Clinical characteristics and laboratory test results
The clinical and laboratory test characteristics of patients at diagnosis are summarized in Table 1, with 189 males and 164 females and with a median age of 45 months (range: 1-213 months).
Assessment of underlying etiology
DNA-targeted sequencing of the HLH related genes was performed in 173 (49.0%, 173/353) patients. Among the 173 patients, 29 were with verified pHLH, including 15 family HLH (FHL) cases (3 FHL-2, 11 FHL-3, 1 FHL-4), 14 other pHLH (4 Chediak-Higashi syndrome, 3 X-Linked Lymphoproliferative syndrome-1, 3 X-Linked Lymphoproliferative syndrome-2, 2 combined immunodeficiency, 1 X-linked chronic granulomatous disease, and 1 X-linked immunodeficiency with magnesium defect, Epstein–Barr virus infection, and neoplasia syndrome); no HLH-causing genes or related gene mutations were found in the remaining 144 patients.
Three hundred thirty-four patients (94.62%, 334/353) had a reported infection at diagnosis, and a patient may be infected by different pathogenic microorganisms (e.g., bacteria, fungi, and viruses): viruses are the most common pathogens, including Epstein-Barr virus (EBV) (n =314), cytomegalovirus (CMV) (n =24), and combined EBV/CMV (n =19) infection. Of the 314 EBV-infected patients, 13 were verified pHLH based genetic testing.
Levels of blood lipids
Among the 353 pediatric patients, 4 had missing lipid data. According to the standard reference ranges used in the clinical laboratory of Children's Hospital of Chongqing Medical University, the normal thresholds of TG, TC, HDL-C, and LDL-C were defined as <1.80 mmol/L, 3.11-5.18 mmol/L, 1.04-2.27 mmol/L, and 1.30-3.40 mmol/L, respectively. For the 349 patients, their median TG, TC, HDL-C, and LDL-C were 3.61 mmol/L, 4.65 mmol/L, 0.38 mmol/L, and 1.13 mmol/L, respectively;91.98% (321/349) of the patients had hypertriglyceridemia (TG≥1.80 mmol/L), 92.84% (324/349) had low HDL-C (≤1.04 mmol/L), 58.74% (205/349) had low LDL-C (≤1.30 mmol/L), and 24.64% (86/349) had low TC (≤3.11 mmol/L), as shown in Fig.2a. There was a moderate positive correlation between LDL-C and HDL-C (r = 0.44, p < 0.0001), between LDL-C and TC (r = 0.41, p < 0.0001), and between TG and TC was moderate (r = 0.30, p < 0.0001); TG was weakly correlated with HDL-C (r = -0.12, p = 0.028); no significant correlation was found between TG and LDL-C (r = 0.06, p = 0.232). The correlation plot of the four lipid indicators was presented in Fig.2b.
The distributions of lipid indicators in different subgroups were further investigated. For the 344 patients tested for both EBV and lipid indicators, no statistically significant differences of the four lipid indicators were identified between the EBV-infected and non-EBV-infected groups (Supplemental Fig.1). For the 172 patients tested for both HLH related gene testing and lipid indicators, the patients with verified pHLH (n=28) had higher TG than those not verified with pHLH (n=144) (5.00 mmol/L vs. 5.18 mmol/L, p = 0.023), while no statistically significant differences were found between the two groups for other lipid indicators (Supplemental Fig.2).
Survival outcomes
The median follow-up time for 353 pediatric patients was 23.9 months (95% CI: 19.4-28.5), with follow-up ending on August 5, 2021; 97 patients were lost to follow-up, 17 of whom were lost within 30 days of diagnosis. A total of 124 patients died, and 132 patients survived at the last follow-up. By Kaplan-Meier analysis, the estimated survival rates at 1, 2, and 3 years were 64.09% (95% CI: 58.97%-69.66%), 62.42% (95% CI: 57.14%-68.20%), and 61.67% (95% CI: 56.27%-67.59%), respectively (Fig.3). Detailed follow‐up information for each of the 353 patients was provided in Supplemental data.
Sixty-four patients died within 30 days, accounting for more than half of the deaths. The observed 30-day mortality rate was 19.05% (64/336, 17 were lost to follow-up). Compared with the survivors (n=272) at 30-day, the non-survivors at (n=64) had lower TC (2.64 mmol/L vs. 3.72 mmol/L, p < 0.001) and HDL-C (0.31 mmol/L vs. 0.38 mmol/L, p = 0.043), while the differences in TG and LDL-C between the two groups were not statistically significant (p = 0.935, p = 0.156, respectively). Fig.4 shows the distributions of lipid indicators in survivors and non-survivors at initial diagnosis. Detailed comparison of clinical and laboratory results between the two groups was presented in Supplemental Table 1. Overall, non-survivors had more abnormalities in liver, kidney, and coagulation functions. In the stratified survival analysis, the estimated 30-day survival rates for EBV-infected and non-EBV-infected children were 82.01% (95% CI: 77.85%-86.39%) and 75.64% (95% CI: 62.28%-91.86%), respectively, showing no significant difference (p = 0.432) (Supplemental Fig.3).
Univariable and multivariable analyses
Table 2 shows the results of univariable and multivariable Cox regression analyses of early mortality in children with HLH. The univariable analysis revealed that Platelet count ≤50×109/L (P = 0.002), Total Protein ≤44g/L (p < 0.001), Albumin ≤ 30g/L (p < 0.001), Total bilirubin ≥21.0μmol/L (p < 0.001), AST ≥550 U/L (p < 0.001), LDH ≥1000 U/L (p = 0.022), BUN ≥7.14 mmol/L (p < 0.001), Plasma Creatinine ≥97 μmol/L (p = 0.032), APTT ≥47s (p < 0.001), PT ≥ 20s (p < 0.001), TT≥25s (p < 0.001), INR Abnormal (p < 0.001), Fibrinogen<1g/L (p < 0.001), TC≤3.11 mmol/L (p < 0.001), Ferritin≥1500 ng/mL (p = 0.038) were significantly associated with early death. To determine the independent risk factors for early death in patients with HLH, a multivariable Cox regression analysis was performed using age, gender, and the significant variables from the univariable analysis (Total Protein, PT, and TT were not included due to their high correlations with Albumin, APTT, APTT, respectively). The results (Table2) showed that BUN ≥7.14 mmol/L [ HR=2.90, 95% CI: (1.48-5.68), p = 0.003] and TC ≤3.11 mmol/L [HR=2.85, 95% CI: (1.46-5.57), p = 0.003] were significantly associated with the survival within 30 days after diagnosis. To validate the robustness of results, a multivariable logistic regression model was also applied, showing that BUN ≥7.14 mmol/L [OR=4.03, 95% CI: (1.53-10.60), p =0.005] and TC ≤3.11 mmol/L [OR=4.66, 95% CI: (2.08-10.48), p < 0.001] were significantly associated with 30-day death after diagnosis (Supplemental Table 2). In addition, similar results were obtained from the Cox and logistic regression models using only complete cases (Supplemental Fig.4 and Fig.5).
Stratified survival analysis
From ROC curve analysis, the AUCs (area under the curve) of TC and BUN were 75.2% (95%CI: 68.5%-81.8%) and 63.0% (95%CI: 57.1%-69.0%), respectively (Fig.5a, Fig.5b). Compared with the normal TC group (3.11 mmol/L -5.18mmol/L), the 30-day survival rate in the low TC group (≤3.11mmol/L) was significantly lower (52.29% vs. 88.22%, p<0.0001) (Fig.5c), while in the high TC group (≥5.18mmol/L), there was no significant difference (93.33% vs. 88.22%, p = 0.157) (Fig.5c). Compared with the normal BUN group (<7.14 mmol/L), the 30-day survival rate in the high BUN group (≥7.14mmol/L) was significantly lower (40.18% vs. 86.13%, p<0.0001) (Fig.5d). Then, we considered the two risk factors simultaneously and classified the patients into three risk groups: low-risk (no risk factors present), intermediate-risk (one risk factor was present ), and high-risk (two risk factors were present,), with the 30-day survival rate of 90.61% (95% CI: 85.48%-96.05%), 52.69% (95% CI: 42.76%-64.91%)% and 39.71% (95% CI: 21.77%-72.40%), respectively (Fig.6). Compared with the low-risk group, the patients had a nearly 10-fold increased risk of early death [HR = 9.98, 95%CI: (4.23, 23.56)] in the high-risk group and a 6-fold increased risk of early death [HR = 6.27, 95%CI: (3.18, 12.22)] in the intermediate-risk group.