Serum C-reactive protein is linked
to cerebral microstructural integrity and cognitive function
"we found higher hs-CRP values to be associated with worse performance in executive functions, including tests of psychomotor speed and attention. Other cognitive domains, i.e., memory and linguistic skills, showed no association with CRP".....Serum hs-CRP levels were higher in subjects with a history of hypertension....Higher high-density lipoprotein levels were associated with lower hs-CRP ....."The phenomenon of cognitive aging is characterized by deficits in executive functions, information processing speed, and attention. Histopathologic studies showed an age-related decrease of myelin and the number of myelinated fibers.31 In vivo, DTI appears to be the most sensitive imaging measure"....."Recent interventional studies using antiinflammatory drugs such as aspirin and statins to lower circulating CRP levels showed a significant reduction in the incidence of cardiovascular events.36,37 There is also evidence of a beneficial effect of lifestyle interventions on cognitive functions, such as physical activity and body weight control,38,39 which have been shown to decrease circulating CRP levels.38,39 Whether lowering of CRP can also prevent cognitive decline and/or microstructural white matter alterations needs to be addressed in upcoming clinical trials."
NEUROLOGY March 2010;74:1022-1029
H. Wersching, MD, T. Duning, MD, H. Lohmann, PhD, S. Mohammadi, PhD, C. Stehling, MD, M. Fobker, MD, M. Conty, J. Minnerup, MD, E.B. Ringelstein, MD, K. Berger, MD, M. Deppe, PhD and S. Knecht, MD
From the Department of Neurology (H.W., T.D., H.L., S.M., M.C., J.M., E.B.R., M.D., S.K.), Institute of Epidemiology and Social Medicine (H.W., K.B.), Institute of Clinical Radiology (C.S.), and Institute of Clinical Chemistry and Laboratory Medicine (M.F.), University of Mčnster, Mčnster, Germany.Address correspondence and reprint requests to Dr. Heike Wersching, Institute of Epidemiology and Social Medicine, University of Mčnster, Domagkstraže 3, 48149 Mčnster, Germany email@example.com
"C-reactive protein (CRP) is a sensitive marker of systemic low-grade inflammation....identified as an independent predictor of clinical endpoints, such as myocardial infarction and stroke,1,2 and has been linked to an increased risk of Alzheimer disease and vascular dementia.....The relation of CRP to subclinical functional changes, however, is less clear.....[in this study] Higher levels of hs-CRP were associated with worse performance in executive function after adjustment for age, gender, education, and cardiovascular risk factors in multiple regression analysis....These data suggest that low-grade inflammation as assessed by high-sensitivity C-reactive protein is associated with cerebral microstructural disintegration that predominantly affects frontal pathways and corresponding executive function.....executive function, the cognitive domain most vulnerable to vascular disease..... These findings suggest the role of hs-CRP as a very sensitive and early marker of cerebral small vessel disease, maybe losing its predictive power in advanced cerebrovascular disease, which might explain previous inconsistent findings.8,9....
.....Regarding pathophysiologic mechanisms, there is an ongoing debate whether CRP is a vascular risk factor itself or rather an epiphenomenon of underlying atherosclerotic disease. Recent genetic analyses argue against a causal role in systemic atherosclerosis25 and cerebral small vessel disease.26 It is also known that concomitant vascular risk factors such as smoking27 and overweight28 directly lead to elevation of CRP. Nevertheless, serum markers of endothelial activation were found to be elevated in patients with WMH, thus arguing for an inflammatory endothelial dysfunction in cerebral small vessel disease.29 Finally, whether being a causal agent or a downstream marker of inflammation, CRP itself is known to be involved in endothelial cell activation, thus actively participating in processes of atherosclerosis"
Objective: C-reactive protein is a marker of inflammation and vascular disease. It also seems to be associated with an increased risk of dementia. To better understand potential underlying mechanisms, we assessed microstructural brain integrity and cognitive performance relative to serum levels of high-sensitivity C-reactive protein (hs-CRP).
Methods: We cross-sectionally examined 447 community-dwelling and stroke-free individuals from the Systematic Evaluation and Alteration of Risk Factors for Cognitive Health (SEARCH) Health Study (mean age 63 years, 248 female). High-field MRI was performed in 321 of these subjects. Imaging measures included fluid-attenuated inversion recovery sequences for assessment of white matter hyperintensities, automated quantification of brain parenchyma volumes, and diffusion tensor imaging for calculation of global and regional white matter integrity, quantified by fractional anisotropy (FA). Psychometric analyses covered verbal memory, word fluency, and executive functions.
Results: Higher levels of hs-CRP were associated with worse performance in executive function after adjustment for age, gender, education, and cardiovascular risk factors in multiple regression analysis (ß = -0.095, p = 0.02). Moreover, higher hs-CRP was related to reduced global fractional anisotropy (ß = -0.237, p < 0.001), as well as regional FA scores of the frontal lobes (ß = -0.246, p < 0.001), the corona radiata (ß = -0.222, p < 0.001), and the corpus callosum (ß = -0.141, p = 0.016), in particular the genu (ß = -0.174, p = 0.004). We did not observe a significant association of hs-CRP with measures of white matter hyperintensities or brain atrophy.
Conclusion: These data suggest that low-grade inflammation as assessed by high-sensitivity C-reactive protein is associated with cerebral microstructural disintegration that predominantly affects frontal pathways and corresponding executive function.
Abbreviations: CRP = C-reactive protein; DTI = diffusion tensor imaging; FA = fractional anisotropy; FLAIR = fluid-attenuated inversion recovery; GMF = gray matter fraction; hs-CRP = high-sensitivity C-reactive protein; ROI = region of interest; SEARCH = Systematic Evaluation and Alteration of Risk Factors for Cognitive Health; TE = echo time; TI = inversion time; TR = repetition time; WM = white matter; WMF = white matter fraction; WMH = white matter hyperintensities.
Inflammatory processes are involved in all stages of the development of atherosclerosis. C-reactive protein (CRP) is a sensitive marker of systemic low-grade inflammation. It has been identified as an independent predictor of clinical endpoints, such as myocardial infarction and stroke,1,2 and has been linked to an increased risk of Alzheimer disease and vascular dementia.3 The relation of CRP to subclinical functional changes, however, is less clear. Histopathologic findings showed that CRP has direct prothrombotic and proatherosclerotic effects, suggesting it is primarily related to executive function, the cognitive domain most vulnerable to vascular disease.4 MRI might provide further insight into the relation between inflammatory markers and cognitive function. However, findings from imaging studies using several methodical approaches are inconsistent. Few reports have demonstrated an association of CRP with cerebral small vessel disease as measured by white matter hyperintensities (WMH) and presence of lacunar infarcts.5-7 However, these findings lack replication.8,9 Recent technological advances now allow for more subtle assessment of structural brain integrity. Diffusion tensor imaging (DTI) is an imaging technique that evaluates white matter (WM) microstructure and allows detecting even subtle pathologic changes of fiber integrity.10 DTI changes correlate with clinical symptoms, histopathologic changes, and neuropsychological deficits in early stages of vascular and degenerative brain damage.11,12
The objective of this study was to evaluate the association between increased CRP serum levels and cognitive function in a population-based sample of elderly stroke-free individuals without dementia. In order to identify primary pathoanatomic correlates of this relationship, we applied different MRI techniques, ranging from semiquantitative measurement of WMHs and automated calculation of brain parenchyma volumes to DTI to detect even subtle neural changes.
In our population-based sample of 447 stroke-free elderly individuals, we found higher hs-CRP values to be associated with worse performance in executive functions, including tests of psychomotor speed and attention. Other cognitive domains, i.e., memory and linguistic skills, showed no association with CRP.
Our findings are in line with previous population- based studies that found elevated CRP levels to be related with cognitive decline.8,18,19 Tests of executive function have recently been identified as predictors of global cognitive impairment20 and instrumental activities of daily living21 in community-dwelling elderly individuals without dementia, thus emphasizing the clinical relevance of deviations in executive function within a normal cognitive range.
We also found a strong association of CRP with global FA and frontal FA, which was triggered by subjects with no or mild WMH, whereas CRP was not independently related to global FA in subjects with moderate to severe white matter disease. Higher CRP levels were also associated with higher semiquantitative scores of WMH and lower brain volume measures, nonetheless not significant. These findings suggest the role of hs-CRP as a very sensitive and early marker of cerebral small vessel disease, maybe losing its predictive power in advanced cerebrovascular disease, which might explain previous inconsistent findings.8,9
We could also show a significant correlation of hs-CRP and FA in motor regions (corona radiata and corticospinal tract). Since no motor task was included in our study, we cannot rule out that motor regions were affected in frontal areas as well, implying a specific effect of hs-CRP on motor fibers. To elucidate this issue, longitudinal data would be necessary with a specific focus on the affection of distinct fibers, the localization of progressive FA changes, and their possible transformation into WMH over time.
DTI in vivo quantifies the extent of diffusivity of water molecules as well as tissue anisotropy, which is the spatial restriction of water movements in certain directions. Normal WM is highly anisotropic. Postmortem studies in humans revealed that lower FA reflects the extent of astrogliosis and loss of myelin and axons.22 In cerebral small vessel disease FA is not only reduced in hyperintense white matter lesions but also in areas of so-called normal-appearing white matter, thus providing a better index of tissue damage than conventional MRI.23 FA was shown to be the most sensitive MRI correlate of impaired executive function,24 a cognitive domain highly susceptible to vascular damage.4 Our findings of a reduction of frontal FA with increasing CRP levels suggest a microvascular damage of white matter projections in frontal-subcortical pathways, leading to executive dysfunction.
Regarding pathophysiologic mechanisms, there is an ongoing debate whether CRP is a vascular risk factor itself or rather an epiphenomenon of underlying atherosclerotic disease. Recent genetic analyses argue against a causal role in systemic atherosclerosis25 and cerebral small vessel disease.26 It is also known that concomitant vascular risk factors such as smoking27 and overweight28 directly lead to elevation of CRP. Nevertheless, serum markers of endothelial activation were found to be elevated in patients with WMH, thus arguing for an inflammatory endothelial dysfunction in cerebral small vessel disease.29 Finally, whether being a causal agent or a downstream marker of inflammation, CRP itself is known to be involved in endothelial cell activation, thus actively participating in processes of atherosclerosis.30
The phenomenon of cognitive aging is characterized by deficits in executive functions, information processing speed, and attention. Histopathologic studies showed an age-related decrease of myelin and the number of myelinated fibers.31 In vivo, DTI appears to be the most sensitive imaging measure to determine these disconnections of the white matter, showing a decrease of FA mainly in the frontal white matter, the internal capsule, and the genu of the corpus callosum.12 Given the similarity in functional and structural effects of increased serum CRP and aging, CRP might contribute to the process of cognitive aging, particularly as aging is directly associated with increased inflammatory activity.32 Supporting this hypothesis in our study cohort, we observed a weak correlation of age and CRP. The association of CRP with executive function as well as FA measures was partly attenuated after adjustment for age, and relations between CRP and FA measures were generally stronger in older subjects. Nevertheless, CRP showed strongest correlations with executive functions in the younger subgroup, thus indicating a mediating though not exclusive role of age-related changes in the present associations.
In the current study, we applied comprehensive neuropsychological testing to detect even subtle differences in cognitive functions. Previous studies that failed to show an association of CRP and cognition mainly used crude measures of cognitive function or a short test battery, often lacking sensitive tests of executive function and psychomotor speed.33,34
Nevertheless, based on the cross-sectional design of the study, causality of associations can only be assumed. CRP values were only measured once, but hs-CRP has shown a high degree of measurement stability over time.35
Recent interventional studies using antiinflammatory drugs such as aspirin and statins to lower circulating CRP levels showed a significant reduction in the incidence of cardiovascular events.36,37 There is also evidence of a beneficial effect of lifestyle interventions on cognitive functions, such as physical activity and body weight control,38,39 which have been shown to decrease circulating CRP levels.38,39 Whether lowering of CRP can also prevent cognitive decline and/or microstructural white matter alterations needs to be addressed in upcoming clinical trials.
The demographic and clinical characteristics of the study population are summarized in table 1. The mean age of participants was 63 years, 55.5% were female, and the median hs-CRP serum level was 0.12 mg/dL. Neuropsychological test performances were widely within normal limits.
Associations between hs-CRP and clinical variables.
Across all subjects, the log-transformed hs-CRP increased with age (r = 0.085, p = 0.072), body mass index (r = 0.311, p = 0.001), systolic blood pressure (r = 0.154, p = 0.002), and low-density lipoprotein (r = 0.147, p = 0.002). Serum hs-CRP levels were higher in subjects with a history of hypertension (p = 0.042), a history of dyslipoproteinemia (p = 0.041), and in subjects using antihypertensive agents (p = 0.001). Higher high-density lipoprotein levels were associated with lower hs-CRP (r = -0.204, p = 0.001). There was no association with HbA1c (r = 0.053, p = 0.267), current (p = 0.298) or former (p = 0.231) cigarette smoking, history of diabetes (p = 0.911), and use of statins (p = 0.509) or antithrombotics (p = 0.448).
Association between hs-CRP and cognitive performance.
Table 2 summarizes the regressions between cognitive domains and log-transformed hs-CRP values. Increasing hs-CRP was significantly associated with decreasing performance in executive functions (figure 2). This correlation was attenuated, but still significant after adjustment for cardiovascular risk factors that showed a relation at the p = 0.1 level with hs-CRP in crude analysis (model 3). Since the variables history of hypertension, history of dyslipoproteinemia, and use of antihypertensive agents were highly correlated with the actually measured blood pressure and cholesterol levels, we only included systolic blood pressure, low-density lipoprotein, and high-density lipoprotein in the regression model to avoid colinearity.
Table e-2 provides an overview of the imaging measures of the 312 participants who completed MRI assessment. According to the Fazekas score, 31.8% showed no WMH (score 0), 53.9% had punctate lesions (score 1), 9.0% showed beginning confluence of lesions (score 2), and 5.3% had large confluent areas of WMH (score 3). Of the 219 subjects with punctuate or confluent WMH, 197 showed WMH in frontal brain regions, 167 showed WMH in temporoparietal regions, and 58 subjects showed WMH in occipital regions. Higher hs-CRP levels were associated with more WMH and less brain parenchyma volumes; however, these correlations were not significant (table 2).
Regions of interest measurement of FA.
Unadjusted regression analyses showed a reduction of mean FA scores associated with increasing CRP values in frontal and motor regions. Significance was set at a level of p = 0.001, to adjust for multiple testing (table 2). Significantly correlated regions were further analyzed with regression models 2 and 3. In summary, even after adjustment for the full set of risk factors, we found highly significant associations of hs-CRP with the frontal lobe, the corona radiata, the corticospinal tract, and the corpus callosum, in particular in its frontal parts (genu). Pearson correlation showed global and regional FA scores to be significantly correlated with executive functions (data not shown).
Voxel-based analysis of FA changes.
The results of a whole-brain, voxel-based analysis of white matter demonstrated significant regions of reduced FA associated with increased CRP values mainly in frontal regions of the brain, in particular in the frontal lobes and in the frontal part of the corpus callosum (figure 3). Adjustment for the full set of covariates (model 3) did not alter the anatomic pattern of correlations significantly. None of the voxels exhibited a significant positive correlation between FA scores and CRP values.
To support the causal role of reduced FA in the pathway of elevated CRP levels and decreased cognitive functions, we performed a partial correlation between hs-CRP and executive function scores adjusting for global FA, which resulted in a strong attenuation of the association (r = -0.069, p = 0.206).
Effect modification by age and the extent of small vessel disease.
To ascertain that our findings were not exclusively due to subjects with evident white matter disease, we re-ran regression analyses in subgroups of participants showing no WMH (Fazekas score 0), mild WMH (score 1), and moderate to severe WMH (score 2 + 3). Strongest associations of CRP with global and regional FA measures were found in the subgroups with no or mild WMH. We also ran separate analyses for subjects with and without WMH in frontal brain regions. In both subgroups, hs-CRP was correlated with frontal FA (ß = -0.233, p = 0.001, and ß = -0.220, p = 0.014).
Finally, we dichotomized the cohort according to age to test for effects of age on the association between CRP and executive functions as well as FA measures. Pearson correlation showed an association (r = -0.129, p = 0.048) of CRP and executive functions in the younger subgroup (mean 57.8 years, range 41-64 years), whereas effects were attenuated (r = -0.103, p = 0.133) in the older subgroup (mean 69.7 years, range 65-85 years).
Estimation of effect size.
In regression analyses using the categorical hs-CRP variable, higher levels of hs-CRP were again associated with worse performance in executive function (ß = -0.092, p = 0.0125, figure 2) and with reduced global and frontal FA (ß = -0.224, p < 0.001 and ß = -0.220, p < 0.001). Compared to the lowest risk level (reference), an increase of hs-CRP to the second risk level was associated with a decrease in frontal FA equivalent to that of 6 years of aging, an increase to the third risk level with a decrease equivalent to12 years of aging.
Study population. Subjects were recruited from the population-based Systematic Evaluation and Alteration of Risk Factors for Cognitive Health (SEARCH) Health Study, which investigates risk factors for cognitive decline in a stroke-free elderly population.13 Community-living individuals aged 40 to 85 years were randomly selected based on dates of birth from the population register of the city of Muenster. Participants received a standardized face-to-face medical interview, a physical examination by a trained study physician, nonfasting blood sampling, a comprehensive neuropsychological assessment, and MRI of the brain at 3.0 Tesla. For details on the assessment of clinical data, see appendix e-1 on the Neurology Web site at www.neurology.org. The time period of baseline recruitment was 53 months (January 2004 to June 2008). The entire SEARCH baseline cohort comprises 729 individuals. Participants who completed both the neuropsychological and the laboratory assessments were eligible for the present investigation. Individuals with dementia or depression and patients with evidence of nonvascular inflammatory disease were excluded (figure 1). MRI data could not be obtained for 126 participants (including cases of claustrophobia and MRI critical metal implants). Thus, all analyses relating to imaging data are based on 321 subjects.
Standard protocol approvals, registrations, and patient consents. The study was approved by the Ethics Committee of the Medical Chamber Westphalia-Lippe (3XKLOSKA1). Each participant gave written informed consent for participation.
C-reactive protein. Plasma hs-CRP was measured from nonfasting blood samples. Plasma hs-CRP levels were assayed by a highly sensitive (detection limit 0.2 mg/L) immunonephelometric method on a BNII Nephelometer (Siemens Healthcare Diagnostics GmbH, Eschborn, Germany). The intraassay coefficient of variation was determined to be 4.7%. The interassay coefficient of variation was determined to be 8%.
Neuropsychological assessment. The neuropsychological test battery was designed to assess a full range of cognitive functions and consists of routinely used standardized neuropsychological tests. A detailed description may be found in Lezak.14 Single tests were grouped into 3 z-transformed composite scores by principal component analysis using oblique (Oblimin with Kaiser normalization) rotation and including coefficients with absolute values above 0.35. The composite scores reflect the cognitive domains of verbal memory, word fluency, and executive functions (table e-1).
MRI acquisition. Image data were obtained on a 3.0 T system (Gyroscan Intera T30, Philips Medical System) with a high-resolution structural T1-weighted 3-dimensional turbo-field-echo sequence for brain volumetry (field of view of 25.6 x 20.5 x 16 cm3, reconstructed after zero filling to 512 x 410 x 320 cubic voxels with an edge length of 0.5 mm), as well as T2-weighted (repetition time [TR] 4,459 msec/echo time [TE] 100 msec) and fluid-attenuated inversion recovery (FLAIR) imaging (TE 126 msec/TR 10 s/inversion time [TI] 2,200 msec).
For DTI, we employed echoplanar imaging (TR/TE/TI = 10 s/90 msec/2.4 s) with two b factors (0 s/mm2 and 1000 s/mm2) in 6 directions, selective averaging of signals obtained at high b (n = 3). Data were acquired with a multislice technique in 72 axial slices 1.8 mm thick with no gap, field of view 230 x 230 mm, reconstructed to 256 x 256 pixels after zero filling, and voxel edge length of 1.80 x 1.81 x 1.80 mm.
Image analysis. Structural MRI analysis was performed by a single experienced observer blinded to clinical data. WMH were identified on FLAIR scans and calculated semiquantitatively based on 15 brain regions in the periventricular and deep white matter of both hemispheres, the left and right cerebellum, and the brainstem. For each region (left and right separately) a severity score extending from 0 = no lesion to 5 = large confluent lesions was applied, thus defining a total WMH score of 0-165. We also calculated regional WMH scores based on the severity scores in frontal, temporoparietal, and occipital brain regions.
WMH were additionally rated on a 3-point scale according to the periventricular score of Fazekas15: 0 = no WMH, 1 = punctate foci of WMH, 2 = beginning confluence of foci of WMH, and 3 = large confluent areas of WMH.
Brain tissue volumes, normalized for subject head size, were calculated from the high-resolution T1-weighted images, using the cross-sectional version of the Structural Imaging Evaluation of Normalized Atrophy software (SIENAx).16 After tissue estimation of volumes of gray matter, white matter, and CSF, a white matter fraction (WMF) and a gray matter fraction (GMF) were calculated by dividing the white/gray matter volume by intracranial volume.
Diffusion tensor and fractional anisotropy (FA) field maps were calculated from spatially normalized images. The method was described in detail previously.10 Voxel-based analysis of the FA image was performed using SPM5 (Wellcome Department of Imaging Neuroscience, London). The linear regression tool of the software was employed to correlate maps of decreased FA with hs-CRP values. Statistical threshold for the correlation analysis was set at p = 0.01 for multiple comparisons (FWE).
To assess the magnitude of regional FA alterations, we also performed additional quantitative region of interest (ROI) analyses. Mean FA values were calculated within 11 defined ROIs (figure e-1), which were derived from an averaged and symmetrized (x-axis) mask of 160 healthy individuals with FA values >0.4 by deleting voxels not associated with the respective structures.
Statistical analysis. Pearson correlation and Student t test were used to assess associations of hs-CRP values with demographic and clinical variables (appendix e-1). As hs-CRP and WMH score were positively skewed, logarithmic transformation was used to obtain normal distribution for statistical analysis. To assess the relationship between hs-CRP and cognitive domains, 3 sets of multiple linear regression models were carried out. First, unadjusted regression (correlation) coefficients were calculated (model 1). The second model adjusted for age, gender, and education. The third model further accounted for possible confounders and mediators by adjusting for factors associated with hs-CRP in crude analyses (p = 0.10). Interactions between factors were considered using stepwise regression analysis. The same set of models was applied to assess the relationship between hs-CRP and imaging data, calculating separate regression analysis with WMH score, brain volume measures, and regional FA values (ROIs) as dependent variables. Regression models 3 were repeated with a categorical variable of hs-CRP reflecting the CDC/AHA risk levels (hs-CRP <0.1 mg/dL, hs-CRP 0.1-0.3 mg/dL, and hs-CRP >0.3 mg/dL).17 The relation of imaging parameters and cognitive domains were analyzed using Pearson correlations. Data analysis was carried out using SPSS 15.0.0. Voxel-by-voxel regression analyses using SPM5 were also calculated for the 3 regression models.