Therapeutics I've helped make.

The whole product is a solubility trick. Levodopa is too poorly water-soluble to deliver a full day's dose in the small volume a wearable subcutaneous pump can push. Phosphorylating the catechol turns each drug into an ionizable prodrug soluble enough to formulate as one concentrated solution, and the body's phosphatases regenerate active levodopa and carbidopa after it's under the skin.

The hard part is keeping two oxidation-prone catechol phosphate prodrugs stable together, in one aqueous solution, at a pH subcutaneous tissue will tolerate, across a 24-hour-plus wear time without discoloration or degradation. Infusion-site tolerability, the dominant adverse event, is itself a formulation constraint, tying the chemistry directly to what the patient feels.

An ~868-dalton molecule, far past the rule-of-five, that is, for practical purposes, insoluble in water (BCS Class IV, logP near 8). Crystalline venetoclax has oral bioavailability under 5%. You cannot dose your way around that; you have to defeat the crystal.

The answer is an amorphous solid dispersion: the drug is molecularly dispersed in a polymer (copovidone) so it never gets the chance to crystallize, a stable crystal traded for a kinetically-trapped amorphous form that actually dissolves. That buys the bioavailability and inherits the burden: physical stability (will it recrystallize on the shelf or in the gut?), the drug-to-polymer ratio, and a manufacturing process controlled tightly enough to reproduce the same dispersion every lot. On top of a long, highly functionalized synthesis.

Here the signature problem is biology, not chemistry. Platelet survival depends on BCL-x_L, so inhibiting it drops the platelet count: thrombocytopenia is the dose-limiting toxicity, and it is the very reason venetoclax was engineered to spare BCL-x_L and hit BCL-2 alone. In Phase 3, navitoclax plus ruxolitinib met its primary spleen-volume endpoint but missed the key symptom-score secondary.

On the CMC side it is venetoclax's harder sibling: larger still (~975 Da), carrying three sulfur atoms and a trifluoromethanesulfonyl group, notoriously lipophilic and poorly soluble. The synthesis is long and the formulation fights the same solubility war, with even less margin.

This is the one that breaks the definition of “a drug.” β-cyclodextrin is a ring of seven glucose units; hydroxypropylating it decorates a large set of available hydroxyls to a variable average degree of substitution. So the active ingredient is not a molecule, it is a statistical distribution of a great many positional isomers.

The CMC challenge is almost existential: you cannot specify a single structure or a lone purity number. You have to control an average degree of substitution and the whole substitution fingerprint, and prove it is the same distribution lot to lot, because different manufacturers' grades differ measurably on exactly that axis, and carry different impurities (some implicated in ototoxicity). It is a textbook case of defining and controlling a complex mixture, the problem Quality by Design was built for.

Here the molecule disappears entirely: the product is living cells, and the factory is a batch of one. Every dose begins with a biopsy of the patient's own, diseased, kidney, so the starting material varies from person to person in a way no small-molecule API ever does.

The CMC problems are the cell-therapy canon at their hardest: an unbreakable chain of identity from biopsy to injection (a mix-up is catastrophic); a potency assay that meaningfully captures what a heterogeneous “selected renal cell” population is supposed to do; viability and sterility for a product that can never be terminally sterilized; validated cold chain in both directions; and comparability for a living, variable product every time the process changes. You scale out, more suites, never up. It's the world my CMC roadmap and the cell-therapy examples in my QbD guide live in.