Vol. 33, No. 5ANTIMICROBIAL AGENTS AND CHEMOTHERAPY, May 1989, p. 726-7300066-4804/89/050726-05$02.00/0Copyright X) 1989, American Society for Microbiology

Interactive Effects of Antifungal and Antineoplastic Agents onYeasts Commonly Prevalent in Cancer Patients

MAHMOUD A. GHANNOUM,' MOHAMED S. MOTAWY,2 MOEEN A. ABU HATAB,1KHALID H. ABU ELTEEN,' AND RICHARD S. CRIDDLE3*

Department of Botany and Microbiology, Kuwait University,' and Kuwait Cancer Control Center,2 Kuwait, Kuwait, andDepartment of Biochemistry and Biophysics, University of California, Davis, California 956163

Received 19 August 1988/Accepted 13 February 1989

The effects of combinations of antifungal and antineoplastic drugs on inhibition of the growth of yeasts whichcommonly infect cancer patients have been analyzed. It was shown that (i) inhibitory drug combinations couldbe selected in which all drugs were at levels far below their individual MICs; (ii) interactive effects amongantineoplastic and antifungal drugs may be very large; (iii) optimum combinations of drugs for inhibition ofyeast growth depended upon both the relative and absolute concentrations of the drugs in the mixture; (iv) drugcombinations which were effective at low levels in inhibiting one test yeast were also generally effective againstother species, but the levels of susceptibilities and, to a lesser extent, the best ratios of drugs in the testcombinations varied with species; and (v) to quantitatively evaluate drug interactions, it is necessary tocarefully define and control all experimental conditions, absolute and relative concentrations of drugs used, andthe organisms tested.

Immunosuppressed patients receiving antineoplastic che-motherapy are specifically susceptible to opportunistic in-fections (6). Treatment of microbial infections requires ei-ther a discontinuation of cancer chemotherapy or the use ofdrugs in combinations which can simultaneously treat boththe neoplasm and the infection. Selection of drugs forcombined treatments has been difficult because of interac-tive effects among test drugs which may enhance or inhibitthe action of each drug alone. Moreover, antimicrobialdrugs, and particularly antifungal agents, may have severetoxic side effects on the host (2). New regimens for com-bined drug treatment must be developed to maximize theireffectiveness.A major goal of studies of combined-drug therapy for

fungus-infected cancer patients is to define mixtures of drugswhich mutually potentiate inhibitory effects on fungalgrowth so that minimal levels of the toxic drugs may be used.This must include not only studies with mixtures of anti-fungal drugs but also evaluations of the antifungal effects ofthe antineoplastic drugs and of combinations of antifungalplus antineoplastic agents. Selection of the best regimens forcombined-drug treatment must rely upon understanding in-teractive effects among the antifungal and antineoplasticdrugs. The number of possible combinations of drugs whichshould be evaluated for treatments thus becomes very large.There are numerous reports of studies of the effects of

combined antifungal agents on the inhibition of Candidaspecies (7, 9). Many of these studies offer encouragingevidence that combined-drug treatments can synergisticallyaffect fungal growth, although the methods used in suchstudies fall short of allowing predictions of optimum condi-tions for inhibition. Little work has been done to defineoptimum conditions for combined antineoplastic and anti-fungal drugs in inhibiting fungal growth.The most frequently isolated yeast species in cancer

patients are Candida albicans, C. tropicalis, C. glabrata, C.parapsilosis, and C. krusei. These accounted for 97.1% ofthe isolates obtained from cancer patients in a study by

* Corresponding author.

Kiehn et al. (5) at Memorial Sloan-Kettering Cancer Center.C. albicans and C. tropicalis accounted for 97.2% of all yeastspecies isolated from cancer patients treated at KuwaitCancer Control Center (4). In addition, 12 other species havebeen less frequently isolated (5).

In this study, we report the results of multifactorial studiesof combinations of antifungal and antineoplastic drugs usedto inhibit the growth of seven yeast species, which includespecies primarily responsible for infections of cancer pa-tients (5). The primary and interactive effects among thesedrugs are quantified, and drug concentrations which arehighly effective in inhibiting fungal growth in vitro aredefined. Effective inhibition was obtained in some trials atdrug levels far below the MICs of individual drugs, suggest-ing combinations that may be selected for their minimal toxicside effects in treatment of infected patients.

MATERIALS AND METHODSOrganisms. Seven species of yeasts were studied: C.

albicans KCCC 14172, C. tropicalis KCCC 13622, C. parap-silosis KCCC 14275, C. krusei KCCC 15312, C. kefyr KCCC13709, C. glabrata KCCC 14801, and Trichosporon cuta-neum KCCC 15286. These yeasts were oral isolates obtainedfrom patients undergoing therapy at the Kuwait CancerControl Center, except T. cutaneum, which was a bloodisolate. Methods of isolation and identification of thesespecies were as described previously (4).

Drugs. The following drugs were used in this study:methotrexate (MT) (Cyanamid International Corp., Basel,Switzerland), cyclophosphamide (CP) (Asta-Werke AG-Chemische Fabrik, Brackwede, Federal Republic of Ger-many), 5-fluorouracil (5-FU) (Hoffmann-La Roche and Co.Ltd., Basel, Switzerland), amphotericin B (AB), flucytosine(FC), and miconazole nitrate (MN). Antifungal agents werepurchased from Sigma Chemical Co., St. Louis, Mo.

Analysis of drug interactions. Tests of multiple drug effectson the growth of seven yeast species were made by using 2"factorial design studies or three-level Box-Behnken studies(1), as reported previously (3). These studies were expandedto include seven test organisms. Linear regression equations

726

by on October 5, 2009

aac.asm.org

Dow

nloaded from

DRUG INTERACTIONS IN ANTIFUNGAL THERAPY 727

TABLE 1. Tests of three-drug combinations in inhibition of yeast growth

Concna Endpoint dilution levelb for:Trial

5-FU MT CP AB MN FC C. kruesi C. parapsilosis T. cutaneum C. tropicalis C. glabrata C. kefyr C. albicans Avg

A1 - - - 3 0 3 0 4.5 4.5 0 2.12 + - - 5.5 1.5 5 5 6.5 6.5 3 4.73 - + - 3 0 2.5 0 3 3 0 1.64 - - + 3 0 2 0 3 3.5 0 1.65 + + - 4.5 1.5 4.5 1 6.5 7 0 3.66 + - + 4.5 2 4 1 5.5 6.5 1 3.57 - + + 1.5 0 2.5 0 2.5 2.5 0 1.38 + + + 4.5 1.5 3 0 4.5 4.5 0 2.69 0 0 0 3 1 3 0.5 4.5 4.5 0.5 2.1

B1 - - - 0 2 2 0 1 4 0 1.32 + - - 1 2 5 1 2 7 1 2.73 - + - 0 2 2 1 1 4 1 1.64 - - + 1 4 2 3 3 7 2 3.15 + + - 1 3 4 2 1 6 1 2.66 + - + 2 4 4 4 3 8 3 4.07 - + + 1 4.5 1 2 2 7 2 2.88 + + + 1 3 3 1 2 8 2 2.99 0 0 0 1 3 3 2 2 6.5 1.5 2.7

a Drug concentrations (in micrograms per milliliter) are as follows (+, 0, - given for each drug): 5-FU, 312, 168, and 78; MT, 1,800, 848, and 400; CP, 1.250,559, and 250; AB, 0.488, 0.0617, and 0.015; MN, 3.9, 0.974, and 0.24; and FC, 24.0, 6.58, and 1.8.

b Fractional dilution levels (e.g., 4.5) for endpoint dilutions are calculated as the averages from replicates of individual samples and are not meant to imply somefractional dilution process.

were developed to describe yeast growth responses to threedrug combinations. Tests of this linear model indicated thatit provided an adequate prediction of yeast responses overthe drug concentration ranges considered (3). Data forfour-factor studies were also fit with first-order polynomials(3).

All inhibition tests were run in random order, and deter-minations were all replicated two or three times. Center testpoints for analysis of fit of equations were replicated fourtimes. Estimates of experimental error were made frompooled standard deviations for all replicated samples. Theeffects of changing various experimental variables in theassay conditions on the levels of drugs required for inhibitionwere evaluated by the Plackett-Burman screening technique(10).The seven clinical yeast isolates were inoculated (by using

a Dently multipoint inoculator) onto agar plates containingtwofold serial dilutions of each test drug combination. Stocksolutions of drugs were prepared as described previously (3).Results were scored as dilutions which caused zero visiblegrowth.

RESULTS

Combinations of three antitumor drugs, 5-FU, MT, andCP, and of three antifungal drugs, AB, MN, and FC, weretested for their abilities to inhibit the growth of seven speciesof yeasts (Table 1). To test a wide range of concentrationsand ratios of drug concentrations in a manageably smallnumber of experiments, we used a 23 factorial design forsimultaneously varying concentrations of each of the anti-fungal or antineoplastic drugs. The drug levels used as theinitial concentrations for each experimental point are shownin Table 1. High levels of the drugs (MIC levels) were codedas +, low levels were coded as -, and medium levels werecoded as 0. The actual drug concentrations corresponding to

these coded values are presented in Table 1, footnote a.Values representing the maximum dilution of each test drugcombination which caused zero visible growth of the variousyeasts are also shown in Table 1. The actual concentrationsof each drug in these mixtures at the lowest levels causinginhibition can be obtained by dividing the appropriate + or -concentration by 2', where n is the value of the endpointdilutions (Table 1). For example, inhibitory concentrationsin trial 1A for C. krusei by antineoplastic drugs were asfollows: 5-FU, (78/23) = 9.75 ,ug/ml; MT, 400/23 = 50 ,ug/ml;and CP, 250/23 = 36.25 ,ug/ml.

Susceptibilities of the organisms to the test combinationsdiffered widely. C. glabrata and C. kefyr were both verysusceptible to inhibition by the antineoplastic drugs as wellas the antifungal drugs, whereas the species of T. cutaneum,C. krusei, C. tropicalis, C. parapsilosis, and C. albicanswere generally more resistant. Yeast susceptibilities to thedrugs varied with the particular drug combination used, andeach individual species tested showed a unique pattern ofinhibition dependent upon the absolute as well as the relativeconcentrations of the three drugs (Table 1).

Qualitative indications of drug interactions are immedi-ately apparent from the data in Table 1. If there were nopositive or negative interactions between drugs, the combi-nation in which all were initially at their highest concentra-tions (i.e., +, +, +) should be inhibitory at the highestdilution. Clearly, this is not uniformly the case. Further-more, if all the yeasts responded similarly to the drugcombinations, the pattern of the particular mixtures of drugsthat are more or are less effective would be expected to bethe same for all species tested. This was not observed. Whenantineoplastic drugs are used, the mixture (+, -, -) (i.e., aninitial mixture high in 5-FU but low in MT and CP) was mosteffective in the inhibition of several of the species, butdifferent drug combinations worked as well if not better forother species.

VOL. 33, 1989

by on October 5, 2009

aac.asm.org

Dow

nloaded from

ANTIMICROB. AGENTS CHEMOTHER.

TABLE 2. Linear regression equation for three-factor drug inhibition'

Drugs and yeast FocC C0 Curva- 2species (measured) ture

Antineoplastic drugsC. krusei a 3.610 1.062 -0.312 -0.312 0.063 0.063 -0.063 0.312 3 0.61 0.965

b 0.733 0.0185 0.0016 0.0038 0.00001 0.00002 0.000003 0C. parapsilosis a 0.883 0.812 -0.063 0.063 -0.063 0.063 -0.063 -0.063 0 0.83 0.994

b -1.153 0.0099 0.00074 0.0024 -0.000006 -0.000007 0 0T. cutaneum a 3.282 0.812 -0.188 -0.438 -0.188 -0.188 0.063 -0.188 3 0.28 0.989

b 1.299 0.0156 0.00076 0.0032 -0.00001 -0.00002 -0.000002 0C. tropicalis a 0.833 0.875 -0.625 -0.625 -0.625 -0.625 0.375 0.375 0 0.83 0.994

b -3.224 0.0347 0.0017 0.0035 -0.00002 -0.00003 -0.000002 0C. glabrata a 4.500 1.254 -0.375 -0.625 0.125 -0.125 0 0.250 3.5 1.0 99.9

b 2.934 0.0158 0.0018 0.0042 -0.00001 -0.00002 0 0C. kefyr a 4.715 1.382 -0.500 -0.500 0.125 -0.125 -0.250 -0.375 4.5 0.22 99.7

b 2.818 0.0145 0.00011 -0.0047 -0.00008 -0.00002 -0.000003 0C. albicans a 0.500 0.500 -0.500 -0.250 -0.500 -0.250 0.250 0.250 0 0.5 0.99

b -1.545 0.0193 0.00085 0.00088 -0.00001 -0.00001 0 0

Antifungal drugsC. krusei a 0.889 0.375 -0.125 0.375 -0.125 -0.125 -0.125 -0.125 0 0.89 0.995

b 0.213 1.507 -0.0642 0.0349 0.163 0.0246 0.00235 -0.0296C. parapsilosis a 3.058 -0.063 0.063 0.813 -0.063 -0.313 -0.188 -0.313 2.5 0.56 0.999

b 2.254 -1.085 -0.116 0.289 1.480 0.141 0.0221 -0.121T. cutaneum a 2.89 1.13 -0.375 -0.375 -0.125 -0.125 -0.125 -0.125 2 0.89 0.995

b 1.95 6.89 0.0495 -0.0963 -1.882 -0.304 -0.0374 0.128C. tropicalis a 1.78 0.250 -0.250 0.750 -0.250 -0.250 -0.750 -0.250 2 -0.22 0.995

b 0.348 0.646 0.165 0.0725 0.982 0.155 -0.0071 -0.111C. glabrata a 2.00 1.250 -0.500 0.500 -0.250 -0.150 -0.500 -0.250 1.5 0.50 0.999

b 1.37 1.054 -0.109 -0.0167 0.858 0.136 0.0159 -0.0824 0C. kefyr a 6.39 0.875 -0.125 1.13 -0.125 -0.275 0.125 0.125 5.5 0.89 0.997

b 4.57 4.99 -0.184 0.063 0.343 -0.0422 0.0177 -0.0352C. albicans a 1.50 0.250 0 0.750 -0.250 0 -0.250 0 0 1.5 0.999

b 0.283 1.081 0.182 0.0293 0.258 0.134 0.00354 -0.057

aMinimum significant coefficient values (95% confidence interval): a = 0.19; b = 0.17. minimum significant curvature (95% confidence interval): a = 0.21; b- 0.37. Regression equations are as follows: for antineoplastic drugs, y = C0 + C1(5-FU) + C2(MT) + C3(CP) + C4(5-FU)(MT) + CS(5-FUXCP) + C6(MT)(CP)+ C7(5-FU)(MT)(CP); for antifungal drugs, y = C0 + C1(AB) + C2(MN) + C3(FC) + C4(AB)(MN) + C5(AB)(FC) + C6(MN)(FC) + C7(AB)(MN)(FC).

Although each of the seven test species responded differ-ently to changes in the drug mixtures, a consistent trend wasnoted among all species, which allowed a definition of betteror poorer combinations for inhibition. This is reflected by theaverages listed in Table 1. Thus, it is possible to selectparticular ratios of drugs in combinations which have gener-ally high levels of toxicity to all species tested.Using the endpoint determination data in Table 1, we

wrote general linear regression equations to describe theinhibition results quantitatively. These equations are pre-sented in two forms in Table 2 for each organism tested. Thefirst fits concentrations in coded form (-1 to +1), and thesecond is converted for use with actual drug concentrations.The value of y is the number of twofold dilutions which canbe made for a test mixture while still obtaining zero visiblegrowth. The magnitudes of the coefficients CO to C7 in theregression equations indicate the magnitudes of primary andinteractive effects among the drugs. The sign of each coef-ficient indicates whether the term has a beneficial (+) or anegative (-) effect on inhibition. For example, from thenegative signs of the coefficients C2 in Table 2, it is evidentthat the primary effect of increasing the levels of MT in theconcentration ranges tested is to decrease the level ofinhibition of all the test species by the drug mixture. Theprimary effects of 5-FU and CP are both positive over therange tested; the two-factor interactive terms are all negativeor not significantly different from zero.The validity of the regression equations in describing drug

responses over the entire range of concentrations was tested

with an R2 test and with the curvature parameter shown inTable 2, in which measured values for inhibition by thecenter point mixture of drugs (CO) are compared with valuespredicted by the equations (3). In studies with antineoplasticdrugs, the curvature values are in all cases between 0.22 and1.0. The minimum significant curvature is 0.19. A value of+1 in curvature indicates that the dilution value predictedfrom the regression equation differs from the measured valueby one dilution. Thus, even when the test results with thelargest curvature are considered, predictions based on theregression equations can be made to within one dilution.Since the predicted values are at the center point in theconcentration ranges tested (i.e., farthest from the testpoints), they would be expected to contain the largest error.It is apparent, for practical applications, that the linearequations quite accurately describe the responses of theyeasts to the three antineoplastic drug combinations and canbe used to predict responses at any selected combination inthe concentration ranges tested.The yeast growth responses to the antifungal drugs are

similarly described quite well by the regression equations(Table 2). Significant coefficients for MN are positive, andfor FC and AB they differ with the species. The curvaturevalues (3), while significant compared with the>minimumsignificant curvature of 0.37, are again small (0.22 to 1.0) andare well within the sensitivities required for most practicalapplication of the equations to prediction of drugs inhibitionsfor laboratory test studies.

Resultsof tests of inhibition in which combinations of two

728 GHANNOUM ET AL.

by on October 5, 2009

aac.asm.org

Dow

nloaded from

DRUG INTERACTIONS IN ANTIFUNGAL THERAPY 729

TABLE 3. Effects of four-drug combinations on growth of yeastsa

Concnb Dilution showing complete inhibition of growth of:Trial

MN AB 5-FC MT C. krusei C. parapsilosis T. cutaneum C. tropicalis C. glabrata C. kefyr C. albicans

1 - 3 2 4 4.5 4 7 4.52 + - - - 4 4 5 5.5 5 7.5 5.53 - + - - 3 2.5 5 4 5 7 54 + + - - 3 5 5 5.5 5 7 5.55 - - + - 5 3 5.5 4 6.5 8.5 46 + - + - 5.5 5 6.5 6.5 7 9 6.57 - + + - 4.5 3 5 3 7 8.5 4.58 + + + - 5 5 5.5 6.5 6.5 8.5 69 - - + + 3 2.5 4.5 4 4 6.5 410 + - - + 3.5 4 4.5 6 4.5 7.5 611 - + - + 3.5 3 3.5 4 4 6.5 3.512 + + - + 3.5 5 5 5.5 5 7.5 513 - - + + 5 3.5 5.5 4 7 8.5 4.514 + - + + 5.5 5 6 7 7.5 9.5 615 - + + + 5.5 3 5.5 6 6.5 9 516 + + + + 6 5 6 7.5 7 9.5 6

a Minimum significant coefficient values (95% confidence interval): a = 0.19; b = 0.17; minimum significant curvature (95% confidence interval): a = 0.21; b= 0.37.

b For concentrations corresponding to +, 0, and -. See Table 1, footnote a.

antifungal plus two antineoplastic agents were used arepresented in Table 3. In this experiment we used a two-levelfactorial design with high (+), low (-), and mid-level (0)concentrations in cornbinations as shown in the table. The+, -, and 0 levels here are the same as shown in Table 1. Afit of the data to a quadratic model was also tested by usinga three-level Box-Behnken design; the linear polynomialmodel, which required fewer tests, provided a somewhatbetter fit of the data over the concentration ranges studied(3).

Sixteen drug combinations were tested, again in a fullyrandomized and replicated fashion. Strong interactions werenoted between antineoplastic drugs, between antifungaldrugs, and among the combinations of antineoplastic plusantifungal drugs. Considering all the trial points, the combi-nation of drugs which is most generally effective for all thespecies (i.e., lowest drug concentrations which were inhibi-tory) was combination no. 6 (Table 3). In this mixture, initialconcentrations were coded (+, -, +, -), i.e., high levels ofMN and 5-FU, with low levels of AB and MT. This combi-nation at a dilution of 25 completely inhibited all the teststrains; at 26 it inhibited five of seven. The actual concen-trations of drugs at each of these inhibitory levels are readilydetermined by inserting concentrations corresponding to thecoded values and calculating dilutions. For example, the (-)level of AB is 0.015 ,ug/ml. The concentration of AB insolutions which inhibit the growth of five of seven testorganisms is 0.015/26 = 2.3 x 10-4 ,ug/ml; for MN, 5-FU,

and MT, the concentrations are 0.061, 28, and 3.9 jig/ml,respectively.Data from the test points in Table 3 were analyzed to

produce polynomial equations which adequately fit the dataand hence allow predictions of inhibitory levels of variouscombinations of these drugs which have not been experi-mentally tested. A polynomial of the form y = bo + b1(Xl) +b2(X2) + b3(X3) + b4(X4) + b5(XlX2) + b6(X1X3) + b7(XlX4)+ b8(X2X3) + bg(X2X4) + b10(X3X4) + bl(XlX2X3) +b12(XlX2X4) + b13(XlX3X4) + b14(X2X3X4) was developed todescribe responses of each test yeast to the drugs. As withthe three-factor studies, the coefficients b1 to b14 indicate themagnitude and direction of individual or interactive effects ofthe drugs on growth and Xl, X2, X3, and X4 correspond toconcentrations (coded) of MN, AB, 5-FU, and MT, respec-tively. The values of the coefficients are summarized inTable 4 for each test organism.

Table 5 summarizes the effects of changes in assay condi-tions on the apparent inhibitory levels of drug combinations.Major effects are noted with changes in medium, assaytemperature, and inoculum size. Not all species were af-fected equally by these parameters. Note that C. krusei isinsensitive to the medium change and C. tropicalis is notsignificantly affected by incubation temperature differences.

DISCUSSIONThere has been considerable discussion in the literature

regarding methods for measuring drug inhibition of fungal

TABLE 4. Polynomial equations describing four-factor effects on yeastsa

Species C0 C1 C2 C3 C4 C5 C6 C7 C8 C9 CIO C11 C12 C13 C14 R2

C. krusei 4.28 0.219 -0.031 0.969 0.156 -0.094 0.031 -0.031 0.031 0.219 0.094 0.094 0.031 0.031 0.031 0.991C. parapsilosis 3.78 0.969 0.156 0.281 0.094 -0.094 -0.031 -0.094 -0.219 -0.031 -0.031 -0.031 -0.031 -0.031 -0.031 0.999T. cutaneum 5.13 0.313 0.063 0.563 -0.063 0 0 0 -0.125 0 0.125 -0.063 0.185 -0.063 0.188 0.971C. tropicalis 5.22 1.03 0.031 0.344 0.281 -0.031 0.281 -0.031 0.156 0.219 0.281 -0.031 -0.281 -0.156 0.219 0.995C. glabrata 5.78 0.156 -0.031 1.09 0.031 -0.031 -0.031 0.031 -0.094 -0.156 0.094 -0.094 0.219 0.094 0.219 0.993C. kefyr 7.97 0.281 -0.031 0.906 0.094 -0.094 -0.031 0.156 0.031 0.094 0.156 -0.031 0.031 -0.031 0.031 0.999C. albicans 5.09 0.719 -0.031 0.219 -0.093 -0.156 0.094 0.031 0.094 -0.094 0.156 -0.031 0.031 -0.219 0.156 0.998

a The equation is as follows: y = C0 + C1(MN) + C2(AB) + C3(5-FU) + C4(MT) + C5(MN)(AB) + C6(MN)(5-FU) + C7(MN)(MT) + C8(AB)(5-FU) +Cg(AB)(MT) + C1O(5-FU)(MT) + CjI(MN)(AB)(MT) + C12(MN)(AB)(CP) + C13(MN)(MT)(CP) + C14(AB)(MT)(CP). Pooled standard deviation, 0.21; minimumsignificant factor effect, 0.17.

VOL. 33, 1989

by on October 5, 2009

aac.asm.org

Dow

nloaded from

ANTIMICROB. AGENTS CHEMOTHER.

TABLE 5. Effects of variables in assay conditionson drug inhibition

Effecta for:Variable

C. albicans C. tropicalis C. glabrata C. krusei

Medium 1.55* 1.70* 4.6* 0.3pH -0.75 -0.80 0.8 0.3Time -0.75 -1.20 -2.2* -1.4*Temp 1.75* 0.90 0.7 2.6*Inoculum -2.05* -2.8* -2.4* -0.5Solvent 0.45 0.50 0.5 -0.3Tween 80 0.45 0.60 0 0.1BSAb -0.15 -0.5 1.0 0.2

a *, Significant effects.b BSA, Bovine serum albumin.c Minimum significant factor effects for C. albicans, C. tropicalis, C.

glabrata, and C. krusei are 0.90, 1.4, 1.5, and 1.0, respectively.

growth, particularly with respect to defining synergism andantagonism. Various scoring methods have been used toanalyze colony growth, viable cell counts, or turbidity.Inhibition levels have been cited as concentrations of drugswhich reduce growth to zero or to some defined percentageof that of controls. A major problem has been the wideconcentration range of some drugs giving only partial inhi-bition. This has led to uncertainties, particularly in drugcombination studies (8). Several careful analyses have beenmade, employing more than one criterion for inhibition ormonitoring growth over the entire time of contact betweendrugs and cells, to reduce the uncertainties (9).The goal of this project was to find drug combinations

which could, at low concentrations, completely inhibit yeastgrowth. Accordingly, a drug mixture was scored as inhibi-tory in these studies only at levels at which zero growth wasobservable. Decreased growth rates and partial inhibitionwere not considered. Complete inhibition of growth seems aproper scoring criterion in seeking drug combinations whichmay be able to block yeast infections or clear up existinginfections in host organisms. This scoring method offersseveral advantages experimentally. For example, it is notnecessary to make repeated measurements of growth versustime or to integrate dose-time effects.Our studies with either three antineoplastic drugs or three

antifungal drugs showed that each yeast species respondedsomewhat differently to the varied drug combinations, butthat there was a general consistency among responses. Acombination which was effective in inhibiting the growth ofone yeast species at low concentrations was generally one ofthe better combinations for effective inhibition of the otherspecies. This study has shown that it is possible to readilyselect drug combinations which are toxic to yeast strains atlevels far below the MICs of any individual drugs in thecombination.To make this selection, we developed polynomial equa-

tions. Polynomial expressions for each organism can besolved, subject to any imposed constraints, to select the bestmixture of drugs for growth inhibition. For example, if aconstraint is imposed that the AB concentration in themixture is not to exceed 0.002 ,ug/ml, one could select thebest mix of the other three drugs to kill the test orgapism. Apractical application might be to set 5-FU and MT at levelscommonly used clinically in animal chemotherapy and usethe equations in Table 4 to find minimal levels of AB or MNwhich would block growth in the presence of these clinicallevels of the antineoplastic drugs. Selection of the bestmixture of drugs for inhibiting yeast growth depends criti-

cally upon the definition of best. One definition with somepossible clinical relevance might be to have the total druglevels as low as possible to avoid toxic side effects. Anothermight be to aim at minimizing levels of one particular drug.Still a third might be to fix one or more drugs, e.g., theantineoplastic drugs, at a clinical level and select minimaleffective levels of the antifungal agents (as a prophylacticantifungal treatment). A fourth definition might be to selectdrug combinations that have a wide spectrum of activity forprophylactic treatments, i.e., to select a drug combinationwhich can inhibit a high percentage of possible infectiveyeasts. The best drug combinations for each of these scenar-ios can be determined by solution of the polynomial equa-tions with appropriately defined limitations on the term best.Although it has long been known that drug synergism is

possible, the critical dependence of synergistic effects notonly upon the drugs selected but also upon their concentra-tion ratios has not previously been emphasized. Obtainingsuch conclusions became possible only when multifactorialanalysis studies were used. It is evident that future experi-ments aimed at optimum combination drug therapy mustsimultaneously treat both absolute and relative concentra-tions of drugs. It would also appear that three- and four-drugcombinations may in some cases prove far superior totwo-drug combinations. It is evident that simple statements(such as drug A is synergistic with drug B) are no longeracceptable. Test conditions, absolute and relative concen-trations, and organisms tested must be clearly defined.

ACKNOWLEDGMENTS

We thank the Ministry of Public Health, Kuwait, for gifts ofdrugs. We also thank J. M. Al-Hassan for his encouragement andsupport and A. Ibrahim for his technical assistance.

This work was supported by Research Council, Kuwait Univer-sity, grants S0031, S0038, S0024, and SBO-17 and by KuwaitFoundation for the Advancement of Science grant 86-4-01.

LITERATURE CITED1. Box, G. E. P., and D. W. Behnken. 1960. Some new three level

designs for the study of quantitative variables. Technometrics2:455-475.

2. Garrod, L. P., H. Lambert, F. O'Grady, and P. Gaterworth.1981. Antibiotic and chemotherapy, 5th ed., p. 238-250.Churchill Livingstone, Edinburgh.

3. Ghannoum, M. A., M. S. Motawy, M. A. Abu Hatab, A. S.Ibrahim, and R. S. Criddle. 1989. Multifactorial analysis ofeffects of interactions among antifungal and antineoplastic drugson inhibition of Candida albicans growth. Antimicrob. AgentsChemother. 33:717-725.

4. Ghannoum, M. A., M. S. Motawy, A. L. Mubarak, and H. A.Al-Awadhi. 1985. Incidence and susceptibility of yeasts isolatedfrom cancer patients in Kuwait. Mykosen 28:619-624.

5. Kiehn, T., F. Edwards, and D. Armstrong. 1980. The prevalenceof yeasts in clinical specimens from cancer patients. Am. J.Clin. Pathol. 73:518-521.

6. Meyers, F. H., E. Jawetz, and A. Goldfien. 1976. Review ofmedical pharmacology, p. 601-605. Lange Medical Publica-tions, San Mateo, Calif.

7. Montgomerie, J. Z., J. E. Edwards, and L. B. Guze. 1975.Synergism of amphotericin B and 5-fluorocytosine for Candidaspecies. J. Infect. Dis. 132:82-86.

8. Odds, F. C. 1979. Problems in the laboratory assessment ofantifungal activity. Postgrad. Med. J. 55:677-680.

9. Odds, F. C. 1982. Interactions among amphotericin B, 5-fluorocytosine, ketoconazole, and miconazole against patho-genic fungi in vitro. Antimicrob. Agents Chemother. 22:763-770.

10. Plackett, R. L., and J. P. Burman. 1946. The design of optimummultifactorial experiments. Biometrika 33:305-325.

730 GHANNOUM ET AL.

by on October 5, 2009

aac.asm.org

Dow

nloaded from