This was an international, blinded, prospective cohort study enrolling patients presenting to two adult large EDs at the Pitié-Salpêtrière APHP-Sorbonne Université hospital in Paris (LPS), France, and the Hospital Universitari Germans Trias i Pujol in Badalona, Spain (UHG). The study assessed MDW’s ability to detect the development of sepsis in consecutive adult patients (18–89 years) presenting to the ED and who had a CBC-DIFF ordered. The study was registered in ClinicalTrials.gov (NCT03588325) and approved by the ethical committees of the respective participating countries.
Patients
The study prospectively enrolled consenting adults (over 18-year-old) presenting to the ED, whose initial evaluation included a CBC-DIFF. Exclusion criterion is as shown in Fig. 1. Emergency physicians were free to order necessary tests according to the standard of care. Treating physicians were blinded to MDW, PCT and CRP values obtained by protocol, but received PCT and/or CRP results they ordered as current practice.
Both PCT and CRP are used routinely at LPS, and CRP is used routinely at UHG.
Blood sampling
Briefly, patients who were identified to have a CBC-DIFF ordered by the physician were asked to participate. After informed consent, baseline blood draws were performed. An additional K3EDTA tube was drawn together with a sample for PCT and CRP measurement as well as routine blood tests at the discretion of the treating physician. Samples for MDW measurement were analyzed on a UniCel DxH 900 analyzer (Beckman Coulter, Inc., Brea, CA) with version 1.0.0.329 software within two hours of collection. This instrument measures specific cell volume variables and the distribution of cell volumes within a group of white blood cells (WBC). Quality control was performed daily with COULTER 6C Plus Cell Control to monitor the DxH 900 system performance. COULTER LATRON CP-X Control was used as part of the daily quality control procedure to monitor volume, conductivity and light scatter measurements. PCT and CRP concentrations were measured on Cobas analyzer (Roche Diagnostics, Meylan, France) at LPS, while UHG utilized Liaison XL (DiaSorin, Saluggia, Italy) and AU5800 (Beckman Coulter, Inc, Brea, CA, USA) analyzers for PCT and CRP measurements, respectively.
Clinical data, follow-up, blinding and adjudication
Clinical data at presentation, including past medical history, assessment of vital signs, symptoms, SIRS criteria [12], qSOFA [13] and SOFA scores [14], microbiological testing and treatments were recorded on an electronic case report form and patients followed up for at least 12 h. The clinical research team and physician adjudicators were blinded to MDW results at the time of clinical data entry and during assignment of the patients to a clinical category. Adjudicators were also blinded to the results of PCT and CRP if not ordered by the treating ED physician.
Study subjects were classified by at least two independent physician adjudicators at each site. Discordances were arbitrated by a third independent physician. Adjudicators categorized subjects based on the “Sepsis-2” consensus criteria [15], such as non-SIRS or case controls (zero or one SIRS criterion and no infection), SIRS (≥ 2 SIRS criteria and no infection), infection (suspected or confirmed infection with 0—1 SIRS criteria), sepsis (infection plus ≥ 2 SIRS) (including sepsis [no organ failures], severe sepsis [sepsis with one or more organ failures] and septic shock [sepsis with refractory hypotension]). Adjudicated categories per Sepsis-3 criteria [16] included controls, infection and sepsis (based on SOFA score criteria). In order to characterize sepsis as being present upon ED admission, sepsis criteria had to be fulfilled within 12 h of the initial CBC-DIFF in patients with suspected infection (as reflected by initiation of diagnostic infection workup) and adjudication was based on the retrospective chart review of tests ordered and clinical data available within the first 12 h of ED presentation. If no infection work-up was performed within 12 h, or if the adjudicator believed that the infection work-up showed no evidence of infection, the patient was categorized as “not infected” or SIRS by the adjudicator. The test results were extracted from the medical records 7–10 days later, including cultures, molecular tests (e.g., polymerase chain reaction), antigens tests (immunoassay) and relevant imaging. A subgroup, consisting of subjects for whom no CRP or PCT was ordered by the emergency physician, was identified (post study) and defined as having low sepsis pretest probability.
Statistical analysis
General descriptive statistics and box plots were calculated for cell population variables. The sample size calculations were based on 95% two-sided confidence interval and 80% power. Proc power of SAS 9.3 was used for calculating sample size based on the approach presented in Johnson et al. [17] A target sensitivity of 75% with the lower limit of the 95% two-sided confidence interval of 65% and a target specificity of 70% with the lower limit of the 95% two-sided confidence interval of 65% were assumed. A minimum of 189 septic subjects and a minimum of 817 non-septic subjects were needed from both sites combined.
Diagnostic ability was evaluated in terms of the area under the ROC curve (AUC), sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), positive and negative likelihood ratios, along with their 95% confidence intervals (CIs). The score approach was used to calculate CI for sensitivity, specificity, PPV and NPV. Differences in AUC were used to demonstrate the added value of MDW in comparison with WBC alone, calculated using a one-predictor variable logistic model with WBC and a two-predictor variables logistic model with both WBC and MDW, as the predictor, and using sepsis status as the response. A similar approach was applied to analyze the diagnostic ability of PCT and CRP, and the combination with WBC and MDW for sepsis detection. AUC comparisons along with their CI were calculated as described by DeLong et al. [18] SAS 9.4 (SAS Institute, Cary, NC) statistical program was used for data analyses.
Cutoff determination
PCT cutoff of > 0.25 µg/mL was based on the literature. The determination of the cutoff for CRP (> 22 mg/L) was based on the Youden index using the data from this clinical study. The proposed MDW cutoff of 21.5 units for K3EDTA was based on previously established MDW cutoff (in K2EDTA) for sepsis identification in the ED and internal testing demonstrating a shift of 1.5 units for blood specimens collected in K3EDTA versus K2EDTA (data not shown). This study validates the new MDW cutoff of 21.5 for K3EDTA.
Probabilities and odds ratios
Identification of sepsis in the ED relies on the physician’s ability to assess probability of disease based on presenting symptoms. To reflect this approach, we analyzed the probability and likelihood ratios for sepsis based upon the values of MDW in combination with SIRS or qSOFA parameters determined during the initial patient encounter in the ED (typically within the first 2 h of ED admission). Predicted probability of a positive sepsis diagnosis was calculated from the positive likelihood ratios (LRs +) as previously described [19, 20]. In this approach, predicted sepsis probability after receiving test results, or posttest probability P1, is based on an estimated pretest probability P0 and LR + and is calculated as:
$$P_{1} = \frac{{P_{1} \times {\text{LR}} + }}{{(1 - P_{0} + P_{0} \times {\text{LR }} +)}}$$
where P0 is the sepsis prevalence of the study cohort. The odds ratios (ratios of posttest probabilities) for sepsis diagnosis between parameter combinations with abnormal and normal MDW values were calculated based on prevalence of sepsis-2 or sepsis-3 for the study population.