Pilot study on thrombosis risk factors in sickle cell anemia

Abstract

Abstract Background:Sickle cell disease (SCD) is a complex hemoglobinopathy affecting approximately 70,000–100,000 individuals in the United States. Patients with SCD are at increased risk for a range of complications, including anemia, vaso-occlusive crisis (VOC), acute chest syndrome, and end-organ damage. Thrombosis, involving both venous and arterial systems, adds disproportionately to disease-related morbidity. Although SCD is recognized as a hypercoagulable state, limited evidence exists to help clinicians identify which patients are at the highest risk for thrombosis. Prior studies of coagulation activation during VOC have yielded inconsistent findings, suggesting that mechanisms beyond transient vascular occlusion may drive thrombosis in this population. In other hypercoagulable conditions, laboratory biomarkers have been used to develop risk scores; however, no validated thrombosis risk score exists for patients with SCD.
 Objective: This study aims to identify laboratory and clinical biomarkers associated with thrombosis in patients with sickle cell anemia (SCA). The goal is to develop a clinically useful risk stratification tool that can identify patients with SCD at highest risk for thrombosis, guiding future prospective studies and targeted prevention strategies.

 Methods: We are conducting a retrospective, matched case-control study using patient encounters from 2013 to 2023. This interim analysis includes 49 adults with sickle cell anemia: 26 with documented thrombotic events (excluding stroke) and 23 matched controls. Controls were matched based on age (±2 years), sex, and genotype (HbSS or HbSC). Laboratory values were collected at steady state (outside of hospitalizations or VOC), using values closest to the index thrombotic event or matched control timepoint. Biomarkers included lactate dehydrogenase (LDH), white blood cell count (WBC), absolute neutrophil count (ANC), platelet count, hemoglobin, reticulocyte count, fetal hemoglobin (HbF%), hemoglobin S percentage (HbS%), D-dimer, and ferritin. Clinical variables included hydroxyurea use, hospitalization frequency, presence of indwelling catheters, chronic kidney disease, and avascular necrosis. Non-parametric testing (Mann–Whitney U for continuous, Fisher's exact for categorical) identified variables with p < 0.05 for entry into multivariable logistic regression. Area under the ROC curve (AUC) was calculated to assess model performance. Statistical analyses were performed using Python (pandas, scipy, and statsmodels libraries). As a pilot study, this interim analysis focuses on hypothesis generation and feasibility assessment.


 Results: Hemoglobin levels were significantly higher in thrombosis cases compared to controls (median 8.6 vs. 7.7 g/dL, p = 0.041). Among thrombosis cases, Hydroxyurea use was significantly lower (p = 0.012). In contrast, hospitalization frequency was higher in patients with thrombosis (median 2.5 vs. 1.0 per year, p = 0.067), but the difference did not reach statistical significance. Other biomarkers, including LDH, platelet count, ANC, reticulocyte count, and HbF%, did not differ significantly between groups. Catheter-associated events comprised 38% of thrombosis cases. Although not statistically significant (p = 0.19), this observation highlights the need to examine procedural risks in future studies. In the final logistic regression model, hemoglobin (OR 3.21, 95% CI 1.07–9.56, p = 0.037) and hydroxyurea use (OR 0.08, 95% CI 0.01–0.57, p = 0.012) were independent predictors of thrombosis. Hospitalizations per year were higher (OR 1.41, 95% CI 0.96–2.06, p = 0.078) but did not reach statistical significance. The model achieved an AUC of 0.83, reflecting strong predictive performance in distinguishing patients with vs. without thrombosis. Conclusion: In this interim analysis of 49 adults with sickle cell anemia, elevated hemoglobin and lack of hydroxyurea use were independently associated with thrombosis. Hospitalization frequency may also reflect underlying disease severity and warrants further investigation. While other candidate biomarkers did not reach statistical significance, further case and control identification is underway. The final model demonstrated strong predictive performance (AUC 0.83) and provides a potential foundation for a clinically useful risk stratification tool in SCD. These preliminary findings point to meaningful directions that warrant validation in large prospective studies.