Smithkline Beecham Plc & Ors. v Apotex Europe Ltd & Ors

The claimant/patentee parties (“SKB”) appeal from a judgment of Pumfrey J of 5^th December 2003 whereby he rejected their claim that the defendant parties (“Apotex”) infringed patent No. 2,297,550, held the patent invalid and ordered its revocation. Mr Waugh QC argued the case for SKB; Mr Watson QC that for Apotex.

In previous litigation, a claim by BASF for revocation of the patent resulted in a finding that most claims of the patent were invalid, but that two claims (10(i) and 11) were valid. These proceedings were before Pumfrey J whose decision [2002] EWHC 1373 (Ch) was upheld by the Court of Appeal [2004] IP&T 846, [2003] EWCA Civ 872. As a consequence of these findings SKB successfully applied to amend the patent to restrict the monopoly to that claimed in those two claims, now renumbered as claims 1 and 2. We are concerned with the validity of these and whether or not they cover the process used by Apotex.

The primary attacks on validity were based on two prior published documents. The more important of these is UK patent application 85-26407 (‘407), which was published in 1987 as part of the file for what became EP 223,403 (“the hemihydrate patent”). The other is US patent No. 2,864,817 (“the erythromycin patent”) published in 1958. Other attacks on validity are also advanced in the nature of a squeeze – it is said that if the claims are wide enough to cover what Apotex do, they also cover that which obvious or are so wide or ambiguous as to be insufficient. It is also contended that the claims are so ambiguous that one is unable to say what their scope is and hence one is unable to say whether the Apotex process falls within them.

The BASF proceedings were more limited in scope than these. Only validity was in issue and only upon the basis of the ‘407 prior art. Neither the erythromycin patent nor the various invalidity arguments based on a “squeeze” were raised. Moreover the evidence was, of course different. It is not suggested that Pumfrey J in this action was, or this court is, any way bound by the previous decision so far as it turned on evidence. This case has to be decided on its own evidence though some of that has been imported via hearsay notices from the earlier case.

The case is concerned with forms of the pharmaceutical paroxetine – it is far from the first (or last) of such cases in this jurisdiction alone. From that one can infer that a great deal of money is involved. A company called Ferrosan found the paroxetine molecule in about 1980 and that it had potential therapeutic effect. SKB (strictly its predecessors) bought up the rights and did all the very great risky work to test it and bring it to market. By the mid-90’s SKB’s paroxetine product, under the trade mark Seroxat, was their biggest selling pharmaceutical drug.

The salt of paroxetine on which SKB did most work was the hydrochloride. This, it is now known, can exist in a number of crystalline forms:

Depending on the particular solvate, the solvent within it may be removed more or less readily. In the case of some solvates it is sufficient to treat them with known drying techniques to remove virtually all the solvent. In the case of others the solvent is more difficult to remove – some at least will not come out even with stringent drying. Because the term “solvate” encompasses a range of possibilities, it is generally necessary to consider its use in context to ascertain the precise meaning intended.

In which crystalline form a drug is actually marketed is of considerable significance. Although at a crude level one might think it does not matter (once it is dissolved in the stomach, it can hardly matter what form the original crystal was) in practice it does. Regulatory authorities are concerned with which form for reasons of bioavailability (some forms dissolve more readily than others), stability and the like. Also, and for more obvious reasons, the authorities are concerned with amounts of impurity. So solvates with large amount of unremoved solvent are (at least where the solvent is one not found in the body naturally) unlikely to get marketing authorisation. We were told that an impurity level of more than 0.5% of any compound is not permitted. Curiously two impurities, each less than 0.5% are permitted. Apotex’s product has about 0.4% of each of IPA (see below) and acetone – so it is permitted.

When SKB first started investigating paroxetine hydrochloride in the early 1980’s they made the anhydrate at pilot scale at their Harlow plant. Their process involved the use of water. At some point, although ostensibly nothing had changed, they discovered that the process was making hemihydrate instead of anhydrate – they could no longer make the anhydrate. This is a known type of phenomenon. It is attributed to the fact that once a particular environment is “seeded” with a particular type of crystal, the seeds will encourage the formation of that type rather than any other type. It is not necessary to go further into this at this stage, though it necessary to discuss it a little more when I come to Mr Watson’s “Bodmin” point.

In 1985 SKB applied for patent ‘407 which, in its broadest aim “provides crystalline paroxetine hydrochloride as a novel material, in particular in pharmaceutically acceptable form.”

‘407 goes on to state that SKB have discovered that paroxetine hydrochloride can exist in at least the hemihydrate and anhydrate forms and that it “can form crystalline solvate with certain solvents …. in particular isopropyl alcohol.” (I shall call this IPA. It has other names, e.g. propan-2-ol, but I shall change all quotations to just IPA for consistency and brevity). Example 1 (of which lots more anon) purports to describe the preparation of the anhydrate. Examples 2-4 purport (and indeed, it is common ground do) describe the production of hemihydrate. Example 5 describes the production of the IPA solvate. Because, say SKB, example 1 in fact produced the hemihydrate and not the anhydrate, ‘407 was used to obtain patent priority only for the hemihydrate. At that point (in the late 1980’s and early 1990’s) therefore, SKB did not have any protection for the anhydrate, nor, self-evidently, for any process for making it.

This was applied for in 1996. Its title is “Paroxetine hydrochloride anhydrate substantially free of bound organic solvent.” In its broadest form it was an attempt to gain patent protection for the anhydrate sufficiently pure to pass regulatory approval. As I have said, following the BASF litigation it was considerably amended – broadly down to particular processes for making the anhydrate free of solvent.

Claims 1 and 2 (as amended) read:

Notwithstanding their brevity, they present formidable problems of construction. At this stage it is merely necessary to quote bits of the specification left following the amendment. The patent begins by referring to Ex.8 of EP ‘403 (the hemihydrate patent) which is the same as ex.1 of 407. It says:

Subsequent repetition of example 8 has failed to yield any type of paroxetine hydrochloride anhydrate, and there is no clear teaching elsewhere in the document of any alternative route or modification of the process which would generate the anhydrate

Paroxetine hydrochloride anhydrate is also purported to be disclosed in [a 1989 document called Buxton & Lynch]. The anhydrate is said to be produced by crystallising paroxetine hydrochloride from anhydrous IPA. Subsequent repetition of this process has resulted in an IPA solvate of paroxetine hydrochloride. That is to say that there is bound IPA in the product. This bound IPA cannot be removed by conventional drying techniques such as vacuum oven drying.

Paroxetine hydrochloride anhydrate substantially free of bound IPA, has not been described in the literature, nor has any method been disclosed which would yield such a product as an inevitable result. A method for preparing paroxetine hydrochloride anhydrate substantially free of bound IPA has now been found.”

So two things: ex. 1 of 407 makes hemihydrate not anhydrate and if you make anhydrate à la Buxton & Lynch you get a solvate from which you cannot remove bound IPA by conventional techniques.

The process of the invention is then said to achieve the anhydrate “substantially free of bound organic solvent.”

This phrase naturally invites the questions “what does that mean?” “How little bound solvent makes the product substantially free?” The patent provides a sort of definition. It says:

“Substantially free of bound organic solvent is to be interpreted to be less than the amount of IPA which would remain solvated, i.e. bound, within the crystal lattice of the product under conventional vacuum oven drying conditions.”

This raises at least one obvious further question: “what are ‘conventional vacuum oven drying conditions’ – a question not of construction but of fact. I proceed with the full context of the specification.

It continues:

“It should be understood that the present invention comprising paroxetine hydrochloride anhydrate substantially free of bound IPA may contain unbound water that is to say water which is other than water of crystallisation.”

So “anhydrate” does not (as it might in other contexts) mean totally dry – water can be present provided it is not actually part of the crystal lattice. As was first taught by ‘407 the anhydrate is hygroscopic – it picks up water from the atmosphere.

Some idea of the level of bound solvent aimed at is given by the next paragraph:

“Typically the amount of bound solvent on a weight for weight basis would be less than 2.0%, preferably less than 1.8%, more preferably less than 1.5%, even more preferably less than 1.0%, yet more preferably less than 0.5% and most preferably less than 0.1%.”

This is a rather clumsy way of saying less than 2% will do but the less the better.

The specification then says:

“The forms of paroxetine hydrochloride anhydrate may be distinguished from each other and the material formed as a result of carrying out the procedures mentioned in [the hemihydrate patent and Buxton & Lynch] by crystalline shape, solvent analysis or techniques such as IR, melting point, X-ray diffraction, NMR, DSC, microscopy and any other analytical techniques which differentiate one form from another.”

There is then a discussion about how various forms (A, B, C, and D) of anhydrate can be distinguished by various analytical data. Melting points, IR bands, and DSC exotherm data are given.

The specification (I continue with the amended form) continues with two statements of invention which are word for word the same as claims 1 and 2 set out above. It then sets about giving the reader more information about the process. It begins by saying you can make the hydrochloride by contacting a solution of the free base with dry HCl gas. This will obviously not introduce water at that stage. Then another waterless route to hydrochloride starting with the hemihydrate (forming an azeotrope and removing water) is described. There is a clear teaching of the avoidance of water implicit here. It is then made explicit by the first sentence of the next paragraph:

“The organic solvents should be substantially free of water to the extent that there is insufficient water present at the time of crystallisation to effect conversion to the hydrochloride hemi-hydrate.”

The importance of making sure there is not too much water in the system (otherwise one will get the hemihydrate) when one comes to crystallisation is here taught. This is of much significance. For no such express teaching is found in the prior art.

The next paragraph of importance is about the solvent used to produce the crystallised anhydrate solvate. The specification says:

“It should be appreciated that an organic solvent or solvents which form a solvate with the crystallised paroxetine hydrochloride and which are not removable by conventional drying techniques may be determined by a matter of routine experimentation. Examples of such organic solvents include, but in no way are limited to, alcohols especially alcohols such as IPA, ethanol and propan-1-ol; organic acids such as acetic acid; organic bases such as pyridine; nitriles such as acetonitrile; ketones such as acetone; ethers such as tetrahydrofuran and chlorinated hydrocarbons such as chloroform.”

So any organic solvent can be found by routine experiment. Note that one of the solvents specifically mentioned is acetone – the specification is specifically saying that you can produce an acetone solvate containing bound acetone not removable by conventional drying techniques.

There then follows an important description of the displacement step:

“The paroxetine hydrochloride solvate produced is suitably isolated and dried by conventional methods such as drying in vacuo to remove some or all of the free or unbound solvent. It should be appreciated that it is preferable and unexpected that the degree of drying is controlled such that only free solvent is removed. The bound solvent is then displaced with a displacing agent such as water or supercritical carbon dioxide. It is possible to use other displacing agents which may be selected by means of routine experimentation.

Preferably gaseous or liquid water may be used as a displacing agent. It is important that the paroxetine hydrochloride solvate is contacted with enough water and for sufficient time to displace the solvent but insufficient to cause conversion to the hydrochloride hemi-hydrate.

The amount of water, the form of the water, e.g., liquid or gaseous and the length of time which the paroxetine hydrochloride solvate is contacted with the water differs from solvate to solvate. This depends largely upon the solubility of the solvate in question.

Particular ratios of paroxetine hydrochloride solvate to water are outlined in the examples hereinafter described (Examples 1, 4 to 6, 9 to 11, 13 and 15). It should be appreciated that the pyridine solvate is believed to be more soluble in water than for example the IPA solvate. Thus the use of the common ion effect when using diluted hydrochloric acid may help prevent dissolution of the solvate and subsequent conversion to the hydrochloride hemi-hydrate.

After contact with water to displace the bound solvent the product is suitably dried, for example, in vacuo at elevated temperature. Suitable drying may be over a desiccant such as phosphorus pentoxide.

When supercritical carbon dioxide is used it should be appreciated that the flow rate, temperature and pressure of the carbon dioxide may be controlled to give optimum solvent removal from the paroxetine hydrochloride solvate. Generally high pressure carbon dioxide may be used for example at about 2,500 psi. Elevated temperatures may also be preferably used such as between 50 to 80C. More preferable between 55 to 75C.”

Nothing more in the general discussion of the specification is germane to this appeal. So I can omit the bits about how to make various forms of anhydrate and what may fairly be called “waffle” about the various medical indications for the use of paroxetine and the preparation of tablets and the like. There are 21 “examples”, though not all are of the claimed process. For instance examples 2 and 3 are for the preparation of a solvate prior to any displacement step. The points worthy of note for the purposes of this appeal which emerge from the examples are these:

The Patents Act 1977 provides that an invention is only patentable if it is, inter alia, “new” (s.1(1)(a)) and “involves an inventive step” (s.1(1)(b)). The language is taken from the EPC, Art. 52.

These are distinct requirements. Novelty is further defined in the following way by s.2, taken from EPC Art.54:

“(1)

An invention shall be taken to be new if it does not form part of the state of the art.

(2)

The state of the art in the case of an invention shall be taken to comprise all matter (whether a product, a process, information about either, or anything else) which has at any time before the priority date of that invention been made available to the public (whether in the United Kingdom or elsewhere) by written or oral description, by use or in any other way.”

“Inventive step” is defined by s.3 (from EPC Art.56):

“An invention shall be taken to involve an inventive step if it is not obvious to a person skilled in the art, having regard to any matter which forms part of the state of the art by virtue only of section 2(2) above (and disregarding section 2(3) above).”

It is settled that if a prior published document contains “clear and unmistakable directions” to do something falling within a later claim, that claim lacks novelty, see the well-known passage in General Tire v Firestone [1972] RPC 457 at 485. It is that “inevitable result” principle which is of importance here.

It may also be that if a prior document merely provides an “enabling disclosure” to make what is later claimed, that is enough to destroy novelty. Laddie J so said in Inhale Therapeutics v Quadrant [2002] RPC 21 at para. 43. An “enabling disclosure” is a lesser test than “inevitable result”: it allows the skilled man to reach the result using his ordinary skills and without undue effort. Provided experiments comply with this requirement, some departure or addition to the prior disclosure would be enough to be an enabling disclosure.

However this Court, in Synthon v SKB [2003] RPC 43 may have somewhat qualified the “enabling disclosure” route to novelty destruction. The case is pending before the House of Lords and it is not necessary for the purposes of this case to go into the difference. This is because here it is open to Apotex to argue obviousness – they can legitimately suggest that even if the prior art document does not comply with the General Tire inevitable result test, the modifications to the prior art needed to make the process work are obvious, i.e. the sort of thing which would occur to the skilled man.

Often the nature of the inquiry here is helped by the well-known Windsurfing steps, [1985] RPC 59 at pp.73-4. I summarise:

No one suggests that this always the appropriate test. In the end the inquiry is simply that posed by the statute, as, for instance, Sir Donald Nicholls V-C pointed out in Mölnyke v P&G [1994] RPC 49 at pp.112-3. He went on to say, at p.113:

“In applying the statutory criterion [i.e. as to whether an alleged inventive step was obvious] and making these findings [i.e. as to obviousness] the court will almost invariably require the assistance of expert evidence. The primary evidence will be that of properly qualified expert witnesses who will say whether or not in their opinions the relevant step would have been obvious to a skilled man having regard to the state of the art.”

In this case, a Windsurfing analysis is not really helpful (and is not suggested to be helpful) when one considers alleged obviousness over ‘407. But I think it does have application to the case based on the erythromycin patent.

These I re-stated in Rockwater v Coflexip [2004] EWCA Civ 381 at paras. 71-74. In brief, the Biogen appeal principle ([1997] RPC 1 at p.45) applies to obviousness but not to novelty, save in special cases, for instance where there is a conflict of expert testimony on the meaning of a technical term or as to what exactly is disclosed by the prior art. The Biogen appeal principle itself is, in short, that an appellate court must exercise caution in differing from the trial judge’s evaluation of the facts unless he has erred in principle.

The Judge, as was entirely logical, dealt with the issues of construction and infringement, before coming to these points. In our pre-reading, having regard to the Biogen appeal principle, it seemed possible that we would not have to go further than validity. So we directed that the issues of novelty or obviousness over ‘407 and obviousness over the erythromycin patent should be heard first. In the event we were persuaded that the Judge may well have made an error of principle and so we heard the rest of the issues. Nonetheless it is convenient to deal with the issues in the order we heard them. It was common ground that they did not (save for one point) turn on the width of the claims.

The Judge was not entirely clear as to which of these two grounds was the basis of his finding of invalidity. Mr Waugh submits that this was an error of principle – the grounds are distinct and the criteria for establishing them are different. He is right in principle. But I do not see that the point, as such, gets him anywhere in this particular case. This is because both questions turn principally on what is disclosed by ‘407, particularly in Ex.1. Nonetheless it is much clearer and better to keep the points distinct, and that is what I propose to do.

Whether one is dealing with novelty or obviousness, the starting point is what is actually said in the piece of prior art in question. So to that I turn. It acknowledges that paroxetine base is known and that the maleic and acetate salts are also known. The hydrochloride in solution is also acknowledged as old. What is said to be new comes next:

“Accordingly the present invention provides crystalline paroxetine hydrochloride as a novel material, in particular in pharmaceutically acceptable form.

It has been discovered that crystalline paroxetine hydrochloride can exist in at least two different pseudo-polymorphic forms,

1)

a hemihydrate

2)

an anhydrate

It has also been discovered that paroxetine hydrochloride can form crystalline solvates with certain solvents such as certain lower alcohols and acetone, in particular isopropyl alcohol.

Accordingly the present invention provides as novel forms of crystalline paroxetine hydrochloride:

1)

paroxetine hydrochloride hemihydrate

2)

paroxetine hydrochloride anhydrate

3)

paroxetine hydrochloride IPA solvate”

There follow three paragraphs devoted to the hemihydrate, the anhydrate and solvate. Analytical data (m.p., X-ray, IR and DSC) for the former two are given. One is told that the anhydrate is hygroscopic, that water is “easily lost on heating” and that “the product contains a variable amount of ‘free’ water depending on drying and storage conditions.” So the anhydrate can co-exist with free water.

Analytical data for the solvate is not given, save that one is told that the m.p. is in the range of 97 - 102⁰C (lower than hemihydrate, 128-132⁰C or anhydrate, 115-119⁰C). One could readily tell the difference between forms by a quick m.p. test.

The specification then goes on to say how crystalline paroxetine hydrochloride can be made:

“The present invention also provides a process for producing crystalline paroxetine hydrochloride which comprises forming a solution of paroxetine hydrochloride and precipitating the crystalline form from solution.

The solution may be formed by dissolution of pre-formed paroxetine hydrochloride or by forming the hydrochloride in situ. The hydrochloride may be formed from a solution of paroxetine free base or a salt other than the hydrochloride by contacting it with hydrogen chloride.

For example a solution of hydrogen chloride, for example concentrated hydrochloric acid or an organic solvent saturated with hydrogen chloride may be added to a solution of paroxetine salt. Alternatively hydrogen chloride gas may be passed through the paroxetine (salt) solution.”

As regards the anhydrate, there is this general teaching:

“The crystalline anhydrate form of paroxetine hydrochloride may be prepared via the initial formation of a crystalline solvate e.g. IPA or acetone solvate, of the hydrochloride and followed by the removal of the solvating solvent. The IPA solvate may be conveniently obtained by crystallisation from IPA, ideally under anhydrous conditions, by adding gaseous or concentrated hydrochloric acid to a solution of the free base or acetate salt in IPA , or by crystallising or recrystallising preformed paroxetine hydrochloride from IPA solution. The solvent of salvation may be removed by drying, typically under vacuum at high temperature e.g. 60C, to give the hygroscopic anhydrate.”

No more of the general teaching is for present purposes relevant and I can pass to the examples. There are five. Only the first purports to describe the preparation of the anhydrate. The title is “paroxetine hydrochloride as anhydrate”. The description:

“Crude paroxetine free base (0.341 kg) was dissolved in diethyl ether (3.5 litres) and stirred with aluminium oxide (ca. 0.3. kg) for about 3 hours. Charcoal (15 g) and filter aid (celite, 15 g) were added and the mixture filtered through a layer of aluminium oxide, the filtered solids being washed with more ether. To the combined ether solutions was added a mixture of acetic acid (66 ml) and ether whereupon the acetate of paroxetine crystallised and was filtered off, washed with ether and dried.

The acetate salt was dissolved in IPA (2.4 litres) and treated with a mixture of concentrated hydrochloric acid (75 ml) and more IPA. After standing at 0C for about 16 hours, the crystals of the hydrochloride salt containing IPA were filtered off and dried. The salt was stirred in distilled water (0.5 litres) for about 20 minutes, filtered off and dried, giving paroxetine hydrochloride anhydrate (m.p. 118C).”

It is around this short recipe, particularly the second paragraph, that the dispute raged. The first paragraph describes the preparation of the acetate. This is treated with conc. HCl to produce the hydrochloride. But how much is “more IPA?” and what is it for? And what is the nature of the “salt containing IPA” which is filtered off and dried (the “intermediate”)? Is it anhydrate or solvate? And what is the 20 minute water slurry for?

The broad position of the parties was as follows:

“More IPA”:

-

a small amount added to the HCl to prevent hot spots as the concentrated acid was added – so a few hundred ml. at most (SKB);

-

enough to produce controlled crystallisation – which after calculation would be about 3l. though a bit less (e.g. 1 or 2 litres) would do (Apotex).

The intermediate:

-

anhydrate (SKB);

-

solvate (Apotex);

The water slurry:

-

merely to wash further (SKB);

-

to convert solvate to anhydrate by displacing the IPA (Apotex).

The Experts’ views

The experts (Dr Lee for SKB and Dr Cunningham for Apotex) were divided. The Judge said of Dr Lee that “he was occasionally less than helpful in his answers” and we indeed saw examples of this. On the other hand Dr Cunningham did not escape some criticism (“all the experts, with exception of Professor McBride, were from time to time argumentative”). In the end, however, that the Judge clearly preferred Dr Cunningham over Dr Lee (“I felt most confidence in the evidence of Dr Cunningham who seemed to me to be much less of an advocate than the others”).

Before us Mr Watson advanced a strong attack on Dr Lee’s evidence going further than the Judge held. It was submitted that he was an “unsatisfactory witness” who kept changing his evidence, particularly his evidence as to the nature of the intermediate. It was submitted that his evidence “flipped.” Mr Watson prepared a convenient sheet identifying passages where he submitted inconsistent evidence was given.

I have read and re-read the passages concerned – together with the passages immediately before to see what assumption Dr Lee was making or being asked to make. I am not prepared to accept that his evidence “flipped”. The problem lies in those assumptions – and whether Dr Lee thought he was being asked about what actually happened with present day knowledge or what he would have understood ‘407 to be teaching without hindsight knowledge. I am bound to say that I do not think the forceful nature of his cross-examination helped.

As regards Dr Cunningham’s evidence – that he would have seen the water slurry step as effecting a displacement of IPA from the crystals, Mr Waugh submitted he was bound to have been affected by hindsight – for he was given along with ‘407 the patent in suit and the earlier judgment. Once you are told that water can remove IPA from anhydrate containing bound IPA you cannot put that out of your mind. The Judge, submitted Mr Waugh, nowhere specifically addressed this question of hindsight and this was an error of principle. To that I will return.

Before I go further, however, it is as well to remember what the key function of an expert witness in a patent action is – as I said in Rockwater (para. 12):

“ Their primary function is to educate the court in the technology – they come as teachers, as makers of the mantle for the court to don. For that purpose it does not matter whether they do or do not approximate to the skilled man. What matters is how good they are at explaining things.”

To that I would add this: although it is inevitable that when an expert is asked what he would understand from a prior document’s teaching he will give an answer as an individual, that answer is not as such all that helpful. What matters is what the notional skilled man would understand from the document. So it is not so much the expert’s personal view but his reasons for that view – these the court can examine against the standard of the notional unimaginative skilled man. There is an analogy here with the well-known Bolam test for professional negligence – what matters is not what the individual expert witness says he personally would have done, but whether the conduct said to be negligent falls short of what a reasonable professional would have done.

Thus in weighing the views of rival experts as to what is taught or what is obvious from what is taught, a judge should be careful to distinguish his views on the experts as to whether they are good witnesses or good teachers – good at answering the questions asked and not others, not argumentative and so on, from the more fundamental reasons for their opinions. Ultimately it is the latter which matter – are they reasons which would be perceived by the skilled man?

The correct and fair way to deal with this question is to endeavour to approach the example without hindsight, with the background knowledge of the skilled man, and to read the example in the context of the rest of the specification.

By the time he comes to the example, the skilled man will have read that the solvent can be removed from the solvate. The general teaching of the specification says this twice (“may removed by heating under vacuum” and “may be removed by drying, typically under vacuum at high temperature e.g. 60⁰C”). When he gets to Ex 5 he is told “the solvate of solution was fairly weakly bound and could be removed by drying at high temperatures”. He is not told, and it is not part of his common general knowledge, that there is residual solvent which cannot be removed by drying. Nor is the phenomenon of displacement of solvent by a slurry with solvent part of his common general knowledge. That much at least was common ground between the experts, both of whom found the phenomenon surprising.

For my part, therefore, I would expect the skilled man to understand that by the time the intermediate has been dried, as taught by Ex. 1, it is already the anhydrate – it is the drying step which removes solvent. On that basis the further water slurry must be just some sort of wash, the point of view espoused by Dr Lee. Mr Watson submitted that the example did not go quite as far as the general teaching in that it merely said “and dried”. But why should the skilled man expect some lesser extent of drying than enough to remove the IPA as taught by the specification? And if at this point the description is ambiguous (how much drying?) there can hardly be a case of “clear and unambiguous instructions.”

So I think a mere comparison of the patent claims and the disclosure of the ‘407 is not novelty destructive. The claims of the patent in suit are about taking solvate and subjecting it to a displacing step. That is not, on merely reading the two documents disclosed in ‘407. Indeed I would go further. It stands to reason that if the authors of ‘407 had realised that the water slurry step (for which they give no explicit reason) was actually affecting the crystals so as to remove solvate they would have said so.

In so holding I do not overlook the fact that to Dr Cunningham personally there was a disclosure of displacement. He expressed the view that this was the only inventive thing about the application. With all due respect I think he was probably applying some hindsight here, careful though he was to try to guard against it. The application as a whole is clearly disclosing for the first time the three forms of paroxetine hydrochloride as well as processes for making them. At best a displacement step is buried deep within Ex. 1. But whether or not Dr Cunningham had hindsight I am quite satisfied that the unimaginative skilled man would not work out for himself that the water slurry step is removing IPA from within crystals of solvate.

I am well conscious that in so holding I am differing from the Judge on a question of expert evidence – and that there must be an error of principle involved. I think there was –that the Judge failed to attempt to read Ex 1 in the context of the disclosure as a whole with the eyes and mind of the skilled man.

So, document for document, there is no anticipation. What about the alternative route – never mind what is said, if you carry out Ex 1 you will inevitably fall within the later claim? A major problem here is “more IPA”. For it is common ground that if you use the small SKB amount you will produce hemihydrate. That is so whether or not the actual SKB experiment (a “literal” repetition of Ex 1) was fair.

I think on this point alone the case for anticipation must fail. It simply is not the case that Ex 1 is “clear and unambiguous” – on the contrary it is, as my old pupil master Anthony Walton QC used to say – “biguous.” There was some debate over whether, if some would carry out a recipe one way, and others another, all having the level of skill of a skilled man, there could be anticipation if one of the ways would fall within a later claim. As I understand Synthon, such an argument would fail as a matter of law. So I need not go into it further from the point of view of anticipation.

But the argument will run as a matter of obviousness. If there are additions or alterations to the recipe which would be made by the ordinary skilled man and which would produce the anhydrate with bound IPA removed, then the claims would be obvious – just non-inventive ways of performing the recipe to make the anhydrate as promised by the recipe. So much was not in dispute.

One starts with this. It is clear that any skilled man asked to conduct Ex 1 to make anhydrate as promised will have to make a decision for himself as to how much IPA to use. SKB say he would use a small amount (to prevent hot spots), Apotex say he would use more. As I see it each route must be considered because Dr Cunningham favoured the latter route and the Judge accepted his position as being reasonable. It must be noted, however, that the reason Dr Cunningham chose the latter route was nothing to do with the nature of the end product – he would use more IPA for controlled crystallisation reasons only.

I start with the SKB contention. It is common ground that if the skilled man followed this route as written he would, as of the date of the Patent, make hemihydrate and not anhydrate. So where does he go next? SKB say he would realise that the problem was water. So he would take all precautions to keep water out. Thus he would not use conc. HCl to displace the acetate but, as told in the specification, dry HCl gas. And he would simply not carry out the water slurry stage for fear of converting his crystals to hemihydrate. If he analysed what he got he would find he had an anhydrous solvate but that some of the IPA was bound – could not be removed by drying. At that point he would try other things to remove IPA such as a recrystallisation in an organic solvent (not water for fear of hemihydrate). This point of view was supported by Dr Lee. Dr Cunningham accepted part of this reasoning – in particular the exclusion of water to make the “intermediate” product. But, upon finding that it was the solvate, he said he would try the water slurry step in Ex 1. Putting on one side any question of knowledge of displacement, his reason was simply that the step was stated in the example and it was supposed to produce the anhydrate. No scientific or technical reason is suggested – just go back to the recipe and see what it does. After all it would not be difficult to do and could be done with a teaspoon of the material made. Moreover it is a step put forward by a mighty and reputable research firm.

I am not convinced of this. To my mind it smacks of ex post facto analysis – the sort of stepwise approach deprecated in Fletcher-Moulton LJ’s judgment in British Westinghouse v Braulik (1910) 27 RPC 209:

“‘I confess that I view with suspicion arguments to the effect that a new combination, bringing with it new and important consequences in the shape of practical machines, is not an invention, because, when it has once been established, it is easy to show how it might be arrived at by starting from something known, and taking a series of apparently easy steps. This ex post facto analysis of invention is unfair to the inventors, and in my opinion it is not countenanced by English Patent Law.’

The difficulty with the argument is perhaps best stated in this way. If the skilled man uses the SKB amount of IPA he gets hemihydrate at the end of the process. If he makes solvate as described in Ex. 5 he finds he cannot remove some IPA contrary to what he is told. So at two places the specification is wrong. Why should he suppose that the slurry step has any effect on the crystals? Having departed from Example 1 by eliminating water why should he go back when he has a fear that it will make hemihydrate and no expectation that it will remove IPA?

The answer - essentially there is no harm in trying though there would seem no point, does not make trying obvious – a word whose Latin root is worth remembering – “in the way”. And here it is again worth remembering that we are in the field of obviousness, not novelty. For the latter a blind following of an apparently pointless or even apparently deleterious step, as part of an unambiguous instruction, will count.

Mr Watson pressed Dr Lee considerably on trying the slurry step. One of Dr Lee’s answers (that he would not risk his hard-won material with water) was the subject of particular scorn. And I agree the hard-won point is an exaggeration since only a small amount of product was needed to do the test. But trying to see the position with the knowledge of the skilled man, I can see no motive to carry out the slurry. It only appears to have a washing purpose at best.

Nor do I think it makes any difference that a skilled man, finding hemihydrate as the ultimate product, might well try to find out where things had gone wrong and analyse the intermediate product. Such a step would be entirely routine. Mr Watson submitted that he would find it to be the hemihydrate and so much seems right. But what the skilled man would do then, without imagination, is not so clear. Doubtless (and this was common ground) he would take steps to keep water out up to that stage – though how far he would go is at least a little speculative since the recipe from which he was working said use conc. HCl, which consists of over 60% water But I am far from convinced that he would think it worth trying putting the product he has kept dry up to that point into water.

So I do not think it was proved that an obvious modification of the recipe was to keep things dry until the intermediate and then to slurry it in water.

What then of the alternative route, using a large amount of IPA? As I have said Dr Cunningham was of the view that to start with such an amount was something he would have done by reason of a desire for a controlled crystallisation – nothing to do with keeping things dry enough that the hemihydrate would not be formed. Dr Lee, under pressure, accepted that this was “not unreasonable” thought he did not believe a skilled man would do this (see the passage of evidence quoted by the Judge at his paragraph 83).

Assuming therefore that it is established that such an amount of IPA would have been used, was it proved that it would indeed produce the anhydrate? Apotex did some experiments with one and two litres. These in part involved taking steps not mentioned in the recipe at all and in part choosing steps consistent with but not spelt out by the recipe. A convenient table showing the recipe, the SKB experiments and the Apotex experiments was set out by the Judge at paragraph 80 of his judgment and I reproduce it here:

The columns that matter are the first and third, i.e. what is said in the recipe and what Apotex actually did. The bits shown in bold are of most importance. SKB submit that use of analytical grade IPA (which will, unlike ordinary IPA contain no water), the use of nitrogen in preparing the intermediate (steps 13-14), the filtering under nitrogen (17) the further water wash (17) and the drying in 3 separate dishes when one could have been used) all call for explanation, both individually and collectively.

SKB invoke was what said by the EPO Board of Appeal in Union Carbide T396/89 [1992] EPOR 312:

“4.7

….. the Board observes that a party attacking the validity of a patent is free to choose his weapons of attack to suit his own convenience, taking into account relevant considerations of cost and effectiveness. If he seeks to establish that an example taken from a prior art document inevitably produces a given result, he thereby assumes the burden of performing his own repetition in such a way as to demonstrate that the repetition is valid. In the light of all the material before it, the Board is not satisfied that a valid repetition of Example 4 of document (1) would lead inevitably to a product falling within Claim 1, and the objection of lack of novelty based on this citation therefore fails.”

Mr Waugh submits that applies exactly here and as much to obviousness as to anticipation. Apotex have decided to rely on experiments which are supposed to show that just choosing one or two litres of IPA will be enough to prevent the formation of hemihydrate but those experiments show other precautions or steps. Why Apotex took them was never explained. Nor was there proof that the unspecified steps made no difference.

I find this most compelling. It is true that perhaps it was proved that the use of analytical grade IPA would make no difference – for the adventitious reason that Canadian conc. HCl contains a little less water than British conc. HCl so that overall the water level was about the same. But it was never explained why an ordinary skilled man, supposing he used one or two litres of IPA would carry out the next steps under nitrogen (whose purpose is to keep out water) when the next thing he is going to do (after drying) is put the product in water. Nor was it explained why the product was divided into 3 drying dishes when it could have been put into one.

The plain fact is that the experiments look contrived – litigation chemistry. In the Synthon case I castigated what appeared to me to be a deliberate departures (by both sides) from the actual recipe of the prior art. I said (para. 61):

“Again for the life of me, I cannot understand why the experimenters were not just given the unembellished disclosure”

That was said in December 2002 when this action had just started. It is most improbable that those concerned with this case were unaware of it. But whether they were or not, it stands to reason that if you want to prove that a skilled man carrying out a recipe will, using his ordinary skills, be able to do so, the thing to do is to give it (without any hints) to a scientist of ordinary skills and ask him to try. Of course lawyers do not like that approach – they lose control and fear that he will fail. But if he does, then probably the attack on that ground is no good. It may be that for a similar situation in future this sort of experiment is better conducted by a court appointed single joint expert. This is a matter which should at least be considered when an application for leave to adduce experimental evidence is given.

The Judge accepted that (with the irrelevant exception of analytical grade IPA) “all the steps taken by Apotex … are technically justified and do not justify the criticisms levelled at them” (para.85). That is a conclusion of fact (or evaluation) which I think is well within the Biogen appeal principle. So before I would interfere with it I must be sure he was wrong in principle. I think he was. His reasons are essentially in paragraphs 81-82:

“81.

The various ticks and crosses do not represent my views, but represent SKB’s contentions on the various steps taken by Apotex. It will be seen immediately that the only substantial differences lie in (1) the quantity of IPA said to represent ‘more IPA’ (2) the use of analytical grade IPA, which is completely anhydrous, and (3) the use of nitrogen to prevent contact with the air. Step 3 is specifically intended to prevent the ingress of moisture. This seems to me to be wholly correct, despite Mr Waugh’s protestations. The conditions are described as ideally anhydrous. This must on any view be within the variations allowed to the skilled man.

82.

The main question was therefore the amount of IPA. Dr Cunningham justified it ex post facto by a calculation…”

The heart of this is the Judge’s acceptance that the general teaching of ‘407 (“that the IPA solvate may be obtained by crystallisation from IPA, ideally under anhydrous conditions”) means that the stringent precaution of the use of nitrogen is a “variation allowed to the skilled man.” There are two errors here:

I therefore conclude that the Judge was wrong to conclude that ‘407 made the claims obvious.

I would add that I reach my conclusions about both novelty and obviousness without reliance on the hearsay documents about work done in Germany at Merck Darmstat and in Japan by Sumika after the date of the patent. They are reports of attempts to perform Ex 1 of ‘407 and show (I need not go into detail) that the workers tried to follow the recipe but produced hemihydrate, using initially small quantities of IPA. The Judge relied upon them in the BASF action (para 61) but says nothing about them in this case. At the very least, submitted Mr Waugh, he ought to have explained why he no longer thought they supported a defence to the novelty attack.

Mr Watson sought to use the documents to support his obviousness attack – saying they showed that the workers eventually did produce hemihydrate. I do not accept this submission – one simply does not have enough information as to what the workers knew, or whether or not they were told about the teaching of the Patent in suit when they did their experiments. They were done with the prospect of attacking the Patent in suit and its foreign equivalents – they were not attempts at original research but litigation chemistry So those who wanted the experiments done, at least, knew the answer. How can one be sure that a hint was not given on the way, at least after the initial experiment failed?

Mr Watson also suggested that SKB had made an admission against interest in US litigation concerned with the hemihydrate patent whose number in the US was ‘723. The Judge accepted this submission. But I do not think it is justified out if one reads the relevant document in full. It is said to be contained in “SB’s post trial reply brief”. The point in issue was whether the hemihydrate patent was enabling, the point being that all the examples, save for example 8, started with a seed of hemihydrate. Example 8 corresponds to example 1 of ‘407 but with an addition saying the anhydrate could be converted to hemihydrate by crushing (the water presumably coming from hygroscopic accretion).

The Reply Brief says:

“There is absolutely no evidence in the records that anyone has ever tried to practice any of the examples of the ‘723 patent and did not obtain paroxetine hydrochloride hemihydrate. This alone should end the enabling enquiries.”

Mr Watson’s point entirely ignores this sentence – which is entirely consistent with the evidence in this case. He relies on what is said in the next paragraph:

“Moreover, if seeds of hemihydrate are needed, example 8 of the ‘723 patent specifically describes how paroxetine hydrochloride hemihydrate will convert to the hemihydrate form under extremely high pressure.”

This ignores the wider context: the document as whole is not saying the only way to make hemihydrate is to make it from anhydrate by crushing. It is saying you get hemihydrate anyway with all examples, but if you got anhydrate you could convert it. In truth the real point at issue in this case simply was not alive in the US proceedings and it would be quite wrong to hold part of the detailed submissions there as an admission against interest. Even if it were it I am satisfied that it is rebutted by the direct evidence in this case.

This is a point raised by the respondents’ notice. It runs like this:

I would reject the argument. First, the reliance on General Tire is misplaced – the passage is about interpretation of the prior document. All it is saying is that its meaning does not change over time. A reader (and I suppose this is true of all documents) who wants to find out what the writer was actually saying, must try to read it with eyes of a contemporary to the publication. If an 18^th century recipe says “use dephlogisticated X” one uses what the writer must have meant – the ash of X after burning. One does not say there is no such thing as phlogiston.

Next the factual basis is too flimsy. Prior to trial the point was not taken in the pleadings or evidence so there was very little on the point (an acceptance by Dr Lee of point vi above) is merely on the assumption that there is a remote place where anhydrate would be produced.

Further one simply does not know whether there would have been a spread of seeding by 1987, the date of publication of ‘407. Nor does one know whether in fact if one started in a remote place one would get anhydrate. Just because that was what happened at Harlow for some years does not tell you what would have happened at a start-up elsewhere. It may be something of a fluke that Harlow ever produced anhydrate.

All that can be said with confidence is that no one, in remote locations or not, has ever tried to conduct Ex. 1 without producing hemihydrate.

Accordingly I do not think the perhaps difficult but probably esoteric question of law about “once available always so” arises in this case and I do not propose to say more about it. The Bodmin point fails.

The case here runs in the following steps:

The Judge dealt with the point in two short paragraphs, 92 and 93. In paragraph 93 he sought to apply the Windsurfing analysis. He said:

“This is a case in which the Windsurfing analysis is helpful. The difference between the inventive concept of the patent (displace the solvent of solvation using water or some other displacing agent) differs from that of the erythromycin patent only in the pharmaceutical concerned and the conditions under which the displacement takes place. The difference between the approaches of the experts under cross-examination was striking. Dr Lee was most unconvincing. He returned to the ‘distance’ between paroxetine hydrochloride and erythromycin on a number of occasions. But there is no difference in problem, and there is a suggested solution. Dr Cunningham was cross-examined on the basis that there were other solvents one would try instead (a contention, incidentally, which when advanced by BASF resulted in a finding of invalidity in the BASF proceeding). This is one of those cases where all the suggested approaches to the problem are obvious to try. The cross-examination of Dr Cunningham proceeded on the basis that a complete and detailed analysis of the thermodynamics would suggest that it was not going to work: and that there is a slight suggestion in the erythromycin patent, fastened on by Dr Lee, that what was happening was a recrystallisation. I do not think that this affects the analysis. It was obvious to try. There was no suggestion that it would not work, if temperatures suitable for paroxetine hydrochloride solvate were selected.”

I think the Judge erred in principle here. The skilled man has his common general knowledge – the mental tools of his trade – but no more. The law of obviousness supposes that he can be given any individual piece of prior art and read it with that knowledge. The piece of prior art forms part of the “state of the art”. What he cannot do is to just link one piece of prior art with another, unless so to do would itself be uninventive. No-one disputes what Lord Reid said in Technograph v Mills & Rockley [1972] RPC 346 at page 355:

“In dealing with obviousness, unlike novelty, it is permissible to make a ‘mosaic’ out of the relevant documents, but it must be a mosaic which can be put together by an unimaginative man with no inventive capacity.”

It is that which must be applied here. If one starts with the Buxton and Lynch paper there is nothing about bound IPA at all. Nor is there anything which would suggest that it should be read with the erythromycin patent. So just putting the two papers together is an obvious illegitimate mosaic.

That is not how Apotex put the case however. They suggest that it is legitimate to suppose that the skilled man tries Buxton and Lynch and discovers for himself the unbound solvent problem. He would then, it is suggested, be entitled to take the information from the erythromycin patent and say “displacement worked there, I should try it here”.

Such an analysis does not comply, as the Judge seemed to think it did, with the Windsurfing analysis:

In short, the argument based on the erythromycin patent involves combining two documents, the separate contents of which are not common general knowledge. Carrying out an experiment on one of them does not make an illegitimate mosaic legitimate.

I am confirmed in my view by the way that Dr Cunningham found the erythromycin patent. Once he knew of displacement he wondered whether there might be an earlier publication of a similar step. He found the obscure erythromycin patent (as it happens via a text book published after the date of the patent in suit). Only when you know that displacement is possible do you find the erythromycin patent. That is to work backwards.

Thus far it has not been necessary to consider the detailed wording of the claims, it being common ground that the attacks based on ‘407 and the erythromycin patent would, if correct, invalidate them on any construction. But having held that the attacks fail, I must turn to the meaning of the claims.

I begin with that aspect of the claims which led the Judge to conclude that Apotex’s process is not covered. What Apotex do is to make the IPA solvate and dry it down to just over 2% IPA. This is then treated with acetone for a time. The product is heated and dried under vacuum. I deliberately do not give details because they are confidential. It is sufficient to observe that the final product contains, as I have said, about 0.4% IPA and about 0.4% acetone. So the acetone has got into the product.

Technically what is happening may well be what SKB described in a now abandoned patent application, WO/32594. What this is about is contacting the IPA solvate with a “transformation solvent”. This changes the IPA solvate to a mixed solvate, “which is more easily desolvated.” The specification goes on to show that heating and drying will now allow removal of solvent. One of the suitable transformation solvents specifically identified is acetone – what Apotex use. The example (10) shows a drying under vacuum at 60⁰C (after 20 hours altogether) down to an amount which must be 0.1% or less IPA and 1.0% acetone.

Apotex say that nothing like that is disclosed in the Patent in suit. When it speaks of “displacing” and “displacing agent” what it is talking about is a process which happens while the solvate is slurried. It is this process which causes the bound IPA or other solvent to come out.

The Judge accepted this submission. He said:

“I conclude that a displacing agent is as its name suggests a material that displaces the unwanted solvent from the material without replacing that solvent over the time during which the material is exposed to it. There is no justification for going beyond the natural meaning of the words, and they were not shown to be words of art” (para. 49)

And:

“”A displacing agent is an agent that displaces the solvent of solvation from the material in which it is bound so as to reduce the amount of solvent of solvation to a level lower than that achievable by drying” (para,51)

Was the Judge right so to conclude? I think he was and essentially for the reasons he gave. Patent claims are to be construed purposively, their meaning to be ascertained from their context as the recent decision of the House of Lords in Kirin-Amgen v Hoechst Marion [2004] UKHL 46 has confirmed. So to find out what the patentee had in mind by “displacing agent” one goes to the body of the specification. The key passage is that I quote in full at para. 25 above. Certain sentences make it clear that the displacement occurs during the slurry step. I pick them out here to emphasise them – they have the same meaning in their context:

“The paroxetine hydrochloride solvate produced is suitably isolated and dried ……The bound solvent is then displaced with a displacing agent such as water or supercritical carbon dioxide.

It is important that the paroxetine hydrochloride solvate is contacted with enough water and for sufficient time to displace the solvent but insufficient to cause conversion to the hydrochloride hemi-hydrate.

The amount of water, the form of the water, e.g., liquid or gaseous and the length of time which the paroxetine hydrochloride solvate is contacted with the water differs from solvate to solvate. This depends largely upon the solubility of the solvate in question.

After contact with water to displace the bound solvent the product is suitably dried, for example, in vacuo at elevated temperature. Suitable drying may be over a desiccant such as phosphorus pentoxide.

When supercritical carbon dioxide is used it should be appreciated that the flow rate, temperature and pressure of the carbon dioxide may be controlled to give optimum solvent removal from the paroxetine hydrochloride solvate.

None of these sentences read on to a process in which solvent is only removed during the drying stage – indeed as Mr Watson pointed out the carbon dioxide removal does not have such a stage (see para. 26(4) above). Moreover the Patent clearly does not regard the drying stage after the slurry as of any significance – it is plainly just to remove surface solvent not to get material out of the crystals themselves.

A yet further indication that the use of acetone was not seen as possible displacing agent lies in the fact that acetone is specifically mentioned as the sort of solvent which forms bound, unremovable solvate – see para. 24 above. The patent is saying acetone is a problem solvent so it can hardly be contemplating it as a displacing agent.

Mr Waugh argued for a wider construction – submitting that a “displacing agent” included anything that got the solvent out of the crystals, whether that happened during the slurry or in the subsequent drying. He pointed out that until the drying was complete the agent would still be in contact with the crystals. What mattered, he submitted, was removal – that was the inventor’s purpose and so purposive construction favoured this construction. I reject that. The Patent in suit is based on a particular discovery – that you can get bound solvent out in a slurry with an appropriate agent. Its claims are correspondingly limited and do not extend to another way of getting the agent out.

Mr Waugh also relied on certain Apotex documents which, he submitted, showed that Apotex themselves regarded the acetone as reducing the bound IPA and that the whole procedure, slurry and drying as a “solvent displacement procedure.” But that is simply to take words in one context and say they have the same meaning in the Patent. That is not legitimate. What I do not read in the documents is that Apotex think that there is any IPA removal going on during the slurry itself.

I also, like the Judge, am impressed with the lack of evidence that any displacement takes place during the Apotex acetone slurry. Mr Waugh said that any attempt to do so would be bound to be criticised because there had to be a subsequent drying and you cannot find out if there is bound solvent until you have dried. I regard that as lame. It was not shown that removed IPA (if any) could not be detected somehow - in acetone drained from a slurry for instance. Besides, SKB’s own experiment 10 in their abandoned transformation application shows that one can (by testing at different drying times) see what is being removed.

I therefore think the Judge was right in his construction of “displacing agent” and in his consequential finding that Apotex do not fall within the claim.

That being so it is unnecessary to go further. Indeed we are not asked to do so. By their respondents’ notice Apotex raised a number of arguments about the construction of the claims, contending that they were not infringed or invalid for insufficiency (either because too broad or because ambiguous). All these arguments were contingent on the judge’s finding of non-infringement not being disturbed. In the event I have upheld that finding. Since I understand that the other members of the Court agree, we are not asked to make any ruling on these points and do not do so.

That is not to say that it should be thought that there is nothing in them. On the contrary it appeared to me that they had considerable force. For the public record (a matter of some importance since, I understand, the remainder of the Court agrees, the patent will be restored to the register) I set forth the most significant of these points, without deciding any of them.

In the result I would reverse the Judge’s finding of invalidity by reason of anticipation/obviousness over ‘407 and uphold his finding of non-infringement.

I agree.

I also agree.