Sero-prevalence of Helicobacter pylori CagA immunoglobulin G antibody, serum pepsinogens and haemoglobin levels in adults

Associations observed of Helicobacter pylori infection with haemoglobin levels are inconsistent. We examined associations of H. pylori sero-prevalence and serum pepsinogens (PGs), as non-invasive markers of atrophic gastritis, with haemoglobin levels. A cross-sectional study was undertaken among 654 Jewish and 937 Arab residents of Jerusalem, aged 25–78 years, randomly selected from Israel’s national population registry in age-sex and population strata. Sera were tested for H. pylori IgG, cytotoxin–associated gene A (CagA) antigen IgG antibody and PGs levels. Multivariable models were fitted to account for confounders. Participants with atrophic gastritis (PGI < 30 μg/L or a PGI: PGII < 3.0) had lower haemoglobin levels than those without: beta-coefficient −0.34 (95% CI −0.59, −0.09); in men −0.27 (95% CI −0.67, 0.12), and in women −0.43 (95% CI −0.74, −0.12). Lower haemoglobin levels were noted in persons with CagA antibody than in those H. pylori sero-negative or H. pylori-CagA sero-negative: beta-coefficient −0.14 (95% CI −0.29, 0.01). Anaemia was more common among women and men with than without atrophic gastritis: adjusted OR 2.58 (95% CI 1.48, 4.48) and 1.52 (95% CI 0.59, 3.95), respectively. In conclusion, independent of known correlates, atrophic gastritis and apparently CagA sero-positivity, a marker of H. pylori virulent strains, are associated with lower haemoglobin levels.

antigen was shown to be important in the pathogenesis of peptic disease and gastric cancer 2,5,16 . H. pylori chronic gastritis changes the physiology in the stomach 17 , including alterations in gastric acidity 17 and ascorbic acid levels [18][19][20] , which are important in the absorption of dietary iron 21,22 . Clearly, if associations of H. pylori infection with haemoglobin and other iron biomarkers is mediated by gastric inflammation, positive associations between H. pylori-related gastric pathology and anaemia are anticipated. However, the association of atrophic gastritis, a severe form of H. pylori-related gastric pathology, with haemoglobin, especially in the general population, was rarely addressed. H. pylori infection can cause atrophic body gastritis, which can lead to deficiencies in vitamin B12 and intrinsic factor, as well as hypochlorhydria; these negatively affect iron absorption 23,24 . Atrophic gastritis can be assessed using serum pepsinogen (PG) I and PGII, proenzymes of the digestive enzyme pepsin, which are secreted to the gastric lumen but which can also be detected in the serum [25][26][27] . With increasing severity of H. pylori gastritis, serum PGI and PGII levels are increased, but when atrophic changes occur in the corpus, PGI and the ratio of PGI: PGII decrease. More severe atrophy is related to a lower PGI: PGII ratio 25,28 .
The aims of the current study were to examine associations of H. pylori immunoglobulin G (IgG) sero-prevalence, CagA IgG sero-positivity and serum PGs, as non-invasive markers of atrophic gastritis, with two outcome variables: haemoglobin levels (continuous variable) and anaemia, in a population-based sample of adult men and women.
Laboratory methods. Sera  . Atrophic gastritis was defined as serum PGI levels of <30 μg/L or a PGI: PGII ratio of <3.0, as recommended by the manufacturer. Higher serum PGI and PGII levels are found in H. pylori infected vs uninfected individuals, while a lower PGI: PGII ratio is found in the former 28 . In our sample, H. pylori sero-status was significantly correlated with PGI level (Spearman's coefficient 0.19, P < 0.001), PGII level (Spearman's coefficient 0.33, P < 0.001) and PGI: PGII ratio (Spearman's coefficient −0.23, P < 0.001). These correlations strengthened the validity of the classification of H. pylori sero-status.
Statistical analysis. Student's t tests and one-way analysis of variance (ANOVA) were used to examine unadjusted differences in mean haemoglobin levels according to sociodemographic variables, H. pylori sero-status, atrophic gastritis and smoking. When more than two strata of a variable were compared, we used a Tukey test to account for multiple comparisons. Multivariable linear regression models with haemoglobin level as the dependent variable were fitted; from these models, we obtained beta (slope) coefficients (and 95% CIs). Categorical independent variables were included in the model as dummy variables. Chi square tests were employed to examine unadjusted associations of sociodemographic variables, H. pylori-CagA IgG antibody sero-status and atrophic gastritis, with anaemia. Multivariable logistic regression models were fitted, from which we obtained adjusted ORs and 95% CIs. Variables associated with the dependent variable (haemoglobin in linear regression and anaemia in logistic regression) in bivariate analysis with P < 0.2 were included in the multivariable analysis, in addition to age, atrophic gastritis (as measured by serum PGs) and H. pylori infection. As haemoglobin levels differ between men and women, the analyses were conducted in sex-specific strata. Interactions between population group, age, sex, H. pylori sero-status and atrophic gastritis were assessed. Data were analysed using IBM SPSS (Armonk, New York, USA) version 23 and Winpepi 33 .

Ethics statement. The study was approved by the Institutional Review Board of the Hadassah Medical
Centre, Jerusalem, and by the ethics committee at Tel Aviv University. All participants signed an informed consent. The study was conducted in accordance with the Declaration of Helsinki ethical principles and regulation of the Ministry of Health.

Haemoglobin levels by demographic variables, H. pylori IgG sero-positivity and atrophic gastritis in men.
The mean haemoglobin level decreased with age (P < 0.001). Differences between the age groups in mean haemoglobin level remained statistically significant after correction for multiple comparisons (by Tukey). A higher mean haemoglobin level was found in smokers than in non-smokers (P < 0.001). No significant difference was found in mean haemoglobin level according to education and according to H. pylori sero-status. Men with evidence of atrophic gastritis had a lower mean haemoglobin level than those without (P = 0.037) ( Table 1).
On multivariable analysis, men with evidence of atrophic gastritis had non-significantly lower haemoglobin levels: beta coefficient −0.27 (95% CI −0.67, 0.12), P = 0.17 (Table 2). No statistically significant difference was found in haemoglobin levels according to CagA IgG sero-status. The associations of age and smoking with haemoglobin level persisted in this model, which also showed lower haemoglobin levels in Arab than Jewish men ( Table 2). No interactions were detected between population group and atrophic gastritis (P = 0.7), population group and CagA IgG antibody sero-status (P = 0.9), and CagA IgG antibody sero-status and atrophic gastritis (P = 0.7).
Haemoglobin levels by demographic variables, H. pylori IgG sero-positivity and atrophic gastritis in women. Lower mean haemoglobin levels were found among Arab than Jewish women. A gradient was observed in relation to education (Table 1); namely, women with some high school education or less had significantly lower haemoglobin levels than did those with an academic degree (P = 0.011 by Tukey test). The average haemoglobin level was lower in women aged 25-44 years than in women aged 45-64 years (P = 0.004 by Tukey test), and than in women aged 65-78 years (P = 0.026 by Tukey test); and it was higher in smokers than in non-smokers (P = 0.03). Women with H. pylori CagA serum IgG antibody had a lower mean haemoglobin level than H. pylori sero-positive women who were lacking serum CagA IgG antibody (P = 0.009 by Tukey test), but the level was similar to that of H. pylori negative women. Women with serological evidence of atrophic gastritis had a lower mean haemoglobin level than women without (P = 0.052) ( Table 1). The differences in haemoglobin levels according to atrophic gastritis persisted in a multivariable analysis (P = 0.007); the association with CagA IgG sero-positivity did not (P = 0.15) ( Table 2). No significant interaction was found between population group and atrophic gastritis (P = 0.2), population group and CagA IgG antibody sero-status (P = 0.6), and CagA sero-status and atrophic gastritis (P = 0.10).
In a pooled multivariable analysis of both sexes, men had a higher mean haemoglobin level than women. Participants with atrophic gastritis had significantly lower haemoglobin levels than those without (P = 0.009). CagA IgG sero-positivity was related to a lower mean haemoglobin level (P = 0.069) ( Table 2). No significant interactions were found between sex and atrophic gastritis (P = 0.8) or by CagA IgG sero-positivity (P = 0.3).   Variables associated with anaemia. Among men, a non-significantly higher prevalence of anaemia was found among those with evidence of atrophic gastritis than those without (P = 0.13). Age was positively related to anaemia prevalence, while a lower prevalence of anaemia was found in smokers than non-smokers (P < 0.001).
No difference was noted in the prevalence of anaemia according to H. pylori IgG sero-prevalence ( Table 3). The results were similar in a multivariable logistic regression model that adjusted for age and H. pylori sero-positivity (Table 4). Among women, the prevalence of anaemia was higher among those who had atrophic gastritis than those without (P = 0.001) ( Table 3). Anaemia prevalence was higher among women with CagA IgG serum antibody than among those who were H. pylori sero-negative, and those who lacked CagA IgG antibodies. The prevalence of anaemia was higher among Arab than Jewish women. These associations persisted in a multivariable analysis that included age and H. pylori CagA sero-positivity in addition to population group, education and smoking (Table 4).
In a secondary analysis, we re-grouped the study participants according to values of the PGI: PGII ratio that correlated with gastritis severity, using the OLGA system 34 : most severe <3.0, moderate 3.0-6.8 and least severe >6.8. With these cut-off values, the mean haemoglobin levels were 14.6 (SD 1.6), 15.0 (SD 1.4) and 15.0 (SD 1.4), respectively, in men (P = 0.2 by ANOVA). In women, the corresponding values were 12.6 (SD 1.6), 12.7 (SD 1.3) and 13.0 (SD 1.2) (P = 0.029 by ANOVA). In a multivariable analysis, the significant gradient between PGI: PGII and haemoglobin level was maintained in women (Supplementary Table S1).

Anaemia sub-type according to H. pylori sero-status and atrophic gastritis.
Overall, the prevalence of microcytic anaemia was 5.2%, 3.1% and 4.5% in persons who were H. pylori sero-negative, H. pylori sero-positive but lacking CagA IgG antibody, and H. pylori sero-positive CagA positive, respectively. The respective prevalences for normocytic anaemia were 7.6%, 8.3% and 10.3%. The prevalence of microcytic anaemia was 9.8% in persons with atrophic gastritis vs 3.6% in those without this condition. The respective prevalence for normocytic anaemia was 8.1% vs 15.4%. All three persons with macrocytic anaemia were positive for H. pylori CagA IgG but without atrophic gastritis (Supplementary Table S2).

Discussion
We examined associations of H. pylori IgG sero-prevalence, CagA IgG sero-positivity and serum PGs, as non-invasive markers of atrophic gastritis, with haemoglobin levels and anaemia, in men and women of two ethnic groups in a general population.
Serologic evidence of atrophic gastritis was associated with a higher prevalence of anaemia, particularly in women, and with lower mean haemoglobin levels (mean difference 0.34 g/dL). A similar trend, although of smaller magnitude, was found in relation to CagA IgG antibody sero-positivity (difference of 0.14 g/dL in mean haemoglobin level). These differences were more evident in women, although no significant interaction was found between sex and atrophic gastritis or between sex and CagA IgG sero-positivity. Notably, these observations were independent of age, population group, education and smoking history. A limited number of studies have addressed associations of H. pylori infection with anaemia or haemoglobin level among adults in the general population [35][36][37][38][39][40] . These mostly showed no significant difference between infected and uninfected individuals in mean haemoglobin levels or in the prevalence of anaemia, except for studies carried out in pregnant women 39,40 . None of these studies has addressed the role of CagA infection or atrophic gastritis. Our findings add a new dimension, suggesting that severe gastric inflammation, even with atrophic gastritis (as evident by serum PGs levels), rather than exposure to H. pylori per se, are involved in decreased haemoglobin levels. Thus, our results improve the risk profiling of low haemoglobin levels in relation to H. pylori infection. In dyspeptic adult patients, lower mean haemoglobin levels and a higher prevalence of anaemia were documented in H. pylori infected  compared to uninfected patients 41 . A case-control study from the United Kingdom 42 showed that adult patients referred for investigation of IDA had significantly more frequent gastric body atrophy, as demonstrated by gastric biopsy, compared to control patients with normal haemoglobin and iron levels. A gradient was observed with increased atrophy grades, whereas H. pylori infection as a main effect was not associated with IDA 42 . Anaemic patients with gastric body atrophy were less likely to have conditions that might be the definite cause of anaemia than were anaemic patients without atrophy. This suggests that gastric body atrophy might be a cause of anaemia in some individuals and a contributory factor in others 42 . Nahon el al. 43 showed that patients with unexplained IDA referred for gastric tract evaluation had a higher prevalence of chronic gastritis than control patients (67% vs 47%), and than patients with atrophic gastritis (15% vs 6%) and H. pylori infection (60% vs 43%) 43 . Similar to our observations, Lee and colleagues 44 showed in a general population sample of 2398 participants, lower haemoglobin levels and higher anaemia prevalence among persons with serological evidence of gastric atrophy (PGI: PGII ratio < 3.0, either with or without H. pylori serum antibodies) compared to those without gastric atrophy 44 .
Collectively, these and our findings confirm the hypothesis that the association of H. pylori infection with lower haemoglobin might be mediated by H. pylori-associated inflammatory and atrophic changes affecting the gastric mucosa. Anaemia is an alarm symptom of gastric cancer that prompts invasive diagnostic techniques. Atrophic gastritis is mostly caused by H. pylori infection, and is a main precursor of gastric cancer 45 . A positive association between anaemia and serologic evidence of atrophic gastritis suggests that anaemia might be a marker of pre-malignant lesions in the stomach. This observation may promote early detection of gastric cancer. Atrophic lesions in the stomach affect gastric acidity 17 and ascorbic acid levels [18][19][20] , which are essential to the absorption of dietary iron 21,22 . Body gastric atrophy is also associated with decreases in secretion of intrinsic factor and in absorption of dietary vitamin B12, which may result in pernicious anaemia 23,24 . CagA is associated with an inflammatory response in the stomach, but also has the potential to induce carcinogenesis irrespective of inflammation (reviewed in 31 ). We found a clear trend of higher prevalence of microcytic anaemia in persons with serological evidence of atrophic gastritis compared to persons without such condition, and all three persons with macrocytic anaemia tested positive for CagA IgG serum antibody.
Consistent with previous reports [46][47][48] , women aged 25-44 years had lower haemoglobin levels than did older women, thus conferring a risk for anaemia among women of child-bearing age; this is likely due to menstrual blood loss and enhanced demands related to pregnancy. In men, however, haemoglobin levels decreased with advanced age. A longitudinal study from Japan demonstrated that haemoglobin levels declined as age increased in men, while in women a decline was observed after the age of 60 years 49 . Age-related patterns in haemoglobin levels likely reflect the ageing process, but might also be related to a birth cohort effect, or both, as has been shown 49 .  Our results also confirmed previous observations of higher haemoglobin levels among smokers than non-smokers 46,47,49,50 . This finding apparently results from compensatory erythropoiesis stimulated by smoking-induced hypoxia.

Men
Lower haemoglobin levels were found among Arab than Jewish patients, and Arabs tended to be more anaemic. These differences might reflect differences in socioeconomic status and dietary habits. Genetic differences may also contribute to the Arab-Jewish gap in anaemia prevalence. Consanguineous marriages are far more common in the Arab population in Israel 51 , as are haemoglobinopathies 52 .
Our study has limitations. The cross-sectional design limits causal inference regarding associations of H. pylori infection and atrophic gastritis with anaemia. Nonetheless, H. pylori infection is typically acquired in early childhood and, without treatment, persists for life. Hence, the H. pylori infection likely preceded the occurrence of anaemia, which presumably developed in adulthood. H. pylori is a main cause of atrophic gastritis, and when the latter ensues, the bacterium usually loses its niche and disappears. Hence, it is likely that some persons who were classified as H. pylori sero-negatives were actually previously infected with the bacterium. Such a scenario would likely result in underestimation of associations of H. pylori infection with haemoglobin levels and anaemia. We cannot determine whether atrophic gastritis resulted from H. pylori infection or autoimmunity. Addressing such a question is especially challenging given the overlap between these two conditions 53 and the evidence showing that H. pylori might play a role in gastric autoimmunity via molecular mimicry 23,54 .
Our use of serum PGs as a surrogate marker to define atrophic gastritis might have limited sensitivity. However, we examined in the serum, both PGI concentration and PGI: PGII ratio; a combination of these parameters has been shown to improve the accuracy of detection. A PGI level of 25-30 µg/L or less and PGI: PGII < 3.0 are commonly used cut-off values 55,56 (reviewed by Zagari et al. 26 ) when using Biohit ELISAs, with sensitivity ranging between 71% and 90%, and high specificity of 90-98% compared to gastric biopsy [56][57][58] ; this usually detects moderate to severe forms of atrophic gastritis.
Our study has a number of strengths. First, it comprises a large sample size of men and women from two general population samples. Second, persons with conditions that might induce anaemia, such as cancer, severe kidney disease, pregnancy and recent birth were excluded from the study. Third, we were able to adjust for potential confounders, which were not available in many of the previous studies that assessed the association between H. pylori infection and anaemia.
In summary, over and above known correlates of haemoglobin levels and anaemia, we found that serological evidence of atrophic gastritis was associated with lower mean haemoglobin levels, mainly in women, and a similar  trend, although of smaller magnitude, was found in relation to CagA IgG antibody sero-positivity. Our results provide new insight regarding populations at risk for low haemoglobin levels in relation to H. pylori infection and its related gastritis, as measured non-invasively by serum pepsinogen levels.

Data Availability
Data can be provided upon request to the corresponding author (KM).