Post chemotherapy FN in patients with hematological malignancies is a condition that carries a high risk of sepsis complications with mortality rates as high as 21% , usually preceded by septic shock. Patients with FN are a heterogeneous group in terms of risks of complications and mortality, and the identification of parameters capable to accurately identify high risk patients is one of the great challenges in their care. Clinical scores such as the MASCC , and laboratory parameters such as C-reactive protein, IL-6, IL-8 and procalcitonin have been recently proved to be useful tools for this purpose . However, it is generally acknowledged that these parameters are more useful to identify patients at low risk for complications, leaving room for refinements in the risk stratification of FN patients. Ideally, one such marker should be easily detected in samples obtained before the development of severe sepsis, and should not be influenced by cytopenias or by the inflammatory milieu associated with disease status. Elements that are directly involved in the pathogenesis of sepsis complications are thus more attractive candidates than non-specific elements of the inflammatory cascade.
Angiopoietins are a family of vascular growth factors with critical roles in embryonic and postnatal angiogenesis. Ang-1 and Ang-2 both act on the Tie2 tyrosine kinase receptor found primarily on endothelial cells, but appear to play antagonist roles . Apparently, Ang-1 promotes vessel stabilization whereas Ang-2 is involved in the destabilization of newly formed vessels. These processes of disassemble and reassemble of the endothelial lining of primitive blood vessels is important because it allows the incorporation of new endothelial cells to growing endothelial tubes, leading eventually to the formation of functional and mature blood vessels . A role of angiopoietins in the pathogenesis of septic shock is supported by multiple lines of evidence. First, increased levels of Ang-2 levels have been demonstrated in adult and pediatric patients with sepsis in intensive care units (ICU) and these levels correlated with sepsis severity [11, 12, 14, 25]. Second, serum from patients with sepsis has been shown to disrupt endothelial architecture, an effect that correlates with Ang-2 levels, is reversed by Ang-1 and is mimicked by recombinant Ang-2 . Finally, Ang-2 levels have been shown to correlate with pulmonary permeability edema and occurrence of acute respiratory distress syndrome in mechanically ventilated patients . More recently, Kumpers et al, studying a population of 43 medical ICU patients demonstrated that Ang-2 levels at the time of ICU admission was a strong predictor of mortality . Ang-1 levels were also evaluated in some of these studies, but so far studies have failed to a relationship between Ang-1 and sepsis outcomes.
In our prospective study we explored the time-course of Ang-1 and Ang-2 in a population of patients with high risk of sepsis complications before the development of severe sepsis. As far as we are aware, the significance of Ang-1 and Ang-2 levels has not been studied in patients with FN. An additional contribution of our study is that we prospectively evaluated the significance of Ang-1 and Ang-2 levels at an earlier time-point in the development of sepsis (which developed in a median time of 4.1 days after fever onset), as opposed to studies that evaluated patients at the ICU. This difference is evidenced by the median SOFA score of our patients (4 range 1-8) compared to higher median SOFA scores (16 range 1-22) from other key studies . Last, Ang-1 and Ang-2 levels were serially evaluated at three time points, thus offering a view of the time-course of Ang-1 and Ang-2 release in the early hours of sepsis.
For this study we only included patients with hematological malignancies and FN after intensive chemotherapy that were treated as inpatients from day one of chemotherapy until the resolution of neutropenia. By doing so, we intended to obtain a representative sample of patients with high risk of sepsis complications. Furthermore, the fact that all patients were treated as inpatients allowed us to standardize important variables that could affect the validity of our results such as the time between fever onset and sample collection. So as to limit as much as possible the influence of different diagnosis and chemotherapy regimens on our results, Ang-1 and Ang-2 levels were also collected immediately before the initiation of chemotherapy, allowing differences between outcome groups to be compared not only as absolute values, but also as fold-increase from baseline levels, which did not alter our results (data not shown). In fact, neither Ang-1 nor Ang-2 levels were significantly different before the initiation of chemotherapy between patients with non-complicated sepsis and septic shock, supporting the uniformity of our group of patients as far as the two study variables were concerned.
The main finding of our study is that the relative concentration of Ang-1 and Ang-2 are different in subgroups of patients with FN that evolve to non-complicated sepsis compared to patients that develop septic shock, and that evaluation of these two proteins within the first 48 hours after neutropenic fever, before the development of any signs and symptoms of septic shock, is a promising tool to discriminate high risk patients with FN. In accordance with previous studies, Ang-2 levels were significantly higher in patients with septic shock compared to patients with non-complicated sepsis. This difference was not present at the time of fever onset, rose sharply and became evident after 48 hours, when levels in the poorer outcome group were 8 times higher than in the good outcome group. Importantly, Ang-2 baseline levels were similar. As in previous studies, we were not able to detect any significant difference of Ang-1 levels between study groups. However, when the relative concentrations of these two antagonistic cytokines were evaluated, we did observe a relative deficiency of Ang-1 compared to Ang-2 levels in patients that developed septic shock, that was evident early at the time of fever onset. It has been known for more than a decade that Ang-1 can protect adult vasculature against VEGF-induced plasma leakage . Ang-1 has also been shown to protect mice from endotoxic shock . In this respect, it is tempting to speculate that the imbalance between Ang-1 and Ang-2 levels present already at the time of fever onset is associated with a poorer outcome of sepsis in FN, and possibly in other patients with sepsis. This hypothesis is well illustrated in our work by the divergent trend of Ang-2/Ang-1 ratio observed in each outcome subgroup of patients. Early in the course of sepsis, patients that evolved to septic shock presented a 7-fold increase in the Ang-2/Ang-1 ratio, whereas patients with non-complicated sepsis presented a 0.8 decrease when compared to baseline (pre-chemotherapy) ratios. After 48 hours, this divergent trend persisted, with patients that evolved to septic shock presenting a 10-fold increase, and patients with non-complicated sepsis with a 0.4-fold decrease. We also evaluated VWF antigen levels, which is increased in patients with sepsis , and is an indicator of endothelial dysfunction and stimulation . In contrast to Ang-2, no difference could be observed between both patient groups at any time-point (table 1). We also demonstrated a significant, though weak correlation of Ang-2 concentrations 48 hours after fever onset and of Ang-2/Ang-1 ratio (both at fever onset and after 48 hours) with the SOFA score of sepsis severity. Based on these results, we estimated diagnostic accuracy of Ang-2 concentrations 48 hours after fever and of Ang-2/Ang-1 ratio at fever onset. Despite wide confidence intervals, area under the ROC curves suggest that both biomarkers should be further investigated as promising tools to discriminate high-risk FN patients in studies with larger sample sizes. The relationship between angiopoietin concentrations and 30-day mortality was also evaluated. Using a multivariate Cox model, we demonstrate that the Ang-2/Ang-1 ratio at the time of fever onset can be an independent predictor of sepsis survival in our population, an observation that will have to be validated in larger studies.
Recently, the time course of Ang-1 and Ang-2 release was evaluated in a model of human endotoxemia , which showed that Ang-2 release is initiated 2.5 hours after LPS challenge, and observed no significant variations in levels of Ang-1. Ang-2 is a Weibel-Palade body-stored molecule that is rapidly released upon endothelial stimulation. In contrast, Ang-1 is believed to exert its vessel-sealing effect by low-level constitutive activation of the Tie2 receptor, in a model in which constitutive Ang-1/Tie2 interactions control endothelial barrier integrity as a default pathway, and Ang-2 acts as a dynamically regulated antagonizing cytokine . Our results showing that patients that evolve to septic shock present an initial relative deficiency of Ang-1 associated with a sharp increase in Ang-2 levels in the first 48 hours of sepsis supports that similar events can be involved in the pathogenesis of septic shock in FN.
There are certain limitations of our study that need to be acknowledged to avoid overinterpretation of its conclusions. Research about new diagnostic tools follow a sequence of phases along which the questions answered by each phase progress from the demonstration that a biomarker behaves differently in diseased and normal individuals (phase I) to the ultimate demonstration of improved clinical outcomes after its incorporation (phase IV) . Ours is a phase II study in which the performance of a new diagnostic assay was tested in patients with a defined condition (febrile neutropenia) and potential different outcomes, but in which several confounding variables were controlled. Phase II studies tell us whether the test shows diagnostic promise under ideal conditions. Therefore, our study design does not allow us to universalize our conclusions to clinical settings where the impact of treatment related and other variables might reject our conclusions. For the formal validation of the assay and its incorporation intro clinical practice, a larger sample size (ideally from distinct centers) under less controlled conditions should be used. The relatively low number of deaths used for the multivariate analysis also deserves to be discussed. Our study design required a very specific population of patients, with stringent inclusion criteria to control confounding variables, so that a further increase in sample size was not feasible. On the other hand, it has been demonstrated that relaxing the rule of 10 events per variable in Cox models is possible without unacceptable increases in confounding bias . With this in mind, we believe that as long as the above-mentioned limits of our study are acknowledged, the association of 30-day mortality with Ang-2/Ang-1 ratio is an important information to the literature.