ANTIMICROBIAL AGENTS AND CHEMOTHERAPY, Apr. 1975, p. 466-480 Copyright 0 1975 American Society for Microbiology
Vol. 7, No. 4 Printed in U.S.A.
Laboratory Evaluation of a Rapid, Automated Susceptibility Testing System: Report of a Collaborative Study CLYDE THORNSBERRY,* T. L. GAVAN, J. C. SHERRIS, ALBERT BALOWS, J. M. MATSEN,' L. D. SABATH,2 FRITZ SCHOENKNECHT, L. D. THRUPP, AND J. A. WASHINGTON II
Center for Disease Control, Atlanta, Georgia 30333*; The Cleveland Clinic Foundation, Cleveland, Ohio 44106; University of Washington, Seattle, Washington 98195; University of Minnesota, Minneapolis, Minnesota 55455; Boston City Hospital, Boston, Massachusetts 02114; University of California, Irvine, California 92668; and The Mayo Clinic and Mayo Clinic Foundation, Rochester, Minnesota 55901 Received for publication 23 December 1974
Seven laboratories participated in a collaborative study to evaluate the Autobac 1 system. Results obtained with this assay system were compared to those obtained by the standardized Bauer-Kirby disk diffusion test, and each of these two methods was compared to the agar dilution technique. Comparison of the Autobac 1 and the disk diffusion results from the seven laboratories showed an overall average of 91.5% interpretive agreement with the 17 antimicrobial agents tested. The distribution in the levels of Autobac 1/disk diffusion agreement was such that with 13 antimicrobial drugs agreement was 90% or higher; with three, between 85 and 90%; and with one, 77% (nitrofurantoin). Comparison of the Autobac 1 and disk diffusion tests with the International Collaborative Study agar dilution test showed that both methods gave levels of agreement with the International Collaborative Study agar dilution technique that were generally high and equivalent. The average overall agreement between the agar dilution test and each of the other two methods was approximately 90%. Disagreements that did occur tended to involve organisms that were drug susceptible by the Autobac 1 system but intermediate or resistant by the other two methods. This was in part due to the narrow intermediate interpretive zone of the Autobac 1 test. In reproducibility studies with the Autobac 1 and disk diffusion methods, no significant differences were observed between the interpretive reproducibility of the two methods.
In 1954, Schneierson (9) described a rapid (4 to 6 h) drug susceptibility test in which prepared disks of known antimicrobial content were used to provide specific drug concentrations in tubes of broth. This approach has been used in at least two automated systems, of which Autobac 1 is one (5, 6, 8, 12). The present report presents the results of a collaborative study designed to compare the Autobac 1 system, the Bauer-Kirby disk diffusion test, and an agar dilution technique. MATERIALS AND METHODS Reagents and media. Identical lots of media, disks, and antimicrobial agents were used in the susceptibility tests performed in each laboratory. The media used were tryptic soy agar, tryptic soy broth, Mueller-Hinton agar, and Eugonic broth, all manufactured by Pfizer Diagnostics and supplied by this firm to the investigators. The antimicrobial drug-con'Present address: University of Utah, Salt Lake City, Utah 84132. 2Present address: University of Minnesota, Minneapolis, Minn. 55455.
466
taining disks used in the Autobac 1 system (Table 1), and those used in the agar diffusion tests (Table 2), were also manufactured by Pfizer, Inc. However, in our study, some Pseudomonas aeruginosa strains were also tested with carbenicillin diffusion disks from Baltimore Biological Laboratories (BBL; Cockeysville, Md.). Antimicrobial powders that were used in the agar dilution susceptibility tests (Table 2) were either Pfizer products or were obtained from other manufacturers by Pfizer for use in these studies. Organisms. The organisms used in the study were current isolates from the clinical laboratories associated with the seven groups of investigators. The organisms were isolated and stored on media in routine use in each laboratory. For susceptibility testing, the organisms were streaked onto tryptic soy agar containing 5% sheep blood; cultures received from other laboratories were subcultured twice on this medium prior to testing. Three control cultures were used throughout the study; two of them were the "Seattle" Escherichia coli (derived from ATCC 25922) and Staphylococcus aureus (derived from ATCC 25923) strains recommended by the Food and Drug Administration and the third was the "Mayo" strain of P. aeruginosa. Susceptibility test methods. (i) Dilution method.
467
AUTOMATED SUSCEPTIBILITY TESTING SYSTEM
VOL. 7, 1975
TABLE 1. Antimicrobial elution disks used in the Autobac I system
The agar dilution method used by the International Collaborative Study group, as described by Ericsson and Sherris (3), was used in this study. Four to five Orga- morphologically similar colonies of each organism AsLabeled nisms sayed Disk Antimicrobial disk ma-s disk testwere inoculated into tryptic soy broth and incubated code agent (g) at 35 C. The turbidity of the suspension was adjusted (M) MaSSa ed8 to that of a 0.5 McFarland Standard, which had been shown to equate to within 5 x 107 to 5 x 108 N 6.7 Ampicillin .......... AM/en 3.6 colony-forming units (CFU) per ml. This suspension Carbenicillin ....... CB/en 120.0 142.0 N 19.4 P, N was diluted 1:10 with tryptic soy broth to contain Cephalothin ...... CL/ea 15.0 N 5.8 C/en 4.0 Chloramphenicol approximately 107 CFU/ml. Agar plates were pre2.3 P Clindamycin ........ CM/ep 2.0 pared to contain varying concentrations of each anN Colistin .......... 10.0 CS/en 13.0 agent by combining molten Mueller-Hintimicrobial 3.3 P Erythromycin ....... E/ep 2.5 ton agar (at 50 C) and the antimicrobial agent (di9.0 11.9 P, N Gentamicin ......... GM/ea luted in sterile, distilled water) in a ratio of nine parts N 19.8 K/en 22.0 Kanamycin ......... agar to one part antimicrobial solution. The cultures 5.0 7.0 P Methicillin ......... SC/ep were inoculated onto the Mueller-Hinton agar surface N 19.4 Nalidixic acidc ...... NA/en 15.0 . FD/en 15.0 21.0 N Nitrofurantoinc with the replicating device of Steers et al. (11). Since Penicillin G ......... P/ep 0.2 (U)d 0.26 P each inoculating point delivered approximately 0.001 N 13.9 Polymyxin B ........ PB/en 12.5 (U)d ml of inoculum, approximately 104 CFU of each P 0.6 Tetracycline ........ TE/ep 0.5 culture was inoculated on the surface of each plate. N 1.2 Tetracycline ........ TE/en 1.2 After the inocula dried, the plates were incubated at 9.9 Vancomycin ........ VA/ep 10.0 35 C for 18 to 24 h. The minimum inhibitory concena Antimicrobial agent eluted from disk and eluate assayed tration was defined as the least amount of antimiby the standard cup-plate method. Assays also performed by crobial agent employed in the test that permitted the the standard diffusion method. growth of no more than one colony. A slight haze, due b N, Gram-negative organisms and enterococcus; P, gram- to inoculum, was ignored. All minimal inhibitory positive cocci excluding enterococcus. concentrations were recorded as micrograms per milliUsed with urinary isolates only. d liter. Penicillin measured in units.
TABLE 2. Interpretive criteria for disk diffusion agar, dilution and Autobac 1 tests Antimicrobial agent
Ampicillin Staphylococci ..... Gram-negative organisms and enterococci Carbenicillin P. aeruginosa. E. coli, Proteus Sp..
Cephalothin. Chloramphenicol Clindamycin.
Disk diffusion zone diameter (mm)
ICS agar dilution minimal inhibitory conc. (tsg/ml)
Ra
Ja
Sa
29
14
> 64
< 12
13-14
> 15
> 256
< 17 < 14 < 12 < 14
18-22 15-17 13-17 15-16
9-10 14-17
Colistin .............
0.60 > 0.60 > 0.60 > 0.60
> 23
> 64
32
< 16
> 18
> 64
32, 16
18 >17
> 64
32 4
< 16
>8
>11 > 18 > 13
> 16
8
>8 > 16
4
< 0.50
8
12
> 16
0.60 > 0.60
......
Polymyxin B .. S
:> coC.))
\
;C
YZZ&E
Z .c . .
0Z
0)
.0
t C1 cc~ ~ ~~~~~~~~~c
0 M 0
00
0)
CC1
0
-{§:
4S
0
0
>
0
0)
0
0 Oco
a
0
0
0
O
0
4-e |.t
S 00)
o
..
_04 (6C ;( ;C o .S~~~~~~C Mroo~o
* _
X
40)
c; ; ; ;
0)
.
W6
0900-4 4c-4 66
t
I
-
co
~
o
0 0).-
I
0.
0
^
0^00 ^ 10--
10-
00
l.-
-
~ Q WC ;rO D o s° ^ co>°
~~.4 -~~ 4
.
-
S. i ~01'-400.-i .-E 4~~~~~~~~~~~~~~C 0N 00O"°o o 01 1 - 0 1
ooi
QA
.4
0)
0
6 6',6646.-
0
0. 0-i >1
eJ~~~~C 0
cL6o a d C6 0- 0( 0Oc -;C( 6~ ~ ~ C C ;~~~
*.
___
~~C'014
___
I 11 z03
I
I
c c 0
01 Er
4
0
~
0
0 '0 z
S.
0)
o.b0
to
2,
._
D la te
~
.0
2
.
:
40e ._
...4
:
:
-
-
0
C I=
I
:
0
._
8
AUTOMATED SUSCEPTIBILITY TESTING SYSTEM
VOL. 7, 1975
TABLE 10. Overall discrepanciesa between the disk diffusion and the agar dilution test results with
gram-positive organisms Antimicrobial agent
S. aureus
mSepidermpdis
Cephalothin Clindamycin . Erythromycin . Gentamicin ....... Methicillin ....... Penicillin G . Tetracycline .
0,0,0/120b 0,0,0/120 1,4,1/80 0,0,0/120 0,0,0/80 0,0,3/80 3,1,2/120
0,0,0,/5ih 0,0,0/36 100.0 100.0 0,0,0/51 93.0 2,0,0/34 99.4 1,0,0/51
Vancomycin .
0,0,0/80
0,0,0/34
0,0,0/34 1,0,6/34 3,4,1/51
Other
0,0,0/3
475
the mean inhibitory zone diameter determined by the collaborating investigators. No laboratory had individual test results outside the limits more frequently than would be expected.
Agreement
100.0 91.2 92.0 100.0
a Very major, susceptible by disk diffusion, resistant by agar dilution; major, resistant by disk diffusion, susceptible by agar dilution; minor, intermediate by only one of the two methods (disk diffusion, agar dilution). ' Numbers of very major, major, and minor discrepancies/ total number of strains tested.
cAverage % agreement, 97.1.
pairs tested by all seven collaborators by both tests (only six collaborators performed tests on colistin; therefore, this antimicrobial agent was eliminated from the equivalence studies). A total of 508 such test combinations were included in this analysis. In more than two-thirds of the 508 tests performed by the seven investigators on the 70 isolates (67.5%), all seven collaborators agreed on a single interpretive category by both tests (Table 16). However, at least five of the collaborators agreed on the same interpretive category by both tests 85.0% of the time (432 of 508 tests). In most of the remaining cases, five to seven of the collaborators agreed on an interpretive category by one of the two methods, while only one to four of the collaborators agreed on the same category by the remaining method. Failure of at least five of the seven collaborators to agree on an interpretive category by both methods occurred in only 1.2% of the tests. In 1% of the tests there was a major or very major discrepancy between the test method results, for example, when five or more collaborators agreed that the organism was resistant by one method, and five or more collaborators agreed it was susceptible by the other method (substantial disagreement, Table 16). Quality control data were available from five of the seven collaborating laboratories for the E. coli and S. aureus reference strains. Control limits (95%) were established for each antimicrobial agent/control organism combination by applying twice the maximum acceptable standard deviation described by the National Committee for Clinical Laboratory Standards (7) to
DISCUSSION These studies, which involved a total of 15,914 tests conducted with the Autobac 1 and the disk diffusion susceptibility methods, showed an overall agreement between the two procedures of approximately 90%. More than one-half the discrepancies were minor, i.e., one test recorded results in the intermediate category and the other recorded results in the susceptible or resistant categories. The majority of the remaining discrepancies involved organisms that were classified as resistant by the diffusion test and susceptible by the Autobac 1 test. A very closely similar pattern was seen when the Autobac 1 was compared with the dilution test. There was 90% agreement with 4.3% major and very major discrepancies. The overall agreement between the diffusion test procedure and the dilution method was 91%, with 2% major and very major discrepancies. Only with nitrofurantoin and ampicillin in the case of the Autobac 1 and with carbenicillin and nitrofurantoin in the case of the disk diffusion test were overall agreements less than 80%. With certain organism/antimicrobial agent combinations, more frequent discrepancies were encountered between the agar dilution results and those obtained by the other tests. More than 10% of the results with Staphylococcus epidermidis tested against penicillin and tetracycline showed substantial discrepancies when the Autobac 1 data were compared with the agar dilution data. Similar results with tetracycline were obtained when the diffusion test was compared with the agar dilution test. The reasons for these discrepancies are not clear, but with penicillin such discrepancies were partly influenced by the restricted definition of minimal inhibitory concentration correlates of susceptibility that were selected for the study. A similar question of definition may account for the high incidence of very major discrepancies in the Autobac 1/agar dilution comparison and of minor discrepancies in the disk diffusion/agar dilution comparison with carbenicillin and Proteus. The susceptibility testing of enterococci with cephalothin and of Pseudomonas sp. with kanamycin has been contra-indicated in the Autobac 1 method. All available data suggest that it is presently not possible to achieve a high level of Autobac 1/disk diffusion agreement for these
476
~ CD
THORNSBERRY ET AL.
-
0o c.)
0 0
-
o o a,
ootr c.)
o.
ANTIMICROB. AGENTS CHEMOTHER.
0 o
Ocs a
QO
0
.bCZ -o "0
c c,
0 Cs 00 N 4~~~~~-
1.
0
4)
0000
M
0
0o
CD
_. ._ U)
bL 0q rA 4-;
--40 0 L)
Z
C
C..
. C;
C
0-"
4
Z e 4)
4 0
00
e _
e
rn
a)
_~. 4)4 LO 0004> et0O>
w~~ O ~~~
cs
^ ^_o o
_ 00 _- 0 _ o c to
^ _
cl t:iS-Ab .2.w~~~~~I
s 0
0
ca
c 06 e e e CD
l
4~~~~~~V-3o oo o- o -4
O)
L N O -4 LOLo U- cq L-4O
O 0
5..
-
. °O0^0^-O0^0^0^~
4)0 0
0
rA
.°0n 0a0 c 4-
d~~~~)a)C
0)SX
. *4)i 4-)
0 44
._ Cu
o"
4)0
bo
._
a)
._
.0
0
.) C0
4
Q
Z
-
.
477
AUTOMATED SUSCEPTIBILITY TESTING SYSTEM
VOL. 7, 1975
TABLE 12. Estimated Standard Deviations (')a for Phase 1 and Phase 3 intra- and interlaboratory reproducibility; Autobac 1 and disk diffusion Autobac 1 (LSI) Organisms
Gram-negative organisms & enterococci Staphylococci
Disk diffusion (mm)
Interlab
Intralab
Intralab
Interlab
Overall
Range
Overall
Range
Overall
Range
Overall
Range
0.09
0.06-0.14
0.10
0.08-0.16
1.2
0.6-2.3
1.6
1.3-2.0
0.08
_b
0.03c
0.02-0.04c
1.4
0.7-2.1
1.4
1.1-2.0
a The ranges of &' listed reflect the different organisms or antimicrobial agents within that category of organisms, whereas the overall a' is a grand pooled estimate within the category. bToo few data to analyze by antimicrobial agent. c Only calculation which included 0.00 and 1.00 LSI values.
TABLE 13. Phase 1 and Phase 3 intralaboratory reproducilibity, by interpretive category, Autobac I and disk diffusion Interpretive agreement
Autobac la No. %
Disk diffusion No. %
91.4 Agreement 362 375 91.7 Substantial shiftb 20 5 1.2 5.1 Minor shiftc 14 3.5 29 7.1 a Cephalothin/enterococcus and kanamycin/ Pseudomonas aeruginosa contra-indications not included in data. b Shift from susceptible to resistant or vice versa on repeat testing. c Shift from susceptible or resistant to intermediate or vice versa on repeat testing.
two particular organism/antimicrobial agent pairs if a single elution disk for each of these antimicrobial agents is used. As a result of these findings, and due to the limited therapeutic value of cephalothin for enterococcus infections and of kanamycin for Pseudomonas infections,
the approach with cephalothin and kanamycin was to strive for good agreement with the other commonly isolated species (Tables 6 to 7) and to contra-indicate the single species in each instance. When these two instances of contraindication were included, an 85.4% agreement for cephalothin and an 86.7% agreement for kanamycin were obtained. Although a small increase in the incidence of minor discrepancies was affected by the inclusion of the contraindicated data, the principal change occurred in the incidences of very major discrepancies which increased from 2.4 and 0.4 to 7.5 and 4.5% with cephalothin and kanamycin, respectively. A considerable improvement in the Autobac 1/disk diffusion agreement was achieved with penicillin G, but not with the other antimi-
crobial agents, if the R < 0.90 LSI interpretive guideline was used. A more accurate estimation of staphylococcal susceptibility to this antibiotic was possible when this guideline was used because it allowed a larger resistant LSI range for Staphylococcus strains which presumably exhibit various levels of penicillinase production in a 3- to 5-h incubation period. Examination of the ampicillin results shows that the interpretative agreement was high and that incidences of substantial discrepancies were small for all of the species except Enterobacter. The principal source of disagreement between the two methods was the occurrence of susceptible Autobac 1 values for certain strains of the ampicillin-resistant population of Enterobacter sp. Two factors contributed to the large number of such discrepancies with Enterobacter and ampicillin. In many cases, there was growth in the cuvette, but it had settled to the bottom of the chamber in spite of continuous shaking. In such cases, the photometer read it as susceptible rather than resistant. This phenomenon also occurred to a lesser extent with Enterobacter and cephalothin. More recent studies at the University of Washington (Seattle) and at the Cleveland Clinic Foundation (Cleveland) have shown that the percentage of very major discrepancies can be approximately halved if the disk mass is kept within 80 to 125% of the nominal content, which was not the case in the study reported herein. Although there was still an average incidence of 15% very major discrepancies, these studies did not include visual observation for settled growth in the ampicillin chamber of the cuvette. Similar results were obtained with nitrofurantoin and Proteus sp. The only significant incidences of very major discrepancies that were encountered with this agent were those found with both P. mirabilis and indole-positive
478
THORNSBERRY ET AL.
ANTIMICROB. AGENTS CHEMOTHER.
TABLE 14. Distribution of interpretive agreement among the seven collaborators for the Autobac 1 and for the disk diffusion test results on 634 organism/antimicrobial agent pairs in phase 3 interlaboratory reproducibility studies Autobac 1 No. of collaborators agreeing on the same interpretive category
Organism/antimicrobial agent pairs
Disk diffusion
Test in
agreementa
No.
%
7
500
78.9
3500
6
64
10.1
5
28
4 3
Tests in dis-
agreementa
Organism/antimicrobial agent pairs
Tests
in agreementb
Tests in dis-
agreementb
No.
%
0
520
82.0
3640
0
384
64
51
8.0
306
51
4.4
140
56
33
5.2
165
66
29
4.6
116
87
21
3.3
84
63
13
2.5
39
52
9
1.4
27
36
aPercentage of Autobac 1 tests in agreement, 94.2; in disagreement, 5.8. b Percentage of disk diffusion tests in agreement, 95.1; in disagreement, 4.9. TABLE 15. Distribution of the interpretive disagreements of individual collaborators with the majority consensus encountered in 4,438 tests with the Autobac 1 and with the disk diffusion on 634 organism/antimicrobial agent pairs in Phase 3 interlaboratory reproducibility studies Individual collaborator disagreements
Type of disagreement with
majority
Autobac 1
Disk diffusion
consensus
Substantial" Minorc Equivocald
No.a
%
No.a
%
73 47 139
28.2 18.1
23 94
53.7
99
10.6 43.5 45.9
aIn this table, individual disagreements from the majority consensus must be considered since in many cases the minority collaborators disagreed with each other (e.g., 5 S; 1 I; 1 R). " Five or six collaborators classify as either susceptible or resistant and one or two collaborators classify the converse. c Either a majority (five or six) or a minority (one or two) of the collaborators classify in the intermediate =
category. d Four
=
=
or fewer collaborators interpretive category.
agree on any one
Proteus sp. The primary reason for these discrepancies was the excess potency of the elution disk used in this study. In more recent studies at the University of Washington and at the Cleveland Clinic Foundation, a nitrofurantoin elution disk was used that was within the recommended potency limits of 80 to 125% of the 15-,ug nominal content. These studies
TABLE 16. Summary Phase 3 Interlaboratory Equivalence, Autobac 1 vs. Disk diffusion Determinations
No.
Perfect agreement (7)b by both methods .343 Good agreement (5 or 6) by both methods .89 Substantial disagreementc between methods .5 Minor disagreemente between methods ..7 Acceptable agreement with disk diffusion, equivocal (.4) with Autobac 1 ........................... 35 Acceptable agreement with Autobac 1 equivocal (
Vol. 7, No. 4 Printed in U.S.A.
Laboratory Evaluation of a Rapid, Automated Susceptibility Testing System: Report of a Collaborative Study CLYDE THORNSBERRY,* T. L. GAVAN, J. C. SHERRIS, ALBERT BALOWS, J. M. MATSEN,' L. D. SABATH,2 FRITZ SCHOENKNECHT, L. D. THRUPP, AND J. A. WASHINGTON II
Center for Disease Control, Atlanta, Georgia 30333*; The Cleveland Clinic Foundation, Cleveland, Ohio 44106; University of Washington, Seattle, Washington 98195; University of Minnesota, Minneapolis, Minnesota 55455; Boston City Hospital, Boston, Massachusetts 02114; University of California, Irvine, California 92668; and The Mayo Clinic and Mayo Clinic Foundation, Rochester, Minnesota 55901 Received for publication 23 December 1974
Seven laboratories participated in a collaborative study to evaluate the Autobac 1 system. Results obtained with this assay system were compared to those obtained by the standardized Bauer-Kirby disk diffusion test, and each of these two methods was compared to the agar dilution technique. Comparison of the Autobac 1 and the disk diffusion results from the seven laboratories showed an overall average of 91.5% interpretive agreement with the 17 antimicrobial agents tested. The distribution in the levels of Autobac 1/disk diffusion agreement was such that with 13 antimicrobial drugs agreement was 90% or higher; with three, between 85 and 90%; and with one, 77% (nitrofurantoin). Comparison of the Autobac 1 and disk diffusion tests with the International Collaborative Study agar dilution test showed that both methods gave levels of agreement with the International Collaborative Study agar dilution technique that were generally high and equivalent. The average overall agreement between the agar dilution test and each of the other two methods was approximately 90%. Disagreements that did occur tended to involve organisms that were drug susceptible by the Autobac 1 system but intermediate or resistant by the other two methods. This was in part due to the narrow intermediate interpretive zone of the Autobac 1 test. In reproducibility studies with the Autobac 1 and disk diffusion methods, no significant differences were observed between the interpretive reproducibility of the two methods.
In 1954, Schneierson (9) described a rapid (4 to 6 h) drug susceptibility test in which prepared disks of known antimicrobial content were used to provide specific drug concentrations in tubes of broth. This approach has been used in at least two automated systems, of which Autobac 1 is one (5, 6, 8, 12). The present report presents the results of a collaborative study designed to compare the Autobac 1 system, the Bauer-Kirby disk diffusion test, and an agar dilution technique. MATERIALS AND METHODS Reagents and media. Identical lots of media, disks, and antimicrobial agents were used in the susceptibility tests performed in each laboratory. The media used were tryptic soy agar, tryptic soy broth, Mueller-Hinton agar, and Eugonic broth, all manufactured by Pfizer Diagnostics and supplied by this firm to the investigators. The antimicrobial drug-con'Present address: University of Utah, Salt Lake City, Utah 84132. 2Present address: University of Minnesota, Minneapolis, Minn. 55455.
466
taining disks used in the Autobac 1 system (Table 1), and those used in the agar diffusion tests (Table 2), were also manufactured by Pfizer, Inc. However, in our study, some Pseudomonas aeruginosa strains were also tested with carbenicillin diffusion disks from Baltimore Biological Laboratories (BBL; Cockeysville, Md.). Antimicrobial powders that were used in the agar dilution susceptibility tests (Table 2) were either Pfizer products or were obtained from other manufacturers by Pfizer for use in these studies. Organisms. The organisms used in the study were current isolates from the clinical laboratories associated with the seven groups of investigators. The organisms were isolated and stored on media in routine use in each laboratory. For susceptibility testing, the organisms were streaked onto tryptic soy agar containing 5% sheep blood; cultures received from other laboratories were subcultured twice on this medium prior to testing. Three control cultures were used throughout the study; two of them were the "Seattle" Escherichia coli (derived from ATCC 25922) and Staphylococcus aureus (derived from ATCC 25923) strains recommended by the Food and Drug Administration and the third was the "Mayo" strain of P. aeruginosa. Susceptibility test methods. (i) Dilution method.
467
AUTOMATED SUSCEPTIBILITY TESTING SYSTEM
VOL. 7, 1975
TABLE 1. Antimicrobial elution disks used in the Autobac I system
The agar dilution method used by the International Collaborative Study group, as described by Ericsson and Sherris (3), was used in this study. Four to five Orga- morphologically similar colonies of each organism AsLabeled nisms sayed Disk Antimicrobial disk ma-s disk testwere inoculated into tryptic soy broth and incubated code agent (g) at 35 C. The turbidity of the suspension was adjusted (M) MaSSa ed8 to that of a 0.5 McFarland Standard, which had been shown to equate to within 5 x 107 to 5 x 108 N 6.7 Ampicillin .......... AM/en 3.6 colony-forming units (CFU) per ml. This suspension Carbenicillin ....... CB/en 120.0 142.0 N 19.4 P, N was diluted 1:10 with tryptic soy broth to contain Cephalothin ...... CL/ea 15.0 N 5.8 C/en 4.0 Chloramphenicol approximately 107 CFU/ml. Agar plates were pre2.3 P Clindamycin ........ CM/ep 2.0 pared to contain varying concentrations of each anN Colistin .......... 10.0 CS/en 13.0 agent by combining molten Mueller-Hintimicrobial 3.3 P Erythromycin ....... E/ep 2.5 ton agar (at 50 C) and the antimicrobial agent (di9.0 11.9 P, N Gentamicin ......... GM/ea luted in sterile, distilled water) in a ratio of nine parts N 19.8 K/en 22.0 Kanamycin ......... agar to one part antimicrobial solution. The cultures 5.0 7.0 P Methicillin ......... SC/ep were inoculated onto the Mueller-Hinton agar surface N 19.4 Nalidixic acidc ...... NA/en 15.0 . FD/en 15.0 21.0 N Nitrofurantoinc with the replicating device of Steers et al. (11). Since Penicillin G ......... P/ep 0.2 (U)d 0.26 P each inoculating point delivered approximately 0.001 N 13.9 Polymyxin B ........ PB/en 12.5 (U)d ml of inoculum, approximately 104 CFU of each P 0.6 Tetracycline ........ TE/ep 0.5 culture was inoculated on the surface of each plate. N 1.2 Tetracycline ........ TE/en 1.2 After the inocula dried, the plates were incubated at 9.9 Vancomycin ........ VA/ep 10.0 35 C for 18 to 24 h. The minimum inhibitory concena Antimicrobial agent eluted from disk and eluate assayed tration was defined as the least amount of antimiby the standard cup-plate method. Assays also performed by crobial agent employed in the test that permitted the the standard diffusion method. growth of no more than one colony. A slight haze, due b N, Gram-negative organisms and enterococcus; P, gram- to inoculum, was ignored. All minimal inhibitory positive cocci excluding enterococcus. concentrations were recorded as micrograms per milliUsed with urinary isolates only. d liter. Penicillin measured in units.
TABLE 2. Interpretive criteria for disk diffusion agar, dilution and Autobac 1 tests Antimicrobial agent
Ampicillin Staphylococci ..... Gram-negative organisms and enterococci Carbenicillin P. aeruginosa. E. coli, Proteus Sp..
Cephalothin. Chloramphenicol Clindamycin.
Disk diffusion zone diameter (mm)
ICS agar dilution minimal inhibitory conc. (tsg/ml)
Ra
Ja
Sa
29
14
> 64
< 12
13-14
> 15
> 256
< 17 < 14 < 12 < 14
18-22 15-17 13-17 15-16
9-10 14-17
Colistin .............
0.60 > 0.60 > 0.60 > 0.60
> 23
> 64
32
< 16
> 18
> 64
32, 16
18 >17
> 64
32 4
< 16
>8
>11 > 18 > 13
> 16
8
>8 > 16
4
< 0.50
8
12
> 16
0.60 > 0.60
......
Polymyxin B .. S
:> coC.))
\
;C
YZZ&E
Z .c . .
0Z
0)
.0
t C1 cc~ ~ ~~~~~~~~~c
0 M 0
00
0)
CC1
0
-{§:
4S
0
0
>
0
0)
0
0 Oco
a
0
0
0
O
0
4-e |.t
S 00)
o
..
_04 (6C ;( ;C o .S~~~~~~C Mroo~o
* _
X
40)
c; ; ; ;
0)
.
W6
0900-4 4c-4 66
t
I
-
co
~
o
0 0).-
I
0.
0
^
0^00 ^ 10--
10-
00
l.-
-
~ Q WC ;rO D o s° ^ co>°
~~.4 -~~ 4
.
-
S. i ~01'-400.-i .-E 4~~~~~~~~~~~~~~C 0N 00O"°o o 01 1 - 0 1
ooi
QA
.4
0)
0
6 6',6646.-
0
0. 0-i >1
eJ~~~~C 0
cL6o a d C6 0- 0( 0Oc -;C( 6~ ~ ~ C C ;~~~
*.
___
~~C'014
___
I 11 z03
I
I
c c 0
01 Er
4
0
~
0
0 '0 z
S.
0)
o.b0
to
2,
._
D la te
~
.0
2
.
:
40e ._
...4
:
:
-
-
0
C I=
I
:
0
._
8
AUTOMATED SUSCEPTIBILITY TESTING SYSTEM
VOL. 7, 1975
TABLE 10. Overall discrepanciesa between the disk diffusion and the agar dilution test results with
gram-positive organisms Antimicrobial agent
S. aureus
mSepidermpdis
Cephalothin Clindamycin . Erythromycin . Gentamicin ....... Methicillin ....... Penicillin G . Tetracycline .
0,0,0/120b 0,0,0/120 1,4,1/80 0,0,0/120 0,0,0/80 0,0,3/80 3,1,2/120
0,0,0,/5ih 0,0,0/36 100.0 100.0 0,0,0/51 93.0 2,0,0/34 99.4 1,0,0/51
Vancomycin .
0,0,0/80
0,0,0/34
0,0,0/34 1,0,6/34 3,4,1/51
Other
0,0,0/3
475
the mean inhibitory zone diameter determined by the collaborating investigators. No laboratory had individual test results outside the limits more frequently than would be expected.
Agreement
100.0 91.2 92.0 100.0
a Very major, susceptible by disk diffusion, resistant by agar dilution; major, resistant by disk diffusion, susceptible by agar dilution; minor, intermediate by only one of the two methods (disk diffusion, agar dilution). ' Numbers of very major, major, and minor discrepancies/ total number of strains tested.
cAverage % agreement, 97.1.
pairs tested by all seven collaborators by both tests (only six collaborators performed tests on colistin; therefore, this antimicrobial agent was eliminated from the equivalence studies). A total of 508 such test combinations were included in this analysis. In more than two-thirds of the 508 tests performed by the seven investigators on the 70 isolates (67.5%), all seven collaborators agreed on a single interpretive category by both tests (Table 16). However, at least five of the collaborators agreed on the same interpretive category by both tests 85.0% of the time (432 of 508 tests). In most of the remaining cases, five to seven of the collaborators agreed on an interpretive category by one of the two methods, while only one to four of the collaborators agreed on the same category by the remaining method. Failure of at least five of the seven collaborators to agree on an interpretive category by both methods occurred in only 1.2% of the tests. In 1% of the tests there was a major or very major discrepancy between the test method results, for example, when five or more collaborators agreed that the organism was resistant by one method, and five or more collaborators agreed it was susceptible by the other method (substantial disagreement, Table 16). Quality control data were available from five of the seven collaborating laboratories for the E. coli and S. aureus reference strains. Control limits (95%) were established for each antimicrobial agent/control organism combination by applying twice the maximum acceptable standard deviation described by the National Committee for Clinical Laboratory Standards (7) to
DISCUSSION These studies, which involved a total of 15,914 tests conducted with the Autobac 1 and the disk diffusion susceptibility methods, showed an overall agreement between the two procedures of approximately 90%. More than one-half the discrepancies were minor, i.e., one test recorded results in the intermediate category and the other recorded results in the susceptible or resistant categories. The majority of the remaining discrepancies involved organisms that were classified as resistant by the diffusion test and susceptible by the Autobac 1 test. A very closely similar pattern was seen when the Autobac 1 was compared with the dilution test. There was 90% agreement with 4.3% major and very major discrepancies. The overall agreement between the diffusion test procedure and the dilution method was 91%, with 2% major and very major discrepancies. Only with nitrofurantoin and ampicillin in the case of the Autobac 1 and with carbenicillin and nitrofurantoin in the case of the disk diffusion test were overall agreements less than 80%. With certain organism/antimicrobial agent combinations, more frequent discrepancies were encountered between the agar dilution results and those obtained by the other tests. More than 10% of the results with Staphylococcus epidermidis tested against penicillin and tetracycline showed substantial discrepancies when the Autobac 1 data were compared with the agar dilution data. Similar results with tetracycline were obtained when the diffusion test was compared with the agar dilution test. The reasons for these discrepancies are not clear, but with penicillin such discrepancies were partly influenced by the restricted definition of minimal inhibitory concentration correlates of susceptibility that were selected for the study. A similar question of definition may account for the high incidence of very major discrepancies in the Autobac 1/agar dilution comparison and of minor discrepancies in the disk diffusion/agar dilution comparison with carbenicillin and Proteus. The susceptibility testing of enterococci with cephalothin and of Pseudomonas sp. with kanamycin has been contra-indicated in the Autobac 1 method. All available data suggest that it is presently not possible to achieve a high level of Autobac 1/disk diffusion agreement for these
476
~ CD
THORNSBERRY ET AL.
-
0o c.)
0 0
-
o o a,
ootr c.)
o.
ANTIMICROB. AGENTS CHEMOTHER.
0 o
Ocs a
QO
0
.bCZ -o "0
c c,
0 Cs 00 N 4~~~~~-
1.
0
4)
0000
M
0
0o
CD
_. ._ U)
bL 0q rA 4-;
--40 0 L)
Z
C
C..
. C;
C
0-"
4
Z e 4)
4 0
00
e _
e
rn
a)
_~. 4)4 LO 0004> et0O>
w~~ O ~~~
cs
^ ^_o o
_ 00 _- 0 _ o c to
^ _
cl t:iS-Ab .2.w~~~~~I
s 0
0
ca
c 06 e e e CD
l
4~~~~~~V-3o oo o- o -4
O)
L N O -4 LOLo U- cq L-4O
O 0
5..
-
. °O0^0^-O0^0^0^~
4)0 0
0
rA
.°0n 0a0 c 4-
d~~~~)a)C
0)SX
. *4)i 4-)
0 44
._ Cu
o"
4)0
bo
._
a)
._
.0
0
.) C0
4
Q
Z
-
.
477
AUTOMATED SUSCEPTIBILITY TESTING SYSTEM
VOL. 7, 1975
TABLE 12. Estimated Standard Deviations (')a for Phase 1 and Phase 3 intra- and interlaboratory reproducibility; Autobac 1 and disk diffusion Autobac 1 (LSI) Organisms
Gram-negative organisms & enterococci Staphylococci
Disk diffusion (mm)
Interlab
Intralab
Intralab
Interlab
Overall
Range
Overall
Range
Overall
Range
Overall
Range
0.09
0.06-0.14
0.10
0.08-0.16
1.2
0.6-2.3
1.6
1.3-2.0
0.08
_b
0.03c
0.02-0.04c
1.4
0.7-2.1
1.4
1.1-2.0
a The ranges of &' listed reflect the different organisms or antimicrobial agents within that category of organisms, whereas the overall a' is a grand pooled estimate within the category. bToo few data to analyze by antimicrobial agent. c Only calculation which included 0.00 and 1.00 LSI values.
TABLE 13. Phase 1 and Phase 3 intralaboratory reproducilibity, by interpretive category, Autobac I and disk diffusion Interpretive agreement
Autobac la No. %
Disk diffusion No. %
91.4 Agreement 362 375 91.7 Substantial shiftb 20 5 1.2 5.1 Minor shiftc 14 3.5 29 7.1 a Cephalothin/enterococcus and kanamycin/ Pseudomonas aeruginosa contra-indications not included in data. b Shift from susceptible to resistant or vice versa on repeat testing. c Shift from susceptible or resistant to intermediate or vice versa on repeat testing.
two particular organism/antimicrobial agent pairs if a single elution disk for each of these antimicrobial agents is used. As a result of these findings, and due to the limited therapeutic value of cephalothin for enterococcus infections and of kanamycin for Pseudomonas infections,
the approach with cephalothin and kanamycin was to strive for good agreement with the other commonly isolated species (Tables 6 to 7) and to contra-indicate the single species in each instance. When these two instances of contraindication were included, an 85.4% agreement for cephalothin and an 86.7% agreement for kanamycin were obtained. Although a small increase in the incidence of minor discrepancies was affected by the inclusion of the contraindicated data, the principal change occurred in the incidences of very major discrepancies which increased from 2.4 and 0.4 to 7.5 and 4.5% with cephalothin and kanamycin, respectively. A considerable improvement in the Autobac 1/disk diffusion agreement was achieved with penicillin G, but not with the other antimi-
crobial agents, if the R < 0.90 LSI interpretive guideline was used. A more accurate estimation of staphylococcal susceptibility to this antibiotic was possible when this guideline was used because it allowed a larger resistant LSI range for Staphylococcus strains which presumably exhibit various levels of penicillinase production in a 3- to 5-h incubation period. Examination of the ampicillin results shows that the interpretative agreement was high and that incidences of substantial discrepancies were small for all of the species except Enterobacter. The principal source of disagreement between the two methods was the occurrence of susceptible Autobac 1 values for certain strains of the ampicillin-resistant population of Enterobacter sp. Two factors contributed to the large number of such discrepancies with Enterobacter and ampicillin. In many cases, there was growth in the cuvette, but it had settled to the bottom of the chamber in spite of continuous shaking. In such cases, the photometer read it as susceptible rather than resistant. This phenomenon also occurred to a lesser extent with Enterobacter and cephalothin. More recent studies at the University of Washington (Seattle) and at the Cleveland Clinic Foundation (Cleveland) have shown that the percentage of very major discrepancies can be approximately halved if the disk mass is kept within 80 to 125% of the nominal content, which was not the case in the study reported herein. Although there was still an average incidence of 15% very major discrepancies, these studies did not include visual observation for settled growth in the ampicillin chamber of the cuvette. Similar results were obtained with nitrofurantoin and Proteus sp. The only significant incidences of very major discrepancies that were encountered with this agent were those found with both P. mirabilis and indole-positive
478
THORNSBERRY ET AL.
ANTIMICROB. AGENTS CHEMOTHER.
TABLE 14. Distribution of interpretive agreement among the seven collaborators for the Autobac 1 and for the disk diffusion test results on 634 organism/antimicrobial agent pairs in phase 3 interlaboratory reproducibility studies Autobac 1 No. of collaborators agreeing on the same interpretive category
Organism/antimicrobial agent pairs
Disk diffusion
Test in
agreementa
No.
%
7
500
78.9
3500
6
64
10.1
5
28
4 3
Tests in dis-
agreementa
Organism/antimicrobial agent pairs
Tests
in agreementb
Tests in dis-
agreementb
No.
%
0
520
82.0
3640
0
384
64
51
8.0
306
51
4.4
140
56
33
5.2
165
66
29
4.6
116
87
21
3.3
84
63
13
2.5
39
52
9
1.4
27
36
aPercentage of Autobac 1 tests in agreement, 94.2; in disagreement, 5.8. b Percentage of disk diffusion tests in agreement, 95.1; in disagreement, 4.9. TABLE 15. Distribution of the interpretive disagreements of individual collaborators with the majority consensus encountered in 4,438 tests with the Autobac 1 and with the disk diffusion on 634 organism/antimicrobial agent pairs in Phase 3 interlaboratory reproducibility studies Individual collaborator disagreements
Type of disagreement with
majority
Autobac 1
Disk diffusion
consensus
Substantial" Minorc Equivocald
No.a
%
No.a
%
73 47 139
28.2 18.1
23 94
53.7
99
10.6 43.5 45.9
aIn this table, individual disagreements from the majority consensus must be considered since in many cases the minority collaborators disagreed with each other (e.g., 5 S; 1 I; 1 R). " Five or six collaborators classify as either susceptible or resistant and one or two collaborators classify the converse. c Either a majority (five or six) or a minority (one or two) of the collaborators classify in the intermediate =
category. d Four
=
=
or fewer collaborators interpretive category.
agree on any one
Proteus sp. The primary reason for these discrepancies was the excess potency of the elution disk used in this study. In more recent studies at the University of Washington and at the Cleveland Clinic Foundation, a nitrofurantoin elution disk was used that was within the recommended potency limits of 80 to 125% of the 15-,ug nominal content. These studies
TABLE 16. Summary Phase 3 Interlaboratory Equivalence, Autobac 1 vs. Disk diffusion Determinations
No.
Perfect agreement (7)b by both methods .343 Good agreement (5 or 6) by both methods .89 Substantial disagreementc between methods .5 Minor disagreemente between methods ..7 Acceptable agreement with disk diffusion, equivocal (.4) with Autobac 1 ........................... 35 Acceptable agreement with Autobac 1 equivocal (
