The Delhi High Court’s decision in Array Biopharma Inc. v. Deputy Controller of Patents highlights a basic but important point: patent objections must be tied to the claims and evidence, not merely asserted. Harsh Kashyap explains how the Court applied this principle to objections on inventive step, Section 3(d), and Section 3(i), while also flagging the questions the judgment leaves unresolved. Harsh is a 3rd-year B.Sc. LLB student at NFSU Gandhinagar.
Merely Citing Isn’t Applying: Looking at the DHC Order in Array Biopharma v. Deputy Controller
By Harsh Kashyap
A combination of three cancer drugs, one patent application, and six years just to properly reject it. If one were to describe Array Biopharma Inc. v. Deputy Controller of Patents in a line, it would be this. The applicant claimed a Patent for a combination of a B-Raf inhibitor (Encorafenib), an EGFR inhibitor (Cetuximab or Erlotinib), and optionally a PI3K-alpha inhibitor (Alpelisib), for treating B-Raf-mutant colorectal cancer. The patent controller rejected it on three grounds, citing lack of inventive step under Section 2 (1) (ja), non-patentability under Sections 3(d) and 3(i). However, the Delhi High Court disagreed with the Controller’s order on all grounds.
Before diving in, it’s important to note that the judgment doesn’t rewrite the law on Sections 3(d) and 3(i). However, for what it’s worth, it clearly shows what happens when a Controller invokes a statutory exclusion without actually applying it. On inventive step, on Section 3(d), on Section 3(i), the pattern in the impugned order is identical: state a conclusion in general, and skip the step where it is required to point to something specific in the claim or the prior art that is actually required to support the conclusion.
Inventive Step: Citing Four Documents isn’t Comparing Them
Let’s start with inventive step, as this is the part where the Court did most of the analysis. The rule on this point — the court asks whether a person skilled in the art, having regard to the prior art, would come to the same conclusion i.e., the claimed invention. The fact that similar drugs were known does not make the claimed combination obvious. Instead, the prior art should have led the PSITA to that particular claimed combination. The Controller’s order lacked this. It merely pointed to four earlier documents, D1 to D4, and simply said that together they taught this combination. It never said which part of which document actually pointed toward combining these particular drugs. So the Court checked: D1 discloses a B-Raf inhibitor, but not Encorafenib. D2 discloses Erlotinib and Cetuximab, but not Encorafenib. D3’s combination is Dabrafenib and Omipalisib, not Encorafenib and Alpelisib. D4 discusses PI3K-alpha inhibition without naming either compound actually claimed. Not even one of these documents, alone or together, discloses the molecules that the Applicant actually claimed.
There’s another issue worth flagging here, the D1 is a patent for a chemical compound, D2 is about diagnosing cancer, specifically figuring out which tumours won’t respond to B-Raf inhibitors, it has nothing to do with teaching how to combine drugs for treatment. Using either of these as proof that someone would have thought to combine these drugs assumes they were written with this exact invention in mind. This exact situation is known as Hindsight bias, or in legalese as ex-post facto reconstruction. The Delhi High Court has repeatedly barred controllers from doing this. In Roche v. Cipla, the court held that obviousness must be assessed without hindsight, and in Avery Dennison Corporation v. Controller, it held that prior-art references cannot simply be put together to establish obviousness without showing why a skilled person would have combined them. The issue is not whether the cited prior art disclosed different drugs. The issue is whether those documents were capable of supporting the Controller’s finding on obviousness at all.
It’s also worth noticing that the Court didn’t use the Roche v Cipla five-step test, which is the usual method courts use to check if an invention is obvious. To simply the five-step test is-
- To identify an ordinary person skilled in the art;
- To identify the inventive concept embodied in the patent;
- To impute to a normal, skilled but unimaginative ordinary person skilled in the art what was common general knowledge in the art at the priority date;
- To identify the differences, if any, between the matter cited and the alleged invention and ascertain whether the differences are ordinary application of law or involve various different steps requiring multiple, theoretical and practical applications;
- To decide whether those differences, viewed in the knowledge of alleged invention, constituted steps which would have been obvious to the ordinary person skilled in the art and rule out a hindside approach.
Instead of applying the whole test, the Court just compared the prior art directly to the contents of the claim and stopped. Other Delhi High Court benches have treated those five steps as the gold standard, and have struck down orders for not following them properly. Further, it’s important to note that the Document-to-Document comparison by the court nearly resulted in the same place as the fifth step would have. However, it didn’t follow the structure, making it harder to tell whether the analysis would hold up in the next case. So it’s still unclear whether Indian courts have settled on one consistent way to test obviousness in pharma cases, or whether it just depends on the bench. This uncertainty isn’t merely speculative. In October 2025, a Delhi High Court Division Bench in Tapas Chatterjee treated the test as nearly mandatory and set aside an order for proceeding directly to Step four without first identifying the PISTA. Approx nine months later, in July 2026, a Division Bench in Sulzer Mixpac AG v. Assistant Controller of Patents and Designs, which notably comprised the same Justice C. Hari Shankar, took a different position, holding that the five steps are “not commandments cast in stone.” Array Biopharma, decided by a different bench three weeks after Sulzer, dispenses with the five-step framework altogether. Therefore, at present, it can be said that no single approach exists.
Section 3(d): Two Different Problems, Run Together
Further, Section 3(d) is where things get more interesting. The issue here can be divided into two parts:-
The first is that the controller never said what the claimed combination was supposedly a derivative of. Section 3(d) only applies when a claim is a “new form” of some specific known substance, a salt, polymorph, or other variant of it. The impugned order asserted that the claimed compounds were “derivatives of the known compounds,” but never named which known compound Encorafenib, or the combination as a whole, was supposed to derive from. That’s the gap the Court actually caught the Controller’s order didn’t give anything concrete to test that claim against. This is in direct violation of the Calcutta High Court’s ruling in Topotarget, as it says that an order that fails to identify the known substance a claim is supposedly a new form of cannot sustain a Section 3(d) objection.
The second problem is a definitional one. The appellants, relying on the IPAB’s decision in Ajantha Pharma argued that the combination altogether falls outside Section 3(d). The Board in that case held that two independently known active drugs can’t be treated as derivatives of each other. Therefore, the word “combinations” in the Explanation only reaches combinations of already listed derivatives, like salts, polymorphs, and a derivative paired with the original substance. Not two unrelated known drugs put together. The Respondent also relied on the Calcutta High Court’s decision in Topotarget, which restated almost the same thing on combination as Ajantha Pharma.
The Court in Array Biopharma never tests that reading anywhere in the judgment. It disposes of the Section 3(d) objection on the ground that the Controller never named the known substance the claim was supposedly derived from, and leaves Ajantha’s framing. It can be said that this was a missed opportunity, as Ajantha’s decision on “combination” isn’t exactly a solid ground to leave standing. Scholars have heavily criticised this decision as flawed (see here). After this judgment, drafting an invention as a formulation, composition, or combination claim has become a go-to move to dodge scrutiny under Section 3(d). Further, the decision reads a word into the Explanation of Section 3(d) that Parliament never intended to include, thereby narrowing “combinations” to mean only combinations of the derivatives listed there, even though the text doesn’t say anything close to that. Keeping both points and the purpose of Section 3(d), which is to stop evergreening, in mind. It can be said that the Ajantha actively works to defeat the very purpose of Section 3(d).
That’s a bigger question it leaves open. If combining API automatically escapes this provision, simply because it’s a combination, then a whole category of pharma patents, ones built from drugs that are already off patent, ends up outside a rule that exists specifically to stop evergreening. Then the only thing left to check such an invention against is altogether a separate question of Obviousness under Section 2(1)(ja).
Section 3(i): The Claim Says ‘Combination,’ not ‘Method’
Section 3(i) has a similar problem. What is actually allowed to claim depends on the wording of the claim itself, not on what’s written elsewhere in the application. A table showing dosage in an example just shows that the invention works in practice. It doesn’t define what’s being claimed. The Controller looked at a dosing schedule from one of the examples, one drug as a capsule, another as a tablet, a third given through an IV, and used that to say the real claim here was actually a method of treatment. But the Court, following its own earlier rulings in the Nestle and Bayer Pharma cases, pointed out that the actual claim describes a product, a pharmaceutical combination comprising, and that the part about how it’s administered just explains how the product might be used, it isn’t claiming that use as a step. This is a mistake in how the claim itself was read, the same kind made under Section 3(d): drawing conclusions based on a general impression of the invention. The pattern juxtaposing the objections reveals that a similar shortcut has been taken every time. Under Section 2(1)(ja), the order named four documents but never showed what in them pointed to this combination. Under Section 3(d), it called the compounds derivative without saying of what. Under Section 3(i), it read a dosing table in a working example as the claim itself. None of this required the Court to find or define anything new. It just needed the Court to actually read what the Controller had skipped over. The case now goes back to the Controller to look again within six months. Whether the fresh decision will actually do the work that was missed the first time, or just reach the same conclusions with better wording, is something only time will tell.
