FormalPara Key Points

We identified 120 documents that reported or cited 65 different “therapeutic” plasma metformin concentrations or ranges.

Although metformin has been available for over 50 years, major methodological and/or conceptual errors have confounded the literature on its therapeutic concentrations.

A dose-efficacy study with measurement of the corresponding plasma metformin concentrations is therefore needed for defining the therapeutic concentration window for metformin.

1 Introduction

For any given drug, the physicians should always prescribe the “therapeutic dose”—a dosage corresponding to the “therapeutic range” that provides sufficient efficacy whilst avoiding overdosing. The “therapeutic range” can also be defined as “an approximation of the average plasma drug concentrations that are safe and efficacious in most patients” [1]. However, when studying the dose–response relationship for various dosage regimens, the relationship between blood concentrations and clinical effects may be evaluated in several different ways: (1) in clinical studies from which pharmacological parameters are measured; (2) by deducing the therapeutic range on the basis of pharmacological data; and (3) by combining clinical and pharmacological data on drug efficacy and concentrations. Once a therapeutic range has been defined, one can subsequently define subtherapeutic and supratherapeutic conditions.

Although metformin has been available since 1957 and is recommended as a first-line treatment for type 2 diabetes mellitus [2], this drug provides an excellent example of why therapeutic concentrations must be characterized. On one hand, metformin should be prescribed more widely because of its beneficial, pleiotropic effects. On the other, it should be prescribed less widely because an increasing proportion of diabetic patients have poor renal function—a contraindication to metformin use because of the risk of so-called “metformin-associated lactic acidosis” [35].

The first reports on a therapeutic dose and therapeutic concentrations for metformin were published in 1962 [6] and 1972 [7], respectively. Over 50 years later, one can legitimately question whether a clear definition of the therapeutic concentrations of metformin is available. To address this issue, we systematically searched the literature for publications that mention “therapeutic concentrations” of metformin.

2 Research Design and Methods

2.1 Data Sources

We performed a systematic search of the scientific literature recorded in the MEDLINE, Scopus, ScienceDirect, and Wiley-Blackwell electronic databases between January 1957 and November 2014. We also searched the Internet using the Google search engine and performed a manual search in our personal libraries (including documents that are not referenced in the aforementioned electronic databases).

2.2 Data Selection

We used the keywords “metformin”, “therapeutic concentration”, and “therapeutic range” but also searched for allusions and related wordings such as “therapeutic level”, “plasma concentration”, and “normal value” with regard to metformin’s antidiabetic effects.

2.3 Data Extraction

We extracted values or ranges described as “therapeutic concentrations” from the retrieved documents and their cited references (Fig. 1). When a blood metformin value or range was not defined exactly as a “therapeutic concentration” or “therapeutic range”, we cite the original wording of the text. When suggested therapeutic values were defined by citing a literature reference, the types of studies in cited references and the concordance of data between the citations and their sources were studied.

Fig. 1
figure 1

Flow diagram of data extraction

2.4 Data Presentation

All of the collected data are presented in a single table (Table 1), which is rather large but it was impossible to subdivide it into study categories (e.g., in vitro studies, animal studies, clinical studies, etc.). Indeed, this type of categorization might have erroneously suggested that the mentioned “therapeutic concentrations” were deduced from the results of the studies themselves. In fact, most of the “therapeutic concentrations” were suggested by citing other data (either personal data or, more often, a previous publication) and thus were not based on the study’s design and results. Furthermore, Fig. 2 serves as a guide to reading Table 1.

Fig. 2
figure 2

A guide to reading and interpreting Table 1

Table 1 Criteria for therapeutic concentrations of metformin in the literature: values, ranges, wording, and sources

In some papers, metformin concentrations were noted in molar or gram units. Here, the concentrations were converted into mg/L if necessary. The exact molecular weight of metformin is 129.1636 [8] (rounded down here to 129 g/mol), and so the equivalences are 10−5 mol/L = 1.29 mg/L and 1 mg/L = 7.75 μmol/L.

3 Results

We identified a total of 120 publications, which reported or cited 65 different therapeutic plasma metformin concentrations or ranges. The individual values ranged from 0.129 to 90 mg/L. When considering concentration ranges, the lowest and highest boundaries were 0 and 1800 mg/L. The narrowest range was 0.000225–0.003 mg/L and the broadest was 150–1800 mg/L. Most (77 %) of the values or ranges proposed were between 0.1 and 4 mg/L.

The collected data are presented in Table 1 [3, 7160] and summarized in Table 2.

Table 2 Summary of the data

None of the studies was performed with the specific objective of defining the therapeutic concentrations for metformin. Fifty-four publications (45 %) cited previous studies as providing therapeutic concentrations, whereas 62 publications (51.7 %) mentioned “therapeutic concentrations” but did not even cite a supporting reference.

Only four original research studies (3.3 %) [8, 47, 80, 119] determined a therapeutic concentration. These studies assayed the plasma metformin concentrations (and the erythrocyte metformin concentrations in two studies) in different populations of diabetic subjects (1) undergoing “well-tolerated chronic metformin treatment” [47, 80]; (2) with chronic kidney disease [119]; and (3) “in steady state not having lactic acidosis” [8].

When a literature value was cited, the references were mostly reviews, pharmacokinetic studies, and in vitro studies. However, none of these cited studies was performed with the specific objective of defining the therapeutic concentration for metformin. In the 54 publications that cited supporting references, concordance between the wording of the citation and the true nature of the source data was observed in only 23 cases (42.6 %).

4 Discussion

The present study is the first to have systematically analyzed literature reports of therapeutic concentrations and therapeutic ranges for metformin. Although one would expect to find a consensus on a drug that has been available since 1957, at least for the antidiabetic properties of metformin, major methodological and/or conceptual errors have confounded this subject.

Firstly, the difficulty of characterizing therapeutic concentrations lies in the fact that a very large number of either values or ranges were identified (n = 65) and that, moreover, there was a very large variation between the different proposals. Indeed, the proposed single values varied from 0.129 to 90 mg/L, and the lowest and highest range boundaries were 0 and 1800 mg/L, respectively.

In this respect, it is surprising (from a biological and physicochemical point of view) that a value of 0 mg/L (i.e., the absence of the drug) could be suggested as the lower limit of a therapeutic range. Likewise, the highest boundary suggested (i.e., 1800 mg/L) is far above the values usually reported for metformin intoxication (i.e., up to 267 mg/L) [149]. Of course, one can consider the putative occurrence of arithmetical errors concerning the format (with a value of zero quoted as a lower limit on a purely statistical basis) and/or the calculation (when converting units into mg/L). However, even the majority of the most frequently cited values were between 0.1 and 4 mg/L, a range that already encompasses a 40-fold variation. In other words, some of these disparate values may be due to procedural (rather than conceptual) errors.

Secondly, most of the definitions of therapeutic concentrations were conceptually flawed. The vast majority of the documents studied (116 of 120) did not directly establish a therapeutic concentration for metformin and therefore cited previous studies. This would not be a problem per se, provided that the cited study provides reliable information. In fact, almost half the 120 publications did not cite a reference when referring to therapeutic concentrations, and when a reference was cited, it was not usually concordant. Even more importantly, the cited references provided were mostly reviews or pharmacokinetic studies: none were studies performed with the specific objective of defining the therapeutic concentration for metformin. The criteria used to define a therapeutic concentration of metformin were predominantly pharmacokinetic parameters [e.g., maximum plasma concentration (C max)]. Furthermore, the use of peak values is flawed: C max refers to a drug’s maximum concentration in a specified compartment (e.g., the blood) after a dose of the drug has been administered, whereas definitions of a therapeutic concentration usually refer to the trough steady-state level achieved by the prescribed dosing regimen [56]. Moreover, the above-mentioned pharmacokinetic parameters were mostly measured after the administration a single dose of metformin (in either diabetic or non-diabetic subjects) rather than during long-term metformin therapy (i.e., a multiple-dosage regimen).

Lastly, only four studies (including two from the same research group) provided putative steady-state plasma metformin concentrations determined in patients receiving long-term metformin treatment [8, 47, 80, 119]. Only two studies also quoted erythrocyte metformin concentrations [47, 80]. This is another critical issue, since plasma values do not necessarily reflect tissue concentrations [3, 80] and thus a drug’s true metabolic effects. In this respect, it has been suggested that erythrocyte metformin concentrations correspond to a deep compartment [80]. However, data on metformin concentrations in other putative deep, major compartments of metformin action (such as the intestine, liver, muscle and adipose tissue), which would be even more informative, are evidently not available in clinical practice.

In addition to the huge variation in the therapeutic concentrations due to flaws, significant “natural”, inter-individual variations in the blood metformin concentration should also be taken into consideration: (1) the various definitions of therapeutic concentrations do not appear to take account of the widely varying metformin dosages used in clinical practice; and (2) more importantly, variations in the genes coding for solute carriers significantly modulate metformin’s pharmacodynamics and pharmacokinetics [161163].

The clinical impact of this type of polymorphism was confirmed in a cohort of type 2 diabetes patients taking 1 g of metformin twice daily over a 24-month period [164]. An 80-fold inter-individual variation in plasma metformin concentrations was documented. Ultimately, this implies that in a particular patient, a given metformin concentration might be therapeutic, subtherapeutic or supratherapeutic, or might even correspond to metformin accumulation. The variations observed thus argue in favor of personalized therapy, as currently recommended by many authors or authorities [165170].

One must also consider whether the present study had any limitations. However, given the nature of our systematic literature search, the only possible limitation would be incomplete identification and retrieval of publications on therapeutic concentrations. In fact, we are confident that we performed an extensive study of the literature by examining nearly 1000 potentially relevant publications. Lastly (in order to avoid subjective interpretations), we also quoted the original wording of the text referring or alluding to therapeutic concentrations.

4.1 Guidance for Clinical Practice

Given the above pitfalls, one can legitimately ask whether any of the literature information on metformin concentrations is of use in clinical practice. In a recent review, Dowling et al. [143] stated that in vivo preclinical and in vitro studies often involve extremely high, non-physiological concentrations of metformin that are far in excess of the doses used in clinical and epidemiological studies. Indeed, in vitro studies have typically featured a metformin dose between 25 and 1000 times higher than that used in clinical studies. For in vivo studies, the values are still between 2 and 45 times higher than in clinical studies. This is of crucial importance when seeking to distinguish between therapeutic and supratherapeutic concentrations and to characterize mechanisms of action. In studies of animal hepatocytes in vitro, metformin concentrations below 50 μmol/L activate AMPK (5′ adenosine monophosphate-activated protein kinase) and suppress dibutyryl-cAMP (cyclic adenosine monophosphate)-stimulated gluconeogenic gene expression (and therefore hepatic glucose production) [171], whereas much higher metformin concentrations (5 mmol/L) are required to inhibit the respiratory chain complex 1 [172]. In an in vitro study in humans, inhibition of the respiratory chain complex 1 [concentration of drug producing 50 % inhibition (IC50)] was obtained at metformin concentrations of between 0.45 and 1.2 mmol/L (i.e., 58–155 mg/L) [173]. These figures should be compared with metformin concentrations reflecting pharmacologic or toxic effects in humans: C max for metformin in healthy subjects is ~8–16 μmol/L (1–2 mg/L) [6] and the metformin concentrations measured in cases of metformin overdose go up to 2 mmol/L (i.e., 210 mg/L) [149].

The only valid way of defining the therapeutic concentration window for metformin would be to relate dose efficacy (in terms of blood glucose control [174]) to the corresponding plasma metformin concentrations in long-term therapy. Since this procedure has never been attempted, the question arises as to what extent it is possible to extract at least some literature data of value for guiding clinical practice (i.e., after avoiding the aforementioned flaws). A pertinent basis for defining the therapeutic concentrations of metformin would be the range of plasma metformin trough concentrations produced by well-tolerated doses. This would also take account of the time interval following metformin administration [175]. On the basis of the very few studies that have actually done this [8], a value of 2.5 mg/L probably corresponds to the upper limit of the therapeutic range. Establishing the lower concentration is more problematic. However, this type of approach cannot answer an additional question: what is the metformin threshold for adverse effects? This is a very difficult question because even major metformin accumulation does not necessarily lead to lactic acidosis [47, 58, 75, 176]. Conversely, some patients may display an idiosyncratic, hyperlactatemic response to metformin therapy—even in the absence of metformin accumulation [60, 176].

It is therefore not possible to define the therapeutic concentrations for metformin purely on a safety basis, i.e., by considering metformin concentrations associated with adverse effects. Nevertheless, Table 3 differentiates between excessive metformin concentrations on one hand and concentrations obtained in patients at therapeutic doses (with no risk factors for metformin accumulation) on the other. On this basis, we suggest that metformin concentrations can be categorized as follows: non-therapeutic (i.e., ineffective), therapeutic and safe (for the great majority of patients), intermediate (with a risk of hyperlactatemia in some individuals), and excessive (with a high risk of hyperlactatemia in most patients).

Table 3 Differentiation between excessive metformin concentrations and those obtained in patients without risk factors for metformin accumulation taking therapeutic doses

5 Conclusion

The huge differences between the various therapeutic metformin concentrations suggested in the literature far exceed the inter-individual variations due to genetic factors (which are already quite large) and thus mainly reflect methodological and/or conceptual errors. Accordingly, it must be acknowledged that the therapeutic concentrations reported to date are of no value in (1) guiding the clinically useful, safe dosage of metformin; (2) distinguishing between therapeutic, supratherapeutic, and subtherapeutic concentrations; and (3) attributing responsibility to metformin in the genesis of so-called “metformin-associated lactic acidosis”.

A dose-efficacy study with measurement of the corresponding plasma metformin concentrations is urgently needed with a view to defining the therapeutic concentration window for metformin.