A case note on the District of Delaware’s August 24, 2026 memorandum opinion holding that Lupin’s ~74% ivacaftor generic does not infringe Vertex’s “80%” and “about 80%” Kalydeco patents, either literally or under the doctrine of equivalents.
Judge Bibas opens Vertex Pharmaceuticals Inc. v. Lupin Ltd. with a line built for quotation: “Words can be twisted, but numbers do not lie.” The line is memorable, but the analysis beneath it turns on something more specific than the flexibility of the word “about.” What the opinion actually holds is that where a patentee narrows a broad, previously disclosed range to a single value in order to overcome the examiner’s rejection — and later fails to convince the same examiner that the same result extends to a wider band — the prosecution record built to win the patent becomes the same record that limits what the patentee can recapture at trial. The number mattered only because the file wrapper made it matter.
The Drug and the Patents
Vertex’s cystic fibrosis drug Kalydeco is built around ivacaftor, a compound that improves the function of the CFTR protein. Ivacaftor’s problem is solubility: it dissolves too poorly for the bloodstream to absorb it in crystalline form. Vertex’s solution was to disperse ivacaftor molecules into other ingredients as an amorphous solid dispersion, which dissolves far more readily during digestion.
Roughly twenty years before this suit, Vertex published its findings, announcing that “a solid dispersion comprising amorphous [ivacaftor]” could treat cystic fibrosis, with an effective formulation comprising “about 10% by weight to about 80% by weight” ivacaftor. That publication became Vertex’s own biggest obstacle at the patent office: a wide, already-public range is difficult to patent, since it isn’t novel or nonobvious relative to what the applicant has already disclosed. The examiner’s first rejection cited exactly that prior disclosure.
Vertex’s response was to narrow, not broaden.
It told the examiner that a dispersion of exactly 80% ivacaftor produced a “surpris[ing]” result — solubility beyond what one would expect from a compound that dense with active ingredient. A second examiner accepted that argument and allowed claims to a drug containing exactly 80% ivacaftor, describing the ratio as “unexpected[ly]” effective and one that “should not work as well as it does.”
Vertex later tried to leverage that same 80% data point into a broader claim, seeking “about 72 wt% to about 88 wt%” ivacaftor in a related application. The examiner rejected the range as obvious over the prior art, reasoning specifically that “unexpected results often do not occur over wide ranges” — without evidence that the surprising solubility observed at 80% would recur anywhere across a sixteen-point band, there was no basis to claim the whole band. Vertex was left claiming “about 80 wt%,” a term the examiner construed narrowly during prosecution, offering 79.9% as an example of what “about 80%” could still encompass, and 0.55% as an example of what “about 0.5%” could encompass. Four patents ultimately issued: two claiming exactly 80% ivacaftor, two claiming “about 80%.” That rejection of the broader range, and the examiner’s own tenth-of-a-percent illustrations of “about,” are not background color — they become the intrinsic evidence the court relies on twice: first to construe the claim term, and later to defeat equivalence.
Lupin’s Generic
Lupin’s ANDA product is an amorphous solid dispersion containing 74.257% or 74.258% ivacaftor, made with different ingredients and a different manufacturing process — wet granulation (spraying a binder fluid onto the ingredients to form granules) rather than Vertex’s dry combination process. That process difference forced different filler ingredients as well. Vertex sued under the Hatch-Waxman Act, § 271(e)(2)(A), asserting all four patents.
Vertex wisely conceded that Lupin’s ~74% product could not literally infringe the two patents claiming exactly 80% ivacaftor, and pursued literal infringement only on the two “about 80%” patents. The case therefore turned on how a person of ordinary skill in the art would understand the word “about” — a term the court had already construed to carry its plain and ordinary meaning.
Literal Infringement: A Battle of Experts, Decided on Credibility
Literal infringement requires that “every limitation set forth in a claim must be found in an accused product, exactly.” Sharing an active ingredient is not enough; the overall claimed composition must match. With the meaning of “about 80%” doing all the work, the case turned on a credibility contest between the parties’ experts — one the court had already signaled its view of at trial.
Vertex’s expert, Dr. Berkland, argued that “80% ivacaftor” functioned as shorthand for a “high drug load,” and that “about 80%” should therefore be read to cover a wide band — “at least 70 to 90 percent as a range.” The court rejected this, and had signaled as much at the close of trial, describing Dr. Berkland as “less careful and less scrupulous and more willing to sign off on extreme statements,” whose “willingness to affirm these things undercuts his overall credibility.” The written opinion adds the substantive gap behind that credibility finding: the phrase “high drug load” appears nowhere in the patents or their prosecution history, and Dr. Berkland offered no scientific or principled basis for treating a numerical claim limitation as a stand-in for a broader conceptual category rather than as a boundary in its own right.
Lupin’s expert, Dr. Donovan, defined “about” as “nearly the same as” and, while declining to give an exact percentage range, limited it to a variance of less than a couple of percentage points — a reading she grounded in Vertex’s own prosecution history, which specified ivacaftor weight down to a tenth of a percentage point. The court found her “completely credible,” “meticulous,” and “admirably careful about what she could or couldn’t say,” and adopted her construction over Dr. Berkland’s.
The court reinforced that reading with the examiner’s own prosecution-history statements, admissible as evidence of how a person of ordinary skill would use the term. As noted above, the examiner treated “about 80%” as reaching 79.9% and “about 0.5%” as reaching 0.55% — variances of roughly a tenth of a percentage point, narrower even than Dr. Donovan’s own estimate, and nothing like Dr. Berkland’s “high drug load” theory. Because Lupin’s ~74% product differs from 80% by several percentage points, not a fraction of one, the court found no literal infringement.
The Doctrine of Equivalents: Three Independent Grounds Against Vertex
Vertex’s fallback theory — that even if not literally identical, Lupin’s drug is the functional equivalent of Vertex’s claimed formulation — failed on three independent grounds, each sufficient on its own.
1. The claim language itself is inherently narrow
The doctrine of equivalents is, in the Federal Circuit’s own description, an “exceptional” path to infringement, available only where language cannot capture “every nuance” of the invention. Courts have recognized that precise numerical claim limitations “warrant little, if any, range of equivalents” because of “the inherent narrowness of the claim language” itself.
Vertex specified its claims down to half a percentage point; Dr. Berkland’s “high drug load” testimony, the court found, would simply erase those numerical limits rather than interpret them. A numerical limitation is not automatically a loose proxy for a broader category — where the intrinsic record shows the number itself was what mattered, the number becomes the boundary.
2. Prosecution history estoppel bars exactly this kind of recapture
The doctrine of equivalents cannot be used to “impermissibly vitiate the limitation[s]” a patentee relied on to obtain its patent. The court drew a direct parallel to a prior case where a claimed ratio was “critical to the invention” during prosecution, foreclosing any later argument that a broader range should be treated as equivalent.
Vertex is in the same position: it won its patent specifically by convincing the examiner that 80% was surprising and different from the rest of its own previously published range, the examiner rejected a 72–88% claim for lack of evidence the surprising result extended that far, and the same examiner then treated even a tenth-of-a-percent variance as the outer edge of “about 80%.” Stretching that claim to cover a product several percentage points away, the court reasoned, “directly conflicts with the patent’s express claim” — and risks the further problem of letting Vertex capture, through litigation, territory it had already disclosed publicly and been refused at the patent office (“ensnar[ing] the prior art”).
What helped Vertex distinguish its invention from the prior art during prosecution is the same record that now limits how far its claims can reach in litigation.
3. Neither recognized equivalence test supports Vertex on the merits
Even setting the claim-language and estoppel problems aside, Vertex could not satisfy either accepted test for equivalence, and the court addressed both independently.
The function-way-result test — whether the accused product does the same work, in the same way, to the same result — was a poor fit on these facts because it would be satisfied by virtually any bioequivalent generic. Vertex’s argument reduced to exactly that: both products use an amorphous ivacaftor dispersion to treat cystic fibrosis.
But bioequivalence alone cannot establish infringement; if it could, selling any generic version of a patented drug would become illegal by definition — the outcome the case law is designed to prevent. For that reason, courts generally treat the insubstantial-differences test as more suitable in the generic-drug context, and the court applied it here as a separate, independently sufficient ground: the intrinsic evidence gave no reason to think a person of ordinary skill would treat 74.257% and 80% as equivalent drug loads, and Vertex’s and Lupin’s different manufacturing processes — wet versus dry granulation — supplied additional, independent evidence against equivalence, not merely background detail about how the products differ.
Vertex’s remaining argument, that similar dissolution profiles show the differences are insubstantial, the court identified as simply the bioequivalence argument restated, and rejected on the same basis.
The Throughline
The opinion’s closing lines make explicit what the whole analysis has been building toward: Vertex is “trying to claim that its patents cover a range of ivacaftor concentrations that it failed to patent.” That is prosecution history estoppel in its most classic form — the statements a patentee made to win a claim can later be used against it to prevent recapturing, through litigation, what an examiner already refused to grant.
Vertex told the examiner 80% was special and surprising; the examiner allowed a term (“about 80%”) that tracked variance in the tenths of a percent; Vertex is now asking a court to read that same term as tracking a generic several percentage points away. The court’s answer was that a
patentee gets “the patent claims that it prosecuted successfully, not the ones that it failed to prove.”
What the Decision Does and Does Not Establish
The opinion should not be read as establishing a general rule about how much variance the word “about” can carry in a pharmaceutical claim, or as holding that a ~74% formulation can never infringe an “about 80%” patent under different facts. Its holding is narrower and tied to this record:
- Where a patentee’s own prosecution history repeatedly treats a specific numerical value as the source of patentability, and a broader range was expressly rejected for lack of evidence that the same effect extended across it, the doctrine of equivalents will not be available to recapture that same rejected range at trial.
- An examiner’s own construction of an approximation term during prosecution — here, tenth-of-a-percent examples for “about 80%” and “about 0.5%” — is intrinsic evidence of ordinary meaning that a court may credit over expert testimony untethered to the specification or file history.
- Bioequivalence between a generic and a branded drug is not, by itself, evidence of patent law equivalence; treating it as such would functionally eliminate the space for lawful generic competition that Hatch-Waxman is designed to preserve.
- A difference in manufacturing process (here, wet versus dry granulation) can independently weigh against a finding of insubstantial differences, separate from any comparison of the drug-load percentages themselves.
The larger lesson for pharmaceutical patent prosecution and litigation is the same one the court states plainly: what a patentee uses to distinguish its invention from the prior art during prosecution can later define — and constrain — what that patent is worth at trial.
Vertex secured a patent by making 80% mean something specific and unexpected. It could not then ask a court to let “about 80%” mean something closer to 70–90% once a generic competitor showed up outside that narrow band.
Reference: Vertex v. Lupin judgment
