Introduction

Vascular inflammation and neuroinflammation have been implicated in increased risks of cognitive impairments, particularly Alzheimer’s disease (AD)1,2,3. Reports have been accumulating showing that systemic inflammation is directly associated with the development of AD2,4,5. Further, vascular risk factors such as hypertension, diabetes mellitus, and chronic kidney disease have been shown to be associated with AD6,7,8, and vascular inflammation is assumed to be behind these vascular risk factors9.

Serum high-sensitivity C reactive protein (hs-CRP) is a conveniently measured biomarker of systemic inflammation10, but there is no consensus on the association between serum hs-CRP and dementia including AD. Although serum hs-CRP has been reported to be significantly lower in AD patients than in healthy subjects11, a recent meta-analysis found that serum hs-CRP was significantly higher in AD patients than in healthy older subjects12. In another meta-analysis, serum hs-CRP was associated with all-cause dementia, but not with AD13. The relationship between serum hs-CRP and dementia has also been investigated in middle and old age groups, and many reports suggest that high serum hs-CRP levels in middle age increase the risk of AD14. However, reports on the relationship between serum hs-CRP and AD in older subjects were controversial15,16.

Regarding the association between serum hs-CRP levels and brain atrophy, there are reports that elevated serum hs-CRP levels were associated with atrophy of the frontal and temporal corticies17, and that serum hs-CRP was not significantly associated with brain atrophy after multivariable adjustment18. Thus, the results are inconsistent.

The Japan Prospective Studies Collaboration for Aging and Dementia (JPSC-AD) was a prospective cohort study of dementia. A total of 11,957 community-dwelling older residents were surveyed in eight sites from 2016 to 2018. The aim of this study was to examine the relationship between serum hs-CRP levels and dementia, especially AD, and their association with brain atrophy using cross-sectional data from this group of older Japanese persons.

Methods

Study population

The JPSC-AD Study was a multisite, population-based, prospective, cohort study of dementia involving eight sites in Japan with a pre-specified protocol and standardized measurement methods across the research sites19,20,21,22. The study was conducted in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) checklist. The Medical Ethics Review Board of Ehime University (approval number 1610004) and Kyushu University Institutional Review Board for clinical research (approval number 686-10) approved this study. Written, informed consent was obtained from all participants. This study was performed in accordance with the Declaration of Helsinki. A total of 11,957 people consented to participate in the study between 2016 and 2018, of which 10,085 were eligible for the present study, excluding 547 participants younger than 65 years and 1325 participants lacking serum hs-CRP levels (Fig. 1A).

Figure 1
figure 1

Flowchart of participants selection for the analysis of the association of serum high-sensitivity C-reactive protein with the presence of dementia (A) and brain volume estimated by brain magnetic resonance imaging (B). Abbreviations: MRI, magnetic resonance imaging; eTIV, estimated total intracranial volume.

Full size image

For the analysis of the association between serum hs-CRP levels and brain volume estimated by brain magnetic resonance imaging (MRI) scan, a total of 8614 participants were included after excluding 1867 participants without available brain MRI data, 17 participants for whom FreeSurfer analysis was errored, 29 participants who had outlier for estimated total intracranial volume (eTIV), 122 participants who had more than 5 outliers for the volume by part, 443 participants aged less than 65 years, 290 participants lacking serum hs-CRP levels, and 575 with missing some covariates data (Fig. 1B).

Measurement of serum hs-CRP

Serum hs-CRP was measured at the central laboratory (LSI Medience Corporation, Tokyo) for all participants using the latex agglutination turbidimetry method. The detailed measurement methods of surum hs-CRP are shown in Supplemental Table 1. Basically, the samples were analyzed once and when abnormal values were found, reassay was carried out. Serum hs-CRP values were categorized into four categories (< 1.0 mg/L, 1.0–1.9 mg/L, 2.0–2.9 mg/L, ≥ 3.0 mg/L) according to clinical cut-off values.

Table 1 Clinical characteristics of the study participants according to serum high-sensitivity C-reactive protein level (n = 10,085).
Full size table

Measurement of other risk factors

Each participant completed a self-administered questionnaire that included education, medical history, alcohol consumption, smoking, medical treatment (antihypertensive medications, diabetes medications), and physical activity. Blood pressure was measured three times with a sphygmomanometer at intervals of at least 5 min, and the average of the three measurements was used for the analysis23. Hypertension was defined by blood pressure greater than 140/90 mmHg or current use of antihypertensive medications. Ischemic heart disease was defined as history of myocardial infarction and/or coronary intervention. Diabetes mellitus was defined by a fasting glucose level ≥ 7.0 mmol/L, glucose level at any time ≥ 11.1 mmol/L, or HbA1c ≥ 6.5%, and/or current use of antidiabetic medications according to the 2010 American Diabetes Association (ADA) criteria24. Serum total cholesterol levels were measured by an enzymatic method. The body mass index (kg/m2) was calculated as an index of obesity. Chronic kidney disease was defined as estimated glomerular filtration rate < 60 ml/min/1.73 m225 and calculated by multiplying the modified Chronic Kidney Disease Epidemiology Collaboration formula by the Japanese coefficient (0.813)26,27. Electrocardiographic abnormalities were defined as ST depression (Minnesota code, 4-1, 2, 3), left ventricular hypertrophy (3-1), and atrial fibrillation (8-3). Regular exercise was defined as physical activity performed for at least 30 min twice a week over the last year. Alcohol consumption and smoking were categorized as ongoing or not. Apolipoprotein E (APOE) ε4 carrier status was defined as the presence of the ε2/ε4, ε3/ε4, and ε4/ε4 alleles. To determine the APOE polymorphism, two single-nucleotide polymorphisms (rs429358 and rs7412) were genotyped using the multiplex PCR-based targeted sequencing method, as previously reported28. Depression was diagnosed according to the Diagnostic and Statistical Manual of Mental Disorders, Third Revised Edition29.

Diagnosis of dementia

Dementia was diagnosed according to the Diagnostic and Statistical Manual of Mental Disorders, Third Revised Edition29. The diagnosis of AD was made based on the following criteria: the National Institute of Neurological and Communicative Disorders and Stroke and the Alzheimer’s Disease and Related Disorders Association criteria30. The diagnosis was made by the study team’s expert psychiatrists and neurologists using a standardized diagnostic system across the eight study sites. If the members of the endpoint determination committee agreed, the diagnosis was confirmed; if not, an endpoint determination committee was convened to confirm the diagnosis through discussion. The diagnostic procedure has been previously reported22. Dementia other than AD (isolated type and mixed type) was defined as non- Alzheimer’s dementia (non-AD).

MRI analysis

Brain MRI equipment was set up according to the Alzheimer’s Disease Neuroimaging Initiative (ADNI) study brain MRI protocols31 and T1-weighted image (T1WI) parameters at all study sites. Details of the MRI at each site are provided in Supplemental Table 2.

Table 2 Age- and sex-adjusted and multivariable-adjusted odds ratios for the presence of all-cause dementia and dementia subtypes according to serum high-sensitivity C-reactive protein level.
Full size table

Segmentation and volumetric measurements of cortical and subcortical brain structures were performed automatically using FreeSurfer software (http://surfer.nmr.mgh.harvard.edu), version 5.3. Total brain volume was calculated from brain segment volumes excluding ventricles. Cortical segmentation was performed using the Desikan-Killiany atlas32. The eTIV of each subject was also calculated using the standard FreeSurfer processing pipeline by exploiting the relationship between the intracranial volume and the linear transformation to the atlas template33. Specifically, FreeSurfer estimates of each brain parameter were fitted with a linear regression model adjusted for age, age squared, and sex. If the residuals of an individual participant’s data were less than − 4 standard deviation (SD) units or greater than + 4 SD units from the linear regression, these data were considered extreme outliers. Extreme outliers in eTIV and volumes of at least five brain regions were excluded.

We selected from the gray matter 4 representative cerebral lobes, i.e., the frontal cortex, parietal cortex, temporal cortex, and occipital cortex, and 3 dementia-related brain regions, i.e., the insular cortex, hippocampus, and amygdala, based on the findings of a previous study34.

Statistical analyses

For patient background, comparisons among the four groups of serum hs-CRP < 1.0 mg/L, 1.0–1.9 mg/L, 2.0–2.9 mg/L, ≥ 3.0 mg/L were evaluated with the logistic regression analysis for binary variables and linear regression analysis for continuous variables. Age- and sex-adjusted prevalence of all-cause dementia and its subtypes was estimated by the direct method with 5-year age groups. Age- and sex-adjusted and multivariable-adjusted odds ratios for presence of all-cause dementia and its subtypes were calculated by logistic regression analysis. Trends between serum hs-CRP levels were calculated using multiple regression analysis. Multivariable adjustment was performed for age, sex, education, hypertension, ischemic heart disease, diabetes mellitus, serum total cholesterol level, BMI, chronic kidney disease, abnormal electrocardiogram (ECG), previous stroke, smoking, alcohol consumption, exercise habits, APOE ε4 carrier status, depression, and research site. Of the participants selected for this analysis, 7.7% (= 774/10,085) were excluded from the multivariable-adjusted analysis due to missing data for any covariates. Mean values of brain volume were adjusted for the same adjustment factors and subjected to analysis of covariance. Two-tailed p value < 0.05 were considered significant. In the analysis for the association between serum hs-CRP levels and each regional brain volume, false discovery rate (FDR)-corrected p values (i.e., q values) was calculated by using the Benjamini–Hochberg methods35 in consideration of multiple comparisons, and a q-value < 0.05 was considered statistically significant. All analyses were conducted using the Statistical Package for the Social Sciences (version 23; SPSS Inc., Chicago, IL, USA).

Results

Baseline characteristics of the study population

The patient background of the study population is summarized in Table 1. Age, low education, current smoker, BMI, hypertension, diabetes mellitus, ischemic heart disease, chronic kidney disease, stroke, and ECG abnormalities were significantly higher with higher serum hs-CRP levels. Frequency of female, serum total cholesterol levels, regular exercise, and frequency of APOE ε4 carriers were significantly lower with higher serum hs-CRP levels.

Associations of serum hs-CRP levels with all-cause dementia and dementia subtypes

First, the age- and sex-adjusted prevalence of all-cause dementia, AD and non-AD are shown in Fig. 2. The age- and sex-adjusted prevalence of all-cause dementia, AD and non-AD increased significantly with rising serum hs-CRP levels (p for trend < 0.001, p for trend = 0.004, p for trend = 0.002, respectively). Moreover, the odds ratios of the presence of all-cause dementia, AD and non-AD according to serum hs-CRP level were calculated; as shown in Table 2. The multivariable-adjusted odds ratios for the presence of all-cause dementia were 1.04 (95% confidence interval [CI] 0.76–1.43), 1.68 (95%CI 1.08–2.61), and 1.51 (95%CI 1.08–2.11) for 1.0–1.9 mg/L, 2.0–2.9 mg/L, and ≥ 3.0 mg/L, respectively, compared to < 1.0 mg/L, with a significant increase with higher serum hs-CRP levels (p for trend < 0.001). Similarly, as shown in Table 2, the multivariable-adjusted odds ratios for the presence of AD were 0.72 (95%CI 0.48–1.08), 1.76 (95%CI 1.08–2.89), and 1.61 (95%CI 1.11–2.35), for 1.0–1.9 mg/L, 2.0–2.9 mg/L, and ≥ 3.0 mg/L, respectively, compared to < 1.0 mg/L, increasing significantly with higher serum hs-CRP levels (p for trend = 0.001). On the other hand, there was no evidence of significant associations between serum hs-CRP levels and the multivariable-adjusted odds ratio of non-AD (p for trend = 0.47). In a sensitivity analysis excluding 491 patients with serum hs-CRP ≥ 5.0 mg/L, the significant association between serum hs-CRP and not only all-cause dementia, but also AD, remained (Supplemental Table 3).

Figure 2
figure 2

The age- and sex-adjusted prevalence of all-cause dementia and dementia subtypes according to serum high-sensitivity C-reactive protein level. The data are expressed as the prevalence after adjusting for age and sex. (a) p < 0.05, compared to the lowest serum high-sensitivity C-reactive protein group.

Full size image
Table 3 Multivariable-adjusted mean values (95% confidence intervals) of total brain volume and each regional brain volume to estimated total intracranial volume ratios according to serum high-sensitivity C-reactive protein level in all participants.
Full size table

Associations of serum hs-CRP levels with brain volume

Next, Table 3 shows the relationships between serum hs-CRP levels and each brain volume/total intracranial volume ratio. In all subjects included in the imaging analysis, the multivariable-adjusted mean value of total brain volume/eTIV ratio decreased significantly with increasing serum hs-CRP levels (p for trend = 0.02) (Table 3). With regard to regional brain volume, higher serum hs-CRP levels were associated significantly with lower multivariable-adjusted mean values of temporal cortex volume/eTIV ratio (p for trend = 0.004). This significant association was also observed, even after FDR correction (q-value = 0.03). Sensitivity analysis after excluding 353 participants with dementia found a similar negative association between serum hs-CRP levels and mean values of temporal cortex volume/eTIV ratio (%): serum hs-CRP < 1.0 mg/L, 4.33 (95%CI 4.24–4.42); 1.0–1.9 mg/L, 4.33 (4.24–4.42); 2.0–2.9 mg/L, 4.36 (4.26–4.46); ≥ 3.0 mg/L, 4.27 (4.18–4.37); p for trend = 0.02 (data not shown).

Discussion

In this large cohort study of Japanese older people, the odds ratios for the presence of all-cause dementia and of AD were positively correlated with serum hs-CRP levels. In addition, the temporal cortex volume/eTIV ratio was inversely correlated with serum hs-CRP levels. This significant relationship also was maintained in a sensitivity analysis after excluding patients with dementia.

Though recent reports that vascular inflammation and neuroinflammation are risk factors for AD have accumulated1,2,3, reports on the association between serum hs-CRP levels and dementia including AD are inconsistent. A recent meta-analysis found that serum hs-CRP was associated with all-cause dementia, but not with AD13. In contrast, the present study showed that elevated levels of serum hs-CRP contributed to an increased odds ratio not only for all-cause dementia, but also for AD alone, and furthermore did not contribute to a significant increased odds ratio for non-AD. The multivariable-adjusted odds ratios for the presence of all-cause dementia and AD for every 1 SD increment in log (serum hs-CRP) were 1.06 (95% confidence interval [CI] 0.97–1.15) and 1.05 (95%CI 0.95–1.15), respectively, which were not significant. The results show that there may not be a linear relationship between hs-CRP and dementia and between hs-CRP and AD, but there is a relationship between elevated hs-CRP and dementia, especially AD. Regarding the association between serum hs-CRP and brain atrophy, previous studies have reported that elevated serum hs-CRP is associated with temporal lobe atrophy17,35, and the present results were consistent with those findings. A longitudinal study using data from the Framingham Offspring Study found no significant association between elevated serum hs-CRP levels and brain atrophy after multivariable adjustment (adjustment factors: age, sex, time to MRI scan, educational history, and APOE ε4 carrier status), and elevated serum hs-CRP levels were significantly associated with temporal lobe atrophy in APOE ε4 carriers18. The present study had a larger sample size than these existing studies, and it included potential confounders not addressed in other studies as adjustment factors.

It is widely known that an inflammatory response occurs in the brains of AD patients. Activation of microglia, immunocompetent cells that release inflammatory cytokines, has been reported to occur around the senile plaque in AD brains36. The association of peripherally derived serum hs-CRP with AD may be explained by the fact that peripheral inflammation transfers into the brain via disruption of the blood–brain barrier37,38. Recent studies have also shown that the migration of immune cells of peripheral origin, such as T cells and neutrophils, into the brain is involved in the brain aging and neurodegeneration39,40,41. Interestingly, serum hs-CRP itself was reported to increase the permeability of the cerebral blood–brain barrier42. In addition, an association between cerebral small vessel disease (white matter lesions, perivascular space, etc.) and risk of dementia, especially AD, has recently been reported43,44,45,46, and vascular risk factors are associated with the development of cerebral small vessel disease47. The background of the involvement of vascular risk factors in AD can be related to vascular inflammation, i.e., inflammatory markers.

The fact that elevated levels of inflammation were associated with temporal cortex atrophy in the present study suggests that the temporal cortex is particularly vulnerable to inflammation among brain regions. Differences in the distribution pattern of inflammatory cytokine receptors may underlie why some brain regions are more vulnerable to inflammation. In AD, the density and expression of inflammatory cytokine receptors has been reported to be increased in neurodegenerative regions, including the prefrontal cortex and medial temporal lobes48. Interestingly, plasma inflammation predicted phenotypic conversion and clinical progression of autosomal dominant frontotemporal lobar degeneration49. It has also been reported that higher serum hs-CRP levels were associated with decreased frontotemporal functional network connectivity in community-dwelling older adults50.

The strengths of this study are the large sample size and the ability to adjust for potential confounding factors. However, there are some limitations. First, because this was a cross-sectional study, the causal relationships among serum hs-CRP, dementia including AD, and brain volume cannot be addressed. In addition, the ability to generalize the results of this study to other racial groups is limited due to differences in lifestyle and background. Longitudinal, world-wide, and basic medical studies are needed to clarify the mechanism of the association between inflammation and AD.

Conclusions

This study’s results suggest that elevated serum hs-CRP levels are associated with greater risk of presence of dementia, especially AD, and of temporal cortex atrophy in a community-dwelling Japanese older population.