The poor aqueous solubility of many modern active pharmaceutical
ingredients (APIs) can severely limit bioavailability. Cocrystals
are emerging as a promising means of overcoming this limitation. Solution
crystallization is a common technique for cocrystal synthesis. The
type of solvent to be used and the operating conditions are important
design decisions. In cocrystal systems, choosing a coformer is also
nontrivial, as it affects process efficiency and possibly enhanced
bioavailability. However, existing model-based crystallization process
optimization approaches focus only on solvent and operating condition
selection. Product performance optimization is typically not addressed
simultaneously. In this work, a new model-based optimization framework
is presented for the integrated selection of coformers, solvents,
and process operating conditions. This approach considers the various
trade-offs arising from the competing process and product performance
criteria of cocrystal systems. A perturbed-chain statistical associating
fluid theory-based equilibrium process model describes the cocrystallization
process. The API dissolution behavior of the cocrystal in a rotating-disk
apparatus is modeled by using a dynamic model. A hybrid branch-and-bound-continuous
mapping approach is proposed as the mixed-integer nonlinear programming
(MINLP) solution strategy, which involves decomposing the original
MINLP problem into a series of computationally tractable nonlinear
programming problems. The results show that the proposed solution
strategy can successfully solve the optimization problem to identify
a list of coformer and solvent candidates. Furthermore, the good predictive
performance of the model is demonstrated experimentally. The optimal
solvent provides a substantially higher solubility for the API, and
thus a higher attainable yield than the ones reported in the literature.
Finally, the impact of various coformer feeding strategies and the
dissolution medium composition on the optimal solution is revealed.
The presented approach is especially impactful during the early stages
of process and product design, as limited experimental input is required.