Dose-response association of metformin use and risk of age-related macular degeneration among patients with type 2 diabetes mellitus: a population-based study

Background: Recent studies have demonstrated that patients with type 2 diabetes mellitus (T2DM) who receive metformin have a decreased risk of developing age-related macular degeneration (AMD). However, other studies have also suggested that metformin may increase the risk of AMD development. Therefore, this study investigated the association between treatment with metformin and the risk of AMD in patients with T2DM by using Taiwan’ National Health Insurance Research Database. Methods: Patients who received a diagnosis of new-onset T2DM between 2002 and 2013 were enrolled in this study. The patients were divided into patients treated and not treated with metformin to evaluate the risk of AMD after 5 years of follow-up. The logistic regression was used to estimate the risk of AMD associated with the intensity of treatment with metformin. Result: A total of 7 517 patients (103.16 patients per 10,000 people) developed AMD in 5 years after DM diagnosis. After adjusting for the relevant variables, patients with T2DM treated with <5 defined daily dose (DDD)/month of metformin had a lower risk of AMD (odds ratios [OR]: 0.93; 95% confidence interval [CI]: 0.88 0.99). Patients treated with >25 DDD/month of metformin had a higher risk of AMD (OR: 1.39; 95% CI: 1.08-1.78). Conclusion: Metformin use may be associated with a risk of AMD among patients with T2DM in a dose-dependent association manner, with the greater benefit at lower DDD/month. However, higher DDD/month exhibited an increased risk of AMD.


Introduction
Age-related macular degeneration (AMD) is the major cause of central irreversible blindness or visual loss among patients aged >50 years in developed countries (Chakravarthy et al., 2010).AMD is typically classified into early and late forms.Patients with early AMD are usually asymptomatic, whereas patients in the late stage of AMD may develop severe progressive vision loss.AMD can be categorized into the 2 following clinical types: nonexudative (dry) and exudative (wet) AMD (Fernandes et al., 2022).Incidence rates of AMD lesions increase substantially with age (Mitchell et al., 2002).
The pathogenesis of AMD is complicated and can be associated with several risk factors, including aging, ocular disorders, systemic diseases, cigarette smoking, diet, body mass index, genetic susceptibility, and environmental conditions (Lim et al., 2012;Ersoy et al., 2014).Studies have investigated whether type 2 diabetes mellitus (T2DM) play a role in AMD development and progression.Several studies have found a positive correlation between T2DM and AMD (Nitsch et al., 2008;Topouzis et al., 2009;Chen et al., 2014;He et al., 2018), whereas some other studies expressed no such effect (Fraser-Bell et al., 2008;Xu et al., 2009).In addition, an inverse association was observed in the Age-Related Eye Disease Study (Clemons et al., 2006).
Several retrospective clinical studies demonstrated that metformin may have a potential role in AMD development (Chen et al., 2019;Lee et al., 2019;Blitzer et al., 2021), while active treatment with metformin is associated an increased risk of dry AMD (Eton et al., 2022).In addition, a meta-analysis study show that metformin is not protective against AMD development (Romdhoniyyah et al., 2021).A study reported that treatment with metformin of low-to-moderate doses is associated with a lower risk of AMD, while higher doses of metformin use did not have reduced risk of AMD development (Blitzer et al., 2021).Conflicting data on the association between metformin exposure dosage and the risk of AMD development.Therefore, we conducted a large-scale nationwide study to determine the association between treatment with metformin and the risk of AMD in patients with T2DM by using data from the National Health Insurance Research Database (NHIRD).

Data source
This study used the Longitudinal Health Insurance Database (LHID) from 2001 to 2018 as the study database provided by the Health and Welfare Data Science Center (HWDC) of the Ministry of Health and Welfare in Taiwan.The LHID encompasses data pertaining to every individual who is registered within Taiwan's National Health Insurance (NHI) program.The NHI contains health insurance claims data for 99% of Taiwan's 23 million residents.Disease diagnoses were coded according to the International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM) and ICD, 10th Revision, Clinical Modification (ICD-10-CM).The NHIRD can be used to obtain real-world evidence to support clinical decisions and healthcare policy-making (Chang et al., 2017;Hsieh et al., 2019;Lai et al., 2020).Therefore, we used data from the LHID to analyze the dose-response association of metformin use and risk of AMD among T2DM patients in Taiwan.

Ethics approval
This study was exempted from informed consent because the personal identification data were encrypted and transformed in the LHID.This study protocol was approved by the Central Regional Research Ethics Committee of China Medical University, Taiwan (No. .

Study participants
This study enrolled patients with new-onset diabetes mellitus (DM) aged ≥50 years from 2002 to 2013.DM (ICD-9-CM: 250) was indicated by the presence of 3 outpatient diagnoses.Metformin of the present study was defined according to the Anatomical Therapeutic Chemical (ATC) code A10BA02.The study participant exclusion criteria contained (Chakravarthy et al., 2010) type 1 DM patients, (Fernandes et al., 2022), patients having a diagnosis of AMD before DM, (Mitchell et al., 2002), patients having a diagnosis of AMD in the first year after DM, and (Lim et al., 2012) patients hospitalized within 1 year after DM diagnosis.After selection (Figure 1).There were a total of 728 703 patients with new-onset DM were included in the study.Patients treated with and without metformin were 377 878 patients and 350 825 patients, respectively.

Statistical analysis
All analyses were performed using SAS version 9.4 (SAS Institute, Cary, NC, United States).The chi-square test was used to evaluate differences between patients treated with and without metformin.Multiple logistic regression was used to estimate the odds ratios (ORs) and 95% confidence intervals (CIs) for AMD risk after adjustment for sex, age, income level, urbanization, diabetes severity, and comorbidities.All statistical results with p < .05were regarded as statistically significant.

Results
Table 1 presents the baseline characteristics of all patients.The average age of all patients was 62.06 ± 8.83 years, and 51.42% of all patients were women.Regarding age groups, 23.72% were 50-54 years old, 22.82% were 55-59 years old, 18.00% were 60-64 years old, 13.96% were 65-69 years old, 10.51% were 70-74 years old, and 10.99% were above 75 years old.In patients treated with metformin, the average age was 61.21 ± 8.43 years.

Discussion
This study found that treatment with metformin may be associated with the risk of AMD among patients with T2DM in a dose-response relationship manner.The results suggest that the intensity of treatment with metformin <5 DDD/month is associated with a lower risk of AMD at 5 years after initial DM diagnosis.However, patients with T2DM treated with >25 DDD/month of metformin experienced higher risks of AMD at 5 years.In addition, we found that among patients T2DM treated with metformin, older patients and patients with a higher DCSI score had a higher risk of AMD.Furthermore, patients with T2DM with comorbid sleep disturbance and hypothyroidism had a higher risk of AMD.
DM may play a significant role in the progression and development of AMD.Previous studies have demonstrated a positive correlation between DM and AMD (Nitsch et al., 2008;Topouzis et al., 2009;Choi et al., 2011;Chen et al., 2014;Khotcharrat et al., 2015;Vassilev et al., 2015;He et al., 2018).Several pathophysiological mechanisms may be associated with DM and AMD.Oxidative stress and chronic inflammation may explain the correlation between DM and the risk of AMD.Oxidative stress causes outer blood-retinal barrier degeneration that contributes to AMD progression (Jung et al., 2022), and oxidative stress is a risk factor for the development of insulin resistance through insulin signal disruption (Houstis et al., 2006;Newsholme et al., 2019).
Metformin achieves its antioxidative and anti-inflammatory effects through the activation of AMP-activated protein kinase (AMPK) (Lee et al., 2013;Zhao et al., 2020) and reduction in reactive oxygen species (Hou et al., 2010).Recent studies have  demonstrated that AMPK plays a major role in the regulation of systemic glucose homeostasis and metabolic stress.AMPK is a conserved energy sensor and master regulator of glucose metabolism, which restores cellular energy balance during metabolic stress (Garcia and Shaw, 2017) and might be involved in AMD pathogenesis (Brown et al., 2019a).Metformin inhibited oxidative stress on human retinal pigment epithelium (RPE) cells by stimulating the AMPK signaling pathway in a mouse model of AMD (Xu et al., 2018).Antioxidant and anti-inflammatory effects of metformin can protect the RPE cells against the lesions of early AMD (Jiang et al., 2022).Our findings demonstrated that patients with T2DM treated with <5 DDD/month of metformin had a lower risk of AMD at 5 years after initial DM diagnosis.Animal studies and physiology studies have suggested that metformin may play a beneficial role in the prophylaxis of AMD (Amin et al., 2022).Several studies suggested that metformin may have a role in AMD development and progression (Romdhoniyyah et al., 2021;Chen et al., 2019;Blitzer et al., 2021).A large-scale study reported the protective outcomes of metformin use in the development of AMD, with a 42% reduction (Brown et al., 2019b).A systematic review and meta-analysis study found that treatment with metformin is not associated with a significant lower risk of AMD (Romdhoniyyah et al., 2021).Another large case-control study reported that treatment with metformin is associated with a lower risk of AMD, with the lowest ORs associated with low-to-moderate doses (Blitzer et al., 2021).This study suggests that metformin use more than 2 years in patients aged 55 years and older is correlated with 5%-10% decreased odds ratio of AMD development.
Our findings revealed that patients treated with >25 DDD/month of metformin exhibited a higher risk of AMD after 5 years of follow-up.A case-control study observed no significant associations between AMD risk and cumulative duration or exposure of treatment with metformin (Lee et al., 2019).Another study with a small sample size found a conflicting relationship between metformin exposure and dry AMD, with the findings based on assessment of metformin cumulative dosage and the intensity of the treatment with metformin (Eton et al., 2022).A study based on medical claims from a large US insurer also indicated that conflicting associations between metformin exposure and development of dry AMD.Cumulative metformin dosage model showed a significant association between the risk of dry AMD with cumulative dosage, with the lowest dosage quartile associated with a decreased risk of dry AMD and the highest dosage associated with an increased risk (Eton et al., 2022).Active treatment with metformin is associated with an increased risk of dry AMD, whereas prior treatment with metformin is associated with decreased risk (Eton et al., 2022).Our findings are similar to a large nationwide case-control study revealed that the use of metformin may protect against AMD development in a dose-dependent manner (Blitzer et al., 2021).This research found that metformin may be useful as a preventive treatment for AMD with strongest at low to moderate doses, while higher dose did not have reduced risk of AMD development.This study reported that doses of greater than 1080 g of metformin use more than 2 years did not have decreased risk of AMD development, while was particularly for low to moderate doses of metformin revealed the greatest potential benefit (Blitzer et al., 2021).The greatest reduction in AMD risk was observed at metformin doses of 271-600 g over 2 years with an OR of 0.91, and doses of 1-270 g and 600-1080 g over 2 years were also correlated with decreased OR, 0.93 and 0.95, respectively (Blitzer et al., 2021).Vitamin B12 deficiency may play a role in AMD development in patients with T2DM receiving long-term treatment with metformin.Treatment with metformin can induce vitamin B12 malabsorption by increasing bacterial overgrowth, altering gut bacterial flora in the enteric canal, and binding to the vitamin B12 intrinsic factor (Zhang et al., 2016).Malabsorption contributes to a decreased serum vitamin B12 plasma level.Current evidence suggests that metformin impairs vitamin B12 status in a dose-dependent and duration-dependent association manner (Infante et al., 2021).A meta-analysis suggest a negative association between metformin use and vitamin B12 plasma levels in T2DM patients (Chapman et al., 2016), and higher cumulative exposure and longer duration of metformin treatment were associated with an increased risk of vitamin B12 deficiency (Khattar et al., 2016;Huang et al., 2022a;Huang et al., 2022b;Huang et al., 2023).Patients received metformin with therapy duration ≥ 5 years and a metformin dose of ≥ 1500 mg/day for a duration of at least 6 month was associated vitamin B12 deficiency, especially the highest risk has been found in patients with a daily metformin dose of ≥ 2000 mg (Infante et al., 2021).T2DM patients received metformin dosage of >2,000 mg/day increased the risk of vitamin B12 deficiency 22 times (Ko et al., 2014).However, the underlying mechanism accounting for metformin-induced vitamin B12 deficiency in patients with long-term and high-dose of metformin use remains unclear.Nevertheless, the proposed underlying mechanisms due to the alteration in small intestine motility, resulting in small intestinal bacterial overgrowth and subsequent B12 deficiency or by directly decreasing vitamin B12 absorption (Ting et al., 2006;Damiao et al., 2016); malabsorption leads to a decreased serum vitamin B12 level.Vitamin B12 and homocysteine may play a role in reducing the risk of AMD.Vitamin B12 deficiencies, folate, or elevated serum homocysteine levels were used as predictors of a high risk of AMD (Gopinath et al., 2013).Vitamin B12 is essential for the conversion of homocysteine to methionine in the methionine cycle (Allen, 2012).Vitamin B12 deficiency can impair the remethylation of homocysteine; moreover, metformin-induced vitamin B12 deficiency is potentially associated with hyperhomocysteinemia (Russo et al., 2011).An animal study found that excess homocysteine levels on the structure and function of retinal pigment epithelial that contribute to the development of AMD-like features (Ibrahim et al., 2016).Human study have reported that plasma homocysteine level was elevated in patients with AMD and highlighted a strong correlation between hyperhomocysteinemia and the development of AMD (Huang et al., 2015).A cross-sectional study found that increased total serum homocysteine and low vitamin B12 concentrations were independently associated with a higher risk of AMD (Rochtchina et al., 2007).The beneficial effects of vitamin B12 and folate on the risk of AMD are partly mediated by lowering the concentration of serum homocysteine (Rochtchina et al., 2007).Although treatment with metformin can decrease the risk of AMD (Brown et al., 2019a;Xu et al., 2018;Jiang et al., 2022), when long-term and high-dose or high cumulative dosage of metformin use were associated with biochemical B12 deficiency and hyperhomocysteinemia (Russo et al., 2011), may offset the protection effect of metformin and could lead to enhance the risk of AMD (Rochtchina et al., 2007).Routine assessment of vitamin B12 levels in individuals treated with metformin should be considered (Aroda et al., 2016;Al-Hamdi et al., 2020).Due to the clinical benefits of metformin use, its associated side effects such as metformin-induced vitamin B12 deficiency is often overlooked in T2DM patients.However, the diagnosis of metformin-induced vitamin B12 deficiency may be difficult (Al-Hamdi et al., 2020).The underlying mechanisms of metformin cumulative dosage and the risk of AMD remain unclear.Thus, further prospective clinical trials are warranted to investigate the protective effect of metformin on AMD, especially regarding duration and dosage of therapy.
Our findings showed that T2DM patients treated with metformin, older patients, and having a higher DCSI score linked to an increased risk of AMD.Previous studies have identified several risk factors for AMD, including aging, ocular disorders, systemic diseases, smoking, diet, genetic susceptibility, and environmental risk factors (Lim et al., 2012), with aging being the strongest risk factor (Aldebert et al., 2018).In the general population, vitamin B12 plasma levels decline with age, and thus, the prevalence of vitamin B12 deficiency increases with age.Age is a strong risk factor for the development of AMD, and individuals aged <50 years have a lower risk of AMD (Jiang et al., 2022) compared with older adults, who also have a higher risk of vitamin B12 deficiency (Gonzalez-Gross et al., 2007).The DCSI is a useful tool for adjusting for baseline severity of disease and predicting mortality and the risk of hospitalization among patients with DM (Young et al., 2008).Our study showed that patients with T2DM treated with metformin with higher DCSI scores had an increased risk of AMD.Thus, DCSI may be used as an indicator for assessing the risk of AMD development.
Our study results demonstrated that patients with T2DM treated with metformin and with comorbid sleep disturbance and hypothyroidism had a higher risk of AMD.A Taiwan populationbased study indicated that insomnia is an independent indicator of an increased risk of AMD (Tsai et al., 2020).Thyroid disease is associated with an increased risk of AMD (Xu et al., 2021).
This study adopted a population-based design and used data from the NHIRD.Because we included the entire Taiwanese population in this study, the sample size is large and sufficient for reducing selection bias and providing high-quality data.Second, the characteristics of the database provide sufficient statistical power for investigating the association between treatment with metformin and the risk of AMD among patients with T2DM.Finally, the intensity of treatment with metformin (DDD/month) was <5, 5-15, 15-25, >25 for determining the relationship between patients with T2DM and the risk of AMD.
This population-based cohort study has several limitations.First, information regarding family histories of AMD among patients with T2DM was unavailable.Second, patients' personal data and their lifestyle habits, such as body mass index, cigarette smoking habits, alcohol consumption, dietary habits, and physical activity (factors that are associated with AMD risk), were unavailable.Due to the limitations of the Taiwan National Health Insurance inpatient medical claims system, the information on the medication dosage during hospitalization was lacking from the NHIRD.Therefore, the use of metformin during hospitalization was not included in the present study, which may result in an underestimation of metformin's DDD in our study.Third, the diagnoses of AMD and other comorbidities were coded in accordance with the ICD-9-CM and ICD-10-CM.Nonetheless, the NHI Bureau of Taiwan randomly reviews the charts and interviews patients to assess the accuracy of the diagnoses, which improves the accuracy and validity of the NHIRD.Fourth, Information regarding biochemical parameters (e.g., fasting glucose, HbA1C, urine protein) is unavailable in the database but may affect developing AMD factors.The severity of DM and the disease duration of DM may also affect developing AMD.Therefore, the present study enrolled the new-onset DM patients as the study subjects and used the DCSI to adjust the severity of DM to reduce the bias.This study was a nationwide population-based study.Thus, the study results have accuracy and representativeness.Finally, this study is a type of epidemiology observational study that analyzes data from a nationwide database.The study result can only provide evidence to demonstrate that metformin is related to incident AMD.It is essential to obtain more information from other databases or questionnaires to conduct a prospective study or randomized controlled trial to analyze the cause-effect relation in future research.

Conclusion
In conclusion, this study provides evidence that treatment with metformin may be associated with the risk of AMD among patients with T2DM in a dose-dependent association manner.Patients treated with <5 DDD/month of metformin had a decreased risk of AMD at 5 years.However, >25 DDD/month exhibited an increased risk of AMD.

FIGURE 1
FIGURE 1Patient selection process.

TABLE 1
The characteristics of patients with diabetes mellitus.

TABLE 1 (
Continued) The characteristics of patients with diabetes mellitus.

TABLE 2
Five-year follow-up of incident age-related macular degeneration.

TABLE 2 (
Continued) Five-year follow-up of incident age-related macular degeneration.