Microbiological culture
Mid-stream, clean catch or catheter urine of patients with bacteremia was collected in a sterile tube (Urin-Monovette®, Sarstedt, Nümbrecht, Germany). Urine samples (10 µl per plate) were cultured on Columbia Blood agar with 5% Sheep Blood (BD, Heidelberg, Germany; ambient air), MacConkey agar (BD, ambient air) and a selective agar for Gram-positive bacteria (Columbia CNA Agar with 5% Sheep Blood, BD, 5% CO2) for a maximum of 48 h at 35 ± 2°C. We defined SABU as detection of S. aureus in urine independent of the concentration (in CFU/ml) (27).
Pairs of aerobic and anaerobic blood culture bottles (BD) were incubated in a Bactec® FX device (BD) for five days (14 days in case of suspected endocarditis) and were subcultured on Columbia Blood agar (BD) and/or chocolate agar (BD, 5% CO2) when the blood culture bottles were flagged positive (28). Species identification was done with matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF MS, Biotyper® Sirius one, Bruker, Bremen, Germany) applying the Biotyper software IVD/Compass (version 4.2).
A screening test was used to detect substances that could inhibit the growth of bacteria in urine (e.g. antibiotics, cytostatic drugs). For that purpose, a blank cellulose disc (ThermoFisher scientific, Wesel, Germany) soaked with 10 µl urine was placed on a Mueller-Hinton II agar (BD, 35 ± 2°C, ambient air) inoculated with a highly susceptible bacterium (ATCC 6031 Bacillus subtilis). If an inhibition zone around the disc was seen after 24 hours (ambient air), the test was interpreted as positive. This screening was recommended in German guidelines until 2020 to aid the interpretation of negative urine cultures in patients with a high suspicion of urinary tract infection.
Nucleic acid amplification test for S. aureus
Urine samples were stored at -20°C until nucleic acid amplification test (NAAT) was done to guarantee equal bacterial cell counts for culture and NAAT. For NAAT, we used the Xpert® SA Nasal Complete G3 test cartridge (Cepheid, Krefeld, Germany) which was developed and validated for Methicillin-resistant S. aureus (MRSA)-Screening using nasal swabs and provides results within one hour. For the current study, we soaked a sterile swab (eSwab, Copan, Brescia, Italy) with urine, washed it in the buffer solution of the test cartridge and removed it afterwards. Then, we performed the test as recommended for nasal swabs. This procedure was chosen to use the test as recommended by the manufacturer with the only variation that urine instead of nasal swabs was sampled.
Analytical sensitivity
The limit of detection (LoD) is defined as the lowest number of colony forming units (CFU) per sample that can be distinguished from negative samples. We calculated the mean volume of urine that can be squeezed out of the swabs into the elution buffer of the cartridge test (= sample per test). Starting from a stock concentration of 2.45 x108/ml of ATCC 29213 S. aureus, we prepared a dilution series in urine with negative inhibitory test to obtain a certain CFU per sample Then, the cartridge test was carried out. The final CFU per sample was calculated from the mean value of the CFU after plating the same volume from the dilution sample that was given into the cartridge test on Columbia Blood agar with 5% Sheep Blood (BD; ambient air).
Statistical analysis
A specific sample size calculation was not done in the absence of any data on the performance of the NAAT in urine samples. We considered a samples size of 100 SAB and 20 non-SAB cases as appropriate for our objective. Differences between groups in categorical variables were compared using Fisher´s exact test or Chi-squared test (RStudio version 4.3.1; (29)) with a significance level at p < 0.05.
To define the optimal Ct-threshold for the prediction of SAB, we performed a receiver operating characteristic (ROC) analysis using the Ct-value as the predictor and SAB as the outcome. For samples with undetermined Ct-values (i.e. no detection after 40 cycles), we imputed a Ct-value and set a maximum value of 40. The ROC curves and optimal cut-off values (Youden-Index) were computed with the R package “cutpointr” (30).
The test performance of NAAT and cultural detection to predict SAB was calculated (sensitivity, specificity, positive predictive value [PPV], negative predictive value, [NPV]) using the optimal Ct-cut-off value as determined by the ROC analysis. The concordance of molecular and cultural test was calculated using Cohen’s kappa coefficient.