Phenytoin II: Limits on the CAT’s “confirmation bias” findings and fairness requirements when remaking CMA infringement decisions

Case: Pfizer v The Competition and Markets Authority
Citation: [2026] EWCA Civ 765 (Court of Appeal (Civil Division), Green LJ, Snowden LJ, Zacaroli LJ)
Date: 19 June 2026
Appeal from: Competition Appeal Tribunal, [2024] CAT 65

1. Introduction

This is the Court of Appeal’s second judgment in the long-running “phenytoin” litigation concerning alleged unfair/excessive pricing under section 18 Competition Act 1998. The Competition and Markets Authority (“CMA”) found that Pfizer (upstream manufacture) and Flynn (downstream distribution) abused dominant positions by imposing unfair selling prices for phenytoin sodium capsules over September 2012 to December 2016, following a de-branding and exclusive distribution arrangement that removed the product from the branded price control environment (PPRS) and into the generic reimbursement framework.

The Competition Appeal Tribunal (“CAT”) in [2024] CAT 65 did two things: (i) it set aside the CMA’s 2022 decision on multiple grounds (including alleged confirmation bias and legal/methodological errors), and then (ii) it remade the infringement decision itself under Schedule 8 CA 1998, again finding abuse (save for one product strength for Pfizer) and largely affirming penalties.

Before the Court of Appeal, the CMA primarily challenged the CAT’s decision to set aside the CMA’s decision. Pfizer and Flynn primarily attacked the CAT’s remade decision as procedurally unfair. The CMA ultimately accepted there was force in the procedural unfairness criticisms of the remade decision.

2. Summary of the Judgment

• The CAT was wrong to set aside the CMA decision. The Court held that the CAT materially misread/mischaracterised the CMA’s reasoning and findings, including by wrongly inferring that the CMA treated the Pfizer/Flynn agreement as an uncharged Chapter I cartel/joint dominance abuse and by wrongly finding the CMA adopted a per se “anything above Cost Plus is unlawful” approach.
• The CAT’s finding of “confirmation bias” failed. The CAT treated perceived analytical errors as proof of bias without applying the established bias/predetermination test and without a fair reading of the CMA decision.
• Flynn ROCE: the CAT’s critique was itself wrong/irrational. In assessing capital employed for ROCE, the CAT wrongly used total expenditure on product acquisition/distribution as “capital employed”; that is operating cost, not the capital required to run the business (working capital/stock/cashflow requirements).
• The CAT’s remade decision was procedurally unfair. It introduced novel constructs (notably an “ultimate consumer” benchmark) and made determinative evaluative findings without giving parties a fair opportunity to address the new theory and factual/economic premises.
• Disposition: The Court granted the CMA permission to appeal, allowed the CMA appeal, set aside the CAT judgment in its entirety, and allowed Pfizer/Flynn appeals against the remade decision. Consequential issues (including reinstatement of the CMA decision and penalties) were left for later submissions and a separate ruling.

3. Analysis

3.1 The legal framework: unfair pricing under section 18 CA 1998

The Court restated that “fairness” is the statutory touchstone under section 18(2)(a) CA 1998, informed (not bound) by EU law. The core reference point remains Case C-27/76 United Brands v Commission EU:C:1978:22, especially its concept that a price may be unfair where it bears no reasonable relation to the economic value of the product.

Building on its earlier synthesis in Phenytoin I ([2020] EWCA Civ 339), the Court reaffirmed:

• There is no single mandatory methodology; authorities have a “margin of manoeuvre/appreciation”.
• Cost Plus is a legitimate proxy; the inquiry does not require proof of a fully specified “workably competitive market” benchmark in every case.
• Prices above Cost Plus are not per se unlawful; if an undertaking advances alternative methods/comparables, the authority must fairly evaluate them.

3.2 Precedents cited and how they shaped the Court’s approach

(a) United Brands and Phenytoin I: flexible evidential routes; Cost Plus as proxy

The Court treated United Brands as anchoring the two-limb logic (excessiveness and unfairness) while accepting modern practice may combine multiple evidential strands. It relied on Phenytoin I’s summary (including that evidence is generally about weight, not admissibility).

(b) Justin Le Patourel v BT Plc [2024] CAT 76 and [2025] EWCA Civ 1061: “cleaner” two-stage structure

The Court endorsed the CAT’s “simplified” structure in Le Patourel: treat Limb 1 (benchmark and significant/persistent excess) as comparatively more “linear”, and evaluate economic value/justifications in Limb 2. The Court’s endorsement matters in this case because the CAT below accused the CMA of treating any price above Cost Plus as unlawful; the Court’s reasoning shows that (properly applied) a Cost Plus-based Limb 1 does not entail any per se rule and can remain consistent with a broad Limb 2 fairness assessment.

(c) Cinven Capital Management and others v CMA [2025] EWCA Civ 578: workable competition is evidential, not mandatory

Cinven was used to rebut the idea that regulators must find “real world competitive market” comparators in every case. The Court reiterated that Cost Plus can itself be a sufficient evidential proxy for fairness, even though it is applied within a non-competitive (dominant) market. This directly undercut key CAT criticisms that the CMA unlawfully relied on Cost Plus, or had to draw a precise “line” between lawful and unlawful “producer surplus” in every case.

(d) Bias and predetermination: Magill v Porter [2001] UKHL 67, BASCO, and R (QA) v Secretary of State for Foreign, Commonwealth and Development Affairs [2024] EWHC 3064 (Admin)

The Court held that the CAT’s “confirmation bias” finding collapsed the distinction between (i) an arguable analytical error and (ii) legally disqualifying bias. The benchmark remained the familiar fair-minded and informed observer test from Magill v Porter. The Court also invoked British Academy of Songwriters, Composers and Authors v Secretary of State for Business, Innovation and Skills [2015] EWHC 1723 (Admin) for the proposition that error is not synonymous with bias. The CAT’s reasoning was criticised for not articulating the bias test and for effectively treating disagreements over economic assessment as proof of closed-mindedness.

(e) Intent evidence and abuse: Case C-549/10P Tomra et ors v Commission and Cinven

In rejecting the CAT’s view that intention/strategy was irrelevant, the Court relied on Tomra et ors v Commission: subjective motives may be a relevant factual element in assessing conduct, even though intent is not a legal precondition for abuse. This supported the CMA’s reliance on internal documents evidencing a deliberate strategy to exploit market conditions.

(f) CAT powers to remake decisions: Imperial Tobacco Group v OFT [2011] CAT 41

Pfizer argued that, under Imperial Tobacco, the CAT could not remake an infringement decision on a materially different factual/legal basis than the CMA’s pleaded case. The Court distinguished Imperial Tobacco on statutory and procedural posture and held that, once a decision is set aside, Schedule 8(3) confers broad remedial powers including to make “any other decision which the CMA could itself have made.”

However, the Court’s key constraint was not jurisdictional but procedural fairness: the broader the departure from the original decision and the more “root and branch” the defects found, the stronger the obligation to pause, hear submissions on next steps, and give proper notice and opportunity to respond to new theories and evidential bases.

3.3 The Court’s legal reasoning on the core appeal issues

(1) The “prior illegality” error: profit-sharing language did not imply an uncharged cartel/joint dominance case

The CAT had held that the CMA’s description of Pfizer and Flynn “sharing profits” necessarily implied a finding of Chapter I illegality or joint dominance abuse, and that this infected the CMA’s analysis (including how Flynn’s costs were assessed). The Court rejected this as a misreading:

• Commercial arrangements can have a joint profit purpose without being cartels.
• The CMA used the agreement as part of the factual matrix explaining how market power was exploited; it did not “overload” the case with unnecessary additional infringements.
• Crucially, the Court held that the CMA did in fact incorporate Pfizer’s input price into Flynn’s cost calculations; the CAT’s contrary suggestion was wrong.

(2) The alleged “per se Cost Plus” approach: the CAT inferred a hidden premise contrary to the Decision’s express text

The CAT concluded the CMA treated prices above Cost Plus as automatically unlawful and failed to “draw the line” between legitimate and illegitimate profit. The Court considered that conclusion unsustainable, pointing to the Decision’s explicit statement that Cost Plus does not set a maximum price and that pricing above Cost Plus can be lawful, together with the extensive treatment of justifications and comparators in the Decision.

The Court also rejected the CAT’s asserted requirement that the CMA must always identify a precise “line” above Cost Plus in every case: that assumes there is always some justified increment, which is inconsistent with the case law.

(3) Confirmation bias: the CAT treated disagreement as disqualification

The Court’s critique was methodological: the CAT upheld most procedural complaints against the CMA, yet then made an extreme finding that the CMA displayed confirmation bias and a “single-minded desire to bring the case home.” The Court held that:

• The CAT did not define “confirmation bias” or apply the bias/predetermination tests.
• The CAT’s bias reasons rested on earlier misreadings (prior illegality; per se Cost Plus) and on the incorrect claim that comparators were disregarded.
• Even if an authority errs on complex economics, that does not equate to bias; bias requires a closed mind or unfair selective evidence handling meeting the legal test.

(4) Flynn ROCE: “capital employed” is not total historic expenditure

One of the judgment’s clearest points of principle is the correction of the CAT’s ROCE analysis. The Court held the CAT’s approach was irrational: the relevant capital for ROCE is the capital required to operate the distribution activity (working capital, stock holding, cashflow timing), not the aggregate amounts spent purchasing and distributing product over years. The Court noted that even Flynn accepted the CAT was wrong on this.

The Court further held the CAT misread the CMA’s WACC reasoning: the Decision relied on multiple evidential anchors and also product-specific risk analysis, and the CMA’s expert evidence before the CAT reinforced the WACC range.

(5) Comparators (tablets; other AEDs): the CAT criticised the CMA without engaging with the Decision’s granular analysis

The Court’s theme recurred: the CAT accused the CMA of “binary” rejection of comparators and the DT, but the Decision contained a lengthy, chronological, market-structure-based analysis of whether tablet pricing reflected market power, as required by the earlier remittal. The CAT did not meaningfully grapple with that reasoning.

The Court also noted an internal inconsistency: in its own remade decision the CAT ultimately gave comparators little or no weight, yet used their supposed importance to condemn the CMA and to infer bias.

(6) EU prices: relevant as “red flag” and viability cross-check, not an unlawful discrimination theory

The Court rejected the CAT’s suggestion that the CMA treated selective pricing across jurisdictions as inherently unlawful discrimination. The CMA’s EU evidence served as supportive context and a check against viability claims: same product, same production, materially lower prices elsewhere (often under regulatory constraints), and even where increases were permitted (e.g., Sweden), they were far below the UK levels.

(7) Price evolution and exploitation of market power: relevant, but not a shortcut

The Court held that dramatic and sustained price increases are not the legal test for unfairness, but are relevant context and may be probative when assessed alongside costs, value, and justifications—especially where the increases are far beyond anything needed for viability. Similarly, intention/strategy evidence is admissible in assessing abuse (per Tomra et ors v Commission), even though intent is not a legal element.

3.4 The remade CAT decision: jurisdiction exists, but the process was unfair

The Court held the CAT did have jurisdiction under Schedule 8(3) CA 1998 to make a new decision after setting aside the CMA decision, distinguishing Imperial Tobacco. The real problem was fairness: the CAT remade the case through a process that deprived parties of proper notice and opportunity to address decisive new constructs and assumptions.

The Court identified multiple fairness failures, notably:

• the CAT’s adoption of a novel “ultimate consumer” benchmark (a hybrid of patient/doctor/payer features) without submissions;
• the introduction of controversial propositions about value, continuity of supply, and portfolio/extraneous costs without evidential and legal ventilation;
• explicit acknowledgements by the CAT that it lacked data/submissions on certain key calculations, and its attempt to “solve” this by making assumptions said to favour the undertakings;
• the quasi-penal nature of infringement/penalty findings, reinforcing the need for statement-of-case style procedural protections when a tribunal effectively becomes the primary decision-maker.

4. Impact

• Merits appeals: fidelity to the decision under appeal. The judgment strengthens the principle that the CAT must start from, and accurately engage with, the CMA’s actual reasoning. Mischaracterisation of a regulator’s decision can constitute an error of law and can unravel a merits judgment.
• Bias allegations in competition enforcement. “Confirmation bias” cannot be inferred from alleged analytical flaws without applying the orthodox tests for bias/predetermination and without identifying a legally sufficient procedural or evidential unfairness.
• ROCE methodology in “asset-light” distribution contexts. The Court clarifies that “capital employed” for ROCE is not cumulative operating spend; it is the capital required to run the activity (working capital/stock/cashflow). This is likely to influence future excessive pricing cases involving wholesalers/distributors and IP-light businesses.
• Remaking powers under Schedule 8: fairness constraints. While the CAT’s remedial jurisdiction is broad after setting aside a decision, the tribunal must not “remake” on novel theories without procedural safeguards—potentially including staged procedure, targeted further submissions, and clear notice of provisional adverse reasoning.

5. Complex concepts simplified

• Dominance: market power enabling an undertaking to behave to an appreciable extent independently of competitive pressures.
• Unfair/excessive pricing (section 18(2)(a)): charging prices that are “unfair”, often analysed by asking whether the price has a reasonable relation to the product’s economic value.
• Cost Plus: a benchmark built from (i) the costs of supplying the product plus (ii) a reasonable rate of return (“Plus”). It is a proxy for what might be expected under workable competition.
• ROS vs ROCE:
  • ROS (return on sales): profit as a percentage of revenue.
  • ROCE (return on capital employed): profit as a percentage of the capital needed to run the activity.
• WACC: weighted average cost of capital; an estimate of the return investors require to fund an undertaking (mix of debt/equity), often used as a benchmark for a reasonable return under ROCE.
• Comparators: prices/returns in other products/markets used to test whether the challenged price looks fair; comparator weight depends on product similarity and whether the comparator market is competitively “clean”.
• Schedule 8 remaking power: after setting aside a CMA decision, the CAT can (in principle) make a decision the CMA could have made—but must do so fairly, especially if changing the theory or evidential basis.

6. Conclusion

The Court of Appeal’s decision is a strong appellate correction of a specialist tribunal’s approach where the tribunal (i) attributed to the regulator positions the regulator did not take, (ii) converted disagreements about economic reasoning into findings of procedural disqualification (“confirmation bias”), and (iii) attempted to “remake” a penal-type infringement decision on novel foundations without a fair process.

The lasting significance lies less in any new substantive test for unfair pricing (the Court reaffirmed existing flexibility) and more in the judgment’s procedural and methodological discipline: accurate reading of administrative decisions, proper legal framing of bias, coherent economic methodology (especially ROCE capital employed), and stringent fairness safeguards when the CAT steps into the regulator’s shoes under Schedule 8.