Key Findings
- Purpose
To identify hematological, biochemical, and metabolic biomarkers associated with sickle cell anemia (SCA) severity and to assess machine learning models’ ability to predict disease severity accurately. - Population
481 participants were enrolled at Korle-Bu Teaching Hospital in Ghana, including 356 with SCA and 125 healthy controls. - Headline Result
Strong correlations emerged between immune cell counts, red cell indices, and bilirubin, reflecting inflammation and hemolysis in SCA. Five biomarkers: direct bilirubin, total bilirubin, reticulocyte count, hydrogen sulfide, and neutrophil count ranked as the most predictive of severity. - Why It Matters
These findings offer a data-driven method to stratify patients by severity using routine biomarkers and machine learning in SCA. - Evidence Gaps
The study was limited to a single-center Ghanaian cohort; the sample size, while meeting minimum requirements for machine learning, limits generalizability. The limited range of hematological biomarkers used did not fully explain inter-group variability, suggesting the need for broader panels that include radiological, laboratory, and clinical markers.
Source: Journal of Sickle Cell Disease – Identification of Hematological Biomarkers and Assessment of Machine Learning Models for Sickle Cell Anemia Severity Classification