The Correlation between Metformin Use and Incident Dementia in Patients with New-Onset Diabetes Mellitus: A Population-Based Study

The evidence of metformin’s effect on dementia is conflicting. This study investigates the association between metformin use and the risk of dementia among patients with diabetes mellitus (DM). This study included patients with new-onset DM between 2002 and 2013. We divided the patients into patients who used metformin and patients who did not. Two models were used to assess metformin use: the cumulative defined daily dose (cDDD) of metformin use and the intensity of metformin use. This study with 3-year and 5-year follow-ups investigated the risk of dementia among patients with DM who used metformin. At the 3-year follow-up, patients who received cDDD < 300 had an odds ratio (OR) of developing dementia of 0.92 (95% confidence interval [CI] = 0.89–0.96); patients who used metformin at intensities <10 and 10–25 DDD/month had ORs of 0.92 (95% CI: 0.87–0.97) and 0.92 (95% CI: 0.85–1.00), respectively. Metformin use at cDDD 300–500 (OR = 0.80, 95% CI = 0.56–1.15) or >500 (OR = 1.48, 95% CI = 0.48–4.60) or at an intensity >25 DDD/month (OR = 0.84, 95% CI = 0.60–1.18) were not associated with an incident of dementia. There were similar results at the 5-year follow-up. Patients with a low intensity of metformin use had a lower risk of dementia. However, higher doses of metformin with higher intensity exhibited no protective role in dementia. Prospective clinical trials are warranted to evaluate the actual underlying mechanisms between metformin dosage and the risk of dementia.


Introduction
Dementia is one of the main neurodegenerative disorders (NDs) in the elderly [1]. Alzheimer's disease (AD) is the most common type of dementia among older adults and is characterized by progressive cognitive impairment and loss of memory, accounting for 60-80% of all cases [2]. The hippocampus is vulnerable to damage at early stages of AD.
Several studies have indicated that insulin signaling is involved in brain functions, such as cognition and memory, which are impaired in the brains of AD patients and AD experimental models. Neuronal insulin resistance (IR) is also present in the brain and can be induced by amyloid β-peptide oligomers in primary cultures of hippocampal neurons in mice and monkeys [3][4][5]. IR in diabetes mellitus (DM) and obesity has been linked with functional and structural brain change in AD brains and is associated with a risk of developing AD [6]. Evidence has indicated a clear positive association between type 2 diabetes mellitus (T2DM) and AD risk [7]; patients with T2DM are at higher risk of dementia. An epidemiological study shows that T2DM patients has been linked to a 1.5-2-fold increased risk for developing mild cognitive impairment [8]. Metformin is an orally effective and insulin-sensitizing antidiabetic medication and is the first-line treatment of T2DM for most patients [9]. Metformin has been suggested to potentially provide neuroprotective effects through the mediation of inflammatory response inhibition [10] and the improvement of cognitive function [11]. Several studies have demonstrated that T2DM patients who receive metformin exhibit a lower risk of dementia [12][13][14][15].
Several possible mechanisms have been proposed that link metformin use with the risk of dementia [16,17] [18]. However, several longitudinal studies have challenged findings on the protective role of metformin during dementia pathogenesis [19][20][21]. More studies have instead indicated that metformin use may increase the risk of dementia [16,[22][23][24]. Therefore, the correlation between metformin use for T2DM and the risk of dementia warrant further research. However, few epidemiological studies have employed a nationwide database to examine the association between metformin use and the risk of dementia. We conducted a population-based study to investigate the association between metformin use and the risk of dementia in patients with T2DM.

Data Source
This study was based on Taiwan's National Health Insurance Research Database (NHIRD) released by the Health and Welfare Data Science Center (HWDC) at the Ministry of Health and Welfare in Taiwan. The NHIRD contains the information of beneficiaries from the Taiwanese National Health Insurance (NHI) program and can serve as a foundation for the procurement of real-world evidence to support clinical decisions and healthcare policymaking [25,26]. The NHI is a government-run, single-payer, national social insurance program that has operated since 1995. As the NHI covers more than 99% of residents in Taiwan, we utilized the NHIRD to examine the risk of dementia among T2DM patients receiving metformin.

Study Subjects
The subjects of this study were patients with new-onset DM aged ≥50 from 2002 to 2013, who were enrolled in this study. DM was defined as receiving a DM diagnosis at least three times a year, in accordance with the International Classification of Diseases, Ninth Revision, Clinical Modification [ICD-9-CM] code 250 [27], and the first DM diagnosis date was defined as the index date. Metformin was defined in accordance with the anatomical therapeutic chemical code A10BA02 and measured based on the first year after the index date. To reduce bias, we excluded the following patients: those with type 1 DM, those who had a diagnosis of dementia before DM or a diagnosis of dementia in the first year after DM, and those who were hospitalized within 1 year after DM diagnosis. A total of 736,473 new-onset patients with DM were included from 2002 to 2013. The study included 382,328 patients in the exposure group who received metformin medication within the first year after their DM diagnosis and 354,145 patients in the comparison group who did not receive any metformin medication. The selection process for study subjects is shown in Figure 1.

Study Design
This was a population-based study that followed patients with DM for 3 and 5 years to investigate the risk of dementia associated with metformin use. We assessed metformin intake using the defined daily dose (DDD), a standard measure of drug use and exposure. The DDD is defined by the World Health Organization as the assumed average maintenance dose per day in adults, although it may not reflect the recommended or prescribed daily dose [28]. We considered a 1-year observation period before initiating metformin treatment, following DM diagnosis, and used a baseline dose of 2 g metformin, based on the DDD [29]. The study subjects were divided into 3 groups according to the following ranges of the cumulative DDD (cDDD) of metformin: 0, <300, 300-500, and >500. In addition, the study subjects were divided into 4 groups according to the following ranges of average monthly DDD: 0, <10, 10-25, and >25. All patients were observed at 3 and 5 years after their DM diagnosis. Dementia was indicated by 3 or more outpatient visits for a dementia diagnosis within 1 year, according to the ICD-9-CM codes 290, 294.1, 331.0, and 331.82 and the ICD-10-CM codes F00-F03, F05.1, G30.0, G30.1, G30.8, and G30.9. The first dementia diagnosis date was defined as the date of incident dementia. The comorbidities consisted of hypertension

Study Design
This was a population-based study that followed patients with DM for 3 and 5 years to investigate the risk of dementia associated with metformin use. We assessed metformin intake using the defined daily dose (DDD), a standard measure of drug use and exposure. The DDD is defined by the World Health Organization as the assumed average maintenance dose per day in adults, although it may not reflect the recommended or prescribed daily dose [28]. We considered a 1-year observation period before initiating metformin treatment, following DM diagnosis, and used a baseline dose of 2 g metformin, based on the DDD [29]. The study subjects were divided into 3 groups according to the following ranges of the cumulative DDD (cDDD) of metformin: 0, <300, 300-500, and >500. In addition, the study subjects were divided into 4 groups according to the following ranges of average monthly DDD: 0, <10, 10-25, and >25. All patients were observed at 3 and 5 years after their DM diagnosis. Dementia was indicated by 3 or more outpatient visits for a de-

Statistical Analysis
We conducted all analyses using SAS version 9.4, and the statistical significance level is a p value of <0.05. We used a chi-squared test to assess whether there were differences in baseline characteristics between patients receiving metformin and patients not receiving metformin. Multiple logistic regression with adjustments for variables was used to estimate the risk of dementia in different cDDDs of metformin use and intensities of metformin use (expressed as DDD/month). The adjusted variables in the logistic model were gender, age, income level, urbanization, DCSI, and comorbidities.

Discussion
There have been limited large-scale epidemiological studies examining the potential link between metformin intake and dementia risk among T2DM patients. However, in our study, we found that metformin use was associated with increased dementia risk among patients with T2DM. Our results indicated that patients receiving <300 cDDD of metformin and <10 and 10-25 DDD/month exhibited a reduced risk of dementia at both the 3-and 5-year follow-ups. Conversely, we observed no protective effect against dementia for cDDD > 300 and >25 DDD/month during the 3-and 5-year follow-ups among patients with T2DM. Additionally, we found that among patients with T2DM who received metformin, an increased risk of dementia was associated with older age and a higher DCSI score.
Our results indicated that patients with DM who received a metformin at cDDD < 300, <10 DDD/month, or 10-25 DDD/month exhibited a lower risk of dementia. Several mechanisms explaining the beneficial role of metformin in the prophylaxis of dementia have been proposed. Metformin may have neuroprotective effects among older patients with T2DM. Activation of adenosine monophosphate-activated protein kinase (AMPK) by metformin could partly help in realizing the minor protective effect on improving cognitive function. Metformin is an AMPK activator that suppresses hepatic glucose production and increases insulin-stimulated glucose uptake [30]. AMPK plays a major physiological role in regulating the plasticity of the hippocampal synapse and in cognitive impairment [31]. Impaired AMPK function is associated with DM and can affect neurological disorders, such as AD [32]. Metformin also plays a neuroprotective role through the AMPK/mTOR signaling pathway to control inflammatory conditions and improve the oxidative status [33]. Metformin decreased histopathological changes in AD [34], improved cognitive performance and neuronal survival in the hippocampus of animals with diabetes, and can significantly reduce neuroinflammation [35]. Metformin may also promote angiogenesis and neurogenesis in the brain [33]. Several epidemiological studies have revealed that metformin use was associated with a lower risk of dementia [13,14] and better cognitive function [15]. A meta-analysis study reported that metformin use was associated with a lower prevalence of cognitive impairment and fewer cases of incident dementia [12].
One study reported that the use of metformin for longer than 2 years was associated with a lower incidence of ND among elderly patients with T2DM; however, metformin exposure did not significantly affect the development of ND during the first 2 years [8]. Our results revealed that the protective role in dementia afforded by metformin was nonsignificant at the 3-year and 5-year follow-ups at a cDDD of 300-500 and >500 and at 10-25 and >25 DDD/month. Studies have suggested possible harmful mechanisms between metformin use and dementia risk [16,17,22]. Metformin also increased the β-amyloid accumulation [22]. However, the claim that metformin plays a protective role during dementia pathogenesis has been challenged in several longitudinal studies. One study from the United Kingdom revealed that long-term metformin use is associated with an increased risk of AD [19]. Another study indicated that metformin use can induce vitamin B12 deficiencies in a dose-dependent manner [36]. A study from Australia found that metformin-induced vitamin B12 deficiencies were associated with cognitive impairment among patients with DM [20]. A retrospective cohort study in Taiwan demonstrated that metformin use among patients with T2DM may be a risk factor for ND, including dementia and Parkinson's disease [21]. Several studies have indicated that this effect could be due to vitamin B12 deficiencies that are potentiated by metformin usage and that contribute to cognitive impairment [18,37,38]. Several studies have expressed that metformin use was associated with lower plasma vitamin B12 levels [18,39,40]. A metaanalysis study revealed that a negative association between metformin use and vitamin B12 levels in patients with T2DM [18], and greater cumulative exposure to metformin and a longer duration of metformin treatment were associated with a higher risk of vitamin B12 deficiency [41]; a metformin dosage of >2000 mg/day increased the risk of vitamin B12 deficiency 22 times [42]. T2DM patients undergoing metformin treatment, particularly those taking metformin at a large dosage (>2000 mg/day) and for a long duration (>4 years), should be regularly screened for serum vitamin B12 levels [43]. However, the mechanism underlying vitamin B12 deficiency in patients with long-term metformin use remains unclear. Nevertheless, the proposed underlying mechanisms include an alteration in the small intestine's motility, leading to small intestinal bacterial overgrowth and subsequent inhibition of vitamin B12 calcium-dependent intrinsic factor complex absorption [44]; malabsorption leads to a decreased serum vitamin B12 plasma level. Although metformin use may lower the risk of dementia [34,35,45], the B12 deficiencies associated with longterm metformin use and high doses of metformin exhibited no protective role in dementia. Therefore, long-term metformin use is associated with vitamin B12 deficiencies, which may counteract the potential protective benefit in dementia. Vitamin B12 deficiencies play a role in the risk of dementia in patients with T2DM who receive long-term metformin treatment and higher doses. The results of our study are consistent with those of an animal study demonstrating that a low metformin dose (100 mg/kg) may improve scopolamine-induced cognitive impairment, whereas higher doses of metformin resulted in no harmful effect [46]. A lower metformin dose may have been associated with milder DM severity, which could have affected the dementia risk. However, the actual underlying mechanisms between the metformin dosage and the risk of dementia remain unclear, and prospective clinical trials should be conducted in the future.
The DCSI is a useful tool for predicting the risk of hospitalization and mortality among patients with DM [47]. The adapted DCSI (aDCSI) has seven categories of complications and is a modified version of a risk assessment scheme that does not include medical laboratory values [47,48]. Patients with new-onset DM and with higher aDCSI scores exhibited a higher risk of developing dementia [49]. Our study indicated that patients with DM who had higher DCSI scores demonstrated a higher risk of developing dementia. The DCSI may be used as an indicator for estimating the risk of dementia.
Our findings indicated that older patients with DM exhibited a higher risk of dementia, especially among patients with DM who were above 75 years old. Aging is the largest risk factor for the development and progression of dementia. Among older patients, multiple pathologies contribute to the progression of dementia, including dementia with Lewy bodies and vascular changes [50]. Metformin-related vitamin B12 deficiencies have been documented for over 40 years [18], and both T2DM and B12 deficiencies increase with age [51].
Depression is more common among patients with cognitive impairments. Depression is a key risk factor of dementia and significant cognitive decline among patients with DM [52]. We determined that depression was associated with an increased risk of incident dementia. These findings were consistent with previous studies, in which depression was associated with accelerated cognitive decline among patients with T2DM [53]. A meta-analysis revealed that anxiety is significantly associated with an increased risk of dementia [54]. Anxiety has also been recognized as a potentially modifiable dementia risk factor. Additionally, CVD is a major cause of cognitive impairment and dementia among older patients [55]. Finally, COPD may also contribute to dementia. A systematic and meta-analysis study indicated that patients with COPD exhibit a higher risk of dementia [56]. COPD may lead to chronic hypoxemia and pulmonary encephalopathy, which may considerably affect brain dysfunction [57]. The prevalence of cognitive decline among patients with COPD was associated with the severity of the COPD case [58].
Our study has several strengths. First, our study used a population-based design. Patients were selected from the total population of Taiwan; thus, the sample was representative. We believe that the combination of the NHIRD with multiple databases could be used as a powerful research engine. The population-based design may also minimize the selection bias that is common in observational studies. Second, the characteristics of the database provided sufficient statistical power to investigate the risk of dementia among patients with T2DM and who were treated with metformin. Third, we assessed patients 3 and 5 years after their T2DM diagnosis. We categorized metformin use based on cumulative defined daily doses (cDDDs) as <300, 300-500, and >500, and based on monthly dose intensity as <10, 10-25, and >25 DDD. Additionally, we explored the influence of various comorbidities on the risk of dementia in patients with T2DM.
Our study has some limitations to note. First, we did not have access to lifestyle data, such as tobacco use, alcohol consumption, and physical activity, which may impact the development and progression of dementia. Second, the ICD-9-CM and ICD-10-CM codes used to identify T2DM severity and dementia do not provide precise information; therefore, we were unable to conduct a severity-based subgroup analysis. For instance, the NHIRD database lacks information on HbA1c. Third, using the NHIRD led to some validity concerns, such as the accuracy of diagnosis codes; the NHIRD does not contain records of medical conditions. However, the NHI implements various measures to maintain the validity and accuracy of the NHIRD. For example, the NHI randomly reviews medical charts and conducts patient interviews to verify the accuracy of diagnoses. These rigorous processes help ensure the reliability and accuracy of the NHIRD.

Conclusions
The present study did not reveal a dose-response relationship of metformin use with incident dementia in DM patients. Specifically, patients with DM who received lower doses of metformin and had a lower metformin use intensity showed a comparatively reduced risk of dementia. However, higher doses of metformin with higher intensity exhibited no protective role in dementia. The mechanisms between metformin dosage and incident dementia remain unclear. It is necessary to conduct further prospective clinical trials with adequate statistical tests to evaluate the actual underlying mechanisms between metformin dosage and the risk of dementia.  The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Institutional Review Board Statement:
The database of the present study was provided by the Health and Welfare Data Science Center, Ministry of Health and Welfare, Taiwan (HWDC, MOHW). The database is an anonymous database, and the HWDC assigns scrambled random identification numbers to protect the privacy of insured patients. The requirement for informed consent was waived. The study protocol was approved by the Central Regional Research Ethics Committee of China Medical University, Taiwan (No. CRREC-109-011).

Informed Consent Statement: Not applicable.
Data Availability Statement: The National Health Insurance Database used to support the findings of this study were provided by the Health and Welfare Data Science Center, Ministry of Health and Welfare (HWDC, MOHW), under license and so cannot be made freely available. Requests for access to these data should be made to HWDC (https://dep.mohw.gov.tw/dos/np-2497-113.html, accessed on 20 February 2023).