Nationwide Analysis of Progressive Kidney Function Decline and Diabetic Macular Edema in Type 2 Diabetes

Abstract

Purpose

To evaluate the impact of renal function on the risk of developing diabetic macular edema (DME) among patients newly diagnosed with type 2 diabetes mellitus (T2DM).

Design

A retrospective cohort study.

Participants

Patients with T2DM without pre-existing ophthalmic diabetic complications, stratified by kidney function (estimated glomerular filtration rate [eGFR]).

Methods

We analyzed electronic health record data from the TriNetX network, including patients diagnosed with T2DM from 2005 to 2025. Patients were grouped based on baseline eGFR levels documented within 6 months of diabetes diagnosis: normal/high (≥90 mL/min), mild chronic kidney disease (CKD; 60–89 mL/min), mild-to-moderate CKD (45–59 mL/min), moderate-to-severe CKD (30–44 mL/min), severe CKD (15–29 mL/min), and end-stage renal disease (ESRD) (<15 mL/min). Propensity score matching balanced covariates including age, sex, race/ethnicity, hemoglobin A1c, hypertension, hyperlipidemia, insulin and oral hypoglycemic agent use, fenofibrate use, prostaglandin analog use, and Diabetes Complications Severity Index components.

Main Outcome Measures

Incidence of DME within 3 years after diabetes diagnosis.

Results

Postmatching, each cohort was balanced in patient characteristics. Compared with patients with normal kidney function, there was a progressively higher risk of DME with declining kidney function: mild CKD (hazard ratio [HR] 1.05, 95% confidence interval [CI] 1.02–1.08), mild-to-moderate CKD (HR 1.41, 95% CI 1.36–1.46), moderate-to-severe CKD (HR 1.78, 95% CI 1.70–1.87), severe CKD (HR 2.35, 95% CI 2.21–2.51), and ESRD (HR 2.53, 95% CI 2.33–2.74). Subgroup analysis restricted to normoalbuminuric patients (urine albumin-to-creatinine ratio ≤30 mg/g) also demonstrated significant associations, highlighting the potential independent effect of declining eGFR on DME risk. Additionally, kidney transplantation among ESRD patients was associated with reduced DME risk (HR 0.65, 95% CI 0.51–0.81).

Conclusions

Our findings reveal a clear, progressive relationship between declining renal function and increased DME risk, independent of albuminuria. These results underscore the need for proactive ophthalmic screening in diabetic patients with impaired renal function and suggest renal improvement may mitigate DME risk.

Financial Disclosure(s)

Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.

Diabetic macular edema (DME) is a major cause of vision loss among individuals with diabetes mellitus (DM), particularly affecting the working-age population.¹ It is characterized by the accumulation of extracellular fluid in the macula due to increased vascular permeability, leading to visual impairment and reduced quality of life. Diabetic macular edema affects approximately 3.8% of individuals with DM over the age of 40 years in the United States, and its prevalence has been rising alongside the global increase in type 2 DM (T2DM) cases.¹ As of 2022, T2DM cases globally amount to an estimated 828 million people, with 445 million (59% of diabetics) being untreated.² Beyond the chronic nature of DME and its impact on visual function, it also bears a significant public health burden, with affected patients requiring significantly higher medical costs and more frequent health care resources compared with patients without DME.³

Current treatment strategies have evolved remarkably over time. The widespread adoption of intravitreal injections of anti-VEGF agents has transformed the management of DME and led to substantial improvements in visual outcomes.⁴ Despite the success of these therapies, a subset of patients continues to experience persistent or refractory disease even after rigorous treatment regimens.⁵ This variability suggests that systemic factors beyond localized retinal pathology may influence DME development and progression. While diabetes duration, hypertension, hyperlipidemia, and hemoglobin A1c levels are well-established risk factors for DME, emerging evidence suggests that systemic factors such as medication use and overall organ health may also contribute to DME pathogenesis.^6,7

Several studies have investigated the relationship between estimated glomerular filtration rate (eGFR) and DME with varying results.7, 8, 9 For example, 1 study found that while decreasing eGFR was associated with DME in age- and sex-adjusted models, this association disappeared after multivariable adjustments.¹⁰ Similarly, a study among Chinese patients reported that impaired renal function was significantly associated with vision-threatening diabetic retinopathy and DME, but the association between eGFR and diabetic retinopathy was only significant in the presence of microalbuminuria.¹¹ These findings reveal a gap in our understanding of the link between renal health and vision loss.

Considering these inconsistencies in the literature, our study uses practice-based clinical data from a nationwide health research network to examine the association between varying degrees of kidney function and the risk of DME among a diverse population of patients with T2DM. Utilizing a large, multicenter dataset, our findings offer greater clarity on how renal impairment, including early-stage disease, influences the development of DME. These insights may inform future research aimed at establishing kidney function metrics as potential indicators of DME risk and prognosis.

Methods

This was a retrospective cohort study utilizing data from TriNetX, a nationwide federated health research network with electronic medical records (EMRs) from 70 health care organizations across the United States. TriNetX provides only aggregate counts and statistical summaries of deidentified information to protect patient health information, allowing it to comply with the Health Insurance Portability and Accountability Act and receive a waiver from the Western Institutional Review Board. This study was conducted in accordance with the principles of the Declaration of Helsinki. Informed consent was not required, as this research involved the analysis of an existing anonymized dataset, ensuring that no identifiable information could be traced back to individual patients. The current study analyzed the TriNetX database for EMRs between February 2005 and February 2025.

Study Sample

The analysis was performed on February 23, 2025. We included patients with T2DM who had no prior history of diabetic ophthalmic complications (eg, diabetic retinopathy, DME). Renal function was determined using the eGFR documented within 6 months of DM diagnosis to accurately reflect baseline kidney function. Patients were grouped by Kidney Disease: Improving Global Outcomes stages based on eGFR: normal or high kidney function (≥90 mL/min) served as the control group; mild chronic kidney disease (CKD) (60–89 mL/min); mild-to-moderate CKD (45–59 mL/min); moderate-to-severe CKD (30–44 mL/min); severe CKD (15–29 mL/min); and end-stage renal disease (ESRD) (<15 mL/min). Estimated glomerular filtration rate was calculated using the Chronic Kidney Disease Epidemiology Collaboration creatinine equation.¹² To ensure adequate ophthalmic surveillance for diabetic eye complications, patients were required to have ≥1 ophthalmology visit within 3 years after DM diagnosis. The patient selection process is detailed in Figure 1.

Figure 1.

Flowchart of the study population. The flowchart illustrates the selection process for the study cohort from the TriNetX US Collaborative Network. The process started with an initial pool of 9 428 960 patients with type 2 diabetes. After applying exclusion criteria for prior diabetic ophthalmic complications (n = 545 522) and requiring ≥3 years of ophthalmic follow-up, a cohort of 571 281 patients remained. Patients were then stratified based on their baseline eGFR and further divided into a normoalbuminuria subgroup. CKD = chronic kidney disease; eGFR = estimated glomerular filtration rate; ESRD = end-stage renal disease; uACR = urine albumin-to-creatinine ratio; T2DM = type 2 diabetes mellitus. ∗Analysis for end stage renal disease in normoalbuminuria group was not conducted due to insufficient event counts.

Subgroup and Sensitivity Analyses

Given prior research suggesting an association of albuminuria and DME, we conducted a subgroup analysis restricted to patients with normal urine albumin levels. For this analysis, we included only patients with a documented urine albumin-to-creatinine ratio <30 mg/g, defined as normal according to Kidney Disease: Improving Global Outcomes guidelines, collected and recorded in the patients EMR within 6 months of diabetes diagnosis, along with their respective eGFR measurement for each study group. This methodology isolated the impact of renal function on DME risk, independent of albuminuria.

Additionally, a sensitivity analysis was conducted comparing patients with ESRD who underwent kidney transplantation to a matched control group of ESRD patients without transplantation to assess the effect of improved renal function after transplantation on DME risk.

Covariates

The TriNetX platform was used to perform 1:1 propensity score matching (PSM) to balance baseline characteristics and comorbidities between groups. Covariates associated with DME development included age, gender, race, ethnicity, hemoglobin A1c (continuous and categorical), hypertension, hyperlipidemia, insulin and oral hypoglycemic agent use, fenofibrate use, and prostaglandin analog use. Additionally, to address diabetes severity, relevant components of the Diabetes Complications Severity Index were matched: diabetic neurological complications (neuropathy), cerebrovascular disease (cerebral infarction and ischemic heart disease), peripheral vascular disease (diabetic circulatory complications and diabetic arthropathy), and metabolic complications (ketoacidosis, hyperosmolarity, and hypoglycemia).¹ Patients with pre-existing diabetic ophthalmic complications were excluded.

Primary Outcome and Statistical Analyses

The primary outcome was the development of DME within 3 years after the index event, defined as the date of diabetes diagnosis. A time-to-event analysis was conducted to estimate the hazard ratio (HR) and 95% confidence interval (CI) for DME development. Patients who did not develop DME by the end of follow-up were right-censored at their last recorded encounter or at the database query end date. Baseline characteristics were summarized as means with standard deviations for continuous variables and as counts with percentages for categorical variables. Propensity score matching results were assessed at each time point to ensure covariate balance, using standardized mean differences as the evaluation metric. A standardized mean difference <0.1 was considered indicative of a well-balanced match between cohorts. A P value <0.05 was considered statistically significant. All coding data are available in Table S1 (available at www.ophthalmologyretina.org).

Results

Primary Analysis

Post-PSM, cohort sizes were matched 1:1 between the study and control groups across all CKD severity levels. The mild CKD group had 134 265 per cohort. The mild-to-moderate and moderate-to-severe CKD groups included 68 165 and 37 854 patients per cohort, respectively. In the severe CKD group, 17 674 patients per cohort were analyzed, and the end-stage group comprised 11 533 per cohort. All baseline characteristics for each CKD severity group, before and after PSM, are detailed in Tables S4–S8 (available at www.ophthalmologyscience.org).

A stepwise increase in the risk of DME was observed with worsening kidney function, as shown in Table 2 and Figure 2. Of patients, 7.42% in the study group developed DME, compared with 7.03% in the control group (HR: 1.05, 95% CI: 1.02–1.08). The risk was higher in mild-to-moderate CKD (9.63% vs. 6.90%; HR: 1.41, 95% CI: 1.36–1.46). In moderate-to-severe CKD, 12.41% of patients developed DME in the study group, compared with 7.19% in the control group (HR: 1.78, 95% CI: 1.70–1.87). Among those with severe CKD, 16.56% were affected, more than double the 7.55% in the control group (HR: 2.35, 95% CI: 2.21–2.51). Finally, in end-stage disease, 17.08% versus 7.17% developed DME (HR: 2.53, 95% CI: 2.33–2.74). These findings underscore a progressive relationship between declining kidney function and DME risk, reinforcing the role of systemic microvascular dysfunction in retinal pathology.

Table 2.

Risk of Diabetic Macular Edema Stratified by Baseline Kidney Function in Patients with Type 2 Diabetes Mellitus

Stages	eGFR Value (ml/min/1.73m2)	CKD Cohort		Control Cohort		Hazard Ratio	95% Confidence Interval
		Total	Events (No. %)	Total	Events (No. %)
G1	Normal or high (eGFR ≥90 mL/min)	Reference	Reference	Reference	Reference	Reference	Reference
G2	Mild (eGFR = 60-89 mL/min)	134 265	9959 (7.42%)	134 265	9434 (7.03%)	1.05	(1.02-1.08)
G3a	Mild to moderate (eGFR = 45–59 mL/min)	68 165	6564 (9.63%)	68 165	4702 (6.90%)	1.41	(1.36-1.46)
G3b	Moderate to severe (eGFR = 30–44 mL/min)	37 854	4697 (12.41%)	37 854	2723 (7.19%)	1.78	(1.70-1.87)
G4	Severe (eGFR = 15–29 mL/min)	17 674	2926 (16.56%)	17 674	1335 (7.55%)	2.35	(2.21-2.51)
G5	End stage (eGFR <15 mL/min)	11 533	1970 (17.08%)	11 533	827 (7.17%)	2.53	(2.33-2.74)

CKD = chronic kidney disease; eGFR = estimated glomerular filtration rate; No. = number.

Figure 2.

Hazard ratios for DME by kidney function categories in patients with type 2 diabetes mellitus. A, Risk of DME stratified by baseline eGFR categories. B, Risk of DME stratified by baseline eGFR categories in patients with normoalbuminuria (urine albumin-to-creatinine ratio ≤30 mg/g). Hazard ratios are presented with 95% CIs; significance is indicated when CI lines do not cross an HR of 1. Kidney function categories are defined per Kidney Disease: Improving Global Outcomes guidelines: mild CKD (eGFR 60–89 mL/min), mild to moderate CKD (eGFR 45–59 mL/min), moderate to severe CKD (eGFR 30–44 mL/min), severe CKD (eGFR 15–29 mL/min), and kidney failure (end-stage renal disease, eGFR <15 mL/min). ∗Analysis for kidney failure in normoalbuminuria group was not conducted due to insufficient event counts. CI = confidence interval; CKD = chronic kidney disease; DME = diabetic macular edema; eGFR = estimated glomerular filtration rate; HR = hazard ratio.

Subgroup and Sensitivity Analyses

In the subgroup analysis restricted to patients with normoalbuminuria, a similar progressive increase in DME risk was observed, as shown in Table 3. Among patients with mild CKD, 6.82% in the study cohort developed DME compared with 6.70% in controls (HR: 1.02, 95% CI: 0.98–1.08). This risk increased in mild-to-moderate CKD, with 8.60% of the study group developing DME versus 6.49% in the control group (HR: 1.34, 95% CI: 1.24–1.46). In patients with moderate-to-severe CKD, the incidence rose to 11.06% compared with 7.17% in the control group (HR: 1.60, 95% CI: 1.42–1.79). The highest risk occurred in severe CKD patients, with 11.38% developing DME, compared with 7.22% among controls (HR: 1.68, 95% CI: 1.38–2.06). Due to insufficient event counts in patients with ESRD, analysis for this group was not conducted. These findings confirm that reduced renal function contributes to DME risk, even in the absence of elevated baseline albuminuria.

Table 3.

Risk of Diabetic Macular Edema Stratified by Baseline Kidney Function in Normoalbuminuric Patients with Type 2 Diabetes Mellitus

Stages	eGFR Value (ml/min/1.73m2)	CKD Cohort		Control Cohort		Hazard Ratio	95% Confidence Interval
		Total	Events (No. %)	Total	Events (No. %)
G1	Normal or high (eGFR ≥90 mL/min)	Reference	Reference	Reference	Reference	Reference	Reference
G2	Mild (eGFR = 60-89 mL/min)	47 623	3247 (6.82%)	47 623	3189 (6.70%)	1.02	(0.98–1.08)
G3a	Mild to moderate (eGFR = 45–59 mL/min)	15 408	1325 (8.60%)	15 408	1000 (6.49%)	1.34	(1.24–1.46)
G3b	Moderate to severe (eGFR = 30–44 mL/min)	6734	745 (11.06%)	6734	483 (7.17%)	1.60	(1.42–1.79)
G4	Severe (eGFR = 15–29 mL/min)	2161	246 (11.38%)	2161	156 (7.22%)	1.68	(1.38–2.06)
G5	End stage (eGFR <15 mL/min)∗

CKD = chronic kidney disease; eGFR = estimated glomerular filtration rate; No. = number.

^∗Too few counts to run.

In the sensitivity analysis, patients who had undergone kidney transplantation showed a lower risk of DME compared with those who had not. Among transplant recipients, 14.6% developed DME, compared with 20.8% in the control group (HR: 0.65, 95% CI: 0.51–0.81).

Discussion

In this retrospective cohort study, we observed a consistent, stepwise increase in the risk of DME with declining renal function among patients with T2DM, after rigorous PSM for relevant covariates. Compared with patients with normal kidney function, those with mild CKD had a modest yet significant increase in DME risk (HR: 1.05, P < 0.05), which escalated progressively with advancing CKD severity, reaching the highest risk in patients with severe CKD (HR: 2.35, P < 0.05) and ESRD (HR: 2.53, P < 0.05). Sensitivity analysis demonstrated that kidney transplantation in ESRD patients was associated with a lower risk of DME (HR: 0.65, P < 0.05), suggesting a potential protective effect from improved renal function. Additionally, our subgroup analysis restricted to patients with normoalbuminuria further reinforced these findings, showing a significant relationship between reduced renal function and increased DME risk for most stages, even in the absence of significant baseline albuminuria. Collectively, these results underscore the importance of renal function as an independent predictor of DME risk, highlighting the need for closer ophthalmic surveillance among patients with declining kidney function, even before significant albuminuria develops.

The observed association between worsening CKD and increased DME risk likely reflects shared pathophysiological mechanisms central to diabetic microvascular complications. The primary driver of DME is VEGF-induced retinal vascular hyperpermeability, resulting in fluid accumulation within the macula. Chronic kidney disease, characterized by systemic inflammation, oxidative stress, and endothelial dysfunction, may exacerbate these retinal microvascular changes by promoting further endothelial permeability and leakage. Specifically, interactions among immune cells, endothelial cells, and retinal parenchyma elevate pro-inflammatory mediators (eg, inflammatory cytokines, interleukin-6, lipids, and VEGF) that contribute directly to retinal vascular injury and permeability.^13,14 Additionally, diabetic kidney disease involves glomerular hyperfiltration mediated by overactivation of the sodium-glucose cotransporter 2 (SGLT2) and the renin-angiotensin-aldosterone system.¹⁵ These alterations induce chronic vascular stress, damage, and endothelial dysfunction, closely mirroring pathological processes observed in retinal vessels during DME progression. Moreover, impaired glucose tolerance and insulin resistance, central metabolic disturbances in diabetes, further perpetuate systemic inflammation, oxidative stress, and vascular damage, creating an environment conducive to simultaneous renal and retinal microvascular injury.¹³ Lastly, impaired renal function may result in increased body fluid volumes, which has been linked to DME.^16,17 Collectively, these interconnected pathways emphasize that CKD severity not only marks advanced or poorly controlled diabetes but also actively contributes to DME pathogenesis. Therefore, therapeutic strategies targeting these common inflammatory, metabolic, and hemodynamic mechanisms may effectively mitigate both renal and retinal diabetic complications.

We performed a subgroup analysis restricted to patients with documented normoalbuminuria and eGFR measurements. Even in this subgroup, the relationship between declining renal function and increased DME risk remained significant, although the overall risks were lower compared with the primary analysis. In normoalbuminuric patients with type 2 diabetes, renal impairment can occur independently of albuminuria due to underlying glomerular hyperfiltration and structural renal changes, such as increased glomerular filtration surface area, which gradually reduce eGFR.¹⁸ Furthermore, damage to endothelial cells promotes increased vascular permeability, inflammation, and compromised microvascular integrity, connecting the pathological pathways of renal and retinal diseases independently of albuminuria status.^19,20 This is in distinction to Wang et al, which found an association between CKD and DME only in the presence of high microalbuminuria.¹¹ This difference in results may be attributed to variations in study design, patient population, and statistical approach. First, our study used PSM to balance baseline characteristics and comorbidities, potentially reducing confounding effects and allowing for a clearer evaluation of eGFR's role in DME risk. Additionally, differences in the outcome observation period may play a role. It is possible that kidney dysfunction begins to impact retinal microvascular changes earlier in the disease course than previously recognized.

Our sensitivity analysis indicated that kidney transplantation was associated with a reduced risk of DME. These findings suggest that improved renal function posttransplantation may lessen systemic microvascular complications, reducing the risk of retinal pathology. In a prospective study by Mavlyanova et al, retinal parameters, including area, perimeter, and circularity index of the foveal avascular zone, showed recovery over the 12-month observation period after kidney transplantation.²¹ Farrah et al used OCT to show that both retinal and choroidal thinning are prominent in CKD and correlate with declining glomerular filtration rate and kidney scarring.²² Importantly, they demonstrated that kidney transplantation is followed by a rapid, sustained increase in choroidal thickness, indicating that restoration of renal function can reverse ocular microvascular changes.²² These findings reinforce our sensitivity analysis; however, further research is needed to delineate the mechanisms underlying this protective effect.

Our study has several strengths, including the use of a large, geographically diverse dataset and PSM to balance baseline characteristics. Data were collected from academic centers, institutions with tertiary ophthalmology departments, community hospitals, and specialty clinics, providing a diverse range of health care settings to enhance accuracy, broaden coverage, and ensure substantial representation of ophthalmic diagnoses and treatments. The use of PSM ensured that the groups were well-balanced at baseline by including hemoglobin A1c and other relevant factors from the Diabetes Severity Index as measures of diabetes control. The inclusion of both eGFR and albuminuria-based measures allowed for a nuanced evaluation of kidney function's role in DME development. This approach reduced the impact of confounding factors and enhanced the validity of the comparative analysis.

Our study has several limitations, primarily related to its retrospective design and reliance on EMRs, which may introduce coding inaccuracies and variability in follow-up care. We did not account for additional kidney-related factors, including tubular dysfunction or novel biomarkers of renal injury, which could independently affect DME risk. Furthermore, our use of PSM was a univariate analysis focused on baseline characteristics. Residual bias may persist, and this statistical approach does not account for changes in kidney function or subsequent treatments that patients may have received after matching, potentially influencing DME risk over time. However, by rigorously matching key baseline variables, we aimed to minimize confounding at the point of diabetes diagnosis. Additionally, the risk did not increase as sharply in the most advanced CKD stages, which may be due to smaller sample sizes in these subgroups and limited event counts, potentially underpowering these comparisons. Competing risks such as high mortality in ESRD patients and potential diagnostic ascertainment bias remain important considerations that could influence incidence estimates. Finally, we did not evaluate the impact of hemodialysis due to insufficient outcome data, as the TriNetX database predominantly captures hospital-based dialysis rather than regular outpatient dialysis sessions. The generalizability of our findings may be limited by variations in International Classification of Diseases coding practices and follow-up patterns across the 70 participating institutions, which could result in misclassification of diagnoses. Furthermore, despite excluding known ophthalmic complications at baseline, some patients may have had undiagnosed DME or retinopathy at study entry; such cases would be misclassified as incident DME, potentially inflating our reported incidence. For example, patients with advanced CKD who receive specialty care might undergo more frequent and detailed ophthalmic examinations (eg, routine OCT imaging), increasing the likelihood of DME detection, whereas those with milder CKD managed in the community might receive less intensive eye screening, leading to potential underdiagnosis of DME. This ascertainment bias may have influenced the observed differences in DME rates among CKD stages. Furthermore, while our data suggest an association between CKD severity and DME, causality cannot be established without prospective validation, for which interventional studies are warranted to establish causality. Despite these limitations, our results provide meaningful insights into the association between renal impairment and DME risk and highlight directions for future prospective studies.

Future studies should investigate whether pharmacologic interventions targeting CKD progression, such as SGLT2 inhibitors or renin-angiotensin-aldosterone system inhibitors, can mitigate DME risk. Recent studies have shown that SGLT2 inhibitors significantly reduce central retinal thickness and decrease the frequency of anti-VEGF injections in patients with DME.^23,24 Ongoing randomized clinical trials are further evaluating these effects (NCT06845163). Additionally, emerging therapies aimed at enhancing endothelial function and reducing systemic inflammation in CKD patients should be explored for their potential role in preventing DME.²⁵

In conclusion, our study provides evidence of an association between CKD severity and increased DME risk, even in the presence of normoalbuminuria. These findings show the importance of comprehensive systemic management in patients with diabetes to address both renal and retinal complications. Patients with T2DM and moderate-to-severe CKD are a high-risk group for DME who may benefit from more frequent retinal screening than the usual annual examination, which suggest considering examinations every 6 to 12 months for earlier DME detection. Moreover, aggressive management of systemic risk factors and interventions to slow CKD progression (eg, optimal glycemic and blood pressure control and use of SGLT2 inhibitors or renin-angiotensin-aldosterone system blockade) should be emphasized, as these may reduce DME incidence and potentially the need for intravitreal anti-VEGF injections. Future prospective studies and interventional trials are needed to explore potential therapeutic strategies that address the intertwined pathophysiology of renal and retinal disease. Enhanced systemic management may reduce the burden of anti-VEGF therapy while simultaneously improving overall patient health.

Manuscript no. XOPS-D-25-00374