Abstract
Background
For hospitalized adults receiving empiric antibiotic therapy, antibiotic de-escalation prevents unnecessary exposure and adverse effects. Whether use of direct-from-blood bacterial testing facilitates earlier antibiotic de-escalation and improves clinical outcomes has never been evaluated in a randomized trial.
Methods
Between December 2023 and December 2024, this pragmatic randomized trial compared direct-from-blood bacterial testing with blood cultures (intervention group) to blood cultures alone (usual care group) in adults with suspected infection receiving empiric intravenous vancomycin in the emergency department of an academic medical center. The primary and secondary outcomes were time from randomization to last dose of intravenous vancomycin and systemic antipseudomonal beta-lactam antibiotics, respectively.
Results
Among 500 patients enrolled, the time from randomization to results of bacterial testing from blood (direct-from-blood test, gram stain, or negative culture) was a median of 5.2 days shorter (95% CI, 5.1–5.2) in the direct-from-blood test group (median 0.4 days; IQR, 0.3–0.5) than the usual care group (median, 5.5 days; IQR, 5.2–5.6). The time between randomization and the last dose of intravenous vancomycin did not differ between the direct-from-blood test group (median, 12.5 hours; IQR, 0.79–57.8) and the usual care group (median, 19.0 hours; IQR 0.9–64.8) (HR, 1.08; 95% CI, .90–1.28; P = .42), nor did the time to last dose of systemic antipseudomonal beta-lactam antibiotics (HR, 1.04; 95% CI .87–1.24). Clinical outcomes were also similar.
Conclusions
Among adults with suspected infection receiving empiric intravenous vancomycin in the emergency department, use of direct-from-blood bacterial testing did not shorten the duration of intravenous vancomycin or antipseudomonal beta-lactam antibiotics.
Keywords: molecular diagnostics, antibiotic stewardship, learning healthcare system, pragmatic clinical trial
Direct-from-blood bacterial testing did not reduce the duration of intravenous vancomycin or antipseudomonal beta-lactam antibiotics compared to blood cultures alone in this randomized trial of 500 adults with suspected infection. Clinical outcomes and antibiotic de-escalation times were similar between groups.
Guidelines recommend early de-escalation of empiric antibiotics to prevent adverse effects and antimicrobial resistance. Prior observational studies indicate rapid bacterial testing may facilitate de-escalation.
Among 500 patients with suspected infection undergoing admission from an emergency department, direct-from-blood testing did not reduce time to last dose of vancomycin or antipseudomonal antibiotics.
(See the Editorial Commentary by Coffey et al on pages e236–8.)
International guidelines recommend immediate empiric antibiotic therapy for patients presenting to the hospital with suspected severe infection [1]. Because of the increasing prevalence of antibiotic-resistant pathogens, including Methicillin-resistant Staphylococcus aureus (MRSA) and Pseudomonas aeruginosa, many adults presenting with suspected severe infection to a hospital in the United States receive empiric treatment for MRSA, such as intravenous vancomycin, and P. aeruginosa, such as antipseudomonal beta-lactams. However, a minority of patients admitted with community onset severe infection grow bacteria requiring broad spectrum antibiotic treatment [2]. As a result, many patients are exposed to unnecessarily broad antibiotic therapy for prolonged durations. Early de-escalation of antibiotic therapy may limit adverse drug events [3], decrease mortality [2, 4], and stem the spread of antimicrobial resistance.
Rapidly identifying or excluding a bacterial pathogen may facilitate early de-escalation of antibiotic therapy. Bacterial blood cultures are the current standard for diagnosing infections involving the blood stream. Blood cultures, however, are limited by suboptimal sensitivity [5, 6] and prolonged time-to-results, which, on average, are up to 3 days for a positive result and 5 days to confirm a negative result [5, 7–9]. Previous randomized trials have found rapid tests using technologies to identify pathogens from positive blood culture broth can decrease the time to antibiotic de-escalation [10–12].
One test has been FDA-cleared for directly detecting bacteria in whole blood (T2Bacteria panel) and more are in development [13]. Observational studies suggest use of direct-from-blood testing is associated with faster time to bacterial detection and potentially earlier de-escalation of antibiotic therapy in patients with bloodstream infections [14–18]. Whether using direct-from-blood testing shortens the time to de-escalation of antibacterial therapy, however, has never been evaluated in a randomized trial.
To address this knowledge gap, we conducted a pragmatic, randomized clinical trial comparing the use of direct-from-blood testing via the T2Bacteria panel in addition to blood cultures versus blood cultures alone with regards to antibiotic therapy and clinical outcomes in adults presenting to the emergency department with suspected infection.
METHODS
Trial Design and Oversight
Between December 2023 and December 2024, we conducted a pragmatic, single-center, unblinded, parallel-group, randomized trial comparing the effects of direct-from-blood testing in addition to blood cultures (intervention group) to blood cultures alone (usual care group) in the emergency department of an academic medical center. Time to last dose of vancomycin and antipseudomonal beta-lactam antibiotics were selected as the primary and secondary outcomes, respectively, given institutional stewardship goals. The trial was approved by the institutional review board at Vanderbilt University Medical Center with a waiver of informed consent and registered before enrollment commenced (NCT06069206). The trial protocol and statistical analysis plan were published before enrollment concluded [19].
Patient Population
Adult patients (≥18 years of age) in the emergency department with orders for blood cultures and intravenous vancomycin within 12 hours of presentation were eligible. Patients were excluded if they had received more than 1 dose of intravenous vancomycin since presentation to the hospital, had a known positive blood culture within the previous 7 days, were known to have a current infection for which at least 7 days of intravenous vancomycin would routinely be administered (eg skin and soft tissue infection, endocarditis, and osteomyelitis), were incarcerated, or were pregnant. An electronic health record (EHR) tool screened all patients for eligibility, and an automated alert within the electronic order entry system requested that clinicians confirm eligibility criteria were met.
Randomization
Using software within the EHR, eligible patients were randomized using simple randomization in a 1:1 ratio.
Trial Interventions
Patients assigned to the usual care group received standard blood cultures only. For patients assigned to the direct-from-blood testing (intervention) group, standard blood cultures were ordered, and additionally a clinical decision support tool in the EHR facilitated placement of an order for the direct-from-blood test. The direct-from-blood test used was the T2Bacteria Panel (T2Biosystems, Lexington, MA), which detects S. aureus, Enterococcus faecium, Klebsiella pneumoniae, P. aeruginosa, and Escherichia coli. Full interpretation of direct-from-blood test results was reported in the EHR. Clinicians simultaneously received the results via a text page (Supplementary Methods). For patients with active orders for intravenous vancomycin after negative direct-from-blood testing for S. aureus, an interruptive alert in the EHR prompted the clinician to either (1) discontinue vancomycin or (2) select a reason for continuing vancomycin.
Patients assigned to both groups underwent blood culture testing per institutional protocols. Gram-stain results for positive blood cultures were provided within an hour of detection and a rapid PCR (ePlex, Roche, Indianapolis, IN) was performed on the blood culture broth to identify bacteria and antimicrobial resistance genes [20]. Gram stain results were reported in the EHR and via electronic page alert which required clinician acknowledgement. Blood cultures were finalized as no growth after a 5-day incubation period. All testing was performed in Vanderbilt Medical Laboratories.
Clinical care, including additional microbiology testing and selection of antibiotics, was managed by treating clinicians for all patients. Antimicrobial Stewardship Program pharmacists performed routine prospective review of patient charts in both groups. For all patients receiving intravenous vancomycin, clinical pharmacists monitored serum drug levels for vancomycin, and the results of both blood culture and direct-from-blood test results were incorporated into recommendations and communication with clinical teams.
Data Collection
Data were generated in routine care and electronically extracted from the EHR of the study institution. Details of data collection are in the Supplementary Material [21–23 ].
Outcomes
The primary outcome was time to last dose of intravenous vancomycin, defined as the time between randomization and the start time for the last documented dose of intravenous vancomycin received by the patient within 14 days of enrollment. The secondary outcome was time to last dose of systemic antipseudomonal beta-lactam antibiotic, defined as the time between randomization and start time of the last dose of systemic antipseudomonal beta-lactam antibiotic received by the patient within 14 days of enrollment (Supplementary Methods and Supplementary Table 1). The primary and secondary outcomes derive from empiric use of vancomycin and antipseudomonal beta-lactam antibiotics (cefepime or piperacillin/tazobactam) at the trial site [24]. Exploratory clinical, safety, and antimicrobial stewardship outcomes and process measures are listed in the Supplementary Methods.
Statistical Methods and Analysis
Sample size estimation details were reported previously [19]. Assuming a 2-sided α of 0.05 and a median time to last dose of intravenous vancomycin of 48 hours in the usual care group based on prior data [24], we calculated that enrollment of 500 patients would provide 90% statistical power to detect a 12-hour difference between trial groups.
The analyses of the primary and secondary outcomes included all randomized patients analyzed in the group to which they were assigned (intention-to-treat) and used an unadjusted Cox proportional hazards model. For the analysis of the primary outcome, a two-sided P value of <0.05 was considered to indicate statistical significance. Between-group differences in the secondary and exploratory outcomes are reported as point estimates and 95% confidence intervals (CI). The widths of the 95% CIs for the secondary analyses were not adjusted for multiplicity and should not be used to infer definitive between-group differences in the treatment effects. Additional information on all analyses is provided in the Supplementary Methods. Analyses were performed with R version 4.3.0 (R Foundation for Statistical Computing).
RESULTS
Trial Population
Among 657 patients screened for trial eligibility, 157 (23.9%) were excluded and 500 met inclusion criteria and were enrolled in the trial, of whom 251 were randomized to the intervention group and 249 to the usual care group (Figure 1). Patients' median age was 57 years and the median time between presentation to the emergency department and enrollment was 37.8 minutes in the intervention group and 26.5 minutes in the usual care group. The baseline characteristics of patients in each group are presented in Table 1. At the time of enrollment, 180 patients (36.0%) met Sepsis-3 criteria, including 103 (41.0%) in the intervention group and 77 (30.9%) in the usual care group. The most common suspected sources of infection were lung, skin and soft tissue, and intra-abdominal. A total of 73 patients (29.1%) in the intervention group and 70 patients (28.1%) in the usual care group were admitted to the ICU.
Figure 1.
Patients were screened for enrollment by the treating clinician upon identification of trial eligibility using a notification in the electronic medical record (EMR). Of the 45 patients for whom clinicians selected “Other” as the exclusion criteria, 18 listed a variation of “Not applicable,” 11 listed unclear reasons, 8 noted order set issues, 6 noted exclusion criteria otherwise listed, and 2 noted reasons relating to the testing laboratory. Patients meeting enrollment criteria were randomly assigned to the direct-from-blood test (intervention) group or the usual care group by software in the EMR using a 1:1 ratio.
Table 1.
Patient Characteristics at Baseline
| Direct-from-blood Test (n = 251) |
Usual Care (n = 249) |
|
|---|---|---|
| Patient characteristicsa | ||
| Age, median (IQR), y | 57 (43–71) | 57 (45–69) |
| Female, no. (%) | 110 (43.8) | 114 (45.8) |
| Race, no. (%) | ||
| White | 175 (69.7) | 179 (71.9) |
| Black or African American | 48 (19.1) | 41 (16.5) |
| American Indian or Alaska Native | 4 (1.6) | 1 (0.4) |
| Unknown | 24 (9.6) | 28 (11.3) |
| Ethnicity, no. (%) | ||
| Hispanic/Latino | 10 (4.0) | 16 (6.4) |
| Non-Hispanic/Latino | 229 (91.2) | 224 (90.0) |
| Unknown | 12 (4.8) | 9 (3.6) |
| BMI, median (IQR) | 26 (23–31) | 27 (22–31) |
| Sepsis,b no. (%) | 103 (41.0) | 77 (30.9) |
| SOFA score ≥ 2,c no. (%) | 31 (12.4) | 47 (18.9) |
| Suspected source of infection,d no. (%) | ||
| Lung | 69 (27.5) | 46 (18.5) |
| Intra-abdominal | 24 (9.6) | 39 (15.7) |
| Genitourinary | 20 (8.0) | 17 (6.8) |
| Skin/soft tissue | 55 (21.9) | 53 (21.3) |
| Other | 36 (14.3) | 31 (12.4) |
| Unknown | 47 (18.7) | 63 (25.3) |
| Transplant recipient, no. (%) | 22 (8.8) | 21 (8.4) |
| Neutropenia,e no. (%) | 12/234 (5.1) | 10/223 (4.5) |
| Chronic kidney disease,f no. (%) | 45/251 (17.9) | 55/249 (22.1) |
| End-stage kidney disease on KRT,g no. (%) | 10 (4.0) | 22 (8.8) |
| Baseline creatinine,h median (IQR)—mg/dL | ||
| Pre-illness | 0.75 (0.63–0.84) | 0.76 (0.63–0.90) |
| Peri-enrollment | 1.07 (0.78–1.67) | 1.05 (0.82–1.74) |
| Charlson Comorbidity Index,i median (IQR) | 4 (1–7) | 4 (2–7) |
| Minutes from hospital presentation to enrollment, median (IQR) | 37.8 (18–110.4) | 26.5 (6–94.8) |
| Admission to ICU during index encounter, no. (%) | 73 (29.1) | 70 (28.1) |
Abbreviations: BMI, body mass index; IQR, interquartile range; KRT, kidney replacement therapy; SOFA, sequential organ failure assessment; mg, milligram; dL, deciliter.
aContinuous data are presented as median (IQR: 25th percentile—75th percentile) Categorical data are presented as number (no.) and percentage (%).
bSepsis is defined according to the Sepsis-3 criteria.
cThe SOFA score is composed of scores from six organ systems, graded from 0 to 4 according to the degree of dysfunction or failure. Scores range from 0 (no evidence of organ dysfunction or failure) to 24 (evidence of severe organ dysfunction or failure).
dThe clinical suspected source of infection on admission was categorized into 21 groups based on previously published data and then aggregated into 6 categories. “Other” includes bone and joint, central nervous system, intravascular catheter, primary bloodstream, or other. “Unknown” means the emergency department clinician did not have a high suspicion for one source and/or did not document a suspected source.
eDefined as an absolute neutrophil count <1500 cells/mcL. 43 patients did not have a complete blood count with differential at the time of enrollment.
fChronic kidney disease was defined as the presence of one or more of the International Classification of Diseases, Tenth Revision (ICD-10) diagnoses codes associated with chronic kidney disease in the calculation of the Charlson Comorbidity Index.
gEnd-stage kidney disease on kidney replacement therapy was adjudicated by trial personnel by chart review.
hPre-enrollment creatinine was defined as the lowest creatinine from 24 h before to 365 days before trial enrollment. Peri-enrollment creatinine was defined as the closest creatinine value within 24 h prior trial enrollment, if not available, then the closest creatinine within 6 h after enrollment.
iThe Charlson Comorbidity Index is a method of categorizing comorbidities of patients based on the International Classification of Diseases (ICD) diagnosis codes. Each comorbidity category has an associated weight (from 1 to 6), based on the adjusted risk of mortality or resource use, and the sum of all the weights results in a single comorbidity score for a patient. A score of zero indicates no comorbidities, with higher values indicating a higher probability of death. Charlson Comorbidity Index data is missing for 18 patients (10 patients in the direct-from-blood-test group, 8 patients in the usual care group).
Testing Characteristics
A total of 235 patients (93.6%) in the intervention group had a direct-from-blood test drawn, of which 26 (11.1%) did not generate a result (most commonly due to insufficient volume of blood) and 209 (88.9% of those drawn; 83.3% of those in the intervention group) generated a result. No patient in the usual care group underwent direct-from-blood testing. The median time from randomization to result for direct-from-blood testing was 7.9 hours (interquartile range [IQR], 7.0 to 10 hours).
Blood cultures were drawn in 246 patients (98.0%) in the intervention group and 245 patients (98.4%) in the usual care group. The median time from randomization to availability of a gram stain of initial blood cultures results was 0.80 days (19.2 hours) in the intervention group and 0.78 days (18.7 hours) in the usual care group. The median time from randomization to final blood culture results was 5.5 days in both trial groups. Blood cultures grew an organism for 49 patients (19.5%) in the direct-from-blood test group and 43 patients (17.2%) in the usual care group.
The time from randomization to initial results of blood specimen bacterial testing (earliest of positive or negative direct-from-blood test, gram stain from positive blood culture, or final negative blood culture) was a median of 5.2 days shorter (95% CI, 5.1 to 5.2) in the intervention group (median 0.4 days; IQR, 0.3 to 0.5 days) than the usual care group (median, 5.5 days; IQR, 5.2 to 5.6 days) (P < .001) (Supplementary Table 6).
Among the 209 patients in the intervention group with valid results for the direct-from-blood test, 208 had blood cultures for comparison, and of these 25 (12.0%) had one or more pathogens result as detected with direct-from-blood testing. Direct-from-blood testing was positive for S. aureus in 7 patients, P. aeruginosa in 6 patients, E. coli in 6 patients, K. pneumoniae in 5 patients, and E. faecium in 2 patients (testing in 1 patient generated detected results for both K. pneumoniae and P. aeruginosa). Results of direct-from-blood testing and blood cultures were concordant for 188 of the 208 patients (90.4%): 177 patients had concordant negative blood cultures and direct-from-blood testing, 11 patients had concordant positive blood cultures and direct from blood testing (S. aureus in 6 patients, E. coli in 3 patients, and K. pneumoniae in 2 patients). Results were discordant for 20 patients (9.6%). Discordant cases included 13 patients with positive direct-from-blood testing (12 patients with 1 target positive, 1 patient 2 targets positive) but negative blood cultures, 4 patients with negative direct-from-blood testing but positive blood cultures for on-panel pathogens, 1 patient with a positive blood culture for an on-panel pathogen but direct-from-blood testing positive for a different pathogen, and 2 patients with incomplete results due to indeterminate E. coli results (one with blood cultures positive for E. coli) (Supplementary Tables 2 and 3). Among the patients with positive direct-from-blood tests and blood cultures without growth of the detected pathogen, 3 had the same organism detected on culture of another body site and 1 in blood at a later time point (Supplementary Table 4). Compared to blood culture as a reference standard, direct-from-blood testing had a sensitivity of 68.8%, a specificity of 98.5%, a positive predictive value of 42.3%, a negative predictive value of 99.5%, and an accuracy of 98.1%.
Primary Outcome
The median time between randomization and the last dose of intravenous vancomycin within 14 days did not differ significantly between the intervention group (median, 12.5 hours; IQR 0.79 hours to 57.8 hours) and the usual care group (median, 19.0 hours; IQR 0.9 hours to 64.8 hours) (hazard ratio, 1.08; 95% CI, .90 to 1.28; P = .42) (Table 2 and Figure 2). Results were similar in all prespecified sensitivity analyses and the modified intention to treat analysis (Supplementary Table 5). Results were similar in each prespecified subgroup (Figure 3), except among patients with neutropenia for whom the time between randomization and the last dose of intravenous vancomycin was shorter in the intervention group (median 0 days; IQR, 0 to 1 day) compared to the usual care group (median, 2 days, IQR, 1 to 6 days).
Table 2.
Outcomesa
| Outcome | Direct-from-blood Test (n = 251) | Usual Care (n = 249) |
Hazard Ratio or Median Difference (95% CI) |
|---|---|---|---|
| Primary outcome | |||
| Time to final dose of intravenous vancomycin, median (IQR), hoursb | 12.5 (0.79–57.8) | 19.0 (0.9–64.8) | 1.08 (0.90 to 1.28) |
| Secondary outcome | |||
| Time to final dose of systemic antipseudomonal beta-lactam antibiotic, median (IQR), daysc,d | 1.6 (0.04–3.6) | 1.7 (0.05–3.80) | 1.04 (0.87 to 1.24) |
| Exploratory outcomes | … | … | … |
| Antimicrobial stewardship | … | … | … |
| Total number of doses of intravenous vancomycin by day 14, median (IQR)e | 4 (2–8) | 4 (2–10) | 0.0 (−1.0 to 1.0) |
| Total number of days gram-positive antibiotic therapy by day 14,f median (IQR)e | 3 (2–6) | 3 (2–6) | 0.0 (−1.0 to 1.0) |
| Total number of days gram-negative antibiotic therapy by day 14,g median (IQR)e | 3 (1–6) | 3 (1–6) | 0.0 (−1.0 to 1.0) |
| Time to effective antibiotic therapy,h median (IQR) (days)e | 0 | 0 | 0.0 (0.0 to 0.0) |
| Clostridioides difficile infection by day 28, no. (%)i | 5 (2.0) | 8 (3.2) | −1.2% (−4.4% to 2.0%) |
| Clinical | |||
| Highest stage of acute kidney injury by day 14, no.(%)j,k | |||
| None | 178 (70.9) | 178 (71.5) | −0.6% (−8.9% to 7.8%) |
| Stage I | 18 (7.2) | 19 (7.6) | −0.5% (−5.4% to 4.5%) |
| Stage II | 14 (5.6) | 14 (5.6) | −0.0% (−4.1% to 4.0%) |
| Stage III | 25 (10.0) | 30 (12.0) | −2.1% (−8.0% to 3.8%) |
| Diedl | 16 (6.4) | 8 (3.2) | 3.2% (−1.0% to 7.3%) |
| Receipt of new kidney replacement therapy by day 14, no. (%)i | 5 (2.0) | 7 (2.8) | −0.8% (−3.9% to 2.3%) |
| Lowest platelet count by day 14, median (IQR)e | 197 (134 to 264) | 198 (138 to 270) | 1.0 (−29.0 to 22.0) |
| Hospital-free days by day 28e,m | 21 (17 to 24) | 22 (17 to 24) | −1.0 (−2.0 to 1.0) |
| Intensive care unit-free days by day 28e,m | 28 (25 to 28) | 28 (26 to 28) | 0.0 (0.0 to 0.0) |
| All-cause, in-hospital mortality by day 28, no. (%)e | 19 (7.6) | 9 (3.6) | 4.0% (−0.5% to 8.4%) |
Abbreviation: IQR, interquartile range.
aContinuous data are presented as median (25th percentile—75th percentile) unless otherwise noted. Categorical data are presented as number (no.) and percentage (%).
bThe median survival time for the direct-from-blood test group was 12.5 h (95% CI 0.12–0.70). The median survival time for the usual care group was 19 h (95% CI 0.50–1.2).
cThe median survival time for the direct-from-blood test group was 1.6 days (95% CI 1.3–1.9). The median survival time for the usual care group was 1.7 d (95% CI 1.6–2.2).
dAntipseudomonal beta-lactam antibiotics include: Piperacillin/Tazobactam, Cefepime, Ceftolozane/Tazobactam, Ceftazidime, Ceftazidime/Avibactam, Cefiderocol, Doripenem, Imipenem, Imipenem/Relebactam, Meropenem, Meropenem/Vaborbactam, and Aztreonam.
eA Wilcoxon-Rank sum test was performed.
fGram-positive therapy was defined using criteria from prior data.
gGram-negative therapy was defined using criteria from prior data.
hEffective coverage for blood stream infections was adjudicated by study personnel. In vitro susceptibility results generated by the microbiology laboratory for final blood culture results collected prior to or at the same time as randomization were compared with prescribed antimicrobial therapy within 14 d of randomization.
iA χ2 test was performed.
jA proportional odds likelihood ratio test was performed.
kAcute kidney injury on enrollment. Stages are defined according to Kidney Disease Improving Global Outcomes (KDIGO) creatinine criteria as a plasma creatinine value at enrollment.
lTwo patients did not have a plasma creatinine value at enrollment.
mDefined as the number of calendar days alive and free of the hospital/intensive care unit between randomization and 28 d after randomization with outcome assessment censored at hospital discharge.
Figure 2.
The time between randomization and the final dose of intravenous vancomycin is shown for patients in the direct-from-blood test group (red) and the usual care group (green).
Figure 3.
Effect modification of the primary outcome. For each prespecified subgroup, the effect of the direct-from-blood test is shown for the primary outcome of time to last dose of intravenous vancomycin. Hazard ratios >1.0 indicate a better outcome in the intervention group compared to the usual care group.
Secondary and Exploratory Outcomes
The median time from randomization to the last dose of systemic antipseudomonal beta-lactam antibiotics within 14 days did not differ significantly between the intervention group (median 1.6 days; IQR 0.04 to 3.6 days) and the usual care group (median 1.7 days; IQR 0.05 to 3.8 days) (hazard ratio 1.04, 95%CI .87 to 1.24) (Table 2).
The median number of doses of vancomycin within 14 days did not differ between the intervention group (median 4 doses; IQR, 2 to 8) and the usual care group (median 4 doses; IQR, 2 to 10). The median number of days of gram-positive antibiotic therapy was 3 days (IQR, 2 to 6) in both groups. The median number of days of gram-negative antibiotic therapy was 3 days (IQR, 1 to 6) in both groups. Antimicrobial stewardship, safety, process measures, and clinical outcomes were also similar between groups (Table 2; Supplementary Tables 6 and 7).
DISCUSSION
Among 500 patients presenting to the emergency department with suspected infection in this randomized trial, use of an FDA-cleared direct-from-blood test did not reduce duration of vancomycin therapy, reduce duration of antipseudomonal therapy, or affect clinical outcomes. These findings suggest that in a setting with rapid identification of bacteria from blood cultures and an active antibiotic stewardship team as part of usual care, the addition of direct-from-blood testing for a limited panel of bacterial pathogens does not impact the assessed outcomes.
Prior observational studies of direct-from-blood testing found varying associations with the duration of antibiotic use. Some report an association between testing and shorter time to optimal antibiotic treatment [14, 18, 25–27], and a recent meta-analysis of observational studies found that testing was associated with shorter time to antibiotic de-escalation [28]. In contrast, our randomized trial found that direct-from-blood testing did not affect antibiotic de-escalation despite providing clinicians with results of bacterial testing a median of 7.9 hours after randomization and 5 days sooner than finalized blood cultures.
Several potential explanations exist for the difference in findings between our trial and prior observational studies. First, the association between receipt of direct-from-blood testing and earlier de-escalation of antibiotics in prior studies may have resulted from indication bias or other confounding that was addressed by use of a randomized trial design. Second, direct-from-blood testing may be more beneficial in patients with higher severity of illness for whom clinicians are resistant to de-escalate antibiotic therapy. Only 40% of patients in our trial met criteria for sepsis and approximately 1-in-3 were admitted to the ICU. However, direct-from-blood testing did not appear to be more beneficial in the subgroups of patients who met sepsis criteria or those with higher baseline severity of illness scores. Third, the median duration of vancomycin therapy was shorter than the median of 48 hours anticipated in the trial design, demonstrating robust stewardship as part of usual care. It is difficult to detect an impact from direct-from-blood testing given the short time between randomization and vancomycin discontinuation in both groups. Interventions involving direct-from-blood testing may be more effective in settings with less robust usual care and stewardship interventions. Though a cost analysis was not performed in this trial, the expense of direct-from-blood testing may add minimal value to robust stewardship practices. Further, effects on antibiotic duration may be greater for patient populations in which the standard duration of therapy is longer. This interpretation is supported by the finding that the intervention appeared to potentially shorten the duration of vancomycin therapy in the subgroup of patients with neutropenia, a group for whom empiric antibiotic therapy duration was longer.
Our trial had several strengths. The design included randomization to balance baseline characteristics and concealment of the trial-group assignment until enrollment to prevent selection bias. The sample size was moderately large, and broad eligibility criteria with few exclusion criteria produced a trial population representative of the patients seen in clinical care. The direct-from-blood test was paired with a multifaceted antibiotic stewardship intervention (including real-time notification of results by text page, detailed information on the interpretation of results in the EHR, interruptive alerting to prompt vancomycin discontinuation, and interaction with ID pharmacy and antibiotic stewardship teams), which prior research suggests is necessary for rapid diagnostic testing to influence antibiotic de-escalation [11, 29]. The results are generalizable as outcomes target stewardship goals of reducing vancomycin and antipseudomonal beta-lactam antibiotic overuse.
Our study also has several limitations. First, despite dedicated clinical research personnel supporting ED clinicians in the collection of the direct-from-blood test, only 83.3% of patients in the intervention group achieved a direct-from-blood test result. Second, the direct-from-blood test did not demonstrate the performance characteristics established for FDA certification [30, 31]. Three false negative S. aureus results occurred, which may have decreased confidence in results and subsequent willingness to de-escalate antibiotics. This is consistent with prior reports, as test performance for the direct-from-blood test used in this trial varied across previous studies [14, 30, 32, 33]. Third, the direct-from-blood test used (the only FDA-approved at the time of trial launch) tested for only five bacterial pathogens, which may have limited willingness to de-escalate vancomycin or antipseudomonal beta-lactam therapy due to anticipation of off-panel pathogens. Fourth, though eligibility criteria excluded patients anticipated to receive at least 7 days of empiric vancomycin, enrollment of patients with skin and soft tissue infections (for whom vancomycin could be appropriate initial therapy) may have limited the power to detect a difference in the primary outcome. Fifth, this trial was not powered to investigate clinical metrics beyond the listed exploratory outcomes. Finally, though this is the largest randomized trial investigating this technology to date, the platform is no longer available, and other testing platforms will differ in performance. Future trials should evaluate direct-from-blood tests that evaluate for a broader range of pathogens paired with multifaceted antibiotic stewardship interventions among patients likely to be exposed to extended durations of empiric intravenous antibiotic therapy.
In summary, among adults presenting to an emergency department with suspected infection, use of direct-from-blood testing did not shorten the duration of intravenous vancomycin or antipseudomonal beta-lactam administration.
Supplementary Material
Notes
Author contributions. Study concept and design: D. G., R. H., A. L., C. L. G., E. T. Q., W. H. S., M. L. D., R. B., M. W. S., G. E. N., M. A. C., J. S., B. E., and L. W. Acquisition of data: E. T. Q. and L. W. Drafting of the manuscript: D. G., R. H., A. L., C. L. G., E. T. Q., W. H. S., M. L. D., and M. W. S. All study authors critically reviewed the manuscript for important intellectual content. Study supervision: D. G., R. H., C. L. G., and M. W. S. All study authors approved the final version of this manuscript.The Vanderbilt Center for Learning Healthcare within the Vanderbilt Institute for Clinical and Translational Research (VICTR) supported the conduct of this trial, with input from the following collaborators (all at Vanderbilt University Medical Center and Vanderbilt University, Nashville, TN): Dan Albert, Adrienne Baughman, Laura Bobbitt, Carleigh Burns, Timothy Duff, Andrea Fletcher, Allyson Hobbie, Austin Ing, Jakea Johnson, Gabriel Kemp, Sheryl Mangrum, Geoff Mavrak, Kelly Moser, David Mulherin, Shannon Pugh, Todd W. Rice, Chrissie Schaeffer, Adam Turner, Sabrina Shipman, Halden Z. VanCleave, Hamilton Wen, and LaKeysha Wiggins.
Data availability. A de-identified dataset may be made available outside of the study team on reasonable request and concurrence with the study team that the data are fit for purpose, in addition to approval from an authorized Institutional Review Board and an executed data use agreement. Data will become available 12 months following publication of outcomes and will remain available for at least 2 years.
Financial support. Supported through an investigator-initiated grant from T2 Biosystems. The sponsor did not participate in protocol design but did review the protocol and provide feedback. The sponsor also did not participate in analysis or manuscript drafting but was provided with the opportunity to review. The sponsor does not have the final decision over any aspect of trial conduct or publication of the outcomes.
Contributor Information
David C Gaston, Department of Pathology, Microbiology, and Immunology, Vanderbilt University Medical Center, Nashville, Tennessee, USA.
Romney M Humphries, Department of Pathology, Microbiology, and Immunology, Vanderbilt University Medical Center, Nashville, Tennessee, USA.
Ariel A Lewis, Vanderbilt Institute for Clinical and Translational Research, Vanderbilt University Medical Center, Nashville, Tennessee, USA.
Cheryl L Gatto, Vanderbilt Institute for Clinical and Translational Research, Vanderbilt University Medical Center, Nashville, Tennessee, USA; Department of Emergency Medicine, Vanderbilt University Medical Center, Nashville, Tennessee, USA.
Li Wang, Department of Biostatistics, Vanderbilt University School of Medicine, Nashville, Tennessee, USA.
George E Nelson, Department of Medicine, Division of Infectious Diseases, Vanderbilt University Medical Center, Nashville, Tennessee, USA.
Joanna L Stollings, Department of Pharmaceutical Services, Vanderbilt University Medical Center, Nashville, Tennessee, USA.
Benjamin J Ereshefsky, Department of Pharmaceutical Services, Vanderbilt University Medical Center, Nashville, Tennessee, USA.
Matthew A Christensen, Department of Medicine, Division of Allergy, Pulmonary and Critical Care Medicine, Vanderbilt University Medical Center, Nashville, Tennessee, USA.
Mary Lynn Dear, Vanderbilt Institute for Clinical and Translational Research, Vanderbilt University Medical Center, Nashville, Tennessee, USA.
Ritu Banerjee, Department of Pediatrics, Division of Pediatric Infectious Diseases, Vanderbilt University Medical Center, Nashville, Tennessee, USA.
Matthew Rodgers, Department of Pathology, Microbiology, and Immunology, Vanderbilt University Medical Center, Nashville, Tennessee, USA.
Alison Benton, Department of Pathology, Microbiology, and Immunology, Vanderbilt University Medical Center, Nashville, Tennessee, USA.
Sharon Glover, Department of Pathology, Microbiology, and Immunology, Vanderbilt University Medical Center, Nashville, Tennessee, USA.
Karen F Miller, Vanderbilt Institute for Clinical and Translational Research, Vanderbilt University Medical Center, Nashville, Tennessee, USA; Department of Emergency Medicine, Vanderbilt University Medical Center, Nashville, Tennessee, USA.
Wesley H Self, Vanderbilt Institute for Clinical and Translational Research, Vanderbilt University Medical Center, Nashville, Tennessee, USA; Department of Emergency Medicine, Vanderbilt University Medical Center, Nashville, Tennessee, USA.
Matthew W Semler, Vanderbilt Institute for Clinical and Translational Research, Vanderbilt University Medical Center, Nashville, Tennessee, USA; Department of Medicine, Division of Allergy, Pulmonary and Critical Care Medicine, Vanderbilt University Medical Center, Nashville, Tennessee, USA.
Edward T Qian, Department of Medicine, Division of Allergy, Pulmonary and Critical Care Medicine, Vanderbilt University Medical Center, Nashville, Tennessee, USA; Department of Biomedical Informatics, Vanderbilt University Medical Center, Nashville, Tennessee, USA.
Vanderbilt Center for Learning Healthcare:
Dan Albert, Adrienne Baughman, Laura Bobbitt, Carleigh Burns, Timothy Duff, Andrea Fletcher, Allyson Hobbie, Austin Ing, Jakea Johnson, Gabriel Kemp, Sheryl Mangrum, Geoff Mavrak, Kelly Moser, David Mulherin, Shannon Pugh, Todd W Rice, Chrissie Schaeffer, Adam Turner, Sabrina Shipman, Halden Z VanCleave, Hamilton Wen, and LaKeysha Wiggins
Supplementary Data
Supplementary materials are available at Clinical Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.
References
- 1. Evans L, Rhodes A, Alhazzani W, et al. Surviving sepsis campaign: international guidelines for management of sepsis and septic shock 2021. Crit Care Med 2021; 49:e1063. [DOI] [PubMed] [Google Scholar]
- 2. Rhee C, Kadri SS, Dekker JP, et al. Prevalence of antibiotic-resistant pathogens in culture-proven sepsis and outcomes associated with inadequate and broad-spectrum empiric antibiotic use. JAMA Netw Open 2020; 3:e202899. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Tamma PD, Avdic E, Li DX, Dzintars K, Cosgrove SE. Association of adverse events with antibiotic use in hospitalized patients. JAMA Intern Med 2017; 177:1308–15. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Webb BJ, Sorensen J, Jephson A, Mecham I, Dean NC. Broad-spectrum antibiotic use and poor outcomes in community-onset pneumonia: a cohort study. Eur Respir J 2019; 54:1900057. [DOI] [PubMed] [Google Scholar]
- 5. Tabak YP, Vankeepuram L, Ye G, Jeffers K, Gupta V, Murray PR. Blood culture turnaround time in U.S. acute care hospitals and implications for laboratory process optimization. J Clin Microbiol 2018; 56. 10.1128/jcm.00500-18. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Gupta S, Sakhuja A, Kumar G, McGrath E, Nanchal RS, Kashani KB. Culture-negative severe sepsis: nationwide trends and outcomes. Chest 2016; 150:1251–9. [DOI] [PubMed] [Google Scholar]
- 7. Melling PA, Noto MJ, Rice TW, Semler MW, Stollings JL. Time to first culture positivity among critically ill adults with methicillin-resistant Staphylococcus aureus growth in respiratory or blood cultures. Ann Pharmacother 2020; 54:131–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Buell KG, Casey JD, Noto MJ, Rice TW, Semler MW, Stollings JL. Time to first culture positivity for gram-negative rods resistant to ceftriaxone in critically ill adults. J Intensive Care Med 2021; 36:51–7. [DOI] [PubMed] [Google Scholar]
- 9. MacBrayne CE, Williams MC, Prinzi A, Pearce K, Lamb D, Parker SK. Time to blood culture positivity by pathogen and primary service. Hosp Pediatr 2021; 11:953–61. [DOI] [PubMed] [Google Scholar]
- 10. Emonet S, Charles PG, Harbarth S, et al. Rapid molecular determination of methicillin resistance in staphylococcal bacteraemia improves early targeted antibiotic prescribing: a randomized clinical trial. Clin Microbiol Infect 2016; 22:946.e9–e15. [DOI] [PubMed] [Google Scholar]
- 11. Banerjee R, Teng CB, Cunningham SA, et al. Randomized trial of rapid multiplex polymerase chain reaction-based blood culture identification and susceptibility testing. Clin Infect Dis 2015; 61:1071–80. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Beuving J, Wolffs PFG, Hansen WLJ, et al. Impact of same-day antibiotic susceptibility testing on time to appropriate antibiotic treatment of patients with bacteraemia: a randomised controlled trial. Eur J Clin Microbiol Infect Dis 2015; 34:831–8. [DOI] [PubMed] [Google Scholar]
- 13. Samuel L. Direct-from-blood detection of pathogens: a review of technology and challenges. J Clin Microbiol 2023; 61:e0023121. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Drevinek P, Hurych J, Antuskova M, et al. Direct detection of ESKAPEc pathogens from whole blood using the T2Bacteria panel allows early antimicrobial stewardship intervention in patients with sepsis. MicrobiologyOpen 2021; 10:e1210. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Kalligeros M, Zacharioudakis IM, Tansarli GS, Tori K, Shehadeh F, Mylonakis E. In-depth analysis of T2Bacteria positive results in patients with concurrent negative blood culture: a case series. BMC Infect Dis 2020; 20:326. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Yu D, Ekwall-Larson A, Özenci V. Performance of T2Bacteria in relationship to blood cultures—a retrospective comparative study. Eur J Clin Microbiol Infect Dis 2024; 43:1977–87. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Peras M, Kuliš T, Mareković I. T2Bacteria panel used simultaneously with blood cultures helps to determine the etiology of bloodstream infections. Acta Microbiol Immunol Hung 2024; 71:280–4. [DOI] [PubMed] [Google Scholar]
- 18. Clodi-Seitz T, Baumgartner S, Turner M, et al. Point-of-care method T2Bacteria®panel enables a more sensitive and rapid diagnosis of bacterial blood stream infections and a shorter time until targeted therapy than blood culture. Microorganisms 2024; 12:967. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Gaston DC, Humphries RM, Lewis AA, et al. Examining the effect of direct-from-blood bacterial testing on antibiotic administration and clinical outcomes: a protocol and statistical analysis plan for a pragmatic randomised trial. BMJ Open 2025; 15:e090263. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. GenMark Diagnostics . Blood Culture Identification (BCID) Panels. 2023. Available at: https://www.genmarkdx.com/panels/eplex-panels/bcid-panels/. Accessed 3 July 2023.
- 21. Leligdowicz A, Dodek PM, Norena M, et al. Association between source of infection and hospital mortality in patients who have septic shock. Am J Respir Crit Care Med 2014; 189:1204–13. [DOI] [PubMed] [Google Scholar]
- 22. Seymour CW, Liu VX, Iwashyna TJ, et al. Assessment of clinical criteria for sepsis: for the third international consensus definitions for sepsis and septic shock (sepsis-3). JAMA 2016; 315:762–74. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Singer M, Deutschman CS, Seymour CW, et al. The third international consensus definitions for sepsis and septic shock (sepsis-3). JAMA 2016; 315:801–10. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. Qian ET, Casey JD, Wright A, et al. Cefepime vs piperacillin-tazobactam in adults hospitalized with acute infection: the ACORN randomized clinical trial. JAMA 2023; 330:1557–67. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Quirino A, Scaglione V, Marascio N, et al. Role of the T2Dx magnetic resonance assay in patients with suspected bloodstream infection: a single-centre real-world experience. BMC Infect Dis 2022; 22:113. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26. Bonura C, Graceffa D, Distefano S, et al. Evaluation of T2 magnetic resonance (T2MR®) technology for the early detection of ESKAPEc pathogens in septic patients. Antibiot Basel Switz 2024; 13:885. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. Giacobbe DR, Crea F, Morici P, et al. T2Bacteria and T2Resistance assays in critically ill patients with sepsis or septic shock: a descriptive experience. Antibiot Basel Switz 2022; 11:1823. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. Giannella M, Pankey GA, Pascale R, Miller VM, Miller LE, Seitz T. Antimicrobial and resource utilization with T2 magnetic resonance for rapid diagnosis of bloodstream infections: systematic review with meta-analysis of controlled studies. Expert Rev Med Devices 2021; 18:473–82. [DOI] [PubMed] [Google Scholar]
- 29. Pliakos EE, Andreatos N, Shehadeh F, Ziakas PD, Mylonakis E. The cost-effectiveness of rapid diagnostic testing for the diagnosis of bloodstream infections with or without antimicrobial stewardship. Clin Microbiol Rev 2018; 31:e00095–17. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Nguyen MH, Clancy CJ, Pasculle AW, et al. Performance of the T2Bacteria panel for diagnosing bloodstream infections. Ann Intern Med 2019; 170:845–52. [DOI] [PubMed] [Google Scholar]
- 31. Maki DG. The T2Bacteria panel had 90% sensitivity for detecting targeted organisms, 43% for any bloodstream infection organism. Ann Intern Med 2019; 171:JC34. [DOI] [PubMed] [Google Scholar]
- 32. Voigt C, Silbert S, Widen RH, et al. The T2Bacteria assay is a sensitive and rapid detector of bacteremia that can be initiated in the emergency department and has potential to favorably influence subsequent therapy. J Emerg Med 2020; 58:785–96. [DOI] [PubMed] [Google Scholar]
- 33. De Angelis G, Posteraro B, De Carolis E, et al. T2Bacteria magnetic resonance assay for the rapid detection of ESKAPEc pathogens directly in whole blood. J Antimicrob Chemother 2018; 73:iv20–6. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.