary version of the list was presented and thoroughly discussed in Lisbon, Portugal, on June 16, 2016, at the “Esophageal and Gastric Cancer Initiative” conference, which was attended by several GCCG members. The final version of the list was presented at two invited sessions at the 12th International Gastric Cancer Congress in Beijing on April 21, 2017, which were attended by most of the GCCG members.

Table 3 Gastrectomy for cancer: the complications list
Full size table
Table 4 The gastrectomy complications list: definitions
Full size table

Results

Following initial agreement on the project design, a list of 27 complication topics was identified, including 3 intraoperative, 14 postoperative general, and 10 postoperative surgical complications. The main findings of the Delphi surveys (Table 2) were as follows.

In Question group 1, most experts agreed that general complications, including medical diseases, should be included. The optimal timing to document the postoperative assessment was initially contentious. After lengthy discussions and several blinded re-submissions, study participants agreed that this assessment should be made electronically via an app at both discharge and 90 days postoperatively.

In Question groups 2, 3, and 4, the term “general” was preferred to “medical” for non-surgical complications, which should include all major neurologic, cardiovascular, respiratory, liver, kidney, and infectious events (Table 3). A precise definition was offered for each of these complications (Table 4), and a consensus was reached.

In Question group 5, the study participants agreed that intraoperative damage to vessels and organs and intraoperative major bleeding should be included. These adverse events may also influence medium- to long-term outcomes. The details of each intraoperative complication were also defined (Table 4), and a consensus was reached. Unexpected adverse events (e.g., cardiac arrest) were also included within this category.

In Question groups 6 and 7, the study participants were asked to comment and agree on 12 postoperative surgical complications that were previously defined by the Italian Research Group on Gastric Cancer (GIRCG) [30]. Ten complications were agreed upon, and a consensus was reached about the definition of each complication (Tables 3, 4).

In Question groups 8 and 9, the study participants were asked to comment on a list of additional quality measures to be added to the end of the Complication Recording Sheet, including in-hospital mortality (yes/no), length of ICU stay (days), length of postoperative hospitalization (days), discharge (home/other facilities), and the Clavien–Dindo complication grade(s) (considering the highest level). They were also asked to provide remarks on which grading system should be adopted (the Clavien–Dindo scale or the Comprehensive Complication Index). Given the wide disagreement on these questions, a consensus was drawn that both topics should be studied more in depth in the next step of the project during the international multicenter validation of the complications list.

In Question group 10, the project working group invited the study participants to reach a final consensus on the complications list (Table 3) and the precise definition of each complication (Table 4). A unanimous consensus was reached.

To summarize, despite disagreement on some questions, the Delphi surveys delivered a strong consensus on the most critical issues (Table 2), especially the list of complications and their specific definitions (Tables 3, 4).

Discussion

A consensus study is not without limitations. The opinions of those involved are based on varying levels of evidence and on clinical and surgical experience. Although this situation can introduce bias, the breadth of practices and health systems covered by the consensus group should minimize any bias. Needless to say, neither the GCCG nor any other single group can determine the ideal international standards for defining and recording gastrectomy complications. However, this project represents a starting point for generating a wider international consensus for standardization of data collection for cancer-related gastric resections. To facilitate the adoption of the proposed list across institutions and countries, Table 5 briefly summarizes the discussions that occurred among the GCCG members and reports the main rationales that guided them in the choice of a given definition for each complication.

Table 5 Executive summary of consensus group’s discussions
Full size table

Key features of the complications list

First, in contrast to previous reports, this study has included intraoperative complications, which are often unreported because their effects are corrected during procedures. However, these interventions, such as unplanned splenectomies or pancreatic resections, portal vein/hepatic artery reconstructions, and bile duct repairs, may also have medium- to long-term clinical consequences. The consensus group also agreed that unplanned blood transfusions should be documented as a marker of an unexpected intraoperative event, because they can adversely affect the oncological outcome [32]. One novel feature of this project is the inclusion of unexpected medical conditions other than intraoperative damage and bleeding in the intraoperative group. Despite being rare, these life-threatening events (e.g., anaphylactic shock, cardiac arrest, myocardial infarction) can occur and have dire consequences if they are not dealt with swiftly and ably.

Second, only major gastrectomy-related complications were included in the list. The difference between major and minor adverse events can be quite confusing and arbitrary, which may be one of the most compelling reasons for the different rates of complications reported in the published literature. Whether a given complication should be defined as “major” or “minor” has not been definitively stated. On the one hand, the grading severity cannot ignore the Clavien–Dindo classification system, which refers to the treatment needed for each adverse event. Hence, for several items in the proposed list, a grade equal to or greater than 3 from the C–D classification system is also considered major. On the other hand, this basic principle does not cover all of the items on the list. For example, the consensus group considered major complications to be those that required the patients to be transferred to a higher level of care (e.g., myocardial infarction should be considered a major complication only when the patient is transferred to the ICU/CCU irrespective of the treatment). In addition, complications that may not necessarily require significant intervention in the C–D system have been classified as major in this study. For example, the consensus group agreed that duodenal and anastomotic leaks after gastrectomy were sufficiently significant and thus should be considered major events irrespective of how they manifested, the method of identification, their clinical consequences, and the required treatment. Other major postoperative complications, such as bowel obstruction, pancreatic fistula, severe pancreatitis, abdominal collections without leakage and delayed gastric emptying, were also included in the proposed list irrespective of the need for treatment.

Third, several common yet minor complications, such as deep venous thrombosis, prolonged urethral catheterization and wound opening at the bedside, have been excluded.

Fourth, the list of postoperative surgical complications also included events such as acute evisceration from wound dehiscence, feeding jejunostomy-related complications, and diaphragmatic hernia. Since these events require re-intervention, they have been considered sufficiently important for inclusion in the list.

Fifth, a consensus among the study participants was also reached on precise yet simple definitions for the proposed complications, which should help improve uniform recording across institutions and countries (Tables 4, 5).

Comparison of the gastrectomy complications list with other complication lists

In the studies of complications after esophagectomy by Low et al. [24, 25], 49 items subdivided into 9 groups were determined using the Delphi process with specialist esophageal surgeons. These groups included both general and specific postsurgical adverse events. Esophagectomy-specific complications were fully defined, because variations in descriptions were present across the different participating centers. Although severity was not graded, the study recommended that complications should be recorded in conjunction with their Clavien–Dindo classifications, because this approach would allow recording of multiple complications in one patient.

Several conditions (atelectasis secondary to mucus plugging, Clostridium difficile infection, urinary retention, peripheral thrombophlebitis, and acute delirium) included in that study were not included as specific gastrectomy complications in this study. These complications can and do occur in patients undergoing gastric resection, but they are not specific to gastrectomy. Additionally, acute aspiration and atrial fibrillation are not included in this study, because they are more commonly related to the thoracic phase of esophagectomy.

A broad study was conducted by the Japanese Clinical Oncologic Group (JCOG) that assessed complications across 9 cancers to ensure consistency of reporting of adverse events in clinical trials [33]. The Clavien–Dindo general rules were used as the guiding principles for grading the severity of each complication. A total of 72 complications were described. A number of complications that were not formally defined could be associated with gastrectomy. The JCOG group included the severity of the complications. However, clear descriptions of the complications were not fully included, which could result in variability in interpretations. Furthermore, events that might be considered fairly minor in terms of severity could actually be specific events after gastrectomy, such as delayed gastric emptying. Finally, although comprehensive, completing the JCOG assessment is not straightforward, since it covers both gastrectomy-related and other abdominal procedures. The assessment also includes six possible grades for 72 adverse events for a total of more than 400 possible combinations.

Optimal timing of complication reporting and implementation of data collection

The wide variation in reporting complications in the literature also pertains to the optimal timing of this reporting. Indeed, as shown in round 1 of the Delphi survey (Table 2), the study participants had strong disagreements on this topic. One-quarter of the experts preferred to report complications at the patient discharge, whereas more than one-third recommended that complications be reported 30 days postoperatively. However, most of the participants felt that the recording of complications and surgical case assessments should be performed at multiple time points to produce a better overall evaluation. After lengthy discussion in round 2, the study participants agreed that the reporting of complications should occur at both patient discharge and 90 days postoperatively and that reporting should be performed electronically through an app. The consensus was that more time should elapse before a more complete outcome assessment can be obtained.

Additionally, although no formal agreement was reached during the Delphi surveys, informal discussions identified the surgeon leading the team performing the surgical procedure as the ideal candidate to complete the Complications Recording Sheet for each patient episode. The GCCG members intend to conduct a more rigorous and thorough discussion to reach a decision on this important point before starting the multicenter validation of the complications list.

Outcome measures and complication severity grading

Although the consensus group discussed which additional quality and outcome measures should be included in the surgical case assessment, they concluded that this topic required more discussion and analysis. Similarly, there was wide disagreement on which grading system to adopt [Clavien–Dindo, Comprehensive Complications Index or other(s)] for recording complications after gastrectomy for cancer. Hence, these two issues, which are clearly important and relevant, will be discussed and studied in depth in the next step of the study.

In conclusion, international comparisons of complications after gastrectomy are unreliable due to differences in the definitions proposed in the literature. These inconsistencies significantly hamper the determination of standards of care and the proposals of quality-improvement initiatives. Therefore, establishment of a common language should be considered a priority. Furthermore, there are indicators (e.g., during esophagectomy for cancer) that perioperative complications have a greater effect on postoperative quality of life than the surgical technique [34].

A consensus approach has identified and rigorously defined a series of 27 complications specific to gastric resection for cancer. At the 12th International Gastric Cancer Congress in Beijing (April 2017) the study received the endorsement of the IGCA executive committee to expand the project from a Europe-based study to an international initiative: surgeons from high-volume centers worldwide will participate in multicenter validation of the comprehensive list of defined complications and collaborate on the development of a large, international database of gastric cancer complications.

In the next step of the project, which is already underway, a Complications Recording Sheet based on a user-friendly electronic application will be developed (a preliminary version of the complications recording sheet is included in Table 6). The incidence of these complications across specialized centers worldwide will be assessed. The study plans to develop a risk model specific to gastrectomy that can be recorded simply using the list of complications on the electronic application. This approach will allow the determination of a benchmark for complications that is both standardized and comparable across institutions and countries. An international assessment of the impact of complications on patient survival, quality of life, and long-term outcomes, as well as that on cost-effectiveness of a novel technology will then be feasible.

Table 6 Complications Recording Sheet (CRS)—preliminary template
Full size table