AAC Accepted Manuscript Posted Online 26 May 2015 Antimicrob. Agents Chemother. doi:10.1128/AAC.00868-15 Copyright © 2015, American Society for Microbiology. All Rights Reserved.
1
Comparative Evaluation of Colistin Susceptibility Testing Methods among Carbapenem-
2
non-Susceptible Klebsiella pneumoniae and Acinetobacter baumannii Clinical Isolates
3 4 5
Konstantina Dafopoulou,1,2 Olympia Zarkotou,1 Evangelia Dimitroulia,1 Christos
6
Hadjichristodoulou,2 Vasiliki Gennimata,1 Spyros Pournaras,1*and Athanasios Tsakris1
7 8 9
Department of Microbiology, Medical School, University of Athens, Athens1; Department
10
of Hygiene and Epidemiology, Medical School, University of Thessaly, Larissa2, Greece
11 12 13 14
Running title: Evaluation of Colistin Susceptibility Testing Methods
15 16 17 18
19
20
21
22
23
24
*Corresponding author. Mailing address: Department of Microbiology, University of
25
Athens, Athens, Greece. Phone: +30 210 7462126; fax: +30 210 7462210; e-mail:
26
[email protected]
27
28
We compared six colistin susceptibility testing (ST) methods in 61 carbapenem-non-
29
susceptible Klebsiella pneumoniae (n=41) and Acinetobacter baumannii (n=20) with
30
provisionally elevated colistin MICs by the routine ST. Colistin MICs were
31
determined by broth microdilution (BMD) as reference method, BMD with 0.002%
32
polysorbate 80 (P80) (BMD-P80), agar dilution (AD), Etest, Vitek2 and MIC test strip
33
(MTS). The EUCAST recommended susceptible/resistant breakpoint of ≤2/>2 μg/ml
34
was applied for both K. pneumoniae and A. baumannii. The proportion of colistin-
35
resistant strains was 95.1/77.0/96.7/57.4/65.6/98.4% by BMD/BMD-
36
P80/AD/Etest/MTS/Vitek2, respectively. Etest and MTS produced excessive rates of
37
very major errors (VMEs; 39.3 and 31.1%, respectively), while BMD-P80 18.0%
38
VMEs, AD 3.3% VMEs and Vitek2 no VME. Major errors (MEs) were rather limited by
39
all tested methods. These data show that gradient diffusion methods may lead to
40
inappropriate colistin therapy. Clinical laboratories should consider using
41
automated systems, such as Vitek2 or dilution methods for colistin ST.
42 43
Keywords: broth microdilution, agar dilution, very major errors, categorical agreement,
44
essential agreement
45
46
The increasing occurrence of multidrug-resistant (MDR) Acinetobacter baumannii,
47
Pseudomonas aeruginosa, and Enterobacteriaceae led to the revival of old and neglected
48
antibiotics that may remain active, such as polymyxins (polymyxin B and colistin) (1, 2).
49
Colistin is increasingly used as last-resort treatment option for infections caused by MDR
50
(2, 3), particularly carbapenem-resistant (CR) Gram-negative bacteria (4). However, during
51
the last years increasing colistin resistance emerged worldwide, especially among
52
Klebsiella pneumoniae and A. baumannii, further limiting treatment options (5-7). In
53
Europe, the evolving colistin resistance is more pronounced in southern countries (notably
54
Greece, Romania and Italy) (8-10).
55
Rapid and reliable colistin susceptibility testing (ST) is needed in routine clinical
56
laboratories to allow appropriate therapeutic decision making. Thus far, few studies
57
assessed the performance of colistin ST methods, displaying controversial results and thus
58
the most accurate one is still challenging (11).
59
Disk diffusion, commonly used in many clinical laboratories, yielded high error rates
60
compared to MIC based methods and is considered unreliable for detecting colistin
61
resistance (12-14). Among commercial methods, gradient diffusion strips are convenient
62
tests determining colistin MICs but their performance is not well established. Some studies
63
demonstrated excellent correlations between Etest (bioMérieux, Marcy l’ Etoile, France)
64
and broth microdilution (BMD) or agar dilution (AD) methods for colistin ST (13-17), while
65
other reports questioned its reliability (18, 19). Another gradient diffusion test, MIC Test
66
Strips (MTS, Liofilchem SRL, Italy) has not been evaluated for colistin ST to the best of our
67
knowledge. Colistin ST using automated methods has been test in limited studies that
68
mainly tested the performance of the Vitek2 system (bioMérieux) (13, 16, 19).
69
As standard methods for MIC ST are widely considered AD and BMD (20), which
70
however are not convenient for routine clinical laboratories. Additionally, for BMD,
71
technical issues, such as the type or surface pre-treatment of microtitre trays have
72
influenced significantly colistin MICs (3, 21). In particular, colistin displays varying
73
adherence to different surfaces used for MIC trays, such as polysterene, resulting in
74
reduced antibiotic concentrations actually present during experimental conditions (22).
75
The addition of the surfactant polysorbate-80 (P80) in BMD (BMD-P80) minimized colistin
76
adhesion to BMD panels and thus significantly reduced colistin MICs, mainly affecting
77
bacteria with relatively low MICs (≤2 μg/mL) (18, 21, 23). Nevertheless, CLSI does not
78
recommend using P80 for colistin ST by BMD (24). In addition, according to recent
79
observations, P80 exhibited synergistic effect with colistin, perhaps enhancing its
80
interaction with the bacterial cell membrane (25). Hence, further studies on the accuracy
81
of BMD-P80 in determining the susceptibility of Gram-negative bacteria are needed.
82
Breakpoints for colistin developed by various organizations differ (24, 26), further
83
complicating the interpretation of antimicrobial ST results. Additionally, most studies
84
investigating the accuracy of colistin ST methods involved predominantly colistin-
85
susceptible Gram-negative isolates, while they have scarcely been tested in colistin-
86
resistant strains. In this study, we evaluated various colistin MIC testing methods, such as
87
BMD, BMD-P80, AD, Etest, MTS and Vitek2 in a collection of carbapenem-non-susceptible
88
K. pneumoniae and A. baumannii strains with provisionally elevated colistin MICs according
89
to the routine ST.
90 91
MATERIALS AND METHODS
92
Bacterial isolates. A total of 61 carbapenem-non-susceptible clinical isolates collected
93
from four tertiary Greek hospitals during 2008-2013 was studied. In particular, this
94
collection contained nonduplicate previously characterized KPC-, OXA-48- or VIM-
95
carbapenemase-producing K. pneumoniae (n=41) and OXA-58- or OXA-23-carbapenemase-
96
producing A. baumannii isolates (n=20). The isolates were recovered from patients with
97
bloodstream (42.6%), respiratory tract (16.4%), urinary tract (16.4%), skin and soft tissue
98
(11.5%) and other infections (13.1%). Colistin was deemed necessary for the treatment of
99
the respective infections however the isolates were identified as colistin non-susceptible
100
(MIC > 2 μg/ml) by the routine laboratory ST, mainly based on automated systems. Due to
101
the existing controversies in colistin ST, to further evaluate the susceptibility results and
102
verify that colistin could not be used for infections caused by these bacteria, the isolates
103
were submitted to the reference laboratory. Prior to use in this study, the isolates had
104
been stored in glycerol stocks at -70°C and subcultured twice before testing.
105
Susceptibility testing. Colistin MICs were determined by BMD, BMD-P80, AD, Etest, MTS
106
and Vitek2. Dilution methods were performed according to the CLSI procedures (20, 24)
107
using colistin sulfate powder (Sigma-Aldrich, St. Louis, MO; Batch number SLBK0713V).
108
Stock solutions of colistin were reconstituted in sterile distilled water in accordance with
109
the manufacturer’s instructions immediately prior to use. BMD and BMD-P80 MIC
110
determination (concentrations range 0.125-128 μg/ml) was conducted on tissue-culture-
111
treated round-bottom polystyrene 96-well trays (Costar 3799; Corning, NY, USA), using a
112
bacterial inoculum of 5 Χ 105 CFU/ml in cation-adjusted Mueller-Hinton broth (Sigma-
113
Aldrich) without/with adding 0.002% P80 (Sigma-Aldrich). For AD, colistin was
114
incorporated in Mueller-Hinton II agar (MHA, Sigma-Aldrich) plates at concentrations
115
0.125-128 µg/ml and final inoculum 104 CFU/spot. Etest and MTS that include the same
116
colistin concentration gradient range (0.016-256 μg/ml) were performed on MHA plates
117
according to manufacturers’ instructions. Etest and MTS MICs between 2-fold dilutions
118
were rounded up to the next 2-fold dilution. The Vitek2 AST-EXN8 susceptibility card
119
(bioMérieux), reporting colistin MICs ≤0.5 to ≥16 μg/ml, was employed according to
120
manufacturer’s recommendations. For data analysis, when needed, MICs by Vitek2 ≥16
121
μg⁄ml were considered 16 μg⁄ml. All methods were performed simultaneously with a
122
single inoculum for each strain, incubated at 35 ± 2°C for 18 to 20 h and results reviewed
123
by two independent observers. Escherichia coli ATCC25922 and P. aeruginosa ATCC27853
124
were used for quality control (24).
125
Interpretation of results and data analysis. CLSI does not provide susceptibility
126
breakpoints for colistin against Enterobacteriaceae, but only for A. baumannii (susceptible,
127
MIC ≤ 2 μg/ml; resistant, MIC >4 μg/ml) (24). For consistency, EUCAST recommended
128
breakpoints, which are available for both species (susceptible, MIC ≤ 2 μg/ml; resistant,
129
MIC >2 μg/ml) (26) were applied.
130
Data were analyzed by comparing results of BMD-P80/AD/Etest/MTS/Vitek2 with
131
those produced by the standard BMD. Essential agreement (EA) was defined as the
132
percentage of MICs within ±1 log2 dilution of MIC determined by BMD. Categorical
133
agreement (CA) was defined as the percentage of isolates classified in the same
134
susceptibility category by BMD and method under evaluation. Very major error (VME)
135
denoted a false-susceptible result, and major error (ME) a false-resistant result (27). It
136
should be noted that because our collection included mainly colistin-resistant isolates and
137
to avoid overestimation of MEs, for the estimation of errors the total number of tested
138
isolates was used as denominator.
139 140
Acceptable performance was evaluated according to criteria established by the International Organization for Standardization: ≥ 90% for essential or category agreement,
141
≤ 3% for VME or ME (28). Statistical analysis was performed with Student’s t test and
142
differences were considered statistically significant at a P value of 3 log2 lower than
163
BMD for 29.5%, 18%, 19.7% and 11.5% of the isolates, respectively, resulting in the lowest
164
rate of EA (50.8% overall; 48.8/55% for K. pneumoniae/A. baumannii). Similarly, MICs
165
obtained by MTS were 1, 2, 3 and >3 log2 dilutions lower for 42.6%, 23%, 6.6% and 4.9% of
166
the isolates, respectively. MTS generated overall EA rate 65.6%, with low EA rate for K.
167
pneumoniae isolates (53.7%) but appropriate EA for A. baumannii (EA 90%). For Vitek2, the
168
EA rate was 78.6% in total; 31.1% of the isolates had MICs equal with those obtained by
169
BMD, while 47.5% of the isolates had one log2 dilution higher or lower MICs (Table 3).
170
In terms of errors (Table 1, Figure 1), high rates of VMEs (18.0%) and no ME were
171
detected for BMD-P80, while AD exhibited 3.3% VMEs and 4.9% MEs. Etest yielded overall
172
the highest rates of VMEs (39.3%) and limited MEs (1.6%); MTS produced high rates of
173
VMEs (31.1%) and also limited MEs (1.6%). No VME were observed for Vitek2; it yielded
174
3.3% MEs.
175 176
DISCUSSION
177
Colistin therapy is commonly necessary for the treatment of serious infections caused by
178
CR Gram-negative bacteria (4). Trends toward elevated colistin MICs have been noted
179
worldwide (5, 8) underlining the importance of accurate colistin susceptibility results. Also,
180
suggestions for the optimal methods to be used for colistin ST have not been formulated
181
by CLSI. In the present study, we evaluated the performance of six colistin ST methods
182
against carbapenem-non-susceptible K. pneumoniae and A. baumannii clinical isolates with
183
provisionally elevated colistin MICs.
184
Much debate on the need to add the surfactant P80 in test systems for polymyxins
185
has been recently arisen. Overall, available evidence for the performance of BMD-P80 in
186
strains with elevated MICs is currently limited. We observed that BMD-P80 showed the
187
highest EA and relatively high CA with BMD, but produced significantly lower MICs
188
(p128 1 1 0 0 0 0
mean 9.2 7.0 18.3 2.8 3.2 11.2
1 0 6 0 1 0
1 0 0 0 0 0
1 1 2 0 0 0
0 0 1 0 0 0
7.5 4.6 14.9 2.1 4.0 6.7
412
Table 3. Differences in log2 dilutions of MICs obtained by BMD-P80, AD, Etest, MTS and Vitek2 compared to BMD No. (%) of isolates showing a MIC difference (in log2 dilutions) of: Testing method and isolate >-3 -3 -2 -1 0 1 2 3 group BMD-P80 All isolates 3 (4.9) 24 (39.3) 34 (55.7) K. pneumoniae 2 (4.9) 12 (29.3) 27 (65.9) A. baumannii 1 (5) 12 (60) 7 (35) AD
All isolates K. pneumoniae A. baumannii
Etest
All isolates K. pneumoniae A. baumannii
MTS
Vitek2
413
3 (4.9) 2 (4.9) 1 (5.0)
3 (4.9) 2 (4.9) 1 (5.0)
15 (24.6) 11 (26.8) 4 (20.0)
16 (26.2) 10 (24.4) 6 (30)
7 (11.5) 12 (19.7) 4 (9.8) 7 (17.1) 3 (15) 5 (25)
11 (18) 10 (24.4) 1 (5)
18 (29.5) 12 (29.3) 6 (30)
12 (19.7) 7 (17.1) 5 (25)
1 (1.6) 1 (2.4)
All isolates K. pneumoniae A. baumannii
3 (4.9) 2 (4.9) 1 (5)
4 (6.6) 3 (7.3) 1 (5)
14 (23) 14 (34.1) 0 (0)
26 (42.6) 15 (36.6) 11 (55)
13 (21.3) 6 (14.6) 7 (35)
1 (1.6) 1 (2.4)
All isolates K. pneumoniae A. baumannii
1 (1.6) 1 (2.4)
2 (3.3) 0 (0) 2 (10)
2 (3.3) 2 (4.9) 0 (0)
11 (18) 7 (17.1) 4 (20.0)
19 (31.1) 11 (26.8) 8 (40)
18 (29.5) 13 (31.7) 5 (25)
18 (29.5) 11 (26.8) 7 (35)
8 (13.1) 7 (17.1) 1 (5)
6 (9.8) 5 (12.2) 1 (5.0)
418 419 420 421
Colistin AD MIC (μg/ml)
417
b)
Colistin BMD-P80 MIC (μg/ml)
a) 414 415 416
422
Colistin BMD MIC (μg/ml)
423 424
c)
Colistin BMD MIC (μg/ml)
d)
427 428 429 430 431
Colistin MTS MIC (μg/ml)
426
Colistin Etest MIC (μg/ml)
425
432 433 434
Colistin BMD MIC (μg/ml)
Colistin BMD MIC (μg/ml)
e)
436 437 438 439
Colistin Vitek2 MIC (μg/ml)
435
440 441
Colistin BMD MIC (μg/ml)
442
Figure 1. Scattergrams showing numbers of isolates (n=61) at colistin MICs obtained
443
by (a) BMD-P80, (b) AD, (c) Etest, (d) MTS, (e) Vitek2 versus BMD as reference. Solid
444
lines represent the 2014 EUCAST breakpoint for susceptibility (≤2 μg/ml). The
445
diagonal line indicates absolute agreement. VMEs are indicated by circles.
1
Comparative Evaluation of Colistin Susceptibility Testing Methods among Carbapenem-
2
non-Susceptible Klebsiella pneumoniae and Acinetobacter baumannii Clinical Isolates
3 4 5
Konstantina Dafopoulou,1,2 Olympia Zarkotou,1 Evangelia Dimitroulia,1 Christos
6
Hadjichristodoulou,2 Vasiliki Gennimata,1 Spyros Pournaras,1*and Athanasios Tsakris1
7 8 9
Department of Microbiology, Medical School, University of Athens, Athens1; Department
10
of Hygiene and Epidemiology, Medical School, University of Thessaly, Larissa2, Greece
11 12 13 14
Running title: Evaluation of Colistin Susceptibility Testing Methods
15 16 17 18
19
20
21
22
23
24
*Corresponding author. Mailing address: Department of Microbiology, University of
25
Athens, Athens, Greece. Phone: +30 210 7462126; fax: +30 210 7462210; e-mail:
26
[email protected]
27
28
We compared six colistin susceptibility testing (ST) methods in 61 carbapenem-non-
29
susceptible Klebsiella pneumoniae (n=41) and Acinetobacter baumannii (n=20) with
30
provisionally elevated colistin MICs by the routine ST. Colistin MICs were
31
determined by broth microdilution (BMD) as reference method, BMD with 0.002%
32
polysorbate 80 (P80) (BMD-P80), agar dilution (AD), Etest, Vitek2 and MIC test strip
33
(MTS). The EUCAST recommended susceptible/resistant breakpoint of ≤2/>2 μg/ml
34
was applied for both K. pneumoniae and A. baumannii. The proportion of colistin-
35
resistant strains was 95.1/77.0/96.7/57.4/65.6/98.4% by BMD/BMD-
36
P80/AD/Etest/MTS/Vitek2, respectively. Etest and MTS produced excessive rates of
37
very major errors (VMEs; 39.3 and 31.1%, respectively), while BMD-P80 18.0%
38
VMEs, AD 3.3% VMEs and Vitek2 no VME. Major errors (MEs) were rather limited by
39
all tested methods. These data show that gradient diffusion methods may lead to
40
inappropriate colistin therapy. Clinical laboratories should consider using
41
automated systems, such as Vitek2 or dilution methods for colistin ST.
42 43
Keywords: broth microdilution, agar dilution, very major errors, categorical agreement,
44
essential agreement
45
46
The increasing occurrence of multidrug-resistant (MDR) Acinetobacter baumannii,
47
Pseudomonas aeruginosa, and Enterobacteriaceae led to the revival of old and neglected
48
antibiotics that may remain active, such as polymyxins (polymyxin B and colistin) (1, 2).
49
Colistin is increasingly used as last-resort treatment option for infections caused by MDR
50
(2, 3), particularly carbapenem-resistant (CR) Gram-negative bacteria (4). However, during
51
the last years increasing colistin resistance emerged worldwide, especially among
52
Klebsiella pneumoniae and A. baumannii, further limiting treatment options (5-7). In
53
Europe, the evolving colistin resistance is more pronounced in southern countries (notably
54
Greece, Romania and Italy) (8-10).
55
Rapid and reliable colistin susceptibility testing (ST) is needed in routine clinical
56
laboratories to allow appropriate therapeutic decision making. Thus far, few studies
57
assessed the performance of colistin ST methods, displaying controversial results and thus
58
the most accurate one is still challenging (11).
59
Disk diffusion, commonly used in many clinical laboratories, yielded high error rates
60
compared to MIC based methods and is considered unreliable for detecting colistin
61
resistance (12-14). Among commercial methods, gradient diffusion strips are convenient
62
tests determining colistin MICs but their performance is not well established. Some studies
63
demonstrated excellent correlations between Etest (bioMérieux, Marcy l’ Etoile, France)
64
and broth microdilution (BMD) or agar dilution (AD) methods for colistin ST (13-17), while
65
other reports questioned its reliability (18, 19). Another gradient diffusion test, MIC Test
66
Strips (MTS, Liofilchem SRL, Italy) has not been evaluated for colistin ST to the best of our
67
knowledge. Colistin ST using automated methods has been test in limited studies that
68
mainly tested the performance of the Vitek2 system (bioMérieux) (13, 16, 19).
69
As standard methods for MIC ST are widely considered AD and BMD (20), which
70
however are not convenient for routine clinical laboratories. Additionally, for BMD,
71
technical issues, such as the type or surface pre-treatment of microtitre trays have
72
influenced significantly colistin MICs (3, 21). In particular, colistin displays varying
73
adherence to different surfaces used for MIC trays, such as polysterene, resulting in
74
reduced antibiotic concentrations actually present during experimental conditions (22).
75
The addition of the surfactant polysorbate-80 (P80) in BMD (BMD-P80) minimized colistin
76
adhesion to BMD panels and thus significantly reduced colistin MICs, mainly affecting
77
bacteria with relatively low MICs (≤2 μg/mL) (18, 21, 23). Nevertheless, CLSI does not
78
recommend using P80 for colistin ST by BMD (24). In addition, according to recent
79
observations, P80 exhibited synergistic effect with colistin, perhaps enhancing its
80
interaction with the bacterial cell membrane (25). Hence, further studies on the accuracy
81
of BMD-P80 in determining the susceptibility of Gram-negative bacteria are needed.
82
Breakpoints for colistin developed by various organizations differ (24, 26), further
83
complicating the interpretation of antimicrobial ST results. Additionally, most studies
84
investigating the accuracy of colistin ST methods involved predominantly colistin-
85
susceptible Gram-negative isolates, while they have scarcely been tested in colistin-
86
resistant strains. In this study, we evaluated various colistin MIC testing methods, such as
87
BMD, BMD-P80, AD, Etest, MTS and Vitek2 in a collection of carbapenem-non-susceptible
88
K. pneumoniae and A. baumannii strains with provisionally elevated colistin MICs according
89
to the routine ST.
90 91
MATERIALS AND METHODS
92
Bacterial isolates. A total of 61 carbapenem-non-susceptible clinical isolates collected
93
from four tertiary Greek hospitals during 2008-2013 was studied. In particular, this
94
collection contained nonduplicate previously characterized KPC-, OXA-48- or VIM-
95
carbapenemase-producing K. pneumoniae (n=41) and OXA-58- or OXA-23-carbapenemase-
96
producing A. baumannii isolates (n=20). The isolates were recovered from patients with
97
bloodstream (42.6%), respiratory tract (16.4%), urinary tract (16.4%), skin and soft tissue
98
(11.5%) and other infections (13.1%). Colistin was deemed necessary for the treatment of
99
the respective infections however the isolates were identified as colistin non-susceptible
100
(MIC > 2 μg/ml) by the routine laboratory ST, mainly based on automated systems. Due to
101
the existing controversies in colistin ST, to further evaluate the susceptibility results and
102
verify that colistin could not be used for infections caused by these bacteria, the isolates
103
were submitted to the reference laboratory. Prior to use in this study, the isolates had
104
been stored in glycerol stocks at -70°C and subcultured twice before testing.
105
Susceptibility testing. Colistin MICs were determined by BMD, BMD-P80, AD, Etest, MTS
106
and Vitek2. Dilution methods were performed according to the CLSI procedures (20, 24)
107
using colistin sulfate powder (Sigma-Aldrich, St. Louis, MO; Batch number SLBK0713V).
108
Stock solutions of colistin were reconstituted in sterile distilled water in accordance with
109
the manufacturer’s instructions immediately prior to use. BMD and BMD-P80 MIC
110
determination (concentrations range 0.125-128 μg/ml) was conducted on tissue-culture-
111
treated round-bottom polystyrene 96-well trays (Costar 3799; Corning, NY, USA), using a
112
bacterial inoculum of 5 Χ 105 CFU/ml in cation-adjusted Mueller-Hinton broth (Sigma-
113
Aldrich) without/with adding 0.002% P80 (Sigma-Aldrich). For AD, colistin was
114
incorporated in Mueller-Hinton II agar (MHA, Sigma-Aldrich) plates at concentrations
115
0.125-128 µg/ml and final inoculum 104 CFU/spot. Etest and MTS that include the same
116
colistin concentration gradient range (0.016-256 μg/ml) were performed on MHA plates
117
according to manufacturers’ instructions. Etest and MTS MICs between 2-fold dilutions
118
were rounded up to the next 2-fold dilution. The Vitek2 AST-EXN8 susceptibility card
119
(bioMérieux), reporting colistin MICs ≤0.5 to ≥16 μg/ml, was employed according to
120
manufacturer’s recommendations. For data analysis, when needed, MICs by Vitek2 ≥16
121
μg⁄ml were considered 16 μg⁄ml. All methods were performed simultaneously with a
122
single inoculum for each strain, incubated at 35 ± 2°C for 18 to 20 h and results reviewed
123
by two independent observers. Escherichia coli ATCC25922 and P. aeruginosa ATCC27853
124
were used for quality control (24).
125
Interpretation of results and data analysis. CLSI does not provide susceptibility
126
breakpoints for colistin against Enterobacteriaceae, but only for A. baumannii (susceptible,
127
MIC ≤ 2 μg/ml; resistant, MIC >4 μg/ml) (24). For consistency, EUCAST recommended
128
breakpoints, which are available for both species (susceptible, MIC ≤ 2 μg/ml; resistant,
129
MIC >2 μg/ml) (26) were applied.
130
Data were analyzed by comparing results of BMD-P80/AD/Etest/MTS/Vitek2 with
131
those produced by the standard BMD. Essential agreement (EA) was defined as the
132
percentage of MICs within ±1 log2 dilution of MIC determined by BMD. Categorical
133
agreement (CA) was defined as the percentage of isolates classified in the same
134
susceptibility category by BMD and method under evaluation. Very major error (VME)
135
denoted a false-susceptible result, and major error (ME) a false-resistant result (27). It
136
should be noted that because our collection included mainly colistin-resistant isolates and
137
to avoid overestimation of MEs, for the estimation of errors the total number of tested
138
isolates was used as denominator.
139 140
Acceptable performance was evaluated according to criteria established by the International Organization for Standardization: ≥ 90% for essential or category agreement,
141
≤ 3% for VME or ME (28). Statistical analysis was performed with Student’s t test and
142
differences were considered statistically significant at a P value of 3 log2 lower than
163
BMD for 29.5%, 18%, 19.7% and 11.5% of the isolates, respectively, resulting in the lowest
164
rate of EA (50.8% overall; 48.8/55% for K. pneumoniae/A. baumannii). Similarly, MICs
165
obtained by MTS were 1, 2, 3 and >3 log2 dilutions lower for 42.6%, 23%, 6.6% and 4.9% of
166
the isolates, respectively. MTS generated overall EA rate 65.6%, with low EA rate for K.
167
pneumoniae isolates (53.7%) but appropriate EA for A. baumannii (EA 90%). For Vitek2, the
168
EA rate was 78.6% in total; 31.1% of the isolates had MICs equal with those obtained by
169
BMD, while 47.5% of the isolates had one log2 dilution higher or lower MICs (Table 3).
170
In terms of errors (Table 1, Figure 1), high rates of VMEs (18.0%) and no ME were
171
detected for BMD-P80, while AD exhibited 3.3% VMEs and 4.9% MEs. Etest yielded overall
172
the highest rates of VMEs (39.3%) and limited MEs (1.6%); MTS produced high rates of
173
VMEs (31.1%) and also limited MEs (1.6%). No VME were observed for Vitek2; it yielded
174
3.3% MEs.
175 176
DISCUSSION
177
Colistin therapy is commonly necessary for the treatment of serious infections caused by
178
CR Gram-negative bacteria (4). Trends toward elevated colistin MICs have been noted
179
worldwide (5, 8) underlining the importance of accurate colistin susceptibility results. Also,
180
suggestions for the optimal methods to be used for colistin ST have not been formulated
181
by CLSI. In the present study, we evaluated the performance of six colistin ST methods
182
against carbapenem-non-susceptible K. pneumoniae and A. baumannii clinical isolates with
183
provisionally elevated colistin MICs.
184
Much debate on the need to add the surfactant P80 in test systems for polymyxins
185
has been recently arisen. Overall, available evidence for the performance of BMD-P80 in
186
strains with elevated MICs is currently limited. We observed that BMD-P80 showed the
187
highest EA and relatively high CA with BMD, but produced significantly lower MICs
188
(p128 1 1 0 0 0 0
mean 9.2 7.0 18.3 2.8 3.2 11.2
1 0 6 0 1 0
1 0 0 0 0 0
1 1 2 0 0 0
0 0 1 0 0 0
7.5 4.6 14.9 2.1 4.0 6.7
412
Table 3. Differences in log2 dilutions of MICs obtained by BMD-P80, AD, Etest, MTS and Vitek2 compared to BMD No. (%) of isolates showing a MIC difference (in log2 dilutions) of: Testing method and isolate >-3 -3 -2 -1 0 1 2 3 group BMD-P80 All isolates 3 (4.9) 24 (39.3) 34 (55.7) K. pneumoniae 2 (4.9) 12 (29.3) 27 (65.9) A. baumannii 1 (5) 12 (60) 7 (35) AD
All isolates K. pneumoniae A. baumannii
Etest
All isolates K. pneumoniae A. baumannii
MTS
Vitek2
413
3 (4.9) 2 (4.9) 1 (5.0)
3 (4.9) 2 (4.9) 1 (5.0)
15 (24.6) 11 (26.8) 4 (20.0)
16 (26.2) 10 (24.4) 6 (30)
7 (11.5) 12 (19.7) 4 (9.8) 7 (17.1) 3 (15) 5 (25)
11 (18) 10 (24.4) 1 (5)
18 (29.5) 12 (29.3) 6 (30)
12 (19.7) 7 (17.1) 5 (25)
1 (1.6) 1 (2.4)
All isolates K. pneumoniae A. baumannii
3 (4.9) 2 (4.9) 1 (5)
4 (6.6) 3 (7.3) 1 (5)
14 (23) 14 (34.1) 0 (0)
26 (42.6) 15 (36.6) 11 (55)
13 (21.3) 6 (14.6) 7 (35)
1 (1.6) 1 (2.4)
All isolates K. pneumoniae A. baumannii
1 (1.6) 1 (2.4)
2 (3.3) 0 (0) 2 (10)
2 (3.3) 2 (4.9) 0 (0)
11 (18) 7 (17.1) 4 (20.0)
19 (31.1) 11 (26.8) 8 (40)
18 (29.5) 13 (31.7) 5 (25)
18 (29.5) 11 (26.8) 7 (35)
8 (13.1) 7 (17.1) 1 (5)
6 (9.8) 5 (12.2) 1 (5.0)
418 419 420 421
Colistin AD MIC (μg/ml)
417
b)
Colistin BMD-P80 MIC (μg/ml)
a) 414 415 416
422
Colistin BMD MIC (μg/ml)
423 424
c)
Colistin BMD MIC (μg/ml)
d)
427 428 429 430 431
Colistin MTS MIC (μg/ml)
426
Colistin Etest MIC (μg/ml)
425
432 433 434
Colistin BMD MIC (μg/ml)
Colistin BMD MIC (μg/ml)
e)
436 437 438 439
Colistin Vitek2 MIC (μg/ml)
435
440 441
Colistin BMD MIC (μg/ml)
442
Figure 1. Scattergrams showing numbers of isolates (n=61) at colistin MICs obtained
443
by (a) BMD-P80, (b) AD, (c) Etest, (d) MTS, (e) Vitek2 versus BMD as reference. Solid
444
lines represent the 2014 EUCAST breakpoint for susceptibility (≤2 μg/ml). The
445
diagonal line indicates absolute agreement. VMEs are indicated by circles.
