- Email: [email protected]
Prior gastroscopy and mortality in patients with gastric cancer: a matched retrospective cohort study
Accepted Manuscript Prior gastroscopy and mortality in patients with gastric cancer: a matched retrospective cohort study Wai K. Leung, MD, Hsiu J. Ho, PhD, Jaw-Town Lin, MD, Ming-Shiang Wu, MD, ChunYing Wu, MD PII:
S0016-5107(17)32036-9
DOI:
10.1016/j.gie.2017.06.013
Reference:
YMGE 10628
To appear in:
Gastrointestinal Endoscopy
Received Date: 7 March 2017 Accepted Date: 12 June 2017
Please cite this article as: Leung WK, Ho HJ, Lin J-T, Wu M-S, Wu C-Y, Prior gastroscopy and mortality in patients with gastric cancer: a matched retrospective cohort study, Gastrointestinal Endoscopy (2017), doi: 10.1016/j.gie.2017.06.013. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT Prior gastroscopy and mortality in patients with gastric cancer: a matched retrospective cohort study Short title: Prior gastroscopy and gastric cancer mortality
RI PT
Wai K Leung, MD1, Hsiu J. Ho, PhD2, Jaw-Town Lin, MD3-4, Ming-Shiang Wu, MD5, Chun-Ying Wu, MD2, 6-9*
Department of Medicine, University of Hong Kong, Hong Kong, China;
2
Division of Gastroenterology, Taichung Veterans General Hospital, Taichung, Taiwan;
3
School of Medicine, Fu Jen Catholic University, Taipei, Taiwan;
4
Institute of Population Health Sciences, National Health Research Institutes, Miaoli, Taiwan;
6
Division of Gastroenterology, National Taiwan University Hospital, Taipei, Taiwan; Faculty of Medicine and Graduate Institute of Clinical Medicine, National Yang-Ming
TE D
University, Taipei, Taiwan; 7
M AN U
5
SC
1
Department of Public Health and Graduate Institute of Clinical Medical Sciences, China
Medical University, Taichung, Taiwan;
National Institute of Cancer Research, National Health Research Institutes, Miaoli, Taiwan;
9
Department of Life Sciences and Rong Hsing Research Center for Translational Medicine,
EP
8
AC C
National Chung-Hsing University, Taichung, Taiwan * Correspondence:
Professor Chun-Ying Wu, MD, PhD, MPH Division of Gastroenterology, Taichung Veterans General Hospital No. 1650, Section 4, Taiwan Avenue, Taichung 407, Taiwan E-mail: [email protected]; Tel: +886-4-23592525 # 3304; Fax: +886-4-23741331
1
ACCEPTED MANUSCRIPT ABSTRACT Background & Aims: The role of prior gastroscopy on outcome of gastric cancer patients remains unknown. This study determines the association between intervals of prior
RI PT
gastroscopy and mortality in patients with gastric cancer. Methods: We identified 20,066 newly diagnosed gastric cancer patients in the national health insurance database of Taiwan between 2002 and 2007. After excluding gastroscopy
SC
performed 6 months before diagnosis of cancer, patients were matched into 3 cohorts
M AN U
according to the intervals of prior gastroscopy: 6 months to 2 years (<2Y cohort), 2 to 5 years (2-5Y cohort), and none within recent 5 years (>5Y cohort). The 3 cohorts were matched for age, curative treatment for gastric cancer, Helicobacter pylori therapy and propensity scores that comprised of gender, comorbidities and concomitant medication usage. The primary
TE D
outcome is the hazard ratio (HR) of all-cause mortality.
Results: After matching, we identified 1,286, 1,286, and 5,144 patients for the <2Y, 2 to 5Y and >5Y cohorts. Compared with the >5Y cohort, the HR of all-cause mortality for the <2Y
EP
and 2 to 5Y cohorts was 0.80 (95% CI, 0.72-0.89; P<0.001) and 0.83 (95% CI, 0.76-0.91;
AC C
P<0.001), respectively. The HRs of gastric cancer-specific mortality was also significantly lower in the <2Y (0.80; 95% CI, 0.71-0.91; P<0.001) and 2 to 5Y cohorts (0.83; 95% CI, 0.75-0.93; P<0.001).
Conclusion: Gastric cancer patients who had gastroscopy performed within 5-year before cancer diagnosis have significantly lower mortality. Our results may support the role of repeat endoscopic examination or surveillance endoscopy in selected patients.
2
ACCEPTED MANUSCRIPT INTRODUCTION Gastric cancer is the third leading cause of cancer death in the world.1 In 2012, there were approximately 950,000 new cases and more than 720,000 deaths from gastric cancer globally.
RI PT
Although the incidences of gastric cancer are generally declining in most countries, it is still having significant impact on many developing countries where 70% of gastric cancer cases occur. In particular, about 50% of gastric cancer cases occur in East Asian countries including
SC
China, Japan and Korea.2,3
M AN U
Whilst screening gastroscopy is rarely performed in most countries except in a few East Asian countries4, majority of gastroscopy are performed in clinical practices as a diagnostic work-up for upper gastrointestinal symptoms. Moreover, it is used for monitoring of gastric ulcer healing and surveillance of pre-neoplastic gastric lesions. Unlike surveillance
TE D
colonoscopy, the optimal interval of repeating gastroscopy in diverse patient groups remains uncertain. As yet, the real impact of routine daily gastroscopy on subsequent outcome of gastric cancer patients has not been properly addressed. In a recent Korean study, it was
EP
shown that gastric cancer patients who had upper endoscopy performed within 3 years or less
AC C
before the diagnosis of cancer had a significantly lower cancer staging as compared with those who never had endoscopy.6 However, the most important clinical outcome, ie, actual reduction in mortality, had not been demonstrated. Given the possible benefits of repeated endoscopic examinations, there is a significant knowledge gap on the role of repeated endoscopic examination on the actual outcome of gastric cancer patients. The current study was a retrospective cohort study of gastric cancer patients with propensity score matching based on a large population-based health insurance database from Taiwan. We 3
ACCEPTED MANUSCRIPT determined the hazard ratio of all-cause mortality in patients who had gastroscopy performed at different intervals before the diagnosis of gastric cancer. Unlike previous studies that focused on screening gastroscopy, we evaluated the role of routine gastroscopies performed
RI PT
for various indications in patients who were subsequently confirmed to have gastric cancer. METHODS
SC
Data Source:
This was a nationwide matched retrospective cohort study based on the Taiwan’s National
M AN U
Health Insurance Research Database (NHIRD). The NHIRD is a compulsory national health insurance program, which covers all 23.7 million Taiwan’s residents. The NHIRD includes comprehensive health care information and the details have been described in our previous studies.7-9 In short, patients’ characteristics, diagnoses, drugs, procedures, and laboratory
TE D
examination items were all available in NHIRD. The International Classification of Disease-9 (ICD-9) codes were used to classify diseases in the database and the accuracy of diagnosis of several major diseases, such as ischemic stroke and acute coronary syndrome, have been
AC C
of Taiwan.
EP
previously validated.10, 11 This study was approved by the National Health Research Institutes
Study Subjects:
We identified all patients who were newly diagnosed to have primary adenocarcinoma of stomach from the NHIRD based on the ICD-9 code (ICD-9 code: 151) between January 1, 2002 and December 31, 2007. To verify the diagnosis, these patients should also be enrolled in the Registry for Catastrophic Illness Patient Database (RCIPD) at the same time. The
4
ACCEPTED MANUSCRIPT RCIPD is a subset of the NHIRD, which requires histological diagnosis based on endoscopic biopsy or surgical specimens for registration as cancer. We excluded patients who had incomplete demographic information or those with other malignancies (ICD codes: 140-150;
RI PT
152-208) before the diagnosis of gastric cancer. We defined the index date as the date of gastric cancer diagnosis, which was the first date of hospitalization for work up and
management of gastric cancer or the first date of enrollment to the RCIPD. Due to the
SC
insurance reimbursement issue, delay in registration in the RCIPD is uncommon.
M AN U
We retrieved the number and date of gastroscopy, performed for any indications, in these patients before the diagnosis of gastric cancer. We excluded gastroscopy performed within 6 months before the diagnosis of gastric cancer as these endoscopies might be used for diagnosis of cancer. The timing of the most recent endoscopy performed before the index
TE D
date was used to classify patients into 3 cohorts: between 6 months and 2 years before the index date (<2Y cohort), between 2 and 5 years (2-5Y cohort) and no endoscopy within recent 5 years (>5Y cohort).
EP
Matching of the 3 Cohorts:
AC C
The baseline data of the 3 cohorts were reported in Supplementary Table 1. To control for possible confounding factors, the 3 groups were matched for age, curative surgery for gastric cancer, Helicobacter pylori therapy, and propensity scores which was calculated on gender, comorbid illnesses and concomitant medications. Patients’ characteristics were defined a priori and analyzed. Because gastric cancer staging data was not available in the insurance database, curative surgery for gastric cancer (total or subtotal gastrectomy with lymph node dissection) and use of chemotherapy were used as surrogate parameters for cancer staging. 5
ACCEPTED MANUSCRIPT The following co-morbidities were included in the propensity score analysis: history of peptic ulcer disease (PUD; ICD codes: 531-533), liver cirrhosis (ICD-9 code: 571⋅5), acute coronary syndrome (ACS; ICD-9 codes: 410-414), cerebrovascular accident (CVA; ICD-9 codes:
RI PT
430-438), chronic obstructive pulmonary diseases (COPD; ICD-9 codes: 490-496); diabetes mellitus (ICD-9 codes: 205), renal failure (ICD-9 codes: 584-586) and hypertension (ICD-9 codes: 401-405). We also analyzed drugs that may be associated with mortality including H
SC
pylori eradication therapy (regimens details described in our previous studies)8, aspirin,
non-steroidal anti-inflammatory drugs (NSAIDs) including COX-2 inhibitors,12 statins and
M AN U
metformin. Concomitant medications were defined as patients who had used these agents for more than one tablet per month before the diagnosis of gastric cancer. H pylori eradication was defined as whether the patient had ever received H pylori eradication treatment. These regimes were then further classified into pre-cancer and post-cancer diagnosis according to
TE D
the timing of H pylori eradication received before or after gastric cancer diagnosis, respectively. A propensity score for each patient was calculated on the logistic regression
EP
model of gender, comorbidities and concomitant medications that was used to estimate the probability of classifying these patients into 3 cohorts. We matched the final cohorts in a
AC C
1:1:4 ratios for the <2Y, 2 to 5Y, and >5Y cohorts (Table 1). Outcome Measures:
The primary outcome of this study is the hazard ratio (HR) of all-cause mortality whereas the secondary outcome is the HR of gastric cancer-specific mortality. All 3 cohorts were followed up from the date of cancer diagnosis to the date of death or the end of 2012. Death was identified from withdrawal from the compulsory National Health Insurance program. Gastric
6
ACCEPTED MANUSCRIPT cancer-specific mortality was defined when the patients had a primary diagnosis of gastric cancer during the last hospitalization before death. Death from other causes before gastric cancer-specific mortality were treated as competing mortality by Gray methods to calculate
RI PT
the modified Cox proportional hazards ratio.
Statistical Analysis
SC
The baseline demographic, comorbidities and concomitant medication uses of the 3 cohorts were compared by the χ2 test or the Student t-test. We matched these 3 cohorts by age (± 2
M AN U
years), curative therapy of gastric cancer, H pylori therapy, and propensity score (± 0.01). Propensity score was calculated by multinomial logistic regression based on the following covariates: gender, liver cirrhosis, acute coronary syndrome, cerebrovascular accident, chronic obstructive pulmonary disease, diabetes mellitus, renal failure, hypertension,
TE D
hyperlipidemia, NSAIDs/aspirin/COX2, statins, and metformin use status. We select the candidate who has the nearest character, but not more than a specific threshold list in
EP
parentheses for continuous variables, age and propensity score.
AC C
The cumulative incidences and HRs of all-cause and gastric cancer-specific mortalities of the 3 groups were compared. The cumulative incidences of all-cause mortality were constructed by Kaplan-Meier method and the cumulative incidences of cancer-specific mortality by Gray method, which is a modified Kaplan-Meier method for competing risk. These 2 methods and comparisons among 3 groups were implemented by the cmprsk package in R software. We used Cox proportional hazards model to determine whether endoscopy interval is independently associated with mortality. All parameters were defined a priori and used in the prediction model to adjust for as many confounders as possible. The assumption of a Cox 7
ACCEPTED MANUSCRIPT proportional hazards model was tested using the method of Grambsch et al13 and was implemented using the functional cox.zph in R package “survival.”
RI PT
RESULTS Patients’ Characteristics
Between January 1, 2002 and December 31, 2007, a total of 20,066 patients were newly
SC
diagnosed to have gastric cancer and fulfilled our inclusion criteria (Figure 1). Among them, 3,303 and 1,785 patients had received upper endoscopy between 6 months to 2 years (<2Y
M AN U
cohort) and between 2 and 5 years (2-5Y cohort) before cancer diagnosis, respectively. The remaining 14,978 patients had not received upper endoscopy in the recent 5 years (>5Y cohort; Supplementary Table 1). There were significant differences in the baseline characteristics among the 3 groups as shown in Supplementary Table 1. After matching, there
TE D
were 1,286, 1,286, and 5,144 patients from the <2Y, 2 to 5Y, and >5Y cohorts, respectively. The 3 cohorts were matched in baseline characteristics and comorbid illnesses except for the
EP
number of endoscopy in recent 5 years, follow-up duration, mortality rates, uses of
AC C
chemotherapy or statins and number of prior hospitalizations (Table 1).
Mortalities of Gastric Cancer Patients Both the <2Y and 2 to 5Y cohorts had significantly lower all-cause as well as gastric cancer-related mortality when compared with the >5Y group (P<0.001; Figures 2A and 2B). However, there was no difference in the non-cancer-related mortality among the 3 groups (Supplementary Figure 1).
8
ACCEPTED MANUSCRIPT Multivariable Cox Proportional Hazards Model Analysis After adjusting for various potential confounding factors, we found that prior gastroscopy is independently associated with a reduced risk of all-cause mortality (Table 2). Compared with
RI PT
the >5Y cohort, the hazards ratios (HRs) of all-cause mortality for the <2Y and 2 to 5Y
cohorts were 0.80 (95% CI, 0.72 to 0.89; P<0.001) and 0.83 (95% CI, 0.76 to 0.91; P<0.001), respectively. The number of endoscopy performed in the recent 5-year was also found to be
SC
significantly associated with a lower all-cause mortality (HR 0.95; 95% CI, 0.92 to 0.99;
model was satisfied (P=0.07).
M AN U
P=0.019). The assumption of our Cox proportional hazard model has been tested and the
Table 3 showed the multivariable Cox proportional hazards model analysis of the gastric cancer-specific mortality. Compared with the >5Y cohort, the HR for cancer-specific
TE D
mortality of the <2Y and 2 to 5Y cohorts was 0.80 (95% CI, 0.71 - 0.91; P<0.001) and 0.83 (95% CI, 0.75 to 0.93; P<0.001), respectively.
EP
Other factors that were found to be significantly associated with both lower all-cause and cancer-specific mortalities included younger patient’s age, curative surgery for gastric cancer,
AC C
no need for chemotherapy, hypertension, H pylori eradication therapy and statins use (Tables 2 and 3).
Sensitivity Analyses
As patients’ age is an important factor associated with mortality, we conducted sensitivity analyses by including patients younger than 80 years only. We found that the <2Y and 2 to 5Y cohorts still had lower mortality when compared with the >5Y group in both younger
9
ACCEPTED MANUSCRIPT (<65 years) and older (65-80 years) age groups (Supplementary Figure 2; both P<0.001). Next, we performed sensitivity analysis to determine the association of prior gastroscopy interval of longer than 5-year interval and all-cause mortality of cancer patients. In this
RI PT
analysis, the >5Y cohort was further divided into 2 groups with one group having prior
endoscopy between 5 and 8 years (5-8Y cohort) and another group with no endoscopy in recent 8 years (>8Y cohort). Consistent with our above findings, gastric cancer patients who
SC
had gastroscopy between 5 and 8 years had a lower mortality than >8Y cohort, but a higher
M AN U
mortality than the <2Y and 2 to 5Y cohorts (Supplementary Figure 3; P<0.001). DISCUSSION
This population-based matched retrospective cohort study from Taiwan demonstrated that upper endoscopy performed for various indications within 5 years before gastric cancer
TE D
diagnosis is associated with a significantly lower all-cause as well as gastric cancer-specific mortalities when compared with those with no endoscopy in recent 5 years. The lower
EP
mortality remains significant even after adjusting for various potential confounding factors such as patient’s age, gender, treatment for gastric cancer, H pylori eradication, comorbid
AC C
illnesses and concomitant medication uses. It is so far the first population-based study that demonstrates the association between routine upper endoscopy, not screening endoscopy, and mortality in gastric cancer patients. The strengths of this study were that we used the national health care insurance database of Taiwan, which covers the whole 25 million populations of Taiwan, to avoid potential selection bias. The comprehensive information available from the health care database allow for control of many potential confounders. To avoid potential difference in baseline 10
ACCEPTED MANUSCRIPT characteristics or health-seeking behavior of the 3 cohorts, we matched the 3 cohorts according to their age, treatment for gastric cancer, H pylori eradication therapy, and propensity scores that included comorbid illnesses and concomitant medications. Moreover,
RI PT
our analysis included both all-cause as well as cancer-specific mortalities. To further support our findings, there was no difference in non-cancer-related mortality among the 3 cohorts, which indicate that the difference in all-cause mortality is mainly driven by difference in
SC
cancer-specific mortality. Third, we have used 2 different sensitivity analyses that looked into
(Supplementary Figures 2-3).
M AN U
the effects of age and longer endoscopy interval to support the robustness of our data
Unlike screening endoscopy14-15, it remains unknown whether upper endoscopy performed for various clinical indications before the cancer diagnosis is associated with lower mortality
TE D
in gastric cancer patients. Moreover, no study has demonstrated the actual reduction in mortality by routine gastroscopy. As the present study included upper endoscopy performed in all age groups regardless of their indications, these findings would be more applicable to
EP
real-life daily clinical practices when endoscopy is performed for diagnostic work up or
AC C
surveillance, rather than screening purpose. It is, however, important to highlight that the majority of our study population are ethnic Chinese (98%). As yet, the latest gastric cancer-related mortality in Taiwan is 12.2 and 7.0 per 100,000 for men and women, respectively, which is much lower than other East Asian countries. The use of population with intermediate cancer risk further increases the external validity of our observations to other non-high risk regions. There are 2 important questions relating to our findings. First, the mechanism underlying the
11
ACCEPTED MANUSCRIPT reduction in mortality by routine clinical gastroscopy remains elusive. Whilst gastric carcinogenesis is generally believed to be a slow process, the recent European Consensus recommended endoscopic surveillance in 3 years for patients with pre-neoplastic gastric
RI PT
lesions 16. We noted that a considerable proportion of patients in both the <2Y and 2 to 5Y groups had more than one endoscopies performed in recent 5 years. In fact, there was a
significant association between the number of endoscopies performed in recent 5 years and
SC
gastric cancer-specific or all-cause mortality (Tables 2 and 3). Hence, some of these patients might be under surveillance for high-risk lesions detected on prior endoscopy such as
M AN U
pre-neoplastic lesions or gastric ulcer, which enable them to have early detection of gastric cancer. On the other hand, gastrointestinal cancer that develops within 3 years of last endoscopy could be regarded as interval cancer or missed lesion. There is also a possibility that some patients may have missed lesions by previous endoscopy. In a study from
TE D
Scandinavia, it was shown that the risk of gastroesophageal cancer was very low (<1% with a median follow-up of 2.7 years) after endoscopy 17. In our study, about 16.5% of patients were
EP
diagnosed to have gastric cancer within 2 year of a recent gastroscopy. Although this number appears to be high when compared with similar figures on colonoscopy, a recent study from
AC C
Korea showed that up to 24.3% of their gastric cancer patients were diagnosed within 2 years of a recent endoscopy6. In a recent meta-analysis of 10 studies5, it was found that the “pooled miss rates” of upper gastrointestinal cancer was 6.4% within 1 year and 11.3% within 3 years. Second, it appears perplexing that patients who had received a more recent (<2Y) endoscopy were not associated with a better survival than those with longer endoscopy intervals (2-5Y). Our further analysis showed that those with endoscopy between 5 and 8 years had in fact a higher mortality than these 2 groups (Supplementary Figure 3). In a previous Japanese study, 12
ACCEPTED MANUSCRIPT it was shown that only 3% of low-grade dysplasia progressed to non-invasive cancer during a median follow-up of 4.7 years, suggesting a slow progression rate of low-grade gastric dysplasia.18 The Dutch nationwide histology registry also showed that the annual incidence
RI PT
of gastric cancer was 0.1% for atrophic gastritis, 0.25% for intestinal metaplasia, 0.6% for mild-to-moderate dysplasia, and 6% for severe dysplasia within 5 years after diagnosis.19 In our recent study in Taiwan population20, we found that the 5-, 10-, and 15-year cumulative
SC
incidences of gastric cancer with intestinal metaplasia were 0.9%, 2.0%, and 3.0%,
respectively. On the other hand, there may be a subgroup of patients with more aggressive
M AN U
tumor requiring repeat endoscopy within 2-year, which account for the lower survival benefits in the <2Y group when compared with the 2 to 5Y group. Apart from prior gastroscopy, we also found that patients who had received H pylori
TE D
eradication therapy, both before and after diagnosis of gastric cancer, were associated with significantly lower all-cause mortality. Previous studies21, 22 suggested that patients with H pylori-negative cancer had poor prognosis as compared with those with H pylori-positive
EP
cancers. Moreover, H pylori eradiation prevents metachronous cancer development after
AC C
endoscopic resection of early gastric cancer23. Further studies may be necessary to investigate the role of H pylori eradication on mortality of gastric cancer patients. This study had some limitations. Our study was based on the NHIRD and some clinical information such as endoscopy finding, pathology reports and imaging results were not available in the database. Therefore, we could not adjust for some tumor characteristics such as histological subtypes, tumor staging and location. To overcome this, we matched the cohorts according to the curative therapies received including surgery and chemotherapy as
13
ACCEPTED MANUSCRIPT surrogate parameters for gastric cancer staging. In countries where gastric cancer screening are not routinely practiced, most patients presented in relatively advanced stages and hence, the proportion of patients with non-advanced cancer who could undergo surgical resection
RI PT
was generally low. In this study, 46.3% of patients had received curative treatment for gastric cancer, suggesting that a considerable proportion of patients presented in a relatively
advanced stage. However, the details of surgery received were lacking in the database and
SC
could not be included in the analysis. In addition, we did not analyze patients who received endoscopy resection only, because these procedures were not reimbursed by insurance before
M AN U
2007. Second, although we have matched and controlled for comorbidities in the study design and analyses, the severity and reversibility of these illnesses are difficult to determine. Intuitively, patients with more severe comorbidities might have more opportunity for endoscopy that lead to earlier diagnosis. However, if this is the case, this would only lead to a
TE D
more conservative estimate of the mortality reduction as these patients with more comorbidities would more likely to succumb to their underlying illnesses. Third, the
EP
indications for endoscopy were lacking in this cohort but all procedures were performed for clinical indications as screening gastroscopy was not covered by national insurance in Taiwan.
AC C
Finally, although death is the most likely reason to withdraw from the compulsive national health insurance system, we could not validate whether withdrawal is entirely due to death. However, it is highly likely that death is the reason to withdraw from this compulsive insurance system, especially when these were cancer patients who would need medical care. Even though there may be a few patients who withdrew from national health insurance did not actually succumb to cancer, this bias would only lead to a more conservative estimation and made our observations more robust. 14
ACCEPTED MANUSCRIPT Conclusion We have observed a significantly lower all-cause and gastric cancer-related mortalities in patients who had received any upper endoscopy within 5-year before the cancer diagnosis.
RI PT
These observed benefits on survival of gastric cancer patients with prior endoscopy might support the role of surveillance endoscopy in patients at risk of gastric cancer and repeat
SC
endoscopy examination in those with recurrent symptoms.
M AN U
Author contributions: WKL and CYW were responsible for the conception and design of the study, and writing of the manuscript. HJH was responsible for the data analysis and drafting of the results. JTL and MSW were involved in the critical review of the manuscript and provided essential intellectual content. CYW is the guarantor of the manuscript. Financial support: This study was supported in part by the Taiwan’s National Health
TE D
Research Institutes (PH-105-PP-22) and the Li Shu Fun Medical Foundation Endowment to WKL. This study was based on data from the NHIRD, provided by the Bureau of National Health Insurance, Department of Health and managed by the National Health Research Institutes. The interpretations and conclusions contained herein do not represent those of the
EP
Bureau of National Health Insurance, Department of Health or the National Health Research Institutes. http://www.nhri.org.tw/nhird/en/index.htm
AC C
Potential Competing interests: None REFERENCES
1. Torre LA, Bray F, Siegel RL, et al. Global cancer statistics, 2012. CA Cancer J Clin 2015;65: 87-108. 2. Wu CY, Lin JT. The changing epidemiology of Asian digestive cancers: From etiologies and incidences to preventive strategies. Best Pract Res Clin Gastroenterol 2015;29: 843-53. 3. Gonzalez CA, Sala N, Rokkas T. Gastric cancer: epidemiologic aspects. Helicobacter 2013;18 Suppl 1: 34-8. 4. Leung WK, Wu MS, Kakugawa Y, et al. Screening for gastric cancer in Asia: current evidence and practice. Lancet Oncol 2008;9:279-87. 15
ACCEPTED MANUSCRIPT 5. Menon S, Trudgill N. How commonly is upper gastrointestinal cancer missed at endoscopy? A meta-analysis. Endosc Int Open 2014;2:E46-50. 6. Nam JH, Choi IJ, Cho SJ, et al. Association of the interval between endoscopies with gastric cancer stage at diagnosis in a region of high prevalence. Cancer 2012;118: 4953-60.
RI PT
7. Wu CY, Chen YJ, Ho HJ, et al. Association between nucleoside analogues and risk of hepatitis B virus-related hepatocellular carcinoma recurrence following liver resection. JAMA 2012; 308: 1906-14. 8. Wu CY, Kuo KN, Wu MS, et al. Early Helicobacter pylori eradication decreases risk of gastric cancer in patients with peptic ulcer disease. Gastroenterology 2009; 137: 1641-8 e1-2.
SC
9. Wu CY, Lin JT, Ho HJ, et al. Association of Nucleos(t)ide Analogue Therapy With Reduced Risk of Hepatocellular Carcinoma in Patients With Chronic Hepatitis B-A Nationwide Cohort Study. Gastroenterology 2014; 147: 143-51 e5.
M AN U
10. Wu CY, Chan FK, Wu MS, et al. Histamine2-receptor antagonists are an alternative to proton pump inhibitor in patients receiving clopidogrel. Gastroenterology 2010; 139: 1165-71. 11. Cheng CL, Kao YH, Lin SJ, et al. Validation of the National Health Insurance Research Database with ischemic stroke cases in Taiwan. Pharmacoepidemiol Drug Saf 2011; 20: 236-42. 12. Wu CY, Wu MS, Kuo KN, et al. Effective reduction of gastric cancer risk with regular use of nonsteroidal anti-inflammatory drugs in Helicobacter pylori-infected patients. J Clin Oncol 2010; 28: 2952-7.
TE D
13. Grambsch PM, Therneau TM, Fleming TR. Diagnostic plots to reveal functional form for covariates in multiplicative intensity models. Biometrics 1995; 51: 1469-82. 14. Dan YY, So JB, Yeoh KG. Endoscopic screening for gastric cancer. Clin Gastroenterol Hepatol 2006; 4: 709-16.
EP
15. Gupta N, Bansal A, Wani SB, et al. Endoscopy for upper GI cancer screening in the general population: a cost-utility analysis. Gastrointest Endosc 2011; 74: 610-24 e2.
AC C
16. Dinis-Ribeiro M, Areia M, de Vries AC, et al. Management of precancerous conditions and lesions in the stomach (MAPS): guideline from the European Society of Gastrointestinal Endoscopy (ESGE), European Helicobacter Study Group (EHSG), European Society of Pathology (ESP),and the Sociedade Portuguesa de Endoscopia Digestiva (SPED). Endoscopy 2012;44:74-94. 17. Lassen A, Hallas J, de Muckadell OB. The risk of missed gastroesophageal cancer diagnoses in users and nonusers of antisecretory medication. Gastroenterology 2005;129:1179-86. 18. Yamada H, Ikegami M, Shimoda T, et al. Long-term follow-up study of gastric adenoma/dysplasia. Endoscopy 2004; 36: 390-6. 19. de Vries AC, van Grieken NC, Looman CW, et al. Gastric cancer risk in patients with premalignant gastric lesions: a nationwide cohort study in the Netherlands. Gastroenterology 2008; 134: 945-52. 20. Lee TY, Wang RC, Lee YC, et al. The Incidence of Gastric Adenocarcinoma Among Patients With Gastric Intestinal Metaplasia: A Long-term Cohort Study. J Clin Gastroenterol 2016;50:532-7 16
ACCEPTED MANUSCRIPT 21. Meimarakis G, Winter H, Assmann I, et al. Helicobacter pylori as a prognostic indicator after curative resection of gastric carcinoma: a prospective study. Lancet Oncol. 2006;7:211–22. 22. Marrelli D, Pedrazzani C, Berardi A, et al. Negative Helicobacter pylori status is associated with poor prognosis in patients with gastric cancer. Cancer 2009;115:2071– 80.
AC C
EP
TE D
M AN U
SC
RI PT
23. Fukase K, Kato M, Kikuchi S, et al. Effect of eradication of Helicobacter pylori on incidence of metachronous gastric carcinoma after endoscopic resection of early gastric cancer: an open-label, randomised controlled trial. Lancet 2008;372:392-7.
17
ACCEPTED MANUSCRIPT Table 1. Baseline characteristics of the 3 study cohorts after matching <2Y*
2-5Y*
>5Y*
N=1,286 68.2±12.9
N=1,286 68.2±13.0
N=5,144 68.2±13.0
439 (34.1) 847 (65.9) 2.1±1.7 4.2±3.5 824 (64.1) 628 (48.8) 450 (35.0)
437 (34.0) 849 (66.0) 1.4±0.8 4.0±3.5 836 (65.0) 647 (50.3) 458 (37.7)
279 (21.7) 224 (17.4) 392 (30.5) 135 (10.5) 157 (12.2) 143 (11.1) 14 (1.1) 182 (14.2) 30 (2.3)
P value >.99 0.91 <.001++ <.001 <.001 0.14 0.002 >.99
1,116 (21.7) 896 (17.4) 1,553 (30.2) 554 (10.8) 661 (12.8) 551 (10.7) 68 (1.3) 743 (14.4) 120 (2.3)
425 (33.0) 450 (35.0) 289 (22.5) 122 (9.5)
479 (37.2) 451 (35.1) 252 (19.6) 104 (8.1)
0 (0.0) 0 (0.0) 0 (0.0) 0 (0.0)
169 (13.1) 460 (35.8) 21 (1.6) 71 (5.5) 1.4±2.1
169 (13.1) 466 (36.2) 18 (1.4) 81 (6.3) 1.4±2.0
676 (13.1) 1,866 (36.3) 45 (0.9) 279 (5.4) 0.8±1.8
TE D
EP
AC C
1,777 (34.5) 3,367 (65.5) 0.0±0.0 3.3±3.4 3,745 (72.8) 2,436 (47.4) 2,068 (40.2)
279 (21.7) 224 (17.4) 386 (30.0) 140 (10.9) 162 (12.6) 142 (11.0) 12 (0.9) 200 (15.6) 37 (2.9)
M AN U
Age (mean ± SD) Gender Female Male Number of endoscopy in recent 5 years+ Follow-up years All deaths Curative surgery for gastric cancer Chemotherapy for gastric cancer Comorbidities^ Peptic ulcer disease Gastric ulcer Duodenal ulcer Hypertension Cerebrovascular accident Acute coronary syndrome COPD Liver cirrhosis Diabetes mellitus Renal failure Gastroscopy performed at Medical Centers Regional hospitals Local hospitals Clinics Concomitant medications^ Anti-HP therapy NSAIDs/Aspirin/COX2^ Statins^ Metformin^ Number of prior hospitalizations
RI PT
+
SC
Cohorts#
0.97 0.95 0.82 0.88 0.47 0.54 0.51 <.001++
>.99 0.94 0.03 0.47 <.001
*<2Y cohort: receiving endoscopy between 6-month and 2 years before the index date; 2-5Y cohort: receiving endoscopy more than 2 years and less than or equal to 5 years before index date; >5Y cohort: never receive endoscopy within recent 5 years before index date #
Cohorts were matched in 1:1:4 ratio for the <2Y, 2-5Y and >5Y cohorts by age, curative therapy for gastric cancer, H pylori eradication and propensity scores. +: Age, No. of endoscopy in 5 years; Follow-up time are treated as continuous variables and reported in years with mean ± SD. ++: P value is tested by the Student t-test between <2Y and 2-5Y cohorts only. 18
ACCEPTED MANUSCRIPT ##: Peptic ulcer disease, cerebral vascular diseases, acute coronary syndrome, COPD, liver cirrhosis, diabetes, renal failure, use of statins, use of NSAIDs or aspirin or COXIBs, and use of metformin were included in the propensity score calculation. ^: Medication users refer to patients who used these drugs at least one day per month on average. Numbers in bracket are percentage unless specified otherwise.
AC C
EP
TE D
M AN U
SC
RI PT
Abbreviations: N, number; COPD: chronic obstructive pulmonary disease; NSAIDs, non-steroidal anti-inflammatory drugs
19
ACCEPTED MANUSCRIPT Table 2. Multivariate Cox proportional hazards model analysis for all-cause mortality
Events (Number)
5,144 1,286 1,286 7,716 5,063
3,745 824 836 5,405 3,695
3,711
2,230
3,003 7,716
2,410 5,405
EP
P value
1 0.80 (0.72 to 0.89) 0.83 (0.76 to 0.91) 1.02 (1.02 to 1.02) 1.09 (1.03 to 1.16) 0.52 (0.49 to 0.55)
<.001 <.001 <.001 0.002 <.001
1.55 (1.46 to 1.64) 0.95 (0.92 to 0.99)
<.001 0.019
2,227 685 1,740 755 694 86 870 169 5,405
1.02 (0.97 to 1.08) 1.31 (1.20 to 1.44) 0.88 (0.83 to 0.94) 0.93 (0.85 to 1.01) 1.13 (1.04 to 1.23) 2.00 (1.62 to 2.48) 1.17 (1.07 to 1.28) 1.59 (1.36 to 1.86) 1.01 (1.00 to 1.03)
0.488 <.001 <.001 0.088 0.005 <.001 <.001 <.001 0.036
488 317 171 2,180 42 321
0.57 (0.51 to 0.63) 0.78 (0.69 to 0.89) 0.40 (0.34 to 0.47) 1.30 (1.23 to 1.37) 0.54 (0.39 to 0.73) 0.83 (0.73 to 0.95)
<.001 <.001 <.001 <.001 <.001 0.006
SC
RI PT
HR (95% CI)
M AN U
3,018 829 2,331 980 836 94 1,125 187 7,716
TE D
Endoscopy interval >5Y <2Y 2-5Y Age (each incremental year) Male Curative surgery for gastric cancer Chemotherapy for gastric cancer No. of endoscopy in 5 years* Comorbidities Peptic ulcer disease Cerebrovascular accident Hypertension Acute coronary syndrome COPD Liver cirrhosis Diabetes Renal failure No. of prior hospitalizations Concomitant drugs H pylori eradication Pre-cancer diagnosis Post-cancer diagnosis NSAIDs/Aspirin/COX2 Statins Metformin
Patients (Number)
1,014 568 446 2,792 84 431
AC C
*No. of endoscopy in 5 years: number of endoscopy in the 5 years before index date, analyzed by each incremental number as a continuous variable Abbreviations: HR: hazard ratio; COPD, chronic obstructive pulmonary disease; NSAIDs, non-steroidal anti-inflammatory drugs
20
ACCEPTED MANUSCRIPT Table 3. Multivariate Cox proportional hazards model analysis for gastric cancer-specific mortality Events (Number)
5,144 1,286 1,286 7,716 5,063
2,692 553 567 3,812 2,581
3,711
1,530
3,003 7,716
1,744 3,812
1,014 568 446 2,792 84 431
AC C
EP
P value
1 0.80 (0.71 to 0.91) 0.83 (0.75 to 0.93) 1.01 (1.01 to 1.02) 1.04 (0.97 to 1.12) 0.58 (0.54 to 0.62)
<.001 <.001 <.001 0.259 <.001
1.49 (1.39 to 1.59) 0.96 (0.91 to 1.00)
<.001 0.052
1,566 1,101 445 538 463 57 589 104 3,812
1.04 (0.97 to 1.11) 0.89 (0.82 to 0.96) 1.08 (0.96 to 1.21) 1.05 (0.95 to 1.16) 1.05 (0.94 to 1.17) 1.36 (0.98 to 1.89) 1.08 (0.97 to 1.21) 1.10 (0.88 to 1.36) 1.00 (0.98 to 1.01)
0.268 0.003 0.222 0.332 0.426 0.063 0.167 0.400 0.639
339 223 116 1,521 29 213
0.60 (0.53 to 0.68) 0.79 (0.68 to 0.92) 0.44 (0.37 to 0.53) 1.22 (1.14 to 1.30) 0.64 (0.45 to 0.93) 0.86 (0.73 to 1.01)
<.001 0.002 <.001 <.001 0.018 0.073
SC
RI PT
HR (95% CI)
M AN U
3,018 2,331 829 980 836 94 1,125 187 7,716
TE D
Screening interval >5Y <2Y 2-5Y Age (each incremental year) Male Curative resection for gastric cancer Chemotherapy for gastric cancer No. of endoscopy in 5 years* Comorbidities Peptic ulcer disease Hypertension Cerebrovascular accident Acute coronary syndrome COPD Liver cirrhosis Diabetes Renal failure No. of prior hospitalizations Concomitant drugs H pylori eradication Pre-cancer diagnosis Post-cancer diagnosis NSAIDs/Aspirin/COX2 Statins Metformin
Patients (Number)
*: No. of endoscopy in 5 years: number of endoscopy in the 5 years before index date, analyzed by each incremental number as a continuous variable Abbreviations: HR: hazard ratio; COPD, chronic obstructive pulmonary disease; NSAIDs, non-steroidal anti-inflammatory drugs
21
ACCEPTED MANUSCRIPT FIGURE LEGENDS: Figure 1. Flow diagram of patient’s selection *A case could be excluded because of more than one criterion. Therefore, the total excluded
RI PT
cases in each step may outnumber the sum of case numbers excluded by individual criteria. <2Y cohort: receiving endoscopy between 6 months and 2 years before the index date; 2-5Y cohort:
SC
receiving endoscopy more than 2 years and less than or equal to 5 years before index date; >5Y cohort: never receive endoscopy within recent 5 years of index date.
M AN U
GCA; gastric cancer
Figure 2. Cumulative incidences of all-cause mortality and gastric cancer-specific
TE D
mortality in gastric cancer patients who had prior gastroscopy at different time intervals. A: All-cause mortality. B: Gastric-cancer specific mortality.
EP
Calculation and comparison of cumulative mortalities were conducted using Kaplan-Meier method. (P<0.001 for both all-cause and gastric cancer-specific mortalities when comparing
AC C
<2Y versus >5Y and 2 to 5Y versus >5Y cohorts. P value was not significant when comparing <2Y versus 2-5Y cohorts)
22
ACCEPTED MANUSCRIPT Supplementary Table 1: Baseline characteristics and outcomes of study cohorts before propensity score matching 2-5Y*
Age (mean±SD)
N=3,303
N=1,785
N=14,978
P-value
67.6±13.7
69.0±13.2
67.0±14.2
<.001
Gender
>5Y*
RI PT
<2Y*
Cohort#
0.01
1,110 (33.6)
598 (33.5)
5,379 (35.9)
Male
2,193 (66.4)
1,187 (66.5)
9,599 (64.1)
Number of gastroscopy in 5yrs (mean±SD)
2.3±2.0
1.4±0.8
0.0±0.0
<.001
Follow-up period (yrs, mean±SD)
4.0±3.5
3.8±3.5
3.3±3.4
<.001
1,201 (67.3)
10,742 (71.7)
<.001
842 (47.2)
6,970 (46.5)
0.19
617 (34.6)
6,115 (40.8)
<.001
2,186 (66.2)
Curative surgery for gastric cancer
1,484 (44.9)
Chemotherapy for gastric cancer
1,076 (32.6)
Comorbidities
##
Peptic ulcer diseases Gastric ulcer diseases Duodenal ulcer diseases Hypertension
<.001
906 (27.4)
442 (24.8)
2,261 (15.1)
945 (28.6)
454 (25.4)
1,554 (10.4)
1,335 (40.4)
671 (37.6)
4,370 (29.2)
<.001
544 (16.5)
274 (15.4)
1,712 (11.4)
<.001
Acute coronary syndrome
718 (21.7)
382 (21.4)
1,940 (13.0)
<.001
COPD
638 (19.3)
339 (19.0)
1,648 (11.0)
<.001
221 (6.7)
98 (5.5)
421 (2.8)
<.001
718 (21.7)
344 (19.3)
2,448 (16.3)
<.001
249 (7.5)
111 (6.2)
588 (3.9)
<.001
Anti-HP therapy
894 (27.1)
335 (18.8)
1,460 (9.7)
<.001
AC C
TE D
Cerebrovascular disease
M AN U
Overall death
SC
Female
693 (21.0)
247 (13.8)
637 (4.3)
<.001
201 (6.1)
88 (4.9)
823 (5.5)
<.001
1,268 (38.4)
669 (37.5)
5,363 (35.8)
0.01
79 (2.4)
45 (2.5)
254 (1.7)
0.003
254 (7.7)
130 (7.3)
977 (6.5)
0.04
2.3±3.7
1.9±2.5
0.8±1.8
<.001
Liver cirrhosis Renal failure
EP
Diabetes Concomitant drugs
##
Before GC After GC
^
NSAIDs/Aspirin/COX2 Statins^
Metformin
^
Number of prior hospitalizations
*<2Y: cohort receiving endoscopy screening between 6-month and 2 years after index date; 2-5Y: cohort 23
ACCEPTED MANUSCRIPT receiving endoscopy screening between 3-5 years after index date; >5Y: cohort never receiving endoscopy screening within 5 years after index date #: Cohorts were subsequently matched in 1:1:4 ratio for the <2Y, 2-5Y and >5Y cohorts by age, curative therapy, H pylori eradication and propensity scores in the final analysis.
RI PT
+: Age and No. of endoscopy in 5 yr are treated as continuous variables
##: Peptic ulcer diseases, cerebrovascular diseases, acute coronary syndrome, COPD, liver cirrhosis, diabetes, renal failure, use of statins, use of NSAIDs or aspirin or COXIBs, and use of metformin were
SC
included in the propensity score calculation. ^
M AN U
: Medication users refer to patients who used these drugs at least one day per month on average.
Abbreviations: N, number; No. of endoscopy in 5yrs, number of endoscopy in the 5 years before index date; COPD: chronic obstructive pulmonary disease; pre-GC: H pylori eradication before gastric cancer diagnosis; post-GC: H pylori eradication after gastric cancer diagnosis; NSAIDs, non-steroidal
AC C
EP
TE D
anti-inflammatory drugs
24
ACCEPTED MANUSCRIPT Supplementary Figure Legends Supplementary Figure 1:
RI PT
Cumulative incidences of non-gastric cancer-related mortality in patients who had prior gastroscopy in different time frames.
SC
Supplementary Figure 2:
M AN U
Cumulative incidences of all-cause mortality in patients according to different age groups.
B: 65-80 years (log rank, P<0.001)
TE D
A: younger than 65 years (log rank, P<0.001).
Supplementary Figure 3:
EP
Cumulative incidences of all-cause mortality in gastric cancer patients who had prior
AC C
gastroscopy in different time frames up to 8 years. The original control cohort was further classified into 6 to 8Y cohort (between 5 and 8 years) and >8Y (no endoscopy in recent 8 years) cohort. The 4 groups were again matched in 1:1:1:4 ratio as described previously (log-rank, P < 0.001).
25
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
ACCEPTED MANUSCRIPT Acronyms
HP, Helicobacter pylori HR, hazard ratio
AC C
EP
TE D
M AN U
SC
RCIPD, Registry for Catastrophic Illness Patient Database
RI PT
NHIRD, National Health Insurance Research Database
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
S0016-5107(17)32036-9
DOI:
10.1016/j.gie.2017.06.013
Reference:
YMGE 10628
To appear in:
Gastrointestinal Endoscopy
Received Date: 7 March 2017 Accepted Date: 12 June 2017
Please cite this article as: Leung WK, Ho HJ, Lin J-T, Wu M-S, Wu C-Y, Prior gastroscopy and mortality in patients with gastric cancer: a matched retrospective cohort study, Gastrointestinal Endoscopy (2017), doi: 10.1016/j.gie.2017.06.013. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT Prior gastroscopy and mortality in patients with gastric cancer: a matched retrospective cohort study Short title: Prior gastroscopy and gastric cancer mortality
RI PT
Wai K Leung, MD1, Hsiu J. Ho, PhD2, Jaw-Town Lin, MD3-4, Ming-Shiang Wu, MD5, Chun-Ying Wu, MD2, 6-9*
Department of Medicine, University of Hong Kong, Hong Kong, China;
2
Division of Gastroenterology, Taichung Veterans General Hospital, Taichung, Taiwan;
3
School of Medicine, Fu Jen Catholic University, Taipei, Taiwan;
4
Institute of Population Health Sciences, National Health Research Institutes, Miaoli, Taiwan;
6
Division of Gastroenterology, National Taiwan University Hospital, Taipei, Taiwan; Faculty of Medicine and Graduate Institute of Clinical Medicine, National Yang-Ming
TE D
University, Taipei, Taiwan; 7
M AN U
5
SC
1
Department of Public Health and Graduate Institute of Clinical Medical Sciences, China
Medical University, Taichung, Taiwan;
National Institute of Cancer Research, National Health Research Institutes, Miaoli, Taiwan;
9
Department of Life Sciences and Rong Hsing Research Center for Translational Medicine,
EP
8
AC C
National Chung-Hsing University, Taichung, Taiwan * Correspondence:
Professor Chun-Ying Wu, MD, PhD, MPH Division of Gastroenterology, Taichung Veterans General Hospital No. 1650, Section 4, Taiwan Avenue, Taichung 407, Taiwan E-mail: [email protected]; Tel: +886-4-23592525 # 3304; Fax: +886-4-23741331
1
ACCEPTED MANUSCRIPT ABSTRACT Background & Aims: The role of prior gastroscopy on outcome of gastric cancer patients remains unknown. This study determines the association between intervals of prior
RI PT
gastroscopy and mortality in patients with gastric cancer. Methods: We identified 20,066 newly diagnosed gastric cancer patients in the national health insurance database of Taiwan between 2002 and 2007. After excluding gastroscopy
SC
performed 6 months before diagnosis of cancer, patients were matched into 3 cohorts
M AN U
according to the intervals of prior gastroscopy: 6 months to 2 years (<2Y cohort), 2 to 5 years (2-5Y cohort), and none within recent 5 years (>5Y cohort). The 3 cohorts were matched for age, curative treatment for gastric cancer, Helicobacter pylori therapy and propensity scores that comprised of gender, comorbidities and concomitant medication usage. The primary
TE D
outcome is the hazard ratio (HR) of all-cause mortality.
Results: After matching, we identified 1,286, 1,286, and 5,144 patients for the <2Y, 2 to 5Y and >5Y cohorts. Compared with the >5Y cohort, the HR of all-cause mortality for the <2Y
EP
and 2 to 5Y cohorts was 0.80 (95% CI, 0.72-0.89; P<0.001) and 0.83 (95% CI, 0.76-0.91;
AC C
P<0.001), respectively. The HRs of gastric cancer-specific mortality was also significantly lower in the <2Y (0.80; 95% CI, 0.71-0.91; P<0.001) and 2 to 5Y cohorts (0.83; 95% CI, 0.75-0.93; P<0.001).
Conclusion: Gastric cancer patients who had gastroscopy performed within 5-year before cancer diagnosis have significantly lower mortality. Our results may support the role of repeat endoscopic examination or surveillance endoscopy in selected patients.
2
ACCEPTED MANUSCRIPT INTRODUCTION Gastric cancer is the third leading cause of cancer death in the world.1 In 2012, there were approximately 950,000 new cases and more than 720,000 deaths from gastric cancer globally.
RI PT
Although the incidences of gastric cancer are generally declining in most countries, it is still having significant impact on many developing countries where 70% of gastric cancer cases occur. In particular, about 50% of gastric cancer cases occur in East Asian countries including
SC
China, Japan and Korea.2,3
M AN U
Whilst screening gastroscopy is rarely performed in most countries except in a few East Asian countries4, majority of gastroscopy are performed in clinical practices as a diagnostic work-up for upper gastrointestinal symptoms. Moreover, it is used for monitoring of gastric ulcer healing and surveillance of pre-neoplastic gastric lesions. Unlike surveillance
TE D
colonoscopy, the optimal interval of repeating gastroscopy in diverse patient groups remains uncertain. As yet, the real impact of routine daily gastroscopy on subsequent outcome of gastric cancer patients has not been properly addressed. In a recent Korean study, it was
EP
shown that gastric cancer patients who had upper endoscopy performed within 3 years or less
AC C
before the diagnosis of cancer had a significantly lower cancer staging as compared with those who never had endoscopy.6 However, the most important clinical outcome, ie, actual reduction in mortality, had not been demonstrated. Given the possible benefits of repeated endoscopic examinations, there is a significant knowledge gap on the role of repeated endoscopic examination on the actual outcome of gastric cancer patients. The current study was a retrospective cohort study of gastric cancer patients with propensity score matching based on a large population-based health insurance database from Taiwan. We 3
ACCEPTED MANUSCRIPT determined the hazard ratio of all-cause mortality in patients who had gastroscopy performed at different intervals before the diagnosis of gastric cancer. Unlike previous studies that focused on screening gastroscopy, we evaluated the role of routine gastroscopies performed
RI PT
for various indications in patients who were subsequently confirmed to have gastric cancer. METHODS
SC
Data Source:
This was a nationwide matched retrospective cohort study based on the Taiwan’s National
M AN U
Health Insurance Research Database (NHIRD). The NHIRD is a compulsory national health insurance program, which covers all 23.7 million Taiwan’s residents. The NHIRD includes comprehensive health care information and the details have been described in our previous studies.7-9 In short, patients’ characteristics, diagnoses, drugs, procedures, and laboratory
TE D
examination items were all available in NHIRD. The International Classification of Disease-9 (ICD-9) codes were used to classify diseases in the database and the accuracy of diagnosis of several major diseases, such as ischemic stroke and acute coronary syndrome, have been
AC C
of Taiwan.
EP
previously validated.10, 11 This study was approved by the National Health Research Institutes
Study Subjects:
We identified all patients who were newly diagnosed to have primary adenocarcinoma of stomach from the NHIRD based on the ICD-9 code (ICD-9 code: 151) between January 1, 2002 and December 31, 2007. To verify the diagnosis, these patients should also be enrolled in the Registry for Catastrophic Illness Patient Database (RCIPD) at the same time. The
4
ACCEPTED MANUSCRIPT RCIPD is a subset of the NHIRD, which requires histological diagnosis based on endoscopic biopsy or surgical specimens for registration as cancer. We excluded patients who had incomplete demographic information or those with other malignancies (ICD codes: 140-150;
RI PT
152-208) before the diagnosis of gastric cancer. We defined the index date as the date of gastric cancer diagnosis, which was the first date of hospitalization for work up and
management of gastric cancer or the first date of enrollment to the RCIPD. Due to the
SC
insurance reimbursement issue, delay in registration in the RCIPD is uncommon.
M AN U
We retrieved the number and date of gastroscopy, performed for any indications, in these patients before the diagnosis of gastric cancer. We excluded gastroscopy performed within 6 months before the diagnosis of gastric cancer as these endoscopies might be used for diagnosis of cancer. The timing of the most recent endoscopy performed before the index
TE D
date was used to classify patients into 3 cohorts: between 6 months and 2 years before the index date (<2Y cohort), between 2 and 5 years (2-5Y cohort) and no endoscopy within recent 5 years (>5Y cohort).
EP
Matching of the 3 Cohorts:
AC C
The baseline data of the 3 cohorts were reported in Supplementary Table 1. To control for possible confounding factors, the 3 groups were matched for age, curative surgery for gastric cancer, Helicobacter pylori therapy, and propensity scores which was calculated on gender, comorbid illnesses and concomitant medications. Patients’ characteristics were defined a priori and analyzed. Because gastric cancer staging data was not available in the insurance database, curative surgery for gastric cancer (total or subtotal gastrectomy with lymph node dissection) and use of chemotherapy were used as surrogate parameters for cancer staging. 5
ACCEPTED MANUSCRIPT The following co-morbidities were included in the propensity score analysis: history of peptic ulcer disease (PUD; ICD codes: 531-533), liver cirrhosis (ICD-9 code: 571⋅5), acute coronary syndrome (ACS; ICD-9 codes: 410-414), cerebrovascular accident (CVA; ICD-9 codes:
RI PT
430-438), chronic obstructive pulmonary diseases (COPD; ICD-9 codes: 490-496); diabetes mellitus (ICD-9 codes: 205), renal failure (ICD-9 codes: 584-586) and hypertension (ICD-9 codes: 401-405). We also analyzed drugs that may be associated with mortality including H
SC
pylori eradication therapy (regimens details described in our previous studies)8, aspirin,
non-steroidal anti-inflammatory drugs (NSAIDs) including COX-2 inhibitors,12 statins and
M AN U
metformin. Concomitant medications were defined as patients who had used these agents for more than one tablet per month before the diagnosis of gastric cancer. H pylori eradication was defined as whether the patient had ever received H pylori eradication treatment. These regimes were then further classified into pre-cancer and post-cancer diagnosis according to
TE D
the timing of H pylori eradication received before or after gastric cancer diagnosis, respectively. A propensity score for each patient was calculated on the logistic regression
EP
model of gender, comorbidities and concomitant medications that was used to estimate the probability of classifying these patients into 3 cohorts. We matched the final cohorts in a
AC C
1:1:4 ratios for the <2Y, 2 to 5Y, and >5Y cohorts (Table 1). Outcome Measures:
The primary outcome of this study is the hazard ratio (HR) of all-cause mortality whereas the secondary outcome is the HR of gastric cancer-specific mortality. All 3 cohorts were followed up from the date of cancer diagnosis to the date of death or the end of 2012. Death was identified from withdrawal from the compulsory National Health Insurance program. Gastric
6
ACCEPTED MANUSCRIPT cancer-specific mortality was defined when the patients had a primary diagnosis of gastric cancer during the last hospitalization before death. Death from other causes before gastric cancer-specific mortality were treated as competing mortality by Gray methods to calculate
RI PT
the modified Cox proportional hazards ratio.
Statistical Analysis
SC
The baseline demographic, comorbidities and concomitant medication uses of the 3 cohorts were compared by the χ2 test or the Student t-test. We matched these 3 cohorts by age (± 2
M AN U
years), curative therapy of gastric cancer, H pylori therapy, and propensity score (± 0.01). Propensity score was calculated by multinomial logistic regression based on the following covariates: gender, liver cirrhosis, acute coronary syndrome, cerebrovascular accident, chronic obstructive pulmonary disease, diabetes mellitus, renal failure, hypertension,
TE D
hyperlipidemia, NSAIDs/aspirin/COX2, statins, and metformin use status. We select the candidate who has the nearest character, but not more than a specific threshold list in
EP
parentheses for continuous variables, age and propensity score.
AC C
The cumulative incidences and HRs of all-cause and gastric cancer-specific mortalities of the 3 groups were compared. The cumulative incidences of all-cause mortality were constructed by Kaplan-Meier method and the cumulative incidences of cancer-specific mortality by Gray method, which is a modified Kaplan-Meier method for competing risk. These 2 methods and comparisons among 3 groups were implemented by the cmprsk package in R software. We used Cox proportional hazards model to determine whether endoscopy interval is independently associated with mortality. All parameters were defined a priori and used in the prediction model to adjust for as many confounders as possible. The assumption of a Cox 7
ACCEPTED MANUSCRIPT proportional hazards model was tested using the method of Grambsch et al13 and was implemented using the functional cox.zph in R package “survival.”
RI PT
RESULTS Patients’ Characteristics
Between January 1, 2002 and December 31, 2007, a total of 20,066 patients were newly
SC
diagnosed to have gastric cancer and fulfilled our inclusion criteria (Figure 1). Among them, 3,303 and 1,785 patients had received upper endoscopy between 6 months to 2 years (<2Y
M AN U
cohort) and between 2 and 5 years (2-5Y cohort) before cancer diagnosis, respectively. The remaining 14,978 patients had not received upper endoscopy in the recent 5 years (>5Y cohort; Supplementary Table 1). There were significant differences in the baseline characteristics among the 3 groups as shown in Supplementary Table 1. After matching, there
TE D
were 1,286, 1,286, and 5,144 patients from the <2Y, 2 to 5Y, and >5Y cohorts, respectively. The 3 cohorts were matched in baseline characteristics and comorbid illnesses except for the
EP
number of endoscopy in recent 5 years, follow-up duration, mortality rates, uses of
AC C
chemotherapy or statins and number of prior hospitalizations (Table 1).
Mortalities of Gastric Cancer Patients Both the <2Y and 2 to 5Y cohorts had significantly lower all-cause as well as gastric cancer-related mortality when compared with the >5Y group (P<0.001; Figures 2A and 2B). However, there was no difference in the non-cancer-related mortality among the 3 groups (Supplementary Figure 1).
8
ACCEPTED MANUSCRIPT Multivariable Cox Proportional Hazards Model Analysis After adjusting for various potential confounding factors, we found that prior gastroscopy is independently associated with a reduced risk of all-cause mortality (Table 2). Compared with
RI PT
the >5Y cohort, the hazards ratios (HRs) of all-cause mortality for the <2Y and 2 to 5Y
cohorts were 0.80 (95% CI, 0.72 to 0.89; P<0.001) and 0.83 (95% CI, 0.76 to 0.91; P<0.001), respectively. The number of endoscopy performed in the recent 5-year was also found to be
SC
significantly associated with a lower all-cause mortality (HR 0.95; 95% CI, 0.92 to 0.99;
model was satisfied (P=0.07).
M AN U
P=0.019). The assumption of our Cox proportional hazard model has been tested and the
Table 3 showed the multivariable Cox proportional hazards model analysis of the gastric cancer-specific mortality. Compared with the >5Y cohort, the HR for cancer-specific
TE D
mortality of the <2Y and 2 to 5Y cohorts was 0.80 (95% CI, 0.71 - 0.91; P<0.001) and 0.83 (95% CI, 0.75 to 0.93; P<0.001), respectively.
EP
Other factors that were found to be significantly associated with both lower all-cause and cancer-specific mortalities included younger patient’s age, curative surgery for gastric cancer,
AC C
no need for chemotherapy, hypertension, H pylori eradication therapy and statins use (Tables 2 and 3).
Sensitivity Analyses
As patients’ age is an important factor associated with mortality, we conducted sensitivity analyses by including patients younger than 80 years only. We found that the <2Y and 2 to 5Y cohorts still had lower mortality when compared with the >5Y group in both younger
9
ACCEPTED MANUSCRIPT (<65 years) and older (65-80 years) age groups (Supplementary Figure 2; both P<0.001). Next, we performed sensitivity analysis to determine the association of prior gastroscopy interval of longer than 5-year interval and all-cause mortality of cancer patients. In this
RI PT
analysis, the >5Y cohort was further divided into 2 groups with one group having prior
endoscopy between 5 and 8 years (5-8Y cohort) and another group with no endoscopy in recent 8 years (>8Y cohort). Consistent with our above findings, gastric cancer patients who
SC
had gastroscopy between 5 and 8 years had a lower mortality than >8Y cohort, but a higher
M AN U
mortality than the <2Y and 2 to 5Y cohorts (Supplementary Figure 3; P<0.001). DISCUSSION
This population-based matched retrospective cohort study from Taiwan demonstrated that upper endoscopy performed for various indications within 5 years before gastric cancer
TE D
diagnosis is associated with a significantly lower all-cause as well as gastric cancer-specific mortalities when compared with those with no endoscopy in recent 5 years. The lower
EP
mortality remains significant even after adjusting for various potential confounding factors such as patient’s age, gender, treatment for gastric cancer, H pylori eradication, comorbid
AC C
illnesses and concomitant medication uses. It is so far the first population-based study that demonstrates the association between routine upper endoscopy, not screening endoscopy, and mortality in gastric cancer patients. The strengths of this study were that we used the national health care insurance database of Taiwan, which covers the whole 25 million populations of Taiwan, to avoid potential selection bias. The comprehensive information available from the health care database allow for control of many potential confounders. To avoid potential difference in baseline 10
ACCEPTED MANUSCRIPT characteristics or health-seeking behavior of the 3 cohorts, we matched the 3 cohorts according to their age, treatment for gastric cancer, H pylori eradication therapy, and propensity scores that included comorbid illnesses and concomitant medications. Moreover,
RI PT
our analysis included both all-cause as well as cancer-specific mortalities. To further support our findings, there was no difference in non-cancer-related mortality among the 3 cohorts, which indicate that the difference in all-cause mortality is mainly driven by difference in
SC
cancer-specific mortality. Third, we have used 2 different sensitivity analyses that looked into
(Supplementary Figures 2-3).
M AN U
the effects of age and longer endoscopy interval to support the robustness of our data
Unlike screening endoscopy14-15, it remains unknown whether upper endoscopy performed for various clinical indications before the cancer diagnosis is associated with lower mortality
TE D
in gastric cancer patients. Moreover, no study has demonstrated the actual reduction in mortality by routine gastroscopy. As the present study included upper endoscopy performed in all age groups regardless of their indications, these findings would be more applicable to
EP
real-life daily clinical practices when endoscopy is performed for diagnostic work up or
AC C
surveillance, rather than screening purpose. It is, however, important to highlight that the majority of our study population are ethnic Chinese (98%). As yet, the latest gastric cancer-related mortality in Taiwan is 12.2 and 7.0 per 100,000 for men and women, respectively, which is much lower than other East Asian countries. The use of population with intermediate cancer risk further increases the external validity of our observations to other non-high risk regions. There are 2 important questions relating to our findings. First, the mechanism underlying the
11
ACCEPTED MANUSCRIPT reduction in mortality by routine clinical gastroscopy remains elusive. Whilst gastric carcinogenesis is generally believed to be a slow process, the recent European Consensus recommended endoscopic surveillance in 3 years for patients with pre-neoplastic gastric
RI PT
lesions 16. We noted that a considerable proportion of patients in both the <2Y and 2 to 5Y groups had more than one endoscopies performed in recent 5 years. In fact, there was a
significant association between the number of endoscopies performed in recent 5 years and
SC
gastric cancer-specific or all-cause mortality (Tables 2 and 3). Hence, some of these patients might be under surveillance for high-risk lesions detected on prior endoscopy such as
M AN U
pre-neoplastic lesions or gastric ulcer, which enable them to have early detection of gastric cancer. On the other hand, gastrointestinal cancer that develops within 3 years of last endoscopy could be regarded as interval cancer or missed lesion. There is also a possibility that some patients may have missed lesions by previous endoscopy. In a study from
TE D
Scandinavia, it was shown that the risk of gastroesophageal cancer was very low (<1% with a median follow-up of 2.7 years) after endoscopy 17. In our study, about 16.5% of patients were
EP
diagnosed to have gastric cancer within 2 year of a recent gastroscopy. Although this number appears to be high when compared with similar figures on colonoscopy, a recent study from
AC C
Korea showed that up to 24.3% of their gastric cancer patients were diagnosed within 2 years of a recent endoscopy6. In a recent meta-analysis of 10 studies5, it was found that the “pooled miss rates” of upper gastrointestinal cancer was 6.4% within 1 year and 11.3% within 3 years. Second, it appears perplexing that patients who had received a more recent (<2Y) endoscopy were not associated with a better survival than those with longer endoscopy intervals (2-5Y). Our further analysis showed that those with endoscopy between 5 and 8 years had in fact a higher mortality than these 2 groups (Supplementary Figure 3). In a previous Japanese study, 12
ACCEPTED MANUSCRIPT it was shown that only 3% of low-grade dysplasia progressed to non-invasive cancer during a median follow-up of 4.7 years, suggesting a slow progression rate of low-grade gastric dysplasia.18 The Dutch nationwide histology registry also showed that the annual incidence
RI PT
of gastric cancer was 0.1% for atrophic gastritis, 0.25% for intestinal metaplasia, 0.6% for mild-to-moderate dysplasia, and 6% for severe dysplasia within 5 years after diagnosis.19 In our recent study in Taiwan population20, we found that the 5-, 10-, and 15-year cumulative
SC
incidences of gastric cancer with intestinal metaplasia were 0.9%, 2.0%, and 3.0%,
respectively. On the other hand, there may be a subgroup of patients with more aggressive
M AN U
tumor requiring repeat endoscopy within 2-year, which account for the lower survival benefits in the <2Y group when compared with the 2 to 5Y group. Apart from prior gastroscopy, we also found that patients who had received H pylori
TE D
eradication therapy, both before and after diagnosis of gastric cancer, were associated with significantly lower all-cause mortality. Previous studies21, 22 suggested that patients with H pylori-negative cancer had poor prognosis as compared with those with H pylori-positive
EP
cancers. Moreover, H pylori eradiation prevents metachronous cancer development after
AC C
endoscopic resection of early gastric cancer23. Further studies may be necessary to investigate the role of H pylori eradication on mortality of gastric cancer patients. This study had some limitations. Our study was based on the NHIRD and some clinical information such as endoscopy finding, pathology reports and imaging results were not available in the database. Therefore, we could not adjust for some tumor characteristics such as histological subtypes, tumor staging and location. To overcome this, we matched the cohorts according to the curative therapies received including surgery and chemotherapy as
13
ACCEPTED MANUSCRIPT surrogate parameters for gastric cancer staging. In countries where gastric cancer screening are not routinely practiced, most patients presented in relatively advanced stages and hence, the proportion of patients with non-advanced cancer who could undergo surgical resection
RI PT
was generally low. In this study, 46.3% of patients had received curative treatment for gastric cancer, suggesting that a considerable proportion of patients presented in a relatively
advanced stage. However, the details of surgery received were lacking in the database and
SC
could not be included in the analysis. In addition, we did not analyze patients who received endoscopy resection only, because these procedures were not reimbursed by insurance before
M AN U
2007. Second, although we have matched and controlled for comorbidities in the study design and analyses, the severity and reversibility of these illnesses are difficult to determine. Intuitively, patients with more severe comorbidities might have more opportunity for endoscopy that lead to earlier diagnosis. However, if this is the case, this would only lead to a
TE D
more conservative estimate of the mortality reduction as these patients with more comorbidities would more likely to succumb to their underlying illnesses. Third, the
EP
indications for endoscopy were lacking in this cohort but all procedures were performed for clinical indications as screening gastroscopy was not covered by national insurance in Taiwan.
AC C
Finally, although death is the most likely reason to withdraw from the compulsive national health insurance system, we could not validate whether withdrawal is entirely due to death. However, it is highly likely that death is the reason to withdraw from this compulsive insurance system, especially when these were cancer patients who would need medical care. Even though there may be a few patients who withdrew from national health insurance did not actually succumb to cancer, this bias would only lead to a more conservative estimation and made our observations more robust. 14
ACCEPTED MANUSCRIPT Conclusion We have observed a significantly lower all-cause and gastric cancer-related mortalities in patients who had received any upper endoscopy within 5-year before the cancer diagnosis.
RI PT
These observed benefits on survival of gastric cancer patients with prior endoscopy might support the role of surveillance endoscopy in patients at risk of gastric cancer and repeat
SC
endoscopy examination in those with recurrent symptoms.
M AN U
Author contributions: WKL and CYW were responsible for the conception and design of the study, and writing of the manuscript. HJH was responsible for the data analysis and drafting of the results. JTL and MSW were involved in the critical review of the manuscript and provided essential intellectual content. CYW is the guarantor of the manuscript. Financial support: This study was supported in part by the Taiwan’s National Health
TE D
Research Institutes (PH-105-PP-22) and the Li Shu Fun Medical Foundation Endowment to WKL. This study was based on data from the NHIRD, provided by the Bureau of National Health Insurance, Department of Health and managed by the National Health Research Institutes. The interpretations and conclusions contained herein do not represent those of the
EP
Bureau of National Health Insurance, Department of Health or the National Health Research Institutes. http://www.nhri.org.tw/nhird/en/index.htm
AC C
Potential Competing interests: None REFERENCES
1. Torre LA, Bray F, Siegel RL, et al. Global cancer statistics, 2012. CA Cancer J Clin 2015;65: 87-108. 2. Wu CY, Lin JT. The changing epidemiology of Asian digestive cancers: From etiologies and incidences to preventive strategies. Best Pract Res Clin Gastroenterol 2015;29: 843-53. 3. Gonzalez CA, Sala N, Rokkas T. Gastric cancer: epidemiologic aspects. Helicobacter 2013;18 Suppl 1: 34-8. 4. Leung WK, Wu MS, Kakugawa Y, et al. Screening for gastric cancer in Asia: current evidence and practice. Lancet Oncol 2008;9:279-87. 15
ACCEPTED MANUSCRIPT 5. Menon S, Trudgill N. How commonly is upper gastrointestinal cancer missed at endoscopy? A meta-analysis. Endosc Int Open 2014;2:E46-50. 6. Nam JH, Choi IJ, Cho SJ, et al. Association of the interval between endoscopies with gastric cancer stage at diagnosis in a region of high prevalence. Cancer 2012;118: 4953-60.
RI PT
7. Wu CY, Chen YJ, Ho HJ, et al. Association between nucleoside analogues and risk of hepatitis B virus-related hepatocellular carcinoma recurrence following liver resection. JAMA 2012; 308: 1906-14. 8. Wu CY, Kuo KN, Wu MS, et al. Early Helicobacter pylori eradication decreases risk of gastric cancer in patients with peptic ulcer disease. Gastroenterology 2009; 137: 1641-8 e1-2.
SC
9. Wu CY, Lin JT, Ho HJ, et al. Association of Nucleos(t)ide Analogue Therapy With Reduced Risk of Hepatocellular Carcinoma in Patients With Chronic Hepatitis B-A Nationwide Cohort Study. Gastroenterology 2014; 147: 143-51 e5.
M AN U
10. Wu CY, Chan FK, Wu MS, et al. Histamine2-receptor antagonists are an alternative to proton pump inhibitor in patients receiving clopidogrel. Gastroenterology 2010; 139: 1165-71. 11. Cheng CL, Kao YH, Lin SJ, et al. Validation of the National Health Insurance Research Database with ischemic stroke cases in Taiwan. Pharmacoepidemiol Drug Saf 2011; 20: 236-42. 12. Wu CY, Wu MS, Kuo KN, et al. Effective reduction of gastric cancer risk with regular use of nonsteroidal anti-inflammatory drugs in Helicobacter pylori-infected patients. J Clin Oncol 2010; 28: 2952-7.
TE D
13. Grambsch PM, Therneau TM, Fleming TR. Diagnostic plots to reveal functional form for covariates in multiplicative intensity models. Biometrics 1995; 51: 1469-82. 14. Dan YY, So JB, Yeoh KG. Endoscopic screening for gastric cancer. Clin Gastroenterol Hepatol 2006; 4: 709-16.
EP
15. Gupta N, Bansal A, Wani SB, et al. Endoscopy for upper GI cancer screening in the general population: a cost-utility analysis. Gastrointest Endosc 2011; 74: 610-24 e2.
AC C
16. Dinis-Ribeiro M, Areia M, de Vries AC, et al. Management of precancerous conditions and lesions in the stomach (MAPS): guideline from the European Society of Gastrointestinal Endoscopy (ESGE), European Helicobacter Study Group (EHSG), European Society of Pathology (ESP),and the Sociedade Portuguesa de Endoscopia Digestiva (SPED). Endoscopy 2012;44:74-94. 17. Lassen A, Hallas J, de Muckadell OB. The risk of missed gastroesophageal cancer diagnoses in users and nonusers of antisecretory medication. Gastroenterology 2005;129:1179-86. 18. Yamada H, Ikegami M, Shimoda T, et al. Long-term follow-up study of gastric adenoma/dysplasia. Endoscopy 2004; 36: 390-6. 19. de Vries AC, van Grieken NC, Looman CW, et al. Gastric cancer risk in patients with premalignant gastric lesions: a nationwide cohort study in the Netherlands. Gastroenterology 2008; 134: 945-52. 20. Lee TY, Wang RC, Lee YC, et al. The Incidence of Gastric Adenocarcinoma Among Patients With Gastric Intestinal Metaplasia: A Long-term Cohort Study. J Clin Gastroenterol 2016;50:532-7 16
ACCEPTED MANUSCRIPT 21. Meimarakis G, Winter H, Assmann I, et al. Helicobacter pylori as a prognostic indicator after curative resection of gastric carcinoma: a prospective study. Lancet Oncol. 2006;7:211–22. 22. Marrelli D, Pedrazzani C, Berardi A, et al. Negative Helicobacter pylori status is associated with poor prognosis in patients with gastric cancer. Cancer 2009;115:2071– 80.
AC C
EP
TE D
M AN U
SC
RI PT
23. Fukase K, Kato M, Kikuchi S, et al. Effect of eradication of Helicobacter pylori on incidence of metachronous gastric carcinoma after endoscopic resection of early gastric cancer: an open-label, randomised controlled trial. Lancet 2008;372:392-7.
17
ACCEPTED MANUSCRIPT Table 1. Baseline characteristics of the 3 study cohorts after matching <2Y*
2-5Y*
>5Y*
N=1,286 68.2±12.9
N=1,286 68.2±13.0
N=5,144 68.2±13.0
439 (34.1) 847 (65.9) 2.1±1.7 4.2±3.5 824 (64.1) 628 (48.8) 450 (35.0)
437 (34.0) 849 (66.0) 1.4±0.8 4.0±3.5 836 (65.0) 647 (50.3) 458 (37.7)
279 (21.7) 224 (17.4) 392 (30.5) 135 (10.5) 157 (12.2) 143 (11.1) 14 (1.1) 182 (14.2) 30 (2.3)
P value >.99 0.91 <.001++ <.001 <.001 0.14 0.002 >.99
1,116 (21.7) 896 (17.4) 1,553 (30.2) 554 (10.8) 661 (12.8) 551 (10.7) 68 (1.3) 743 (14.4) 120 (2.3)
425 (33.0) 450 (35.0) 289 (22.5) 122 (9.5)
479 (37.2) 451 (35.1) 252 (19.6) 104 (8.1)
0 (0.0) 0 (0.0) 0 (0.0) 0 (0.0)
169 (13.1) 460 (35.8) 21 (1.6) 71 (5.5) 1.4±2.1
169 (13.1) 466 (36.2) 18 (1.4) 81 (6.3) 1.4±2.0
676 (13.1) 1,866 (36.3) 45 (0.9) 279 (5.4) 0.8±1.8
TE D
EP
AC C
1,777 (34.5) 3,367 (65.5) 0.0±0.0 3.3±3.4 3,745 (72.8) 2,436 (47.4) 2,068 (40.2)
279 (21.7) 224 (17.4) 386 (30.0) 140 (10.9) 162 (12.6) 142 (11.0) 12 (0.9) 200 (15.6) 37 (2.9)
M AN U
Age (mean ± SD) Gender Female Male Number of endoscopy in recent 5 years+ Follow-up years All deaths Curative surgery for gastric cancer Chemotherapy for gastric cancer Comorbidities^ Peptic ulcer disease Gastric ulcer Duodenal ulcer Hypertension Cerebrovascular accident Acute coronary syndrome COPD Liver cirrhosis Diabetes mellitus Renal failure Gastroscopy performed at Medical Centers Regional hospitals Local hospitals Clinics Concomitant medications^ Anti-HP therapy NSAIDs/Aspirin/COX2^ Statins^ Metformin^ Number of prior hospitalizations
RI PT
+
SC
Cohorts#
0.97 0.95 0.82 0.88 0.47 0.54 0.51 <.001++
>.99 0.94 0.03 0.47 <.001
*<2Y cohort: receiving endoscopy between 6-month and 2 years before the index date; 2-5Y cohort: receiving endoscopy more than 2 years and less than or equal to 5 years before index date; >5Y cohort: never receive endoscopy within recent 5 years before index date #
Cohorts were matched in 1:1:4 ratio for the <2Y, 2-5Y and >5Y cohorts by age, curative therapy for gastric cancer, H pylori eradication and propensity scores. +: Age, No. of endoscopy in 5 years; Follow-up time are treated as continuous variables and reported in years with mean ± SD. ++: P value is tested by the Student t-test between <2Y and 2-5Y cohorts only. 18
ACCEPTED MANUSCRIPT ##: Peptic ulcer disease, cerebral vascular diseases, acute coronary syndrome, COPD, liver cirrhosis, diabetes, renal failure, use of statins, use of NSAIDs or aspirin or COXIBs, and use of metformin were included in the propensity score calculation. ^: Medication users refer to patients who used these drugs at least one day per month on average. Numbers in bracket are percentage unless specified otherwise.
AC C
EP
TE D
M AN U
SC
RI PT
Abbreviations: N, number; COPD: chronic obstructive pulmonary disease; NSAIDs, non-steroidal anti-inflammatory drugs
19
ACCEPTED MANUSCRIPT Table 2. Multivariate Cox proportional hazards model analysis for all-cause mortality
Events (Number)
5,144 1,286 1,286 7,716 5,063
3,745 824 836 5,405 3,695
3,711
2,230
3,003 7,716
2,410 5,405
EP
P value
1 0.80 (0.72 to 0.89) 0.83 (0.76 to 0.91) 1.02 (1.02 to 1.02) 1.09 (1.03 to 1.16) 0.52 (0.49 to 0.55)
<.001 <.001 <.001 0.002 <.001
1.55 (1.46 to 1.64) 0.95 (0.92 to 0.99)
<.001 0.019
2,227 685 1,740 755 694 86 870 169 5,405
1.02 (0.97 to 1.08) 1.31 (1.20 to 1.44) 0.88 (0.83 to 0.94) 0.93 (0.85 to 1.01) 1.13 (1.04 to 1.23) 2.00 (1.62 to 2.48) 1.17 (1.07 to 1.28) 1.59 (1.36 to 1.86) 1.01 (1.00 to 1.03)
0.488 <.001 <.001 0.088 0.005 <.001 <.001 <.001 0.036
488 317 171 2,180 42 321
0.57 (0.51 to 0.63) 0.78 (0.69 to 0.89) 0.40 (0.34 to 0.47) 1.30 (1.23 to 1.37) 0.54 (0.39 to 0.73) 0.83 (0.73 to 0.95)
<.001 <.001 <.001 <.001 <.001 0.006
SC
RI PT
HR (95% CI)
M AN U
3,018 829 2,331 980 836 94 1,125 187 7,716
TE D
Endoscopy interval >5Y <2Y 2-5Y Age (each incremental year) Male Curative surgery for gastric cancer Chemotherapy for gastric cancer No. of endoscopy in 5 years* Comorbidities Peptic ulcer disease Cerebrovascular accident Hypertension Acute coronary syndrome COPD Liver cirrhosis Diabetes Renal failure No. of prior hospitalizations Concomitant drugs H pylori eradication Pre-cancer diagnosis Post-cancer diagnosis NSAIDs/Aspirin/COX2 Statins Metformin
Patients (Number)
1,014 568 446 2,792 84 431
AC C
*No. of endoscopy in 5 years: number of endoscopy in the 5 years before index date, analyzed by each incremental number as a continuous variable Abbreviations: HR: hazard ratio; COPD, chronic obstructive pulmonary disease; NSAIDs, non-steroidal anti-inflammatory drugs
20
ACCEPTED MANUSCRIPT Table 3. Multivariate Cox proportional hazards model analysis for gastric cancer-specific mortality Events (Number)
5,144 1,286 1,286 7,716 5,063
2,692 553 567 3,812 2,581
3,711
1,530
3,003 7,716
1,744 3,812
1,014 568 446 2,792 84 431
AC C
EP
P value
1 0.80 (0.71 to 0.91) 0.83 (0.75 to 0.93) 1.01 (1.01 to 1.02) 1.04 (0.97 to 1.12) 0.58 (0.54 to 0.62)
<.001 <.001 <.001 0.259 <.001
1.49 (1.39 to 1.59) 0.96 (0.91 to 1.00)
<.001 0.052
1,566 1,101 445 538 463 57 589 104 3,812
1.04 (0.97 to 1.11) 0.89 (0.82 to 0.96) 1.08 (0.96 to 1.21) 1.05 (0.95 to 1.16) 1.05 (0.94 to 1.17) 1.36 (0.98 to 1.89) 1.08 (0.97 to 1.21) 1.10 (0.88 to 1.36) 1.00 (0.98 to 1.01)
0.268 0.003 0.222 0.332 0.426 0.063 0.167 0.400 0.639
339 223 116 1,521 29 213
0.60 (0.53 to 0.68) 0.79 (0.68 to 0.92) 0.44 (0.37 to 0.53) 1.22 (1.14 to 1.30) 0.64 (0.45 to 0.93) 0.86 (0.73 to 1.01)
<.001 0.002 <.001 <.001 0.018 0.073
SC
RI PT
HR (95% CI)
M AN U
3,018 2,331 829 980 836 94 1,125 187 7,716
TE D
Screening interval >5Y <2Y 2-5Y Age (each incremental year) Male Curative resection for gastric cancer Chemotherapy for gastric cancer No. of endoscopy in 5 years* Comorbidities Peptic ulcer disease Hypertension Cerebrovascular accident Acute coronary syndrome COPD Liver cirrhosis Diabetes Renal failure No. of prior hospitalizations Concomitant drugs H pylori eradication Pre-cancer diagnosis Post-cancer diagnosis NSAIDs/Aspirin/COX2 Statins Metformin
Patients (Number)
*: No. of endoscopy in 5 years: number of endoscopy in the 5 years before index date, analyzed by each incremental number as a continuous variable Abbreviations: HR: hazard ratio; COPD, chronic obstructive pulmonary disease; NSAIDs, non-steroidal anti-inflammatory drugs
21
ACCEPTED MANUSCRIPT FIGURE LEGENDS: Figure 1. Flow diagram of patient’s selection *A case could be excluded because of more than one criterion. Therefore, the total excluded
RI PT
cases in each step may outnumber the sum of case numbers excluded by individual criteria. <2Y cohort: receiving endoscopy between 6 months and 2 years before the index date; 2-5Y cohort:
SC
receiving endoscopy more than 2 years and less than or equal to 5 years before index date; >5Y cohort: never receive endoscopy within recent 5 years of index date.
M AN U
GCA; gastric cancer
Figure 2. Cumulative incidences of all-cause mortality and gastric cancer-specific
TE D
mortality in gastric cancer patients who had prior gastroscopy at different time intervals. A: All-cause mortality. B: Gastric-cancer specific mortality.
EP
Calculation and comparison of cumulative mortalities were conducted using Kaplan-Meier method. (P<0.001 for both all-cause and gastric cancer-specific mortalities when comparing
AC C
<2Y versus >5Y and 2 to 5Y versus >5Y cohorts. P value was not significant when comparing <2Y versus 2-5Y cohorts)
22
ACCEPTED MANUSCRIPT Supplementary Table 1: Baseline characteristics and outcomes of study cohorts before propensity score matching 2-5Y*
Age (mean±SD)
N=3,303
N=1,785
N=14,978
P-value
67.6±13.7
69.0±13.2
67.0±14.2
<.001
Gender
>5Y*
RI PT
<2Y*
Cohort#
0.01
1,110 (33.6)
598 (33.5)
5,379 (35.9)
Male
2,193 (66.4)
1,187 (66.5)
9,599 (64.1)
Number of gastroscopy in 5yrs (mean±SD)
2.3±2.0
1.4±0.8
0.0±0.0
<.001
Follow-up period (yrs, mean±SD)
4.0±3.5
3.8±3.5
3.3±3.4
<.001
1,201 (67.3)
10,742 (71.7)
<.001
842 (47.2)
6,970 (46.5)
0.19
617 (34.6)
6,115 (40.8)
<.001
2,186 (66.2)
Curative surgery for gastric cancer
1,484 (44.9)
Chemotherapy for gastric cancer
1,076 (32.6)
Comorbidities
##
Peptic ulcer diseases Gastric ulcer diseases Duodenal ulcer diseases Hypertension
<.001
906 (27.4)
442 (24.8)
2,261 (15.1)
945 (28.6)
454 (25.4)
1,554 (10.4)
1,335 (40.4)
671 (37.6)
4,370 (29.2)
<.001
544 (16.5)
274 (15.4)
1,712 (11.4)
<.001
Acute coronary syndrome
718 (21.7)
382 (21.4)
1,940 (13.0)
<.001
COPD
638 (19.3)
339 (19.0)
1,648 (11.0)
<.001
221 (6.7)
98 (5.5)
421 (2.8)
<.001
718 (21.7)
344 (19.3)
2,448 (16.3)
<.001
249 (7.5)
111 (6.2)
588 (3.9)
<.001
Anti-HP therapy
894 (27.1)
335 (18.8)
1,460 (9.7)
<.001
AC C
TE D
Cerebrovascular disease
M AN U
Overall death
SC
Female
693 (21.0)
247 (13.8)
637 (4.3)
<.001
201 (6.1)
88 (4.9)
823 (5.5)
<.001
1,268 (38.4)
669 (37.5)
5,363 (35.8)
0.01
79 (2.4)
45 (2.5)
254 (1.7)
0.003
254 (7.7)
130 (7.3)
977 (6.5)
0.04
2.3±3.7
1.9±2.5
0.8±1.8
<.001
Liver cirrhosis Renal failure
EP
Diabetes Concomitant drugs
##
Before GC After GC
^
NSAIDs/Aspirin/COX2 Statins^
Metformin
^
Number of prior hospitalizations
*<2Y: cohort receiving endoscopy screening between 6-month and 2 years after index date; 2-5Y: cohort 23
ACCEPTED MANUSCRIPT receiving endoscopy screening between 3-5 years after index date; >5Y: cohort never receiving endoscopy screening within 5 years after index date #: Cohorts were subsequently matched in 1:1:4 ratio for the <2Y, 2-5Y and >5Y cohorts by age, curative therapy, H pylori eradication and propensity scores in the final analysis.
RI PT
+: Age and No. of endoscopy in 5 yr are treated as continuous variables
##: Peptic ulcer diseases, cerebrovascular diseases, acute coronary syndrome, COPD, liver cirrhosis, diabetes, renal failure, use of statins, use of NSAIDs or aspirin or COXIBs, and use of metformin were
SC
included in the propensity score calculation. ^
M AN U
: Medication users refer to patients who used these drugs at least one day per month on average.
Abbreviations: N, number; No. of endoscopy in 5yrs, number of endoscopy in the 5 years before index date; COPD: chronic obstructive pulmonary disease; pre-GC: H pylori eradication before gastric cancer diagnosis; post-GC: H pylori eradication after gastric cancer diagnosis; NSAIDs, non-steroidal
AC C
EP
TE D
anti-inflammatory drugs
24
ACCEPTED MANUSCRIPT Supplementary Figure Legends Supplementary Figure 1:
RI PT
Cumulative incidences of non-gastric cancer-related mortality in patients who had prior gastroscopy in different time frames.
SC
Supplementary Figure 2:
M AN U
Cumulative incidences of all-cause mortality in patients according to different age groups.
B: 65-80 years (log rank, P<0.001)
TE D
A: younger than 65 years (log rank, P<0.001).
Supplementary Figure 3:
EP
Cumulative incidences of all-cause mortality in gastric cancer patients who had prior
AC C
gastroscopy in different time frames up to 8 years. The original control cohort was further classified into 6 to 8Y cohort (between 5 and 8 years) and >8Y (no endoscopy in recent 8 years) cohort. The 4 groups were again matched in 1:1:1:4 ratio as described previously (log-rank, P < 0.001).
25
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
ACCEPTED MANUSCRIPT Acronyms
HP, Helicobacter pylori HR, hazard ratio
AC C
EP
TE D
M AN U
SC
RCIPD, Registry for Catastrophic Illness Patient Database
RI PT
NHIRD, National Health Insurance Research Database
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT