Integrative modeling to improve bleeding risk prediction in adult female hemophilia A carriers

Background Although bleeding phenotypes and factor (F)VIII activity (FVIII:C) in men with hemophilia A correlate highly with F8 mutations, FVIII:C in female HA carriers is more variable due to mosaicism arising from a second wild-type F8 allele and X-chromosome inactivation (XCI). Conventionally, only female HA carriers with low FVIII:C were considered at risk for bleeding. However, hemostatic FVIII:C levels in females remain unestablished. Moreover, HA carriers with normal FVIII:C levels can also have reproductive tract and other bleeding sequelae. Objectives We sought to develop a statistical model for predicting abnormal bleeding risk. Methods Potential HA carriers were enrolled in an observational cross-sectional study. F8 genotyping confirmed the carrier status. Bleeding severity was quantified by bleeding assessment tools. FVIII levels were assessed using 1-stage and chromogenic activity assays, as well as FVIII antigen ELISA. XCI skewing was also quantified. Logistic regression analyses were performed to determine the relationship between abnormal bleeding score and covariates. Results Of 92 adult females, 62 were confirmed F8 mutation carriers; 55% of HA carriers had abnormal bleeding (Condensed Molecular and Clinical Markers for the Diagnosis and Management of Type 1 VWD Bleeding Questionnaire 1 score ≥ 4), despite only 19% with FVIII:C < 40%. Logistic regression modeling incorporating XCI skewing, FVIII antigen, and activity performed better than those with FVIII:C alone. Highest model performance was seen for carriers with nonsevere F8 mutations. Conclusion Bleeding tendency in HA carriers is common, yet FVIII:C predicts only 38% of those with abnormal bleeding. Logistic regression modeling is highly promising and demonstrates superior performance in predicting bleeding risk compared with baseline FVIII activity alone.