To the Editor:


Haemophagocytic lymphohistiocytosis (HLH) is a life-threatening clinical syndrome of excessive immune activation [1]. Multiple predisposing factors influence the likelihood of developing HLH; children frequently have inherited defects in cytotoxic lymphocyte function, younger adults have viral infections and/or autoimmunity whilst older adults most likely have underlying malignancy. Therapies are directed towards both acute inflammation and underlying trigger factors. The incidence of HLH has been increasing [2, 3] and high mortality rates, particularly those with underlying malignancy, well established [1, 4]. Published data are derived from single or a small number of large academic centres [1, 5], or specialty-centred populations such as paediatrics [1, 6, 7] over extended time-periods. We examined 1-year survival of HLH in England for all cases diagnosed 2003–2018 and modelled interactions between demographics and comorbidities with survival.

Linked electronic health records from English Hospital Episode Statistics (HES) [8], the National Cancer Registration Dataset (NCRD) [9] and Office for National Statistics (ONS) death certification data were used. Patients diagnosed with HLH were identified using our validated approach [2, 3, 10]. They included patients of all ages admitted to hospital or died between 1 January 2003 and 31 December 2018. Date of diagnosis was the first day of the admission in which HLH was coded. The presence/absence of comorbidities was identified from available HES and NCRD records prior to diagnosis of HLH and up to three months after. Where there was overlap of non-infectious comorbidities, patients were classified with a mutually exclusive hierarchy: haematological malignancy, rheumatological disease/inflammatory bowel diseases (IBD), non-haematological malignancy and none recorded. As there was no access to serological or genetic tests, the type (acute, chronic, reactivation) of viral illness nor genetic cause could not be ascertained. The number at risk from the date of HLH diagnosis and 1-year mortality frequencies were calculated by age, gender, calendar period and comorbidity.

A Cox regression model was fitted, adjusted by age, sex, comorbidities and calendar time period to assess if observed differences in survival were confounded. We fitted interactions between age and calendar year, and between comorbidity and calendar year, and tested them with general likelihood ratio tests. We used R (version 4.1.2) [11] for statistical analyses (see Additional file 1: Methods).

A total of 1628 patients were identified (Additional file 2: Fig. S1). Characteristics of the cases and 1-year survival are described in Table 1. Of the whole cohort, 461 (28.3%) had a recorded haematological malignancy, 378 (23.2%) a non-malignant comorbidity (rheumatological disease/IBD), and 107 (6.6%) a non-haematological malignancy. Overall, crude 1-year survival was 50% (95% CI 48–53%) which varied with age (0–4, 61%; 5–14, 76%; 15–54, 61%; > 55, 24% p < 0.01), gender (males, 46%, females, 55% p < 0.01) (Additional file 3: Fig. S2) and comorbidity (rheumatological/IBD, 69%, haematological, 28%, or other malignancy, 37% p < 0.01) (Fig. 1). Most deaths occurred within two months of diagnosis. Those aged 0–14 and 15–54 years had a threefold increased risk of death at 1-year among HLH associated with haematological or non-haematological malignancy versus rheumatological disease/IBD (Additional file 4: Fig. S3, Additional file 5: Table S1). Outcomes did not depend upon underlying cancer subtype, when split into B-cell, T-cell and Hodgkin lymphoma sub-groups or by solid organ malignancy subtype (Additional file 6: Fig. S4, Additional file 7: Fig. S5). In the adjusted model, no change in survival was observed over time.

Table 1 Characteristics of the HLH cohort and 1-year survival (95% CI)
Full size table
Fig. 1
figure 1

1-year survival estimates by hierarchical comorbidity

Full size image

Our study provides the first estimates of survival for HLH by associated trigger factors using a population-based cohort. In terms of age, sex distribution and proportions with haematological malignancy and rheumatological diseases, the characteristics of our cohort are similar to a prior CPRD study [2] and comparable studies of HLH elsewhere [1]. The 461 patients with HLH and haematological cancer and > 500 cases of HLH in patients aged 0–14 years represent the largest respective series reported, with no comparable sized cohorts of auto-immune related HLH. Notably, patients aged 0–5 years have a greater risk of death than those 5–14 years. This is potentially due to the youngest paediatric patients having more profound genetic T- and NK-cell dysfunction, and presenting with more florid cytokine storms. Haematological malignancies complicated by HLH have dismal outcomes regardless of disease subtype, as do all patients over 55 years regardless of trigger. Potential reasons include reduced ability to tolerate acute cytokine storm and, for auto-immune disease, an assumption that cases only occur in younger people. We also critically show survival outcomes have remained static for the study-period.