External Validation of the HLH-2004 Diagnostic Criteria and H-Score for Diagnosis of Hemophagocytic Lymphohistiocytosis in Adults
Bibliographic record
Abstract
Introduction: Hemophagocytic lymphohistiocytosis (HLH) is a hyperinflammatory syndrome associated with multi-organ failure and death. Diagnosis in adult patients is currently based on the HLH-2004 diagnostic criteria; however, these criteria were developed for pediatric HLH and have not been formally validated in adults. An alternative diagnostic score, the H-score, was developed for adults with reactive HLH. There have been few external validation studies comparing the diagnostic accuracy between the HLH-2004 criteria and H-score, mostly in critically ill patients. In this external validation study, we aimed to compare the discriminatory power of these two diagnostic criteria in predicting HLH in a multicenter cohort of adults with suspected HLH. Methods: We identified all adult inpatients (≥18 years) with an International Classification of Diseases (ICD) code for HLH in the province of Alberta, Canada from January 1999 to December 2019. Following independent chart review by two reviewers, cases were classified as positive, negative, or indeterminate cases of HLH. The HLH-2004 diagnostic criteria and H-score were determined for each case. The following performance characteristics of the diagnostic scoring systems were calculated: sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV). C-statistic was calculated from the receiver operator curve (ROC) analysis. Results: Data on 101 patients were collected. The median age was 46 years (range 18-88), 66 (65%) were male. Active malignancy within 6 months of presentation, infections, and autoimmune diseases were present in 31%, 36%, and 18%, respectively. Of the 96 patients who underwent bone marrow aspirate and biopsy, hemophagocytosis was present in 79 (82%). Using chart review as the gold standard, 89 (88%) patients were determined to have HLH, whereas 5 (5%) and 7 (7%) were reclassified as negative and indeterminate cases, respectively. 83 (82%) patients met ≥5 of the 8 HLH-2004 diagnostic criteria; using the HLH-2004 criteria missed 8 cases of HLH. The sensitivity and specificity of HLH-2004 criteria was 91.0% and 83.3%, respectively, with a PPV of 97.6% and an NPV of 55.6%. The median overall H-score was 238 (range 129-337). An H-Score cut-off of >169 (established as the optimal cut-off from the original derivation and validation paper) predicted HLH with a sensitivity of 95.5%, specificity of 16.7%, PPV of 89.5%, and NPV of 33.3%. The discriminatory power improved with an H-Score cut-off >210, with a sensitivity of 86.5%, specificity of 66.7%, PPV of 95.1%, and NPV of 40.0%. The c-statistics for the HLH-2004 criteria and the H-Score were 0.872 and 0.737, respectively. In the malignant subgroup, all patients fulfilled the HLH-2004 criteria and all were deemed to have HLH upon chart review. All but one had an H-Score >169, with a median H-Score of 238 (range 168-304) in the malignant group. Conclusion: In our highly selected patient population with suspected HLH, the HLH-2004 had better discriminatory power than the H-Score, with a high sensitivity and specificity and excellent c-statistic >0.80. An H-Score cut-off of >169 was highly sensitive but non-specific in identifying adults with HLH. The validity and applicability of our study is limited by a high prevalence of chart-confirmed HLH in our sample. We are completing ongoing data collection to expand our validation study in a cohort of adults with hyperferritinemia who underwent bone marrow evaluation and/or sCD25 testing. Disclosures Sun: Sanofi: Other: Advisory board; Octapharma: Other: Advisory board; Octapharma: Research Funding; Novo Nordisk: Other: Advisory board; Pfizer: Other: Advisory board.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.002 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".