Department of Clinical Laboratory, Virga Jesseziekenhuis, Stadsomvaart 11, B-3500 Hasselt, Belgium
Received 1 November 2002/ Returned for modification 30 December 2002/ Accepted 2 June 2002
The Osiris and Sirscan 2000 systems are two semiautomated systems
that can be used to read and interpret the results on disk diffusion
agar plates. They are both used for determination of susceptibility
to antimicrobial agents. The present study compared both systems
versus the NCCLS standard method of visual reading with a ruler.
Both inpatient and outpatient samples with a total of 315 nonfastidious
gram-negative strains were obtained. In total, 3,724 organism-antimicrobial
agent combinations that fulfilled the NCCLS guidelines for disk
diffusion susceptibility testing were evaluated prospectively.
The results obtained with both systems in comparison with those
obtained by the classical nonautomated means of interpretation
were excellent, with correlation coefficients of 0.96 for both
systems. The overall agreements for susceptibility interpretation
were 96.56 and 96.24% with the Osiris and Sirscan systems, respectively.
Very major errors were obtained for 8 (1.07%) and 10 (1.34%)
organism-antimicrobial agent combinations with the Osiris and
Sirscan systems, respectively. In addition, major errors were
obtained for 2 (0.07%) and 6 (0.21%) combinations with the Osiris
and Sirscan systems, respectively. Minor errors were obtained
for 118 and 124 organism-antimicrobial agent combinations with
the Osiris and Sirscan systems, respectively. Overall, both
the Osiris system and the Sirscan system are comparable and
reliable systems for determination of interpretative categories
from the zone diameters of standard disk diffusion test plates.
An important task for the diagnostic clinical microbiology laboratory
is the detection of clinically relevant antimicrobial resistance
in individual isolates (
4). Many different techniques are available
for the testing of susceptibility, including the disk diffusion
(Kirby Bauer), broth microdilution, agar dilution, and antibiotic
gradient methods (
5). A variety of automated instruments based
on these different methodologies have been introduced. The potential
advantages of automation include standardization, which results
in increased accuracy; the faster availability of results; improved
data management; and the possibility for the use of artificial
intelligence.
Kirby Bauer disk diffusion susceptibility testing merely provides susceptibility category results (susceptible, intermediate, and resistant). On the other hand, measurement of zone sizes is tedious, time-consuming, and fraught with transcription errors. Disk diffusion nevertheless offers certain advantages over other methods: low cost, the ability to test large numbers of organisms, and the ease with which different antimicrobial agents can be chosen (2). In addition, there is no evidence that determination of MICs is superior to determination of susceptibility category in patient management (4).
The Osiris system (Sanofi Pasteur) and the Sirscan 2000 system (i2a, Montpellier, France) are two semiautomated data management systems developed to combine the advantages of automation and disk diffusion. These video systems measure the inhibition zone diameter and interpret the results for disk diffusion susceptibility agar plates. The agar plate is placed onto a sliding tray, and the reading of the zone diameter is initiated by a keystroke. A clear image with fully calculated zone diameters appears on the video screen. Both manufacturers recommend a review of each plate by an experienced technologist so that adjustments can be made, if necessary. A user-programmed expert system screens the results for each isolate.
Three hundred fifteen gram-negative aerobic bacilli of 9 genera
and 12 species were included in this study (Table
1). In total,
3,724 organism-antimicrobial agent combinations were tested
prospectively. The organisms were isolated from inpatients and
outpatients at our hospital. The organisms tested are representative
of the rapidly growing nonfastidious gram-negative aerobe organism
mix recovered at our hospital. The isolates were obtained from
different sources: urine samples (41%), sputum and tracheal
aspirates (34%), blood (3%), and miscellaneous sources. The
antimicrobial agents tested depended on both the identification
and the source of the organism. Members of the family
Enterobacteriaceae isolated from urine were tested with amikacin, amoxicillin-clavulanic
acid, ampicillin, aztreonam, cefazolin, cefotaxime, ceftazidime,
cefuroxime, netilmicin, nitrofurantoin, norfloxacin, piperacillin,
and trimethoprim-sulfamethoxazole.
Escherichia coli strains
obtained from urine were also tested with fosfomycin.
Pseudomonas aeruginosa and
Acinetobacter sp. strains isolated from urine
were tested with amikacin, aztreonam, ceftazidime, netilmicin,
norfloxacin, and piperacillin. Members of the family
Enterobacteriaceae from sites other than urine were tested with amikacin, amoxicillin-clavulanic
acid, ampicillin, aztreonam, cefazolin, cefotaxime, ceftazidime,
cefuroxime, ciprofloxacin, meropenem, netilmicin, piperacillin,
and trimethoprim-sulfamethoxazole.
P. aeruginosa and
Acinetobacter sp. strains isolated from sources other than urine were tested
with amikacin, aztreonam, ceftazidime, ciprofloxacin, meropenem,
netilmicin, and piperacillin. Three thousand seven hundred twenty-four
organism-antimicrobial agent combinations met the criteria and
guidelines for susceptibility testing established by the NCCLS,
and category interpretations were made according to those guidelines
(
8).
| fig.ommitted |
TABLE 1. Organism groups, organisms, and number of isolates tested
| |
Disk diffusion susceptibility testing was performed according
to the NCCLS guidelines (
8). All inocula were prepared from
the growth of pure cultures of bacteria cultivated for approximately
18 to 24 h on MacConkey agar II (Becton Dickinson Bioscience)
or Columbia agar with 5% sheep blood (Becton Dickinson Bioscience)
at 35°C. Bacterial isolates were suspended in 5 ml of brain
heart infusion broth so that the turbidity was equivalent to
that of a 0.5 McFarland standard. All organisms were tested
on one square Mueller Hinton II agar plate (Becton Dickinson
Bioscience), as advised by Acar et al. (
1). Antibiotics were
applied to the plates by using antibiotic dispensers. These
susceptibility testing plates were incubated in ambient air
at 35°C. Inhibition zones were measured after 18 to 24 h
of incubation by one person only. For each susceptibility test,
visual ruler and video readings by both systems were made within
a time period of 2 h. For both semiautomated systems, the procedure
followed was that outlined in the instructions of each manufacturer.
Because both manufacturers recommend a review of the results,
zone diameters automatically measured by the video systems were,
if necessary, adjusted on screen by the reader (reviewed readings).
The visual reading obtained with a ruler was considered the
"gold standard." The zone diameters calculated with the Osiris
and Sirscan systems were compared to the zone diameters measured
with a ruler according to the NCCLS standard method (
8). Differences
in zone diameter measurements (in millimeters) were recorded.
In order to be able to compare our evaluation to the studies
of Acar et al. (
1) and Haddad-Prots et al. (
3), we adopted the
arbitrary threshold of 3 mm. That is, consensus was achieved
when the results of the various systems did not differ by more
than 3 mm. Interpretative categories (susceptible, intermediate,
and resistant) were calculated for each zone measurement for
each organism-antimicrobial agent combination tested. Discrepancies
in interpretative categories were noted. A very major error
was defined as a ruler reading interpretation of resistant and
a video reading interpretation of susceptible. A major error
was defined as a ruler reading interpretation of susceptible
and a video reading interpretation of resistant. The following
discrepancies were defined as minor errors: a ruler reading
interpretation of resistant and a video reading interpretation
of intermediate, a ruler reading interpretation of intermediate
and a video reading interpretation of susceptible, a ruler reading
interpretation of intermediate and a video reading interpretation
of resistant, and a ruler reading interpretation of susceptible
and a video reading interpretation of intermediate.
In total, 3,724 organism-antimicrobial agent combinations were
evaluated prospectively.
For the interpretation, we divided the 315 organisms into six groups (Table 1): E. coli (101 strains), Klebsiella species (42 strains), members of the family Enterobacteriaceae with an inducible ß-lactamase (64 strains), Proteus species (30 strains), P. aeruginosa (67 strains), and Acinetobacter species (11 strains).
The zone diameters were normally distributed for all organism-antimicrobial agent combinations. Because the visual ruler reading was considered the gold standard, the measure of agreement was determined with a population correlation coefficient in a constrained bivariate model (6). There were very good and similar correlations for the zone diameters of the reviewed readings for both the Osiris and the Sirscan systems versus the visual ruler readings. The analysis of variance for the Osiris and Sirscan systems resulted in intercepts of 0.51 and 0.51, respectively; slopes of 1.02 and 1, respectively; and correlation coefficients of 0.96 and 0.96, respectively. For all organism groups and antibiotics tested, the correlation coefficient exceeded 0.88 (Table 2) except for Osiris system readings with fosfomycin and Osiris and Sirscan system readings with amikacin.
| fig.ommitted |
TABLE 2. Correlation of visual ruler readings versus reviewed video readings
| |
Differences in zone measurements (in millimeters) between the
video readings and the visual ruler readings were determined.
Two thousand five hundred thirty-three (68.01%) reviewed readings
for the Osiris system and 2,533 (68.01%) reviewed readings for
the Sirscan system differed from the visual ruler readings.
The mean sizes of these differences for the Osiris and Sirscan
system readings versus the visual ruler readings were 2.31 ±
1.79 mm (standard deviation) and 2.19 ± 1.68 mm, respectively,
with zone diameters systematically being slightly larger for
the reviewed video readings. Differences that exceeded 3 mm
were noted for 340 (9.12%) and 319 (8.56%) of the organism-antimicrobial
agent combinations for the Osiris and Sirscan systems, respectively.
Interpretative category discrepancies were classified as very major errors, major errors, or minor errors (Table 3). The overall agreements for susceptibility interpretation were 96.82 and 94.59% for the Osiris and Sirscan systems, respectively, with simple kappa coefficients of 0.92 and 0.91 for the two systems, respectively.
| fig.ommitted |
TABLE 3. Interpretative category discrepancies for visual ruler readings versus reviewed video readings
| |
Eight (1.07%) and 10 (1.34%) organism-antimicrobial agent combinations
tested were found to be resistant by ruler reading and susceptible
by video reading with the Osiris and the Sirscan systems, respectively,
and were classified as very major errors. Two (0.07%) and 6
(0.21%) organism-antimicrobial agent combinations tested were
found to be susceptible by ruler reading and resistant by video
reading with the Osiris and the Sirscan systems, respectively,
and were classified as major errors. The results for 118 and
124 organism-antimicrobial agent combinations tested were classified
as minor errors with the Osiris and the Sirscan systems, respectively.
The Osiris and Sirscan semiautomated data management systems
read, interpret, and report the antimicrobial agent disk diffusion
susceptibility results.
The correlation of the zone diameter measurements of the reviewed Osiris system readings and the reviewed Sirscan system readings versus the visual ruler readings was good. The mean difference in the results obtained with both semiautomated systems in comparison with the results obtained by visual ruler reading was less than 2.5 mm. Zone diameters were systematically slightly larger for the video reading systems. In order to be able to compare our evaluation to the studies of Acar et al. (1) and Haddad-Prots et al. (3), we adopted the arbitrary threshold of a 3-mm zone diameter difference. In this setting, the results for 9.12 and 8.56% of the organism-antimicrobial agent combinations tested appeared to be discordant with the Osiris and Sirscan systems, respectively. These results are comparable to those reported by Acar et al. (1) and Haddad-Prots et al. (3).
In addition, these differences rarely affected the classification of the organisms as susceptible or resistant. The overall frequency of very major errors was low, i.e., 1.07% (8 of 742 combinations) and 1.34% (10 of 742 combinations) for the Osiris and Sirscan systems, respectively.
The study of Haddad-Prots et al. (3) compared Osiris system readings to manual caliper readings, and the studies of Medeiros et al. (7) and by Acar et al. (1) compared Sirscan system readings to manual (caliper) readings. The rates of very major errors reported in those three studies were comparable to those reported here (0.7, 0.3, and 1.76%, respectively). The differences between these studies are possibly due to the mix of species and antibiotics tested. Moreover, except for the Acinetobacter species strains, the discrepancies in results between bacterial species appeared to be random. A possible explanation for the discrepancies seen with the Acinetobacter species is the difficulty that the video systems have in visualizing the faint growth that sometimes occurs with this species. Another possible reason could be failure to swab the susceptibility plate thoroughly. If the growth of the organism is not confluent, the Osiris and Sirscan systems may read between the growth. An experienced technician can minimize these errors by reviewing the automatic readings prior to validation and adjusting them as needed. In this study the most likely reason for the discrepancies with the Acinetobacter species was the limited number of Acinetobacter-antibiotic combinations tested. More extensive evaluations with Acinetobacter species are recommended.
The antibiotics most often responsible for discrepant results were aztreonam, ciprofloxacin, meropenem, and norfloxacin. The inhibition zones obtained with aztreonam and ciprofloxacin were large, which may be a possible source of the discrepant readings. The positions of the meropenem and norfloxacin disks in the corner of the petri dish also hamper precise measurement. These reasons for discrepant results for antibiotics with large inhibition zones and because of the position of the disk on the plate were also previously suggested by Haddad-Prots et al. (3).
In conclusion, the performances of the Osiris and Sirscan image analyzers are very similar and satisfactory. We think that both systems will be of great value in clinical practice.
We are grateful to Biostatistics, Limburgs Universitair Centrum,
Diepenbeek, Belgium, for performing expert statistical analysis.
- Acar, J., O. Bellon, H. Chardon, J. Charrel, F. Goldstein, R. Labia, and J. Sirot. 1997. Statistical study of the Sirscan computerized camera. Pathol. Biol. 45:383-388. (In French.)
- Berke, I., and P. M. Tierno, Jr. 1996. Comparison of efficacy and cost-effectiveness of BIOMEC video and Vitek antimicrobial susceptibility test systems for use in the clinical microbiology laboratory. J. Clin. Microbiol. 34:1980-1984.
- Haddad-Prots, L., D. Assayag, V. Jarlier, and J. Nguyen. 1999. Evaluation of the Osiris video reader system for interpretation of disk diffusion susceptibility tests. Clin. Microbiol. Infect. 5(Suppl. 3):92.
- Jorgensen, J. H. 1993. Selection criteria for an antimicrobial susceptibility testing system. J. Clin. Microbiol. 31:2841-2844.
- Korgenski, E. K., and J. A. Daly. 1998. Evaluation of the BIOMEC video reader system for determining interpretative categories of isolates on the basis of disk diffusion susceptibility results. J. Clin. Microbiol. 36:302-304.
- Laurent, R. T. 1998. Evaluating agreement with a gold standard in method comparison studies. Biometrics 54:537-545.
- Medeiros, A. A., and J. Crellin. 2000. Evaluation of the Sirscan automated zone reader in a clinical microbiology laboratory. J. Clin. Microbiol. 38:1688-1693.
- National Committee for Clinical Laboratory Standards. 1999. Performance standards for antimicrobial disk susceptibility testing. NCCLS document M2-A6, M100-S9. National Committee for Clinical Laboratory Standards, Wayne, Pa.
作者：
A. Nijs R. Cartuyvels A. Mewis V. Peeters J. L 2007-5-10