F. Hoffmann-La Roche v. Cipla: Section 3(d) is a patent-eligibility provision at grant, not an infringement defence

F. Hoffmann-La Roche Ltd v. Cipla Ltd is the doctrinal companion to Novartis. The decade-long Delhi High Court litigation over Roche's Indian patent on Erlotinib Hydrochloride — the molecule marketed by Roche as Tarceva for the treatment of non-small-cell lung cancer and pancreatic cancer — produced a Division Bench judgment that, on 27 November 2015, resolved two related questions that had been left unanswered by Novartis. The first was whether the rejection of a downstream polymorph patent application under Section 3(d) could narrow the scope of the parent compound patent at the infringement stage. The second — and the more doctrinally consequential — was whether Section 3(d) of the Patents Act 1970 operates at the grant stage only, or whether it also operates as a defence available to the accused infringer at the Section 48 infringement stage. The Bench, comprising Pradeep Nandrajog, J. and Mukta Gupta, J., answered the first in the negative and the second in favour of grant-stage operation. The Bench affirmed the patent's validity, found infringement, imposed ₹5 lakh in costs on Cipla, declined a permanent injunction because the patent was expiring in March 2016, and remanded the matter for an accounts inquiry on damages.

The judgment is reported at 2015 SCC OnLine Del 13619. The reasoning is structured around an extended treatment of patent-construction principles (covered versus disclosed; compound claims versus polymorph claims), the validity defences mounted by Cipla, and the doctrinal location of Section 3(d) within the architecture of the 1970 Act. The decision sits at the patent-infringement-defence end of the post-TRIPS pharmaceutical-patent triptych — Novartis at the grant gate, Bayer Corporation at the compulsory-licence post-grant gate, and Roche at the infringement-defence gate.

The statutory architecture

Five provisions hold the weight of the decision.

Section 48 confers on the patentee the exclusive right to prevent third parties from making, using, offering for sale, selling or importing the patented product in India for the term of the patent. The right is the operative monopoly that grounds the infringement action.

Section 25(1) governs pre-grant opposition and Section 25(2) governs post-grant opposition. These are the procedural vehicles by which the patent-eligibility grounds of Section 3, including Section 3(d), can be agitated against a granted patent.

Section 64 governs revocation. The grounds for revocation are comprehensive and include lack of novelty, obviousness, insufficiency of disclosure, the patented invention being one falling within Sections 3 or 4 (the patent-eligibility exclusions), and several others. Section 64, the Bench held, is the procedural channel through which Section 3(d) operates against a granted patent in a litigation context.

Section 107 provides that, in an infringement suit, every ground on which a patent may be revoked under Section 64 is available as a defence. This is the textual link the Bench had to engage with: if Section 64 grounds include Section 3(d), does Section 107 make Section 3(d) a freestanding defence at the infringement stage?

Section 8 requires the patentee to disclose to the Patent Office, before and during prosecution, information regarding corresponding foreign applications. Cipla raised a Section 8 non-compliance defence as one of its validity attacks.

The Section 3(d) text and its Explanation — the heart of the Novartis gloss — remained the doctrinal centrepiece of the litigation. The Bench's central interpretive move was the structural reading that Section 3(d) operates at the patent-eligibility level (the Section 3 enumeration is captioned "What are not inventions") and is to be deployed through the opposition and revocation channels — not as a standalone Section 107 defence.

The factual matrix

Roche held Indian patent IN 196774, granted on 23 February 2007 with priority from a US application of 6 March 1996. The patent was titled "A novel quinazoline derivative" and claimed Erlotinib and its pharmaceutically acceptable salts including Erlotinib Hydrochloride. The patent contained two principal claims that mattered to the litigation: Claim 1 covered the compound Erlotinib (and Erlotinib Hydrochloride); the patent did not, on its face, confine the claim to any particular polymorph or crystalline form of the salt.

Two polymorphic forms of Erlotinib Hydrochloride had been identified in the prosecution history — Polymorph A and Polymorph B. Polymorph A is the form initially obtained in the original synthesis; Polymorph B is a more thermodynamically stable form. Roche had separately filed an Indian application (IN/PCT/2002/00507/DEL) on the Polymorph B form of Erlotinib Hydrochloride. That application was rejected by the Patent Office in December 2008 on Section 3(d) grounds — the Polymorph B form was held not to demonstrate enhanced therapeutic efficacy over the salt covered by IN 196774.

Cipla launched its generic Erlotinib Hydrochloride product under the brand name Erlocip in 2008. Cipla's product was, on the chemistry record adduced in the litigation, Polymorph B. The Erlocip product was priced at approximately one-third of Tarceva's price.

Roche instituted suit CS(OS) 89/2008 before the Delhi High Court seeking a permanent injunction against Cipla's manufacture and sale of Erlocip on the ground that Erlocip infringed IN 196774. Cipla counter-claimed for revocation of IN 196774 under Section 64 on multiple grounds including anticipation by prior art, obviousness, insufficiency of disclosure, Section 3(d) (lack of enhanced therapeutic efficacy of the claimed compound over disclosed prior-art compounds), and Section 8 non-compliance.

Manmohan Singh, J., sitting as the single judge of the Delhi High Court, dismissed Roche's suit on 7 September 2012 after a full trial. The single judge held the patent valid but held that Roche had not proved infringement — the suit-patent claim was on a non-Polymorph-B form of Erlotinib Hydrochloride, and Cipla's Polymorph B did not infringe. Roche appealed by RFA(OS) 92/2012; Cipla cross-appealed against the validity finding. The matter was heard by the Division Bench of Nandrajog, J. and Mukta Gupta, J. and judgment was reserved.

The Court's reasoning

The Bench's reasoning is structured in three steps: validity (taking up Cipla's revocation grounds in cross-appeal); infringement (the Polymorph B / parent compound question); and the doctrinal location of Section 3(d).

Validity — IN 196774 holds up

The Bench worked through Cipla's revocation grounds methodically.

Anticipation and obviousness. The principal prior art was a European Roche group patent (EP 0566226) on a closely related quinazoline derivative. The Bench held the prior art disclosed a different molecule — substituent differences at specific ring positions produced a different compound with different pharmacological properties. Obviousness was rejected on the same record: the choice of substituents that produced Erlotinib's specific selectivity for the EGFR (epidermal growth factor receptor) tyrosine kinase was held not obvious to a skilled artisan at the priority date.

Section 3(d) at the validity stage. Cipla pleaded Section 3(d) under Section 64 — arguing that IN 196774's claimed compound was not shown to have enhanced therapeutic efficacy over disclosed prior-art compounds. The Bench held the attack did not succeed on the facts: Section 3(d)'s "new form of a known substance" architecture did not apply because Erlotinib was not a new form of the prior-art compound — it was a different compound.

Section 8 non-compliance. Cipla argued Roche had breached Section 8's foreign-application-disclosure obligations. The Bench held that even if made out, on the standard later confirmed by the Supreme Court in Sukesh Behl v. Koninklijke Philips (2014), the breach was not material — it had not concealed information that would have changed the prosecution outcome.

Insufficiency. Rejected on the specification's detailed synthesis examples.

The validity of IN 196774 was affirmed.

Infringement — the polymorph question

The single judge's dismissal had rested on the proposition that IN 196774, on the prosecution history, was confined to a non-Polymorph-B form of Erlotinib Hydrochloride. Cipla's Polymorph B, accordingly, did not infringe. The Bench reversed that finding.

The reasoning ran on patent-construction principles. Claim 1 of IN 196774 covered Erlotinib and its pharmaceutically acceptable salts — Erlotinib Hydrochloride being the principal salt. The claim was on the compound as a class. The specification disclosed the synthesis but did not confine the claim to any polymorphic form. Polymorph A and Polymorph B are both forms of Erlotinib Hydrochloride — both fall within the compound class covered by Claim 1.

The fact that Roche had separately applied for a Polymorph B patent and that application had been rejected on Section 3(d) grounds did not — and could not — narrow the scope of the parent compound patent. The Bench drew the structural distinction between covered and disclosed: the parent patent covered Erlotinib Hydrochloride (including both polymorphs); the downstream Polymorph B application sought to specifically disclose and claim Polymorph B as a separate invention; the rejection of the downstream application meant Roche could not get a separate patent on Polymorph B, but the parent patent's scope was unaffected.

Cipla's Polymorph B Erlocip, accordingly, fell squarely within the scope of Claim 1 of IN 196774. Infringement was found.

The doctrinal location of Section 3(d)

The Bench's most consequential analytical move was the structural reading of Section 3(d). The textual chain Cipla pressed ran as follows: Section 64 lists, among revocation grounds, that the patented invention is one falling within Sections 3 or 4; Section 107 makes available, in an infringement suit, every ground on which a patent may be revoked under Section 64; therefore Section 3(d) is available as a freestanding defence at the Section 48 infringement stage.

The Bench rejected the chain. Section 3 of the 1970 Act, captioned "What are not inventions", enumerates exclusions from patent eligibility — public-morality, frivolous inventions, mere discovery of a scientific principle, mere admixtures, and at clause (d) new forms of known substances lacking enhanced efficacy. The provision operates at the patent-eligibility level and is engaged at the prosecution stage by the Patent Office, at the pre-grant opposition stage under Section 25(1), at the post-grant opposition stage under Section 25(2), and at the revocation stage under Section 64. The Section 48 infringement court is not the appropriate forum for a patent-eligibility re-examination; the accused infringer who wishes to attack the grant on Section 3(d) grounds has the procedural channel of revocation under Section 64 (which can be raised by counterclaim within the infringement proceeding itself).

The Bench was careful to clarify what the holding does and does not do. Section 3(d) is not unavailable in litigation — it is available through Section 25(2) and Section 64. What is closed off is a freestanding Section 3(d) defence disconnected from the Section 64 procedural architecture. The analytical consequence is to keep the Novartis heightened-patentability filter within its proper procedural locus: the post-grant infringement court applies the patent as granted; the proper challenge to a Section 3(d)-non-compliant grant is the revocation channel, which the infringement defendant can pursue by counterclaim.

Relief — costs, accounts, no injunction

The relief portion turned on the practical constraint that IN 196774 was within four months of expiry. The Bench declined a permanent injunction — the patent was expiring on 23 March 2016. It imposed ₹5 lakh in costs on Cipla and remanded the matter to the single judge for an accounts inquiry on damages for the 2008–2016 infringement period.

The doctrinal contribution

Roche v. Cipla contributes to Indian patent law on three axes. The Section 3(d) eligibility-versus-defence distinction — the structural reading that Section 3(d) operates at the patent-eligibility level and reaches the patent through opposition and revocation channels, not as a freestanding Section 107 defence — stabilises the post-grant architecture: the infringement court applies the patent as granted; the patent-eligibility challenge is the revocation court's business. The covered-versus-disclosed distinction in polymorph patents — that a downstream polymorph application's rejection under Section 3(d) does not narrow the parent compound patent — supplies an important construction principle: compound claims cover the API as a class; polymorph claims, where granted, cover specific forms; the failure to obtain a polymorph claim does not retrospectively narrow the parent's class. The validity-defences architecture in infringement — Cipla's revocation grounds engaged on the merits and disposed of with reasoned findings — supplies the working Section 64 template and has informed subsequent Delhi High Court infringement-validity practice through the AstraZeneca AB v. Intas Pharmaceuticals (2021) line and the recent Bansal v. Philips Division Bench reversal (May 2026) on SEP infringement.

What the judgment did not decide

The Bench did not foreclose all paths by which Section 3(d)'s logic might inform claim construction at the infringement stage; what it foreclosed is the freestanding deployment of Section 3(d) as a Section 107 defence disconnected from Section 64. It did not lay down procedural standards for a Section 64 counterclaim in an infringement action — pleading, evidentiary base, sequencing relative to the main trial — and subsequent practice has worked these out case-by-case. It did not address the biologics dimension of the polymorph / compound class distinction; the holding is anchored in small-molecule pharmaceutical chemistry. And the Supreme Court has not authoritatively re-examined the framework — the Special Leave Petition was admitted in 2016 and withdrawn by Roche under a global settlement with Cipla in June 2017. The Delhi Division Bench reasoning remains the operative law.

The doctrinal arc

The first phase, immediately after the judgment, was consolidation in pharmaceutical infringement practice. Delhi High Court single-judge benches in AstraZeneca AB v. Intas Pharmaceuticals (2021) and in a series of crystalline-form actions worked the Roche framework into the standard architecture — Section 3(d) as a revocation-channel ground; the Section 64 counterclaim as the standard vehicle. The second phase extended the covered-versus-disclosed distinction into selection-patent and divisional-patent contexts. The third phase, beginning in the late 2020s, is intersection with the SEP / FRAND architecture: the Bansal v. Koninklijke Philips Division Bench reversal (Delhi HC, 18 May 2026, C. Hari Shankar, J. and Om Prakash Shukla, J.) supplies the contemporary architecture for SEP infringement defence — proof-of-essentiality, FRAND-rate evidentiary burden, Section 107A(b) patent exhaustion — within the Roche-grounded Section 64 validity-defence template.

What practitioners take

For the patentee in infringement. The pleading should be organised around the Section 48 claim with a clean validity narrative under Section 64. Where the patent is a compound patent with downstream polymorph or salt applications, the parent claim's scope as a class is the operative claim — the patentee should plead claim construction explicitly and tender the prosecution history with care. The Roche covered-versus-disclosed distinction operates in the patentee's favour where the accused product is a polymorph covered by the parent.

For the accused infringer. The validity attack should be framed as a Section 64 revocation counterclaim. Grounds — anticipation, obviousness, insufficiency, Section 3 exclusions, Section 8 — should each be pleaded with particulars and evidentiarily supported. The freestanding Section 3(d) defence at the Section 48 stage is no longer available; the Section 3(d) attack must be routed through Section 64. Where the accused product is a recognised polymorph of the patented compound, the better path is the validity attack on the parent itself, not the scope-of-coverage argument.

For drafting and prosecution. Prosecution strategy should be coordinated across the parent compound patent and the downstream polymorph and salt applications. The Section 3(d) exposure on a downstream polymorph application — significant given the Novartis gloss — does not, after Roche, bleed into the parent compound patent. Downstream applications should be pursued with full therapeutic-effect data; the failure of a downstream application is a setback but not a structural threat to the compound monopoly.

The doctrinal residue

F. Hoffmann-La Roche v. Cipla is the third pillar of the post-TRIPS Indian pharmaceutical patent canon. The Section 3(d) eligibility-versus-defence distinction sustains the procedural integrity of the 1970 Act: Section 3(d) operates at the grant stage and reaches the patent through the Section 25 / Section 64 channels, not as a freestanding Section 107 defence. The covered-versus-disclosed distinction in polymorph patents preserves the parent compound's class scope against the downstream-rejection narrowing argument. The validity-defence architecture in Section 64 counterclaim — engaged on the merits in the judgment — supplies the procedural template for subsequent infringement actions. Read together with Novartis (grant-stage) and Bayer Corporation (post-grant compulsory licensing), the Roche judgment completes the triptych that defines the working architecture of Indian pharmaceutical patent law in the TRIPS era. The Supreme Court has not re-examined the framework; the Delhi Division Bench reasoning is, after nearly a decade of consistent application, the settled doctrine.

Related editorial pieces

• Novartis AG v. Union of India: the enhanced-therapeutic-efficacy gloss on Section 3(d)
• Bayer Corporation v. Union of India: India's first compulsory licence
• Intex v. Ericsson: the Delhi HC DB's SEP / FRAND framework
• Delhi HC in Hindware v. Google Ads: keyword-bidding, trademark infringement and intermediary liability