Introduction

Hemophagocytic lymphohistiocytosis (HLH; also known as hemophagocytic syndrome) is a life-threatening excessive immune activation with cytokine storm and multi-organ dysfunction. HLH can occur as a sporadic disorder or due to a genetic predisposition. Infection is a common trigger in both entities, especially when induced by human herpes virus (HHV), influenza virus, human immunodeficiency virus (HIV), Epstein-Barr virus (EBV), cytomegalovirus (CMV), or parvovirus or varicella-zoster virus (VZV) (1–7), but also COVID-19, caused by SARS-CoV-2, can be associated with HLH (8). So far, in a collective of 669 patients with severe COVID-19, Alam et al. have reported a rate of 2% with an H-Score ≥ 169 (9, 10), and Tang et al. have described a HLH prevalence of 7% in severe COVID-19 patients in a systematic review (8).

Rapid initiation of HLH treatment is crucial for a favourable outcome, but diagnosis of HLH is challenging due to the wide range of possible laboratory findings, the clinical gestalt, and the rarity of the syndrome. While the so-called HLH-94 protocol (11) or its updated version (HLH-2004) (12) have been widely used in the past, the H-Score (13, 14), which was first described in 2014, is nowadays commonly utilized for estimating an individual’s risk of having HLH. The score gives points for a known underlying immunosuppression, temperature/fever, organomegaly, number of cytopenias, levels of ferritin, triglycerides, fibrinogen and alanine aminotransferase (AST), as well as hemophagocytosis features on bone marrow aspirate. (13) As the score increases, so does the probability of HLH: ≤ 90 points confers a < 1% probability of HLH, whereas a score of ≥ 241 (up to a maximum of 337) corresponds to a > 99% probability. (13) There is no uniform cut-off in the literature, but a value around 169 (93% sensitivity, 86% specificity, accurate classification in 90% of patients (15)) is usually seen as diagnostic for HLH (9).

It is known that higher H-Scores seem to be associated with a higher risk of ICU admission, an extended hospitalization, and higher mortality rates in COVID-19 patients. (16) However, the prognostic value of the H-score has not yet been assessed in detail in the spotlight of secondary HLH in severe COVID-19.

Methods

Study population and design

In this retrospective cohort study, COVID-19 patients treated at the intensive care unit (ICU) of the Department of Infectious Diseases and Tropical Medicine, Clinic Favoriten, Vienna Healthcare Group, Vienna, Austria, were eligible for inclusion. Patients were consecutively enrolled between February 2020 and April 2021. Exclusion criteria were transfer to another intensive or intermediate care unit (due to lack of adequate follow-up), age < 18 years, and missing relevant data.

Ethics statement

Ethical approval for this study (EK 20–079-VK) was provided by the Ethical Committee of the City of Vienna. The study protocol complies with the Declaration of Helsinki, and data reporting was performed according to STROBE guidelines.

Collection of demographic and medical data

Data including patient characteristics, medical history, clinical information, laboratory parameters, and further ICU data were assessed retrospectively by document review and were then inserted into a predefined record abstraction form.

Statistical analysis

Categorial data are presented as absolute and relative frequencies and were compared among subgroups using χ2-test. Continuous variables are described using means ± standard deviations (SD) or medians and the respective interquartile ranges (IQR), and were compared via Mann-Whitney-U test. We used a cut-off of ≥ 169 for the diagnosis of HLH by using the H-Score (‘positive’ H-Score). (9, 13, 17) We assumed statistical significance through two-sided p-values of < 0.05, and performed calculations with the statistical software R (RStudio Version 1.2.5033, RStudio Inc., Boston, MA, U.S.A.).

Results

Between February 2020 and April 2021, we enrolled 254 COVID-19 patients (defined as new onset of symptoms and positive for SARS-CoV-2 by RT-PCR test (CT value < 30) from tracheal secretion, nasal or nasopharyngeal swab). After excluding 46 subjects, the data of 208 (81.9%) patients could be analyzed. Figure 1 gives an overview of the study participants and the classification in terms of H-Score.

Fig. 1
Fig. 1
Full size image

Flow chart of study participants and classification in terms of H-Score. ICU = intensive care unit; IMCU = intermediate care unit.

Basic characteristics

The average age was 61.8 ± 13.6 years, around a quarter were women (28.4%), and more than half were obese (29.9 ± 6.3 kg/m2). There were no significant differences in the aforementioned parameters between the two groups H-Score ≥ 169 and H-Score < 169. The groups also hardy differed regarding known chronic diseases (only chronic kidney disease (CKD) was more frequent in the group with a H-Score ≥ 169). Table 1 gives a more detailed overview of the basic study population characteristics, known chronic diseases, and laboratory findings. During the inclusion period, the entire study cohort was unvaccinated.

Table 1 Basic study population characteristics. Categorial data are presented as counts and percentages, continuous data as means and standard deviations, or median and interquartile range, respectively. Categorial data are compared via χ² test, continuous data via Mann-Whitney-U test. CAD = coronary artery disease; CKD = chronic kidney disease; CRP = c-reactive protein, IL-6 = interleukin-6, DDQ = ??, LDH = lactate dehydrogenase, ICU = intensive care unit, NIV = non-invasive ventilation, CRRT = continuous renal replacement therapy; ECMO = extracorporeal membrane oxygenation.
Full size table

Established ICU scores

Data regarding the patients’ APACHE II, SAPS II, and SAPS III scores at ICU admission can be found in Table 1. Only APACHE II showed significant differences between the two H-Score groups. In patients with a “negative” H-Score, the mortality rates fit to the mortality rates regarding the ICU scores. However, in patients with a H-Score of ≥ 169, the mortality rates were much higher. 28-day mortality: For H-Score < 169: 26.3% vs. APACHE II 16 (25%), SAPS II 34 (15.3%), and SAPS III 54 (23.9%); whereas for H-Score ≥ 169: 66.7% vs. APACHE II 19 (25%), SAPS II 38 (21.3%), and SAPS III 59 (33.5%).

(Predicted) mortality

With special regard to the then initial stage of pandemic and the lack of overall knowledge, especially regarding infectiousness and the according fears, autopsies were rarely carried out at this time. Based on the clinical presentation, however, we assume that sepsis and subsequently multiple organ failures were the main cause of death.

H-Score

The average H-score for the entire study population was 94 ± 51 points, despite the fact that no one received a bone marrow aspiration (up to 35 points) and several patients had no imaging documented (up to 38 points can be given for hepatosplenomegaly). The proportion of patients with an H-Score ≥ 169 was 8.7% (18 individuals) or 4.8% (10 individuals; all laboratory parameters included in the score from the same day), respectively. Table 2; Fig. 2 give an overview of the exact distribution.

Table 2 Overview of the points achieved in the H-Score. Positive values are highlighted in blue.
Full size table
Fig. 2
Fig. 2
Full size image

Overview of the points achieved in the H-Score.

H-Score in detail

Almost all patients (93.3%) received points for an elevated aspartate aminotransferase (AST), 70.7% for high triglyceride levels, 55.3% had a temperature of > 38.4 °C, more than a quarter (27.9%) had ferritin levels ≥ 2,000 ng/mL, and 21.2% had at least two lines of cytopenia (Fig. 3; Table 3). Less than 5% were scored for immunosuppression and fibrinogen levels. No statement can be made about organomegaly or pathological bone marrow aspiration due to missing data.

Fig. 3
Fig. 3
Full size image

Points awarded according to the H-Score in detail; light or dark blue depending on the number of points given. WBC = white blood cell; AST = aspartate aminotransferase.

Table 3 Points awarded according to the H-Score in detail. WBC = white blood cell; AST = aspartate aminotransferase.
Full size table

H-Score and ICU data

On average, patients required intensive care from day eight after symptom onset. Two thirds needed support in the form of non-invasive ventilation, 45.7% even required invasive mechanical ventilation, with an average ventilation duration of 15 ± 8 days. About 5% needed continuous renal replacement therapy, and 4% ECMO treatment. Furthermore, superinfection was very common, with 100% of the patients who also had an H-Score of > 169.

H-Score and mortality

ICU survival was significantly lower in the higher H-score group (38.9% vs. 75.8%). The same trend was observed for 28-day (66.7 vs. 26.3%, p < 0.001, Figs. 4) and 90-day mortality (72.2 vs. 27.9%, p < 0.001). In particular, increased temperature and ferritin levels were associated with mortality, but there were also significant differences in cytopenia and triglyceride levels (Table 4).

Fig. 4
Fig. 4
Full size image

H-Score and 28-day-mortality.

Table 4 Parameters of the H-Score associated with mortality. WBC = white blood cell; AST = aspartate aminotransferase.
Full size table

Patients with a ‘positive’ H-Score

Eighteen patients of the study cohort (8.7%) had a H-Score of ≥ 169 points and are shown in detail in Table 5. All of them got points due to increased temperature, for at least two lineages of cytopenia, and high ferritin, triglyceride, and AST levels. Patients who did not survive died at 22 [14–31] days after symptom onset.

Table 5 Details of patients with ‘positive’ H-Score. Hb = hemoglobin; WBC = white blood cell; AST = aspartate aminotransferase; ICU = intensive care unit † = dead.
Full size table

Discussion

Our study assessed the rate of secondary HLH (sHLH) in critically ill COVID-19 patients as well as their mortality, and found an average H-Score of 94 ± 51 points; 8.7% had a score of ≥ 169, thus testing positive for sHLH. A positive score was associated with increased mortality rates.

Viral infections, sHLH, and mortality

Secondary HLH can occur associated with a variety of triggers, but the most important is viral infection. In 74%, sHLH is triggered by Epstein-Barr virus (EBV), which also has the poorest prognosis, followed by infections with Cytomegalovirus (CMV) and human herpes virus 8 (HHV-8) as main triggers. Regarding mortality, data are very heterogeneous: Ishii et al. reported a 5-year overall survival rate of over 80% for patients with EBV- or other infection-associated HLH in Japan (19); a recently published study reported an overall HLH mortality rate of 40% (20); and even higher mortality rates of around 57% (regardless of HLH aetiology or specific treatment) were reported in a large ICU cohort of patients diagnosed with HLH in France (10). Within this wide range lie the results of our study with a 28-day mortality of two-thirds of all patients with COVID-19 and a positive score.

(S)HLH and (severe) COVID-19

In contrast to other published studies, our patient population shows an average H-Score in the entire study collective of 94 ± 51 points, with 8.7% scoring ≥ 169 and thus testing positive for sHLH. In addition, we report significantly higher mortality rates (two-thirds of all patients). In hospitalized paediatric and adult patients with a proven diagnosis of COVID-19 (n = 193), Bordbar et at. described a median H-Score of 43.0 (IQR 0–63), and, in general, a higher score in those who need intensive care and died of COVID-19. (16) Data from 19 patients with COVID-19 requiring an ICU stay showed 2% with a H-Score ≥ 169, and 16% of all subjects in this study died. (9) The significantly higher mortality rates also suggest that the population analysed in our study was likely to have been much more severely affected by the underlying disease. High rates of invasive mechanical ventilation and bacterial superinfection also support this. Of note, when assessing established ICU scores such as APACHE II, SAPS II, and SAPS III, the respective mortality rates fit patients without HLH but not those with HLH (in whom mortality was much higher than predicted by the scores). Accordingly, the H-score could potentially be a better predictor of ICU mortality in severe COVID-19 than the mentioned ICU scores. This must be further investigated.

sHLH versus severe COVID-19

COVID-19 and sHLH (with cytokine release syndrome (CRS) or hyperinflammatory syndrome (HIS)) have parallels but are two different entities; sHLH may or may not be present in COVID-19. Moreover, the H-Score without directing it at evaluating HLH is likely not appropriate for detecting hyperinflammatory states in “pure” COVID-19 and also not suitable to make statements regarding outcome. (21–26) Temperature peaks of > 39 °C, leukopenia, very high levels of ferritin, hypertriglyceridaemia, and low levels of fibrinogen (all paramters from the H-Score) are, however, no typical clinical features of COVID-19: Common laboratory findings among hospitalized COVID-19 patients include lymphopenia (27) with a varying total WBC (28) and elevated inflammatory markers (e.g., ferritin). In a large study with 5700 hospitalized COVID-19 patients, only 31% had a temperature of > 38 °C at presentation, the ferritin was 798 (411–1515; reference range 15–400) ng/mL, and the WBC 7.0 (5.2–9.5) G/L (29). Other authors published ferritin levels of 139 (66–403) ng/mL, fibrinogen of 386 (129) mg/dL (= 3.86 g/L) (30), and median triglyceride concentrations between 0.96 and 1.94 mmol/L (= 85.0 or 171.7 mg/dL, respectively). (31–33)

Bone marrow aspiration and potential hemophagocytosis features

As in previous literature (16), no bone marrow punctures were performed in our study collective (see Limitations). However, the limited available data predict a very high rate of hemophagocytosis in patients with both HLH and severe COVID-19: Bichon et al. reported a bone marrow hemophagocytosis for 91% of patients in a general ICU-HLH collective (10), and haemophagocytes with erythrophagocytosis were detected in 67% of a collective of deceased individuals with severe COVID-19 (18). If we assume the same 2/3 rate in our subjects, we would have a rate of patients with an H-score ≥ 169 of 16.8% (N = 35). An adjustment of the score could be considered in the future, considering among other things that a puncture is time-consuming, can be associated with complications and is not very specific.

Clinical implications and future outlook

The findings support regular collection of the H-Score, especially for COVID-19 ICU patients. An automatic warning of an ICU IT solution when a certain limit value is exceeded could be considered to be mindful of potential worse outcomes and a potential trigger for specific therapeutic options. Automatic HLH screening could be conducted daily by artificial intelligence, for instance.

Of note, our findings should be validated in another COVID-19 cohort which is vaccinated against SARS-CoV-2-.

Limitations

The main limitation of this study was the lack of bone marrow aspiration and imaging data, as no one had a puncture, and some imaging was not documented. Therefore, up to 73 points (21.7% of a maximum of total score of 337) are missing in the score results. The lack of bone marrow analyses and knowledge about organomegaly was due to the beginning of the pandemic, where little was known about COVID-19, let alone respective sHLH, and these diagnostics were simply not performed routinely.

Furthermore, 37 patients had to be excluded because of transfers to other ICUs or IMCUs. These patients had a maximum H-Score of 126 at the time of transfer, which is significantly higher than the average of the collective examined in this study (94 ± 51 points). Also, the percentage with a positive score was higher (13.5 vs. 8.7%). A possible reason could be the lack of availability of veno-venous extracorporeal membrane oxygenation in the ICU of the study site, which is often the last therapeutic option in the most severe COVID-19 cases. In general, the single-centre nature of a study is always fraught with disadvantages.

Moreover, the timepoint of patients presenting to the emergency medical services, the emergency department, and the ICU naturally varied, depending on incubation time, clinical courses, and also the patients’ subjective feelings. Therefore, the H-score was determined at different stages of the disease.

With regard to mortality, there could also be confounding factors that were not taken into account and could therefore play a role. In particular, we did not have detailed information about the treatment, how much the patients had previously been affected in their lives, or the causes of mortality.

Conclusion

In our cohort study, critically ill COVID-19 patients with a H-Score ≥ 169 and thus testing positive for sHLH during their ICU stay had increased mortality rates at 28 and 90 days. Thus, specific attention should be paid to individuals with high H-Scores in ICUs. Therapeutic options and their impact on mortality for patients with COVID-19-associated sHLH should be evaluated in further studies.