Young-Il Kim and So Young Kim contributed equally to this work.
The aim of this study was to investigate whether aspirin use can reduce the incidence of gastric cancer in patients with hypertension or type 2 diabetes.
A total of 200,000 patients with hypertension or type 2 diabetes were randomly selected from the Korean National Health Insurance claim database. Of these, 3,907 patients who used 100 mg of aspirin regularly (regular aspirin users) and 7,808 patients who did not use aspirin regularly (aspirin non-users) were selected at a frequency of 1:2, matched by age, sex, comorbid illnesses (type 2 diabetes and hypertension), and observation periods. The incidence of gastric cancer in this cohort was then assessed during the observation period of 2004 to 2010.
In the matched cohort, the incidence rates of gastric cancer were 0.8% (31/3,907) for regular aspirin users and 1.1% (86/7,808) for aspirin non-users, but the cumulative incidence rates were not significantly different between groups (p=0.116, log-rank test). However, in multivariate analysis, regular aspirin users had a reduced risk of gastric cancer (adjusted hazard ratio [aHR], 0.71; 95% confidential interval [CI], 0.47 to 1.08; p=0.107). Duration of aspirin use showed significant association with reduction of gastric cancer risk (aHR for each year of aspirin use, 0.85; 95% CI, 0.73 to 0.99; p=0.044), particularly in patients who used aspirin for more than 3 years (aHR, 0.40; 95% CI, 0.16 to 0.98; p=0.045).
Long-term low-dose aspirin use was associated with reduced gastric cancer risk in patients with hypertension or type 2 diabetes.
Aspirin has been widely used as an anti-platelet drug for the primary or secondary prevention of cardiovascular (CV) events, including ischemic heart disease and stroke. Guidelines recommend prescription of low-dose aspirin for hypertensive patients with history of or at high risk for CV events [
In addition to its preventive effects against CV events, aspirin has been widely investigated as an anti-cancer drug since the first report of its reduction of colorectal cancer incidence [
Thus, we investigated the association between long-term, low-dose aspirin use and gastric cancer incidence using data from a large cohort of patients with hypertension or type 2 diabetes obtained from the Korean National Health Insurance (KNHI) claim database.
Because a large number of regular aspirin users were necessary to achieve statistical power, only patients with hypertension or type 2 diabetes were considered; use of regular aspirin was above average for both groups due to their increased risk of CV events. The KNHI Corporation provided data on 200,000 patients randomly sampled from all claimants with a diagnosis of hypertension or type 2 diabetes (100,000 patients with hypertension and 100,000 with type 2 diabetes) from its database using SAS ver. 9.2 software (SAS Institute Inc., Cary, NC). Using drug codes, patients prescribed 100 mg of commercially available aspirin products for at least 6 consecutive months were selected. A control group of aspirin non-users, defined as patients who had never made claims for aspirin prescription or those who had claimed payments for less than 6 months of aspirin prescriptions was also identified. However, to avoid the possible confounding effects of NSAIDs, patients who had claims for regular use of NSAIDs (6 or more consecutive months) were excluded from both groups. To match aspirin non-users with regular aspirin users, a frequency-matched sampling with regular aspirin users was performed, grouped by the distribution of hypertension and type 2 diabetes, age, sex, and starting time of observation periods and death at a ratio of 1:2 (regular aspirin users: aspirin non-users). Matched sampling enabled selection of units from a large reservoir of potential controls to produce a control group of modest size with covariate distributions similar to that of the focal case group [
Aspirin use or non-use from 1 January 2004, to 31 December 2010 was investigated. For aspirin users, the observation period for gastric cancer was started only when patients had claimed payments for aspirin use for 6 consecutive months. Those observation periods were continued until 31 December 2010 (
This study was approved by the Institutional Review Board of the National Cancer Center, Korea (NCCRE-11-003). Informed consent was waived because the study was based only on routinely collected administrative data.
All data used in the current study were obtained from the KNHI claim database. In Korea, medical services are provided by healthcare providers on a fee-for-service basis, with reimbursement by the KNHI Corporation, a single payer, after a claim has been submitted. Therefore, all data necessary for reimbursement for the claimed payments of medical services were recorded in the KNHI claim database. Those data included patients’ socio-demographic information, including sex, age, residential area, disease for which payment is claimed, costs incurred, comorbid diseases, a detailed list of diagnostic tests, procedures, and prescription provided, and outcomes (deaths). Codes from The International Classification of Diseases 10th edition [
Descriptive analyses were performed in order to clarify the distributions of regular aspirin users and aspirin non-users; the chi-square test was used for categorical variables and Student’s t test for continuous variables. The Kaplan-Meier method for life-table estimates and the log-rank test were used for comparison of the incidence of gastric cancer during the observation periods. For regular aspirin users, the duration of aspirin use was calculated as the time between the beginning and end of continuous payment claims for aspirin. Time to gastric cancer occurrence was defined as a period from the first date of observation to the date of gastric cancer diagnosis. All patients were followed until 31 December 2010, except for 1,653 patients who were censored because of death before that date (566 patients [14.5%] in regular aspirin users vs. 1,087 patients [13.9%] in aspirin non-users; p=0.407).
A multivariate Cox proportional hazards regression analysis was performed to evaluate the effects of aspirin use on the incidence of gastric cancer. Before the analysis, the assumption of proportionality was confirmed by plotting the log (-log) hazard estimates against observation periods. In addition, Schoenfeld residuals for each covariate used for adjustment indicated that the assumption for proportional hazards was met (p > 0.10 for all covariates). For the multivariate Cox proportional hazards model, the effects of aspirin use were evaluated using three different models. In model 1, status of aspirin use was incorporated as a categorical variable and evaluated according to the regular use of aspirin (regular aspirin users vs. aspirin non-users). In model 2, duration of aspirin use was incorporated as a continuous variable, an approach that meets the assumption of linearity. In model 3, duration of aspirin use was divided into four groups (0.5-1.0 year, 1.1-2.0 years, 2.1-3.0 years, and > 3.0 years) and incorporated as a categorical variable. All statistical analyses were performed using SAS ver. 9.2 software (SAS Institute Inc.). The criterion for statistical significance was p < 0.05.
The median age of all included patients was 64 years (interquartile range [IQR], 56 to 70 years) and the percentage of male patients was 50.5%. The proportions of patients with hypertension and type 2 diabetes were 56.9% and 43.1%, respectively. The mean duration of aspirin use was 2.3 years for regular aspirin users. No significant differences in age, sex, or comorbidities were observed between regular aspirin users and non-users. However, higher rates of urban residency were observed for regular aspirin users compared with aspirin non-users (45.0% vs. 42.2%, respectively; p=0.014). Detailed characteristics of cohort patients are summarized in
During the observation periods (median, 6.4 years; IQR, 4.6 to 7.0 years), 117 patients were diagnosed with gastric cancer (31 patients who were regular aspirin users and 86 who were aspirin non-users). The rate of diagnosis with gastric cancer was lower for regular aspirin users (0.8%) than aspirin non-users (1.1%), but without statistical significance (p=0.114) (
Univariate and multivariate analyses were performed to examine the effects of regular aspirin use on gastric cancer incidence in the cohort. In model 1, regular aspirin users tended to have reduced gastric cancer risk in univariate (crude hazard ratio [cHR], 0.72; p=0.117) and multivariate analysis (adjusted HR [aHR], 0.71; p=0.107), but these results were not statistically significant. However, results of multivariate analysis of the duration of aspirin use showed a significant reduction of gastric cancer risk. In model 2, longer duration of aspirin use showed association with a reduced risk of gastric cancer in univariate (cHR for each year of aspirin use, 0.84; 95% confidence interval [CI], 0.72 to 0.98; p=0.028) and multivariate analyses (aHR for each year of aspirin use, 0.85; 95% CI, 0.73 to 0.99; p=0.044). In model 3, regular aspirin users who used aspirin for more than 3 years had significantly reduced risk of gastric cancer in univariate (cHR, 0.37; 95% CI, 0.15 to 0.90; p=0.029) and multivariate analyses (aHR, 0.40; 95% CI, 0.16 to 0.98; p=0.045) (
The results of univariate and multivariate analyses for other covariates are shown in
Historically, studies of the chemopreventive effects of aspirin on gastrointestinal cancers have focused on its prevention of colorectal cancer. In recent meta-analyses of both case-control and cohort studies aspirin use was protective against colorectal cancer risk [
Case-control [
In contrast, our study has several comparative strengths. First, only patients who were prescribed 100 mg aspirin were included, and patients prescribed 300-500 mg aspirin or NSAIDs were excluded. Second, our study showed long-term effects of regular aspirin use on gastric cancer incidence, especially when continued for more than 3 years. Our analyses might be sensitive to this phenomenon because we carefully defined aspirin users as patients who made claims for aspirin for at least 6 consecutive months; previous studies have not made this distinction.
Duration and continuity of aspirin use are crucial variables when examining the effect of aspirin on gastrointestinal cancers because several years are necessary for the clinical detection of disease [
The main mechanism of the cancer-preventing properties of aspirin and NSAIDs is associated with the blockade of the COX pathway, which influences various cellular processes, including inflammation, thrombosis, angiogenesis, apoptosis, and cell proliferation and migration [
Previous studies on the short-term use of aspirin as a chemotherapeutic agent showed that its activity in gastric cancer depends on anatomic location,
The current study has several limitations. First, this study is a retrospective cohort study. Although the results were derived from the analyses of a large, matched population, the possibility of selection bias from the KNHI claim database was unavoidable. Second, our findings may not be applicable to other geographic areas; the incidence of gastric cancer in Korea is the highest in the world [
In conclusion, among Korean patients with hypertension or type 2 diabetes, long-term, low-dose aspirin use was associated with a reduced risk of gastric cancer. Administration of low-dose aspirin may be a candidate regimen for prevention of gastric cancer in areas where its incidence is high. However, regarding its use as a chemopreventive agent, further studies evaluating the risk of adverse events may be needed.
Conflict of interest relevant to this article was not reported.
This work was supported by grant 1510530 from the National Cancer Center, Korea.
Study flowchart. Patients with hypertension or type 2 diabetes (100,000 patients each) were classified as aspirin users or non-users. Non-users were matched with users based on demographic and comorbidity covariates. KNHI, Korean National Health Insurance; NSAIDs, non-steroidal anti-inflammatory drugs.
Observation periods of aspirin users. Claims for aspirin use were investigated in all cohort patients from 2004-2010. For aspirin users, observation periods for gastric cancer began after 6 consecutive months of payment claims for aspirin and ended on 31 December 2010. Non-user observation periods began immediately after any claim was made for hypertension or diabetes. Duration of observation was matched between users and non-users to prevent sampling bias.
Cumulative incidence of gastric cancer according to aspirin use. The cumulative gastric cancer incidence rate in aspirin users was not statistically different from that of non-users (p=0.116, log-rank test).
Baseline characteristics of the study population
Characteristic | Regular aspirin users (n=3,907) | Aspirin non-users (n=7,808) | p-value |
---|---|---|---|
64 (56-70) | 64 (56-70) | 0.765 | |
Male | 1,980 (50.7) | 3,935 (50.4) | 0.774 |
Female | 1,927 (49.3) | 3,873 (49.6) | |
Metropolitan | 2,148 (55.0) | 4,513 (57.8) | 0.014 |
Urban | 1,168 (29.9) | 2,198 (28.2) | |
Rural | 591 (15.1) | 1,097 (14.1) | |
Hypertension | 2,208 (56.5) | 4,452 (57.0) | 0.603 |
Type 2 diabetes | 1,699 (43.5) | 3,356 (43.0) | |
0-0.49 | 0 | 7,808 (100) | < 0.001 |
0.50-1.00 | 1,211 (31.0) | - | |
1.01-2.00 | 1,038 (26.6) | - | |
2.01-3.00 | 586 (15.0) | - | |
> 3.00 | 1,072 (27.4) | - | |
Mean±standard deviation | 2.323±1.854 | 0.003±0.031 | < 0.001 |
6.4 (4.6-7.0) | 6.4 (4.6-7.0) | 0.977 | |
31 (0.8) | 86 (1.1) | 0.114 |
Values are presented as number (%) unless otherwise indicated. IQR, interquartile range.
Cumulative incidence rate of gastric cancer according to each year of observation periods
Regular aspirin users (n=3,907) |
Aspirin non-users (n=7,808) |
p-value | |||
---|---|---|---|---|---|
Cumulative gastric cancer cases | Cumulative rate of gastric cancer (95% CI, %) | Cumulative gastric cancer cases | Cumulative rate of gastric cancer (95% CI, %) | ||
FU year | |||||
At 1 yr | 3 | 0.08 (0.03-0.24) | 13 | 0.17 (0.10-0.29) | 0.028 |
At 2 yr | 10 | 0.27 (0.14-0.49) | 25 | 0.33 (0.22-0.48) | |
At 3 yr | 15 | 0.41 (0.25-0.68) | 40 | 0.54 (0.40-0.74) | |
At 4 yr | 19 | 0.53 (0.34-0.83) | 56 | 0.79 (0.61-1.02) | |
At 5 yr | 22 | 0.63 (0.42-0.96) | 65 | 0.94 (0.73-1.19) | |
At 6 yr | 28 | 0.88 (0.60-1.27) | 77 | 1.17 (0.93-1.46) | |
At 7 yr | 31 | 1.05 (0.73-1.51) | 86 | 1.41 (1.13-1.75) |
CI, confidence interval; FU, follow-up.
Univariate Cox-proportional hazard regression analysis.
Univariate and multivariate Cox proportional hazard model for incidence of gastric cancer
Variable | No. | Person-years | Unadjusted |
p-value | Adjusted for Covariates |
p-value | ||
---|---|---|---|---|---|---|---|---|
cHR | 95% CI | aHR | 95% CI | |||||
Male | 5,915 | 32,877 | 1.00 | 1.00 | ||||
Female | 5,800 | 32,054 | 0.39 | 0.26-0.58 | < 0.001 | 0.39 | 0.26-0.59 | < 0.001 |
11,715 | 64,931 | 1.02 | 1.00-1.04 | 0.066 | 1.01 | 0.99-1.03 | 0.512 | |
Metropolitan | 6,661 | 36,954 | 1.00 | 1.00 | ||||
Urban | 3,366 | 18,622 | 1.10 | 0.71-1.69 | 0.676 | 1.07 | 0.69-1.64 | 0.773 |
Rural | 1,688 | 9,355 | 1.84 | 1.17-2.91 | 0.009 | 1.82 | 1.15-2.88 | 0.011 |
Hypertension | 6,660 | 37,351 | 1.00 | 1.00 | ||||
Type 2 diabetes | 5,055 | 27,580 | 1.38 | 0.96-1.98 | 0.081 | 1.26 | 0.87-1.81 | 0.219 |
Model 1 (regular aspirin users) | 11,715 | 64,931 | 0.72 | 0.48-1.09 | 0.117 | 0.71 | 0.47-1.08 | 0.107 |
Model 2 (continuous variable, yr) | 11,715 | 64,931 | 0.84 | 0.72-0.98 | 0.028 | 0.85 | 0.73-0.99 | 0.044 |
Model 3 (categorical variable) | ||||||||
0.5-1.0 yr | 1,211 | 6,134 | 0.91 | 0.49-1.71 | 0.769 | 0.82 | 0.44-1.54 | 0.541 |
1.1-2.0 yr | 1,038 | 5,489 | 0.74 | 0.36-1.52 | 0.410 | 0.71 | 0.34-1.46 | 0.349 |
2.1-3.0 yr | 586 | 3,234 | 1.09 | 0.51-2.36 | 0.823 | 1.11 | 0.52-2.41 | 0.783 |
> 3.0 yr | 1,072 | 6,802 | 0.37 | 0.15-0.90 | 0.029 | 0.40 | 0.16-0.98 | 0.045 |
cHR, crude hazard ratio; CI, confidence interval; aHR, adjusted hazard ratio.