Sepsis-Related Outcomes of Patients with Philadelphia-Negative Myeloproliferative Neoplasms

Abstract We analyzed the National Inpatient Sample (NIS) database to study the sepsis-related outcomes in patients with Philadelphia negative myeloproliferative neoplasms (MPN). A total of 82,087 patients were included, most had essential thrombocytosis (83.7%), followed by polycythemia vera (13.7%), and primary myelofibrosis (2.6%). Sepsis was diagnosed in 15,789 (19.2%) patients and their mortality rate was higher than nonseptic patients (7.5% vs 1.8%; p < .001). Sepsis was the most significant risk factor of mortality (aOR, 3.84; 95% CI, 3.51–4.21), others included liver disease (aOR, 2.42; 95% CI, 2.11–2.78), pulmonary embolism (aOR, 2.26; 95% CI, 1.83–2.80), cerebrovascular disease (aOR, 2.05; 95% CI, 1.81–2.33), and myocardial infarction (aOR, 1.73; 95% CI, 1.52–1.96).


Introduction
The 2016 version of the World Health Organization (WHO) classification system for hematopoietic and lymphoid tumors categorized the myeloproliferative neoplasms (MPNs) with Philadelphia chromosome-negative MPNs or JAK2/CALR/MPL mutation-related MPNs as: polycythemia vera (PV), essential thrombocytosis (ET), and primary myelofibrosis (PMF) (1). The most commonly reported complications from Philadelphia-negative MPNs are thrombosis, bleeding, and leukemic transformation, all of which have been well documented. Nonetheless, the outcomes of severe infectious complications, such as sepsis, have been seldomly described at a population level (2)(3)(4).
The 2016 third international consensus defined sepsis as a life-threatening organ dysfunction caused by a dysregulated host response to infection (5). Patients with hematological malignancies are prone to develop sepsis due to immunosuppression, particularly those with acute myeloid and lymphocytic leukemia, non-Hodgkin lymphomas, and multiple myeloma (6,7). However, a large Swedish epidemiologic study showed a twofold increase in the rate of sepsis among patients with MPN compared to the normal population, even in those not treated with cytoreductive medications (8).
Patients with hematologic malignancies have a 43% sepsis-related mortality (9,10). Furthermore, a recent study reported that if sepsis progresses to septic shock, the 90-day mortality rate can be as high as 80.6% (11). Despite the rising concern about infections and sepsis on MPN patients, the overall effect of sepsis in-hospital mortality is not well studied. Hence, we aimed to study the impact of sepsis on mortality and length of stay in hospitalized Philadelphia chromosome-negative MPNs patients and to identify the risk factors associated with mortality.

Design and data source
We queried the National Inpatient Sample (NIS) database from 2016 to 2018. The NIS is part of a family of administrative databases developed by the Healthcare Cost and Utilization Project (HCUP) and sponsored by the Agency for Healthcare Research and Quality (AHRQ) (12). It was designed to produce U.S. regional and national estimates of inpatient utilization, charges, and outcomes. It contains patient demographic characteristics (e.g., age, sex, race), International Classification of Diseases, Tenth Revision, Clinical Modification/Procedure Coding System (ICD-10-CM/PCS) diagnosis and procedures, severity and comorbidity measures, hospital characteristics, discharge status, and length of stay (LOS). Weighted, it represents 35 million hospitalizations nationally and approximately 20% of all discharges from U.S community hospitals. The 2016 NIS database included 46 states plus the District of Columbia and 4,575 hospitals. The 2017 was composed of 47 states plus the District of Columbia and a total of 4,584 hospitals. Finally, the 2018 database represented 47 states plus the District of Columbia with 4,450 hospitals. During each year, approximately seven million unweighted discharges were recorded. We completed a data user agreement with the AHRQ prior to using the NIS databases. The databases used are consistent with the definition of limited data sets under the HIPAA (Health Insurance Portability and Accountability Act) Privacy Rule and contain no direct patient identifiers. Therefore, since the NIS is a publicly available and de-identified database, the present study was not submitted for approval to an institutional review board.

Study variables
Patient and hospital variables were pooled from the NIS databases. We obtained demographic variables such as gender, age, race, healthcare payer from NIS Core Data Elements. Hospitalrelated data included bed capacity, hospital control, geographical region, and teaching status. To summarize the comorbidities, we used the Elixhauser comorbidity index (ECI), first described in 1998 as significant predictors of outcomes. To create and analyze the comorbidities from the ECI, we used a previously validated method (14). We also identified common MPN complications such as myocardial infarction (I21.x, I22.x, I25.2), cerebrovascular disease (G45.x, G46.x, H34.0, I60.x -I69.x), and pulmonary embolism (I26.x). A hospitalization complicated by sepsis was defined as an admission having any of the sepsis codes as a discharge diagnosis.

Outcomes
Our primary objective was to determine the factors associated with in-hospital mortality in patients with MPN. The secondary objective was to describe the inpatient mortality, LOS, and cost of hospitalization in patients with and without sepsis. Inpatient mortality, LOS, and cost of hospitalization were obtained through the NIS database coding for 'DIED', 'LOS' and 'TOTCHG' respectively.

Statistical analyses
We used descriptive methods to report baseline patient and hospital characteristics. Continuous variables are represented as the median and interquartile range (IQR). Categorical variables are shown as frequencies with absolute numbers and percentages. To evaluate differences between the patients with sepsis and those without sepsis, we computed the Pearson chi-squared test and the Mann-Whitney U-test for categorical and continuous variables, respectively. We used a univariate and multivariate logistic regression to estimate the association between clinical characteristics, including sepsis, with in-hospital mortality. The multivariate model included clinically relevant variables significantly associated with mortality in the univariate analysis (p < .05). We adjusted for sex, race, age, hospital characteristics, median income, and payer type. All tests were two-sided, and statistical significance was defined as p < .05. All analyses were performed using Stata Statistical Software version 14 (StataCorp, College Station, TX, USA).

Baseline characteristics
A total of 21,400,282 patients were identified in the NIS database from 2016 to 2018. After excluding pediatric patients and those with more than one MPN, 82,087 patients with Philadelphia-negative MPN were included in the study ( Figure 1). The baseline characteristics of included patients are described in Table 1. The median age of the entire cohort was 63 (IQR 49-76); the majority of them were female (54.6%). By race, 67.4% of the patients were white, 17.1% black, 9.6% Hispanic, and 5.9% other. The most common type of MPN was ET (83.7%), followed by PV (13.7%) and PMF (2.6%). The median ECI was 4 (IQR 2-5). Nearly three-quarters of the admissions were in urban teaching hospitals (70.1%). Most of the patients (72.6%) were admitted to private, not-for-profit hospitals (Supplemental eTable 1). More than half of the admissions were covered by Medicare (52.3%), followed by private insurances (22.4%) and Medicaid (17.8%). Almost a third of the patients were in the first quartile of median income (30%), followed by a quarter of them that were in the second quartile (25.8%).

In-hospital mortality
There were a total of 2,360 in-hospital deaths, 1185 (1.8%) in the nonseptic patients, and 1,175 (7.5%) in the septic patients. In the multivariate analysis, several factors were independently associated with in-hospital mortality. Among them, sepsis was associated with the highest odds for mortality (aOR, 3 (Table 2). Among the MPNs, PMF was associated with higher mortality (aOR, 1.7; 95% CI, 1.37-2.10). A second multivariate model was developed to assess the effect of the three sepsis severity categories, sepsis, severe sepsis, and severe sepsis with septic shock (Supplemental eTable2). The second model showed a marked incremental risk of mortality as the severity of sepsis increased.
To visually assess the influence of the number of comorbidities in the main outcome, we plotted the predicted mortality by ECI (Figure 2). We observed an incremental effect on the ECI on mortality in the entire cohort. The effect was more pronounced in the sepsis patients than those who did not develop sepsis and was independent of the type of MPN. For instance, the predicted mortality of an ECI of 10 was 24.1% in the septic patients compared to 7.5% in the nonseptic counterparts.

Discussion
We identified over 80,000 patients with Philadelphia-negative MPNs, and nearly one-fifth   were baseline characteristics such as advanced age and liver disease, thrombotic complications such as pulmonary embolism, cerebrovascular disease, and myocardial infarction, and acute nonthrombotic complications such as fluid and electrolyte disorders, and neutropenia. Among these risk factors, sepsis had the highest association with poor outcomes, increasing mortality risk by three-fold. To the best of our knowledge, this is the largest study evaluating sepsis outcomes in patients with MPN in a U.S. population. Sepsis and septic shock are among the main causes of intensive care admission and mortality among patients with malignancies, (11,15,16) particularly in non-Hodgkin lymphomas, acute myeloid leukemia, and multiple myeloma patients (17). However, data describing the outcomes of patients with MPN and sepsis are scarce, particularly in critically ill patients (18). In our cohort, sepsis was diagnosed in approximately 19% of MPN patients. The increased risk for sepsis was also described by Landtblom et al. (8) in the Swedish population-based matched cohort study that identified PMF as the MPN with the highest risk of developing sepsis (HR, 6.2).
There are potential mechanisms that could explain the susceptibility of sepsis within the MPNs population, particularly in the PMF subset. First, patients with PMF have a dysregulated immune system with increased production of pro-inflammatory cytokines that causes a chronic inflammation state (8). Second, the presence of a low burden of JAK2V617F has been associated with the PMF's cytopenic phenotype, which in turn, entails a decreased overall survival than patients without cytopenias due to bone marrow failure (19,20). Similarly, Guglielmi et al. found that patients with low JAK2V617F burden have a higher probability of dying from infections and sepsis than those with a high JAK2V617F burden (20). Third, cytopenic patients with transfusion dependency have an increase burden of infections. A recent retrospective study showed that transfusion-dependent patients with myelofibrosis had an increased risk for infections (HR, 2.13; P ¼ .019) compared to nontransfusion-dependent patients (21). Whether the increased risk to infections is related to the transfusions or the cytopenic phenotype of myelofibrosis warrants further investigation. Fourth, treatment with JAK inhibitor drugs like ruxolitinib has been associated with an increased frequency of infections or reactivation of latent diseases like tuberculosis (22)(23)(24). Inhibition of the JAK-STAT pathway, negatively affects the host immunity and leading to immunosuppression. In particular, ruxolitinib causes impaired cytokine signaling (IL1, IL6, TNFa, and IFN-c), dendritic cell dysfunction, reduced natural killer cell levels, and decreased T-cell response (22,25). Additionally, in certain cases, treatment with ruxolitinib reduces the JAK2V167F allele burden (26,27).
In our study, sepsis was identified as a risk factor for mortality. A similar association was described previously. In a large Swedish population-based study, patients with MPN and infection had an increased risk of death than non-MPN patients that developed infections (HR, 2.7; 95% CI, 2.4-3.1) (28). Interestingly, the risk of dying with sepsis was greater than life-threatening thrombotic complications like pulmonary embolism, cerebrovascular disease, and myocardial infarction, considered serious complications in MPNs patients. For instance, arterial and venous thromboembolisms cause approximately 45% of all disease-associated fatal events (29). Furthermore, patients with MPNs have higher mortality from cardiovascular and cerebrovascular disease compared to the general population (28). The latter being more evident among the younger patients (28).
We cannot discard that liver disease, fluid and electrolyte disorders, and coagulopathy, which were also associated with mortality, could have resulted from sepsis-related multiorgan dysfunction. However, since patients with sepsis represented less than 20% of the entire cohort, this effect is less likely. We did, however, conducted a second multivariate analysis removing these factors and classifying sepsis by severity (Supplemental eTable 2). As expected, patients with septic shock had the worse outcomes. Furthermore, we also found an increase in mortality in patients with higher ECI with and without sepsis. This effect was similar in the three subtypes of MPN.
In addition to higher mortality, sepsis also increased the LOS, which was twice longer than non-septic patients (8 vs 4; p < .001). This finding correlates with studies in septic patients and other malignancies (10,30,31). Furthermore, septic patients had almost double hospitalization charges (73,861 vs 40,119; p < .001). Increased hospitalization cost is related to medication costs, staff, and the need for ICU level of care in severe cases (30,32).
Some limitations should be taken into consideration when interpreting our results. First, due to the data's administrative nature, we were limited by the ICD-10 codes, which can be subject to mistakes. For instance, the high prevalence of ET in the present cohort (83%) could be a result of a coding inaccuracy. This phenomenon is not unique to our study, previous studies using the NIS database also showed similar prevalence of ET (13,33). Second, we were unable to gather information regarding the clinical status of the MPN, the histologic characteristics, and management. Similarly, we could not ascertain the main cause of admission in the present cohort. This critical limitation should be evaluated to confirm our findings. Despite the complexities of carrying out research in patients with cancer and sepsis, (18) further research in this area is warranted. Future studies should report granular clinical data and develop prognostic scores specific for this population. Identification of MPN patients at a higher risk could help future investigators evaluate whether there is a role for antimicrobial prophylaxis in the most vulnerable subpopulations.
In this cohort of MPNs patients, sepsis was identified as the risk factor with the greatest odds of in-hospital mortality in MPN patients. The risk was even higher than common MPNs complications like pulmonary embolism, cerebrovascular disease, and myocardial infarction. Therefore, sepsis should be considered a concerning complication of MPN patients. Strategies and protocols to promote early detection and prevention of sepsis should be developed for this population. Decreasing the incidence of sepsis could also lessen the healthcare system's burden by lowering the length of hospitalization and costs. Further research should assess if the increased risk is secondary to disease-specific characteristics or associated with the treatment regimens.

Declaration of interest
No potential conflict of interest was reported by the author(s).

Funding
This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Data availability statement
The data that support the findings of this study are available from the corresponding author upon reasonable request.

Meeting presentation
Preliminary findings of this work were presented in part as abstract e19038 at the virtual Annual Meeting of the American Society of Clinical Oncology, June 4-8, 2021.