Upload
others
View
2
Download
0
Embed Size (px)
Citation preview
CLINICAL AND ECONOMIC OUTCOMES FOR GASTRIC CANCER
PATIENTS TREATED WITH GASTRECTOMY AT CENTRES WITH
AND WITHOUT CANCER SURGERY CENTRE DESIGNATION IN
ONTARIO
by
Yunni Jeong
A thesis submitted in conformity with the requirements for the degree of Master of Science in Health Services Research
Institute of Health Policy, Management and Evaluation University of Toronto
© Copyright by Yunni Jeong 2019
ii
CLINICAL AND ECONOMIC OUTCOMES FOR GASTRIC CANCER PATIENTS
TREATED WITH GASTRECTOMY AT CENTRES WITH AND WITHOUT CANCER
SURGERY CENTRE DESIGNATION IN ONTARIO
Yunni Jeong
Master of Sciences in Health Services Research
Institute of Health Policy, Management and Evaluation University of Toronto
2019
Abstract
Gastric cancer places high clinical and economic burden on healthcare systems. This
study compares patient outcomes between centres with and without cancer surgery centre
designation (CSCD). Gastric cancer patients diagnosed 2002-2014, and treated with gastrectomy
were identified using Ontario’s linked administrative databases. Outcomes included 90-day
mortality, overall survival, and healthcare costs. Institutions with CSCD (n =1,436) were
associated with lower 90-day mortality (OR 0.70, 95% CI 0.52-0.94, p=0.02), and similar overall
survival (HR 0.94, 95% CI 0.85-1.04, p=0.24) as institutions without CSCD (n = 1,494). The
cost analysis included 1,143, and 1,256 patients treated at institutions with and without CSCD,
respectively. Adjusted mean monthly costs were $3,310 (95% CI $2,384-$4,595) for the CSCD
group, and $3,430 (95% CI $2,469-$4,765) for the non-CSCD group (p=0.36). Treatment at
institutions with CSCD may result in lower 90-day mortality, and similar overall survival, and
costs of care for gastric cancer patients.
iii
Co-Authorship
This thesis was the product of Yunni Jeong in collaboration with her supervisors Dr. Natalie G.
Coburn, Dr. Nicole Mittmann, Dr. Julie Hallet, and Dr. Alyson L. Mahar. Study design was
completed by Yunni Jeong, Dr. Coburn, Dr. Mittmann, Dr. Hallet, and Dr. Mahar. The algorithm
utilized to identify patients without metastatic disease was applied with guidance from Dr.
Alyson Mahar and Mr. Brandon Zagorski. Data linkage and macro application was performed by
Ms. Qing Li, Senior Research Analyst, Institute for Clinical Evaluative Sciences.
iv
Acknowledgments
First and foremost, I would like to thank my thesis supervisor Dr. Natalie Coburn, for her
mentorship, wisdom, and unwavering support.
Much gratitude to my Thesis Committee members for their patient time, effort, and knowledge:
Dr. Nicole Mittmann, Dr. Julie Hallet, and Dr. Alyson Mahar. This work would not have been
possible without you.
Last but not least, thank you to my family and friends for all their support and encouragement.
v
Table of Contents
Co-Authorship......................................................................................................................iii
Acknowledgments.................................................................................................................iv
Table of Contents...................................................................................................................v
List of Tables........................................................................................................................ix
List of Figures.......................................................................................................................xi
List of Appendices...............................................................................................................xii
Chapter 1...............................................................................................................................1
Introduction.....................................................................................................................11
1.1 Gastric Cancer Overview......................................................................................................1
1.1.1 Descriptive Epidemiology.......................................................................................................1
1.1.2 Risk Factors.............................................................................................................................3
1.1.3 Diagnosis and Workup............................................................................................................5
1.1.4 Staging....................................................................................................................................6
1.1.5 Treatment................................................................................................................................8
1.2 Gastric Cancer Outcomes....................................................................................................10
1.2.1 Short-term Mortality.............................................................................................................10
1.2.2 Long-term Survival...............................................................................................................10
1.2.3 Costs of Care.........................................................................................................................11
1.2.4 Factors Associated with Gastric Cancer Clinical and Economic Outcomes.........................13
1.3 Cancer Surgery Centre Designation....................................................................................16
1.3.1 Introduction to Cancer Surgery Centre Designation.............................................................16
1.3.2 Controversies with CSCD.....................................................................................................17
1.3.3 Cancer Surgery Centre Designation in Ontario, Canada.......................................................18
1.4 Ontario Healthcare System.................................................................................................19
1.5 Summary of Evidence and Rationale..................................................................................19
Chapter 2.............................................................................................................................21
Methods..........................................................................................................................212
vi
2.1 Study Objectives..................................................................................................................21
2.2 Hypotheses...........................................................................................................................21
2.3 Study Design........................................................................................................................21
2.4 Study Population.................................................................................................................22
2.4.1 Inclusion Criteria...................................................................................................................22
2.4.2 Exclusion Criteria..................................................................................................................23
2.4.3 Restrictions............................................................................................................................23
2.5 Study Perspective and Timeframe.......................................................................................23
2.5.1 90-dayMortalityandOverallSurvival...................................................................................23
2.5.2 CostsofCare.........................................................................................................................24
2.6 Study Setting........................................................................................................................25
2.7 Data Sources........................................................................................................................26
2.7.1 Ontario Cancer Registry........................................................................................................26
2.7.2 Canadian Institute for Health Information-Discharge Abstract Database and Same Day Surgery..............................................................................................................................................26
2.7.3 National Ambulatory Care Reporting System.......................................................................26
2.7.4 Home Care Database.............................................................................................................27
2.7.5 Ontario Health Insurance Plan..............................................................................................27
2.7.6 Registered Persons Database.................................................................................................27
2.7.7 Activity Level Reporting.......................................................................................................27
2.8 Study Variables...................................................................................................................27
2.8.1 PrimaryExposure..................................................................................................................29
2.8.2 Outcomes..............................................................................................................................31
2.8.3 Covariates..............................................................................................................................33
2.8.4 Confounders..........................................................................................................................38
2.9 Statistical Analysis...............................................................................................................39
2.9.1 Objective 1: CSCD and 90-Day Mortality............................................................................39
2.9.2 CSCD and Overall Survival..................................................................................................39
2.9.3 CSCD and Costs of Care.......................................................................................................40
2.10 Sensitivity and Subgroup Analyses....................................................................................41
2.10.1 Cohort Excluding 90-Day Decedent Patients......................................................................41
2.10.2 Cohort Restricted to Patients Treated with Distal Gastrectomy Only.................................41
2.10.3 Cohort Restricted to Patients with Complete Cost Data......................................................41
vii
2.10.4 Cohort Including Patients with Missing Institution Information.........................................41
2.10.5 Cohort Restricted to Non-Decedent Patients.......................................................................41
2.10.6 Cluster Analyses..................................................................................................................42
2.10.7 Missing Data.......................................................................................................................42
2.11 Ethical Considerations.......................................................................................................42
Chapter 3.............................................................................................................................44
Results............................................................................................................................443
3.1 Cohort Selection..................................................................................................................44
3.2 Cohort Description..............................................................................................................45
3.2.1 OverallStudyCohort.............................................................................................................45
3.2.2 CostStudyCohort.................................................................................................................47
3.3 Objective 1...........................................................................................................................49
3.4 Objective 2...........................................................................................................................52
3.5 Objective 3...........................................................................................................................56
3.6 Sensitivity and Subgroup Analyses......................................................................................60
3.6.1 Exclusion of 90-day Decedent Patients.................................................................................60
3.6.2 Cohort of Patients Who Underwent Distal Gastrectomy Only..............................................61
3.6.3 Complete Costs.....................................................................................................................62
3.6.4 Cohort of Missing Institution Information versus Non-Missing Institution Information......63
3.6.5 Cohort Restricted to Non-Decedent Patients.........................................................................64
3.6.6 Cluster Analyses....................................................................................................................64
Chapter 4.............................................................................................................................66
Discussion.......................................................................................................................664
4.1 Summary of Study Findings................................................................................................66
4.2 Summary of Study Cohorts.................................................................................................67
4.3 Centralization to Institutions with Cancer Surgery Centre Designation............................68
4.3.1 Objective1............................................................................................................................69
4.3.2 Objective2............................................................................................................................71
4.3.3 Objective3............................................................................................................................72
4.4 Study Limitations and Strengths.........................................................................................74
4.5 Implications.........................................................................................................................77
viii
4.6 Future Directions.................................................................................................................79
References or Bibliography.................................................................................................81
Appendices...........................................................................................................................89
ix
List of Tables
Table 1. Sources and definitions of patient and disease characteristics
Table 2. Cancer Care Ontario Institutions with CSCD, by type
Table 3. Definitions of type of gastrectomy performed
Table 4. Patient and disease characteristics of gastric cancer patients who underwent gastrectomy
in Ontario (n=2,930)
Table 5. Patient and disease characteristics of gastric cancer patients diagnosed between June 1,
2002 and March 31, 2013 who underwent gastrectomy in Ontario (n=2,399)
Table 6. Univariable and multivariable analyses of 90-day mortality for patients who underwent
gastrectomy (n=2,930)
Table 7. Median survival in days and results of Log-Rank tests comparing survival curves of
categories of variables
Table 8. Cox Proportional Hazards Analysis of overall survival in patients who underwent
gastrectomy in Ontario (n=2,930)
Table 9. Linear regression of log-transformed monthly costs for patients who underwent
gastrectomy in Ontario (n=2,399)
Table 10. Cox proportional hazards analyses of overall survival in patients treated with
gastrectomy, excluding patients who died within ninety days of surgery (n=2,718)
Table 11. Linear regression of log-transformed monthly costs of care for patients who underwent
gastrectomy, excluding patients who died within ninety days of surgery (n=2,216)
Table 12. Frequency of 90-day mortality by type of gastrectomy for centres with and without
CSCD
Table 13. Logistic Regression of 90-day mortality in patients who underwent distal gastrectomy
without multivisceral resection or preoperative therapy
x
Table 14. Cox Proportional Hazards Regression of overall survival in patients treated with distal
gastrectomy without multivisceral resection or preoperative therapy
Table 15. Linear regression of log-transformed monthly costs of care for patients who underwent
distal gastrectomy without multivisceral resection or preoperative therapy (n=1,183)
Table 16. Linear regression with monthly costs of care for patients who underwent gastrectomy
diagnosed between 2006 and 2013 (n=1,506)
Table 17. Logistic Regression of 90-day mortality in patients with and without missing
institution information (n=3,103)
Table 18. Linear regression of log-transformed monthly costs of care for patients with and
without missing institution information
Table 19. Linear regression of log-transformed monthly costs of care for non-decedent patients
(n=851)
xi
List of Figures
Figure 1. Outline of study timeframe for analyses of 90-day mortality and overall survival in
patients treated with gastrectomy in Ontario
Figure 2. Outline of study timeframe for cost analysis of patients treated with gastrectomy in
Ontario
Figure 3. Conceptual model depicting the relationship of patient, disease, and healthcare system
characteristics with outcomes
Figure 4. Cohort selection process for analyses of 90-day mortality and overall survival
Figure 5. Cohort selection process for cost analysis
Figure 6. Survival curves for patients who underwent gastrectomy, stratified by CSCD
Figure 7. Distribution of cost per person per 30 days alive
Figure 8. Distribution of log-transformed cost per patient per month
Figure 9. Proportions of patient-level mean total costs by contributing cost parameters for
patients treated at institutions with CSCD
Figure 10. Proportions of patient-level mean total costs by contributing cost parameters for
patients treated at institutions without CSCD
Figure 11. Distribution of inpatient costs
Figure 12. Median survival by institution, with institutions ordered by descending volume
Figure 13. Median costs per institution, with institutions ordered by descending volume
Figure 14. Survival curves for patients who underwent gastrectomy with exclusion of 90-day
decedents, stratified by cancer surgery centre designation
xii
List of Appendices
Supplemental Table 1. ICD-10 codes defining gastric cancer
Supplementary Table 2. ICD-O-3 codes defining adenocarcinoma
Supplementary Table 3. Test for multicollinearity of independent variables included in the
models
Supplementary Table 4. Patient and disease characteristics of gastric cancer patient cohort
restricted to patients who lived beyond 90 days following gastrectomy (n=2,718)
Supplementary Table 5. Patient and disease characteristics of gastric cancer patient cohort
restricted to patients treated with distal gastrectomy only (n=1,346)
Supplementary Table 6. Patient and disease characteristics of gastric cancer patient patients
diagnosed between June 1, 2006 and March 31, 2013 (n=1,506)
Supplementary Table 7. Patient and disease characteristics of gastric cancer patients with and
without missing institution information (n=3,103)
Supplementary Table 8. Patient and disease characteristics of non-decedent gastric cancer
patients June 1, 2002 and March 31, 2013 (n=851)
1
Chapter 1
Introduction 11.1 Gastric Cancer Overview
Gastric cancer is the fifth most common cancer and third leading cause of cancer-related
mortality in the world (1). In 1975, gastric cancer was the most common neoplasm in the world.
Since that time, the global incidence of gastric cancer has been on the decline with increased
food refrigeration, decreasing prevalence of Helicobacter pylori (H. pylori), and reduced tobacco
consumption (1, 2). Advances in diagnostics and therapy for gastric cancer have occurred, and
processes of care that improve gastric cancer outcomes have been identified (3-5). Despite these
developments, and the decreasing incidence, gastric cancer remains both clinically and
economically burdensome to healthcare systems (6, 7).
Gastric adenocarcinoma is the most common histopathologic subtype of gastric cancer.
Other cancers arise much less frequently in the stomach, including gastrointestinal stromal
tumours and other sarcomas, lymphomas, and neuroendocrine tumours (8). Henceforth, gastric
cancer will be used synonymously with gastric adenocarcinoma.
1.1.1 Descriptive Epidemiology
Gastric cancer incidence has significant geographic variability (1). Regions such as
Eastern Asia, Central and Eastern Europe, and South America have a high incidence of gastric
cancer with age-standardized incidence rates of 35.4 per 100,000 for males and 13.8 for females,
20.3 per 100,000 for males and 8.9 per 100,000 for females, and 14.2 per 100,000 for males and
7.0 per 100,000 for females, respectively. Conversely, regions such as Northern Europe, Western
Europe, Africa, and North America have a low incidence of gastric cancer with age-standardized
incidence rates of 7.4 per 100,000 for males and 3.7 per 100,000 for females, 8.8 per 100,000 for
males and 4.3 per 100,000 for females, 4.5 per 100,000 for males and 3.2 per 100,000 for
females, and 5.5 per 100,000 for males and 2.8 per 100,000 for females, respectively (3, 8).
World cancer data from 2012 estimated more than 70% of new cases occur in developing
countries, with an age-standardized incidence of 17.4 per 100,000 for males and 7.5 per 100,000
for females (1).
2
Gastric cancer mortality also varies significantly throughout the world. Overall, an
estimated 723,000 deaths due to gastric cancer occur annually throughout the world, making it
the fifth most common cause for cancer-related death (1). Similar to incidence, deaths
attributable to gastric cancer occur more frequently in males (age-standardized mortality rate of
12.8 per 100,000 in 2012) than in females (age-standardized mortality rate of 5.7 per 100,000 in
2012), and shows variability between countries (1).
In Canada, incidence rates are significantly less than the global incidence rates. In 2017,
the estimated age-standardized incidence rates for males was 11.8 per 100,000 for males and 5.9
per 100,000 for females (9). In the same year, an estimated 2,100 deaths occurred in Canada due
to gastric cancer—age-standardized mortality rates were 6.9 per 100,000 in males and 3.6 per
100,000 deaths in females (9). Ontario, the largest province in Canada, has age-standardized
incidence rates slightly higher than the national equivalent at 12.0 per 100,000 for males and 6.5
per 100,000 for females, and age-standardized mortality rates of 6.7 per 100,000 for males and
3.6 per 100,000 for females (9).
Although a higher incidence of the gastric cancer exists in Asian countries such as Japan
and Korea, the survival rate has improved over time. However, the same improvement in stage-
matched survival has not been observed in low-incidence regions such as North America (3, 4).
High mortality rates are also observed in Central and Eastern Europe as well Central and South
America (1). In low-incidence countries such as the United States and Canada, presentation with
advanced disease, and high recurrence rates of 40-65% have been described as contributing
factors to the poor prognosis of patients with gastric cancer (3). Theories to explain the
differences in the incidence and outcomes for gastric cancer patients in the East versus the West
include biologic (tumour location, tumour type), patient comorbidities, and treatment differences
(10).
Although the incidence of gastric cancer in Canada and Ontario is low, the high mortality
of Canadian patients diagnosed with gastric cancer behooves investigation into factors affecting
outcomes for these patients.
3
1.1.2 Risk Factors
A wide range of risk factors for gastric cancer have been described, including ethnicity,
infectious, dietary, environmental, medical, and hereditary factors.
The remarkable differences in incidence and mortality of gastric cancer in Asia and
Western countries suggest an ethnic, or environmental origin as a possible risk factor for the
development of gastric cancer (3). In particular, individuals of Korean, Vietnamese, Japanese,
Native American, and Hawaiian descent have been identified as a high risk group; Latino,
Chinese, and black individuals as an intermediate risk group; and Filipino and white individuals
as a low risk group for gastric cancer development (3). Interestingly, the first generation of
individuals who migrate from high-incidence to low-incidence regions retain the risk rate of their
native country, while subsequent generations acquire the risk rate of their new environment,
suggesting an environmental etiology over an ethnic etiology (3, 11).
A very significant factor in the etiology of gastric cancer is H. pylori. In 1994, the
International Agency for Research on Cancer declared the gram-negative microaerophilic
spirochete as a human carcinogen for its role in gastric cancer carcinogenesis (12). The pattern of
H. pylori infection affects outcomes; patients with antrum-predominant gastritis have increased
acid production, predisposition to duodenal ulcer formation, and decreased risk of gastric cancer
formation, while chronic corpus-predominant gastritis is associated with lower acid secretion,
multifocal atrophic gastritis, and increased risk of intestinal type gastric cancer formation (12,
13). H. pylori virulence factors such as cytotoxin-associated gene (Cag)A, CagL, vacuolating
cytotoxin A (VacA), outer inflammatory protein A (OipA), and H. pylori peptidoglycan, as well
as the inflammatory response to, the oxidative stress from, and the E-cadherin gene induction by
H. pylori infection have been described in gastric tumorigenesis (12, 13). Factors such as diet
and lifestyle behaviours may also play a role in the gastric cancer pathogenesis mediated by H.
pylori infection. A synergistic effect on the development of gastric cancer is observed with a
high-salt diet in combination with H. pylori infection and smoking in the context of H. pylori
infection may also enhance gastric cancer formation. On the other hand, concurrent helminth
infection and antioxidant intake with H. pylori infection may have a protective effect against
H.pylori-mediated gastric cancer (13). H. pylori eradication through a prophylactic regimen has
4
been shown effective for the prevention of metachronous lesions in early stage gastric cancer
(14).
The etiology of gastric cancer also includes dietary factors. High consumption of salt
including cured or salt-preserved meat and fish, as well as pickled vegetables, high starch, and
low protein have been associated with increased gastric cancer risk (15, 16). Gastric
carcinogenesis is thought to occur through enhanced N-nitroso compounds formed by the
nitrosation of amides or amines in the stomach by nitrites, but is inhibited by Vitamin C in
gastric juice (15). High fruit and vegetable intake, therefore, has been described as protective
against gastric cancer formation (15, 16).
The environment in which a person lives is also described as playing a potential role in
gastric cancer development. Environmental risk factors for gastric cancer include lack of
refrigeration of food, occupation hazards such as mining, farming, refining, and fishing, as well
as rubber, timber, and asbestos processing. Lifestyle factors have also been shown to be
associated with gastric carcinogenesis. Alcohol, by acting as a gastric irritant, and smoking, by
decreasing prostaglandins important to maintain gastric mucosal integrity are involved in gastric
cancer development (16).
Other medical diagnoses are also associated with increased risk of developing gastric
cancer. Pernicious anemia is a condition characterized by achlorhydria, which leads to the
destruction of chief and parietal cells in an autoimmune reaction. Gastric mucosa then becomes
atrophic, leading to antral and intestinal metaplasia (17). Rare inherited disorders such as Li-
Fraumeni syndrome (tumour suppressor gene p53 mutation), Peutz-Jeghers syndrome, hereditary
diffuse gastric cancer (E-cadherin or CDH1 mutation), or hereditary nonpolyposis colorectal
cancer or Lynch syndrome (microsatellite instability) have also been linked to gastric cancer (17,
18). While familial adenomatous polyposis is associated with benign fundic gland polyps, a new
autosomal dominant gastric polyposis syndrome called gastric adenocarcinoma and proximal
polyposis of the stomach (GAPPS), characterized by fundic gland polyposis of the gastric body
and fundus, has been reported as progressing to dysplasia and intestinal type gastric
adenocarcinoma (17, 19).
Modifiable risk factors of gastric cancer necessitate investigation to identify targets for
the potential prevention of gastric cancer development. Furthermore, factors strongly correlated
5
with the development of gastric cancer warrant investigation to identify those at risk for
development of gastric cancer, facilitate early diagnosis, and may improve outcomes for those
diagnosed with gastric cancer.
1.1.3 Diagnosis and Workup
To guide diagnostic workup, a comprehensive history should be taken to elicit potential
symptoms of gastric cancer, including dyspepsia, dysphagia, involuntary weight loss, vomiting,
early satiety, and bleeding (8). Patients should also be screened for a family history of gastric
cancer (20). Physical examination should be conducted with a focus on detection of features
suggestive of advanced disease, including palpable abdominal masses, ascites, hepatomegaly,
left supraclavicular lymphadenopathy (Virchow’s node), left axillary lymphadenopathy (Irish’s
node), periumbilical lymphadenopathy (Sister Mary-Joseph’s node), or cachexia (3). Bloodwork,
including a complete blood count to assess for iron deficiency anemia, and comprehensive
chemistry profile to assess renal and liver function, should also be performed (8, 20). Diagnosis
of gastric cancer should be made by biopsy from endoscopic examination, or from a metastatic
disease site. Biopsies with adequate tissue should be taken to collect 6-8 specimens.
In high-incidence countries such as Korea and Japan, systematic screening programs exist
to facilitate gastric cancer detection at earlier stages (21, 22). Early gastric cancers also have a
favourable prognosis, further supporting screening programs (22). In addition to increased
diagnosis of gastric cancer at earlier stage, screening has afforded other benefits of early gastric
cancer detection, such as more minimally invasive therapeutic treatments such as endoscopic
mucosal resection and endoscopic submucosal resection (23). These interventions have also
resulted in increased long-term outcomes and quality of life for patients with early gastric cancer
(23). In high incidence regions such as China, Japan, and Korea, screening programs have been
found to be cost-effective (24).
In Canada, screening is not considered cost-effective given the low incidence of gastric
cancer in the country, as population-level screening for rare diseases is costly (25). A lack of
population-based screening programs and the nonspecific symptomatology of gastric cancer thus
lead to presentation with the disease at advanced stages, and contribute to the poor prognosis of
gastric cancer patients in these low-incidence regions (8, 21). The insidious onset and lack of
early detection perpetuated by an absence of systematic screening further contributes to high
6
mortality for patients diagnosed with gastric cancer in Canada, and poses questions regarding the
processes of care that affect outcomes for these patients.
1.1.4 Staging
Worldwide, the most commonly used system for staging is that which is maintained by
the Union for International Cancer Control (UICC) and the American Joint Committee on
Cancer (AJCC). This system classifies the anatomic extent of cancer by description of primary
tumour (T), status of regional lymph nodes (N), and presence or absence of distant metastases
(M) in the TNM classification (26). Accurate clinical staging is important in the guidance of
clinical decision-making and appropriate administration of treatment. Preoperative evaluation
has also been described as a specific process of care for gastric cancer that influences the
outcome of health following the care received by a patient (4)
Appropriate clinical investigations need to be conducted to obtain accurate disease
staging. Endoscopic examination is important for establishing the location of the gastric tumour
(20). Cross-sectional imaging with a CT scan of the chest, abdomen, and pelvis, is the
preoperative staging investigation recommended by many gastric cancer guidelines, including
those from the National Comprehensive Cancer Network (NCCN), the European Society for
Medical Oncology (ESMO), and Cancer Care Ontario’s Program in Evidence-Based Care
(PEBC) (8, 20, 27). The accuracy, sensitivity, and specificity of CT varies from 60-100%, 60-
100%, and 92-100%, respectively, so adjunct investigations may be employed to compensate for
deficiencies in CT performance, including endoscopic ultrasound (EUS), positron emission
tomography (PET) scan, magnetic resonance imaging (MRI), and diagnostic laparoscopy (27,
28). These alternative investigations have specific indications, however, and should only be
utilized if the results of the adjunct investigations will change the management plan (29).
EUS may be employed in patients with planned curative-intent treatment with suspicion
for local invasion or involvement of regional lymph nodes, and, if technically feasible and safe,
combined with fine-needle aspiration to further inform stage and subsequent management (29).
A systematic review and meta-analysis of the utility of EUS for preoperative staging for gastric
cancer by Cardoso et al. reported EUS may be most useful for advanced cancers with a pooled
accuracy of 85% for T3 staging and 69% for T4 staging (30). From the same study, the pooled
7
accuracy, sensitivity, and specificity of EUS for N staging was 64%, 74%, and 80%, respectively
(30).
PET scan may be helpful in the detection of anatomically small or occult foci of
metastastic disease (8, 31). For N staging in particular, PET has been reported to have low
sensitivity (40%) but high specificity (98%). However, PET has poor spatial resolution, and
limited utility in the context of mucinous carcinoma, signet ring cell carcinoma, and poorly
differentiated adenocarcinomas as these diseases are less PET avid (31). Thus, PET is not
routinely recommended for gastric cancer staging (5).
MRI scans may be useful to aid in further characterization of liver lesions (29). MRI has
also been reported to have improved performance accuracy over CT with an overall accuracy of
83% for T staging. MRI scans, however, are limited by motion artifacts from respiration,
prolonged examination times, high costs, and a lack of a standardized gastric protocol. MRI
scans, like PET scans are therefore also not recommended to be routinely performed unless
finding of metastatic disease will change the treatment plan (29).
Diagnostic laparoscopy is recommended for patients under consideration for neoadjuvant
therapy prior to initiation of chemotherapy (8, 20, 29). Additionally, patients with clinical
suspicion of T3 or T4 disease and patients at higher risk for metastatic disease—those with
poorly differentiated tumours and those with higher nodal disease burden should undergo
diagnostic laparoscopy prior to initiation of therapy (29). Cytologic examination of ascitic fluid
or instillation of 200 mL of normal saline for peritoneal washings at the time of diagnostic
laparoscopy may be used to increase the informative yield (27). The aim of diagnostic
laparoscopy should be to increase staging accuracy beyond that of the CT scan to potentially
avoid the morbidity of a laparotomy without curative intent in these patients (29, 32). In a
systematic review of the accuracy and indications for diagnostic laparoscopy prior to curative-
intent gastric cancer resection by Leake et al., diagnostic laparoscopy was particularly useful in
the detection of metastatic disease, especially peritoneal carcinomatosis, with accuracy of
detecting overall metastatic disease reported between 90-100% (27). Diagnostic laparoscopy has
also been reported to change the management plan in up to 60% of cases (27).
Substantial evidence outlines the appropriate staging of patients with gastric cancer and
how accurate staging guides treatment for these patients. However, in a population-based study
8
exploring the processes of care for patients with gastric cancer in Ontario, the underuse of both
preoperative radiology and endoscopy was found (4). Furthermore, relevant information,
including tumour size and pathologic markers were not consistently recorded in pathology or
operative reports (33). This variation in the application of guidelines that affect the management
of and outcomes for patients with gastric cancer warrants further investigation.
1.1.5 Treatment
Many treatment options exist for gastric cancer patients depending on stage and treatment
intent, including therapeutic endoscopy, surgery, chemotherapy, radiotherapy, targeted therapies,
or any combination of these. Before initiation of treatment, all patients with a confirmed
diagnosis of gastric cancer and completed appropriate workup should be discussed at a
multidisciplinary tumour board (MTB) meeting (29). MTB meetings are regularly occurring,
formal meetings whereby networks of cancer care specialists prospectively review individual
patients with cancer, discuss the diagnosis, and formulate management plans in an evidence-
based approach to improve diagnostic accuracy, clinical practice guideline adherence, and
clinical outcomes (34). The care of patients with gastric cancer should also be delivered by a
multidisciplinary team, which may include surgeons, medical oncologists, radiation oncologists,
radiologists, pathologists, gastroenterologists, nurses, social workers, palliative care specialists,
and dietitians (5).
1.1.5.1 Surgery
Surgical resection is the cornerstone of curative therapy for gastric cancer. Surgical
interventions for gastric cancer include local resection, segmental gastrectomy, proximal
gastrectomy, pylorus-preserving gastrectomy, distal gastrectomy, and total gastrectomy (35).
Non-resectional surgical options include bypass surgery, gastrostomy, and jejunostomy.
Selection of gastrectomy is dependent on tumour location, and stage. For clinically node positive
tumours, or T2 to T4a tumours, the standard oncologic surgical procedures are total or distal
gastrectomy (35). Distal tumours may be treated with distal gastrectomy if a satisfactory
proximal margin can be obtained. Total gastrectomy should be considered for proximal tumours,
or tumours along the greater curvature with involvement of 4sb lymph nodes along the left
gastroepiploic artery (35). T1a lesions with no positive nodes should be considered for removal
by endoscopic methods (5).
9
Beyond the appropriate type of surgery performed, quality oncologic surgery for gastric
cancer should focus on complete removal of the cancer, or R0 resection, and adequate lymph
node dissection (5, 35). To assist with achievement of R0 resection, intraoperative assessment of
margin status through frozen section is an important consideration (5). For curative-intent
resection in patients with advanced, non-metastatic gastric cancer, more aggressive (D2)
lymphadenectomy is the preferred method of lymph node dissection, and for patients with early
gastric cancer, or patients with significant comorbidities, a less aggressive (D1)
lymphadenectomy is the preferred method of lymph node dissection (5). Furthermore, a
minimum of 16 lymph nodes need to be assessed for patients treated with curative-intent
resection for gastric cancer for improved accuracy of staging and improved survival (5).
1.1.5.2 Adjuvant Therapy
All gastric cancer patients with curatively resectable tumours should be considered for
neoadjuvant and/or adjuvant therapy (36). The two most common regimens were the
postoperative 5-fluorouracil (5-FU)-based chemoradiotherapy (Macdonald) approach or
perioperative epirubicin/cisplatin/5-FU (ECF)-based chemotherapy (Cunningham/Medical
Research Council Adjuvant Gastric Infusional Chemotherapy (MAGIC)) approach (37, 38). If
contraindications to either the Macdonald or MAGIC approaches exist, adjuvant chemotherapy is
considered a reasonable alternative (36). A recent multicenter randomized phase 3 trial
comparing perioperative chemotherapy with docetaxel, oxaliplatin, and fluorouracil/leucovorin
(FLOT) versus ECF or epirubicin/cisplatin/capecitabine (Xeloda) (ECX) revealed improved 30-
and 90-day mortality rates, progression-free survival, and overall survival with FLOT compared
to ECF/ECX (39). Within the study period of this project, the standard treatments were
perioperative chemotherapy (MAGIC), or postoperative chemoradiotherapy, as FLOT was not
introduced as standard therapy until after the end of the study period, in 2017.
Despite guidelines on the surgical and non-surgical treatments recommended for
curative-intent resectable gastric cancer tumours, variation in management exists. A population-
based study in Ontario investigated variation in practice patterns within the universal, publicly
funded healthcare system for the treatment of patients with early stage, resectable gastric cancer
and found significant variation in the types of surgical and non-surgical treatments provided (33).
The rare incidence of potentially curative gastric cancer in combination with variability in the
10
care provided to patients with curable gastric cancer raises interesting questions regarding the
optimal organization of care for patients, with consideration of providers and healthcare systems.
1.2 Gastric Cancer Outcomes
Many important outcomes exist for patients with gastric cancer. This study focuses on
short-term mortality (Section 1.2.1), long-term survival (Section 1.2.2), and costs of care
(Section 1.2.3).
1.2.1 Short-term Mortality
The short-term mortality following gastrectomy for patients diagnosed with gastric
cancer ranges from 0.6 to 5.2% (33, 40-42). In the United Kingdom, where centralization of
esophagogastric cancer surgery has occurred, the National Oesophago-Gastric Cancer Audit,
evaluating the care delivered by all English hospitals performing esophagogastric cancer surgery
was interrogated for patients treated with curative surgery between April 1, 2011 and March 31,
2013. From these data, the overall 30-day mortality was found to be 2.3%, and the overall 90-day
mortality to be 4.4% (40). From the nationally maintained American College of Surgeons
National Surgical Quality Improvement Program (ACS-NSQIP) database, patients treated with
gastrectomy between 2011 and 2013 for both benign and malignant pathologies were found to
have a 30-day mortality rate of 5.2% (41). In a population-based cohort study of resected, non-
metastatic gastric cancer patients diagnosed between April 1, 2005 and March 31, 2008 treated in
the province of Ontario, the in-hospital mortality was 5.1% (33).
1.2.2 Long-term Survival
Despite the variety of treatment options available to patients with gastric cancer, in low-
incidence countries, the overall prognosis following diagnosis with the disease is poor (8).
Survival varies across disease stage and geography. High-incident Asian countries have reported
improved survival compared to low-incident North American countries. In Japan, five-year
survival has been described as ranging from 90.2% in Stage I disease, 75.2% in Stage IIA
disease, 59.3% in Stage IIB disease, 43.4% in Stage III disease, to 14.1% in Stage IV disease
(26). Five-year survival in the United States from the National Cancer Database (NCDB) have
been described as ranging from 56.7% for Stage I disease, 47.3% in Stage IIA disease, 33.1% in
Stage IIB disease, 25.9% in Stage III disease, to 5% in Stage IV disease (26). In non-metastatic,
11
resected gastric cancer patients in Ontario, five-year overall survival was 71% for patients with
Stage I disease, 51% for patients with Stage II disease, and 27% for patients with Stage III
disease (33).
Disease-level, patient-level, provider-level, and healthcare system-level factors are
implicated in the determination of outcome for patients with cancer (43). Even within a
universal, publicly-funded healthcare system, variation in practice patterns, and outcomes has
been described (33). Identifying modifiable factors in gastric cancer care is fundamental to
optimizing outcomes for patients with gastric cancer.
1.2.3 Costs of Care
1.2.3.1 Introduction to Health Economics
Health economics is the study of efficiency and effectiveness of health procedures,
services, or programs, as well as the optimization of available resource allocation within the
healthcare sector (44). The study of health economics includes the monetary valuation of direct
or indirect resource utilization related to medical procedures, services, or programs in a given
timeframe, from an assumed evaluation viewpoint or perspective (e.g. patient, institution,
healthcare system, societal) (44). The field of health economics encompasses the identification,
measurement, and valuation of costs and consequences through economic evaluation—the
comparative analysis of both costs and consequences of alternative courses of action (44). Within
a climate of escalating demands on a healthcare system with increasingly scarce resources, health
economics and economic evaluation have been sought for the provision of evidence-based
frameworks to assist in healthcare decision-making (45). Information regarding the economic
impact of medical programs, services, or programs may also be valuable for forecasting
healthcare system use, planning of resource allocation, and budgeting at the political (macro)
level (45).
At the level of the physician or patient, however, the role of health economics and
economic evaluation is controversial. While economic evaluation helps to identify alternatives
that are good value for money from a population standpoint, physicians are tasked with
providing optimal patient-centred care at an individual level. What may benefit one patient may
not align with the best interests of society as a whole, or with the remainder of patients seeking
12
care once resources have been committed to the first patient (45). An understanding of health
economics and economic evaluation, then, may help in overall decision-making by viewing
every decision made by a clinician for a patient as an opportunity cost in terms of what cannot be
provided for another patient. Furthermore, training in health economics would facilitate
appropriate and efficient health resource allocation within the current constraints of the
healthcare system, particularly as clinicians expand their roles beyond healthcare provision to
join decision-making bodies and committees involved in health policy creation (45).
1.2.3.2 Gastric Cancer Costs
Gastric cancer care is expensive and gastric cancer healthcare costs are generally higher
than that of other cancers (46). The costs of treating gastric cancer increase as the disease
progresses, with the highest costs of care associated with the treatment of patients with
metastatic gastric cancer (46). Similar to that of incidence and mortality, the economic burden
associated with gastric cancer care also varies across geographic regions with annual costs
ranging from 27 million in Iran to 8.5 billion United States (US) dollars in Japan (46). The
estimated annual cost of gastric cancer and gastroesophageal cancer care in 2017 for Europe
(France, Germany, Italy, Spain, United Kingdom), Asia (Iran, Japan, China), North America
(United States and Canada), and Australia was 20.6 billion US dollars (46).
In a Canadian study estimating the mean net costs for the 21 most prevalent cancers,
gastric cancer was identified as the third most expensive cancer to treat (7). Models predicting
the costs of gastric cancer have identified primary treatment strategy as the explanatory variable
accounting for the greatest amount of variance within the models, and surgery described as the
most expensive treatment strategy (47). Of the highest costs of care, inpatient hospital days are
the highest contributor to costs in the healthcare system (47, 48). Despite evidence suggesting the
high economic burden of gastric cancer care, particularly related to surgical care, health
economic literature providing detailed costing of gastric cancer phases of care or information on
selecting cost-effective gastric cancer treatment at the level of the population, or individual
patient, is sparse (49). Investigating the clinical and economic outcomes of gastric cancer care
collectively is essential to identifying potential areas for sustainable improvement in the care of
patients with gastric cancer.
13
1.2.4 Factors Associated with Gastric Cancer Clinical and Economic Outcomes
Various predictors of cancer outcomes have been identified. Factors associated with
cancer outcomes have previously been classified as tumour-related, host-related, and
environment-related in acknowledgement of the factors beyond the tumour itself that affect the
clinical outcomes of patients (43). The factors associated with cancer outcomes relevant to this
study include patient characteristics (age, sex, comorbidities, rurality, socioeconomic status),
disease characteristics (tumour location, burden of disease), and healthcare system characteristics
such as processes of care (50-54).
Age
Older age of gastric cancer patients has been reported to be independently associated with
mortality, worse overall survival and early recurrence (41, 50-52). Conversely, patients with
younger age experienced more aggressive disease and poorer outcomes compared to older
patients diagnosed with breast, as well as colorectal cancer (55, 56). Similarly, patients
diagnosed with gastric cancer at a young age had a higher incidence of advanced disease
compared to that of the general patient population (57)
Sex
Gastric cancer incidence and mortality is much higher in male patients than female
patients worldwide (1). Male sex has also been independently associated with worse overall
survival in gastric cancer (50). Furthermore, the absolute lifetime costs of gastric cancer care in
Ontario was higher for male patients than for females, but the third most expensive cancer to
treat for both sexes (7).
Comorbidity
In a joint report from the American Cancer Society, Centers for Disease Control and
Prevention, the National Cancer Institute, and the North American Association of Central Cancer
Registries, level of comorbidity was described as an important influence on the probability of
death for patients diagnosed with lung, colorectal, breast, or prostate cancer (58). Increased
comorbidity burden has also been associated with decreased participation in screening,
prolonged length of stay, and increased readmission in patients with breast, cervical, and
14
colorectal cancer (58, 59). Independent association between comorbidity and overall survival for
gastric cancer patients has also been reported with the presence of preoperative comorbidity
conferring a higher risk of death compared to those without preoperative comorbidity (60).
Rurality of Residence
Rurality presents a challenge in cancer care as patients living in rural areas may face
transportation barriers, financial issues, limited access to clinical trials, and reduced availability
of cancer treatments and providers (61). Overall, the impact of rurality on screening rates,
receipt of recommended treatment and follow-up, and cancer outcomes is unclear. (61).
However, survival rates of gastric cancer patients in rural versus urban China have been found to
be significantly different, and improvement in survival observed with regional economic
development (62). Improved survival of gastric cancer patients living in urban versus rural areas
has also been reported in Russia (63).
Socioeconomic Status
Socioeconomic status impacts cancer incidence and prognosis (64). High socioeconomic
status has been associated with higher risk for patients diagnosed with breast cancer and
malignant melanoma, whereas low socioeconomic status is associated with adverse prognosis in
lung, prostate, and ovarian cancers (64). Specific to gastric cancer patients, high individual
socioeconomic status is associated with a lower risk of mortality (65, 66).
Tumour Location
Tumour location has been associated with differences in overall survival with distal
tumours associated with the highest 5-year survival rate, of 50.7%, followed by corpus or fundus
tumours at 28.6%, and lastly, tumours in the proximal stomach, with a 5-year survival rate of
21.4% (p
15
Tumour Burden
Increased tumour burden necessitating resection of adjacent organs has also been
identified as independently associated with overall survival, and an increased risk of death
compared to no resection of adjacent organs is described (50).
Processes of Care
Processes of care involve specific steps leading to a particular outcome metric, and have
become a target for quality improvement in healthcare delivery (68, 69). Processes, with
structures and outcomes, also comprise the Donabedian model of healthcare quality
conceptualization to understand potential sources of perioperative variation, and to outline
modifiable sources of risk for surgical quality optimization (69). Processes of care have been
emphasized as excess variation in care has been identified as a potential source of medical errors,
and standardization of care delivery processes as a solution for improved healthcare outcomes
and quality (69). In the treatment of patients with breast and colorectal cancer, processes of care,
including timing of diagnosis, sequence of treatment, and appropriateness of treatment have been
identified as targets for care quality improvement (68, 70).
Gastric cancer outcomes are also influenced by processes of care such as preoperative
evaluation, resection criteria, operative techniques, referral patterns, and administration of
adjuncts to surgical treatment (4). While the relationship between genetics, surgical techniques,
and stage migration and worse survival particular to Western gastric cancer patients has been
broadly studied, studies examining the processes of care for gastric cancer are sparse (4). In a
study describing the processes of gastric cancer care associated with improved survival in a
population of gastric cancer patients diagnosed in Ontario between April 1, 2000 and March 31,
2005, age, surgical type, and resection in an institution with higher volumes was found to be
associated with survival (4). Furthermore, an international RAND/UCLA Expert Panel created
tenets for optimal gastric cancer management, including processes of gastric cancer surgical care
such as the performance of gastric cancer surgery by a surgeon experienced in gastric cancer
management with an annual volume of greater than 6 cases per year, and at a centre with
sufficient support to prevent or manage complications of gastric cancer surgery with an
institutional volume of greater than 15 gastric cancer resections per year (5). In the Netherlands,
16
where gastric cancer care has been centralized, processes of care, including centralization, and
clinical audits were identified as potential methods of gastric cancer outcome improvement (71).
1.3 Cancer Surgery Centre Designation
1.3.1 Introduction to Cancer Surgery Centre Designation
Regionalization of surgical healthcare delivery, establishment of surgical centres of
excellence, or cancer surgery centre designation (CSCD), has been described as a specific
process of care that improves patient outcomes, including morbidity and mortality (72, 73).
Through concentration of expertise and relevant resources centred on a specific area of medicine,
care is anticipated to be delivered in a comprehensive and interdisciplinary fashion to afford the
best possible outcomes to patients (74). High procedure volumes, greater access to
multidisciplinary consultation and tumour boards, computerized physician order entry, high
nursing-to-patient ratios, earlier adoption of newer, beneficial therapies, increased use of
standard treatment protocols and clinical pathways, and less practice variation are proposed as
reasons for improved outcomes to patients receiving treatment at these specialized centres (72,
75).
Minimum operative volume standards have been the hallmark of regionalization or
establishment of designated specialty surgery centres (76). Higher volumes of select procedures
are suggested to enhance proficiency in surgeons and team-based familiarity in the healthcare
system, and in turn, improve patient outcomes (77, 78). Another theory to explain the volume-
outcome relationship is the tendency of patients to be referred to institutions already seen to
provide high quality care, thereby increasing the case volume of the institution (78). A third
hypothesis to describe the causal model of the volume-outcome relationship is the concept of an
ecosystem, in which system characteristics of institutions, such as technology, staffing, and
expertise in other surgical procedures may equip institutions to optimize the healthcare provided
to patients and effect improved healthcare outcomes (78).
Specific to surgical oncology patients, higher numbers of specialist staff and higher
procedural volume have been associated with reduced perioperative mortality rates (79-82). In
particular, volume-outcome relationships exhibiting high volumes resulting in improved
outcomes are strong for rare, high-risk surgical conditions, such as pancreatic cancer, and weaker
17
for common conditions, such as bariatric surgery for obesity (75). Rare and complex diseases,
through disparate and dispersed treatment, have been found to have variation in treatment that
may affect outcomes (33). To improve care and outcomes to these patients of low-incident
disease, regionalization to specialty centres has been proposed and implemented (83). For
example, in the Netherlands, Lemmens et al. described increased resection rates, reduced
postoperative mortality, and improved survival in the post-centralization period compared to the
pre-centralization period for pancreatic cancer surgery (84). As well, Wouters et al. described
reduced postoperative morbidity, shorter length of stay, and improved mortality following
centralization compared to pre-centralization for patients undergoing esophagectomy (84, 85).
The costs of care at specialized centres have also been evaluated. Higher costs of care
have been attributed to centres designated as teaching hospitals (86). The higher care costs may
be owing to more complex case mixes, the location of these teaching institutions in urban areas,
where labour and overhead costs are higher, costs associated with the provision of graduate
medical education, and the use of more sophisticated forms of technology (86). However, a study
evaluating hospital volume and clinical as well as economic outcomes following complex
gastrointestinal surgical procedures suggested an inverse relationship between hospital volume
and in-hospital death and lower costs associated with high-volume centres (87). A similar study
aiming to investigate the effect of regionalization and cost specific to
pancreaticoduodenectomies revealed high-volume regional centres were associated with lower
total hospital charges (88). However, while the costs specific to particular procedures may
decrease as the procedural volume increases, the overall costs will increase at tertiary centres for
the delivery of high-cost specialty care (88).
1.3.2 Perspectives Regarding CSCD
Movements towards volume-based surgical performance, however, have been met with
controversy. Examples include advocacy for volume-based referral by Leapfrog Group, a
consortium of large corporations and public agencies involved in purchasing healthcare, and the
“Take the Volume Pledge” by Dartmouth-Hitchcock Medical Center, Johns Hopkins Hospital
and Health System, and the University of Michigan Health System, asking healthcare systems to
join in restricting the performance of specific surgical procedures to hospitals and surgeons with
minimum volume standards (76, 89). This approach to restricting surgical care to high-volume
18
hospitals aimed to have the performance of complex elective surgery conducted by providers and
hospitals performing the highest numbers of the complex elective surgeries. However, the
question of performance of complex surgery in small and rural hospitals has not been answered,
and low-volume surgeons were offended by their portrayal as hobbyists fueled by pride and
professional autonomy to continue performing complex and rare procedures irrespective of
clinical and economic consequences (89). Furthermore, patient preferences for local care,
financial incentives for smaller hospitals to retain surgical cases, and lack of access to high-
volume centres in some regions are barriers to regionalization (76). Increased travel demands,
and lack of support from nearby family and friends, are other possible disadvantages of
centralization that could lead to decreased access to care (90).
1.3.3 Cancer Surgery Centre Designation in Ontario, Canada
In Ontario, CSCD has been assigned to hospitals meeting safety and quality standards for
the delivery of specific procedures or particular types of care, including hepato-pancreatic
biliary, thoracic, head and neck, and gynecological cancers (91). The establishment of CSCD
mandates the fulfillment of surgeon and hospital criteria (e.g. knowledge, training, minimum
number of surgeons) as well as physical (e.g. operating room, diagnostic modalities), system
(e.g. multidisciplinary cancer conferences, regional cancer program participation) and human
(e.g. surgeon, radiology, pathology) resources with guidelines for quality assurance measures to
be met (91).
In Canada, regionalization of both esophageal and pancreas surgery based solely on
hospital volumes has not had a clear relationship to postoperative outcomes. An investigation of
the effect of hospital volume on in-hospital mortality for patients undergoing esophagectomy in
Canada revealed no significant reduction in in-hospital mortality with increasing hospital
volume, suggesting hospital volumes alone did not explain the improved outcomes observed at
higher-volume institutions (92). Another study investigating the volume-outcome hypothesis in
patients undergoing pancreas surgery in Ontario and Quebec revealed improved operative
mortality in Ontario, but not in Quebec, despite similar increases in annual pancreas surgery
volume over the study period. The results of this study do not support the volume-outcome
hypothesis, suggesting more complexity to the relationship between volume and outcome than
simply increased surgeon experience or selective referral to providers perceived to have better
19
outcomes resulting in improved outcomes (93). These studies suggested regionalization may lead
to better outcomes through mechanisms other than increased volume, including optimization of
the use of limited resources, increased clinical trial accrual, or regional quality improvement
program facilitation (93).
Although pancreatic and esophageal cancers have similar morbidity and mortality to that
of gastric cancer with similarly challenging resections necessary to effect cure, no
regionalization of gastric cancer has been established in North America, including Canada.
1.4 Ontario Healthcare System
Healthcare in Canada is provided through a publicly funded system with a mandate to
maintain and improve the health of Canadians. The Canada Health Act governs this mandate
through its five pillars of public administration, comprehensiveness, universality, portability, and
accessibility (94). The provincial and territorial health systems, supported by the federal
government, have the responsibility and accountability for the organization, management,
delivery, quality, safety, and performance for acute, chronic, long-term, and palliative, end-of-
life needs of their respective residents (94). The provincial and territorial governments in turn
distribute the provision and planning of health services to regional governance structures such as
health authorities and local health integration networks (94).
In Ontario, the Ministry of Health and Long-Term Care is tasked with medically
necessary healthcare provision to Ontario residents through the Ontario Health Insurance Plan
(95). The Ontario government has a principal cancer advisor in Cancer Care Ontario. Through
accountability, data, planning, and performance, Cancer Care Ontario works towards cancer
prevention and cancer care delivery (91).
1.5 Summary of Evidence and Rationale
In North America, gastric cancer care places a relatively substantial clinical and
economic burden on the healthcare system (7, 8, 33). Curative-intent gastrectomy with
corresponding lymph node dissection is the cornerstone of gastric cancer therapy, but is
technically challenging and complex, and is associated with considerable morbidity, mortality,
and cost (33, 47). While diseases with similar levels of surgical care complexity, including
esophageal and pancreatic cancer, have had care regionalized to institutions with CSCD, the
20
same has not been implemented for gastric cancer surgical care. Gastric cancer care has also
been delivered with significantly variability in care patterns (33). Currently established
specialized centres, however, through standardized processes of care, concentration of relevant
expertise and resources, and increased access to newer, beneficial therapies, may effect improved
outcomes on similar non-regionalized procedures (72, 74).
Despite evidence suggesting gastric cancer care is very expensive, a paucity of literature
exists regarding the economic burden of gastric cancer care (7, 49). While regionalization of care
has been suggested in the literature as a method of improving outcomes to patients, it may result
in increased costs to the healthcare system for an already costly disease (88).
This thesis therefore compares the clinical and economic outcomes of curatively resected
gastric cancer patients between those treated at institutions with and without CSCD with a view
to evaluate whether CSCD is an organizational system that can result in improved short-term
mortality and overall survival, and at how much more cost to the healthcare system compared to
institutions without CSCD.
21
Chapter 2
Methods 22.1 Study Objectives
In a population-based, retrospective cohort study of resected, non-metastatic gastric cancer
patients in Ontario, the study objectives were to
1. Compare 90-day mortality between patients who underwent gastrectomy at an institution
with CSCD and patients who underwent gastrectomy at an institution without CSCD.
2. Compare overall survival between patients who underwent gastrectomy at an institution
with CSCD and patients who underwent gastrectomy at an institution without CSCD.
3. Compare costs of care between patients who underwent gastrectomy at an institution with
CSCD and patients who underwent gastrectomy at an institution without CSCD.
2.2 Hypotheses
The hypotheses of this thesis include:
1. 90-day mortality is lower for patients treated with gastrectomy at institutions with CSCD
compared to patients treated with gastrectomy at institutions without CSCD.
2. Overall survival is improved for patients treated with gastrectomy at institutions with
CSCD compared with patients treated with gastrectomy without CSCD.
3. Costs of care are higher for patients treated with gastrectomy at institutions with CSCD
compared to patients treated with gastrectomy at institutions without CSCD.
2.3 Study Design
This study followed a population-based, observational, retrospective cohort study design.
The design was chosen to describe the clinical (90-day mortality and overall survival) and
economic (costs of care) outcomes of patients diagnosed with gastric cancer in Ontario and to
22
compare the outcomes of patients who underwent gastrectomy at centres with cancer surgery
centre designation (CSCD) to patients who underwent gastrectomy at centres without CSCD.
2.4 Study Population
This study included a target population of individuals diagnosed and treated with curative-
intent gastrectomy for gastric adenocarcinoma in Ontario. Patients with a diagnosis of gastric
adenocarcinoma in the Ontario Cancer Registry (OCR) between January 1, 2002 and December
31, 2014 were included.
2.4.1 Inclusion Criteria
Patients aged 18 years or greater with a valid Ontario Health Insurance Plan (OHIP) number,
a valid Institute for Clinical Evaluative Sciences (ICES) key number (IKN) traceable in the
Registered Persons Database (RPDB), a confirmed gastric adenocarcinoma diagnosis within the
study period identified using the International Classification of Diseases-10 (ICD-10) codes
(C16.x) (Supplemental Table 1), an ICD for Oncology-3 (ICD-O-3) histology code for
adenocarcinoma (Supplemental Table 2), treated with a gastrectomy within a year of the date of
diagnosis, and those with non-metastatic disease were included.
Complete and accurate stage data for gastric cancer was not available in the Ontario Cancer
Registry. Metastatic disease was therefore identified using the algorithm with the highest
accuracy (73%) and specificity (82%) from a series of algorithms created using administrative
healthcare data captured in the peri-diagnostic period for identification of metastatic gastric
cancer in the absence of population-based stage data, and validated using American Joint
Committee of Cancer (AJCC) 7th Edition TNM stage data from patient hospital charts (96). Data
from one or more hospitalization, emergency department visits, or outpatient physician visits
were collected from the Canadian Institute of Health Information-Discharge Abstract Database
and the Same Day Surgery Database, the National Ambulatory Care Reporting System, and the
OHIP database at ICES were collected, and identified metastatic disease from a conservative list
of ICD diagnosis codes in the six months before or after diagnosis (96). Patient hospital chart
information was collected in a province-wide chart review including data from operative,
endoscopy, radiology, and pathology reports, and were the reference standard for validation of
the algorithms (96).
23
2.4.2 Exclusion Criteria
Patients were deemed ineligible for inclusion and excluded from this study if they died
before the date of diagnosis, if their date of last contact was missing or before the date of
diagnosis, were missing an institution number associated with a gastrectomy, had a tumour
location in the esophagus (ICD-10 codes C15.x), had previous or multiple primary cancers, or
had metastatic disease.
2.4.3 Restrictions
The cost analysis was limited to those patients diagnosed before March 31, 2013, as they did
not have the opportunity to accrue costs for the entirety of the time horizon selected.
2.5 Study Perspective and Timeframe
2.5.1 90-day Mortality and Overall Survival
The study timeframe for the analyses of 90-day mortality and overall survival are outlined in
Figure 1. For the comparison of 90-day mortality between patients treated at centres with and
without CSCD, the index event date was the date of surgery. For the comparison of overall
survival between patients treated at centres with and without CSCD, the index event date was the
date of diagnosis.
24
Figure 1. Outline of study timeframe for analyses of 90-day mortality and overall survival
in patients treated with gastrectomy in Ontario
2.5.2 Costs of Care
The study perspective for the analysis of costs of care was the perspective of the Ministry of
Health and Long-Term Care (MOHLTC), or the healthcare system payer. The costing
perspective chosen for an economic evaluation should be directly related to the decision
problem, and the costs and outcomes included in the analysis defined by the selected costing
perspective for inclusion in the study (97). The healthcare system payer perspective is often
chosen as it accounts for direct medical costs for their relevance to health policy planners and
healthcare system decision-makers (98). Within a publicly funded healthcare system such as that
of Canada, relevant sources of costs include drugs, medical procedures, equipment, facilities,
overhead, healthcare provider fees, hospital services, investigations, rehabilitation, and
community-based services such as home care (97).
For economic evaluations, the study perspective and time horizon need to be relevant to the
research question and appropriate for the outcomes of an economic study (44, 97). The median
survival of gastric cancer patients in Ontario who have undergone gastrectomy for non-
metastatic disease is 32 months (IQR 25-38) (33). The reported median time to recurrence for
gastric cancer patients in the United States who have undergone curative-intent gastrectomy is 28
months (IQR 23-36). Therefore, a time horizon of 38 months—2 months prior to date of
Look-back Window for
Comorbidity Data Collection :
2 years
Observation Window:
January 1, 2002-June 30, 2015
Index Event Date
Accrual Window: January 1,
2002-December 31, 2014
Max Follow-up Date: March 31,
2017
25
diagnosis and up to 3 years following date of diagnosis was chosen to accommodate costs related
to diagnostic workup, treatment, and end-of-life care (49, 99). The outline of the study timeframe
for the cost analysis is outlined in Figure 2.
Figure 2. Outline of study timeframe for cost analysis of patients treated with gastrectomy
in Ontario
2.6 Study Setting
This research took place at the ICES Central research facility. Conduct of the study at this
facility uses the housed administrative databases facilitated the study of all patients with a
confirmed diagnosis of gastric cancer in Ontario within the study period.
The data holdings at ICES uniquely positions researchers with the advantages of a
repository of all routinely collected, publicly funded healthcare administrative data within a
single-payer universal healthcare system. Furthermore, a population-based cohort study could be
conducted with a large sample size, and the provision of statistical power. This type of study will
realistically reflect the outcomes of a heterogeneous group of patients as they are treated across
multiple institutions, with variation, and with minimal loss to follow-up, whereas controlled
clinical trials reflect the outcomes of a homogeneous group of patients as they are treated using
strict guidelines and protocols. Moreover, the low incidence of gastric cancer in Canada renders
randomized controlled trials unfeasible.
Look-back Window for
Comorbidity Data Collection
2 years
Observation Window:
Index Date to (Index Date + 1095 days)
Index Event Date
Accrual Window: January
1, 2002-March 31, 2013
Max Follow-up Date: March
31, 2016
26
2.7 Data Sources
Administrative healthcare databases are held at ICES and linked by individual patients’
IKNs. The following databases were accessed for this project.
2.7.1 Ontario Cancer Registry
The Ontario Cancer Registry database is updated annually and is a source of information on
the burden of cancer in the province (91). This database was utilized to derive the study cohort
through identification of patients with gastric cancer registered within the database and their
respective diagnosis dates. Although stage capture rates for breast, prostate, colorectal, lung and
cervical cancer are greater than 90%, the stage capture rate for all cases of cancer diagnosed is
57% (100).
2.7.2 Canadian Institute for Health Information-Discharge Abstract Database and Same Day Surgery
Data from the Canadian Institute for Health Information (CIHI)-Discharge Abstract Database
(DAD) and Same Day Surgery (SDS) database are updated on an annual basis. They are a source
of national patient- and facility-level data on all acute, chronic, rehabilitation, and same day
surgery facilities in Canada (101). Data from CIHI-DAD & SDS were utilized to identify
patients who underwent gastrectomy, the institutions at which they underwent gastrectomy, and
the costs associated with inpatient hospital care. The Johns Hopkins Resource Utilization Band
for each patient to provide a metric for comorbidity was also derived using this database.
2.7.3 National Ambulatory Care Reporting System
The National Ambulatory Care Reporting System (NACRS) is a database that is updated
annually and is a source for patient and facility-level information on emergency department
visits, day surgery, dialysis, and outpatient clinic visits including cancer visits (101). NACRS
was used in the derivation of comorbidity measures and in the algorithm to identify patients with
metastatic disease and to derive costs for dialysis, emergency department visits, and cancer
visits.
27
2.7.4 Home Care Database
The Home Care Database (HCD) is updated on a yearly basis and is a source of community-
based in-home services provided to Ontario residents (101). The services provided are
coordinated by Community Care Access Centres and include services such as personal support
work, nursing, and social work. The HCD was used to identify costs associated with home care
provision.
2.7.5 Ontario Health Insurance Plan
The Ontario Health Insurance Plan (OHIP) Claims Database is a updated bi-monthly and
is a source of claims submitted by healthcare providers able to claim under OHIP (physicians
both in and out of province, and laboratories) for remuneration from the MOHLTC (101).
Information derived from the OHIP Claim Database included identification of patients who
underwent gastrectomy, and the institution at which patients underwent gastrectomy,
chemotherapy use, radiotherapy use, and costs associated with physician billings.
2.7.6 Registered Persons Database
The Registered Persons Database (RPDB) files are updated bi-monthly and are a source of
patient-level demographic data. The database is coordinated by the Ontario MOHLTC and
includes basic demographic information about all recipients of an Ontario health card number, as
well as geographic, contact, and death information from other ICES data holdings (101). The
RPDB was used to derive patient demographic information, the Local Health Integration
Network of residence, rurality of residence, death date, date of last contact with the healthcare
system, and aggregate-level data on socioeconomic status.
2.7.7 Activity Level Reporting
The cancer Activity Level Reporting (ALR) database is updated annually and is a source of
provincial patient-level data on radiotherapy and systemic therapy services (91). The ALR
database was used to derive information regarding patient receipt of radiotherapy.
2.8 Study Variables
The following sections describe all variables used to describe the curatively resected gastric
cancer population, including the primary exposure, covariates, and outcomes. Patient and disease
28
variables to describe the cohort and to compare patients treated at centres with and without
CSCD were identified a priori using literature, and based on clinical relevance. Variables were
limited to those available within the administrative healthcare data holdings at ICES. Variables
with their respective sources and definitions are outlined in Table 1.
Table 1. Sources and definitions of patient and disease characteristics
Variable Type Source Definition
Patient Characteristic
Age Categorical RPDB 74 years
Sex Categorical RPDB Female/Male
Johns Hopkins Resource Utilization Band
Categorical CIHI-DAD 0 or 1/2/3/4/5
Rurality of Residence
Categorical RPDB Rural/Urban
Income Quintile Categorical RPDB Lowest/2/3/4/Highest
Disease Characteristic
Type of Gastrectomy Categorical CIHI-DAD, OHIP
Distal/Proximal/Total/Unknown
Multivisceral Resection
Categorical CIHI-DAD, OHIP
Yes/No
Healthcare System Characteristics
CSCD Categorical CIHI-DAD, OHIP
Yes/No
Preoperative Chemotherapy
Categorical OHIP Yes/No
Preoperative Radiotherapy
Categorical OHIP, ALR Yes/No
Length of Stay Continuous CIHI-DAD Number of Days in Hospital
RPDB=Registered Persons Database; CIHI-DAD=Canadian Institute for Health
29
Information-Discharge Abstract Database; OHIP=Ontario Health Insurance Plan; ALR=Activity Level Reporting
2.8.1 Primary Exposure
The primary exposure of interest was gastrectomy at an institution with Cancer Surgery
Centre Designation (CSCD). CSCD was defined according to current status of CSCD as outlined
by Cancer Care Ontario (91). These institutions are hospitals that have received designation for
meeting specified safety and quality standards for hepato-pancreatic biliary (HPB), thoracic,
head and neck, and gynecological cancers. All types of CSCD were included as the required
standards include surgeon criteria such as subspecialty training, and hospital criteria such as
specialty-specific diagnostic and therapeutic resources, and volume, in addition to benchmarks
for quality assurance such as a ta