We found that in approximately two thirds of selected patients who underwent a PET/CT, the radio-diagnostic procedure resulted in one or more consequences for their management. New foci of infection were discovered in 51% of patients, and these new findings led to a new intervention and/or increase of antimicrobial treatment dosage and/or duration in 30.8% of patients. The absence of new diagnostic findings on the PET/CT regularly led to alterations in antimicrobial treatment dosage and duration as well, i.e., mostly de-escalations. This can be explained by the observation that in clinical practice, PET/CT was often used in patients who had one or several risk factors for complicated disease, but no apparent clinical manifestations of endocarditis or metastatic infection foci. If no new foci were detected by PET/CT, physicians may have felt assured that a shorter duration of therapy was safe. This also shows that, while PET-CT results can be a key driver of the medical team’s decisions to perform an intervention or change therapy, other variables, such as the patient’s clinical condition, play a relevant role as well.

The detection of a new infection focus did not lead to any traceable intervention or therapy alteration in 20.3% of patients. Although there may not be a direct objectifiable consequence of detection of a new metastatic focus of infection, its discovery may still have contributed to the clinicians’ understanding of the disease in the individual patient and, thereby, to the overall care that was provided. This is a variable that is difficult to scientifically quantify.

Results from a previous study already suggested that treatment duration can be safely de-escalated in ‘high-risk’ SAB patients without signs of metastatic infection on PET/CT and absence of signs of endocarditis on echocardiography [15]. Hence, the use of PET/CT to guide the medical management in SAB patients with risk factors for complicated disease, but in whom complicated disease has not yet been proven nor excluded, may offer a niche for the application of PET/CT.

In another study [16], PET/CT was recommended for patients with risk factors for complicated disease based on similar criteria as in our study. PET/CT was performed in 66.9% of patients with risk factors for complicated disease. Here, the authors reported treatment modifications in 74.6% of patients. Alterations in therapy dosage and/or duration as well as surgical or radiological interventions were considered as treatment modifications. Surgical or radiological interventions ensued in 19.2% of patients, compared to 9.8% in the current study. Despite that similar risk factor criteria were used, fewer surgical or radiological interventions ensued in the current study for which - apart from case-mix heterogeneity - we could find no clear explanation.

While there are guidelines that offer recommendations for the use of PET/CT, these are not strongly supported by scientific evidence. Due to the complex nature of SAB, other factors can influence the decision to perform a PET/CT, such as overall clinical presentation or the clinical course after admission. For example, from our data it was derived that PET/CT was performed less frequently in ICU patients with SAB (data not shown), which can be explained by challenges associated with patient mobility due to medical equipment and compromised health conditions.

Previous studies reported that new infection foci were found by PET/CT in 70.8% and 45.8% of patients [9, 16]. In both studies, bone- and joint infections (38.4% − 33.3%) comprised the largest groups of newly detected foci. In the current study, although osteomyelitis and arthritis were also found often, PET/CT mostly led to the discovery of soft tissue and lung foci (Table 2a). Another study from the Netherlands, also most frequently found tissue and lung foci [17]. This shows that the frequency and/or type of newly discovered infectionfoci is dependent on indications used for PET/CT use, as well as diagnostic work-up prior to PET/CT, but possibly also on geographical location influencing the case-mix.

To optimize the utilization of PET/CT in the context of personalized diagnostic assessment of patients with SAB, associations between baseline clinical factors and clinically relevant outcomes of PET/CT need to be considered. Preceding studies found a positive association between CRP and the detection of new foci, which this current study was able to replicate [16]. In this study, high CRP levels led to more new discoveries, while low CRP levels were protective for the detection of new foci. Notably, in our study, an association was found between the use of corticosteroids at presentation and fewer new infection foci discovered by PET/CT. This could be partly due to the anti-inflammatory properties of corticosteroids that could have complicated the detection of infectionfoci by PET/CT.

Regarding the implementation of PET-CT for patients with SAB, some previous studies have showed an association between performing PET/CT and a lower mortality rate [9, 17]. However, this difference in mortality can be attributed to immortal time bias because patients who passed away in the first week after SAB acquisition probably did not yet underwent PET/CT scanning [10]. When assessing mortality after the first week in the current study, no considerable difference in mortality rate was observed (Supplementary Figs. 1, 2).

The main strength of this research is its novel and extensive approach to investigating the revenues of PET/CT for SAB patients by considering three different clinically relevant aspects. Clinical data and revenues of PET/CT were evaluated by two independent expert clinicians and cross-checked, but subjectivity in determining what was a focus of infection or not, can never be fully excluded. The use of an AI based clinical data collection tool allowed for rapid, reliable, and systematic collection of patient data. Multiple steps of thorough data validation have been implemented to minimize the amount of missing data and capture errors, resulting in a complete and comprehensive database. The results of this study were obtained in a single academic tertiary care center and therefore may not be generalizable to other settings with different patient populations. Another potential limitation is the selection introduced by the real-life practice setting of the study, were not all patients with indications for PET/CT scans ultimately received a scan for various reasons e.g., non-adherence to ID-consultation recommendations or death, and not in all patients new BC cultures were obtained 48 h after start of treatment. This could lead to an overestimation of the information obtained- and actions instigated by PET/CT, as the selected patients may not represent the entire population of eligible SAB patients according to the local protocol or guideline. On the other hand, the design may also enhance the generalizability of the results, as findings were based on a real-life clinical practice setting. The reported 90-day relapse rate (1%) may be underestimated because patients with a relapse may incidentally have presented in a different hospital, without notification.

One of the aims of this study was to understand which clinical variables are associated with a consequential outcome of PET/CT. Despite this study being limited to showing an association, and not causation, the identified associations could still be applicable in a clinical setting. To draw causal conclusions, future research is warranted in the form of a diagnostic randomized controlled trial (RCT), with a group of patients being randomly divided between getting a PET/CT or not. This would better allow patient-specific predictions on PET/CT benefit and increase specificity of guidelines for PET/CT use. Furthermore, by performing a RCT, mortality and relapse rates can be more fairly assessed, as well as quality of life and cost-efficiency.

PET/CT led to the identification of new infection foci (NNT-scan = 2), the performance of new interventions (NNT-scan = 4), and alterations in therapeutic dosage and/or duration (NNT-scan = 3) in a substantial number of high-risk patients. The NNT-scan, although estimates from a single center study, were clinically acceptable, as relatively few patients had to undergo PET/CT to result in a new positive finding in one or more of these categories. Dialysis dependency and high CRP levels at baseline showed a strong association with interventions following PET/CT. To confirm these findings and to increase predictability of actionable PET/CT outcomes, a prospective diagnostic trial is needed.