Capillaroscopy and diabetes: microangiopathy, type 2 diabetes, and microvascular risk

Why capillaroscopy is of interest in diabetes

The pathophysiological rationale is strong. Diabetes damages the endothelium, alters perfusion, and promotes microvascular remodeling, making it reasonable to investigate whether some of that damage can be detected at the nailfold [1, 2, 3, 4].

The distinction from systemic sclerosis is essential. In diabetes, the aim is to identify a set of signs of chronic microvascular damage rather than a unique capillaroscopic pattern. This changes how the test is interpreted: the question is whether the microvascular profile is compatible with a greater burden of microangiopathy, rather than whether a “diabetic pattern” is present [1, 6].

A recent meta-analysis summarizes the current evidence. Across 10 studies and 1,423 participants, it found significant associations between diabetes and tortuosity, abnormal morphology, ectasia, bushy capillaries, hemorrhages, avascular areas, bizarre capillaries, and meandering capillaries. Subgroup analysis suggested more pronounced findings in T2DM than in other settings [1]. This does not make the technique a clinical standard, but supports its consideration as a promising emerging biomarker.

Recurring findings

Although the literature contains some disagreement, several abnormalities recur consistently enough to form a recognizable set of findings [1, 3, 6, 7, 8, 9].

The most frequent abnormalities are tortuosity, ectasia or dilation, hemorrhages, avascular areas, abnormal morphology, and neoangiogenesis. Earlier individual studies already described this combination: in 2017, for example, Maldonado et al reported predominantly tortuous capillaries, avascular areas, and ectasias in people with diabetes, with greater damage in those with more vascular comorbidity [6]. A recent meta-analysis added support by confirming that these associations are not isolated observations [1].

A practical interpretation separates mild findings from more markedly abnormal profiles. Slight isolated tortuosity in a patient with diabetes changes little. In contrast, marked tortuosity combined with repeated hemorrhages, meandering capillaries, new vessel formation, and avascular areas suggests a greater microangiopathic burden [1, 2, 6, 7, 8, 9].

A. New vessel formation

B. Avascular area

C. Hemorrhage

Results from a meta-analysis estimating the increased odds of microvascular abnormalities detectable by capillaroscopy in patients with diabetes [1].

*Bushy capillaries. CI, confidence interval; OR, odds ratio.

Retinopathy, nephropathy, and other complications

The area of greatest clinical interest is currently the relationship between capillaroscopy and established microvascular complications, particularly retinopathy and nephropathy [2, 3, 4, 7, 8, 9].

Diabetic retinopathy

One of the two 2025 meta-analyses cited above examined capillaroscopic findings in this complication. It included 1,349 patients with diabetes, 628 of whom had retinopathy. Patients with retinopathy consistently had larger arteriolar diameters and more bushy capillaries, new vessel formation, giant capillaries, tortuosity, hemorrhages, meandering capillaries, and avascular areas [2]. Some associations were substantial: the odds ratios (ORs) with 95% confidence intervals (CIs) were 8.37 (5.07–13.80) for giant capillaries and 7.92 (2.68–23.38) for avascular areas [2].

These findings align with earlier clinical studies. In 2018, Bakirci et al found more hemorrhages, ectasias, giant capillaries, and neoangiogenesis in patients with T2DM and retinopathy than in those without retinopathy [8]. In a later study, Okabe et al analyzed several variables longitudinally and observed progressive reductions in capillary number, length, and width with increasing retinopathy severity [9]. Together, these studies suggest a real association, but remain insufficient to replace fundus examination or standard ophthalmological screening techniques [2, 8, 9].

Diabetic nephropathy

The literature in this area has grown rapidly in recent years. A 2022 study found more capillary aneurysms, hemorrhages, tortuosity, new vessel formations, and bizarre capillaries in patients with albuminuria than in controls, and also found a correlation between albuminuria and tortuosity [3]. A recent larger cohort confirmed this finding, documenting significant differences according to eGFR and KDIGO categories, with more comma-shaped configurations and a higher proportion of abnormal shapes in groups with worse kidney function [4].

These observations do not establish capillaroscopy as a diagnostic test for retinopathy or nephropathy. They suggest that it may reflect part of the systemic microangiopathic burden and could eventually help improve multimodal risk models [2, 3, 4].

T1DM, T2DM, and prediabetes

Diabetes evidence does not all point in the same direction. It currently appears more robust in T2DM, while findings in T1DM have historically been less consistent [1, 7].

The 2025 meta-analysis found particularly pronounced associations in T2DM [1]. In a notable 2023 study, Lisco et al used computerized videocapillaroscopy in 102 patients with T2DM and observed an association between capillaroscopic abnormalities, clinical profile, and metabolic burden [7]. Earlier studies focused on T1DM did not always demonstrate clear morphological differences from controls, even with established retinopathy or vasculopathy [10].

Prediabetes also deserves attention as an intermediate setting. A study from the previous decade found more capillaroscopic distribution and morphological abnormalities in people with prediabetes than in healthy individuals, although the changes were less marked than in T2DM [11]. This suggests that early microvascular damage may be detectable at the nailfold, but evidence remains too limited to use the technique for population screening.

AI and datasets in diabetes

Diabetes is one of the settings in which combining capillaroscopy and artificial intelligence (AI) may offer the greatest potential, given the large number of patients and quantifiable microvascular outcomes [5].

A study recruiting 120 patients analyzed 5,236 images to develop a convolutional neural network model. Although its ability to predict cardiovascular events, retinopathy, albuminuria, or hypertension was modest, it achieved an area under the receiver operating characteristic curve (AUROC) of 0.84 for identifying diabetes [5]. This does not establish AI-assisted capillaroscopy as a clinical screening tool, but shows that nailfold images contain information that can be analyzed computationally.

Methodologically, diabetes offers both an advantage and a challenge. The potential data volume is enormous, allowing much larger datasets to be built and compared than in many rare rheumatic diseases. However, clinical labels are heterogeneous and depend on composite outcomes, disease duration, treatment, and comorbidities. As in other biomedical applications of AI, quality control and standardization are therefore as important as the algorithm itself [5].

Using the technique without overstating its capabilities

The main risk in diabetes is presenting capillaroscopy as an established screening test for retinopathy or nephropathy. The literature does not support that step [1, 2, 3, 4, 5].

• Use it as a complementary microangiopathy biomarker, not as a replacement for standard tests.
• Describe specific domains and avoid presenting a “diabetic pattern” as a validated construct.
• When relating findings to retinopathy or nephropathy, frame the conclusion as an association rather than a diagnosis.
• Prioritize T2DM with microvascular complications when selecting patients for research or exploratory assessment.
• When using AI, require image quality, prospective design, and external validation before inferring clinical utility.

Short cases

Case 1. T2DM with nonproliferative retinopathy

A patient with longstanding T2DM and retinopathy has marked tortuosity, hemorrhages, and avascular areas. Capillaroscopy is consistent with a high microvascular burden, but does not add more than fundus examination for retinopathy classification. Its value lies in supporting the systemic phenotype and, in cohort studies, improving risk models [2, 8, 9].

Case 2. A T2DM cohort with albuminuria

A research group analyzes tortuosity, abnormal shapes, and capillary density in patients with T2DM and different albuminuria levels. Capillaroscopy can quantify microangiopathy noninvasively and support subgroup comparisons, but interpretation still depends on eGFR, the albumin-to-creatinine ratio, and other conventional clinical variables [3, 4].

FAQ

Glossary

Diabetic microangiopathy

Small-vessel damage associated with chronic hyperglycemia, endothelial inflammation, and vascular remodeling, affecting the retina, kidneys, peripheral nerves, and other tissues.

Bushy capillary

A capillary with an abnormal shape or marked branching, interpreted as a sign of neoangiogenesis or disorganized remodeling.

Prediabetes

A metabolic state preceding established diabetes, in which early signs of microvascular abnormalities may appear.

KDIGO

A chronic kidney risk classification system based on glomerular filtration rate and albuminuria.

Research biomarker

A variable useful for exploring associations or building prognostic models, even without an established clinical role.