Advertisement

Antihypertensive drugs and the risk of gastrointestinal bleeding∗

  • Samy Suissa
    Correspondence
    Requests for reprints should be addressed to Samy Suissa, PhD, Division of Clinical Epidemiology, Royal Victoria Hospital, 687 Pine Avenue West, Ross 4.29, Montreal, Québec, Canada, H3A 1A1
    Affiliations
    Department of Epidemiology and Biostatistics, the McGill Pharmacoepidemiology Research Unit, the Department of Medicine, McGill University, the Royal Victoria Hospital and Montreal General Hospital, Montreal, Canada
    Search for articles by this author
  • Chantal Bourgault
    Affiliations
    Department of Epidemiology and Biostatistics, the McGill Pharmacoepidemiology Research Unit, the Department of Medicine, McGill University, the Royal Victoria Hospital and Montreal General Hospital, Montreal, Canada
    Search for articles by this author
  • Alan Barkun
    Affiliations
    Department of Epidemiology and Biostatistics, the McGill Pharmacoepidemiology Research Unit, the Department of Medicine, McGill University, the Royal Victoria Hospital and Montreal General Hospital, Montreal, Canada
    Search for articles by this author
  • Odile Sheehy
    Affiliations
    Department of Epidemiology and Biostatistics, the McGill Pharmacoepidemiology Research Unit, the Department of Medicine, McGill University, the Royal Victoria Hospital and Montreal General Hospital, Montreal, Canada
    Search for articles by this author
  • Pierre Ernst
    Affiliations
    Department of Epidemiology and Biostatistics, the McGill Pharmacoepidemiology Research Unit, the Department of Medicine, McGill University, the Royal Victoria Hospital and Montreal General Hospital, Montreal, Canada
    Search for articles by this author

      Abstract

      Purpose: Calcium channel blockers have been reported to increase the risk of gastrointestinal bleeding. We tested this hypothesis, and also assessed whether beta blockers decrease this risk.
      Subjects and methods: A nested case-control design within a population-based cohort of all 34,074 new users of beta blockers, angiotensin-converting enzyme (ACE) inhibitors, or calcium channel blockers in Saskatchewan, from 1990 to 1993 and followed up to March 1995, was used. We identified all 311 subjects hospitalized because of gastrointestinal bleeding during this period, each of whom was matched to 10 randomly selected controls.
      Results: The rate of hospitalization for gastrointestinal bleeding was 3.0 per 1,000 per year. The adjusted rate ratio of gastrointestinal bleeding for current use of calcium channel blockers was 1.1 (95% confidence interval [CI] 0.8 to 1.4) and 0.66 (95% CI 0.44 to 0.98) for beta blockers compared with no current use of anti-hypertensive drugs. The adjusted rate ratio for ACE inhibitor use was 1.0 (95% CI 0.7 to 1.3) while that for diuretic use was 1.4 (95% CI 1.0 to 2.0).
      Conclusions: The use of calcium channel blockers does not appear to increase the risk of gastrointestinal bleeding in the first five years of treatment, while beta blockers may prevent this adverse event. The unexpected elevated risk associated with the use of diuretics needs to be investigated further.
      Calcium channel blockers achieved widespread use for the treatment of hypertension and angina in the 1980s primarily because they appeared to have fewer side effects than other antihypertensive drugs. A recent cohort study in an elderly population reported an elevated risk of gastrointestinal bleeding associated with the use of calcium channel blockers when compared with beta blockers (
      • Pahor M.
      • Guralnik J.M.
      • Furberg C.D.
      • et al.
      Risk of gastrointestinal hemorrhage with calcium channel antagonists in hypertensive persons over 67 years old.
      ). That study was prompted by case reports proposing that the antiplatelet effects of calcium channel blockers could be associated with such hemorrhage (
      • Wagenknecht L.E.
      • Furberg C.D.
      • Hammon J.W.
      • et al.
      Surgical bleeding unexpected effect of calcium antagonist.
      ,
      • Gore J.M.
      • Sloan M.
      • Price T.R.
      • et al.
      Intracerebral hemorrhage, cerebral infarction, and subdural hematoma after acute myocardial infarction and thrombolytic therapy in the Thrombolysis in Myocardial Infarction study. Thrombolysis in myocardial infarction, phase II, pilot and clinical trial.
      ,
      • Kario K.
      • Imiya M.
      • Ohat Y.
      • Shimada K.
      Gastrointestinal bleeding induced by amlodipine.
      ).
      A hypothesis proposed to explain the increased risk found for calcium channel blockers is that it could be an artifact resulting from a protective effect of beta blockers against gastrointestinal bleeding (
      • Gordon R.D.
      Calcium channel blockers and gastrointestinal haemorrhage the balancing act.
      ). The study design used by Pahor et al (
      • Pahor M.
      • Guralnik J.M.
      • Furberg C.D.
      • et al.
      Risk of gastrointestinal hemorrhage with calcium channel antagonists in hypertensive persons over 67 years old.
      ) could not provide an assessment of the risk specifically associated with beta blockers, since the users of these medications formed the reference group. In that study, the measure of antihypertensive drug use was limited to only two points in time, so that changes that occurred in patients from one drug class to the other during the 6-year follow-up could not be determined. Moreover, the timing of drug use and gastrointestinal bleeding could not be assessed.
      We therefore determined whether the use of calcium channel blockers increases the risk of gastrointestinal bleeding in a large population-based cohort of subjects who were treated with an antihypertensive medication. We also tested the hypothesis that beta blockers prevent gastrointestinal bleeding.

      Methods

      We used a population-based cohort of first-time users of three antihypertensive drug classes, angiotensin-converting enzyme (ACE) inhibitors, beta blockers, and calcium channel blockers, in Saskatchewan, Canada. The study period spanned 5 years, from January 1, 1990, to March 31, 1995. The data were analysed using a nested case-control design to permit the assessment of the independent risks associated with the use of each drug.

      Sources of data

      Data were obtained from the computerized databases of the Saskatchewan Health Branch, responsible for administering universally insured health care services for this Canadian province. The identification database includes demographic data, such as date of birth, date of death, gender, social assistance status, as well as coverage initiation and termination dates. The prescription drugs database includes information on all out-patient prescription medications dispensed in the province, including drug name, date of dispensing, number of units administered and dose. The hospitalization database includes primary and secondary discharge diagnoses, classified according to the International Classification of Diseases, 9th revision (ICD-9) codes, and dates of admission and discharge. The insurance program covers 95% of the Saskatchewan population. These databases are linked by a unique identifier, and are accurate and comprehensive (
      • Strand L.M.
      • Downey W.
      Health databases in Saskatchewan.
      ,
      • Rawson N.S.
      • Malcolm E.
      Validity of the recording of ischaemic heart disease and chronic obstructive pulmonary disease in the Saskatchewan health care datafiles.
      ).

      Study cohort

      We identified all subjects who received an ACE inhibitor, a beta blocker, or a calcium channel blocker for the first time between January 1, 1990, and December 31, 1993. This was verified by checking that no such prescriptions were dispensed as far back as January 1, 1989. We designated the date of this first prescription the cohort entry date. All subjects with the outcome event (hospitalization for gastrointestinal bleeding) from January 1, 1989, to the cohort entry date were excluded. All cohort members were followed up until either March 31, 1995, the date of the outcome event, the date of death, or the date of emigration from the province.

      Case definition

      The study outcome was hospitalization for gastrointestinal bleeding. We selected, consistent with the outcomes examined by Pahor et al (
      • Pahor M.
      • Guralnik J.M.
      • Furberg C.D.
      • et al.
      Risk of gastrointestinal hemorrhage with calcium channel antagonists in hypertensive persons over 67 years old.
      ), hospitalizations with one of the following primary discharge diagnoses: gastric, duodenal, gastrojejunal, or peptic ulcer with bleeding (ICD-9 codes 531.0, 531.2, 531.4, 531.6, 532.0, 532.2, 532.4, 532.6, 533.0, 533.2, 533.4, 533.6, 534.0, 534.2, 534.4, and 534.6) and other bleeding of the gastrointestinal tract (ICD-9 codes 578.0, 578.1, and 578.9). We also included hospitalizations with one of these diagnoses listed as the secondary discharge diagnosis, if the primary diagnosis was iron deficiency and other anemias, other hemorrhagic conditions, diseases of the esophagus, gastritis and duodenitis disorders of function of stomach and duodenum, symptoms involving the digestive system, abdomen and pelvis, nonspecific findings on blood examination and function studies, or other ill-defined and unknown causes of morbidity. If a cohort member was hospitalized more than once during the study period, the first eligible event was used. We designated the date of hospitalization as the index date.

      Selection of controls

      Because of the large size of the cohort and the time-dependent nature of antihypertensive drug use, we used a nested case-control design within the cohort to analyze the complex patterns of drug use (
      • Suissa S.
      Novel approaches to pharmacoepidemiology study design and statistical analysis.
      ,
      • Lubin J.H.
      • Gail M.H.
      Biased selection of controls for case-control analyses of cohort studies.
      ,
      • Robins J.M.
      • Prentice R.L.
      • Blevins D.
      Designs for synthetic case-control studies in open cohorts.
      ). For each case, we formed a risk set of all potential controls consisting of all cohort members with the same year and month of cohort entry, and with follow-up at least as long as that of the case. A random sample of 10 controls was selected from each risk set and matched to the case according to the index date. This scheme permits future cases to be selected as controls before they become cases (
      • Lubin J.H.
      • Gail M.H.
      Biased selection of controls for case-control analyses of cohort studies.
      ).

      Antihypertensive and other drug use

      All prescriptions for beta blockers, ACE inhibitors and calcium channel blockers dispensed during the follow-up period were identified, along with all prescriptions for diuretics. Data on all other drugs, including nonsteroidal anti-inflammatory drugs (NSAIDs) and anti-ulcer medications, dispensed in the year before cohort entry and during follow-up were also obtained. Other dispensed drugs were used to measure chronic disease status based on the Chronic Disease Score developed by Van Korff et al (
      • Von Korff M.
      • Wagner E.H.
      • Saunders K.
      A Chronic Disease Score from automated pharmacy data.
      ).

      Statistical analysis

      We estimated the rate of gastrointestinal bleeding in the entire cohort using cumulative person-time. Subjects were defined as currently exposed to an antihypertensive drug if it was dispensed in the 90-day time window before the index date. Subjects who received more than one antihypertensive drug were considered currently exposed to each of them. Nonuse of any antihypertensive drug in the 90-day time window was the reference category. Rate ratios for gastrointestinal bleeding were estimated for all antihypertensive drug classes using conditional logistic regression. Models included adjustments for the concurrent use of other antihypertensive drugs, gender, age (in years) at index date, social assistance status at cohort entry, use of NSAIDs, anti-ulcer medications or anticoagulants, and the Chronic Disease Score during the year before index date. Effect modification by age and concurrent use of anti-ulcer medications was assessed. We report rate ratios (RR) with 95% confidence intervals (CI). Deviance residuals were examined to assess goodness-of-fit.
      To study the effect of the duration of current use of antihypertensive drugs, prescriptions were assumed to last 45 days so that successive prescriptions within 45 days of each other defined continuous use. Such continuous use must have overlapped the 90-day period preceding the index date. Three time intervals for the duration of use of each class were considered, namely, the first 45 days, 45 to 120 days, and more than 120 days of continuous use. For each of these intervals, a rate ratio was estimated with the reference group consisting of subjects unexposed during the interval.

      Results

      The cohort included 34,074 subjects, 76 of whom were excluded because of a hospitalization for gastrointestinal bleeding before cohort entry, leaving 33,998 eligible subjects. During follow-up, 311 cases were identified, of whom 288 were hospitalized with a primary diagnosis of gastrointestinal bleeding. An additional 23 cases had gastrointestinal bleeding as the secondary discharge diagnosis. Gastric and duodenal ulcers were the most common diagnoses (Table 1), although the site of bleeding was not specified in the discharge diagnosis for more than 40% of the patients.
      Table 1Hospital Discharge Diagnoses of 311 Cases of Gastrointestinal Bleeding
      ICD-9 = International Classification of Diseases, 9th edition.
      Hospital Discharge DiagnosisICD-9 Codesn%
      Gastric ulcer531, 531.4, 531.65317.0
      Duodenal ulcer532, 532.4, 532.65417.4
      Peptic ulcer533, 533.4, 533.682.6
      Gastrojejunal ulcer534, 534.4, 534.610.3
      Hematemesis578196.1
      Melena578.14915.8
      Hemorrhage of gastrointestinal tract, Unspecified578.912740.8
      legend ICD-9 = International Classification of Diseases, 9th edition.
      The rate of gastrointestinal bleeding for the entire cohort was 3.0 cases per 1,000 subjects per year. For subjects 68 years of age and older, the rate was 5.9 per 1,000 per year. At cohort entry, the subjects were 57 ± 17 years old (mean ± SD) and 50% were women. At entry, 11,961 subjects were initially treated with beta blockers, 11,014 with ACE inhibitors, and 11,023 with calcium channel blockers.
      Table 2displays the characteristics of cases and controls in the nested case-control sample. Cases were 12 years older than controls and less likely to be women. The cases used more medications than controls in the year before cohort entry, particularly anti-ulcer medications, NSAIDs, and anticoagulants.
      Table 2Characteristics of Cases and Controls at Index Date or During the Year Prior to Index Date
      Plus-minus values are mean ± SD.
      NSAIDs = Nonsteroidal anti-inflammatory drugs.
      CharacteristicCases (n = 311) PercentageControls (n = 3,110) Percentage
      Age, years69.4 ± 14.557.8 ± 17.3
      Female34.750.7
      Social assistance5.85.6
      Using medications for
      Respiratory illness11.65.8
      Diabetes11.66.1
      Asthma, rheumatism10.66.4
      Peptic ulcer disease38.614.9
      Subjects taking
      NSAIDs46.632.9
      Psychotropic drugs37.625.1
      Anticoagulants7.12.5
      Chronic Disease Score
      Based on Von Korff et al (11).
      4.2 ± 2.42.9 ± 2.2
      06.417.0
      1–211.621.7
      3–1382.061.4
      Plus-minus values are mean ± SD.
      Based on Von Korff et al
      • Von Korff M.
      • Wagner E.H.
      • Saunders K.
      A Chronic Disease Score from automated pharmacy data.
      .
      legend NSAIDs = Nonsteroidal anti-inflammatory drugs.
      Table 3presents the prevalence of current use of each antihypertensive drug class in the 90 days before the index date. In crude analyses, current use of calcium channel blockers, relative to no use of antihypertensive drugs within this 90-day time window, was associated with an increased risk of gastrointestinal bleeding (RR 2.0, 95% CI 1.5 to 2.7), as was current use of diuretics (RR 2.4, 95% CI 1.7 to 3.4), and current use of ACE inhibitors (RR 1.7, 95% CI 1.2 to 2.3). Current use of beta blockers was associated with a lower risk (RR 0.8, 95% CI 0.5 to 1.2). Adjustment for concurrent use of other antihypertensive drugs led to a reduction in these rate ratios. Models that also adjusted for other confounding factors, such as age and gender, disclosed that there was no longer an association between use of calcium channel blockers or ACE inhibitors and hospitalization for gastrointestinal bleeding. However, use of beta blockers was associated with a reduced risk and use of diuretics was associated with an increased risk of gastrointestinal bleeding.
      Table 3Use of Antihypertensive Medications among Cases (n = 311) and Controls (n = 3,110) and Rate Ratios
      Adjusted for current use of other antihypertensive drugs, age (in years) at index date, gender, social assistance status at cohort entry, use of nonsteroidal anti-inflammatory drugs, anti-ulcer drugs or anticoagulants as well as the Chronic Disease Score during the year prior to index date.
      for Hospitalization for Gastrointestinal Bleeding Associated with Current Exposure to Antihypertensive Drugs at Index Date
      ACE inhibitor = angiotensin-converting enzyme inhibitor.
      Antihypertensive MedicationCases (n = 311) %Controls (n = 3,110) %Unadjusted Rate RatioRate Ratio Adjusted for Other Antihypertensive AgentsFully Adjusted Rate Ratio (95% Confidence Interval)
      Beta blockers10.617.80.800.610.66 (0.44–0.98)
      ACE inhibitors26.721.41.661.260.95 (0.70–1.31)
      Calcium channel blockers31.521.12.031.671.06 (0.78–1.43)
      Diuretics21.511.72.421.941.42 (1.00–2.01)
      None (Reference)32.542.0
      Adjusted for current use of other antihypertensive drugs, age (in years) at index date, gender, social assistance status at cohort entry, use of nonsteroidal anti-inflammatory drugs, anti-ulcer drugs or anticoagulants as well as the Chronic Disease Score during the year prior to index date.
      legend ACE inhibitor = angiotensin-converting enzyme inhibitor.
      To assess biases induced by interruption of antihypertensive drug use, we determined whether there were differences in the previous use of medications among cases and controls who were not current users of antihypertensive medications, as might have occurred if previous use led to abdominal symptoms. Among cases no longer using antihypertensive medications, 33% had last used beta blockers, 25% had last used ACE inhibitors, 40% had last used calcium channel blockers, and 17% had last used diuretics (these proportions sum to more than 100% since multiple use was possible). Among controls no longer using antihypertensive medications, these rates were 46% for beta blockers, 20% for ACE inhibitors, 34% for calcium channel blockers, and 10% for diuretics. These patterns were proportional to those among cases and controls who were current users of antihypertensive medications (Table 3).
      Several confounders affected the adjusted rate ratios. To understand this effect, Table 4examines selected characteristics of the controls according to their current use of antihypertensive drugs. Nonusers of antihypertensive drugs in the 90-day period before the index date were more than 10 years younger than current users of ACE inhibitors, calcium channel blockers and diuretics, but only 3 years younger than current users of beta blockers. Nonusers of antihypertensive drugs and current users of beta-blocking agents were also generally less ill than the others: they were dispensed fewer medications in the year prior to treatment initiation, as reflected by their lower use of NSAIDs and anti-ulcer medications, as well as by a lower mean Chronic Disease Score.
      Table 4Selected Characteristics of the Control Group According to Current Exposure to Antihypertensive Drugs
      NSAIDs = nonsteroidal anti-inflammatory drugs; ACE inhibitor = angiotensin-converting enzyme inhibitor.
      CharacteristicNonusersACE InhibitorsCalcium Channel BlockersBeta BlockersDiuretics
      Mean age at index date (years)52.962.663.856.265.3
      Mean Chronic Disease Score
      Based on Von Korff et al (11).
      1.84.34.32.74.8
      NSAIDs use (%)29.930.039.732.736.0
      Anti-ulcer drug use (%)13.514.019.113.917.6
      Based on Von Korff et al
      • Von Korff M.
      • Wagner E.H.
      • Saunders K.
      A Chronic Disease Score from automated pharmacy data.
      .
      legend NSAIDs = nonsteroidal anti-inflammatory drugs; ACE inhibitor = angiotensin-converting enzyme inhibitor.

      Discussion

      We found that the use of calcium channel blockers was not associated with an elevated risk of hospitalization for upper gastrointestinal bleeding. Although the crude relative risk was elevated, adjustment for confounding factors such as age and use of NSAIDs and anti-ulcer medications abolished this effect. On the other hand, we found that beta blocker use was associated with a 33% reduction in the risk of hospitalization for gastrointestinal bleeding.
      Our results for calcium channel blockers are similar to those previously reported, but our conclusions are different. Pahor et al (1) reported that users of calcium channel blockers were 1.9-fold (95% CI 1.2 to 2.8) more likely to develop gastrointestinal bleeding than users of beta blockers. We found, however, that the principal explanation for this elevated relative risk is that beta blockers, the reference drug class, appear to protect against gastrointestinal bleeding. A recent case-control study (
      • Garcia Rodriguez L.A.
      • Cattaruzi C.
      • Troncon M.
      • Agostinis L.
      Grazia
      Risk of hospitalization for upper gastrointestinal bleeding associated with Ketorolac, other nonsteroidal anti-inflammatory drugs, calcium antagonists, and other antihypertensive drugs.
      ) examining the association of antihypertensive drug use and gastrointestinal bleeding found an increased risk associated with current use of calcium channel blockers (RR = 1.7) whereas beta blockers were associated with the lowest risk (RR = 1.0) of all antihypertensive agents when compared with nonusers. These results are not discordant with our conclusions since that study used subjects without hypertension as the reference group. Subjects with past use of antihypertensive drugs had a 1.3-fold increased risk compared with never users.
      Some in vitro and animal studies support a biologically plausible mechanism by which calcium channel blockers may predispose to bleeding (
      • Rostagno C.
      • Abbate R.
      • Gensini G.F.
      • et al.
      In vitro effects of two novel calcium antagonists (nitrendipine and nisoldipine) on intraplatelet calcium redistribution, platelet aggregation and thromboxane A2 formation Comparison with diltiazem, nifedipine and verapamil.
      ,
      • Blache D.
      • Ojeda C.
      Comparative inhibitory effects of dihydropyridines on platelet aggregation, calcium uptake and cyclic AMP concentration.
      ,
      • Pietraszek M.H.
      • Chabielska E.
      • Buczko W.
      Effects of verapamil, diltiazem and nifedipine on some parameters of hemostasis in the rat.
      ). However, available human studies are conflicting (
      • Wagenknecht L.E.
      • Furberg C.D.
      • Hammon J.W.
      • et al.
      Surgical bleeding unexpected effect of calcium antagonist.
      ,
      • Ding Y.A.
      • Han C.L.
      • Chou T.C.
      • et al.
      Effects of the calcium antagonist isradipine on 24-hour ambulatory blood pressure, platelet aggregation, and neutrophil oxygen free radicals in hypertension.
      ,
      • Rostagno C.
      • Prisco D.
      • Paniccia R.
      • et al.
      Effects of calcium channel blockers on platelet aggregation and thromboxane A2 formation an in vivo double blind randomized study.
      ,
      • Feinberg W.M.
      • Bruck D.C.
      Effect of oral nimodipine on platelet function.
      ,
      • Pickard J.D.
      • Murray G.D.
      • Illingworth R.
      • et al.
      Effect of oral nimodipine on cerebral infarction and outcome after subarachnoid haemorrhage British aneurysm nimodipine trial.
      ,
      • Kario K.
      • Imiya M.
      • Ohta Y.
      • Shimada K.
      ,
      • Grimm I.
      • Shaheen N.
      Calcium channel blockers and gastrointestinal hemorrhage more bad press?.
      ). The largest clinical series to date, published in abstract form, demonstrated no significant increase in the risk of hemorrhage following cardiac surgery in 2,248 patients treated with calcium channel blockers (
      • Grodecki-De Franco P.V.
      • Steinbul S.R.
      • Taylor P.C.
      • et al.
      Calcium antagonist use and perioperative bleeding complications: an analysis of 5,157 patients.
      ).
      These results support the hypothesis that beta blockers may prevent nonvariceal gastrointestinal bleeding (
      • Gordon R.D.
      Calcium channel blockers and gastrointestinal haemorrhage the balancing act.
      ). Beta blockers have been shown to prevent esophageal variceal bleeding (
      • Poynard T.
      • Cales P.
      • Pasta L.
      • et al.
      Beta-adrenergic- antagonist drugs in the prevention of gastrointestinal bleeding in patients with cirrhosis and esophageal varices. An analysis of data and prognostic factors in 589 patients from four randomized clinical trials. Franco-Italian Multicenter Study Group.
      ,
      • Bernard B.
      • Lebrec D.
      • Mathurin P.
      • et al.
      Beta-adrenergic antagonists in the prevention of gastrointestinal rebleeding in patients with cirrhosis a meta-analysis.
      ), and to decrease gastric mucosal blood flow in patients with hypertensive portal gastropathy (
      • Bernstein D.E.
      • Phillips R.S.
      Portal hypertensive gastropathy.
      ). Propranolol has been reported to exert a protective effect against ethanol-induced gastric hemorrhagic lesions in mice (
      • Bhandare P.
      • Diniz-D’Souza R.
      • Mainker A.
      • Dhume V.
      Protective effect of propranolol on ethanol-induced gastric lesions in mice.
      ), as well as ethanol-induced contraction of the circular muscle of rat fundic strip (
      • Rataboli P.V.
      • Bhandare P.N.
      • D’Souza R.S.
      • Dhume V.G.
      Protective effect of propranolol on ethanol-induced gastric lesions in rats probable mechanism of action.
      ), an in vitro animal model of intestinal function. These effects may be mediated by prostaglandins and nitric oxide, or a membrane stabilizing action (
      • Brandsborg O.
      • Brandsbrog M.
      • Christesen N.J.
      The role of beta-adrenergic receptor in the secretion of gastrin studies in normal subjects and patients with duodenal ulcers.
      ). Propranolol has also been shown to block isoproterenol stimulated gastrin production (
      • Brandsborg O.
      • Brandsbrog M.
      • Christesen N.J.
      The role of beta-adrenergic receptor in the secretion of gastrin studies in normal subjects and patients with duodenal ulcers.
      ,
      • Daly M.J.
      The classification of adrenoceptors and their effects on gastric acid secretion.
      ) as well as hydrochloric acid and gastric juice secretions in humans (
      • Asnchelewitsch J.
      • Okun K.
      Adrenegic processes and partial secretion of the stomach.
      ).
      An unexpected finding of our study is that diuretics may increase the risk of gastrointestinal bleeding. Few data are available on the effects of diuretics on gastrointestinal bleeding. Early studies reported an association between certain diuretics and gastrointestinal bleeding (
      • Jick H.
      • Porter J.
      Drug-induced gastrointestinal bleeding. Report from the Boston Collaborative Drug Surveillance Program, Boston University Medical Center.
      ). More recently, potassium supplements (often administered to patients taking potassium-wasting diuretics) were found to cause upper gastrointestinal lesions (
      • Ryan J.R.
      • McMahon F.G.
      • Akdamar K.
      • et al.
      Mucosal irritant potential of a potassium-sparing diuretic and of wax-matrix potassium chloride.
      ). Hydrochlorothiazide has been reported to cause an in vitro decrease in platelet activity (
      • Gleerup G.
      • Peterson J.R.
      • Mehlsen J.
      • Winther K.
      Effect of spirapril and hydrochlorothiazide on platelet function and euglobin clot lysis time in patients with mild hypertension.
      ), while amiloride significantly inhibited thrombin-stimulated platelet response and suppressed the inhibitory effect of endothelin-1 on thrombin-induced platelet aggregation and intracellular free calcium concentration (
      • Touyz R.M.
      • Schiffrin E.L.
      Role of protein kinase C in the anti-aggregatory effects of endothelin-1 on human platelets.
      ). Other studies have yielded contradictory findings (
      • Rendu F.
      • Bachelot C.
      • Molle D.
      • et al.
      Indapamide inhibits human platelet aggregation in vitro comparison with hydrochlorothiazide.
      ).
      Our study, which was based on computerized claims databases, has several strengths and limitations. First, our study design permitted an analysis of drug exposure at or near the time of the admission for gastrointestinal bleeding. We could therefore estimate the independent effects of different antihypertensive drugs, as compared with no antihypertensive drug use, during the 90-day period before the index date. Second, the drug exposure data for our study come from medications dispensed by a pharmacist, thus reducing the degree of exposure misclassification. Finally, we studied a population-based cohort using data that have been validated with excellent accuracy (
      • Rawson N.S.
      • Malcolm E.
      Validity of the recording of ischaemic heart disease and chronic obstructive pulmonary disease in the Saskatchewan health care datafiles.
      ).
      We did not, however, have data on over-the-counter use of NSAIDs, particularly aspirin, as a possible confounding factor. This would not, however, affect our study results unless use of NSAIDs varied by class of antihypertensive medication. Indeed, it seems likely that aspirin use may have been greater among users of beta blockers, who were nonetheless at reduced risk of gastrointestinal bleeding. Second, our reference group of current nonusers could have stopped using antihypertensive drugs for a variety of reasons, including abdominal symptoms. However, the distributions of the last use of antihypertensive drugs for current nonusers were proportional to those among users. Thus, cessation of use of antihypertensive medications appears to be independent of the drugs that had been previously used. Third, although the outcome definition we used for gastrointestinal bleeding was identical to that used by Pahor et al, our rates were only about one third of those observed previously among people aged 68 years or more. This is surprising since we used the same ICD-9 codes for hospital discharges and the age distributions were similar. One explanation is that gastrointestinal bleeding is underdiagnosed in Saskatchewan. Although medical records were not available to validate this outcome in the current study, it has been validated previously (
      • Garcia Rodriguez L.A.
      • Walker A.M.
      • Perez Gutthann S.P.
      Nonsteroidal antiinflammatory drugs and gastrointestinal hospitalizations in Saskatchewan a cohort study.
      ,
      • Guess H.A.
      • West R.
      • Strand L.M.
      • et al.
      Fatal upper gastrointestinal hemorrhage or perforation among users and nonusers of anti-inflammatory drugs in Saskatchewan, Canada, 1983.
      ). Instead, we believe differences may have arisen because the cohort of 1,636 elderly patients used in the previous study was selected after excluding 6,569 subjects, and thus may have been sicker. In our study, subjects were selected from the entire population without restriction, thus eliminating potential selection bias.
      In conclusion, we found that calcium channel blockers do not increase the risk of gastrointestinal bleeding, whereas beta blockers may prevent against this adverse event. The finding of an elevated risk of diuretics is unexpected and requires further investigation.

      References

        • Pahor M.
        • Guralnik J.M.
        • Furberg C.D.
        • et al.
        Risk of gastrointestinal hemorrhage with calcium channel antagonists in hypertensive persons over 67 years old.
        Lancet. 1996; 347: 1061-1065
        • Wagenknecht L.E.
        • Furberg C.D.
        • Hammon J.W.
        • et al.
        Surgical bleeding.
        BMJ. 1995; 310: 776-777
        • Gore J.M.
        • Sloan M.
        • Price T.R.
        • et al.
        Intracerebral hemorrhage, cerebral infarction, and subdural hematoma after acute myocardial infarction and thrombolytic therapy in the Thrombolysis in Myocardial Infarction study. Thrombolysis in myocardial infarction, phase II, pilot and clinical trial.
        Circulation. 1991; 83: 448-459
        • Kario K.
        • Imiya M.
        • Ohat Y.
        • Shimada K.
        Gastrointestinal bleeding induced by amlodipine.
        J Hum Hypertens. 1995; 9: 206-207
        • Gordon R.D.
        Calcium channel blockers and gastrointestinal haemorrhage.
        Lancet. 1996; 347: 1056
        • Strand L.M.
        • Downey W.
        Health databases in Saskatchewan.
        in: Strom B.L. Pharmacoepidemiology. John Wiley & Sons, New York1994: 217-229
        • Rawson N.S.
        • Malcolm E.
        Validity of the recording of ischaemic heart disease and chronic obstructive pulmonary disease in the Saskatchewan health care datafiles.
        Stat Med. 1995; 14: 2627-2643
        • Suissa S.
        Novel approaches to pharmacoepidemiology study design and statistical analysis.
        in: Strom B.L. Pharmacoepidemiology. John Wiley & Sons, New York1994: 629-645
        • Lubin J.H.
        • Gail M.H.
        Biased selection of controls for case-control analyses of cohort studies.
        Biometrics. 1984; 40: 63-75
        • Robins J.M.
        • Prentice R.L.
        • Blevins D.
        Designs for synthetic case-control studies in open cohorts.
        Biometrics. 1989; 45: 1103-1116
        • Von Korff M.
        • Wagner E.H.
        • Saunders K.
        A Chronic Disease Score from automated pharmacy data.
        J Clin Epidemiol. 1992; 45: 197-203
        • Garcia Rodriguez L.A.
        • Cattaruzi C.
        • Troncon M.
        • Agostinis L.
        • Grazia
        Risk of hospitalization for upper gastrointestinal bleeding associated with Ketorolac, other nonsteroidal anti-inflammatory drugs, calcium antagonists, and other antihypertensive drugs.
        Arch Intern Med. 1998; 158: 33-39
        • Rostagno C.
        • Abbate R.
        • Gensini G.F.
        • et al.
        In vitro effects of two novel calcium antagonists (nitrendipine and nisoldipine) on intraplatelet calcium redistribution, platelet aggregation and thromboxane A2 formation.
        Thrombos Res. 1991; 63: 457-462
        • Blache D.
        • Ojeda C.
        Comparative inhibitory effects of dihydropyridines on platelet aggregation, calcium uptake and cyclic AMP concentration.
        Pharmacology. 1992; 45: 250-259
        • Pietraszek M.H.
        • Chabielska E.
        • Buczko W.
        Effects of verapamil, diltiazem and nifedipine on some parameters of hemostasis in the rat.
        Thrombos Res. 1988; 50: 325-328
        • Ding Y.A.
        • Han C.L.
        • Chou T.C.
        • et al.
        Effects of the calcium antagonist isradipine on 24-hour ambulatory blood pressure, platelet aggregation, and neutrophil oxygen free radicals in hypertension.
        J Cardiovasc Pharmacol. 1992; 19: S32-S37
        • Rostagno C.
        • Prisco D.
        • Paniccia R.
        • et al.
        Effects of calcium channel blockers on platelet aggregation and thromboxane A2 formation.
        Thrombos Res. 1990; 59: 531-539
        • Feinberg W.M.
        • Bruck D.C.
        Effect of oral nimodipine on platelet function.
        Stroke. 1993; 24: 10-13
        • Pickard J.D.
        • Murray G.D.
        • Illingworth R.
        • et al.
        Effect of oral nimodipine on cerebral infarction and outcome after subarachnoid haemorrhage.
        BMJ. 1989; 298: 636-642
        • Kario K.
        • Imiya M.
        • Ohta Y.
        • Shimada K.
        J Hum Hypertens. 1995; 9 (Letter): 206-207
        • Grimm I.
        • Shaheen N.
        Calcium channel blockers and gastrointestinal hemorrhage.
        Gastroenterology. 1996; 111: 1391-1393
        • Grodecki-De Franco P.V.
        • Steinbul S.R.
        • Taylor P.C.
        • et al.
        Calcium antagonist use and perioperative bleeding complications: an analysis of 5,157 patients.
        Circulation. 1996; : I-476
        • Poynard T.
        • Cales P.
        • Pasta L.
        • et al.
        Beta-adrenergic- antagonist drugs in the prevention of gastrointestinal bleeding in patients with cirrhosis and esophageal varices. An analysis of data and prognostic factors in 589 patients from four randomized clinical trials. Franco-Italian Multicenter Study Group.
        NEJM. 1991; 324: 1532-1538
        • Bernard B.
        • Lebrec D.
        • Mathurin P.
        • et al.
        Beta-adrenergic antagonists in the prevention of gastrointestinal rebleeding in patients with cirrhosis.
        Hepatology. 1997; 25: 63-70
        • Bernstein D.E.
        • Phillips R.S.
        Portal hypertensive gastropathy.
        Gastrointest Endosc Clin North Am. 1996; 6: 697-708
        • Bhandare P.
        • Diniz-D’Souza R.
        • Mainker A.
        • Dhume V.
        Protective effect of propranolol on ethanol-induced gastric lesions in mice.
        Eur J Pharmacol. 1990; 191: 167-172
        • Rataboli P.V.
        • Bhandare P.N.
        • D’Souza R.S.
        • Dhume V.G.
        Protective effect of propranolol on ethanol-induced gastric lesions in rats.
        Ind J Physiol Pharmacol. 1992; 36: 35-38
        • Brandsborg O.
        • Brandsbrog M.
        • Christesen N.J.
        The role of beta-adrenergic receptor in the secretion of gastrin.
        Eur J Clin Invest. 1976; 6: 395-401
        • Daly M.J.
        The classification of adrenoceptors and their effects on gastric acid secretion.
        Scand J Gastroenterol. 1984; 89: 3-9
        • Asnchelewitsch J.
        • Okun K.
        Adrenegic processes and partial secretion of the stomach.
        Zeitschrift fur die Gesamte Innere Medizin und ihre Grenzgebiete. 1976; 31: 277-280
        • Jick H.
        • Porter J.
        Drug-induced gastrointestinal bleeding. Report from the Boston Collaborative Drug Surveillance Program, Boston University Medical Center.
        Lancet. 1978; 2: 87-89
        • Ryan J.R.
        • McMahon F.G.
        • Akdamar K.
        • et al.
        Mucosal irritant potential of a potassium-sparing diuretic and of wax-matrix potassium chloride.
        Clin Pharmacol Therapeut. 1984; 35: 90-93
        • Gleerup G.
        • Peterson J.R.
        • Mehlsen J.
        • Winther K.
        Effect of spirapril and hydrochlorothiazide on platelet function and euglobin clot lysis time in patients with mild hypertension.
        Angiology. 1996; 47: 9541-9545
        • Touyz R.M.
        • Schiffrin E.L.
        Role of protein kinase C in the anti-aggregatory effects of endothelin-1 on human platelets.
        Clin Sci. 1995; 88: 277-283
        • Rendu F.
        • Bachelot C.
        • Molle D.
        • et al.
        Indapamide inhibits human platelet aggregation in vitro.
        J Cardiovasc Pharmacol. 1993; 22: S57-S63
        • Garcia Rodriguez L.A.
        • Walker A.M.
        • Perez Gutthann S.P.
        Nonsteroidal antiinflammatory drugs and gastrointestinal hospitalizations in Saskatchewan.
        Epidemiology. 1992; 3: 337-342
        • Guess H.A.
        • West R.
        • Strand L.M.
        • et al.
        Fatal upper gastrointestinal hemorrhage or perforation among users and nonusers of anti-inflammatory drugs in Saskatchewan, Canada, 1983.
        J Clin Epidemiol. 1988; 41: 35-45