Chapter 2. Materials and Methods
2.4 Laboratory assay
A fasting venous blood sample was collected from each participant in the morning after an 8-hour fasting period.
Serum H. pylori IgG level
H. pylori infection was assessed by a serology test of IgG using a commercial enzyme
immunoassay (EIA) (IMMULITE 2000, Siemens, Germany) with a reported sensitivity and specificity of > 90%. All assays were performed and interpreted by a single laboratory investigator who was unaware of the sample status. The antibody titer ranged from 0.4 to 8.0 U/mL with a clinical cutoff point of 1.1 U/mL. In addition, the titer of H.
pylori serology does not provide a method for predicting the presence of ulcer in patients with H. pylori infection, but may imply the severity of histological changes [60].
Other serum biomarkers
The plasma levels of CRP and homocysteine were determined by TBA-200FR (Toshiba Medical Systems, Japan); vitamin B12 and folate levels were determined by an ABBOTT Architect i2000 analyzer (Abbott Park, Illinois, U.S.A.). Serum creatinine, high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C), triglyceride (TG) and fasting glucose were included in the package of the elderly health checkup program. The presence of IL-6 was tested by enzyme linked immunosorbent assay (ELISA) and was determined by a Tecan Infinite M200 multimode reader (Tecan, Männedorf, Switzerland).
Genotyping assay
Genomic DNA was extracted from buffy coat using a QuickGene-Mini 80 system
doi:10.6342/NTU20190252
16
(Fujifilm, Tokyo, Japan) and then stored in a -80°C freezer. The APOE ε4 status was determined by two single nucleotide polymorphisms (SNPs) s42938 and rs7412 [61]
using TagMan Genomic Assays by the ABI7900HT fast real-time PCR system (Applied Biosystems Inc., California, U.S.A.). Quality control samples were obtained from 5% of the internal samples provided by the manufacturers and then duplicated and genotyped together with all of the other samples, and the concordance rate was 100%. The Hardy-Weinberg equilibrium (HWE) test was performed among controls to examine possible genotyping error and selection bias for each SNP.
doi:10.6342/NTU20190252
17
2.5 Statistical analyses
Participants were quartiled according to their serum H. pylori IgG level. For descriptive analyses, analysis of variance (continuous variables) and chi-square tests (categorical variables) were used to compare different exposure variables across the quartiles of serum H. pylori IgG level. Multivariable linear regression models were used to examine the
change of global and domain-specific cognition scores per one unit increase of serum IgG level (i.e., the regression coefficient β). Multivariable logistic regression models were performed to estimate the odds ratio (OR) and 95% confidence interval (CI) for the association between serum H. pylori IgG level and cognitive impairment (i.e., MoCA-T score < 24 for global cognition; the lowest tertile (T1) of Z-score for specific cognitive domain test; or the lowest tertile (T1) of factor score of each cognitive factor). The covariates that were adjusted in the model included factors with biological importance for cognition (e.g., age, sex, years of education, and APOE ε4 status) and variables that were selected by the stepwise model selection (SLENTRY=0.15, SLSTAY=0.15). Stratified analyses were performed by important covariates [age groups (65-74 vs. ≥75), sex, and APOE ε4 status] for the association between the quartiles of serum H. pylori IgG and
cognitive impairment. A trend test (Ptrend) was performed for testing the relationship between the four quartiles of serum H. pylori IgG level and cognitive tests. We also evaluated how those covariates modified the association between H. pylori IgG level and the risk of cognitive impairment by comparing the model with main effect terms and interaction terms to the model with main effect terms only (Pinteraction < 0.10 indicated the existence of an interaction). A risk estimate (i.e., OR) reaches statistical significance when its corresponding 95% CI does not include 1. Statistical analyses were performed using SAS software version 9.3 (SAS Institute Inc., Cary, NC, U.S.A.) and all of the tests were
doi:10.6342/NTU20190252
18
two-sided.
Statistical power
EpiInfoTM 7 (Centers for Disease Control and Prevention, Atlanta, U.S.A.) was used to estimate the statistical power. According to previous studies, given the ratio of uninfected participants to infected participants as 1.2 and the effect size was 0.10 and 0.18 for the ratio of global cognitive impairment in uninfected and infected participants respectively, a sample size of 583 gave a statistical power of 80 % at a significance level of 0.05. This study included 587 participants, which provided sufficient statistical power to assess the association between H. pylori infection and cognitive impairment.
doi:10.6342/NTU20190252
19
2.6 Mediation and moderation analyses
To further understand the relationship between H. pylori and cognitive impairment, we performed mediation and moderation analyses to test if folic acid, vitamin B12, homocysteine, CRP or IL-6 play a role in the aforementioned association.
Mediation analyses
Mediation analyses were firstly described by Baron & Kenny in 1986 to explore the mechanism through which the causal variable affects the outcome. The mediator has been called an intervening or process variable [62]. The mediation model was shown in Figure 2-2 and included the following four steps to establish mediation:
• Step 1: to check if the exposure is associated with the outcome in a regression equation (path c)
• Step 2: to check the exposure is associated with the mediator in a regression equation (path a)
• Step 3: to check if the mediator affects the outcome variable in a regression equation (path b), exposure should be controlled in establishing the effect of the mediator on the outcome.
• Step 4: to check if the mediator completely mediates the exposure-outcome relationship, i.e., all four steps should be met. Partial mediation indicates the essential steps in establishing mediation, i.e., Steps 2 and 3. In addition, mediators should be either dichotomous or continuous variables. If path c’ is equal to 0, which indicates no contribution of the exposure to the outcomes.
doi:10.6342/NTU20190252
20
Moderation analyses
The framework and variables in the moderation analyses are summarized in Figure 2-3. The moderation model included three causal paths that feed into outcome variable of cognitive function [62]. The impact of H. pylori infection was treated as the exposure (path a), the impact of folate, vitamin B12, homocysteine, CRP or IL-6 were treated as the moderator (path b), and the product of the exposure and the moderator indicated the interaction between them. The interaction (path c) exists if the interaction term reach statistical significance.
Each of these 5 variables was dichotomized into lower- and higher-level group.
Higher-level group was defined as the highest tertile (T3), and lower-level group was defined as the combination of the1st and 2nd tertile (T1+T2).
doi:10.6342/NTU20190252
21
Chapter 3. Results
3.1 Characteristics of the study population
Among the 587 study participants, 104 of them (17.7%) showed cognitive impairment (MoCA-T score < 24). Table 3-1 summarizes the demographic characteristics of the study population according to the H. pylori IgG quartiles. Participants showed borderline significant differences for depressive symptoms (P=0.05) and HDL level (mg/dl, P=0.05) across the H. pylori IgG quartile. Moreover, 262 (44.6%) participants were seropositive, defined as having serum H. pylori IgG ≥ 1.1 U/mL for H. pylori infection, which was similar to the findings of previous studies[19, 20]. For global cognition, H. pylori seropositivity was 47.1% and 44.1% in cognitive impaired and normal elders, respectively (P=0.57).
doi:10.6342/NTU20190252
22
3.2 Identification of cognitive factors
Four cognitive factors were identified by EFA (Table 3-2). The first factor is the
“memory” factor, which includes 4 domain-specific cognitive variables, i.e., logical memory-thematic I & II and recall I & II. The second factor is the “verbal fluency” factor, which includes 3 verbal fluency tests, i.e., tests for naming fruits, fish and vegetables. The third factor is the “executive” factor, which includes the trail making tests- A & B. The fourth factor is the “attention” factor, which includes digit spans-forward & backward.
The variances of the four cognitive factors accounted for 78.4% of the total variance in cognitive tests.
doi:10.6342/NTU20190252
23
3.3 Association between H. pylori IgG quartiles and cognitive function
Significant associations were found between serum H. pylori IgG levels in quartiles and some cognitive variables (Table 3-3). For the continuous cognitive outcome, multivariable linear regression models showed that, compared with the lowest quartile of H. pylori IgG (Q1), the highest quartile (Q4) was associated with poor performance of verbal fluency (vegetables: β=-0.24; 95% CI = -0.45, -0.03; Ptrend =0.03), domain-specific attention (digit span-forward: β=-0.19; 95% CI = -0.42, -0.04; Ptrend =0.17), and attention factor (β=-0.20; 95% CI = -0.37, -0.02; Ptrend =0.07).
For dichotomous cognitive impairment, multivariable logistic regression models showed that, compared with the lowest H. pylori IgG quartile (Q1), the highest H.
pylori IgG quartile (Q4) was associated with impairment on both domain-specific
attention (digit span-forward test: OR= 1.83, 95% CI = 1.03-3.24; Ptrend = 0.14, Table 3-3) and the attention factor (OR= 2.67, 95% CI = 1.51-4.73; Ptrend = 0.001). No
significant associations were observed between H. pylori IgG quartiles and global or other domain-specific cognitive function.
doi:10.6342/NTU20190252
24
3.4 Stratified analyses by selected covariates
For significant associations identified between serum H. pylori IgG and cognitive impairment (Table 3-3), stratified analyses were performed by age group (<75 and ≥ 75 years old), sex and APOE ε4 status. No significant interaction (Pinteraction <0.10) was observed between serum H. pylori IgG and age, sex or APOE ε4 status on cognitive function. After stratifying by these variables, significant associations were observed in some subgroups (Figure 3-1). Compared with the lowest quartile of serum H. pylori IgG (Q1), the highest quartile of serum H. pylori IgG (Q4) was associated with impaired verbal fluency-vegetables in men (β=-0.38, 95% CI=-0.66,-0.09; OR=3.01, 95% CI=1.42-6.38), in APOE ε4 non-carriers (β=-0.25, 95% CI=-0.48, -0.09) and in elderly people < 75 years of age (β=-0.26, 95% CI=-0.51, -0.02). In addition, the 2nd quartile (Q2) and the 3rd quartile (Q3) of serum H. pylori IgG was associated with impaired verbal fluency-vegetables in men (OR=2.34, 95% CI=1.12-4.90 and OR=2.36, 95% CI=1.13-4.91).
Compared with the lowest quartile of serum H. pylori IgG (Q1), the highest quartile of serum H. pylori IgG (Q4) was associated with impaired attention function (digit span-forward) in elderly people < 75 years of age (β=-0.23, 95% CI=-0.44, -0.02), in men (β=-0.03, 95% CI= -0.59, -0.01) and in APOE ε4 non-carriers (β=-0.27, 95%
CI=-0.49, -0.05; OR=1.83, 95% CI=1.00-3.34, Figure 3-1).
Moreover, compared with the lowest quartile of serum H. pylori IgG (Q1), the highest quartile of serum H. pylori IgG (Q4) was associated with impaired attention factor (digit span-forward & backward) in elderly people ≥ 75 years of age (β=-0.31, 95% CI=-0.61, -0.02; OR=4.55, 95% CI=1.74-11.9), in men (β=-0.39, 95% CI=-0.61, -0.08; OR=2.80, 95% CI=1.26-6.22) and in APOE ε4 non-carriers (β=-0.24, 95% CI=
-0.43, -0.05; OR=2.41, 95% CI=1.35-4.27, Figure 3-1).
doi:10.6342/NTU20190252
25
3.5 Mediation analyses of factors involved in one carbon metabolism and inflammatory markers for the association between H. pylori infection and cognitive impairment
For factors involved in one carbon metabolism (i.e., folate ,vitamin B12,
homocysteine, CRP or IL-6), mediation analysis was performed for the association between H. pylori infection and cognitive impairment. First, before exploring the association between H. pylori infection and each potential mediator (folate ,vitamin B12, homocysteine, CRP or IL-6), log transformation was performed to transform the right skewness of these variables into normality. Second, multivariable linear regression was performed (Table 3-4) to compare higher quartiles (Q4, Q3 or Q2) with the lowest quartile of H. pylori IgG (Q1). We found that higher H. pylori IgG level was associated with higher level of log(vitamin B12) ( H. pylori IgG Q4 vs. Q1: β=0.14; 95% CI =0.03, 0.24; Q2 vs. Q1: β=0.12; 95% CI =0.16, 0.23) and log(CRP) (β= 0.26; 95% CI =0.02, 0.50). Because the direction of the association between quartile of H. pylori IgG and log(vitamin B12) was not compatible with our hypothesis; further analysis was not performed. Similarly, we performed multiple linear regression (Table 3-5) and logistic regression (Table 3-6) for testing the association of serum H. pylori IgG level, CRP level with cognitive function. Compared with the lowest quartile of serum H. pylori IgG (Q1), the highest quartile of serum H. pylori IgG (Q4) was associated with poor
performance on attention domain (digit span-forward: β= -0.22; CI =-0.42, -0.03).
However, log(CRP) was not significant (β= -0.01; 95% CI =-0.08, 0.06) in this model.
Further mediation analysis was terminated due to non-significant finding in path b (i.e., association between mediator and outcome in Figure 2.2).
doi:10.6342/NTU20190252
26
3.6 Moderation analysis of factors involved in one carbon metabolism and inflammatory markers for the association between H. pylori infection and cognitive impairment
For each significant association identified between serum H. pylori IgG and cognitive impairment (Table 3-3), stratified analyses were performed by high and low serum level of folate, vitamin B12, homocystenine (factors involved in one carbon metabolism), CRP and IL6 (inflammatory marker).
No significant interaction (Pinteraction <0.10) was observed between the quartiles of serum H. pylori IgG and folate, vitamin B12, homocysteine, CRP or IL-6 on different cognitive function (Table 3-7, Table 3-8 and Table 3-9). After stratifying by these factors, significant associations were observed in some subgroups. Compared with the lowest quartile of serum H. pylori IgG (Q1), the highest quartile of serum H. pylori IgG (Q4) was associated with impaired digit span-forward in low folate group (β=-0.29, 95% CI= -0.52, -0.06) and high CRP group (β= -0.51, 95% CI= -0.86, -0.17, Table 3-8).
A significant association was found for the attention factor (digit span-forward &
backward) in high vitamin B12 group (β=-0.38, 95% CI= -0.68, -0.09; OR=4.47, 95%
CI=1.71-11.7), low homocysteine group (β=-0.32, 95% CI= -0.54, -0.12; OR=3.33, 95% CI=1.72-6.48) and low IL-6 group (β= -0.28, 95% CI= -0.50, -0.05; OR=2.54, 95% CI=1.30-4.95, Table 3-9).
doi:10.6342/NTU20190252
27
Chapter 4. Discussion
4.1 Main finding
This study found a significant association between H. pylori infection (assessed by IgG level) and cognitive impairment, especially for verbal fluency-vegetables, attention domain (digit span-forward) and the attention factor (digit span-forward & backward) in a community-dwelling elderly population. Stratified analyses showed that a significant association remained for the attention factor in the old-old group (age ≥ 75) and for attention domain (digit span-forward) in men, which has not been reported previously.
In this study, H. pylori seropositivity, defined as IgG ≥ 1.1 U/mL (i.e., clinical cutoff point), was not associated with global cognitive impairment as in previous studies [63, 64]. It is possible that the cognitive performance was relatively good in this healthy elder population. The clinical cut-off for H. pylori IgG level may not be suitable for this population. Therefore, this study quartiled the IgG level to better demonstrate its relationship with cognitive function.
Our result revealed that the attention function of the elderly was mostly affected by H. pylori infection. This may be due to the characteristic of our study population, i.e.,
our study participants were relatively healthy and highly educated. Attention function is the crucial component of the organization, management and completion of complex cognitive activities [65]. Impaired attentional function is especially damaging to complicated daily activities, such as shopping for different item in different stores [66]
or keeping track of conversation with more than one person [67]. In recent years, there has been increasing evidence for impaired attention function early in the MCI and during AD disease progress [68, 69]. In addition, the performance of attention task can be a surrogate of AD at preclinical stage and the impairment increased gradually during the preclinical phase of AD [65]. This may explain why in our study the attention
doi:10.6342/NTU20190252
28
function of our study participants was affected most significantly while the function of most other cognitive domains were preserved.
Theoretically, H pylori infection cause atrophic gastritis and then decreased the absorption of vitamin B12. However, we observed that the serum vitamin B 12 level was higher in participants with the 2nd and 4th quartile of H. pylori IgG level compared with the participants with 1st quartile. Our patients are relatively healthier; the vitamin B12 level was 706 pg/ml for Q1, 783 pg/ml for Q2, 749 pg/ml for Q3 and 798 pg/ml for Q4 of H. pylori IgG level respectively (normal range: 200-900 pg/ml). The possible mechanism may be due to the higher vitamin B12 supplement in in participants with the 2nd and 4th quartile of H. pylori IgG level. The ratios of participants with high vitamin B12 supplement (≥ one pill/day) were 20.5%, 32.9%, 29.0% and 35.6%, respectively, for 1st to 4th quartile (p < 0.05). The supplement intake of vitamin B12 may interfere the result of our study.
doi:10.6342/NTU20190252
29
4.2. Sex difference for the association between H. pylori infection on cognitive function
We found significant associations between elevated H. pylori IgG level and attention impairment in men, which has not been reported previously. It is possible that the intrinsic sex differences, i.e., the neuroprotective effects of estrogen, modulated the H. pylori-induced neuroinflammation. For example, estrogen promotes the formation of dendritic spines and synapses [70], modulates the neurotransmitter system [71] and the inflammatory response [72], reduces accumulation of Aβ [73], and acts as an antioxidant [74]. Most H. pylori infections are acquired in early childhood [75]; therefore, the neuronal damage is chronic and persistent. This may explain why H. pylori infection has a more significant impact on cognitive impairment in men than in women, who are protected by estrogen until menopause. From the finding of our study, we postulated a mechanism that sex hormone may play a protective role on the association between H, pylori infection and cognitive impairment as shown in Figure 4-1. However, investigations into the interactions between H. pylori infection and estrogen on cognitive performance are limited and further research is necessary.
doi:10.6342/NTU20190252
30
4.3. Comparison with previous studies
Most of the previous studies mainly investigated the global cognitive function, our study not only assessed the global cognition but also assessed the domain specific cognition. Beydoun et. al. demonstrated poorer verbal memory among those H. pylori seropositive people in the U.S. who aged 60-90 years old [76]. However, our result did not show an association between H. pylori IgG level and impaired logical-memory function, which is the typical manifestation of MCI or AD. Our finding showed that H.
pylori infection mainly influenced the attention and executive function, which is the
manifestation of subcortical type dementia (e.g., vascular dementia, dementia with Lewy bodies and Parkinson's disease-dementia) [77-79].
Xu et. al. studied 173 vascular dementia patients, they found that H. pylori infected patients had significant lower MMSE and MoCA scores than non-infected patients. In addition, significantly higher inflammatory cytokine (IL-1β, IL-6 and TNF-α) was found in H. pylori infected vascular dementia patients, which suggests inflammatory response may be involved in the pathophysiology of dementia [80]. In addition, one large community-based study showed that those who had H. pylori infection had a significantly increased risk of stroke, independent of potential confounders and known risk factors for stroke compared with those without these conditions. H. pylori infection is likely to affect gastric physiological status, including alterations in metabolic
hormones (e.g., ghrelin and leptin) and micronutrient absorption, which leads to
increased homocysteine concentration and causes vascular damage [81]. Nonetheless, a Korean study did not found the relationship between H. pylori infection and cerebral small-vessel disease (i.e., silent brain infarction and cerebral microbleeds). On the other hand, the association between H. pylori infection and Parkinson's diseases was reported in case-control, cross-sectional, cohort study and meta-analysis [82, 83]. The postulated
doi:10.6342/NTU20190252
31
mechanism for this association was that endotoxin of bacteria stimulates microglia to release inflammatory markers and then leads to neuroinflammation [83].
The results of our study revealed that attention function and executive function was impaired mostly by chronic H. pylori infection compared to other cognitive domains, which is similar to the presentation of subcortical vascular dementia. To clarify the character of cognitive impairment of our study population, we performed additional analysis to evaluated the relationship of small-vessel diseases and cognitive function. In our study, 390 out of 587 participants received brain Magnetic Resonance Imaging (MRI) and 324 participants showed lacunar infarct or white matter hyperintensities.
Multivariable linear regression and multivariable logistic regression were used to evaluate the association between quartiles of H. pylori IgG and global cognition or domain-specific cognitive tests among those 324 participants with small-vessel diseases. The results were summarized as Table 4-1. In comparison with the lowest quartile of H. pylori IgG (Q1), the highest quartile (Q4) was associated with poor performance of attention (digit span-forward: β=-0.32, 95% CI = -0.59, -0.05; OR
=3.23, 95% CI = 1.45, 7.21, digit span-backward: β=-0.31, 95% CI = -0.63, -0.01; OR
=2.61, 95% CI = 1.27, 5.35), and attention factors (β=-0.38, 95% CI = -0.62, -0.13;
OR= 3.01, 95% CI = 1.44, 6.30). No significant association was observed for global cognition. The above results verify the cognitive impairment identified in our study population are more related to subcortical vascular dementia.
However, because this study aims to assess the cognition in the preclinical phase of dementia, the specific subtype of dementia for our study population could not be precisely ascertained.
doi:10.6342/NTU20190252
32
4.4. Strengths
Our study has several strengths. First, we utilized an EFA to identify cognitive factors (the combination of domain-specific cognitive variables). Second, multivariable linear regression and multivariable logistic regression were used to adjust for important potential confounders (e.g., age, sex, years of education, and APOE ε4 status), which were not fully adjusted for in prior studies. Third, our sample size was relatively large, which allowed us to perform a stratified analysis by important covariates (sex, age, and APOE ε4 status) and to assess their interactions with H. pylori infection on the risk of cognitive
impairment. Moreover, serology tests reduce false negative results in people with gastric atrophy or under concurrent proton pump inhibitor use as they tend to have negative results from the C13 urea breath tests and gastric biopsy (rapid urease test & histology).
In addition, compared with histological analysis, a serology test is relatively non-invasive, which makes it a better approach for population-based studies [76].
doi:10.6342/NTU20190252
33
4.5. Limitations
There were some limitations in our study. First, because this study used a cross-sectional design, causal inferences between H. pylori infection and cognitive impairment could not be ascertained. However, because the acquisition of most H. pylori infections usually occurs in the early childhood [75], it is highly possible that chronic infection with this microorganism will lead to cognitive impairment in late life. Second, a serology test of H. pylori infection was unable to differentiate between current and past infection as the antibody remained detectable after the eradication of H. pylori [84]. A previous study revealed that serology titers of H. pylori decreased by 67% about 18 months after H.
pylori eradication [85]. Therefore, quartiled the H. pylori IgG titer may allow us to
identify the subpopulation with current infection, i.e., the highest IgG quartile. This may also explain why the highest quartile was associated with cognitive impairment. Finally, our participants were recruited from the EHC program and are community dwelling
identify the subpopulation with current infection, i.e., the highest IgG quartile. This may also explain why the highest quartile was associated with cognitive impairment. Finally, our participants were recruited from the EHC program and are community dwelling