Causal association between circulating leukocyte characteristics and diabetic retinopathy based on two‐sample Mendelian randomization

ABSTRACT

Aims/Introduction

Leukocytes are implicated in the inflammatory cascades of diabetic retinopathy (DR), but their causal roles remain ambiguous. This study employed a two‐sample Mendelian randomization (MR) analysis to dissect the causal effects of circulating leukocyte counts on DR risk.

Materials and Methods

We utilized summary statistics from large‐scale genome‐wide association studies (GWAS) for five leukocyte subtypes and DR in European‐ancestry populations. The inverse‐variance weighted (IVW) method was primary, supported by comprehensive sensitivity analyses including MR‐Egger, weighted median, and the MR‐Pleiotropy Residual Sum and Outlier (MR‐PRESSO) test to ensure result robustness.

Results

A total of 2,136 leukocyte‐related SNPs were extracted as instrumental variables for causal inference. MR analysis revealed that increased lymphocyte counts are associated with reduced DR risk (IVW OR = 0.93, 95% CI = 0.86–0.99, P = 0.03), while the initial association between higher eosinophil counts and DR risk (IVW OR = 1.11, 95% CI = 1.03–1.19, P < 0.01) was attenuated following correction for outliers. No significant associations were observed for basophil, monocyte, or neutrophil counts. Sensitivity analyses found no evidence of pleiotropy or substantial influence from single SNPs.

Conclusions

Our findings provide genetic evidence supporting a potential causal association between lymphocyte counts and diabetic retinopathy risk, while the association for eosinophil counts was attenuated after correction for outliers. These results highlight the importance of further investigating the physiological role of lymphocytes in diabetic retinopathy to inform effective prevention and treatment strategies.

A schematic representation of the two‐sample Mendelian randomization framework used to investigate the causal association between circulating leukocyte traits and diabetic retinopathy.

INTRODUCTION

Diabetic retinopathy (DR) stands as a significant complication of diabetes mellitus, presenting in two distinct stages: the nonproliferative and the proliferative phases. Initially, symptoms like microaneurysms and intraretinal hemorrhages may be subtle but can deteriorate vision over time. In the proliferative phase, abnormal vessel and tissue growth can lead to retinal detachment and increased blindness risk. This condition poses formidable challenges, especially in resource‐constrained settings, due to inadequate screening and delayed treatment. ¹

The prevalence of DR exhibits variability, with certain regions recording up to 45% of diabetic individuals developing the condition over their lifetimes. ² Noteworthy risk factors encompass the duration of diabetes, suboptimal glycemic control, hypertension, and dyslipidemia. While metabolic irregularities primarily underpin the onset of diabetic retinopathy (DR), contemporary research underscores the substantial involvement of the immune system, particularly the participation of white blood cells. Leukocytes play a pivotal role in the inflammatory cascades contributing to DR. They demonstrate functional abnormalities ³ which enhance inflammation and vascular damage, ⁴ releasing substances like TNF‐α, IL‐6, and chemokines that worsen the condition by increasing vascular permeability and promoting new vessel and fibrous tissue growth. ⁵ Although increased white blood cell counts have been associated with a higher risk of microvascular diabetic complications, ⁶ and lower monocyte levels have been linked with DR risk in Chinese patients after adjusting for confounders, ⁷ most prior studies were cross‐sectional or case–control in design, limited by small sample sizes and the inability to establish causality.

To overcome the limitations of observational studies, Mendelian randomization (MR) utilizes genetic variants identified from genome‐wide association studies (GWAS) as instrumental variables (IVs), which are theoretically independent of confounding factors and reverse causality. ⁸ Several recent studies have successfully employed MR to elucidate risk factors for DR. ⁹ , ¹⁰ , ¹¹ However, the causal association between circulating leukocyte counts and DR remains underexplored. Given that leukocyte measurements are routinely available in clinical practice, confirming a causal relationship could improve the feasibility of DR screening and offer novel insights into prevention and treatment.

Therefore, our study aims to fill this significant gap by leveraging large‐scale GWAS data and MR analysis to investigate the potential causal associations between circulating leukocyte subtypes and DR risk, thus providing valuable genetic evidence to inform future research and clinical management.

METHODS

Study design

This study utilizes a large data set from a previous GWAS to investigate the relationship between leukocyte characteristics and DR. Single‐nucleotide polymorphisms (SNPs) served as instrumental variables (IVs) in the MR investigation. To ensure reliable results, three critical assumptions pertaining to the IVs were addressed: 1. Relevance: The chosen IVs must be strongly linked to leukocyte characteristics. 2. Independence: The IVs should not be related to any factors that could influence the results. 3. Exclusion Restriction: The IVs must influence DR only through their effect on leukocytes, not through any other pathways (Figure 1).

Figure 1.

Study design and three corresponding assumptions of this MR study. SNPs, single nucleotide polymorphism.

Data sources

The summary statistics of the GWAS data on circulating leukocyte characteristics, including neutrophil count, lymphocyte count, monocyte count, basophil count, and eosinophil count, were derived from a study that integrated data from trans‐ethnic meta‐analyses for 15 hematological traits in 746,667 participants, including 184,535 non‐European individuals. ¹² To mitigate potential confounding from population stratification, all GWAS summary statistics for both exposures and the outcome were derived exclusively from large‐scale consortia of individuals of European ancestry. For our study, we utilized only the aggregated European data to ensure consistency and minimize residual confounding. This focused approach facilitates a more streamlined examination of genetic influences on leukocyte characteristics within a homogeneous population, thereby offering clearer insights into the genetic determinants relevant to diabetic retinopathy in European populations (Table 1).

Table 1.

GWAS data on circulating leukocyte characteristics and diabetic retinopathy

Exposure	GWAS ID	Sample size (case/control)	SNPs	Population
Neutrophil count	GCST90002351	519,288	46,379,899	European
Lymphocyte count	GCST90002316	524,923	46,394,491	European
Monocyte count	GCST90002340	521,594	46,391,183	European
Basophil count	GCST90002292	474,001	46,167,033	European
Eosinophil count	GCST90002298	474,237	46,133,037	European
Diabetic Retinopathy	finn‐b‐DM retinopathy	14,584/202,082	16, 380,459	European

FINNGEN is a large public–private partnership aiming to collect and analyze genome and health data from 500,000 Finnish biobank participants. The genetic summary data related to DR explored in this study were derived from a GWAS conducted on the Finn cohort. The study population included a DR case group (n = 14,584 cases) along with a substantial healthy control group (n = 202,082 cases). A total of 9,289,492 single nucleotide polymorphisms (SNPs) were analyzed, offering a rich resource for detailed investigation into the genetic risk factors associated with the disease.

Instrumental variable selection

The IVs included in this study needed to meet the following criteria: (i) autosomal duplex sequence SNPs were first screened for significant genome‐wide correlation with circulating blood leukocytes, that is, meeting the criteria of P < 5 × 10⁻⁸; (ii) SNPs with a minimum allele frequency (MAF) > 0.01 were screened; and (iii) according to the criteria of R ² < 0.001 and window size = 10,000 kb, the linkage disequilibrium (LD) effect among SNPs was eliminated; (iv) when the screened IV did not exist in the summary data of the ending, SNPs with high LD (R ² > 0.8) to the IV were sought out and replaced as proxy SNPs for replacement; (v) We harmonized the exposure and outcome data sets to eliminate ambiguous SNPs with nonconcordant alleles and SNPs with intermediate allele frequencies. The F value of each SNP in the IV was calculated to assess the IV strength and exclude possible weak instrumental variable bias between the IV and exposure factors, calculated as follows: F = R ²*(N − 2)/(1 − R ²), R ² is the proportion of variance in the exposure explained by the SNPs in the IV, and an F value of >10 is considered a strong genetic IV. ¹³

MR analysis

Inverse‐variance weighted (IVW) was used as the primary analytical method ¹⁴ to assess the causal association between circulating leukocyte characteristics and DR by calculating the odds ratio (OR) and 95% confidence interval (CI). IVW is the predominant method for interpreting MR results, which calculates a weighted average of effect sizes by calculating the inverse variance of each SNP as a weight, and is considered to be the most robust metric in the absence of evidence of directional multidirectionality between the selected IVs. ¹⁵ In addition, we applied multiple complementary approaches, including the MR‐Egger, weighted median and weighted mode methods. ¹⁶ , ¹⁷ The weighted median method assumes that half of the instrumental variables are valid and analyzes the causal association between exposure and outcome.

Sensitivity analysis

In this study, heterogeneity between IVs was assessed by Cochran's Q test, and heterogeneity was considered low when P > 0.05, i.e., the valuation between instrumental variables varied randomly and produced little effect on IVW results. Also considering the impact of pleiotropy of genetic variation on the estimation of the association effect, this study used MR‐Egger regression to explore the existence of horizontal pleiotropy, and when the intercept term of MR‐Egger regression tends to zero or is not statistically significant, it suggests that there is no pleiotropy, and the exclusion assumption can be regarded as valid. ¹⁸ In addition, the MR pleiotropy residual sum and outlier (MR‐PRESSO) test was used in this study to detect possible outliers (i.e., SNPs with P < 0.05) and to re‐estimate the causal associations after eliminating them, thus correcting for horizontal pleiotropy. ¹⁹ Scatter plots, forest plots, funnel plots, and leave‐one‐out plots were used for visualization. All statistical tests were two‐sided and were performed using the TwoSampleMR and MR‐PRESSO packages in R software 4.0.5.

RESULTS

Instrumental variables

In this study, a comprehensive screening identified 2,136 SNPs related to circulating leukocytes. The mean F‐statistic was robust at 129.47, ranging from 29.78 to 5100.97. However, during the MR analysis with diabetic retinopathy as the outcome, 78 SNPs not matching the summary data were removed, and 108 proxy SNPs were employed (Table 2). The remaining SNPs are shown in the appendix. After harmonizing the exposure and outcome data sets, the final counts of IVs included were 393 SNPs related to Neutrophils, 475 SNPs related to Lymphocytes, 485 SNPs related to Monocytes, 191 SNPs related to Basophils, and 432 SNPs related to Eosinophils.

Table 2.

Details of instrumental variables

Exposure	Outcome	N.SNPs	R 2 mean(min, max)	F mean(min, max)	N. Proxy SNPs	N. Not matched SNPs
Neutrophil count	Diabetic retinopathy	430	0.0002 (0.00006, 0.007)	111.94 (29.85, 3750.82)	19	23
Lymphocyte count		519	0.0002 (0.00006, 0.004)	114.73 (29.78, 2294.96)	24	24
Monocyte count		524	0.0003 (0.00006, 0.009)	174.89 (29.86, 5100.97)	17	18
Basophil count		207	0.0002 (0.00006, 0.004)	90.37 (29.79, 1201.23)	4	15
Eosinophil counts		456	0.0003 (0.00006, 0.005)	128.34 (29.80, 2652.76)	14	25

Mendelian randomization analysis

In our MR analysis, a significant positive association was observed between eosinophil count and the risk of DR for IVW methods (OR = 1.1195% CI 1.03–1.19, P < 0.01). This association suggests that increased eosinophil counts contribute to a higher risk of developing DR. However, the other three methods (MR‐Egger, weighted median, and weighted mode) did not find a significant causal relationship (MR‐Egger: OR = 1.12, 95% CI: 0.98–1.27, P = 0.09; weighted median: OR = 1.06, 95% CI: 0.96–1.18, P = 0.23; weighted mode: OR = 1.06, 95% CI: 0.91–1.22, P = 0.45), although the direction of the OR was consistent with that of the IVW method.

Conversely, lymphocyte counts were inversely correlated with DR risk for IVW methods (OR = 0.93; 95% CI: 0.86–0.99, P = 0.03) in the IVW analysis. The other three methods did not find a significant causal relationship, even though the direction of the OR was consistent with the IVW method (MR‐Egger: OR = 0.97, 95% CI: 0.85–1.12, P = 0.68; weighted median: OR = 0.92, 95% CI: 0.83–1.03, P = 0.14; weighted mode: OR = 0.83, 95% CI: 0.67–1.02, P = 0.08).

For the other leukocyte types—basophils, monocytes, and neutrophils—IVW analysis did not reveal any statistically significant associations with DR risk (Basophil count: OR = 0.95, 95% CI: 0.86–1.06, P = 0.36; monocytes: OR = 0.96, 95% CI: 0.92–1.02, P = 0.17; neutrophils: OR = 1.07, 95% CI: 0.97–1.18, P = 0.15). Similarly, the other three methods did not find a significant causal relationship. The MR estimates of different methods were presented in Table 3. The scatter plot for effect sizes of SNPs for leukocyte count and DR is shown in Figure 2; Single SNPs MR effect size for leukocyte count on DR is shown in Figure 3.

Table 3.

MR analysis of circulating leukocyte characteristics in diabetic retinopathy

Exposure	Outcome	N.SNPs	Methods	OR (95% CI)	P
Basophil count	Diabetic retinopathy	191	IVW	0.95 (0.86–1.06)	0.36
			MR‐Egger	0.89 (0.73–1.08)	0.24
			Weighted Median	0.92 (0.79–1.06)	0.25
			Weighted Mode	0.93 (0.78–1.12)	0.45
Eosinophil counts	Diabetic retinopathy	432	IVW	1.11 (1.03–1.19)	0.00
			MR‐Egger	1.12 (0.98–1.27)	0.09
			Weighted Median	1.06 (0.96–1.18)	0.23
			Weighted Mode	1.06 (0.91–1.22)	0.45
Lymphocyte count	Diabetic retinopathy	475	IVW	0.93 (0.86–0.99)	0.03
			MR‐Egger	0.97 (0.85–1.12)	0.68
			Weighted Median	0.92 (0.83–1.03)	0.14
			Weighted Mode	0.83 (0.67–1.02)	0.08
Monocyte count	Diabetic retinopathy	485	IVW	0.96 (0.92–1.02)	0.17
			MR‐Egger	0.93 (0.86–1.01)	0.10
			Weighted Median	0.94 (0.87–1.02)	0.16
			Weighted Mode	0.94 (0.87–1.01)	0.09
Neutrophil count	Diabetic retinopathy	393	IVW	1.07 (0.97–1.18)	0.15
			MR‐Egger	1.11 (0.92–1.34)	0.27
			Weighted Median	1.01 (0.91–1.13)	0.85
			Weighted Mode	1.03 (0.89–1.19)	0.71

Figure 2.

Scatter plot of circulating leukocyte properties in diabetic retinopathy. (a) Neutrophil count. (b) Lymphocyte count. (c) Monocyte count. (d) Basophil count. (e) Eosinophil count.

Figure 3.

Forest plot of circulating leukocyte properties in diabetic retinopathy. (a) Neutrophil count. (b) Lymphocyte count. (c) Monocyte count. (d) Basophil count. (e) Eosinophil count.

Sensitivity analysis

The Cochran's Q test revealed significant heterogeneity (P < 0.001) among the instrumental variables for all five leukocyte subtypes, as illustrated in funnel plots (Figure 4). However, MR‐Egger regression analysis indicated no significant horizontal pleiotropy for any of the leukocyte traits (P > 0.05 for all, Table 4).

Figure 4.

Funnel plot of circulating leukocyte properties in diabetic retinopathy. (a) Neutrophil count. (b) Lymphocyte count. (c) Monocyte count. (d) Basophil count. (e) Eosinophil count.

Table 4.

Multiple and heterogeneous analysis of circulating leukocyte characteristics on the level of diabetic retinopathy

Exposure	Outcome	Heterogeneity		Pleiotropy
		Q statistic (IVW)	P value	MR‐Egger Intercept	P value
Basophil count	Diabetic retinopathy	280.37	<0.001	0.002	0.415
Eosinophil counts		676.58	<0.001	0.000	0.858
Lymphocyte count		676.14	<0.001	−0.001	0.436
Monocyte count		627.05	<0.001	0.001	0.323
Neutrophil count		903.64	<0.001	−0.001	0.662

In the MR‐PRESSO analysis, after exclusion of outlier SNPs, the association between lymphocyte count and the risk of diabetic retinopathy remained statistically significant (outlier‐corrected OR = 0.91, 95% CI: 0.85–0.97, P = 0.005), suggesting a robust inverse association. In contrast, the previously significant association between eosinophil count and diabetic retinopathy became nonsignificant after outlier correction (outlier‐corrected OR = 1.04, 95% CI: 0.97–1.10, P = 0.253). The results for the other leukocyte subtypes (neutrophils, monocytes, and basophils) remained nonsignificant after outlier adjustment (Table 5). Additionally, leave‐one‐out sensitivity analyses demonstrated that no individual SNP had a disproportionate influence on the causal estimates for any leukocyte subtype (Figure 5).

Table 5.

MR‐PRESSO analysis of circulating leukocyte characteristics in diabetic retinopathy

Exposure	Outcome	Raw		Outlier corrected		Global P	Number of outliers	Distortion P
		OR (CI %)	P	OR (CI %)	P
Neutrophil count	Diabetic retinopathy	1.07 (0.97–1.18)	0.154	1.00 (0.93–1.07)	0.946	<0.001	4 (rs150861794, rs28588142, rs62360185, rs78738581)	0.004
Lymphocyte count		0.93 (0.86–0.99)	0.028	0.91 (0.85–0.97)	0.005	<0.001	3 (rs1523178, rs28588745, rs62262391)	0.660
Monocyte count		0.96 (0.92–1.02)	0.169	0.97 (0.92–1.02)	0.219	<0.001	5 (rs12055642, rs1587222, rs2343551, rs2957873, rs79086185)	0.730
Basophil count		0.95 (0.86–1.06)	0.358	0.97 (0.88–1.06)	0.492	<0.001	3 (rs2286599, rs2916193, rs33931987)	0.513
Eosinophil counts		1.11 (1.03–1.19)	0.004	1.04 (0.97–1.10)	0.253	<0.001	6 (rs11703539, rs152197, rs159963, rs2817399, rs310747, rs62183994)	0.012

Figure 5.

Leave‐one‐out plot of circulating leukocyte properties in diabetic retinopathy. (a) Neutrophil count. (b) Lymphocyte count. (c) Monocyte count. (d) Basophil count. (e) Eosinophil count.

DISCUSSION

This study uniquely employs a two‐sample MR approach to rigorously investigate the potential causal relationships between circulating leukocyte characteristics and DR. Notably, the causal effect of lymphocytes on DR risk was robust, whereas the initially observed positive association for eosinophil count was attenuated after outlier correction. In contrast, basophil, monocyte, and neutrophil counts did not demonstrate a direct causal link with DR. To enhance the reliability of our findings, MR‐Egger regression analysis was utilized to effectively exclude any pleiotropic bias, thus strengthening the credibility of our results.

Our genetic evidence highlights a potentially protective effect of lymphocytes against DR, but the underlying mechanisms remain to be fully elucidated. We hypothesize that this effect may be mediated through the modulation of the retinal inflammatory microenvironment by lymphocyte subtypes. Specifically, higher circulating lymphocyte counts—possibly reflecting increased regulatory T cell activity—could promote an anti‐inflammatory cytokine profile (e.g., increased IL‐10 and decreased TNF‐α/IL‐6), thereby attenuating microvascular damage in DR. For example, a greater proportion of regulatory T cells may foster such a favorable cytokine milieu, protecting the retina from chronic inflammation and vascular injury. ²⁰ , ²¹ , ²²

However, directly validating this cytokine‐mediated pathway was beyond the scope of our genetic study. We acknowledge as a key limitation that actual cytokine levels (such as TNF‐α and IL‐6) were not measured in our study, due to the use of GWAS summary statistics rather than individual clinical data. Therefore, future clinical and translational studies are needed to verify the relationship between lymphocyte counts and specific cytokine profiles in patients with DR. Such research would be crucial for interrogating the mechanistic link our findings suggest and for confirming the immunological pathways underlying DR pathogenesis.

Our study advances the field by providing genetic evidence for the distinct roles of five major leukocyte subtypes in DR pathogenesis. The significant inverse association between lymphocyte counts and DR risk, and the positive (but outlier‐sensitive) association for eosinophils, add granularity to prior findings on systemic immune‐inflammation indices such as the NLR and PLR in DR ²⁰ , ²¹ The involvement of eosinophils and lymphocytes in DR may be linked to their distinct immune and inflammatory functions. Eosinophils, for example, are implicated in tissue remodeling and fibrosis, processes potentially harmful in the microvascular settings characteristic of DR ²³ , ²⁴ Conversely, higher lymphocyte counts may reduce the occurrence of DR by enhancing immune surveillance and inhibiting chronic inflammatory processes. ²² The role of monocytes in retinal inflammation has also been recognized ²⁵ , ²⁶ although our study did not find a causal association between monocyte levels and DR risk—possibly because the mechanistic changes are insufficient to translate into a clinical effect, or due to differences in population or study design. Notably, our finding differs from a previous cross‐sectional Chinese study, which found lower monocyte levels associated with increased DR risk, but could not establish causality—a limitation overcome in the present MR design. No significant associations were observed for basophil and neutrophil counts, although some studies have indicated that a higher NLR or neutrophil albumin percentage (NPAR) is associated with DR risk ²¹ , ²⁷ and higher circulating alkaline granulocytes have been linked to severe myopic retinopathy. ²⁸

Circulating leukocyte counts, particularly lymphocyte and eosinophil levels, are readily available and cost‐effective biomarkers in clinical practice. ²⁹ However, given that the eosinophil association was not robust after outlier correction, lymphocyte count emerges as a more reliable biomarker candidate for DR risk prediction. If validated in further studies, lymphocyte counts could be incorporated into DR risk prediction models alongside established factors such as HbA1c and diabetes duration, especially in genetically susceptible individuals.

Importantly, the mechanistic hypothesis described above provides a logical basis for potential future strategies against DR. Rather than merely surveilling leukocyte counts, future interventions could focus on modulating lymphocyte function to establish a protective cytokine environment. For example, therapies designed to promote regulatory T cell activity (such as low‐dose IL‐2 therapy) or to inhibit key pro‐inflammatory cytokine pathways might represent promising avenues for DR prevention or treatment. These targeted interventions, derived from our genetic findings, warrant systematic evaluation in both preclinical DR models and clinical settings.

Regarding the eosinophil finding, while an elevated eosinophil count was initially associated with increased DR risk in our primary analysis, this association was attenuated and became non‐significant after correction for outliers. This underscores the rigor of our analytical approach and further strengthens the credibility of the lymphocyte signal.

Our study was based on data from a European population, which may limit the applicability of the findings to other geographic populations. The genetic architecture and environmental factors influencing diabetic retinopathy may vary significantly across different populations. ³⁰ However, the global prevalence of diabetic retinopathy is high, with approximately 45% of diabetic individuals developing the condition over their lifetimes. ² Given the similarity in DR pathophysiology and leukocyte involvement across populations, our findings may still provide valuable insights for other regions. Future research should include diverse ethnic populations to validate and expand upon these findings.

In summary, the pathogenesis of DR, a microvascular complication of diabetes, involves complex pathophysiologic processes in which chronic inflammatory responses play a key role, involving key members of the circulating leukocyte family, including neutrophils, lymphocytes, and monocytes/macrophages. These cells exhibit significant changes in their activity, functionality, and their ability to adhere to and migrate within vascular walls. ²⁷ First, the hyperglycemic environment enhances leukocyte activation, making them more likely to adhere to the damaged vascular endothelium and exacerbate inflammatory responses. Second, there is an upregulation in the expression of adhesion molecules like selectins and integrins on both leukocytes and vascular endothelial cells, intensifying their interactions and contributing to further inflammation and vascular damage. ⁴ Third, leukocytes release a substantial number of inflammatory mediators, such as cytokines (e.g., TNF‐α, IL‐6), which play crucial roles in increasing vascular permeability, promoting neovascularization, and fibrosis. Lastly, the enhanced migration capability of leukocytes leads them to accumulate in the microvascular regions of the damaged retina and penetrate the blood‐retinal barrier, thus aggravating local inflammatory responses and tissue damage.

Available evidence increasingly suggests that the innate immune system may play an important role in the pathogenesis of DR, ³¹ whereas the specific role of the adaptive immune system in the development of DR is of limited understanding. ⁵ The pathogenesis of type 1 diabetes mellitus (T1D), as an autoimmune disease, is closely related to the dysfunction of the immune system during pathophysiologic processes. However, the specific association between diabetes disease course and circulating immune cell changes has lacked in‐depth investigation in the context of T1D. ³² Our findings suggest that genetic changes influencing circulating blood eosinophils and lymphocytes are associated with DR risk, and highlight the need for future research to clarify their physiological functions and mechanisms. Targeted modulation of these immune cell populations and their associated cytokine profiles may provide innovative therapeutic approaches to delay or prevent DR progression. Regular fundoscopic exams and monitoring of specific leukocyte subpopulations, such as eosinophils and lymphocytes, may facilitate early detection and intervention, potentially reducing the severity of DR and the risk of blindness. ³³ It also highlights the need for future research to explore the specific mechanisms through which these leukocytes influence DR progression. This could involve detailed molecular studies and developing animal models that more accurately replicate the human condition of DR. Such investigations will not only deepen our understanding of DR pathogenesis but also guide the development of more effective treatments.

This study had several strengths. It is the first to utilize GWAS summary‐level statistics along with two‐sample MR analysis to investigate the causal links between circulating blood leukocyte characteristics and DR. The two‐sample MR method leverages extensive summary‐level genetic data, minimizing the impact of potential confounders and reverse causation, and enhancing the reliability of the findings. Additionally, through rigorous sensitivity analyses—including multiple MR methods—this study systematically evaluated potential biases from outliers and multidirectionality, confirming the stability and reliability of the results. However, there are some limitations. First, the selected validated SNPs explained only a small portion of the circulating leukocyte variation and may not capture associations with weak effects. Second, our exclusive use of data from European populations, while strengthening the internal validity by minimizing population stratification bias, limits generalizability to other ethnic groups. Third, the use of pooled data precludes exploration of nonlinear associations or stratification effects. Lastly, we only examined individual white blood cell counts; future studies should address composite indices (e.g., NLR, PLR) and measure cytokine levels directly in clinical cohorts. Nevertheless, our findings provide new perspectives and valuable clues for the etiology of DR, and lay the foundation for further investigation of its pathogenesis and for optimizing prevention and treatment strategies.

In conclusion, our study provides genetic evidence that abnormalities in eosinophils and lymphocytes within circulating leukocyte subpopulations are causally associated with diabetic retinopathy. These findings not only highlight new avenues for mechanistic research and the development of targeted therapeutic and preventive strategies for DR, but also suggest that clinical interventions aimed at modulating lymphocyte function—such as promoting regulatory T cell populations—may represent a promising direction for the precise prevention or delay of DR progression in the future.

FUNDING

This work was supported by the Mayinglong Pharmaceutical Group Co., Ltd. (303‐02‐317) and Major Central Government Projects with Significant Budgetary Impact, Sustainable Utilization Capacity Building Project for Precious Chinese Medicinal Resources (2060302‐2303‐21).

DISCLOSURE

The authors declare no conflict of interest.

Approval of the research protocol: This Manuscript follows the STROBE‐MR guidelines. ³³ According to Article 32 of the Ethical Review Measures for Life Science and Medical Research Involving Human Subjects of the People's Republic of China, this study qualifies for exemption from ethical review because the data used in this study is publicly available and legally accessible. It does not cause harm to humans or involve sensitive personal privacy or commercial secrets. Additionally, the GWAS data sets utilized in this study had obtained ethical approval in their respective studies. The FINNGEN project has been reviewed and approved by an ethical review board in its original study (number: HUS/990/2017; date: August 2017). Therefore, no additional ethical approval or informed consent was required for this study.

Informed Consent: N/A.

Registry and the registration no. of the study: N/A.

Animal studies: N/A.

Supporting information

Appendix S1. Detailed information of instrumental variables utilized in the Mendelian Randomization analysis of circulating leukocyte characteristics and diabetic retinopathy risk.

Data Availability Statement

All data generated or analyzed during this study are included in this published article and its supplementary information files.