Accurate identification of S. pneumoniae using MALDI-TOF mass spectrometry, still a challenge for clinical laboratories?

Abstract

To the Editor Due to its high reliability and convenience for micro-organism identification, MALDI-TOF mass spectrometry is now implemented in a large number of clinical laboratories. However, the Microflex LT and the Compass software (Bruker Daltonics, Bremen, Germany) lack of performance for the accurate identification of Streptococcus pneumoniae, S. pseudopneumoniae, and other Streptococciwithin themitis/oralis group (SMOG). Several reports have proposed to solve this issue by implementing the supplier database with additional spectra [1], performing visual analysis of spectra [2, 3], or using algorithms [1, 4]. In comparison to routine use of MALDI-TOF mass spectrometry, most of these strategies are time-consuming, uneasy to perform, or still insufficiently accurate. Furthermore, most reports include a few numbe r o f c l i n i c a l s t r a i n s and mos t o f t en no S. pseudopneumoniae. Moreover, most of the strains are frozen collection strains that do not reflect routine use conditions. Consequently, none of these strategies were implemented by the manufacturer. The Compass database is continuously enriched with references spectra, and the software algorithm was improved in 2017. Since the Compass version 4.2., the software algorithm uses only the 2 first spectra with highest scores to provide identification, while previously the 10 first reference spectra were used. However, the supplier did not recommend to identify S. pneumoniae, S. pseudopneumoniae, or SMOG without additional tests such as optochin susceptibility or bile solubility. Considering these changes, we assessed the Compass software with the IVD7712 (V8) database performances for S. pneumoniae identification in the routine use, i.e., using species and scores of the 10 closest reference spectra provided by the software for each identification. This retrospective multicentric study was performed by the clinical laboratories of Hospital Foch (Suresnes, France), Nice Teaching Hospital (France), and Reims Teaching Hospital (France) from November 1st, 2018, to June 30th, 2019. All colonies cultured from respiratory samples, blood cultures or cerebrospinal fluids, showing morphology likely to be S. pneumoniae and an identification best score ≥ 2.0 were included. Clinical samples were cultured on Columbia plus 5% sheep blood or Chocolate agar plates incubated in anaerobic or aerobic enriched with 5% CO2 atmosphere. For each strain, we collected the following: (i) the date and hour of platting on agar media as well as MALDI-TOF identification; (ii) the score and identification of the two higher-ranked reference spectra matching; (iii) the number of occurrences of each species within the 10 reference spectra. Four algorithms for identification were tested: (i) algorithm 1 (manufacturer’s algorithm): 2 first reference spectra belonging to the same species or group; (ii) algorithm 2: 10 occurrences belonging to the same species/ group; (iii) algorithm 3: 8 occurrences belonging to the same species/group; (iv) algorithm 4: 2 first reference spectra and at least 6 occurrences belonging to the same species/group. Reference identification was provided by the optochin test as previously described [5]. MALDI-TOF identification was considered (i) correct if concordant with the reference test; (ii) misidentification, if discordant with the reference test; (iii) or absence of identification to the species level (no identification) if not reaching the algorithm criteria. Overall, 416 clinical strains were included: 169 (40.6%) S. pneumoniae, 236 (56.7%) SMOG, and 11 (2.6%) S. pseudopneumoniae. For S. pneumoniae identification, the algorithms displayed a sensitivity of 88.9%, 90.0%, 94.1%, and * Eric Farfour [email protected]; [email protected]