A high angiopoietin-2/angiopoietin-1 ratio is associated with a high risk of septic shock in patients with febrile neutropenia

Septic shock is one of the most severe complications of sepsis, frequently preceding multi-organ dysfunction syndrome and death. Cancer patients are at an increased risk of sepsis complications for several reasons, including the need for frequent invasive procedures, the effects of cancer on nutritional status and immunity, and the immunosuppressive effects of chemotherapy [1]. As life expectancy increases, so does the prevalence of age-related morbidities, such as cancer, considered one of the most important contributing factors for the increasing incidence of sepsis [2]. In this context, the discovery and validation of new biomarkers for septic shock development is regarded as a major research objective in the field [3, 4]. However, the implementation of a biomarker usually involves a stepwise process, in which results from initial studies need to be confirmed in larger studies in independent populations, ideally performed in experimental environments that more closely resemble the "real-world" use of the biomarker [3, 5].

In the last decade, the cellular and molecular mechanisms that regulate endothelial barrier integrity have been described [6]. In the embryo, VEGF-A, sFlt-1, angiopoietin (Ang)-1 and Ang-2 play critical roles in the physiological process of angiogenesis, during which the endothelial barrier is constantly assembled and disassembled, to allow the incorporation of new vessel sprouts. While VEGF-A and Ang-2 are involved in the destabilization of endothelial cell junctions, sFlt-1 and Ang-1 present distinct functions, stabilizing the endothelial barrier. After the initial demonstration that the mechanisms that regulate endothelial barrier formation in the embryo are also involved in the pathological disruption of the endothelial barrier in adults during sepsis [7], several studies explored the role of these proteins as biomarkers of septic shock development. Encouraging results were reported by several groups, and Ang-1 and Ang-2 emerged among the most promising endothelial-associated sepsis biomarkers [8‐19].

Patients with cancer and febrile neutropenia (FN) are one of the populations with the highest risks of sepsis-related mortality [20, 21]. In these patients, septic shock can evolve as a fulminant complication, despite aggressive treatment with broad-spectrum antibiotics and best supportive care. Nevertheless, these patients are usually excluded or largely underrepresented in studies about sepsis biomarkers and management. In a small exploratory study, we previously demonstrated that levels of modulators of endothelial permeability, such as VEGF-A and Ang-2 were altered in patients with FN that developed septic shock [22, 23]. In the present work, we sought to validate the use of VEGF-A, sFlt-1, Ang-1 and Ang-2 levels as biomarkers of septic shock development in a larger and independent cohort of patients with cancer-associated FN, in a more clinically relevant scenario regarding the use of a sepsis biomarker.

Among the several factors that contribute to a higher risk of sepsis complications in patients with cancer, chemotherapy-associated neutropenia is probably the single-most important one [29]. Although the risk is higher in patients with severe and long-lasting neutropenia, risk stratification tools are limited, and the majority of patients end up treated aggressively with hospitalization and broad-spectrum antibiotics [29]. The most accepted risk stratification tool for cancer-related FN is the MASCC score [27]. However, this score offers limited information for patients categorized as high-risk (MASCC ≤21), which encompasses most patients submitted to more intensive myelotoxic chemotherapy protocols. Accordingly, management of patients with cancer-related FN remains a challenging situation, for which the availability of informative biomarkers would be welcome.

In the last decade, mechanisms by which the endothelial barrier is established in the embryo have been described in detail [6]. As they were being described, some authors demonstrated that the same mechanisms were involved in the pathological disruption of the endothelial barrier observed during sepsis [7]. Soon after that, encouraging results were published about the use of VEGF-A, sFlt-1, Ang-1 and Ang-2 as biomarkers of septic shock development. In general, these studies demonstrated a consistent pattern in patients with septic shock, characterized by increased levels of proteins that loosen endothelial cell junctions during vessel formation, such VEGF-A and Ang-2, and decreased levels of barrier stabilizing proteins, such as Ang-1. Indeed, imbalances in Ang-1/Ang-2 ratio, expressed by a high Ang-2/Ang-1 ratio, recently emerged as one of the most promising biomarkers for septic shock development in several populations [11, 16, 30]. However, because patients with cancer were excluded or underrepresented in these studies, it is not known whether these results may be applied to patients with cancer.

In a preliminary study, we demonstrated that VEGF-A, sFlt-1, Ang-1 and Ang-2 were associated with sepsis severity in patients with FN [22, 23]. So, the objective of the present study was to validate the use of these biomarkers in a larger and independent cohort of patients with cancer and chemotherapy-associated FN, in experimental conditions more closely mimicking the scenario in which a sepsis biomarker would be ordered. This was achieved by: (1) sequential enrollment of patients; (2) limitation of exclusion criteria; (3) sample collection and initial processing performed by hospital staff; and (4) sampling time mimicking the local use of a biomarker (CRP) already used for patients with FN [14].

Not infrequently, results obtained in well-designed studies of diagnostic accuracy are not reproduced in similar independent populations [31], due to differences in time of sample collection and even in sample type and manipulation. In fact, we were not able to confirm previous reports of VEGF-A or sFlt-1 as informative biomarkers of septic shock development in patients with FN [23, 32]. We speculate that these discrepancies are associated with differences in the time-point of sample collection, and with the potential influence of severe neutropenia and thrombocytopenia in the variation of VEGF-A, as neutrophils and platelets are major sources of these mediator [33]. In addition, we cannot rule out the possibility that differences in baseline levels of VEGF-A between different types of malignancies [34] could have also played a role. Whatever the reasons for these discrepancies, they suggest that assays for VEGF-A and sFlt-1 possibly lack the robustness required for a sepsis biomarker to overcome the challenge of "real world" conditions, based on our local-practice, in the context of FN.

On the other hand, the main finding of our study was the confirmation that the Ang-2/Ang-1 ratio is independently associated with septic shock development in patients with cancer and high-risk FN, for whom other risk stratification tools provide very limited information. Although it should be noted that the relatively low sensitivity and specificity of the Ang-2/Ang-1 ratio do not qualify it as a biomarker that could influence decision-making by itself, it should also be pointed out that it presented a higher estimated diagnostic accuracy than CRP, Ang-1, Ang2 or the SOFA and MASCC scores, with the latter being the most widely used risk stratification tool for this population. In addition, the sensitivity and specificity of the Ang-2/Ang-1 ratio was similar to that of procalcitonin for the diagnosis of sepsis in critically ill patients [35]. Finally, our findings are important not only because a biomarker with a negative predictive value of 85% deserves to be evaluated as part of decision-making algorithms that incorporate clinical and laboratory markers in patients with FN, but also because they provide additional insights into the pathogenesis of septic shock in this particular population.

The use of the Ang-2/Ang-1 ratio rather than Ang-2 as a biomarker has been a matter of recent debate [36]. As opposed to Ang-2, which is exclusively stored in endothelial cells, Ang-1 is also expressed in platelets [37], so that ex vivo platelet activation can result in falsely elevated Ang-1 levels, limiting the accuracy of the Ang-2/Ang-1 ratio. Accordingly, the effect of pre-analytical variables, such as sample type and manipulation, have been cited as possible explanations for discrepant results of Ang-1 levels as a sepsis biomarker [36]. On the other hand, it is tempting to speculate that "true" imbalances in Ang-1 and Ang-2 levels do represent a biologically relevant determinant of endothelial barrier integrity, as suggested by several lines of evidence [38]. It is not known whether the effect of ex vivo platelet activation on Ang-1 levels is influenced by sample platelet count. However, our data showing a significant correlation between platelet count and Ang-1 levels only in patients with severe thrombocytopenia, but not in patients with higher and more variable platelet counts, suggest that severe hypoplastic thrombocytopenia could attenuate the influence of the pre-analytical variables that normally limit the determination of "true" Ang-1 levels. Therefore, although we acknowledge the technical limitations of Ang-1 determination, we speculate that more reliable Ang-1 levels can be obtained in patients with chemotherapy-induced severe thrombocytopenia.

One of the main objectives of our study was to determine whether Ang-1 and Ang-2 measurements would remain as relevant sepsis biomarkers in a population with different characteristics than those from studies of non-neutropenic patients. Neutrophils are critical components of the innate immune response, releasing cytokines and chemokines, engulfing pathogens and contributing to microbial killing by a range of peptides and by neutrophil extracellular trap formation [39]. It was interesting to note that the optimal cut-off of the Ang-2/Ang-1 ratio associated with septic shock in our population was very similar to the mean peak Ang-2/Ang-1 ratio observed in non-survivors in a study with non-neutropenic sepsis patients (5.0 vs 5.4, respectively) [11]. From a clinical standpoint, a significant difference between our population and non-neutropenic patients is the earlier collection of blood samples in the course of infection, when clinical signs of sepsis were mild or absent (as illustrated by the lower SOFA score in our patients). The reason for this characteristic lies in the fact that patients with FN are instructed to look for medical care immediately after fever onset, allowing earlier admission. In fact, we believe that the relatively lower Ang-2 levels in our patients are another consequence of this earlier evaluation. Of note, we have previously shown a five-fold increase in Ang-2 levels during the first 48 hours of FN in patients that evolved to septic shock [22]. Another important difference in septic shock between neutropenic and non-neutropenic patients is the high frequency of invasive fungal infections (IFI) in the former group, often associated with a protracted course of fever, and a longer time to septic shock. In fact, although most of our patients evolved to septic shock within a week from fever onset, four of the five patients in whom septic shock developed from 7 to 17 days presented probable or proven IFI. Together, the fact that a similar value of the Ang-2/Ang-1 ratio was associated with worse outcomes in these two different populations strengthens the concept that imbalances in Ang-2 and Ang-1 concentrations are indeed a relevant biomarker of septic shock development. Moreover, the fact that imbalances in this pathway remain relevant for the pathogenesis of septic shock in patients with severe neutropenia indicates the independence of the Tie/Angiopoietin pathway from neutrophil function during sepsis. Indeed, this concept is reinforced by a series of other recent studies showing that manipulation of angiopoietin pathways can improve sepsis outcomes in different animal models [15, 40, 41].

This study has several limitations. First, because of a higher than expected coefficient of variation of VEGF-A levels in our patients (CV% of 193%), and with only 20 patients with septic shock, our study had power to detect only extreme differences of VEGF-A levels between the outcome groups. Second, we did not evaluate baseline Ang-2 levels. So, it is conceivable that differences in Ang-2 levels present before chemotherapy, which have been shown to be associated with disease-free survival in high-risk myeloid malignancies [42], might have influenced our results. Arguing against this possibility is the fact that we and others have not found significant differences in pre-chemotherapy Ang-2 levels between different disease groups, even in cancers with significant angiogenic activity, such as multiple myeloma [22, 42]. A third limitation refers to the importance of the SOFA score as a readout of systemic organ failure in our population, since thrombocytopenia was the only category in which 61% of our patients scored, most of them with high scores. Since thrombocytopenia in our patients is simply a transient effect of chemotherapy, it is fair to assume that the diagnostic accuracy of this score is more limited in our population. We tried to minimize this limitation by avoiding the use of the score (mainly driven by thrombocytopenia) in our multivariate analysis, in which platelet count had already been used for adjustment. Fourth, we did not evaluate other outcomes that have also been associated with alterations of the Ang/Tie pathway, such as acute lung injury [43], so that our results can only be applicable to the context of septic shock development. Fifth, we did not control the effect of fluid management, which has been shown to influence Ang-2 levels [8], so that it is not possible to exclude the idea that more aggressive fluid therapy could have resulted in higher Ang-2 levels in some patients. However, extreme values of Ang-2 were observed in both outcome groups, a consistent finding in studies with this biomarker. Furthermore, fluid management before sample collection was fairly standardized in our study. Last, because three of our patients developed septic shock between enrollment and sample collection, our study does not support the use of the Ang-2/Ang-1 ratio as a predictive tool, but rather as a biomarker of septic shock development in a high-risk population.

Finally, we believe that an additional use of the Ang-2/Ang-1 ratio in sepsis in the future could be the identification of septic patient subgroups, in which disruption of the endothelial barrier could play a more prominent pathogenic role. In theory, the identification of these more "leaky" patients could rationally guide the use of new therapeutic strategies targeting pathways involved in endothelial barrier regulation [13, 44]. As targeted-therapies directed to different pathogenic mechanisms of sepsis begin to reach the phase of human clinical trials, the availability of biomarkers capable of identifying discrete patient subgroups among the highly heterogeneous population of patients with sepsis, could well be the difference between failure and success of new therapies for sepsis [4].

In conclusion, our study confirms that the measurement of the Ang-2/Ang-1 ratio is a biomarker of septic shock development in patients with cancer and chemotherapy-related FN.