[18F]FDG-PET/CT in Staphylococcus aureus bacteremia: a systematic review

This systematic review was performed according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement [13]. The study protocol was prospectively registered with PROSPERO (CRD42021238077). We included all randomized controlled trials, clinically controlled trials, prospective and retrospective cohort studies, and case–control studies investigating the effects of [18F]FDG-PET/CT in adult patients with SAB. Case reports, systematic reviews, and meta-analyses were excluded. We also excluded studies lacking a control group without [18F]FDG-PET/CT, studies including less than 20 patients, duplicate studies, and studies without full text available.

Primary outcomes were all-cause mortality rate as defined by the authors and new diagnostic findings on [18F]FDG-PET/CT related to SAB. Secondary outcomes included infection relapse, classification of SAB as complicated or uncomplicated as defined by the authors, change of antibiotic regimen or antibiotic treatment duration, source control interventions, and rate of non-infection related accidental findings on [18F]FDG-PET/CT. Studies which did not investigate at least one of the prespecified outcomes were excluded. In the qualitative synthesis we included only outcome measures for which a formal comparison was made with a group that did not undergo [18F]FDG-PET/CT.

A systematic search was performed on 29th of January 2021 (by GBL and DTPB), using the databases PubMed, Embase.com, Clarivate Analytics/Web of Science Core Collection and the Wiley/Cochrane Library. The search included keywords and free text terms for (synonyms of) ‘Staphylococcus aureus’ combined with (synonyms of) ‘positron emission tomography’. Animal studies were excluded. A full overview of the search terms per database can be found in the supplementary information (see Additional file 1: Appendix SA). No limitations on date or language were applied. The references of the articles included in the qualitative synthesis were searched manually for relevant publications. We manually searched www.​clinicaltrials.​gov and the International Clinical Trials Registry Platform for ongoing studies. We approached authors of studies that fulfilled all inclusion criteria, but did not report study outcomes separately for SAB patients to obtain these data. Two investigators (DTPB and WYH) independently evaluated all identified studies based on the in- and exclusion criteria as defined in the review protocol. Any discrepancies were resolved by discussion. If no consensus was reached the final decision was made by a third investigator (ES). We used a standardized electronic form for data extraction to collect data concerning study design, population, intervention characteristics, and results. Two investigators (DTPB and WYH) extracted the data using this form and reported independently if data planned to extract was missing. Any discrepancies were resolved by discussion.

Two investigators (DTPB and WYH) independently assessed the certainty of the body evidence at the outcome level using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) methodology. Risk of bias of non-randomized studies was assessed at the study level using the ROBINS-I tool by two investigators (DTPB and WYH) independently. In case of discrepancies between the two authors, a third investigator (ES) was consulted. The ROBINS-I is the preferred tool for risk of bias assessment in non-randomized studies according to the Cochrane Collaboration [14]. This tool distinguishes seven different bias domains: bias due to confounding, bias in selection of participants, bias in classification of intervention, bias due to deviations from intended interventions, bias due to missing data, bias in measurement of the outcome, and bias in selection of the reported result. The full methods of our risk of bias assessment is set out in Additional file 2: Appendix SB. We did not assess publication bias with funnel plot inspection, because of the limited number of studies included in the qualitative synthesis.

We planned to perform a meta-analysis including sufficiently homogenous studies with regard to study design and study outcomes, i.e. prospective studies with a control group, performance of [18F]FDG-PET/CT within 14 days of diagnosis of SAB, and similar reporting of study outcomes. Since only one study fulfilled these criteria we did not perform a quantitative synthesis [15].

We identified 1956 records through database searches: 373 through PubMed, 1002 through Embase, 569 through Web of Science, and 12 through the Cochrane Library (Fig. 1). After removal of duplicates we screened 1437 records and assessed 44 full text articles for eligibility. We included five studies in our final qualitative synthesis. The most common reason for exclusion during the full text assessment was lack of a control group without [18F]FDG-PET/CT (n = 26). Review of references of articles included in the qualitative syntheses did not result in additional studies. We retrieved one relevant ongoing study via manual searching of www.​clinicaltrials.​gov and the International Clinical Trials Registry Platform (ICTRP) [16]. We obtained data on 3 month mortality in SAB patients by approaching the authors of one of the studies [17].

Table 1 displays the characteristics of the five included studies, including a total of 880 patients [15, 17‐20]. Three of the studies were performed in The Netherlands, one in Belgium and one in Israel. All included studies were non-randomized. 3 studies were retrospective cohort studies, one a prospective cohort study, and one a prospective cohort study with historical controls. Prevalence of methicillin resistant Staphylococcus aureus varied between 0 and 22%. Three studies only included SAB patients with risk factors for metastatic infections. Risk factors in these studies included community acquisition, signs of infection more than 48 h before initiation of appropriate treatment, fever more than 72 h after initiation of appropriate treatment, positive blood cultures more than 48 h after initiation of appropriate treatment or presence of foreign body materials. In the other two studies the prevalence of risk factors was 46 and 80%, respectively. In all five studies, median or mean time between diagnosis of SAB and performance of [18F]FDG-PET/CT was 11 days or less (range 7–11 days). Follow-up duration ranged between 3 months and one year.

The main results of included studies are shown in Table 2. Four studies compared 3-month mortality rates between an intervention group who underwent [18F]FDG-PET/CT and a control group who did not[15, 17‐19]. All four reported lower 3-month mortality rates in the intervention group. Three of these studies reported respectively a 7, 15 and 21% reduction in mortality and one did not provide an effect estimate [15‐18]. The results of these studies are depicted in Fig. 2. Moreover, three of these studies observed an association between performance of [18F]FDG-PET/CT and lower mortality in multivariate regression analyses with adjustment for confounding variables. Two of these three studies also performed analyses with mortality at other time points as outcome, i.e. 1, 6-month and one-year mortality, which all showed consistent lower mortality rates in the intervention group [15, 18].

One study applied a different approach, comparing mortality between two groups with different characteristics [20]. The authors retrospectively compared patient outcomes between cases with risk factors for metastatic infections but without signs of metastatic infection on [18F]FDG-PET/CT and normal echocardiography with controls without these risk factors and no known metastatic disease. This control group did not undergo [18F]FDG-PET/CT. This study showed no statistically significant difference in mortality (19.4 vs. 15.0%, p = 0.64) and SAB-specific mortality (0 vs. 2.5%, p = 1.00) between the two groups [20].

Three of the included studies reported new diagnostic findings related to SAB as an outcome measure. However, only one study compared this outcome between the intervention and the control group and reported more detected metastatic foci in the participants who underwent [18F]FDG-PET/CT. This study reported 49 diagnostic foci detected in 48 patients in the PET-CT group and 13 foci in 54 patients in the control group (p < 0.00001) [18]. The other two studies only reported the rate of new findings (in 71 and 70% of patients, respectively) in the [18F]FDG-PET/CT group, but not in the control group [15, 19]. Additional file 3: Table S1 describes diagnostic findings on [18F]FDG-PET/CT per organ system in individual studies.

Four studies reported on risk of infection relapse as outcome [15, 17, 19, 20]. One study found a lower cumulative incidence of relapse in the intervention group (1.4% vs. 8.9%, p = 0.04) and two studies found no difference between both groups (3 vs. 3%, p = 0.74; 0 vs. 3%, p = 0.56) [15, 17, 19]. The study that compared cases with risk factors for complicated bacteremia without signs of metastatic infection on [18F]FDG-PET/CT and normal echocardiography with controls without risk factors and no known metastatic disease showed no statistically significant difference in cumulative incidence of infection relapse (2.8 vs. 5.0%, p = 1.00 [20].

One study compared duration of appropriate antibiotic treatment between the intervention and the control group and reported a higher (42 vs. 19 days, p = 0.001) duration of appropriate antibiotic treatment in the group that underwent [18F]FDG-PET/CT [15]. One study reported more interventions in patients who received [18F]FDG-PET/CT (22 vs. 12%, p = 0.021), but it was unclear whether these interventions included only source control interventions specifically for SAB and which proportion was performed after [18F]FDG-PET/CT [15]. Only one study reported non-infection related accidental findings on [18F]FDG-PET/CT, which were detected in 7.1% of the patients [19]. None of the studies reported final disease classification of complicated versus uncomplicated SAB as a separate outcome.

We identified three important potential confounding domains for studies included in our review: demographics and comorbidities of the study population, severity of disease, and use of co-interventions. Using ROBINS-I assessment, we judged three studies to be at overall serious risk of bias and two studies at overall moderate risk of bias, as shown in Fig. 3. The main reasons for judgment of high risk of bias were bias due to confounding and bias in selection of the reported result. Two studies used matching and two performed multivariate regression analyses to adjust for confounding variables. One study did not employ methods to avoid confounding or adjust for it. Additional file 4: Appendix SC displays the risk of bias assessment per domain for the individual studies.

Table 3 shows the GRADE assessments of certainty of evidence for the primary and secondary outcomes. Certainty of evidence for all outcomes ranged from very low to low. For the primary outcomes of our review, we found low certainty evidence that [18F]FDG-PET/CT is associated with lower mortality and very low certainty evidence that it leads to more new SAB-related diagnostic findings. Most common reasons for downgrading of level of evidence were risk of bias and imprecision of the effect estimates.