Introduction

Diabetic retinopathy (DR) is a leading cause of blindnessin adults and a major microvascular complication of dia-betes [1]. Because DR often remains asymptomatic until advanced, vision-threatening stages, proactive surveillance is essential to avoid delayed diagnosis of complications such as macular edema or neovascularization. Since struc-tured screening programs can prevent up to 98% of severe vision loss through early detection, they remain a corner-stone of modern diabetes management [1, 2].

    The pathogenesis of DR is multifactorial, with prevalence and severity strongly correlated with diabetes duration [3]. Alongside established modifiable risk factors such as poor glycemic control, hypertension, and dyslipidemia, recent research has increasingly focused on the clinical impact of long-term glycemic variability, measured via serial HbA1c, as a potential predictor for the disease [4–6].

    Glycemic variability is a heterogeneous cluster of gly-cemic dysregulations defined by either short- or long-term fluctuations of glucose levels [7–9]. The long-term glycemic variability quantifies glycemic fluctuations over extended periods, typically months or years. It is commonly assessed through visit-to-visit changes in glycemic markers, includ-ing serial measurements of HbA1c, fasting plasma glucose, and postprandial glucose. Metrics such as the standard devi-ation (SD) and the CV are frequently used to calculate this long-term variability [6, 10, 11].

    It is now understood that people with diabetes may exhibit significant long-term glycemic variability even with similar HbA1c values. This visit-to-visit variability has emerged as an important predictor for adverse outcomes in diabe-tes, including micro- and macro-vascular complications and mortality [6, 10–16]. Interestingly, recent systematic reviews and meta-analyses showed that HbA1c variability is positively associated with the risk of developing microvas-cular complications and cardiovascular diseases, even inde-pendently of mean HbA1c [6, 16]. Furthermore, the impact of HbA1c variability in individuals with type 2 diabetes on the development of vascular complications appears to be comparable to that of major established risk factors, includ-ing age, male sex, and a history of hypoglycemia. This sug-gests that HbA1c variability may exert a greater influence on the disease progression than the HbA1c levels themselves [16]. With regard to the predictive role of HbA1c on thedevelopment of diabetic retinopathy, emerging evidence from observational studies suggests a link between glyce-mic variability and DR, with the association appearing more consistent in type 1 than in type 2 diabetes [6, 7, 12, 17–19], although its definitive role remains to be fully elucidated.

    Therefore, taking advantage of a joint eye screening pro-gram at a Tertiary Diabetes Center, the present study aims to investigate the independent association between long-term glycemic variability—specifically measured as HbA1c CV over a 5-year period—and the presence of DR in a real-world cohort of adults with type 1 and type 2 diabetes. This approach allows us to clarify the role of HbA1c CV as a potential metric for refined risk stratification in clinical practice.

Subjects, materials and methods Study design and population

This was a monocentric, retrospective study carried out at the tertiary Diabetes Center of the Unit of Endocrinology and Metabolic Diseases of University of Campania “Luigi Vanvitelli” (Naples, Italy). The project was developed within a joint DR screening program in collaboration with the University Hospital’s Ophthalmology Unit. The study protocol was conducted in conformity to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) Statement [20], and the corresponding checklist is reported in the Online Only Supplemental Material.

    We consecutively enrolled all adults with a confirmed diagnosis of type 1 or type 2 diabetes from June 2023 to June 2025 who have a minimum continuous follow-up of at least 5 years at the Diabetes Center and did not met the following exclusion criteria: 1) diabetes diagnosis not clear  nor confirmed, 2) acute or chronic conditions (assessed with an accurate anamnesis and laboratory tests) that could influ-ence glucose variability, 3) incomplete clinical records, 4) follow-up discontinuity during the observation period; 5) fewer than 10 HbA1c measurements over the preceding five years, or 6) a previous diagnosis of DR. All participants underwent an ophthalmological evaluation that included dilated fundus examination and color fundus photogra-phy, and HbA1c measurements to calculate the variability and retrospectively collected from the medical records of the previous 5-years follow-up, as summarized in Supple-mentary Figure S1. The study was approved by the Medical Ethics Committee of University of Campania “Luigi Vanvi-telli” (Prot. 0003239/I, 01/02/2023) and complies with the Declaration of Helsinki and the International Conference on Harmonization/Good Clinical Practice Guidelines. All participants signed informed consent before the enrollment.

Clinical variables

All clinical and laboratory data were collected for each par-ticipant during the recruitment visit, which coincided with the ophthalmological screening. This baseline assessment included age, sex, type and duration of diabetes, current medication, and smoking status. Anthropometric measure-ments, including body weight and height, were used to cal-culate the body mass index (BMI). Overweight and obesity were defined according to the World Health Organization (WHO) criteria [21]. Blood pressure was measured accord-ing to standard procedures, and hypertension was defined as an average blood pressure≥140/90 mmHg or the cur-rent use of antihypertensive medication. Fasting plasma glucose, HbA1c, and full lipid profile (total, HDL, and LDL cholesterol, and triglycerides) were collected from medical records exhibited by patients. Hyperlipidemia was diag-nosed in the presence of LDL cholesterol>160 mg/dL, HDL cholesterol<40 mg/dL for men or<50 mg/dL for women, or triglycerides>150 mg/dL, or current use of lipid-lowering therapy. Data regarding the presence of atherosclerotic dis-ease and other diabetes-related micro and macrovascular complications were also recorded.

Assessment of glycemic variability

Long-term glycemic variability was assessed using the CV of serial HbA1c measurements, a well-established metric for visit-to-visit fluctuations. [11]. For each participant, two HbA1c measurements per year from the five-year follow-up period preceding the DR assessment were retrospectively collected from their electronic medical records. These val-ues were compiled into a dedicated internal database by study physicians. The CV was calculated for each patient using the following formula: CV (% ) = (SD/mean HbA1c) × 100, where SD corresponds to standard deviation [9]. A CV value equal or superior to 5% was used as the threshold to iden-tify the participants as having high long-term glycemic vari-ability, a cut-off supported by evidence linking it to adverse clinical outcomes [12].

DR assessment

For each enrolled patient, dilated fundus examination was performed, and color fundus photographs were acquired in both eyes using a digital non‐mydriatic camera (Cen-terVue® DRS Automatic Retinal Camera) during the one-time recruitment visit. DR was retrospectively graded by two independent experienced ophthalmologist (C.G. and C.I.) basing on the International Clinical Diabetic Reti-nopathy Disease Severity Scale [22] as follows: no apparent retinopathy; mild non-proliferative DR (NPDR), featuring microaneurysms only; moderate NPDR, with microaneu-rysms, dot and blot hemorrhages, hard exudates, or cot-ton wool spots but less than severe NPDR; severe NPDR, characterized by minimum 20 intraretinal hemorrhages in each of the 4 quadrants, venous beading in 2 quadrants, or intraretinal microvascular abnormalities (IRMA) in 1 quad-rant; and proliferative DR (PDR), featured by findings of

Statistical analysis

The primary objective of this study was to investigate the independent association between long-term glycemic vari-ability —calculated as the HbA1c CV over the five-year follow-up period—and the presence of diabetic retinopa-thy. As a secondary outcome, we assessed the incidence of newly diagnosed DR within the cohort of patients through the systematic eye screening program. The sample size was originally calculated to estimate the prevalence of diabetic retinopathy in our population with sufficient precision. Based on an expected prevalence of 22% [2], a minimum sample size of 264 participants was required to achieve a 95% confidence interval with a 5% margin of error. The final study cohort included 379 participants, exceeding this minimum requirement. This sample size provided sufficient statistical power to perform the multivariable logistic regres-sion analysis, maintaining an adequate ratio of events (new DR diagnoses) per predictor variable, in line with current recommendations for robust clinical association studies.

     Descriptive statistics were used to summarize the baseline characteristics of the study population. The Kol-mogorov–Smirnov test was used to analyze the distribution of the continuous variables. Data were presented as median and interquartile ranges (IQ) for continuous variables and as frequencies and percentages for categorical variables. The cohort was then stratified based on HbA1c CV into two groups: high variability (CV≥5%) and low variability (CV<5%). Differences between groups were analyzed using the Mann–Whitney rank sum test for continuous variables and the Chi-Square test for categorical variables, respec-tively. A binary logistic regression model was employed to evaluate the association between the independent variable (HbA1c CV) and the dependent variable (incidence of dia-betic retinopathy). To control for confounding, the model was adjusted for covariates known to be associated with both dependent and independent variables, including dia-betes duration, age, hypertension status, current smoking habit, LDL cholesterol, total cholesterol, and BMI). The predicted probabilities were calculated using the standard logistic regression formula: P(Y = 1|X) = e (β0 + β1X) + e (β0 + β1X)

Results

Study cohort characteristics

A total of 574 patients attending the Diabetes Center received the fundus examination in the eye screening pro-gram for diabetic retinopathy. Of these, 195 participants were excluded from the final analysis because they lacked a sufficient number of serial HbA1c measurements to calcu-late glycemic variability; therefore, the final study cohort, comprised 379 participants. The demographic and clinical characteristics of the study population are summarized in Table 1. The study population was equally distributed for sex, with 165 females (43.5%) and 214 males (46.5%), and the median age was 60.0 years old. The majority of patients had type 2 diabetes (60.7%), with a median diabetes dura-tion of 12.0 years. The median BMI resulted 27.4 kg/m2 , with 36.4% of the study population being overweight/obese. HbA1c was 7.0% (53 mmol/mol), and the median HbA1c

CV was also 7.0%. The overall prevalence of microvas-cular complications was 19.0%, with diabetic neuropathy being the most common (10.0%), whereas macrovascular complications were present in 70 subjects (18.5%) (Supple-mental Figure S2). Regarding diabetes treatment, among participants with type 1 diabetes the most common regimen (47%) was continuous subcutaneous insulin infusion (CSII) combined with continuous glucose monitoring (CGM), fol-lowed by multiple daily insulin injections (MDI) and CGM (36.7%) (Supplemental Figure S3). In participants with type 2 diabetes, metformin monotherapy was the most frequent treatment (35.4%), followed by metformin plus injective drugs (27.6%), and metformin plus oral drugs (16.8%) (Supplemental Figure S4).

Incidence and characteristics of DR

The screening program identified a new diagnosis of dia-betic retinopathy (DR) in 47 participants (12.4% of the total cohort). Among these newly diagnosed cases, the majority presented with mild non-proliferative DR (NPDR) (72.3%), followed by moderate NPDR (21.2%). Severe NPDR and proliferative DR were less frequent (2.3% and 4.2%, respec-tively) (Table 1, Fig. 1). The incidence of new DR diagno-sis according to the HbA1c CV was described in Table 2. The 273 (72%) participants with high glycemic variabil-ity (HbA1c CV≥5%) had a significantly longer diabetes duration [median (IQ), 13.0 (6.8–20.0) vs 9.0 (4.0–15.0)years, P=0.002] and lower total cholesterol levels [155.0

(133.0–181.0) vs 167.0 (143.0–196.0), P=0.040] compared to the 106 subjects with low glycemic variability (HbA1c CV<5%). The two groups were similar for the other explored clinical parameters. Although the overall incidenceof newly diagnosed DR did not differ significantly between the high- glycemic variability and low- glycemic variability groups (10.6% vs. 17.0%, P=0.131), the incidence of mild NPDR was significantly lower in the high-glycemic vari-ability group (7.0% vs. 14.2%, P=0.046).

Association between HbA1c CV and diabetic retinopathy

In the binary logistic regression analysis, after adjusting for potential confounders, a higher HbA1c CV was indepen-dently associated with an increasing likelihood of having DR (β coefficient=0.643, P=0.042). Other factors significantly associated with DR were longer diabetes duration (β coef-ficient=0.051, P<0.001), older age (β coefficient=0.041, P<0.001), and serum lipid levels (LDL and the total cho-lesterol) (Table 3).

Discussion

To the best of our knowledge, this is the first study to inves-tigate the incidence of DR within the framework of a joint screening initiative at a Tertiary Diabetes Center in South-ern Italy. A unique aspect of our work was the simultaneous investigation of the association with long-term glucose vari-ability, calculated over a minimum follow-up of 5 years, and the incidence of DR in this cohort.

    Our study focused on a group of adults with DM, with the majority having type 2 diabetes, and characterized by a fair glyco-metabolic control. The study population reported a low incidence of diabetes-related microvascu-lar complications, with diabetic neuropathy being the most representative. Our screening program revealed a DR inci-dence of 12.4%, with most cases classified as mild NPDR. This rate is considerably lower than the global prevalence of approximately 22% and the Italian published preva- lence, which ranges from 15.5 to 27.6% [2, 23]. This find-ing characterizes the clinical context of our cohort, since we hypothesized that this difference reflects the efficacy of the comprehensive and specialized care provided at our ter-tiary-level diabetes center. This interpretation is supported by the cohort’s overall fair glyco-metabolic profile, includ-ing well-managed dyslipidemia despite a high prevalence of hypercholesterolemia. The observed low incidence of DR is therefore consistent with the well-established evidence that structured, long-term follow-up and intensive glycemic control significantly reduce the risk of microvascular com-plications [24–27].

    Our assessment of long-term glycemic variability was based on serial HbA1c measurements collected two times a year over five years. We observed that the median HbA1c CV in our cohort exceeded the proposed target of 5% despite fair overall glycemic control as indicated by the median HbA1c level [12]. A potential explanation for this elevated glycemic variability is that our observation period partially overlapped with the COVID-19 pandemic. It is well docu-mented that public health restrictions during this time led to significant disruptions in routine diabetes care, includ-ing discontinuation of face-to-face visits, which may have impacted glycemic stability [28].

    A key finding is that after adjusting for traditional risk factors, higher long-term glycemic variability, measured by HbA1c CV, was independently associated with the presence of DR in our cohort. This result is particularly notewor-thy when considering our initial bivariate analysis, which revealed a counterintuitive higher incidence of new DR cases (specifically mild NPDR) in the low- glycemic vari-ability group (CV<5%). This apparent paradox is under-scored by the fact that the low- glycemic variability group also had a significantly shorter diabetes duration, a factor that is typically protective against the development of reti-nopathy. This discrepancy strongly suggests that the unad-justed analysis is influenced by significant confounding. Therefore, the multivariable logistic regression model—by adjusting for duration, age, and other covariates—was essential to isolate the true, independent detrimental effect of high HbA1c CV [15, 19, 29].

    The association between high glycemic variability and DR is supported by several proposed pathogenetic mecha-nisms, as summarized by Kilpatrick [29]. First, high gly-cemic variability may act through a mechanism analogous to the “metabolic memory”, whereby transient but repeated hyperglycemic episodes induce persistent vascular stress and epigenetic changes [30, 31]. Second, because the risk of microvascular complication rises exponentially with increasing HbA1c, individuals with high variability of HbA1c spend more time at the upper, more damaging end of their HbA1c range, which disproportionately elevates their overall risk [32]. Third, the phenomenon of “early worsen ing” of DR, after rapid improvement in glycemic control, suggests that large glucose fluctuations themselves can be detrimental to the retinal vasculature [33].

    Nevertheless, we acknowledge that the role of long-term glycemic variability as an independent risk factor for DR remains a subject of debate. The findings in literature are inconsistent, with several studies suggesting the role of HbA1c CV as an independent risk factor for the occurrence of DR [15, 19, 29]. On the other hand, there is evidence from observational studies that the association between visit-to-visit HbA1c variability and DR is attenuated or becomes non-significant after adjusting for mean HbA1c levels, particularly in cohorts with type 2 diabetes [20, 34]. Our findings contribute to the growing body of evidence supporting the clinical relevance of monitoring long-term glycemic variability as a potentially modifiable risk fac-tor. We suggest that integrating HbA1c CV into routine care could provide clinicians with a refined tool for risk strati-fication, potentially guiding decisions on the frequency of ophthalmological follow-up and the intensification of glu-cose-lowering therapy. In our study, high variability was defined using a threshold of CV equal or higher than 5%, calculated over a 5-year data window with a minimum of 10 measurements. While these parameters proved effective for identifying patients at higher risk of DR in our cohort,we acknowledge that it is currently not possible to estab-lish a universal clinical protocol. Moreover, since the retro-spective calculation of HbA1c CV may be time-consuming in a typical follow-up visit, clinicians might prioritize this metric for a specifical subset of those patients experienc-ing oscillations between periods of glycemic stability and decompensation. Towards a clinical application, further evidence is required to standardize the optimal timing of assessment, define a validated ‘high-risk’ cutoff, and deter-mine specific management changes—such as earlier referral or personalized screening intervals—to be implemented in routine clinical practice.

    This study has many strengths, including 1) a well-sized cohort, with a similar proportion of female and male par-ticipants; 2) the inclusion of participants who had an history of at least five-years follow-up at our Diabetes Unit, that allowed us to assess the HbA1c CV based on a considerable number of HbA1c measurements; 3) the comprehensive col-lection of clinical and biochemical data, that allowed us to adjust the analyses for a wide range of potential confounders.

    Our findings should be interpreted also considering sev-eral limitations. The primary limitation is the retrospective study design, which allows us to identify an association, but does not permit the inference of causality between high HbA1c CV and DR. Furthermore, the design precludes a formal evaluation of the predictive value of glycemic vari-ability for and the future progression of DR. Second, the biochemical data, including the HbA1c measurements, were sourced from routine clinical records rather than a central laboratory. This introduces the possibility of inter-assay variability, although it also enhances the real-world applica- bility of our findings.

    In conclusion, in a cohort of individuals with both type 1 and type 2 diabetes with a low incidence of DR, long-term glycemic variability, estimated as HbA1c CV, is indepen-dently associated with an increased risk of this complica-tion. Our findings come from an eye screening program for early DR detection and highlight the importance of integrat-ing glycemic variability assessment into clinical practice for enhanced risk stratification. Further prospective studies are warranted to confirm the predictive role of long-term gly-cemic variability in the development and progression of the microvascular diabetes-related complications, particularly DR, also analyzing possible associations with recognized ocular biomarkers.

Supplementary Information The online version contains supplementary material available at https://doi.org/10.1007/s00592-0 26-02711-7.

Acknowledgements We thank all investigators, study teams and patients for participants in the trial.

Author contributions M.I.M., C.G., S.R., F.S., K.E. were involved in the conceptualization, methodology, project administration, investiga-tion, validation, visualization, writing—review & editing. S.A. and C.I. were involved in data curation, formal analysis and writing—original draft. N.D.M., M.L., L.S. and G.B. were involved in the data curation and formal analysis. All authors edited, critically reviewed,and approved the final version of the manuscript. K.E. is the supervisor and 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.

Funding Open access funding provided by Università degli Studi della Campania Luigi Vanvitelli within the CRUI-CARE Agreement. This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Data availability All data generated or analyzed during this study are included in the published article (and its online supplementary files).

Declarations

Conflict of interest MIM has given lectures for Eli Lilly and Sanofi. KE received a consultancy fee from Eli Lilly and GSK and has given lectures from Eli Lilly, Sanofi, Novo Nordisk, Roche and Bayer. No other potential conflicts of interest relevant to this article were report-ed

Ethical approval Study protocol was approved by the Medical Eth-ics Committee of University of Campania “Luigi Vanvitelli” (Prot. 0003239/I, 01/02/2023). All procedures performed in studies involv-ing human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. Written informed consent was obtained from all par-ticipants included in the study.

Open Access

This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unlessindicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.o rg/licenses/by/4.0/.

This article is excerpted from the 《Acta Diabetologica》 by Wound World.