Abstract Background Sepsis is a life-threatening syndrome requiring rapid identification and appropriate management, including timely treatment of bacterial infections. Biomarker-based tests aid in determining infection etiology and guiding clinical decisions for patients with suspected sepsis. MeMed BV (MMBV), a host-protein test for differentiating between bacterial and viral infection, is based on computational integration of the circulating levels of three biomarkers (TRAIL, IP-10, CRP). This study evaluates MMBV’s accuracy in identifying bacterial versus viral infections in hemodynamically stable patients with suspected sepsis. Methods A post-hoc analysis of patients recruited prospectively in three studies (Apollo NCT04690569, Observer NCT03011515, Curiosity NCT01917461). Patients were recruited in urgent care centers, emergency departments or internal wards. Patients meeting the eligibility criteria of the parent studies, aged =18 years and with suspected sepsis (defined as at least two SIRS criteria) were included in this analysis. Reference standard infection etiology was determined based on the parent studies, where three independent adjudicators, blinded to MMBV results, assigned bacterial or viral labels after reviewing comprehensive patient data. A unanimous label with =90% confidence was required. MMBV scores (0-100) were interpreted as bacterial/viral/equivocal according to manufacturer’s instructions. MMBV results were compared to the reference standard infection etiologies. Area under the receiver operating characteristic curve (ROC-AUC) was calculated including patients with equivocal results across all possible thresholds. When calculating other diagnostic accuracy parameters, equivocal cases were removed. Results Out of 527 potentially eligible patients, 134 were included (Figure 1A). Median age was 42 years (interquartile range (IQR): 30-65); 69 (52%) were female. 111 patients (82.8%) had two SIRS criteria; 22 (16.4%) had three; and one patient (0.7%) had four. The most common source of infection was the respiratory tract (73.9%). 53% of the patients were hospitalized with a median duration of 4 days (IQR: 3-7 days). Most patients (54%) had no microbiological confirmation in their medical record (even after the ED visit), 21% had a bacterial detection, 11% had viral detection and 14% had both bacterial and viral detections. Reference standard adjudication determined that 56% had a bacterial infection. MMBV results were viral for 53 (39.6%), bacterial for 72 (53.7%) patients, and equivocal for 9 (6.7%) patients. MMBV achieved AUC of 0.98 (95% confidence interval: 0.96-1.00). Among patients with non-equivocal MMBV results (Figure 1B), MMBV attained sensitivity of 97.2% (95%CI: 89.8-99.8), specificity of 96.2% (95%CI: 86.5-99.7), PPV of 97.2% (95%CI: 89.8-99.8) and NPV of 96.2% (95%CI: 86.5-99.7). Notably, the positive likelihood ratio of MMBV was 25.76 (95%CI: 6.61-100.39) and the negative likelihood ratio was 0.03 (95%CI: 0.01-0.11). Conclusion MMBV demonstrated high diagnostic accuracy in distinguishing bacterial from viral infections in adult patients with suspected sepsis. The use of this biomarker-based tool alongside clinical judgment may support timely decision-making in this high-risk population. Figure 1. Top: Flow of adult patients included in the analysis. Bottom: MeMed BV diagnostic performance. Patients with equivocal MMBV results (n=9, 6.7%) are excluded for diagnostic performance calculations.
Navon et al. (Wed,) studied this question.