Antifungal Susceptibility Testing

15 Pages·2006·3.14 MB·English

CLINICALMICROBIOLOGY REVIEWS, Oct. 1993, p. 367-381 Vol. 6, No. 4
0893-8512V93/040367-15$02.00/0
Copyright X 1993, American Society for Microbiology
Antifungal Susceptibility Testing
JOHN H. REX,'* MICHAELA. PFALLER,2 MICHAEL G. RINALDI,3
ANAMARIE POLAK,4 AND JOHN N. GALGIANIs
CenterforInfectious Diseases, University ofTexasMedical SchoolatHouston, Houston, Texas 770301;
DepartmentofPathology 471, Oregon Health Sciences University, Portland, Oregon 97201-20982;
DepartmentofPathology, University ofTexasHealth Science Center, Laboratory Service (113),
AudieL. MurphyMemorial Veterans'Hospital, SanAntonio, Texas 78284-77503;
F. Hoffman-La Roche, Ltd., CH-4002Basel, Switzerland4; andMedicalService (111),
VA Medical Center, andDepartmentofMedicine,
University ofArizona, Tucson, Arizona 857235
INTRODUCTION 367
.............................................................
EARLY STUDIES 367
.............................................................
FACTORS THAT INFLUENCE ANTIFUNGAL SUSCEPTIBILITY TESTING.................................369
Endpoint Definition 369
.............................................................
Inoculum Size 370
.............................................................
Inoculum Preparation............................................................. 370
Incubation Time and Temperature............................................................. 370
Media 370
.............................................................
Results Obtained with the Proposed NCCLS Reference Method...................................................371
OTHER ORGANISMS 372
.............................................................
CORRELATION OF IN VITRO TESTING WITH IN VIVO OUTCOME........................................373
Animal Studies.............................................................. 373
Clinical Demonstrations ofBoth Inherent and Acquired Resistance...............................................375
Clinical Trials 376
.............................................................
SUMMARY 376
.............................................................
Implications for the Clinical Laboratory............................................................. 377
REFERENCES 377
.............................................................
INTRODUCTION dardization, the various methods have produced widely
discrepant results because of substantial dependence on
The frequency of serious fungal infections is rising, and suchfactorsaspH,inoculumsize, liquidversussolidmedia,
this trend has been attributed to such factors as the increas- medium formulation, time ofincubation, andtemperatureof
ing use of cytotoxic and immunosuppressive drugs to treat incubation. Evenwhenagroupoflaboratoriesagreesupona
both malignant and nonmalignant diseases, the increasing general technique, widely discrepant results may be ob-
prevalence of infection due to human immunodeficiency tainedforsome antifungal agentsifallprocedural details are
virus type 1, and the widespread use of newer and more not precisely defined (19). Finally, once a technique is
powerful antibacterial agents (5, 7). Fortunately, this in- selected and the MIC for an isolate is determined, there are
crease in fungal infections has been accompanied by the fewuseful data available to guide MIC interpretation. Many
development ofnew, less toxic, systemically active alterna- authors have sought in vivo-in vitro correlations in animal
tives to amphotericin B such as fluconazole, itraconazole, models or with isolates from patients: some have found a
and the various amphotericin B lipid formulations (129). correlation (3, 14, 39, 72, 85, 93, 104, 105, 119, 120, 122, 127,
With this proliferation of antifungal agents, therapeutic 128, 136, 139, 149), but others have not (20, 47, 55, 73, 86,
options are more numerous, and the clinician must now 127, 134). This situation is obviously unsatisfactory, and a
choose amongthem. This decision is mademore difficultby substantial amount ofwork has been undertaken to resolve
the steady stream of reports of putative drug resistance to these difficulties. While acomplete solution is not available,
one or more antifungal agents (6, 13, 16, 26, 32, 33, 38, 39, considerable progress has been made, and it is the purpose
46, 47, 62, 69, 70, 72, 75, 89, 93, 103, 105, 119, 120, 126, 136, ofthis review to bring this information into perspective.
138, 146, 150-152). Asthesereportsarenotalwaysconvinc-
ing or well controlled, making an informed antifungal ther-
EARLY STUDIES
apy choice is not easy.
Aswithbacterial infections, the clinicianwould like tobe
guided by knowledge of local epidemiological patterns de- Antifungal susceptibility testingwas notrelevant until the
rivedfromdrugsusceptibilitytestingresults. Unlikeantibac- introduction in the 1950s of amphotericin B. That this was
terial susceptibilitytesting, however, reliable antifungal sus- nearly 30 years after the discovery of the first antibacterial
ceptibility testing is still largely in its infancy. A prodigious agents explains in part the immaturity of antifungal suscep-
tibility testing today. Even then, antifungal susceptibility
array of techniques has been described, but without stan-
testing lay fallow for many years because, while not all
fungal infections responded to amphotericin B, there were
no alternatives. It was onlywith the development of5-fluo-
* Corresponding author. rocytosine and, more recently, the azole antifungal agents
367
368 REX ET AL. CLIN. MICROBIOL. REV.
that differences within and between species started to be- make this approach practical has not been developed fur-
comeapparent. Evenso, onlyasmallnumberoflaboratories ther.
routinely performed antifungal susceptibility testing. In Agar-based techniques have been used extensively by a
1986, Calhoun et al. (19) surveyed 350 hospital laboratories. few laboratories because they are simple, economical, and
Only 41 of the 210 respondents were performing antifungal easy to perform simultaneously on large numbers of organ-
susceptibility tests, with 45% of these laboratories testing isms. Either the fungus can be placed in the agar and the
only one to five isolates during the previous year. Avariety antifungal agent can be placed on the surface of the agar
of methods was being used: most commonly, a broth tech- (disk or well diffusion) or vice versa (agar dilution). These
nique that had been derived from an antibacterial suscepti- techniques sufferfrom substantial dependence on inoculum,
bility testing method. When seven laboratories used this temperature, and duration ofincubation (15, 66, 67, 75, 94).
common, published methodology to test five Candida albi- The estimation ofthe size ofinhibitory zones canbe difficult
cans isolates against amphotericin B and 5-fluorocytosine, because of partial growth inhibition (37, 67, 139) or the
results thatvaried by as much as 512-foldwere obtained. Of presence of"persistor" [sic] colonieswithin otherwise clear
significant interest, however, was the observation thatwhile zones of inhibition (145). Finally, the absolute MICs of the
the absolute values of the MICs for each isolate varied azoles tend to be lower than those produced bybroth assays
greatly among laboratories, the relative susceptibilities of (15, 94). Some of these problems might be overcome by
the isolates varied little among laboratories. In other careful standardization (59), but the physical-chemical prop-
words, the rank order of the isolates was quite constant erties ofthe antifungal agents to be tested and their interac-
despite wide variability in endpoint results. In a follow-up tionwith the agar cannotbe altered. Agarisnot achemically
study conducted by three laboratories (51), another set of defined material, and, as has been noted in assays of the
Candida isolates was tested against amphotericin B, 5-fluo- aminocyclitol antibiotics, agars from different sources can
rocytosine, and ketoconazole. In this study, however, no produce differentresultsbecauseofvariations insuch simple
methodology was prescribed. Rather, each group used the properties as cation concentration (66). Such variations are
method with which theywere most familiar, and the imple- also potentially relevant to antifungal assays (94). In addi-
mented techniques included broth dilution, agar dilution, tion, some (61, 68), but not all (130), authors havefound that
and turbidimetric techniques. Again, while the actual MICs amphotericin B and the azoles tend to deteriorate when
varied by as much as 50,000-fold, the rank order of the stored indilute formordriedon antibiotic assaydisks. Some
isolates produced bythe results from each laboratoryvaried of the azoles, especially miconazole, ketoconazole, and
only slightly. itraconazole, are relatively insoluble and may diffuse poorly
Given these data and the growing interest in antifungal in an agar diffusion system (61, 121, 129). Finally, some
susceptibility testing, it was clear that standard, reproduc- species ofCandida otherthanC. albicans areinhibitedwhen
ible techniques for antifungal susceptibility testing were grown in agar (97). Despite these problems, it has been
needed. Manyapproachesto antifungal susceptibility testing possible to develop routine testing methods that use an agar
have been reported, but not all readily lend themselves to diffusion format (133). Reproducible results that correlate
standardization. Tests that measure the rate ofelongation of well with broth dilution MICs have been obtained (37, 115,
germ tubes (75, 136, 151) are tedious to perform and are 130), and standard techniques forperforming this assay have
useful onlyforC. albicans, the one species in the genus that been described (111, 137). In addition, many reports of
forms germ tubeswithin 3 h in serum. An automated system amphotericin B resistance have used agar dilution methods
based on the dynamic growth ofa single hypha as measured (32, 33, 62, 87, 89, 103). Whether currently proposed broth
by microscope, camera, television monitor, video tape re- methods can detect this resistance needs to be determined
corder, and microcomputer was recently described (102, (see reference 50 and below).
154), butit appears impractical for routine use. Examination Brothdilution methods forantifungal susceptibility testing
of even more subtle morphologic effects is possible and were the most commonly used techniques in the United
would be applicable to awider range offungi (121, 136), but States in the previously mentioned survey (19). Because of
such approaches are highly subjective. Measurements of this and the other factors just discussed, broth dilution
uptake ofradiolabeled metabolites (30, 31, 35, 56) or reduc- methods have been the focus of the most recent efforts at
tion ofcolored substrates (79, 124) have been used to assess test standardization. A number of collaborative studies of
phagocyte-mediated damage to C. albicans, Aspergillusflu- broth dilution methods have now been carried out both
migatus,Rhizopusarrhizus (=R. oryzae), and Coccidioides independently (43, 63, 111, 135, 144) and under the aegis of
immitis and could readilybe adapted to measuring the effect the Subcommittee on Antifungal Susceptibility Testing of
of antifungal drugs on fungi (88, 140, 143). However, these the National Committee for Clinical Laboratory Standards
tests are indirect since lack ofmetabolic uptake ofthe given (NCCLS) (19, 44, 48, 51, 107, 108). As a result, manyofthe
substrate does not necessarily imply death. Moreover, they factors that influence (often profoundly) this technique have
require special material-handling procedures in the case of been identified, and a standard method that minimizes their
theuseofradiolabeled substratesandrelativelylargeinocula adverse effects has been proposed (92). This method has
in order to produce a measurable amount of product in the been proposed exclusively for testing Candida spp. and
case ofcolored substrates. Approaches that use flow cytom- Cryptococcus neoformans; reproducible methods for other
etry(60, 116-118) orviable colonycounts (71, 95) havebeen organisms are still under development (see below). In the
described but are applicable only to organisms that disperse nextsection,wereviewthevariablesinvolvedinimplement-
freely and are technically or physically demanding. Mea- ing a broth-based susceptibility test, with special emphasis
surementofbiomassbybioluminescence spectrophotometry onhoweachvariableaffectstheproposedNCCLSreference
hasbeen described (95). This technique used measurements method (Table 1). Knowledgeofthesevariablesishelpfulin
at multiple points along a dose-response curve to allow understanding the relative importance ofthevarious details
computation of a relative inhibition factor (100), and al- of the assay and in interpreting studies that use variant
though the method is intriguing, instrumentation thatwould methodologies.
O
VOL. 6, 1993 ANTIFUNGAL SUSCEPTIBILITY TESTING 369
Iinn
0
o Z0 0 - 0ID 00 80 W/o
(D 0 0 ''. .......'''7 - -. ------ -- ...-. ................................................................
a
c0nn
40
CCD z CD
z 20
COD 0 0
00 CD n.
= (D0 Cl
CD 0.125 0.25 0.5 1 2 4 8 16
gNOg- ° O 0
CD 0. 0D DrugConcentration,,ug/ml
0~~~ _). FIG. 1. Trailingendpoints. A C. albicans isolatewas tested for
00 susceptibilitybyusingtheproposedNCCLSreferencemethod,and
03A 1-1 percent transmission at 530nmwas determined. The lines marked
Ut 0Q 80and90%arethepercenttransmissionof1:5and 1:10dilutionsof
<0
thedrug-free controltube, respectively. The amphotericin Bcurve
CD (O) makes a very rapid transition from marked turbidity to 100%
transmission, while the fluconazole curve (x) trails markedlyover
CD
thisrange.TheamphotericinBMICforthisisolateis0.5 ,ug/ml(first
tubewith100%transmission),andthefluconazoleMICis0.25 ,ug/ml
N uC. D_ (firsttube ator above 80% transmission).
CD - o '-A0
0. FACTORS THAT INFLUENCE ANTIFUNGAL
CD SUSCEPTIBILITY TESTING
_.
0 C.DDD 0
oZ 0 0O0-0 0Cr Endpoint Definition
0 O Endpoint determination is probably the most significant
,0 CD r0 0. source ofinterlaboratoryvariability for the azole antifungal
cCDD agents and, under some conditions, 5-fluorocytosine. For
CD these agents, inhibition does not develop abruptly over a
3 00
ZP smallconcentration range. Rather, afteran initialprominent
0_C; (0 drugeffect, smallamountsofturbiditymaypersistformany,
CD0 _~ 00 if not all, drug concentrations tested (Fig. 1). As a result,
3.1 laboratories that insist on stringent reduction of turbidity
report much higher MICs than do laboratories that adopt
04 P-0
toCD 0 o'0 0 endpoints that are tolerant ofsmall amounts ofturbidity.
CsD Incontrast, endpointswithamphotericinB andcilofungin
(DZ. 00CD qCD do not show this phenomenon, and comparison of the
U) 0 kinetics of the onset of drug effect has provided some
00~ la C_CDDF explanationfordifferencesbetween these drugs (54, 85). At
x. x rC x -1. odrimaebtroivce smoeamseucrreitmiecnaltsampdheomtoenrsitcrianteB ctohnactengtrroawtitohn,cteuarsbeis-
0' CD . CD soon after exposure to the drug. However, growth does not
00 CA c)p begin to slow until approximately one doubling time after
_ -4'I=1 exposure to 5-fluorocytosine or the azole antifungal agents,
0 andgrowthisnotfullyarrested until some time later. These
-' o A 0 _t observations demonstrate that this trailing endpoint is the
0 result of growth occurring for a period of time prior to the
o~ 0 onset ofcomplete drug effect. Predictably, the trailing end-
point problem has been shown to worsen with increasing
00o
t initialinoculum(127), andthismaybethereasonthatrecent
collaborative trials have found better reproducibility with a
smallerinoculum (27,44, 48). Ithasbeen suggestedthatthis
problem couldbe reduced bythe addition to the medium of
a protein synthesis inhibitor such as doxycycline (96, 101),
but this approach has not been exploited further.
0
Giventheseobservations, several approaches toendpoint
definition are possible. One is to describe the endpointwith
5 carefully selected phrases. This was done in two recent
tz collaborative studies in which endpoint determinations of
"optically clear," "slightly hazy," and "prominent de-
crease in turbidityfromcontrol" were used (44, 48). (These
descriptions are also referred to as 0, 1+, and 2+, respec-
CDa tively.) It was found that the endpoints that allowed some
370 REX ET AL. CLIN. MICROBIOL. REV.
turbidity gave more reproducible results and also produced Inoculum Preparation
better agreementwith known patterns ofinvivo resistance.
While some trade-offs were involved, the "prominent de- In a recent collaborative study of inoculum preparation,
crease in turbidity from control" endpoint was the most four techniques were compared by three laboratories: a
consistently reproducible. spectrophotometric method, the Wickerhamcardmethod, a
Communicating the precise meaning of "prominent de- hemacytometer method, and the Prompt inoculation system
crease in turbidity from control" is, however, problematic. (107). Yeast suspensions were adjusted in an agreed upon
An approach to the resolution of this problem has been fashion, using each system in order to produce aninoculum
described byEspinel-Ingroffet al. (44). These authors dem- of 1 x 106 to 5 x 106 per ml, and then plated to assay for
onstrated that the simple method of diluting the drug-free viable CFUs. The Wickerham card method and the Prompt
control 80% after incubation (8 parts medium, 2parts yeast inoculationsystemwerethemostvariableandoftenfailedto
suspension) provided averygood approximation of"prom- produce the desired inoculum. Adjustment by matching the
inent decrease in turbidity." To illustrate this relationship, turbidityat530nmofa0.5McFarlandstandardwasthebest
Fig. 1 shows the turbidity of a strain of C. albicans after method; this was followed closely by hemacytometer
incubation with different concentrations of fluconazole and counts. The hemacytometer method displayed greatervari-
amphotericin B. As can be seen, an 80% dilution is easily ability, apparently because of consistently aberrant results
distinguished from the residual turbidity and for this reason from one laboratory. On the basis of these results, the
is likely to result in agreement when read visually by proposed NCCLS reference method has adopted spectro-
different observers. On the other hand, the turbidity pro- photometric standardization of the inoculum. While such
ducedwith a 90% dilution ofthe drug-free control provides techniquesasstainingwithafluorescentdye(25)andtheuse
alesssatisfactoryendpointdefinitionsinceitisverynearthe of a Coulter counter (41) have been suggested, these tech-
residual trailing endpoint. Similar trailing endpoints were niques require equipment that maynot be readily available.
observedwiththesulfonamides inantibacterialtestinganda
similar solution was applied (8), thus lending precedent to Incubation Time and Temperature
this approach.
Another approach has used spectrophotometric determi- Although the magnitude of the effect varies from drug to
nations to quantitate turbidity (4, 9, 11, 53, 54). Similar in drug, MICs tend to increase with increasing incubation
spirit tovisual estimations ofrelative turbidity, spectropho- periods (11, 12, 15, 36, 75, 94). For example, Brass et al.
tometric measurements lend themselves better to numeric found that the ketoconazole MICfor a strain ofC. albicans
manipulation aswell as holding the potential foreliminating increased 10-foldinabrothmethodand2,000-foldinanagar
subjectiveinterpretations. Intheseapproaches, theturbidity dilution assay during 7 days of observation, while the
ofthe drug-free controls is defined as 100%, and the turbid- nystatin MICs roseby, at most, one dilution (15). MICs for
ities of the test dilutions are scaled to this value. The yeasts are usually stable by 4 days (94). Temperature
concentration atwhichtheturbidityisreducedtoaspecified changes are also significant. The MIC hasbeendescribedto
percentage, typically 50 or 30% ofthe drug-free control, is both rise and fallwith temperature changes from 22 to 37°C
then taken as the endpoint and called the 50 or 30% inhibi- (9, 12, 99). The standard temperature for antibacterial sus-
tory concentration (IC50 or IC30, respectively). Several ceptibility testing in the United States is 35°C, and this
studies have found that such an approach eliminates inocu- appears to be a good choice because it generally supports
lum dependence (4, 9, 53, 54), but this method requires fungal growth better than 30°C (108) and because MIC
specific instrumentation to perform the measurement. resultsarelessvariablethanat37°C. Inonesurprisingstudy,
Endpoint determinations that examine killing have been macro- and microbroth procedures for fluconazole and the
described. Such studies can be implemented as kinetic investigational triazole SCH39304 in RPMI 1640 were com-
turbidimetric and time-kill curves or as determinations of pared at two inocula (102 and 104 CFU/ml) and two temper-
minimum lethal concentrations (MLCs) (147, 149). A few atures (35 and37°C) (27). Whilethevarious combinationsof
reports have suggested that MLCs may be more important temperature and inoculum gave nearly identical results at
thanMICsinpredictinginvivoresistance (24, 83, 149). This 35°C, therewas asmuch as an eightfold difference inresults
general area has not yet been pursued in detail. If the at 37°C. Finally, a collaborative study of 100 isolates in 13
experience with bacteria is any guide (142), defining a laboratories found that, for tests of amphotericin B, keto-
reproducible technique for MLCs will likely require more conazole, and 5-fluorocytosine in RPMI 1640 at 35°C, the
work than has already been required to define a technique most consistent overall results were obtained after incuba-
for MICs. tion at 35°C for 48 h for Candida spp. or 72 h for Crypto-
coccus neofonnans (48). In both cases, these times of
reading correspond to the second day on which growth is
Inoculum Size readily apparent in the drug-free control tube.
On the basis of these results, the proposed NCCLS
Starting concentrations for yeasts in broth assays have reference method specifies incubation at 35°C for 48 h for
rangedfrom 102to106perml.Unlessarelativeturbidimetric Candida spp. or 72 h for Cryptococcus neofonnans.
endpointisused, MICsformostdrugsincreasewithincreas-
ing inoculum over this range, regardless of test format,
Media
sometimes as much as 512-fold (12, 15, 53, 57, 75, 85, 94).
Both intra- and interlaboratory reproducibilities tend to Aswithtestingofbacteria,useofdifferentmediamaygive
improvewith smaller inocula (27, 44, 48). Recent collabora- substantially different results (27, 36, 37, 57, 66, 108). When
tive efforts found that an inoculum in the range of0.5 x 103 conventional, undefined media (e.g., Sabouraud agar) were
to 2.5 x 103 per ml produced the most interlaboratory critically compared with the totally synthetic medium
agreement, and thisvalue hasbeen adopted in the proposed SAAMF (synthetic amino acid medium, fungal), substantial
NCCLS reference method (92). differences in 5-fluorocytosine MICs were noted (66), and
VOL. 6, 1993 ANTIFUNGAL SUSCEPTIBILITY TESTING 371
these differences turned out to be due to antagonism of mans were studied in 13 independent laboratories for sus-
5-fluorocytosine action by purines and pyrimidines in the ceptibilityto amphotericin B, 5-fluorocytosine, and ketocon-
media (66, 113, 114). Other factors may affect other antifun- azole, while Espinel-Ingroffs study made use of the same
gal agents: sterols in the medium can interfere with the isolates in 5 of the same 13 laboratories, but these 5
effects of polyenes (78), and as yet undefined factors have laboratories also tested fluconazole. Thus, Espinel-Ingroff's
beenpostulated tointerferewiththeeffectsofazoles in some data for fluconazole are unique, but the other data from this
media (67, 94). Even well-defined media can hold surprises: study are a subset of the Fromtling study and will not be
the use of morpholinepropanesulfonic acid (MOPS)-Tris to discussed further.
buffer SAAMF was found to inhibit the action of 5-fluoro- Many of the results from these studies have been alluded
cytosine, independently of its effect on pH (18). This effect to during the previous discussion, but there are additional
was later determined to be due entirely to the Tris (80). noteworthy points. First, the choice ofan inoculum of0.5 x
Differences that are not due to the presence of antagonistic 103 to 2.5 x 103 per ml, an endpoint of"prominent decrease
substances may be due to pH. Not all media have the same in turbidity," and reading at 48 h for Candida spp. and 72 h
native pH or a similar buffering capacity, and the rate of for Cryptococcus neoformans constitute a compromise
growth of the test organism may be affected by pH (18). made to optimize interlaboratory agreement for all of the
Further, acidicconditions havebeenfound toraisetheMICs study drugs. A result was said to be in agreement if it was
of amphotericin B (9, 40), fluconazole (85, 127), SCH39304 within 1 dilution of the modal result for that strain by all
(85), ketoconazole (40, 90, 127), miconazole (40), and cil- laboratories, and bythese criteria, amphotericin Bproduced
ofungin (86) but to lower 5-fluorocytosine MICs (40). This >85% interlaboratory agreement for a variety of starting
effect on MICs can be substantial: ketoconazole MICs are conditions and endpoint definitions. Ketoconazole, on the
64- to 1,000-fold higher at pH 3.0 than at pH 7.4 (90, 127). other hand, required precisely the above conditions in order
Thus, it is clearly desirable to use a buffered medium. The to achieve even 75% interlaboratory agreement. Second, the
optimumpH, however, isless clear. While the neutralpHof interlaboratory agreement produced with this technique is
7.0isintuitivelyattractive, McIntyre andGalgiani (86) found better than has been achieved in the past but is still not
thatcilofungin MICs forC. albicans spanned awide range at perfect. The agreement for amphotericin B is 90%; for
pH 7.0 (0.08 to 2.5,ug/ml) but a much narrower range at pH 5-fluorocytosine, 85%; for fluconazole, 88%; and for keto-
3.0 (5 to 10,ug/ml). When isolates forwhich MICswere 0.08 conazole, 75%. On the basis ofthese and other studies, it is
and 1.25,ug/ml at pH7.0, but identical (5.0,g/ml) atpH3.0, clear that this rank order for reproducibility is a feature of
were tested in a rat model of disseminated candidiasis, the the drugs and not the assay system. For example, when
50% effective doses of cilofungin were identical, thus sug- Shawaretal. (135) recentlytested avarietyofbroth microdi-
gesting that the pH 3.0 MIC results were more meaningful. lution and semisolid agar dilution methods in two laborato-
While MICs for some isolates ofthree other species studied ries, they found that results for amphotericin B and 5-fluo-
by these authors were >80 ,ug/ml at pH 7.0, these isolates rocytosine were highly reproducible, fluconazole had good
were not tested in the animal model. but not perfect reproducibility, and itraconazole and
The insolubility ofsome compounds maybe a final source SCH39304 had poor reproducibility. Third, this method
of difficulty. When working from highly concentrated stock produces abroadrangeofMICsof5-fluorocytosine (0.125 to
solutions, it is important that all dilutions be made in media 64 ,ug/ml), ketoconazole (0.03 to 2 ,ug/ml), and fluconazole
thatwill maintain the drug in a soluble form. When working (0.12 to 64 ,ug/ml), but not amphotericin B (0.25 to 1.0
with bifonazole, an imidazole derivative, Plempel et al. ,ug/ml). These results were obtained despite inclusion in the
demonstrated that drugconcentrations couldbe artifactually study of a number of C. lusitaniae isolates, a species for
lowered by as much as 10-fold if this principle was not which elevated MICs by agar dilution methods in cases of
observed (110). To address this problem, the proposed amphotericin B resistance have frequently been described
NCCLS reference method suggests that amphotericin B and (62, 87, 89, 103). These datasuggest that eithertherewereno
ketoconazole be dissolved in dimethyl sulfoxide. resistant C. lusitaniae isolates in the test collection or the
On the basis of these results, a completely synthetic, assay system is relatively insensitive to amphotericin B
bufferedmediumwould appeartobeoptimal. Even thoughit resistance.
maybesuboptimalforsomedrugs, pH7.0hasbeen themost This apparent inability to detect resistance to amphoteri-
commonly used pH. Most current investigations, including cin B may in part be a function of the choice of a broth
the proposed NCCLS reference method, have adopted the system rather than an agar system. While MICs by agar
syntheticcell culture medium RPMI 1640,buffered topH7.0 dilution can range overmore than 2orders ofmagnitude (33,
with MOPS. RPMI 1640 is readily available commercially 89), results by broth dilution methods have usually varied
andisrelativelyinexpensive. This mediumandbufferarenot over narrower ranges. For example, in the relatively large
sacrosanct: similar results with other buffers (80) and the series of amphotericin B-resistant Candida infections re-
synthetic HR (high-resolution) medium (108) can clearly be ported by Powderly et al. (120), MICs for the 29 described
obtained. isolates were between 0.4 and 7.5 ,ug/ml by a macrobroth
technique (yeast extract broth, 106yeast cells per ml, 30°C,
48 h). Other authors have also described amphotericin B
Results Obtained with the Proposed NCCLS
MICs by macrobroth techniques that varied over a similar
Reference Method
range forbothCryptococcus neoformans (9)andC. albicans
Fromtling et al. (48) and Espinel-Ingroff et al. (44) have (19, 122). The use of RPMI 1640 in the proposed NCCLS
recently reported results from the first major studies to use method may have further narrowed the range of amphoter-
the NCCLS assay. Of note, these two studies present icin B MICs produced by the proposed method. Ghannoum
somewhat overlapping sets ofdata: in Fromtling's study, 35 et al. (57)recentlyhave described results ofMICtestingfor
strainsofC. albicans, 15strainsofC. tropicalis, 15strainsof 21 isolates of Cryptococcus neofonnans, using 104 yeast
C. parapsilosis, 10 strains of C. lusitaniae, 10 strains of cells per ml and a 48-h incubation at 35°C in a microtiter
Torulopsisglabrata, and 15 strains ofCryptococcus neofor- format. Four media (MOPS-buffered pH 7.0 RPMI 1640,
372 REX ET AL. CLIN. MICROBIOL. REV.
MOPS-buffered pH 7.0 yeast nitrogen base [YNB], MOPS- the macrobroth method. Agitation of the microtiter plates
buffered pH 7.0 SAAMF, and pH 5.4 YNB) were studied. prior to reading may be useful to help eliminate these
The data demonstrate compression of the amphotericin B discrepancies and to improve reproducibility (4, 57, 85).
MIC range when RPMI 1640 is used, and these authors Even more innovative simplifications could be developed.
conclude that pH 7.0 MOPS-buffered YNB is preferable to For example, the relative turbidity of the growth of an
RPMI 1640 for testing amphotericin B susceptibility of organism in a single concentration of a drug was used to
Cryptococcus neofornans. As a final demonstration of this classify organisms by comparison to the typical growth of
phenomenon, when 234 blood isolates ofCandida spp. were the given species in drug-free media (98). While this tech-
tested by the proposed NCCLS method, the fluconazole nique does not give an actual MIC, it could, with carefully
MICs ranged from 0.125 to >64 ,ug/ml, but all amphotericin chosen drug concentrations, prove to be a simple way to
B MICs were between 0.125 and 1 ,ug/ml (123). distinguish resistant and susceptible strains.
Taken together, these observations suggest that, while Finally, achievingconsensus onareference procedurewill
broth systems can detect amphotericin B resistance, the make itpossible to begin to assess the importance ofinvitro
proposed NCCLS method does not appear to be well suited drug resistance in therapeutic failures. As will be seen
to detect resistance to this particular drug. Two approaches below, it is highly likely that a correlation based on isolate
to resolving this problem can be proposed. First, if the rank order will exist, but it is much less certain that the
proposed NCCLS reference method produces a narrow actual MIC generated by the test will have a meaningful or
range of values, more dilution steps within the expected predictable relationship to achievable serum levels of that
range might be required to demonstrate a distribution of drug.
values and to consistently allow discrimination between
isolates. This concept is supported byrecent data presented
OTHER ORGANISMS
by Dignani et al. (34). With a microplate adaptation of the
proposed NCCLS method (4), four strains of C. lusitaniae Testing of organisms other than Candida spp. and Cryp-
with MICs ranging from 0.312 to 2.5 ,ug/ml were identified, tococcusneoformans presentsaspecialsetofproblems (49).
andthese MICswere found tocorrelatewiththeoutcome of While Candida spp. may exhibit some hyphal forms, Can-
amphotericin B therapy in a mouse intravenous model of dida spp. and Cryptococcus neoformans exist predomi-
candidiasis. When these isolates were tested by one of us nantly in the laboratory and during infection in a small,
(J.H.R.)with the proposedNCCLS method (the testingwas round yeast form. These yeast cells are easily counted,
done in a blinded fashion during the testing of the just- diluted, and transferred from onevessel to another, making
described 234 blood isolates of Candida spp.), a MIC of 2 a wide variety of test procedures possible. And, in part
,ug/ml on five occasions and a MIC of 1 ,ug/ml on two other becausetheinvasiveformoftheorganismiseasilystudiedin
occasionswere found forthe isolatewith amicrobroth MIC the laboratory, it seems likely that susceptibility testing
of 2.5 ,ug/ml. Thus, for this isolate, a modal MIC that was would predict clinical response. It is for these reasons, as
higher than that for any of the other tested isolates was well asthefactthatthemajorityoffungalinfections are due
found, but given the expected one-tube, between-run varia- totheseorganisms, thatmostpublished studiesofantifungal
tion in MIC results, it is not reasonable to expect the susceptibility testing and all of the current efforts towards
proposed NCCLS method to consistently detect such a standardization have focusedexclusivelyon studiesofCan-
difference. Alternatively, if smaller concentration steps do dida spp. and Cryptococcus neoformans.
not resolve this problem, some modification ofthemediaor The dimorphic fungi which exist in nature as molds but
test conditionswill need to be developed. This is an area of which haveyeast forms during infection (Histoplasma cap-
active investigation. sulatum, Blastomyces dermatitidis, and Sporothrx
Despite these difficulties, the proposed NCCLS reference schenckii) can be maintained as yeasts in the laboratory.
methodology should be useful as a touchstone during the Special laboratory precautions are required, and the tech-
next stages in the evolution of antifungal susceptibility nique would need to be modified in order to accommodate
testing. Inadditiontothepointsmentionedabove,twoother the typically slower growth of these organisms (24, 45, 49,
areas need to be addressed. First, the proposed NCCLS 64),buttheycouldpotentiallybestudiedinthesamefashion
reference method is a macrobroth method that uses 1.0-ml as Candida spp. and Cryptococcus neofonnans. In making
volumes in test tubes. This technique is cumbersome and suchadaptations, itwill, however, be criticaltopayheed to
should be replaced by a microbroth method that gives the individual properties ofthe various fungi. For example,
identical results as soon as possible. Even though the some strains of S. schenckii are exquisitely sensitive to
difference is only a change in scale of the assay, MICs by temperatures just above 35°C, and it would be critical that
microbroth methods have not been the same as those ob- this temperature never be exceeded (132).
tainedwith a macrobroth technique: microbroth MICs have Filamentous fungi such as Aspergillus spp. present a
been reported to be both higher (43, 44, 85) and lower (27) different picture. These organisms do have a small, round,
thanmacrobroth levels. In arecent studythat comparedthe conidialformthatiseasilycounted, diluted, andtransferred,
proposed NCCLS reference macrobroth method with a but the invasive form ofthe organism is a hyphal form that
microbrothmethodthatsimplyreducedthe 1-mlmacrobroth cannot be accurately counted, diluted, or transferred be-
volumes to 0.2 ml (43), interlaboratory agreementwas opti- causeofboththereadyadherenceofhyphaetosurfacesand
mal and similar for the two techniques with the same the macroscopic, interconnected nature ofthe hyphal mat.
inoculum, incubation, and endpoint conditions. However, Because of these difficulties, most studies of molds pub-
the resulting microbroth MICs were higher than the macro- lished to date have startedwith an inoculum ofconidia and
broth MICs. Of note, this study compared only results terminated upon detection of the presence or absence of
obtained by identical methods (e.g., microbroth and macro- hyphalgrowth(29, 121). Lackofgrowthinsuchstudiesmay
broth with the same starting inoculum, time to reading, and be due to inhibition ofgermination rather than inhibition of
reading endpoint), rather than seeking to define an indepen- hyphal growth, and it is less certain that such approaches
dent microbroth method that produced the same results as wouldcorrelatewithclinicalresponse. Susceptibilitytesting
VOL. 6, 1993 ANTIFUNGAL SUSCEPTIBILITY TESTING 373
ofaspergilli has been reviewed recently (29), and the issues 139). ThemostthoroughstudyisthatofStilleretal. (139), in
involved in defining a reproducible procedure have only which 40 isolates withvarying in vitro susceptibilities were
beguntobedefined(42). Asnotedabove, allofthesestudies examined. This study demonstrated a clear-cut statistical
suffer from potential problems related to starting the assay correlation, but unfortunately there was substantial overlap
with an inoculum of conidia or asexual propagules rather between the amount of5-fluorocytosine required to control
than with one ofhyphae. infectioncausedbyisolateswith low5-fluorocytosine MICs
PreliminarydataobtainedbyMartinet al. suggestthatthe versus thosewith high MICs. Thus, despite the presence of
distinctionbetween startingwithconidiaversushyphaemay an obvious in vivo-in vitro correlation, no clear breakpoint
be more than academic (82). In their study, germinating todistinguish5-fluorocytosine-susceptible and-resistantiso-
conidia were observed and harvested when discrete germ lates could be established (Fig. 2). These authors conclude
tubes, but not intertwined hyphae, were present. Such an that their isolates did not contain a distinct subset ofresis-
inoculum should, in theory, exhibit behavior that more tantisolates butratherwerecomposedofagroupofsubiso-
closelyapproximatesthatofmaturehyphae. Similarampho- lates with a continuous spectrum of susceptibilities. The
tericin B MICs forAspergillus spp.,Fusanum spp., andthe otherthreestudiesof5-fluorocytosineareeithertoosmallor
dematiaceousfungiwere foundwithbothresting andgermi- asyetincompletelypublishedand donotshedmore lighton
nating conidia, while higher amphotericin B MICs for R this problem.
arrhizus were found with germinated sporangiospores. Itra- The largest number of studies are those of the azole
conazole MICs were either the same or only slightly higher antifungals. While these studies have usually demonstrated
for the germinated forms. However, MLCs ofitraconazole, an in vivo-in vitro correlation, they also provide several
but not amphotericin B, for the dematiaceous fungi and useful lessons. First, as with the above-mentioned studyby
Aspergillus spp. were much higher in the germinated forms. O'Dayet al. (93), the degreeofcorrelation depends strongly
Taken together, these data suggest that substantial work to on the particular in vitro test. Ryley et al. (128) examined
define appropriate testing procedures for the molds remains two isolates that were clearly resistant to ketoconazole in
to be done. vivo inboth humans and three different animal models. The
isolateswerefoundtoberesistantbytwobrothmethodsand
CORRELATION OF IN VITRO TESTINGWITH two agar diffusion methods, but by an agar dilution method
IN VIVO OUTCOME the MICs for the isolates were the same as those for the
control isolates. Polak et al. (112) studied the ketoconazole
While it is obviously important to have a reproducible susceptibility of 58 C. albicans isolates, using 16 different
assay,anassayisneededonlywhenclinicalresistanceexists agar and broth methods, and correlated these results with
and is predictedbythe assay. Adequate studies tocorrelate the dose of ketoconazole required to keep 50% of the
invitro test resultswithinvivo efficacyin humans have not animals alive (50% effective dose [ED50]). Onlyone method
yet been done. However, manyindependentpiecesofinfor- (disk zone diameter on YNB agar) demonstrated a correla-
mation suggest that it will be possible to demonstrate a tion, but this correlation could not be used to establish a
correlation. Three types of data are available. First, there MIC breakpoint because of the relatively narrow range of
are animal studies in which the in vitro-in vivo correlation ED50valuesseenintheanimalmodelaswellasdifficultiesin
hasbeen directly addressed. Second, there are a substantial reading the zone diameters.
number ofclinical observations inwhichclinical recurrence Second, the in vitro test must demonstrate differences
or failure during drug therapy has been correlated with between isolates in order to be predictive. In the study by
reduced in vitro susceptibility. Finally, there are a small Fisher et al. (47), only three Candida isolates were tested.
numberofclinicaltrialsduringwhichisolateswerecollected For all three the fluconazole MICs were >100 ug/rml. Flu-
and tested for susceptibility and for which the results were conazolewas active invivo against two ofthethree strains,
correlated with clinical outcome. butwithout differences in MICs it is not possible to discern
acorrelation. Itisnotsufficienttoassumethatanumerically
large MIC implies resistance: the above review oftechnical
Animal Studies
variablesclearlydemonstratesthatthetestconditionscanbe
The relationship between in vitro results and in vivo manipulated to produce MICs in any desired range.
efficacy forthe polyenes has been examinedwith a number Third, interpretation of the results must take drug phar-
of animal models (Table 2). The most exhaustive study is macology into account. Rogers and Galgiani (127) studied
that of O'Day et al. (93). Microbroth MICs produced an twoC. albicansisolates: theMICsofbothketoconazoleand
excellentcorrelationforbothamphotericinB andnatamycin fluconazole were low for one isolate and high for the other.
(another polyene), while microbroth MLCs and an agar Fluconazole was effective therapy for the isolate with the
diffusionassaywerelesspredictiveandtendedtoincorrectly lower MIC but not for the isolate with the higher MIC. On
classify isolates unless the tests were performed multiple the other hand, despite having a 16-fold lower MIC than
times. Similar resultswere obtained in a mouse intravenous fluconazoleagainstthesusceptibleisolate,ketoconazolewas
infection model (3). In the study by Fisher et al. (47), three ineffectiveinthisintravenous infectionmodel. This discrep-
Candida isolateswithamphotericinBMICsof0.2, 0.39, and ancycouldbe duetothegreaterproteinbindingandshorter
0.78 jig/ml were tested. Invivo, amphotericin Bwas active half-life ofketoconazole. Moreover, since the kidney is the
against all three isolates, but the reduction in kidney and primarytargetorganinthisexperimentalmodel,thefactthat
liverCFU pergramwasleastforthe isolatewiththehighest fluconazole is concentrated and excreted in the urine while
MIC.Thisresult,however, isdifficulttointerpretbecauseof ketoconazole is not may also be important.
otherdifferencesbetweentheisolates: differentinoculawere In anotherdemonstrationoftherelevanceofhostfactors,
required to produce infection, and the organ CFU pergram Boyle et al. (14) studied a large number ofvariously substi-
in the untreated controlsvaried over a 3-log range. tutedazole derivativesbothinvivo andinvitro. Theyfound
Allfourpublishedstudiesofinvivo-invitrocorrelationfor that,when groupedbyclassofsubstitution, those classesof
5-fluorocytosine have demonstratedacorrelation(3, 34, 122, molecules that would be expected to be relatively resistant
374 REX ET AL. CLIN. MICROBIOL. REV.
TABLE 2. Relationship between invitro susceptibility test results and invivo efficacyofantifungal agents in animal models
Drug(reference) Model(s)andfungus (no. ofstrains) Invitromethod,medium,inoculum(CFU/ml), Invitro-invivocorrelation
temp(O'C), time
Amphotericin B (3) Mouse/i.v.,a C. albicans (9) Microbroth, other details not stated Yes
Amphotericin B (47) Mouse/i.v., C. tropicalis, T. glabrata, Macrobroth, antibiotic medium 3, 106/ml, Not clearly
and C. krusei (1 each) tempnot stated, 48 h
Amphotericin B(34) Mouse/i.v., C. lusitaniae (4) Microbroth, RPMI 1640, 104/ml, 35°C, Yes
24h
Amphotericin Band Rabbitmeningitis, C. neoformans (1); Macrobroth, Sabouraud agar, 103-104/ml, No
amphotericin B andrabbit endocarditis and pyelone- 24h, 30°C
methylester (104) phritis, C. albicans (1)
Amphotericin B and Rabbit keratitis, C. albicans (17) Microbroth, antibiotic medium 3, 105/ml, Yesformicrobroth MIC;
natamycin (93) 24h, 37°C; alsobyan agardilution microbroth MFC' and
method agardilution methods
not asgood
5-Fluorocytosine (3) Mouse/i.v., C. albicans (9) Microbroth, otherdetails not stated Yes
5-Fluorocytosine (34) Mouse/i.v., C. lusitaniae (4) Microbroth, RPMI 1640, 104/ml, 35°C, Yes
24h
5-Fluorocytosine (122) Mouse/i.v., C. albicans and Microbroth, RPMI 1640, 103/ml, 24h, Yes
C. tropicalis (1 each) 350C
5-Fluorocytosine (139) Mouse/i.v., C. albicans (40) Macrobroth, YNB, 103/ml, 7days, 37°C Yes
Ketoconazole (122) Mouse/i.v., C. albicans and Microbroth, RPMI 1640, 103/ml, 24h, Yes
C. tropicalis (1 each) 350C
Ketoconazole (128) Mousevaginosis, ratvaginosis, and Various agar andbroth methods Yes for allbut one agar
mouse/i.v., C. albicans (43) dilution method
Ketoconazole (112) Mouse/i.v., C. albicans (58) 16different agar and brothmethods Slight correlationwith one
method; nonewithoth-
ers
Ketoconazole and Rabbit/subcutaneous chamber, Macrobroth, YNB, 105/ml, 48 h, 370C Yes
fluconazole (72) C. albicans (5)
Fluconazole (34) Mouse/i.v., C. lusitaniae (4) Microbroth, RPMI 1640, 104/ml, 350C, Yes
24h
Fluconazole (47) Mouse/i.v., C. tropicalis, T. glabrata, Macrobroth, SAAMF, 106/ml, 48 h, temp Not clearly
and C. krusei (1 each) notstated
Fluconazole (127) Rat/i.v., C. albicans (2) Macrobroth, buffered SAAMF, 103/ml, Yes
22h, 37°C
Fluconazole (148) Mouse/intracerebral, Cryptococcus Macrobroth, SAAMF, 0.5 x 103-2.5 x Yes at24hbut no at48h
neoformans (20) 103/ml, 24 and48h
Fluconazole and Rabbit pyelonephritis, C. albicans (2) Macrobroth, SAAMF, 105/ml, 24h, 300C Yes
itraconazole (105)
SCH39304(85) Rat/i.v., C. albicans (2) Macrobroth, buffered SAAMF, 103/ml, Yes
22h, 370C
BayN 7133 (109) Mouse/i.v., C. albicans (59) Agardilution, Isotonic broth, 104/ml, No
48h, 28°C
Multiple mono- and Mousevaginosis, C. albicans (1) Macrobroth, YNB, 105/ml, 48 h, 370C Yes
bis-triazoles (14)
Cilofungin (86) Rat/i.v., C. albicans (2) Macrobroth, buffered SAAMF, 103/ml, No
22h, 370C
ai.v., intravenous.
bMFC, minimalfungicidalconcentration.
VOL. 6, 1993 ANTIFUNGAL SUSCEPTIBILITY TESTING 375
cY 1ooo- for the posttreatment isolate was found to have risen four-
cm fold but to also be less virulent in a mouse model. These
-.4
correlations are not overwhelming, nor have they been
ac)) 100-
0 consistently observed byothers. Hughes et al. (73) recently
a) tested 60 Candida bloodstream isolates for amphotericin B
15 10- susceptibilitybyabroth method in RPMI 1640with 5 x 104
0 yeast cells per ml at 37°C for 48 h. While the MICs for all
LELJ isolates were 0.5 to 2.0 ,ug/ml, those for strains of C.
EE 1.0- 00 parapsilosis were 1.0 and 2.0 ,ug/ml and MICs for strains of
0* C. albicanswere0.5or1.0 ,ug/ml. Despitethistrendtowards
01 lower susceptibility of the C. parapsilosis strains, the pa-
> 0.1- tientsinfectedwithC.parapsilosis hadahigherfrequencyof
._
resolution of infection. It is difficult to interpret these con-
._ flicting data on polyene resistance. While amphotericin B
Ill
invitroSusceptibilityGroup MICs,byatleastsometechniques, doseemtorisefollowing
therapy (9, 32, 119), the more resistant organisms may be
FIG. 2. Correlation of 5-fluorocytosine MIC with minimum ef-
lessvirulent (9, 119), and it is often difficult to truly ascribe
fective dose of 5-fluorocytosine. Responses to 5-fluorocytosine
therapyinmiceinfectedwithfourdifferentgroupsofC. albicansare failuretothe druginthefaceofthemultiple immune system
shown.SusceptibilitygroupIcontainsthemostsusceptibleisolates, defects present in the typical patient.
while susceptibility group IV contains the most resistant isolates. Althoughexcellent animalmodeldataareavailable (Table
DataarefromStilleretal.(139)andarereproducedwithpermission 2), 5-fluorocytosine is now rarely used alone to treat pa-
ofthe publisher. tients, and well-studied cases describing the correlation
between invitro and invivo results are not available. In an
unpublished doctoral thesis (28), Deldicque described a
to metabolic degradation by the host had a good invitro-in small numberofcases inwhich5-fluorocytosine susceptibil-
vivo correlation whereas compounds thatwere predicted to ity results appeared to correlate with outcome of 5-fluoro-
be readilydegraded did not demonstrate agoodcorrelation. cytosine monotherapy. Unfortunately, clinical details are
Only one study has sought an invivo-in vitro correlation limited and it is hard to draw a firm conclusion.
for cilofungin (86). This study was discussed above, and Azole resistance has often been described. In an early
therewas nocorrelation despite demonstration ofatantaliz- report, initial clinical response was followed by relapse
ing range ofcilofungin MICs at one pH but not another. during prolonged use of oral miconazole: the sequential
isolates demonstrated a 200- to 500-fold rise in MIC by an
unspecified method (69). In another set of reports, three
Clinical Demonstrations ofBoth Inherent and
isolates from patients who relapsed after prolonged contin-
Acquired Resistance
uous ketoconazole therapyfor chronic mucocutaneous can-
Polyene resistance has been said to be rare, but its exact didiasis (70, 136) and one from a patient who never re-
incidence is hard to determine. A number of studies have sponded (151) were described and studied. These C.
demonstrated rises in amphotericin B MICs for isolates albicans isolates are clearly more resistant to ketoconazole
obtained during prolonged polyene therapy, and these rises bytests ofinhibition ofhyphal elongation (151) and byboth
have sometimes appeared to correlate with clinical failure agardilution andmacrobroth techniques (75, 105, 127, 128).
(13,33,38, 46, 103). Inoneoft-citedstudy(32),isolatesfrom Three of the four isolates were studied in more detail and
6 of70 cancer patients exhibited amphotericin B resistance found to be cross resistant to miconazole, ketoconazole,
(defined as a MIC of 22 ,ug/ml) by an agar dilution tech- itraconazole, and fluconazole by agar dilution, macrobroth,
nique. Five of these six patients had received extensive inhibition of germ tube elongation, and relative inhibition
polyene therapy, but additional clinical details were not factor techniques (136). Two of these isolates were also
provided. In a later report, Powderly et al. (120) examined studied in animal models and found to be probably more
bloodstream isolatesfrompatients undergoingbonemarrow resistant to therapy in several different animal models (136)
transplantation. Using a macrobroth technique in yeast anddefinitelymoreresistantinarabbitpyelonephritismodel
extract broth at 300Cwith 106yeast cells perml and read at (105). It is interesting to note that inboth ofthese studies a
48 h, theyfound a higher mean amphotericin B MICfor the larger inoculum of the resistant isolate was required to
bloodstream isolates than for the control isolates from non- produce infection, implying a lowered virulence, and it is
patients, even though many of these patients had not re- this difference invirulence that makes the animal data from
ceived amphotericin B previously. Further, they found that the first study (136) hard to interpret. This findingofappar-
isolates for which the MIC was >0.8 ,ug/ml more often ently reduced virulence is not universal, however. When
caused fatal infections. While this study is limited by its Rogersetal. (127)andMcIntyreandGalgiani(85)testedone
retrospectivenatureanddifficultiesinclassifyingoutcome, a of the above-described resistant isolates in a rat model of
similar retrospective study also found a higher crude mor- fluconazoleorSCH39304therapy,respectively,therequired
tality among patients with bloodstream infection due to inoculum was identical to that needed for the susceptible
Candida isolates with an amphotericin B MIC of >0.78 strain ofC. albicans.
,ug/ml (17). Finally, the isolates from an AIDS patient with Radetsky et al. (122) collected 84 Candida isolates and
cryptococcalmeningitiswhorelapsedwhile onmaintenance determineddrugMICsinamicrobrothsystemwithavariety
fluconazole (119) andthen failed torespond to amphotericin ofmediaandantifungalagents, astartinginoculumof103per
B were studied for amphotericin B susceptibility by a ml, and incubation at 350Cfor 24 h. Use ofRPMI 1640was
macrobroth technique in yeast extract at 30°Cwith 5 x 105 found to produce a broad spectrum of MICs. The authors
yeast cells per ml and read at 24, 48, and 72 h. By compar- then present a small amount of data that correlates clinical
isonwith the pretreatment isolate, the amphotericin B MIC outcome with MIC results (using RPMI 1640) obtained for
376 REX ET AL. CLIN. MICROBIOL. REV.
the isolates from seven patients who hadbeen treatedwith a of MICs was quite narrow and not predictive of outcome
single antifungal agent (ketoconazole, 5-fluorocytosine, or (20).
amphotericin B). While a correlation appears to exist, insuf-
ficient data are presented to support a firm conclusion.
Because of its ready oral and intravenous administration, SUMMARY
fluconazole is widely used for therapy of candidiasis and
cryptococcosis. As with ketoconazole, reports of clinical The prospect for meaningful, reproducible antifungal sus-
resistance have begun to appear, and studies of some of ceptibilitytesting, atleastofCandida spp. andCryptococcus
theseisolates have demonstrated relative invitro resistance. neofonnans, appearsbright. Itisclear from awidevarietyof
Willocks et al. (152) described four AIDS patients with data that a correlation between invitro susceptibility testing
oropharyngeal candidiasis who failed fluconazole therapy. and in vivo efficacy is likely to exist when the proper test
The MICs for these patients' isolates as determined by an conditions are selected. Although the absolute MICs ob-
unspecified broth methodwere four- to eightfold higher than
tained vary with technique, the rank order of the isolates
those for a group of control isolates. Using the proposed
appears tobe constant andverytelling (18, 19, 51, 108, 152).
NCCLS reference methodology, Rodriguez-Tudela et al.
Thus, no matterwhat method is used, its range ofMICs will
(126) have brieflyreported on strains from sixAIDS patients
need to be correlated with animal model data or clinical
with oropharyngeal candidiasis who failed fluconazole ther-
outcomes: it is simply not possible to try to predict the
apy, and MICs for these isolates were 216 ,ug/ml. Finally,
elevated MICs for isolates from a group of AIDS patients meaning of a MIC on the basis of whether it is higher or
with persistent oropharyngeal candidiasis despite flucona- lowerthan achievable serumlevels ofthatdrug. To facilitate
zole therapy were found by a microtiter method, using HR cooperative testing and to allow analysis ofclinical isolates,
media (other details unspecified), and these isolates were a reproducible reference method is at hand. While cumber-
also morevirulent in an undescribed mouse model (39). some, itwill provide atouchstone duringcooperative studies
Of particular interest are the multiple reports of both todefine improved techniques. Several othermajor areas for
prophylactic and therapeutic failures of fluconazole against immediate work are apparent. First, large numbers of iso-
C. krusei (1, 2, 10, 23, 84, 106, 125, 153). Thelargest ofthese lates from well-designed clinical therapy trials need to be
studies is a retrospective study of 463 bone marrow trans- tested by the reference method, and the results need to be
plant or leukemia patients (153). There was a sevenfold- correlated with clinical outcome. Such studies are already
greater (8.3 versus 1.2%) incidence of bloodstream or vis- under way, and the results should be available in the near
ceralinfectionwithC. krusei in the 84patientswhoreceived future. As there is no guarantee that results obtained by the
fluconazole prophylaxis in comparison to the 355 patients proposed reference method will be predictive of clinical
whoreceivedotherformsofprophylaxis, includingketocon- outcome, variant procedures should also be examined. By
azole, miconazole, and amphotericin B, or no prophylaxis. designing variations that follow the spirit of the proposed
Thisfinding hasbeen reported in some (106, 141), but not all reference method, investigators are reasonably assured that
other, comparable studies (22, 58). Taken together, these a new method could be reproduced in other laboratories.
data strongly suggest thatwhile there maybe local epidemi- Second, it appears that the reference method may not be
ologic factors that affect the incidence ofC. krusei infection sensitive to amphotericin B resistance, and carefulworkwill
(131), C. kruseiisoftenresistant to fluconazole. While MICs be needed to resolve this issue. Third, the relevance of
for individual C. krusei isolates have been reported as MLCs ratherthan MICswill needtobe evaluated, and there
elevated by some authors, only one report directly compar-
will likely be many technical issues that must be resolved
ingthe MICs forC. kruseiwith those forother Candida spp.
before a reproducible method for MLC determination is
is available (81). It demonstrated higher MICs for C. krusei
available. Fourth, increasingly detailed studies ofthe mech-
thanforotherspecies. Comparabledataarenotyet available
anisms of antifungal resistance of selected isolates have
for the proposed NCCLS reference method.
become available (65, 76, 77, 89, 147), and such isolates
should prove to be useful for further validating any MIC
methodology. Finally, all of these studies may benefit from
Clinical Trials
the application ofmolecular genetic analysis. For example,
Casadevall et al. studied sequential isolates from five pa-
In only a few therapy trials have isolates been collected
tients with recurrent cryptococcal meningitis treated with
and tested and data correlated with clinical outcome, and
fluconazole (21). These authors found that the isolateswere
noneprovideusefulresults. In atrialof400versus 800mgof
clonally related and all had the same MIC by the proposed
ketoconazole per day for progressive forms of coccidioid-
omycosis, the isolates were tested by a broth dilution NCCLS reference method, leading the authors to conclude
method (55, 74). The MICs for all of the isolates were that the clinical failures in these patients were because of
similar, and no correlation with therapeutic outcome could changes in immune function or lack of compliance rather
be drawn. In a second study, 69 isolates from 50 of 52 than due to failure offluconazole therapy.
patients enrolled in a collaborative trial of ketoconazole as Susceptibility testing ofthe filamentous fungi is currently
therapyforblastomycosis (n = 21),coccidioidomycosis (n = poorlydeveloped. Standardizedmethodsfortestingofasex-
24), histoplasmosis (n = 9), nonmeningeal cryptococcosis (n ual conidia could probably be extrapolated from those used
= 7), andsporotrichosis (n = 8)weretestedbyamacrobroth for yeasts, but it may not be clinically relevant to test this
systemwithavarietyofmedia, aninoculumof2.5 x 104per phase of the organism. Tests of antifungal susceptibility of
ml, and incubation at 30°C for 48 to 96 h (134). The invitro hyphae need tobe developed. While such tests are likelyto
results were not predictive of clinical outcome, even when be technically challenging, correlation ofthese results with
isolates obtained from patients who had relapsed during testing ofconidia and other asexual propagules and clinical
therapy were tested. Finally, in a retrospective review of outcome could then take place, allowing an appropriate
cases ofsporotrichosis treatedwithketoconazole, the range testingprocedure tobe defined.