Longitudinal Change in Serum Neurofilament Light Chain in Type 2 Diabetes and Early Diabetic Polyneuropathy: ADDITION-Denmark Free

Diabetic polyneuropathy (DPN) is a common complication of types 1 and 2 diabetes, affecting up to 50% of people with diabetes during the course of their disease (1). Continuous evaluation of the peripheral nerve status in people with diabetes is therefore important for preventive purposes. Nerve function is usually assessed by subjective or semiobjective bedside tests such as vibration and pinprick sensation (2,3). Objective measures, like nerve conduction studies (NCS) or intraepidermal nerve fiber densities, enable continuous quantification of peripheral nerve function and morphology, but are generally time consuming and require specific technical skills, limiting their use on a larger scale in routine diabetes control and as clinical outcome measures (2). Therefore, accurate objective disease progression tools are needed in order to follow peripheral nerve function and to allow timely preventative action against DPN development.

The neuron-specific axonal protein neurofilament light chain (NfL) has been widely studied in the field of neurology (4). As a biomarker for disorders of the peripheral nervous system, it has been shown to correlate with disease severity, clinical outcome, and therapy response in various peripheral neuropathies (5–9). However, little evidence exists on the role of NfL in DPN or diabetes in general, and longitudinal levels of plasma or serum NfL (s-NfL) in neuropathies of any etiology are scarcely explored (6,9–13). As a structural protein, NfL has a strictly controlled turnover rate, and intraindividual variation of NfL levels is small in people without neurological disease, strengthening the potential of NfL as a marker for individual longitudinal disease monitoring (14–17). The first and, so far, only longitudinal observations of NfL in diabetes are from the Treatment Options for Type 2 Diabetes in Adolescents and Youth (TODAY) cohort of youth-onset type 2 diabetes, where a steeper increase in NfL levels was found in people developing DPN (11). Conversely, longitudinal data from patients with hereditary polyneuropathy show no change in NfL levels over time (12). Hence, the potential of NfL in longitudinal monitoring of polyneuropathy remains unclear.

In this retrospective longitudinal nested case-control study of the Danish arm of the Anglo-Danish-Dutch Study of Intensive Treatment in People with Screen-Detected Diabetes in Primary Care (ADDITION) cohort, we studied the longitudinal development of s-NfL in people with screen-detected type 2 diabetes with and without DPN and compared these changes with a large reference database of s-NfL in neurologically healthy individuals (18). Additionally, we explored the predictive potential of s-NfL as a biomarker in DPN and assessed the association between changes in s-NfL levels and changes in levels of covariates associated with DPN and s-NfL.

Here we present results from 1) nested longitudinal case-control analysis based on longitudinal data from an unselected subgroup of 200 participants of the ADDITION-Denmark (ADDITION-DK) cohort who attended the clinical 10-year follow-up examination at the Aarhus study site and 2) comparison of s-NfL levels from these participants to a large Swiss reference database of s-NfL in healthy individuals (18,19).

ADDITION-DK is the Danish arm of the ADDITION-Europe trial, a cluster-randomized, pragmatic trial designed to investigate the effect of intensive multifactorial treatment in people with screen-detected type 2 diabetes in primary care (20). After the end of the 5-year trial period in 2009 (referred to as the 5-year follow-up), participants were followed observationally. An extensive clinical examination for neuropathy was conducted at the 10-year follow-up, between 2015 and 2016. ADDITION and its outcomes have previously been described in detail (19–23).

Participant characteristics were collected according to standardized study protocols at both the 5- and 10-year follow-ups (20). Records of alcohol consumption, smoking habits, and previous chemotherapy (only at 10 years) were self-reported. Records of prescribed medication were provided by the general practitioners. Diagnoses of central nervous system (CNS) disease, potentially contributing to s-NfL levels, were obtained from Danish national registers. The diagnoses were chosen to cover the most common CNS disorders, and the diagnosis codes used were selected based on previous validation. See Supplementary Table 1 for details. Cross-sectional characteristics and s-NfL for the 10-year follow-up of the current subgroup have previously been published (24).

The presence of DPN was defined cross-sectionally according to the Toronto criteria for a confirmed diagnosis of DPN at the clinical 10-year follow-up examination (25). Participants were required to have abnormal NCS in combination with at least one symptom and/or symmetrical sign indicative of DPN to be defined as having confirmed DPN (24). Participants with subclinical DPN (defined according to Toronto criteria as abnormal NCS without symptoms or signs) were pooled together with the DPN-free participants. In brief, symptoms of neuropathy were collected and evaluated by participant interviews using the Michigan Neuropathy Screening Instrument questionnaire (MNSIQ) and the Douleur Neuropathique en 4 Questions questionnaire and were considered present when reported in the feet and/or legs (26,27). Clinical signs of neuropathy were evaluated by vibration, pinprick, and light touch sensation on the dorsal aspect of the great toe, and by ankle reflexes and position sensation of the interphalangeal joint of the great toe as described in the MNSI physical assessment and the Utah Early Neuropathy Scale (26,28). The evaluated signs were required to be present bilaterally in the feet. Available DPN measures from both 5- and 10-year follow-ups included the MNSIQ and vibration detection thresholds (VDT), evaluated as percentiles according to standard protocol with the CASE IV system (WR Medical Electronics Co.), which is an automated device for measuring sensory thresholds (26,29).

NCS were performed according to standard protocols by a neurophysiologist using Keypoint.NET equipment (Dantec, Skovlunde, Denmark) (30). Briefly, bilateral sural sensory NCS, right-sided peroneal and tibial motor NCS, and right-sided median sensory and motor NCS were performed as previously described (31). The NCS were compared against in-house reference material yielding age- and height-matched z-scores of the assessed NCS parameters. NCS were considered abnormal when the sum z-score of the six most DPN-sensitive NCS parameters was >2.0 (criteria number 8 by Dyck et al. [32]). An additional axonal sum z-score was constructed to reflect axonal damage and consisted of bilateral sural amplitudes, right median sensory nerve action potential amplitude, and tibial, peroneal, and median compound muscle action potential amplitudes (24).

Biobank serum samples were collected by trained laboratory personnel at the ADDITION-DK 5- and 10-year follow-ups and stored at −80°C until s-NfL analysis in May 2021 (10-year follow-up) and May 2022 (5-year follow-up). s-NfL was measured by a certified laboratory technician, blinded to all participant information, at the Department of Clinical Biochemistry, Aarhus University Hospital, Aarhus, Denmark (international laboratory accreditation no. DS/EN ISO/IEC 15189) by the NF-light assay using the ultrasensitive Simoa HD-1 platform (Quanterix, Billerica, MA) according to previously described methods (33). The intra-assay and interassay coefficients of variation in the laboratory were 5% and 10% (human serum pools around 30.0 ng/L), respectively.

In the absence of a control group, we converted s-NfL levels to z-scores based on a Swiss reference database of 4,532 control individuals (aged 18–81 years, with no documented CNS disease), from four population-based studies and control groups from multiple sclerosis studies in Europe and the U.S. (18). Details on the reference database, providing age- and BMI-corrected NfL z-scores, are given in Benkert et al. (18).

The original ADDITION-DK study was approved by the Committee on Health Research Ethics in the Central Denmark Region (file nos. 20000183 and 1-10-72-63-15) and the Danish Data Protection Agency (file no. 2005-57-0002, ID185), and was conducted in accordance with the principles of the Declaration of Helsinki, version 1996. All study participants gave written informed consent for storage of blood samples in the biobank. Renewed ethical approval was obtained prior to analysis of biobank samples (file no. 1-10-72-316-20).

Participant characteristics at the 5- and 10-year follow-ups are reported by DPN status at the 10-year follow-up and for the total study cohort. Data are presented as medians and interquartile range (IQR) for continuous variables and as frequencies and proportions for categorical variables. Comparison of covariates and unadjusted comparison of s-NfL and NfL z scores between participants with DPN (+DPN) and without DPN (−DPN), as determined by the clinical 10-year follow-up examination, was done using Pearson χ² test, Fisher exact test, McNemar test, and unpaired and paired Wilcoxon Mann-Whitney tests as appropriate.

Group-wise comparisons of the longitudinal change in s-NfL (Δs-NfL) and NfL z-scores (ΔNfL-z) from the 5- to the 10-year follow-up between +DPN and −DPN participants were done using Wilcoxon Mann-Whitney tests. Absolute s-NfL values were transformed to z-scores based on the NfL reference database (18). Multiple linear regression models were used to estimate the adjusted group differences in Δs-NfL (log[10-year s-NfL] − log[5-year s-NfL]) and ΔNfL-z (expressed as the absolute change). Absolute s-NfL was log-transformed to improve linearity and normal distribution in all linear regression models.

The risk of DPN by 5-year s-NfL and by yearly change in s-NfL was estimated with multivariable logistic regression models. To examine the potential additional value of s-NfL in the prediction of DPN, receiver operating characteristic (ROC) curves derived from these models, with and without the inclusion of 5-year s-NfL as an independent variable, were compared by the DeLong test. Additionally, ROC analysis was performed for s-NfL, the MNSIQ, and VDT at both 5- and 10-year follow-ups against a DPN diagnosis at the 10-year follow-up. Multiple linear regression was used for prediction of nerve integrity (as reflected by the two NCS sum z-scores) by 5-year s-NfL. Associations between Δs-NfL and change in levels of covariates associated with DPN or s-NfL were also estimated by multiple linear regression. The risk factor changes were defined as absolute differences with 10-year levels as the minuend, except for estimated glomerular filtration rate (eGFR) where the minuend was the 5-year level. Changes in HbA_1c and in urine albumin-creatinine ratio were log-transformed (log[10-year covariate] − log[5-year covariate]).

All models were adjusted using a predefined step-by-step approach for selected covariates including sex, randomization group, age, and BMI. Moreover, eGFR was included in models of Δs-NfL, and HbA_1c was included in predictive models of DPN risk and NCS sum z-scores. Additionally, adjustments for 5-year s-NfL and the time difference between the 5- and 10-year follow-ups was made in models of Δs-NfL. Other time-dependent variables in models of Δs-NfL were from the 10-year follow-up. Models predicting DPN risk or NCS sum z-scores only included covariate levels from the 5-year follow-up, except when DPN risk was modeled by yearly change in s-NfL.

Sensitivity analyses with inclusion of participants with subclinical DPN in the +DPN group, and exclusion of participants with previous registry-based CNS diagnoses, were performed. For all analyses, the significance level was set to 0.05. All statistical analyses were performed in R (version 4.2.3 [2023-03-15]).

Of the initial subgroup of 200 participants examined clinically for DPN at the 10-year follow-up at the Aarhus study site, 193 (96.5%) had enough biobank serum from both 5- and 10-year follow-ups for s-NfL analysis. NCS were complete in 178 (89.0%) participants who were included in the analyses requiring DPN status. Details of the participant flow from baseline to the 10-year follow-up are illustrated in Supplementary Fig. 1. In total, 39 (21.9%) participants had confirmed DPN and 10 (5.6%) had subclinical DPN out of the 178 with complete NCS at the 10-year follow-up. Participants developing DPN at the 10-year follow-up were heavier and had a larger waist circumference and higher systolic blood pressure at 5 years than participants who stayed free of DPN (Table 1). Differences in weight and waist circumference persisted from 5 to 10 years, but differences in HbA_1c and MNSIQ scores were present only at 10 years. Characteristics for all 200 participants are provided in Supplementary Table 2.

Figure 1 illustrates the absolute s-NfL levels at the 5- and 10-year follow-ups in +DPN and −DPN participants on a group level and as individual trajectories. s-NfL increased in both +DPN (from 11.3 ng/L [IQR 9.54; 15.6] to 18.8 ng/L [IQR 14.4; 27.9], P < 0.001) and −DPN participants (from 10.2 ng/L [IQR 7.49; 13.7] to 15.4 ng/L [IQR 11.7; 20.1], P < 0.001). There was a significant difference in s-NfL between +DPN and −DPN participants at both 5 and 10 years (P = 0.044 for comparison between 5-year levels; 10-year data reported earlier [24]). Differences at 5 years, however, became nonsignificant with adjustment (8.1% [95% CI −5.8; 24.0] higher s-NfL in +DPN compared with −DPN). Absolute median Δs-NfL was higher in +DPN than −DPN participants (7.36 ng/L [IQR 3.78; 12.9] vs. 4.69 ng/L [IQR 2.77; 8.07]). NfL z-scores also increased between the 5- and the 10-year follow-ups in both +DPN (from 0.08 SD [IQR −0.95; 0.64] to 1.02 SD [IQR 0.36, 1.74], P < 0.001) and −DPN participants (from −0.48 SD [IQR −1.23; 0.26] to 0.50 SD [IQR −0.25; 1.08], P < 0.001 [Supplementary Fig. 2]) corresponding to a median ΔNfL-z of 1.15 SD (IQR 0.54; 1.38) in +DPN participants and 0.84 SD (IQR 0.23; 1.50) in −DPN participants. Raw estimates of ΔNfL-z were not significantly different between the groups. The adjusted group differences in Δs-NfL and ΔNfL-z from the 5- to the 10-year follow-up are presented in Table 2. Overall, +DPN participants had a steeper increase in s-NfL, with a larger Δs-NfL from the 5- to the 10-year follow-up corresponding to a 17.4%, or 0.31 SD, higher increase in s-NfL and NfL z scores, respectively, compared with −DPN participants (P < 0.01 and P = 0.03). Post hoc analysis grouping +DPN and subclinical DPN participants as “severe DPN” (N = 13) and “DPN” (N = 36) (definition: original NCS sum z-score > 3.3 SD, corresponding to the 75th percentile in the post hoc analysis group), showed a larger Δs-NfL and ΔNfL-z in participants with severe DPN compared with DPN (data not shown).

Prediction of DPN status by 5-year s-NfL and by averaged yearly s-NfL change is presented in Table 3. Five-year s-NfL was not predictive of a DPN diagnosis at 10 years when adjustments for age and other covariates were implemented. This was reflected in the corresponding ROC curves, where removal of 5-year s-NfL from the prediction model for DPN had no impact on the area under the curve (AUC) (AUC for model including s-NfL, age, sex, HbA_1c, BMI, and randomization group as covariates: 0.72 [95% CI 0.62; 0.81]; AUC for the same model without s-NfL: 0.71 [95% CI 0.62: 0.80], P = 0.72). Higher yearly s-NfL increase was associated with a higher risk of DPN throughout adjustment (Table 3). Higher s-NfL at the 5-year follow-up was associated with higher continuous original and axonal NCS sum z-scores at the 10-year follow-up (Supplementary Table 3).

ROC curve analysis of s-NfL, the MNSIQ score, and VDT for the 5- and 10-year follow-ups against a DPN diagnosis at the 10-year follow-up is presented in Supplementary Fig. 3 and Supplementary Table 4 (10-year data for s-NfL were reported earlier [24]). AUCs for s-NfL were not inferior to AUCs for the MNSIQ score or VDT.

The level of Δs-NfL was associated with the level of change in weight, BMI, total cholesterol, and eGFR in exploratory analysis (Supplementary Table 5). Change in HbA_1c was not associated with Δs-NfL. Sensitivity analysis with inclusion of participants with subclinical DPN in the +DPN group rendered the absolute s-NfL change nonsignificant but had no effect on the adjusted difference in Δs-NfL between +DPN and −DPN. Exclusion of participants with registered CNS diagnoses had no effect on the interpretation of the results.

This is the first study to investigate longitudinal changes in s-NfL levels in adults with type 2 diabetes. We demonstrate that s-NfL increases over time both in people with type 2 diabetes who develop DPN and in those who do not develop DPN at levels above the age-induced s-NfL increase seen in healthy people. The change in s-NfL was significantly greater in people who developed DPN than in people who stayed free of DPN. s-NfL levels 5 years prior to DPN evaluation were not predictive of a DPN diagnosis, but they were prospectively associated with a larger degree of nerve damage as reflected by NCS sum z-scores. Additionally, higher yearly increase in s-NfL significantly increased the risk of DPN. Taken together, our findings suggest that s-NfL trajectories may have biomarker value in DPN monitoring. Further work is, however, needed to determine the clinical value of s-NfL trajectories in DPN management.

The observation that s-NfL increased more over time in people who developed DPN compared with people who did not develop DPN is supported by the only previous study of longitudinal s-NfL levels in diabetes and DPN, which is based on 12 years of follow-up of the TODAY cohort of youth-onset type 2 diabetes (11). The TODAY study reported higher cross-sectional s-NfL levels in +DPN compared with −DPN participants and a steeper s-NfL increase over time in people who developed DPN. There are, however, certain differences that complicate direct comparison between the two studies. In the TODAY study, DPN diagnosis was based on the MNSI physical examination, while, in the current study, it was based on electrophysiological criteria in combination with symptoms and/or signs of DPN (25,26). The clinical characteristics of TODAY participants are different from the ADDITION cohort, as participants in the former were diagnosed with type 2 diabetes in their teenage years, as opposed to the screen-detected type 2 diabetes of ADDITION participants, diagnosed at an age of around 60 years. Finally, the TODAY study was based on a selected subset of +DPN and −DPN participants, whereas the participants in our study represent an unselected subgroup, defined only be their continued participation in ADDITION-DK up to the 10-year follow-up. Despite these differences, our results point in the same direction.

Transformation of our participants’ s-NfL levels to z-scores based on existing reference material provides the possibility of comparing the observed s-NfL increase with the expected age-related s-NfL increase. NfL z-scores are expected to stay unchanged over time, as age and BMI are accounted for in the reference material (18). NfL z-scores, however, increased in both +DPN and −DPN participants, and, as for s-NfL, the increase was larger in +DPN than in −DPN, although the difference between the groups was limited. This may be due to the combined effect of the large s-NfL variation in the elderly and the fact that our cohort consist of people with less severe DPN. Larger differences in ΔNfL-z might be expected in cohorts of people with more advanced DPN, as indicated by our post hoc analysis of Δs-NfL and ΔNfL-z in DPN compared with severe DPN. Additionally, the s-NfL reference database we used contains a small number of individuals with diabetes (N = 177 of 4,533), potentially diminishing our NfL z-scores.

The fact that NfL z-scores also increase in −DPN participants suggests that factors other than DPN, such as unmeasured small nerve fiber damage or subclinical CNS degeneration, may influence s-NfL levels in our cohort. Although there is a body of evidence coupling diabetes with cognitive dysfunction and structural changes of the brain, population-based s-NfL levels in diabetes have not been investigated (34,35). While some studies have found associations between NfL levels and a diabetes diagnosis, details about the diabetes type and duration, glycemic control, or DPN presence have not been accounted for (18,36,37). Our observation of increasing s-NfL above what is explained by normal aging in −DPN individuals is hence noteworthy and suggests that diabetes, or factors associated with diabetes, influences s-NfL even in relatively early and well-controlled type 2 diabetes.

We did not find cross-sectional s-NfL 5 years prior to DPN evaluation to be prospectively associated with the dichotomous DPN diagnosis used in our study. This is in contrast to findings from the TODAY cohort, where s-NfL 8 years prior to DPN diagnosis differed between +DPN and −DPN and increased the risk for a DPN diagnosis (11). This underlines the difficulty of using cross-sectional measures, both for diagnostic purposes and for prediction, especially in the elderly where the inherently large interindividual variation in s-NfL is considerable (18). Despite this, the discriminative ability of s-NfL for a DPN diagnosis was not inferior to the discriminative ability of the MNSIQ or VDT, which are widely used clinical measures for DPN screening (2). This is noteworthy, considering that there is an element of incorporation bias in the ROC analysis of the MNSIQ and VDT. Regardless of the challenges of interpreting cross-sectional s-NfL levels, knowing an individual’s s-NfL trajectory may have prognostic value, as modeled yearly s-NfL change significantly increased the risk of DPN. Although not a truly prospective association, as the second time point of s-NfL measurement coincides with DPN diagnosis, this suggests that s-NfL trajectories may have potential for the monitoring of DPN risk. This is strengthened by the fact that higher s-NfL at 5 years was associated with a higher degree of nerve impairment at 10 years.

Our observation of associations between s-NfL and weight, BMI, cholesterol levels, and eGFR is in line with previous cross-sectional studies and strengthens these findings by demonstrating longitudinal associations (38). Interestingly, no association was seen between the change in HbA_1c and the change in s-NfL, in contrast to previous reports (11,39). This could be because of the high degree of glucose control in the ADDITION-DK cohort. Besides these initial associations, further insights into the dynamics of NfL release from damaged neurons as well as the turnover and excretion of NfL are needed to correctly interpret s-NfL levels in both healthy individuals and people with diabetes.

The strength of this study is in the use of longitudinal data from an unselected subgroup of a cohort of people with screen-detected type 2 diabetes to illustrate s-NfL changes in an early stage of type 2 diabetes and DPN. We use a robust definition for confirmed DPN requiring NCS abnormality (25). Additionally, we report both absolute s-NfL values and age- and BMI-standardized NfL z-scores to provide comparison with a healthy control group (18). Our study also has some limitations. First, we cannot exclude that some of the participants already had DPN at the 5-year follow-up, as we do not know the exact time of DPN onset due to a lack of longitudinal NCS measures, meaning that presence of DPN was only determined cross-sectionally at the 10-year follow-up. Because of this and the fact that the second s-NfL measurement coincides with the DPN evaluation, we cannot state that we have investigated s-NfL trajectories prior to DPN onset. Yet, because of the early and well-controlled diabetes detected by screening in the present cohort, a high prevalence of DPN prior to the 10-year follow-up is less likely. Additionally, exclusion of participants with a 5-year MNSIQ score >3 (N = 22), indicating DPN, had no effect on the group difference in s-NfL change or on the estimates of DPN risk (data not shown). Second, because of the unselected nature of our study population, only a relatively small number of participants had DPN, which, in combination with the mild and early nature of DPN and the slowly progressing course of DPN, may be a reason for not finding a larger difference in s-NfL change between +DPN and −DPN participants. Healthy survivor bias may accentuate this effect, as nonattenders at the ADDITION-DK 10-year follow-up were older and had more diabetes complications (40). Yet, the estimates of both s-NfL and NfL z-score changes show a consistent difference between +DPN and −DPN participants, which warrants closer investigation in prospective studies before any conclusion about the clinical utility of s-NfL can be made. Third, because of the lack of objective measures of small nerve fiber function, we may have missed DPN cases with isolated small fiber dysfunction. Small fibers, however, contain less NfL than large fibers because of their smaller diameter (16). Lastly, CNS disorders, including subclinical or undiagnosed CNS pathology, remain a potential source of s-NfL. Exclusion of participants with existing or previous CNS diagnoses, however, had no impact on the results, and the relationship between s-NfL and NCS points toward a relevance of peripherally derived s-NfL (6,24). Although subject to possible underdiagnosis, the registry-based CNS diagnoses we used have overall moderate to high validity (Supplementary Table 1).

In conclusion, this study suggests that s-NfL trajectories differ slightly between people who develop DPN and people who do not develop DPN and implies a possible biomarker value of s-NfL trajectories in DPN. Individual s-NfL trajectories coupled with repeated objective evaluation of peripheral nerve status from the time of diabetes diagnosis are needed to clarify whether s-NfL has clinical value in DPN management.

Acknowledgments. The authors express their deepest gratitude to Doctor Pascal Benkert, Department of Clinical Research, University of Basel, Basel, Switzerland, for providing access to the NfL reference database. Furthermore, the authors thank laboratory technicians Charlotte Nørby Pedersen, Katrine Bremer, and Arnaq Hammeken at the Department of Clinical Biochemistry, Aarhus University Hospital, Aarhus, Denmark, for their skillful laboratory analysis of the ADDITION biobank blood samples. Lastly, the authors are grateful for the support and collaboration of the ADDITION-Denmark steering committee.

Funding and Duality of Interest. This work was supported by research grants from the Danish Diabetes and Endocrine Academy, which is funded by the Novo Nordisk Fonden (grant no. NNF17SA0031406), and from Steno Diabetes Center Aarhus, Aarhus University Hospital. The work is also part of the International Diabetic Neuropathy Consortium research program, which is supported by a Novo Nordisk Foundation Challenge Program grant (grant no. NNF14OC0011633). Additional support was received from the A.P. Møller Foundation, the Riisfort Foundation, and the Department of Clinical Biochemistry, Aarhus University Hospital. ADDITION-Denmark was funded by the National Health Services in the former counties of Copenhagen, Aarhus, Ringkøbing, Ribe and the county of Southern Jutland in Denmark, the Danish Council for Strategic Research, the Danish Research Foundation for General Practice, the Novo Nordisk Foundation, the Danish Center for Evaluation and Health Technology Assessment, the Danish Foundation of the National Board of Health, the Danish Medical Research Council, the Aarhus University Research Foundation, Novo Nordisk Scandinavia AB, Novo Nordisk UK, ASTRA Denmark, Pfizer Denmark, GlaxoSmithKline Pharma Denmark, Servier Denmark A/S, and HemoCue Denmark A/S. M.C. has received speaker/expert testimony honoraria from Novo Nordisk Denmark, Boehringer Ingelheim Denmark, and Abbott Rapid Diagnostics. J.K. received speaker fees, research support, and travel support from, and/or served on advisory boards of, Swiss MS Society, Swiss National Research Foundation (320030_189140/1), University of Basel, Progressive MS Alliance, Alnylam, Bayer, Biogen, Bristol-Myers Squibb, Celgene, Immunic, Merck, Neurogenesis, Novartis, Octave Bioscience, Quanterix, Roche, Sanofi, and Stata DX. D.R.W. reports a grant from the European Foundation for the Study of Diabetes and Sanofi, consulting fees from the European Commission, and meeting support from the Novo Nordisk Foundation. D.R.W. additionally holds stock in Novo Nordisk A/S and is a member of the external advisory board for the Maastricht Study. T.S.J. is funded by the Novo Nordisk Foundation (grant no. NNF14OC0011633) and is part of the Advisory Committee for the Danish Diabetes Association. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. L.L.M. carried out the statistical analysis, wrote the manuscript, collected the NfL-data, and contributed to the design of the study. S.T.A., T.P., D.R.W., and T.S.J. contributed to the design of the study and the drafting of the manuscript. S.T.A., T.P., C.V.B.H., E.S., D.R.W., and T.S.J. initiated the idea of the study. S.T.A., L.B., M.C., and D.R.W. designed ADDITION-DK and/or collected the data. M.A.K. and H.T. performed the neurophysiological examinations, and H.T. provided input for the utilization of these. D.R.W. provided input on the statistical analysis. J.K. is the initiator, collector, and guarantor of the NfL reference database. All authors reviewed/edited the manuscript and approved the final version for publication. L.L.M. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this study were presented at the Peripheral Nerve Society 2023 Annual Meeting, Copenhagen, Denmark, 17–20 June 2023, and at the 33rd Annual Meeting of the Diabetic Neuropathy Study Group, Thessaloniki, Greece, 28 September to 1 October 2023.

Handling Editors. The journal editors responsible for overseeing the review of the manuscript were Elizabeth Selvin and Rodica Pop-Busui.