“…In our previous work, we developed computational models to analyze the biomarker profiles for 85 subjects from the HUSTLE clinical trial (NCT00305175; total of 260 participants) and we used the pharmacy refill record to incorporate the nonadherence in the dosing profile 17,27 (Text S1 and S2). In this analysis, we retrospectively analyzed the longitudinal data of 40 subjects that includes hematologic parameters prior to and during HU therapy leaving out participants without sufficient data points or having large time gaps, participants with adherence "adequately" inferred from pharmacy data, and participants receiving blood transfusion for our nonadherence study.…”