Low‐dose aspirin and risk of gastric and oesophageal cancer: A population‐based study in the United Kingdom using The Health Improvement Network

Abstract There is increasing interest regarding potential protective effects of low‐dose aspirin against various gastrointestinal cancers. We aimed to quantify the association between use of low‐dose aspirin and risk of gastric/oesophageal cancer using a population‐based primary care database in the UK. Between January 2005 and December 2015, we identified a cohort of 223 640 new users of low‐dose aspirin (75‐300 mg/day) and a matched cohort of nonusers at the start of follow‐up from The Health Improvement Network. Cohorts were followed to identify incident cases of gastric/oesophageal cancer. Nested case‐control analyses were conducted and adjusted odds ratios (ORs) with 95% confidence intervals (CIs) were calculated for current vs nonuse of low‐dose aspirin using logistic regression. Current use was defined as when low‐dose aspirin lasted 0 to 90 days before the index date (event date for cases, random date for controls) and previous duration was ≥1 year. We identified 727 incident cases of gastric cancer and 1394 incident cases of oesophageal cancer. ORs (95% CIs) were 0.46 (0.38‐0.57) for gastric cancer and 0.59 (0.51‐0.69) for oesophageal cancer. The effect remained consistent with no clear change seen between previous duration of low‐dose aspirin use of 1‐3, 3‐5 or >5 years. The reduced risks was seen with 75 mg/day, and effects were consistent in lag‐time analyses. In conclusion, our results indicate that use of low‐dose aspirin is associated with a 54% reduced risk of gastric cancer and a 41% reduced risk of oesophageal cancer as supported by mechanistic data.

analysis of CVD randomised controlled trials (RCTs) and from observational studies that daily low-dose aspirin reduces the risk of colorectal cancer (CRC) by about 30% to 40%. [1][2][3][4][5][6][7] Several lines of evidence indicate that the central mechanism for low-dose aspirin's protective effect against CRC-the permanent inhibition of cyclooxygenase (COX)-I enzyme in platelets, leading to suppression of thromboxane A 2 synthesis and reduced platelet activation-is the same through which its cardiovascular benefits are mediated. 8,9 Although most evidence for low-dose aspirin effects in the prevention of cancer is for CRC, there is increasing interest in the chemoprotective effects of low-dose aspirin against other gastrointestinal cancers, in particular gastric and oesophageal cancer. These effects may be mediated through a similar mechanism of action on the gastrointestinal mucosa as with CRC. 8 Survival rates of these two cancers-the fifth and seventh most commonly diagnosed cancers worldwide, respectively, 10 are poor [11][12][13][14] with diagnosis commonly at advanced stage at presentation. In the United Kingdom (UK), over half of patients have stage III or IV disease at diagnosis 15,16 and only around half are still alive 1 year postdiagnosis. 12,13 Further chemoprotective benefits of low-dose aspirin, in addition to the established benefits in CVD and CRC prophylaxis, could therefore potentially further improve the overall balance of benefits and bleeding risks for low-dose aspirin. In an analysis of three aspirin CVD trials with 20-year posttrial follow-up, [17][18][19] Rothwell et al 20 found that low-dose aspirin use at baseline was associated with a 60% reduction in death from oesophageal cancer and a 30% reduction in death from gastric cancer.
The majority of observational studies similarly support substantial reductions in risk of mortality 3 and incidence 3,21-25 of gastric/ oesophageal cancer with low-dose aspirin use, yet confidence in the existing data would be strengthened by similar findings from further population-based studies with a robust study design to minimise bias and confounding. Furthermore, questions remain regarding the effects of low-dose aspirin duration and dose on gastric/oesophageal cancer risk.

| Study design
Using primary care data from the UK, we conducted a populationbased cohort study involving the follow-up of a cohort of new-users of low-dose aspirin (75-300 mg/day) and a matched cohort of nonusers of low-dose aspirin at the start of follow-up to identify incident cases of gastric and oesophageal cancer. The matching served to minimise bias caused by differences between users and nonusers of low-dose aspirin at the time that treatment was initiated. As exposure to low-dose aspirin can change during a long follow-up period, we subsequently performed a nested case-control analysis to more accurately evaluate the effect of low-dose aspirin exposure over time. Our study design has been implemented in previous studies evaluating the associations between low-dose aspirin use and clinical outcomes. 4

| Data sources
We used The Heath Improvement Network (THIN) primary care database, which is validated and representative of the UK demographic. 28,29 The database contains the longitudinal electronic health records (EHRs) of approximately 6% of the UK population, with approximately 3 million individuals currently alive and registered with a participating general practice. The database captures information recorded by general practitioners (GPs) and other practice staff as part of routine patient care. Patient data are entered using the Read code system-the comprehensive coded thesaurus used by clinicians in the UK. 30 Information from secondary care received via email or letter is also recorded. The Health Improvement Network is a valid data source for studying low-dose aspirin use in the UK because all prescriptions issued by GPs are automatically recorded in the patient's EHR, and while low-dose aspirin is available over-the-counter (OTC) in the UK, we have previously shown that misclassification of lowdose aspirin in THIN due to unrecorded use of OTC low-dose aspirin is minimal. 31

| Identification of the study cohorts
A flowchart depicting the study design is shown in Figure S1. We identified all individuals aged 40 to 89 years in THIN between January 1, 2005 and December 31, 2015 (study entry period) with at least 2 years' registration with their GP and at least 1 year of recorded prescription history. The date an individual met these eligibility criteria was the study entry date. Individuals with a prescription for low-dose aspirin (n = 417 256) or with a record of cancer before this date

What's new?
Low-dose aspirin may help protect against the development of gastric and oesophageal cancers. This population-based study using data from primary care electronic health records in the United Kingdom lends further support to that idea.
Compared with no use of low-dose aspirin, daily use of lowdose aspirin (75-300 mg) for at least one year was associated with a 54 percent reduction in gastric cancer risk and a 41 percent reduction in oesophageal cancer risk. No clear difference in effect was observed between low-dose aspirin use for one to three years versus more than three years.
(n = 223 274) were excluded. We subsequently identified patients (n = 223 640) with a first prescription for low-dose aspirin (75-300 mg/day; start date) and matched each 1:1 to a nonuser of low-dose aspirin on their start date by age, sex, time since study entry, and number of GP visits in the previous year. This process resulted in two cohorts: new users of low-dose aspirin and nonusers of low-dose aspirin at the start of follow-up. The start date for a member of the latter cohort was the start date of their matched partner in the lowdose aspirin cohort. All members of both cohorts were assigned to either the primary or secondary CVD prevention population based on whether they had a Read code for CVD before the start date, as described previously. 32

| Follow-up and outcome identification
We performed two independent follow-ups, one to identify first time cases of gastric cancer and the other to identify first time cases of oesophageal cancer. Individuals were followed until the earliest of the following endpoints: a recorded diagnosis of gastric/oesophageal cancer (based on codes specific for these cancers; see Tables S1 and S2 for Read codes), a recorded diagnosis of another cancer, death or the end of follow-up (31 December 2017). A patient was designated as confirmed/unconfirmed case following a stepwise process (Supporting Information Methods) involving manual review of patient records. A fatal case was defined as death from any cause within 1 year after the index date. The index date for all confirmed cases was the date of the recorded diagnosis of gastric/oesophageal cancer.

| Selection of controls
Controls were selected from both the low-dose aspirin and matched nonuser study cohorts using incidence density sampling and were frequency matched to cases by age, sex and calendar year (n = 5000 each for cases of both cancers). The index date for controls was a random date during the individual's observation period.

| Low-dose aspirin exposure
The length of supply in days of a low-dose aspirin prescription was calculated as the number of tablets prescribed divided by the prescribed daily posology. Although we did not have information on adherence to treatment, we assumed that patients took their medication daily as prescribed. Current use of low-dose aspirin among cases and controls was defined as when supply of the most recent prescription before the index date lasted until/over the index date or ended 0 to 30 days before the index date. In our analysis of "current use", we required current users to have at least 1 year of low-dose aspirin use because those with short durations (ie, those who recently initiated low-dose aspirin) of use may be prone to confounding by contraindication; early signs of gastric cancer manifesting as gastrointestinal symptoms that may cause the GP to withhold prescribing low-dose aspirin. Remaining current users were those with a treatment duration of less than 1 year. Other categories of low-dose aspirin exposure were recent use, when supply of the most recent prescription ended 31 to 365 days before the index date; past use was when supply of the most recent prescription ended ≥365 days before the index date, and nonuse, when there was no recorded use at any time before the index date. Duration of low-dose aspirin was calculated by summing the duration of individual prescriptions before the index date. Lowdose aspirin dose was computed based on the posology of the last prescription before the index date. H. pylori infection (as a separate variable to the proxy variable for H. pylori described above) was identified any time before the index date.

| Statistical analysis
For both gastric and oesophageal cancers, we performed nested casecontrol analyses using cases and controls from both cohorts to estimate the association between use of low-dose aspirin (as well as the demographics, lifestyle factors, comorbidities and co-medication variables as described above) with the risk of gastric/oesophageal cancer.
Odds ratios (ORs) with 95% confidence intervals (CIs) were calculated using unconditional logistic regression adjusted for confounders (smoking, number of GP visits in the year before the index date and low-dose aspirin [in analyses of all other variables] in addition to the matching factors of age, sex and calendar year). We used a stepwise approach adding potential confounders to the unadjusted model, retaining those that changed the estimate by at least 10%. As controls were selected using incidence density sampling, the ORs were regarded as an unbiased estimate of the incidence rate ratio. Stratified Odds ratios (95% CIs) for the association between patient characteristics and the risk of gastric cancer

| Baseline characteristics of cases and controls
Baseline characteristics of cases and controls are shown in Table 1 for gastric cancer and Table 2

T A B L E 2
Odds ratio (95% CIs) for the association between patient characteristics and the risk of oesophageal cancer

| Low-dose aspirin use and risk of gastric cancer
Associations between low-dose aspirin and risk of gastric cancer are shown in Table 3

| Low-dose aspirin use and risk of oesophageal cancer
Associations between low-dose aspirin and risk of oesophageal cancer are shown in Table 4. Compared to nonuse of low-dose aspirin, current use of low-dose aspirin was associated with a 41% reduced risk of oesophageal cancer (OR 0.59, 95% CI: 0.51-0.69).
Findings relating to dose, duration and primary/secondary CVD prevention were similar to those seen for gastric cancer, including in the lag-time analyses (Table S4) Note: Comorbidities were identified any time before the index date except for anaemia, where we included recorded diagnoses in the year before the index date, and pernicious anaemia, where we included recorded diagnoses in the 5 years before the index date. Odds ratios (95% CIs) for the association between low-dose aspirin use and the risk of gastric cancer  4,5,8 We observed the substantial reduction in gastric/oesophageal cancer risk in both primary and secondary CVD prevention populations, as we have also shown previously for CRC. 4,5 Whether increasing evidence in support of low-dose aspirin for chemoprophylaxis of gastric and oesophageal cancer in addition to CRC will further improve the overall balance of benefits and harms toward favouring low-dose aspirin in primary CVD prevention is a question that will be answered in time. Expert opinion and clinical guidelines on this topic have been evolving in recent years as new robust data on cancer, CVD and bleeding outcomes emerge. [36][37][38][39][40] Incorporating data on gastric/oesophageal cancer prophylaxis into benefit-risk assessments of low-dose aspirin for both CVD and cancer prevention would require greater consensus around any duration of effects. Data applicable to patient subgroups would also be beneficial. Interestingly, the ASCEND RCT of 100 mg daily aspirin vs placebo among individuals with diabetes but no evident CVD found no significant difference in the incidence of gastrointestinal cancers as a group after 7.4 years' follow-up. However, the trial only had 60% power to detect the 30% hypothesised difference between study arms, 41 and results from longer-term follow-up are awaited.
Strengths of our study include the large sample representative of the population from which it was drawn, meaning our results are generalizable to the UK population. The study population reflected users and nonusers of low-dose aspirin in clinical practice including the elderly and those with multiple comorbidities. Our matched cohort design helped to minimise bias from differences between low-dose aspirin users and nonusers at the start date. Nested case-control analyses enabled low-dose aspirin exposure to be accurately assessed, and lag-time analyses found our findings to be robust. We explored many potential confounders with adjustment made for variables found to be confounders in our analyses, yet we acknowledge that residual confounding cannot be completely ruled out, for example, Odds ratio (95% CIs) for the association between low-dose aspirin and the risk of oesophageal cancer from unknown confounders. The stepwise process for case confirmation involved manual review of patients' EHRs, and although we did not validate our cases of gastric/oesophageal cancer through linkage to hospital data, in a previous study for CRC restricted to individuals also in the Hospital Episode Statistics (HES) database, the false negative rate for CRC (using HES as gold standard) was only 6.1%. 4 Hence, it is likely that THIN is also a valid data source to identify cases of gastric/oesophageal cancer. Linkage to UK cancer registry data was not possible because this is not currently available for THIN. We were unable to evaluate associations according to histologic type of cancer or stage because this information is not systematically recorded in THIN. Although an increased risk of both gastric and oesophageal cancer was seen among low-dose aspirin users with concomitant PPI use, this finding should be interpreted very cautiously because of the likelihood of protopathic bias. Proton pump inhibitors are commonly prescribed for acid-reflux or similar symptoms, which are also early indicators of these two gastrointestinal cancers.
Our findings add to the existing evidence that low-dose aspirin is associated with a substantial reduction in the risk of gastric/ oesophageal cancer in the general population.