The Chemistry Controls the Calendar: Predictive Testing, Design Around Reformulations & the Cost of Racing to Trial in Kaneka Corp. v. Designs for Health

“Waiting was unacceptable to a patent owner watching what it believed was ongoing infringement capturing market share while the evidence ‘matured’ on the laboratory shelf. . . .”

Kaneka Corporation won.

The Japanese manufacturer of reduced coenzyme Q10 had spent years asserting U.S. Patent No. 7,829,080 (“the ’080 patent”) against two American companies selling a competing ubiquinol product without a license. In the summer of 2024, Judge William C. Bryson of the United States Court of Appeals for the Federal Circuit, sitting by designation in the District of Delaware, conducted a bench trial in late May and early June of 2024 and found that the defendants’ original products infringed claims 5 and 15 of the ‘080 patent. The patent was valid. The infringement was proven. Kaneka had its judgment.

The defendants did not appeal. They reformulated instead. Internal documents that later became evidence at trial show how they deliberately limited the amount of reducing agent in their product to ensure the critical chemical conversion at the heart of the infringement case would stop short of the patented threshold. The reformulated products soon reached the market.

Kaneka was not inclined to concede market share to the defendants’ design-around. Its patent was valid, its claim construction adopted, and it believed the reformulation was itself infringing. Or, at least, it would convert to an infringing form sometime within its commercial shelf life. To prove it, Kaneka pressed for an early July 2025 second trial, relying on accelerated stability testing and predictive regression modeling rather than waiting for complete ambient real-time data that would only become available after the products had run their full shelf life. When the defendants sought additional time before trial, Kaneka objected in writing. When the court offered a postponement, Kaneka declined. It raced to verdict on the predictive evidence it had. See Kaneka Corp. v. Designs for Health, Inc. and American River Nutrition LLC, Civil Action No. 21-209-WCB (D. Del. Mar. 23, 2026) (Bryson, J., sitting by designation) (“Kaneka, slip op.”).

It should have waited.

I. The Patent and the First Win

The ’080 patent addresses a fundamental problem in the nutritional supplement industry. Ubiquinol—reduced Coenzyme Q10—is marketed as an antioxidant. It is commercially valuable. It is also chemically unstable. Left to its own devices, it oxidizes back into ubiquinone, the non-reduced form of Coenzyme Q10. Kaneka’s patent covers both an ubiquinol composition and a method of stabilizing it, specifically by including related compounds reduced CoQ9 and/or reduced CoQ11 alongside the ubiquinol. See the ’080 patent, col. 2, ll. 36-41.

The key limitation in both asserted claims requires that the proportion of reduced CoQ10 relative to the total CoQ10 in the composition be not less than 90 percent by weight. Specifically, claim 5 recites a “reduced coenzyme Q10-containing composition” wherein “the proportion of reduced coenzyme Q10 relative to the total amount of coenzyme Q10 is not less than 90 wt %.” The ’080 patent, col. 16, ll. 55-65. Claim 15 recites the same limitation in the context of a method for producing such a composition. Id., col. 18, ll. 4-21. The parties called this the “QH ratio.” It was the only limitation in genuine dispute throughout the second phase of the litigation, and it became the axis around which everything turned.

The defendants, Designs for Health, Inc. and its affiliate American River Nutrition LLC, sold a ubiquinol product under the trademark DuoQuinol. Their formulation used three components: geranylgeraniol (“GG”) as a solvent to dissolve ubiquinone, and ascorbyl palmitate (“AP”) as a reducing agent to convert ubiquinone into ubiquinol. The original DuoQuinol formulation used a 4:2:4 ratio of GG to AP to ubiquinone. Kaneka, slip op. at 3 (citing Trial Tr. vol. 2 at 315:13-16). That formulation produced a QH ratio exceeding 90 percent within the products’ shelf life. The first bench trial established as much, and a finding of direct infringement followed. Under settled Federal Circuit precedent, “[a]n accused product infringes a claim if it embodies each claim element or its equivalent.” TEK Glob., S.R.L. v. Sealant Sys. Int’l, Inc., 920 F.3d 777, 784 (Fed. Cir. 2019). The original formulation of DuoQuinol products did.

In the face of that loss, the defendants made a shrewd decision. They did not appeal. They did not seek a license. They instead went back to the laboratory.

II. The Design-Around

The documentary record on what happened next is unusually candid. Beginning in June 2023, before the first bench trial had even concluded, the defendants were already exploring whether they could reformulate to ensure the QH ratio would never reach 90 percent. Kaneka, slip op. at 4 (citing Plaintiff’s Trial Exhibit (“PTX”) 268). The objective was explicit: reduce the amount of ascorbyl palmitate sufficiently to cap the conversion of ubiquinone into ubiquinol before it crossed the patent’s threshold. Id. (citing PTX 269).

The defendants tested six different component ratios and ultimately settled on 4:1.5:4—the same ratio of GG to ubiquinone as the original, but with AP reduced from 2 parts to 1.5 parts. Id. (citing PTX 270; Defendants’ Trial Exhibit (“DTX”) 489). Less reducing agent meant less reduction. At least in theory, the arithmetic was straightforward.

In October 2023, the defendants pulled their original products from the market and replaced them with the reformulated versions. They had a chemical argument for why the design-around would hold. If a fixed ratio of AP were the only reducing agent in the reformulated products, the maximum QH ratio those products could ever achieve would be 78.3 percent. That fixed quantity of AP would be consumed entirely in the reduction reaction, and once exhausted, no further reduction would be possible. Both parties’ experts agreed that, if AP were the only reducing agent, the theoretical ceiling was 78.3 percent. Kaneka, slip op. at 41-43 (citing Trial Tr. vol. 2 at 412:14-20 (Dr. Myerson); Trial Tr. vol. 4 at 921:21-25 (Dr. Banakar)). And 78.3 percent is comfortably below 90 percent.

It turned out to be more complicated than that.

III. The Chemical Mystery

When Kaneka began testing the reformulated products, something unexpected appeared. The QH ratios were rising above 78.3 percent. Not dramatically, not quickly—but rising. Something in the formulation was continuing to drive reduction even after the AP had been consumed.

Nobody could explain precisely what that something was.

Kaneka’s expert, Dr. Allan Myerson, advanced a theory at trial. The GG component—characterized by the defendants as a passive solvent—was actually reacting. Its concentration in the tested products was declining over time. Dr. Myerson hypothesized that GG, or something GG decomposes into, was acting as a secondary reducing agent, pushing the QH ratio upward even after the AP was gone. His candor about the limits of that hypothesis was striking. “We just don’t know,” he testified, because the mechanism “hasn’t been studied in any level of detail.” Kaneka, slip op. at 43 (quoting Trial Tr. vol. 2 at 416:23-417:3). He could not identify the secondary reducing agent with certainty. “I’m not even sure exactly what the chemistry is of this continued reduction,” Dr. Myerson acknowledged, because “as we said, these ascorbyl palmitates [AP] are all gone.” Id. (quoting Trial Tr. vol. 2 at 456:11-15).

The defendants’ expert, Dr. Umesh Banakar, was unequivocal in response. GG cannot act as a reducing agent. Kaneka, slip op. at 43 (citing Trial Tr. vol. 4 at 924:24-925:5). The defendants offered an alternative explanation for the rising QH ratios—mathematical rather than chemical reduction. They asserted that the product’s CoQ10 was degrading over time, which would reduce the total denominator in the QH calculation and thereby increase the ratio without any additional reduction occurring. Dr. Myerson disputed that explanation by pointing to the absence of any discernible downward trend in total CoQ10 in the ambient testing data. Kaneka, slip op. at 43-44 (citing Trial Tr. vol. 2 at 446:5-8; PTX 226).

Two things were simultaneously true in a practical sense. Nobody at trial could explain the mechanism with confidence, but the data showed the QH ratio continuing to climb beyond what AP alone could produce. The legal question was whether it would climb above 90 percent before the products’ commercial shelf life expired. That question drove the parties’ strategic choices for the second bench trial, and it sent both sides to their predictive models.

IV. Kaneka’s Strategic Choice

Here is where the story takes a critical turn.

Kaneka faced a genuine timing problem. The reformulated products had initially been assigned a 36-month shelf life. Waiting for real-time ambient testing to establish that the QH ratio would cross 90 percent within that timeframe meant waiting roughly three years from the original October 2023 reformulation. Waiting was unacceptable to a patent owner watching what it believed was ongoing infringement capturing market share while the evidence “matured” on the laboratory shelf.

There was a path around the wait. The defendants had conducted “accelerated” stability testing on their products, storing them at elevated temperature and humidity—96 degrees Fahrenheit, 80 to 85 percent relative humidity—to simulate extended real-time storage. Kaneka, slip op. at 36 (citing Trial Tr. vol. 2 at 361:12-20). Their records indicated that one week of accelerated conditions was equivalent to one month at ambient conditions. Id. (citing PTX 226). If that equivalence held, 36 weeks of accelerated testing could substitute for 36 months of real-time data.

Kaneka chose that path. And it pushed hard for an early trial.

Everything that followed flowed from that choice.

When the defendants sought additional time ahead of the scheduled June 2025 trial date, citing the expanded scope of Kaneka’s infringement allegations, Kaneka objected in writing that the defendants’ “delay tactics compound the harm to Kaneka, which continues to suffer from Defendants’ infringement.” Kaneka, slip op. at 14 (quoting D.I. 273 at 2). At a March 2025 status conference, Kaneka’s counsel was direct: “Kaneka’s position is clear that we don’t want any further delay. We want to move forward.” Id. (quoting D.I. 289 at 54:10-11). At no point before trial did Kaneka suggest it needed additional time to complete its long-term shelf-life testing program.

The second bench trial ran from July 21 through July 24, 2025.

The defendants had strategically shortened the shelf life of their reformulated products to 18 months in January 2025. They gambled that the reformulated products would expire before the QH ratio surpassed 90 percent. That gamble came at a potential strategic cost. Products reformulated in October 2023 would reach the end of their new 18-month shelf life in April 2025—three months before the second bench trial. This meant that conducting long-term shelf-life testing to confirm whether those products crossed 90 percent within 18 months was achievable before trial.

Kaneka chose not to wait for its own ambient testing to reach that endpoint before trial.

V. The Evidence at Trial

A plaintiff in a patent infringement case bears the burden of proving infringement “by a preponderance of the evidence.” Advanced Cardiovascular Sys., Inc. v. Scimed Life Sys., Inc., 261 F.3d 1329, 1336 (Fed. Cir. 2001). That burden required Kaneka to show that the reformulated products would satisfy the QH limitation’s 90 percent threshold within their designated shelf life. The defendants did not contest any other patent claim limitation. Kaneka, slip op. at 3 (citing D.I. 352 at 6-9). Everything turned on that single number.

Kaneka arrived at trial with three categories of evidence.

The first was ambient real-time testing by Triclinic Labs in Indiana, conducted under the direction of Dr. Myerson between March and May 2025. Kaneka, slip op. at 33 (citing PTX 159; PTX 225). None of the tested lots—with one notable exception addressed below—showed a QH ratio above 90 percent. The tested lots ranged from roughly 82 to 86 percent under ambient conditions. Kaneka, slip op. at 33-35 (citing D.I. 335 at 3).

The second was the defendants’ own stability data for Lot 35550, a CoQnol 100 lot tracked under both ambient and accelerated conditions beginning in October 2023. Kaneka, slip op. at 36 (citing PTX 226). Under ambient conditions, 18 months of testing produced a peak QH ratio of 87.7 percent—measured at the April 2025 endpoint. Id. The 90 percent threshold was never reached. Under accelerated conditions, the picture was different. The QH ratio exceeded 90 percent after 10 weeks, dropped briefly below 90 percent, then rose above it again and remained elevated through week 36, reaching a peak of 98.5 percent. Id. Kaneka’s argument was that the accelerated results predicted ambient performance on the one-week-equals-one-month conversion. The products would infringe within their shelf life.

The third category was predictive modeling. Kaneka’s expert analyzed the rate of QH ratio increase in the Lot 35550 ambient data and extrapolated that rate to other lots with higher starting QH ratios, projecting that those lots would cross 90 percent within 18 months.

A. The Evidentiary Pillars Fall

Judge Bryson worked through each category with precision. This was not a Daubert exclusion case. It was a burden-of-proof case on the substantive issue of patent infringement. His analysis represents a methodical accounting of what happens when predictive evidence is asked to carry a burden that direct evidence would carry without effort. The case ultimately draws a bright line between possible infringement and provable infringement.

The accelerated testing failed first. The problem was not that accelerated stability testing is inherently unreliable. It is a standard pharmaceutical development tool that has earned its place in the science. The problem was the specific data presented in this case. The accelerated and ambient results converged early and diverged significantly over time.

At five months ambient and five weeks accelerated—the timepoints corresponding under the one-week-equals-one-month equivalence—the QH ratios were closely aligned—83.5 percent ambient, 83.2 percent accelerated. Kaneka, slip op. at 39 (citing PTX 226). So far, so good. But that alignment gave way entirely at extended timeframes. At 10 weeks accelerated—which should correspond to 10 months ambient under the defendants’ own equivalence formula—the accelerated QH ratio was 90.1 percent while the ambient measurement was 85.8 percent—a gap of more than four percentage points. Id. And at 18 weeks accelerated versus 18 months ambient, the gap widened further—9 2.8 percent versus 87.7 percent. Id. The divergence was systematic and ran in exactly the wrong direction for Kaneka’s theory.

Triclinic’s independent testing of Lot 35550 in April and May 2025, conducted roughly contemporaneously with the defendants’ 18-month ambient measurements, confirmed the ambient picture. Triclinic found QH ratios of 85.6, 85.93, and 85.2 percent. Kaneka, slip op. at 39 (citing D.I. 335 at 3; PTX 225). Consistent with the defendants’ ambient data. Inconsistent with the accelerated data.

It was beginning to appear that the ambient data was telling the true story.

At trial, Dr. Myerson highlighted the early convergence but did not address the subsequent divergence. He acknowledged that “the relationship between ambient testing and accelerated testing varies depending on the product being tested,” and that “[s]ome products are very temperature-sensitive.” Kaneka, slip op. at 39-40 (quoting Trial Tr. vol. 2 at 435:7-21). He offered no specific evidence regarding the temperature sensitivity of the defendants’ products that would validate the divergence at extended timeframes. The defendants’ belief that one week of accelerated testing was generally indicative of one month at ambient conditions did not establish a validated correlation for these specific products over extended periods. Kaneka, slip op. at 39-40.

The predictive modeling failed on grounds that were damaging in part because Kaneka’s own expert acknowledged the underlying weaknesses. Dr. Myerson performed two analyses of the Lot 35550 ambient data.

The first was a least-squares linear regression across all 42 data points, producing a slope indicating a 0.0158 percent per day increase in the QH ratio. He acknowledged on cross-examination that the data was “very scattered” and the correlation coefficient was only 0.7879, well below the threshold of analytical confidence. Kaneka, slip op. at 46 (quoting Trial Tr. vol. 2 at 450:17-19). He further testified that he had not relied on this analysis in his expert report because “I decided that looking at it, that the fit wasn’t very good.” Id. (quoting Trial Tr. vol. 2 at 490:10-14).

The second analysis, the one he did rely on, used only two data points—the first measurement and the last. He drew a straight line between them and used the slope of that line. His rationale was that it “seemed to be a better way to do the analysis. It kind of eliminated all the scatter.” Kaneka, slip op. at 49 (quoting Trial Tr. vol. 2 at 490:17-21).

The defendants’ expert was withering in response. The first requirement of any statistical modeling is to “use all data available,” and “picking and choosing” two endpoints while ignoring all intermediate measurements “is against the principles and practices” of statistical analysis. Kaneka, slip op. at 50 (quoting Trial Tr. vol. 3 at 863:3-12; Trial Tr. vol. 4 at 943:8-16). Statistical literature in evidence established that shelf-life predictions become increasingly error-prone when based on fewer than eight data points. Kaneka, slip op. at 49 (citing DTX 479 at 468). Dr. Myerson had relied on only two. A line between two points eliminates scatter by definition. It discloses nothing about the behavior of the data in between.

Judge Bryson credited the critique. The court identified a further problem the data itself exposed. Dr. Myerson’s model predicted that Lot 35950 would infringe within 18 months because of its starting QH ratio. Kaneka’s own Triclinic testing showed Lot 35950 did not reach 90 percent at 18 months. The model had failed a test the evidentiary record could itself run. Kaneka, slip op. at 47-48.

The deepest problem ran beneath all of it. The QH ratio in the reformulated products was exceeding 78.3 percent—the theoretical maximum if AP were the only reducing agent. Something else was driving reduction. Nobody could identify what, with confidence. Without understanding the mechanism, there was no reliable basis for a model predicting how far that secondary reduction would go or when it would cross any particular threshold. “Given the fundamental lack of clarity regarding the chemistry resulting in the amount of reduction of ubiquinone leading to a QH ratio in excess of 78.3 percent,” Judge Bryson wrote, “the support for any specific model, including a model assuming a constant rate of reduction, is speculative.” Kaneka, slip op. at 44.

B. The Anomalous Lot

There was one lot where Kaneka’s case succeeded.

Lot 45818 was a CoQnol 200 lot that Triclinic tested in March 2025 under ambient conditions. All three samples registered QH ratios above 90 percent — 94.3, 93.6, and 93.9 percent. Kaneka, slip op. at 34 (citing PTX 159 at 16). The patent’s QH limitation was met. The defendants conceded infringement at the post-trial hearing.

The defendants’ explanation was also uncontested. A mechanical failure in the softgel melter tank had kept the DuoQuinol blend for Lot 45818 under pressurized, heated conditions for more than eight weeks—against the 5.5 days specified by the defendants’ protocol. Kaneka, slip op. at 34 (citing D.I. 352 at 6; Trial Tr. vol. 2 at 367:2-25). A production employee was disciplined for not recognizing the deviation before the lot was released for sale in December 2024. Kaneka, slip op. at 34-35 (citing DTX 455 at 2). When the problem was identified in March 2025, the defendants placed remaining inventory on hold and destroyed it in April 2025, less than a month after discovery. Kaneka, slip op. at 73 (citing DTX 464).

The court found infringement proven for Lot 45818. But it also found Lot 45818 entirely unrepresentative of the reformulated product line. The equipment failure had produced reaction conditions dramatically different from anything occurring in ordinary production. The higher QH ratio was the predictable result of a reaction running for eight weeks instead of 5.5 days. It was not evidence that normal lots would reach 90 percent within their shelf life.

Lot 45818 gave Kaneka an isolated finding of infringement. It did not give Kaneka the injunction, the broad damages award, or the precedent for ongoing liability that the second bench trial was intended to produce.

C. The Attempted Do-Over

Kaneka recognized the problem quickly.

On August 26, 2025—the same day Judge Bryson denied the preliminary injunction based on the second trial’s record—Kaneka began shipping product samples to Triclinic and its own laboratories for further testing. According to Kaneka’s own records, shipments went out on August 27, September 6, and September 15. Kaneka, slip op. at 13 (citing D.I. 426 at paras. 5, 8, 10). The results, Kaneka maintained, showed reformulated product lots crossing the 90 percent threshold within their shelf life. Kaneka sought to put those results before the court through three post-trial motions: (1) a motion to reopen the record to admit the new test data; (2) a motion for judicial notice of four publicly available documents bearing on whether GG or its derivatives could act as a secondary reducing agent; and (3) a motion for judgment on partial findings under Federal Rule of Civil Procedure 52(c).

The court denied all three.

1. Reopening a closed record

The decision to reopen the record after a bench trial “to submit additional proof is addressed to [the court’s] sound discretion.” Zenith Radio Corp. v. Hazeltine Research, Inc., 401 U.S. 321, 331 (1971). The court, however, “must avoid perpetrating any type of injustice in so doing.” Gibson v. Mayor & Council of City of Wilmington, 355 F.3d 215, 229 (3d Cir. 2004). The Third Circuit has identified three factors governing that analysis: (1) the burden on the parties and their witnesses; (2) the undue prejudice that may result from admitting or not admitting the new evidence; and (3) considerations of judicial economy. Rochez Bros., Inc. v. Rhoades, 527 F.2d 891, 894 n.6 (3d Cir. 1975).

All three factors weighed against Kaneka. On burden, the new test results had not been disclosed to the defendants, some were obtained using different protocols from those used at trial, and others were produced by Kaneka’s own Quality Assurance Team using an undisclosed procedure. Kaneka, slip op. at 8. If the record were reopened, the defendants would require new discovery, new expert reports, new testimony, and a new opportunity for cross-examination. What Kaneka framed as supplemental evidence would amount, in all likelihood, to a new trial on infringement. Kaneka, slip op. at 10. On prejudice, the defendants correctly observed they had been given no opportunity to challenge test results that “were not the subject of fact or expert discovery” and had not been subject to cross-examination at trial. Kaneka, slip op. at 7 (quoting D.I. 448 at 1). On judicial economy, reopening the record would have required a third round of pretrial proceedings in a case already more than five years old. Kaneka, slip op. at 17-18.

The analysis of justification was harder on Kaneka still. It argued it could not have submitted the new evidence earlier because the evidence did not exist at the time of trial. Judge Bryson was unforgiving. Kaneka had begun shipping samples for testing within days of the preliminary injunction denial and had not explained why those same products could not have been submitted for testing before the July 2025 trial. Kaneka, slip op. at 13. The court had offered the parties an opportunity to postpone the trial. Kaneka had declined. It had never suggested additional testing was anticipated or necessary.

On that record, Kaneka faced waiver-like consequences.

“A party cannot request relief from the court for its own errors in judgment, when the strategy of litigation was based on conscious and informed choices.” Bell Tel. Lab’ys, Inc. v. Hughes Aircraft Co., 73 F.R.D. 16, 23 (D. Del. 1976). In effect, Bryson wrote, “Kaneka is asking for a do-over.” Kaneka, slip op. at 6. That was one thing the court declined to provide.

2. Judicial notice.

Kaneka’s parallel motion asked the court to take judicial notice of four documents: two patents owned by defendant ARN (U.S. Patent Nos. 7,989,006 and 9,949,938), a white paper posted on ARN’s own website discussing its “GG-Gold” product, and a 2024 journal article on the biological properties of GG-related compounds. Kaneka, slip op. at 19.

Federal Rule of Evidence 201(b) permits judicial notice of facts that are “not subject to reasonable dispute” because they are “generally known” within the court’s jurisdiction or “can be accurately and readily determined from sources whose accuracy cannot be questioned.” Fed. R. Evid. 201(b). The problem was that the scientific propositions for which Kaneka sought notice were squarely in dispute. Dr. Myerson had offered a tentative version of the GG-reduction theory at trial. Dr. Banakar had flatly rejected it. The Federal Circuit has made clear that judicial notice of disputed scientific premises is inappropriate because doing so “would displace the required fact finding with what amounts to a rule of law.” Amgen Inc. v. Sanofi, 872 F.3d 1367, 1378 (Fed. Cir. 2017). This was not a case for a different result.

There was a more fundamental problem. All four documents had been publicly available before the second bench trial. The ARN patents had issued in 2011 and 2018, more than seven years before trial. Kaneka had no explanation for why they were not offered into evidence when the trial occurred. Rule 201(d) permits a court to take judicial notice at “any stage of the proceeding,” but that power “is limited by ‘the requirement of fairness.’” Colonial Leasing Co. of New England v. Logistics Control Grp. Int’l, 762 F.2d 454, 461 (5th Cir. 1985). Granting Kaneka’s motion for documents it could have used at trial for scientific propositions that were actively contested would have required reopening discovery and reconvening trial. “To press the doctrine of judicial notice to the extent attempted in this case and to do that retroactively after the case had been submitted,” wrote Justice Cardozo in a formulation that Judge Bryson reached back nearly ninety years to apply, “would be to turn the doctrine into a pretext for dispensing with a trial.” Ohio Bell Tel. Co. v. Pub. Utils. Comm’n of Ohio, 301 U.S. 292, 302 (1937). The motion was denied.

3. Rule 52(c).

Federal Rule of Civil Procedure 52(c) permits a court in a bench trial to “enter judgment against [a] party on a claim or defense that, under the controlling law, can be maintained or defeated only with a favorable finding on that issue” after the party “has been fully heard on that issue.” Fed. R. Civ. P. 52(c). The court “weighs the evidence and assesses the credibility of witnesses” without viewing the record in the light most favorable to the nonmoving party. United Techs. Corp. v. Chromalloy Gas Turbine Corp., 105 F. Supp. 2d 346, 355-56 (D. Del. 2000), aff’d, 30 F. App’x 980 (Fed. Cir. 2002). Kaneka’s Rule 52(c) motion rested substantially on the post-trial test results and the judicial notice documents. Both had been excluded. What remained in the record was insufficient, for the same reasons that had led Judge Bryson to deny the preliminary injunction in August. The motion was denied. Kaneka, slip op. at 28-31.

VI. The “Inexorable” Damages Problem

With the exception of Lot 45818 and the original products at issue in the first bench trial, Kaneka failed to prove infringement in the second bench trial. For the first bench trial original products sold between October 1, 2020 and September 30, 2023, the court calculated Kaneka’s lost profits at $2,832,501, a figure derived from the volume of the defendants’ infringing sales multiplied by Kaneka’s per-kilogram profit margin. That amount was reduced further by the profits Kaneka earned when defendants purchased a large quantity of ubiquinol from Kaneka to seed their own product development—"a purchase that would not have occurred but for the defendants’ infringement.” Kaneka, slip op. at 57 (citing Trial Tr. vol. 2 at 538:21-24, 548:19-549:11).

Under settled Federal Circuit precedent, a patent owner establishes entitlement to lost profits by showing: (1) demand for the patented product; (2) the absence of acceptable non-infringing substitutes; (3) manufacturing and marketing capability to exploit that demand; and (4) the amount of profit the patentee would have made. Rite-Hite Corp. v. Kelley Co., 56 F.3d 1538, 1545 (Fed. Cir. 1995) (en banc) (citing Panduit Corp. v. Stahlin Bros. Fibre Works, Inc., 575 F.2d 1152, 1156 (6th Cir. 1978)). The court found each factor met with respect to the original products. Kaneka and the defendants were the only suppliers of ubiquinol in the American market during the infringement period, and where that is true “it is reasonable to infer that the patent owner would have made the sales made by the infringers.” Del Mar Avionics, Inc. v. Quinton Instrument Co., 836 F.2d 1320, 1327 (Fed. Cir. 1987).

Translating that figure into a damages award for Kaneka Japan presented its own complications. Kaneka owns the ’080 patent. Kaneka North America (“KNA”), its United States subsidiary, is the entity that actually sells ubiquinol products in this country under a nonexclusive license. A nonexclusive licensee lacks constitutional standing to sue for patent infringement. Schreiber Foods, Inc. v. Beatrice Cheese, Inc., 402 F.3d 1198, 1203 (Fed. Cir. 2005). For Kaneka to recover KNA’s lost profits, Kaneka had to show those profits would “inexorably flow” from KNA through an intermediate holding company (Kaneka Americas Holdings, or “KAH”) to Kaneka itself.

The Federal Circuit has acknowledged the potential availability of this “inexorable flow” theory of recovery while declining to define precisely what evidence establishes it. Mars, Inc. v. Coin Acceptors, Inc., 527 F.3d 1359, 1367 (Fed. Cir. 2008). A corporate parent relationship alone is not enough. Poly-Am., L.P. v. GSE Lining Tech., Inc., 383 F.3d 1303, 1311 (Fed. Cir. 2004). As recently as March 2025, the Federal Circuit applied that rule in rejecting a jury’s lost profits award as unsupported by substantial evidence where the plaintiff had not shown more than corporate ownership. See Roland Corp. v. inMusic Brands, Inc., No. 2023-1327, 2025 WL 926703, at *12 (Fed. Cir. Mar. 27, 2025) (nonprecedential).

The court found the first leg of the flow—KNA to KAH—established. KNA has no bank account of its own; all revenues flow directly into accounts KAH maintains for both entities. Kaneka, slip op. at 63-64 (citing Trial Tr. vol. 1 at 164:21-165:3, 167:24-168:3). Both parties’ damages experts characterized the nominal dividend arrangement between KNA and KAH as a “paper exercise.” Id. (citing Trial Tr. vol. 3 at 620:24-621:13; Trial Tr. vol. 4 at 1021:7-11). The second leg—KAH to Kaneka—presented more difficulty. During the three years of the original product infringement period, KAH suffered large net losses in two of three relevant fiscal years and paid no dividends to Kaneka in those years. The dividend policy Kaneka relied upon to establish flow did not guarantee that profits earned in one year would survive intervening losses and reach Kaneka in later years. Kaneka, slip op. at 68-71. The court ordered an accounting proceeding and directed the parties to meet and confer regarding next steps by April 10, 2026. Id. at 75.

The permanent injunction claim required no extended analysis. Having failed to prove infringement as to the reformulated products generally, Kaneka could not demonstrate the irreparable injury that forms the first of the four factors a patentee must satisfy to obtain permanent injunctive relief. eBay Inc. v. MercExchange, L.L.C., 547 U.S. 388, 391 (2006); Jazz Pharms., Inc. v. Avadel CNS Pharms., LLC, 136 F.4th 1075, 1082 (Fed. Cir. 2025). As for Lot 45818, monetary damages were adequate and no further sales were anticipated. Injunctive relief based on a single anomalous, unrepresentative lot would impermissibly expand the scope of relief beyond “the particular adjudicated infringing activity.” Jazz Pharms., 136 F.4th at 1082 (citing Aspex Eyewear, Inc. v. Marchon Eyewear, Inc., 672 F.3d 1335, 1344 (Fed. Cir. 2012)). The request was denied.

VII. Unforgiving Calendars and Indeterminate Chemistries

Strip away the procedural complexity and the CoQ10 biochemistry, and this case’s lesson becomes clear.

Kaneka illustrates a recurring but underappreciated principle in patent litigation. When infringement turns on time-dependent chemical behavior, the timing of trial can be outcome-determinative. A plaintiff who proceeds on prediction rather than proof assumes the risk that the science will not cooperate on the court’s schedule.

At the end of the day, Kaneka gambled on an infringement theory that depended entirely on a chemical transformation occurring over time inside a softgel capsule. The 90 percent QH threshold would either be reached before the product’s expiration date or it would not. Nobody could know for sure until the laboratory observations ran their course.

That is not a litigation problem. It is a laboratory problem. Laboratory problems run on laboratory timelines.

Kaneka had the option to wait. Its competitor had shortened the shelf life of the reformulated products to 18 months in January 2025. Products reformulated in October 2023 would reach the 18-month mark in April 2025 — three months before the July 2025 trial. The post-trial testing that Kaneka desperately sought to put before the court was generated beginning in August 2025. A few weeks of earlier diligence, and those results would have been admitted trial evidence.

Reliance on the accelerated testing and predictive modeling strategy was not unreasonable in concept. Both are recognized tools in pharmaceutical stability analysis. The problem was that the accelerated and ambient results diverged materially at extended timeframes. The regression model rested on two endpoints plucked from scattered data. The underlying secondary reduction chemistry was unexplained, which undermined every model’s claim to reliability. Each weakness was foreseeable before trial. Each was exploited at trial.

The lesson is not that predictive evidence is insufficient. It is that predictive evidence must be validated, product-specific, and internally consistent—especially when it is offered as a substitute for real-world data.

Kaneka’s miscalculation likely became visible during trial itself. When cross-examination exposed the 0.79 correlation coefficient, the two-point methodology, and the failure of the Lot 35950 prediction, the gap between what the predictive evidence could support and what Kaneka needed to prove became unmistakable. The post-trial motions were an after-the-fact response to that realization. They came too late.

Had Kaneka postponed the trial until after the completion of full ambient shelf-life studies, the predictive testing would have served a different and far stronger function—corroboration. Ambient data confirming that multiple lots crossed 90 percent within 18 months, supported by accelerated testing and regression modeling consistent with both, would have presented a convergent evidentiary structure genuinely difficult to attack. The predictive evidence, discredited on its own, would have been confirmatory evidence within an evidentiary portfolio that did not depend upon it. Instead, it was asked to carry the entire weight of the infringement case. It was not strong enough to do so.

“A party cannot request relief from the court for its own errors in judgment, when the strategy of litigation was based on conscious and informed choices.” Bell Tel. Lab’ys, 73 F.R.D. at 23. Here, Kaneka resisted postponement, declining the court’s own offer to delay trial, and never once suggesting its testing program was incomplete. The scheduling record made that observation nearly impossible to contest.

VIII. Conclusion: We Have Seen This Movie Before

Pharmaceutical and chemical patent practitioners who have litigated at the intersection of legal theories and laboratory timelines will recognize the pattern.

The paroxetine hydrochloride hemihydrate litigation of the early 2000s offers an instructive parallel. SmithKline Beecham held patents on a specific crystalline form — Form II — of paroxetine hydrochloride hemihydrate, the active ingredient in Paxil. See SmithKline Beecham Corp. v. Apotex Corp., 403 F.3d 1331 (Fed. Cir. 2005). The infringement theory against generic manufacturers rested in part on a seeding and conversion mechanism: Form II crystals, once present in a manufacturing environment, would spontaneously nucleate and convert other crystalline forms into the thermodynamically stable, patented Form II. The theory was chemically sound in principle.

The rate of conversion was slower and more variable than the litigation required. Establishing that sufficient infringing Form II was present in a generic manufacturer’s product to constitute infringement became a contest of analytical chemistry and threshold arguments—the same kind of contest that Judge Bryson conducted over the QH ratio in Kaneka. Although infringement by crystalline conversion was ultimately shown, the conversion theory created another problem.

On the validity side, the same conversion dynamics created an anticipation problem. Prior art batches of paroxetine had unsuspectingly produced Form II through exposure to ambient conditions during manufacture, and the question of whether that prior art therefore disclosed the patented form before the patent’s priority date became a fight about the timing of the conversion. The legal theory and the laboratory reality were traveling at different speeds and toward different consequences for the litigation. The consequences ran in both directions simultaneously—in favor of the infringement case and against the validity case.

In Kaneka, the consequences ran in one direction only—against Kaneka. There was no validity challenge during the second bench trial. The design-around was deliberate and documented from the beginning. The evidentiary challenge it created was real and foreseeable. The solution required only a little strategic patience. The litigation strategy chose speed instead.

When the infringement theory depends on a transformation that takes months or years to complete, the trial date is not a scheduling question. It is a chemistry question. The calendar should follow the chemistry.

In Kaneka, it did not. And the record of what followed speaks for itself—a two-trial case stretching past five years, a $2.8 million lost profits calculation whose full translation to the Japanese parent remains unresolved in accounting proceedings, and a post-trial do-over attempt dispatched with a nod to both an unconvincing accelerated product testing protocol and an unnecessarily accelerated trial schedule.

Postponing a five-year battle only a few months longer might have changed everything.