We have demonstrated that the IG count discriminates between SIRS patients with and without infection. We observed a significant difference within the first five days after the initial SIRS alert, and particularly within the first 48 hours (P < 0.0001). Polymorphonuclear neutrophil granulocytes (PMN) are the first line effectors of host defense against bacteria and other pathogens representing about 50% to 60% of all circulating white blood cells (WBC). Induced by granulocyte-colony stimulating factor (G-CSF), PMN develop from progenitor cells and maturate in the bone marrow over several stages into mature segmented neutrophils. After a maturation period of seven to ten days they migrate into the peripheral blood. Healthy individuals do not have immature granulocytes present in their peripheral blood and consequently the incidence of IG in the periphery indicates substantially increased bone marrow activation as seen in sepsis. Besides the microbiological evidence for infection there are no other generally accepted accurate clinical parameters defined by the ACCP/SCCM Consensus Conference Committee to distinguish sepsis from SIRS so far. Therefore we investigated changes in immature granulocyte numbers in patients with infections. In a previous study, Selig et al. found significantly elevated numbers of myeloid progenitor cells in patients with bacterial infections compared with healthy controls. In another study, Ansari-Lari et al. found a significantly higher percentage of immature granulocytes in infected than in non-infected patients and designated a percentage of IG of more than three (IG% > 3) as a predictor of sepsis, with a specificity of more than 90%. In that study patients were enrolled who were admitted to the Emergency Department or already were hospitalized and IG% was measured over five consecutive days. In the present study the focus was on patients who were admitted to the ICU and IGs were measured over twenty-one consecutive days. In line with these previous findings, our study confirms that the absolute number of immature granulocytes is significantly higher in infected patients than in non-infected controls. We also demonstrated that the IG count had greater sensitivity and specificity (89.2% and 76.4% respectively) than other parameters (CRP, LBP and IL-6) which all had sensitivities of less than 68%.
We also compared the performance of the IG count with CRP, LBP and IL-6, which are parameters commonly evaluated in patients with suspected infection or sepsis. Although all parameters except CRP were significantly increased in infected patients, the highest discriminative power for infection in the first 48 hours was shown by the total number of immature granulocytes (P < 0.0001). ROC analyses showed the highest area under the curve (AUC) for the IG count together with the highest sensitivity (89.2%). Only the AUC for IGs showed significance compared to all other clinical parameters in the first 48 hours.
In later periods the absolute numbers of IGs showed the highest significance for positive prediction of infection during period 5 (days 3–5). While LBP and IL-6 levels were not meaningful, the CRP level was significantly elevated. Synthesis of the acute phase protein CRP is activated by the pro-inflammatory cytokine IL-6, which is in accordance with our findings since we observed a significant discriminatory power of CRP only during period 5 whereas elevated IL-6 levels occurred within the first 48 hours. By period 10 (days 6–9) none of the evaluated parameters were meaningful and no significant difference in the AUC of different parameters could be observed. This may have been due to beneficial effects of early antibiotic treatment given to almost every patient suspected of having an infection. We conclude that the IG count displays superior discriminative power for infection over CRP, LBP and IL-6, particularly within the decisive first 48 hours after the onset of SIRS.
In addition, for the first 48 hours a high diagnostic odds ratio (DOR) for infection of 26.7 could be calculated for the IG count. All other parameters such as CRP, LBP and IL-6, had a DOR lower than 10 at any point in time.
Secondly, we examined the ability of IG count, CRP, LBP, IL-6 and the SOFA score to predict ICU mortality. As expected, the SOFA score, which is well-known and widely used as a predictor of mortality in patients with severe sepsis[31, 32], showed the greatest significant correlation to ICU mortality during the entire observation period. A significant differentiation between survivors and non-survivors was obtained from IL-6 levels in periods 1 and 5 but surprisingly also at later stages during periods 15 (days 10–14) and 20 (days 15–21), as well. In early stages a correlation of high IL-6 levels with mortality in critically ill patients is well-known[33, 34]. However, the correlation is not usually sustained and IL-6 is considered to have a short half-life. The significant values observed even during periods 15 and 20 in the present study may be explained by the increasing relative number of infected patients during these periods because of the earlier discharge of non-infected patients from the ICU. Thus, the balance of the two examined groups is skewed towards the infected population. The significant values of the IG count for periods 15 and 20 may be interpreted in the same manner. Therefore, the IG count may not have prognostic power for mortality as is the case for CRP and LBP. For CRP these findings are consistent with the literature where no correlation was observed between CRP concentrations and the severity and mortality of sepsis. The prognostic power of LBP for the occurrence of sepsis is controversial. In one study the LBP level was significantly elevated in non-survivors vs. survivors on days 2 and 7. In contrast, although a higher LBP concentration was found in infected than in non-infected patients, a correlation between LBP and sepsis mortality could not be detected in another study.
Recently, many potential biomarkers to identify sepsis and predict disease outcome have been examined. In particular, acute phase proteins such as procalcitonin (PCT) have been investigated. PCT is now used as a supportive clinical parameter to clarify diagnosis, but due to its high negative predictive value (99%) it is more frequently used to rule out sepsis. A comparison of the IG count with PCT in terms of positive predictive values for sepsis should be considered for future studies.
While our findings suggest that the IG count is not suitable as a prognostic marker for mortality it is a good and useful early marker for distinguishing infected and non-infected patients. Our study population included a wide range of patients with diverse causes for ICU admission and various disease stages. The distinct correlation of the IG count with the presence of infection shows that the former parameter is representative in general and not only valid for a special patient subgroup. The IG parameter on the XE-2100 has been previously validated. In this study, the automated IG count was compared to routine manual microscopy using the National Committee for Clinical Laboratory Standards (NCCLS) document H20-A, resulting in very good agreement between the two methods with a correlation coefficient of r = 0.9. In addition, reference values for IGs measured in the “differential channel” were generated and a 5th percentile of zero and a 95th percentile of 0.03x109/L with neither gender nor age dependency was detectable. Taken together, the automated IG count was found to be superior to manual microscopy whilst providing fast, inexpensive, accurate and reliable quantification, particularly for very small proportions with counts less than 5% of total WBC. Our findings are in line with a recent work by Mardi et al. who also demonstrated the ability of white cell morphology to predict sepsis.
In summary, we have shown that, within the first 48 hours after the first SIRS alert, an elevated number of immature granulocytes indicates infection. Measuring the IG count with a fully automated hematology analyzer provides new possibilities for rapid and early screening of SIRS patients on the ICU with suspected infections.