Introduction

In diabetes mellitus, diabetic retinopathy (DR) is considered a prevalent microvascular complication and stands as a primary contributor to vision impairment among diabetic patients (1, 2). The DR in the early stages is typically asymptomatic (3). However, if left untreated, DR can lead to significant vision impairment and eventually lead to blindness. In addition to adverse visual outcomes that might compromise mobility, lower life quality, and depression, there is also an association between DR and increased risk of vascular diseases, putting significant costs on patients, families, and health care. DR is a progressive complication and can be categorized into the non-proliferative and proliferative stages according to severity (4, 5). Non-proliferative DR is accompanied by cotton-wool spots, microaneurysms, venous beading, internal microvascular illness, and hard exudes. On the other hand, proliferative DR is characterized by vitreous and pre-retinal hemorrhage and neovascularization of the optic disc (1, 6). Proliferative DR is less prevalent but has more devastating visual effects compared to non-proliferative DR (6, 7).

Despite the availability of recent diagnostic and therapeutic advancements, including optimal diabetes management and early diagnosis that can considerably lower the risk of vision loss, DR has remained a significant risk factor for vision impairment and blindness worldwide (8, 9). In 2010, 3.7 million were impaired visually, among which DR was responsible for 1.9% of visual impairments and globally made up 2.6% of all cases of blindness (10), while even more recent statistical data shows that 1.07 million people became blind as a result of DR in 2020 (11). Unless substantial preventive and treatment improvements are achieved, DR prevalence and burden can only be expected to escalate as the global diabetes epidemic increases. It is, therefore, necessary to evaluate optimal clinical management and treatment strategies to curb the surging cases of DR.

Proper diabetes management is crucial, and it requires regular follow-up and monitoring by measuring glycated hemoglobin (HbA1c) levels (12, 13). Moreover, the risk of diabetes-associated comorbidities, including diabetic retinopathy, will be minimized by ensuring that HbA1c is within the recommended range (14–16). Studies imply that the likelihood of developing DR and delaying its progress in those affected is significantly lower with optimal control of HbA1c (17–19). However, the extent to which HbA1c control affects the occurrence and worsening of DR differs among various groups of people. That highlights the need for a thorough analysis of existing research. This study aimed to assess the influence of glycemic control on the occurrence and progression of DR.

Methodology

Study Design

The current study is a systematic review and meta-analysis that followed the guidelines highlighted in the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement to ensure quality and comprehensiveness. The study aimed to systematically review and meta-analyze the published articles evaluating the effect of glycemic control on the development and progression of DR in patients with diabetes.

Literature Search Strategy

The electronic databases where the search for articles was conducted were Cochrane Library, Google Scholar, and PubMed: the articles were retrieved through an extensive and systematic approach. The search terms included relevant keywords and MeSH (Diabetic Retinopathy) AND (Glycemic Control OR HbA1c) AND (Incidence OR Progression). Filtering ensured they only included articles published between 2008 and 2024. Other sources included manual searching on the bibliographies of the papers selected for review.

Eligibility Criteria

The eligibility criteria for study inclusion were defined using the PICO (Population, Intervention, Comparison, Outcome) framework (20). The criteria for including or excluding articles are summarized in Table 1.

Table 1: PICOS Eligibility Criteria that was applied in Identifying Studies for Inclusion
Item	Criteria for Inclusion	Criteria for Exclusion
Population	Studies that included patients diagnosed with diabetes (Type 1 or Type 2) were eligible. No restrictions were placed on age, gender or geographic location.	No diabetes mellitus.
Intervention	Studies assessing glycemic control, primarily measured by HbA1c levels, were included. Studies with different glycemic control strategies or interventions aimed at maintaining or improving HbA1c levels were eligible.	Studies that did not assess glycemic control and how it influences DR were excluded.
Comparators	Studies that compare different levels of glycemic control such as good vs. poor, intensive Vs standard/conventional, optimal Vs suboptimal control or any intervention that affected glycemic levels.	Did not compare different levels of glycemic control.
Outcomes	Studies that reported on the occurrence or worsening of DR as the primary or secondary outcome were included.	Did not provide sufficient data on HbA1c levels or diabetic retinopathy outcomes.
Design	Both RCTs and observational studies (case-control, cohort, and cross-sectional studies) were considered for inclusion.	Conference abstracts, reviews or editorials without original data.
Language	English	Not available in English.
Date published	2008 or later	Published before 2008

Study Selection

Studies were selected in adherence to PRISMA recommendations and were performed in two stages. First, an initial screening was conducted, during which the titles and abstracts of all identified records were screened. This process was carried out independently by two reviewers to ensure the exclusion of non-relevant records. Arising discrepancies between discussed or resolved by engaging a third reviewer. The second phase involved a full-text review. Relevant studies were identified in a PRISMA chart, which numerically outlines the records obtained during the search, taken through screening, examined for relevance, and selected for the final review and meta-analysis.

Data Extraction

The data extraction process was conducted by two reviewers, who independently entered the relevant information from the articles into a customized data extraction form. The extracted data included study features such as author, publication year, study design, sample size, duration of follow-up, and outcomes of interest, including the occurrence of diabetic retinopathy (DR) and its worsening or progression.

Risk of Bias Assessment

The Newcastle-Ottawa Scale (NOS) and the Cochrane Risk of Bias tool (ROB2) were employed to evaluate the studies for any risk of bias. The NOS was applied to assess the observational studies on selection, comparability, and outcome. On the other hand, RCTs were evaluated for bias across the selection, performance, detection, attrition, and reporting domains on ROB2.

Statistical Analysis

Review Manager Version 5.4 is used to perform all calculations. Odds Ratios (ORs) and the 95% confidence interval (95%CI) were determined for the incidence and progression of diabetic retinopathy, and estimations of pooled effect were visually represented on forest plots using the random-effects method due to the anticipation of a possibility of the studies being significantly heterogeneous. The pooled mean difference in HbA1c levels between the diabetes patients who developed DR and those who did not have DR was also calculated. The I² statistic was deployed in estimating the degree of heterogeneity. Studies had no heterogeneity if the I2 value < 50%. 50% ≤ I2 < 75% implied the studies were moderately heterogeneous, while studies were considered highly heterogeneous if I2 ≥75%.

Results and Findings

Search Results

Overall, 13155 potential papers were yielded by searches conducted on the databases (Google Scholar = 7550, PubMed = 6341, and Cochrane Library = 1064). After removing 9876 duplicate records, 3279 studies remained. Screening for the titles and abstracts led to the dropping of 2671 studies. Of the 608 studies considered for retrieval, 487 full texts could not be obtained. The 121 full-text studies that were successfully retrieved were then screened for relevance based on the PICOS criteria described earlier. Ultimately, 99 studies were discarded for not meeting the eligibility criteria, leaving 22 eligible papers from which data were extracted for analysis. The sequential procedure of selecting studies is illustrated in Figure 1, while the key features and outcomes are summarized in Table 2 (21–41).

Figure 1: The Process of Identifying Relevant Records Illustrated Using PRISMA Diagram.

Table 2: Summarizing of Important Features of the Relevant Studies, Patient Demographics, and Retinopathy Outcomes
Study ID	Study Design	Follow-up Period	Population (N)	Age (Years, Mean±SD)	Type of Diabetes	Duration of Diabetes (Years, Mean±SD)	HbA1c (%) (Mean±SD)	Glycemic Control Groups	Method of Retinopathy Diagnosis
ACCORDION 2016 (21)	Randomized trial	8 years	1,310	61.3±5.8	DM2	9.9± 6.8	Baseline (8.2± 1.0)	Intensive Standard	Fundus photography
ADVANCE Collaborative Group 2008 (22)	Factorial randomized, controlled trial	5 years	11,140	66±6	DM2	Standard Control (7.9±6.3) Intensive Control (8.0±6.4)	Baseline (Intensive Control = 7.51±1.57 Standard Control = 7.52±1.54) End of follow-up (Intensive Control = 6.53±0.91 Standard Control = 7.30±1.26)	Intensive Control (N=5571) Standard Control (N=5569)	Fundus photography
Alemayehu et al. 2024 (23)	Hospital-based cross-sectional study	30 May to 15 July 2022	391	49 (Median)	Both DM1 and DM2	6 (median)	-	Poor glycemic control, Good glycemic control	Fundus photography
Al-Rubeaan et al. 2015 (24)	Cross-sectional		50 464	59.7±12.78	DM2	13.4±8.24	Overall (8.9±2.32) No retinopathy (8.8±2.32) NPDR (9.5±2.17) PDR (9.55±2.34)	-	Ophthalmoscopy
Azad et al. 2014 (25)	Open-label prospective, randomized controlled trial	5 Years	1,791 (858 included in analysis)	60±9	DM2	11±7	Baseline = 9.4±1.5	Intensive (n=433) Standard (n=425)	Fundus photographs
Beulens et al 2009 (26)	2× 2 factorial, randomized controlled trial	2001 to 2008	1,602	65.6± 5.8	DM2	6 (Median)	Baseline = 7.4±1.5	Intensive glucose control (n=630) Standard glucose control (n=611) Placebo (n=618) Perindopril–indapamide (n=623)	Retinal photography
DCCT 2015 (27)	Randomized trial	18 years	1,441	27	DM1	5.7	Baseline (9.1) At DCCT closeout (7.2)	-	Ophthalmoscopic exam and seven-field color stereo fundus photography
Duckworth et al. 2009 (28)	Open-label study	5 years	1791	60.4	DM2	11.5	Baseline (9.4±2.0)	Standard Therapy (N=899) Intensive Therapy (N=892)	The 23-point Early Treatment Diabetic Retinopathy Study grading scale
Hatz et al. 2019 (29)	Cross-sectional study	5.8 (Median)	151	39.9 ± 14.0	DM1	14.3 ± 5.8	7.3 (Median)	_	7-field color fundus (CF) evaluation
Huang et al. 2010 (30)	Population-based survey	2 years	768	62.5±9.4	Both DM1 and DM2	12.1±8.7	All (8.0±2.0) No DR (8.0±2.0) DR (8.9±2.0)	Optimal glycemic control Suboptimal glycemic control	Standardized dilated retinal photography
Lu et al. 2018 (31)	Prospective study	2005 – 2012	3,262	60.4 ± 12.0	DM2	8.1 6 ±6.8	All subjects (8.9 ±2.2) No DR (8.8±2.2) Mild NPDR (9.0±1.8) Moderate NPDR (9.4±2.0) VTDR (9.4±2.1)	­_	_
Lu et al. 2019 (32)	Retrospective	2005 – 2015	2,927	57.7±10.1	DM2	7.7±6.3	Overall (8.9±2.1) No DR (8.8±2.2) DR (9.2±2.0)	_	Fundus photography
Mersha et al. 2022 (33)	Institution-based cross-sectional study	September 07/ 2020-November 06/2020	331	45.48±16.88	Both DM1 and DM2	5 (Median)	-	Poor control group and good control group	Clinical Examination
Raman et al. 2012 (34)	Population-based, cross-sectional study		1414	56.5 ± 10.6 (Optimal group) 56.2 ± 9.7 (Suboptimal group)	DM2	4.6 ± 6.0 for optimal group 5.9 ± 6.3 for suboptimal group	HbA1c < 7% (N = 487) HbA1c ≥ 7% (N = 927)	Optimally controlled (HbA1c < 7%) Suboptimal control (HbA1c ≥ 7%)	Fundus photography
Romero-Aroca et al. 2017 (35)	prospective population-based study	9 years	15396	35.19±10.03	DM1 (366) DM2 (15030)	13.63 ±8.42	DM1 (8.38±1.16) DM2 (7.38±1.29) Any DR (8.57±1.41) No DR (7.94±1.19)	_	One 45° field retinography
Sartore et al. 2013 (36)	Cross-sectional Study		68	19 – 69	DM1 = 35 DM2 = 33	Retinopathy = 19.1 ± 9.2 No Retinopathy = 12.1 ± 7.7	Retinopathy = 8.3 ± 1.2 No Retinopathy = 7.9 ± 1.8	Stable Glycemic control	Fundus photography
Shumye et al. 2024 (37)	Multicenter, hospital-based, cross-sectional study	May 8 to June 15, 2023	1219	53 (Median)	Both DM1 and DM2	_	_	Poor control, Good Control	Fundus photography
Shurter et al. 2013 (38)	Retrospective, case-control study	September 2005 to August 2007	68	Intensive (52±2.2) Standard (52.3±1.95)	DM2	Intensive (8.4±1.12) Standard (13.1±1.52)	Baseline (intensive group = 10.7±0.35; and control group = 7.9±0.27) Max decline (Intensive = 4.0±0.41, standard = 0.2±0.11)	Intensive (34) Standard (34)	Retinal imaging
Surowiec et al. 2022 (39)	Retrospective	4.9 ± 1.4 (Mean)	384	34 ± 9.2	DM1	20.5 ± 7.9	6.9 ± 1	Intensive (n = 658) Standard (n = 652)	Fundus photography
Tan et al. 2018 (40)	Population-based	8 years	2877	61.5±10.2	DM1 (51) DM2 (2826)	No DR (5.47±7.68) DR (12.3±10.1)	No DR = 7.49 ±1.54) DR = 8.26±1.85	_	Two-field, 45-degree digital retinal photography
White et al. 2018 (41)	Diabetes Control and Complications Trial	10 years	1211	27± 7	DM1	1 to 15 years	-	Intensive Conventional	Fundus photography
Wu et al. 2024 (42)	Intervention study	_	466 (no DR) 136 (DR)	66.0 ± 8.5 for No DR, 65.2 ± 8.5 for DR	DM2	11.2 ± 7.7 For no DR, and 16.6 ± 8.7 for DR	<8.5 (DR = 116, No DR = 443) ≥8.5 (DR = 26, No DR = 34)	_	Fundus photography

DR – Diabetic Retinopathy, PDR – Proliferative DR, NPDR – Nonproliferative DR, DM – Diabetes mellitus, VTDR – Vision threatening DR, DM1 – Type 1 Diabetes mellitus, DM2 – Type 2 Diabetes mellitus

Quality Appraisal

The 16 observation studies were evaluated for quality using NOS, as shown in Table 3. NOS tests the studies for quality using three domains: selection of participants, comparability, and outcome. Overall, the studies included were moderate to high quality (Moderate quality = 13 studies, High quality = 3 studies).

Table 3: Quality appraisal of Observational Studies Using NOS
Study	Selection (Max = 4)	Comparability (Max = 2)	Outcome (Max = 3)	Total Score (Max = 9)	Quality
Alemayehu et al. 2024 (23)	2	2	2	6	Moderate
Al-Rubeaan et al. 2015 (24)	3	1	2	6	Moderate
Duckworth et al. 2009 (28)	2	1	2	5	Moderate
Hatz et al. 2019 (29)	3	2	1	6	Moderate
Huang et al. 2010 (30)	2	1	2	5	Moderate
Lu et al. 2018 (31)	3	1	1	5	Moderate
Lu et al. 2019 (42)	3	1	2	6	Moderate
Mersha et al. 2022 (32)	3	2	2	7	High
Raman et al. 2012 (33)	3	2	2	7	High
Romero-Aroca et al. 2017 (34)	3	1	2	6	Moderate
Sartore et al. 2013 (35)	2	1	2	5	Moderate
Shumye et al. 2024 (36)	3	2	2	7	High
Shurter et al. 2013 (37)	3	1	2	6	Moderate
Surowiec et al. 2022 (38)	3	1	2	6	Moderate
Tan et al. 2018 (39)	3	1	1	5	Moderate
Wu et al. 2024 (41)	3	2	1	6	Moderate

On the other hand, assessment of the 6 RCTs was achieved using ROB2 (Figure 2).

Figure 2: ROB2 Plot for the Assessment of the 6 RCTs

Figure 3: Visual Representation of the Incidence of DR in Intensive Vs. Standard Glycemic Control Groups on a Forest Plot.

Figure 4:Visual Representation of the progression of DR in the Intensive vs. standard Control on a Forest Plot.

Figure 5: Difference in HbA1c Levels between the DR and No-DR Patients on a Forest Plot.

Comparing the Incidence of DR Based on Different Glycemic Interventions

The risk of DR was also compared for patients receiving glycemic therapies. The meta-analysis of 4 articles that gave data on the occurrence of DR in patients on insulin therapy and those on anti-glycemic tablets was performed. The results revealed no significant distinction in the likelihood of DR between patients receiving insulin therapy and oral tablets (OR = 1.96 95%CI [0.93, 4.16]). However, the studies showed significantly high heterogeneity (I2 = 82%, p = 0.0007) (Figure 6A).

Similarly, the data analyzed from 3 studies that presented data comparing insulin therapy with combined (both insulin and tablet) treatment revealed no substantial difference in the likelihood of DR (OR = 0.81 95%CI [0. 21, 3.03]) (Figure 6B). Additionally, combined therapy was revealed to decrease the likelihood of DR more significantly than oral tablets alone (OR = 0.48 95%CI [0. 29, 0.81]) (Figure 6C).

Discussion

This systematic review and meta-analysis assessed the impact of glycemic control on DR in diabetes mellitus patients by comparing the risk of occurrence and DR in patients as reported in primary studies that investigated DR outcomes in patients undergoing different methods and interventions for glycemic control. Moreover, the mean difference in percentage HbA1c % between DR cases and those who did not develop DR was evaluated to determine the relevance of HbA1c as a DR risk predictor. The intensive and standard glycemic control methods were compared to determine the difference in the risk of DR occurrence and progress based on the intensity of glycemic control. Intensive glycemic control aims to reduce and keep the HbA1c levels to near normal/optimal levels, while standard/ suboptimal glycemic control always seeks to maintain the HbA1c to acceptable levels.

The findings of this study showed that intensive glycemic control can result in a significant reduction in the risk of DR incidence (OR = 0.50 95%CI [0.38, 0.67]), its worsening/new onset (OR = 0.91 95%CI [0. 79, 1.04]), and the risk both ≥2 steps (OR = 0.77 95%CI [0. 63, 0.94]) and ≥3 steps (OR = 0.59 95%CI [0. 42, 0.82]) progression of DR on the ETDRS. The incidence of proliferative DR was significantly lower with glycemic control groups than the standard groups (OR = 0.37 95%CI [0. 30, 0.45]) while the risk of less advanced stages of non-proliferative (moderate (OR = 0.40 95%CI [0. 13, 1. 21]) or severe (OR = 0.57 95%CI [0. 15, 2.10])) and the difference in mild stages of DR was not significant. Further analysis revealed that significantly higher HbA1c levels were common in diabetes patients who had DR than those with no DR (Mean Difference = 0.64 95%CI [0.52, 0.75]). However, no significant relationship was identified between the occurrence of DR and the type of glycemic treatment used (insulin therapy alone, oral tablets alone or combination of both insulin therapy and oral tablets).

Figure 6: A: Visual Representation of the Occurrence of DR in Insulin Therapy Vs Tablet on a Forest Plot. B: Visual Representation of the Occurrence of DR in Patients on Insulin Therapy vs. Combined Treatment on a Forest Plot. C. Visual Representation of the Occurrence of DR with Oral Tablets Alone vs. Combined Treatment on a Forest Plot.

These findings reveal the desirable effect of a strict control method for blood glucose in managing the risk of DR among patients with diabetes mellitus. Suboptimal control of blood glucose and pressure are two of the risk escalators for DR and microvascular problems in diabetes patients (43, 44). Intensive glucose control modestly reduces HbA1c levels, which has been reported to substantially decrease the risk of diabetic retinopathy (DR) in patients with diabetes.

These results align with a previous study, highlighting the significant benefit of intensive glycemic control in limiting DR progression (Risk ratio = 0.77) (45). The landmark DCCT study reported a 54% and 76% reduction in the risk of DR occurrence and worsening in diabetes patients in the intensive group who achieved a mean 7.2% HbA1c compared to the 9.2% mean HbA1c achieved by those in the standard control. The EDIC follow-up survey demonstrated that these benefits persisted after four years, with those in the intensive group maintaining a risk reduction for the progression of DR even though there had been gradual convergence in HbA1c levels between the groups (46). A different study on newly diagnosed cases of type 2 diabetes mellitus also showed a 21% decrease in DR risk of progression in the intensive control (mean HbA1c = 7%) than the group within the standard control (mean HbA1c = 7.9%) (47). These observations also agree with the report of a Cochrane review performed on DM2 patients on intensive control (48). Despite the revealed benefits of intensive glucose control on DR occurrence and its progression, the main disadvantage has been reported to be the frequent hypoglycemic episodes and initial worsening of DR (27). However, as revealed in the DCCT study, these events were transient, while the benefits of intensive glycemic control were long-term (27). Considering these revelations, the current study emphasizes the need for diabetes patients to maintain optimal levels of HbA1c (lower than 7.0%) to reduce the risk of DR and other long-term diabetes mellitus complications.

Glycemic variability (blood glucose levels change throughout the day), as opposed to HbA1c as an indicator of glycemic control level, has been the focus of some recent surveys (49). Glucose fluctuations have been proven to be a more accurate indicator of oxidative stress-triggering effects (49). Picconi et al. revealed a relationship between early damage to the neuroretina cell structures and glucose fluctuations. They proposed addressing glycemic variability even in cases with reasonable glycemic control (50). This discovery was also reported in Virk et al.’s study that confirmed that significant variability in HbA1c is a predisposing factor for DR. Contrarily, the DCCT study reported that glycemic variability within the day has no apparent influence on the occurrence of microvascular problems outside the effect of the mean glucose level (46). However, the current study did not find sufficient data to perform a meta-analysis to comprehensively determine and comment on the role of glycemic variability on the risk of development and progression of DR in diabetes mellitus patients. Therefore, more clinical trials or observational studies with large sample sizes are required to investigate and bring a consensus on the role of within-day glycemic variation on microvascular complications, including DR, in patients with diabetes mellitus.

Study Strengths and Limitations

This systematic review and meta-analysis give an elaborative analysis of the impact of glycemic control on the development and progression of DR using quality data from peer-reviewed and approved primary studies. A key strength is the comprehensive search strategy employed across various electronic databases, which enhanced the likelihood of capturing significant glycemic control studies and DR studies. However, despite all the effort put into this study to enhance the accuracy of the conclusions reached, it is essential to consider potential limitations before interpreting and using the results obtained from this elaborate study. First, the data pooled across various studies might have suffered from publication bias. Publication bias emerges typically due to the likelihood of studies with significant results being published as opposed to those with non-significant results. The heterogeneous nature of the studies used is also a notable limitation since it was not easy to pinpoint the origin of study heterogeneity due to the limited number of studies for specific outcomes to perform subgroup and meta-regression analyses.

Conclusion

This study demonstrates that glycemic control significantly influences the likelihood of DR occurrence and further development into more sight-threatening phases. Intensive glycemic control to maintain the HbA1c to optimal levels has been found to have beneficial effects in reducing the risk of occurrence and worsening status of DR. In contrast, poor glycemic control exacerbates the likelihood of DR complications in patients with diabetes. However, other indicators of the degree of glycemic control, such as glycemic variability, also affect DR development and other diabetes mellitus complications. Further observational studies are needed to evaluate how glycemic control and glucose variability influence the occurrence and progression of DR to more severe stages.

Disclosures

Author contributions

The authors have reviewed the final version to be published and agreed to be accountable for all aspects of the work.

Ethics statement

Not applicable.

Conflict of interest

The authors declare no competing interest.

Funding

The authors have declared that no financial support was received from any organization for the submitted work.