ORIGINAL RESEARCH

Population Pharmacokinetics of Semaglutide forType 2 Diabetes

Rune V. Overgaard . Philip H. Delff . Kristin C. C. Petri .

Thomas W. Anderson . Anne Flint . Steen H. Ingwersen

Received: February 4, 2019 / Published online: February 20, 2019� The Author(s) 2019

ABSTRACT

Introduction: The aim of the present analysiswas to characterise the absorption, distributionand elimination of semaglutide by means ofpopulation pharmacokinetic (PK) models usingdata from nine clinical pharmacology trialsconducted in both healthy subjects and thosewith type 2 diabetes.Methods: Data were obtained from trials withsubcutaneous and intravenous administrationof semaglutide that utilised frequent PK sam-pling and included a total of 353 subjects with10,573 concentration values.

Results: Semaglutide PK properties across trials,drug product strengths and populations werewell characterised by a two-compartment modelwith first-order absorption and elimination. Fora typical subject with type 2 diabetes, clearancewas estimated to be 0.0348 L/h [95% confidenceinterval (CI) 0.0327–0.0369 L/h], and the cen-tral and peripheral volumes were estimated tobe 3.59 L (95% CI 3.28–3.90 L) and 4.10 L (95%CI 3.78–4.42 L), respectively (i.e. a total volumeof distribution of 7.7 L). Interindividual varia-tion was low (* 15%) for both clearance andvolumes of distribution, with low residual error(\5%). Clearance and the total volume of dis-tribution were approximately proportional tobody weight. Minor differences were identifiedbetween healthy subjects and subjects withtype 2 diabetes with respect to clearance andabsorption rate, and between injection siteswith respect to bioavailability.Conclusions: A novel two-compartment modelwas developed to provide the general charac-teristics of semaglutide absorption followingsubcutaneous administration, and of distribu-tion and elimination across administrationroutes. Semaglutide PK was shown to be pre-dictable across populations and administrationroutes and within subjects, and was primarilyinfluenced by body weight.Funding: Novo Nordisk, Bagsværd, Denmark.

Enhanced Digital Features To view enhanced digitalfeatures for this article go to https://doi.org/10.6084/m9.figshare.7667777.

Electronic supplementary material The onlineversion of this article (https://doi.org/10.1007/s13300-019-0581-y) contains supplementary material, which isavailable to authorized users.

R. V. Overgaard (&) � P. H. Delff � K. C. C. Petri �S. H. IngwersenQuantitative Clinical Pharmacology, Novo NordiskA/S, Søborg, Denmarke-mail: RUVO@novonordisk.com

T. W. AndersonDevelopment PK/PD, Novo Nordisk A/S, Maløv,Denmark

A. FlintClinical Pharmacology, Novo Nordisk A/S, Søborg,Denmark

Diabetes Ther (2019) 10:649–662

https://doi.org/10.1007/s13300-019-0581-y

Keywords: Clinical trials; Diabetes; Pharma-cokinetics; Pharmacotherapy; Population analy-sis; Semaglutide

INTRODUCTION

Semaglutide is a receptor agonist of the natu-rally occurring peptide hormone glucagon-likepeptide-1 (GLP-1). As with native GLP-1,semaglutide acts to stimulate insulin secretionand inhibit glucagon release depending on theblood glucose concentration, leading toimproved blood glucose levels together with areduced risk of hypoglycaemia [1, 2]. Semaglu-tide is approved for the treatment of type 2diabetes mellitus, using the subcutaneous routeof administration. With 94% amino acidsequence homology to native GLP-1, semaglu-tide has three structural modifications thatmake it less susceptible to degradation bydipeptidyl peptidase-4 (DPP-4) enzymes andprolong its half-life to approximately 1 week[1, 3]. Semaglutide is thereby appropriate foronce-weekly administration.

The prolonged exposure of semaglutide ascompared with native GLP-1 is mainly drivenby slow systemic elimination and, to a lesserextent, by delayed absorption [4], in line withthe observation that more than 99% of thesemaglutide is bound to plasma albumin [5]. Ina metabolism trial, semaglutide was the primarycomponent circulating in plasma and, prior toexcretion, it was metabolised by proteolyticcleavage of the peptide backbone and sequen-tial beta-oxidation of the fatty acid side chain[4]. Semaglutide metabolites are primarilyexcreted in the urine and faeces [4].

Semaglutide population pharmacokinetics(PK) and exposure in terms of average concen-tration in the dosing interval have been previ-ously investigated for subcutaneous dosingusing phase 3 steady-state data from subjectswith type 2 diabetes [6]. Of the covariatesinvestigated, only body weight had a relevanteffect on the exposure of semaglutide at eitheronce-weekly maintenance dose of 0.5 or 1.0 mg[6].

In addition to an assessment of semaglutideexposure levels, the aim of the present analysis

was to provide additional details of the PKproperties of semaglutide in terms of absorp-tion, distribution and elimination and theirrelationships to covariate factors, in order todevelop a novel disposition model that wouldbe generally applicable across different routes ofadministration. This was achieved by means ofpopulation PK models using frequently sampledPK profiles from nine clinical pharmacologytrials. Inclusion of data from intravenousadministration provided information about thestructural two-compartment disposition model,allowing for separate estimation of clearanceand bioavailability.

The analysis was applied to identify andvalidate the use of a one-compartment modelpreviously applied to analyse sparsely sampledphase 3 data [6], and subsequently to investi-gate the consistency between PK data fromphase 1 and phase 3 trials.

METHODS

Clinical Data Included in the Analysis

A two-compartment PK model was estimatedusing data from nine clinical pharmacologytrials employing single as well as multiplesemaglutide doses and conducted in subjectswith type 2 diabetes or obesity, or in healthyvolunteers (Table 1). The included trials werethose with administration of recombinantsemaglutide in which PK samples, assessed vialiquid chromatography and tandem mass spec-trometry (LC–MS/MS) [3], were available at thetime of the analysis. One trial provided data forthe intravenous administration of semaglutideand also used different drug product strengths(see further descriptions in the ‘‘Methods’’ sec-tion of the Electronic supplementary material,ESM).

In order to compare the current analysis withthe analysis conducted using phase 3 data, aone-compartment model was estimated usingdata from seven of the clinical pharmacologytrials (Table 1) which employed multiple dosesof the semaglutide drug product used in thephase 3 trials. Data from the two trials with

650 Diabetes Ther (2019) 10:649–662

Table1

Overviewof

clinicaltrialsincluded

intheanalysis

Trial

[reference]

Description

Num

ber

of subjects

Pop

ulation

PK

data

Num

berof

samples/sub

ject

Doses

Drug

prod

uct

strength

(mg/mL)

1[18]*

Japanese

safety

and

PK

32Health

ymale

Japanese

and

Caucasians

SS?

SD

profiles

460.25

mg[

0.5mg[

1.0mg

1.34

2[19]*

Beta-cellfunction

37Males

andfemales

withT2D

SSprofiles

240.25

mg[

0.5mg[

1.0mg

1.34

3[5]

Hepatic

impairment

44Health

ymales

and

females

andthose

withhepatic

impairment

SDprofiles

200.5mg

1.34

4[20]*

QTstudy

82Health

ymales

and

females

Trough

samples

?SS

profiles

500.25

mg[

0.5mg[

1.0mg[

1.5mg

1.34

5[21]*

Hypoglycaem

ic

coun

terregulation

37Males

andfemales

withT2D

Trough

samples

?SS

profiles

240.25

mg[

0.5mg[

1.0mg

1.34

6[22]*

Energyintake

30Males

andfemales

withobesity

SSprofiles

160.25

mg[

0.5mg[

1.0mg

1.34

7a(dataon

file)

Drugproduct

strengths

(equivalence)

42Health

ymales

and

females

TwoSD

profiles

20(s.c.)

37(i.v.)

0.5mgs.c.

0.25

mgi.v.

1,3and10

8[23]*

DDIstudy1

23Health

ymales

and

females

Trough

samples

?SS

profiles

270.25

mg[

0.5mg[

1.0mg

1.34

Diabetes Ther (2019) 10:649–662 651

single dose administration were not included inthis comparison.

Population Pharmacokinetic Modelling

Datasets containing plasma concentration andcovariate data from each trial were converted toa format compatible with the nonlinear mixed-effects modelling (NONMEM) software (ICONDevelopment Solutions, Ellicott City, MD,USA), version 7.1.2. Out of a total of 10,573concentration values, 43 values (i.e.\1%) wereexcluded because the subjects had an inade-quate dosing history. Conditional weightedresiduals (CWRES) were used as a diagnostictool to identify outliers, and an additional 24concentration values (23 in the one-compart-ment model) were excluded from the finaldataset as they were associated with CWRESgreater than 6 in order to limit bias from influ-ential data items. The software package R (ver-sion 3.2.3, obtained from the R Foundation:http://www.r-project.org/foundation) was usedfor data file processing, explorative data analysisand to plot figures.

Model Development

Model development was conducted in NON-MEM as follows:1. The structural disposition model, which

was supported by graphical analysis, inparticular of intravenous data, was a two-compartment model with first-order elimi-nation. Different absorption models wereinvestigated, including parallel absorption,but a simpler first-order absorption modelfor subcutaneous administration was foundto adequately describe the data, and nosignificant improvement was obtainedbased on other absorption models.

2. An additive residual error model on log-transformed concentration values wasassessed to adequately describe the data.

3. Inter-individual variability in the absorp-tion rate (ka), clearance (CL), intercompart-mental clearance (Q), central volume (Vc),and peripheral volume (Vp) was exploredusing a log normal distribution in a model

Table1

continued

Trial

[reference]

Description

Num

ber

of subjects

Pop

ulation

PK

data

Num

berof

samples/sub

ject

Doses

Drug

prod

uct

strength

(mg/mL)

9[23]*

DDIstudy2

26Health

ymales

and

females

Trough

samples

?SS

profiles

270.25

mg[

0.5mg[

1.0mg

1.34

Alltrialswereused

forthetwo-compartmentmodel,and

trialsmarkedwithan

*wereused

fortheone-compartmentmodel

Doses

wereadministeredusingthePD

S290

injectionpen,

except

fortrial7,

which

used

NovoP

en4fors.c.adm

inistrationandasyringefori.v.adm

inistration

Clin

icalTrials.gov

identifiers

for

the

above

trials:

NCT02146079;NCT02212067;NCT02210871;NCT02064348;NCT02147431;NCT02079870;

NCT02231684;NCT02022254;andNCT02243098

DDIdrug–d

ruginteraction,

i.v.intravenous

dosing,P

Kpharmacokinetic(s),s.c.subcutaneousdosing,S

Dsingle-dose,SS

steady-state,T

2Dtype

2diabetes

aDataon

file,partlyreported

inthepresentpublication

652 Diabetes Ther (2019) 10:649–662

that included covariates with clear effects:drug product strength on the absorptionrate constant and body weight as a covariatefactor on CL, Vc and Vp. A correlationbetween the central volume and CL wassignificant and included in the model. Dueto the limited amount of intravenous data,less information was available for individualestimation of Q and for the separation of Vc

and Vp. Various implementations for theseparameters were investigated and selectedbased on objective function values; i.e.using individual (eta) values for CL toidentify differences across subjects for Q,as well as for Vc and Vp.

4. The covariate effects of sex, race, ethnicity,age, hepatic impairment and glycaemicstatus were investigated for CL, Vc and kabased on a graphical analysis of individualparameter estimates versus body weight,and glycaemic status was identified as acovariate factor for CL and ka. Similarly,injection site was investigated with respectto F and ka. Covariates with 90% confidenceintervals which were within the 80–125%equivalence interval were not consideredimportant.

5. Covariate effects were investigated using afull model approach with inclusion of allinvestigated covariates in one step (Table 2).The final model included only those covari-ates that were significant (10% alpha level).To assess the amount of variabilityexplained by covariate factors, a base modelwas investigated, including only covariatesfor the effects of drug product strength onthe absorption rate constant.

Model Description

The full model for subcutaneous administrationwas a model with first-order absorption, two-compartment disposition and first-order elimi-nation, whereas intravenous data were mod-elled as a bolus injection into the centralcompartment. Inter-individual variability wasestimated for ka, and CL/Q (with a commonfactor), and for Vc/Vp (with a common factor).

The covariate models were parameterised asfollows for the ith individual:

CLi ¼ CLref � Eweight;CL � Esex � Erace � Eethnicity

� Eglycaemicstatus � Eagegroup � expðgCL;iÞ

Qi ¼ Qref � Eweight;CL � expðgCL;iÞ

Vc;i ¼ Vc;ref � Eweight;V � expðgV;iÞ

Vp;i ¼ Vp;ref � Eweight;V � expðgV;iÞ;

and the effect of drug product strength on therate of absorption was described as

ka ¼ ka;ref� �1:34 mg� hka; 1 mg

� �1mg� hka;3mg

� �3mg�

� hka;10mg

� �10mg�� hka;T2DT2D � exp gka:i

� �:

Table 2 Covariates included in the full two-compartmentmodel

PK Parameter Covariates includeda

Clearance (CL) Body weight (continuous)

Sex (female, male)

Race (white, black, Asian)

Ethnicity (Hispanic, non-

Hispanic)

Glycaemic status

(normoglycaemia, T2D)

Age group (B 65,[ 65 years)

Hepatic impairment

Bioavailability (F) Injection site (abdomen, thigh)

Intercompartmental

clearance (Q)Body weight

Central volume (Vc) Body weight

Peripheral volume

(Vp)

Body weight

Absorption rate

constant (ka)Drug product strength (1, 1.34, 3

and 10 mg/mL)

Glycaemic status

(normoglycaemia, T2D)

T2D type 2 diabetesa Assessed at baseline

Diabetes Ther (2019) 10:649–662 653

Covariate relationships were described asfollows:

F ¼ Fref hthigh� �thigh

Eweight;CL ¼ weighti85 kg

� �hweigth;CL

Eweight;V ¼ weighti85 kg

� �hweight;V

Esex ¼ hmaleð Þmale

Erace ¼ hblackð Þblack� hAsianð ÞAsian

Eethnicity ¼ hHispanic� �Hispanic

Eglycaemicstatus ¼ hT2Dð ÞT2D

Eagegroup ¼ h[ 65years

� �[ 65years:

In the above equations, h represents anestimated value for a covariate and g representsbetween-subject variability. The otherexponents are indicator variables assigned avalue of 1 for the specific covariate values and0 otherwise. CLref, Qref, Vc,ref, Vp,ref, Fref and ka, refare parameter values for the reference subjectprofile, which is a healthy, white, non-Hispanicfemale aged B 65 years with a body weight of85 kg, using the abdomen as the injection siteand dosed with a 1.34 mg/mL semaglutide drugproduct. The residual error model was additivefor log-transformed observations.

Steady-State Exposure Evaluationfor Subcutaneous Semaglutide

A one-compartment model, similar to the oneused to evaluate phase 3 data, was investigatedusing steady-state data from subcutaneous dos-ing only in clinical pharmacology trials; hence,the fraction absorbed (F) could not be estimatedseparately in this model. Therefore, the modelwas parameterised by the apparent clearance(CL/F), volume of distribution (V/F) and theabsorption rate constant (ka). The inter-indi-vidual variability was estimated for each of

these. The residual error model was additive forlog-transformed observations.

The one-compartment model containedthree covariate relationships: body weight andglycaemic status (healthy, type 2 diabetes) onCL, and body weight on volume of distribution.These were parameterised as for the two-com-partment model.

Compliance with Ethics Guidelines

This article is based on previously conductedstudies and does not contain any studies withhuman participants or animals performed byany of the authors.

RESULTS

Population Characteristics

A total of 353 subjects with 10,573 semaglutideconcentrations were included in the populationPK model analysis. For the comparison to thephase 3 modelling approach, 267 subjects con-tributed with 8292 semaglutide concentrationvalues. The demographic characteristics for thesubjects included in each of the analyses aresummarised in Table 3.

Semaglutide Pharmacokinetics

The final population PK model provided anaccurate description of the PK time courseacross the nine clinical pharmacology trials.Figure 1 shows the consistency between themodel and data across the eight trials whichemployed both single and multiple doses ofmarketed semaglutide with a drug productstrength of 1.34 mg/mL. Figure 2 shows that thebiphasic PK profile obtained from the two-compartment model is consistent with intra-venous data and, furthermore, that the rate ofabsorption as well as the maximum concentra-tion following subcutaneous doses of differentstrengths were well captured by the model.

The model-based analysis suggested thatdrug product strength influences the rate of

654 Diabetes Ther (2019) 10:649–662

absorption but not the bioavailability (Table 4),indicating that the total exposure of semaglu-tide is not affected by the product strength.These model-based results are consistent withthe noncompartmental results for trial 7 (see

Table S1 in the ESM), in which equivalence wasdemonstrated for exposure (AUC0–?) for allpairwise comparisons between the threestrengths of semaglutide. For Cmax, the com-parison between 1 mg/mL and 3 mg/mL met

Table 3 Demographic characteristics of subjects included in the two- and one-compartment models

Category Group Total 2-compartment Total 1-compartment

All N 353 267

Population Normoglycaemia 277 193

T2D 76 74

Race White 327 243

Asian 16 16

Other/missing 7 5

Black/African American 3 3

Ethnicity Not Hispanic 349 264

Hispanic 4 3

Sex Male 226 180

Female 127 87

Dosing Multiple dose 267 267

Single dose 86 –

Dose, mg 0.25 5 5

0.5 105 19

1.0 166 166

1.5 77 77

Injection site Abdomen 286 244

Thigh 67 23

Age, years Mean (SD) 44.6 (11.8) 44.1 (11.5)

Range [19–70] [19–70]

BMI, kg/m2 Mean (SD) 26.9 (4.3) 27.0 (4.3)

Range [18.7–42.8] [20.0–42.8]

Body weight, kg Mean (SD) 81.9 (15.1) 83.0 (14.7)

Range [51.9–121.2] [54.1–121.2]

Hepatic impairment status Mild, moderate or severe 25 25

No impairment 19 19

BMI body mass index, N number of subjects, SD standard deviation, T2D type 2 diabetes

Diabetes Ther (2019) 10:649–662 655

the equivalence criterion but equivalence wasnot shown for 1 mg/mL relative to 10 mg/mL orfor 3 mg/mL relative to 10 mg/mL, as Cmax

increased with increasing strength.All parameters were successfully estimated

(relative standard error B 11%) in the finalmodel using data from the nine clinical trials(Table 4). In accordance with findings based on

phase 3 semaglutide data investigating covari-ate effects on CL/F [6], body weight was foundto be the only important covariate with respectto CL and, hence, average semaglutide con-centrations under steady-state conditions with90% CIs outside the equivalence interval.There were also clear relationships betweenbody weight and the Vc and Vp, whereas the kawas independent of body weight (Table 4 andFig. S1c in the ESM). Less important thoughstatistically significant effects included aneffect on semaglutide CL of glycaemic status,with a 12% higher CL (95% CI 9–16%)observed in subjects with type 2 diabetescompared to healthy subjects (Table 4 andFig. S1a in the ESM). Moreover, ka was reducedin subjects with type 2 diabetes compared tohealthy subjects (Table 4 and SupplementaryFig. S1c in the ESM). For a typical subject withtype 2 diabetes and a body weight of 85 kg,clearance was estimated to be 0.0348 L/h (95%CI 0.0327–0.0369 L/h) and the total volume ofdistribution was estimated to be 7.7 L (Vc ? Vp;Table 4). A small effect of injection site wasfound with respect to the bioavailability,which was 12% lower when using the thighcompared with the abdomen (95% CI0.84–0.92). The effects of other covariates werenot relevant as they were found to lie withinthe 80–125% equivalence interval. No effects

Sem

aglu

tide

(nm

ol/L

)

Time (days)

0 1 2 3 4 5 6 7

1

2

5

10

201 mg/mL s.c.10 mg/mL s.c.1 mg/mL i.v.3 mg/mL s.c.

a b

Sem

aglu

tide

(nm

ol/L

)

Time (days)

0 5 10 15 20 25

1

2

5

10

201mg/mL s.c.10 mg/mL s.c.1 mg/mL i.v.3 mg/mL s.c.

Fig. 2a–b Observed and model-predicted subcutaneousand intravenous semaglutide PK profiles over 7 days(a) and 30 days (b). Data are from trial 7 with semaglutidedoses of up to 0.5 mg. Data points with error bars areobserved geometric means and 95% CIs. Model predic-tions (lines) are population predictions from the final two-

compartment model of semaglutide PK. In b, two valuesabove the lower limit of quantification obtained at day 35were not included in the figure. CI confidence interval, i.v.intravenously, PK pharmacokinetic(s), s.c subcutaneously

0 5 10 15 20

0.5

1.0

2.0

5.0

10.0

20.0

50.0

Sem

aglu

tide

(nm

ol/L

)

Time (weeks)

Trial 1, SD & SSTrial 2, SSTrial 3, SDTrial 4, SSTrial 5, SSTrial 6, SSTrial 8, SSTrial 9, SS

Fig. 1 Observed and model-predicted semaglutide phar-macokinetic profiles from eight clinical pharmacologytrials. All trials employed semaglutide doses of up to1.0 mg, except trial 3, which was a single dose (0.5 mg)trial, and trial 4, which used a supratherapeutic dose of1.5 mg. Data points with error bars are observed geometricmeans and 95% CIs. Model predictions (lines) arepopulation predictions from the final two-compartmentmodel of semaglutide PK. CI confidence interval, PKpharmacokinetics, SD single dose, SS steady state

656 Diabetes Ther (2019) 10:649–662

of sex, race, ethnicity, age, or hepatic impair-ment on CL were found, and these covariateswere excluded from the final model.

Trial 7, which provided data from both sub-cutaneous and intravenous administration,allowed the absolute bioavailability for subcu-taneous administration of semaglutide to beestimated. Using noncompartmental analysis,the absolute bioavailability after the subcuta-neous administration of semaglutide was esti-mated to be 89% (95% CI 0.83–0.94). This wascaptured well in the final model, which esti-mated the bioavailability to be 85% (95% CI0.80–0.90) (Table 4).

Inclusion of covariates in the final two-compartment model explained 58%, 56% and24% of the interindividual variance of CL, vol-ume of distribution and ka, respectively.

Model Qualification

Parameter estimates (Table 4) were estimatedwith a low (B 11%) relative standard error inthe final model; shrinkage values were all below10%, supporting the accuracy of the individualparameter estimates.

The standard visual predictive checks indi-cated that the model was able to reproduce thedata used for estimation in terms of both med-ian concentrations and variability betweensubjects (Fig. S2 in the ESM). In addition, anillustration of variability in a dosing intervalwas included for the dose-normalised steady-state concentrations from the target semaglu-tide doses 0.5 and 1.0 mg (Fig. 3a). Subjects atother dose levels and those who did not com-plete the study were excluded from this analy-sis. Thus, the figure was limited to include data

Table 4 Parameter estimates obtained from the final two-compartment model

Parameter Estimate [95% CI] RSE (% CV) IIV (% CV) Shrinkage (%)

Absorption rate constant (ka), h-1 0.0253 [0.0236–0.027] 3.42 37.9 9.4157

Clearance (CL), L/h 0.0348 [0.0327–0.0369] 3.06 15.2 1.5202

Central volume (Vc), L 3.59 [3.28–3.9] 4.44 15.4 6.6167

Intercompartmental clearance (Q), L/h 0.304 [0.249–0.359] 9.19 15.2 1.5202

Peripheral volume (Vp), L 4.10 [3.78–4.42] 3.97 15.4 6.6167

Absolute bioavailability (F) 0.847 [0.798–0.896] 2.96 NA NA

Body weight effect on CL and Q 1.01 [0.912–1.1] 4.71 NA NA

Body weight effect on Vc and Vp 0.923 [0.833–1.01] 4.96 NA NA

T2D effect on CL 1.12 [1.09–1.16] 1.66 NA NA

Injection site effect on F 0.883 [0.844–0.921] 2.24 NA NA

Drug product strength effects on ka, 1 mg 0.0346 [0.0301–0.039] 6.53 NA NA

Drug product strength effects on ka, 3 mg 0.0526 [0.0438–0.0615] 8.58 NA NA

Drug product strength effects on ka, 10 mg 0.139 [0.109–0.168] 11.0 NA NA

T2D effect on ka 0.544 [0.485–0.603] 5.49 NA NA

Residual error 0.103 NA NA NA 4.2914

A covariance of 0.0172 was found between CL and Vc. Data from all 9 trials were included in the modelCI confidence interval, CL clearance, CV coefficient of variation, F bioavailability, IIV interindividual variability, NA notapplicable, Q intercompartmental clearance, RSE relative standard error, T2D type 2 diabetes, Vc central volume, Vp

peripheral volume

Diabetes Ther (2019) 10:649–662 657

from the last PK profile in each trial to illustratethe variability at steady-state conditions.

The dose-normalised time profiles acrossbody weight in 10-kg intervals (Fig. 3b) spanned

a similar range to the overall 90% range, illus-trating that body weight does explain most ofthe variability. Furthermore, the close fit of themodel to data in each of the 10-kg intervalsillustrates that the influence of body weight onthe different parameters was adequately descri-bed by the model.

The above findings indicate that the two-compartment PK model provided an accuratedescription of the PK of semaglutide acrossclinical pharmacology trials, including differentpopulations, administration routes and drugproduct strengths, and was found suitable forpredictive purposes, i.e. for the simulation ofsemaglutide concentration–time profiles invarious populations. Moreover, the lowinterindividual variation of * 15% for both CLand Vc as well as the low residual error (\5%;Table 4) indicate that the PK of semaglutide ishighly predictable.

Estimation of a One-Compartment Modelfor Steady-State Evaluationof Subcutaneous Data

The population PK model was compared to aone-compartment model similar to the oneused in the phase 3 analysis. The two modelsprovided almost identical apparent exposureestimates for individual subjects (Fig. S3a in theESM). In addition, mean simulated PK profilesduring subcutaneous administration werealmost identical, with similar exposure levels aswell as peak-to-trough fluctuations (Fig. S3b inthe ESM). The PK parameter estimates in theone-compartment model used for comparisonwere similar to those obtained with the two-compartment model (Table S2 in the ESM).

Comparison with the Model Estimatedwith Phase 3 Data

Parameter values obtained in the present anal-ysis for a reference subject profile were com-pared to the corresponding parameter valuesobtained with phase 3 data [6] (Table 5). Due tolimitations in the phase 3 data, that analysisapplied a fixed value of ka. Although the totalvolume of distribution was lower in the present

Sem

aglu

tide/

dose

(nm

ol/L

*mg-1

)a

b

Sem

aglu

tide/

dose

(nM

/mg)

0 1 2 3 4 5 6 7

10

20

50

100

Time after last dose (days)

0 1 2 3 4 5 6 710

20

50

100

Time after last dose (days)

50−60 kg60.1−70 kg70.1−80 kg80.1−90 kg90.1−100 kg100.1−110 kg

Fig. 3a–b Variability of semaglutide PK under steady-state conditions (a) and model fits to dose-normalisedsemaglutide concentration–time profiles at steady state fordifferent body weight quantiles (b). For a, points witherror bars are observed steady-state PK data (geometricmean and 90% range) for 173 completing subjects atselected time points with PK sampling across trials.Overlaid are 58 simulated replicates of data (n = 10,034)in a dosing interval after the latest treatment dose. Thelight blue line represents the geometric mean of theobserved data and the shaded area represents the 90%prediction interval. The figure includes data from six trialswith steady-state PK and subjects treated with 0.5 mg and1.0 mg semaglutide as well as the model from nine clinicalpharmacology trials of semaglutide. For b, data points witherror bars are observed geometric means and 95% CIs fromseven clinical pharmacology trials with semaglutide atsteady state concentrations after weekly dosing of 0.5 or1.0 mg. Model predictions are population predictions fromthe final two-compartment model of semaglutide PK. CIconfidence interval, PK pharmacokinetic(s)

658 Diabetes Ther (2019) 10:649–662

analysis than in the phase 3 analysis, profilesfrom these models were similar (Fig. 4), indi-cating that the higher volume of distribution inthe phase 3 model had a relatively small impacton the overall PK profile. High consistencybetween the simulated profiles indicate that thePK in the phase 3 trials could be predicted usingthe two-compartment model that was devel-oped using clinical pharmacology data, sup-porting the overall predictability of semaglutidePK. Furthermore, the estimated inter-individualvariability in clearance (15.2%; Table 4) wasconsistent with phase 3 data (12.9%) [6].

DISCUSSION

The PK of semaglutide were well characterisedby a two-compartment model with first-orderabsorption and elimination, and were shown tobe predictable across clinical pharmacology andphase 3 trials and across subjects with lowinterindividual variation and low residual error.This model provided a joint description ofsemaglutide PK across clinical pharmacologytrials in both healthy subjects and those withtype 2 diabetes with frequently sampled PKprofiles following subcutaneous and intra-venous administration in well-controlledsettings.

The total volume of distribution (Vc ? Vp)was approximately 8 L, similar to the volume ofextracellular water in the body and supportingthe idea that semaglutide distributes intoplasma and peripheral tissues to the sameextent as albumin [7, 8]. The terminal half-lifewas driven by the disposition parameters ratherthan the rate of absorption. The population PKmodel for semaglutide was similar overall tothat developed for dulaglutide [9], whereasliraglutide had much faster elimination [10],and exenatide extended-release had nonlineardisposition and a substantially slower absorp-tion [11]. Compared to semaglutide, dulaglutidehas approximately 50% greater CL and a 20%lower volume of distribution, as well as

Table 5 Comparison of pharmacokinetic parameter values from different semaglutide models

Parameter Two-compartmentfinal model

One-compartmentmodel

One-compartment modelfrom phase 3a trialsa

Absorption rate constant ka, h-1 0.0253 0.0296 0.0286 (fixed)c

CL/F for subject with T2D and body

weight of 85 kg, L/h

0.0460 0.0473 0.0478

CL/F for healthy volunteer with body

weight of 85 kg, L/h

0.0411 0.0415 –

Total volume/Fb, L 9.08 9.77 12.2

CL/F apparent clearance, F bioavailability, T2D type 2 diabetes, Vc central volume, Vp peripheral volume, ka absorption rateconstanta Values obtained from [6]b Total volume/F is (Vc ? Vp)/F for the two-compartment model and V/F for the one-compartment modelc Due to sparse sampling in phase 3 trials, ka was fixed at the value obtained from a previous version of the one-compartment model based on clinical pharmacology data

0 1 2 3 4 5

0

10

20

30

40

50

Sem

aglu

tide

(nm

ol/L

)

Time (weeks)

Two−compartment modelOne−compartment model (Phase 3)

Fig. 4 Simulated PK profiles under steady-state conditionsfor a reference subject profile based on the final two-compartment model developed from clinical pharmacologydata and the one-compartment model developed fromphase 3 data

Diabetes Ther (2019) 10:649–662 659

approximately threefold slower absorption [9].The interindividual variability was, however,greater for dulaglutide (40.5% coefficient ofvariation (CV) for CL) compared to semaglutide(CV for CL was 15.2% in the present analysisand 12.9% in the phase 3 analysis [6]). Similarly,both exenatide extended-release and liraglutidehad large variabilities between subjects com-pared to semaglutide [10, 11].

Based on the present steady-state data, wecompared the developed population PK modelto a one-compartment PK model similar to theone estimated in the phase 3 analysis [6] withalmost identical results. Moreover, we verifiedthat the typical steady-state profiles obtainedfrom phase 3 data [6] were consistent with thosepredicted based on clinical pharmacology data,and similar estimates for clearance CV% alsoverified that variability was consistent com-pared to phase 3 trial results.

In the phase 3 population PK analysis, onlybody weight had a relevant effect on the expo-sure of semaglutide, a finding which did notwarrant dose adjustment [12]. No effects of sex,age, race, ethnicity, renal function or injectionsite used on semaglutide exposure wereobserved [6]. The present analysis confirmed thefinding that only body weight was a relevantcovariate for exposure. Compared to the phase 3evaluation, the current analysis provides a moregeneral description of semaglutide PK that isapplicable across different administrationroutes, and informs on covariate relationshipsnot only for exposure, but also for absorption,distribution and elimination. Additionalcovariate effects were identified for body weighton the Vc and Vp and glycaemic status on CLand the ka. Other investigated covariates (sex,race, ethnicity, age group, hepatic impairment)were not found to be influential and were notincluded in the final model. The impact ofcovariates on the PK of semaglutide supportprevious findings [6, 12] that no adjustment ofthe dose is needed across population subgroups.

With once-daily liraglutide, another GLP-1receptor agonist approved for the treatment oftype 2 diabetes at doses of up to 1.8 mg, bodyweight and sex were found to be the mostimportant covariates for exposure which, how-ever, did not influence liraglutide PK to any

clinically relevant degree [13]. Similarly, neitherage, body weight, sex, race, ethnicity norinjection site influenced once-weekly semaglu-tide or dulaglutide PK to any clinically relevantextent [6, 9]. For twice-daily exenatide andonce-weekly exenatide extended-release, bodyweight was also found to have a significantrelationship with the volume of distribution,and a significant relationship between renalfunction and elimination was found [11, 14].

The slightly higher CL of semaglutide insubjects with type 2 diabetes compared tohealthy subjects has also been seen previouslywith liraglutide 1.8 mg [10] and with liraglutidefor weight management at a higher 3.0 mg dose[15]. The mechanism behind this difference isnot known.

The present analysis showed that injectionsite had a small effect on the bioavailability ofsemaglutide, which was lower following injec-tion in the thigh compared to abdominalinjection. This observation was in accordancewith previous findings that the injection sitehas a minor and not clinically relevant effect onthe CL/F of semaglutide [6]. Similarly, thebioavailability of liraglutide was found to beslightly lower after administration in the thighversus the abdomen [16]. Differences betweeninjection sites are generally considered to be ofno clinical relevance across GLP-1 receptoragonists [9, 17]. A limitation of the presentanalysis is that we cannot exclude the possibil-ity that other demographic factors that we didnot evaluate were influential in the targetpatient population.

CONCLUSIONS

In summary, semaglutide PK was shown to bepredictable across clinical pharmacology andphase 3 trials, with limited interindividualvariability that was primarily influenced bybody weight. A novel two-compartment modelprovided a general characterisation ofsemaglutide absorption following subcutaneousadministration and of distribution and elimi-nation across administration routes. A one-compartment model structure was found to besuitable for conducting covariate analyses of

660 Diabetes Ther (2019) 10:649–662

exposure levels based on sparse PK sampling inlate-stage trials for subcutaneous semaglutide.

ACKNOWLEDGEMENTS

We thank the participants whose data con-tributed to this study, as well as Niels RodeKristensen, PhD (Novo Nordisk A/S, Søborg,Denmark) for taking part in the planning of theanalysis.

Funding. Funding for the trials included inthe analysis, the journal’s article processingcharges as well as the trial products was pro-vided by Novo Nordisk, Bagsværd, Denmark. Allauthors had full access to all of the data in thisstudy and take complete responsibility for theintegrity of the data and accuracy of the dataanalysis.

Medical Writing and Editorial Assis-tance. Editorial and medical writing services inthe preparation of this manuscript were pro-vided by Angela Stocks, PhD, of Larix A/S,Copenhagen, Denmark. Support for these ser-vices was funded by Novo Nordisk.

Authorship. All named authors meet theInternational Committee of Medical JournalEditors (ICMJE) criteria for authorship for thisarticle, take responsibility for the integrity ofthe work as a whole, and have given theirapproval for this version to be published.

Authorship Contributions. Philip H. Delffand Rune V. Overgaard planned the analysiswith contributions from Kristin C.C. Petri andSteen H. Ingwersen. Philip H. Delff and Rune V.Overgaard were responsible for the data collec-tion and statistical analysis. All authors con-tributed to the interpretation of the data. Allauthors were involved in the writing, reviewingand editing of the manuscript, gave finalapproval and agreed to be accountable for allaspects of the work.

Disclosures. Rune V. Overgaard is employedby and holds stock in Novo Nordisk. Philip H.Delff holds stock in Novo Nordisk. He was

employed by Novo Nordisk when the work wascarried out. His current affiliation is: Quantita-tive Clinical Pharmacology, Early ClinicalDevelopment, IMED Biotech Unit, AstraZeneca,Boston, USA. Kristin C.C. Petri is employed byand holds stock in Novo Nordisk. Thomas W.Anderson is employed by and holds stock inNovo Nordisk. Anne Flint is employed by andholds stock in Novo Nordisk. Steen H. Ingw-ersen is employed by and holds stock in NovoNordisk.

Compliance with Ethics Guidelines. Thisarticle is based on previously conducted studiesand does not contain any studies with humanparticipants or animals performed by any of theauthors.

Data Availability. The datasets generatedduring and/or analysed during the currentstudy are available from the correspondingauthor on reasonable request.

Open Access. This article is distributedunder the terms of the Creative CommonsAttribution-NonCommercial 4.0 InternationalLicense (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercialuse, distribution, and reproduction in anymedium, provided you give appropriate creditto the original author(s) and the source, providea link to the Creative Commons license, andindicate if changes were made.

REFERENCES

1. Lau J, Bloch P, Schaffer L, Pettersson I, Spetzler J,Kofoed J, et al. Discovery of the once-weekly glu-cagon-like peptide-1 (GLP-1) analogue semaglutide.J Med Chem. 2015;58:7370–80.

2. Meloni AR, DeYoung MB, Lowe C, Parkes DG. GLP-1 receptor activated insulin secretion from pancre-atic beta-cells: mechanism and glucose depen-dence. Diabetes Obes Metab. 2013;15:15–27.

3. Kapitza C, Nosek L, Jensen L, Hartvig H, Jensen CB,Flint A. Semaglutide, a once-weekly human GLP-1analog, does not reduce the bioavailability of thecombined oral contraceptive, ethinylestradiol/levonorgestrel. J Clin Pharmacol. 2015;55:497–504.

Diabetes Ther (2019) 10:649–662 661

4. Jensen L, Helleberg H, Roffel A, Jaap van Lier J,Bjornsdottir I, Pedersen PJ, et al. Absorption,metabolism and excretion of the GLP-1 analoguesemaglutide in humans and nonclinical species. EurJ Pharm Sci. 2017;104:31–41.

5. Jensen L, Kupcova V, Arold G, Pettersson J, Hjerp-sted JB. Pharmacokinetics and tolerability ofsemaglutide in people with hepatic impairment.Diabetes Obes Metab. 2018;20:998–1005.

6. Petri KCC, Ingwersen SH, Flint A, Zacho J, Over-gaard RV. Semaglutide s.c. once-weekly in type 2diabetes: a population pharmacokinetic analysis.Diabetes Ther. 2018;9:1533–47.

7. Peters TJ. All about albumin: biochemistry, genet-ics, and medical applications. New York: Academic;1995.

8. Diao L, Meibohm B. Pharmacokinetics and phar-macokinetic-pharmacodynamic correlations oftherapeutic peptides. Clin Pharmacokinet.2013;52:855–68.

9. Geiser JS, HeathmanMA, Cui X, Martin J, Loghin C,Chien JY, et al. Clinical pharmacokinetics ofdulaglutide in patients with type 2 diabetes: anal-yses of data from clinical trials. Clin Pharma-cokinet. 2016;55:625–34.

10. Watson E, Jonker DM, Jacobsen LV, Ingwersen SH.Population pharmacokinetics of liraglutide, a once-daily human glucagon-like peptide-1 analog, inhealthy volunteers and subjects with type 2 dia-betes, and comparison to twice-daily exenatide.J Clin Pharmacol. 2010;50:886–94.

11. Cirincione B, Edwards J, Mager DE. Populationpharmacokinetics of an extended-release formula-tion of exenatide following single- and multiple-dose administration. AAPS J. 2017;19:487–96.

12. Petri KCC, Ingwersen SH, Flint A, Zacho J, Over-gaard RV. Exposure-response analysis for evaluationof semaglutide dose levels in type 2 diabetes. Dia-betes Obes Metab. 2018;20:2238–45.

13. Ingwersen SH, Khurana M, Madabushi R, Watson E,Jonker DM, Le Thi TD, et al. Dosing rationale forliraglutide in type 2 diabetes mellitus: a pharma-cometric assessment. J Clin Pharmacol.2012;52:1815–23.

14. Cirincione B, Mager DE. Population pharmacoki-netics of exenatide. Br J Clin Pharmacol.2017;83:517–26.

15. Overgaard RV, Petri KC, Jacobsen LV, Jensen CB.Liraglutide 3.0 mg for weight management: a pop-ulation pharmacokinetic analysis. Clin Pharma-cokinet. 2016;55:1413–22.

16. Kapitza C, Zdravkovic M, Zijlstra E, Segel S, Heise T,Flint A. Effect of three different injection sites onthe pharmacokinetics of the once-daily humanGLP-1 analogue liraglutide. J Clin Pharmacol.2011;51:951–5.

17. Calara F, Taylor K, Han J, Zabala E, Carr EM, WintleM, et al. A randomized, open-label, crossover studyexamining the effect of injection site on bioavail-ability of exenatide (synthetic exendin-4). ClinTher. 2005;27:210–5.

18. Ikushima I, Jensen L, Flint A, Nishida T, Zacho J, IrieS. A randomized trial investigating the pharma-cokinetics, pharmacodynamics, and safety of sub-cutaneous semaglutide once-weekly in healthymale Japanese and Caucasian subjects. Adv Ther.2018;35:531–44.

19. Kapitza C, Dahl K, Jacobsen JB, Axelsen MB, Flint A.Effects of semaglutide on beta cell function andglycaemic control in participants with type 2 dia-betes: a randomised, double-blind, placebo-con-trolled trial. Diabetologia. 2017;60:1390–9.

20. Demmel V, Sandberg-Schaal A, Jacobsen JB, GolorG, Pettersson J, Flint A. No QTc prolongation withsemaglutide: a thorough QT study in healthy sub-jects. Diabetes Ther. 2018;9:1441–56.

21. Korsatko S, Jensen L, Brunner M, Sach-Friedl S, TarpMD, Holst AG, et al. Effect of once-weeklysemaglutide on the counterregulatory response tohypoglycaemia in subjects with type 2 diabetes: arandomised, placebo-controlled, double-blind,crossover trial. Diabetes Obes Metab.2018;20:2565–73.

22. Blundell J, Finlayson G, Axelsen M, Flint A, Gib-bons C, Kvist T, et al. Effects of once-weeklysemaglutide on appetite, energy intake, control ofeating, food preference and body weight in subjectswith obesity. Diabetes Obes Metab.2017;19:1242–51.

23. Hausner H, Karsbøl JD, Holst AG, Jacobsen JB,Wagner F, Golor G, et al. Effect of semaglutide onthe pharmacokinetics of metformin, warfarin,atorvastatin and digoxin in healthy subjects. ClinPharmacokinet. 2017;56:1391–401.

662 Diabetes Ther (2019) 10:649–662