Abstract
Abstract 4726
Neonatal sepsis remains an important clinical syndrome despite advances in neonatology. Early diagnosis and adequate antibiotic treatment are required because of high rates of mortality and morbidity. Current hematology analysers can determine cell volume (V), conductivity for internal composition of cell (C) and light scatter for cytoplasmic granularity and nuclear structure (S) and standart deviations. This VCS measurements were used to diagnose sepsis particularly in adults and higher levels of V and lower levels of S were seen in septic patients. We investigated these parameters in diagnosis of neonatal sepsis.
We investigated these parameters in secreening of neonatal sepsis, defined as a systemic infection, validated by a positive blood culture, diagnosed beyond the first day of life. We used LH780 hematological analyzer (Beckman Coulter, Fullerton, CA). We combined these parameters with interleukin-6 (IL-6) and C-reactive protein (CRP), and developed models to diagnose sepsis by Effective Modelling of Moleculer Activity (EMMA). It uses combinatorial algorithm of the selection parameters for regression equation based on modified stepwise procedure. It allows compute a number of “best” regression equations with different combinations of parameters.
A total of 237 newborn, 61 proven sepsis, 108 clinical sepsis and 68 control, were enrolled the study. Gram positive and negative bacteria including 12 types of microorganism were isolated from 34 and 27 patients, respectively. Mean neutrophil volume (MNV) and volume distribution width (VDW) were found to be statistically increased both in proven and clinical sepsis groups. We developed models using MNV, VDV, IL-6 and CRP (Table 1). We used these models both for proven sepsis and sepsis (proven and clinical sepsis) groups. Table 2 and 3 summarize the cut-off levels of MNV, VDV, IL-6, CRP and models' performance of proven sepsis and sepsis groups. These models gave more sensitivity and specificity than usage of MNV, VDW, IL-6 and CRP alone.
Test results and models' performance of proven sepsis group
| Parameter . | Cut-off . | Sens . | Spec . | AUC . | 95% Confidence Interval . | Significance level P (Area=0.5) . |
|---|---|---|---|---|---|---|
| MNV (au) | >159.6 | 75.00 | 86.73 | 0.847 | 0.785 to 0.897 | <0.0001 |
| VDW (au) | >36.9 | 69.74 | 73.47 | 0.788 | 0.719 to 0.846 | <0.0001 |
| IL6 (pg/mL) | >22 | 77.05 | 94.12 | 0.887 | 0.819 to 0.936 | <0.0001 |
| CRP (mg/dL) | >5.2 | 78.95 | 94.12 | 0.918 | 0.861 to 0.957 | <0.0001 |
| Model 1 | >0.3372 | 85.53 | 95.59 | 0.951 | 0.902 to 0.980 | <0.0001 |
| Model 2 | >0.3615 | 90.79 | 92.65 | 0.956 | 0.909 to 0.983 | <0.0001 |
| Model 3 | >0.3659 | 86.89 | 97.06 | 0.963 | 0.914 to 0.988 | <0.0001 |
| Parameter . | Cut-off . | Sens . | Spec . | AUC . | 95% Confidence Interval . | Significance level P (Area=0.5) . |
|---|---|---|---|---|---|---|
| MNV (au) | >159.6 | 75.00 | 86.73 | 0.847 | 0.785 to 0.897 | <0.0001 |
| VDW (au) | >36.9 | 69.74 | 73.47 | 0.788 | 0.719 to 0.846 | <0.0001 |
| IL6 (pg/mL) | >22 | 77.05 | 94.12 | 0.887 | 0.819 to 0.936 | <0.0001 |
| CRP (mg/dL) | >5.2 | 78.95 | 94.12 | 0.918 | 0.861 to 0.957 | <0.0001 |
| Model 1 | >0.3372 | 85.53 | 95.59 | 0.951 | 0.902 to 0.980 | <0.0001 |
| Model 2 | >0.3615 | 90.79 | 92.65 | 0.956 | 0.909 to 0.983 | <0.0001 |
| Model 3 | >0.3659 | 86.89 | 97.06 | 0.963 | 0.914 to 0.988 | <0.0001 |
Test results and models' performance of sepsis group
| Parameter . | Cut-off . | Sens . | Spec . | AUC . | 95% Confidence Interval . | Significance level P (Area=0.5) . |
|---|---|---|---|---|---|---|
| MNV (au) | >157.1 | 78.64 | 81.63 | 0.852 | 0.807 to 0.890 | <0.0001 |
| VDW (au) | >37.4 | 59.71 | 77.55 | 0.740 | 0.687 to 0.789 | <0.0001 |
| IL6 (pg/mL) | >18 | 81.76 | 92.65 | 0.912 | 0.869 to 0.945 | <0.0001 |
| CRP (mg/dL) | >7.5 | 71.57 | 98.53 | 0.894 | 0.852 to 0.928 | <0.0001 |
| Model 1 | >0.3099 | 88.73 | 92.65 | 0.954 | 0.921 to 0.975 | <0.0001 |
| Model 2 | >0.3615 | 87.75 | 92.65 | 0.946 | 0.912 to 0.970 | <0.0001 |
| Model 3 | >0.2429 | 95.86 | 91.18 | 0.978 | 0.950 to 0.992 | <0.0001 |
| Parameter . | Cut-off . | Sens . | Spec . | AUC . | 95% Confidence Interval . | Significance level P (Area=0.5) . |
|---|---|---|---|---|---|---|
| MNV (au) | >157.1 | 78.64 | 81.63 | 0.852 | 0.807 to 0.890 | <0.0001 |
| VDW (au) | >37.4 | 59.71 | 77.55 | 0.740 | 0.687 to 0.789 | <0.0001 |
| IL6 (pg/mL) | >18 | 81.76 | 92.65 | 0.912 | 0.869 to 0.945 | <0.0001 |
| CRP (mg/dL) | >7.5 | 71.57 | 98.53 | 0.894 | 0.852 to 0.928 | <0.0001 |
| Model 1 | >0.3099 | 88.73 | 92.65 | 0.954 | 0.921 to 0.975 | <0.0001 |
| Model 2 | >0.3615 | 87.75 | 92.65 | 0.946 | 0.912 to 0.970 | <0.0001 |
| Model 3 | >0.2429 | 95.86 | 91.18 | 0.978 | 0.950 to 0.992 | <0.0001 |
We suggest to use combination of MNV and VDW with markers such as CRP and IL-6, and use diagnostic models created by using EMMA including these markers.
Table 1. Developed models to diagnose sepsis.
Model 1: Sepsis=-1.17+0.015*[CRP]+0.009*[MNV].
Model 2: Sepsis=-1.35+0.0136*[CRP]+0.0074*[MNV]+0.0123*[VDW].
Model 3: Sepsis=-0.94+0.0043*[IL6]+0.011*[CRP]+0.0069*[MNV]
No relevant conflicts of interest to declare.
Author notes
Asterisk with author names denotes non-ASH members.
