How strong is each kind of pharmaceutical patent? A field guide, with the biologics twist

The question the register does not answer

A patent register tells you that a patent exists, what it claims and when it expires. It does not tell you whether the patent would survive a challenge or whether a competitor could simply go around it, and those two questions decide the launch date far more often than the expiry date does. The earlier explainer on the types of patent says what each kind covers. This one says how much weight each kind usually deserves, from the European cases, and why the answer is different for a biologic.

Small molecules: the order of strength

Compound patent: the wall. A patent on the molecule blocks every salt, form, dose and use of it, and nothing a generic does by way of design gets around it. It falls only if it is invalid, and the ground on which compound patents have fallen in Europe lately is plausibility: the application did not make it credible that the compound worked. That is how apixaban and dapagliflozin lost their UK protection years early. The counter-examples are as instructive: enzalutamide and sacubitril/valsartan were attacked on the same grounds and held, in the UK and elsewhere. A compound patent is strong, and when it has survived one serious attack it is very strong; the date on its SPC is then the date.

Manufacturing process: a nuisance, not a wall. A generic buys its active ingredient from whichever manufacturer has a route that avoids the patent, and for most small molecules there are several. A process patent matters only where the only practical synthesis is patented, which is rare, or where it covers an intermediate the whole industry uses.

Formulation: designed around, except where it cannot be. For an ordinary tablet or capsule a generic reaches bioequivalence with its own excipients and the formulation patent never bites. It bites where the formulation is the product: a sustained-release profile that is hard to match, a stable injectable solution. The dabigatran case, where one generic company’s formulation patent kept another off the German market, shows the exception.

Salt, polymorph, enantiomer: a chemistry question first. Whether a generic can use a different salt or crystalline form that is stable and bioequivalent decides the matter before any court does. Where it can, these patents are irrelevant; where it cannot, they are attacked as obvious, and often successfully, as apremilast’s enantiomer patent was at the EPO.

Use and dosing regimen: the weakest common layer. A second-medical-use or dosing patent protects only the use, so a generic can carve the indication out of its label; and the patents themselves fall often, because the claimed regimen was what the published clinical trial already tested. Rivaroxaban’s once-daily regimen fell in the UK, Germany, Sweden, the Netherlands and elsewhere; cabazitaxel’s use patent was revoked by the UPC three months after the EPO upheld it; nivolumab’s flat-dose patent and Tecfidera’s dosage patent fell at the EPO for added matter. They are not nothing: nalmefene’s patient-subgroup claim survived a central attack at the UPC in 2026, and enforcing a carved-out label against actual prescribing is possible. But a launch plan that treats a use patent as a wall is usually leaving money on the table.

Combination: fragile, and the SPC more so. A patent on two actives together blocks the combination product, but the Court of Justice has made combination SPCs hard to sustain unless the combination was the patent’s own invention, and national courts have been refusing them: dapagliflozin with metformin lost its French certificate in 2026. The mono-product’s protection is the one to plan around.

Device: seldom decisive for a small molecule, which is usually a tablet.

Biologics: the order reverses

A generic of a small molecule is the same molecule, bought from any competent chemist. A biosimilar is not the same molecule: it is a protein made by a different cell line, in a different process, and it is similar rather than identical. That one fact rearranges the whole table.

The process is the product. A biosimilar developer must make its own antibody, and the way the originator makes it, the cell line, the culture conditions, the purification steps that set the glycosylation profile and the impurity levels, is patented in layers. Those patents cannot be side-stepped by buying from another supplier, because there is no other supplier, and they are hard to design around without changing the product’s characteristics, which the regulator will notice. Manufacturing patents are therefore the first thing a biosimilar’s lawyers look at and the last thing an originator lets go, and they are routinely among the patents asserted when a biosimilar is enjoined.

Formulation patents are strong for biologics. A high-concentration antibody in a stable buffer, an injection that can be given under the skin instead of by infusion, a presentation that does not aggregate on the shelf: these are real inventions, hard to match, and the European courts have been enforcing them. Omalizumab’s liquid-formulation patent held in the UK and kept the biosimilar out until it expired. Aflibercept’s formulation patent was enforced from Munich across twenty countries and from The Hague in the Netherlands while the UK found it not infringed, which is a split, not a weakness. The enzyme that makes subcutaneous Keytruda possible is a third-party patent that has kept MSD’s own product off the German market. For a biosimilar planner the formulation and delivery patents are the real runway after the sequence patent, and they are usually the last to expire.

The sequence patent is strong but no longer untouchable. Broad first-generation antibody claims, which claimed every antibody that bound a target, are being revoked at the EPO and the UPC for lack of inventive step or sufficiency; the EPO’s own 2025 reviews describe the trend. Evolocumab’s antibody patent was revoked by the UPC at first instance in 2024 and reinstated on appeal in 2025, which tells you both that the attack is real and that the outcome is uncertain. Eculizumab’s remaining patent was found not infringed in the UK and too doubtful to support an injunction at the UPC.

Use and dosing patents are as weak for biologics as for small molecules, and they are being cleared early: nivolumab’s dosing patents at the EPO years before any biosimilar exists; the PCSK9 lipoprotein(a) use patent held valid at the UPC but not infringed, because the label said nothing about the use.

Devices matter for substitution. Where a pharmacy may substitute a biosimilar only if it matches the reference product’s device and presentation, as Germany now requires, the autoinjector patents and the pen design decide whether the biosimilar can take the prescription at the counter.

And the regulatory layer is often the binding one. For many biologics the eight years of data protection and ten of market protection, or orphan exclusivity, end after the sequence patent and its SPC. The clock that matters is whichever ends last, and for biologics it is frequently not a patent at all.

How to read a patent in the register, then

• Ask what kind it is before asking when it expires.
• For a small molecule: the compound patent and its SPC set the earliest date; everything else is a question of what your product does and whether the patent would survive, and for use, dosing and combination patents the honest answer is often “probably not”.
• For a biologic: the sequence patent sets the first date, but the manufacturing and formulation patents set the real one, and they cannot be bought around. Plan on them, and on the regulatory clock, before the sequence patent.
• Either way, a patent that has already survived a serious challenge in one European court is worth more than one that has never been tested, and one that has fallen in the UK may still be standing in Germany.

The register entries, the SPC status and the litigation record for each product and country are what MPA Search holds; which patents would actually hold is what its support is for.