The Burden of Hyperbilirubinemia in Critically Ill Patients with Hematological Malignancies: Post-hoc Analysis of a Prospective Multicenter Multinational Study

Background The evidence on the clinical significance of hyperbilirubinemia (HB) in critically ill patients with hematological malignancies is scarce. We therefore studied its burden in a 2010-2011 Franco-Belgian multicenter prospective study designed to evaluate the prognosis of these patients. The cohort comprised 893 patients from 17 centers, 61% men, with a median (interquartile range) age of 60 (49 – 70) years, and preferentially with underlying non-Hodgkin lymphoma (32%) or acute myeloid leukemia (27%). HB was defined as a total serum bilirubin ≥ 33 µmol/L at intensive care unit (ICU) admission. Our main goal was to evaluate the relationship between HB and outcome of critically ill hematological patients. Causes and management of HB in the ICU were analyzed as secondary end points.

systematically explored and treated.
Diagnosis of liver injury in critically ill hematological patients is challenging as clinical, biological and radiological findings are, in general, nonspecific, and because of the absence of systematic guidelines, including on the use of liver biopsies. Total serum bilirubin level, a biomarker of liver function, has been incorporated in several organ dysfunction scores (20,21) to grade liver injury in intensive care patients.
Hyperbilirubinemia (HB), defined as an increased total serum bilirubin level ≥ 68 µmol/L, has been associated with a doubled risk of mortality in a large cohort allo-HSCT patients (22,23). In pediatric patients, a total serum bilirubin ≥ 33 µmol/L one month after allo-HSCT has been associated with higher non-relapse mortality (24). HB seems more associated with mortality than hepatocellular injury, defined as elevated aminotransferase levels without HB in hematological patients (25).
Overall, the level of evidence, including incidence, risks, causes and management, on HB is scarce in critically ill hematological patients. In order to decipher its burden, we analyzed a large prospective multicenter sample of critically ill hematological patients and report on the management of HB by caring physicians.

Study population
We performed a post hoc analysis of a Franco-Belgian multicenter prospective study assessing the prognosis of patients with hematological malignancies admitted in 17 ICU between January 2010 and May 2011 (26). Among the 1011 patients enrolled in the original study, only patients with total serum bilirubin measurement at admission were included. Two groups were created according to liver SOFA score (21) at ICU admission: one with HB defined as a liver SOFA score >1 (total bilirubinemia ≥ 33 μmol/L) and one without HB defined as a liver SOFA score ≤ 1 (total bilirubinemia < 33 μmol/L) (24). The appropriate ethics committees approved the study (26) and all patients or relatives were informed and consented to participate in the study.

Data collected in the prospective cohort
All patients had a diagnosis of initial/relapsed hematological malignancy within 5 years before ICU admission. The performans status (27) and the Charlson comorbidity index (28) were determined at ICU admission. History of mild, moderate or severe liver diseases was defined by a hepatic Charlson comorbidity index ≥ 1 (severe = cirrhosis and portal hypertension with variceal bleeding history, moderate = cirrhosis and portal hypertension but no variceal bleeding history, mild = chronic hepatitis or cirrhosis without portal hypertension). Underlying malignancies, disease status at ICU admission, radiotherapy and chemotherapy received in the last month and HSCT were recorded. In allo-HSCT patients, the type of donor, type and intensity of conditioning regimen and GvHD prophylaxis and treatment were collected. Chemotherapy, systemic corticosteroids, hematopoietic growth factors and immunosuppressive drugs prescribed by the hematologist in charge of each patient and administered during ICU stay were recorded.
Neutropenia was defined as a neutrophil count of less than 0.5 G/L and presence of haemophagocytosis features on myelogram was informed. Data from clinical examination at ICU admission were reported such as digestive, neurological, cutaneous, renal and hematological symptoms. Organ dysfunctions, presence of sepsis and life sustaining therapies at ICU admission and during the ICU stay were also recorded.

Data collection
Medical hospital records of patients with HB at ICU admission were selected and reviewed by an expert panel including an intensivist (MB), a hematologist (CS) and hepatologists (CM and VM). The five following questions were assessed: (1) was HB at ICU admission mentioned by the attending physician in the medical records? (2) was a diagnostic strategy carried out to explore HB, including a liver biopsy? (3) which cause of HB was suspected by the attending physician or confirmed by hepatic histology when liver biopsy was performed just before ICU admission or during ICU stay? (4) which specific treatments were initiated? (5) which cause of HB was considered a posteriori by the expert panel?

Statistical analysis
Results were reported as median and interquartile ranges or counts and proportions (%).
Qualitative variables were compared using the chi-square test or Fisher's exact test as appropriate and continuous variables using the Mann-Whitney test. Multivariate analyses were performed using logistic regressions. Variables yielding P-values < 0.20 in the univariate analyses or considered clinically relevant were entered in backward stepwise logistic regression models. Survival curves were constructed according to the Kaplan-Meier method and compared with the log-rank test. P-values <0.05 were considered statistically significant. Statistical analyses were done using SPSS software (version 20).

Characteristics of patients
Among the 1011 patients included in the TRIALOH study (26), 118 (11.7%) were excluded because total serum bilirubin level was not measured at ICU admission ( Figure 1). Among the 893 remaining patients, 185 (20.7%) patients had HB, defined as a total bilirubinemia ≥ 33 µmol/L at admission, which corresponds to a SOFA score > 1.
Characteristics of the cohort by HB are presented in Table 1. Median (interquartile range) age was 60 (49 -70) years and there was a majority (61%) of men. Less than 5% of patients had a history of liver diseases. Non-Hodgkin lymphoma was the most frequent underlying hematological malignancy (31.5%), followed by acute myeloid leukemia (27.1%) and myeloma (11.9%). There were 133 (14.7%) allo-HSCT patients, including 6.8% with a previous myeloablative conditioning regimen. About two-third of patients (64.9%) had been treated with antibiotics before ICU transfer. Acute respiratory failure concerned almost two-third (62.4%) of patients at ICU admission, followed by cardiovascular failure (43.1%), acute kidney injury (31%) and acute liver failure (8.8%). A total of 497 (55.7%) patients had at least two organ failures at ICU admission. Digestive, hematological (such as lymphadenopathy, mucositis, hepatomegaly, splenomegaly, purpura, and intraoral hemorrhage) and/or neurological symptoms were found on clinical examination in onequarter of patients. Renal symptoms (such as decreased urine output, anuria, hematuria, and pyuria) were present in one-third of patients (32.2%). Nearly one-third of patients (30.8%) were neutropenic, sepsis was diagnosed for 578 (64.7%) patients, and a high majority of patient received antimicrobial therapy at ICU admission and during ICU stay.
About 10% of patients were receiving chemotherapy at ICU admission. Few patients had positive Cytomegalovirus (CMV) or Herpes simplex virus (HSV) nucleic acid testing (NAT).
Haemophagocytosis features were described on myelograms in 16 patients (1.8%). Half of patients required invasive mechanical ventilation and/or vasoactive drugs during ICU stay, and one-third needed renal replacement therapy.

Risk factors for HB in critically ill hematological patients at ICU admission
Patients with HB were younger than patients without HB; had more frequently hepatic comorbidities; had received more anthracyclins, aracytin, busulfan but not cyclophosphamide, thalidomide and fludarabine; had undergone more frequently allo-HSCT; were treated with cyclosporine, mycophenolate mofetil or antimicrobial therapy; presented with more organ failures and digestive clinical symptoms; had higher serum concentrations of hepatic enzymes, a longer prothrombin time, lower platelet and higher lactate levels; and required more life sustaining therapies during ICU stay. Factors independently associated with HB at ICU admission are presented in Table 2. The presence at ICU admission of cyclosporin treatment, antimicrobial therapy, digestive symptoms, ascites, history of liver diseases, neutropenia and increased serum creatinine level, were associated with an increased risk of HB. Myeloma and its treatments were not associated with a higher risk of HB.

Discussion
In a large multicentric cohort, HB, defined as a total serum bilirubin level ≥ 33 µmol/L (twice the upper limit of normal range), concerned 21% of critically ill patients with hematological malignancy at ICU admission and was associated with outcome. The first underlying causes of HB were cancer progression, hypoxic hepatitis, cancer treatment toxicities, and sepsis. Liver workups for HB, including liver biopsies, led to high rates of treatment modifications, including cancer treatment, which was associated with reduced mortality. In spite of that, HB was overlooked by the intensive care teams in more than half of patients.
This is the largest study reporting on the burden of HB in critically ill hematological patients. Our findings demonstrate that HB is closely related to the prognosis hematological patients, as for other critical care patients, including severely burned patients (32), patients with acute respiratory distress syndrome (ARDS) (33) and trauma (34). An identical serum bilirubin threshold, twice the upper limit of normal range, has been associated with an increased risk of mortality in patients with severe sepsis and septic shock (35).
There is, to date, no robust literature on the underlying causes of HB in critically ill hematological patients. In this series, almost one third of treatable liver diseases contributed to HB. In the rare subgroup of patients who underwent a throughout liver workup, cancer, cancer drug toxicities, heart failures and infections led frequently to therapeutic adjustments, including antimicrobial treatment modifications, chemotherapy initiations, cancer drug modifications, hemodynamic management, and defibrotide in patients with SOS.
We were surprised by the rather high number of patients in whom HB was overlooked in this cohort: total serum bilirubin was not measured in 12% of patients at ICU admission, was unrecognized in half of HB patients, and was poorly investigated (63%), diagnosed (59%) and treated (40%) when it was taken into account by the intensive care team. HB is also frequently ignored in other immunocompetent critically ill patients, although it is closely associated with outcome (36,37). Our findings support the idea that HB is not a satellite of the multi-organ failure syndrome and that it should be actively treated as other organ failures. Echocardiography, bacterial, virologic and fungal workups, liver ultrasonography, CT-scans, and bone marrow aspirate or biopsy should be performed immediately at bedside of HB patients. We acknowledge the absence of guideline to support empirical treatments or decide when to perform liver biopsies in hematological HB patients. The few numbers of liver biopsies performed in the cohort could have modified treatments. Percutaneous liver biopsy is contraindicated in the context of severe thrombopenia and was associated with mortality in this series. Transjugular liver biopsy is safe (38,39) and more than 20 years ago, Shulman at al. demonstrated that it significantly improved the management of allo-HSHC patients (40).
The presence at ICU admission of cyclosporin treatment, antimicrobial therapy, digestive symptoms, ascites, history of liver diseases, increased serum creatinine level, neutropenia were associated with HB. Cyclosporin treatment was strongly associated with HB, which confirmed the relationship between allo-HSCT and outcome of hematological patients (41).
Other classical factors were neutropenia, favoring sepsis and requiring antimicrobial therapy. In hematological patients, HB and acute kidney injury (AKI) seem to share similar risk factors such as sepsis, antimicrobials and cyclosporin nephrotoxicity, tumor lysis syndrome (42), except chemotherapy which did not impact HB in our cohort. Ascites can be explained in these patients by SOS (17), engraftment syndrome in allo-HSCT patients (43), capillary leak syndrome (44)  Availability of data and materials: The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.
Competing interests: The authors declare that they have no competing interests.
Funding: The authors declare that they have no funding for this research.