Search

Background and Objectives: With the current advanced data-driven approach to health care, machine learning is gaining more interest. The current study investigates the added value of machine learning to linear regression in predicting anastomotic leakage and pulmonary complications after upper gastrointestinal cancer surgery. Methods: All patients in the Dutch Upper Gastrointestinal Cancer Audit undergoing curatively intended esophageal or gastric cancer surgeries from 2011 to 2017 were included. Anastomotic leakage was defined as any clinically or radiologically proven anastomotic leakage. Pulmonary complications entailed: pneumonia, pleural effusion, respiratory failure, pneumothorax, and/or acute respiratory distress syndrome. Different machine learning models were tested. Nomograms were constructed using Least Absolute Shrinkage and Selection Operator. Results: Between 2011 and 2017, 4228 patients underwent surgical resection for esophageal cancer, of which 18% developed anastomotic leakage and 30% a pulmonary complication. Of the 2199 patients with surgical resection for gastric cancer, 7% developed anastomotic leakage and 15% a pulmonary complication. In all cases, linear regression had the highest predictive value with the area under the curves varying between 61.9 and 68.0, but the difference with machine learning models did not reach statistical significance. Conclusion: Machine learning models can predict postoperative complications in upper gastrointestinal cancer surgery, but they do not outperform the current gold standard, linear regression

Background: In the randomized Asian REGATTA trial, no survival benefit was shown for additional gastrectomy over chemotherapy alone in patients with advanced gastric cancer with a single incurable factor, thereby discouraging surgery for these patients. The purpose of this study was to evaluate treatment strategies for patients with metastatic gastric cancer in daily practice in five European countries, along with relative survival in each country. Methods: Nationwide population‐based data from Belgium, Denmark, the Netherlands, Norway and Sweden were combined. Patients with primary metastatic gastric cancer diagnosed between 2006 and 2014 were included. The proportion of gastric resections performed and the administration of chemotherapy (irrespective of surgery) within each country were determined. Relative survival according to country was calculated. Results: Overall, 15 057 patients with gastric cancer were included. The proportion of gastric resections varied from 8·1 per cent in the Netherlands and Denmark to 18·3 per cent in Belgium. Administration of chemotherapy was 39·2 per cent in the Netherlands, compared with 63·2 per cent in Belgium. The 6‐month relative survival rate was between 39·0 (95 per cent c.i. 37·8 to 40·2) per cent in the Netherlands and 54·1 (52·1 to 56·9) per cent in Belgium. Conclusion: There is variation in the use of gastrectomy and chemotherapy in patients with metastatic gastric cancer, and subsequent differences in survival.

Contains fulltext : 193343.pdf (Publisher’s version ) (Open Access) BACKGROUND: For esophageal cancer, the number of retrieved lymph nodes (LNs) is often used as a quality indicator. The aim of this study is to analyze the number of retrieved LNs in The Netherlands, assess factors associated with LN yield, and explore the association with short-term outcomes. This is a population-based study on lymph node retrieval in patients with esophageal cancer, presenting results from the Dutch Upper Gastrointestinal Cancer Audit. STUDY DESIGN: For this retrospective national cohort study, patients with esophageal carcinoma who underwent esophagectomy between 2011 and 2016 were included. The primary outcome was the number of retrieved LNs. Univariable and multivariable regression analyses were used to test for association with >/= 15 LNs. PATIENTS AND RESULTS: 3970 patients were included. Between 2011 and 2016, the median number of LNs increased from 15 to 20. Factors independently associated with >/= 15 LNs were: 0-10 kg preoperative weight loss (versus: unknown weight loss, odds ratio [95% confidence interval]: 0.71 [0.57-0.88]), Charlson score 0 (versus: Charlson score 2: 0.76 [0.63-0.92]), cN2 category (reference: cN0, 1.32 [1.05-1.65]), no neoadjuvant therapy and neoadjuvant chemotherapy (reference: neoadjuvant chemoradiotherapy, 1.73 [1.29-2.32] and 2.15 [1.54-3.01]), minimally invasive transthoracic (reference: open transthoracic, 1.46 [1.15-1.85]), open transthoracic (versus open and minimally invasive transhiatal, 0.29 [0.23-0.36] and 0.43 [0.32-0.59]), hospital volume of 26-50 or > 50 resections/year (reference: 0-25, 1.94 [1.55-2.42] and 3.01 [2.36-3.83]), and year of surgery [reference: 2011, odds ratios (ORs) 1.48, 1.53, 2.28, 2.44, 2.54]. There was no association of >/= 15 LNs with short-term outcomes. CONCLUSIONS: The number of LNs retrieved increased between 2011 and 2016. Weight loss, Charlson score, cN category, neoadjuvant therapy, surgical approach, year of resection, and hospital volume were all associated with increased LN yield. Retrieval of >/= 15 LNs was not associated with increased postoperative morbidity/mortality.

Aims Seven countries (Denmark, France, Ireland, the Netherlands, Poland, Sweden, United Kingdom) collaborated to initiate a EURECCA (European Registration of Cancer Care) Upper GI project. The aim of this study was to identify a core dataset of shared items in the different data registries which can be used for future collaboration between countries. Methods Item lists from all participating Upper GI cancer registries were collected. Items were scored ‘present’ when included in the registry, or when the items could be deducted from other items in the registry. The definition of a common item was that it was present in at least six of the seven participating countries. Results The number of registered items varied between 40 (Poland) and 650 (Ireland). Among the 46 shared items were data on patient characteristics, staging and diagnostics, neoadjuvant treatment, surgery, postoperative course, pathology, and adjuvant treatment. Information on non-surgical treatment was available in only 4 registries. Conclusions A list of 46 shared items from seven participating Upper GI cancer registries was created, providing a basis for future quality assurance and research in Upper GI cancer treatment on a European level.

A D2 lymphadenectomy can be considered standard of surgical care for advanced resectable gastric cancer. Currently, several multimodality strategies are used, including postoperative monochemotherapy in Asia, postoperative chemoradiotherapy in the United States, and perioperative chemotherapy in Europe. As the majority of gastric cancer patients are treated outside the framework of clinical trials, quality assurance programs, including referral to high-volume centers and clinical auditing are needed to improve gastric cancer care on a nationwide level. J. Surg. Oncol. 2013;107:298-305. © 2012 Wiley Periodicals, Inc.

Research described in this thesis focuses on several aspects of gastric cancer care: staging and prognostication, multimodality treatment, and surgical quality assurance. PART I - STAGING AND PROGNOSTICATION Cancer staging is one of the fundamental activities in oncology.6,7 For over 50 years, the TNM classification has been a standard in classifying the anatomic extent of disease.8 In order to maintain the staging system relevant, the International Union Against Cancer (UICC) and the American Joint Committee on Cancer (AJCC) have collaborated on periodic revisions of this staging system, leading to the 7th edition in 2010.65 In Chapter 2, differences between the 6th and 7th edition TNM classification for gastric cancer are described, and both staging systems are compared with regards to complexity and predictive accuracy. In the 7th edition TNM classification, nodal status cut-off values were changed, leading to a more even distribution for the redefined nodal classification groups. This increased the predictive accuracy of N-classification. Overall, the TNM staging system became more complex, with an increase in the number of TNM groupings from 56 to 80, which did not result in an increased predictive accuracy. Future refinements of the TNM-classification should consider whether increased complexity is balanced by improved prognostic accuracy. Another change that was incorporated in the 7th edition TNM classification was the addition of tumor grade as an independent determinant of stage grouping in early stage tumors. With the significantly lower prognosis of poorly differentiated early stage adenocarcinomas, these tumors might become candidate for neoadjuvant therapy, given an accurate identification of these tumors with preoperative staging. In Chapter 3, the accuracy of preoperative histopathologic grading in adenocarcinomas of the gastroesophageal junction (GEJ) was evaluated. The overall accuracy of tumor grade assessment was 73%. However, in early stage tumors the sensitivity to detect a poorly differentiated tumor was only 43%, and 21% of patients with an early stage GEJ tumor were assigned to an incorrect stage/prognostic group based on preoperative tumor grading. Caution should therefore be exhibited in staging patients with esophageal adenocarcinoma based on preoperative biopsy data. Although the TNM classification can be used to assess a patient__s prognosis, tools for individual patient prognostication have been developed that significantly outperform the TNM-classification in prognostic accuracy. For gastric cancer, a nomogram has been developed based on a single US-institution database,12,13 and has been validated in several international patient cohorts.14-16 Chapter 4 describes the development of a new gastric cancer nomogram that not only can predict survival for patients directly after an R0 gastrectomy, but also for patients alive at time points after surgery. This conditional probability of survival nomogram was highly discriminating (concordance index: 0.772), and surviving one, two, or three years from surgery showed a median improvement of 5-year disease-specific survival of 7.2%, 19.1%, and 31.6%, as compared to the baseline prediction directly after surgery. This nomogram was based on variables available directly after surgery, while variables available with follow-up (such as weight loss and performance status) did not further improve the predictive accuracy of this nomogram. In Chapter 5, the performance of the original gastric cancer nomogram, which was based on patients who underwent surgery without multimodality therapy, was assessed in a group of patients who received postoperative chemoradiotherapy after an R0 resection for gastric cancer. The nomogram significantly underpredicted 5-year survival for patients who received postoperative chemoradiotherapy, indicating a benefit in survival for patients who receive postoperative chemoradiation after an R0 resection for gastric cancer. Furthermore, this study stresses the need for updating nomograms that incorporate multimodality therapy use. PART II - MULTIMODALITY TREATMENT Over the past decade, many trials have been performed in which the effect of multimodality treatment on survival for resectable gastric cancer was evaluated. In Chapter 6, an overview of the literature on the treatment of gastric cancer is presented, and the available multimodality strategies are discussed. Currently accepted regimens include postoperative monochemotherapy with S-1 in Asia,66 and perioperative chemotherapy and postoperative chemoradiotherapy in the Western world.57,58 In Chapter 7, patterns of recurrence and survival of patients who received postoperative chemoradiotherapy were compared to recurrence and survival patterns of patients who only underwent surgery. The local recurrence rate was significantly lower in the chemoradiotherapy group (5% versus 17%, P = 0.0015). Subgroup analysis revealed that this difference was even stronger in patients who underwent a gastrectomy with a limited (D1) lymph node dissection (2% versus 18%, P = 0.001), while no difference was found for patients who underwent an extended (D2) lymph node dissection. Additional analysis with prolonged follow-up showed a higher 2-year overall survival for patients who received postoperative chemoradiotherapy after a D1 lymphadenectomy compared to surgery alone, and no difference in overall survival for patients who received a D2 dissection. Postoperative chemoradiotherapy was also significantly associated with higher two-year overall survival for patients who underwent a microscopically irradical (R1) resection (66% versus 29%, P = 0.02). Results from this study indicate that, especially after a gastrectomy with a limited lymph node dissection, postoperative chemoradiotherapy has a major impact on local recurrence and overall survival. Postoperative chemoradiotherapy should be offered to patients who undergo a microscopically irradical (R1) resection. In Chapter 8, the results of a study on lymph node yield after gastric cancer resections are described. While it is suggested that more than 15 lymph nodes (LNs) should be evaluated for accurate staging of gastric cancer, LN yield in Western countries is generally low. The effect of preoperative chemotherapy on LN yield in gastric cancer is unknown. In this study, LN yields of patients who received preoperative chemotherapy and patients who only underwent surgery were compared. Preoperative chemotherapy was not associated with a decrease in LN yield, indicating that evaluating more than 15 LNs after gastrectomy is feasible, also after administration of preoperative chemotherapy. In Chapter 9, the final chapter of part II of this thesis, the study protocol of the currently accruing Dutch-Swedish-Danish CRITICS trial is described. This trial was initiated to determine which of the two currently used standard regimens for the multimodality treatment of gastric cancer in the Western world, postoperative chemoradiotherapy, or perioperative chemotherapy, should be preferred. In this trial, all patients receive three cycles of preoperative ECC (epirubicin, cisplatin, and capecitabine), followed by D1+ surgery (D2 dissection without splenectomy or pancreatectomy). Postoperative therapy consists of another three cycles of ECC, or chemoradiotherapy with capecitabine and cisplatin without epirubicine. Results of this study will play a key role in the future management of patients with resectable gastric cancer. PART III - SURGICAL QUALITY ASSURANCE As an introduction to part III of this thesis, in Chapter 10, the results of a systematic review of the literature on quality of care indicators for gastric cancer surgery are described. The availability of specific literature on quality of care indicators was limited, but several indicators could be identified in more general literature on gastric cancer surgery. High hospital volume was found to be strongly related to lower postoperative mortality and higher long-term survival. Several quality indicators regarding operative technique were identified, including the performance of an extended lymphadenectomy, avoiding a routine spleen and pancreatic tail resection, and the use of a pouch reconstruction. Free resection margins were also found to be strongly associated with improved long-term survival. In Chapter 11 and Chapter 12, incidence and survival patterns for tumors of the esophagus, GEJ, and stomach in the Netherlands over the past 20 years are described. While the incidence of esophageal adenocarcinoma has doubled, the incidence of both tumors of the GEJ and stomach has decreased. These findings most likely reflect true changes in disease burden, rather than being the result of changes in diagnosis or reclassification. The increasing incidence of esophageal adenocarcinoma can be attributed to the increasing incidence of obesity and gastroesophageal reflux disease.67,68 Over the study period, five-year survival for non-metastatic esophageal cancer strongly improved (12% to 25% for adenocarcinoma, 12% to 19% for squamous cell carcinoma), while five-year survival for non-metastatic GEJ cancer (20%) and stomach cancer (32%) remained stable. In Chapter 13, patterns of care for gastric cancer in the Netherlands over the past 20 years are described. Whereas resection rates for stage I-III gastric cancer have remained stable at about 85%, the use of preoperative and/or postoperative chemotherapy has strongly increased since 2005. In 2008, nearly 40% of the patients with stage I-III gastric cancer received preoperative or postoperative chemotherapy with curative intent, and it is likely that since then, this percentage has further increased. In Chapter 14, the results of a study on hospital volumes, mortality, and long-term survival for esophagogastric cancer surgery in the Netherlands between 1989 and 2009 are described. In the Netherlands, a minimum hospital volume standard of at least 10 esophagectomies per year was introduced in 2006, while during the study period, no such standard was present for gastrectomies. During the study period, esophagectomy was effectively centralized in the Netherlands, and in 2009, 64% of all esophagectomies were performed in annual volumes of __21/year. Gastrectomy has not been centralized, and in 2009 only 5% of all gastrectomies were performed in annual volumes of __21/year. Whereas short-term and long-term survival after esophagectomy and gastrectomy improved over the years, this improvement was significantly stronger for esophagectomy. High hospital volume was associated with lower 6-month mortality (HR 0.48, P < 0.001) and longer 3-year survival (HR 0.77, P < 0.001) after esophagectomy, but not after gastrectomy. However, for gastrectomy, the number of high volume resections in the current study was too low to detect a statistical significant difference in outcomes when compared with low volume resections. This study indicates an urgent need for improvement in the treatment of resectable gastric cancer in the Netherlands. Chapter 15 describes the results of a study on the effect of hospital type on outcomes after esophagectomy and gastrectomy in the Netherlands. Hospitals were categorized into university hospitals, teaching non-university hospitals, and non-teaching hospitals. Three-month mortality after esophagectomy in university hospitals was 2.5%, compared to above 4% in non-university hospitals (P = 0.006). After gastrectomy, three-month mortality was 4.9% in university hospitals, and 8.7% in non-university hospitals (P < 0.001). Both after esophagectomy and gastrectomy, three-year survival was higher in university hospitals compared to non-university hospitals. No differences in mortality or survival were found between teaching and non-teaching non-university hospitals. However, when analyzing differences between individual hospitals, there were non-university hospitals with excellent outcomes. Therefore, it can be concluded that centers of excellence can not be designated solely by hospital type, and that detailed information on case-mix and outcomes is needed to identify centers of excellence. In Chapter 16, the results of an international study on esophagogastric cancer surgery between 2004 and 2009 in several European countries are described. Differences in resection rates, postoperative mortality, survival and hospital volumes were compared between the Netherlands, Sweden, Denmark, and England. In the Netherlands, postoperative mortality was average after esophagectomy (4.6%), but significantly higher after gastrectomy (6.9%) when compared to the other countries. Although increasing hospital volume was associated with lower 30-day mortality both after esophagectomy and gastrectomy, differences in outcomes between countries could not just be explained by existing differences in hospital volumes. To further investigate the differences in outcomes, a European upper GI audit is currently initiated.