Association between Modifiable Risk Factors and Levels of Blood-Based Biomarkers of Alzheimer’s and Related Dementias in the Look AHEAD Cohort

Background Emerging evidence suggests that a number of factors can influence blood-based biomarker levels for Alzheimer’s disease (AD) and Alzheimer’s related dementias (ADRD). We examined the associations that demographic and clinical characteristics have with AD/ADRD blood-based biomarker levels in an observational continuation of a clinical trial cohort of older individuals with type 2 diabetes and overweight or obesity. Methods Participants aged 45-76 years were randomized to a 10-year Intensive Lifestyle Intervention (ILI) or a diabetes support and education (DSE) condition. Stored baseline and end of intervention (8-13 years later) plasma samples were analyzed with the Quanterix Simoa HD-X Analyzer. Changes in Aβ42, Aβ40, Aβ42/Aβ40, ptau181, neurofilament light chain (NfL), and glial fibrillary acidic protein (GFAP) were evaluated in relation to randomization status, demographic, and clinical characteristics. Results In a sample of 779 participants from the Look AHEAD cohort, we found significant associations between blood-based biomarkers for AD/ADRD and 15 of 18 demographic (age, gender, race and ethnicity, education) and clinical characteristics (APOE, depression, alcohol use, smoking, body mass index, HbA1c, diabetes duration, diabetes treatment, estimated glomerular filtration rate, hypertension, and history of cardiovascular disease) . Conclusions Blood-based biomarkers of AD/ADRD are influenced by common demographic and clinical characteristics. These factors should be considered carefully when interpreting these AD/ADRD blood biomarker values for clinical or research purposes.


Introduction A
l z h e i m e r ' s d i s e a s e ( A D ) b l o o d -b a s e d biomarkers have evolved rapidly over the last decade with the development of more sensitive technologies, such as single molecular array (SIMOA) immunoassays, to more accurately measure biomarkers of AD pathology (Aβ42/Aβ40 ratio and phosphorylated tau at threonine 181 [ptau181]), and neurodegeneration (neurofilament light chain, [NfL] and glial fibrillary acidic protein [GFAP]) (1)(2)(3).However, with the increasing availability of blood biomarker assays for clinical use, there is an urgent need to understand their utility in large, heterogeneous populations.Several recent studies have reported that multiple chronic conditions can affect the interpretation of the blood biomarkers, causing increases or decreases in levels due to peripheral physiological factors such as obesity or renal function (4)(5)(6)(7)(8).Understanding the impact of multiple chronic conditions on AD biomarker levels in diverse populations will inform their use and interpretation in clinical settings.
The Action for Health in Diabetes (Look AHEAD) MIND cohort provides an opportunity for a unique contribution to this emerging body of literature in the high AD and Alzheimer 's related dementia (ADRD) risk group of persons with type 2 diabetes (T2D) (9).Look AHEAD participants represent a diverse group of individuals with T2D and overweight or obesity, who were recruited from 16 sites across the US who were randomized 1:1 to a lifestyle intervention vs. a control condition (10).To date there have been few, if any, studies that have comprehensively examined the AD/ADRD blood-based biomarkers in this type of cohort.Also, associations between diabetes duration and the AD/

Setting, Participants, and Intervention
Participants (n=5,145) aged 45-76 years were recruited from 16 sites across the United States and were randomized to either an Intensive Lifestyle Intervention (ILI) or a Diabetes Support and Education (DSE) condition.The ILI was a multidomain intervention including dietary modification and physical activity with a goal of inducing an average of ≥7% weight loss at one year and maintenance of weight loss over the course of the study (12).Participants in the ILI arm had a daily calorie goal of 1200 to 1800 kcal based on initial weight.The diet specified <30% total calories from fat (<10% saturated fat) and a minimum of 15% total calories from protein.The physical activity goal was similar in intensity to brisk walking for at least 175 minutes per week.The participants randomized to the DSE condition were invited, but not required, to attend three yearly group sessions.These sessions focused on diet, physical activity, and social support (13).There were no specific instructions or goals for weight loss, physical activity, or dietary modification.The current analyses include those who participated in blood draws at baseline and in 2010-2013 participated in cognitive interviews.Preference for inclusion was given to those who also participated in a brain MRI ancillary study.Compared to those included in the study, participants who were not included were younger, more were women, and more often selfidentified as American Indian/Native American/Alaskan Native or Hispanic.Excluded participants had higher levels of education, higher baseline BMI, and fewer had a baseline history of cardiovascular disease (CVD), as seen in Supplemental Table 1.

Biomarkers
Blood samples were drawn at baseline and proximal to the end of the intervention, approximately 8-13 years later.All assays were performed using EDTA plasma samples stored at Medpace Reference Laboratories, Cincinnati, OH.Concentrations of Aβ42, Aβ40, GFAP, and NfL were measured using the Simoa Human Neurology 4-Plex E Advantage Kit on a Quanterix Simoa HD-X bead-based immunoassay analyzer (Quanterix Corporation, Billerica, MA).The determination of the concentration of the pTau-181 was performed using the Simoa pTau-181 V2 Advantage kit.For all assays, withinrun coefficients of variation (CVs) ranged from 3-19% and between-run CVs ranged from 6-13%.Biomarkers were converted to z-scores using baseline sample means and standard deviations for comparability.

B a s e l i n e d e m o g r a p h i c a n d c l i n i c a l characteristics
Demographic characteristics were collected at baseline and included age; self-identified gender, race, and ethnicity; years of education; alcohol use; and smoking status.Participants provided self-reports of clinical characteristics including the duration of their diabetes, type of diabetes treatment (diet, oral medications, oral medications and insulin), hypertension, high cholesterol, history of CVD (including MI, heart bypass surgery, coronary artery bypass graft, carotid endarterectomy, lower leg angioplasty, aortic aneurysm, congestive heart failure, or stroke), peripheral neuropathy, and use of antidepressants.Additional variables measured at baseline included BMI, systolic and diastolic blood pressure, and depressive symptoms (Beck Depression Inventory,BDI (14)).Blood specimens were analyzed centrally at the Northwest Lipid Metabolism and Diabetes Research Laboratories at the University of Washington for HbA1c and creatinine.Estimated glomerular filtration rate (eGFR) was calculated.APOE ε4 carrier status was determined for participants who provided consent (80% of women versus 86% of men, p<0.001), using TaqMan genotyping (rs7412 and rs429358) (15).

Statistical Methods
Demographic and clinical characteristics of the sample were compared by intervention assignment.Continuous variables were evaluated using t-tests and categorical variables were evaluated with χ 2 tests.Spearman correlations between individual biomarkers at baseline and change from baseline to the end of the intervention were calculated.Spearman correlations were also calculated for associations between each of the biomarkers and age, BMI, and diabetes duration at baseline and for change between these two times.For each characteristic, we evaluated associations with biomarker levels at baseline and change from baseline to the end of the intervention, adjusted for age and gender using linear regression models.We present these data in two ways, forest plots with each individual measure compared to a 'standard' referent category that most resembles existing biomarker studies, and in tables where the least squares means are reported without fixed reference categories.

Results
There were 779 participants (383 DSE and 396 ILI) with available blood samples and cognitive data for this  2a and 2b.At baseline, all biomarkers significantly correlated with each other, with two exceptions: the Aβ42/40 ratio only correlated with Aβ42 and Aβ40; and ptau181 and GFAP were not correlated.Next, we examined correlations between change in biomarker levels from baseline to the end of the intervention (Table 2b); nearly all biomarkers correlated, except for pTau181 or GFAP with Aβ42/40.We then examined correlations between baseline biomarker levels and baseline values of age, BMI, and diabetes duration; we also examined correlations between these variables and longitudinal change in biomarker levels (Table 3).At baseline, Aβ42, Aβ40, NfL and GFAP positively correlated with age; Aβ42/40, NfL, and GFAP were inversely correlated with baseline BMI; and diabetes duration positively correlated with NfL.Change in NfL and GFAP positively correlated with age, while change in Aβ42/40 negatively correlated with age.In Figures 2a and 2b, forest plots depict the associations between each biomarker by age groups (45-55, 56-65, 65-76).Both Aβ42 and Aβ40 levels differed significantly among age groups at baseline and change over time with higher biomarker levels in older age groups at baseline and greater increases in biomarker levels over time among the youngest and oldest age groups.Age group was not significantly associated with Aβ42/40 ratio at baseline, but it was associated with change in Aβ42/40 ratio from baseline to years 8-13 (p<0.001).Figure 2b shows no significant differences in pTau181 by age groups at baseline, but significant differences by age group emerged when considering change from baseline (p=0.005).NfL and GFAP demonstrated significant associations with age groups at baseline and change (all p<0.05).
Figure 3a and 3b show forest plots of each biomarker by gender at baseline and change.Women had significantly lower levels of Aβ40 compared to men (p=0.02) at baseline.Women had lower NfL levels compared to men at baseline and change (all p<0.05).However, women had higher levels of GFAP at baseline (p=0.009; Figure 3b).
Given the effect of age and gender on biomarker levels, subsequent analyses were adjusted for these demographics.Results are presented two ways: 1) visually for baseline and change values (from baseline to end of intervention) as forest plots (Figures 4-9) with comparisons to a common reference category set at zero; and 2) as tables in which least squares means of the z-scores are reported without reference categories and including baseline and change from baseline to the end of the intervention (Tables 4-9).

Aβ42
Aβ42 levels differed by race and ethnicity in change from baseline (p=0.007);Hispanic and Black participants had lower levels of change than White or other/ mixed race participants (Table 4 and Figure 4).Never smokers had the lowest increase in Aβ42 levels (p=0.004)compared to past or current smokers.Participants with elevated baseline HbA1c levels (≥7%) had lower baseline levels of Aβ42 (p=0.002) and a greater increase over time (p=0.004).Participants with lower eGFR had higher levels of Aβ42 at baseline (p<0.0001).Hypertension at baseline was associated with higher Aβ42 levels (p=0.0002).Those with a history of CVD at baseline had greater increase in Aβ42 over time than those without a baseline history of CVD (p=0.021).

Aβ40
White participants and those reporting other/mixed race and ethnicity had higher Aβ40 levels than Black or Hispanic participants at baseline (p=0.018); and greater increases from baseline (p=0.003)(Table 5 and Figure 5).Participants who never smoked had had less of an increase in Aβ40 over time (p=0.008)than past or current smokers.Participants with HbA1c levels ≥7% had lower Aβ40 levels at baseline (p=0.002) and a greater increase over time (p=0.025).Participants with lower baseline eGFR had higher levels of Aβ40 (p<0.0001) at baseline compared to those with higher eGFR.Participants with hypertension at baseline had higher levels of Aβ40 at baseline than those without hypertension (p=0.004);participants with a baseline history of CVD had a greater increase in Aβ40 levels than those without a baseline history of CVD (p=0.009).

Aβ42/40
At baseline, Black participants had a higher Aβ42/40 ratio compared to White participants and those reporting other/mixed race (p=0.029)(Table 6 and Figure 6).APOE ε4 carriers had a lower Aβ42/40 ratio at baseline (p=0.003).Participants who had diabetes treated with medications had a higher Aβ42/40 ratio compared to those who were not treated with oral medications or insulin (p=0.038).Higher eGFR was associated with a lower Aβ42/40 ratio at baseline (p=0.018).Participants who self-reported hypertension at baseline had greater decrease in Aβ42/40 ratio compared to those who did not (p=0.012).

pTau181
At baseline, there were no significant associations between any characteristic and ptau181 levels, adjusting for age and gender (Table 7 and Figure 7).In the time from baseline to the end of the intervention, White participants had the greatest increase in pTau181 levels over time compared to the other groups (p=0.0002).

NfL
Baseline NfL levels (Table 8 and Figure 8), adjusted for age and gender, are associated with race and ethnicity, with Black participants having the lowest NfL levels at baseline (p=0.010).Participants who had diabetes <5 years at baseline had lower NfL levels than those who had diabetes ≥5 years (p<0.0001).Participants with lower HbA1c levels at baseline had less increase in NfL levels by the end of the intervention (p=0.032).Participants who treated their diabetes without medication had the lowest NfL levels at baseline (p=0.003).Lower eGFR levels were associated with higher NfL levels at baseline (p<0.0001).

GFAP
Lower GFAP levels (Table 9, Figure 9) were associated with lower education levels at baseline (p=0.006).BDI scores ≥10 were associated with lower GFAP levels at baseline (p=0.001).Alcohol use at baseline ≥21 oz/week was associated with higher GFAP levels at baseline (p=0.019).Higher baseline BMI was associated with lower GFAP levels at baseline (p=0.002).Participants with baseline eGFR of 60-90 had the highest GFAP levels at baseline (p=0.006).Finally, having a history of CVD at baseline was associated with higher GFAP levels at baseline (p=0.032).

Intervention
As noted above, there were no significant intervention group differences at baseline in biomarker levels.

Discussion
Blood-based AD biomarkers have complex associations with each other, with age and gender, and with a number of common conditions that are themselves risk factors for dementia.Some of these conditions may influence blood biomarker levels because they are risk factors for AD whereas others may influence the biomarker levels because they physiologically impact peripheral levels (4,8).Most studies of blood-based AD biomarkers conducted to date, have focused on relatively healthy and homogenous populations which do not reflect the reality of older populations with multiple chronic conditions and cognitive impairment, who will be most likely to have blood biomarker testing.There is an urgent need to understand the impact of factors that may affect the interpretation of the blood markers in the general population, especially now that some of these biomarkers are clinically available for use in forming diagnoses at the population level.In this study, we examined associations between AD-related plasma biomarkers (Aβ42, Aβ40, Aβ42/Aβ40, pTau181, NfL, and GFAP) and common conditions in older adults with T2D and overweight or obesity.The cohort comprised clinical trial participants, so we also evaluated the legacy effect of the intensive lifestyle intervention on biomarker levels.Of the 18 factors tested, 15 had significant associations with at least one of the AD blood-based biomarkers.We discuss and place into context our results for each factor studied below.We have chosen to focus on comparisons to studies with similar methods, age ranges, and those that mostly included participants who were cognitively normal at their baseline assessment.See Supplemental Table 2 for basic information on the comparison studies including biomarker methods, biomarkers targeted, sample size, age, and cognitive status.

Age
At baseline, Aβ42, Aβ40, NfL, and GFAP were positively associated with age.At the end of the intervention, Aβ42/Aβ40 was inversely associated with age, while NfL and GFAP were positively associated with age.Our findings largely correspond with some, but not all other studies.For example, the Mayo Clinic Study of Aging (MCSA) showed higher levels of Aβ42 with older ages (4); but the Health, Aging and Body Composition Study (Health ABC) study found no association between Aβ42 and age (16).MCSA also found increases in Aβ40 levels with age, as did Health ABC ( 16) and the Systolic Blood Pressure Intervention Trial (SPRINT) trial ( 5).In the Australian Imaging, Biomarker & Lifestyle Flagship Study of Aging (AIBL) (17), similar to our study, there was no association between baseline Aβ42/Aβ40 and age whereas MCSA showed a negative association between Aβ42/Aβ40 and age.We found no association between pTau181 and baseline age, but the AIBL study found higher pTau181 levels with older ages.Our NfL and GFAP findings were more consistent with findings from the Action to Control Cardiovascular Risk in Diabetes (ACCORD) study (18), AIBL (17), the Austrian Stroke Prevention Family Study (ASPS-Fam) (19), the National Health and Nutrition Examination Survey (NHANES) (20), and SPRINT ( 5), all showing higher NfL and GFAP levels with increased age.Taken together, these findings suggest that Aβ42, Aβ40, NfL, and GFAP levels increase with age.Relationships between Aβ42/Aβ40, pTau181, and age are less consistent at this time.Of note, one study reported an increase of pTau181 with age only among individuals who were amyloid PET positive (8).Thus, the relationship between pTau181 and age may appear stronger among cohorts who have a higher prevalence of amyloid positivity.

Gender
In the present study, women had lower levels of Aβ40 than men at baseline.Our results corresponded with the lower Aβ40 levels found in women in the UK Medical Research Council (MRC) National Survey for Health and Development (NSHD) (21).Conversely, there were no gender differences in Aβ40 levels between women and men in the Health ABC cohort (16).While we reported no gender differences in baseline or levels of change over time in Look AHEAD, AIBL (17) similarly showed no differences in Aβ42/Aβ40 by gender; but the MCSA showed lower levels of Aβ42/Aβ40 among women (4).These findings do not paint a clear picture and suggest that more work is needed to understand gender and sex differences in blood-based biomarkers for AD.The various studies we have cited have adjusted for different sets of comorbidities and samples were drawn from different populations with different characteristics which could contribute to different findings.

Race and Ethnicity
Lower levels of Aβ42 were observed among Black and Hispanic participants compared to White participants.Aβ40 at baseline and change in biomarker levels over the interval followed the same pattern with Black and Hispanic participants having lower levels and less change over time.The opposite trend emerged for the Aβ42/ Aβ40 ratio at baseline and in change over time with White and other/mixed race groups having lower levels and less change over time.Ptau181 levels increased the most over time among White participants compared to other groups.Black participants had the lowest NfL levels on average at baseline.The only biomarker in our study that was not associated with race and ethnicity was GFAP.Others have found similar associations between race and biomarker levels.Hajjar et al. (22), found race differences in biomarkers Aβ42, Aβ40, pTau181, and NfL with African Americans having lower levels than their White counterparts in the Brain, Stress, Hypertension, and Aging Research Program (B-SHARP).In the Health ABC study, investigators similarly found lower levels of Aβ42 and Aβ40 in African American participants compared to White participants (16).Mexican Americans were found to have lower levels of Aβ40, total tau, and Aβ42/Aβ40, in the Health & Aging Brain study among Latino Elders (HABS-HD) (7).However, a study using the NHANES data found no differences in NfL levels by race in models that controlled for age, sex, stroke, diabetes, eGFR, and alcohol consumption (20).More work needs to be done to get a better grasp on these associations as race and ethnicity are social constructs and represent proxies for systemic racism.Associations between biomarkers and race and ethnicity could change depending on the confounders or social determinants of health that are addressed in any given analysis.

Education
Education levels were associated with GFAP at baseline with those in the lowest education group having the lowest GFAP levels.In the Health ABC study, low education was associated with lower Aβ42 levels (16).It is unclear whether associations between education level and blood-based biomarkers are biologically based or whether higher education is an effective proxy for health advantages due to socio-economic status.

APOE
APOE ε4 carrier status was only associated with lower Aβ42/Aβ40 ratio at baseline in Look AHEAD.In Health ABC, APOE ε4 carrier status was associated with lower   Aβ42 levels (16).In the AIBL study, APOE ε4 carrier status was not associated with biomarker levels after adjustment for age, sex, diagnosis, and amyloid PET Aβ status (17).

Depressive Symptoms and Antidepressant Usage
Depressive symptoms as measured by the BDI (stratified as BDI<10 vs BDI ≥10), were associated with GFAP levels at baseline.A recent systematic review of GFAP studies indicated associations between GFAP and a number of neurological diseases and disorders, including major depressive disorder as well as Alzheimer's disease (23).Although there seems to be some association with depression, we were unable to find studies that also investigated antidepressant usage.

Alcohol and Smoking
Higher alcohol use was associated with higher baseline GFAP levels in our study.In the NHANES study, alcohol use was associated with NfL levels instead, in models that controlled for age, sex, stroke, race, diabetes, and eGFR (20).A history of smoking was associated with an increase in Aβ42 and Aβ40 levels over time in our cohort.We were only able to find one other study that referenced smoking and blood-based AD biomarkers.That study, conducted among patients with heart failure, reported former smokers had lower GFAP levels (24).More work should be done to better understand the interplay between alcohol use, smoking and AD biomarker levels as these are common exposures in older cohorts.

BMI
Given that Look AHEAD recruited participants with overweight or obesity (mean baseline BMI=34.8±5.3 kg/ m 2 ), there was less variation in BMI compared to other studies so we expected to have limited ability to identify differences by BMI.Higher BMI was only significantly associated with lower GFAP levels at baseline, and this association was consistent with a report using BioFINDER data where BMI was inversely correlated with NfL and GFAP (25).In a cohort of 327 community dwelling participants in the ASPS-Fam study, Koini et al. found that BMI was a significant predictor of lower NfL among   younger participants (<60) (19).In the AIBL study, lower BMI was correlated with higher pTau181, NfL, and GFAP levels (17).

Diabetes Duration
More than half of our participants (57%) had diabetes for 5 years or more at baseline.A longer duration of diabetes was associated with higher NfL levels at baseline.Serum NfL levels were also elevated among diabetics in an HNANES analysis compared to those without diabetes (20).Using data from the HABS-HD study, O'Bryant et al. found relationships between the presence of diabetes (not duration) and higher levels of Aβ42, Aβ40, and NfL (7).The NHANES study examined the association between NfL and the presence of diabetes (also not considering duration), and found that diabetes was associated with higher NfL levels in models that controlled for age, sex, stroke, race, diabetes, eGFR, and alcohol consumption (20).

HbA1c
HbA1c levels <7% were associated with higher Aβ42 and Aβ40 at baseline and lower levels of increase from baseline to end of intervention.HbA1c levels ≥7% were associated with greater NfL increases from baseline.In the ACCORD Study, NfL was similarly positively associated with HbA1c (18).In the HABS-HD study, higher HbA1c was also related to higher Aβ42, Aβ40, and NfL (7).

Diabetes Treatment
Participants not using diabetes medications had lower Aβ42/Aβ40 and NfL levels at baseline.While few studies have reported specifically on the relationship between diabetes treatment and blood-based biomarkers for AD, several studies have investigated the associations between diabetes treatment and cognitive outcomes (26).Future work should examine more closely the relationship between specific diabetes medication classes and blood-based biomarkers as they have different effects (27).

Peripheral Neuropathy
We found no significant relationships between any of the biomarkers and neuropathy.In contrast, a metaanalysis of 36 studies found significant associations between neuropathy and higher NfL levels (28).It is possible our data failed to reflect this association because the generally higher BMI levels may dilute NfL levels in our cohort.eGFR Estimated glomerular filtration rate (eGFR) has been consistently associated with biomarker levels in the literature (8).A higher filtration rate (i.e., higher eGFR) indicates better kidney function and tends to be associated with lower levels of biomarkers in the blood.In Look AHEAD, higher eGFR was associated with lower baseline levels of Aβ42, Aβ40, Aβ42/Aβ40 ratio, NfL, and GFAP.In the ASPS-Fam study, ACCORD, and NHANES, renal function measured by eGFR was similarly associated with NfL levels (18)(19)(20).A higher eGFR was also significantly related to lower Aβ42, Aβ40, Tau, and NfL; and higher levels of Aβ42/Aβ40 in the HABS-HD study (7).In SPRINT, after adjustment for age, both Aβ40, and NfL were negatively associated with eGFR.Of all the factors studied here, eGFR seems to have consistent associations with biomarkers across studies similar to increased age.

Blood Pressure
Self-reported hypertension was associated with higher baseline Aβ42 and Aβ40 and greater decline in the Aβ42/Aβ40 ratio from baseline to end of intervention in our study.In the HABS-HD and ACCORD studies, hypertension was related to NfL (7,18).In SPRINT, greater increases in NfL were seen in the intensive treatment (blood pressure lowering) arm of the study, however the association was attenuated when adjusted for eGFR (5).

Dyslipidemia
We examined potential associations with self-reported high cholesterol and found no significant associations.In the HABS-HD Study, dyslipidemia was related to Aβ42, Aβ40, and NfL; and higher levels of Aβ42/Aβ40 in models that were adjusted for age, sex, and education (7).

History of CVD
A history of CVD at baseline in Look AHEAD, defined as self-reported myocardial infarction, heart bypass surgery, coronary artery bypass graft, carotid endarterectomy, lower leg angioplasty, aortic aneurysm, congestive heart failure and stroke, was associated with a greater increase in Aβ42 and Aβ40 over time.History of CVD was also associated with higher baseline GFAP levels.Syrjanan et al. examined history of stroke and myocardial infarction and found that stroke was related to higher baseline NfL and total tau, while myocardial infarction was only related to total tau (4).
There are several key findings from this exercise.First, blood-based biomarker levels at baseline in our cohort were associated with 15 out of 18 variables, many of which represent conditions that are very common in older adults.Interestingly, in this study of participants with diabetes and overweight or obesity, diabetes treatment was associated with Aβ42, Aβ40, Aβ42/Aβ40, and NfL, while BMI was associated only with GFAP.As we have outlined, these associations are in general agreement with the limited number of studies that have evaluated these biomarkers in other cohorts (4, 5, 7, 16-22, 25, 29, 30).For example, when considering our work and others, we observe general agreement that Aβ42, Aβ40, NfL, and GFAP levels go up with age.There is a strong suggestion that levels of Aβ42, Aβ40, pTau181 are lower among persons with self-reported Black and Hispanic race and ethnicity.Second, biomarker levels were not associated with legacy effects of our ILI intervention.Third, changes in the biomarker levels over time are associated with race (Aβ42, Aβ40, Aβ42/Aβ40, and NfL), smoking (Aβ42, Aβ40), HbA1c levels (Aβ42, Aβ40, pTau181), and history of CVD (Aβ42, Aβ40).Evaluation of variables that are associated with change in biomarker levels over time suggests potential for malleability.Whether this translates into modifiable lifestyle behaviors is currently unknown.
There has been much progress in the technology to evaluate blood-based biomarkers for neurodegenerative disease.As blood samples are more distal than CSF, the assays for blood samples have to be particularly sensitive and are subject to additional complexities inherent in their measurement in blood (31).As the field begins to use these markers as screening tools for trials or as clinical indicators of brain health, it is critically important to understand all the patient/participant characteristics that can influence their measurement.In our study, all the participants have T2D and all of them had overweight or obesity at baseline, both of which are characteristics associated with increased risk of cognitive impairment and dementia in later life.

Limitations
This study is not without limitations.There was no baseline cognitive assessment as it was not a primary focus of the original Look AHEAD clinical trial.Therefore, participants were not excluded based on the presence of cognitive impairment at baseline.Cognitive impairment was not assessed until after the end of the intervention.Nonetheless, our rigorous screening procedures, designed to ascertain whether participants could execute the protocol, would have effectively excluded those with clear impairment.Our self-reported designations of race and ethnicity were not collected in a mutually exclusive fashion, thereby conflating the two.Randomization facilitated comparable demographic and health characteristics across study arms at baseline, and there is no reason to suspect that the two groups would have differed in cognitive performance at baseline either.
There may be other age-related chronic diseases that influence plasma levels of biomarkers for which we did not account.Our findings are perhaps only generalizable to a high-risk subset of the population, i.e., older adults with T2D and overweight/obesity.However, this group represents a growing at-risk portion of the population.While we tried to put our results in context with other studies, we acknowledge that there are differences in study designs including the methods used to ascertain biomarker level, the proportions of people of different races and ethnicities included, and assumptions about cognitive status at baseline.We also acknowledge that while we are looking at differences among important clinical subgroups, we are unable to determine the degree to which these reflect underlying differences in neuropathology.Finally, a strength of the work is the fact that Look AHEAD was a randomized controlled clinical trial, conducted using rigorous methods.

Conclusions
Blood-based biomarkers for neurodegenerative disorders have great potential for identifying individuals at risk for cognitive decline and dementia, however more work needs to be done in this area before they can be used clinically.Much of the work to date in biomarker studies has focused on clinical samples, comparing participants with and without clinically diagnosed dementing illness, often with confirmatory and Aβ and Tau PET imaging.Emerging work similar to results we present suggests that a number of participant or patient characteristics need to be considered when interpreting blood-based biomarkers.Our study of biomarkers in the Look AHEAD cohort reveals a number of participant characteristics that correlate with biomarker levels in a sample of individuals with diabetes and overweight or obesity, which should be considered when developing clinical applications for these biomarkers.

Figure 8 .
Figure 8. NfL levels at baseline and change over 8-13 years

Figure 9 .
Figure 9. GFAP levels at baseline and change over 8-13 years

Table 1 (
continued).Baseline Characteristics of Participants by Randomization Group

Table 2b .
Spearman Correlations between Biomarker Change from Baseline to End of the Intervention

Table 3 .
Spearman Correlations between Biomarkers, Age, BMI, and Diabetes Duration at Baseline, and Change from Baseline to the End of the Intervention

Table 4 .
Aβ42 levels by individual baseline characteristics adjusted for age and gender

Table 5 .
Aβ40 levels by individual baseline characteristics adjusted for age and gender

Table 1 )
. Interactions between intervention arm and age, gender, BMI, history of CVD, and APOE ε4 carrier status were tested.There was a significant interaction between intervention group and age such that those over age 65 at baseline and randomized to DSE had greater increases in NfL than those under age 65.Those over age 65 at baseline and randomized to DSE also had a greater increase in NfL than those in the ILI group who were aged 65 or older.No other interactions by randomization group emerged.

Table 6 .
Aβ42/Aβ40 levels by individual baseline characteristics adjusted for age and gender

Table 8 .
NfL levels by individual baseline characteristics adjusted for age and gender Each categorical group of characteristics was evaluated individually, controlling for age and gender.Significant associations are bolded; p-values are presented for each group.Abbreviations: Aβ, Amyloid beta; APOE ε4, Apolipoprotein E gene, ε4 carrier status; BDI, Beck Depression Inventory; BMI, body mass index; CVD, cardiovascular disease; GFR, glomerular filtration rate; HbA1c, hemoglobin A1c; Hx, history; kg/m², kilogram per meters squared; LDL, low density lipoprotein; LS, least-squares; OM, oral medication; oz, ounces; Trt, treatment; wk, week; w/, with; w/out, without; Yrs, years