Risk of uncomplicated peptic ulcer disease in a cohort of new users of low-dose acetylsalicylic acid for secondary prevention of cardiovascular events

A retrospective cohort study was performed using data from The Health Improvement Network (THIN), a computerized primary care database containing anonymized records for individuals currently registered with participating primary care practices in the UK. THIN is age, sex and geographically representative of the UK population [17] and has been extensively validated for use in epidemiological studies [18]. Data recorded in THIN include patient demographics, details of consultations with PCPs, information about consultant referrals and hospitalizations, laboratory test results, diagnoses and prescriptions. The Read classification is used to code specific diagnoses [19],[20], and a drug dictionary based on data from the Gemscript classification is used to record prescriptions [21]. Ethical approval for the collection of data in THIN database was obtained from a Multicentre Research Ethics Committee (NHS; MREC reference number: 08/H0305/49).

A previously identified cohort of patients who received a first-ever prescription for low-dose ASA (75-300 mg/day) for secondary prevention of cardiovascular ischaemic diseases or cerebrovascular ischaemic events and who were aged 50-84 years between 1 January 2000 and 31 December 2007 was used for the present study [11]. Patients were required to have been registered for at least 2 years with their PCP, and to have at least 1 year of computerized prescription history. The date of their first-ever recorded prescription for low-dose ASA for secondary prevention of cardiovascular events was defined as their start date. All patients who had a prescription for low-dose ASA recorded before their start date were excluded, as were patients with a recorded diagnosis of alcohol abuse or cancer. Individuals aged 70 years or older with a follow-up longer than 1 year and less than two health contacts during their follow-up period were also excluded (proxy for incomplete data recording). The total study cohort consisted of 38,975 individuals, who were followed up to identify incident cases of uncomplicated PUD.

All study cohort members were followed up for a mean of 5.6 years from the first day after their start date to the earliest of the following endpoints: first recorded diagnosis of uncomplicated PUD, cancer, alcohol abuse or alcohol-related disease; reaching the age of 85 years; date of the last practice data collection; death; or end of the study period (30 September 2011).

To confirm the validity of our case ascertainment further, we sent a questionnaire to the corresponding PCPs requesting confirmation and copies of paper-based records for 100 patients randomly sampled from the definite cases (n = 96) and possible cases (n = 4). We received records for 98 patients. The uncomplicated PUD diagnosis was confirmed by the PCPs for 76 patients. The confirmation rate among the definite cases was 80% and only 25% among possible cases. We retained as cases all definite cases confirmed by a questionnaire (n = 75) and those definite cases for which we did not have a questionnaire (n = 233). Due to the low confirmation rate among possible cases, we only retained the single patient initially classified as possible, whose diagnosis was confirmed by the PCP. The majority of patients who were not retained as uncomplicated PUD cases had a discharge letter with a diagnosis of a complication (e.g. bleeding) not recorded in their computerized file.

Following this two-step review process, 309 patients were considered to be incident cases of uncomplicated PUD: 308 individuals from those initially classified as definite and 1 from the possible cases (Figure 1). The index date was defined as the date of the computer-recorded diagnosis (n = 144) or the date of the first symptom leading to the diagnosis of PUD (n = 165), whichever occurred first. When the date of the first symptom was used as the index date, the mean time to the computer-recorded diagnosis of uncomplicated PUD was 35 days. The peptic ulcer was located in the stomach for 188 patients (61%), the duodenum for 103 (33%) and multiple sites (stomach and duodenum) for 18 (6%).

A date within the study period was generated at random for each member of the study cohort, excluding PUD cases. If the random date was included in the individual’s person-time contribution, we marked that person as an eligible control. The random date for each control individual was used as the index date in the nested case–control analysis. In total, 2000 controls, frequency-matched to cases by age (within 1 year), sex and follow-up time (interval between start date and index date), were randomly selected from the pool of eligible controls.

Information on patient demographics, baseline characteristics, comorbidities and comedications was collected from THIN. In the cohort analyses, we investigated the following potential risk factors: age, sex, year of start date, ASA indication (myocardial infarction, unstable angina, ischaemic heart disease and cerebrovascular disease), history of PUD and use of proton pump inhibitors (PPIs). A history of PUD any time before the start date was defined as any record of peptic ulcer symptoms, or a diagnosis of uncomplicated or complicated PUD. We defined PPI users at their start date as those who had received at least one prescription for a PPI in the month before or in the week after their first ASA prescription (start date).

For nested case–control analyses, information on comorbidities and other potential risk factors, such as smoking, alcohol use and body mass index (BMI), were collected from patients’ records at any time before the index date. We assessed gastroesophageal reflux disease (GERD) and symptoms related to PUD, including vomiting, nausea, epigastric pain, dyspepsia, heartburn and gastritis between the start date and the index date. In addition, data on the number of PCP visits, referrals and hospitalizations recorded for each patient were collected for the year before their index date. Finally, for the case–control analyses, information on patients’ drug exposure was assessed between their start date and index date, and was categorized as follows: current use, when the supply of the most recent prescription lasted until the index date or ended in the 30 days preceding the index date; past use, when the supply of the most recent prescription ended 31-365 days before the index date; and non-use, when the most recent prescription ended more than 365 days before the index date or there was no recorded use at any time between the start date and the index date.

The overall incidence of uncomplicated PUD and associated 95% confidence interval (CI) was determined along with age- and sex-specific estimates. We also calculated the incidence of uncomplicated PUD in subgroups of ASA users with and without a history of PUD before their start date. The incidence of uncomplicated PUD in new users of low-dose ASA who were exposed to a PPI at their start date was also compared with the incidence in those who were not exposed to a PPI at their start date. Nelson–Aalen cumulative hazard estimates were calculated for ASA users with and without a history of PUD and compared using a log-rank test. Hazard ratios (HRs) and associated 95% CIs were calculated using Cox regression analyses adjusted for age, sex, year of start date, ASA indication, PPI use and history of PUD. All variables were ascertained at the start date.

Nested case–control analyses were performed to estimate the contribution of various risk factors to the development of uncomplicated PUD during follow-up. Odds ratios (ORs) and associated 95% CIs were calculated by unconditional multiple logistic regression models. All estimates were adjusted for frequency-matched variables (age, sex, follow-up time) and for health service utilization (PCP visits and referrals), smoking, and use of acid-suppressing drugs, non-steroidal anti-inflammatory drugs (NSAIDs), ASA and paracetamol.

Statistical analyses were performed using Stata® version 12.0 (StataCorp LP, College Station, TX, USA).

The crude incidence of uncomplicated PUD in new users of low-dose ASA was 1.41 per 1000 person-years (95% CI, 1.26-1.58). For women and men, the incidences were 1.56 per 1000 person-years (95% CI, 1.33-1.83) and 1.30 per 1000 person-years (95% CI, 1.11-1.52), respectively. When stratified by age, women aged 50-59 years had a higher incidence of uncomplicated PUD (2.19 per 1000 person-years, 95% CI, 1.43-3.34) than men in the corresponding age group (1.16 per 1000 person-years, 95% CI, 0.77-1.76). For patients aged 80-84, the opposite was observed (Figure 2). A higher incidence of uncomplicated PUD was observed in new users of ASA with a history of PUD (3.03 per 1000 person-years, 95% CI, 2.26-4.08) than in those without such a history (1.30 per 1000 person-years, 95% CI, 1.15-1.46). In a Nelson–Aalen cumulative hazards analysis, patients with a history of PUD had a significantly greater probability of developing uncomplicated PUD than patients without such a history (log-rank test, p < 0.001; Figure 3). New users of ASA exposed to a PPI at their start date had a higher incidence of uncomplicated PUD (1.84 per 1000 person-years, 95% CI, 1.45-2.32) than those not exposed (1.32 per 1000 person-years, 95% CI, 1.16-1.50).

In an adjusted Cox regression analysis, sex, year of start date, ASA indication and use of PPIs did not appear to be associated with an increased risk of uncomplicated PUD (Table 1). Older patients (aged 80-84 years) were more likely to develop uncomplicated PUD than patients aged 50-59 years (adjusted HR, 1.69; 95% CI, 1.05-2.74). New users of low-dose ASA with a history of PUD were about twofold more likely to develop uncomplicated PUD than those without such a history (adjusted HR, 2.22; 95% CI, 1.60-3.09).

In nested case–control analyses, smoking was a predictor of uncomplicated PUD development (adjusted OR, 1.96; 95% CI, 1.37-2.80) (Table 2). We further examined the association between smoking and uncomplicated PUD according to the ulcer site (stomach or duodenum). Smokers had a threefold greater risk of duodenal ulcer than non-smokers (adjusted OR, 3.40; 95% CI: 1.94-5.97), whereas the risk of gastric ulcer was increased by less than twofold (adjusted OR, 1.64; 95% CI, 1.04-2.59). The risk of developing uncomplicated PUD was twofold greater for patients with four or more referrals to a specialist during the year preceding their index date than for those who had not been referred (adjusted OR, 1.98; 95% CI, 1.30-3.00). Individuals living in urban areas were more likely to experience uncomplicated PUD than those living in rural areas (adjusted OR, 3.21; 95% CI, 1.46-7.04). Patients living in the most deprived areas (as measured by the Townsend deprivation index) had an increased risk of uncomplicated PUD compared with those living in the least deprived areas (adjusted OR, 1.56; 95% CI, 1.01-2.39). Alcohol use and BMI did not have a statistically significant effect on the risk of uncomplicated PUD. Among the comorbidities investigated, an increased risk of uncomplicated PUD in new users of low-dose ASA was associated with anaemia (adjusted OR, 2.53; 95% CI, 1.82-3.53), stress (adjusted OR, 1.58; 95% CI, 1.06-2.33) and depression (adjusted OR, 1.38; 95% CI, 1.04-1.83). The presence of uncomplicated PUD symptoms (e.g. nausea, vomiting and epigastric pain) after initiation of ASA treatment was a predictor of uncomplicated PUD (adjusted OR, 2.09; 95% CI, 1.56-2.81), whereas a diagnosis of GERD during follow-up was not (adjusted OR, 1.13; 95% CI, 0.81-1.57).

Overall, current use of NSAIDs (selective cyclooxygenase-2 [COX-2] inhibitors and traditional NSAIDs) was associated with a significantly increased risk of uncomplicated PUD (adjusted OR, 1.50; 95% CI, 1.06-2.13). When considered separately, the association was significant for COX-2 inhibitors (adjusted OR, 2.33; 95% CI, 1.13-4.79) but was not for traditional NSAIDs (adjusted OR, 1.36; 95% CI, 0.93-1.99). Current use of oral steroids and current use of paracetamol were also associated with an increased risk of PUD (adjusted OR, 1.88; 95% CI, 1.09-3.25 and adjusted OR, 1.45; 95% CI, 1.08-1.95, respectively) (Table 3). An association between acid-suppressing drugs and uncomplicated PUD was observed: current users of PPIs and current users of histamine-2 receptor antagonists (H₂RAs) were significantly more likely to have uncomplicated PUD than non-users of these drugs (adjusted OR, 1.68; 95% CI, 1.26-2.25 and adjusted OR, 2.26; 95% CI, 1.34-3.82, respectively). When current users of PPI therapy were divided into subgroups, current use of PPI therapy initiated more than 30 days after the start date was significantly associated with an increased risk of uncomplicated PUD (adjusted OR, 2.38; 95% CI, 1.65-3.42), whereas current use of PPI intitiated before the start date or within the 30 days after the start date was not significantly associated (adjusted OR, 1.25; 95% CI, 0.86-1.80).

In a secondary analysis, regression models were computed only for current users of low-dose ASA at the index date. Estimates of risks for the development of uncomplicated PUD in current low-dose ASA users were comparable to those observed for the whole cohort for all variables (Additional file 1: Table S1 and Additional file 2: Table S2).

We used The Health Improvement Network (THIN) primary care data for this study. The company that owns THIN (Cegedim Strategic Data Medical Research) has received ethical approval from the South East Research Ethics Committee (REC) to supply anonymized, pre-collected primary care data for scientific research. Patients can opt out of having their depersonalized records collected. Therefore, patient consent is not required when working with anonymized records from THIN database.