1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. Study Limitations
  7. Conclusion
  8. Acknowledgments and disclosures
  9. References

Calcium channel blockers (CCBs) are an important class of medication useful in the treatment of hypertension. Several observational studies have suggested an association between CCB therapy and gastrointestinal (GI) hemorrhage. Using administrative databases, the authors re-examined in a post-hoc analysis whether the Antihypertensive and Lipid-Lowering Treatment to Prevent Heart Attack Trial (ALLHAT) participants randomized to the CCB amlodipine had a greater risk of hospitalized GI bleeding (a prespecified outcome) compared with those randomized to the diuretic chlorthalidone or the angiotensin-converting enzyme inhibitor lisinopril. Participants randomized to chlorthalidone did not have a reduced risk for GI bleeding hospitalizations compared with participants randomized to amlodipine (hazard ratio [HR], 1.09; 95% confidence interval [CI], 0.92–1.28). Those randomized to lisinopril were at increased risk of GI bleeding compared with those randomized to chlorthalidone (HR, 1.16; 95% CI, 1.00–1.36). In a post-hoc comparison, participants assigned to lisinopril therapy had a higher risk of hospitalized GI hemorrhage (HR, 1.27; 95% CI, 1.06–1.51) vs those assigned to amlodipine. In-study use of atenolol prior to first GI hemorrhage was related to a lower incidence of GI bleeding (HR, 0.69; 95% CI, 0.57–0.83). Hypertensive patients on amlodipine do not have an increased risk of GI bleeding hospitalizations compared with those taking either chlorthalidone or lisinopril.

Calcium channel blockers (CCBs) are a commonly prescribed class of antihypertensive medications.[1] Numerous studies have demonstrated the efficacy of CCBs in controlling blood pressure (BP) and preventing the adverse effects of chronic hypertension, except for heart failure.[2-9] In the previous 2 decades, several investigators have raised concern about possible prohemorrhagic activity of CCBs, citing an association between use of CCBs and the occurrence of cerebral,[10] surgical,[11] or gastrointestinal (GI) hemorrhage.[12-17] Laboratory studies have also suggested that both dihydropyridine and nondihydropyridine CCBs impair platelet activity, providing a possible mechanistic basis for the putative prohemorrhagic activity of CCBs.[18-21]

Nevertheless, the evidence regarding risk of bleeding associated with CCBs is inadequate for several reasons. First, there is a lack of randomized clinical trial evidence supporting the association between CCBs and GI hemorrhage, relative to other classes of antihypertensive agents. Second, there has been a failure to account for the differences between the 2 classes (dihydropyridine vs nondihydropyridine) of CCBs or specific agents.

Because uncontrolled hypertension is a prevalent and significant risk factor for cardiovascular disease, appropriate treatment of hypertension with proven safe medications is a critically important aspect of efforts to improve public health. Adequately powered, randomized clinical trials provide the best available evidence for determining the efficacy and safety of therapeutic agents. The Antihypertensive and Lipid-Lowering Treatment to Prevent Heart Attack Trial (ALLHAT)[8] is the largest single randomized clinical trial undertaken to compare the relative efficacy of antihypertensive therapy in preventing hypertension-related complications. The trial, which compared lisinopril-, amlodipine-, and doxazosin-based regimens to a chlorthalidone-based reference population, has provided significant data regarding the safety and efficacy of these specific drugs, representing 4 classes of antihypertensive agents. ALLHAT previously reported the risk of hospitalization with GI hemorrhage for the prespecified comparisons of amlodipine[8] (hazard ratio [HR], 0.92; 95% confidence interval [CI], 0.82–1.03) or lisinopril[8] (HR, 1.11; 95% CI, 0.99–1.24) with chlorthalidone, and for a post-hoc comparison of lisinopril with amlodipine[22] (HR, 1.20; 95% CI, 1.06–1.37).

The purpose of this report is to re-examine these results utilizing a cohort defined by a revised set of International Classification of Diseases (ICD) codes for GI bleeding hospitalizations and to explore in detail the issue regarding the association between selected classes of antihypertensive therapy and hospitalizations with GI hemorrhage.


  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. Study Limitations
  7. Conclusion
  8. Acknowledgments and disclosures
  9. References

The rationale and design of ALLHAT have been described previously.[23-25] Briefly, eligible participants for ALLHAT were men and women 55 years or older who had systolic BP (SBP) of at least 140 mm Hg and/or diastolic BP (DBP) of at least 90 mm Hg or took medication for hypertension and had at least 1 additional risk factor for coronary heart disease (CHD). These risk factors included previous myocardial infarction (MI) or stroke (>6 months or age indeterminate), other atherosclerotic cardiovascular disease (ASCVD), ischemic changes on electrocardiography within the past 2 years, left ventricular hypertrophy (LVH) by electrocardiography or echocardiography, type 2 diabetes, current cigarette smoking, and low levels of high-density lipoprotein (HDL) cholesterol (<35 mg/dL). Details of the inclusion and exclusion criteria have been described previously.[24, 25] Eligible participants were randomized into the trial at visit 2, at which time all prior antihypertensive medications were discontinued. From February 1994 to January 1998, 42,418 participants were recruited in 623 centers in the United States, Canada, Puerto Rico, and the US Virgin Islands.[25, 26] All participants signed an informed consent form, and all centers received institutional review board approval.

Participants were assigned by a computer-generated randomization schedule to 1 of 4 treatments: chlorthalidone, amlodipine, lisinopril, or doxazosin, in a ratio of 1.7:1:1:1, respectively. Randomization was stratified by center and blocked over time to maintain the ratio. More participants were assigned to chlorthalidone in accordance with Dunnett's[27] multiple comparison procedure for comparing 3 treatment groups to a single control group. The treatment goal in each study arm was a BP <140/90 mm Hg. Because the doxazosin study arm was terminated early, resulting in differential follow-up time, we have not included data from the doxazosin arm of ALLHAT.

All study drugs were identical in appearance and masked at each of 3 dosage levels: 12.5 mg/d, 12.5 mg/d (sham titration), and 25 mg/d for chlorthalidone; 2.5 mg/d, 5 mg/d, and 10 mg/d for amlodipine; and 10 mg/d, 20 mg/d, and 40 mg/d for lisinopril. If participants did not meet the BP goal at the maximum tolerated dosage for the initial medication, an open-label step 2 agent (but not a medication from the same class as any of the study drugs) was titrated in 3 doses until goal was reached. The step 2 agents were atenolol (25–100 mg/d), reserpine (0.05–0.2 mg/d), and clonidine (0.1–0.3 mg twice per day). If goal BP was still not achieved, an open-label step 3 agent, hydralazine (25–100 mg twice per day), could be added. After initial titration visits, participants were seen routinely every 3 months during the first year and every 4 months thereafter for an average of 4.9 years (range: 3 years, 8 months to 8 years, 1 month) of follow-up.

Nonstudy open-label drugs could also be added to or substituted for step 2 or step 3 open-label medications to improve tolerance or BP control. However, use of open-label medications from one of the masked classes of drugs was to be avoided unless a compelling indication, such as heart failure, arose for one of the masked classes of drugs. A medication dose could be decreased or a medication stopped if it was believed to be causing adverse effects.

Data regarding aspirin use were collected through patient self-report at baseline and every 2 years throughout the active study. Information regarding the use of nonsteroidal anti-inflammatory drugs (NSAIDs), anticoagulants and other antiplatelet agents, and alcohol use was not collected.

The occurrence of hospitalizations with GI bleeding (a prespecified secondary outcome[23, 28]) was not directly collected by ALLHAT clinics. Data regarding hospitalized GI bleeding (yes or no) were obtained through passive surveillance of participants treated as Medicare beneficiaries or within the Veterans Affairs (VA) health system for 1994 to 2001. Our initial cohort included 24,783 individuals who entered the Medicare system between 1994 and 2001 or were still in the VA system at the close of ALLHAT. Our new cohort consists of 20,844 individuals who were eligible for Medicare or in the VA systems at their entry into ALLHAT. As such, our new analysis includes 62% of participants randomized to the chlorthalidone-, lisinopril-, or amlodipine-based treatment arms in ALLHAT. Events were identified through CMS and VA data using relevant ICD-Ninth Revision (ICD-9) codes. In this analysis, GI bleeding was identified if the hospital records had one of the following ICD-9 codes as either a primary or secondary diagnosis code: 531.0/2/4/6, 532.0/2/4/6, 533.0/2/4/6, 534.0/2/4/6, 535.01/11/21/31/41/51/61, and 578.x.[29] The previous definition used in ALLHAT also used codes 459.0, 997.02, 998.1, 998.11, 998.12, 99.03, 99.04, and 99.05, which captured unspecified and procedural hemorrhages and transfusions.[8]

Statistical Methods

Baseline characteristics were compared for participants with and without hospitalized GI bleeding using Student's t test for continuous variables and contingency table analyses for categorical data, overall and by treatment group. Hospitalized GI bleeding data were analyzed according to participants' randomized treatment assignments regardless of their subsequent medication status (ie, intention-to-treat analysis). Cumulative event rates were calculated using the Kaplan-Meier procedure. Cox proportional hazards models were used to obtain HRs and 95% CIs for the time to the first hospitalized GI bleed, adjusting for both baseline and time-dependent covariates. No corrections are made for multiple comparisons.


  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. Study Limitations
  7. Conclusion
  8. Acknowledgments and disclosures
  9. References

Table 1 presents the baseline characteristics of the study population by occurrence of hospitalization with GI bleeding. During a mean of 4.9 years of active follow-up, we identified 915 participants who were hospitalized with GI bleeding as the primary or secondary diagnosis. Participants who developed hospitalized GI bleeding were older, more likely to be male and non-Hispanic, and had lower educational attainment. Compared with those without GI bleeding hospitalization, they were similarly likely to be taking aspirin at baseline (40.5% vs 39.4%) and more likely to have diabetes, a history of MI or stroke, a history of ASCVD, and/or LVH on baseline electrocardiography. They were less likely to be taking estrogen (among women) and/or to be enrolled in the ALLHAT Lipid-Lowering Trial (ALLHAT-LLT), likely because of the ALLHAT-LLT eligibility criteria. Aspirin use at baseline was not different between participants randomized to any of the treatment groups (data not shown). Sixty percent of participants who were not diabetic at baseline also took aspirin, compared with 40% of diabetic participants (data not shown).

Table 1. Baseline Characteristics by Occurrence of Hospitalized GI Bleeding
 Participants, No. (%)
GI BleedNo GI BleedP Value
  1. Abbreviations: CHD, coronary heart disease; CVD, cardiovascular disease; GI, gastrointestinal; HDL, high-density lipoprotein; LVH, left ventricular hypertrophy; MI, myocardial infarction; SD, standard deviation.

  2. a

    History of diabetes at baseline or fasting glucose ≥126 mg/dL.

Participants randomized, No.a91519,929 
Age, mean (SD), y72.0 (7.1)70.3 (6.8)<.001
55–64119 (13.0)3210 (16.1).012
65+796 (87.0)16,719 (83.9)
Women340 (37.2)8375 (42.0).004
Men575 (62.8)11,554 (58.0)
Black340 (37.2)6800 (34.1).06
Non-black575 (62.8)13,129 (65.9)
Hispanic119 (13.1)3278 (16.5).007
Non-Hispanic787 (86.9)16,561 (83.5)
Education, mean (SD), y10.3 (4.0)10.8 (4.1).001
Antihypertensive treatment
Treated849 (92.8)18,208 (91.4).13
Untreated66 (7.2)1720 (8.6)
Aspirin use at baseline371 (40.5)7853 (39.4).72
Women taking estrogen33 (10.0)1190 (14.5).03
HDL, mean (SD), mg/dL46.3 (15.5)46.6 (14.7).58
HDL <35 mg/dL103 (11.3)2470 (12.4).31
Diabetes classificationa
Diabetes409 (48.3)7845 (42.4).001
Nondiabetes437 (51.7)10,666 (57.6)
Body mass index, mean (SD), mg/kg228.8 (6.0)29.1 (5.8).13
Cigarette smoker (yes/no)186 (20.3)3713 (18.6).20
History of CHD281 (31.3)5699 (28.8).12
Atherosclerotic CVD558 (60.1)11,234 (56.8).006
History MI or stroke309 (33.8)5289 (26.5)<.001
History coronary revascularization143 (15.6)2985 (15.0).59
Other atherosclerotic CVD275 (30.1)5124 (25.7).003
ST-T wave88 (9.8)2080 (10.6).48
LVH by Minnesota code65 (8.2)913 (5.3)<.001
Lipid trial participants184 (20.1)4688 (23.6).02

The Figure1 shows the cumulative hospitalized GI bleeding rates by randomized treatment group, and Cox regression analysis (Table 2) revealed no significant difference between the chlorthalidone and amlodipine treatment groups (HR, 1.09; 95% CI, 0.92–1.28). However, when compared with amlodipine- or chlorthalidone-, the lisinopril-treated participants had a significantly higher risk of hospitalized GI bleeding (HR, 1.27; 95% CI, 1.06–1.51 and HR, 1.16; 95% CI, 1.00–1.36, respectively). There were no significant differences in treatment effect (ie, interactions) when analyzed by race, sex, ethnicity, aspirin use, or smoking at baseline. When limiting the sample to participants taking monotherapy, the findings were similar for the amlodipine vs chlorthalidone (HR, 1.05; 95% CI, 0.80–1.39) comparison. The HRs for the monotherapy cohort were higher, however, for the lisinopril vs amlodipine comparisons (HR, 1.52; 95% CI, 1.12–2.07) and for the lisinopril vs chlorthalidone comparisons (HR, 1.44, 95% CI, 1.12–1.87) (data not shown). Table 3 provides the cumulative incidence of participants, by treatment arm, with a hospitalized GI bleed by the end of years 1, 3, and 5. Of the 3 treatment arms, participants assigned to the amlodipine group experienced the lowest bleeding incidence over the entire trial, although it was statistically indistinguishable from the chlorthalidone group.

Table 2. Cox Regressions for Hospitalized GI Bleeding
 AmlodipineChlorthalidoneLisinoprilUnadjusted HR (95% CI)aLisinopril vs Chlorthalidone
Chlorthalidone vs AmlodipineLisinopril vs Amlodipineb
  1. Abbreviations: CI, confidence interval; GI, gastrointestinal.

  2. a

    Total hazard ratios (HRs) adjusted for race, sex, ethnicity, aspirin use at baseline, age, and smoking (not shown) are similar.

  3. b

    Post-hoc comparison.

Total226/5674410/9534279/56311.09 (0.92–1.28)1.27 (1.06–1.51)1.16 (1.00–1.36)
Black89/1917146/3216105/19280.98 (0.75–1.28)1.19 (0.90–1.58)1.22 (0.95–1.56)
Non-black137/3713264/6228174/36421.15 (0.94–1.42)1.31 (1.05–1.64)1.13 (0.94–1.37)
P value for interaction   .34.61.66
Male140/3281255/5472180/32871.10 (0.89–1.35)1.30 (1.04–1.62)1.19 (0.98–1.43)
Female86/2349155/397299/22831.07 (0.82–1.40)1.20 (0.90–1.61)1.12 (0.87–1.45)
P value for interaction   .90.70.75
Hispanic26/73848/123730/7551.10 (0.69–1.78)1.13 (0.67–1.92)1.03 (0.65–1.62)
Non-Hispanic200/4892362/8207249/48151.08 (0.91–1.29)1.28 (1.06–1.55)1.18 (1.01–1.39)
P value for interaction   .94.67.58
Aspirin at baseline92/2253159/3712120/22271.07 (0.82–1.38)1.35 (1.03–1.77)1.27 (1.00–1.61)
No aspirin at baseline133/3323247/5633156/32851.09 (0.88–1.35)1.19 (0.95–1.50)1.09 (0.89–1.33)
P value for interaction   .68.50.34
Age 55–64 y27/91457/151735/8631.30 (0.82–2.06)1.39 (0.84–2.30)1.07 (0.70–1.63)
Age 65+ y199/4716353/7927244/47071.06 (0.89–1.26)1.24 (1.03–1.50)1.18 (1.00–1.39)
P value for interaction   .41.69.68
Nonsmoker188/4577323/7672218/45451.03 (0.86–1.23)1.18 (0.97–1.44)1.15 (0.97–1.36)
Smoker38/105287/177261/10251.36 (0.93–1.99)1.68 (1.12–2.52)1.24 (0.89–1.71)
P value for interaction   .20.13.70
Table 3. Cumulative Proportion of Participants With Hospitalized GI Bleeding by Year and Treatment Group by Race, Sex, Ethnicity, Age, and Smoking
Treatment GroupCumulative Incidence of Hospitalized GI Bleed per 1000 Participants (95% CI) at the End of the Specified Year
Year 1Year 3Year 5
  1. Abbreviations: CI, confidence interval; GI, gastrointestinal.

Chlorthalidone7.9 (6.2–9.9)28.9 (25.6–32.6)46.6 (42.1–51.5)
Amlodipine7.9 (5.8–10.6)26.0 (22.0–30.7)44.4 (38.9–50.7)
Lisinopril11.1 (8.6–14.3)35.6 (30.9–41.0)54.0 (47.9–60.9)
Chlorthalidone8.1 (6.1–10.7)28.7 (24.7–33.3)45.1 (39.8–51.2)
Amlodipine9.1 (6.5–12.8)23.8 (19.3–29.5)40.6 (34.2–48.2)
Lisinopril10.7 (7.8–14.7)34.1 (28.5–40.8)51.8 (44.5–60.2)
Chlorthalidone7.4 (4.9–11.1)29.3 (23.8–36.1)49.4 (41.8–58.4)
Amlodipine5.4 (2.9–10.0)30.2 (23.2–39.4)51.9 (42.1–64.0)
Lisinopril11.9 (7.8–18.0)38.4 (30.4–48.5)58.3 (48.0–70.9)
Chlorthalidone8.3 (5.9–11.7)28.0 (23.1–33.9)42.5 (36.1–49.9)
Amlodipine9.6 (6.3–14.6)25.3 (19.5–32.7)41.6 (33.5–51.5)
Lisinopril10.9 (7.3–16.2)29.8 (23.3–38.0)50.0 (40.8–61.1)
Chlorthalidone7.5 (5.5–10.2)29.6 (25.3–34.6)49.5 (43.5–56.3)
Amlodipine6.6 (4.3–10.1)26.5 (21.3–32.9)46.6 (39.3–55.1)
Lisinopril11.2 (8.1–15.5)39.6 (33.3–47.2)56.9 (49.0–66.1)
Chlorthalidone8.4 (6.6–10.6)28.4 (24.9–32.4)46.4 (41.7–51.6)
Amlodipine7.6 (5.5–10.5)25.7 (21.4–30.7)44.5 (38.6–51.2)
Lisinopril11.5 (8.8–15.0)36.6 (31.5–42.6)54.9 (48.3–62.3)
Chlorthalidone4.3 (1.8–10.3)32.4 (23.5–44.6)47.4 (35.2–63.8)
Amlodipine10.0 (4.8–20.9)28.1 (18.0–43.7)45.4 (29.7–69.1)
Lisinopril8.3 (3.7–18.4)28.9 18.7–44.4)47.1 (32.8–67.3)
Aspirin at baseline
Chlorthalidone8.8 (6.2–12.5)29.9 (24.8–36.2)44.0 (37.4–51.8)
Amlodipine7.3 (4.5–11.9)24.1 (18.4–31.6)41.5 (33.5–51.4)
Lisinopril10.1 (6.7–15.3)40.0 (32.4–49.4)57.4 (47.9–68.7)
No aspirin at baseline
Chlorthalidone6.8 (4.9–9.4)28.0 (23.8–32.9)48.3 (42.4–54.9)
Amlodipine8.4 (5.8–12.2)27.4 (22.1–33.8)46.7 (39.4–55.4)
Lisinopril11.9 (8.7–16.4)32.5 (26.7–39.4)51.5 (43.8–60.4)
Age 55–64 y
Chlorthalidone4.1 (1.8–9.1)27.1 (19.8–37.1)38.5 (29.4–50.3)
Amlodipine1.1 (0.2–7.8)17.4 (10.5–28.7)30.4 (20.6–44.8)
Lisinopril8.3 (4.0–17.3)23.3 (14.9–36.2)42.5 (30.1–59.9)
Age 65+ y
Chlorthalidone8.6 (6.7–10.9)29.3 (25.6–33.4)48.1 (43.2–53.7)
Amlodipine9.2 (6.8–12.4)27.7 (23.2–33.0)47.2 (41.0–54.4)
Lisinopril11.6 (8.9–15.2)37.9 (32.6–44.0)56.2 (49.4–63.8)
Chlorthalidone7.8 (6.0–10.0)28.4 (24.8–32.5)44.9 (40.1–50.2)
Amlodipine8.5 (6.2–11.7)27.1 (22.6–32.4)45.1 (39.0–52.1)
Lisinopril11.8 (9.0–15.4)32.6 (27.7–38.4)51.5 (44.9–58.9)
Chlorthalidone8.2 (4.9–13.8)31.2 (23.7–40.9)54.4 (43.6–67.9)
Amlodipine4.9 (2.1–11.8)21.1 (13.6–32.5)41.8 (30.0–58.1)
Lisinopril8.1 (4.1–16.1)49.5 (37.2–65.8)65.8 (50.9–84.7)

Figure 1. Hospitalizations with gastrointestinal bleed by treatment group.

Download figure to PowerPoint

Table 4 provides data on comparative effects of antihypertensive therapy on the occurrence of hospitalizations with GI bleeding after adjusting for selected baseline characteristics and subsequent (in-trial) use of aspirin and atenolol. The adjustment essentially yielded similar results to those in Table 2. Baseline aspirin and atenolol use had no significant effect on hospitalized GI bleeding risk; however, in-trial atenolol use significantly reduced the risk for hospitalized GI bleeding (HR, 0.69; 95% CI, 0.57–0.83) whereas in-trial aspirin use did not affect subsequent risk. Age, current smoking, and male sex were also significant risk factors for hospitalized GI bleeding.

Table 4. The Effect of Antihypertensive Treatment on Risk of Hospitalized GI Bleeding Adjusting for Baseline Characteristics and In-Trial Use of Aspirin and Atenolol
CovariatesHazard Ratio (95% CI)
  1. Abbreviations: CI, confidence interval; GI, gastrointestintal.

  2. a

    Time-dependent covariate.

Chlorthalidone vs amlodipine1.06 (0.89–1.26)
Lisinopril vs amlodipine1.26 (1.04–1.53)
Lisinopril vs chlorthalidone1.20 (1.01–1.41
Baseline aspirin use (yes/no)1.07 (0.90–1.27)
Aspirin use ever prior to GI bleed (yes/no)a0.92 (0.78–1.08)
Baseline atenolol use (yes/no)1.04 (0.69–1.55)
Atenolol use ever prior to GI bleed (yes/no)a0.69 (0.57–0.83)
Age (per year)1.05 (1.04–1.06)
Black (yes/no)1.03 (0.97–1.32)
Male (yes/no)1.27 (1.04–1.48)
Hispanic (yes/no)0.99 (0.78–1.25)
Smoker (yes/no)1.35 (1.12–1.61)


  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. Study Limitations
  7. Conclusion
  8. Acknowledgments and disclosures
  9. References

During the previous 2 decades, several investigators have presented data associating calcium channel antagonists with GI hemorrhage,[12] “life-threatening bleeding,”[30] increased perioperative blood transfusion requirements,[16] and declining hemoglobin concentrations among hospitalized patients.[15]

This report demonstrates that participants randomized to the amlodipine-based antihypertensive regimen have a similar occurrence of hospitalizations with GI bleeding to that of participants randomized to the chlorthalidone-based therapy. When the lisinopril and amlodipine treatment arms were compared in a post-hoc comparison, participants randomized to the amlodipine regimen had a significantly lower occurrence of GI bleeding hospitalizations. In a protocol-specified comparison of lisinopril vs chlorthalidone treatment groups, participants in the chlorthalidone arm had a marginally lower risk of hospitalized GI bleed (HR, 1.16; 95% CI, 1.00–1.36), similar to that previously reported (HR, 1.11; 95% CI, 0.99–1.24).[8] The population in this manuscript consists of the 915 participants who were hospitalized for GI bleeding during ALLHAT. These participants were drawn from the 20,844 who were on Medicare or in the VA system from time of randomization and throughout the study. The previous analyses[8] were based on the 24,783 ALLHAT participants who entered Medicare at some time during the study or were still in the VA system at the end of ALLHAT. The smaller cohort in this study ensures a more consistent and complete follow-up time for all of the participants, thus allowing us to present a more in-depth analysis than that which was published previously.

These new analyses, which used a more stringent definition and more restricted cohort, were quite similar to those initially reported. Subgroup analyses by race, sex, ethnicity, and baseline aspirin use also did not identify any differences in risk for hospitalized GI bleeding between the chlorthalidone- and amlodipine-based treatment arms.

Data examining the association between CCB therapy and GI hemorrhage are sparse. Two randomized clinical trials have reported a significant association between CCBs and increased hemorrhage. However, these trials did not examine GI hemorrhage, and instead reported non–stroke-related bleeding[9] and major surgical bleeding[11] and were terminated prematurely[9, 11] or had a small number of endpoints.[11] In contrast to the clinical trials suggesting an association between CCB therapy and hemorrhage, ALLHAT has a longer follow-up, a much larger sample size, and a larger number of hospitalized GI bleeding outcomes.

The relationship between CCB therapy and GI bleeding in the cohort, case-control studies, and previously reported clinical trials is inconsistent. Our results are in agreement with the majority of published data that do not support an association between CCB therapy and GI hemorrhage. The cumulative evidence includes multiple case-control studies,[31, 35] a large retrospective cohort study involving more than 100,000 patients,[36] and a surgical study that did not demonstrate an association between perioperative CCB therapy and major bleeding (HR, 0.92; 95% CI, 0.65–1.28).[37] In addition, there are several lines of evidence that suggest that the association between GI bleeding and CCBs described in previous studies may be explained by factors that are not causal.

First, the Pahor[12] and Kaplan[13] studies compared CCB-treated patients to a reference group on β-adrenergic antagonist therapy. β-Adrenergic antagonists are known to prevent bleeding from gastric and esophageal varices.[38] In addition, because variceal bleeding has been estimated to be responsible for up to 30% of all upper GI tract bleeding,[39] it is possible that increased associations of bleeding observed relative to a β-adrenergic antagonist reference group[12, 14] reflect a therapeutic benefit of β-blockers rather than a prohemorrhagic effect of CCBs. Our analysis of data from ALLHAT suggests that the use of atenolol was associated with a lower rate of GI bleeding, although, in this trial, assignment to atenolol was not random and thus no definitive conclusion can be drawn. However, our finding is consistent with the previous observation that β-blockers reduce hospitalizations with GI bleeding.[38]

Second, it is possible that exposure misclassification could result in a spurious association between CCBs and bleeding. For example, Pahor and colleagues[12] comment that both “exposure and outcome may have been misclassified” in their study because medications were verified only at the inception of the study and 3 years afterward. This concern is supported by the observation that in their analysis,[12] the association between NSAID use and bleeding was much lower than that reported by other investigators.[32, 33]

Third, because a trend of decreasing association between GI bleeding and CCBs after adjustment for potential confounders was observed in multiple studies,[13-15, 31, 35] the association between CCB therapy and GI bleeding may have also represented an observation that participants taking CCBs were significantly more ill than the reference populations with which they were compared, as has been previously suggested in several observational studies.[14, 15, 31, 33] In addition, among studies that describe an association between CCBs and bleeding, no specific agent has been consistently associated with bleeding.[11-16]

The 5-year rates of hospitalized GI bleeding in ALLHAT were 44.4 events, 46.6 events, and 54.0 events per 1000 participants for amlodipine, chlorthalidone, and lisinopril (9.1, 9.5, and 11.0 events per 1000 person-years, respectively) (data not shown). These values are slightly higher relative to an ancillary study of hypertensive patients older than 65 years enrolled in the Cardiovascular Health Study (CHS) (9.9 and 6.0 events per 1000 person-years in individuals using CCBs or other antihypertensive agents, respectively).[13] The total incidence of GI bleeding across the CHS cohort was 6.8 events per 1000 person-years.[30] Several differences between ALLHAT and CHS may account for the different rates of GI bleeding between the 2 studies. First, the CHS cohort was likely to be healthier than ALLHAT participants at baseline. Second, ALLHAT obtained data from Medicare and VA databases. In contrast, CHS did not include VA patients, leaving the possibility that difference between the 2 study populations may exist. Third, although both studies reviewed hospital discharge records for selected ICD-9 codes related to GI bleeding, ALLHAT used a different number of ICD-9 codes and as such could have had a broader definition of GI bleeding (CHS used 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, 534.6, 569.0, 578.0, 578.1, and 578.9, respectively).[40] ALLHAT did not include code 569.0, but included the other codes used by CHS, as well as 459.0, 535.01, 535.11, 535.21, 535.31, 535.41, 535.51, and 535.61. The ICD-9 coding of GI bleeding employed by Pahor and colleagues[12] was more restrictive than that used by ALLHAT but cannot explain the differences in results with respect to hospitalized GI bleeding.[29]

Finally, the reasons behind the increased risk of bleeding among diabetic participants are not clear and are likely multifactorial but cannot solely be attributed to aspirin use.[41] In a recent paper, De Berardis and colleagues,[42] looking at a population-based cohort in Italy, used administrative data to determine that patients with diabetes had a high rate of bleeding, although it was not independently associated with aspirin use. In patients with acute coronary syndrome, GI hemorrhage is associated with diabetes as well as hypertension.[43] Further, diabetic patients with peptic ulcer bleeding have poorer outcomes, including greater mortality, than nondiabetic patients.[44] Diabetic angiopathy, impaired healing, autonomic neuropathy, and increased susceptibility to infection may be at least partially responsible for the poor outcomes and may contribute to the cause of the bleeding itself.[41, 44]

Surprisingly, our analysis found that participants randomized to a lisinopril-based regimen had the highest frequency of hospitalized GI bleeding events of the 3 randomized groups. Angiotensin-converting enzyme (ACE) inhibitor therapy has been associated with decreased fibrinogen and von Willebrand factor levels in patients with heart failure,[45] and has been associated with decreased levels of plasminogen-activating inhibitor-1 (PAI-1)[40, 46] and fibrinogen,[47] compared with patients treated with amlodipine. Amlodipine, in contrast, while not being associated with decreases in PAI-1 or fibrinogen, has been associated with an increase in tissue plasminogen activator levels.[40] The clinical significance of these measurements as related to the frequency of GI bleeding is not certain, and our review did not identify other clinical trials or case reports suggesting a relationship between ACE inhibition and increased GI bleeding.

Study Limitations

  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. Study Limitations
  7. Conclusion
  8. Acknowledgments and disclosures
  9. References

The present study does have limitations. Because amlodipine was the only CCB studied in ALLHAT, the results are limited in addressing the safety of other CCBs. However, our primary goal is not to prove the safety of other CCBs, but, in reporting the ALLHAT findings, to add to the large pool of existing data regarding CCB therapy and GI hemorrhage. Second, our analysis of VA and CMS data identified participants who were hospitalized with a GI hemorrhage, although we cannot determine whether the bleeding was a cause or consequence of hospitalization. However, any error related to this would be observed across the entire study and would not be expected to be the basis of a differential comparison in the 3 study arms included in this report. Finally, information on aspirin usage was collected only biannually and by patient report. Information on the use of NSAIDs or anticoagulation therapy was not collected.


  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. Study Limitations
  7. Conclusion
  8. Acknowledgments and disclosures
  9. References

In ALLHAT, amlodipine therapy did not increase the risk of hospitalized GI hemorrhage when compared with the participants treated with a diuretic. Furthermore, in a post-hoc comparison, participants treated with amlodipine had significantly fewer occurrences of hospitalized GI bleeding than participants randomized to a lisinopril-based regimen. The apparent increase in hospitalizations with GI bleeding in those treated with lisinopril has not been observed previously in trials of ACE inhibitors. The relationship between ACE inhibitor therapy and hospitalized GI hemorrhage may warrant further investigation in other studies.

Acknowledgments and disclosures

  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. Study Limitations
  7. Conclusion
  8. Acknowledgments and disclosures
  9. References

This research was supported by Health and Human Services contracts N01-HC-35130 and HHSN268201100036C from the National Heart, Lung, and Blood Institute, National Institutes of Health, US Department of Health and Human Services, Bethesda, MD. The ALLHAT investigators acknowledge contributions of study medications supplied by Pfizer, Inc (amlodipine and doxazosin), AstraZeneca (atenolol and lisinopril), and Bristol-Myers Squibb (pravastatin) and financial support provided by Pfizer, Inc. Dr Alderman has received research support from Daiichi Sankyo. Dr Davis has consulted for Amgen and Takeda. Dr Grimm has consulted for Merck, Pfizer, and Roche, and has received honoraria from Takeda. Dr Oparil has consulted for Schering Plough, Eli Lilly, Daiichi Sankyo, Boehringer Ingelheim, Novartis, NicOx, Pfizer, Omron Healthcare, and Forest Laboratories, and has received research support from Gilead, Daiichi Sankyo, Merck, and Amgen. Dr Probstfield has received research support from Abbott Laboratories, GlaxoSmithKline, and Sanofi Aventis. Drs Einhorn, Ford, Furberg, Jafri, Phillips, Piller, Whittle, and Williamson have no financial interests to report. Editorial assistance was provided by Ellen Breckenridge, PhD, of The University of Texas School of Public Health, Houston, TX, and was funded by NHLBI.


  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. Study Limitations
  7. Conclusion
  8. Acknowledgments and disclosures
  9. References
  • 1
    Stafford RS, Monti V, Furberg CD, Ma J. Long-term and short-term changes in antihypertensive prescribing by office-based physicians in the United States. Hypertension. 2006;48:213218.
  • 2
    Staessen JA, Fagard R, Thijs L, et al. Randomised double-blind comparison of placebo and active treatment for older patients with isolated systolic hypertension. The Systolic Hypertension in Europe (SYST-EUR) trial investigators. Lancet. 1997;350:757764.
  • 3
    Hansson L, Zanchetti A, Carruthers SG, et al. Effects of intensive blood-pressure lowering and low-dose aspirin in patients with hypertension: principal results of the Hypertension Optimal Treatment (HOT) Randomised Trial. HOT study group. Lancet. 1998;351:17551762.
  • 4
    Brown MJ, Palmer CR, Castaigne A, et al. Morbidity and mortality in patients randomised to double-blind treatment with a long-acting calcium-channel blocker or diuretic in the international nifedipine GITS study: intervention as a Goal In Hypertension Treatment (INSIGHT). Lancet. 2000;356:366372.
  • 5
    Hansson L, Hedner T, Lund-Johansen P, et al. Randomised trial of effects of calcium antagonists compared with diuretics and beta-blockers on cardiovascular morbidity and mortality in hypertension: the Nordic Diltiazem (NORDIL) study. Lancet. 2000;356:359365.
  • 6
    Hansson L, Lindholm LH, Ekbom T, et al. Randomised trial of old and new antihypertensive drugs in elderly patients: cardiovascular mortality and morbidity the Swedish Trial in Old Patients with hypertension-2 study. Lancet. 1999;354:17511756.
  • 7
    Rosei EA, Dal Palu C, Leonetti G, et al. Clinical results of the Verapamil in Hypertension and Atherosclerosis Study. VHAS investigators. J Hypertens. 1997;15:13371344.
  • 8
    ALLHAT Officers and Coordinators for the ALLHAT Collaborative Research Group. Major outcomes in high-risk hypertensive patients randomized to angiotensin-converting enzyme inhibitor or calcium channel blocker vs diuretic: the Antihypertensive and Lipid-Lowering Treatment to Prevent Heart Attack Trial (ALLHAT). JAMA. 2002;288:29812997.
  • 9
    Black HR, Elliott WJ, Grandits G, et al. Principal results of the Controlled Onset Verapamil Investigation of Cardiovascular Endpoints (CONVINCE) trial. JAMA. 2003;289:20732082.
  • 10
    Gore JM, Sloan M, Price TR, 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;83:448459.
  • 11
    Legault C, Furberg CD, Wagenknecht LE, et al. Nimodipine neuroprotection in cardiac valve replacement: report of an early terminated trial. Stroke. 1996;27:593598.
  • 12
    Pahor M, Guralnik JM, Furberg CD, et al. Risk of gastrointestinal haemorrhage with calcium antagonists in hypertensive persons over 67 years old. Lancet. 1996;347:10611065.
  • 13
    Kaplan RC, Heckbert SR, Koepsell TD, et al. Calcium channel blocker use and gastrointestinal tract bleeding among older adults. Age Ageing. 2002;31:217218.
  • 14
    Kaplan RC, Heckbert SR, Koepsell TD, et al. Use of calcium channel blockers and risk of hospitalized gastrointestinal tract bleeding. Arch Intern Med. 2000;160:18491855.
  • 15
    Zuccala G, Pedone C, Cocchi A, et al. Use of calcium antagonists and hemoglobin loss in hospitalized elderly patients: a cohort study. Gruppo Italiano di Farmacoepidemiologia nell'Anziano (GIFA) Investigators. Clin Pharmacol Ther. 2000;67:314322.
  • 16
    Zuccala G, Pahor M, Landi F, et al. Use of calcium antagonists and need for perioperative transfusion in older patients with hip fracture: observational study. BMJ. 1997;314:643644.
  • 17
    Pitt B, Byington RP, Furberg CD, et al. Effect of amlodipine on the progression of atherosclerosis and the occurrence of clinical events. PREVENT investigators. Circulation. 2000;102:15031510.
  • 18
    Pales J, Palacios-Araus L, Lopez A, Gual A. Effects of dihydropyridines and inorganic calcium blockers on aggregation and on intracellular free calcium in platelets. Biochim Biophys Acta. 1991;1064:169174.
  • 19
    Wallen NH, Held C, Rehnqvist N, Hjemdahl P. Platelet aggregability in vivo is attenuated by verapamil but not by metoprolol in patients with stable angina pectoris. Am J Cardiol. 1995;75:16.
  • 20
    Blache D, Ojeda C. Comparative inhibitory effects of dihydropyridines on platelet aggregation, calcium uptake and cyclic AMP concentration. Pharmacology. 1992;45:250259.
  • 21
    Berkels R, Klaus W, Boller M, Rosen R. The calcium modulator nifedipine exerts its antiaggregatory property via a nitric oxide mediated process. Thromb Haemost. 1994;72:309312.
  • 22
    Leenen FH, Nwachuku CE, Black HR, et al. Clinical events in high-risk hypertensive patients randomly assigned to calcium channel blocker versus angiotensin-converting enzyme inhibitor in the Antihypertensive and Lipid-Lowering Treatment to Prevent Heart Attack Trial. Hypertension. 2006;48:374384.
  • 23
    Davis BR, Cutler JA, Gordon DJ, et al. Rationale and design for the The Antihypertensive and Lipid-Lowering Treatment to Prevent Heart Attack Trial (ALLHAT) [design]. Am J Hypertens. 1996;9:342360.
  • 24
    Grimm RH Jr, Margolis KL, Papademetriou V, et al. Baseline characteristics of participants in the The Antihypertensive and Lipid-Lowering Treatment to Prevent Heart Attack Trial (ALLHAT). Hypertension. 2001;37:1927.
  • 25
    Wright JT Jr, Cushman WC, Davis BR, et al. The Antihypertensive and Lipid-Lowering Treatment to Prevent Heart Attack Trial (ALLHAT): clinical center recruitment experience. Control Clin Trials. 2001;22:659673.
  • 26
    Pressel S, Davis BR, Louis GT, et al. Participant recruitment in the antihypertensive and lipid-lowering treatment to prevent heart attack trial (ALLHAT). Control Clin Trials. 2001;22:674686.
  • 27
    Dunnett CW. A multiple comparisons procedure for comparing several treatments with a control. J Am Stat Assoc. 1955;50:10961121.
  • 28
    The Antihypertensive and Lipid-Lowering Treatment to Prevent Heart Attack Trial protocol.
  • 29
    Centers for Disease Control and Prevention (CDC). International classification of diseases, ninth revision, clinical modification (ICD-9-CM). Accessed May 9, 2012.
  • 30
    Kaplan RC, Heckbert SR, Koepsell TD, et al. Risk factors for hospitalized gastrointestinal bleeding among older persons. Cardiovascular Health Study investigators. J Am Geriatr Soc. 2001;49:126133.
  • 31
    Suissa S, Bourgault C, Barkun A, et al. Antihypertensive drugs and the risk of gastrointestinal bleeding. Am J Med. 1998;105:230235.
  • 32
    Pilotto A, Leandro G, Franceschi M, et al. Antagonism to calcium antagonists. Lancet. 1996;347:17611762.
  • 33
    Lanas A, Serrano P, Bajador E, et al. Risk of upper gastrointestinal bleeding associated with non-aspirin cardiovascular drugs, analgesics and nonsteroidal anti-inflammatory drugs. Eur J Gastroenterol Hepatol. 2003;15:173178.
  • 34
    Kelly JP, Laszlo A, Kaufman DW, et al. Major upper gastrointestinal bleeding and the use of calcium channel blockers. Lancet. 1999;353:559.
  • 35
    Garcia Rodriguez LA, Cattaruzzi C, Troncon MG, Agostinis L. Risk of hospitalization for upper gastrointestinal tract bleeding associated with ketorolac, other nonsteroidal anti-inflammatory drugs, calcium antagonists, and other antihypertensive drugs. Arch Intern Med. 1998;158:3339.
  • 36
    Smalley WE, Ray WA, Daugherty JR, Griffin MR. No association between calcium channel blocker use and confirmed bleeding peptic ulcer disease. Am J Epidemiol. 1998;148:350354.
  • 37
    Wijeysundera DN, Beattie WS, Rao V, et al. Calcium antagonists are associated with reduced mortality after cardiac surgery: a propensity analysis. J Thorac Cardiovasc Surg. 2004;127:755762.
  • 38
    Bernard B, Lebrec D, Mathurin P, et al. Beta-adrenergic antagonists in the prevention of gastrointestinal rebleeding in patients with cirrhosis: a meta-analysis. Hepatology. 1997;25:6370.
  • 39
    Laine L. Upper gastrointestinal tract hemorrhage. West J Med. 1991;155:274279.
  • 40
    Fogari R, Preti P, Lazzari P, et al. Effect of benazepril amlodipine combination on fibrinolysis in hypertensive diabetic patients. Eur J Clin Pharmacol. 2003;59:271275.
  • 41
    Sellin JH, Chang EB. Therapy insight: gastrointestinal complications of diabetes–pathophysiology and management. Nat Clin Pract Gastroenterol Hepatol. 2008;5:162171.
  • 42
    De Berardis G, Lucisano G, D'Ettorre A, et al. Association of aspirin use with major bleeding in patients with and without diabetes. JAMA. 2012;307:22862294.
  • 43
    Al-Mallah M, Bazari RN, Jankowski M, Hudson MP. Predictors and outcomes associated with gastrointestinal bleeding in patients with acute coronary syndromes. J Thromb Thrombolysis. 2007;23:5155.
  • 44
    Thomsen RW, Riis A, Christensen S, et al. Diabetes and 30-day mortality from peptic ulcer bleeding and perforation: a Danish population-based cohort study. Diabetes Care. 2006;29:805810.
  • 45
    Gibbs CR, Blann AD, Watson RD, Lip GY. Abnormalities of hemorheological, endothelial, and platelet function in patients with chronic heart failure in sinus rhythm: effects of angiotensin-converting enzyme inhibitor and beta-blocker therapy. Circulation. 2001;103:17461751.
  • 46
    Pahor M, Franse LV, Deitcher SR, et al. Fosinopril versus amlodipine comparative treatments study: a randomized trial to assess effects on plasminogen activator inhibitor-1. Circulation. 2002;105:457461.
  • 47
    Fogari R, Zoppi A, Malamani GD, et al. Effects of different antihypertensive drugs on plasma fibrinogen in hypertensive patients. Br J Clin Pharmacol. 1995;39:471476.