Population Modeling Is a Critical Element of Bridging Pharmacokinetic Data From Dried Blood Spot and Plasma Across Clinical Programs

Purpose: To demonstrate through two case studies how population pharmacokinetic (PK) modeling should be leveraged for bridging plasma and dried blood spot (DBS) PK data across clinical development programs. Methods: In two case studies (MK-X and MK-Y), population PK models initially developed from Ph1 plasma data were updated to include plasma and DBS data from healthy subjects (Phase 1 setting) and patient (late-stage trials) plasmaDBS bridging studies. DBS samples were collected via in-clinic venipuncture (MK-X and MK-Y) and in-clinic and at-home fingerstick (MK-Y). An estimated population slope converted between DBS and plasma concentrations. Separate residual errors for plasma and DBS data were estimated. For MK-Y, residual error was further partitioned based on in-clinic vs at-home DBS sampling. Models were qualified through standard diagnostics and qualification approaches. Interchangeability of matrices was evaluated through various approaches including a comparison of post-hoc predicted exposures from plasma vs DBS data alone (using slope as a conversion factor). Results: In both cases, two-compartment population PK models were developed. Population PK parameter estimates were similar with and without DBS data. The slope parameters were well estimated and consistent with the DBS-plasma linear regression slopes and in vitro blood:plasma ratio data. Residual error for in-clinic DBS was low and generally comparable to that for plasma; however, high residual error (113% CV) was observed for at-home DBS (MK-Y). For MK-X, DBS and plasma based post-hoc estimates of plasma exposures were interchangeable and lacked bias. Phase 1 results were used to inform Phase 2 analysis plans and development of a DBS Go/No Go decision tree for later phase implementation. Conclusion: The literature cites simplified approaches with generally arbitrary cut-offs (ISR criteria, regressions, Bland-Altman plots, etc) to bridge plasma and DBS concentrations. We have shown that pop PK modeling with prospective model-informed analysis plans should be a critical element of plasma-DBS bridging strategies. These approaches directly address the development question of whether DBS sampling supports pharmacometric aspects of regulatory submissions if incorporated in larger-scale patient studies.