In the present study, we assessed the performance of BCs and the T2Bacteria assay for the detection of bacterial pathogens in BSI, as well as TAT. Notably, the combined in-panel bacterial detection rate attributable to both methods was 15.8%, higher than bacteremia rates found in previous studies [9, 11]. Furthermore, T2 was positive in a higher proportion of cases than the BCs. Of particular interest is that most episodes - nearly three-quarters of all positive detections of in-panel bacteria - were characterized by discordant results where one test yielded a positive result while the other did not.

Almost half of all episodes were positive by T2 only, with no BCs sampled within the 72-hour window surrounding the T2 testing that were positive for the same isolate. For the episodes that had other non-BC microbiological samples for comparison, only 23% of the episodes had the same pathogen isolated from the other samples and could therefore be regarded as having a true positive T2 result.

As noted in previous studies [9, 11], finding a “gold standard” reference in T2 positive/BC negative cases is problematic, as there is no method of confirming that the test represents a true positive finding. This is demonstrated by the fact that, although higher than in the T2 positive/BC negative group, the proportion of episodes that had matching non-BC microbiological samples were only 59% in the BC positive/T2 negative group and 75% in the group with concordant BC and T2 results. This implies that the method of determining significance of isolated positive findings with T2 (or any novel BSI diagnostic method) used in this study as well as previous studies is not entirely reliable. Previous studies used a composite reference standard consisting of non-BC cultures and clinical adjudication. The latter, however, is influenced by the test itself which might lead to misclassification. The bacteria identified by T2 are among the most implicated in clinically significant infections and consequently, a positive T2 result, even in the absence of a supporting BC or a corresponding microbiological sample from another site, could indicate a true infection rather than a false positive. This observation is particularly crucial in settings in which prior antibiotic administration may have reduced the performance of BCs.

From the other viewpoint, around a quarter of all episodes were T2 negative but BC positive. The proportion of T2 negative/BC positive episodes are higher than reported in previous studies with a smaller sample size [11, 12, 14]. The reason for this difference is not obvious. One aspect is that most previous reports only included BCs sampled at the very same occasion as the T2 sampling. In addition, in several studies including the largest study by Nguyen et al. [9, 12], only one BC set was included as the comparator. It is well known that the practice of collecting only one BC set (e.g. solitary BC) limits the detection capabilities of BCs substantially, and a minimum of two BC sets are recommended in the guidelines [2, 18]. In contrast, the present study included all BCs during a time frame surrounding the T2 test, reaching a median number of three BCs per T2 test included. Therefore, the use of BCs as the comparator in the present study is used more closely in-line with the guidelines for BC sampling. This increases the diagnostic power of the comparator by both increasing the BC sampling volume and the number of BC sets. In our results, this was reflected by higher detection rate by BC in episodes with more BCs taken. Notably, in the 115 episodes that only included a single BC set, no additional diagnostic yield was achieved by BC.

When expanding the scope of detection to microorganisms not included in the T2 panel, BCs detected pathogens in additional 32 episodes, corresponding to a relative increase of 58%. This raises concerns about T2’s limited repertoire in clinical practice and underscores the need for a more comprehensive panel to capture a broader spectrum of bacterial infections as well as the complementary benefit of BCs. Polymicrobial BSI is increasingly common and ranges from 2 to 20% in current literature [19]. In the present study, the majority of polymicrobial BSIs included microorganisms not included in the T2 panel, suggesting that the limitation in detecting polymicrobial episodes is partly due to the limited repertoire with T2 bacteria. However, the polymicrobial sample size is small and warrants further study.

TAT is an important metric in determining the utility of a diagnostic method for BSI. As TAT for BCs still remains high due to the culture step, varying between 12 and 48 h for most cases [5, 12], molecular methods directly from blood have the potential to reduce TAT. As expected, we found that the total TAT was shorter with T2 than BC, however, the difference was partially balanced out by a substantial delay in the workflow for T2. Previous studies have shown a TAT around 3–7 h for T2 [9, 11, 12], however, providing only a portion of the actual TAT as the metric in these studies were defined as the sample processing time.

One concern has been that a longer TAT could negatively affect the performance of the T2 assay. However, this study shows that the TAT for positive and negative T2 samples did not significantly differ, providing reassurance against this concern. In most centers including the present study setting, T2 is a novel method and only available during work hours, which can lead to very long times to arrival. In addition, in settings centralized to one laboratory such as in the present study, delays imposed by geographic distances can be substantial. When examining how different sampling times affect TAT, the analysis revealed a significant difference between weekday and weekend sampling, as well as variations based on the time of day. This was attributed to a wider spread and more outliers during weekends and non-work hours. Importantly, median TAT was around 20 h even during weekday work-hour sampling.

The present study was a retrospective study, including only patients that underwent T2 sampling. Potentially, this approach might introduce a bias in patient selection as well as a limited panorama of studied infections. As this was a laboratory data study, we did not obtain comprehensive clinical data, and therefore the results are limited by lack of information about ongoing antibiotic use, indications for T2 sampling, and pertinent clinical data that could differentiate true from false positive results in the case of T2 positive/BC negative episodes. However, we used a similar approach as previous studies to assess the likelihood of true infection, considering additional microbiological samples in our assessment. The present study was performed at a single laboratory, limiting the generalizability of the data study and warrants caution when interpreting the results in an international context. However, the samples were collected from six large hospitals across Stockholm with a broad representation of patient categories, which in turn increases the external validity of the results.

There are several strengths to this study. Compared to prior studies of T2Bacteria [12,13,14, 20], the sample size is large, and the bacteremia rate is high. We have provided a conservative approach to the definition of positive BCs, taking the aggregate results of all BC samples within 72 h of T2 sampling as the comparative measure, with the majority of episodes including three or more BC sets. This served to ensure a proper BC sample volume and reduced the impact of potential intermittent bacteremia. In contrast to the controlled setting of “one T2 sample versus one BC sample” already studied in previous reports [9, 12], we have taken in account that clinical practice often includes several BC samplings across several time points during suspicion of infection. The TAT was for the first time described for T2Bacteria using an in-depth analysis, which reflect the actual sample-to-report time and not only the time for assay completion.