The effect of F8 missense variants on desmopressin response in people with nonsevere hemophilia A investigated using machine learning.

BACKGROUND: Interindividual variability in factor FVIII peak levels after desmopressin administration in people with nonsevere hemophilia A underscores the need for individual testing before clinical application. The FVIII gene (F8) variant is a strong determinant of the desmopressin response. OBJECTIVES: To investigate the effect of F8 missense variants on FVIII peak levels in response to desmopressin. METHODS: Machine learning (Shapley Additive exPlanations) was applied to a population pharmacokinetic model. Data from 1441 people with nonsevere hemophilia A, including 55 different F8 variants, who underwent desmopressin testing, were analyzed. RESULTS: Ten F8 missense variants had a significant effect on FVIII peak levels, while the remaining variants were classified as reference, including patients with unknown variants. Variants Tyr450Asn, Gly477Val, Pro149Arg were associated with a reduced FVIII increase of 24%, 25%, and 56%, respectively, compared with the reference population (100%); Asn637Ser, Phe2146Ser, Glu132Asp, Arg550His, Ser2030Asn, Thr74Met, and Arg717Leu were associated with an augmented FVIII increase of 199%, 219%, 231%, 252%, 265%, 296%, and 336%, respectively. Following subcutaneous and intranasal administration, FVIII peak levels were 88% and 73%, respectively, compared with the intravenous level (100%). The inclusion of all covariates reduced FVIII peak level variability from 141.6% to 78.1%. CONCLUSION: This population pharmacokinetic study, using machine learning, enabled a comprehensive analysis of the effect of F8 variants on FVIII peak levels in response to desmopressin, accounting for other covariates and illustrating the complexity of interindividual variability. These findings provide insights into the underlying molecular mechanisms that influence FVIII peak levels in response to desmopressin.