Diabetes Mellitus is Associated With Higher Serum Neurofilament Light Chain Levels in the General US Population Free

Abstract

Context

Serum neurofilament light chain (sNfL) levels are biomarkers of neuroaxonal injury in multiple neurological diseases.

Objective

Given the paucity of data on the distribution of sNfL levels in the general population, in the present study we identified predictors of sNfL levels in a community setting and investigated the association between diabetes and sNfL.

Methods

sNfL levels were measured in 2070 people aged 20 to 75 years from the general US population (275 with and 1795 without diabetes) that participated in the 2013-2014 cycle of the National Health and Nutrition Examination Survey. We evaluated the association between diabetes and sNfL levels after adjustment for age, sex, race-ethnicity, alcohol use, and kidney function using a multivariable linear regression model. Cognitive function was evaluated in a subset of participants aged 60 to 75 years using the Consortium to Establish a Registry for Alzheimer's Disease-Word Learning test, the Animal Fluency test, and the Digit Symbol Substitution test.

Results

The weighted prevalence of diabetes was 10.4% (95% CI, 9.0-11.9). In each age stratum, patients with diabetes exhibited higher sNfL levels compared with nondiabetic participants. Age, proportion of males, prevalence of diabetes, and homeostatic model of insulin resistance increased progressively across quartiles of sNfL levels in the overall population, whereas estimated glomerular filtration rate (eGFR) showed an opposite trend. In the multivariable model, age, sex, eGFR, alcohol use and diabetes were significantly associated with sNfL levels. Moreover, higher sNfL levels were associated with worse performance in all 3 cognitive function tests.

Conclusion

Diabetes is associated with higher sNfL. Further large-scale and prospective studies are needed to replicate our results and evaluate the ability of sNfL to predict the incidence of neuropathy and dementia in this patient population.

Neurofilaments are neuron-specific type IV intermediate filament heteropolymers composed of light, medium, and heavy chains (1). Neurofilaments are the dominant proteins of the neural cytoskeleton and are released into the extracellular space following neuroaxonal damage and have thus been proposed as putative biomarkers of neuroaxonal injury in multiple neurological diseases (2, 3).

Neurofilament light chain (NfL) has especially been shown to be a promising biomarker because of its high solubility (4). Until recently, neurofilament studies were limited to the cerebrospinal fluid (CSF), because detection systems were not sensitive enough to quantitate the physiologically lower levels of NfL in the peripheral blood and this restricted clinical applicability. Conversely, to obtain CSF requires lumbar puncture, which is an invasive procedure, requiring stringent indication for diagnostic purposes (5). Importantly, several studies have demonstrated that CSF and serum NfL (sNfL) levels are highly correlated, and this has given a reason to study sNfL in a wide range of neurologic disorders (6). Nonetheless, to correctly interpret sNfL levels in disease states, it is essential to know if and how the concentration of this protein changes with age, sex, and other traits in the general population. Recent studies showed that sNfL increases with normal aging and reduced estimated glomerular filtration rate (eGFR) in the absence of neurological diseases (7–9).

Both patients with type 1 and type 2 diabetes are at higher risk of developing neurological complications (10). Apart from different forms of diabetic neuropathy, including generalized symmetric polyneuropathies and focal and multifocal neuropathies (11), several studies highlighted an association between diabetes and central nervous system diseases. In particular, meta-analyses of both cross-sectional and cohort studies showed that diabetes increases the risk of mild cognitive impairment and dementia (12, 13). It is also a well-defined risk factor for ischemic stroke (14).

Given the absence of population-based studies on the topic, the aim of the present study is to identify the main contributors to sNfL levels in the general population and to evaluate whether patients with diabetes display higher levels, indicative of subclinical neuronal damage. To achieve these goals, we analyzed data obtained from the 2013-2014 cycle of the National Health and Nutrition Examination Survey (NHANES).

Materials and Methods

This is an analysis of data from the 2013-2014 cycle of NHANES, which is conducted in the United States by the National Center for Health Statistics. It is an ongoing cross-sectional complex survey aimed at including individuals representative of the general, noninstitutionalized population of all ages. To this end, it applies a stratified, multistage, clustered probability sampling design with oversampling of non-Hispanic black and Hispanic persons, people with low income and older adults. The survey consists of a structured interview conducted in the home, followed by a standardized health examination that includes a physical examination as well as laboratory tests. Full methodology of data collection is available elsewhere (15). The original survey was approved by the Centers for Disease Control and Prevention Research Ethics Review Board; written informed consent was obtained from all adult participants. The present analysis was deemed exempt by the institutional review board at our institution because the dataset used in the analysis was completely deidentified.

Laboratory Tests and Clinical Data

Participants self-reported age, sex, race-ethnicity (categorized as non-Hispanic White, non-Hispanic Black, Hispanic, or other), education, smoking status, and medical history. Body measurements including height (cm), weight (kg), and waist circumference (cm) were ascertained during the mobile examination center visit; BMI was calculated as weight in kilograms divided by height in meters squared.

Alcohol consumption was estimated based on self-reported data on the amount and frequency of alcohol use within the previous year. It was considered significant if >1 drink per day for women and >2 drinks per day for men on average (16).

Diabetes was defined in accordance with the American Diabetes Association criteria if any of the following conditions were met: (1) A self-reported diagnosis of diabetes; (2) use of antidiabetic drugs; (3) hemoglobin A1c (HbA1c) level ≥6.5% (48 mmol/mol); (4) fasting plasma glucose ≥126 mg/dL; and (5) a random plasma glucose ≥200 mg/dL (17, 18).

Laboratory methods for measurements of HbA1c, glucose, lipid profile, alanine aminotransferase, aspartate transaminase, gamma-glutamyl transferase, platelet count, creatinine, cholesterol, and triglycerides levels are reported in detail elsewhere (19). The eGFR was computed according to the Chronic Kidney Disease Epidemiology Collaboration equation and chronic kidney disease was defined as an eGFR < 60 mL/min/1.73 m². Based on the measured urine albumin to creatinine ratio (UACR) participants were defined as having normo-albuminuria (UACR < 30 mg/g), micro-albuminuria (UACR between 30 and 300 mg/g), or macro-albuminuria (UACR ≥ 300 mg/dL). Diagnoses of heart failure, coronary artery disease, and stroke were based on self-report. Patients reporting either coronary artery disease or stroke were classified as having cardiovascular disease.

Serum Neurofilament Light Chain Measurement

Sera from stored surplus specimens from participants aged 20 to 75 years in a half-sample from NHANES 2013-2014 who consented testing their samples for future research and had stored surplus or pristine serum samples were eligible.

Measurements were performed using a highly sensitive NfL immunoassay that uses acridinium ester (AE) chemiluminescence and paramagnetic particles and may be run on an existing, high-throughput, automated platform (Attelica).

Initially, the sample is incubated with AE-labeled antibodies, which bind to the NfL antigen. Following this step, paramagnetic particles coated with capture antibody are added to the sample, forming complexes of antigen bound to AE-labeled antibodies and paramagnetic particles. Unbound AE-labeled antibodies are then separated and removed, following which acid and base are added to initiate chemiluminescence and light emission is measured. All steps were performed on the fully automated Attelica immunoassay system. The lower limit of quantification (LLOQ) of the assay is 3.9 pg/mL, which was determined by replicate testing (n = 44) of low-concentration NfL samples. The LLOQ was defined as the concentration at which the coefficient of variation was ≤ 20%. For analytes with analytic results below the LLOQ (n = 36 participants), an imputed fill value was placed in the analyte results field equal to the LLOQ divided by the √2. No participants had values above the upper limit of quantification (500 pg/mL).

Cognitive Function

Three cognitive function tests were performed in participants aged 60 years or older. These tests were originally included in the NHANES survey as they are valuable instruments to estimate the prevalence of cognitive dysfunction in the general population as well as to examine its association with the many medical conditions and risk factors measured during the NHANES examination.

The Consortium to Establish a Registry for Alzheimer's Disease Word Learning (CERAD W-L) test assesses immediate and delayed learning ability for new verbal information (20, 21). The CERAD W-L consists of 3 consecutive learning trials and a delayed recall. For the 3 learning trials, participants were instructed to read aloud 10 unrelated words. Immediately following the presentation of the words, participants recalled as many words as possible. The delayed recall occurred approximately 10 minutes after the start of the word learning trials. The maximum score on each trial is 10; the maximum score for the total word list is 40 (sum of the 3 trials plus the delayed recall).

The Animal Fluency Test examines verbal category fluency, a component of executive function, as well as other functions such as semantic memory and processing speed (22). Participants were asked to name as many animals as possible in 1 minute with a point given for each named animal.

The Digit Symbol Substitution Test, a global measure of brain health, relies on processing speed, visual scanning, sustained attention, and short-term memory (23). The test is conducted using a paper form with a key at the top containing 9 numbers paired with distinct symbols. Participants had 2 minutes to copy the corresponding symbols in the 133 boxes that adjoin the numbers.

Statistical Analysis

All analyses were conducted using Stata version 16 (StataCorp, College Station, Texas, USA), accounting for the complex survey design of NHANES. We used appropriate weighting for each analysis, as suggested by the National Center for Health Statistics to obtain estimates that were generalizable to the US population. Data are expressed as number and weighted proportions for categorical variables and as weighted means (SE) for continuous variables. Participants’ characteristics according to sNfL quartiles were compared using linear regression for continuous variables and the design-adjusted Rao–Scott χ² test for categorical variables.

The association between diabetes and sNfL levels was modeled using multivariate linear regression analysis. sNfL levels were log-transformed to achieve a normal distribution. The choice of covariates to include in the model was based on data from previous studies showing that higher age, lower kidney function, male sex, and history of stroke were associated with higher sNfL levels. To evaluate whether race-ethnicity affects sNfL levels, it was included in the model. A 2-tailed value of P < 0.05 was considered statistically significant.

Results

Features of the Study Population

The study population consisted in 1795 participants without diabetes and 275 patients with diabetes (weighted prevalence 10.4%; 95% CI, 9.0-11.9). The distribution of sNfL across different ages was modeled. The resulting sNfL percentiles in the entire population, and separately, in patients with and without diabetes are presented in Table 1. For each age class, participants with diabetes showed higher mean sNfL levels compared with their nondiabetic counterparts.

Clinical and biochemical features of the entire population divided by quartiles of sNfL are shown in Table 2. Participants with higher sNfL levels were older, more frequently male of non-Hispanic White ethnicity, had higher blood pressure, and with a lower proportion of never-smokers. They also had lower eGFR and higher HbA1c levels. Most comorbid conditions were more prevalent in quartile 4 compared with quartile 1, including diabetes, cardiovascular disease, stroke, and albuminuria.

sNfL and Cognitive Performance

The association between sNfL levels and cognitive performance was evaluated in a subgroup of 521 participants aged 60 to 75 years with available data on both variables. The weighted prevalence of diabetes in this subgroup was 21.8% (95% CI, 18.1-26.0). Mean values on the CERAD, animal fluency, and digit symbol score tests across quartiles of sNfL levels are shown in Table 3. For all tests, a progressive decrease in the participants score was identified with increasing sNfL levels. This trend remained significant after adjustment for age, sex, BMI, and educational status.

Independent Predictors of sNfL Levels in the General Population

The association between diabetes and sNfL levels was modeled using multivariate linear regression analysis. Results are shown in Table 4. Independently of other factors, higher age, presence of diabetes, and heavy alcohol consumption were positively associated with sNfL, whereas an independent, inverse relationship was identified for eGFR and female sex. Finally, no independent association was identified for race-ethnicity and a history of stroke.

Discussion

In the present study, performed on a large and representative sample of the multiethnic US population, we made a series of observations. First, we showed that in each age class, patients with diabetes are characterized by higher sNfL levels compared with their nondiabetic counterparts. Second, we explored the association between sNfL levels and cognitive performance (as estimated by 3 well-validated tests) in a community setting, showing a significant, indirect relationship, which persisted after adjustment for age, sex, BMI, and educational status. Third, to provide data to interpret sNfL levels in different disease states, we studied which features had an impact on sNfL. We identified age, eGFR, sex, alcohol consumption, and diabetes as independent predictors.

Although several studies investigated sNfL levels in patients with different neurological disorders, this is to our knowledge the largest study reporting the distribution of sNfL levels in a truly representative sample of the general population with a wide age distribution.

The percentile distribution in our sample is generally in line with the one reported by Disanto et al, who measured sNfL levels in 254 healthy subjects aged 18 to 70 years from Switzerland (9). As shown in that and other studies, we also confirm that sNfL levels increase progressively with increasing age, even in the absence of neurological disorders. The large study population recruited in the 2013-2014 cycle enabled us to perform multivariable analysis to investigate the role of different variables on sNfL. Apart from age, we confirmed previous observations from Akamine et al (24). and Koini et al (7), who showed that renal function is an important contributor in sNfL levels in smaller samples of older individuals. These results were confirmed in more recent studies in both Swiss (25) and Belgian cohorts (26).

Although the mechanisms underlying this association have not been elucidated, several hypotheses exist. sNfL may be cleared by the kidneys. This hypothesis suggests that there is a risk of overestimating the extent of neuroaxonal damage among adults with low renal function. Another possible explanation is related to the decline in different kidney functions, such as production of erythropoietin and activation of vitamin D. Interestingly, both hormones are reported to have neuroprotective effects (27, 28).

Notably, we show that the association between diabetes and sNfL remains significant after taking these aspects into account, a finding that was not previously reported. As in the case of renal function, many factors might explain it.

Neuropathy is a well-known microvascular complication of diabetes and is present in a large proportion of patients, even though estimates vary substantially by method of detection (29). Because several studies showed that sNfL levels are elevated in patients with peripheral neuropathy of different origins (30, 31), presence of this diabetic complication might account, to some extent, for the observed association.

Furthermore, a link between diabetes and neuronal degeneration in the central nervous system has been described (32). Indeed, although the exact connection between Alzheimer's disease and diabetes is still in debate, studies have shown that poorly controlled blood sugar may increase the risk of developing Alzheimer's (33).

This relationship is so strong that some have called Alzheimer's disease “diabetes of the brain” or “type 3 diabetes” (34). Given that sNfL is significantly associated with cognitive function as demonstrated in the present study, its measurement in patients with diabetes might help identify patients at risk for neuropathy and dementia.

Our study has several strengths. It is a large study performed in an unselected sample of US adults with and without diabetes, including both sexes and participants of different ethnic background. The high number of participants included yielded high statistical power to perform sensitivity analyses and evaluate the impact of several predictors in multivariable models. Being based on NHANES data, our results have a high degree of external validity because the purpose of the survey is to be representative of the overall US population. Acquisition of clinical, laboratory, and anthropometric data was standardized and homogenous. Finally, sNfL levels were measured with a highly reproducible method.

Several limitations should also be acknowledged. First, the cross-sectional nature of the study enables us to identify associations with prevalent, rather than incident disease. Second, although cognitive function tests were performed in a subgroup of subjects, the cohort has not been thoroughly studied from a neurological point of view, both in terms of physical findings and imaging techniques. Related to this point, no objective measurement of sensorimotor neuropathy was available in the dataset, preventing us from evaluating to which degree neuropathy might account for the observed associations. Additional limitations include lack of adjustment for some possible confounders (such as BMI, educational status, blood pressure, years since diabetes diagnosis, and HbA1c levels) and the use of only 3 specific cognitive function tests, but not other well-established function tests (eg, Mini-Mental State Examination, the Montreal Cognitive Assessment Test).

Finally, despite adjustment for several variables, the possibility of residual confounding cannot be completely excluded.

In conclusion, in this large, population-based cross-sectional study, we show that patients with diabetes are characterized by higher sNfL levels compared with their nondiabetic counterparts. Moreover, sNfL is strongly associated with cognitive performance (as estimated by 3 well-validated tests) in a community setting. Finally, the relationship between diabetes and sNfL remained significant after adjustment for several covariates. Further large-scale and prospective studies are needed to replicate our results and evaluate the ability of sNfL to predict the incidence of neuropathy and dementia in patients with and without diabetes.

Acknowledgments

None.

Author Contributions

S.C. and G.P. designed the study, wrote, reviewed, and edited the manuscript. S.C. researched and analyzed data. E.M., E.B., R.C., S.P., and F.Z. participated in manuscript writing and revision. All authors approved the final version of the manuscript to be published. S.C. is the guarantor of this work.

Disclosures

No potential conflicts of interest related to this article were reported.

Data Availability

Some or all data generated or analyzed during this study are included in this published article or in the data repositories listed in the References. NHANES data are available publically at https://wwwn.cdc.gov/nchs/nhanes/Default.aspx

References

Abbreviations

 
• AE

  acridinium ester

 
• BMI

  body mass index

 
• CERAD

  Consortium to Establish a Registry for Alzheimer's Disease

 
• CSF

  cerebrospinal fluid

 
• eGFR

  estimated glomerular filtration rate

 
• HbA1c

  hemoglobin A1c

 
• LLOQ

  lower limit of quantification

 
• NfL

  neurofilament light chain

 
• NHANES

  National Health and Nutrition Examination Survey

 
• sNfL

  serum neurofilament light chain

 
• UACR

  urine albumin to creatinine ratio