Introduction
Hemophagocytic lymphohistiocytosis (HLH) is a hyperinflammatory syndrome caused by excessive cytokine release, triggered by genetic or acquired overactivation of macrophages, T and natural killer (NK) cells [
1]. Clinical presentation may include fever, cytopenias, organomegaly, and hyperferritinemia, none of which are specific for this rare though life-threatening condition [
2‐
4]. As HLH shares similarities with other inflammatory states, e.g., sepsis, its diagnosis is challenged by clinical overlap particularly in the intensive care unit (ICU) [
2]. Consequently, HLH is likely to be under-recognized in critically ill patients where evidence for clear definition and correct diagnostic workup is lacking [
5]. So far, diagnosis largely relies on data derived from studies conducted in pediatric patients [
6]. Henter et al. developed the HLH-2004 criteria whereby a diagnosis of HLH is confirmed if five out of eight criteria are fulfilled [
7]. However, these guidelines lack prospective validation in adult HLH patients. Moreover, the specificity of some criteria has been questioned. According to HLH-2004 guidelines, a ferritin ≥ 500 μg/L meets the criterion of hyperferritinemia [
7]. However, markedly higher ferritin levels have been seen in adult HLH patients [
8]. In fact, we detected best prediction accuracy at a ferritin cutoff level of 9083 μg/L with 92.5% sensitivity and 91.9% specificity for HLH in critically ill patients, thereby providing satisfying discrimination of HLH patients [
8].
The HScore published by Fardet et al. [
9] calculates a sum score of nine variables allowing to assess the probability of HLH. Each variable was assigned a maximum number varying between 18 and 64 points. The authors found the best discriminatory performance at an HScore of 169 with a sensitivity of 93.0% and specificity of 86.0%. Unlike the HLH-2004 criteria which are composed of parameters derived from a pediatric population, the HScore was developed in an adult cohort including patients aged ≥ 18 years. Yet, only non-ICU patients were included, possibly limiting the tool’s generalizability to critically ill patients. Hence, it is unclear whether HLH-2004 criteria and HScore reliably detect and discriminate HLH in adult critically ill patients. We therefore calculated the sensitivity and specificity of HLH-2004 criteria and HScore, respectively, in a cohort of adult patients admitted to ICUs at an academic medical center.
Discussion
This is the largest study investigating the diagnostic performance of HLH-2004 criteria and HScore in an adult intensive care population. We found the best prediction accuracy of HLH diagnosis for a cutoff of 4 fulfilled HLH-2004 criteria (95.0% sensitivity and 93.6% specificity) and an HScore cutoff of 168 (100% sensitivity and 94.1% specificity). Analyses of each single HLH-2004 criterion revealed good sensitivity and specificity for a ferritin cutoff of 9083 μg/L described previously [
8] while all other HLH-2004 criteria had unsatisfying predictive ability in our study, including sIL-2R, which was also described previously [
8]. The combination of 4 fulfilled HLH-2004 criteria provided better diagnostic accuracy compared to each single criterion. By adjusting HLH-2004 criteria cutoffs of both hyperferritinemia to 3000 μg/L and fever to 38.2 °C, sensitivity and specificity increased to 97.5 and 96.1%, respectively. Our approach to analyze fixed combinations of fulfilled HLH-2004 criteria showed less prediction accuracy compared to independent combinations of at least 4 fulfilled HLH-2004 criteria. Both HLH-2004 criteria and HScore were independently associated with in-hospital mortality.
The HLH-2004 criteria, currently the standard in HLH diagnosis, have been developed in pediatric populations but so far have not been validated in adult patients. According to current recommendations for HLH in adults, HLH diagnosis requires ≥ 5 fulfilled HLH-2004 criteria which should be considered along with patient’s history and clinical presentation [
6]. In daily practice, clinical presentation might be suggestive of HLH, while less than 5 out of 8 HLH-2004 criteria are present. Moreover, the diagnostic value of some criteria, e.g., fever, is limited, particularly in critically ill patients where the use of antipyretic agents and devices such as extracorporeal membrane oxygenation (ECMO) and hemodialysis are frequently seen rendering body temperature an unreliable or even invalid parameter. Of note, the cutoff of 4 fulfilled HLH-2004 criteria had the best sensitivity and specificity possibly allowing faster HLH diagnosis, prompt treatment, and thus improved survival. Yet, these findings need further confirmation in prospective studies to validate safe HLH diagnosis in adults with only 4 fulfilled HLH-2004 criteria. Ongoing studies could contribute to improve safe HLH diagnosis in adult critically ill patients [
10].
Our analysis of sensitivity and specificity of single HLH-2004 criteria is broadly in line with data reported in pediatric HLH patients. Hypofibrinogenemia is known to have high specificity but rather low sensitivity as only 53% of children with HLH had fibrinogen levels < 1.5 g/L [
11]. Also, we found sensitivity for fibrinogen of 1.5 g/L to be at 43.6% while specificity was at 91.1%. In contrast to studies in pediatric populations, sIL-2R proves to be of insufficient diagnostic value in adult patients [
8] whereas levels ≥ 2400 U/L in children provided good sensitivity and excellent specificity of 93.0% and 100%, respectively [
11]. The diagnostic value of ferritin in the present cohort has been described previously by our research group and appeared as a good screening marker [
8]. Importantly, the presence of 4 fulfilled HLH-2004 criteria provides higher sensitivity and specificity for HLH diagnosis than ferritin alone. Of note, sensitivity and specificity of 4 fulfilled HLH-2004 criteria increased to 97.5% and 96.1%, respectively, when cutoffs of both hyperferritinemia and fever were adjusted to 3000 μg/L and 38.2 °C, respectively. In this context it is noteworthy that 5 out of 8 HLH-2004 criteria can be fulfilled in critically ill non-HLH patients. NK cell activity was assessed in four patients only in whom HLH was likely considered as a differential diagnosis. For practical guidance, we recommend assessment of body temperature, cytopenias, ferritin, triglycerides, fibrinogen, splenomegaly, and wherever available sIL-2R. Hemophagocytosis, even though the eponymous feature of HLH with high specificity, is an unreliable diagnostic marker with only poor sensitivity, again particularly in critically ill patients with sepsis [
12]. However, the latest recommendations for HLH in adult patients advise bone marrow investigation as it helps to detect occult hemato-oncological malignancies and to differentiate between cytopenias caused by chemotherapy from patients who have actually underlying HLH [
6]. The HScore developed by Fardet et al. [
9] provides a tool to predict the probability of HLH diagnosis in adults. The authors found the best cutoff at an HScore of 169 yielding 93.0% sensitivity and 86.0% specificity in a cohort of non-ICU patients. Our present study included ICU patients only and revealed an HScore of 168 to have the best sensitivity and specificity of 100% and 94.1%, respectively, thereby providing slightly superior prediction accuracy compared to the HLH-2004 criteria. Importantly, the similar found cutoff underlines the value of the HScore for HLH diagnosis and its reliability in critically ill patients.
One previous study by Meena et al. [
13] also analyzed the diagnostic performance of HLH-2004 criteria and HScore in critically ill patients. The authors included 445 patients with ferritin assessment among whom ten were diagnosed with HLH. They reported an HScore of 143.5 for best possible classification and found 5 out of 6 criteria to be the cutoff for HLH-2004 criteria with 70% and sensitivity and 97.2% specificity. However, we present a larger cohort of 2623 patients including 40 HLH cases. Yet, the work by Meena et al. and our study are currently the only data available investigating the diagnostic standard for HLH diagnosis in the adult ICU population.
Both HLH-2004 criteria and HScore were associated with in-hospital mortality suggesting that both indicate disease severity. This relationship has been reported previously by Gualdoni et al. who found increased 30-day mortality correlating with HLH-2004- or HScore-positive patients [
14].
Our study has several limitations. As this is a retrospective study, data availability had to rely on patients who had a ferritin assessment during their ICU stay. This might constitute an important selection bias as patients with ferritin assessment might have been more severely ill. For instance, suspicion of inflammation or diagnostic of anemia was likely when ferritin assessment was considered. Thus, our findings might not be generalizable to ICU patients without hyperferritinemia. In addition, not all variables of HLH-2004 criteria and HScore were available in all patients which reflects clinical practice where rather rare diagnostic tests such as NK cell activity might be unavailable. Our study bears a considerable risk that HLH cases could have remained undiagnosed depending on physicians’ expertises, particularly in patients with ≥ 5 fulfilled HLH-2004 criteria.
Acknowledgements
We thank the Department of Cardiovascular Surgery, the Department of Surgery CCM/CVK, the Medical Department, Division of Nephrology and Internal Intensive Care Medicine (CVK/CCM), the Medical Department, Division of Infectiology and Pneumonology, the Medical Department, Division of Cardiology (CVK), the Department of Cardiology (CBF), the Department of Neurology with Experimental Neurology, and the Department of Anesthesiology and Operative Intensive Care Medicine (CBF) for being part of our study, providing the data and excellent collaboration. We are grateful to Oguzhan Mizrak for his help with data acquisition.
We acknowledge support from the German Research Foundation (DFG) and the Open Access Publication Funds of Charité – Universitätsmedizin Berlin.
Open AccessThis article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit
http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (
http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.