A novel prognostic model for predicting mortality in adults with sickle cell disease and severe sinusoidal obstruction syndrome

Abstract Background: Liver involvement in sickle cell disease (SCD) is often underdiagnosed, with hepatic complications such as sinusoidal obstruction syndrome (SOS) often masked by hemolysis or gallbladder disease (PMID: 31808845). While SOS is well described in the post-stem cell transplant setting, its significance in SCD remains poorly defined. This study characterizes the clinical, laboratory, imaging, and survival features of severe SOS in adults with SCD, using bilirubin thresholds to define severity in the absence of mechanical biliary obstruction (PMID: 30967906, 37981193). Furthermore, we developed a prognostic model tailored to this population, improving upon general liver risk scores like the MELD score, which performs suboptimally in settings with complex hepatic pathophysiology (PMID: 31394020, 31553133). Methods: This retrospective cohort study included adults (≥18 years) with confirmed SCD receiving care at the University of Illinois, Chicago, from January 1, 2010, to March 31, 2025. Data were extracted from the Clinical Data Warehouse, and additional validation was performed through manual chart review in EPIC and Cerner. Patients with any recorded total serum bilirubin level ≥15 mg/dL were classified as having pronounced SOS, while those with levels ≥30 mg/dL were designated as major SOS (PMID: 30967906, 37981193), in the absence of extrahepatic biliary obstruction or liver transplantation. Demographic, clinical, and laboratory data were compared across sexes and SOS severity using appropriate statistical tests. All-cause mortality was the primary outcome. Survival analysis was done with Kaplan-Meier curves and Cox proportional hazards models. Patients with invalid or missing follow-up durations were excluded. A multivariable Cox model was constructed using a 70/30 training/test data split, ridge penalization, and backward stepwise selection from 21 clinically relevant covariates. Model performance was evaluated using concordance index (C-index) and Schoenfeld residuals. Results: Among 114 SCD patients with total bilirubin ≥15 mg/dL, 100 (87.7%) met criteria for pronounced SOS and 32 (28.1%) for major SOS. Patients with mechanical biliary obstruction (n=11) or post-liver transplant (n=3) were excluded. The median age at follow-up was 39 years (IQR: 32–51), and 61(53.5%) were male. Most had HbSS genotype (91%), with a median follow-up period of 2.76 years (IQR: 0.33–6.57). The overall mortality rate was 44.7% (51/114); excluding 12 patients lost to follow-up, mortality reached 50%. There were no significant sex-based differences in bilirubin, liver enzymes, or imaging. Survival was significantly worse in major SOS, especially with imaging-confirmed cirrhosis (log-rank p<0.01). In multivariable Cox regression (n=77), elevated lactate dehydrogenase (HR 1.0012, 95% CI 1.0009–1.0012), alanine aminotransferase (ALT; HR 1.0014, 95% CI 1.0005 – 1.0024), alkaline phosphatase (ALP; HR 1.0013, 95% CI 1.0004 – 1.0022), and the presence of major SOS (HR 1.81, 95% CI 1.06–3.10) were independently associated with mortality. Protective factors included higher serum albumin (HR 0.43, 95% CI 0.33–0.57) and hemoglobin levels (HR 0.88, 95% CI 0.79–0.97). The model achieved a C-index of 0.83 in the training set and 0.88 in the test set, outperforming the MELD score, whose published C-index is 0.7–0.8 (PMID: 17326206, 34333100). In our cohort, MELD components—bilirubin, creatinine, and INR—did not significantly predict mortality, underscoring their limited utility in SCD. By incorporating SCD-specific clinical variables, our model enhances the ability to predict outcomes in this population. Conclusions: Pronounced and major SOS are common but underrecognized complications in adults with SCD and are associated with significantly reduced survival, particularly when coexisting with radiographic cirrhosis. Standard liver risk scores such as MELD may misclassify mortality risk in this population. Our multivariable Cox model, which integrates markers of hemolysis, liver injury, nutritional status, and disease severity, provides a more nuanced and clinically relevant tool for risk stratification. These findings highlight the need for systematic hepatic monitoring in SCD and suggest that early identification of SOS could inform targeted interventions and transplant referral decisions.