Clinical features and outcome of Aeromonas sobria bacteremia in pediatric and adult patients with hematologic malignancies: A single-center retrospective study in Peru

Background Previous studies have found that healthcare-associated bacteremia (HAB) by Aeromonas species is associated with mortality. However, there is limited data on this outcome in patients with hematologic malignancies. This study aimed to identify the clinical features of patients with malignant hematologic diseases diagnosed with Aeromonas sobria bacteremia and to evaluate whether the type of bacteremia, community-acquired bacteremia (CAB) or HAB, is associated with mortality. Methods We retrospectively reviewed the clinical records of pediatric and adult patients between January 2000 and December 2017. Clinical characteristics were compared between CAB and HAB. Additionally, we stratified based on age group. Survival outcomes were assessed with Kaplan-Meier curves and a multivariate Cox regression analysis. Results A total of 37 patients (median age 24 years) were identified; 23 (62%) had HAB and 14 (38%) had CAB. Overall, the most common presenting symptom was abdominal pain (41%). Acute lymphoblastic leukemia (n = 12/15, 80%) and acute myeloid leukemia (n = 8/22, 36%) were the primary hematologic malignancies in pediatric and adult patients, respectively. CAB patients had worse overall survival (OS) rates at 30 days in all (43% versus HAB 91%, p = 0.006) and adult patients (30% versus HAB 92%, p = 0.002). Cox regression analysis found that quick Sequential Organ Failure Assessment and CAB were statistically significant factors associated with mortality. Low antimicrobial-resistant was noted, except for ciprofloxacin (n = 5/37, 14%). Conclusion Our study found a worse OS among patients with hematologic malignancies and CAB by Aeromonas sobria. Our results suggest that patients with CAB present with a worse disease severity. These findings should aid clinicians to determine the survival prognosis in this population.


Introduction
Aeromonas species, a gram-negative heterotrophic bacteria mainly found in warm climates, cause human disease through contamination of water, seafood, meat, and vegetables [1,2]. Incidence of Aeromonas infection varies from 20 to 76 cases per 1,000,000 persons. The most common species isolated are A. hydrophila, A. sobria, and A. caviae, accounting for about 86%, 11%, and 3% of the cases, respectively [1,3,4].
These bacteria cause a broad spectrum of human infections, such as gastrointestinal tract infection, septicemia, acute respiratory tract infection, soft-tissue infection, among others [5][6][7]. Although the disease is more commonly seen in immunocompromised individuals, it can develop in immunocompetent hosts [6]. Factors associated with mortality by A. sobria include an APACHE II score of 20 or more points, thrombocytopenia, and the presence of diarrhea [8]. Similarly, other related factors include hypotension, impaired renal function, liver cirrhosis, lower limb infection, and altered consciousness [9][10][11].
Nosocomial infection is a typical factor for poor prognosis and mortality, especially in patients with decreased immunity [12]. Similarly, a previous study in patients with A. sobria infection found that healthcare-associated bacteremia (HAB) was related to mortality compared to community-acquired bacteremia (CAB) [13]. However, there is scarce literature on this association in patients with hematologic malignancies and A. sobria bacteremia [8,9,14]. This population is commonly immunocompromised, and outcomes might differ from previous studies that included patients with a variety of comorbidities apart from hematological malignancies. Therefore, the present study aims to describe the clinical features and outcome of patients with malignant hematologic diseases diagnosed with A. sobria bacteremia and to identify the factors associated with survival in this population. Specifically, we assess whether the type of bacteremia (CAB or HAB) is related to mortality in all patients and by age group (pediatric and adults) stratification.

Patients and inclusion/exclusion criteria
We conducted a retrospective study of pediatric and adult patients diagnosed with malignant hematologic diseases and A. sobria bacteremia at the National Institute of Neoplastic Diseases (Instituto Nacional de Enfermedades Neoplásicas-INEN) in Lima, Peru, between January 2000 and December 2017. Malignant hematologic diseases included in this study were: the acute phase of myeloid and lymphoid leukemias, high-grade lymphomas (i.e., diffuse large B-cell lymphoma, or Burkitt lymphoma), and multiple myeloma. Patients developed bacteremia by A. sobria while on active cancer treatment or within 30 days from their last dose. Bacteremia was diagnosed by clinical and laboratory findings, with A. sobria identified in blood cultures. We excluded patients with a previous history of a bone marrow transplant, solid tumors, and those in which outcomes were not assessable (e.g., lost to follow-up or incomplete medical records). Medical records were identified using the databases from the Microbiology Department and the Hematology and Medical Oncology Department.

Study variables
We defined pediatric patients to those aged 0-19 years old, based on the recommendation of the International Agency for Research on Cancer [15]. CAB was defined as the presence of clinical symptoms 72 hours before hospital admission, while HAB was defined as the presence of symptoms during hospitalization or the development of symptoms within 72 hours after hospital discharge. Data can be found in the following link: https://github.com/valcarcelb/ AeroS. Patients were followed for 30 days after isolation of A. sobria from the bloodstream. Grade 3 to 4 cytopenias (i.e. leukopenia, anemia, and thrombocytopenia) were defined as absolute leucocyte count of �1.0 × 10 3 leukocytes/μL, hemoglobin <8 g/dL, and platelets <50 × 10 3 cells/μL, respectively. Death by A. sobria was defined as the one occurring within the first 30 days after its isolation from the bloodstream, or if death occurred during the acute course of sepsis with post-mortem isolation of A. sobria. Infection resolution was considered when there was an improvement in clinical symptoms coupled to negative blood cultures for two consecutive times. Cancer-directed therapy response of hematologic malignancies (complete response, partial response, or disease progression) was assessed using established clinical guidelines [16][17][18].

Statistical analysis
We performed descriptive statistics to present the characteristics of the patients. Age had a non-normally distribution based on the visual assessment of histograms and the Shapiro-Wilk test (p = 0.003). Therefore, we present the age at diagnosis using the median and the range. Categorical variables are shown as frequency and percentages. Based on the low expected count, categorical variables were compared using Fisher's exact test and the age with the Mann-Whitney U test.
The influence of the type of bacteremia (CAB and HAB) was assessed using a multivariate Cox regression analysis and the results are reported with Hazard Ratios (HRs). The model includes variables that were associated with mortality in previous studies [9,11], such as age and clinical characteristics, and factors that we considered relevant in the clinical course of bacteremia in patients with hematologic malignancies. We evaluate the effect of shock at presentation in the multivariate model for overall survival; however, the proportional hazard assumption was violated (global test: p = 0.021). Because the database has few observations, we could not perform a stratified analysis based on the presence of shock at diagnosis. The quick Sequential Organ Failure Assessment (qSOFA) score was used as a surrogate to adjust for the severity of clinical presentation. Our final model includes age group (adult vs. pediatric), comorbidities (yes vs. no), qSOFA score (�2 vs. 0-1), type of bacteremia (CAB vs. HAB). This model did not violate the hazard assumption based on the visual assessment of the Schoenfeld residuals and the global test (p = 0.350).
We estimated the survival probabilities in all patients and by age group stratification using the Kaplan-Meier method. Survival curves between groups (HAB vs. CAB) were compared with the Log-rank test. Although we did not find an interaction between age group and type of bacteremia on the mortality outcome (p = 0.428), we stratified based on age group because of the different clinical presentation and disease outcome that is typically found between pediatric and adult patients. Statistical significance was considered with a p-value of less than 0.05. We used the R software for analysis.

Ethical statement
The Institutional Review Board (IRB) of the INEN approved the development of this study with the code 'INEN 18-25'. Because of the retrospective design, the IRB waived the need for informed consent. We censored the patients' personal information during the data extraction from the medical records by assigning them numbers for identification. The data extraction was between January 2018 -April 2018.

Clinical features in the overall population
A total of 37 patients with malignant hematologic diseases and A. sobria isolation from the bloodstream were included. Table 1 summarizes the clinical features of these patients and according to the type of bacteremia. The median age at the time of bacteremia diagnosis was 24 years (range: 2-74 years), with 22 (59%) patients being adults (�20 years old). Most patients were diagnosed with acute lymphoblastic leukemia (n = 19/37, 51%) and acute myeloblastic leukemia (n = 10/37, 27%). Abdominal pain (41%) was the most common presenting symptom, followed by diarrhea (30%) and pain in the lower limbs (22%). A total of 82% (n = 27/33) patients presented with fever (temperature of >100.4˚F or 38˚C).
The source of bacteremia was identified in 30 (81%) patients, of whom 77% (n = 23/30) had the gastrointestinal tract as the primary source. Eleven (30%) patients presented with septic shock at diagnosis. Grade 3 and 4 leukopenia, thrombocytopenia, and anemia were found in 78%, 78%, and 35%, respectively. All patients received empiric antimicrobial therapy at the time of diagnosis, with the antimicrobial regimen modified according to the results of the antimicrobial susceptibility testing. Meropenem was the most common empiric antibiotic used (n = 24/37, 65%). Regarding the treatment response of hematologic malignancies, most patients developed bacteremia while on complete remission (n = 20/37, 54%) as opposed to patients in partial response (n = 3/37, 8%) or in disease progression (n = 14/37, 38%).

Survival outcomes
The overall case-fatality rate was 27% (n = 10). Subjects with CAB had a higher fatality rate at 30-days compared to HAB (n = 8/14, 57% versus n = 2/23, 9%, respectively; p = 0.002) ( Table 1). When stratified by age group, adults had a higher fatality rate at 30-days compared to pediatric patients, without statistical significance (36% versus 13%, respectively; p = 0.153) ( Table 2). To explore the effect of A. sobria bacteremia on survival, we plotted Kaplan-Meier curves according to the type of bacteremia in the total population (Fig 1A), pediatric (Fig 1B), and adult (Fig 1C) patients. In all and adult patients, individuals with CAB had worse overall survival at 30 days (43% versus 91%, p = 0.006 and 30% versus 92%, p = 0.002, respectively). Most deaths occurred during the first ten days after isolation of A. sobria in all (n = 7/8, 88% versus n = 0/2, 0%; p<0.001) and adult patients (n = 6/7, 86% versus n = 0/1, 0%, p = 0.003). In contrast, there were two deaths in the pediatric group, one for each type of bacteremia (CAB and HAB) without a statistical difference in survival rates (91% versus 75%, p = 0.390). The only death in the CAB pediatric group occurred within the first ten days after bacteria isolation. In the univariate analysis, older age (� 65-years, p = 0.010), qSOFA score �2 (p = 0.025), and CAB (p = 0.004) were associated with increased mortality (Table 3). In the multivariate analysis, a qSOFA score of �2 and CAB were predictors of mortality. Table 4 shows the antimicrobial susceptibility testing results. Overall, A. sobria was susceptible to most of the antibiotics used; patterns of resistance were found in patients treated with ciprofloxacin in both the adult (4/22, 18%) and pediatric (1/15, 7%) population.

Discussion
Bacteremia caused by the Aeromonas species is associated with high mortality rates. However, data regarding the outcomes in patients with malignant hematologic diseases is limited. This study reports the influence on survival of the type of bacteremia (CAB versus HAB) by A. sobria on patients with malignant hematologic diseases and their clinical course. By collecting data over 17 years, we observed that patients with CAB have an apparent worse survival compared to patients with HAB. Furthermore, we found distinct clinical outcomes on patients with A. sobria bacteremia based on clinical presentation. Abdominal pain was the most common presenting symptom of A. sobria bacteremia in this study, with the gastrointestinal tract as the most common primary source of sepsis. This goes in line with previous reports that describe abdominal pain as the most common presenting symptom [13]. Other common symptoms include diarrhea and fever [8,9,11,14,[19][20][21], which were also presented in our population.
Our findings add to the literature the 30-day overall survival rates in patients with hematological malignancies and bacteremia with A. sobria. In contrast to previous reports, we included pediatric and adult patients and analyzed them separately given the difference in the incidence of hematologic malignancies seen according to age group (i.e., acute lymphoblastic leukemia is more common in pediatric than adult patients), and its distinctive clinical course and outcome (i.e., pediatric acute leukemia is often associated to better outcomes compared to acute leukemia during adulthood). Case-fatality rates were higher in adult patients compared to pediatric patients. As opposed to pediatric patients, adults had a higher percentage of preexisting comorbidities (31.8% versus 6.7%; p = 0.108); this may have influenced the intensity of cancer-directed therapy the adults received, consequently increasing the fatality rates seen in this age group, an observation that has been reported on a previous study [13]. However, in the multivariate analysis, age could not be established as an independent prognostic factor for mortality.
Our study found CAB as a negative prognostic factor for survival in both the univariate and multivariate analysis. This finding differs from previous reports that described either HAB as a risk factor for mortality, or no significant differences in mortality between both types of bacteremia [8,13]. Community patients presented with higher percentages of septic shock and qSOFA scores in our cohort, implying a higher disease severity. Although we adjusted for qSOFA in the multivariate model, CAB remains a prognostic factor for mortality. We highlight some explanations for this outcome. First, patients with CAB had a higher median age at diagnosis compared to patients with HAB. Second, community patients may have experienced a delay in bacteremia diagnosis. Hospitalized patients are daily assessed through quantification of vital signs and periodic blood work, which increases the likelihood of early detection and subsequent treatment of an evolving infection. In contrast, these procedures are less consistent in the outpatient setting, increasing the disease severity at presentation. Third, patients in the CAB cohort could have had worse performance status or comorbidities with a worse prognosis than patients in the HAB subgroup. These factors might have increased the mortality in the CAB cohort; however, we lack data to corroborate these explanations.
Prior studies have described the clinical course and the associated factors for mortality in patients with bacteremia by A. sobria. However, we attempted to find the influence of the type of bacteremia in our patient population using conventional methods for survival analysis. One small report drew inference using a parametric statistical test, and others used regression techniques to identify whether certain factors increase the odds of mortality [9,11]. We opted for a more tailored approach by using non-parametric tests (Fisher's exact and Mann-Whitney U test) to make inferences on the relatively small count and distribution of our population. Furthermore, we used time-to-event analysis (Kaplan-Meier and Cox regression) to identify the survival differences and prognosis between both types of bacteremia.
Most of the strains in this study were resistant to ciprofloxacin, which correlates with previous findings [8,13]. It is important to mention that nearly all the tested isolates were susceptible to antibiotics in the pediatric population, most of whom had HAB. Therefore, we can hypothesize that bacteremia caused by resistant strains of A. sobria happens in patients infected in the community setting, most of whom are adults. All community patients in this study received antimicrobial prophylaxis with levofloxacin upon discharge. Thus, it is unclear to what extent this approach may have contributed to the observed pattern of resistance and survival outcomes.
Our study has some limitations. First, this is a single-center study, hence, our results should be taken with caution. Second, the results of the antimicrobial susceptibility testing in this study were heterogeneous, due to the lack of a standardized protocol regarding which drugs should be tested when A. sobria is isolated from the bloodstream. However, this report is one of the largest cohorts of patients with A. sobria infection and hematologic malignancies. A. sobria bacteremia in these patients is a rare condition that limits the gathering of large sample size and statistical analysis. Third, we were unable to include septic shock in our model because of the violation of the hazard assumption and we could not perform a stratify analysis by septic shock status at presentation given the low count between patients with HAB or CAB. As a result, we used qSOFA as a proxy for severity at presentation. Despite these shortcomings, our chosen Cox regression model did not violate the proportional hazard assumption, making our results reliable. Moreover, we were able to provide survival probabilities and case-fatality rates, in contrast to a prior study that only reported death from any cause [13]. Therefore, our approach gives a more extensive analysis of the risk of mortality and survival prognosis from A. sobria.
In conclusion, CAB is associated with higher mortality risk and worse OS in individuals with hematologic malignancies and A. sobria bacteremia. These findings suggest that individuals with CAB present with worse disease severity. Our results should aid clinicians to determine the survival prognosis in this population.