Digestive Diseases and Sciences

, Volume 56, Issue 8, pp 2349–2353

Association Between Proton Pump Inhibitor Use and Anemia: A Retrospective Cohort Study


    • Department of MedicineMichigan State University
  • Chethan Puttarajappa
    • Department of Medicine, Renal-Electrolyte DivisionUniversity of Pittsburgh
  • Yan Xie
    • Center for Statistical Training and ConsultingMichigan State University
  • Madhusudan Grover
    • Division of Gastroenterology and HepatologyMayo Clinic
  • Heather Laird-Fick
    • Department of MedicineMichigan State University
Original Article

DOI: 10.1007/s10620-011-1589-y

Cite this article as:
Sarzynski, E., Puttarajappa, C., Xie, Y. et al. Dig Dis Sci (2011) 56: 2349. doi:10.1007/s10620-011-1589-y



Proton pump inhibitors (PPIs) are widely prescribed to treat gastrointestinal diseases. However, concerns have been raised regarding their long-term use. Gastric acid suppression may decrease iron absorption, and it remains uncertain whether iron-deficiency anemia may result from chronic PPI therapy.


We aimed to explore the association between chronic PPI use and iron-deficiency anemia.


We conducted a retrospective cohort study of adult patients in an academic outpatient setting who received PPI therapy for at least 1 year between January 1, 2004 and January 1, 2006. We compared the change in hematologic indices among patients receiving PPI therapy for at least 1 year with matched controls.


Of the 98 patients on chronic PPI therapy who met inclusion criteria, 35% had no documented indication for such therapy. At baseline, demographics and hematologic indices were similar between PPI-users and controls. Among patients on PPI therapy, all hematologic indices decreased from baseline, including hemoglobin (−0.19 g/dL, P = 0.03), hematocrit (−0.63%, P = 0.02), and mean corpuscular volume (−0.49 fL, P = 0.05). PPI users had significant decreases in mean hemoglobin and hematocrit (P < 0.01 for both) compared with matched controls. After adjustment for confounders, including rates of esophagogastroduodenoscopy, colonoscopy and remote cancer status, the odds ratio of decreasing hemoglobin by 1.0 g/dL while on chronic PPI therapy was 5.03 (95% CI, 1.71–14.78, P < 0.01), while the odds ratio of decreasing hematocrit by 3% was 5.46 (95% CI, 1.67–17.85, P < 0.01).


Among adult patients receiving chronic PPI therapy, there is a significant decrease in hematologic indices from baseline.


Proton pump inhibitorAnemiaMalabsorptionIron deficiency


Since their advent in the late 1980s, proton pump inhibitors (PPIs) have become widely used for the treatment of many upper gastrointestinal disorders, including gastroesophageal reflux disease, peptic ulcer disease and stress ulcer prophylaxis. In 2009 in the United States alone, 119.4 million prescriptions for PPIs were dispensed totaling sales of $13.6 billion [1, 2]. Overutilization of PPIs is well documented [3]. In one series of hospitalized patients, rates of PPI use increased six-fold from admission to discharge, presumably because of failure to discontinue PPIs started for stress ulcer prophylaxis [4]. More recently, adverse effects of long-term PPI therapy such as interactions with clopidogrel that affect efficacy after acute coronary syndrome, risk for hip fracture, and association with community-acquired pneumonia have gained significant scientific and public attention [57]. In May 2010, the U.S. Food and Drug Administration revised the prescription and over-the-counter labels for PPIs to include new safety information about a possible increased risk of fractures of the hip, wrist, and spine [8]. Understanding the effects of chronic acid-suppression caused by PPI therapy is of great clinical value.

In the normal physiologic state, gastric acid facilitates the release of cyanocobalamin from its protein-bound form in the diet, and several studies have demonstrated a link between long-term PPI use and decreased cyanocobalamin absorption [9, 10]. Gastric acid also facilitates absorption of dietary non-heme iron by converting it from a non-absorbable ferric form to an absorbable ferrous form [11]. Hypochlorhydric states, including vagotomy and partial gastrectomy, have been associated with iron deficiency [1214].

While it seems intuitively possible that chronic PPI use may reduce dietary iron absorption and lead to iron deficiency anemia, only a few studies have examined this potential association, with conflicting results. In one study by Koop and Bachem, patients on chronic omeprazole therapy for refractory peptic ulcer disease did not develop decreased serum iron or ferritin levels over a 4 year follow-up [15]. However, the study was small and baseline iron and ferritin levels were not assessed. Similarly, patients with Zollinger-Ellison Syndrome receiving long-term PPI therapy did not develop iron deficiency anemia or deplete their iron stores [16]. More recently, Hutchinson et al. observed that chronic use of PPIs in patients with hereditary hemochromatosis suppressed dietary iron absorption and resulted in significantly fewer phlebotomies over time [17]. In a case report of two patients with known iron-deficiency anemia who failed oral iron replacement therapy while concurrently taking PPIs, iron status improved once PPIs were discontinued [18]. Given the conflicting results and poor generalizability of prior studies, we designed a retrospective cohort study to see if the sustained use of PPIs leads to iron deficiency anemia in adults seen in a primary care practice.


Study Design

We conducted a retrospective cohort study to determine whether chronic PPI use is associated with development of iron deficiency anemia. We used the Michigan State University Health Team electronic medical record database to identify study participants. This database is a computerized medical record system of approximately 150,000 patients cared for by general medicine, family medicine, pediatrics, surgery, obstetrics and gynecology, and subspecialty clinics. The database is representative of the urban population of Lansing, Michigan and surrounding suburban areas. The electronic medical record contains patient demographics, current and previous diagnoses classified by ICD-9 codes, current and previous medications and dosages, hospitalization records, and sub-specialty consultation records. Our study was approved by the Michigan State University Institutional Review Board.

Study Cohort

We identified 1,700 adult (age 18–80) patients who were initiated on PPIs (including esomeprazole, lansoprazole, omeprazole, pantoprazole, and rabeprazole) between January 1, 2004 and December 31, 2006 in one of our primary care practices. Of these subjects, 1,140 had at least one hemoglobin value documented in their chart. Subjects were excluded because of co-morbidities known to cause anemia as identified by ICD-9 codes in their problems lists (i.e., gastrointestinal bleeding, dysfunctional uterine bleeding, chronic kidney disease, inherited hemoglobinopathies, chronic anticoagulation with coumadin or low-molecular weight heparin, hemolysis, vitamin B12 or folate deficiency, pregnancy, or active cancer) (n = 261) or lack of hematologic studies drawn prior to and 1 year after initiation of PPI therapy (n = 781). Some patients met several exclusion criteria. The remaining 98 subjects (8.6%) composed the eligible study cohort (Fig. 1). Using a random number generator, age and sex-matched controls were identified using the same inclusion and exclusion criteria.
Fig. 1

Flow chart of study design

Data Extraction

The primary exposure of interest was PPI use for more than 1 year. Duration and dose of medication, as well as changes in medications, were tabulated. Hematologic indices (including hemoglobin, hematocrit, red blood cell count, mean corpuscular volume, white blood cell count, and platelet count) were recorded prior to and after at least 1 year of PPI therapy. Additional laboratory data, including ferritin and iron studies, were recorded if available. If multiple hematologic indices were available, we recorded those values in closest proximity to the first year of PPI therapy. Dates of laboratory results were recorded in order to assess the length of time between collection and initiation of PPI therapy.

Statistical Analysis

We performed descriptive statistical analyses on baseline demographics, as well as a list of potential confounders that could bias any association between PPI use and anemia, including body mass index, medications (e.g., aspirin, clopidogrel, non-steroidal anti-inflammatory drugs, and oral iron replacement therapy), surgery during PPI therapy, H. pylori status, rates of peptic ulcer disease (PUD), hematuria, epistaxis, and results of endoscopic studies (esophagogastroduodenoscopy and colonoscopy). A general linear model was used to compare the change in hematologic indices from baseline between PPI users and controls, with and without adjustment for identified confounders (variables with P ≤ 0.05). Matched paired t-test was used to compare the change in hematologic indices. Univariate and multivariate logistic regression was used to calculate the odds ratios (ORs) and 95% confidence intervals for changes in hematologic indices between PPI-users and matched controls. Data analysis was implemented using Statistical Analysis Software (SAS) version 9.2.


There were 98 subjects who met inclusion criteria and 98 controls. Baseline hematologic indices and medication use were similar for both groups (Table 1). Only one PPI-user and two controls had serial ferritin levels; only one PPI-user and two controls had serial iron studies documented in their charts, and therefore these data are not reported. Rates of aspirin and NSAID use were similar between PPI-users and controls (P = 0.11 and P = 0.67, respectively). Similarly, clopidogrel use was not significantly different between PPI-users and controls (P = 0.50, Table 1). Rates of documented PUD, epistaxis, hematuria, and H. pylori infection were rare and not significantly different among PPI-users and controls (Table 1). However, documented endoscopic procedures, including EGD and colonoscopy, were significantly higher among PPI-users than non-users (Table 1), and therefore these confounders were adjusted for in the odds ratio calculation.
Table 1

Cohort characteristics and baseline hematologic indices


PPI users


P value

Age, mean (SD)

56.67 (12.40)

55.21 (12.73)


BMI, mean (SD)

30.47 (7.40)

29.45 (6.09)


Gender, n (%)


57 (58.16)

57 (58.16)



41 (41.84)

41 (41.84)


Hgb (g/dL), mean (SD)

14.30 (1.33)

14.29 (1.13)


HCT (%), mean (SD)

42.04 (3.64)

41.83 (2.87)


MCV (fL), mean (SD)

90.03 (4.64)

89.24 (4.69)


NSAID use, n (%)

41 (41.84)

37 (37.76)


ASA use, n (%)

33 (33.67)

22 (22.45)


Clopidogrel use, n (%)

6 (6.12)

3 (3.06)


Documented EGD

34 (34.69)

2 (2.04)


Documented colonoscopy

40 (40.82)

25 (25.51)



4 (4.08)

0 (0.00)



0 (0.00)

0 (0.00)



3 (3.06)

0 (0.00)


H. pylori

2 (2.04)

0 (0.00)


History of cancer

14 (14.29)

5 (5.10)


Documented bleeding event

0 (0.00)

1 (1.02)


Iron supplementation

3 (3.06)

2 (2.04)


BMI body mass index, Hgb hemoglobin, HCT hematocrit, MCV mean corpuscular volume, NSAID non-steroidal anti-inflammatory drug, ASA aspirin, EGD esophagogastroduodenoscopy, PUD peptic ulcer disease

On average, hematologic indices were collected 6.63 months prior to and 20.16 months after commencing PPI therapy. For controls, the two comparison hematologic indices were drawn on average 24.3 months apart. Of the patients receiving chronic PPI therapy, only 65% had a clearly documented indication in their problem list. Among PPI-users, 34.7% used omeprazole, 32.7% esomeprazole, 15.3% lansoprazole, 13.3% pantoprazole, and 4.1% rabeprazole (Table 2).
Table 2

Proton pump inhibitor (PPI) use by brand

Brand of PPI












Among patients on PPI therapy, all hematologic indices decreased from baseline, including hemoglobin (−0.19 g/dL, SE ± 0.09, P = 0.03), hematocrit (−0.63%, SE ± 0.27, P = 0.02), and mean corpuscular volume (−0.49 fL, SE ± 0.25, P = 0.05), while those for controls did not (Fig. 2). Compared with controls, PPI users had significant decreases in mean hemoglobin and hematocrit (P < 0.01 for both), and this effect persisted after controlling for confounders (Fig. 2).
Fig. 2

Change in hematologic indices (± SEM) in patients before and after initiating proton pump inhibitor (PPI) therapy, compared with patients not receiving PPI therapy. SEM standard error of mean

Among PPI-users, 21 subjects (21.4%) had a decrease in their hemoglobin >1.0 g/dL while on therapy for more than a year (Table 3). The adjusted odds ratio (OR) for a 1.0 g/dl decrease was 5.03 (95% CI, 1.71–14.78). Similarly, the adjusted odds ratio of decreasing hematocrit by 3% was 5.46 (95% CI, 1.67–17.85). The odds ratio of decreasing mean corpuscular volume by 2 fL was 2.49 (95% CI, 1.20–5.17), yet after adjusting for confounders, this change was no longer significant.
Table 3

Odds ratios of decreasing hematologic indices after 1 year of proton pump inhibitor (PPI) therapy, before and after adjusting for confounders (rates of EGD, colonoscopy, and remote cancer status)

Hematologic index



OR (95% CI)


OR (95% CI)


Unadjusted data

Adjusted data

Hgb decrease >1 g/dL



5.07 (1.83–14.08)


5.03 (1.71–14.78)






HCT decrease >3%



5.29 (1.72–16.27)


5.46 (1.67–17.85)






MCV decrease >2 fL



2.49 (1.20–5.17)


1.77 (0.77–4.05)






Hgb hemoglobin, HCT hematocrit, MCV mean corpuscular volume, EGD esophagogastroduodenoscopy, OR odds ratio, CI confidence interval


We found a significant decrease in both hemoglobin and hematocrit in patients taking PPIs for more than 1 year, compared with age- and sex-matched controls, even after controlling for potential confounders. The changes we detected exceeded one standard deviation of the mean for our regional laboratory. More than 20% of subjects had a clinically significant decrease (hemoglobin ≥ 1.0 g/dL and/or hematocrit ≥ 3%) after receiving PPI therapy for at least 1 year. Given the widespread use of PPIs, even a modest increase in adverse events such as anemia could have significant clinical and economic implications. To our knowledge, this is the first study to systematically evaluate hematologic indices of patients on chronic PPI therapy.

Our study utilized a primary care population of urban and suburban subjects, thereby increasing the generalizability of our results. Hematologic indices available through the electronic medical record are comprehensive, as automatic electronic updates link the medical record with regional laboratories. We stringently applied exclusion criteria to limit confounding, and adjusted for potential confounders in our odds ratio analyses.

Our study limitations stem primarily from its retrospective design and small sample size. We attempted to balance the exclusion of subjects for potential confounders with the need for adequate sample size. We included subjects using aspirin, clopidogrel, and NSAIDs, despite the associated risk for occult gastrointestinal blood loss, and then compared the rate of use between groups. Although rates of ASA, NSAIDs, and clopidogrel use were slightly higher among PPI-users than controls, the differences were not statistically significant. Despite this strategy, our sample size was still small. Many potential subjects were eliminated because they did not have blood counts drawn prior to and 1 year after initiation of PPI therapy (n = 781 of 1,140 potential subjects).

Other potential confounders, including over-the-counter medications such as H2 blockers, antacids, and vitamins were difficult to assess because of inconsistent documentation in the EMR. We did not collect data on use of other prescription drugs that may alter gastric pH or decrease platelet aggregation, such as sucralfate or selective serotonin reuptake inhibitors. Similarly, we were unable to assess subjects’ medication adherence or dietary iron consumption. Therefore we cannot exclude a non-random distribution of these characteristics between the two groups. Another pertinent limitation is that menstrual blood loss is inconsistently documented in the EMR. Therefore we excluded female patients with a history of excessive menstrual blood loss (as documented in their problem lists, ICD-9 code 626.X) and used age- and sex-matched controls to minimize this potential confounder.

Lastly, our data does not confirm iron deficiency as the cause for the observed decline in hematologic values, and serial ferritin levels were available for only a few subjects. The trend of decreasing MCV for PPI-users and the pathophysiologic mechanisms for impaired iron absorption suggests but does not confirm this condition [11, 19].

Despite the limitations of our study, we feel our results warrant attention and provide an impetus for additional studies. A larger prospective study could include additional variables (ferritin levels, over-the-counter medication use, documentation of menstrual blood loss, and medication adherence, etc.), ensure availability of measures for all subjects, and include a longer observation period to confirm or refute our findings.

Although PPIs have revolutionized the management of many gastrointestinal disorders, there is growing concern about the potential adverse effects associated with the chronic use of these medications. Several review articles highlight these concerns [20, 21]. Our results demonstrate that anemia may be another potential consequence of chronic PPI therapy. In our study of adult patients receiving chronic PPI therapy, we found a significant decrease in hemoglobin and hematocrit levels. The widespread use of PPIs warrants further exploration to determine if iron-deficiency anemia may result from chronic PPI therapy.


We are grateful to Adriano Tonelli, MD for providing the inspiration for this research. We also thank Abhimanyu Beri, MD and Sahibzada Usman Latif, MD for their critical review of the manuscript.

Conflict of interest

None reported for any authors.

Copyright information

© Springer Science+Business Media, LLC 2011