Abstract

A single point mutation in the β hemoglobin gene causes sickle cell disease (SCD), but patients have extremely variable phenotypes. Hemolysis-related complications include pulmonary hypertension (PHT), priapism, stroke and leg ulceration; blood viscosity and sickle vasoocclusion are associated with painful episodes, acute chest syndrome and osteonecrosis. Predicting who is at highest risk of death would be useful therapeutically and prognostically. Applying Bayesian network modeling that describes complex interactions among many variables by factorizing their joint probability distribution into modules, to data from 3380 SCD patients, we constructed a disease severity score (DSS: 0, least severe; 1, most severe), defining severity as risk of death within 5 years. A network of 24 variables described complex associations among clinical and laboratory complications of SCD. The analysis was validated in 140 patients whose SCD severity was assessed by expert clinicians and 210 adults where severity was also assessed by the echocardiographic diagnosis of PHT and death. Information about PHT allowed a comparison of the DSS with the tricuspid regurgitant jet velocity (TRJV), an objective marker of PHT and an independent risk factor for death. DSS and three indices of clinical severity (severity ranking of individuals by expert clinicians; objective measurement of the presence and severity of PHT; risk of prospective death) were correlated. Among living subjects, the median score was 0.57 in 135 patients without PHT, 0.64 in 40 patients with mild PHT and 0.86 in 15 patients with severe PHT. The difference in average score between living patients with and without PHT is significant. The same increasing trend was noticeable in the subjects who died during follow-up: 0.60 in subjects without PHT; 0.68 in subjects with mild PHT; 0.79 in subjects with severe PHT. The utility of the DSS is also supported by the ability to assign a score to subjects for whom the TRJV cannot be measured. Surprisingly, besides known risk factors like renal insufficiency and leukocytosis, we identified the intensity of hemolytic anemia and clinical events associated with hemolytic anemia as contributing to risk for death. Priapism, an excellent reflection of the hemolytic anemia-related complications of SCD, is associated with PHT and its association with death was unexpected. Laboratory variables predictive of disease severity included LDH and reticulocytes that reflect the intensity of hemolytic anemia. Elevated systolic blood pressure increased the odds of death by 3.4, consistent with hypertension as a marker of early death in SCD. Subjects with sickle cell anemia are at greatest risk compared with subjects with sickle cell anemia-α thalassemia and with subjects with HbSC disease. Our model suggests that the intensity of hemolytic anemia, estimated by LDH, reticulocyte count and AST, and shown previously to be associated with PHT, priapism, leg ulceration and possibly stroke, is an important contributor to death. This model can be used to compute a personalized measure of disease severity that might be useful for guiding therapeutic decisions and designing clinical trials.

Disclosures: Consulted for industry.; Dr. Sebastiani is Chief Scientific Officer of Bayesware.; NIH support.; For consulting, speaking.; Medicolegal proceedings.

Author notes

*

Corresponding author