Intercontinentally validated diagnostic criteria for secondary hemophagocytic lymphohistiocytosis—So welcome!

Abstract

In this issue of the Journal of Internal Medicine, Lachmann et al. report on a multicontinental effort to validate diagnostic criteria for secondary hemophagocytic lymphohistiocytosis (sHLH) [1].

The report, the first multicenter diagnostic validation study for sHLH, is important for critically ill inflamed patients and their physicians because rapidly making the diagnosis of sHLH can markedly reduce its mortality and morbidity. The diagnostic criteria that performed best were the original HLH-2004 criteria (using the cutoff 4 of 8 criteria) and the recently revised HLH-2004 criteria, also termed the HLH-2024 criteria, in which NK-cell activity is removed (cutoff 4 of 7 criteria), followed by the HScore (cutoff 169 points) [2-4]. Moreover, ferritin is confirmed as a reliable biomarker to screen for sHLH; ferritin ≥ 500 µg/L had a mean sensitivity of 94.0% overall in 13 validation cohorts.

HLH is a severe inflammatory syndrome that comes in a primary (Mendelian-inherited) form (pHLH) and a secondary (non-Mendelian) form (sHLH). sHLH is one of the most critical clinical disorders in adults; the overall mortality rate has been reported as 41% in 1109 adults [5]. It is most often triggered by infections, malignancies, and autoimmune diseases. Other triggering factors include transplantation and novel drugs, such as chimeric antigen receptor (CAR) T-cells, bispecific T-cell engagers, and checkpoint inhibitors [5].

Prompt and accurate diagnosis of HLH is important in order to initiate adequate therapy early and thereby decrease mortality and morbidity (such as inflammatory-induced brain damage). Diagnostic criteria for pHLH in children have been available ever since 1991 [6] and were revised in 2004 (“the original HLH-2004 criteria”) [2] and then further revised through data-driven case–control analysis of international pediatric trial databases in 2024 (in this commentary referred to as “the HLH-2024 criteria”) [3]. Notably, although the original HLH-2004 criteria and the HLH-2024 criteria were developed for children with Mendelian-inherited HLH, they have also been commonly used for adults with non-Mendelian HLH, that is, sHLH. An obvious question has therefore been: How accurate are these diagnostic criteria, and other diagnostic criteria, in sHLH?

In the present article, Lachmann et al. set out to answer this question and, more specifically, aimed to systematically optimize and validate diagnostic criteria of sHLH using an ambitious multicenter approach. First, they used a dataset of their own containing 2623 critically ill adult patients, of whom 40 were diagnosed with sHLH [7, 8], trained a model on this dataset to systematically optimize diagnostic criteria (termed “their optimized criteria” in this commentary). Second, they conducted a systematic literature search of PubMed according to a well-defined strategy for the acquisition of suitable validation cohorts. They received requested data from 17 studies, of which 13 (from 4 continents) were considered suitable as validation cohorts: 6 from Europe (France = 2, Belgium = 2, Spain = 1, and the Netherlands = 1), 4 from Asia (China = 3, South Korea = 1), 2 from America (USA = 1, Canada = 1), and 1 from Africa (Morocco). Third, in these 13 cohorts, they validated several sets of diagnostic criteria as well as 2 biomarkers, ferritin, and soluble interleukin-2 receptor (sIL-2R), which they also had evaluated earlier [8, 9]. In addition, as a post hoc analysis, they evaluated the value of ferritin as an HLH screening marker.

Results of the study for various diagnostic criteria presented as the mean Youden's indices over all 13 cohorts are shown in Table 1. Youden's index captures the performance of a diagnostic test. The index is defined as = sensitivity + specificity − 1 =  ($\frac{TruePositive}{TruePositive+FalseNegative}$) + ($\frac{TrueNegative}{TrueNegative+FalsePositive}$) – 1. Thus, its value ranges from −1 to +1. A value of zero indicates that the diagnostic test gives the same proportion of positive results for groups with and without the disease, that is, the test is useless, whereas a value of +1 indicates that there are no false positives or negatives, that is, the test is perfect. The index gives equal weight to false positive and false negative values.

The results are both interesting and reassuring. The best performing diagnostic criteria were actually the original HLH-2004 criteria (cutoff 4 of 8 criteria; mean sensitivity 86.5%, mean specificity 86.1%, mean Youden's index 0.725) and the HLH-2024 criteria (cutoff 4 of 7 criteria; mean sensitivity 83.8%, mean specificity 87.8% mean Youden's index 0.716) followed by the HScore (cutoff 169 points; mean sensitivity 82.4%, mean specificity 87.6%, mean Youden's index 0.700) [1]. The “optimized HLH inflammatory” (OHI) index developed for malignancy-associated HLH and based solely on ferritin (>1000 µg/L) and sIL-2R (>3900 U/mL) showed a mean Youden's index of 0.235, whereas it was 0.366 in a cohort of 348 patients aged ≥13 years with lymphomas only [1, 10]

The clinical implications of this report are wide. A first step to identify patients with sHLH is to think of and screen for the diagnosis. Lachmann et al. show that a ferritin of 500 µg/L showed high mean sensitivity (94.0%) over all cohorts and 100% in ICU cohorts. Raising the ferritin cutoff to 1000 or 3000 µg/L markedly decreased mean sensitivity (to 85.7% and 57.8%, respectively). The next clinical step is to establish the diagnosis. In contrast to primary (Mendelian) HLH, where genetics can confirm the diagnosis and cellular cytotoxicity assays showing that absent perforin expression or defective lymphocyte exocytosis can strongly suggest it [3], the diagnosis of sHLH still has to rely on diagnostic criteria. Therefore, this, the first multicenter validation study of diagnostic criteria for sHLH, is tremendously valuable. It actually confirms and extends the authors’ previous single-center study, in which the original HLH-2004 criteria and HScore proved to be of good diagnostic accuracy for HLH diagnosis in critically ill patients. Notably, as the current study covers four continents and both children and adults, as well as the three main underlying diseases triggering sHLH (infections, malignancies, and autoimmune diseases), there are reasons to believe that the results can be generalized to many different clinical settings.

It is interesting and a bit surprising that the original HLH-2004 criteria and the HLH-2024 criteria, both developed for children with Mendelian-inherited (primary) HLH, performed so well also in adults with non-Mendelian (secondary) HLH. A possible explanation could be that the biological mechanisms in the hyperinflammatory phase of pHLH and sHLH are quite similar, even though they are not identical.

Notably, as the original HLH-2004 criteria and the HLH-2024 criteria both recently were shown to have high diagnostic accuracy in pHLH [3], both these sets of diagnostic criteria can be used for making the diagnosis of HLH without knowing in advance whether the patient has pHLH or sHLH, which also facilitates making the diagnosis of HLH and initiation of its treatment. The clinical situation, such as the age of the patient, genetic factors, background inflammation, underlying immunosuppression, and infectious triggers, can help to discriminate between pHLH and sHLH, as well as genetic and cellular cytotoxicity assays detailed in reference [3]. Importantly, the syndrome of HLH may have various possible causes, and it is therefore of vital importance to always search for the underlying trigger(s) for the HLH. Moreover, as always in clinical medicine, making a diagnosis should be based on the entire evaluation of the patient and cannot only be based on predefined clinical criteria.

Finally, the study also serves as one successful example of many in science where collaboration over multiple continents can improve the well-being of individuals in all countries in the world. Our world currently faces many urgent needs where scientific collaboration can provide guidance for decisions that politicians and other decision-makers need to make.

None.