Hemophagocytic Lymphohistiocytosis in Critically Ill Patients

Abstract

Background: 

Hemophagocytic lymphohistiocytosis (HLH), an uncontrolled overactivation of the immune system, is well characterized in pediatric patients, yet, much less is known about this life-threatening condition in adult patients. As HLH is often complicated by organ failure, patients will require admission to the intensive care unit for organ support therapy. However, recognition of HLH patients in the intensive care unit (ICU) is challenged by the clinical overlap with sepsis. Here, we analyze HLH patients to better understand its clinical presentation, diagnosis, and treatment.

Methods: 

For the purpose of this retrospective observational study, we searched for suspected and diagnosed adult HLH of all patients admitted to at least one adult surgical, anesthesiological or medical ICU between January 2006 and August 2018 at the university hospital Charité — Universitätsmedizin Berlin. All cases were reviewed by two HLH experts, who confirmed or declined the diagnosis.

Results: 

Of 6,340 ICU patients with ferritin measurement, 40 suffered from HLH (0.63%). Of these, in-hospital mortality was 60.0% over all cases, which was highest in malignancy-associated HLH (71.4%). Infections were identified as most common triggers (42.5%). A variety of 19 different treatment strategies were applied. Non-survivors showed higher ferritin at diagnosis compared with survivors (P = 0.021), which was also seen in multivariable analyses. A minimum ferritin of 4083 μg/L after diagnosis was most predictive for 30-day mortality (AUC 0.888, 95% CI 0.771–1.000; sensitivity 93.8%, specificity 78.9%).

Conclusions: 

Mortality in adult HLH patients in the ICU is high, particularly in malignancy-associated HLH. Infections are the most frequent HLH triggers in critically ill patients. At present, there is no standardized treatment for HLH in adult patients available. Assessment of ferritin is valuable for diagnosis, prognosis, and treatment monitoring.

Trial registration: The study was registered with www.ClinicalTrials.gov (NCT02854943) on August 1, 2016.

INTRODUCTION

Hemophagocytic lymphohistiocytosis (HLH) is a hyperinflammatory syndrome caused by excessive cytokine release from dysregulated T cells leading to overactivation of macrophages with subsequent organ infiltration and life-threatening multiple organ dysfunction ⁽¹⁾. Mortality rates are high ranging from 43% in pediatric patients to 75% in adults and up to 90% in untreated patients ^(2–4). Although there is a substantial body of research in pediatric patients, where genetic defects are a frequent cause of HLH, much less is known about adult HLH, particularly in critically ill patients. For those, only few available studies, mostly small in sample size and retrospective in nature, systematically analyzed precipitating factors, specific treatment, and outcome ^(5–7). Clinical features include multiple organ involvement among which respiratory failure, coma, shock, and acute liver and kidney failure are the most common causes for intensive care unit (ICU) admission ^(5–7). Moreover, symptoms as fever, cytopenia, organomegaly, and hyperferritinemia, all of which are part of the HLH-diagnostic criteria, are frequently seen in the critically ill and are not specific for HLH ⁽⁸⁾. Due to this clinical overlap, particularly with sepsis, HLH remains a diagnostic challenge and is likely to be underrecognized in ICU patients ⁽⁹⁾. This is of particular concern, given that early identification of the trigger and prompt treatment initiation is critical for survival ⁽¹⁰⁾. Treatment of HLH consists of immunosuppression and cytotoxic agents to control the hyperinflammatory state and cell proliferation. Additionally, specific therapies targeted at the suspected trigger are required to eliminate the cause of uncontrolled inflammation ⁽¹¹⁾. Dexamethasone, etoposide, and cyclosporine A are considered standard HLH treatment according to the HLH-1994 protocol ⁽¹²⁾. In addition, cytokine adsorption and plasmapheresis have been applied in critically ill HLH patients to eliminate cytokines from the blood stream ^{(13, 14)}.

Additionally, supportive treatment is needed as these patients are at high risk of fatal organ failure and bleeding due to HLH-related liver and kidney failure as well as coagulopathy and cytopenias ⁽¹⁵⁾. Yet, data on both diagnosis and treatment of secondary HLH in adults is limited ^{(5, 6, 16)}. Treatment strategies rely on clinical experience reported in a small number of case series. Current guidelines for diagnostics and treatment in adults are based on the HLH-1994 pediatric protocols which have not been validated in an adult HLH cohort ⁽¹²⁾. In view of the high mortality associated with HLH in the critically ill ^{(5, 6, 9, 16)}, accurate and timely diagnosis is all the more important. However, diagnosis is hampered by a lack of clinical knowledge due to the absence of larger studies of critically ill HLH patients. Therefore, detection of HLH requires high vigilance of those taking care of HLH patients. If understanding of secondary HLH increases, improvement in its diagnosis and development of specific therapies might follow. The primary aim of this investigation was to evaluate in-hospital mortality among 40 cases of adult HLH patients admitted to ICUs of a university hospital. As secondary outcomes, we determined associations between maximum serum ferritin levels and 30-day mortality and prognostic performance of minimum ferritin levels after diagnosis to predict 30-day mortality. Additionally, we analyzed triggers and HLH-specific treatment.

METHODS

Patients

This retrospective observational study was conducted at the university hospital Charité — Universitätsmedizin Berlin. All clinical data including survival data and parameters required for HLH diagnosis according to HLH-2004 criteria and the HScore were retrieved from two electronic patient data management systems operated at the Charité — Universitätsmedizin Berlin (COPRA, Sasbachwalden, Germany and SAP, Walldorf, Germany). Medical records of all patients who were admitted to at least one adult surgical, anesthesiological, or medical ICU between January 2006 and August 2018 with at least one ICU ferritin value and hyperferritinemia of at least 500 μg/L were searched for clinically diagnosed or suspected HLH. In parallel, we searched for all adult ICU patients diagnosed with ICD 10 codes for HLH (D76.1, D76.2, and D76.3). Two HLH experts reviewed and retrospectively confirmed or rejected the diagnosis, which was based on HLH-2004 criteria, currently the gold standard for HLH diagnosis (Supplement Table 1, Supplemental Digital Content 1, https://links.lww.com/SHK/A953), and HScore ⁽¹⁹⁾ (Supplement Table 2, Supplemental Digital Content 2, https://links.lww.com/SHK/A954). In case of disagreement, a third HLH expert was consulted. Of note, this is a subgroup analysis of a cohort of 2623 ICU patients ⁽¹⁷⁾, which also includes seven previously undiagnosed HLH cases that have been retrospectively diagnosed and published by our research group ⁽⁹⁾. The study period was defined from admission to the ICU until hospital discharge, transfer, or death. We assumed the date of HLH diagnosis when it was diagnosed by the clinicians.

Statistical analysis

Continuous variables are summarized as medians with quartiles and categorical variables as frequencies and percentages. Characteristics of survivors and non-survivors and between the three different trigger groups were compared using non-parametric tests for independent groups. We conducted multivariable logistic regression analysis adjusting for age and maximum sequential organ failure assessment (SOFA) score to analyze the impact of ferritin at diagnosis on 30-day mortality. For the purpose of this analysis, ferritin levels were divided by 1,000 to maintain clinical meaningfulness (estimates correspond to increments of 1,000 μg/L). Receiver operating characteristics (ROC) analysis was performed to analyze best minimum ferritin levels after diagnosis to predict 30-day mortality. All tests should be understood as constituting exploratory data analysis, so that no adjustment for multiple testing was made. All numerical calculations were performed with IBM SPSS Statistics, Version 25, Copyright 1989, 2010 SPSS Inc. A P value < 0.05 was considered statistically significant.

Ethics

Ethics approval was obtained from the institutional review board (Ethikkommission der Charité — Universitätsmedizin Berlin, EA1/176/16). The study was registered with www.ClinicalTrials.gov (NCT02854943) on August 1, 2016.

RESULTS

Study population and characteristics

A total of 116,310 patients were admitted to the ICU within the study period of which 109,970 had no ferritin assessment. HLH incidence in 6,340 patients with ferritin assessment was 0.63%. Considering 2,623 patients with hyperferritinemia only, 40 patients (1.52%) developed HLH (Fig. 1). Twenty-four out of 40 HLH patients died during their ICU stay (60.0%). HLH was diagnosed before ICU admission in eight patients (20.0%). Thirty-two patients (80.0%) were diagnosed with HLH during their ICU stay. Main reasons for ICU admission were respiratory failure (45.0%) and sepsis (37.5%). During intensive care, sepsis or septic shock was diagnosed in 35 HLH patients (87.5%). Non-survivors were significantly older, required more vasopressors, mechanical ventilation, and dialysis, while showing higher maximum SOFA scores (Table 1). For the diagnosis of HLH, median 5 out of 8 HLH-2004 criteria were fulfilled, while median HScore was 224. Importantly, HScores during ICU stay exceeded 169 in all patients. Non-survivors showed significant higher aspartate aminotransferase (ASAT), higher rates of hemophagocytosis, and fulfilled more HLH-2004 criteria during their ICU stay (Table 2). In four out of the 16 survivors (25.0%), HLH recurred in two patients after 1 week (deceased), 2 months (survived), and repetitively in the same patient after 1.25 and 2.5 years (deceased). Genetic testing to rule out a genetic component was not performed in these patients. Parameters of each single patient are shown in Supplement Table 3, Supplemental Digital Content 3, https://links.lww.com/SHK/A955.

Fig. 1: Consort diagram.

Table 1: Patient characteristics and outcome parameters

Table 2: HLH-2004 criteria, HScore, and supportive markers

Triggers and treatment

Seventeen patients had infectious HLH triggers (42.5%). Overall, main triggers (Table 3) were lymphoma (20.0%), HIV (12.5%), and leukemia (10.0%). Mortality was highest in malignancy triggered HLH (71.4%), while it was lowest in autoimmune triggered HLH (44.4%). Bacterial sepsis was the trigger in three patients (7.5%). Twenty-eight patients received HLH-specific immunosuppressive treatment (70.0%) without significant delay after diagnosis (median: the day of diagnosis). Of these, nine patients were treated with corticosteroids alone, 10 patients with additional etoposide or cyclosporine. In three patients, anakinra was applied instead of etoposide and cyclosporine. Six patients received other treatments (Supplement Table 1, Supplemental Digital Content 1, https://links.lww.com/SHK/A953). Overall, 19 different treatment strategies were used (Tables 3 and 4). Treatment strategies between survivors and non-survivors are shown in Table 4. Of note, seven patients (17.5%) survived without HLH-specific treatment. HLH in these patients was triggered by infections (4×), autoimmune disease (1×), and malignancy (2×), while one of the latter patients was at low-dose hydrocortisone as adjuvant sepsis therapy.

Table 3: Patient characteristics, treatments, and outcome parameters between trigger groups

Table 4: HLH-specific treatment strategies between survivors and non-survivors

Associated factors with mortality

Thirty-day mortality rate was 50%, whereas in-hospital mortality rate amounted to 60.0%. Non-survivors had higher ferritin at diagnosis and higher minimum ferritin after diagnosis (Table 2; Fig. 2). In multivariable analysis, a trend was seen for ferritin at diagnosis (OR 1.036 (95% CI 0.996–1.078); P = 0.077) while maximum SOFA score (OR 1.395 (95% CI 1.103–1.763; P = 0.005)) was significantly associated with 30-day mortality. ROC analysis identified a minimum ferritin after diagnosis of 4083 μg/L as predictive for 30-day mortality with 93.8% sensitivity and 78.9% specificity (AUC 0.888, 95% CI 0.771–1.000).

Fig. 2: Ferritin at diagnosis, minimum after diagnosis, and last measurement between survivors and non-survivors.

DISCUSSION

In this mixed cohort of critically ill patients with ferritin measurement, HLH incidence was 0.63% and 1.52% when only patients with hyperferritinemia were considered. In-hospital mortality among HLH patients was high with 60.0%. Relapse of HLH occurred in 25.0% of patients who survived initial remission. Infections were the most common triggers. Mortality was highest in malignancy-related HLH with 71.4%. Treatment strategies varied greatly depending on the suspected trigger. Ferritin at diagnosis was higher in non-survivors and was also associated with 30-day mortality after adjustment for age and maximum SOFA score. A minimum ferritin after diagnosis of 4,083 μg/L was most predictive for 30-day mortality. Among 6,340 ICU patients with ferritin measured, we found 40 (0.63%) patients with HLH. Previous reports on HLH in critically ill adults are scarce and often restricted to medical ICU patients. In the study of Buyse et al. ⁽⁶⁾, 1.49% of 5,027 patients admitted to an ICU fulfilled diagnostic HLH criteria. Meena et al. ⁽¹⁸⁾ detected HLH in 10 patients (2.2%) in their cohort of 445 medical ICU patients with ferritin assessment. Although only 6,340 of 116,310 patients (5.5%) of our cohort had a ferritin assessment and only those with hyperferritinemia (n = 2,623; 2.3%) were included, our study is currently the largest report of HLH incidence in a mixed ICU cohort. Importantly, HLH was clinically diagnosed at physicians’ discretion, which was retrospectively confirmed or declined by expert consensus. Even though 11 patients did not meet a minimum of five HLH criteria, HScores beyond 169 were detected in all patients.

Of 40 HLH patients, 24 (60.0%) died before hospital discharge. This is in line with numbers reported in previous studies. In their reports of critically ill HLH patients, Buyse et al. ⁽⁶⁾ and Barba et al. ⁽⁵⁾ found in-hospital mortality rates of 51.8% and 68.0%, respectively. Multisystem organ failure is one of the common features of HLH in the ICU ⁽⁵⁾. A median SOFA score of 9 at ICU admission in our population is suggestive for multisystem organ failure. In fact, the use of vasopressors, transfused blood products, mechanical ventilation, and hemodialysis in the majority of patients reflects the need for organ support therapy. Among the common causes for initiation of organ supportive care in adult HLH are respiratory failure (30%–36%), shock (18%–29%), coma (21%), renal (16%), and liver failure (7%) ^{(5, 6)}. Of note, eight out of 40 patients had been diagnosed with HLH prior to ICU admission. Unexplained organ failure in the presence of fever, cytopenias, organomegaly, and unresponsiveness to anti-infective treatment should prompt intensivists to consider HLH as a differential diagnosis, both at admission and during critical care. The fact that ASAT at diagnosis was markedly higher in non-survivors possibly implies more severe liver failure, respectively organ dysfunctions, ultimately leading to death. As one of the diagnostic criteria included in the HScore, elevated ASAT is a common feature of HLH while normal transaminases render a diagnosis of HLH less likely ^{(15, 19)}. In HLH patients, time to recognition and initiation of adequate therapy with elimination of the trigger is the key to improved outcome ⁽²⁰⁾. In addition, hemophagocytosis on bone marrow aspirates was also more frequent in non-survivors. Yet, hemophagocytosis is not specific for HLH as it is found in as many as 64% of ICU patients with sepsis and thrombocytopenia and in 65% of non-survivors in the ICU ^{(21, 22)}. Given that hemophagocytosis is rarely present on initial bone marrow sampling but may occur with progression of the disease, its performance as a diagnostic marker might be limited but could be indicative of unfavorable outcome.

Overall, infections account for the majority of precipitating factors triggering HLH, both in our cohort (42.5%) and in the general population ^{(23, 24)}. In critically ill patients, however, malignancies have been reported to outnumber infections as the underlying cause ^{(5, 6)}. Differentiating between distinct triggers in our cohort, lymphoma (20.0%), HIV (12.5%), and leukemia (10.0%) were the main precipitating factors. Mortality was highest in malignancy-triggered HLH, which is in line with the findings reported by Barba et al. ⁽⁵⁾, whereas presence of B cell lymphoma was associated with better survival in the cohort of Buyse et al. ⁽⁶⁾ likely due to early targeted chemotherapy. Prompt initiation of specific therapy to eliminate the trigger is the key to ensure survival. However, determining the underlying trigger is not always successful. In the populations described by Buyse et al. and Barba et al., in 7.1% and 26%, respectively, no precipitating factor could be identified ^{(5, 6)}. Importantly, elimination of the trigger is the only causal and sustainable therapy preventing recurrence of HLH. If the trigger persists, relapse may occur which might probably have been the cause for HLH recurrence in 25.0% of the initial survivors in our cohort.

In total, 19 different treatment strategies have been applied, most of which consisted of a combination of corticosteroids and another immunosuppressive drug. This large variability in therapeutic approaches reflects the dilemma of appropriate treatment of this rare condition. According to the most recent recommendations by an interdisciplinary working group in collaboration with the Histiocyte Society ⁽¹¹⁾, a tailored approach based on the etiology is mandatory to effective treatment of secondary HLH. The current strategies are standardized with immunosuppressive agents forming the basis of therapy with few exceptions in infection-related HLH where anti-infective treatment may be sufficient. Additional components to control hyperinflammation include polyvalent immunoglobulins (IVIG) and etoposide in case of severe presentation ⁽¹¹⁾. To account for the challenge of individualized therapy of a rare and variable condition, a quality improvement project initiated at the Johns Hopkins Hospital provides advice to physicians taking care of patients with suspected HLH ⁽²⁵⁾. In consequence, diagnostic procedures improved with a reduction in erroneous testing and timely trigger identification.

Regarding treatment strategies in our cohort, we have seen higher mortality rates in patients treated with etoposide compared with those who received anakinra (anti-IL-1 antibody) and tocilizumab (anti-IL-6 antibody). However, whether these findings in fact result from an improved treatment effect related to anakinra and tocilizumab cannot be ascertained due to small sample size. In addition, more severe presentation of HLH could have led to treatment with etoposide and thus higher mortality due to higher rate of organ dysfunctions. Yet, recent studies demonstrated beneficial effects associated with the use of biologicals. Tocilizumab leads to inflammatory control both in autoimmune-triggered HLH and in patients suffering from HLH after Chimeric Antigen Receptor (CAR) T-cell therapy ^{(11, 26)}. Likewise, anakinra might be considered in cases where autoimmune diseases are the underlying condition and even in sepsis patients with clinical feature of HLH, survival was improved with IL-1-blocking therapy ^(26–29). Of note, 17.5% of patients survived without specific immunosuppressive treatment. Remission in these patients might be due to the successful elimination of the trigger as seen in secondary HLH when infections are underlying triggers ^{(11, 30)}. Low-dose hydrocortisone as adjuvant sepsis therapy could also have played a role in these patients.

Ferritin at diagnosis was associated with 30-day mortality independent of age and maximum SOFA score suggesting that presence of a higher grade of inflammation contributes to higher severity of organ failure and subsequently increases mortality. This might have implications for treatment management and planning. Our results suggest that immunosuppressive treatment should be maintained and in case of unresponsiveness escalation needs to be considered, until ferritin levels are below this threshold. Moreover, this threshold might serve as a marker of treatment response indicating when reduction of immunosuppressants is appropriate. This would align with the findings reported by Kyriazopoulou et al. ⁽³¹⁾. The authors demonstrated in their cohort of patients with macrophage activation-like syndrome that a reduction of ferritin of less than 15% after diagnosis is related to unfavorable outcome after 10 days. However, these results need confirmation in larger studies. Recommendations of the Histiocyte Society state that duration and intensity of immunosuppression depend on HLH severity ⁽¹¹⁾.

Several limitations of our study deserve consideration. First, this is a retrospective investigation. Data availability was thus limited by documentation during patients’ clinical course. Second, excluding patients with ferritin levels < 500 μg/L bears the risk of missing HLH cases without hyperferritinemia. However, this is a supposedly rare occurrence ⁽³⁾. We therefore assume the number of missed HLH cases to be negligible.

CONCLUSIONS

HLH in critically ill patients is a rare though life-threatening condition. Mortality rates are high even with treatment, particularly in malignancy-associated HLH. Infections and malignancies are the main precipitating factors. At present, there is no evidence available for a single treatment strategy but rather tailored approaches are recommended based on the underlying trigger. Treatment largely relies on a combination of immunosuppressive drugs and additionally measures aiming at the elimination of the trigger. Ferritin should be considered as one of the key diagnostic and prognostic markers. Suspected HLH should prompt physicians to initiate further HLH diagnostics. Moreover, reassessments of ferritin are valuable for treatment monitoring and prediction of prognosis.