Abstract
Background:
Hemostasis is a key factor in cerebrovascular disease, but the association of hemostatic factors with cognitive decline is unclear.
Objective:
To prospectively evaluate associations of 20 hemostatic factor levels with changes in cognition over ≥8 years of follow-up in the Cardiovascular Health Study (CHS) of older adults.
Methods:
We included participants of an existing CHS cross-sectional substudy (n=400) with hemostatic factors measured in 1989/1990. Between 1989/1990 and 1998/1999, cognitive function was measured using the Modified Mini-Mental State Examination (3MSE) and Digit Symbol Substitution Tests (DSST). Mixed-effects linear regression models estimated change in cognitive function over time, adjusting for sociodemographic and clinical factors and APOE genotype, using Bonferroni adjustment. We also derived principal components to account for the interrelationship among factors.
Results:
Of 20 factors evaluated individually, only higher levels of plasmin-α2-antiplasmin complex (PAP), tissue factor pathway inhibitor (TFPI), and lower factor X (FXc) levels were associated with faster cognitive decline, estimated by annual change in 3MSE points (1-standard deviation PAP β=−0.65, 95%CI: −1.08, −0.21; TFPI β=−0.55, 95%CI: −0.90, −0.19; FXc β=0.52, 95%CI: 0.21, 0.84). One of four principal components, loading positively on D-dimer, prothrombin fragment 1.2 (F1.2), and PAP was significantly associated with change in 3MSE.
Conclusions:
Levels of PAP, TPFI, and FXc and a combination of factors driven by PAP, D-dimer, and F1.2 were associated with cognitive decline. Whether these findings can be used to improve dementia prevention or prediction requires further study.
Keywords: Aged, Blood Coagulation, Cognition, Fibrinolysis, Hemostasis
Introduction
Dementia is a multifactorial disease, with vascular dysregulation thought to contribute to cognitive decline and dementia onset.[1] Hemostasis, which involves the balance of coagulation and fibrinolysis, is associated with both arterial and venous thrombotic event risk, including stroke,[2, 3] and cardiovascular disease is intimately related to dementia. Despite this, many hemostatic factors important to vascular function have not been evaluated in relation to cognitive impairment, have been evaluated in only a single study, or have been inconsistently associated with cognitive impairment and dementia risk.[3–6] Determining whether levels of hemostatic factors are associated with cognitive decline offers an opportunity to better understand the biologic mechanisms associated with cognitive change and to potentially identify markers for use in earlier detection of changing cognition.
Of the hemostatic factors previously studied in relation to cognition, fibrinogen and D-dimer have been studied most frequently.[4] Results were inconsistent between studies, potentially due to study design and differences in cognitive impairment definitions among cohorts.[4] Higher fibrinogen levels were associated with vascular cognitive impairment risk in a prospective study[5], but not with non-vascular cognitive impairment in this same prospective cohort[5], nor with cognitive impairment measured by a telephone-based assessment of cognitive function in a case-control study.[6] Higher D-dimer levels were not associated with cognitive impairment in either of these studies.[5, 6]
To address the relation of hemostatic factors with cognitive decline more comprehensively, we prospectively evaluated cognitive decline in relation to baseline measures of 22 hemostatic factors measured in the Cardiovascular Health Study (CHS), a population-based cohort study of older adults. We hypothesized that higher levels of procoagulant and fibrinolytic factors and lower levels of anticoagulant factors would be associated with a faster decline in cognition over follow-up from 1989/1990 to 1998/1999.
Methods
Study Design and Participants
CHS is a prospective cohort study of community-dwelling adults aged 65 years and older, with detailed methods described previously.[7] At CHS baseline (1989/1990), 5,201 participants were recruited in four United States communities using Medicare eligibility lists. Demographic, medical history, and cognitive function data were collected at annual clinic visits from 1989/1990 until 1998/1999.
Men and women included in this prospective analysis were participants of an existing substudy that included 400 participants randomly selected from the 3,352 CHS participants who did not have prevalent coronary heart disease, cerebrovascular disease, and peripheral vascular disease at baseline,[3, 8] for whom hemostatic factor levels had been previously measured in plasma samples collected in 1989/1990. Individuals using anticoagulant medications in 1989/1990 were excluded (n=4), resulting in 396 individuals eligible for this analysis. Participants provided informed consent and all participating sites received institutional review board approval.
Assessment of Cognitive Function
To assess cognitive function, the 100-point Modified Mini-Mental State Examination (3MSE)[9] and Digit Symbol Substitution Test (DSST)[10] were first administered in 1990/1991 and 1989/1990, respectively, with administration of these tests continuing annually until 1998/1999.[11] The 3MSE assesses memory, orientation, fluency, problem solving, and ability to follow instructions, resulting in a global score ranging from zero (worst) to 100 (best). The DSST assesses information processing speed, resulting in a score ranging from zero (worst) to 90 (best).[12, 13]
For some participants who could not attend in-person cognitive assessments, from 1995/1996 to 1998/1999, the Telephone Interview for Cognitive Status (TICS) and the Informant Questionnaire on Cognitive Decline in the Elderly (IQCODE) were used annually to estimate missing 3MSE scores (1995/1996 n=11 estimated; 1996/1997 n=24; 1997/1998 n=39; 1998/199 n=26), using equations previously developed and used for CHS analyses.[12]
Assessment of Hemostatic Factor Levels
Levels of hemostatic factors were measured in fasting plasma samples collected in 1989/1990 as part of a CHS substudy; all hemostatic parameter measures related to coagulation or fibrinolysis that were measured as part of this substudy and found to be measured reproducibly were included in this analysis. Pairwise correlations for all hemostatic factors are included in Supplemental Table 1; none were correlated at a level >0.07. Detailed methods have been previously reported for included hemostatic factors, with all CVs as follows being inter-assay, generated using control pools prepared from healthy individuals: D-dimer (CV: 7.0%)[8, 14, 15], factor VII antigen (FVII antigen) (CV: 6.6%)[16], factor VII (FVIIc) (CV: 5.3%)[3], factor VIII (FVIIIc) (CV: 9.7%)[3], factor IX (FIXc) (CV: 5.8%)[17, 18], factor X (FXc) (CV: 4.7%)[17, 18], fibrinogen (CV: 3.0%)[3], fibrinopeptide A (CV: 12.4% at 1.5 ng/ml)[17, 19] [20], prothrombin fragment 1.2 (F1.2) (CV: 8.5%)[17] [20], plasminogen activator inhibitor antigen (PAI-1) (CV: 8.4%)[17] [18], plasmin-α2-antiplasmin complex (PAP) (CV: 3.0%)[17, 18], plasminogen (CV: 3.6%)[8], tissue plasminogen activator antigen (tPA) (CV: 7.0%)[18], tPA-PAI-1 complex (CV: 14.3%)[18], antithrombin activity (CV: 4.5%)[8], protein C antigen (CV: 2.5%)[8, 21], total protein S antigen (CV: 6.7%)[8, 22], free protein S antigen (CV: 9.9%)[8, 22], and tissue factor pathway inhibitor activity (TFPI) (CV: 9.5%)[8]. Platelet count was measured at CHS field centers by Coulter counters.[23]
A priori, we categorized hemostatic factors into 10 procoagulant factors (D-dimer, FVII antigen, FVIIc, FVIIIc, FIXc, FXc, fibrinogen, fibrinopeptide A, platelet count, and F1.2); 5 fibrinolytic factors (PAI-1, PAP, plasminogen, tPA, and tPA-PAI-1 complex); and 5 anticoagulant factors (antithrombin activity, protein C antigen, protein S antigen [total and free], and TFPI).
Covariates and Other Measures
Participants were enrolled from four field center locations: Washington County, MD; Pittsburgh, PA; Forsyth County, NC; Sacramento County, CA. At baseline, participants self-reported their sex, age, educational level, race (categorized as white vs. nonwhite), current and former cigarette smoking, alcohol consumption, and use of hormone therapy (in women only).[7] Prevalent diabetes was defined as a fasting glucose concentration ≥126 mg/dl or use of diabetes medications. Hypertension was defined by systolic blood pressure ≥140 mm Hg or diastolic blood pressure ≥90 mm Hg. Antihypertensive medication use was self-reported. Body mass index (BMI) was calculated as weight in kg divided by height in meters squared. Total kcal/week from leisure-time activity was calculated using participant responses regarding types, frequency, and duration of 15 different activities during the prior two weeks. Mild cognitive impairment (MCI) at baseline was defined as a score <88 on the 3MSE.[24] Total cholesterol levels were measured in fasting plasma.[7] Among participants consenting to genetic analyses, apolipoprotein E (APOE) e4 genotyping was performed using the methods of Hixson and Vernier[25]; data regarding presence of the APOE-e4 allele was unavailable for 43 participants (10.9%).
Statistical Analysis
Analyses were conducted using Stata 15 (Stata Corp, College Station, TX, USA). Hemostatic factors were winsorized at the 99th and 1st percentiles. To evaluate the association between a 1-standard deviation (SD) difference in each hemostatic factor, separately, and cognitive function over time (as estimated by annual rate of change in mean 3MSE and DSST measures over follow-up), we used mixed-effects linear regression models. We included all measures of cognition available during follow-up, with follow-up in DSST analyses beginning in 1989/1990 and in 3MSE analyses beginning in 1990/1991, with follow-up to 1998/1999. Models included terms for each hemostatic factor individually, time under observation (i.e. years from enrollment to each exam during follow-up), the interaction between the hemostatic factor level and time under observation, and potential confounders. Two multivariable models included confounders measured at baseline, identified a priori; model 1 confounders included age, sex, education, race, and field center, and model 2 included model 1 covariates + BMI, physical activity, alcohol intake, current and former smoking, hormone therapy use in women, prevalent diabetes, systolic blood pressure, hypertensive medication use, total cholesterol, and the presence of the APOE-e4 allele (including a missing category). Models separately evaluated annual rates of change in 3MSE and DSST scores. We evaluated statistical significance using a Bonferroni-adjusted alpha level of 0.005 for procoagulant factors (alpha of 0.05/10 factors evaluated), 0.01 for fibrinolytic factors (0.05/5), and 0.01 for anticoagulant factors (0.05/10).
To evaluate whether baseline cognitive impairment changed our results, we conducted a sensitivity analysis in which we excluded all individuals scoring in the approximately lowest 10th percentile on the 3MSE at baseline (i.e. ≤77 on the 3MSE) (n=33), as well as all individuals missing their baseline 3MSE measurements (n=33). In a second sensitivity analysis that aimed to determine whether the use of TICS and IQCODE data to estimate missing 3MSE values impacted our findings, we conducted a sensitivity analysis in which we did not use TICS or IQCODE data to estimate missing values.
Principal Component Analysis.
Given the established interrelationships of hemostatic factors due to common biologic sources and functions, we conducted a principal component analysis (PCA) to reduce our data into uncorrelated components. PCA identifies patterns among groups of variables that can then be used to reduce them to principal components, which are linear, uncorrelated combinations of variables that account for incrementally more of the variability in the original variables.[26] Hemostatic factors missing for >5% of participants were excluded. We selected the number of components based on graphical inspection of scree plots; we identified the leftmost point on this plot that was approximately on a straight line, and all points to the left of this were retained in our models (components 1–4). These components were added to our mixed-effects linear regression models, and their significance was evaluated using an alpha level of 0.05. As a sensitivity analysis, we retained all components with an eigenvalue ≥1.0 (components 1–5) in our mixed-effects linear regression model.
Interaction by APOE-e4 Allele Carrier Status.
In secondary analyses, we excluded individuals missing APOE-e4 allele carrier status (n=43) and separately evaluated the presence of interaction by APOE-e4 allele carrier status with each hemostatic factor, using covariates included in model 2, and using the Bonferroni-adjusted alpha levels as defined. In addition to our basic model terms defined previously, models testing for interaction also included terms for APOE-e4 allele carrier status, the interaction between APOE-e4 allele carrier status and time under observation, the interaction between APOE-e4 allele carrier status and the hemostatic factor of interest, and the three-way interaction between APOE-e4 allele carrier status, the hemostatic factor of interest, and time under observation. To test for interaction, we used the likelihood ratio test for interaction.
Results
Eligible participants (n=396) had a median age of 77 years at baseline in 1989/1990 (Table 1) and 50% were women. In Table 2, we report the association of a 1-SD difference in each hemostatic factor level with annual rate of change in mean 3MSE scores over 8-years of follow-up. For example, every 1-SD higher PAP level measured in 1989/1990 was associated with a 0.65-point faster decline in predicted 3MSE score over a 1-year period (β=−0.65; 95% CI: −1.08, −0.21), after adjusting for additional confounders (model 2). To place these annual rates of change into context, each additional year of age at baseline was associated with a 0.17-point faster decline in predicted 3MSE scores (β=−0.17; 95% CI: −0.21, −0.13), suggesting that the decline in 3MSE associated with 1-SD higher levels of PAP is similar to that of 4 additional years of age at baseline, on average. After accounting for multiple comparisons, evidence of a faster decline in 3MSE remained for higher PAP levels, higher TFPI levels, and lower FXc levels in adjusted analyses. We did not observe evidence of a non-linear association between any hemostatic factor and rate of change in 3MSE, after adjusting for multiple comparisons (all p≥0.01) (Supplemental Table 2).
Table 1.
Baseline (1989/1990) characteristics of participants, by sex (n=396).
| Characteristic* | Men (n=197) | Women (n=199) |
|---|---|---|
| Age, median (IQR), years | 77 (12) | 77 (13) |
| Non-white race, % | 3.1 | 10.1 |
| Beyond high school education, % | 49.8 | 49.3 |
| Current smoking, % | 9.6 | 7.0 |
| BMI, median (IQR), kg/m2 | 25.9 (4.8) | 25.1 (6.6) |
| Physical activity, median (IQR), kcal/week | 1348 (2130) | 840 (1403) |
| Number of alcoholic drinks per week, median (IQR) | 0.3 (3.0) | 0.0 (0.5) |
| Current hormone therapy use, % | - | 13.6 |
| Diabetes mellitus, % | 9.6 | 3.0 |
| Hypertension, % | 30.5 | 40.2 |
| Total cholesterol, median (IQR), mg/dL | 192.8 (49.0) | 214.8 (51.0) |
| Hypertension medication use, % | 29.4 | 37.7 |
| Lipid-lowering medication use, % | 0.5 | 1.5 |
| APOE-e4 allele carrier, % | 17.8 | 19.1 |
| Mild cognitive impairment†, % | 36.0 | 22.6 |
| Cognitive scores, median (IQR) | ||
| 3MSE | 90.5 (12.0) | 92.0 (8.0) |
| DSST | 32 (20) | 35 (18) |
| Procoagulant Hemostatic Factors, median (IQR) | ||
| D-dimer, ng/mL | 163 (153) | 178 (173) |
| FVII antigen, % normal | 111 (33) | 129 (48) |
| FVIIc, % normal | 111 (28) | 128 (35) |
| FVIIIc, % normal | 116.5 (44.0) | 126.5 (49.0) |
| FIXc, % normal | 114 (23) | 121 (34) |
| FXc, % normal | 112 (23) | 119 (29) |
| Fibrinogen, mg/dL | 311 (80) | 328 (76) |
| Fibrinopeptide A, ng/mL | 2.5 (2.4) | 3.3 (3.0) |
| Platelet count, x109 L−1 | 222.5 (76) | 253.5 (86) |
| F1.2, nmol/L | 0.31 (0.16) | 0.39 (0.21) |
| Fibrinolytic Hemostatic Factors, median (IQR) | ||
| PAI-1, ng/mL | 31 (32) | 29 (28) |
| PAP, nmol/L | 5.6 (2.8) | 6.1 (2.9) |
| Plasminogen, % normal | 98 (19) | 107 (17) |
| tPA, ng/ml | 10.2 (4.2) | 9.5 (4.5) |
| tPA-PAI-1 complex, nmol/L | 2.0 (1.3) | 1.8 (1.4) |
| Anticoagulant Hemostatic Factors, median (IQR) | ||
| Antithrombin activity, % normal | 101 (18) | 110 (18) |
| Protein C antigen, mg/L | 3.2 (0.7) | 3.6 (0.9) |
| Free Protein S antigen, mg/L | 5.1 (0.97) | 4.8 (1.1) |
| Total Protein S antigen, mg/L | 22.2 (4.2) | 23.1 (4.7) |
| TFPI, U/ml | 125 (29) | 122.5 (33) |
3MSE = modified mini-mental state; BMI = body mass index; DSST = digit symbol substitution test; FVII antigen = factor VII antigen; FVIIc = Factor VII; FVIIIc = factor VIII; FIXc = factor IX; FXc = factor X; F1.2 = prothrombin fragment 1.2; kcal = kilocalories; PAI-1 = plasminogen activator inhibitor-1; PAP = Plasmin-alpha2-antiplasmin complex; SD = standard deviation; tPA = tissue plasminogen activator antigen; TFPI = tissue factor pathway inhibitor activity; y = years.
Characteristics missing in <1% of participants except for current hormone therapy use (11.6% missing among women), diabetes mellitus (5.3% missing), APOE-e4 carrier status (10.9% missing), mild cognitive impairment (8.3%), 3MSE (8.3%), DSST (3.8%), D-dimer (1.3%), FVIII (1.0%), fibrinopeptide (1.3%), platelet count (1.5%), free protein S (9.8%), total protein S (9.3%), TFPI (35.1%).
Baseline mild cognitive impairment defined as a score <88 on 3MSE.
Table 2.
Annual rate of change in mean 3MSE scores over an 8-year period (1990/1991–1998/1999) associated with 1-SD differences in hemostatic factor levels in 1989/1990.
| Linear Exposure Model 1* | Linear Exposure Model 2† | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| SD | n | Beta‡ | 95% CI | p value | n | Beta‡ | 95% CI | p value | |
| Age (1-year difference) | 7.2 | 376 | −0.16 | (−0.20, −0.12) | <0.001 | 333 | −0.17 | (−0.21, −0.13) | <0.001 |
| Procoagulant Hemostatic Factors | |||||||||
| D-dimer, ng/mL | 420.0 | 371 | −0.18 | (−0.60, 0.24) | 0.39 | 329 | −0.22 | (−0.66, 0.23) | 0.34 |
| FVII antigen, % normal | 132.0 | 375 | −0.12 | (−0.93, 0.68) | 0.77 | 332 | 0.031 | (−0.84, 0.90) | 0.94 |
| FVIIc, % normal | 27.7 | 375 | −0.11 | (−0.42, 0.21) | 0.51 | 332 | −0.087 | (−0.41, 0.24) | 0.61 |
| FVIIIc, % normal | 36.5 | 372 | −0.36 | (−0.66, −0.047) | 0.024 | 329 | −0.34 | (−0.66, −0.022) | 0.036 |
| FIXc, % normal | 23.1 | 376 | −0.053 | (−0.38, 0.27) | 0.75 | 333 | −0.029 | (−0.36, 0.30) | 0.86 |
| FXc, % normal | 20.7 | 376 | 0.55 | (0.24, 0.87) | 0.001 | 333 | 0.52 | (0.21, 0.84) | 0.001 |
| Fibrinogen, mg/dL | 70.5 | 374 | −0.13 | (−0.45, 0.19) | 0.41 | 331 | −0.079 | (−0.41, 0.25) | 0.64 |
| Fibrinopeptide A, ng/mL | 38.6 | 371 | −0.30 | (−0.75, 0.16) | 0.20 | 329 | −0.40 | (−0.91, 0.12) | 0.13 |
| Platelet count, x109 L−1 | 70.7 | 370 | 0.22 | (−0.10, 0.53) | 0.18 | 328 | 0.22 | (−0.094, 0.54) | 0.17 |
| F1.2, nmol/L | 0.39 | 375 | −0.27 | (−0.86, 0.33) | 0.38 | 333 | −0.43 | (−1.03, 0.17) | 0.17 |
| Fibrinolytic Hemostatic Factors | |||||||||
| PAI-1, ng/mL | 30.1 | 376 | 0.40 | (0.052, 0.75) | 0.024 | 333 | 0.39 | (0.040, 0.74) | 0.029 |
| PAP, nmol/L | 3.1 | 375 | −0.63 | (−1.06, −0.20) | 0.004 | 332 | −0.65 | (−1.08, −0.21) | 0.004 |
| Plasminogen, % normal | 16.8 | 376 | 0.10 | (−0.22, 0.43) | 0.53 | 333 | 0.16 | (−0.17, 0.49) | 0.35 |
| tPA, ng/ml | 6.7 | 374 | 0.057 | (−0.38, 0.50) | 0.80 | 331 | 0.083 | (−0.35, 0.52) | 0.71 |
| tPA-PAI-1 complex, nmol/L | 1.6 | 374 | 0.28 | (−0.19, 0.75) | 0.25 | 331 | 0.31 | (−0.16, 0.78) | 0.20 |
| Anticoagulant Hemostatic Factors | |||||||||
| Antithrombin activity, % normal | 15.1 | 376 | 0.0093 | (−0.31, 0.33) | 0.95 | 333 | 0.063 | (−0.26, 0.38) | 0.70 |
| Protein C antigen, mg/L | 0.7 | 375 | 0.12 | (−0.19, 0.43) | 0.46 | 332 | 0.22 | (−0.093, 0.54) | 0.17 |
| Free Protein S antigen, mg/L | 0.9 | 339 | 0.29 | (−0.043, 0.62) | 0.088 | 299 | 0.26 | (−0.075, 0.59) | 0.13 |
| Total Protein S antigen, mg/L | 3.2 | 341 | 0.091 | (−0.23, 0.42) | 0.58 | 301 | 0.084 | (−0.24, 0.41) | 0.62 |
| TFPI, U/ml | 23.4 | 246 | −0.36 | (−0.73, 0.012) | 0.058 | 215 | −0.55 | (−0.90, −0.19) | 0.003 |
CI = confidence interval; FVII antigen = factor VII antigen; FVIIc = Factor VII; FVIIIc = factor VIII; FIXc = factor IX; FXc = factor X; F1.2 = prothrombin fragment 1.2; PAI-1 = plasminogen activator inhibitor-1; PAP = Plasmin-alpha2-antiplasmin complex; SD = standard deviation; tPA = tissue plasminogen activator antigen; TFPI = tissue factor pathway inhibitor activity.
Model 1 adjusted for age, sex, education, race, and field center
Model 2 adjusted for model 1 covariates + BMI, physical activity, alcohol intake, current and former smoking, hormone therapy use (in women only), prevalent diabetes, systolic blood pressure, hypertensive medication use, total cholesterol, APOE genotype.
Beta coefficients estimate the association between a 1-SD difference in each hemostatic factor and annual rate of change in mean 3MSE measures over follow-up.
In Table 3, we report the annual rate of change in mean DSST scores over a 9-year period of follow-up, as associated with a 1-SD difference in hemostatic factor levels measured in 1989/1990 baseline. In models 1 and 2, there was no evidence of a linear association between levels of any hemostatic factors and annual rate of change in mean DSST scores (all p-values>0.05). Evaluation of a quadratic association also revealed no significant associations after adjusting for multiple comparisons (all p≥0.01) (Supplemental Table 2).
Table 3.
Annual rate of change in mean DSST scores over a 9-year period (1989/1990–1998/1999) associated with 1-SD differences in hemostatic parameter levels in 1989/1990.
| Linear Exposure Model 1* | Linear Exposure Model 2† | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| SD | n | Beta‡ | 95% CI | p value | n | Beta‡ | 95% CI | p value | |
| Age (1-year difference) | 7.2 | 359 | −0.068 | (−0.095, −0.041) | <0.001 | 316 | −0.066 | (−0.094, −0.039) | <0.001 |
| Procoagulant Hemostatic Factors | |||||||||
| D-dimer, ng/mL | 420.0 | 354 | −0.063 | (−0.32, 0.19) | 0.63 | 312 | −0.110 | (−0.39, 0.17) | 0.43 |
| FVII antigen, % normal | 132.0 | 358 | −0.21 | (−0.69, 0.26) | 0.38 | 315 | −0.19 | (−0.72, 0.34) | 0.48 |
| FVIIc, % normal | 27.7 | 359 | −0.047 | (−0.23, 0.14) | 0.62 | 316 | −0.012 | (−0.21, 0.19) | 0.91 |
| FVIIIc, % normal | 36.5 | 356 | −0.017 | (−0.21, 0.17) | 0.86 | 313 | −0.041 | (−0.25, 0.16) | 0.69 |
| FIXc, % normal | 23.1 | 359 | −0.0072 | (−0.20, 0.18) | 0.94 | 316 | −0.022 | (−0.22, 0.17) | 0.83 |
| FXc, % normal | 20.7 | 359 | 0.073 | (−0.12, 0.26) | 0.45 | 316 | 0.081 | (−0.12, 0.28) | 0.43 |
| Fibrinogen, mg/dL | 70.5 | 357 | −0.039 | (−0.23, 0.16) | 0.70 | 314 | −0.058 | (−0.26, 0.15) | 0.58 |
| Fibrinopeptide A, ng/mL | 38.6 | 354 | −0.000052 | (−0.27, 0.27) | 1.00 | 312 | −0.22 | (−0.56, 0.13) | 0.23 |
| Platelet count, x109 L−1 | 70.7 | 353 | −0.13 | (−0.32, 0.050) | 0.15 | 311 | −0.10 | (−0.29, 0.10) | 0.35 |
| F1.2, nmol/L | 0.39 | 358 | 0.077 | (−0.26, 0.42) | 0.65 | 316 | −0.086 | (−0.45, 0.27) | 0.64 |
| Fibrinolytic Hemostatic Factors | |||||||||
| PAI-1, ng/mL | 30.1 | 359 | 0.064 | (−0.14, 0.27) | 0.54 | 316 | 0.076 | (−0.13, 0.28) | 0.47 |
| PAP, nmol/L | 3.1 | 358 | −0.20 | (−0.47, 0.067) | 0.14 | 315 | −0.23 | (−0.51, 0.047) | 0.10 |
| Plasminogen, % normal | 16.8 | 359 | 0.031 | (−0.17, 0.23) | 0.76 | 316 | 0.038 | (−0.17, 0.24) | 0.72 |
| tPA, ng/ml | 6.7 | 357 | 0.070 | (−0.19, 0.33) | 0.60 | 314 | 0.12 | (−0.14, 0.38) | 0.36 |
| tPA-PAI-1 complex, nmol/L | 1.6 | 357 | 0.15 | (−0.13, 0.43) | 0.28 | 314 | 0.23 | (−0.064, 0.51) | 0.13 |
| Anticoagulant Hemostatic Factors | |||||||||
| Antithrombin activity, % normal | 15.1 | 359 | 0.029 | (−0.16, 0.22) | 0.77 | 316 | 0.023 | (−0.17, 0.22) | 0.81 |
| Protein C antigen, mg/L | 0.7 | 358 | 0.027 | (−0.15, 0.21) | 0.77 | 315 | 0.051 | (−0.13, 0.24) | 0.59 |
| Free Protein S antigen, mg/L | 0.9 | 322 | 0.0085 | (−0.18, 0.20) | 0.93 | 282 | −0.046 | (−0.24, 0.15) | 0.65 |
| Total Protein S antigen, mg/L | 3.2 | 324 | 0.1200 | (−0.059, 0.30) | 0.19 | 284 | 0.180 | (−0.0032, 0.37) | 0.054 |
| TFPI, U/ml | 23.4 | 235 | −0.086 | (−0.30, 0.13) | 0.44 | 204 | −0.13 | (−0.35, 0.094) | 0.26 |
CI = confidence interval; FVII antigen = factor VII antigen; FVIIc = Factor VII; FVIIIc = factor VIII; FIXc = factor IX; FXc = factor X; F1.2 = prothrombin fragment 1.2; PAI-1 = plasminogen activator inhibitor-1; PAP = Plasmin-alpha2-antiplasmin complex; SD = standard deviation; tPA = tissue plasminogen activator antigen; TFPI = tissue factor pathway inhibitor activity.
Model 1 adjusted for age, sex, education, race, and field center
Model 2 adjusted for model 1 covariates + BMI, physical activity, alcohol intake, current and former smoking, hormone therapy use (in women only), prevalent diabetes, systolic blood pressure, hypertensive medication use, total cholesterol, APOE genotype.
Beta coefficients estimate the association between a 1-SD difference in each hemostatic factor and annual rate of change in mean DSST measures over follow-up.
Results of our sensitivity analysis in which individuals with baseline 3MSE scores in the bottom 10th percentile were excluded were similar to those from primary analyses. Point estimates for FXc, PAP, and TFPI were similar to those in primary analyses; FXc and TFPI were no longer statistically significant after our Bonferroni correction (FXc β =0.33, 95%CI: 0.020, 0.64, p=0.037; PAP β =−0.56, 95%CI: −0.98, −0.14, p=0.009; TFPI β =−0.40, 95%CI: −0.74, −0.066, p=0.019). No hemostatic factors became significant that had not been significant in primary analyses. In sensitivity analyses that did not use TICS and IQCODE data to estimate missing 3MSE values, results were also not meaningfully different.
In PCA, three hemostatic factors were missing values for >5% of eligible participants and were excluded (free protein S antigen [9.8% missing], total protein S antigen [9.3% missing], and TFPI [34.6% missing]). Four components explained 60% of the total variance and were further tested in a mixed-effects linear regression model that included all four components. The third component, which explained 11.7% of the total variance, was significantly associated with annual rate of change of 3MSE (β=−0.49; 95% CI: −0.74, −0.25; p<0.001); no components were significantly associated with the annual rate of change of DSST. Loadings in component 3 were largest in size for the procoagulant factors D-dimer (0.43) and F1.2 (0.33) and for fibrinolytic factor, PAP (0.55) (Supplemental Table 3). In a sensitivity analysis that included component 5, which had an eigenvalue >1.0, component 5 was not significant at an alpha level of 0.05, and results were similar to those of our primary analysis.
In secondary analyses evaluating interaction by APOE-e4 carrier status, there was no evidence of an interaction with any measured hemostatic factor in relation to annual rate of change in mean 3MSE (all p-interaction≥0.12) or mean DSST (all p-interaction >0.017) after adjusting for multiple comparisons.
Discussion
In this prospective cohort study of older adults, higher PAP and TFPI levels, and lower FXc levels were positively associated with faster cognitive decline as estimated by change in 3MSE scores. The association with PAP was in the direction as hypothesized a priori, but associations with TFPI and FXc were in directions opposite from what we hypothesized. To contextualize the findings from our mixed-effects linear regression models, each additional year of age at baseline was associated with a 0.17-point faster decline in 3MSE scores. Thus, 1-SD higher levels of PAP were similar to 3.8 additional years of age at baseline, on average; 1-SD higher levels of TFPI were similar to 3.2 additional years of age at baseline; and, 1-SD lower levels of FXc were similar to 3.1 additional years of age at baseline, on average. While there were no associations of a wide range of other individual hemostatic factors with cognitive decline, as measured by the 3MSE and DSST, principal component analysis suggested an association with one component that loaded positively on D-dimer, F1.2, and PAP, potentially representing procoagulation and fibrinolysis.
Hemostatic Factor Levels and Cognitive Decline
Prior studies have evaluated some hemostatic factor levels in relation to cognitive impairment, but most factors included in our study have been less commonly evaluated. The lack of prior studies that evaluated many of the hemostatic factors included in this study, coupled with other methodologic differences between studies including age of study participants, measurement of cognitive decline, and duration of follow-up, make a direct comparison of results difficult. The Caerphilly Study reported that higher levels of fibrinogen (HR=1.4; 95% CI: 1.0, 2.0) and PAI-1 (HR=1.8; 95% CI: 1.1, 2.8) were positively associated with risk of vascular but not non-vascular cognitive impairment[5], and that higher levels of FVII were negatively associated with the risk of non-vascular (HR=0.7; 95% CI: 0.6, 1.0) but not vascular cognitive impairment. They reported no association with impairment for FVIII, von Willebrand factor (vWF) antigen, vWF activity, activated factor XII, activated partial thromboplastin time, activated protein C ratio, reaction clotting time, fibrin clotting time, F1.2, thrombin-antithrombin complex, tPA, or D-dimer[5]. In contrast to the Caerphilly Study, in CHS we did not find evidence of an association between fibrinogen, PAI-1, or FVIIc with cognitive decline, and in PCA analyses, we did find evidence that one component that loaded positively on D-dimer, F1.2, and PAP may be associated with cognitive decline. However, these differences may be due the dissimilarities in our two studies’ measurement of cognitive decline and impairment.
In a recent case-control study of men and women ≥45 years of age, nested in the Reasons for Geographic and Racial Differences in Stroke (REGARDS) study, quartiles of baseline D-dimer, fibrinogen, FVIII, and protein C levels were not significantly associated with incident cognitive impairment as measured by three-test cognitive battery administered longitudinally by telephone.[6] In REGARDS, authors reported that presence of at least two elevated biomarkers was associated with incident cognitive impairment (ORadj=1.7; 95% CI: 1.1–2.7).[6] In our study, when modeled separately, we also did not find evidence of an association of D-dimer, fibrinogen, FVIIIc, or protein C antigen levels with cognitive decline. Results may not be comparable between REGARDS and CHS, however, due to a younger population in REGARDS, a shorter period of follow-up, and differences in the assessment of cognitive function.
A 2011 systematic review and meta-analysis identified twenty-one studies that evaluated hemostatic factor levels in relation to dementia risk or cognitive impairment[4], with a total of 17 hemostatic factors evaluated across these studies. Individual factors that were associated with cognitive decline in our study (PAP, TFPI, and FXc) were not included in that analysis and we are not aware that the included studies have since reported on these.
Biologic Plausibility
Despite the general lack of association of individual hemostatic factors with cognitive decline observed in our study, circulating levels of these factors are associated with the risk of both arterial and venous cardiovascular outcomes, including stroke and venous thromboembolism[3, 27]. Given the role of stroke in vascular cognitive decline and vascular dementia, the relatively well-characterized association between markers of fibrinolysis and coagulation with stroke risk increases plausibility for a role by fibrinolysis and coagulation in cognitive decline and dementia risk, by way of vascular changes. Previously, structural equation models used by the Caerphilly Study tested the hypothesis that specific coagulation pathways would be associated with the risk of vascular dementia. These investigators identified associations with latent variables representing “FVIII/vWF complex”, implicating the formation of the platelet and fibrin plug, and of “clot formation and lysis”, implicating the formation and lysis of the fibrin plug.[5] Somewhat similarly, PCA in the present study highlighted one combination of procoagulant and fibrinolytic factors in relation to change in 3MSE and DSST scores, although the specific factors (D-dimer, F1.2, and PAP) identified as important contributors to the associated component differed between this study and the Caerphilly Study. In our case, this component reflected higher levels of D-dimer and PAP, both of which are byproducts of activation of both the clotting and fibrinolytic systems, suggesting a potential role specifically for this relatively late-stage aspect of thrombosis in cognitive deterioration in older adults.
In our study, when modeled separately, higher PAP and TFPI levels and lower FXc levels were positively associated with faster cognitive decline, with associations for the procoagulant FXc and the anticoagulant TFPI being in the directions opposite from hypothesized. To our knowledge, prior studies have not evaluated TFPI and FXc in relation to cognitive decline, and we can only speculate as to the mechanism of these associations. When bound together, TFPI and FXa inhibit the procoagulant tissue factor-factor VIIa complex, which triggers blood coagulation.[28, 29] The inhibition of tissue factor-factor VIIa by TFPI results in the formation of the quaternary complex, TF-FVIIa-FXa-TFPI.[30, 31] It is possible that lower levels of FXc, a factor that is required for TFPI to inhibit tissue factor-FVIIa complex[29], is associated with greater amounts of blood coagulation and subsequent cognitive decline by way of reduced inhibition of tissue factor-factor FVIIa by TFPI.
Limitations and Strengths
This study has several limitations. Hemostatic factors were measured only once, and so they do not reflect within-person variation in levels or changes in levels over time. Given that hemostatic factors were measured only once, unmeasured changes in hemostatic factors during follow-up may have biased estimates towards the null.[32] Some individuals were missing 3MSE and DSST scores; however, missing 3MSE scores were estimated from the TICS and IQCODE[12], when available, which reduced this missingness. In a sensitivity analysis in which TICS and IQCODE data was not used to estimate missing 3MSE scores, estimates and conclusions were not meaningfully different, suggesting that this estimation did not alter our findings. At baseline in our study, individuals had varying levels of cognitive impairment; however, results were not meaningfully different in a sensitivity analysis in which we excluded the individuals with the most severe levels of cognitive impairment at baseline. As an observational study, CHS did not test any specific anti-thrombotic or anticoagulant interventions and thus further study is needed to determine if these findings can be used to identify or target new approaches to dementia prevention. As strengths, this study is population-based and prospective, and included a long follow-up and repeated measures of cognitive function over the study’s period of follow-up. To our knowledge, our study also includes the largest series of hemostatic factors tested in relation to cognitive decline in a single study to date, providing a comprehensive overview of these biomarkers; given the number of hemostatic factors evaluated, we corrected for multiple testing.
Conclusions
In conclusion, in this prospective cohort study of older adults, a combination of factors driven predominantly by PAP, D-dimer, and F1.2 is associated with decline in 3MSE; levels of PAP, TFPI, and FXc may be individually associated with cognitive decline based on the 3MSE. To fully understand the role of hemostasis in relation to changing cognition, a further understanding of clotting and fibrinolytic systems rather than individual hemostatic factors in relation to cognitive decline may be needed.
Supplementary Material
Essentials.
It is uncertain whether hemostasis markers are associated with cognitive decline.
Among 400 older adults, 20 measures of hemostasis were measured in samples from 1989/1990.
Levels of PAP, TFPI, and FXc were associated with cognitive decline by change in 3MSE scores.
Principal component analysis identified a component depicting procoagulation and fibrinolysis.
Acknowledgements
This research was supported by contracts HHSN268201200036C, HHSN268200800007C, HHSN268201800001C, N01HC55222, N01HC85079, N01HC85080, N01HC85081, N01HC85082, N01HC85083, N01HC85086, and grants U01HL080295 and U01HL130114 from the National Heart, Lung, and Blood Institute (NHLBI), with additional contribution from the National Institute of Neurological Disorders and Stroke (NINDS). Additional support was provided by R01AG023629, R01AG15928, and R01AG20098 from the National Institute on Aging (NIA). A full list of principal CHS investigators and institutions can be found at CHS-NHLBI.org. This study was further supported by NINDS grant R01NS089638 and NHLBI grants R01HL046696, T32HL098048, and K01HL139997. This content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
Footnotes
Conflicts of Interest
All authors (L.B. Harrington, A.N. Ehlert, E.L. Thacker, N.S. Jenny, O. Lopez, M. Cushman, A. Fitzpatrick, K.J. Mukamal, and M.K. Jensen) report having no conflicts of interest.
Bibliography
- 1.Strickland S Blood will out: vascular contributions to Alzheimer’s disease. J Clin Invest. 2018;128:556–63. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Zakai NA, Katz R, Jenny NS, Psaty BM, Reiner AP, Schwartz SM, Cushman M. Inflammation and hemostasis biomarkers and cardiovascular risk in the elderly: the Cardiovascular Health Study. J Thromb Haemost. 2007;5:1128–35. [DOI] [PubMed] [Google Scholar]
- 3.Cushman M, Yanez D, Psaty BM, Fried LP, Heiss G, Lee M, Polak JF, Savage PJ, Tracy RP. Association of fibrinogen and coagulation factors VII and VIII with cardiovascular risk factors in the elderly: the Cardiovascular Health Study. Cardiovascular Health Study Investigators. Am J Epidemiol. 1996;143:665–76. [DOI] [PubMed] [Google Scholar]
- 4.Quinn TJ, Gallacher J, Deary IJ, Lowe GD, Fenton C, Stott DJ. Association between circulating hemostatic measures and dementia or cognitive impairment: systematic review and meta-analyzes. J Thromb Haemost. 2011;9:1475–82. [DOI] [PubMed] [Google Scholar]
- 5.Gallacher J, Bayer A, Lowe G, Fish M, Pickering J, Pedro S, Dunstan F, White J, Yarnell J, Ben-Shlomo Y. Is sticky blood bad for the brain?: Hemostatic and inflammatory systems and dementia in the Caerphilly Prospective Study. Arterioscler Thromb Vasc Biol. 2010;30:599–604. [DOI] [PubMed] [Google Scholar]
- 6.Gillett SR, McClure LA, Callas PW, Thacker EL, Unverzagt FW, Wadley VG, Letter AJ, Cushman M. Hemostasis biomarkers and incident cognitive impairment: the REGARDS study. J Thromb Haemost. 2018. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Fried LP, Borhani NO, Enright P, Furberg CD, Gardin JM, Kronmal RA, Kuller LH, Manolio TA, Mittelmark MB, Newman A, et al. The Cardiovascular Health Study: design and rationale. Ann Epidemiol. 1991;1:263–76. [DOI] [PubMed] [Google Scholar]
- 8.Sakkinen PA, Cushman M, Psaty BM, Kuller LH, Bajaj SP, Sabharwal AK, Boineau R, Macy E, Tracy RP. Correlates of antithrombin, protein C, protein S, and TFPI in a healthy elderly cohort. Thromb Haemost. 1998;80:134–9. [PubMed] [Google Scholar]
- 9.Teng EL, Chui HC. The Modified Mini-Mental State (3MS) examination. The Journal of clinical psychiatry. 1987;48:314–8. [PubMed] [Google Scholar]
- 10.Weschsler D Adult Intelligence Scale-Revised Manual. Corporation P, editor. New York, NY, 1981. [Google Scholar]
- 11.Lopez OL, Jagust WJ, DeKosky ST, Becker JT, Fitzpatrick A, Dulberg C, Breitner J, Lyketsos C, Jones B, Kawas C, Carlson M, Kuller LH. Prevalence and classification of mild cognitive impairment in the Cardiovascular Health Study Cognition Study: part 1. Arch Neurol. 2003;60:1385–9. [DOI] [PubMed] [Google Scholar]
- 12.Arnold AM, Newman AB, Dermond N, Haan M, Fitzpatrick A. Using telephone and informant assessments to estimate missing Modified Mini-Mental State Exam scores and rates of cognitive decline. The cardiovascular health study. Neuroepidemiology. 2009;33:55–65. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Haan MN, Shemanski L, Jagust WJ, Manolio TA, Kuller L. The role of APOE epsilon4 in modulating effects of other risk factors for cognitive decline in elderly persons. Jama. 1999;282:40–6. [DOI] [PubMed] [Google Scholar]
- 14.Lawler CM, Bovill EG, Stump DC, Collen DJ, Mann KG, Tracy RP. Fibrin fragment D-dimer and fibrinogen B beta peptides in plasma as markers of clot lysis during thrombolytic therapy in acute myocardial infarction. Blood. 1990;76:1341–8. [PubMed] [Google Scholar]
- 15.Cushman M, Lemaitre RN, Kuller LH, Psaty BM, Macy EM, Sharrett AR, Tracy RP. Fibrinolytic activation markers predict myocardial infarction in the elderly. The Cardiovascular Health Study. Arterioscler Thromb Vasc Biol. 1999;19:493–8. [DOI] [PubMed] [Google Scholar]
- 16.Howard PR, Bovill EG, Pike J, Church WR, Tracy RP. Factor VII antigen levels in a healthy blood donor population. Thromb Haemost. 1994;72:21–7. [PubMed] [Google Scholar]
- 17.Sakkinen PA, Wahl P, Cushman M, Lewis MR, Tracy RP. Clustering of procoagulation, inflammation, and fibrinolysis variables with metabolic factors in insulin resistance syndrome. Am J Epidemiol. 2000;152:897–907. [DOI] [PubMed] [Google Scholar]
- 18.Sakkinen PA, Cushman M, Psaty BM, Rodriguez B, Boineau R, Kuller LH, Tracy RP. Relationship of plasmin generation to cardiovascular disease risk factors in elderly men and women. Arterioscler Thromb Vasc Biol. 1999;19:499–504. [DOI] [PubMed] [Google Scholar]
- 19.Sakkinen PA, Macy EM, Callas PW, Cornell ES, Hayes TE, Kuller LH, Tracy RP. Analytical and biologic variability in measures of hemostasis, fibrinolysis, and inflammation: assessment and implications for epidemiology. Am J Epidemiol. 1999;149:261–7. [DOI] [PubMed] [Google Scholar]
- 20.Cushman M, Psaty BM, Macy E, Bovill EG, Cornell ES, Kuller LH, Tracy RP. Correlates of thrombin markers in an elderly cohort free of clinical cardiovascular disease. Arterioscler Thromb Vasc Biol. 1996;16:1163–9. [DOI] [PubMed] [Google Scholar]
- 21.Howard PR, Bovill EG, Mann KG, Tracy RP. A monoclonal-antibody-based radioimmunoassay for measurement of protein C in plasma. Clin Chem. 1988;34:324–30. [PubMed] [Google Scholar]
- 22.Bovill EG, Landesman MM, Busch SA, Fregeau GR, Mann KG, Tracy RP. Studies on the measurement of protein S in plasma. Clin Chem. 1991;37:1708–14. [PubMed] [Google Scholar]
- 23.van der Bom JG, Heckbert SR, Lumley T, Holmes CE, Cushman M, Folsom AR, Rosendaal FR, Psaty BM. Platelet count and the risk for thrombosis and death in the elderly. J Thromb Haemost. 2009;7:399–405. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Sink KM, Craft S, Smith SC, Maldjian JA, Bowden DW, Xu J, Freedman BI, Divers J. Montreal Cognitive Assessment and Modified Mini Mental State Examination in African Americans. Journal of aging research. 2015;2015:872018. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Hixson JE, Vernier DT. Restriction isotyping of human apolipoprotein E by gene amplification and cleavage with HhaI. Journal of lipid research. 1990;31:545–8. [PubMed] [Google Scholar]
- 26.Jolliffe IT, Cadima J. Principal component analysis: a review and recent developments. Philosophical transactions Series A, Mathematical, physical, and engineering sciences. 2016;374:20150202. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Zakai NA, McClure LA, Judd SE, Safford MM, Folsom AR, Lutsey PL, Cushman M. Racial and regional differences in venous thromboembolism in the United States in 3 cohorts. Circulation. 2014;129:1502–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Sanders NL, Bajaj SP, Zivelin A, Rapaport SI. Inhibition of tissue factor/factor VIIa activity in plasma requires factor X and an additional plasma component. Blood. 1985;66:204–12. [PubMed] [Google Scholar]
- 29.Hubbard AR, Jennings CA. Inhibition of the tissue factor-factor VII complex: involvement of factor Xa and lipoproteins. Thromb Res. 1987;46:527–37. [DOI] [PubMed] [Google Scholar]
- 30.Girard TJ, Warren LA, Novotny WF, Likert KM, Brown SG, Miletich JP, Broze GJ, Jr. Functional significance of the Kunitz-type inhibitory domains of lipoprotein-associated coagulation inhibitor. Nature. 1989;338:518–20. [DOI] [PubMed] [Google Scholar]
- 31.Carlsson K, Freskgård PO, Persson E, Carlsson U, Svensson M. Probing the interface between factor Xa and tissue factor in the quaternary complex tissue factor-factor VIIa-factor Xa-tissue factor pathway inhibitor. European journal of biochemistry. 2003;270:2576–82. [DOI] [PubMed] [Google Scholar]
- 32.Hutcheon JA, Chiolero A, Hanley JA. Random measurement error and regression dilution bias. Bmj. 2010;340:c2289. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.