Predictive PK/PD models for hematological toxicities to inform MTD determination and regimen selection for ADCs.

2568 Background: Off-target hematological toxicities have primarily been dose-limiting for ADCs. Here, we present a PK/PD model framework where the incidences of hematological toxicities can be predicted across ADCs if they share the same linker/payload (LP), and can be predicted across regimens within an ADC. Having prior knowledge of the MTD can provide a safe and efficient dose-escalation design, reduce the number of patients receiving sub-efficacious doses and aid in regimen selection. Methods: A semi-mechanistic PK/PD model of neutropenia (NP) for SGN-35 was constructed using a previously established structural model. Published ADC thrombocytopenia (TCP) PK/PD models were also utilized. Results: Predicting across regimens:Due to the availability of the clinical data in multiple regimens for T-DM1 and SGN-35, models developed using data from Q3W regimen were utilized to predict the toxicity incidences of QW regimen. T-DM1 Q3W model predicted 10% grade 3+4 TCP in patients at QW MTD, matching with the clinically observed incidence of 11%. SGN-35 Q3W model predicted 19% grade 3+4 NP at QW MTD, which is consistent with the clinically observed incidence of 10%. Predicting across ADCs: Since SGN-35, CDX-011 and PSMA-ADCs share the same LP (vcMMAE), we utilized the SGN-35 model to predict incidences for CDX-011 and PSMA-ADC. SGN-35 Q3W model predicted 18% and 50% grade 3+4 NP incidences for the 1.9 mg/kg CDX-011 and 2.8 mg/kg PSMA-ADC, respectively, again in concordance with the observed incidences of 26% and 55%. Predicting across ADCs, regimens and populations: As CMC-544 and GO share the same LP (abCali), we predicted TCP incidences for GO in AML patients from the Q4W CMC-544 model developed with data from NHL patients. The model predicted grade 3+4 TCP incidence of 98% for GO regimen of 9 mg/m2 (2 doses, 14-days apart), which agrees with the observed incidence of 99%. Conclusions: Robust predictability of hematological toxicity is demonstrated across 5 ADC, 2 toxicities and 3 regimens. With prospective validation, the approach presented here can help construct a safe and efficient dose-escalation scheme, provide a projected MTD, and assist in optimal regimen selection for ADCs entering clinical development.