Metformin disposition—A 40‐year‐old mystery

1.

Metformin (N,N‐dimethylbiguanide) is a first‐line medication for the treatment of type 2 diabetes. It is an old drug, having been discovered in 1922 and first used in France in 1957. ¹ The purpose of this commentary is to emphasise that, despite its age, aspects of its disposition remain obscure and controversial. This reflects the fact that it is a highly polar cationic compound with a log P of −1.4 and a pKa of 12.4, which would predict that it is unlikely to cross biological membranes easily by passive diffusion, thereby requiring transporters for distribution and excretion, and that it would not be expected to undergo metabolism.

The first studies of the pharmacokinetics of metformin were reported in 1978–1981 and involved both intravenous and oral administration. One of these studies ² measured radioactivity after giving ¹⁴C‐labelled drug and two used specific assays (mass fragmentography and gas chromatography, respectively). ³ , ⁴ Congruent findings from these studies were an oral bioavailability of 50%–60% and a negligible plasma protein binding. After intravenous injection ‘terminal’ apparent half‐lives of 1.5–4.5 h and of 9–19 h were reported based on measurement of plasma drug concentrations over 8–12 h after dosage and urinary excretion rates over 40–60 h, respectively. A slow accumulation in erythrocytes was observed such that whole blood drug concentrations were eventually considerable in excess of those in plasma, with a prolonged decay. ⁴ An extended elimination phase was also indicated by an increment in the trough plasma drug concentration between 7 and 14 days of continuous dosage. ⁴ This has implications for the time to achieve steady state on continuous dosing, the time to wash out drug after stopping long‐term therapy and the linkage of exposure to effect. ⁴

A major difference between the findings of the radiolabel study ² and the others ³ , ⁴ is apparent when comparing the data on urinary excretion. Whereas Pentikainen et al ² reported 100% recovery of unchanged drug by 48 h after intravenous injection (n = 3), Tucker et al ⁴ could only recover 79% of the dose in this form in the urine at 72 h (n = 4), with seemingly little further accumulation beyond that time, and Sirtori et al ³ reported a mean urinary recovery after intravenous injection of 87 ± 22% at 48 h, beyond which time drug was not detectable in urine (n = 5). The latter figure differs from a mean of 77% evident from the ratios of their estimates of renal to total clearance. Pentikainen et al ² concluded that metformin is not metabolised, based on their recovery and a radio‐chromatographic analysis of the urine. In contrast, a degree of metabolism can be postulated to account for the incomplete recovery of unchanged drug by Tucker et al ⁴ and Sirtori et al. ³ However, while the compound is known to be biodegraded to guanylurea by microbes, ⁵ metabolic products have never been identified in animals or humans. A search for specific metabolites in rabbits after oral administration of ¹⁴C‐metformin HCl gave negative results, and concentrations of unchanged drug, as determined by mass fragmentography, were concluded to be essentially identical to those calculated from plasma radioactivity. ³ In contrast, Choi and Lee ⁶ and Choi et al ⁷ claim a degree of cytochrome P450 (CYP)‐mediated transformation based on in vivo studies in rats with enzyme inhibitors and inducers and in vitro disappearance studies with expressed rat and human enzymes. The latter indicated metabolism by CYP3A4. Lennard et al ⁸ recovered over 95% of the dose, albeit less than 100%, as unchanged metformin in tissue and perfusate at 1 h in a perfused rat liver preparation. Clearly, there is significant inconsistency in the literature with regard to whether or not metformin is metabolised to some degree, the conflicting evidence being ignored in many reviews of metformin and references to its elimination.

Can the conflicting data be reconciled? Pentikainen et al ² clearly claim complete urinary recovery of unchanged drug after intravenous injection of metformin. There are several issues in regard to this. To justify their claim of 100% renal clearance at 48 h, Pentikainen et al ² set total clearance equal to renal clearance. However, they did not substantiate that equality numerically. Although their average value of total clearance was only 5.7 ml/min greater than renal clearance, the 95% confidence interval for this difference is from −173 to 184 ml/min, such that their insignificant difference in clearance by Student's t test did not exclude a huge potential range of clinically significant difference. In any case, the total clearance would have been overestimated when determined by extrapolation using a ‘terminal’ apparent half‐life in plasma of 1.7 h, a value that is less than those from studies with longer sampling times. Also, there are several problems with calculating renal clearance from rapidly changing plasma drug concentrations and urinary excretion rates, ⁹ as was done by Pentikainen et al. ² Values obtained under non‐equilibrium conditions when the extent of distribution in the rest of the body is increasing would underestimate renal clearance. ¹⁰ Furthermore, there is a discordance between the mean cumulative excretion data of Pentikainen et al ² and their mean urinary excretion rate versus time plot. Using a digitiser on the latter data and assuming that they approximated mid‐point urinary excretion rates by dividing the amount excreted in each urine collection interval by the time of the associated collection, the estimated recovery at 48 h is 440 mg. Assuming that the mass dose was 500 mg and the amount recovered in the urine was expressed as non‐radioactive metformin in the same way, the recalculated recovery would be 88% at 48 h, which is much closer to the figure of 79% at 72 h reported by Tucker et al. ⁴

As evidence of a lack of metabolism of metformin, Pentikainen et al ² show a single radioactive spot following two‐dimensional thin‐layer chromatography of extracted urine, although it is possible that a structurally very similar product, possibly formed by N‐hydroxylation or N‐demethylation, was not separated from parent drug? However, the data in rabbits mentioned by Sirtori et al ³ would seem to corroborate a lack of identifiable metabolites.

The changes in the total clearance of metformin after intravenous injection in the rat observed by Choi and Lee ⁶ and Choi et al ⁷ following administration of CYP inducers and inhibitors are mostly modest. Reported changes in non‐renal clearance of 24%–79% in the expected directions were revealed only after subtraction of estimates of predominant renal clearance from estimates of total clearance and assume that these clearance estimates were correct. ¹⁰ In the limit, any error in calculating clearances would reduce to an absence of metabolic clearance, in which case, the changes in total clearance might be due to interaction at the level of renal transport. Larger differences observed after oral administration may be magnified by competition for active transport also in the gut wall. The disappearance of drug from the incubation medium in the in vitro studies might reflect incomplete extraction, although differences in residues were apparent, for example, between incubations with and without itraconazole. ⁷ In comparing data from studies with expressed enzymes, ⁷ the isolated perfused liver ⁸ and in vivo measurements ⁶ , ⁷ differences in access to the enzymes and the time of exposure could confound interpretation.

Viewing the literature as a whole, it seems that the jury must still be out regarding whether or not metformin undergoes a degree of metabolism.

Recently, a further twist to the tale/tail has been added. Wesolowski et al ¹⁰ proposed an alternative explanation for the incomplete urinary recovery of unchanged metformin at 72 h following intravenous injection, ⁴ based on modelling of metformin disposition kinetics in dogs with a gamma‐Pareto convolution that predicts extremely slow, continuing distribution. Potential mechanisms for protracted tissue uptake include active transport and membrane potential‐driven import, especially into mitochondria. ¹⁰ The model would indeed predict 79% urinary excretion of unchanged drug over 72 h in dogs, similar to humans, with very slow further accumulation ⁴ and without assuming any metabolism. Accordingly, it reinforces the need for fractal models to explain ‘strange’ or ‘anomalous’ pharmacokinetic behaviour. Such models challenge the common use of sums of exponentials to describe pharmacokinetic data, particularly with respect to under‐extrapolation of AUC and hence over‐estimation of clearance. ¹⁰ A test of the new metformin model would be to recover the ‘washout’ of the drug from patients after long‐term therapy.

In conclusion, we emphasise that there are inconsistencies apparent in the literature regarding the pharmacokinetics and potential metabolism of metformin, with clinical implications for its accumulation and elimination, and fundamental implications for how its disposition and that of other drugs are described by both classical and physiologically based pharmacokinetic models.

COMPETING INTERESTS

There are no competing interest to declare.

Tucker GT, Wesolowski CA. Metformin disposition—A 40‐year‐old mystery. Br J Clin Pharmacol. 2020;86:1452–1453. 10.1111/bcp.14320