Identification of Clinical Sub-Phenotypes of Sickle Cell Disease Using Latent Class Analysis

Introduction: Sickle cell disease (SCD), characterized by the presence of abnormal hemoglobin (HbS), is the most common monogenic blood disorder in the world. Chronic hemolytic anemia and recurrent vaso-occlusion, both hallmarks of SCD, confer high risk of disease-related complications and end-organ damage to affected individuals. Although SCD arises from a single point mutation on the beta-globin gene, the onset and severity of disease-related complications are highly variable. Two clinical sub-phenotypes of SCD have been previously described in the literature: (1) hemolysis/endothelial dysfunction, characterized by symptomatic anemia, ischemic stroke, pulmonary hypertension, leg ulcers and priapism; and (2) increased viscocity/vaso-occlusion with recurrent pain episodes, acute chest syndrome, and osteonecrosis. To the best of our knowledge, these SCD sub-phenotypes are based on data from single-institution, cross-sectional studies, and review articles. Using latent class analysis (LCA), we tested the hypothesis that disease-related complications would aggregate into classes in a longitudinal cohort of people living with SCD in California, stratified by age.

Methods: LCA is a statistical technique more commonly used in social sciences to identify unobserved subgroups within a population based on patterns of responses to certain categorical variables. Using data from the California Department of Health Care Access and Information (CA HCAI) database, including both patient discharge and emergency department utilization data, we derived an observational SCD cohort study and collated disease-related complications using ICD-9/10-CM codes. We included common SCD complications such as: recurrent acute chest syndrome (ACS, ≥3 episodes over the study duration), severe recurrent vaso-occlusive pain episodes (VOE, ≥3 episodes per year), sepsis, chronic kidney disease, essential hypertension, pulmonary hypertension, venous-thromboembolism, osteonecrosis, liver disease, gallstones, leg ulcers, retinopathy, priapism, and stroke. Using LCA, we identified distinct classes of individuals with SCD with similar patterns of complications. We derived multiple models with different number of classes and complications included and used standard model performance measures (G², AIC, BIC, and ABIC) to select the best model. A separate analysis was performed on individuals ≤25 years old to determine class differences among the younger cohort.

Results: Among 7,636 individuals with SCD living in California and followed between 1991-2019, we identified four distinct classes based on patterns of disease related complications. Class 1 comprised 7.7% of the cohort, who had high incidence of severe VOE (96.3%), recurrent ACS (99.7%), osteonecrosis (76.5%), and pulmonary hypertension (82.5%). Class 2 (19.4% of the cohort) included individuals with high incidence of severe VOE (59.2%), recurrent ACS (79.9%), and osteonecrosis (59.6%), but much lower incidence of pulmonary hypertension (21.0%). Class 3 (9.2% of the cohort) had high incidence of chronic kidney disease (81.0%), pulmonary hypertension (50.2%) and stroke (32.3%), but lower incidence of severe VOE (10.0%). A majority of the cohort fell into class 4 (63.7%) with a low incidence of most complications, the highest being severe VOE (12.5%) and recurrent ACS (13.1%). Follow up time was similar among classes 1, 2, and 3 (median 25.0 years) while individuals in class 4 had significantly less follow up time (median 21.0 years).

When we examined 2,210 individuals with SCD age ≤25 years of age, we again found 4 divergent classes. Class 1 (7.6%) had high severe VOE (95.8%), recurrent ACS (73.8%) and osteonecrosis (100%). Class 2 (11.5%) had high recurrent ACS (100%) but less frequent VOE (35.3%). Class 3 (3.2%) had increased chronic kidney disease (34.3%), liver disease (54.3%), pulmonary hypertension (41.4%), and stroke (27.1%). Similar to the entire cohort, most young individuals with SCD were in class 4 (77.7%) with low rates of complications overall.

Conclusions: Individuals with SCD have distinct patterns of disease related complications, which differ by age. Results of this analysis support the hypothesis of SCD sub-phenotypes. Analysis on how class status is associated with sex at birth, socio-economic status, and mortality are ongoing.