Metformin’s Effects on Cognitive Function from Biovariance Perspective

1. Introduction

Metformin, a drug belonging to the biguanide class, is a widely prescribed medication for managing non-insulin-dependent type 2 diabetes mellitus (T2DM) and has garnered significant scientific attention due to its multifaceted effects on cellular metabolism and its emerging therapeutic potential in diverse areas of medicine. Its well-documented mechanisms, such as enhancing insulin sensitivity, reducing hepatic gluconeogenesis, and improving overall energy metabolism, have paved the way for research into a broad range of possible therapeutic applications [1], spanning from antitumor effects to roles in neuroplasticity and cognitive function. Additionally, metformin’s therapeutic potential has been evaluated in a variety of metabolic and endocrine disorders, such as polycystic ovary syndrome [2].

This growing interest in metformin has spurred numerous studies exploring its effects across various human pathologies, including cardiovascular and metabolic diseases [3,4], inflammatory conditions [5], processes associated with senescence and anti-aging properties [6], and psychiatric conditions [7]. This review examines the intriguing neuropsychiatric effects of metformin, with a focus on its influence on cognitive function and the underlying biocellular mechanisms that contribute to interindividual variation in its effects.

Cognitive decline is a well-recognized complication of diabetes [8], hypothesized to be driven by glycemic fluctuations and disruptions in neuronal metabolism. Similarly, there is documented evidence of a direct link between diabetes and structural changes in white matter integrity [9]. Existing studies, however, present mixed findings on metformin’s effect on cognitive function in both human and non-human subjects, with notable variations across specific population subgroups.

The present review aims to reconcile the conflicting effects of metformin on cognitive function by identifying the studies that report opposing results and proposing a possible biochemical explanation. Documentation of studies was performed through systematic searches of the PubMed and Google Scholar databases. To identify relevant studies, combinations of subject-specific terms were used: “metformin,” “diabetes,” “T2DM,” “white matter,” “cognition,” “dementia,” ”memory,” “inflammation,” “mechanism of action,” “neuro,” “molecular,” “cellular target,” “stroke,” and “endothelium.”

The search focused on articles published in recent years, employing truncation operators to broaden terminological coverage (e.g., “metformin,” “diabet*” “neuro*”, “cognit*”). The selection of studies was based on the following eligibility criteria:

Inclusion criteria: studies published after 2018 and written in English; studies specifically examining the impact of metformin treatment on cognitive function or neurodegenerative disorders; studies analyzing changes in brain structures; clinical trials; relevant preclinical studies; review articles and meta-analyses; studies investigating metformin’s mechanisms of action on inflammation and neurodegeneration.

Exclusion criteria: studies not analyzing the impact of metformin treatment on brain function, studies lacking clear differentiation between the effects of metformin and other oral antidiabetic drugs, studies in languages other than English, small observational studies without adequate control, or studies with notable methodological errors.

Studies investigating metformin’s mechanisms of action on inflammation and neurodegeneration, with emphasis on cellular signaling pathways and neuroprotective processes, were also included. In particular, meta-analyses and review articles were prioritized to provide a comprehensive overview of metformin’s effects.

Relevant articles were selected through a manual screening of search results, facilitating an in-depth evaluation of design characteristics of the identified studies (Figure 1). This approach allowed for a thorough assessment of inclusion and exclusion criteria, while also helping to mitigate the risk of overlooking relevant studies due to database indexing limitations.

In addition to the primary articles identified through the systematic search, additional references cited within the bibliographies of selected studies were reviewed. Preclinical animal studies examining metformin’s effects on neurobiological structures and cellular mechanisms were also included, as these studies offer complementary mechanistic insights to clinical data. Study assessments took into account design characteristics, statistical analysis methods, population sizes, and follow-up durations where applicable, with primary emphasis placed on the authors' conclusions. The Table 1 below provides an overview of the key studies exploring the impact of metformin.

2. Effects of Metformin on Metabolism and the Brain

Metformin is a first-line drug used in the treatment of type 2 diabetes. Its hypoglycemic effect is primarily attributed to a reduction in endogenous glucose production at the hepatic level by inhibiting gluconeogenesis [10]. However, other mechanisms have also been well documented, such as the improvement of insulin sensitivity in peripheral tissues, which promotes glucose utilization by muscles and other tissues. Among the proposed molecular mechanisms are the stimulation of cellular expression of the insulin-sensitive glucose transporter GLUT4 and the inhibition of SHIP2 enzyme activity [11,12,13,14]. The complete basis for the hypoglycemic effect of metformin has yet to be fully elucidated.

The bioavailability of metformin is relatively modest, ranging between 50-60% after oral administration [15]. Its absorption occurs primarily in the small intestine, is influenced by diet [16], and the drug is largely excreted unchanged by the kidneys. An essential characteristic of metformin is that it does not bind to plasma proteins and is not Mmetabolized by the liver. Metformin is widely distributed throughout body tissues. Furthermore, the unabsorbed portion of metformin (approximately 50%) can influence the intestinal microflora, with evidence showing that metformin affects the microbiome [17,18,19]. One study even suggests a possible correlation between an individual’s intestinal microbiome and their response to metformin [20].

Regarding metformin’s action on the brain, it is important to note that animal studies confirm its ability to cross the blood-brain barrier [21,22]. We emphasize that the permeability of the blood-brain barrier varies between species and, additionally, between individuals. Metformin concentrations in cerebrospinal fluid (CSF) are lower than in plasma, and studies measuring CSF concentrations in human subjects following oral administration are limited and address this issue only indirectly.

Metformin can influence cognitive function both directly and indirectly.

Indirectly, it can do so by reducing blood glucose and influencing metabolic functions. Chronic hyperglycemia is a major risk factor for cognitive decline and structural brain damage [23]. Numerous studies correlate diabetes with an increased risk of developing dementia or other cognitive disorders [24,25,26,27]. It should be noted that beyond the impact of glycemic fluctuations, diabetes is associated with other metabolic dysfunctions that negatively affect the brain. Additionally, metformin exhibits significant anti-inflammatory and antioxidant effects [28,29,30,31], which may partially counteract the oxidative stress and inflammation associated with high glycemic variability.

Diabetes is also a significant cardiovascular risk factor [32,33], and metformin has demonstrated a protective role on cardiovascular health as well as on vascular endothelium [34,35,36], which may contribute to its indirect effects on cognition.

Directly, there is evidence that metformin may exert neuroprotective effects by activating the AMPK pathway [37], which regulates cellular energy homeostasis. AMPK activation in brain neurons can promote neurogenesis, reduce inflammation, and improve mitochondrial function, all of which are important for cognitive health.

3. Effects of Metformin on Cognitive Function

Numerous studies have demonstrated the beneficial effects of metformin on cognitive function in patients with type 2 diabetes mellitus (T2DM), suggesting improvements in short-term memory, a reduced risk of developing dementia, and enhanced executive function [38,39,40,41,42,43,44,45,46,47]. One study identified interesting correlations between metformin, the gut microbiome, and the risk of developing dementia [48].

Regarding its multimodal impact on cognitive function, a recent animal study found a positive correlation between metformin and cognitive function (as measured by attention and inhibitory control) in young subjects [49], which shifted to a negative correlation in older subjects, suggesting an age-related effect on cognitive performance. Notably, two human studies reported more pronounced effects of metformin in Western populations [43,46], thereby raising the hypothesis that racial factors may influence the effects of metformin.

Interestingly, positive cognitive effects were also observed in patients with neuropsychiatric conditions. One study reported improvements in cognitive function and psychotic symptoms in patients with schizophrenia [50]. Similarly, studies indicate a possible anti-epileptic effect of metformin, suggesting it may modulate multiple factors implicated in the etiopathogenesis of epilepsy [51,52].

However, some research has suggested that the cognitive benefits of metformin may be time-limited, with a tendency to diminish or even reverse with long-term use. Additionally, some studies have shown no cognitive benefits from metformin treatment, including in the prevention of dementia [53,54], and others have even suggested an increased risk of Alzheimer's dementia associated with metformin use [55]. Furthermore, one study found a dose-related correlation, with lower doses linked to a reduced risk of developing dementia, while higher doses showed no benefit [56]. A similar correlation was observed in another study [57], led by the same primary author, specifically in the context of Parkinson’s disease.

Regarding other neurodegenerative conditions, results remain inconclusive. Some studies did not find any correlation between metformin use and the risk of developing Parkinson's disease [58], while one study suggested a possible association between metformin and an increased risk of Parkinson’s disease [59]. This finding partially contradicts other studies that suggest metformin may reduce the incidence and severity of neurodegenerative diseases [46,60]. The conflicting findings on metformin’s impact on Parkinson’s disease have been investigated in several studies [61], none of which have demonstrated a definitive effect. Interestingly, one study in animal models of Parkinson’s disease demonstrated a positive impact on motor function, potentially mediated by metformin’s effects on astrocytes [62].

Several studies have reported that metformin has a positive impact on cerebral white matter in diabetic individuals, helping to preserve white matter microarchitecture [60]. Long-term metformin use has been associated with increased white matter integrity and reduced diabetes-related brain volume loss in metformin-treated patients [63], particularly in regions such as the parietal lobes and cingulate cortex, which are involved in spatial orientation and memory processes. Moreover, a study based on fractional anisotropy measurements showed increased cerebral parenchymal integrity in diabetic patients treated with metformin [60]

In a murine model, one study hypothesized that metformin stimulates oligodendrocyte progenitor cells (OPCs) under hypoxic conditions [63]. Another animal study demonstrated metformin’s anti-inflammatory and neuroprotective properties against the neurotoxic effects of streptozotocin [64], a compound used to mimic Alzheimer’s dementia in animal experiments. Additionally, metformin administration counteracted neurotoxic effects from repeated alcohol exposure, indicating neuroprotective potential under toxic conditions [65].

The hypothesis that metformin acts as a cardiovascular protective factor has also been explored, and several studies have demonstrated its beneficial effects on endothelial function and its role in reducing oxidative stress at this level [66,67,68]. These findings may be crucial in understanding metformin's role in cognitive protection, particularly through the enhancement of cerebral circulation and the prevention of ischemic strokes. Similarly, several studies have reported a decrease in the incidence and severity of ischemic attacks among patients already undergoing treatment with metformin, as well as a positive association between metformin use and improved clinical outcomes [69,70,71].

Additionally, studies aimed at elucidating metformin’s mechanism of action at the biocellular level were identified. The primary cellular targets of metformin are illustrated in the Table 2. Although other cellular targets are not included due to challenges in defining metformin’s specific effects on them, these targets certainly play a role in explaining and understanding metformin's mechanism of action.

4. Biovariance: A New Frontier in Personalized Medicine

Interindividual variability is a well-documented concept in medical sciences, explaining differences in response to a particular treatment between individuals. In another sense, this reflects the axiom "each individual is unique". This variability accounts for the fact that a drug can exhibit a different action profile from one patient to another and may also cause adverse effects or allergic reactions in certain individuals. Clinical trials primarily assess general effects within a population, allowing conclusions to be drawn about efficacy, optimal dosage, and safety profile. Ultimately, a risk-benefit ratio is calculated, determining if the drug qualifies for therapeutic application

Interindividual variability arises from numerous factors, ranging from age, genetics, environmental influences to the intricate dynamics of chemical reactions within cells. Unlike other factors, cellular factors are challenging to use as population selection filters due to their stochastic and partly unpredictable nature. Biovariance addresses this type of variability, which cannot be eliminated by selection criteria but can be reduced if the criteria are adapted to target metabolic function. This approach enables a personalized therapeutic strategy and maximizes the therapeutic potential of drugs by identifying subgroups with a strong positive response.

Biovariance refers to the heterogeneity of a drug’s effect, characterizing the response to treatment rather than the substance itself. Its biological foundation lies in the complex dynamics of metabolic reactions occurring at the cellular level, which are influenced by the pre-existing biochemical context, competition for receptors, and subtle molecular interactions that generate stochastic behavior. Therefore, biovariance is fundamentally tied to how living organisms process and respond to a range of substances, reflecting the complexity of metabolic interactions that vary among individuals and even within the same individual. While biovariance is directly related to the complexity of a biological system, this relationship is not linear; as complexity increases, biovariance tends to increase exponentially. The brain is a striking example: due to the unique and highly varied organization of neuronal networks, synaptic plasticity, and the dynamic interactions between neurons, the central nervous system introduces an additional layer of variability in response to treatments. This is consistent with the findings of our study, which demonstrate the difficulty in establishing a clear effect of metformin on cognitive function based on the existing literature.

4.1. Role of Biovariance in Shaping Clinical Trial Design

We believe that one of the most significant consequences of accepting and validating the concept of biovariance would be a reorientation of clinical trial design. This would involve adapting selection criteria to include specific markers of cellular metabolism and selecting patients based on the characteristics of their metabolome. Such an approach would enable the development of new, relevant evaluation criteria for metabolic status, providing a more accurate picture of how patients respond to treatments according to their metabolic profiles and bringing us closer to the goal of truly personalized medicine.

Most medical studies aim to identify a correlation between the administration of a substance and the occurrence of a particular effect within a population. While these studies provide valuable clinical insights, their methodology often falls short when investigating effects with high biovariance. Ignoring inherent interindividual variability can lead to misinterpretations of the data and frequently results in the conclusion that further, larger studies are necessary. Acceptance of the biovariance concept could, therefore, help halt the "race" for ever-larger, costlier studies, as the biovariant drug effects of a drug will continue to differ among individuals. Consequently, even studies with substantial sample sizes, numbering in the hundreds of thousands of participants, may still fail to yield conclusive results. In contrast, a single, highly relevant study that identifies distinct or even opposing effects within a homogeneous sample, carefully selected based on well-defined criteria, could effectively reveal the biovariant nature of the effects.

The Figure 2 (reproducible via the functions described above right) illustrates a bimodal distribution model in comparison to a normal distribution, highlighting the limitations of statistical analysis when applied to abnormal distributions in identifying clinically valuable data.

4.2. Metabolic Profiling Drugs. Metabolic Markers

The most accessible methods for studying cellular metabolism involve measuring metabolic products in blood samples or, when feasible, in the pericellular space. In certain contexts, radioisotopes can be employed to trace metabolic pathways and provide detailed insights into cellular biochemical dynamics, although their use is typically restricted to preclinical research or specialized laboratory settings due to safety and regulatory limitations. Additional methods, such as advanced imaging techniques and fluorescence-based assays, can also provide indirect but valuable insights into metabolic activity.

Relatively new concepts, such as fluxomics [80], complement these methods by synthesizing data from multiple -omics fields .(e.g., genomics, proteomics, microbiomics etc.). Together, these methods provide a comprehensive toolkit for understanding the metabolic status of individuals, offering insights into intracellular activity, metabolic abnormalities associated with various pathologies, and the potential for customized treatment approaches. The totality of small molecules resulting from metabolic processes, known as metabolites, constitute the metabolome. Metabolome constitutes a promising resource in medical research [81,82,83], providing information about the metabolic status of a subject and enabling the understanding of the metabolic abnormalities associated with various pathologies, which can be used to customize treatment.

The metabolome also serves as a mirror of intracellular activity, reflecting changes that occur following the administration of a drug and providing an effective means of studying individual metabolic responses. Additionally, prior to drug administration, biologically inert substances that produce no effects in the body but allow for the evaluation of individual metabolic characteristics through the measurement of metabolic by-products could be utilized.

We could refer to these substances as metabolic profiling drugs. By using them, we could identify population subgroups that exhibit common metabolic patterns, subgroups which could later be selected for clinical studies adapted to their metabolic profile. In this way, conclusions could be drawn about the effects of a drug by referring to the metabolic response that the population subgroup presents to the administration of the profiling drug, a process illustrated in Figure 3:

An example of a drug that could be used for metabolic profiling is codeine. Codeine (methylmorphine) is a medication used to treat cough and pain. Its analgesic effect is primarily due to its conversion to morphine in the liver, a process mediated by the enzyme CYP2D6. This enzyme catalyzes the demethylation process, converting the methoxy group at position 3 of the aromatic ring into a hydroxyl group, resulting in morphine. Another metabolic pathway involves the enzyme CYP3A4, which transforms codeine by removing the methyl group from the nitrogen atom in codeine’s structure, producing norcodeine, an inactive metabolite in terms of analgesic effect. If CYP3A4 activity is predominant, less morphine is generated, leading to a weaker analgesic effect. Conversely, if CYP2D6 activity is dominant, more morphine is produced. This variability can result in significant heterogeneity in codeine’s effects, depending on individual metabolic characteristics.

CYP2D6 and CYP3A4 are part of the cytochrome P450 enzyme family, which is widely distributed across living organisms. These two enzymes play a role in the metabolism of a broad range of drugs [84,85,86], and differences in their activity contribute to variability in individual responses to a drug. Codeine’s role as a metabolic profiler is particularly important for drugs that are substrates for both CYP2D6 and CYP3A4, as identifying the predominance of one enzyme’s activity can significantly aid in treatment personalization. Enzyme activity is subject to multiple variations, and this must be interpreted in the context of the specific substrate, as activity depends on affinity, which describes the enzyme-substrate interaction.

It is important to acknowledge the limitations of this approach, as each enzyme has distinct affinities for different chemical structures. Therefore, a patient with predominant CYP2D6 activity for one drug may still experience more intensive metabolism via CYP3A4 for another drug, depending on the chemical structure of the drug. To address this challenge and provide a more comprehensive view of overall metabolism, it may be necessary for a metabolic profiling drug to exhibit enzyme polyspecificity, allowing metabolic variations to be correlated more accurately with changes in metabolite concentrations.

5. Conclusions

Based on our findings, we recommend reorienting the selection criteria in clinical trials for metformin and other medications that display a wide range of observed effects to include specific markers of cellular metabolism. Additionally, we call for the development of metabolic profiling drugs designed to tailor treatments to individual metabolic profiles.

Given that sufficient large-scale studies have confirmed variable effects of metformin on cognition, instead of undertaking more extensive research with large sample sizes, a focused strategy may more effectively elucidate underlying mechanism of the distinct effects of metformin across different subgroups. This approach could optimize resources by reducing the need for large, inconclusive studies, redirecting them towards investigations of individual responses.

By integrating biovariance into study design and developing metabolic profiling drugs to assess cellular metabolism, research can advance towards personalized medicine, maximizing both treatment effectiveness and the predictability of therapeutic outcomes. This approach could pave the way for more precise patient stratification in future metformin research and, more broadly, contribute to a deeper understanding of variability in drug response.

6. Limitations

This review has several limitations that may affect the interpretation of findings. First, the selection of studies in this review is limited due to our selection criteria, database limitations, or the exclusion of publications in languages other than English. Second, the heterogeneity of cognitive tests used in the analyzed studies complicates direct comparisons, as varied assessments may differ in sensitivity to metformin’s cognitive effects. Additionally, linguistic and cultural differences in test adaptations may introduce biases, affecting cross-study consistency.

The new concepts proposed to explain the effects of metformin, like biovariance, have not been validated through experimental or clinical studies, highlighting the need for further research to confirm and apply these ideas in practice. The interpretation of data and development of these concepts may be influenced by our own perspectives, introducing a potential subjective bias on the part of the authors.