The main objective of our study was to find indicators that allow us to make an early prediction of Gr-BSI or SIRS episodes, which we have called high-risk episodes (HRE). These two types of episodes were chosen as the outcome to be predicted due to the fact that they are two of the most serious infectious conditions that a pediatric cancer patient can present during treatment, the latter frequently being the consequence of the former.
Among the possible indicators, analytical biomarkers are of special interest. An ideal biomarker for use in febrile oncologic patients should be able to predict, identify, and thus stratify risk in febrile patients early in their clinical course. In addition, the biomarker should be able to provide robust discrimination of all parameters between mild and serious infections .
Many studies have tried to determine the best predictive biomarkers of infection in cancer with contradictory results in many cases [5, 8–12]. This fact may be due to differences in the cut-off points used, type of episodes to be predicted, and use of only univariate rather than both uni- and multivariate models.
In our study, in addition to the static values of biomarkers, we added their kinetics in the first 24 hours in order to try to approximate our predictive model to the changing reality that this type of infectious disease entails. We have not found in the literature any publication that has used these types of variables in pediatric oncology patients.
Among the biomarkers analyzed, those that proved to be good predictors of HRE in the multivariate analysis were an IL6-1 value greater than 164 pg / ml (OR = 26.03) and a CRP-2vs1 value greater than 291% (OR = 19.62).
IL6-1, the best predictor according to our study, had high sensitivity (92.8%), specificity (82.5%), positive predictive value (56.1%) and negative predictive value (98.1%) with an AUC of 0.89 (0,791-0,989). IL6-2 and IL6-2vs1 did not prove to be good predictors, although IL6-2 did have statistical significance in univariate analysis.
This finding concurs with other studies that show that IL6 is useful as an infection marker in initial moments of fever, especially in detecting patients with a low risk of presenting an HRE given its high negative predictive value [18, 20]. On the other hand, there are others studies where its utility has not been demonstrated . It is worth pointing out that the IL6 value loses its accuracy within 12 to 24 hours from the beginning of antibiotic treatment due its rapid decrement. This is shown by the fact that median IL6-2vs1 is a negative value.
Regarding CRP, neither the CRP-1 nor CRP-2 values showed statistical significance in predicting HRE in the multivariate analysis. However, CRP-2vs1 was clearly significant in the multivariate analysis. The increase of 291% present between CRP-1 and CRP-2 implied a risk of HRE that was 19.62 times greater than in the absence of this increase. This may be due to the fact that elevated CRP values can occur in these patients due to other types of less serious and slower-evolving infections, as well as non-infectious causes such as mucositis or the oncological disease itself. In these cases, even if the absolute values were high, they would not vary much in a period as short as 12 to 24 hours, as could happen in the HRE. In this case, we can observe that the dynamic approach to the infectious phenomenon provides more information than the assessment of static parameters.
Regarding PCT, neither the absolute values nor the variations in biomarkers showed statistical significance in predicting HRE in the multivariate analysis. In the initial analysis, cut-off points of 0.32 ng/ml for PTC-1, 0.94 ng/ml for PCT-2, and an 113% increase between PCT-1 and PCT-2 appeared to show a good capacity for discriminating HRE with AUC of 0.8, 0.83 and 0.81 respectively and remained significant in the univariate analysis. However, this statistical significance was not maintained in the multivariate analysis, so PCT could not be considered to be a good independent predictor if IL6 and CRP are available within the first 24 hours from the fever onset. This fact is also shown in a study with a large sample size and multivariate analysis carried out by Santolaya et. al., in which the predictive value of CRP, PCT and IL8 for severe sepsis was analyzed . In this study, the authors concluded that the use of PCT does not provide a significant benefit in the early detection of severe sepsis compared to CRP and IL8. Similarly, in our study, PCT does not appear to provide a significant benefit in the early detection of Gr-BSI or SIRS when IL6 and CRP are used. Conversely, this lack of usefulness of PCT seems discordant with the results from another study carried out by Mian et. al., which concluded that CRP and PCT were useful in risk stratification of febrile neutropenia episodes in pediatric oncology patients . This study had a slightly smaller sample size than our study and a multivariate analysis was carried out. Comparative analysis between studies must be performed cautiously as populations, statistical analysis and outcomes are different.
In clinical practice according to our results, if we only had one biomarker available in the laboratory, PCT would be the elecction because it had good diagnostic accuracy during the first 24 hours. IL6 had the best performance in the first hours from the fever onset and CRP-2vs1 after 12-24 hours. Therefore, if we have IL6 and CRP available, PCT did not add any value to IL6 and CRP combination.