Non-Steroidal Anti-Inflammatory Drug Use and the Risk of Acute Myocardial Infarction in the General German Population: A Nested Case–Control Study

Data Source

This study was based on data from the German Pharmacoepidemiological Research Database (GePaRD), which has been described elsewhere [31, 32]. For the present study, claims data for about 17 million insurance members from four statutory health insurance providers (SHIs) from all geographical regions of Germany were included using the years 2004–2009. Besides demographic data, the database contains inpatient and outpatient diagnoses coded according to the German Modification of the International Classification of Diseases (ICD-10 GM), inpatient and outpatient diagnostic and therapeutic procedures, and outpatient drug dispensing. Inpatient data include information on admission and discharge dates. Dispensing data are available for all outpatient dispensing reimbursed by the SHIs and contain the dispensed drugs characterized by central pharmaceutical number (PZN), the dates of prescription and dispensing, as well as information on the prescribing physician. These data are linked to a pharmaceutical reference database, adding information on the defined daily dose (DDD), the anatomical-therapeutical-chemical (ATC) code, strength, packaging size, and the generic and brand names.

In Germany, utilization of health insurance data for scientific research is regulated by the code of Social Law. All contributing SHIs and their regulatory authorities approved the use of the data for this study. Informed consent was not required by law, since the study was based on pseudonymous data, and the necessary permissions were granted.

Study Design

A case–control study nested in a cohort of new NSAID users (ATC code M01, more detailed ATC codes available in Online Resource 1, see electronic supplementary material [ESM]) was conducted. Cohort members had to be continuously insured for at least 12 months before the first NSAID dispensing and had to be at least 18 years of age. In order to avoid bias by the inclusion of prevalent NSAID users, patients with an NSAID prescription within these 12 months were excluded [33]. Furthermore, patients with a diagnosis of malignant cancer (ICD-10-GM code D23.-) within these 12 months were excluded (except non-melanoma skin cancer). Cohort entry was defined as the patient’s first dispensing of an NSAID between January 01, 2005 and December 31, 2009. All patients were followed from their first NSAID dispensing in the study period until either interruption of insurance status for more than 3 days, end of insurance including death, diagnosis of malignant cancer, or the end of the study period/longest available follow-up in the database, whichever came first.

Definition of Cases and Controls

The outcome was a first hospitalization with a main discharge diagnosis of AMI (ICD-10 codes I12.-) or subsequent MI (ICD-10 codes I22.-) after cohort entry. The hospital admission day was defined as the index date of the case. Recurrent events were not examined.

From the cohort of new NSAID users, we randomly selected up to 100 controls for each case, matched by age at index date, sex, and SHI using risk set sampling. An index date was assigned to each control that resulted in the same duration of follow-up as the corresponding case, that is, cases and controls were also matched by length of follow-up. Cohort members who were hospitalized for any reason at the index date of the case were excluded from the set of potential controls, since they were not at risk of being hospitalized because of an AMI event. Patients might have served as controls for more than one case and were eligible to be selected as controls until they became a case [34].

Exposure Assessment

Classification of exposure to NSAIDs was based on the period (in days) between the index date and the end of supply of the most recent dispensing before the index date. Use status at the index date was categorized as follows: (1) current: if the supply overlapped the index date or ended within the 14-day period before the index date, (2) recent: if the supply ended between 15 and 183 days before the index date, or (3) past: if the supply ended 184 or more days before the index date. Past users of any NSAIDs were used as reference. If user numbers of individual NSAIDs were too small for analysis, they were grouped into the category of ‘all other NSAIDs’. The cumulative amount of NSAID use was calculated as the sum of dispensed DDDs between cohort entry and the last dispensing before the index date. The cumulative amount was categorized into low (0–90 DDDs), medium (91–180 DDDs), and high use (≥181 DDDs).

Risk Factor and Confounder Assessment

Potential confounders were assessed in the 12 months preceding cohort entry. The following potentially confounding co-morbidities (obtained from inpatient as well as outpatient diagnoses) were considered in our analysis: myocardial infarction, chronic ischemic heart disease, heart failure, stroke, hypertension, atrial fibrillation and flutter, peripheral arterial disease, other cardiovascular disease (cardiac arrhythmia/conduction disorder and arrest, cardiomyopathy, valvular disorder and endocarditis, myocarditis and pericarditis, arterial embolism and thrombosis), hyperlipidemia, osteoarthritis, rheumatoid arthritis, diabetes mellitus, chronic liver disease, kidney failure, alcohol abuse, obesity. The following drugs were assessed in the 12 months before the index date: angiotensin-converting-enzyme (ACE) inhibitors + angiotensin (AT) II antagonists, calcium channel blockers, β-blocking agents, diuretics, other antihypertensive drugs. The following confounding drugs were assessed in the 90 days before index date: aspirin, glucocorticoids, anticoagulants, statins, nitrates, platelet aggregation inhibitors, oral contraceptives, postmenopausal hormone therapy. The exact definition of co-morbidities and co-medication in terms of ICD-10 GM codes and ATC codes is available as Online Resource in S2–S5 Tables (see ESM).

Statistical Analyses

Conditional logistic regression was conducted to obtain confounder-adjusted odds ratios (ORs) with corresponding 95% confidence intervals (CIs) for current as well as recent use of individual NSAIDs, using past use of any NSAID as reference. If patients were simultaneously exposed to more than one NSAID, they were combined into the category ‘multiple NSAID use’, which was included as an additional dichotomous variable in the multivariable model.

A priori-specified confounders (Online Resource 2–5, see ESM) were always included in the model. Each potential confounder with a prevalence of 5% in controls was added to the multivariable model if the Wald test was significant (p value <0.05). Confounder selection was then done by a backward selection approach (p < 0.05). To evaluate potential effect modification, analyses were stratified by sex, age (<60 years/≥60 years), use of aspirin, platelet aggregation inhibitors, anticoagulants, ACE inhibitors + AT II, calcium channel blockers, β-blockers, nitrates, postmenopausal hormone therapy, glucocorticoids, other antihypertensive drugs, hypertension, other cardiovascular disease (cardiac arrhythmia/conduction disorder and arrest), atrial fibrillation and flutter, myocardial infarction, heart failure, peripheral arterial disease, stroke, diabetes mellitus, prior chronic ischemic heart disease, and hyperlipidemia (Online Resource 6 and 7, see ESM).

Confounder inclusion and selection were done in the same way as described above.

All statistical analyses were done using SAS 9.3 (SAS Institute Inc., Cary, NY, USA).

Baseline Characteristics

During the study period, 3,476,931 new NSAID users were identified with a median follow-up time of 886 days (Q1: 424 days, Q3: 1368 days). The median age at cohort entry was 48.0 years (Q1: 37 years, Q3: 61 years), and 44% of the cohort was male.

Of those new NSAID users, 17,236 cases had an AMI and 1,714,006 matched controls were randomly selected from the cohort risk set. Table 1 shows the baseline characteristics of cases and controls and also the respective unadjusted and adjusted ORs. In the case–control sample, the median age at index date was 68 years (Q1: 57 years, Q3: 77 years), and 66% were male.

Relative Risks Associated with Individual NSAID Use

Current use of fixed combinations of diclofenac with misoprostol (OR 1.76, 95% CI 1.26–2.45, Table 2) and of indometacin (1.69, 1.22–2.35) showed the highest relative AMI risks followed by the most frequently used NSAIDs in Germany, ibuprofen (1.54, 1.43–1.65), diclofenac (1.43, 1.34–1.52), and the COX-2 selective NSAID etoricoxib (1.52, 1.24–1.87). There was an increased relative AMI risk among current users of dexketoprofen (1.31, 0.80–2.16), naproxen (1.28, 0.86–1.90), piroxicam (1.21, 0.85–1.70), aceclofenac (1.21, 0.54–2.71), and dexibuprofen (1.19, 0.56–2.53), but the confidence intervals were wide and included the null value. No association was seen for the tNSAIDs meloxicam and acemetacin or for the COX-2 selective NSAIDs celecoxib and lumiracoxib. However, the number of current users was low for celecoxib and lumiracoxib and both confidence intervals include the null value.

Relative Risk Associated with Cumulative Amount of Individual NSAID Use

Low cumulative amount of NSAID use (0–90 DDDs) of indometacin (1.68, 1.12–2.51), fixed combinations of diclofenac with misoprostol (1.62, 1.04–2.52), diclofenac (1.33, 1.16–1.53), and ibuprofen (1.37, 1.19–1.58) was associated with an elevated relative risk of AMI (Table 3). This indicates that already short-term use of these NSAIDs was associated with an increased relative risk of AMI. Except for acemetacin (0.89, 0.48–1.62), celecoxib (0.86, 0.49–1.50), and phenylbutanzone (0.76, 0.18–3.27), relative risk estimates of other NSAIDs were elevated, but numbers of users were low and CIs wide, showing no statistically significant association.

High cumulative amount of NSAID use (≥181 DDDs) of etoricoxib (1.64, 1.05–2.55) and ibuprofen (1.26, 1.01–1.57) was associated with an increased relative risk of AMI. For indometacin (1.46, 0.54–3.98), fixed combinations of diclofenac with misoprostol (1.26, 0.54–2.93), and acemetacin (1.06, 0.36–3.06), elevated relative risks were observed, but again the confidence intervals were wide and included the null value (Table 3).

Subgroup Analyses of Cardiovascular Risk Factors

Relative AMI risk estimates were similar in males and females for current use of diclofenac, fixed combinations of diclofenac with misoprostol, etoricoxib, and ibuprofen, whereas current use of indometacin seems to be associated with a higher relative AMI risk in males than in females (Fig. 1).

Fig. 1

Adjusted relative risk of acute myocardial infarction associated with current use of NSAIDs compared with past use stratified by sex. NSAIDs non-steroidal anti-inflammatory drugs

The relative AMI risk seems to be higher in patients <60 years of age for current use of the respective NSAIDs with the exception of indometacin (Fig. 2).

Fig. 2

Adjusted relative risk of acute myocardial infarction associated with current use of NSAIDs compared with past use stratified by age. NSAIDs non-steroidal anti-inflammatory drugs

Patients with or without prior use of aspirin, anticoagulants, or platelet aggregation inhibitors seemed to be at similar enhanced relative risk when using the examined individual NSAIDs (Fig. 3a).

Fig. 3

Adjusted relative risk of acute myocardial infarction associated with current use of NSAIDs compared with past use and stratified by subgroups of cardiovascular risk factors: a with and without prior use of aspirin, anticoagulants, or platelet aggregation inhibitors; b with and without prior use of ACE inhibitors, AT II antagonists, calcium channel blockers, β-blockers, other antihypertensive drugs or diagnosis of hypertension; c with prior history of chronic ischemic heart disease or other cardiovascular disease; d with and without prior atrial fibrillation, myocardial infarction or heart failure; e with and without prior history of peripheral atrial diseases or stroke. NSAIDs non-steroidal anti-inflammatory drugs

Prior use of ACE inhibitors, AT II antagonists, calcium channel blockers, β-blockers, other antihypertensive drugs, or a history of hypertension appears to lower the relative risk of AMI associated with current use of NSAIDs, with the exception of indometacin. However, most CIs overlapped (Fig. 3b).

Patients with prior history of chronic ischemic heart disease or other cardiovascular disease, with atrial fibrillation, myocardial infarction or heart failure, or prior history of peripheral atrial diseases or stroke, had a similar relative risk associated with current use of the respective NSAIDs as patients without such conditions (Fig. 3c–e).

Below is the link to the electronic supplementary material.