Do statins protect against upper gastrointestinal bleeding?

Abstract

AIMS

Recently, an apparent protective effect of statins against upper gastrointestinal bleeding (UGB) was postulated in a post hoc analysis of a randomized trial. We aimed to evaluate the effect of statin use on acute nonvariceal UGB alone or in combinations with low-dose aspirin and other antithrombotic drugs.

METHODS

A population-based case–control study was conducted in the County of Funen, Denmark. Cases (n = 3652) were all subjects with a first discharge diagnosis of serious UGB from a hospital during the period 1995 to 2006. Age- and gender-matched controls (10 for each case) (n = 36 502) were selected by a risk set sampling. Data on all subjects' drug exposure and past medical history were retrieved from a prescription database and from the County's patient register. Confounders were controlled by conditional logistic regression.

RESULTS

The adjusted odds ratios (ORs) associating use of statins with UGB were 0.94 (0.78–1.12) for current use, 1.40 (0.89–2.20) for recent use and 1.42 (0.96–2.10) for past use. The lack of effect was consistent across most patient subgroups, different cumulative or current statin doses and different statin substances. In explorative analyses, a borderline significant protective effect was observed for concurrent users of low-dose aspirin [OR 0.43 (0.18–1.05)].

CONCLUSION

Statins do not prevent UGB, except possibly in users of low-dose aspirin.

WHAT IS ALREADY KNOWN ABOUT THIS SUBJECT.

• An apparent protective effect of statins against upper gastrointestinal bleeding was postulated in a post hoc analysis of a randomized trial.

• We aimed to evaluate the effect of statin use on acute nonvariceal upper gastrointestinal bleeding alone or in combinations with low-dose aspirin and other antithrombotic drugs.

WHAT THIS STUDY ADDS

• Our study could not demonstrate an inverse association between use of statins and upper gastrointestinal bleeding.

• The lack of effect was consistent across most subgroups, across different cumulative or current doses and different statin substances.

• In explorative analyses, there appeared to be protective effect only in users of low-dose aspirin.

• There was no interaction with use of other antithrombotic drugs.

Introduction

3-Hydroxy-3-methylglutaryl-coenzyme A reductase inhibitors, popularly referred to as ‘statins’, inhibit the rate-limiting step in cholesterol synthesis [1], and are used to lower blood cholesterol levels and prevent atherosclerosis [2]. Statins are amongst the most extensively investigated pharmaceutical agents in current clinical use. Apart from their lipid-lowering effects, many beneficial effects independent of cardiovascular risk-lowering properties have been stated as well: reduced risk of fractures [3], cancer [4], dementia [5], neuroprotective properties in cerebral ischaemia and stroke [6], their potential role for respiratory diseases [7], reduced incidence of cataract [8], etc.

Recently, it was found that statins could reduce the risk of postoperative bleeding [9]. The authors reported that treatment with pravastatin before coronary artery bypass grafting surgery using cardiopulmonary bypass significantly reduced the systemic inflammatory response as well as postoperative mediastinal blood loss in statin-treated patients compared with those receiving placebo, and this finding was probably attributed to a protective effect against endothelial cell injury. Recently, statin use has been associated with less overall gastrointestinal (GI) bleeding as well as major or severe bleeding in in-hospital GI bleeding due to use of antiplatelet therapy in acute coronary syndromes (ACS) [10]. The authors performed a post hoc analysis on 10 288 patients with ACS included in the OPUS-TIMI 16 trial and concluded that age >65 years, orbofiban treatment, Killip class >1, history of cerebrovascular disease, and calcium channel-blocker use were associated with a higher risk of GI bleeding, whereas statin therapy was associated with a lower risk [odds ratio (OR) 0.68, 95% confidence interval (CI) 0.45, 1.04; P = 0.079]. It was speculated that this could be mediated by an altered prostaglandin (PG) level from an increase in the expression and activity of c-phospholipase A2, cyclooxygenase-2, PGI₂ and PGE₂ synthases.

We conducted this population-based case–control study to evaluate the association between statin use with acute nonvariceal upper gastrointestinal bleeding (UGB) alone or in combinations with low-dose aspirin (ASA) and other antithrombotic drugs. Although a gastroprotective effect would hardly justify statin use in itself, it could have a substantial impact on our understanding of the risk–benefit balance and would affect the outcome of health economic assessments of statin therapy.

Subjects and methods

Setting

Data were retrieved from three different sources: the Patient Registry of the County of Funen (FPAS), the Danish Civil Registry, and the Odense University Pharmacoepidemiological Database (OPED).

The FPAS contains data on all discharges from hospitals in the County of Funen (population 470 000) since 1977. Diagnoses have been encoded using International Classification of Diseases, 8th revision (ICD8) from 1977 to 1993 and ICD10 thereafter [11]. ICD9 was never used in Denmark.

OPED is described elsewhere. Briefly, OPED is a population-based research database derived from administrative prescription data used for clearing of reimbursement and covers the County of Funen (470 000 persons) [12]. The substances and quantities are registered according to the World Health Organization's anatomical-therapeutic-chemical (ATC) system and defined daily doses (DDD) methodology [13].

The Danish Civil Registry was used to identify the source population, to extract controls, and to ensure that all subjects were residents of the County of Funen for ≥12 months before their index date.

All these data can be linked by using the mutual person identifier, the Central Person Registry code, which is shared with virtually all other health-related registries in Denmark, thereby allowing record-linkage studies.

Cases and controls

A validated complete dataset of cases of bleeding (n = 3652) along with an age- and sex-matched control group of 10 subjects per case (n = 36 502) was generated as in a previous study by our group [14]. Cases were defined by fulfilment of all of the following criteria: admission with peptic ulcer or gastritis as main diagnosis within one of the County's hospitals during the period August 1995 to July 2006; significant bleeding defined either by melena, subnormal haemoglobin or the need for transfusions; and a potential bleeding source in the stomach or duodenum identified by endoscopy or surgery. Excluded were gastric varices and persons who were <18 years old.

Cases that fulfilled the above-mentioned criteria were identified by manual review of all 12 607 discharge summaries with a main diagnosis of peptic ulcer (complicated or not) or gastritis (ICD10-code K25-29) within the study period. Cases were assigned an index date equivalent to their first registered date of a UGB diagnosis within the study period. Age- and sex-matched controls, 10 for each case, were sampled by use of a risk set sampling technique. In brief, controls for a given case were randomly selected among those within the county who matched the case with respect to gender and exact birth year. A few of the very old cases had <10 eligible control subjects, hence the control : case ratio deviated slightly from 10 : 1. The controls were assigned an index date identical to the admission date of the corresponding case. For both cases and controls, we required that they had been residents of the county for ≥1 year on the index date. Cases were eligible as control subjects until their first admission with UGB. By this sample technique, the generated ORs will be unbiased estimates of the incidence rate ratios [15].

Exposure definition

All relevant statin doses are available on the Danish market as single tablets. We therefore based our exposure definition on the assumption that the vast majority of patients would take one tablet per day. For each prescription we assigned an exposure period in days equivalent to the number of tablets in the dispensed package plus 20% to allow for minor noncompliance and for irregular prescription refills. The exposure period started on the day of redeeming the prescription, and the exposure clock was reset with each new prescription. The first 90 days after the expiry of an exposure period was termed ‘recent exposure’ and the period following that ‘past exposure’. As a sensitivity analysis we repeated the analyses with the grace period assigned to each prescription set to 10% and 30%. It had essentially no effect on the estimated ORs (results not shown).

Unless otherwise stated, analyses were based on current exposure, and the reference was person-time never exposed to statins. ‘Statins’ refer to simvastatin (ATC code C10AA01), lovastatin (ATC code C10AA02), pravastatin (ATC code C10AA03), fluvastatin (ATC code C10AA04), atorvastatin (ATC code C10AA05), cerivastatin (ATC code C10AA06), and rosuvastatin (ATC code C10AA07). The antithrombotic drugs (ATC code B01A) included in this study are clopidogrel (ATC code B01AC04), low-dose ASA (ATC code B01AC06), dipyridamole (ATC code B01AC07) and vitamin K antagonists (ATC code B01AA).

Data analysis

The crude and adjusted ORs with 95% confidence intervals were calculated by conditional logistic regression with adjustment for the mentioned confounders. Potential confounders included were current use of low-dose ASA (a prescription of low-dose ASA within the past 90 days), current use of any antithrombotic agents, current use of nitrate vasodilators (ATC code C01D), current use of anti-acid drugs (ATC code A02B), current use of selective serotonin reuptake inhibitors (SSRIs) (ATC code N06AB), current use of nonsteroidal anti-inflammatory drugs (NSAIDs) including cyclooxygenase (COX)-2 inhibitors (ATC code M01A), ever use of antidiabetics (ATC code A10) or a diagnosis of diabetes (ICD8 250, ICD10 E10-E14), previous diagnosis of severe GI bleeding (ICD8 531.9, 532.9, 533.9, 534.9; ICD10 K25.4, 26.4, 27.4, 28.4), previous diagnosis of peptic ulcer disease (ICD8 530-534; ICD10 K25-28), previous Helicobacter pylori eradication (all triple regimens obtained from OPED), hypertension (ICD8-40, ICD10-I10) or ever use of antihypertensive agents (ATC codes C03A, C07, C08, C09), previous diagnosis of ischaemic heart disease or acute myocardial infarction (ICD8 412-414, ICD10 I20-I25), previous diagnosis of heart failure (ICD8 427; ICD10 I50), previous diagnosis of stroke (ICD8-431, 433-435; ICD10 I61, I63 and I64 excluding I631 and I641), previous diagnosis of chronic obstructive lung disease (COLD) (ICD8 490-491, ICD10 J44) or current use of systemic corticosteroids (ATC code H02AB), previous history of hepatic cirrhosis (ICD8 571; ICD10 K74), previous history of renal failure (ICD8 580-583; ICD10 N17-N19), and any history of alcohol-related disorder (ICD8 303, ICD10 F10) or use of disulfiram (ATC code P03AA04) or psychiatric disorder (ICD8 295-300; ICD10-F20, 30-33). The potential confounding effect of age, sex and calendar year was handled by the matching procedure. Previous nongastroenterological bleedings were not included in the model.

The dose–response-like relationship was evaluated both in terms of daily dose taken and cumulative dose before the index date. To account for the difference in potency between statins, the Prescribed Daily Dose was expressed in units of DDD. Cut-off points were <1 DDD day⁻¹, 1–1.5 DDD day⁻¹ and ≥1.5 DDD day⁻¹. When analysing for use of different statin substances, only the last statin prescription before the index date was considered.

Results

We identified 3652 cases fulfilling our criteria. Men accounted for 1869 cases (51.2%) and the median and interquartile range for the age of cases was 75 years and 63–83 years. The other characteristics of cases and controls are shown in Table 1.

Table 1.

Characteristics of 3652 cases of upper gastrointestinal bleeding and their 36 502 control subjects

	Cases (n = 3652)	Controls (n = 36 502)
Age, mean, years (SD)
<60	707 (19.4)	7 070 (19.4)
≥60	2945 (80.6)	29 432 (80.6)
Male	1869 (51.2)	18 672 (51.2)
Current drug use (ATC code)
Statins	159 (4.4)	1 702 (4.6)
Low-dose aspirin (B01AC06)	621 (17.0)	3 039 (8.3)
Anticoagulants (B01AA)	164 (4.5)	761 (2.1)
Drug acting on the renin system (C09)	678 (18.6)	4 531 (12.4)
NSAIDs (M01A)	1213 (33.2)	3 887 (10.6)
Systemic corticosteroids (H02AB)	352 (9.6)	1 447 (4.0)
SSRIs (N06AB)	406 (11.1)	1 908 (5.2)
PPIs (A02B)	776 (21.2)	2 730 (7.5)
Spironolactone (C03DA01)	204 (5.6)	550 (1.5)
History of
Upper GI bleeding	91 (2.5)	214 (0.6)
Peptic ulcer	372 (10.2)	1 304 (3.6)
Helicobacter pylori eradication	180 (4.9)	519 (1.4)
Ischaemic heart disease	599 (16.4)	3 430 (9.4)
Heart failure	384 (10.5)	1 885 (5.2)
Hypertension	489 (13.4)	2 384 (6.5)
Stroke	315 (8.6)	1 562 (4.3)
Diabetes mellitus	246 (6.7)	1 170 (3.2)
COLD	265 (7.3)	1 213 (3.3)
Renal failure	80 (2.2)	157 (0.4)
Hepatic cirrhosis	54 (1.5)	84 (0.2)
Alcohol-related diagnosis or drug use	242 (6.6)	475 (1.3)

Unless otherwise indicated, data are shown as numbers (%). NA, not applicable; PPI, proton pump inhibitor; GI, gastrointestinal; COLD, chronic obstructive lung disease; ASA, acetyl salicylic acid; SSRIs, selective serotonin reuptake inhibitors; NSAIDs, nonsteroidal anti-inflammatory drugs.

Among the 3652 cases and 36 502 controls, 159 (4.4%) and 1702 (4.6%) were current users of statins, respectively. The adjusted ORs associating use of statins with UGB was 0.94 (95% CI 0.78, 1.12) for current use, 1.4 (95% CI 0.89, 2.20) for recent use, and 1.42 (95% CI 0.96, 2.10) for past use (Table 2). We found no protective effect with particular high or low cumulative or current doses or with any particular statin substance.

Table 2.

Association between exposure to statins or and upper gastrointestinal bleeding

Exposure	Cases exposed/ unexposed	Controls* exposed/ unexposed	Crude odds ratio (95% CI)	Adjusted odds ratio (95% CI)**
Statins, current use	159/3433	1 702/34 372	0.94 (0.79, 1.11)	0.94 (0.78, 1.12)
Statins, recent use	25/3433	180/34 372	1.38 (0.90, 2.10)	1.40 (0.89, 2.20)
Statins, past use	35/3433	248/34 372	1.42 (0.99, 2.02)	1.42 (0.96, 2.10)
Dose–response (current use, PDD)
Low dose (<1 DDD day−1)	32/3433	322/34 372	0.99 (0.69, 1.43)	0.98 (0.66, 1.45)
Medium dose (1–1.5 DDD day−1)	70/3433	780/34 372	0.90 (0.70, 1.15)	0.91 (0.70, 1.19)
High dose (≥1.5 DDD day−1)	57/3433	600/34 372	0.95 (0.72, 1.25)	0.95 (0.71, 1.27)
Cumulative dose of ever users of statins
Low dose (<500 DDD)	32/3433	417/34 372	0.78 (0.54, 1.12)	0.82 (0.56, 1.20)
Medium dose (500–1000 DDD)	54/3433	617/34 372	0.88 (0.66, 1.16)	0.88 (0.65, 1.19)
High dose (>1000 DDD)	73/3433	667/34 372	1.10 (0.86, 1.40)	1.07 (0.82, 1.39)
Statin substance
Simvastatin (C10AA01)	111/3433	1 147/34 372	0.97 (0.79, 1.18)	0.99 (0.80, 1.23)
Pravastatin (C10AA03)	18/3433	186/34 372	0.99 (0.61, 1.61)	0.89 (0.53, 1.50)
Atorvastatin (C10AA05)	23/3433	243/34 372	0.97 (0.63, 1.49)	0.98 (0.62, 1.54)

^*Matched for age, sex and index date.

^**Adjusted for age, sex, a previous discharge diagnosis of UGB, Helicobacter pylori eradication, COLD, peptic ulcer, ischaemic heart disease, liver cirrhosis, alcohol-related diagnoses, renal failure, diabetes, heart failure, hypertension, stroke and for current use of low-dose ASA, oral anticoagulants, clopidogrel, dipyridamole, anti-acid drugs (PPIs and H2RAs), SSRIs, NSAIDs, nitrate vasodilators, systemic corticosteroids. UGB, upper gastrointestinal bleeding; PDD, prescribed daily dose; DDD, defined daily dose; CI, 95% confidence interval; PPIs, proton pump inhibitors; H2RA, histamine-2 receptor antagonists; COLD, chronic obstructive lung disease; ASA, acetyl salicylic acid; SSRIs, selective serotonin reuptake inhibitors; NSAIDs, nonsteroidal anti-inflammatory drugs.

Table 3 shows the stratum-specific crude and adjusted ORs for the association between exposure to statins alone or combined with antithrombotic drugs and UGB. The lack of effect was consistent across most relevant subgroups.

Table 3.

Stratum-specific odds ratios for the association between current use of statins and upper gastrointestinal bleeding (UGB)

Stratum	Cases exposed/ unexposed	Controls exposed/ unexposed	Crude odds ratio (CI)	Adjusted odds ratios (CI)*
Men	89/1751	895/17 575	1.00 (0.80, 1.25)	1.04 (0.82, 1.32)
Age (years)
<60	34/664	325/6 665	1.04 (0.73, 1.49)	1.14 (0.76, 1.71)
≥60	125/2769	1 377/27 707	0.91 (0.76, 1.10)	0.89 (0.73, 1.09)
Current drug use:
Low-dose ASA users	21/587	126/2 877	0.62 (0.30, 1.28)	0.43 (0.18, 1.05)
Low-dose ASA non-users	138/2846	1 576/31 495	0.99 (0.83, 1.19)	0.97 (0.80, 1.17)
NSAID users	46/1141	186/3 655	0.83 (0.51, 1.33)	0.80 (0.47, 1.35)
NSAID non-users	113/2292	1 516/30 717	0.99 (0.81, 1.21)	0.97 (0.78, 1.19)
History of:
Peptic ulcer	11/354	64/1 226	0.93 (0.28, 3.09)	0.52 (0.05, 5.55)
Ischaemic heart disease	28/555	138/3 259	1.91 (0.96, 3.81)	1.68 (0.72, 3.92)
Heart failure (+)	22/356	84/1 782	1.65 (0.64, 4.27)	2.09 (0.59, 7.45)
Heart failure (−)	137/3077	1 618/32 590	0.89 (0.74, 1.06)	0.93 (0.77, 1.12)
Hypertension	20/460	103/2 260	0.93 (0.42, 2.02)	0.59 (0.22, 1.57)
Stroke	17/290	63/1 483	1.86 (0.52, 6.70)	3.43 (0.43, 27.54)
Diabetes mellitus	6/231	56/1 102	0.50 (0.05, 5.51)	7.11 (0.18, 280.96)
COLD	20/243	69/1 131	2.16 (0.70, 6.64)	1.09 (0.14, 8.30)
No history of alcohol-related diagnosis or drug use	150/3203	1 679/33 925	0.95 (0.80, 1.12)	0.93 (0.77, 1.11)

^*Adjusted for age, sex, a previous discharge diagnosis of UGB, Helicobacter pylori eradication, COLD, peptic ulcer, ischaemic heart disease, liver cirrhosis, alcohol-related diagnoses, renal failure, diabetes, heart failure, hypertension, stroke and for current use of low-dose ASA, oral anticoagulants, clopidogrel, dipyridamole, anti-acid drugs (PPIs and H2RAs), SSRIs, NSAIDs, nitrate vasodilators and systemic corticosteroids. CI, 95% confidence interval; PPIs, proton pump inhibitors; H2RA, histamine-2 receptor antagonists; COLD, chronic obstructive lung disease; ASA, acetyl salicylic acid; SSRIs, selective serotonin reuptake inhibitors NSAIDs; nonsteroidal anti-inflammatory drugs.

A borderline significant protective effect was observed for concurrent users of low-dose aspirin (OR 0.43, 95% CI 0.18, 1.05).

Discussion

Our large case–control study could not demonstrate an inverse association between use of statins and UGB. The lack of effect was consistent across most subgroups, across different cumulative or current doses and different statin substances. In explorative analyses, there appeared to be a protective effect only in users of low-dose aspirin [OR 0.43 (0.18–1.05)]. There was no interaction with use of other antithrombotic drugs.

In a post hoc analysis of 10 288 patients with ACS included in the OPUS-TIMI 16 trial, Atar et al. reported that statin therapy was associated with a lower risk of UGB (0.68; 0.45–1.04; P = 0.079) [10]. All participants in the OPUS-TIMI 16 trial took low-dose aspirin. We were able to confirm a suggested protective effect in low-dose ASA users, but can add that this protective effect is not generalizable to other patient subgroups or statin users as a whole. Prostaglandins reduce the risk of GI bleeding by protecting the gastric mucosa [16], and Chen et al. [17] have shown that statins induce COX-2 gene expression and PGE₂ release in murine macrophages. Although there is no well-established reason, an increase in systemic PGI₂ and PGE₂ production by statins could lead to a protective effect exerted on GI bleeding by statins. Although such a mechanism should also protect against GI bleedings caused by NSAIDs, we found only a weak, statistically nonsignificant protective effect of statins among users of NSAIDs [OR 0.80 (0.47–1.35)].

It should also be kept in mind that the OPUS-TIMI 16 study was randomized on orbofiban, not on statins. Therefore, it has the evidence attributes of an observational study, i.e. with an inherent risk of confounding. For example, persons of high socioeconomic status (SES) may choose to take statins and may have a low bleeding risk. To our knowledge, there are no case–control or cohort studies reporting on the association between statin use and UGB other than the present study.

Among the strengths of our study, the most important is the retrieval and ascertainment of a large number of cases, all of which were reviewed manually. Second, we used a true population-based approach with full coverage of admissions and statin prescriptions, which minimizes the risk of selection bias. Finally, we had the possibility to adjust for important potential confounders such as previous UGB or ulcers, H. pylori eradication, and drugs such as NSAIDs, corticosteroids, ulcer drugs, anticoagulants and SSRIs [18–20].

Among the limitations are the potential for misclassifying use of aspirin (high or low dose) or ibuprofen, as these drugs are available over the counter. The coverage of our prescription database is 98, 7 and 86% for low-dose aspirin, high-dose aspirin (i.e. tablet strengths of 500 mg) and NSAIDs [14], respectively. We would expect statin use to be associated with low-dose aspirin, but there is no good reason to expect a strong link with use of NSAIDs or high-dose aspirin. As our coverage for low-dose aspirin was very high, we find it unlikely that we would have substantial residual confounding with use of ulcerogenic drugs.

We did not have data on smoking and alcohol abuse, both risk factors for UGB that potentially could confound the statin–UGB association. Instead, we adjusted for history of COLD, which could be a crude marker of smoking, and for an alcoholism diagnosis or use of disulfiram. We performed some analyses with these variables excluded from the model, which had no impact on the estimated OR for the statin–UGB association (data not shown). We concede, however, that COLD and alcohol diagnoses are not very sensitive markers, and that there could be some residual confounding.

A limitation could be lack of data on socioeconomic gradient in use of statins among subjects. Socioeconomic inequalities in statin use are a public health concern, but they may also confound observational studies of statins' effectiveness. A recent Danish study by Thomsen et al. reported a weak association between statin use and SES in men with cardiovascular disease and no association in women [21]. It is unlikely that a weak association between SES and statin use would strongly confound our results. Furthermore, the direction of this confounder would probably be to lower the OR; preventive strategies like statins are more often adopted by people of high SES who, for other reasons, have a lower incidence rate of UGB. Therefore, it would not explain the general absence of an apparent protective effect in our study. In addition, if SES did substantially confound our result in men but not in women, we would observe a higher OR in women and not a lower one as we have actually seen.

In conclusion, we find it unlikely that statins would substantially reduce the risk of UGB, except possibly in users of low-dose ASA.

Competing interests

J.H. has received fees for teaching from AstraZeneca, MSD and the Danish Association of the Pharmaceutical Industry and research grants from Novartis, MSD, Pfizer and Nycomed. M.A. has participated in research projects funded by AstraZeneca, Lundbeck, Novartis and Nycomed and has received fees for teaching from the Danish Association of the Pharmaceutical Industry.

Data were provided free of charge by the University of Southern Denmark and County of Funen.