If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Specific Nutritional Biomarker Profiles in Mild Cognitive Impairment and Subjective Cognitive Decline Are Associated With Clinical Progression: The NUDAD Project
Address correspondence to Francisca A. de Leeuw, MD, Department of Neurology, Alzheimer Center Amsterdam, Amsterdam UMC, PO Box 7057, 1007 MB Amsterdam, the Netherlands.
Department of Neurology, Alzheimer Center Amsterdam, Amsterdam Neuroscience, Vrije Universiteit Amsterdam, Amsterdam UMC, Amsterdam, the NetherlandsDepartment of Epidemiology and Biostatistics, Amsterdam Public Health Research, Vrije Universiteit Amsterdam, Amsterdam UMC, Amsterdam, the Netherlands
Neurochemistry Laboratory and Biobank, Department of Clinical Chemistry, Amsterdam Neuroscience, VU University Medical Center Amsterdam, Amsterdam, the Netherlands
Nutritional insufficiencies have been associated with cognitive impairment. Understanding whether nutritional biomarker levels are associated with clinical progression could help to design dietary intervention trials. This longitudinal study examined a panel of nutritional biomarkers in relation to clinical progression in patients with subjective cognitive decline (SCD) or mild cognitive impairment (MCI).
Design, setting and participants
We included 299 patients without dementia (n = 149 SCD; age 61 ± 10 years, female 44%, n = 150 MCI; age 66 ± 8 years, female 38%). Median (interquartile range) follow-up was 3 (2-5) years.
Methods
We measured 28 nutritional biomarkers in blood and 5 in cerebrospinal fluid (CSF), associated with 3 Alzheimer's disease pathologic processes: vascular change (lipids), synaptic dysfunction (homocysteine-related metabolites), and oxidative stress (minerals and vitamins). Nutritional biomarker associations with clinical progression to MCI/dementia and cognitive decline based on the Mini-Mental State Examination score were evaluated using Cox proportional hazard models and linear mixed models. We used partial least squares Cox models (PLS-Cox) to examine nutritional biomarker profiles associated with clinical progression.
Results
In the total group, high high-density lipoprotein (HDL) levels were associated with clinical progression and cognitive decline. In SCD, high folate and low bilirubin levels were associated with cognitive decline. In MCI, low CSF S-adenosylmethionine (SAM) and high theobromine were associated with clinical progression to dementia and high HDL, cholesterol, iron, and 1,25(OH)2 vitamin D were associated with cognitive decline. PLS-Cox showed 1 profile for SCD, characterized by high betaine and folate and low zinc associated with clinical progression. In MCI, a profile with high theobromine and HDL and low triglycerides and a second profile with high plasma SAM and low cholesterol were associated with risk of dementia.
Conclusion and Implications
High HDL was most consistently associated with clinical progression. Moreover, different nutritional biomarker profiles for SCD and MCI showed promising associations with clinical progression. Future dietary (intervention) studies could use nutritional biomarker profiles to select patients, taking into account the disease stage.
Rising dementia numbers worldwide are a global health concern. Previous studies have identified lifestyle-related, modifiable risk factors for dementia, for example, hypertension, diabetes, and obesity.
These risk factors are closely related to dietary intake and cause peripheral metabolic changes. Nutrition also plays an important role in patients with dementia; for example, involuntary weight loss and nutritional biomarker insufficiencies have been reported.
These nutritional biomarker changes may already occur in predementia stages, as lower levels of some nutritional biomarkers have also been observed in mild cognitive impairment (MCI).
Nutritional biomarkers required for phospholipid synthesis are lower in blood and cerebrospinal fluid in mild cognitive impairment and Alzheimer's disease dementia.
As such, dietary interventions could be a relatively safe and cheap approach to prevent cognitive decline. Trials have tested the possible preventive effect of nutritional interventions. For example, the LipiDiDiet trial investigated the effect of a multinutrient intervention (Fortasyn Connect) in prodromal AD patients. Beneficial effects were found for functional performance (clinical dementia rating—sum of boxes) and hippocampal atrophy, whereas effects for cognition (the primary endpoint) were suggested.
24-month intervention with a specific multinutritional biomarker in people with prodromal Alzheimer's disease (LipiDiDiet): A randomised, double-blind, controlled trial.
The FINGER trial tested a multidomain intervention (including dietary advice) in an elderly population at risk for dementia and found beneficial effects on cognition.
A 2 year multidomain intervention of diet, exercise, cognitive training, and vascular risk monitoring versus control to prevent cognitive decline in at-risk elderly people (FINGER): A randomised controlled trial.
Dietary changes and cognition over 2 years within a multidomain intervention trial-The Finnish Geriatric Intervention Study to Prevent Cognitive Impairment and Disability (FINGER).
These results seemingly show that interventions might be beneficial in populations at increased risk. The specific blood or cerebrospinal fluid (CSF) nutritional biomarker profiles that relate to clinical progression, however, remain unclear. We hypothesize that nutritional biomarkers involved in AD-related pathologic processes are of special interest; that is, lipids and glucose have been related to vascular change, homocysteine-related nutritional biomarkers to synaptic dysfunction, and minerals and vitamins to antioxidant function.
Moreover, multiple interrelated metabolic pathways might be involved, and therefore nutritional biomarkers should not only be studied as single nutritional biomarker shifts but preferably as nutritional biomarker profiles.
In this study, we aimed to examine the association between a panel of nutritional biomarkers measured in the blood and CSF and clinical progression in memory-clinic patients with subjective cognitive decline (SCD) or MCI. First, we assessed associations of individual nutritional biomarkers with clinical progression to MCI or dementia and with rate of decline on the Mini-Mental State Examination (MMSE). Next, as nutritional biomarkers are likely to be highly interdependent, we identified nutritional profiles of multiple nutritional biomarkers associated with clinical progression.
Methods
Patient Cohort
From the Amsterdam Dementia Cohort, we retrospectively included patients with a baseline diagnosis of SCD (n = 149) or MCI (n = 150), with sufficient blood/CSF volumes and ≥1 year of clinical follow-up.
To increase statistical power, we oversampled patients with clinical progression to MCI or dementia during follow-up. All patients underwent standardized cognitive screening (between 2001 and 2015), including neurologic and cognitive examination, blood sampling, a lumbar puncture, and magnetic resonance imaging. Diagnoses were made in a multidisciplinary consensus meeting. Subjects with SCD presented with memory complaints but appeared normal on clinical examinations; that is, criteria for MCI, dementia, or psychiatric diagnosis were not fulfilled. The diagnosis of MCI was based on Petersen's criteria until 2012 and the National Institute on Aging-Alzheimer's Association criteria from 2012 onwards.
The diagnosis of mild cognitive impairment due to Alzheimer's disease: recommendations from the National Institute on Aging-Alzheimer's Association workgroups on diagnostic guidelines for Alzheimer's disease.
Diabetes mellitus, hypertension, hypercholesterolemia, and myocardial infarction were represented as self-reported medication use or medical history for these conditions at baseline. Self-reported use of lipid-lowering agents was also assessed at baseline. Cardiovascular disease was defined as having hypertension, hypercholesterolemia, diabetes mellitus, or myocardial infarction. All participants gave written informed consent to use their clinical data for research purposes and to collect blood and CSF for biobanking.
Follow-up
At annual follow-up visits, neuropsychological testing and medical examination was repeated. Median follow-up (interquartile range) was 3 (2-5) years. Main outcome was clinical progression, defined as a follow-up syndrome diagnosis of MCI (for SCD) or dementia based on commonly used criteria.
The diagnosis of mild cognitive impairment due to Alzheimer's disease: recommendations from the National Institute on Aging-Alzheimer's Association workgroups on diagnostic guidelines for Alzheimer's disease.
Clinical diagnosis of Alzheimer's disease: Report of the NINCDS-ADRDA Work Group under the auspices of Department of Health and Human Services Task Force on Alzheimer's Disease.
The diagnosis of dementia due to Alzheimer's disease: Recommendations from the National Institute on Aging-Alzheimer's Association workgroups on diagnostic guidelines for Alzheimer's disease.
Time to clinical progression was defined as time between biofluid sampling and change of diagnosis to MCI or dementia (any type). In addition, all available MMSE measurements were used to estimate annual cognitive decline over time. Median (interquartile range) MMSE measurements per participant was 2 (1-4). In total, 1108 MMSE scores were available.
Blood and CSF Collection
Blood and CSF sample were obtained between 2001 and 2015 and within 6 months of baseline diagnosis. In 6 SCD cases, blood and CSF samples were obtained >6 months from baseline diagnosis; these cases, however, had a stable SCD diagnosis before and after biofluid sampling. Nonfasting serum and EDTA plasma samples were collected in 6-mL tubes. CSF was obtained by lumbar puncture using a 25-gauge needle and collected in 10-mL polypropylene tubes (Sarstedt, Nümbrecht, Germany). Within 2 hours, blood and CSF was centrifuged at 1800×g for 10 minutes at 4°C (before 2014) or at room temperature (since 2014), aliquoted in vials of 0.5 mL and stored at −80°C until further analysis.
Biomarkers for AD
Amyloid β peptide 1-42 (Aβ42), total tau (t-tau), and phosphorylated tau (p-tau) were measured in CSF on a routine basis using commercially available enzyme-linked immunosorbent assays (Innotest β-amyloid(1-42), Innotest hTAU-Ag, and Innotest phosphotau (181P); Innogenetics, Ghent, Belgium) as described.
We composed a panel of 33 nutritional biomarkers (Table 1) that can be associated with 3 pathologic mechanisms of AD—vascular change, synaptic dysfunction, and oxidative stress.
All 33 nutritional biomarkers were measured in serum or plasma, except for theobromine, which was measured only in CSF. Homocysteine, uridine, S-adenosylmethionine (SAM), and S-adenosylhomocysteine (SAH) are associated with synaptic dysfunction and were measured both in plasma and CSF to assess systemic and central availability.
Nutritional biomarkers required for phospholipid synthesis are lower in blood and cerebrospinal fluid in mild cognitive impairment and Alzheimer's disease dementia.
For most nutritional biomarkers, missing values were <5%. Causes for missing data were lack of volume, measurements exceeding the lower or upper limit of detection, or disturbed chromatogram. Docosahexaenoic acid and total omega-3 fatty acids were available in a subset of the population (n = 185, 62%). Glucose and vitamin B12 measurements were included in the diagnostic workup for most participants (n = 198, 66%, and n = 224, 75%, respectively).
Table 1Baseline Characteristics According to Baseline and Follow-up Diagnosis
P < .05. P values correspond to χ2 tests, t tests, and Mann-Whitney U tests when appropriate.
APOE, apolipoprotein E; Aβ42, amyloid β peptide 1-42; BMI, body mass index; IQR, interquartile range; SD, standard deviation.
Unless otherwise noted, values are n (%). Groups were compared on their outcome (stable vs progression) in the total cohort and within subgroups of their baseline syndrome diagnosis (SCD/MCI).
∗ P < .05. P values correspond to χ2 tests, t tests, and Mann-Whitney U tests when appropriate.
† 61 participants had missing (20%) BMI values.
‡ 30 participants had missing (10%) alcohol intake values; all other baseline characteristics have <5% missing values.
Nutritional biomarkers required for phospholipid synthesis are lower in blood and cerebrospinal fluid in mild cognitive impairment and Alzheimer's disease dementia.
for the remaining patients, these nutritional biomarkers were measured in 2017. Previous studies showed albumin and bilirubin to have low interassay coefficients of variation, allowing direct comparison.
For homocysteine, choline, betaine, and folate, 3 to 10 samples of the old batch (2014) were reanalyzed in the new batch (2017) to assess variability between old and new measurements. Variation in concentrations of all nutritional biomarkers, except for homocysteine in plasma, did not differ from normal interassay differences, allowing direct comparison. (See Supplementary Text 1 for interassay differences.) Reanalyzed homocysteine measurements in plasma were lower (mean 11%) in the new batch. All samples measured in the old batch were therefore normalized to the new values for homocysteine plasma. Supplementary Text and Supplementary Table 1 contain detailed information on measurement methods and reference values of the nutritional biomarkers.
Statistical Analysis
Nutritional biomarkers, Aβ42, t-tau, and p-tau levels were log-transformed when not normally distributed. Subsequently, all nutritional biomarkers, Aβ42, t-tau, and p-tau were converted to z-scores to facilitate parametric comparison of effect sizes. Groups were compared on their outcome (stable vs clinical progression) in the total cohort and stratified for syndrome diagnosis (SCD/MCI).
Clinical characteristics were compared using χ2 tests, t tests, and Mann-Whitney U tests when appropriate. We present an unadjusted model (model 1) and a model adjusted for sex, age, diagnosis (total cohort models), lipid-lowering medication, and cardiovascular disease (model 2). Nutritional biomarker levels were compared using linear regression analysis. Cox proportional hazard models were used to investigate if nutritional biomarkers (continuous determinants) were associated with time to clinical progression (outcome). Hazard ratios (HRs) are presented with a 95% confidence interval (CI). Pearson correlations assessed correlations between biomarkers. To identify profiles of the 33 nutritional biomarkers associated with clinical progression over time (outcome), we used PLS-Cox, stratified for syndrome diagnosis.
PLS-Cox is especially useful when investigating many interrelated predictors simultaneously, as it reduces dimensionality in the data by combining predictors that show a similar relationship with the outcome measure into components (=profiles). PLS-Cox is an extension of the PLS method for Cox proportional hazard analyses, allowing the identification of profiles associated with time to event. We determined the number of profiles to retain based on 5-fold cross-validation using the integrated area under the curve by Song and Zhou criterion.
A profile score for each participant was calculated. Each extracted profile is characterized by weights assigned to each nutritional biomarker. We determined the stability of assigned weights in each profile using leave-one-out training sets. Nutritional biomarker weights were calculated for each profile in all training sets (n – 1). We report the variance in weights across the training sets in mean and standard deviation.
Associations of profile scores with clinical progression [HRs (95% CIs)] are reported for an unadjusted model (model 1) and a model adjusted for covariates as described before (model 2). Next, we studied how profile scores associate with clinical characteristics using linear regression analysis. For visualization purposes, profile scores were divided in quartiles and plotted in Kaplan-Meier curves. Finally, we used linear mixed models to assess the associations of nutritional biomarkers with rate of decline on the MMSE. Linear mixed models included nutritional biomarkers, time, the interaction between nutritional biomarkers and time (model 1), and the covariates as described before (model 2). The dependent variable consisted of all MMSE scores. A random intercept and slope with time were included. In these models, the main effect of nutritional biomarker [B (SE)] represents the association with baseline MMSE, while the interaction term [B (SE)] represents the association of the nutritional biomarker with annual decline in MMSE. Analyses were performed with SPSS for Windows, version 22. PLS-Cox and correlation analyses were performed using the corrplot and plsRcox package in R, version 3.4.2.
A probability level of P < .05 was considered statistically significant.
Results
Descriptives
In the total cohort and in SCD and MCI separately, participants who showed clinical progression were older and had lower Aβ42 values and higher t-tau and p-tau values compared with those who were stable (Table 1). In the total group, no differences in nutritional biomarker levels were found between clinical progression and stable. Within SCD, bilirubin levels were lower, and within MCI, homocysteine plasma and theobromine levels were higher in individuals showing clinical progression compared with those who were stable (Supplementary Table 2).
Nutritional Biomarkers and Clinical Progression
Next, we used Cox proportional hazard analyses to estimate the association of the individual nutritional biomarkers with clinical progression (Table 2). High levels of high-density lipoprotein (HDL) were associated with clinical progression [HR (95% CI), 1.4 (1.0-1.8), P = .02, model 2]. When we stratified the analyses for baseline diagnosis, effects sizes for HDL remained similar in SCD and MCI but lost significance (model 2). Additionally, some novel associations were found. Within MCI, low CSF SAM levels and high theobromine levels were associated with clinical progression [HR (95% CI) 0.8 (0.6-1.0), P < .05; 1.4 (1.0-2.0), P = .02, model 2].
Table 2Cox Proportional Hazard Models for the Association of Nutritional biomarkers With Clinical Progression During Follow-up
Cox proportional hazard models. Model 1 was unadjusted, and model 2 was adjusted for age, sex, diagnosis, cardiovascular disease, and lipid-lowering medication. Data are presented as HR (95% CI). Nutrients were log-transformed if not normally distributed, and all nutrients were transformed to z scores prior to analysis.
∗ P < .05.
† Coefficients did not converge. Data should be interpreted with caution.
Nutritional Biomarker Profiles and Clinical Progression
Nutritional biomarkers show considerable interdependencies; see for correlations between nutritional biomarkers and of nutritional biomarkers with CSF AD biomarkers in Supplementary Text 2, Supplementary Figures 1 and 2. Therefore, we used PLS-Cox analyses to detect nutritional biomarker profiles associated with clinical progression (Kaplan-Meier curves in Figure 1).
Fig. 1Kaplan-Meier curves according to nutritional biomarker profile scores in quartiles: red = quartile 1 (lowest), green = quartile 2, blue = quartile 3, purple = quartile 4 (highest). Kaplan-Meier curves depict the number of participants who still have their baseline diagnosis at a certain follow-up, with 1.0 indicating 100% and 0.8 indicating 80%. *HRs were adjusted for sex, age, cardiovascular disease and lipid-lowering medication.
Within SCD, PLS-Cox revealed 1 nutritional biomarker profile that was associated with clinical progression [HR (95% CI) SCD 1.5 (1.1-2.0), P < .01, model 2]. This profile was characterized by high positive weights for folate and betaine and high negative weights for zinc (Figure 2). When labeling SCD participants according to higher or lower profile scores, those with high scores were older, more often had hypertension and hypercholesterolemia, more often used lipid-lowering medication and had higher t-tau and p-tau levels (Supplementary Table 3).
Fig. 2Mean weights of the 33 nutritional biomarkers for the different nutritional biomarker profiles. Mean weights based on leave-one-out cross-validation are shown as bars, with the gray whiskers denoting standard deviation of nutritional biomarker weights for each nutritional biomarker profile characteristics. 25(OH) D, 25-hydroxyvitamin D; 1,25(OH)2D, 1,25-dihydroxy vitamin D; DHA, docosahexaenoic acid; HDL, high density lipoprotein; LDL, low density lipoprotein;Omega-3-FA, omega 3 fatty acids; SAM, S-adenosylmethionine; SAH, S-adenosylhomocysteine
Within MCI, we found 2 different profiles associated with clinical progression [HR (95% CI) 2.0 (1.4-2.7), P < .01; 1.7 (1.3-2.2), P < .01, model 2). The first profile was characterized by high positive weights for theobromine and HDL and high negative weights for triglycerides (Figure 2). The second profile was characterized by high positive weights for plasma SAM and high negative weights for total cholesterol (Figure 2). MCI participants with high scores for the first profile were more often female, had less follow-up, and had less often hypertension and myocardial infarction. Moreover, they showed less use of lipid-lowering medication, had lower Aβ42 levels, and had higher t-tau and p-tau levels. MCI participants with high scores for the second profile were more often male, older, and had higher body mass index and more often hypertension (Supplementary Table 3).
Nutritional Biomarkers and Cognitive Decline
Finally, we used linear mixed models to investigate the association of nutritional biomarkers with MMSE (Table 3). High triglycerides and low HDL, folate, CSF SAH, vitamin B6, and plasma and CSF uridine levels were associated with lower baseline MMSE scores. Only high HDL levels were associated with decline on the MMSE [B (SE) −0.12 (0.06), P = .04, model 2). Next, we stratified for baseline diagnosis. In SCD, low vitamin B6 levels were associated with baseline performance on the MMSE [B (SE) 0.30 (0.14), P = .03, model 2], and high folate and low bilirubin levels were associated with decline on the MMSE [B (SE) −0.16 (0.07), P = .03; 0.15 (0.06), P = .01, model 2]. In MCI, high homocysteine and low plasma and CSF uridine levels were associated with baseline MMSE scores [B (SE) −0.41 (0.20), P = .04, 0.40 (0.19), P = .03, 0.38 (0.19), P = .04, model 2] and high levels of iron, HDL, total cholesterol, and 1,25(OH)2 vitamin D were associated with decline on the MMSE [B (SE) −0.34 (0.12), P = .01; −0.18 (0.09), P < .05; −0.18 (0.09), P < .05; −0.17 (0.08), P < .05, model 2].
Table 3Linear Mixed Models for the Association of Nutritional Biomarkers With Cognitive Decline During Follow-up
Data are presented as standardized effect estimates with standard error. Linear mixed models were used to assess the association between nutritional biomarkers and the rate of cognitive decline as measured with the MMSE score. A random intercept and random slope for time (in years) were assumed. The model included nutritional biomarker, time, the interaction between nutritional biomarker and time (model 1) and sex, age, diagnosis (for total cohort models), cardiovascular disease, and lipid-lowering medication (model 2). The given effect estimates represent the difference in baseline MMSE score per standard deviation (SD) nutrient level and the difference in annual MMSE score change per SD nutritional biomarker level.
In this retrospective, longitudinal cohort study, we found promising stage-dependent nutritional biomarker profiles associated with clinical progression. Most consistent findings include high-serum HDL levels associated with clinical progression and cognitive decline.
High HDL levels were associated with clinical progression and cognitive decline. Moreover, the first nutritional biomarker profile in MCI showed a cardiovascular favorable cholesterol profile; that is, high HDL and low triglyceride levels. This is in apparent contrast with previous studies showing that cardiovascular risk factors in midlife increase the risk of late-life dementia.
Our finding is, however, consistent with studies showing that “shortly” before onset of dementia, lower cholesterol levels (suggesting a cardiovascular favorable cholesterol profile) are associated with progression to dementia.
In our study, high HDL levels were associated with clinical progression in MCI but not in SCD, suggesting that this change in cholesterols might only occur in a symptomatic stage. Our finding for HDL might also be explained by the different MCI nutritional subtypes. One MCI nutritional biomarker profile, characterized by high HDL, was associated with less cardiovascular disease, lower Aβ42, and higher t-tau and p-tau levels, whereas the second MCI profile was associated with hypertension and higher body mass index but not with AD biomarkers. This suggests 1 AD-related subgroup with a better cardiovascular health and another subgroup with increased cardiovascular risks in MCI. This is consistent with another study, showing that MCI patients with low risk on clinical progression to AD had more often obesity and hypercholesterolemia than those with high risk on AD.
Plasma homocysteine levels were higher in MCI clinical progression vs stable. Moreover, lower CSF SAM levels were associated with clinical progression. In SCD, high folate levels and a nutritional biomarker profile high in folate were associated with cognitive decline and clinical progression. The results for SCD are in contrast with those in MCI, as folate lowers homocysteine levels.
In contrast, the second nutritional biomarker profile in MCI showed high plasma SAM levels, indicating perhaps that the role of SAM alters with different compositions of nutritional biomarker synergy reflected in a profile. As homocysteine metabolism is important for the formation of synaptic membranes, these findings confirm homocysteine metabolism as an interesting, but stage- and subgroup-dependent, target for dietary intervention studies.
Consistent with previous reports on lower levels of antioxidants in AD, we found low bilirubin and zinc levels associated with cognitive decline and the nutritional biomarker profile in SCD.
We also showed some contra-intuitive findings for antioxidants; that is, high iron, 1,25(OH)2 vitamin D levels were associated with cognitive decline in SCD and high theobromine levels were related to clinical progression and the first nutritional biomarker profile in MCI. Iron and 1,25(OH)2 vitamin D levels show diurnal variation and short half-lives, which might have biased our results.
Clinical review: The role of the parent compound vitamin D with respect to metabolism and function: Why clinical dose intervals can affect clinical outcomes.
Our findings could, however, also indicate that lower levels of antioxidants are not detectable in earlier stages.
As nutritional change is likely to involve different metabolic pathways, an integrative approach that examines nutritional biomarker profiles is desirable. The best approach to study nutritional biomarker profiles is not yet clear. One previous study also used PLS-Cox to identify nutritional biomarker profiles associated with dementia risk.
Unfortunately, the overlap in measured nutritional biomarkers is small, hampering direct comparison of results. In this study, we made important first steps in testing the validity of PLS-Cox models. Replication and further validation of these nutritional biomarker profiles is however needed to define the (clinical) value.
The major limitation of this study is that our nutritional biomarker findings cannot elucidate the mechanistic role these markers have in the disease progress of dementia. Nutritional biomarker levels could indicate a nutrient insufficiency. We cannot exclude the possibility of reverse causality, however, as predementia disease could have altered the nutritional status. Therefore, this study cannot make any recommendations on whether alterations of these nutritional biomarkers with dietary interventions would be beneficial. Our findings do suggest, however, that these nutritional biomarkers might aid in the identification of patients at (metabolic) risk for clinical progression. In future studies, it might be interesting to stratify patients based on (combinations) of nutritional biomarkers; patients at (metabolic) risk for progression might benefit from dietary interventions, whereas patients without metabolic risk might be selected for physical activity or drug interventions. Another potential limitation is that we used nonfasting blood samples and had no data on dietary intake or supplement use. Dietary intake and medication possibly influenced nutritional biomarker levels.
We have corrected our results, however, for lipid-lowering medication. The nutritional biomarker findings reported in this study are subtle, because as an exploratory study we did not control for multiple comparisons; however, the results of this study need to be replicated. A strength of this study is the use of a large and diverse panel of nutritional biomarkers. Moreover, cognitive follow-up and clinical data, for example, CSF biomarkers, were available. Group sizes in this study are relatively small in comparison to population-based cohorts. To mitigate this, we oversampled participants with clinical progression during follow-up to maximize our power for risk of clinical progression.
Conclusion and Implications
In conclusion, we observed stage- and subtype-dependent nutritional biomarker profiles associated with clinical progression in memory-clinic patients with SCD and MCI. These results show that nutritional biomarker changes in predementia stages are subtle, but studied as nutritional biomarkers profiles, combinations of nutritional biomarkers are highlighted that can potentially be of value to select patients for dietary intervention trials.
Acknowledgments
We acknowledge all members of the NUDAD project team:
Amsterdam University Medical Center location VUmc: Wiesje van der Flier, Maartje Kester, Philip Scheltens, Charlotte Teunissen, Marian de van der Schueren, Francien de Leeuw, Astrid Doorduijn, Jay Fieldhouse, Heleen Hendriksen, José Overbeek, and Els Dekkers; VU University: Marjolein Visser; Wageningen University & Research: Ondine van de Rest and Sanne Boesveldt; DSM: Peter van-Dael and Manfred Eggersdorfer; Nutricia Research: John Sijben, Nick van Wijk, Amos Attali, and Martin Verkuijl; FrieslandCampina: Rolf Bos, Cecile Singh-Povel, Ellen van den Heuvel, and Martijn Veltkamp.
The authors would like to thank Desirée Smith, Rob Barto, Sigrid de Jong, Suzanne Weijers, Wjera Wickenhagen, Mariska van der Wal, and Laila Hanna-Hana from the VUmc, Amsterdam, the Netherlands and Astrid Braad from Medlon, Enschede, the Netherlands, for their technical support.
Supplementary Text 1
Most analyses were performed in the department of Clinical Chemistry of the Vrije Universiteit Amsterdam, Amsterdam UMC, Amsterdam, the Netherlands, except for the zinc, selenium, and copper analyses, which were performed in Medlon, Enschede, the Netherlands; theobromine measurements, which were performed at the Center for Neuroscience and Cell Biology, University of Coimbra, Cantanhede, Portugal; uridine measurements, which were performed at Maastricht UMC+, Maastricht, the Netherlands; and docosahexaenoic acid (DHA) and omega-3 fatty acids, which were measured at Nightingale, Helsinki, Finland.
High-density lipoprotein (HDL) cholesterol, total cholesterol, and triglycerides were measured in serum with a colorimetric assay on a Cobas 8000 system (HDL-Cholesterol plus, Cholesterol gen, Triglycerides; Roche Diagnostics, Roche, Basel, Switzerland). For these lipids, interassay coefficients of variations (CVs) were 0.9% to 2.0%. The lower limit of quantitation (LLOQ) was 0.08 mmol/L for HDL, 0.1 mmol/L for triglycerides, and 0.1 mmol/L for total cholesterol. Low-density lipoprotein (LDL) cholesterol was calculated from total cholesterol using the Friedewald formula LDL cholesterol = total cholesterol – HDL cholesterol – (0.45 × triglyceride). The Friedewald formula is valid for triglyceride levels ≤5 mmol/L. For 3 cases, triglyceride levels were >5 mmol/L, and these cases were imputed using the Friedewald formula with the highest valid value for triglycerides (= 5 mmol/L). DHA and omega-3 fatty acids were measured in plasma using high-throughput proton nuclear magnetic resonance metabolomics as described previously.
Serum insulin was measured by sandwich luminescence immunoassay on a Centaur XP analyzer (Siemens). Interassay CV for insulin in serum is 7.3%, and LLOQ is 10 pmol/L. Plasma glucose was tested with ultraviolet (UV) wave detection on a Cobas 8000 system (Glucose HK Gen; Roche Diagnostics). Interassay CV for glucose was 1.1% to 1.3%; LLOQ was 0.11 mmol/L.
Plasma homocysteine, serum folate, and serum vitamin B12 were measured by competitive luminescence immunoassay on an Architect analyzer (Abbott Diagnostics, Abbott Laboratories). Interassay CVs were 2% to 4% for homocysteine in plasma, 9% for folate, and 6.3% for vitamin B12. LLOQ was 1 μmol/L for homocysteine in plasma, 2 nmol/L for folate, and 44 pmol/L for vitamin B12. For cerebrospinal fluid (CSF) homocysteine, to 100 μL of CSF, 250 pmol [D4]-homocysteine (Cambridge Isotopes) was added, and the samples were reduced using 25 mmol/L dithiothreitol for 30 minutes at 4°C. Subsequently, sample clean-up was performed using cation exchange solid-phase extraction, and the homocysteine was determined using liquid chromatography–tandem mass spectrometry (LC-MS/MS). Details of the method are described previously.
The interassay CV for this assay was 5.1% to 28.1%. The lower LLOQ for CSF homocysteine was 0.005 μmol/L. Choline and betaine were detected in plasma using normal-phase chromatography and positive electrospray mass spectrometry (MS) as described by Holm et al.
Determination of choline, betaine, and dimethylglycine in plasma by a high-throughput method based on normal-phase chromatography-tandem mass spectrometry.
A plasma pool spiked at 3 different levels was used for assessing the precision. Interassay CVs for choline and betaine were 0.8% to 1.9% and 1.1% to 2.2%, respectively. The lower limit of quantification for choline and betaine in plasma were 0.3 and 0.8 μmol/L. Plasma and CSF S-adenosylmethionine (SAM) and SAH were measured by positive electrospray LC-MS/MS. 2H3-SAM, and 13C5-SAH were used as internal standards. SAM and SAH were detected using m/z transitions 399.2 → 136.1 for SAM, 402.2 → 136.1 for 2H3-SAM, 385.2 → 136.1 for SAH, and 390.2 → 136.1 for 13C5-SAH. The interassay CVs for SAM are 3.2% in CSF and 7.6% in plasma and for SAH 8.6% in CSF and 6.0% in plasma. Vitamin B6 in plasma was determined using LC-MS/MS adapted from a previously published method.
Plasma was diluted 10 times with saline and deproteinized using 10% trichloroacetic spiked with stable isotope labeled internal standard (PLP-d2). Quantification was carried out with reverse-phase chromatography and positive electrospray ionization mode. Interassay variation calculated for 3 quality control samples was 1.4% to 2.0%, and the LLOQ was 1.2 nmol/L. Plasma and CSF uridine were measured by ultraperformance LC-MS/MS (UPLC-MS/MS). 15N2-uridine was used as an internal standard. Samples were deproteinized with acetonitrile before quantification. Interassay CV for uridine in plasma and CSF was 4.0% to 14%. LLOQ for plasma and CSF uridine was estimated at 0.2 μmol/L.
Nutrients required for phospholipid synthesis are lower in blood and cerebrospinal fluid in mild cognitive impairment and Alzheimer's disease dementia.
Plasma zinc, copper, and selenium were detected by electron multiplier using inductively coupled mass spectrometry (ICP-MS) (NexION 300D; Perkin Elmer, Waltham, MA). The LLOQ was 0.27 μmol/L for copper, 0.3 μmol/L for zinc, and 0.15 μmol/L for selenium. Interassay CVs were 3.1% for copper, 2.9% for zinc, and 4.7% for selenium. Vitamins A and E were measured in plasma using high-performance liquid chromatography with UV wave detection.
Interassay variation was determined at 2 different concentrations and ranged from 0.7% to 1.1 % for vitamin A and 0.8% to 1.6% for vitamin E. The LLOQ was 0.1 μmol/L for vitamin A and 1.0 μmol/L for vitamin E. Serum albumin concentrations were determined by rate nephelometry on a Beckman Coulter IMMAGE 800 immunochemistry system (ALB test, Beckman Coulter, Danaher, Washington, DC). The interassay CV for serum was 2.3% to 4.3%. Serum bilirubin, magnesium, and iron were measured by a colorimetric assay on a Cobas 8000 system (Bilirubin Total Gen, Magnesium Gen, Iron Gen; Roche Diagnostics). The interassay CV for serum bilirubin, iron, and magnesium was 1.3% to 3.3%. Five cases had bilirubin levels below the limit of detection (2.5 μmol/L); these were imputed with half of the lowest limit of detection (1.25 μmol/L). Theobromine in CSF was analyzed on a CTC-xt (PAL System)–LC Nexera system (Shimadzu) coupled to a 4000 QTRAP mass spectrometer operated by Analyst 1.6.1 (Sciex). Theobromine and theobromine-d6 were quantified using LC-MS/MS. The intra- and interday CVs were <10% for low, medium and high levels. The LLOQ determined based on the calibration curve was 0.04 μM. LC-MS/MS was used to detect 1,25-dihydroxyvitamin D and 25-hydroxyvitamin D levels in serum. Interassay CVs for 25-hydroxyvitamin D and 1,25-dihydroxyvitamin was 7.5%. LLOQ was 10 pmol/L for 1,25-dihydroxyvitamin D and 4 nmol/L for 25-hydroxyvitamin D.
Supplementary Text 2
Interdependency of nutrients was evaluated using Pearson correlations (Supplementary Figure 1). As expected, we found correlations between nutritional biomarkers sharing metabolic pathways such as lipids and nutrients from the homocysteine metabolism. Additionally, fat-soluble vitamin E was positively correlated to LDL and total cholesterol (r = 0.63, P < .01, and r = 0.70, P < .01), whereas SAM and uridine were negatively correlated both in plasma and CSF (r = −0.39, P < .01, and r = −0.35, P < .01, in CSF). The above-mentioned correlations remained intact when analyzing subjective cognitive decline (SCD) and mild cognitive impairment (MCI) separately, except for the negative association between SAM and uridine plasma that lost significance in both groups (data not shown). Next, associations between nutritional biomarkers and AD biomarkers (ie, Aβ42, tau, and p-tau) were examined in SCD and MCI (Supplementary Figure 2). In both groups, the strongest correlations were found between SAH CSF and tau and p-tau levels (tau r = 0.38, P < .01 SCD; r = 0.38, P < .01 MCI; p-tau r = 0.45, P < .01 SCD; r = 0.46, P < .01 MCI). Additionally, some correlations were specific for SCD and MCI. In SCD, bilirubin was negatively correlated with tau and p-tau (r = −0.27, P < .01; r = −0.25, P < .01). In MCI, triglycerides were negatively correlated, and HDL was positively correlated with tau and p-tau levels (r = −0.35, P = .02; r = −0.35, P = .02; r = 0.28 P < .01; r = 0.27 P < .01).
Supplementary Figure 1Correlation between nutritional biomarkers. Pearson correlation plot between nutritional biomarkers. Blue color depicts positive correlation coefficients, whereas red depicts negative correlation coefficients. Insignificant (P > .05) correlation coefficients are left blank. 25(OH) D, 25-hydroxyvitamin D; 1,25(OH)2D, 1,25-dihydroxy vitamin D; DHA, docosahexaenoic acid; HDL, high density lipoprotein cholesterol; LDL, low density lipoprotein cholesterol; Omega-3-FA, omega 3 fatty acids; SAH, S-adenosylhomocysteine; SAM, S-adenosylmethionine.
No reference values available, but levels in our cohort are not materially different from large healthy cohorts, except for choline levels that were higher in our cohort.7–9
No reference values available, but levels in our cohort are not materially different from large healthy cohorts, except for choline levels that were higher in our cohort.7–9
No reference values available, but levels in our cohort are not materially different from large healthy cohorts, except for choline levels that were higher in our cohort.7–9
No reference values available, but levels in our cohort are not materially different from large healthy cohorts, except for choline levels that were higher in our cohort.7–9
No reference values available, but levels in our cohort are not materially different from large healthy cohorts, except for choline levels that were higher in our cohort.7–9
∗ Reference values as reported by local laboratories that performed measurement of the nutritional biomarkers.
† LDL cholesterol was calculated using the Friedewald formula: LDL cholesterol = total cholesterol – HDL cholesterol – (0.45 × triglycerides).
‡ No reference values available, but levels in our cohort are not materially different from large healthy cohorts, except for choline levels that were higher in our cohort.
Determinants of the essential one-carbon metabolism metabolites, homocysteine, S-adenosylmethionine, S-adenosylhomocysteine and folate, in cerebrospinal fluid.
P < .05. P values correspond to linear regression analysis on (log-transformed) z-scores.
25(OH) D, nmol/L
64.0 (44.9–84.0)
65.0 (48.7–79.6)
66.0 (47.5–85.5)
64.8 (47.1–79.1)
60.9 (42.1–81.2)
65.2 (49.4–79.9)
1,25(OH)2 D, pmol/L
118 (98–144)
118 (95–136)
119 (101–140)
111 (90–131)
115 (93–146)
124 (100–143)
25(OH) D, 25-hydroxyvitamin D; 1,25(OH)2D, 1,25-dihydroxy vitamin D; CSF, cerebrospinal fluid; DHA, docosahexaenoic acid; HCy, homocysteine; HDL, high density lipoprotein cholesterol; LDL, low density lipoprotein cholesterol; Omega-3-FA, omega 3 fatty acids; SAH, S-adenosylhomocysteine; SAM, S-adenosylmethionine; TC, total cholesterol.
Data in median (interquartile range).
Groups were compared on their outcome (stable vs progression) in the total cohort and within subgroups of their baseline syndrome diagnosis (SCD/MCI). Nutritional biomakers level models were adjusted for sex, age, diagnosis, cardiovascular disease, and lipid-lowering medication.
∗ P < .05. P values correspond to linear regression analysis on (log-transformed) z-scores.
Aβ42, amyloid beta-42; APOE, apolipoprotein E; BMI, body mass index; MMSE, Mini-Mental State Examination; p-tau, phosphorylated tau; t-tau, total tau.
Unless otherwise noted, values are B (SE). Associations of clinical characteristics with profile scores were tested with linear regression analyses. Aβ42, t-tau, and p-tau scores were log-transformed and standardized into z scores.
∗ P < .05.
† 61 participants (20%) had missing values in BMI values.
Nutritional biomarkers required for phospholipid synthesis are lower in blood and cerebrospinal fluid in mild cognitive impairment and Alzheimer's disease dementia.
24-month intervention with a specific multinutritional biomarker in people with prodromal Alzheimer's disease (LipiDiDiet): A randomised, double-blind, controlled trial.
A 2 year multidomain intervention of diet, exercise, cognitive training, and vascular risk monitoring versus control to prevent cognitive decline in at-risk elderly people (FINGER): A randomised controlled trial.
Dietary changes and cognition over 2 years within a multidomain intervention trial-The Finnish Geriatric Intervention Study to Prevent Cognitive Impairment and Disability (FINGER).
The diagnosis of mild cognitive impairment due to Alzheimer's disease: recommendations from the National Institute on Aging-Alzheimer's Association workgroups on diagnostic guidelines for Alzheimer's disease.
Clinical diagnosis of Alzheimer's disease: Report of the NINCDS-ADRDA Work Group under the auspices of Department of Health and Human Services Task Force on Alzheimer's Disease.
The diagnosis of dementia due to Alzheimer's disease: Recommendations from the National Institute on Aging-Alzheimer's Association workgroups on diagnostic guidelines for Alzheimer's disease.
Clinical review: The role of the parent compound vitamin D with respect to metabolism and function: Why clinical dose intervals can affect clinical outcomes.
Funding sources: The NUDAD project is funded by NWO-FCB (project number 057-14-004), and Francisca de Leeuw and Maartje Kester are appointed on this project. Research of the Alzheimer Center Amsterdam is part of the neurodegeneration research program of Amsterdam Neuroscience. The Alzheimer Center Amsterdam is supported by Stichting Alzheimer Nederland and Stichting VUmc fonds. The clinical database structure was developed with funding from Stichting Dioraphte. Docosahexaenoic acid and total omega-3 fatty acids measurements were funded by Biobanking and Biomolecular Resource Research Infrastructure (BBMRI)-NL (NWO 184.021.007). Theobromine measurements were supported by Fundação para a Ciência e a Tecnologia (FCT) [POCI-01-0145-FEDER-007440 (strategic project UID/NEU/04539/2019), UID/BIM/04773/2013, PEst-C/SAU/LA0001/2013-2014, POCI-01-0145-FEDER-016428 (ref.: SAICTPAC/0010/2015), POCI-01-0145-FEDER-30943 (ref.: PTDC/MEC-PSQ/30943/2017)], PTDC/MED-NEU/27946/2017 and co-financed by “COMPETE Programa Operacional Factores de Competitividade,” QREN; the European Union (FEDER—Fundo Europeu de Desenvolvimento Regional) and by the National Mass Spectrometry Network (RNEM) (POCI-01-0145-FEDER-402-022125).
N.v.W. is an employee of Nutricia Research; E.H. is an employee of FrieslandCampina, a dairy company; H.M. is an advisor to the life science and medical industry and a lecturer at the University of Zurich. During the course of the present work, he was an employee of DSM. P.S. has received consultancy/speaker fees (paid to the institution) from Novartis, Vivoryon, Genentech, and EIP Pharma. C.T. received grants from the European Commission, the Dutch Research Council (ZonMW), Association of Frontotemporal Dementia / Alzheimer's Drug Discovery Foundation, The Weston Brain Institute, Alzheimer Nederland. C.T. has a collaboration contract with ADx Neurosciences, performed contract research, or received grants from Probiodrug, Biogen, Esai, Toyama, Janssen Prevention Center, Boehringer, AxonNeurosciences, Fujirebio, EIP Farma, PeopleBio, and Roche. Research programs of W.v.d.F. have been funded by ZonMw, NWO, EU-FP7, EU-JPND, Alzheimer Nederland, CardioVascular Onderzoek Nederland, Health∼Holland, Topsector Life Sciences & Health, Stichting Dioraphte, Gieskes-Strijbis fonds, Stichting Equilibrio, Pasman Stichting, Biogen MA Inc, Boehringer Ingelheim, Life-MI, AVID, Roche BV, Janssen Stellar, and Combinostics. W.v.d.F. holds the Pasman chair. W.v.d.F. has performed contract research for Biogen MA Inc and Boehringer Ingelheim. W.v.d.F. has been an invited speaker at Boehringer Ingelheim and Biogen MA Inc. W.v.d.F. is recipient of a donation by Stichting Equilibrio and of a ZonMW Memorabel grant (#733050814). All funding is paid to her institution. F.d.L., B.T., V.M., B.M., J.B., and M.K. report no conflict of interest.
30869-2&id=gr1.jpg){kind=link}
30869-2&id=gr2.jpg){kind=link}
30869-2&id=fx1.jpg){kind=link}
30869-2&id=fx2.jpg){kind=link}