Your Blood Sugar Isn’t Diabetic Yet — But Your Arteries May Already Know
A landmark proteomic study of 389 apparently healthy adults identifies novel protein biomarkers connecting prediabetes to silent arterial disease — before any symptoms appear. Three proteins emerge as particularly compelling candidates for early cardiovascular risk detection.
Prediabetes is often described as a warning sign — a moment when blood sugar has crept above normal but before the threshold of a formal diabetes diagnosis. What is less appreciated is how early vascular damage may already be underway during this supposedly pre-disease window. A new Polish study using cutting-edge proteomic technology and advanced multivariate modelling has gone looking for the molecular fingerprints of this silent convergence — and found something genuinely surprising: the proteins linking prediabetes to arterial wall thickening include markers associated with neurological function and ageing, not just the inflammation and lipid pathways that dominate conventional cardiovascular thinking.
Published in Cardiovascular Diabetology, this study from researchers at the Medical University of Białystok profiled over 1,000 proteins simultaneously in 389 apparently healthy adults — excluding anyone with existing diabetes, cardiovascular disease, or lipid-lowering therapy — and used sophisticated machine learning techniques to identify which protein signatures distinguished people with prediabetes from those without, and which of those proteins were independently linked to early arterial damage.
The Problem: A Silent Convergence Nobody Is Screening For
Cardiovascular disease and type 2 diabetes are the two most prevalent chronic diseases in the developed world — and they are deeply intertwined. People with diabetes have two to four times the cardiovascular risk of those without it. But the damage begins well before a diabetes diagnosis. Prediabetes — defined as impaired fasting glucose, impaired glucose tolerance, or HbA1c between 5.7% and 6.4% — is already associated with early vascular changes including thickened arterial walls, increased arterial stiffness, and the earliest formation of atherosclerotic plaques.
The challenge is that these changes are entirely asymptomatic. No chest pain, no breathlessness, no warning. The person feels well. Standard blood tests show borderline numbers. And yet, at the level of the vessel wall, a disease process has already begun.
In this study of 389 apparently healthy Polish adults aged 35–70, 232 (60%) had prediabetes and 169 (43%) had preclinical atherosclerosis detected on carotid ultrasound. These were people with no known diabetes, no diagnosed cardiovascular disease, and no medications that would influence lipids or blood sugar. The prevalence of these “silent” conditions in an ostensibly healthy population is striking — and it is the central motivation for developing better early detection tools.
Who Was Studied and How
The study drew participants from the Białystok PLUS cohort — a carefully characterised, population-based sample of urban adults in eastern Poland. The researchers applied stringent exclusion criteria to ensure they were studying the metabolic-vascular interface in as clean a sample as possible: no type 1 or type 2 diabetes, no inflammatory diseases, no cancer, no prior heart attack or stroke, no steroidal medications, no cholesterol-lowering treatment.
- Prediabetes classification: Oral glucose tolerance test (75g glucose), fasting glucose, 2-hour post-load glucose, and HbA1c — using American Diabetes Association 2025 criteria
- Preclinical atherosclerosis: Carotid ultrasound measuring intima-media thickness (IMT) and plaque presence; IMT >0.9mm was considered subclinical atherosclerotic disease
- Arterial stiffness: Carotid-femoral pulse wave velocity (PWV) and augmentation index (AIx) — both established markers of vascular ageing and early cardiovascular risk
- Proteomic profiling: 1,050 proteins measured simultaneously using Olink Reveal proximity extension assay with next-generation sequencing readout — one of the most comprehensive commercial proteomics platforms available
- Statistical modelling: Block sparse partial least squares discriminant analysis (block-sPLS-DA) — a machine learning approach that integrates clinical and proteomic data simultaneously to identify the best-discriminating features
The Two-Layer Discovery: What the Model Found
The core analytical innovation is a multiblock model that simultaneously analyses clinical data and protein data as separate “blocks,” allowing the researchers to identify what is already captured by conventional clinical measures versus what proteomics adds independently. The model produced two components — and they told very different stories.
This component largely recapitulates what standard clinical markers already capture. It aligns with BMI, LDL cholesterol, and male sex — the established cardiometabolic risk factors.
Key proteins here include CDCP1, HSPB6, FURIN, IFI30, and IL-6 — proteins connected to metabolic stress, inflammation, and lipid biology.
This component adds limited new information beyond what a standard cardiovascular risk assessment would reveal.
This is where the study breaks new ground. Component 2 is dominated by the proteomic block, with minimal contribution from clinical variables. It captures information about prediabetes that standard clinical measures completely miss.
Key proteins include EDA2R, C16orf89, VGF, MOG, and NEFL — many with neurological or structural roles rather than traditional cardiovascular ones.
These proteins discriminate prediabetic from non-prediabetic individuals independent of BMI, lipids, blood pressure, and kidney function.
The block weights confirm this story quantitatively: Component 1 had roughly equal clinical (0.357) and proteomic (0.338) contributions, while Component 2 was dominated by proteomics (0.163) with near-negligible clinical contribution (0.039). Component 2 is, in effect, a signal that exists in the blood but is invisible to standard clinical assessment.
The Three Key Proteins — What They Are and Why They Matter
After applying multivariable regression analyses and false discovery rate correction, three proteins from Component 2 emerged with the most robust and statistically durable associations with early arterial damage:
A member of the tumour necrosis factor receptor superfamily. EDA2R is a transcriptional target of the tumour suppressor p53, and its activation promotes apoptosis (programmed cell death). It has been identified as a hallmark of cellular ageing and mediates pro-inflammatory responses.
EDA2R is elevated in obesity, insulin resistance, and type 2 diabetes. In the heart, it contributes to ischaemia/reperfusion injury — and experimental knockdown reduces cardiac damage. It has also been shown to predict incident coronary artery disease across multiple time points in large biobank studies.
An actin filament nucleator highly enriched in smooth muscle tissues — including major arteries. Genetic variation at the LMOD1 locus is one of 64 loci identified in genome-wide association studies for coronary artery disease risk, specifically through arterial wall mechanisms.
Evidence suggests LMOD1 has insulin-responsive phosphorylation sites linked to glucose uptake, and it shows strong genetic associations with glycaemic traits. One study found reduced LMOD1 expression in individuals with prediabetes after an exercise intervention.
A relatively understudied protein whose function remains incompletely characterised — which makes its emergence here particularly intriguing. In the correlation heatmap, C16orf89 was associated with a more favourable vascular profile (higher HDL and GFR, lower AIx and IMT at lower expression levels).
Its robust association with carotid IMT, independent of all clinical covariates and surviving FDR correction, positions it as a genuinely novel candidate biomarker whose biology in the context of prediabetes and vascular disease deserves dedicated investigation.
The Unexpected Finding: A Neurological Signal in Cardiovascular Disease
Perhaps the most intellectually striking finding is the enrichment analysis result. When the researchers mapped Component 2 proteins onto a protein-protein interaction database and tested for biological pathway enrichment, the result was not inflammation, not lipid metabolism, not the usual cardiovascular suspects.
The most significantly enriched pathway was synapse maturation (GO:0060074, FDR = 0.045), driven by three proteins: BCAN (brevican), SEZ6L2 (seizure 6-like protein 2), and NEFL (neurofilament light chain). This points toward a neurovascular signature — connections between neuronal processes and vascular biology that may be mechanistically relevant to prediabetes.
Neurofilament light chain (NEFL) — a structural protein normally found in neuronal axons — showed a nominal positive association with atherosclerotic plaque presence. While this association did not survive FDR correction, its appearance is not surprising in context. Prior studies have found elevated NEFL in people with prediabetes, proposed it as an early marker of prediabetic peripheral neuropathy, and associated it with greater odds of cardiovascular disease in population-based analyses. Longitudinal data from the PESA cohort shows that subclinical atherosclerosis progression is associated with accelerated decline in cerebral glucose metabolism — with circulating NEFL mediating approximately 20% of this pathway. This suggests NEFL may reflect a broader neurovascular risk signal connecting glucose dysregulation, arterial disease, and brain health simultaneously.
What the Correlation Patterns Reveal
The heatmap analysis adds important texture to the protein findings. Two distinct protein modules emerge with opposing vascular associations:
The existence of a “favourable” protein module is an important and underappreciated finding. It raises the possibility that some individuals with prediabetes may have a proteomic profile that partially counteracts vascular risk — and that understanding what drives that protective signature could eventually inform therapeutic strategies.
The Study’s Strengths — and Its Honest Limitations
The researchers are commendably explicit about what their study can and cannot claim. Strengths include the carefully phenotyped, medication-free cohort that eliminates major confounders, the use of one of the most comprehensive proteomics platforms commercially available, and the application of a multiblock machine learning framework that genuinely separates proteomic signal from clinical noise.
Limitations are equally important to understand. This is a cross-sectional study — it captures a single snapshot in time, meaning causality cannot be established. The cohort is ethnically homogeneous (Central European urban population), limiting generalisability to other populations. No independent external validation cohort exists yet. Prediabetes subtypes (IFG, IGT, elevated HbA1c) were combined into a single group, which may mask differences between subtypes. And the Olink platform, while extensively validated, is an affinity-based approach — independent confirmation using targeted assays like ELISA has not been performed.
The protein biomarkers identified here — EDA2R, LMOD1, C16orf89 — are hypothesis-generating candidates that require replication in independent, ethnically diverse cohorts before any clinical application can be considered. The cross-sectional design means the study cannot determine whether elevated protein expression precedes vascular damage or reflects it. Longitudinal follow-up tracking progression to overt diabetes and clinical cardiovascular events is the essential next step.
Why This Research Direction Matters
The broader significance of this study extends beyond the specific proteins identified. It demonstrates a proof-of-concept that is genuinely important for preventive cardiology: that the molecular processes linking prediabetes to arterial disease are detectable in blood before any clinical symptoms appear, and that these processes include pathways not currently captured by any standard risk assessment tool.
If even one of these protein markers can be validated in larger prospective studies, the implications are substantial. A blood test that flags early arterial damage in people with prediabetes — before standard risk scores trigger concern — could enable far earlier intervention. Lifestyle modification, metabolic optimisation, and targeted monitoring could begin at the right moment, before the disease process has progressed to a clinically harder-to-reverse stage.
Key Takeaways from the Research
- Prediabetes is already a vascular disease: In this cohort of apparently healthy adults, prediabetes was significantly associated with measurable arterial wall thickening and stiffness — the silent precursors of heart attack and stroke
- Proteomics sees what clinical measures miss: The Component 2 proteomic signature distinguished prediabetic from non-prediabetic individuals independent of BMI, lipids, blood pressure, and kidney function — information invisible to standard tests
- Three proteins stand out: EDA2R (arterial IMT, ageing/inflammation), LMOD1 (arterial stiffness, smooth muscle/glycaemic genetics), and C16orf89 (IMT, mechanism unclear) showed FDR-robust associations with vascular damage after full multivariable adjustment
- The biology is unexpectedly neurological: Enrichment analysis points to synapse maturation and neurovascular pathways — including NEFL (neurofilament light chain) — as candidate mechanisms linking early glucose dysregulation to arterial disease
- This is discovery science, not clinical application — yet: Independent replication in larger, diverse cohorts is essential before any of these biomarkers can inform clinical practice. Longitudinal studies tracking progression to diabetes and cardiovascular events are urgently needed
- The protective protein module deserves equal attention: The identification of a protein signature associated with better vascular health in prediabetic individuals raises the possibility of endogenous protective mechanisms that could eventually be targeted therapeutically
Prediabetes is not a waiting room. It is an active biological process. The more precisely we can characterise that process at the molecular level — and the earlier we can detect its vascular consequences — the better our chances of interrupting it before the damage becomes irreversible.
