Comparison between clinical features and prognosis of malignancy- and non-malignancy–associated pediatric hemophagocytic lymphohistiocytosis

Background The differences between the clinical characteristics and survival time in malignancy- and non-malignancy–associated secondary hemophagocytic lymphohistiocytosis (HLH) are unclear. Here, we describe the clinical characteristics, prognostic factors, and survival outcomes of malignancy-associated HLH compared to that of non-malignancy–associated HLH. Methods We retrospectively analyzed 91 pediatric patients with HLH (age < 14 years) at the Affiliated Hospital of Qingdao University Pediatric Department between January 2005 and October 2016. The patients were divided into the malignancy-associated group (n = 22) and non-malignancy–associated group (n = 69, also considered the control group). The clinical features were compared using the Mann–Whitney U and χ2 tests. The overall survival time was compared using log rank and Mann–Whitney U tests. Results Hemoglobin (HGB; p = 0.004), alanine aminotransferase (ALT; p = 0.002), and aspartate aminotransferase (AST; p = 0.001) levels in the malignancy-associated group differed from that in the non-malignancy–associated group. The mean survival times were 26.9 ± 3.82 months (malignancy-associated HLH) and 35.03 ± 2.19 months (non-malignancy–associated HLH). The overall survival time between the two groups was not statistically significantly different (p = 0.055). Univariate analysis showed that disseminated intravascular coagulation (DIC) score > 5 (p = 0.001), albumin < 25 g/L (p = 0.000), HGB < 60 g/L (p = 0.001), and platelet count (PLT) < 30 × 109/L (p = 0.042) correlated with prognosis. Multivariate Cox analysis showed that albumin < 25 g/L (p = 0.017), HGB < 60 g/L (p = 0.027), and bone marrow hemophagocytosis (p = 0.034) correlated with worse prognosis. Conclusions Patients with non-malignancy–associated HLH do not have better survival, although their prognosis is relatively better in clinical practice. A higher DIC score at diagnosis and lower albumin, HGB, and PLT levels are negative prognostic factors in malignancy-associated HLH.


Background
Hemophagocytic lymphohistiocytosis (HLH) is characterized by aberrant activation and proliferation of polyclonal CD8 + T lymphocytes and macrophages that infiltrate multiple organs and overproduce inflammatory cytokines [1]. HLH has been characterized as primary HLH or secondary HLH (SHLH). Primary HLH is often associated with inherited gene defects.
SHLH generally occurs later in life than primary HLH. SHLH is an acquired phenomenon that develops in response to severe infections, autoimmune or rheumatologic disorders, drugs, and in association with concurrent malignancies [2]. Malignancy-associated HLH is most commonly seen in acute leukemia in children. HLH can occur as initial HLH or concomitantly with malignant disease. The mechanism of malignancy-associated hemophagocytosis is not well understood, and may be associated with cytokine secretion (including interferon-γ and interleukin-6), persistent antigen stimulation by malignant cells, chemotherapyinduced immunosuppression, hematopoietic stem cell transplantation (HSCT), or infection [3].
The difference between malignancy-and nonmalignancy-associated (i.e., causes other than tumor) HLH remains unclear. We retrospectively reviewed patients with malignancy-and non-malignancy-associated HLH at our institution and compared the clinical characteristics, treatment response, overall survival time, and prognostic factors of the two groups. In this study, we also share our limited experiences with allogeneic HSCT in malignancy-associated HLH. We aimed to distinguish between malignancy-associated and non-malignancy-associated HLH, compare their clinical features and prognostic significance, and analyze the difference between the two conditions.

Participants
The clinical data of 91 patients aged < 14 years who had been diagnosed with HLH from January 2005 to October 2016 at the Children Medical Center of the Affiliated Hospital of Qingdao University were included and retrospectively reviewed after we had received University of Qingdao Ethics Committee approval. A retrospective historical cohort study was carried out. Primary HLH was excluded.

Statistical analysis
The clinical features of the two groups were compared using Mann-Whitney U tests for skewed continuous data, and with χ 2 tests for categorical data. All results are presented as the median and range (min-max) as indicated. Overall survival time, defined as the time from HLH diagnosis to death from any cause, was estimated with the Kaplan-Meier method. The overall survival time of the two groups was compared using log rank and Mann-Whitney U tests. The relationship of prognostic factors associated with malignancy-associated HLH was evaluated using Pearson correlation for two normally distributed variables and with Spearman rank correlation for non-normally distributed variables. Prognostic factors associated with malignancy-associated HLH were evaluated using univariate and multivariate Cox proportional hazard models. Multiple group corrections were performed using the false discover rate. In all tests, p < 0.05 was considered significant. All statistical analyses were performed using SPSS version 13.

Clinical characteristics
There was no significant difference between age, sex, splenomegaly, ANC, PLT, presence of EBV infection, albumin, presence of jaundice, and presence of cholecystitis in the two groups (Table 1). Serum TG, ferritin, Fib, and LDH levels and presence of DIC were also not significantly different between the two groups. However, patients with nonmalignancy-associated HLH had significantly higher HGB (p = 0.004), ALT (p = 0.002), and AST (p = 0.001) levels than the patients with malignancy-associated HLH.
Thirteen patients (59.1%) were treated according to HLH-2004 as initial therapy for HLH, and four patients received chemotherapy as initial therapy. The overall

Outcome
The mean survival time in the malignancy-associated HLH and non-malignancy-associated HLH groups was 26.9 ± 3.82 and 35.03 ± 2.19 months, respectively. The difference in the overall survival time between the two groups was not statistically significant (p = 0.055) (Fig. 1).

Role of allogeneic HSCT in malignancy-associated HLH
In our cohort, only five patients (three patients with leukemia; two patients with lymphoma) received allogeneic HSCT. Four patients achieved complete remission after allogeneic HSCT. One patient with HD achieved partial remission, but died of HD relapse, complicated by infection.

Discussion
We performed a retrospective analysis of 22 children with malignancy-associated HLH and 69 children with nonmalignancy-associated HLH at a single institution. The pathogenesis of SHLH remains unclear. Delavigne et al. [6] proposed extended 18-point diagnostic criteria that are more easily and rapidly available in smaller institutions and primary care settings than the HLH-2004 variables. Non-malignancy-associated HLH is common in acute self-limited infectious mononucleosis (IM), rheumatic or autoimmune diseases, immunodeficiency diseases, and CAEV. EBV is the most frequent antigen activator of SHLH [7]. The pathological changes in IM and CAEV differ. In contrast to B cell infection in IM, CAEV features the proliferation and infection of polyclonal, predominantly non-CD8 + (CD4 + CD8 − and CD4 + CD8 + ) T cells, and CD16 + natural killer (NK) lymphocytes [8]. In CAEV, mortality generally results from the subsequent development of HLH and/or T/NK lymphoproliferative neoplasm [9]. The prognosis is poor once CAEV develops into HLH [9]. Chronic granulomatous disease is an inherited disorder of phagocyte nicotinamide adenine dinucleotide phosphate (NADPH) oxidase, which may be associated with HLH [10,11]. HLH is characterized by impaired function of T cell-mediated inflammation, which is partly regulated by NADPH oxidase. This pathophysiological cooperation may account for the increased severity.
EBV-related HLH is an acquired, infection-related HLH that typically represents a fulminant presentation of acute EBV infection of CD8 + T cells, and has a mortality rate of 30-50% [8]. In the present study, 30.4% of the children with non-malignancy-associated HLH were EBV-positive; eight patients (36.4%) with malignancyassociated HLH were positive for EBV infection. Ahn et al. [12] suggested that patients with high EBV DNA viral load have poor prognosis. In the present cohort, a 12-year-old girl with γδ T cell lymphoma relapsed and developed hemophagocytic syndrome after receiving chemotherapy for 6 weeks; she died of severe EBV infection. Strenger et al. [13] found that malignancy-induced HLH concurrent with EBV infection might be a possible trigger in immunocompromised patients.  In children, HLH may be associated with SLE, a systemic autoimmune disorder involving multiple visceral organs. In HLH due to SLE, corticosteroids and immunosuppressive agents, including cyclosporine, cyclophosphamide, intravenous immunoglobulin, and etoposide, have been used with variable success [14].
In the present study, patients with non-malignancyassociated HLH had significantly higher HGB (p = 0.004), ALT (p = 0.002), and AST (p = 0.001) levels than the patients with malignancy-associated HLH. The cause of low HGB in malignancy-associated HLH may be associated with the inhibition of hematopoiesis by malignance and chemotherapy. However, the cause of high ALT and AST levels in non-malignancy-associated HLH is unclear. Damage to liver function is characterized by severe inflammation and immune-mediated organ damage. Inflammatory cell proliferation and infiltration into organs and tissues and uncontrolled hypercytokinemia in non-malignancy-associated HLH may be more obvious than that in malignancy-associated HLH [15].
The difference in survival time was not statistically significant between the two groups (p = 0.055). However, previous studies have confirmed that patients with malignancy have worse survival than those without malignancy [16][17][18]. Celkan et al. [19] reported 54% overall survival in 13 children and adolescents with malignancy-associated HLH. The 13 children included five patients with leukemia; eight patients with rhabdomyosarcoma, neuroblastoma, or lymphoma; and one patient with Langerhans cell histiocytosis. Another study reported that the 2-year survival rate of 25 children with malignancy-associated HLH was 40.9%, and survival was 56% following the acute phase of HLH; a 5-year survival rate of 36% has also been reported [20]. We did not detect significant differences in outcome between malignancy-associated HLH and non-malignancy-associated HLH. Our results show that the low survival rate in the latter group might be due to CAEV, immunodeficiency diseases, and autoimmune diseases, as the prognosis of the abovementioned underlying diseases was poor.
The factors that affect survival are not well known, and there are currently no standard outcome predictors for SHLH. Recent data have shown that ferritin reduction is a prognostic variable for mortality in children with HLH [21]. Following univariate analysis, Park et al. [22] found that serum Fib ≥166 mg/dL at the initial visit was significantly associated with survival time. Low histiocyte proportion in the bone marrow and early initiation of treatment are generally correlated with favorable outcome. However, we did not find any relationship between the decline in serum ferritin, Fib level, and survival. Univariate analysis demonstrated that DIC score > 5 at diagnosis (p = 0.001) and albumin < 25 g/L (p = 0.000), HGB < 60 g/L (p = 0.001), and PLT < 30 × 10 9 /L (p = 0.042) were negative prognostic factors for patients with malignancy-associated HLH. The multivariate analysis showed that HGB < 60 g/L (p = 0.027), albumin < 25 g/L (p = 0.017), and bone marrow hemophagocytosis (p = 0.034) were negative prognostic factors for both groups. A previous comprehensive study involving 52 Turkish children with HLH suggested that high DIC score (≥5) and age < 2 years at diagnosis are important risk factors that can increase the mortality risk by twofold [23]. We also found that high DIC score (≥5) and low albumin, HGB, and PLT levels predict poor progress.
Most cases of non-malignancy-associated HLH should be treated aggressively with standard HLH protocols. In particular, the prognosis for such cases has improved dramatically with chemotherapy and immune-modifying agents such as corticosteroids, intravenous immunoglobulins, cyclosporin A, and etoposide.
The treatment of malignancy-associated HLH has not been prospectively studied. In the present study, the malignancy-associated cohort received malignancy-directed treatments, HLH-directed treatments, or a combined approach to overcome HLH. However, we could not determine which approach was superior. In a murine model of familial HLH type 2, mice that received etoposide, cyclophosphamide, or methotrexate survived [24]. Regimens involving the three agents can possibly treat malignancyassociated HLH. Moreover, clinical experience is greatest for etoposide [25]. In malignancy-associated HLH, 13 patients (59.1%) received etoposide as initial therapy for HLH, and six children are alive; the survival rate was 46.2%. Therefore, we found that etoposide is effective.
Stem cell transplantation might be considered consolidation in malignancy-associated HLH. The complete remission rate of malignancy-associated HLH was 90%. Therefore, stem cell transplantation is an effective therapy.

Conclusions
We analyzed a single center's experience with malignancyassociated HLH and non-malignancy-associated HLH in children. The etiology of malignancy-associated SHLH is unclear. New research has demonstrated that an underlying polygenic inheritance defect should be suspected in acquired HLH [4]. Prospective studies with a randomized controlled design and involving large cohorts of children are warranted to evaluate the exact incidence, clinical features, and appropriate treatment protocols of this association. Mutation analysis, effective diagnostic criteria, and new targeted treatments for HLH for improving the survival rate of patients with malignancy-associated HLH are necessary.