Diabetes affects 29.3 million people in the United States [
Metformin has been used to treat hyperglycemia since 1950s and becomes the first-line and the most widely used oral medication for type 2 diabetes (T2D) recently [
In this retrospective cross-sectional study, we assessed the relationship between long-term oral metformin and severity of DR in patients with T2D for 15 years or longer. Metformin-treated patients were found less likely to have SNPDR/PDR versus nonmetformin-treated patients, regardless of gender and race, and independent of glycated hemoglobin (HbA1c) levels. The association between metformin and a significant lower rate of SNPDR/PDR was not confounded by the use of insulin or sulfonylurea.
This study was a retrospective, cross-sectional study that included all patients with a diagnosis of DR and a history of T2D ≥ 15 years within Henry Ford Health System from January 1990 to September 2013. Cases were identified with an International Classification of Diseases- (ICD-) 9 diagnosis of DR. The index date of cases was the date of the most recent visit. Two groups were included in this study. One was the metformin-treated group that comprised all those who received oral metformin for the last 5 or more consecutive years. The other was the nonmetformin control group that consisted of cases that had no record of metformin use for T2D therapy. Concurrent use of insulin or other hypoglycemic agents was not excluded in both groups. Cases with intermittent metformin use, history of retinal detachment, and coexistence of other retinal vascular disorders, such as retinal vein occlusion and wet-type age-related macular degeneration, were excluded. A history of T2D ≥ 15 years and metformin treatment ≥ 5 years were chosen because prevalence of visual complication was notably increased when duration of diabetes exceeds 15 years [
From each case, the following information were collected: (1) demographic features, including age, gender, and race; (2) general information of T2D progression and management, such as date of T2D diagnosis, 5-year HbA1c levels (including the lowest, median, and highest values), and medications for T2D (including metformin, insulin, and other hypoglycemic agents); (3) grading of DR and prior DR treatments such as laser photocoagulation and intravitreal antivascular endothelial growth factor (VEGF) agents; and (4) ophthalmic history such as retinal vein occlusion and age-related macular degeneration.
Diagnosis of DR was made by a staff retinal specialist. The grading of DR severity and the presence of clinically significant macular edema (CSME) were determined according to the Early Treatment Diabetic Retinopathy Study (ETDRS) grading standards [
The rate of SNPDR/PDR was compared between the metformin group and the nonmetformin control group, in all cases as well as in subgroups categorized according to sex and race. The odds ratio (OR) of SNPDR/PDR was calculated to assess the association between hypoglycemic treatment and severity of DR in all cases as well as in treatment-stratified cohorts.
Statistical analysis was performed using SAS software, version 9.2 (SAS Institute Inc., USA). Group difference of categorical variables was determined using standard chi-square test in the presence of nonsparse data, and Fisher exact test in the presence of sparse data. Sparsity was defined as the presence of expected cell counts less than 5. Group comparisons of numeric variables were made using two-sample
This retrospective cross-sectional study included 335 DR cases with T2D for 15 years or longer. Of all the cases, 193 (58%) had used metformin for at least 5 consecutive years and 142 (42%) had no record of metformin use. Demographic features were comparable between the two groups regarding age, gender, and race (Table
Demographic features and clinical characteristics of the patients.
Characters | Nonmetformin treated ( |
Metformin treated ( |
|
---|---|---|---|
Age (years) | 74.2 ± 9.6 | 73.8 ± 10.6 | 0.73 |
Gender—number (%) | |||
Male | 69 (49) | 99 (51) | 0.67 |
Female | 73 (51) | 94 (49) | |
Race—number (%) | |||
Black | 90 (63) | 104 (54) | 0.06 |
White | 45 (32) | 81 (42) | |
Other | 7 (5) | 8 (4) | |
Duration of diabetes (years) | 15.7 ± 7 | 15.1 ± 6.7 | 0.39 |
HbA1c (%) | |||
5-year low | 6.9 ± 1.1 | 7 ± 1.3 | 0.40 |
5-year high | 9.4 ± 2 | 9.5 ± 1.8 | 0.81 |
5-year median | 8.2 ± 1.3 | 8.2 ± 1.4 | 0.58 |
Treatment for diabetes—number (%) | |||
Insulin | |||
Yes | 132 (93) | 141 (73) | |
No | 10 (7) | 52 (27) | |
Other oral hypoglycemic agent | |||
Sulfonylurea | 54 (38) | 142 (74) | |
Other oral hypoglycemic agents | 3 (2) | 6 (3) | |
No other oral hypoglycemic agents | 85 (60) | 45 (23) | |
Treatment for diabetic retinopathy—number (%) | |||
Focal/grid laser photocoagulation | |||
Yes | 64 (45) | 73 (38) | 0.22 |
No | 78 (55) | 120 (62) | |
Pan-retinal photocoagulation | |||
Yes | 61 (43) | 48 (25) | |
No | 81 (57) | 145 (75) | |
Intravitreal anti-VEGF reagent | |||
Yes | 12 (8) | 23 (12) | 0.37 |
No | 130 (92) | 170 (88) |
Fewer percentage of patients used insulin in the metformin group than that in the nonmetformin group (141 of 193, 73% versus 132 of 142, 93%) (
Local therapies for DR including focal/grid laser photocoagulation, pan-retinal photocoagulation (PRP), and intravitreal anti-VEGF injection were applied in both groups. Significantly fewer patients in the metformin group received PRP versus the nonmetformin group (48 of 193, 25% versus 61 of 142, 43%) (
SNPDR/PDR was diagnosed among 48 (25%) of the patients who used metformin for 5 years or longer, compared with 67 (47%) of those who never used metformin (
Logistic regression analysis between patients with mild/moderate NPDR and those with SNPDR/PDR for variables associated with the use of metformin, insulin, or sulfonylurea.
Diabetes treatment | Severity of DR | |||
---|---|---|---|---|
Mil/Mod NPDR number (%) | SNPDR/PDR number (%) | OR (95% CI) | ||
Effect of metformin | ||||
Nonmetformin users ( |
75 (53) | 67 (47) | ||
Metformin users ( |
145 (75) | 48 (25) | ||
Effect of insulin | ||||
Noninsulin users ( |
54 (87) | 8 (13) | ||
Insulin users ( |
166 (61) | 107 (39) | ||
Effect of sulfonylurea | ||||
Nonsulfonylurea users ( |
75 (54) | 63 (46) | ||
Sulfonylurea users ( |
143 (73) | 54 (27) |
DR: diabetic retinopathy; Mil/Mod NPDR: mild/moderate nonproliferative diabetic retinopathy; SNPDR/PDR: severe nonproliferative diabetic retinopathy/proliferative diabetic retinopathy; OR: odds ratio;
Long-term metformin treatment was associated with significantly reduced rate of SNPDR/PDR in type 2 diabetes patients. 25% of 193 metformin users were found to have SNPDR/PDR, while 47% of 142 cases that never used metformin had SNPDR/PDR (
Since there was notable difference in the use of insulin and sulfonylurea between the two groups (both
Further, comparison with stratification for the hypoglycemic treatments was performed to identify possible confounding effects. The association between metformin and lowered frequency of SNPDR/PDR was persistent in sulfonylurea cohort [OR 0.35 (95% CI, 0.18–0.68),
Odds ratios that indicate the association between the rate of SNPDR/PDR and hypoglycemic therapy in all cases and within each stratum of metformin, sulfonylurea, and insulin users. Metformin- or sulfonylurea-treated patients were significantly less likely to have SNPDR/PDR when all cases were assessed, while insulin users were more likely to have SNPDR/PDR versus those did not use insulin. When stratified for the use of each hypoglycemic agent, metformin showed independent association with lowered frequency of SNPDR/PDR within each stratums of hypoglycemic treatment, with statistical significance in sulfonylurea users, nonsulfonylurea users, and insulin users. However, sulfonylurea-associated significantly lower rate of SNPDR/PDR was only observed in insulin users. Insulin was associated with higher rate of SNPDR/PDR, which persisted in all the four cohorts stratified by metformin or sulfonylurea. Three of the four ORs had borderline
When stratified for the use of metformin, sulfonylurea treatment lost its significant association with reduced rate of SNPDR/PDR [OR 0.54 (95% CI, 0.26–1.09) and 0.69 (95% CI, 0.35–1.36),
It was interesting to note that patients who used insulin were more likely to have SNPDR/PDR in cohorts stratified by either metformin or sulfonylurea [OR 1.96 (95% CI, 0.88–4.38),
Focal/grid laser photocoagulation was the major form of therapy for CSME before anti-VEGF therapy was available. CSME requiring focal/grid laser was identified in 83 of 193 (43%) cases among metformin users, and 69 of 142 (49%) cases among nonmetformin users. No significant difference was found between the two groups (
It is known from the United Kingdom Prospective Diabetes Study (UKPDS) that metformin significantly reduced the risk of many diabetes-related macrovascular and microvascular events when compared with other hypoglycemic therapies in diabetic patients [
As a retrospective study, we were unable to control the use of insulin and sulfonylurea among the patients. A higher percentage of nonmetformin users received insulin treatment versus that of metformin users (93% versus 73%). On the contrary, more patients of the metformin group received sulfonylurea compared with those of the nonmetformin group (74% versus 38%). Insulin therapy is generally advocated when oral medications are insufficient in controlling blood glucose. It is likely that nonmetformin users had poorly control glycemia at some point, which led to prescription of insulin and might contribute to a higher risk of DR in this group. The homogeneity in the duration of diabetes and 5-year HbA1c levels among the two groups helped provide a relatively equal comparison regarding the risk factors of DR in this study, since these two clinical parameters are believed to be the most important factors that impact DR progression [
The initiation of DR was associated with hyperglycemia, nonenzymatic glycosylation, and local oxidative and inflammatory stresses [
One major molecular target of metformin is AMP-activated protein kinase (AMPK), a crucial cellular energy sensor that regulates lipid and glucose metabolism. Activation of AMPK was intimately associated with metformin’s action in regulating the metabolic processes [
Metformin did not change the rate of CSME in this study. Macular edema is mediated in substantial part by VEGF. Currently, the most successful treatment of macular edema is intravitreal injection of anti-VEGF agents [
Comparison with stratification for use of insulin or sulfonylurea treatment confirmed that the association between metformin and reduced DR severity was independent of either hypoglycemic therapy. On the other hand, although sulfonylurea was also correlated with a lower rate of SNPDR/PDR in all cases, this correlation disappeared in cohorts stratified by the use of metformin. Since 74% of sulfonylurea users also used metformin in this study, it is possible that the effect of sulfonylurea on DR severity was modified by metformin. Our data previously supported findings by the ADVANCE trial that sulfonylurea does not prevent or reduce DR progression despite reaching the glucose control goals in T2D patients with vascular complications or risk factors of vascular diseases [
In contrast to metformin and sulfonylurea, insulin was associated with a higher rate of SNPDR/PDR in all patients with an OR of 4.35. Stratified analysis reduced the level of significance but did not reverse this trend, with the OR values varied from 1.96 to 6.91. Similar to what we found, some other clinical studies also associated insulin with increased risk of DR in T2D patients [
Our finding that long-term metformin treatment was associated with significantly reduced rate of SNPDR/PDR was supported by previous clinical and basic research reports on the protective effects of metformin against microvascular complications of diabetes. In particular, this study adds first-hand clinical data on the glycemic-independent effect of metformin on the severity of DR in patients with established T2D. Although there was an imbalance in the use of insulin and sulfonylurea between the two groups, we used stratified comparison to adjust for potentially confounding effects of both agents. The consistence of the crude and adjusted OR for metformin treatment helps to rule out the confounders. The difference between the crude OR and adjusted OR for sulfonylurea suggested that its effect was modified by metformin and insulin. Insulin exhibited an opposite effect on DR severity compared with metformin and sulfonylurea. As a single-centered, retrospective, cross-sectional study, we could only control a few most prominent risk factors of DR. Unmeasured confounding may remain, such as blood pressure, blood lipid, and degree of microproteinuria. Future large-scale prospective studies are warranted for a better understanding of how metformin influences the progression of DR in T2D patients.
Our results should be interpreted as a significant association between long-term metformin treatment and reduced severity of DR in patients with established T2D. This association exists in the lack of a different HbA1c level, persists across the gender and racial cohorts, and is not confounded by sulfonylurea or insulin treatment. Insulin users were more likely to have severe DR versus noninsulin users. Metformin might be used for the purpose of reducing DR progression in patients with long history of T2D. Our investigation is a conceptual study which calls for large-scale future studies for a full evaluation of the exact role of metformin in the progression of DR.
Part of the results in the manuscript was presented as a meeting abstract [
The authors declare that there is no conflict of interests.
This work was supported by grants from the Alliance for Vision Research and fund from the Henry Ford Health System (Xiaoxi Qiao). The authors are also grateful to Dr. Lee Jampol from Jesse Brown Veterans Affairs Medical Center and Northwestern Memorial Hospital, Chicago, IL, USA, for his valuable comments on the manuscript. The authors appreciate Mr. Gordon Jacobsen at Public Health Sciences at Henry Ford Health System for his assistance on statistical analysis.