JOURNAL OF CLINICAL MICROBIOLOGY, July 1986, p. 1-60095-1137/86/070001-06$02.00/0Copyright © 1986, American Society for Microbiology

Selection of a Reference Lot of Mueller-Hinton AgarHELEN M. POLLOCK,'* ARTHUR L. BARRY,' THOMAS L. GAVAN,3 PETER C. FUCHS,4 SHARON HANSEN,5

CLYDE L. THORNSBERRY,6 HARRY FRANKEL,7 AND SARAH B. FORSYTHE8tDepartment of Pathology, University of South Alabama, Mobile, Alabama 36617'; The Clinical Microbiology Institute,

Inc., Tualatin, Oregon 970622; Departments of Microbiology3 and Biostatistics,8 The Cleveland Clinic Foundation,Cleveland, Ohio 44106; Department of Pathology, St. Vincent's Hospital, Portland, Oregon 972254; Laboratory Service,Veterans Administration Medical Center, Baltimore, Maryland 212185; Clinical Bacteriology Section, Centers for Disease

Control, Atlanta, Georgia 303336; and Pfizer Pharmaceuticals, Inc., New York, New York 100177

Received 14 January 1986/Accepted 18 March 1986

A collaborative study was undertaken to evaluate the performance of currently marketed Mueller-Hintonagars from seven manufacturers by replicate disk diffusion tests with standard quality control strains.Identification of the manufacturers was concealed, and the resulting data were evaluated for the selection of a

physical reagent standard against which the performance of future production lots would be tested and madeto conform. A medium was selected which was sufficiently close to existing National Committee for ClinicalLaboratory Standards quality control limits that current interpretive criteria would require minimummodification. Two of the seven lots were eliminated from further consideration because the final pHs were

outside acceptable limits. The remaining four lots had 96% of mean zone diameters '2 mm from those of thechosen lot and 65% of the means were <1 mm from those of the chosen lot for all 28 antimicrobialagent-organism combinations. Manufacturers then attempted to produce new lots of Mueller-Hinton agarwhich performed within the prescribed limits of the chosen lot. One lot performed in close conformity with theselected standard, but the overall performance of the media was essentially the same as that of the randomlychosen lots in the initial study. It was concluded that one of the original seven lots demonstrated propertieswhich made it a tentative candidate for a physical reagent standard and that the use of a physical reagentstandard in evaluating production lots might aid in stabilizing the performance of Mueller-Hinton agar.

The Bauer-Kirby disk diffusion susceptibility test wasdeveloped to provide a simple routine procedure whichwould reliably predict the efficacy of a given antibioticregimen in the treatment of a specific infectious disease (3).Many technical factors affect the size and clarity of zones

around the disks, including inoculum density, temperature,depth of agar, disk potency, reading of zones, and medium(1, 2, 4, 5, 7-14, 16-20).

Ericsson and Sherris reported results from a large collab-orative study in Europe, which included a number of insti-tutions in different countries (8). Variations in results werefound with all antibiotics, especially tetracycline. Althoughthe methods and media used in these laboratories were notalways the same, the study pointed to a need for improvedstandardization of the methods used in disk diffusion sus-ceptibility testing.During the past decade a great deal of professional time

has been expended to develop a standard method whichwould take into account the technical factors that contributeto variable results (15). These efforts included developmentof quality control limits for standardized control strains anddelineation of interpretive standards. As these standardshave been used, it has become apparent that the originalquality control limits did not adequately describe the per-formance of media in the field (12).

Variations in Mueller-Hinton agar became most apparentwith tetracyclines and aminoglycoside antibiotics (5, 9, 11,16). Cation content contributed to many of these variations(14), and some workers have suggested that the magnesiumand calcium content of Mueller-Hinton agar be adjusted to a

predetermined level at the time that powder lots are manu-

* Corresponding author.t Present address: 1238 Armacost Ave., Los Angeles, CA 90025.

factured (17). Subsequent studies revealed that this was nota practical approach since the cation contribution by the agaris unpredictable (11). In addition, it appeared that only theavailable soluble cation interfered with antibiotic activity,not the total cation content. The addition of one cation mayunpredictably move more of another cation into solutionfrom the agar matrix (11). Recent attempts to reduce thethymidine content of Mueller-Hinton agar for clearersulfonamide and trimethoprim zones had an inhibitory effecton the growth of some staphylococci. Preliminary evidencesuggested that this phenomenon may also be related toproblems in detecting resistance to penicillinase-resistantpenicii4ins in staphylococci.More subtle differences began to emerge which were

reflected in the performance of different media with specificclasses of antibiotics. These variations appeared to be man-ufacturer related, with each manufacturer producing lotswith a particular pattern of performance for each class ofantibiotics. With cephalosporins and penicillins, media fromsome manufacturers gave results that were consistently atthe low extreme of published quality control limits with thestandard strains; others were at the upper extreme of theacceptable range. It was not unusual for an antibiotic to givea zone 1 to 2 mm outside the quality control range. Studieswere undertaken through the National Committee for Clini-cal Laboratory Standards (NCCLS) consensus mechanismto adjust the quality control ranges to better reflect thespectrum of performance given by lots of Mueller-Hintonagar that were being marketed and develop a new perform-ance test for evaluating newly manufactured lots. The needfor tighter control on medium performance seemed appar-ent.The study was designed to determine the extent of varia-

tion in the performance of Mueller-Hinton agar lots currently

1

Vol. 24, No. 1

on May 6, 2020 by guest

http://jcm.asm

.org/D

ownloaded from

2 POLLOCK ET AL.

available in the marketplace and the relationship of thatperformance to published control limits. Since the number ofvariables inherent in performing the test was high, this studywas undertaken to determine whether it was possible toselect an ideal lot from those in production to serve as aphysical reagent standard against which the performance offuture lots of Mueller-Hinton agar could be compared.Subsequently, a trial was undertaken to determine whethernewly manufactured lots could be adjusted to perform sim-ilarly to the chosen standard.Performance criteria for selecting a physical reagent stan-

dard were set up along with a test protocol for defining areference medium. The results of these studies are reportedhere.

MATERIALS AND METHODS

Standard strains. New vials of lyophilized quality controlstrains of Staphylococcus aureus ATCC 25923, Escherichiacoli ATCC 25922, and Pseudomonas aerguinosa ATCC27853 were obtained from the American Type Culture Col-lection. There was one set per participating laboratory. Asupplementary study was performed by one of us (A.L.B.)with Streptococcus faecalis ATCC 29212 and ATCC 33186and two methicillin-resistant strains of Staphylococcus au-reus chosen for their particular sensitivity to the influence ofagar media in detecting resistance to the penicillinase-resistant penicillins by disk diffusion.

Media. Powdered Mueller-Hinton agars (20 kg of each)were donated by the seven manufacturers that provideMueller-Hinton agar for sale in the United States. Theseincluded Acumedia, Baltimore, Md.; BBL MicrobiologySystems, Cockeysville, Md.; Difco Laboratories, Detroit,Mich.; Gibco Laboratories, Madison, Wis.; Inolex SpectrumDiagnostics, Glenwood, Ill.; Oxoid, Columbia, Md.; andScott Laboratories, Fiskeville, R.I. All lots of agar wereshipped by each manufacturer to the Centers for DiseaseControl (CDC).

Distribution of media. Bottles for transporting media lotswere donated by Difco Laboratories and shipped to the CDCBiologic Products Division. Through the courtesy of MorrisSuggs and R. Knox Harrell, this division served as thecollection and distribution point for all media lots. Eachmanufacturer submitted 20 kg of a representative lot fortesting. At CDC these lots were repackaged in the uniformbottles donated by Difco Laboratories and given a codenumber. Each participating laboratory had a different set ofcode numbers to ensure that investigators could not compareresults. In addition, the name of the manufacturer of eachcoded lot was not revealed to any of the investigators orcommittee members. The coded powder lots were shipped toeach of the five participating laboratories from CDC.

Disks. Disks were provided through the courtesy of BBLMicrobiology Systems. All disks were assayed at greaterthan 100% stated potency, except for cefoperazone andoxacillin, which had 93 and 97% of the stated potency,respectively. Antibiotics were selected to represent thespectrum of agents known to be affected by medium varia-tions (Table 1).

Laboratories. To eliminate as much variation as possible,independent laboratories at different institutions with expe-rience in doing studies of this type and with a history ofperforming the disk test in a reproducible manner wereselected. To firm up the performance of the standardizeddisk susceptibility test, technologists from each of the par-ticipating laboratories went to the CDC Antibiotic Investi-

gational Laboratory for a final briefing on the details of theprotocol and the need to follow the protocol precisely and tocorrect any misconceptions about the study design.

Antimicrobial susceptibility test. The standardized diskdiffusion test was performed with strict adherence to theprocedures and time restrictions listed in NCCLS M2-A2(15). All plates were poured within 24 h of use. Tworepresentative plates for each lot of agar medium were testedat the same time with the same adjusted inoculum for eachquality control strain. Tests with each strain were initiatedand completed before the next strain was tested to minimizetime delays. The diameter of each zone of inhibition wasmeasured to the nearest 0.1 mm after 16 to 18 h of incubationat 35°C. All zones were measured independently by twoobservers from the back of the plate. The plates wereilluminated with reflected light and read against a blackbackground.

Criteria for selecting the primary reference reagent. Thecontrol limits established in NCCLS Standard M2-A2 andsupplements were assumed to be the best estimate of howthe ideal medium should perform. The primary referencereagent was selected by the following criteria: (i) closenessto the existing midpoint of current NCCLS control limits; (ii)low variability in zone readings reflecting clarity of zonesand distinctness of zone edges; (iii) ability of medium to beproduced in plated form without adjustment of pH; and (iv)ability to detect methicillin-resistant staphylococci and mea-sure trimethoprim and sulfonamide susceptibility of entero-cocci.

Statistical evaluation. Statistical analyses of all data weredone separately for each of the 28 microorganism-drugcombinations. The basic study design was a complexmultifactorial design. The model contains the following fivefactors: (i) institution, a random factor with five levels; (ii)agar lot, a fixed factor with seven levels; (iii) reader, arandom factor with two levels; (iv) day, a random factor withfive levels; and (v) plate, a random factor with two levels.The seven lots of Mueller-Hinton agar were from different

manufacturers. Five institutions (investigator laboratories)participated: CDC, the Cleveland Clinic Foundation, theClinical Microbiology Institute, Tualatin, Oreg., and St.Vincent's Hospital, Portland, Oreg., and the Veterans Ad-ministration Medical Center, Baltimore. Two plates for eachlot of agar were tested each day; each test was repeated onfive days; each of two readers from each of the five institu-tions measured and recorded zone diameters on all plates foreach day. A total of 100 measurements was taken for eachmicroorganism-drug combination on each lot of agar.The mathematical model for the analysis of variance

(ANOVA) may be written explicitly as

Yijklm = t + Oti + P + 'Yfi)k + 8Wi + 4(ijI)m + (OLI)1i + (P-Y)(i3yk+ (38))(i)jl + (Y8)(i)kl + (.Yy)(ijl)km + (P-Y8)(i)jkl

where ,u, overall mean; ai, institution effect; P, lot effect;'Y(i)k, reader effect; 8(j)y, day effect; (ùjl)m, plate effect; (OP)ij,interaction effect between institution and lot; (p-y)(i2k, inter-action effect between lot and reader; ( interactioneffect between lot and day; (Y8)(i)kl, interaction effect be-tween reader and day; (7f)(ffl)km, interaction effect betweenreader and plate; and (I78)()kI, interaction effect betweenlot, reader, and day. We assume that j'1 pj = 0 and that(Xi, 'Y(i)kg 8(i)lg 4ij/)m, (afflii, (P5^Y)(i)ji" S)()l (-Yb)(i)kl, (-Y()(ijl)km,and (8Y)(i)jkl are independent samples from Gaussian popu-lations with zero means and variances o2, C Y28,a2"2)or(t (Ir2(1, (y2(P ),(Y2(-Y)2(yg,) and o2(y), respectively.

J. CLIN. MICROBIOL.

on May 6, 2020 by guest

http://jcm.asm

.org/D

ownloaded from

MUELLER-HINTON AGAR REFERENCE LOT SELECTION 3

TABLE 1. Mean zone diameters from 100 measurements for each antimicrobial agent-organism combination

Mean zone diam (mm) with NCCLS lot no.: Range ofStrain and drug 1234567meansi 2 3 4 5 6 7men

Staphylococcus aureus ATCC 25923Cephalothin 33.2 32.6 32.5 32.5 33.0 32.6 31.5 1.7Chloramphenicol 22.3 23.6 22.8 22.8 23.8 22.9 23.4 1.5Clindamycin 23.8 26.1 24.9 25.8 26.1 25.0 25.7 2.3Erythromycin 23.7 26.3 25.2 25.8 25.8 24.6 26.1 2.6Gentamicin 22.8 21.2 24.4 25.1 22.7 22.5 22.3 3.9Oxacillin 21.1 21.1 21.2 21.2 20.0 19.4 20.2 1.8Penicillin G 32.8 31.9 32.2 32.4 32.4 31.8 31.5 1.3Tetracycline 25.4 26.4 25.4 25.8 25.7 25.5 23.5 2.9Vancomycin 17.6 17.1 16.5 17.2 17.9 17.4 17.2 1.4

Escherichia coli ATCC 25922Amikacin 21.4 21.4 21.0 23.3 22.7 21.5 21.4 1.9Ampicillin 18.4 18.8 17.5 18.7 19.1 10.0 18.7 2.5Carbenicillin 22.6 24.3 23.1 23.8 25.1 25.7 24.5 3.1Cefotaxime 29.4 30.2 29.4 30.4 30.9 31.3 30.3 1.9Cefoxitin 23.0 24.5 23.6 24.7 25.4 25.6 24.9 2.6Cephalothin 16.6 18.9 17.1 19.3 19.2 19.5 19.1 2.9Chloramphenicol 23.4 23.3 22.5 23.7 23.5 24.2 23.0 1.7Gentamicin 20.8 20.8 21.3 23.7 22.8 21.5 22.2 2.9Moxalactam 28.5 29.2 28.6 29.5 30.7 30.2 29.5 2.2Sulfisoxazole 18.9 20.9 19.2 19.5 19.7 23.3 20.3 4.4Tetracycline 22.4 21.0 20.8 21.5 21.9 23.0 19.1 3.9Trimethoprim-sulfamethoxazole 25.4 26.7 25.7 25.4 26.8 27.2 25.7 1.8

Pseudomonas aeruginosa ATCC 27853Amikacin 21.9 20.5 22.4 21.2 22.0 20.7 20.6 1.9Carbenicillin 20.6 20.9 20.3 20.3 20.1 20.8 20.8 0.8Cefoperazone 25.2 25.8 25.1 25.1 25.0 25.1 25.0 0.8Cefotaxime 20.5 20.2 20.3 20.2 20.1 20.1 20.2 0.4Gentamicin 18.7 17.4 19.8 19.1 19.1 17.8 17.7 2.4Moxalactam 20.4 21.7 21.1 20.7 20.3 21.0 20.9 1.4Piperacillin 28.3 28.9 28.3 28.0 28.3 28.4 28.2 0.9

This model was used to test the equality of the mean zonediameters for the levels of each model component as well asto obtain an estimate of the zone diameter variance for eachmicroorganism-drug combination. To address the issue ofvariability differences between the lots, the ANOVA wasdone separately for each lot (the components involving lot Pjwere eliminated from the model specified above). All analy-ses were done with the BMDP statistical software programP8V, which performs an ANOVA for general mixed models.For each lot of agar, the mean zone diameter, 95%

confidence intervals, and control limits were calculated. Thezone diameter variance used in calculating the confidenceintervals was estimated by the mean square of the interac-tion effect between institution and lot. The control limitswere obtained by using the concept of a tolerance limit(which estimates a range containing a high proportion ofindividual values in a population). The following formula forthe control limits uses a variance estimate that involvesintralaboratory variability and the appropriate degrees offreedom associated with the study design: yi + 2.24v'@7,with yi equal to the mean zone diameter of the ith lot and si2being the ith lot zone diameter variance estimated by thesum of the variance component estimates for reader, day,plate, reader by day, and reader by plate. These controllimits covered 95% of all zone diameters in the populationwith 95% confidence.

Trial production of lots matching performance of proposedreagent reference standard. Each of the seven manufacturersreceived samples of the proposed reagent reference standard

which was selected by the above criteria. Each manufacturerthen attempted to adjust the performance of a newly manu-factured lot of agar to match that of the standard. Evaluationof the newly manufactured lots was accomplished by testingeach new lot in parallel with the reference reagent with thesame adjusted inoculum and lot of disks. The differencesbetween the means of 30 replicates tested on the same dayfor each of the 28 microorganism-drug combinations werestatistically analyzed by a paired t test. The tentative criteriafor accepting a lot of agar as essentially identical to thereference material were that the means of the lot for all 28microorganism-drug combinations be within 1 mm of thoseobserved with the proposed reference medium observed atthe same time. Each manufacturer submitted the lot whichthey felt most closely fit these criteria to CDC for evaluationby the same protocol. Thirty replicates were selected, sinceit was determined, by using variance estimates from the firststudy, that this would be the number required for a highprobability of detecting a 1-mm difference in means betweentwo agar lots run in parallel (at = , = 0.01).

RESULTS

pH measurements. The five independent laboratories mea-sured the pH of each batch of agar poured. The measure-ments were remarkably similar and revealed that lots 1 and7 should be eliminated from consideration because they didnot come within the required pH range of 7.2 to 7.4 (Table2).

VOL. 24, 1986

on May 6, 2020 by guest

http://jcm.asm

.org/D

ownloaded from

4 POLLOCK ET AL.

TABLE 2. Pooled pH measurements for seven lots ofMueller-Hinton agar at five independent laboratories

pH (n = 25 determinations)Lot no.

Mean SD Minimum Maximum

1 7.05 0.05 6.95 7.202 7.28 0.05 7.20 7.403 7.28 0.04 7.20 7.354 7.36 0.05 7.25 7.455 7.30 0.06 7.15 7.406 7.20 0.06 7.10 7.337 7.41 0.06 7.30 7.52

Mean zone diameters. Mean zone diameters for eachmicroorganism-drug combination were calculated by poolingdata from all five institutions, representing 100 readings foreach of the 28 microorganism-drug combinations (five insti-tutions x two readers x 5 days x 2 plates). The means of thezone diameters for each antibiotic and organism combinationare found in Table 1 together with the range of means.Greater differences occurred due to medium variation withsome antibiotics than with others.With S. aureus ATCC 25923, gentamicin, erythromycin,

clindamycin, and tetracycline had the greatest spread ofmean zone diameters between agar lots. Variations in themeans for E. coli ATCC 25922, attributable in part todifferences in agar lots, were greater with sulfisoxazole,tetracycline, carbenicillin, gentamicin, and cephalothin.Gentamicin and amikacin demonstrated the greatest varia-tion in means associated with differences between lots ofagar with P. aeruginosa ATCC 27853. The spread betweenthe means on the seven lots was never less than 1 mm withS. aureus ATCC 25923 and E. coli ATCC 25922; the spreadof zone diameters was as great as 4.4 mm with sulfisoxazoleand E. coli.Ranking of lots. The seven lots were ranked by the sum of

the absolute differences between the midpoint for each of the28 drug and organism combinations and the respectiveNCCLS midpoint. These ranks are presented in Table 3,with a rank of 1 indicating the smallest midpoint deviation.Lot 5 gave the closest fit, while lots 3 and 7 were least closeto existing midpoints (lots 7 and 1 had pHs outside theacceptable range).The sums of the calculated control interval widths for the

28 microorganism-drug combinations were ranked for theseven lots, with the lot having the smallest sum given a rankof 1 (Table 3). Lot 7 exhibited the smallest amount ofvariation, while lot 1 showed the greatest.The sum of the control interval widths was derived from

control limits calculated with a formula for a varianceestimate involving intralaboratory variability with 95% con-

fidence limits. The sum of the widths of the control limits forall 28 microorganism-drug combinations derived by thismethod are given (Table 3).Comparison of performance of other lots to the one best

meeting the criteria for selection. All lots were evaluated forsimilarities and differences and for determination of the lotwhich best met the current NCCLS midpoint standard andhad the least variability (Table 3). Lot 5 appeared to fit thesecriteria better than any of the other six lots and was

tentatively selected as the new NCCLS reagent standard.The means for the other six lots were compared to the

mean for lot 5 to determine the extent of their variation fromthis lot (Table 4). Of the 28 means, 13 were significantlydifferent from those of lot 5 for lots 1 and 2, 7 from those of

lot 3, and 9 from those of lot 4, and 11 of the 28 means for lots6 and 7 were significantly different from the means for lot 5.Although some of the means were statistically different fromthose of lot 5, 92% of all means were within 2 mm and 65%were within 1 mm of the means for lot 5. A comparison of thefour lots with the correct pH revealed that 65% of the meanswere within <1 mm of the mean of lot 5 and 96% were withinc2 mm of this lot.Streptococcus faecalis and Staphylococcus aureus testing.

There was wide variation in the performance of the sevenlots of agar with trimethoprim-sulfamethoxazole and Strep-tococcus faecalis (ATCC 29212 and ATCC 33186). Lots 3and 4 gave no zone, while lots 2, 5, and 7 gave zones .20mm and lots 1 and 6 gave smaller zones (<20 mm). Otherspecies produced large clear zones of inhibition when tri-methoprim-sulfamethoxazole disks were tested on lot 5.Two strains of methicillin-resistant Staphylococcus au-

reus were selected by one of us (C.T.) for their particularsensitivity to variations in agar media that may preventdetection of methicillin resistance. Lot 5 demonstrated def-inite growth up to the edge of the oxacillin disk, whereas theother lots demonstrated various amounts of growth within adefinite zone of inhibition. With one of the strains, methicil-lin disks tested as sensitive on all media. The other strainwas resistant on lots 1 and 5 but gave intermediate (10 to 13mm) zones on the other lots.

Trial production of lots matching performance of reagentstandard. After selection of lot 5 as the proposed physicalreagent standard, each ofthe seven manufacturers was givensome of this lot to experimentally manufacture a new pro-duction lot with essentially similar performance. The per-formance of the five new agar lots was compared with that oflot 5 by performing 30 replicate tests on each medium in oneof the reference laboratories (CDC). One manufacturer didnot participate, and one lot presented technical difficulties.The number of antimicrobial agent-organism combinationswhich had greater than 1 mm difference between the meanzone diameters had not clearly decreased with media sub-mitted by four of the five manufacturers (Table 5). Onemanufacturer, however, improved its product sufficiently toallow it to be considered a possible second reference stan-dard.

DISCUSSION

Variation in the performance of lots of Mueller-Hintonagar led us to develop a tightly controlled protocol whichcompared the performance of agar media currently on themarket. The study was designed to determine whether a

TABLE 3. Sums and ranks of the control interval widths andmidpoint deviations for the 28 microorganism-drug combinations

Widtha Midpoint deviationsLot no.

Sum Rank Sum Rank

1 153 7 44.5 6.52 143 4 33.5 43 139 2 44.5 6.54 147 6 31.5 35 140 3 25.0 16 146 5 30.0 27 136 1 35.0 5

a Width of control interval calculated from an estimate of variance. Controlinterval width is the difference between the low and high limit of the controllimits (in millimeters).

b Absolute difference between midpoint of existing NCCLS control inter-vals and midpoint of control interval determined by using method III.

J. CLIN. MICROBIOL.

on May 6, 2020 by guest

http://jcm.asm

.org/D

ownloaded from

MUELLER-HINTON AGAR REFERENCE LOT SELECTION 5

TABLE 4. Absolute value of difference of means compared with the proposed reference medium (lot 5)

Absolute differences vs lot 5 (mm)Organism Drug

1 2 3 4 6 7

Staphylococcus aureus Cephalothin 0.2 0.4 0.5 0.5 0.4 1.5*Chloramphenicol 1.5* 0.2 1.0* 1.0* 0.9 0.4Clindamycin 2.3* 0.0 1.2 0.3 1.1 0.4Erythromycin 2.1* 0.5 0.6 0.0 1.2* 0.3Gentamicin 0.1 1.5* 1.7* 2.4* 0.2 0.4Oxacillin 1.1* 1.1* 1.2* 1.2* 0.6 0.2Penicillin G 0.4 0.5 0.2 0.0 0.6 0.9Tetracycline 0.3 0.7 0.3 0.1 0.2 2.2*Vancomycin 0.3 0.8* 1.4* 0.7* 0.5 0.7*

Escherichia coli Amikacin 1.3* 1.3* 1.7* 0.6 1.2* 1.3*Ampicillin 0.7 0.3 1.6* 0.4 0.9* 0.4Carbenicillin 2.5* 0.8* 2.0* 1.3* 0.6 0.6Cefotaxime 1.5* 0.7 1.5* 0.5 0.4 0.6Cefoxitin 2.4* 0.9 1.8* 0.7 0.2 0.5Cephalothin 2.6* 0.3 2.1* 0.1 0.3 0.1Chloramphenicol 0.1 0.2 1.0* 0.2 0.7* 0.5Gentamicin 2.0* 2.0* 1.5* 0.9* 1.3* 0.6Moxalactam 2.2* 1.5* 2.1* 1.2* 0.5 1.2*Sulfisoxazole 0.8 1.2* 0.5 0.2 3.6* 0.6Tetracycline 0.5 0.9* 1.1* 0.4 1.1* 3.9*Trimethoprim- 1.4* 0.1 1.1* 1.4* 0.4 1.1*

sulfamethoxazole

Pseudomonas Amikacin 0.1 1.5* 0.4 0.8* 1.3* 1.4*aeruginosa Carbenicillin 0.5* 0.8* 0.2 0.2 0.7* 0.7*

Cefoperazone 0.2 0.8 0.1 0.1 0.1 0.0Cefotaxime 0.4 0.1 0.2 0.1 0.0 0.1Gentamicin 0.4 1.7* 0.7* 0.0 1.3* 1.4*Moxalactam 0.1 1.4* 0.8* 0.4 0.7* 0.6*Piperacillin 0.0 0.6 0.0 0.3 0.1 0.1

a *, Significant at overall a = 0.05 by the Bonferroni inequality test for each lot.

physical reagent standard tested in parallel with newlyproduced lots would enable manufacturers to control theirproducts with greater precision, resulting in greater uniform-ity in the performance of products from different manufac-turers. Written control limits which define zone size limitsfor each antimicrobial agent-organism combination must bebroad enough to account for a variety of the technicalvariables that influence measurement of zone diameters aswell as inherent differences in performance of Mueller-Hinton agar from different manufacturers. We propose a

performance test which involves replicate testing of a refer-ence medium in parallel with a test lot. Differences betweenmean zone sizes would be used to evaluate medium variabil-ity and reduce variability resulting from other factors.The data demonstrated that some changes in quality

control protocols were necessary at the manufacturing level.First, the measurements for control strains on currentlymarketed media were slightly different from existing stan-dards. Secondly, media produced by five manufacturersperformed within 2 mm of the proposed standard for mostmicroorganism-drug combinations. Statistical analyses ofthe variances indicated that a manufacturer would have toperform 30 replicates on the same day to have 99% confi-dence of detecting 1-mm differences between two lots of agarrun in parallel. This number of replicates exceeded the usualnumber of tests performed in manufacturing agar media. Inaddition, current quality control practices compare testedperformance against a written rather than a physical stan-dard. The use of a physical standard with an increasednumber of replicates could theoretically provide manufac-turers with a basis for producing more uniform lots of media

and the ability to detect smaller drifts in performance than iscurrently possible.

Reliable detection of methicillin resistance in Staphylo-coccus aureus presents a number of technical difficulties. Oncurrently marketed Mueller-Hinton agar, some of thesestrains appear to be susceptible with methicillin disks butresistant (double zones) with oxacillin disks. The characterof the growth within the zones of inhibition appears to bemedium dependent with the strains that were tested. Thisproblem requires careful consideration in setting any stan-

TABLE 5. Comparison of means with those of lot S

No. of means fallingFirst or second outside standard by:Manufacturer lot submitteda

>1 mm >2 mm

A 1 12 72 9 2

B 1 9 02 7 3

C 1 16 22 10 3

D 1 6 12 9 2

E 1 8 22 0 0

a First lot, Mean zone size of 100 readings on the proposed reference lotcompared with that for lot 5 (5 days x 2 readers x 5 institutions x 2 days);second lot, mean zone size of 30 replicate tests on the same day comparedwith that for lot 5 at CDC.

b The standard was the mean zone of inhibition obtained with lot 5 of thefirst set of agar lots submitted.

VOL. 24, 1986

on May 6, 2020 by guest

http://jcm.asm

.org/D

ownloaded from

6 POLLOCK ET AL.

dard for Mueller-Hinton agar. Trimethoprim-sulfamethox-azole susceptibility of Streptococcus faecalis could be de-termined on only two of the seven lots of Mueller-Hintonagar.The major question in developing a physical reagent

standard for Mueller-Hinton agar was whether additionallots could be produced which performed essentially the sameas the selected reference lot. In the initial study, overallperformance for the media which had the proper pH wasrelatively uniform for production lots, but there were no datato demonstrate how closely it was possible to duplicate theperformance of a new lot of agar medium with a referencereagent available for comparison. Additional studies wereperformed to answer that question.When additional lots of agar were produced and tested in

parallel with the selected physical standard, one manufac-turer submitted a medium which was essentially the same inperformance as the reference lot, that is, within 1 mm of themean zone size for all microorganism-drug combinations.Four manufacturers submitted lots with performances whichwere not much closer to the selected standard than those ofrandom lots that met the written standard.

Changing performance associated with variations in rawmaterials over a period of time may be eliminated bymaintaining a standard reference medium with which manu-facturers may compare the performance of future lots. Whileit appears that different manufacturers may not be able toproduce lots of Mueller-Hinton agar with greater similarityin performance than is currently accomplished, the problemof drift in behavior of these media may be circumvented bycomparing the performance of new lots with that of aphysical reagent as well as against written criteria. A statis-tically valid performance test based on this principle is beingdeveloped.

ACKNOWLEDGMENTS

We wish to thank Les Cunningham, Acumedia; David Power andGeorge Evans, BBL Microbiology Systems; Aaron L. Lane, Difco;Mary Nichols, Gibco; Vernon L. Olson, Spectrum; Eric Bridson,Oxoid; and Tom Scott, Scott Laboratories, for their participation inand contributions to this study. We thank the Association ofMicrobiologic Media Manufacturers for their support and contribu-tions. We appreciate the help of Morris Suggs and his group at CDC,without whose assistance this entire project would not have beenpossible. We also thank George Williams, Cleveland Clinic, forreviewing the manuscript. The administrative assistance of NCCLSwas invaluable, in particular that of John Bergen and John McCon-nell.Funds for support of this research were donated to the NCCLS

Microbiology Development Fund.

LITERATURE CITED1. Acar, J. F. 1980. The disc susceptibility test, p. 24-54. In V.

Lorian (ed.), Antibiotics in laboratory medicine. The Williams& Wilkins Co., Baltimore.

2. Barry, A. L. 1980. Procedure for testing antibiotics in agarmedia: theoretical considerations, p. 1-23. In V. Lorian (ed.),

Antibiotics in laboratory medicine. The Williams & WilkinsCo., Baltimore.

3. Bauer, A. W., W. M. Kirby, J. C. Sherris, and M. Turck. 1966.Antibiotic susceptibility testing by a standardized single diskmethod. Am. J. Clin. Pathol. 45:493-496.

4. Brenner, V. C., and J. C. Sherris. 1972. Influence of differentmedia and bloods on the results of diffusion antibiotic suscep-tibility tests. Antimicrob. Agents Chemother. 1:116-122.

5. D'Amato, R. F., C. Thornsberry, C. N. Baker, and L. Kirsten.1975. Effect of calcium and magnesium ions on the susceptibilityof Pseudomonas spp. to tetracycline, gentamicin, polymixin B,and carbenicillin. Antimicrob. Agents Chemother. 7:596-600.

6. Dixon, W. J. (ed.). 1981. BMDP statistical software, p. 427-436.University of California Press, Los Angeles.

7. Drew, W. L., A. L. Barry, R. O'Toole, and J. C. Sherris. 1972.Reliability of the Kirby-Bauer disc diffusion method for detect-ing methicillin-resistant strains of Staphylococcus aureus. Appl.Microbiol. 24:240-247.

8. Ericsson, H. M., and J. C. Sherris. 1971. Antibiotic sensitivitytesting: report of an international collaborative study. ActaPathol. Microbiol. Scand. Sect. B Suppl. 217.

9. Garrod, L. P., and P. M. Waterworth. 1969. Effect of mediumcomposition on the apparent sensitivity of Pseudomonasaerguinosa to gentamicin. J. Clin. Pathol. 22:534-538.

10. Hanus, F. J., J. G. Sands, and E. O. Bennett. 1967. Antibioticactivity in the presence of agar. Apple. Microbiol. 15:31-34.

11. Kenny, M. A., H. M. Pollock, B. H. Minshew, E. Casillas, andF. D. Schoenknecht. 1980. Cation components of Mueller-Hinton agar affecting testing of Pseudomonas aerguinosa sus-ceptibility to gentamicin. Antimicrob. Agents Chemother. 17:55-62.

12. Knowles, R. C., and B. F. Gilmore. 1981. Quality control of agardiffusion susceptibility tests. Am. J. Clin. Pathol. 76:590-596.

13. Kunin, C. M., and W. P. Edmondson. 1968. Inhibitor of antibi-otics in bacteriologic agar. Proc. Soc. Exp. Biol. Med. 129:118-122.

14. Medeiros, A. A., T. F. O'Brien, W. E. C. Wacker, and N. F.Yulug. 1971. Effect of salt concentration on the apparent in vitrosusceptibility of Pseudomonas and other gram-negative bacillito gentamicin. J. Infect. Dis. (Suppl.) 124:559-564.

15. National Committee for Clinical Laboratory Standards. 1982.NCCLS Approved Standard M2-A2 (performance standards forantimicrobic susceptibility tests) and 1982 supplement. NCCLS,Villanova, Pa.

16. Pollock, H. M., B. H. Minshew, M. A. Kenny, and F. D.Schoenknecht. 1978. Effect of different lots of Mueller-Hintonagar on the interpretation of the gentamicin susceptibility ofPseudomonas aerguinosa. Antimicrob. Agents Chemother. 14:360-367.

17. Reller, L. B., F. D. Schoenknecht, M. A. Kenny, and J. C.Sherris. 1974. Antibiotic susceptibility testing of Pseudomonasaerguinosa: selection of a control strain and criteria for magne-sium and calcium content in media. J. Infect. Dis. 130:454-463.

18. Thornsberry, C., and L. K. McDougal. 1983. Successful use ofbroth microdilution in susceptibility lists for methicillin-resistant (heteroresistant) staphylococci. J. Clin. Microbiol.18:1084-1091.

19. Yousef, R. T., M. A. El-Nakeeb, and M. H. El-Masrl. 1967.Binding of antibiotics to agar and its relation to the inhibitionzone. Acta Pharm. Suecica 4:253-260.

20. Waterworth, P. 1978. Uniformity of sensitivity test media. J.Antimicrob. Chemother. 4:4-6.

J. CLIN. MICROBIOL.

on May 6, 2020 by guest

http://jcm.asm

.org/D

ownloaded from