In the present study of sepsis alert patients, we found that the novel biomarker p-calprotectin was elevated in patients with established infection and was superior to CRP, PCT, and NLR for detecting patients in need of ICU or HDU care.
Calprotectin has been shown to be a potential biomarker for discriminating sepsis patients from other patients with infection (16–19). Due to the heterogeneity of sepsis patients, some researchers have focused on the role of calprotectin in multi-organ failure (27). We wanted to study calprotectin in a well characterized cohort of patients with clinical suspicion of sepsis already in the ED and evaluate whether calprotectin may be of value in the early assessment of sepsis. We found that patients with infection who the multidisciplinary team in the ED decided to transfer to the ICU or HDU had significantly higher calprotectin than other patients (Fig. 2). This was not explained by higher comorbidity (as measured by Charlson comorbidity score) or disease severity (as measured by delta-SOFA) on hospital arrival. There was no significant difference in delta-SOFA score between the group ending up direct in ICU or HDU and the group transferred to ward. This shows the difficulties in discrimination patients with sepsis so severe that they need ICU care by only using delta-SOFA score. In this case the patients were also assessed by three different doctors. This also shows the need for better biomarkers, where calprotectin could have a place.
CRP, PTC, and NLR could not discriminate between those who were admitted to ICU/HDU and those admitted to wards, in line with findings from similar studies (10, 18, 27, 28). ROC curve analysis thus showed that calprotectin was superior to CRP, PCT, and NLR in assessing the need for ICU/ HDU care (Fig. 4, Table 4).
As shown in Fig. 4, p-calprotectin had an AUC of 0.65 for predicting the for ICU/ HDU care. Larsson et al. recently studied p-calprotectin in ICU patients and found an AUC of 0.67 for discrimination between sepsis and non-sepsis in ICU, and 0.64 for prediction of ICU death, i.e. similar findings as in our study (20). However, when looking at sensitivities, specificities, and predictive values, we considered p-calprotectin 4.0 mg/L to be the best cut-off value for predicting the need for ICU/ HDU care, with a sensitivity of 42% and a specificity of 83%. Larsson et al. found the best p-calprotectin cut-off to be 1.3 mg/L for discrimination between sepsis and non-sepsis, with sensitivity 81% and specificity 56%, and 1.1 mg/L for prediction of ICU death, with sensitivity 80% and specificity 46% (20). Other studies have shown optimal calprotectin cut-offs of 3.4 mg/L for differentiating sepsis from post-operation inflammation (PPV47% and NPV 69%), e.g. in a study by Simm et al, and 2.2 (unknown unit) with sensitivity 62.5% and specificity 69.7% for blood culture positive sepsis in children in a study by Decambrinos et al. (16, 19). This might be related to patient selection, analytic method for p-calprotectin or the nature of the study question.
P-calprotectin and the other biomarkers did not differ considerably in performance based on sex, age, co-morbidity and source of infection and there was no clear difference in p-calprotectin levels between bacterial and viral infection. This was also the case for CRP, PCT, and NLR, and may be due to the small number of patients with viral infection in the present study. Also, to be noted, p-calprotectin was not higher among patients with verified bacteremia (positive blood culture), compared to those with negative blood cultures in our study (Table 2).
Faecal calprotectin has been shown to be affected by immune suppression (29), but we find no difference in p-calprotectin levels in patients with or without immunosuppression.
On repeated samples, mostly NLR but also calprotectin decreased on day 2–3, whereas CRP and PTC increased, and had started to decrease on day 5–7. This may reflect different inherent kinetics of the biomarkers. Our findings differ from those of Bartakova et al, and Simm et al (16, 18). In Bartakova´s study all biomarkers hade the same dynamics as we found for calprotectin and NLR (18). However, in the study by Simm et al., only CRP had the same pattern as p-calprotectin (16). This may reflect different timing of sampling and a more heterogenous group of patients in our study compared with the others.
Since calprotectin increases early during infection, it could be potentially be useful in the pre-hospital care setting. Further studies will be needed to define its exact role.
The present study has several strengths. First, our study cohort was well characterized and already clinically stratified for suspected sepsis, since it was part of a sepsis alert used in routine practice in the ED. All patients, even those who did not have any infection, could be assumed to be severely ill due to the inclusion criteria in the alert.
The study also has some limitations. For ethical reasons we did not try to obtain consent from relatives to patients who were already deceased. This was a concern for the 35 patients who died before consents were obtained, and we did not study how many of the deceased patients who were transferred to ICU or HDU before death. Thus, many severely ill patients were not included in the study and it is hard to assess how this might have affected the outcome of the study.
In conclusion, in sepsis alert patients, p-calprotectin was elevated in those who a multidisciplinary team decided to transfer immediately to the ICU/ HDU. P-calprotectin was superior to traditional biomarkers for predicting the need for ICU/ HDU care.