
To evaluate the predictive power of screening microdilution tests for detection of ESBL phenotype using MIC criteria of Ceftriaxone, Ceftazidime and Cefotaxime in clinical isolates of enterobacteriaceae.
Methods:
Results of MIC microdilution test for Ceftriaxone, Ceftazidime and Cefotaxime (MIC > = 1 ug was considered as ESBL predictor), were compared to confirmation phenotypic criteria (using CLSI 2010 criteria: 3 fold increase comparing Ceftazidime´s MIC 0,25-128 ug/ml with Ceftazidime /Clavulanate 0,25-4 y 128-4 ug/ml) , 3 multivariate logistic regression models (One model for each antibiotic) were tested as predictors for screening test in order to detect ESBL phenotype (dependent variable). Independent variables were: 1) 1832 Clinical isolates of E.coli, K.pneumoniae or K.Oxitoca collected from patients with intraabdominal or urinary tract infection (Venezuela SMART project) in the period 2009-2014 2) epidemiologic data: patient age, year of sample collection (YOC) ,gender, location and sample type . Each model was evaluated separately using forward stepwise multivariate logistic regression methodology, estimating sensitivity, specificity and ROC (Receiver Operating Characteristic). Sub analysis for E.coli and K.pneumoniae prediction model was also evaluated. General significance criteria were 0.05. STATA 11.0 was used as statistical package
Results:
Even although every model has a good prediction capacity, Ceftazidime model shows the best ROC (0,98) p 0,001, independently of other covariates ( gender , age, YOC, location and type of sample
Model |
Variables |
|
P |
Sensib |
Specif |
ROC |
Ceftriaxone |
Ceftriaxone, Age, Gender , YOC |
|
0,001 |
91 |
97 |
94 |
Ceftazidime |
Cefatizidime, age,gender YOC |
|
0,001
|
99 |
96 |
98 |
Cefotaxime |
Cefotaxime , age , gender, YOC , |
|
0,001
|
92 |
96 |
95 |
Conclusion:
In settings were confirmation with secondary test for ESBL detection are not available, MIC >=1 of Ceftazidime is a good predictor of ESBL phenotype.

J. Castro, None