In vitro pharmacodynamic models to determine the effect of antibacterial drugs

16 Pages·2010·0.46 MB·English

J Antimicrob Chemother 2010; 65: 186–201
doi:10.1093/jac/dkp434 Advance publication 21 December 2009
In vitro pharmacodynamic models to determine the effect
of antibacterial drugs
Julia Gloede1, Christian Scheerans1,2,Hartmut Derendorf3 andCharlotte Kloft1,2*
1Departmentof Clinical Pharmacy, Institute ofPharmacy,Martin-Luther-UniversitaetHalle-Wittenberg,Halle, Germany;
D
2Department of ClinicalPharmacy,Institute of Pharmacy,Freie Universitaet Berlin,Berlin,Germany; o
w
3Departmentof Pharmaceutics, College of Pharmacy, UniversityofFlorida, Gainesville,FL,USA n
lo
a
d
e
*Correspondingauthor.DepartmentofClinicalPharmacy,InstituteofPharmacy,Martin-Luther-UniversitaetHalle-Wittenberg,Wolfgang-Langenbeck- d
Str.4,06120Halle,Germany.Tel:þ49-345-5525190,Fax:þ49-345-5527257;E-mail:[email protected] fro
m
h
In vitro pharmacodynamic (PD) models are used to obtain useful quantitative information on the effect of ttp
s
either single drugs or drug combinations against bacteria. This review provides an overview of in vitro PD ://a
models and their experimental implementation. Models are categorized on the basis of whether the drug ca
d
concentration remains constant orchanges and whether there is a loss of bacteriafrom the system. Further e
m
subdifferentiation is based on whether bacterial loss involves dilution of the medium or is associated with ic
.o
dialysis or diffusion. For comprehension of the underlying principles, experimental settings are simplified and u
p
schematically illustrated, including the simulations of various in vivo routes of administration. The different .c
o
model types are categorized and their (dis)advantages discussed. The application of in vitro models to m
specialorgans,infectionsandpathogensiscomprehensivelypresented.Finally,therelevanceandperspectives /jac
of invitro investigationsin drugdiscoveryand clinical research areelucidatedand discussed. /artic
le
Keywords: invitro models, antibiotics,dilution models, dialysis/diffusion models, static models -a
b
s
tra
c
t/6
5
Introduction infection is clearly defined (cure or death) and comparable to /2
that in humans.4,5A significant disadvantage of animal models /18
6
Thedosingregimensofantibioticsareoftennotoptimalandthe is differences in the PK,1,5 e.g. in metabolism, which limit or /6
dose–response relationships not well known.1 One important necessitate sophisticated scaling methods for transferring data 86
3
reason is that in the patient the pure antibiotic effect, i.e. the from animalsto humans.5 32
pharmacodynamic(PD)characteristics,2cannotclearlybesepar- In contrast, in vitro models can mimic human PK1 and are by
atedfromotherfactorsdeterminingtheresponsetotheantibac- thus better suited for the investigation of antibiotic activity.6 In gu
e
terial treatment. The effect also has to be regarded along with addition, they allow resistance analyses,7,8 determination of s
pharmacokinetic (PK) properties,2 such as the ability of the time–kill behaviour, and the identification and optimization t on
drugtoreachitstarget.Thus,thePK,andthePD,2arecharacter- of PK/PD indices and breakpoints.9–13Although a large number 19
istics of an antibacterial agent and should be considered in the ofmodelshavebeendeveloped,inpracticetheyareallvariants N
o
development and prediction of the efficacy of the antibacterial of 10 basic experimental set-ups. Rather than discuss all the ve
m
therapy. By linking the concentration–time course (at the site modelsreportedintheliterature,thisarticleprovidesageneral- b
e
of action) to the drug effect (PK/PD), various dosing regimens izedoverviewofthemostfrequentlyusedandnewlydeveloped r 2
for different pathogens can be investigated in silico, enabling invitro PDmodels. Thehistorical development ofinvitro models 01
the identification of potentially effective dosing regimens. has been already reviewed by Grasso until 198514 and 8
However, there is no standardized procedure for PK/PD evalu- others.15–17 Grasso divided in vitro models according to their
ation for antibiotics, although the European Medicines Agency working principle into two basic groups: (i) models based on
(EMEA)1 and the FDA3 clearly recommend these investigations dilution; and (ii) those based on diffusion or dialysis. MacGowan
for new compounds. et al.18,19 described the information and conclusions obtained
For characterizing the PD of an antibiotic, bacterial growth frominvitromodels.Theimpactofinvitromodelshasbeendis-
and death under antibiotic exposure have to be investigated. cussed by Li and Zhu,17 and others.16 PK modelling of in vitro
Since these are difficult to measure in human tissue, animal models has beenbasicallydescribedbyBlaser,20RoweandMor-
and in vitro models have been developed. Animal models ozowich,21andFirsovetal.;15detailedmathematicaldescriptions
providesimilargrowing conditions forbacteria,closely imitating for the interpretation of PK/PD analyses and PK/PD modelling
thecharacteristicsofahumaninfection,andtheendpointofan can be found in the work of Derendorf and Meibohm,22 Czock
#TheAuthor2009.PublishedbyOxfordUniversityPressonbehalfoftheBritishSocietyforAntimicrobialChemotherapy.Allrightsreserved.
ForPermissions,pleasee-mail:[email protected]
186
JAC
Review
and Keller,23 and others.24–37 PK/PD parameters for antibiotics, III models with changingdrug concentration and no bacterial
the PK/PD indices, are well defined in the literature from loss.
Mouton et al.10,11
The first models can be more accurately named static in vitro
models(Table1,No.I).Thistermisalreadyadvisedinthelitera-
Fundamentals ture of microbiology and biotechnology for models with a con-
stant environment, i.e. constant antibiotic exposure, with
Characteristics of in vitro models unchanged medium;39,40 thus, no in- and outflow of medium
occurs in these systems.41,42 These models have been used
The two main characteristics of in vitro PD models are drug
extensively;28,43–48however, these models had either not been
exposure and bacterial concentration. The literature does not
named before or have been described as ‘models investigating D
comprise a uniform and complete definition of these main o
constant concentrationsof antibiotics’.42 w
characteristics. Hence, we characterize these terms as follows: n
Theothertwogroupsofmodelsareknownasdynamicinvitro lo
constantdrugexposureisachievedbynotreplacingorchanging a
modelsandarefurtherdifferentiatedonthebasisofwhetheror de
themedium,whilechangingdrugconcentrationsareobtainedin d
systems with flowing medium. Consequently, we will focus on nteortiablalocstes3ri8al(Nloos.sIIo)cicsunrsot(Tinabtelend1e,dNaonsdIIcaanudseIsIIb)i.aUs,suwahlliyc,hbcaacn- from
thetermsconstantandchangingtodescribethedrugexposure. be corrected,40,49,50whereas the dilution of toxic waste cannot h
ThebacterialconcentrationrepresentsthemagnitudeofthePD ttp
be considered. To avoid bacterial loss, appropriate technical s
effect.Alossofbacteriaduetotheexperimentalsetting,which arrangements have to be carried out (No. III), e.g. by a mem- ://a
is observed in some models, may therefore have a substantial c
braneorfiltersystem.38ModelsNosIIandIIIcanbesubclassi- a
influence on the results. Thus, to define the loss of bacteria in de
imnovditerolsmalolodwelsthweeesxucghgaensgtethoef btearcmtesriaopwenithantdhecleonsevidro:3n8moepnetn; fideiludtidoenpe(NnodsinIgIoanndwhIIeItah)eorrthdeiamlyseics/hdaifnfiussmionof(Ndrou.gIIlIobs)s.iDnivluotlvioens mic.o
modelswithbacterialloss(No.II)workbystepwisesubstitution u
and closed models have no bacterial exchange. As a result, an orcontinuousdilution ofmedium.Dilutionmodelswithoutbac- p.c
open model always has a flowing medium (i.e. changing drug om
terialloss(No.IIIa)operatebystepwiseorcontinuousdilution,or
exposure), whereas closed models can have an unchanging or /ja
stepwise substitution of medium through afilter system. In the c
flowingmedium(i.e. constant orchanging drugexposure). /a
special case of dilution models without bacterial loss (No. IIIa), rtic
medium is added, with a resulting increase in volume, i.e. no le
loss of bacteria, although their concentration is reduced due to -a
b
Classification dilution (see the ‘Stepwise simple dilution’ section below and stra
Table 2). Dialysis/diffusion models can be further classified by c
Abceisonatticrcasotcnioasnneqabuneedcnlwcaeshsoeififthethedreasocercdonerodfitinntightieotronest,ihsaebllcaihncatvenitrgiraeolomlofosthdsee(lTdsarfbuolgerac1no)tn:i-- theOtuyrpereovifemwesmhobwrasntehautsethde(eciuthrreernatrctilfiacsisaifilcoartnioant1u4ra(ld).ilutionor t/65/2/1
dialysismodels)doesnotencompassallmodels(e.g.intracellu- 86
I modelswithaconstantdrugconcentrationandnobacterial lar models), although they could still be integrated into the /68
6
loss; classification scheme. We adapted and revised the existing 3
3
II models with changing drug concentration and bacterial classification, and focused on in vitro models, which mimic PK 2
b
loss; and profilesin plasmaand other biological matrices, and thus allow y g
u
e
s
t o
Table1. Revisedclassificationofinvitromodels n
1
9
N
Bacterialloss o
v
e
m
Drugconcentration yes(opensystems) no(closedsystems) b
e
r 2
Constant — I staticmodels 0
1
8
Changing II dynamicdilutionmodels IIIa dynamicdilutionmodels
† viastepwisesubstitution(withoutfilters) † viastepwisesimpledilutiona
† viacontinuoussimpledilutionb(withoutfilters) † viastepwisesubstitution(withfilters)
† viacontinuousdilution
W withoutoutleta
W withfiltersb
IIIb dynamicdialysis/diffusionmodelsb
† withartificialmembranes
† withnaturalmembranes
aWithbacterialdilution.
bMulticompartmentspossible.
187
1 R
8 e
8 v
ie
w
D
o
w
n
lo
a
d
e
d
fro
Table2. Specialapplicationsofinvitromodels m
h
ttp
Applicationandrationale TypeofPDinvitromodel Specialfeatures Ref. s
://a
c
Modelsaccountingfortheimmunesystem a
d
e
realisticcell/fluidenvironment,examinethe staticmodel usingimmunecellsandhumanserum 46,109 m
penetrationintoimmunecells,46and dilutionmodelwithstepwisesubstitution usinghumanserum 133 ic.o
includetheeffectofimmunecells dilutionmodelwithcontinuousdilution usingserum 80,81 up
dialysismodel usingblood 93 .co
m
Biofilmmodels /ja
c
bacteriainbiofilms:propertiesdistinctfrom staticmodel a 47,135–138 /a
singleorganisms,e.g.increased dialysismodel a 100,132,139 rtic
le
antimicrobialresistance134 -a
b
s
Intracellularmodels tra
PKandPDofintracellularbacteriadifferent staticmodel b 140,141 ct/6
fromextracellular dialysismodel b,specialinterest:Mycobacterium105,141 103–108,141 5/2
/1
Siteofinfections/diseases 8
6
differentPKprofilesorbacterialgrowing /6
8
conditions41 63
3
bladder/bacterialcystitis dilutionmodelwithcontinuousdilutionwithoutoutlet usefultodescribeconditionsinemptying 77,142–145 2
b
compartments y
g
otitismedia dialysismodel disease-specificinfectiousbacteriaused 92,146 u
e
endocarditis dialysismodel 131 st o
chronicpneumonia staticmodel 140 n
1
infectedfibrinclots dialysismodel bacteriasuspendedinafibrinclot 131,147,148 9
N
dentalinfections staticmodel bacteriaformbiofilmonartificialteeth 149 o
v
dialysismodel 149–151 em
skininfections staticmodel 152 be
tuberculosis dialysismodel 105,141 r 2
0
1
Multicompartmentalmodels 8
simulatePKofdrugswithnumerous dilutionmodelwithcontinuousdilutionwithoutandwith seealso‘Modeldevelopments’andFigure2(candf) 21,40,52,84
theoreticalcompartments,no.of filters
compartmentsdeterminesno.of(culture)
vessels
JAC
Review
48,1537315420,155–15751,117,158–160 161,16212116372164162,165 nfectionispositive.totheshortesthalf- idcnttuniroeesoavenwpnetdideeso,intdnetmidigxomeaotnaentdasnintoeiddnlnasbespcadcbhpcaaoceatncnelaratarrbesicatsrethilsiaefieecglrcapirasatroctteeriwoicagolansttonthcerioeo(l.izysfnneeSit/dnpeirsna.eTvtcoaiiTitofifbarnoclbe,clmoeo1wnr)ogch2idaseeennbr(lest‘saArb,aspayiet.pneidold.aincodla-lntrhtuicabeogooninrmtdcsho’s)mpn.tmiecomTceenhinnealey---l
id
ee
enthdjust Experimental settings of in vitro models
ha
we D
withfilterswithoutfilters withoutfilterswithfilter withartificialmembraneswithnaturalmembranes 135onasaeruginosaonvarioussurfaces.cultureisincubatedandinvestigationsstartmedium,whichdecreasesthedrug,hastob PaiAhCvaspeetneeiKfnalraontglferlor/trseemioPatiwanccamtDlamtomtnnv)cepfettiiofodooniiptinubnbrrndldntoliiiaofidccsoteichs.aemltweltespiiensbtccTsooeaptlf,iehoerecdirnelasekxtxreecstuvppidfilaianoiocoobtwanlarhrgnssantorinuuti–ecdbaapharsrtqtieselyerloicluimy.sitsrto1ansisitirp8abnceireecs,muel1sydmmic9aropihdap,cot3feliaioiiuouo3esnfnbvnf,snrn3steamseecc4dssrecos,e,ee4bo(ts,mn1Tnreeen,at(5.treatipgtrit1bhaainnmoo.mletlednridienegoeet3etfdrn–insheom)vo.ktcooew—iNm(tevldlaoliteou–shtsepcrnmhku(eaitsTeltrdceialnvhm/ia.crbebndeouNneslauedls)octauspeutm4tanneeadrn)aetdc.ritimdilifhiTbealfifvehramldiecfcreocdesceoctslurinblefooideorvftarnmevlewei—caielrldnys----,l ownloaded from https://academic.oup.com/ja
cModelsforcombinationtherapysynergisticeffectsstaticmodeldilutionmodelwithstepwisesubstitution(withfilters)dilutionmodelwithcontinuousdilution dialysismodelModelsforfungiandanaerobicorganismsstaticmodeldilutionmodelwithcontinuousdilution dilutionmodelwithstepwisesimpledilutiondialysismodel a47Productionofbiofilmsbygrowingslime-producingbacteria,e.g.StaphylococcusaureusandPseudombHostcellsgrownuntilastablecultureappears(continuouslayer),bacterialsuspensiondirectlyadded,cIfantibioticswithdifferenthalf-livesaresimultaneouslyinvestigatedindynamicmodels,theflowofthe41lifeor,moreappropriately,thedrugwiththelongerhalf-lifehastobesubstitutedtothereservoir. mstmttStt(mcdtIntobTsfwoaiccatmamsiruhrhhinhetuaeoihiebrdpnienoom)alurdooeootieeevnbusneospmcsumsilisAddaaieddddleolesttsepuetaoeitcoirnemmrt.isaeeieebinrceelausdansuupievau,dodtallllta.oldtrTmasss.tiseliyauiuts.tstrstccmiarTlhrtdbflTodhtuetmutyTtibeethohnsiireiahsihinoiendeoghreoo,hyinlsauoefsedoeupfeewmrrneaenvatmtrbicdrcatvbgiaerensilmpavoidsrvhoisedoinifvopyunrreltr.fisweu-ottncratonsarrfrpneelimcpeusheoucpuduloiAcnidtinuufrkttrpsturgoturpsgoleneihlgmooiouhetcmtmeioawletiflninmlsn.uonioovudcteldlnnoioc.oceorctioecrenonrtdmfmtTse.itgreefoefarwoocEfeotisdngeahne.Tiinnamonmn4pfisetnnoed,,cltaevstothmh1tctmslndhcnliehphefhrasreio,r.sreerets4rueecatieon,sreeoeexn,htovtpd2oaon.gtsrilottxe.dmsbebdstdemoaliiccrcaehmtueohpsxaTpiaa(rancrruuhelTtaeeidummpnaeehtaana)ccotaspslmihmaddwigliretetttctnnosaamoiepuvsiedeicidmvmeidoerubitcneinrraiidctehrrrenriefa,detsneeeimnheiniiarmaoreeeammaaverdiesowfidnaigntnvnlmtpinlod(eviacnnmiteeisdilaaeioutrhecngvur.orogctahegrmeebaneilvbairmheseeanhtsrnrvsndomk.rolfittetsoiarr.esotaaiosti,4dihfuflametinlbniwoinhirtotloao2nnnnoimvecullloiseayhgndedswnrwniegtio-buFpmnwccdcnmesshpcarilyalheosbe.sniu(tetatiulmacbde(spdiornveeu.emhwgsmsm(FnullicnaodoriomdoprdtarheiuaodoiTlesigrduenpleldnrpwhtteemueluluhnfumfehurataydrcvllevrasemdiessoetsunrsorettbroeiffr,epoplepeetoehtruirtgaticusimsvaeeszmorioxoledadeedtnasloe)et1esetsnimfltnc,itiesvcihnfevsouelsh)diaticttoauuoedcwlo.esshlednfhm4olnhnonmtssweuayenrma1ceritteeiimosimtabcftsdnoullrh,ohn)rhxliieeIeteitopebsineogesnawapnphdneltth.napddidd.ftahnpenaoodepioithIeroydinmtyyiwxbucedigrutsriSunenimnheunrnnitnatiiaoumeeclumcstimooiutmlfaattnfibllrstonccclbbvireiiditvaemp)mcm.otearooootegitasiavietesnoaehbbinsvlnnnttIntoerrciieltriilsihnndndhnaoooeeeeyccss--------tr.. c/article-abstract/65/2/186/686332 by guest on 19 November 2018
189
Review
Table3. Requirementsforinvitromodels
Parameter Rationale Implementation
Growthmedium appropriategrowingconditions
Temperature choiceandcontroltomimicinvivogrowingcondition waterbath,incubator
Mixing quickhomogeneousdistributionandaerationofbacterialsuspension shakingorstirring
are separated by a semi-permeable membrane, i.e. permeable et al.58 demonstrated that bacteria might adhere to the vessel Do
w
fordrugandmediumbutnotforbacteria.Freshmediumiscon- wall, forming a biofilm, which protects the bacteria from n
tinuouslypumpedfromareservoirintothecentralcompartment (out)flow and from antibiotics. Hence, only released bacteria loa
d
and then into the waste. Thus, the medium in the peripheral can be counted in the medium and higher bacterial concen- e
d
compartment is continuously renewed by diffusion (from the trations can be found. Nevertheless, due to its simplicity the fro
centralcompartment),whilethedrugandbacteriacaninteract, Grasso model was adapted by several groups.59–68In a similar m
but the bacteria cannot leave this compartment (Figure 1). The model, air pressure instead of peristaltic pumps has been http
circulationofmediumintheperipheralcompartment—ascoun- used.69 Bergan et al.70 introduced a second peristaltic pump s
terflowtowardsthecentralcompartment—canhelptooptimize for the out-flowing medium. Later, a computer was added to ://a
c
the diffusion.52,53 controlthree pumpsets inparallel.71 ad
Murakawa et al.40 developed a two-compartment model em
Model developments bdarusgedisoandmthienisGteraressdoams aodbeollu(Fsiginutroet2hce).fiArsttt(hceenbtreagli)ncnoinmgp,athrte- ic.oup
Static models (No. I) mentcontainingthebacteria,withthesecond(peripheral)com- .com
partment remaining drug free. Fresh medium is pumped from /ja
Staticmodelsconsistofaclosedculturevessel(Figure2a).These the reservoir into the first compartment. A second pump c
vfleassskesl,s43a,5r4ecaevllaciluabltluereinflaasvkasr4i4etoyrosfpsinhnaeprefls,assuksc,h55aasntdubmesa,y46b,4e7 eexxcchhaannggeesisthneotmperdeviuemntebdetawnedenbabcotethriacoamrepealritmmineanttes.dBinatcotetrhiael /article
made of glass43 or polystyrene.47 The first time–kill investi- waste;mathematical corrections havebeen applied. -ab
gations in static models were established by Garrett et al.43 stra
in1966. c
t/6
5
Dynamicdilutionmodelswithoutbacterialloss(No.IIIa) /2
/1
Dynamic dilution models with bacterial loss (No. II) 8
Stepwise simple dilution 6
/6
Stepwise substitution (without filters) 8
Inastepwisesimpledilutionmodelthemediumisnotremoved 63
Nishidaetal.56describedadilutionmodelwithstepwisesubsti- from the system. Fresh medium is added periodically and the 32
tutionofthemedium(Figure2b).Inthismodel,atubecontains drugconcentrationdeclinesovertime,inrelationtotheincrease by
thebacteriainmedium.Freshmediumisperiodicallyaddedand inthevolumeofthemedium(Figure2d).72Simultaneously,bac- gu
e
atthesametimethesamevolumeofusedmediumisdiscarded, teria will be diluted; hence, bacterial concentrations have to be s
leading to a stepwise decline of the drug and a removal of the corrected. t on
bacteria. 19
N
o
v
e
Continuous simple dilution (without filters) Stepwise substitution (with filters) m
b
e
Models with continuous simple dilution reflect the in vivo con- Nolting et al.45 developed a model (‘syringe model’) where the r 2
0
ditions of a drug much more closely than a stepwise decline of drug concentration is decreased by stepwise substitution, but 1
8
the drug. The decisive improvement in this field was made by the bacterial loss is prevented by a filter. A syringe needle is
Grasso et al.57 The Grasso model consists of a flask containing stuck into a cell culture flask containing bacteria and medium
the bacteria (culture vessel), a reservoir and a waste container (Figure 2e). The needle is connected with a filter unit and a
(Figure 2b). Fresh medium is continuously pumped from the syringe. Used medium is withdrawn at regular intervals from
reservoir into the flask and used medium leaves the culture thecellcultureflask(in contrasttothe stepwisesimple dilution
vessel by the pressure of the incoming medium. Drug and bac- model)and replacedbyfresh medium.45,73
terial samples can be taken from the vessel. A magnetic stirrer Another stepwise substitution model was introduced by
ensures a homogeneous distribution of the drug and bacteria. Haller,65 and comprises a Teflon-coated ultrafiltration unit filled
In the Grasso model, the bacteria are diluted by the incoming with medium and bacteria. After adding the drug, air pressure
medium and flow out with the outgoing medium, which is applied. Thus, medium is continuously eluted and discarded.
demands a mathematical correction for bacterial counts. In Fresh medium, however, is replaced at intervals (Figure 2e).65,74
theory, flow rates that are faster than the bacterial growth The elution can also be performed by centrifugation of the fil-
rateswould lead to acomplete loss of bacteria; however, Haag tration unit.75
190
R
e
v
ie
w
D
o
w
n
Table4. Quantificationmethodsforbacteria lo
a
d
e
Properties d
fro
m
differentiation h
online direct detectionoflive betweenlive ttp
Method Principle measurementa measurementb cellsonly anddeadcells Comment Ref. s://a
c
a
Viablecellcounts incubationofbacterialsamplesonagar, 2 2 þ NA mostfrequentlyused, 43,56,57,97 d
e
followedbycounting avoidanceofantibiotic m
ic
carry-overeffect .o
u
necessary123–125,c p
.c
Turbidimetry measurementofopticaldensityof þ þ 2 2 discussedasnotreliable79 76,77,84 o
m
bacteriainmedium(correlateswith /ja
c
bacterialconcentration) /a
Impedance measurementofimpedanceofbacterial 2 2 2 þ 43 rtic
cells(correlateswithbacterial le-a
concentration) bs
Bioluminescence determinationofATPcontentof 2 2 2 2 126 tra
c
bacterialcellsreleased(correlates t/6
5
withbacterialconcentration) /2
Microscope determinationofbacteriabyaphase þ 2 2 þ hastobecarriedoutwithin 126 /18
6
contrastmicroscopy onegenerationtimeof /6
8
bacteriatoavoiddoubling 6
3
Fluorescence determinationofreleaseoffluorogenic þ 2 þ NA 127 3
2
quantification substancesfromasubstrateby b
y
bacterialphosphatases gu
e
RNAprofiling quantitativePCRofRNA(correlateswith 2 2 þ NA 128,129 s
bacterialconcentration) t on
1
9
NA,notapplicable. N
o
aWithin10min. ve
bIntheculturevessel. mb
cSpecialcautionforcarry-overeffectshouldbepaidforquinolones. er 2
0
1
8
J
A
1 C
9
1
Review
(Figure 2f). Several vessels are connected in series and the
Dynamic models number of vessels depends on the number of compartments
of the underlying mathematical model. The model
operates as the Grasso model. Bacterial loss is prevented by
Working principle a special filtration unit, which is placed between each vessel.
Dynamic dialysis/diffusion models (No. IIIb)
In the majority of dialysis/diffusion models (for simplification,
Dialysis/
Dilution models further called ‘dialysis models’), the setting is as follows: fresh
diffusion models
medium is pumped from a reservoir into the central compart- D
o
ment and then into the waste (except for the model by w
n
Al-Asadi et al.85), thereby decreasing the drug concentration. lo
Prevention of bacterial loss The hallmark of these models is that the drug (and medium) ade
d
has to diffuse through a membrane to reach the bacteria in fro
Yes No Yes the peripheral compartment. In consequence, bacterial growth m
and fresh medium flow happen in two compartments. Two h
different settings of the central and peripheral compartment ttps
Direct adding or Adding and Membrane material havebeen employed: adjacent and embedded. ://ac
direct removing removing Dialysis models can be subclassified by the nature of their ad
of medium of medium membrane, i.e. models with artificial and natural membranes em
(Table 5). In this way, the classification presented here includes ic
.o
models that previously have not been named dialysis models. u
p
dSiilmutpiolen Substitution Abratirfircieiarl Nbaatrurirearl Aprtinficrisptlegloafnctheetsheismmoidgehltssiseedmiffuusniounsuaacl,robsustasinmceemthberawneo,rktihnigs .com
classification seems meaningful. /jac
Figure1. Detailedoverviewondynamicinvitromodels. /a
rtic
Artificial membranes le-a
CCoonnttiinnuuoouussddiluiltuiotinonmodelsoperatingwithoutoutlet,resultingin Ivnesdseialslywsisermeuosdeedls,swucithhaasrttiufibceiasl,8m5(esmqubararen)evse,sdseiflfse,r8e6nstepbaarcatteirniag bstrac
increasingvolumes,weredescribedbySanfilippoandMorvillo,76 funnels,87,88 artificial kidneys,89–91 a plexiglass chamber with t/65
andO’GradyandPennington.77Theymostlyreflectonlyselected changeable membrane filters,92,93 a hollow t-tube94,95 and /2/1
aspects of the in vivo situation. Pumps transport the medium hollow fibres.96 These models can have adjacent peripheral 86
from a reservoir into the culture flask (Figure 2d). Since there is and central compartments or compartments embedded in /6
8
6
nooutlet,thevolumeofthesecondflaskcontinuouslyincreases, eachother. 3
3
the drug concentration changes and the bacteria will be In the adjacent setting, as in the model of Drugeon et al.,87 2
diluted.76 The increasing volume does not necessarily allow the upper and lower part of a separating funnel are part of an by
g
exact exponential decline of the drug concentrations (see also entire loop (peripheral compartment), which also runs through u
e
‘bladder/bacterial cystitis’ inTable 2).77 one part of the dialysis unit. The other part of the dialysis unit st o
Continuousdilutionwithoutbacteriallosscanalsobeachieved forms the central compartment; the dialysis unit enables the n
withfilters.ThemodelbyGreenwoodandTupper78consistsofa exchange of the drug and the medium. Continuous dilution of 19
N
vessel separated by a filter membrane in two chambers, with the central compartment decreases the drug concentration o
the bacteria and drug being added to the upper one (Figure 2g).87,88 Toothaker et al.86 horizontally separated a ve
m
(Figure 2e).78 However, this model has not been further used. vesselbyahaemodialysismembraneintwoparts.Onepartcon- b
e
Instead, for the already mentioned decisive Grasso model tains fresh medium and the antibiotic, and the other part r 2
(dilution,butbacterialloss),57differentmodificationshavebeen includesthe bacteria(Figure2g). 01
8
suggestedtoimprovetheaccuracy.Filtersareinsertedbetween Intheembeddedsetting,Guggenbichleretal.89workwithan
the culture vessel and the waste, and the outlet has moved artificialkidney,theinnerpartofwhichisconnectedtothebac-
fromthesidepaneltothebottomoftheflask.79–81Apracticalsol- teria vessel (Figure 2h). Shah utilizes a plexiglass chamber with
utiontopreventthebacteriallossintheGrassomodelwasfound changeablemembranefiltersatbothendsasthebacteriacom-
byLowdinetal.55(Figure2e).Thebaseofaspinnerflaskismodi- partment(peripheralcompartment).93Thischamberisplacedin
fied,includinganoutletandaperforatedmetalsupport,onwhich an outer chamber with medium (central compartment;
afiltermembraneandapre-filterareadjusted.Abovethemem- Figure 2h). The same model was adapted by Garrison et al.,94
brane,amagneticstirrerisplacedtopreventmembraneblockage. who modified the inner chamber to a hollow t-tube. In the
Freshmediumispumpedinviaonesidearminthespinnerflask. ‘hollow fibre model’ by Zinner et al.,96 the tubing of the central
Theotherarmispreparedwithasiliconmembraneforrepeated compartment includes a bundle of artificial capillaries. These
sampling.55,82,83 capillaries consist of polysulphone fibres, permeable for drug
A multicompartment model based on the Grasso model and medium, and are continuously flushed with medium. The
with retention of bacteria was presented by Navashin et al.84 fibres pass through a vessel with bacteria (peripheral
192
JAC
Review
compartment; Figure 2i). The hollow fibre model was further are ‘captured’ behind a barrier and the drug has to pass the
developed by Blaser et al.,52,97who added a second and more barrier to reach the bacteria. Haller102 described a ‘tissue
bacteria vessels.20,52,97–101 Al-Asadi et al.85 use two tubes culturemodel’,wheretissuecellsaregrownonadialysisultrafil-
clamped together, separated bya membrane, with drug in one teruntilacontinuouslayerisformed.Themembrane,consisting
(centralcompartment)andbacteriaintheothertube(peripheral of the filter and the tissue cell layer, is placed on a cylinder
compartment).Afterafinitetimeofdrugdiffusionfromthedrug (central compartment). Another cylinder located above serves
to the bacteriatube, fresh mediumis pumpedinto the bacteria as the peripheral compartment with bacteria. The antibiotic is
tubeandleadstoadecreaseofthedrugconcentration.Incon- administered in the lower part of the chamber by syringes and
trarytoallotherdialysismodels,herethebacteriacompartment diffuses through the cells to the upper part (Figure 2g). The
itself is flushed(Figure2j). model was suggested to investigate the penetration of the
drugthroughintercellularspacesandwaslaterusedbynumer- D
o
ousgroupstoinvestigatedrug–bacteriaeffects.103–107Anintra- w
Natural membranes cellularmodelimplementingtissueculturesinPDinvitromodels nlo
a
The principle of dialysis models with natural membranes is is presented by Hulten et al.103 Tissue cells are grown in inserts de
d
almostthesameasforthosewithartificialmembranes.Bacteria in a glass chamber, similar to a closed Petri dish (central fro
m
h
ttp
s
(a) Static Model (No. I) ://a
c
a
d
e
B m
ic
.o
u
p
Dynamic dilution models (No. II) .co
m
(b) stepwise substitution (1.) (c) continuous simple dilution, /ja
continuous simple dilution (2.) multi compartments c/a
1. rtic
2. or le-a
R B W b
s
tra
c
t/6
5
R B W /2
B /1
8
6
/6
8
Dynamic dilution models (No. IIIa) 6
3
(d) stepwise simple dilution (1.) 32
continuous dilution without outlet (2.) b
y
g
u
1. e
2. or st o
n
1
9
N
o
v
R B e
m
b
e
r 2
(e) stepwise substitution with filters, e.g. (f) continuous dilution with filters, 01
8
models by Nolting (1.+3.) or Haller multi compartments

(1.+4.)

continuous dilution with filters, model
by Lowdin (2.+4.)

1. 3.
or or R B W
2. 4.
R B W
Figure2. Schematicdepictionofsettingsofinvitromodelsatthebeginningofanexperiment.
193
Review
Dynamic dialysis/diffusion models (No. IIIb)
(g) Models with adjacent peripheral and central (h) Models with embedded peripheral
compartments with compartments in central compartments with
artificial membranes, models by e.g. artificial membranes, e.g.
Drugeon, Toothaker Guggenbichler, Shah
natural membranes, e.g. tissue culture natural membranes, e.g. intracellular
model by Haller model by Hulten, fibrin clot model
by McGrath
B Central
Peripheral
D
o
R Central W R B W w
Peripheral nlo
a
d
e
d
fro
(i) Hollow fibre model with artificial membrane (j) Special case: m
Model by Al Asadi with artificial membrane h
ttp
BPeripheral s://a
c
R Central W R B W ad
Peripheral Central em
ic
.o
u
Caption: p.c
o
m
B culture vessel with bacteria /ja
R reservoir c
/a
W flwoaws tdei rection rticle
stepwise medium flow -a
continuous medium flow bs
filter tra
c
semi-permeable membrane, i.e. permeable for drug and medium, not for bacteria t/6
5
/2
/1
Figure2. Continued 8
6
/6
8
6
3
3
2
Table5. Typesofmembranesindialysismodels(alphabeticalorder) Experimental implementation of in vivo routes by
g
of administration u
e
Artificialmembranes Naturalmembranes Invitromodelscanbeusedtosimulatedifferentroutesofdrug st on
material ref. material ref. administrationinpatients.Generally,indilutionmodelsthedrug 1
9
canbeaddeddirectlytotheculturevesselorintoanadditional N
o
celluloseacetate 85 agarosegel 130 vessel between the reservoir and the culture vessel, simulating ve
haemodialysismembranes 86 cells 102 no(i.e.bolusadministration) orfirst-orderabsorption(i.e.extra- mb
ppoollyyscualrpbhoonnaete 5926,94,97 cfieblrlinmembranes 102–110351 vthaescdurluagraisdmtrainnissptroarttieodn)w,riethsptehcetivmeelyd.iFurmomintthoetahdedcituioltnuarelvveesssseel,l er 20
1
regeneratedcellulose 89 slime 47,132 and into the waste. Zero-order absorption (i.e. infusion) of the 8
syntheticregeneratedcellulose 95 drugcanbeachievedbyaddingthedrugtothereservoir.Drug-
ester containingmediumistransportedtotheculturevesselandfrom
there into the waste. The end of absorption in this case can be
realizedbyexchangeofthedrug-containingreservoirtoadrug-
compartment). The cells had previously been infected with free reservoir (Figure3a).
intracellular-growing bacteria (peripheral compartment). A Simulation of in vivo routes of drug administration in dialysis
metal rack for permeable cell culture inserts facilitates the modelsisthesameasindilutionmodels.Thedrugistransported
tissue cell growth in the glass chamber. The cell membranes withtheflowingmediumtothecentralcompartment(Figure3b).
operate as dialysis membranes and the cells are continuously Fromthereitdiffusestotheperipheralcompartment.Inallscen-
flushed with fresh medium (Figure 2h). The drug has to pass arios the drug concentrations are suggested to follow in vivo
the cell membrane to reach the bacteria. The bacteria can be absorption/PK.21,57,62 Determinations of drug concentrations in
counted afterdestruction of the cells.103–109 samplesfromtheculturevesselshouldsupportthisassumption.
194
JAC
Review
(a) In dilution models with the example of a continuous simple dilution model
Infusion Extravascular Bolus administration
administration
0-order absorption 1st-order absorption No absorption
D D D
A
D
R or B W ow
n
lo
a
d
e
d
fro
m
(b) In dialysis/diffusion models with the example of an embedded peripheral h
compartment ttps
://a
c
Infusion Extravascular Bolus administration a
d
administration e
m
0-order absorption 1st-order absorption No absorption ic
.o
u
p
D D D .c
o
m
Central /ja
c
/a
A rtic
le
R or B W -ab
Peripheral s
tra
c
t/6
5
/2
/1
8
6
D drug /6
8
R reservoir 63
A additional vessel, mimicking absorption (optional) 32
B culture vessel with bacteria b
y
W waste g
u
e
Figure3. Schematicdepictionofinvitroimplementationofthedifferentinvivoroutesofadministration. st o
n
1
9
Applications todifferentconditions,andarelesscost-andresource-intensive. No
v
Additionally, the relatively high inocula and volumes in in vitro e
m
AsubstantialnumberofinvitroPDmodelshavebeendeveloped modelsallowbetterstudiesofresistance,becauseofthehigher b
e
to simulate specific conditions. Even if not all of these models mutationfrequencythaninanimals.110ThePKpropertiesofthe r 2
can imitate the designated PK profiles, theyare useful tools for drug of interest can be applied in vitro and the time course of 01
specific conditions. In Table 2, the models are grouped by their an antimicrobial agent can be monitored exactly. On the 8
main aspects andmayappear indifferentcategories. other hand, in vitro models need special conditions, such as
atemperature-controlledenvironment,andtheriskofcontami-
nation oftheculturevessel withexternalbacteriaincreasesthe
Relevance and perspectives
longer the experiment lasts.111 Since in vitro models cannot
Fortheapprovalandrationaluseofantibioticsinpharmacother- mimic all in vivo conditions,112 such as immunological factors
apy, pre-clinical investigationswill have to focusmore on PK/PD (e.g. host defence mechanisms), the pathologyof the infection,
investigations in the future. In this respect, in vitro models and the virulence and metabolic behaviour of a pathogen,1 the
mightpresentavaluable predictivetool.1Astandardized meth- derived PD parameters cannot directly be transferred to the
odologyforuseinpre-clinicalresearchwouldprovideavaluable invivosituation.Theinvivogrowthenvironmentisdifferentfrom
toolfor the optimizationof dosing strategies. the in vitro one. This may lead to phenotypic differences
Generally,invitromodelshaveseveraladvantagescompared betweenbacteriagrowninvitroandinvivo.113Ingeneral,invitro
withinvivoanimalstudies:theyaremoreflexibleandadaptable bacterialgrowthismuchfasterthanthatinvivo.38,114,115Hence,
195