Iron Deficiency and Helicobacter pylori Infection in the United States

Abstract

Using data from the current National Health and Nutrition Examination Survey (1999–2000), the authors assessed whether Helicobacter pylori infection is associated with iron deficiency and iron-deficiency anemia (IDA) in the United States. Iron deficiency was defined as at least two abnormal results out of three biomarkers of iron stores. IDA was defined as a low hemoglobin level in the presence of iron deficiency. H. pylori infection was measured by serology. Complex survey estimators were used in the analysis. For 7,462 survey participants aged ≥3 years, H. pylori infection was associated with decreased serum ferritin levels (percent change = −13.9%, 95% confidence interval (CI): −19.5, −8.0) but not with levels of free erythrocyte protoporphyrin, transferrin saturation, or hemoglobin (percent change = 1.5%, −2.8%, and −1.1%, respectively). Multinomial logistic regression analyses indicated that H. pylori infection was associated with the prevalence of IDA (prevalence odds ratio (POR) = 2.6, 95% CI: 1.5, 4.6) and, to a lesser degree, other types of anemia (POR = 1.3, 95% CI: 1.0, 1.7). H. pylori infection was associated with a 40% increase in the prevalence of iron deficiency (POR = 1.4, 95% CI: 0.9, 2.0) after controlling for relevant covariates. In the United States, H. pylori infection was associated with iron deficiency/IDA regardless of the presence or absence of peptic ulcer disease.

Iron deficiency, defined as decreased total body iron content, is among the most common nutritional deficiencies in the world. Iron deficiency results in impairments in immune, cognitive, and reproductive functions, as well as decreased work performance. Iron deficiency develops through three stages: 1) iron depletion, 2) iron-deficient erythropoiesis, and 3) iron-deficiency anemia (IDA) (1). Although the mechanisms remain unclear, clinical and epidemiologic studies suggest that infection with Helicobacter pylori is associated with iron deficiency and IDA (2, 3). Barabino (2) hypothesized that gastritis increased levels of neutrophil-derived lactoferrin, and since H. pylori has a lactoferrin-binding protein receptor, the infection would result in increased iron losses related to bacterial turnover. A study of children in Alaska showed that anemia responded to oral iron replacement but recurred when iron therapy was discontinued, suggesting that mild chronic bleeding was involved (4). It seems likely that the pathogenesis is multifactorial, including combinations of reduced iron absorption related to decreased acid secretion, increased iron loss from microbleeding, and utilization by bacteria.

Successful interventions in the United States targeting high-risk groups for iron supplementation led to a significant reduction in iron deficiency and IDA in the 1960s (5–7). However, the impact of such programs has leveled off, particularly among women of childbearing age and toddlers (6). We used data from the current (1999–2000) National Health and Nutrition Examination Survey (NHANES) to assess the association between H. pylori infection and the prevalence of iron deficiency, IDA, and anemia from other causes in the most current large representative sample of the US population. Specifically, we tested whether mean levels of different markers of iron stores or prevalences of iron deficiency, IDA, and anemia from other causes differed according to H. pylori infection status. We also examined whether this relation was limited to adults with a history of peptic ulcer disease.

MATERIALS AND METHODS

Details on the design and methods of NHANES have been published elsewhere (8). In brief, the National Center for Health Statistics collects data on a representative sample of the civilian, noninstitutionalized US population in 2-year cycles. In the 1999–2000 NHANES, questionnaire, laboratory, and examination data were collected from 9,965 survey participants to obtain national estimates of the prevalence of selected health conditions. This cycle was the first phase of NHANES IV (1999–2004). By design, ethnic minorities are oversampled in NHANES to allow evaluation of the health status of specific racial/ethnic groups and comparisons between them.

Since 1976–1980 (i.e., NHANES II), the case definitions of iron deficiency and IDA used by the National Center for Health Statistics have been based on age- and gender-specific cutoff values (6). Anemia was considered present if hemoglobin levels were low according to these cutoff values. Iron deficiency was defined as at least two abnormal results in three tests of iron status measuring levels of free erythrocyte protoporphyrin, transferrin saturation, and serum ferritin. The specific laboratory methods used to measure levels of hemoglobin and iron indicators have been described elsewhere (6). Briefly, hemoglobin was measured using an automated electronic counter, while immunoradiometric, colorimetric, and fluorescence extraction assays were used to measure serum ferritin, iron and iron binding capacity, and free erythrocyte protoporphyrin, respectively (6). In the 1999–2000 NHANES, H. pylori infection was measured using the Wampole enzyme-linked immunosorbent assay (9) among 7,493 participants aged ≥3 years. Thus, the sample population for this cross-sectional study consisted of participants aged ≥3 years. We excluded the records of participants without valid information on anemia and iron-store status; this left an analytic sample of 7,462 participants.

In our analyses, we took into account the complex survey design of NHANES, using replicate weight jackknife estimators available in SAS-callable SUDAAN (version 8.1; Research Triangle Institute, Research Triangle Park, North Carolina) to characterize the distribution of indicators of iron stores and the prevalences of iron deficiency and IDA by H. pylori infection status and other selected variables. Mean values and standard errors for serum ferritin, percent transferrin saturation, and erythrocyte protoporphyrin were compared using multiple linear regression analysis that controlled for several potential confounders. We examined as potential confounders selected factors that have been previously found to be associated with iron deficiency or H. pylori infection in NHANES. Forty percent of the US population uses multivitamin and mineral supplements (10); information on iron intake from supplements was not readily available in the current NHANES. We also found evidence of confounding by channel: that is, persons with anemia or at high risk of anemia seemed more likely to report higher iron intakes in 24-hour dietary recalls. Therefore, we decided to limit the analysis to actual levels of biomarkers of iron stores.

Previous studies with NHANES data have used the poverty income ratio as a proxy for socioeconomic status (11). The poverty income ratio is computed as reported household income divided by the US poverty threshold, which is determined annually by the Bureau of the Census and adjusted for inflation. The National Center for Health Statistics calculated these values for each record. For assessment of renal insufficiency in adults, we used cutoff values of estimated glomerular filtration rates derived from serum creatinine levels as done in NHANES III (12). Levels of vitamin C have been found to be inversely related to H. pylori infection (13) and to enhance iron uptake (14). We selected variables for these linear and logistic models by entering those that separately or jointly changed the crude point estimates in a meaningful way (presented in the reduced model) (15). We also obtained results for all potential confounders (presented in the full model).

Thus, the analyses controlled for age, gender, ethnicity, place of birth, education, poverty income ratio, current pregnancy status, number of previous pregnancies, vitamin C intake, diabetes, hypertension, and renal insufficiency.

Multiple linear regression analysis was used to take advantage of the full range of values for the outcome variables. The levels of all markers of iron deficiency were logarithmically transformed, and the percent change was obtained by retransforming the values to the original scale. The logarithmic transformation centered the distribution of data on some markers and simplified the process of estimating differences between means, which were expressed as percent change. Multiple logistic regression analysis was used to model the prevalence of iron deficiency (determined to be present or absent using age- and gender-specific cutoff values). Prevalence odds ratios and their 95 percent confidence intervals were calculated in bivariate, stratified, and multivariate analyses of the association between H. pylori infection and iron deficiency, while multiple multinomial logistic regression was used to assess the relation between H. pylori infection and a three-level dependent variable: 1) IDA, 2) anemia from other causes, and 3) no anemia and normal values for iron-store indicators. Multinomial logistic regression allowed us to compare two categories of anemia (IDA and anemia from other causes) using a common referent.

Records that had missing data on covariates or were not applicable (e.g., for missing data on current pregnancy status, females other than those aged 12–59 years and all males) were treated as a separate category in multivariate analyses.

To address the specific question of whether the H. pylori association was explained by ulcers and bleeding resulting from this infection, we used data on history of peptic ulcer disease available for the subset of adult NHANES participants. Specifically, we created a four-level composite exposure variable that combined H. pylori infection status and history of peptic ulcer, using persons exposed to neither as the reference category. We assessed effect modification by relevant covariates, such as race/ethnicity, using this composite exposure variable approach. Statistical interaction in the context of linear regression was tested using a chi-squared test for product terms. Population attributable fraction estimates of the prevalence of iron deficiency due to H. pylori infection were calculated using Greenland estimators (16).

RESULTS

In the study population, 27.1 percent (95 percent confidence interval (CI): 24.8, 29.4), 6.6 percent (95 percent CI: 5.7, 7.6), 1.9 percent (95 percent CI: 1.4, 2.6), and 3.3 percent (95 percent CI: 2.7, 4.1) had H. pylori infection, iron deficiency, IDA, and anemia from other causes, respectively. The distribution of the correlates of these three conditions is shown in table 1. No results of statistical testing are shown in table 1; its main purpose is to depict the patterns of occurrence of the dependent and independent variables according to select covariates. H. pylori infection increased with age and was more prevalent among racial/ethnic minorities, persons born outside the United States, non-breastfed children, the poor, and persons in the lowest quintile of vitamin C intake. The prevalence of iron deficiency was increased among females, ethnic minorities, pregnant women, and women with multiple previous pregnancies. Some of the stratum-specific estimates of IDA were unreliable because of small numbers; however, increases in prevalence were noted among the same groups that experienced iron deficiency, as well as other subgroups (females aged 16–69 years and the foreign-born), while no differences in the prevalence of IDA were found by poverty income ratio. Anemia from other causes was more prevalent among non-Hispanic Blacks, non-Mexican-American Hispanics, poor persons, women who were currently pregnant or had had previous multiple pregnancies, persons who had been medically diagnosed with diabetes or high blood pressure, and persons with serum creatinine levels compatible with renal impairment. We found no relation between iron deficiency/IDA and smoking or regular alcohol intake, and these factors were not considered in further analyses as covariates.

Table 2 presents the results of multiple linear regression analysis of the biomarkers of iron deficiency. H. pylori infection was associated with an important decrease in levels of serum ferritin across the entire US population, after simultaneous adjustment for relevant covariates. The associations between H. pylori infection and other indicators of iron stores were of comparatively small magnitude or were null.

As table 3 shows, results of the multinomial analysis indicated that the prevalence of IDA among H. pylori-infected participants was increased. Point estimates were meaningfully changed by age, gender, ethnicity, and gestational history (current and past). In a reduced model, we added poverty to these variables as a proxy for socioeconomic status. The relations between the prevalence of H. pylori infection and IDA were similar in the reduced model after adjustment for this set of variables and in the fully adjusted model. Table 3 also shows the H. pylori infection prevalence odds ratio for anemia from other causes: prevalence was increased by 30 percent in both the reduced model and the full model. In addition, after adjustment for all covariates via multiple logistic regression analysis, the prevalence of iron deficiency increased by 40 percent among H. pylori-infected participants. Measures of association from the reduced model were consistent with such an increase.

Ten percent of US adults reported ever having been diagnosed with peptic ulcers, and half of them tested positive for immunoglobulin G H. pylori antibodies, while 88 percent of H. pylori-infected adults in our study population had no history of medically diagnosed ulcers. The association between H. pylori infection and iron deficiency/IDA seemed slightly stronger among adults who reported a history of medically diagnosed peptic ulcer disease, as shown in table 4, while serum ferritin levels and prevalences of iron deficiency or IDA among persons with a history of peptic ulcer disease but without antibodies to H. pylori were similar to those of persons who had neither H. pylori infection nor a history of peptic ulcer.

The relations between H. pylori infection and iron deficiency and IDA seemed stronger among non-Hispanic Whites and persons of other/mixed ethnicity and seemed weaker or nonexistent among other persons in other racial/ethnic groups (data not shown). The percent change in serum ferritin level associated with H. pylori infection status among non-Hispanic Whites and persons of other race/ethnicity was −21.5 percent, while among non-Hispanic Blacks, Mexican Americans, and other Hispanics, it was 2.9 percent (χ² test of product term: p = 0.007).

On the basis of our findings, if the association was causal, we estimate that 32.2 percent (95 percent CI: 15.6, 45.2) of cases of IDA and 13.6 percent (95 percent CI: 0, 26.3) of cases of iron deficiency in the United States might be related to H. pylori infection.

DISCUSSION

We found, in the most current cross-sectional data on the US population (1999–2000), that on average H. pylori-infected persons aged ≥3 years had a mean serum ferritin deficit of 9 μg/liter. On the basis of quantitative phlebotomy studies (17), this translates to a loss of 72–90 mg of storage iron. We found a weak-to-moderate increase in prevalences of iron deficiency and IDA among H. pylori-infected persons in the US population. We also found a weaker increase for anemia from other causes, suggesting that H. pylori produces anemia by altering iron metabolism. Overall, our data suggest that H. pylori may cause microbleeding and/or affect iron uptake and thus deplete iron stores in persons with iron deficiency and IDA, independently of ulcer disease. However, the associations between H. pylori infection and iron deficiency and IDA seem to be stronger among persons with a history of peptic ulcer, suggesting that ulcers may play a role in that small subpopulation. We did not observe differences in iron deficiency/IDA among uninfected persons with a history of peptic ulcer disease. The pathogenesis of iron deficiency is likely to be complex and possibly different in different age groups. For example, in children, iron loss seems likely, whereas in elderly persons with atrophic gastritis, low acid secretion and reduced gastric juice ascorbic acid could result in impairment of iron absorption.

One clinical trial of H. pylori eradication and/or iron supplementation found an effect on anemia in the H. pylori eradication treatment group that was comparable to the effect in persons who received only iron supplementation (18).

Our study controlled for several potential confounders of the iron deficiency-H. pylori association. We found no evidence of confounding by socioeconomic status using the poverty income ratio. However, there might still have been some degree of uncontrolled confounding by socioeconomic status, since this measurement of poverty is based on current income. Given that approximately one third of our sample comprised children under 18 years of age, data on occupation and education of the head of the household could have provided a better measurement of socioeconomic status but were not examined in this study. We were able to control for major predictors of anemia in the United States, such as age, race/ethnicity, socioeconomic status, current pregnancy, gestational history, and kidney disease, including those associated with major chronic diseases such as diabetes and hypertension. The most important confounder in these data was not socioeconomic status but race/ethnicity. Differences in the H. pylori prevalence odds ratios reported here deserve further investigation, since African Americans seem to have lower hemoglobin concentrations (19–21) after results are controlled for income. That could also be the case for Mexican Americans. This issue merits further research. Given the cross-sectional nature of this study, caution should be taken when interpreting the results, particularly since serology may misclassify H. pylori infection status, as it measures past status rather than current status. Data on the cytotoxin-associated gene A (cagA) type of H. pylori infection in the 1999–2000 NHANES was not available to the public when we conducted this study. However, at least one study found the association between IDA and H. pylori to be unrelated to cagA type (19). We did not explore the potential relations between infections with other agents and iron deficiency, which could possibly have confounded these results. These issues deserve further examination in NHANES.

As Dubois and Kearney (3) have noted, most epidemiologic studies conducted so far have been cross-sectional. Two of them, carried out in Denmark (22) and Germany (23), were national health surveys targeting the adult population. These surveys found decreased levels of serum ferritin among persons infected with H. pylori (22, 23). The German survey also found a 16 percent decrease in levels of serum ferritin among adults (23), while the Danish study found associations among men and postmenopausal women (22). Another large cross-sectional study conducted in Alaska (24) found that H. pylori infection was associated with decreasing levels of serum ferritin and that this association seemed stronger among adolescents and women of childbearing age. Children and adolescents were also studied in population-based surveys in South Korea (25–27); these studies found an association between H. pylori and iron deficiency. A study conducted in Australia found an association among women (28), and a cross-sectional study conducted in New Zealand reported differences in serum iron levels among men (29). A recent study of pregnant women in Germany reported an association between current H. pylori infection and hemoglobin levels (30). A recent abstract on a cross-sectional study based on NHANES III data also reported significant differences in serum ferritin and iron deficiency and IDA by H. pylori infection status that seemed to hold irrespective of cagA type (31), which agrees with the results of the German study (23).

Besides being the most recent large national survey, the 1999–2000 NHANES provided data on children, adults, and pregnant women and used validated indicators of the prevalences of iron deficiency and IDA. NHANES collects detailed information on study participants, which allowed us to consider several potential confounders. Our findings are consistent with those of previous studies that examined the association between H. pylori infection and serum ferritin level. It also adds information on the relations between infection and the prevalences of IDA and anemia from other causes.

IDA is a condition with important health consequences regarding reproduction, immunity, work performance, and possibly cognitive development (1). Thirty-five percent of the cases of iron deficiency and 51 percent of the cases of IDA in our study population involved H. pylori infection. Significant fractions of the prevalences of iron deficiency and IDA (14 percent and 32 percent, respectively) could be related to H. pylori infection. The trend of decreasing prevalences of iron deficiency and IDA from 1960 to 1980, and its subsequent leveling off (6, 7), does not lead to a straightforward interpretation of trends in H. pylori infection, since it was first measured in NHANES III. Moreover, the prevalence of iron deficiency in our study was based on fewer subjects than those of NHANES III, limiting the statistical power of comparisons of time trends for specific high-risk groups. It is possible that the dramatic decline in the prevalence of iron deficiency seen since the 1960s and 1970s was entirely due to public health interventions that focused on iron supplementation and fortification. Achieving further declines might require addressing other causes of iron deficiency, including infection, such as that from H. pylori. The overall evidence is consistent with a continuing decline in H. pylori infection, particularly in developed countries. However, the available data on the prevalence of H. pylori infection in the US population are consistent with a slow decline. This may reflect the significant increase in the number of foreign-born persons, particularly Mexican Americans, who immigrated to the United States in the 1990s and now constitute the main reservoir of H. pylori infection and related disease in the United States (32, 33).

Emerging evidence seems to place H. pylori infection next to helminthiasis as a communicable cause of anemia (34). Therefore, new initiatives designed to further decrease prevalences of iron deficiency and IDA in high-risk groups may need to address the eradication of H. pylori infection, a class A carcinogen that also causes peptic ulcer and gastritis.

No funding was specifically designated for this project. The University of Texas–Houston School of Public Health and Baylor College of Medicine provided salary support.

Conflict of interest: none declared.

References