Disposition of metformin (N,N-dimethylbiguanide) in man*
`
`Kinetic parameters ol metformin (N,N-dimethylbiguanide), an anti-diabetic reported to be
`associated with a 10IVer number of episodes of lactic acidosis than phenformin, lVere determined
`in volunteers with normal renal function and in patients with different degrees ol renal
`impairment. Drug in body fluids was measured by a highly specific and sensitive mass
`fragmentographic method, alter the formation of a triazine derivative, obtained with
`heptafluorobutyric anhydride. The ha(l-life (t1h.) for the elimination ol drug from plasma after
`intravenous injection in 5 normal subjects (1.52 ± 0.3 hr) (mean ± SD) was shorter than that
`reported for phenformin by a similar assay method (7 to 15 hr). The mean tV2 in 5 renal
`patients was 4.94 ± 1.11 hr, and a correlation was observed between t'/2 ol drug from plasma
`and creatinine clearance. After oral administration of metlormin tablets, drug recovery in urines
`was only 37.6%, possibly not as a consequence of low bioavailability (a similar low recovery was
`found after oral administration of the metformin solution usedlor the intravenous studies), but of
`binding to the intestinal wall, as shown in animal and clinical studies lVith metlormin and other
`biguanides. Me(lormin is rapidly eliminated through active secretion by the kidney (mean renal
`clearance, 440.8 ml/min)-it is neither metabolized nor protein-bound in plasma. The very brief
`plasma t'/2 makes significant cumulation, with a standard tid regimen, unlikely. These findings may
`help explain the lower incidence of toxic effects, particularly lactic acidosis, than after phenformin.
`
`Cesare R. Sirtori, M.D., Ph. D., Guido Franceschini, M.S., Marzia Galli-Kienle, Ph.D.,
`Giuliana Cighetti, M.S., Gianni Galli, Ph.D., Alighiero Bondioli, M.D., and
`Franco Conti, M.D. Milan, Italy
`Center E. Grossi Paoletti, Institute ol Chemistry, School ol Medicine, Institute of Pharmacology
`and Pharmacognosy, School of Pharmacy, University of Milan, and Divisione Medica Vergani,
`Maggiore Hospital
`
`The kinetics of biguanide anti-diabetic drugs
`has received little attention, possibly because of
`the lack of adequate specific and sensitive assay
`procedures. 5 Recently developed new methods
`for the specific determination of biguanides 20 in
`
`Received for publication Feb. 2, 1978.
`Accepted for publication July 22, 1978.
`Reprint requests to: Dr. Cesare Sirtori, Center E. Grossi Paoletti,
`Via A. Del Sarto 21, 20129 Milano, Italy.
`'This study was carried out with partial support from a Re(cid:173)
`search Grant of the Consiglio N azionale delle Ricerche of
`Italy.
`
`biologic fluids have made possible more accu(cid:173)
`rate definition of kinetic parameters. By the use
`of these methods, it has been shown that sig(cid:173)
`nificant toxic effects ofbiguanides, in particular
`lactic acidosis (LA), are often correlated with
`very high plasma drug levels. 4, 9, 28
`Metformin (N,N-dimethylbiguanide) is pres(cid:173)
`ently the most widely used drug of this type
`in Western Europe.!.') Interest in the use of
`metforrnin has recently been stimulated by
`the observation of significant hypolipidemic
`effects, particularly in hypertriglyceridemic pa-
`
`0009-9236/78/120683+ 11$01.10/0 © 1978 The C. V. Mosby Co.
`
`683
`
`AUROBINDO EX. 1021, 1
`
`

`

`684 Sirtori et al.
`
`Clin. Pharmacol. Ther.
`December 1978
`
`Table I. Participating subjects with normal and altered renal excretory functions
`
`Subject
`
`Normals
`1
`2
`3
`4
`5
`Renal patients
`1
`2
`3
`4
`5
`
`M
`M
`F
`M
`M
`
`F
`M
`M
`M
`M
`
`Age (yr)
`
`Weight (kg)
`
`Creatinine
`clearance (mllmin)
`
`36
`39
`46
`60
`43
`
`40
`72
`72
`59
`29
`
`70
`64
`68
`81
`77
`
`70
`67
`53
`64
`96
`
`98
`108
`105
`92
`94
`
`48
`45
`34
`22
`20
`
`tients, 10, 23 and of changes in lipoprotein struc(cid:173)
`ture and metabolism in both experimental ani(cid:173)
`mals and in man. 24
`In spite of the very wide use of metfonnin in
`diabetes and in hyperlipidemia, a considerably
`lower incidence of LA has been reported after
`the use of this drug than after phenfonnin. I In a
`recent review of 330 cases of LA in biguanide(cid:173)
`treated diabetics, 19 12 were associated with
`metfonnin and 281 with phenfonnin. A detailed
`clinical analysis of this large patient sample
`showed that control of diabetes did not appear
`to be a significant factor in the development of
`LA, whereas impairment of kidney function
`was a frequent finding, particularly in the case
`of metfonnin, all 12 toxic patients exhibiting
`creatinine plasma levels above 3 mg/ 100 m!.
`Recent surveys in France and Switzerland com(cid:173)
`paring relative uses of biguanide drugs and
`market distribution indicated an incidence of
`LA with phenfonnin 20 and 40 times that with
`metfonnin.6, 15
`In the present report, analysis of the most
`relevant kinetic parameters of metfonnin was
`carried out in volunteers with nonnal and al(cid:173)
`tered renal excretory functions. Comparison of
`these data with findings previously reported for
`phenfonnin and bufonnin2, 17 may help explain
`differences in the incidence of LA in patients
`treated with various biguanides.
`
`Methods
`Patients. The oral and intravenous kinetic
`studies were carried out in 5 patients (4 male
`
`and I female) with nonnal blood urea nitrogen,
`plasma creatinine (PCr) , and creatinine clear(cid:173)
`ance (CrCI). All the five patients had elevated
`plasma lipid levels, 2 of them also having a
`mild chemical diabetes. The kinetic studies
`were perfonned prior to the beginning of a
`study on the effects of metformin on plasma
`lipoproteins. An intravenous test was carried
`out in 5 additional patients (4 male and 1
`female) with significantly impaired renal ex(cid:173)
`cretory function, as detennined by CrCI and
`para-aminohippurate clearances (Table I).
`Steady-state plasma levels of metformin were
`determined in patients chronically treated with
`different doses of the drug for hyperlipidemia or
`diabetes. All these patients had a nonnal renal
`function as assessed from the PCr and CrCI
`tests.
`Protocol of the pharmacokinetic studies.
`All the subjects were fasted for 12 hr before the
`oral or intravenous administration of metfor(cid:173)
`min. In the intravenous test, 1 gm of metformin
`HCI (Spemsa) was rapidly injected intraven(cid:173)
`ously and plasma samples were collected from
`an indwelling venous catheter into heparinized
`tubes. Collection of samples was at 5, 10, 30,
`60,90, 120, and 240 min, and 6,8,24,36, and
`48 hr. Urine was collected after 1, 3, and 6 hr
`and every 6 hr thereafter. Plasma samples and a
`small sample of metformin solution used for
`injection were kept frozen (- 20° C) until
`analysis.
`The oral test was carried out by administering
`a tablet of metformin (Glucophage) equivalent
`
`AUROBINDO EX. 1021, 2
`
`

`

`Volume 24
`Number 6
`
`Disposition of metformin 685
`
`to 850 mg of metformin HCI. Samples were
`collected after I, 2, 3, 4, 6,10,16,24,30,36,
`48, and 72 hr. Urine collections were at 2, 4,
`12, 24, 30, 36, 48, 54, 60, and 72 hr. The
`schedule of plasma and urine collections was
`simplified and shortened after the first two pa(cid:173)
`tients, when it became clear that the drug was
`not detectable in blood or urine after 48 hr.
`In the five patients with impaired renal func(cid:173)
`tion, only the intravenous kinetic study was
`carried out. A dose of metformin two thirds of
`that used for the normal subjects was injected
`(the correct concentration was later calculated
`by mass fragmentography). Plasma samples
`were collected after I, 2, 4, 8, 12, and 24 hr.
`The steady-state plasma samples from 55 pa(cid:173)
`tients treated with metformin for hyperlipidemia
`or diabetes were collected after an overnight
`fast. The last drug administration had taken
`place the night before; i.e., 12 to 14 hr before
`sampling. Daily drug dosage and interval be(cid:173)
`tween drug intake and blood sampling were
`recorded for each patient.
`Determination of metformin in biologic
`fluids. Concentrations of unchanged metformin
`in plasma and urinary samples were determined
`by mass fragmentography after the preparation
`of triazine derivatives essentially as described
`by Matin, Karam, and Forsham,20 with deuter(cid:173)
`ated metformin as the internal standard. Deu(cid:173)
`terated metformin was prepared by fusion of d6-
`dimethylammonium chloride with
`l-cyano(cid:173)
`guanidine as described by Werner and Bell. 26
`The obtained d6-metformin analyzed by mass
`spectrometry using
`the direct
`inlet system
`showed 98% isotopic enrichment.
`For the quantitative drug determination in
`biologic samples, an alkaline solution of d6-
`metformin (between 1 and 10 /-tg of d6-met(cid:173)
`formin based on the predicted drug concen(cid:173)
`trations) was added to the plasma or urine
`samples. After deproteinization with 10% tri(cid:173)
`chloroacetic acid in I N HCI followed by al(cid:173)
`kalinization with 10 N NaOH, samples were
`extracted with methylene chloride. 20 Heptaflu(cid:173)
`orobutyric anhydride (1 ° /-tl) was added to the
`
`extracts, and, after refluxing for I. hr, the
`methylene chloride was washed with 1 N NaOH
`followed by water. The residue, obtained after
`evaporation of the solvent, and containing the
`
`1.5
`
`0.5
`
`0.5
`
`1.0 melformin
`~ • metformin
`
`1.5
`
`Fig. 1. Standard curves showing the ratio of signal
`intensities of the metformin derivative (m/e 307) and
`of the deuterated metformin derivative (m/e 313)
`plotted against the ratio of {Lg metformin to {Lg of
`deuterated metformin.
`
`2 -amino- 4-dimethylamino- 6 -heptafluoro -s -tri(cid:173)
`azine' was dissolved in ethyl acetate and ali(cid:173)
`quots of the solution were injected for fragmen(cid:173)
`tographic analysis into a Varian 112 S spec(cid:173)
`trometer connected to a Varian SS 100 data
`system.
`Conditions of the analysis were the fol(cid:173)
`lowing:
`
`Glass column packed with I % SE 30 on 100 to
`120 mesh Gaschrom P. The oven temperature was
`160° C; injector and detector were at 250° C; the
`helium flow was 20 ml/min. The electron energy was
`set at 70 e V and other operating parameters were:
`emission current, 1.5 mA; electron multiplier, 2.5
`kV; molecular separator and ion source were at 280°
`C. For the fragmentographic analysis, the molecular
`ion (M+) at m/e 307 of the triazine, and the peak at
`m/e 313 corresponding to the molecular ion of the
`deuterated triazine, were focused.
`
`A standard curve was prepared by adding to
`ml of plasma or urine a standard amount
`of deuterated metformin HCI and increasing
`amounts of metformin HCI, and by treating
`samples as described for plasma analysis. The
`ratios of peak intensities were plotted against
`the theoretical values of /-tg metformin/ /-tg d6-
`metformin (Fig. 1). Concentrations of metfor(cid:173)
`min in the analyzed fluids were calculated from
`the standard curve.
`
`AUROBINDO EX. 1021, 3
`
`

`

`686 Sirtori et al.
`
`Clin. Pharmacal. Ther.
`December 197R
`
`Table n. Two-compartment disposition constants for metformin following intravenous
`administration in 5 normal and 5 renal patients
`
`A
`( f.Lglml)
`
`f3
`(f.Lg Iml)
`
`Normal
`
`Mean ± SEM
`
`Renal
`
`Mean ± SEM
`
`I
`2
`3
`4
`5
`
`I
`2
`3
`4
`5
`
`28.7
`900
`44.7
`960
`912
`57.2
`45.5
`912
`950
`60.1
`927 ± 12 47.2 ± 5.6
`680
`158.5
`689
`39.8
`680
`33.9
`686
`17.2
`684
`15.8
`684 ± 2
`53.0 ± 27
`
`1.59
`26.1
`1.80
`23.1
`4.80
`8.7
`2.53
`16.4
`4.04
`10.3
`2.95 ± 0.6 16.9 ± 3.4
`4.22
`9.8
`2.35
`17.7
`1.29
`32.2
`1.02
`40.8
`0.26
`160
`1.83 ± 0.7 52.1 ± 27
`
`6.31
`10.80
`9.40
`4.32
`6.99
`7.56 ± 1.15
`14.56
`17.08
`17.07
`14.84
`10.95
`14.90 ± 1.12*
`
`0.45
`0.55
`0.56
`0.35
`0.44
`0.47 ± 0.04
`0.25
`0.20
`0.10
`0.10
`0.15
`0.16 ± 0.03t
`
`Significances of the differences between values of normal and renal patients:
`*p < 0.005.
`tp < 0.001.
`:j:p < 0.02.
`§p < 0.01.
`
`Plasma concentration at zero time: C?, = A + B.
`Rate constant:
`- q. K
`0'{3. K
`-
`K
`12 =
`el - AUC.'
`21 - K-'
`eI
`
`0' + f3 + K
`el·
`
`IV
`
`Total clearance rate: Dose/ AUC i \" = Cl tot.
`
`Renal clearance rate:
`
`Drug excreted in urine
`Dose
`
`x Cltot.
`
`D
`Volume of distribution: V d{3 = -
`q
`Half life: 0.~3
`
`.
`
`Kel
`f3
`
`Areas up to time !after intravenous injections
`were calculated from the infinite area according
`to the equations:
`
`where
`
`Analysis of data. Plasma concentration-time
`curves after intravenous administration of met(cid:173)
`formin were determined after calculating the log
`regression
`lines of the
`,8-phase
`Cp = f(t)
`(elimination phase) and of the residuals of the
`a-phase (distribution phase), according to a
`two-compartment open model as suggested by
`Lintz and co-workers.17 From the curves of
`each studied subject, the hybrid constants A, a,
`B, and ,8 were computed:
`A = intercept at time zero of the first order plot of
`slope -0' obtained from the "feathered" plot of
`In C against time, where Band f3 are obtained
`and then the residuals plotted to obtain A and 0'.
`0' = apparent first order rate constant for distribution
`of the drug in the body.
`B = intercept at time zero of the first order plot of
`slope - f3 obtained from the plot of In C against
`time.
`f3 = apparent first order rate constant for elimination
`of the drug from the body.
`
`From these constants pharmacokinetic pa(cid:173)
`rameters were then determined according to the
`following equations 17:
`
`Total area under plasma concentration curve:
`
`iCpdt = ~ + ~ and ct = B e-f3t
`
`0'
`
`f3
`
`P
`
`.
`
`Bioavailability after oral administration of met(cid:173)
`formin was determined according to Equation I:
`
`F = AUCpo Div
`AUCiv Dpo '
`
`(I)
`
`AUROBINDO EX. 1021, 4
`
`

`

`Volume 24
`Number 6
`
`Disposition of metformin 687
`
`tl/2
`(hr-I)
`
`kef
`(hr-I)
`
`k21
`(hr-I)
`
`kl2
`(hr-I)
`
`Cl tot
`(mllmin)
`
`Cl ren
`(mllmin)
`
`467
`62.0
`0.29
`0.66
`255
`1.09
`1.54
`359
`39.3
`0.31
`0.79
`304
`1.26
`1.25
`56.3
`1.24
`1.17
`513
`530
`1.88
`2.30
`0.54
`286
`334
`85.8
`0.70
`1.64
`1.98
`317
`514
`70.1
`0.82
`1.58
`1.48
`2.18
`0.80 ± 0.11
`335 ± 46
`441 ± 40
`0.93 ± 0.32 62.7 ± 7.7
`1.69 ± 0.24
`1.52 ± 0.13
`118
`28.4
`0.58
`65.3
`2.08
`1.81
`2.77
`78.4
`112
`33.9
`0.84
`1.15
`0.56
`3.47
`37.4
`57.5
`34.7
`0.50
`0.63
`0.26
`6.93
`47.1
`40.7
`0.54
`69.3
`0.39
`6.93
`0.19
`85.1
`73.7
`0.01
`34.2
`0.20
`0.20
`4.62
`4.94 ± 0.86* 0.60 ± 0.3U 0.53 ± 0.10 0.85 ± 0.36 34.4 ± 1.9§ 88.4 ± 11.8t 60.4 ± 7.8t
`
`the AUC being estimated by the trapezoidal
`rule.
`All statistical and pharmacokinetic calcula(cid:173)
`tions were performed with the help of a
`Hewlett-Packard 65 calculator. Mean values
`and standard deviations were calculated accord(cid:173)
`ing to standard formulas; significance of the dif(cid:173)
`ferences was determined by Student's t test.
`Regression analysis was carried out by the
`least-squares method.
`Protein binding study. Protein binding was
`estimated in vitro by equilibrium dialysis as de(cid:173)
`scribed by Keen14 using 14C uniformly labeled
`metformin (Amersham-Searle). The radiochem(cid:173)
`ical purity of the 14C-metformin was checked by
`radio-gas chromatography of the triazine de(cid:173)
`rivative, prepared as previously described.
`More than 98% of the radioactivity was shown
`to be associated with the derivative.
`The percentage of binding was calculated
`after the serum samples, taken from 5 healthy
`laboratory co-workers and kept in dialysis tub(cid:173)
`ings (Thomas), had been in contact for 18 hr at
`4° C with a 0.15 M phosphate buffer solution
`containing the labeled metformin, at concen(cid:173)
`trations of 0.5, 0.05, and 5 fLg/ml, i.e., approx(cid:173)
`imately equivalent to steady-state plasma con(cid:173)
`centration of chronically treated patients, 1/ 10
`to 10 times this value.
`Protein concentrations18 and radioactivity
`were estimated both inside and outside the
`
`dialysis bags and albumin content of each serum
`sample was determined by electroimmunodif(cid:173)
`fusion against anti-albumin serum. 16 For radio(cid:173)
`activity determination, a I: 2 Lumasolve (Lu(cid:173)
`mac GmbH-Essen)-isopropanol solution was
`added to plasma followed by bleaching with
`H 20 2 . Samples were then added with a 9: I
`Lumagel-0.5 M HCI solution, and counted on a
`3385 Packard B Counter. Values were cor(cid:173)
`rected for efficiency and background.
`Ancillary measurements. All subjects un(cid:173)
`dergoing the oral and intravenous tests, in par(cid:173)
`ticular patients with renal excretory dysfunc(cid:173)
`tion, were carefully monitored for any changes
`in vital signs (EKG, blood pressure) during the
`experiments. In addition, plasma lactate levels
`were determined enzymatically (Boehringer(cid:173)
`Mannheim kit) in all patients undergoing the
`intravenous kinetic studies, and in patients in
`chronic treatment with metformin.
`
`Results
`Kinetic parameters in subjects with normal
`renal excretory function. After intravenous
`administration of metformin (the injected dose
`was calculated to be 926 ± 26 mg), a biexpo(cid:173)
`nential plasma decay curve was determined in
`all patients. Correlation coefficients for both the
`a and f3 phase regression lines were always
`higher than 0.7, with a good fitting (p < 0.001)
`of the observed values with the calculated lines
`
`AUROBINDO EX. 1021, 5
`
`

`

`688 Sirtori et al.
`
`c/in. Pharmacal. Ther.
`December 1978
`
`ng/ml
`
`•
`
`•
`
`3
`
`4
`
`5
`
`6
`
`7
`
`8
`
`9
`
`10
`hrs
`
`11
`
`Fig. 2. Plasma metfonnin decay curve in a volunteer with normal renal function, following the
`intravenous injection of 1.0 gm metfonnin.
`
`for all the subjects under study. Fig. 2 shows
`the curves obtained in one normal volunteer.
`Pharmacokinetic parameters calculated for each
`subject from the A, et, B, and {3 hybrid con(cid:173)
`stants derived from the plasma curves are
`shown in Table 11. A mean value of 1.52 ± 0.3
`hr was calculated for the t11z of the unchanged
`drug (Fig. 3). A mean 86 ± 2l.9% of the ad(cid:173)
`ministered dose was recovered unchanged in
`urines. Despite the variability of elimination,
`the areas under the plasma decay curves and the
`urinary excretion at fixed time intervals for all
`subjects were found to correlate (p < 0.001).
`Plasma concentration curves after oral admin(cid:173)
`istration of metformin to the same subjects who
`received the drug by intravenous injection are
`shown in Fig. 4. Data from only four volunteers
`were available, due to the loss of several sam(cid:173)
`ples from the fifth subject (pattern of the plasma
`curve, estimated from the available samples,
`however, was not different from that of the
`others). Urinary excretion after oral adminis-
`
`tration of the tablets was 37.6 ± 10.6% of
`the dose after 48 hr. This recovery differs
`(p < 0.02) from that observed after intravenous
`administration to the same SUbjects. Bioavaila(cid:173)
`bility, as expressed by the ratio between the
`AUC's, was 0.32.
`To evaluate the hypothesis of a low bioavail(cid:173)
`ability of the metformin tablets, the solution
`used for the intravenous experiments, as sug(cid:173)
`gested by Huffman, Manion, and Azarnoff,12
`was given to 4 normal male subjects. One gram
`of the metformin Hel solution (the exact con(cid:173)
`centration was later controlled by mass frag(cid:173)
`mentography) was administered, and urines
`were collected for the following 48 hr. The uri(cid:173)
`nary recovery of metformin in these subjects
`was 29.7 ± 10.3%, which does not differ from
`that after oral administration of tablets.
`Kinetic parameters in patients with altered
`renal excretory function. The results after in(cid:173)
`travenous injection of metformin to patients
`with impaired renal function are reported in
`
`AUROBINDO EX. 1021, 6
`
`

`

`Volume 24
`Number 6
`
`Disposition of metformin 689
`
`IO~
`
`ng/ml
`
`ra • 0.94
`
`r~ • 0.97
`
`7
`:3
`2
`4
`5
`6
`Fig. 3. Plasma metformin decay curve in 5 volunteers with normal renal function; X ± SD.
`Regression lines of the ex and {3 phase were fitted by the least-squares method from the mean values
`at each time interval.
`
`Table 11 and Fig. 5. The mean plasma tV2 for
`these patients was 4.94 ± 1.11 hr, i.e., longer
`(p < 0.005) than that of normal volunteers.
`Analysis of the different disposition constants
`(Table II) shows reduced ,B -coefficient, with
`decreased rate of elimination (Kel)' total and
`renal clearance (Cltot and Clren), and volume of
`distribution (V d)/l' There was significant corre(cid:173)
`lation between the plasma tV2 of metformin and
`the CrCl in the normal and nephropathic pa(cid:173)
`tients (Fig. 6). The extrapolated tV2 at CrCl = 0
`was found to be 7 hr; it was 10.9 hr when calcu(cid:173)
`lated from the extrarenal elimination in normal
`subjects.
`Relationship between dose, plasma levels,
`and plasma lactic acid. Plasma levels after
`chronic treatment were available for 55 pa(cid:173)
`tients. All of these had plasma creatinine levels
`below 1.2 mg/ 100 ml and a CrCl within normal
`limits. The range of steady-state plasma levels
`was between 60 and 2,240 ng/ml. Means ±
`
`SEM for different dose levels were as follows:
`
`500 mg/day (n = 6): 161.7 ± 91 ng/ml;
`1,000 mg/day (n = 4): 384 ± 62.4 ng/ml;
`1,700 mg/day (n = 13): 382.3 ± 76.9 ng/ml;
`2,550 mg/day (n = 12): 1,038 ± 192.4 ng
`
`There was a correlation between doses adminis(cid:173)
`tered and the steady-state plasma levels (r =
`0.50; P < 0.001). Plasma lactate levels were
`determined in 43 patients concomitantly with
`the drug level. Only in two was the lactate level
`above 2 mM/I; both of these had a low-normal
`drug level (160 and 570 ng/ml). There was no
`correlation between plasma lactic acid and drug
`levels (r = 0.21, P < 0.1). Lactic acid levels
`were also determined during the intravenous
`kinetic studies. In none of these were there
`significant elevations of lactacidemia.
`Protein binding. Protein binding in serum
`samples from 5 healthy volunteers (albumin
`
`AUROBINDO EX. 1021, 7
`
`

`

`690 Sirtori et al.
`
`Clin. Pharmacal. Ther.
`December 1978
`
`10'
`
`ngAnI
`
`10'
`
`ra =0.94
`rf3 = 1.00
`
`I~L-~-r--.-------.--------'-------'~
`hrs
`40
`10
`20
`4 6
`30
`
`Fig. 4. Plasma levels of metformin in 4 normal volunteers following administration of one 850-mg
`tablet.
`
`10
`
`5
`
`• •
`
`r = 0.88
`
`10
`
`20
`
`hrs
`
`Fig. 5. Plasma metformin decay curve in patients
`with impaired renal excretory functions. Regression
`lines were fitted as in Fig. 3.
`
`Fig. 6. Correlation between half-lives of the ,a-phase
`of the metformin plasma decay curves and the
`creatinine clearance (CrCI) of subjects with normal
`and impaired renal excretory functions.
`
`50
`
`100 ...!!!L CrCI
`min
`
`150
`
`concentrations ranging between 3.23 and 4.77
`gm/lOO ml) was found to be practically negli(cid:173)
`gible. As shown in Table Ill, labeled metformin
`at all three concentrations was perfectly equili(cid:173)
`brated in the dialysis system.
`
`Discussion
`
`Our study reports on kinetic parameters of
`metformin in subjects with normal and altered
`renal excretory functions. A highly specific and
`sensitive fragmentographic method was de(cid:173)
`veloped for
`the determination of drug
`in
`plasma. The use of this method was necessary
`because preliminary tests described by other au(cid:173)
`thors20 for gas-liquid chromatography did not
`provide reproducible results in our hands. By
`
`using a biguanide analog, buformin, as the in(cid:173)
`ternal standard, we noted excessive variability
`in the relative recoveries after extraction of the
`biguanides from basic solutions with methylene
`chloride, as well as in the formation of the
`chlorodifiuoroacetic anhydride derivative.
`Mass fragmentography has the advantage of
`distinguishing between the compound to be
`tested and a deuterated standard, with identical
`physicochemical properties. The measurement
`of metformin levels in plasma and urines is thus
`independent of loss of material during sample
`preparation. This method indicates excellent
`correlation of the experimental values with the
`line calculated by regression analysis (Fig. 1)
`and allows the measurement of metformin con-
`
`AUROBINDO EX. 1021, 8
`
`

`

`Volume 24
`Number 6
`
`Disposition of metformin 691
`
`Table Ill. Results of equilibrium dialysis of 14C-metformin between normal plasma and buffer
`
`Metformin in buffer
`before dialysis (ng Iml)
`
`50
`500
`5,000
`*x :!: so of 5 subjects.
`
`Proteins (mglml)
`
`Metformin (nglml)*
`
`I
`
`Plasma
`
`79.8 ± 5.1
`84.2 ± 6.4
`8l.8 ± 5.6
`
`Buffer
`
`Plasma
`
`l.7 ± l.6
`l.5 ± 1.4
`1.5 ± l.5
`
`29.5 ± 1.5
`282.8 ± 7.3
`293.5 ± 1.37
`
`I
`
`Buffer
`
`30.0 ± l.6
`292.4 ± 4.3
`305.4 ± 4.9
`
`centrations in biologic samples up to nanogram
`levels.
`After intravenous injection to healthy sub(cid:173)
`jects, metformin was shown to be rapidly
`cleared from plasma, with a t1f2 (1.52 hr)
`significantly lower than that reported for phen(cid:173)
`formin (from 7 to 15 hr) and for buformin (4 to
`6 hr). 2, 8, 11, 17 The almost complete urinary
`recovery of unchanged drug after intravenous
`injection indicates that met form in is not me(cid:173)
`tabolized by man21 and that as buformin it is
`significantly secreted by the renal tubule. 17
`A comparison of the disposition constants for
`metformin after intravenous administration to
`normals and renal patients, as reported in Table
`11, shows that the tYz of the f3-phase differs
`significantly between the two groups. Similar
`differences were observed in the rate of elimi(cid:173)
`nation and in the total and renal clearances.
`Renal patients also have significant reduction in
`the apparent volume of distribution. The mark(cid:173)
`edly reduced V d/3, found in renal patients for
`other drugs with a short plasma t1f2 and predom(cid:173)
`inant renal excretion, has been attributed to a
`different plasma/tissue distribution of drugs in
`our patients. 9
`In normals and in renal patients, the t1fz of
`the unchanged drug correlated with creatinine
`clearance, thus indicating the possibility of
`cumulation in severe kidney impairment. This
`would, however, be far less likely than with
`phenformin or buformin.
`Metformin is most often administered three
`times daily (i.e., at approximately 8-hr inter(cid:173)
`vals), and cumulation of unchanged drug after
`chronic treatment
`
`( Fcum =
`
`1
`1-2-<
`
`, where E = r*)
`t1f2
`
`*Interval of drug administration in hours.
`
`will be approximately 20% higher in a patient
`with a t1Iz of 3 hr (i.e., with a creatinine clear(cid:173)
`ance between 30 and 60 ml) than in one with a
`normal t1f2 of 1.5 hr. Doubling of cumulation
`would occur only with tllz above 7 hr, i.e., in
`patients with CrCl of 30 ml/min or less.
`In patients with normal renal function chron(cid:173)
`ically treated with metformin, there was sig(cid:173)
`nificant correlation between the doses adminis(cid:173)
`tered and the steady-state plasma levels. No
`correlation was found in these same subjects
`between the steady-state plasma drug levels and
`PCr or CrCl. In chronically treated patients with
`no evidence of LA, plasma metformin levels
`contrast strikingly to plasma levels of 500 to
`800 ng/ml in acidotic patients treated with
`phenformin or buformin. 8, 13, 28 This difference
`is, however, to be ascribed to the higher daily
`doses of metformin used in therapy (recom(cid:173)
`mended daily doses for phenformin or buformin
`are respectively 50 to 100 mg and 250 to 350
`mg). The relatively low toxicity of met form in in
`association with high plasma levels may pos(cid:173)
`sibly depend on the lower affinity of metformin
`for mitochondrial membranes, recently shown
`by Schiifer22 to be part of the mechanism of
`action of biguanides, as well as a likely expla(cid:173)
`nation for the increased plasma lactic acid levels
`induced by this type of drugs.
`Reliable data are unfortunately not available
`for plasma metformin levels in patients with
`LA. A recent report by Assan and co-workers4
`on several cases of this syndrome in metfor(cid:173)
`min-treated diabetics indicates drug levels 10-
`to lOO-fold those found by us. Conlay and co(cid:173)
`workers8 also noted a discrepancy between their
`findings and those of Assan and co-workers 4
`on phenformin plasma levels in patients with
`LA.3 These discrepancies are probably depen(cid:173)
`dent on the colorimetric method for guanidine
`
`AUROBINDO EX. 1021, 9
`
`

`

`692 Sirtori et al.
`
`Clin. Pharmacal. Ther.
`December 1978
`
`compounds used by Assan and co-workers,3, 4
`not adequately specific for measuring very low
`drug levels in plasma. Monitoring of plasma
`lactic acid levels never brought out significant
`elevations during our kinetic studies. Only
`in two chronically treated subjects was lac(cid:173)
`tacidemia slightly above normal; both of these
`subjects had plasma metformin levels in the low
`range.
`In the study by Conlay and co-workers 8 on
`phenformin, it was noted that plasma levels of
`unchanged drug in toxic patients do not appear
`to bear a significant correlation with
`lac(cid:173)
`tacidemia. This can be due to the fact that a
`considerable interval had elapsed between diag(cid:173)
`nosis of LA and sampling of plasma drug levels
`and, perhaps more important, to the presence of
`phenfonnin metabolites. Thirty percent of a
`dose of phenformin is present in plasma as a
`hydroxylated derivative,I:3 for which no analyt(cid:173)
`ical methods are presently available. In con(cid:173)
`trast, metfonnin is apparently not metabolized
`and the only excretory means is the kidney.
`Thus, measuring metformin in plasma may be a
`more useful parameter for predicting toxicity.
`Moreover, metformin, unlike phenformin, is
`not plasma protein bound; this factor would thus
`become unimportant in uremic patients, who
`may have a lower plasma protein binding
`affinity for drugs. 7
`After oral administration, urinary recovery of
`metfonnin was very low both when the drug
`was given in tablet fonn and when metfonnin
`was given as the solution used in the intrave(cid:173)
`nous study (which provides maximal bioavail(cid:173)
`ability). 12 If a poor bioavailability of metformin
`tablets is excluded, there must be other expla(cid:173)
`nations for the poor absorption. Although the
`metabolism of metfonnin was not the object of
`this study, parallel animal investigations, in
`mice and rabbits, published elsewhere,25 were
`carried out with 14C-metfoimin to examine the
`distribution of the drug and the possible pres(cid:173)
`ence of metabolites. In both animal species,
`after oral and intravenous administration, sig(cid:173)
`nificant binding of drug to the intestinal wall
`was detected. Liver/intestine ratios (per mg of
`protein) of 1. 44 and 1. 51 were found in rabbits
`6 and 4 hr after the oral and intravenous admin(cid:173)
`istrations of labeled metformin; in mice, the
`
`liver/intestine ratios, 2 hr after the oral and in(cid:173)
`travenous administrations, were 0.29 and 0.34.
`Kinetic parameters, as determined in the rabbits
`after
`intravenous administration,
`resembled
`those in man, with a tY2 of the plasma ,a-phase
`of 1.47 hr. The search for specific drug metab(cid:173)
`olites in rabbits after oral administration of 14C_
`metforrnin gave negative results: levels of un(cid:173)
`changed drug, as determined by the fragmen(cid:173)
`tographic technique, were essentially identical
`to levels calculated from plasma radioactivity at
`different time intervals after drug administra(cid:173)
`tion. The hypothesis that the intestine is a major
`site of action for metformin is supported by
`similar findings with buformin 17 and by the re(cid:173)
`sults of several experimental studies on the ef(cid:173)
`fects of biguanides on intestinal absorptive
`mechanisms. 27
`In conclusion, kinetic studies on metfonnin
`demonstrate rapid excretion of unchanged drug
`with no protein binding. Excretion is slower in
`nephropathic patients, but only with extremely
`impaired renal excretory function is significant
`cumulation likely to occur. These results con(cid:173)
`finn the clinical observation that only in some
`patients with a very severe kidney impainnent,4
`metformin may be retained for a long enough
`period of time to induce LA. This may explain
`the lower incidence of LA reported with met(cid:173)
`formin than with other biguanides.
`
`We thank Or. Daniel Azarnoff for his generous
`help in reviewing this manuscript, Drs. GiancarIo
`Nesi and Enzo Cocuzza (Spemsa-Aron, Florence) for
`kindly providing the different pharmaceutical prep(cid:173)
`arations of metformin and for constant advice and
`support of this study, Drs. Co hen and A vramoussis
`(Aron, Paris, France) for supplying the intermediates
`for the synthesis of deuterated metformin and for
`generous support throughout the study, Or. Vittore
`Tamassia (Milan) who kindly revised the protocols
`and the results and allowed ample time for helpful
`discussion, and Or. M. Fattorini (Centro Anti(cid:173)
`diabetico, Lodi) for collecting samples fr