A customer in (b)(6) notified biomérieux of discrepant results associated with vitek® ms acquisition station ((b)(4)) related to misidentification of an organism.The customer reported on (b)(6) 2016 a strain from a blood culture identified as listeria grayi (99.9%).However, on (b)(6) 2016 the same patient sample strain from the heart valve was identified as leuconostoc mesenteroides (99.9%).The strains are confirmed with 16s method.Both strains were identified as actinotignum schaalii (formerly actinotignum schaalii).The results were discussed with the physician and patient treatment was started on (b)(6) 2016.On (b)(6) 2016 the results were also discussed with the physician and a special treatment for listeria-endocarditis was started.The drug administered also covers the actinotignum schaalii, but the interpretation of the susceptibility was done with a listeria.The actinotignum schaalii is a urine-pathogen, and listeria is not, therefore there was a delay of a few days in examination of the urinary tract infection.This treatment did not harm the patient.An internal biomérieux investigation will be initiated.
|
An internal biomérieux investigation was performed with results as follows: conclusion on the system: regarding of fine tuning acceptance criteria defined in the mar (mandatory action required) 3262 sent on december 20th 2016: fine tuning was needed during customer test.Conclusion on the identification: the expected species (actinobaculum schaalii) is not included in the vitek® ms database v2.0.Vitek® ms system identification is based on a species pattern classification.The system limitation is due to the use of a predictive modeling based on a supervised learning.Typically, such a model includes an algorithm that learns certain properties (peaks presence) from a training spectra dataset in order to make those predictions.When the microorganism tested is not part of the training dataset, no specific species pattern will be available in the database for comparison.Consequently, the system can give: no identification (most probable answer) when the spectrum acquired doesn't match with any species pattern.An incorrect single choice identification to the nearest pattern species (often same genus than expected) when the spectrum acquired presents high level of similarity with a specific species pattern present in the database.A low discrimination identification (often the same genus than expected) when the spectrum acquired presents high level of similarity with multiple specific species patterns present in the database.Customer's spectra obtained with knowledge base (kb) v2.0 were reprocessed with next vitek® ms kb v3.0 and results were improved: all samples gave single choice to actinobaculum schaalii.Suspected root cause of the issue: vitek ms system limitation, non-optimal fine tuning.Recommended actions : update the system to vitek® ms v3.0.Fine tuning status have to be monitored periodically (refer to the service manual).
|