NAPP Pharmaceutical Holdings Ltd v Dr Reddy's Laboratories (UK) Ltd & Anor

The Claimant (“Napp”) is the proprietor of European Patent (UK) No. 2 305 194, the specification of which is entitled “A buprenorphine transdermal patch for use in the treatment of pain for a dosing interval of at least 7 days” (“the Patent”). Napp is also the proprietor of European Patent (UK) No. 1 731 152 (“152”). Napp contends that the First Defendant (“Dr Reddy’s”) and the Second Defendant (“Sandoz”) threaten to infringe the Patent and 152 by marketing branded generic 7 day buprenorphine transdermal patches which Dr Reddy’s and Sandoz propose to launch. It is common ground that the allegation of infringement of 152 adds nothing to the allegation of infringement of the Patent, and accordingly attention may be concentrated on the Patent.

The Patent and 152 are two members of a family of (so far) one parent and eight divisional patents and applications. The parent patent, European Patent No 0 964 677, was revoked by the Technical Board of Appeal of the European Patent Office, although there are outstanding petitions to the Enlarged Board of Appeal for review on the ground of alleged serious procedural irregularity. Another granted patent, European Patent No 1 570 823, has been revoked by the Opposition Division, a decision which is under appeal to the Technical Board of Appeal. On the other hand, three oppositions to the Patent were rejected by the Opposition Division for the reasons given in a written decision dated 22 December 2014, which is also under appeal. Five further applications remain in prosecution. In the present proceedings there is no challenge to the validity of the Patent save for an allegation of insufficiency advanced by Dr Reddy’s by way of a squeeze on construction. Nor is there any challenge to the earliest claimed priority date of 24 February 1997. The Patent and 152 will expire on 24 February 2018.

Buprenorphine is an opioid indicated for the treatment of non-malignant pain of moderate intensity when an opioid is necessary for obtaining adequate analgesia. Examples of the clinical uses of 7 day buprenorphine transdermal patches include the treatment of chronic back pain and osteoarthritis. The fact that such patches last for 7 days makes them more convenient and cost-effective to use than patches which last a shorter period.

Napp is a company within the Mundipharma network of independent but associated companies. Napp’s subsidiary Napp Pharmaceuticals Ltd (“NPL”) has marketed a 7 day buprenorphine transdermal patch under the brand name BuTrans since 2005. There is no dispute that this has been a very successful product, with total sales now approaching 10 million patches a year. Since 1 February 2016 another Napp subsidiary, Qdem Ltd (“QL”), has sold a 7 day buprenorphine transdermal patch under the brand name Butec. For regulatory reasons, buprenorphine transdermal patches are required to be sold under brand names. Butec is identical to BuTrans and is in effect an originator’s own branded generic which had been launched to protect Napp’s commercial position in anticipation of loss of exclusivity. Both BuTrans and Butec are manufactured by LTS LOHMANN Therapie-Systems AG (“LTS”) in Germany.

Sandoz intends to market a 7 day buprenorphine transdermal patch which is manufactured by Hexal AG in Germany under the brand name Reletrans. Sandoz obtained a UK marketing authorisation for Reletrans on 10 February 2016, and it has been ready to launch the product since about mid March 2016.

Dr Reddy’s intends to market a 7 day buprenorphine transdermal patch manufactured by an unidentified manufacturer in an unidentified country. On 15 January 2016 Dr Reddy’s solicitors wrote to the previous proprietor of the Patent setting out a description of the patch which Dr Reddy’s proposed to market and seeking an acknowledgement that such marketing would not infringe the Patent. At that time, Dr Reddy’s anticipated receiving a marketing authorisation “shortly” and planned to launch around the end of May 2016. As at the last day of the trial, however, Dr Reddy’s had not yet obtained a marketing authorisation. Moreover, Dr Reddy’s had only just got to the point of starting production. As I will explain, this has prompted Dr Reddy’s to adopt an approach to this dispute which has led to certain complications.

There is no dispute that Sandoz has obtained its marketing authorisation, and Dr Reddy’s is seeking its marketing authorisation, using BuTrans as the reference product. It follows that Sandoz’s and Dr Reddy’s products are intended to be bioequivalent to BuTrans. It is debatable whether BuTrans falls within the claims of the Patent, however. In any event, Sandoz contends that it has, and Dr Reddy’s contends that it is able to, work around the Patent. There is nothing wrong in working around a patent. On the contrary, part of the point of patent claims is to enable third parties to know what they have to do to avoid infringement. As is often the case, the issue between the parties is whether the attempt to do so has been successful. As is also often the case, this turns primarily upon the correct interpretation of claim 1 of the Patent. But in the present case, it may also depend upon the variability in the products which are to be manufactured on behalf of Sandoz and Dr Reddy’s. It is common ground that there is some degree of inherent variability, but the parties are disagreed as to the extent of such variability and as to the consequences of it.

Napp’s claim against Sandoz is primarily a quia timet one and its claim against Dr Reddy’s is a wholly quia timet one. I shall explain what is meant by quia timet, and the significance of this, below. In that regard, Napp relies on the fact that, if Sandoz and Dr Reddy’s are permitted to market their respective products, it is likely that they will sell substantial quantities of patches before expiry of the Patent. Thus it is not unrealistic to suppose that they could each sell as many as 2-3 million patches in that period, although it is inevitably uncertain how many they will in fact sell.

Napp commenced the present claim on 19 February 2016. On the same date Napp launched an urgent application for an interim injunction against Sandoz, which led to Sandoz giving temporary undertakings on 22 February 2016. On 22 February 2016 Napp launched an application for an interim injunction against Dr Reddy’s. On 15 March 2016 Dr Reddy’s counterclaimed for a declaration of non-infringement. On 16 March 2016 I made an order for an expedited trial of the claim on 7 and 8 June 2016, as a result of which the Defendants gave undertakings not to launch their products pending the determination of the claim. At that stage, it was envisaged that expert evidence would only be required from experts in transdermal delivery systems (“TDSs”).

On 21 April 2016 Dr Reddy’s applied for permission to adduce a report from an expert statistician. On 9 May 2016 I gave all the parties permission to adduce evidence from experts in statistics and enlarged the trial to three days. With the benefit of hindsight, it should have been appreciated by the parties that statistical evidence would be required at an earlier stage. Be that as it may, the parties and their respective statisticians had to work very hard at very short notice in order to put the required statistical evidence before the court. This process was further complicated by the fact that data relating to Dr Reddy’s proposed product became available shortly before the trial started. This not only required the experts to consider that data under considerable time pressure, particularly so in the case of Napp’s expert, but also led to Dr Reddy’s changing its case in certain respects. In those circumstances it is not surprising that, through no fault of the experts, the statistical evidence was not as well considered as it ideally would have been. I have borne this in mind in my assessment of that evidence.

In the result, the parties and their representatives are to be congratulated on preparing what turned out to be a more complicated case than originally envisaged for trial in less than four months.

Both Sandoz and Dr Reddy’s regard the details of their respective products and manufacturing processes as confidential. This also complicated both the preparation for trial and the trial somewhat, although the complexity was minimised by a sensible agreement that all the lawyers and experts should have access to all the confidential information. In my view it is important that as much of my reasoning as possible should be contained in a public judgment, and I have approached the preparation of this judgment on that basis. Certain details are contained in two confidential annexes, however, one for Dr Reddy’s and one for Sandoz.

As explained above, each side called an expert in TDSs and an expert in statistics.

Napp’s expert was Dr Kenneth Miller II. Dr Miller obtained a BS in Chemical Engineering from Carnegie Mellon University in 1986, an MS in Chemical Engineering from the University of West Virginia in 1988 and a PhD on topical local anaesthetic formulations from the University of Florida in 1991. After two years as a post-doctoral researcher at the University of Florida and the University of Utah, he was employed as a Senior Research Scientist developing novel drug delivery systems, including transdermal systems, from 1993 to 1995. From 1995 to 1998 he was Adjunct Professor in the Department of Chemical Engineering at Pennsylvania State University, where he oversaw the creation of a transdermal research laboratory. During the same period he was employed as an Associate Director of the Franz Dermatopharmacology Research Laboratory in the McClellan VA Hospital in Arkansas. From 1997 to 1998 he was also Assistant Professor in the Department of Pharmaceutics at the University of Arkansas, where he was again responsible for a transdermal research laboratory. From 1998 to 2012 he was employed in a series of increasingly senior transdermal research and development and project management roles in the Mylan group of companies, which is a major developer and manufacturer of transdermal patches. During his career at Mylan he was responsible for the development of products from project inception through to commercial launch. From 2012 to 2013 he was Head of Research and Development at IGI Laboratories Inc. Since then, he has run a consultancy firm.

Counsel for Sandoz submitted that Dr Miller’s evidence with respect to the construction of the Patent should be treated with caution. Napp initially instructed Professor Jens Hansen of Copenhagen University as its expert. After the expedition of the trial, Prof Hansen was unable to continue to give evidence on behalf of Napp due to previous commitments. Accordingly, Napp instructed Dr Miller instead. Parts of Dr Miller’s main report were copied from a report prepared by Prof Hansen without any acknowledgement. It appears from Dr Miller’s evidence that he adopted this wording, albeit with some modifications and additions, after he had been presented with it in the form of a draft of his own report. Although he stated that he agreed with it, it also appears from his evidence that it did not represent the opinion he formed when he first read the Patent. This approach to the preparation of expert evidence must be deprecated. I agree with counsel for Sandoz that it follows that Dr Miller’s evidence on this topic must be treated with some caution. What is more important, however, is the substance of Dr Miller’s reasoning, which I shall consider below.

The Defendants’ expert was Professor Adrian Williams. Prof Williams obtained a BSc in Biology and Chemistry in 1987 and a PhD on penetration enhancers for human skin under the supervision of Professor Brian Barry, a leader in the field of transdermal drug delivery, from the University of Bradford in 1990. From 1990 to 2002 he was successively Lecturer, Senior Lecturer and Reader in Pharmaceutical Technology, and from 2002 to 2004 Professor of Biophysical Pharmaceutics, at the University of Bradford. From 2004 to 2015 he was successively Professor of Pharmaceutics, Head of the School of Pharmacy and Head of the School of Chemistry, Food and Pharmacy at the University of Reading. Since 2015 he has been Research Dean for Health at the University of Reading. He has researched and published extensively in the field of transdermal delivery. He is the author of a textbook entitled Transdermal and Topical Drug Delivery: From Theory to Clinical Practice (2003) and has contributed chapters to numerous other books, including the chapter on “Topical and transdermal drug delivery” in the core undergraduate textbook Aulton’s Pharmaceutics: The Design and Manufacture of Medicines (ed Aulton and Taylor, 4^th ed, 2013). He has published 92 original peer-reviewed papers, of which 58 were directly related to transdermal or topical drug delivery. He has acted as a consultant to a number of pharmaceutical companies, which has involved him visiting the industrial facilities of such companies, and he sits on the Scientific Advisory Board of MedPharm, a pharmaceutical development company that specialises in topical and transdermal product development.

Counsel for Napp rightly accepted that Prof Williams had done his best to assist the court with measured answers, but pointed out that Prof Williams had fairly said that he would defer to Dr Miller’s experience when it came to issues over the industrial manufacture of transdermal patches.

Napp’s expert was Dr Kevin Carroll.After obtaining a BSc in Mathematics and Statistics from Leeds University in 1999, Dr Carroll was employed by AstraZeneca UK Ltd for 23 years as a statistician in a series of increasingly senior roles, rising to the most senior statistical position in that company. During that time he obtained an MSc in Applied Statistics from Sheffield Hallam University in 1993 and a PhD in Statistics from the University of East Anglia in 2013. He now runs a consultancy company specialising in statistics within the pharmaceutical industry.

Counsel for Dr Reddy’s submitted that Dr Carroll had approached the case with preconceptions which he was unwilling to depart from, while counsel for Sandoz submitted that Dr Carroll had been an advocate for Napp’s case. I do not accept these submissions. In my judgment Dr Carroll did his best to assist the court, although a couple of the points he raised were of more theoretical concern than practical importance.

The Defendants’ expert was Professor Peter Sasieni. Prof Sasieni obtained a BA in Mathematics from Cambridge University and a PhD in Biostatistics from the University of Washington in Seattle. He is currently Professor of Biostatistics of Cancer Epidemiology and Deputy Director of Cancer Prevention at the Wolfson Institute of Preventive Medicine at Queen Mary, University of London. Although the focus of Prof Sasieni’s recent work has been on cancer prevention, he has undertaken statistical research and provided statistical analysis in other fields. He has acted as an expert witness in a number of patent cases in England and elsewhere. When considering Prof Sasieni’s evidence, it needs to be understood that he was initially instructed solely on behalf of Dr Reddy’s, and only subsequently was he instructed on behalf of Sandoz and asked to analyse the data in Sandoz’s PPD.

Counsel for Napp criticised Prof Sasieni’s evidence on two grounds. First, he submitted that Prof Sasieni had stepped outside his expertise in stating in Annex A to his second report that it was “an entirely reasonable assumption” that the variability within a batch did not vary from batch to batch. I do not accept this criticism. As Prof Sasieni explained, and as is obvious, statisticians have to make assumptions in their work. One would hardly expect a statistician to make an unreasonable assumption. Thus Prof Sasieni’s statement that he considered the assumption a reasonable one merely made explicit what would otherwise have been implicit. What matters is that Prof Sasieni clearly identified the assumption he was making, thereby enabling the validity of that assumption to be tested. Furthermore, Prof Sasieni had expressly pointed out, in the context of stating another assumption in his first report, that he was not an expert in manufacturing or chemical engineering.

Secondly, counsel submitted that Prof Sasieni had ignored evidence which undermined the assumption that the variability within a batch did not vary from batch to batch. I do not accept this criticism either. In fact Prof Sasieni expressly stated in Annex B to his third report that it was clear from his own analysis of the levulinic acid data in the Sandoz PPD that the estimated within-batch variation was considerably less than the between-batch variation. Counsel suggested that he ought to have considered the buprenorphine data as well, but Prof Sasieni explained that this was due to his instructions. In any event, this would have added little to the conclusion to be drawn from the levulinic acid data.

Each party adduced evidence from at least one factual witness, none of whose evidence was challenged by cross-examination. Napp relied on a witness statement of Steve Smith, Regulatory Affairs Manager, Chemistry Manufacturing and Control at Mundipharma Research Ltd, which provides services to NPL and QL. Mr Smith described the composition of BuTrans and Butec and certain tests undertaken by LTS to validate the manufacturing process and exercise quality control.

Dr Reddy’s relied on the declarations of Tony Mayhall, Director and Qualified Person/Head of Quality (Europe Generics) at Dr Reddy’s, who verified Dr Reddy’s Product and Process Description (“PPD”) and its Supplemental PPD.

Sandoz relied on three witness statements of Thomas Kohr, a scientist in the formulation department at Hexal, which verified Sandoz’s (second) PPD, described tests undertaken by Hexal to validate the manufacturing process and verified additional data which were added to the PPD as Annexes 8-11.

Human skin is a multi-layered organ that prevents the ingress of foreign organisms and substances whilst regulating heat and water loss from the body. The skin is made up of three main layers: the subcutaneous fatty tissue, the dermis and the epidermis. The epidermis is itself multi-layered. The outermost layer of the epidermis is the stratum corneum, which provides most of the skin’s barrier properties.

Blood vessels extend into the dermis, and thus molecules which pass through the epidermis are rapidly carried away systemically by the blood. This creates a concentration gradient which enables transport of a drug which can permeate the epidermis to the systemic circulation. Drug permeation can be enhanced by the use of an occlusive backing film, which retains moisture and thus softens the stratum corneum, and in other ways described below.

By 1997 transdermal administration was an established method of delivering drugs. Transdermal drug delivery provides the ability to deliver medications over a long time and at a relatively constant rate. This reduces the variability associated with cyclic blood levels which accompany frequent periodic dosing such as that provided orally or by injection. TDSs are also simple to apply, non-invasive and painless. Furthermore, dosage can be controlled simply by changing the size of the patch. There are relatively few drugs which are suitable for delivery across the skin, however.

The rate-limiting step in the delivery of such drugs is usually the rate of diffusion across the stratum corneum. In order to drive drug transport across the skin by maintaining a concentration gradient, it is necessary for there to be more active pharmaceutical ingredient (“API”) in the patch than will be delivered to the patient, and there will usually be a significant excess or “overage”. Drug permeation may be improved with the use of permeation (or penetration) enhancers or other methods. The rate of diffusion per unit area is known as the flux.

TDSs fall into two main types: liquid/gel reservoir patches and solid matrix patches. The former have a cavity filled with liquid or solid drug between a backing layer and an impermeable membrane. The latter mix the adhesive and the drug in the same layer: the surface that adheres to the patient also delivers the drug. This case is concerned with solid matrix patches.

Although there were a number of TDSs approved on the market by the priority date, including patches containing the opioid fentanyl, this was a very small number when compared with the number of drugs available on the market generally. TDSs nearly always contained well-characterised and familiar drugs with an established history.

No TDS containing buprenorphine had been approved by a regulatory authority by February 1997.

Among the factors that need to be borne in mind when designing and developing a TDS are the following:

These factors must be balanced so as to result in a TDS displaying:

In developing a TDS, a skilled formulator will initially produce trial patches on a laboratory scale and test their release properties using standard in vitro methods. Once a patch has been developed which appears satisfactory in vitro, it will usually be tested in an animal model. After that it will tested in the usual sequence of clinical trials, namely Phase I to evaluate safety and metabolism in healthy volunteers, Phase II to establish an effective dose and Phase III to test whether the product is effective on a statistically significant basis. For Phase III trials, as well as subsequent commercial manufacture, the manufacturing process will need to be scaled up to industrial methods and quantities.

Although the skilled formulator will attempt to devise a robust formulation that can be consistently manufactured on a commercial scale, he will be aware that there is inherent manufacturing variability in TDSs for the reasons explained below. Accordingly, he will not rely upon exceedingly tight tolerances. On the contrary, the skilled formulator would usually anticipate that reasonably broad tolerances would be acceptable for the following reasons:

Matrix TDSs are manufactured by mixing all the components of the matrix in one or more solvents to produce a drug adhesive blend. This is then coated onto a substrate, which is usually a disposable polyester release liner. The coated substrate is then dried in a convective drying system of one or more zones with differing temperatures and air speeds to remove the solvents. Once dried, a cover, which is usually an occlusive backing film, is continuously laminated to the coated surface to produce a drug adhesive laminate. This is then cut into individual patches. Production is carried out in batches rather than continuously.

There are a number of sources of variability inherent in the process including the following:

Testing will ensure that the expected flux rates are achieved. Hence, the skilled formulator can be assured that any variation is not impacting on the flux rate of the drug, which is what is important for delivery.

Regulators require the consistency of the manufacturing process to be validated. This will include assaying certain aspects of the composition of the product, but not necessarily all aspects (e.g. assaying of some excipients may not be required). Regulators do not require absolute consistency even in those respects which they required to be tested, however.

As will appear, the Patent refers in a number of places to International Patent Application No. WO 96/19975 entitled “Transdermal resorption of active substances from supercooled masses” in the name of LTS (“Hille”). It is common ground that the skilled reader would read the Patent together with Hille. It is convenient to begin with the disclosure of Hille before turning to the Patent. Hille was published in German, and I shall quote from the agreed English translation.

Hille explains the problem to which the invention is directed in the following way (at page 1):

“Experiments have shown that dodecanol has a penetration-enhancing effect … Since its melting point is at 24^oC, dodecanol is solid at normal room temperature. The provision of dodecanol out of a matrix is made more difficult by the solid physical state. This shortcoming limits the possible uses of dodecanol in polymer matrices, because the delivery rate of a substance out of polymer matrices is greater, the more easily it can diffuse in the matrix. This applies not only to dodecanol, but also to other penetration-enhancing substances which are solid at room temperature. The task of the invention is therefore to improve the delivery of excipients, which are solid at room temperature, out of a matrix.”

The invention solves this problem by the use of an excipient, such as dodecanol or levulinic acid, which forms a “supercooled melt”, that is to say a compound with a melting point that lies above room temperature, but which remains in the liquid state “after a melting process during cooling to room temperature”.

The invention is illustrated by a number of examples. Hille first describes “Example 1” as follows (at page 3):

“1.139

g of a 47.83 w/% polyacrylate solution of a self-crossing acrylate copolymer of 2-ethylhexylacrylate, vinyl acetate, acrylic acid (solvent: ethyl acetate:heptane:iso-propanol:toluene:acetylacetonate in the ratio of 37:26:26:4:1), 100 g levulinic acid, 150 g oleyl oleate, 100 g polyvinylpyrrolidone, 150 g ethanol, 200 g ethyl acetate and 100 g buprenorphine base are homogenised. The mixture is stirred for about 2 hours and checked visually to see if all the solids are dissolved. The evaporation loss is controlled by weighing, and, if necessary, replenishing the loss of solvent with ethyl acetate.

The mixture is then spread onto a 420 mm wide transparent polyester film, so that the surface weight of the dried adhesive layer is 80 g per m². The polyester film, which can be removed again with a silicon treatment, serves as a protective layer.

The solvent is removed by drying with heated air which is passed over the damp track. The heat treatment does not only cause the solvents to evaporate, but also melts the levulinic acid. The adhesive film is subsequently covered with a 15 μ polyester film. A 16 cm² area is punched out with appropriate cutting tools, and the edges remaining between the individual systems are removed.”

It is common ground that:

Hille then states that “Examples 2 to 5 were conducted in a manner analogous to Example 1”. It goes on to describe the columns of Table 1, which include the relative quantity of buprenorphine that diffuses through mouse skin in vitro over 24 hours. It then states (at page 4):

“The composition of Examples 2-5 are shown in Table 1.

The results of the in vitro penetration tests and the qualitative and quantitative compositions of buprenorphine TTS [trans-dermal therapeutic systems] according to Example 1-5 are shown in Table 1”.

I reproduce Table 1 below:

A note to Table 1 states that “Cetiol = oleyl oleate DAB 10”. It is common ground that:

It is common ground that the skilled reader would understand that the “%” figures for the components in Table 1 are percentage weights. Hille does not state in terms whether these figures refer to the ingredients in the recipe for the matrix (i.e. the inputs into the manufacturing process, excluding the solvent which is removed during drying) or the composition of the finished product (i.e. the outputs from the manufacturing process). Nor does it explain how they were calculated (if inputs) or measured (if outputs). The Defendants contend, however, that the skilled reader would understand from Hille’s statement that Table 1 shows “quantitative compositions of buprenorphine TTS” that the figures are output %-wts. I agree with this for similar reasons to those I shall give in relation to the Patent (see paragraphs 104-110 below).

Either way, it can be seen that Example 1 in Table 1 is not the same as Example 1 in the text of Hille. Example 1 in Table 1 has 10% glutaric acid monomethyl ester, whereas the text specifies levulinic acid in an amount which can be calculated to be 10 %-wt of the inputs, and it has 10% oleyl oleate, whereas the text specifies oleyl oleate in an amount which can be calculated to be 15 %-wt. For this reason, I will refer to Example 1 in the text of Hille as “Example 1a”, and Example 1 in Table 1 as “Example 1b”. Hille does not report any penetration test results for Example 1a.

It can also be seen that the example which gave the highest penetration result was Example 3, which gave 38.6%. The next highest was Example 1b, which gave 28.6%. The next highest was Example 5, which gave 27.3%. Examples 2 and 4 were rather lower, at 5.9% and 7.7% respectively.

Hille comments on these results as follows (at page 5):

“From Column 4 it can be seen that the dissociation constants of the carboxylic acids used are quite similar. If, in each case, the emollient oleyl oleate is used in 10% concentration, however, it is evident from Examples 1 and 3 or 2 and 4 that only the carboxylic acids, which tend to form supercooled melts, cause significant penetration enhancement under in vitro conditions. As can be seen from the dissociation constants, undecenoic and octanoic acid are weaker acids than glutaric acid monomethyl ester or levulinic acid. This was offset by the fact that with 15% the concentration of the two weaker acids is higher than the concentration of the two stronger acids. At the same time, however, it is evident that the dissociation constants are so close together, that the resorption-promoting effect cannot be explained by the dissociation constants. Example 5 indicates that, even with an acid that obviously has little impact on the penetration-promoting effect of buprenorphine base, an increase of resorption can be achieved when a neutral substance, namely dodecanol, which like levulinic acid or glutaric acid monomethyl ester forms supercooled melts, is used. The physical effect of the listed excipients is sufficiently substantiated by the 5 examples.”

Thus Hille attributes the difference between Examples 1b, 3 and 5 on the one hand, and Examples 2 and 4 on the other hand, to the fact that glutaric acid in Example 1b, levulinic acid in Example 3 and dodecanol in Example 5 formed supercooled melts, whereas octanoic acid in Example 2 and undecenoic acid in Example 4 did not. Hille also explains that octanoic acid and undecenoic acid were included at 15% rather than 10% to compensate for the fact that they are weaker acids than glutaric acid and levulinic acid (although it can be seen from the pKa values that undecenoic acid is only slightly weaker than levulinic acid).

The specification of the Patent begins by explaining the benefits of sustained release analgesics such as sustained release oral formulations. The specification then states at [0004]:

“Another approach to sustained delivery of a therapeutically active agent are transdermal delivery devices, such as trans-dermal patches. Generally, transdermal patches contain a therapeutically active agent (e.g., an opioid analgesic), a reservoir or matrix containing the opioid or other active ingredient(s) and an adhesive which allows the transdermal device to adhere to the skin, allowing for the passage of the active agent from the device through the skin of the patient. Once the active agent has penetrated the skin layer, the drug is absorbed into the blood stream where it can exert a desired pharmacotherapeutic effect, such as analgesia.”

The specification acknowledges at [0005] the commercial availability of a fentanyl patch which is said to provide adequate analgesia. It then states at [0006] that buprenorphine transdermal delivery devices are of particular interest because buprenorphine is a potent, partial agonist opioid analgesic with desirable therapeutic properties. For example, buprenorphine is 50 to 100 times more potent than morphine, but has a much safer therapeutic index than morphine.

The specification goes on to explain at [0008] that buprenorphine has a low oral bioavailability and, like other opioids, carries an addiction risk which may be more significant where the form of administration results in spikes in the levels in the blood. Both of these issues are addressed by using transdermal patches as the delivery method.

Having acknowledged a prior art buprenorphine transdermal delivery device, the specification states at [0010]:

“Despite these advances in the art, there remains a need for methods of treating patients with buprenorphine that provide effective analgesic levels of buprenorphine for prolonged periods of time while eliminating or minimizing dependence, tolerance, and side effects, thus providing a safe and effective method of pain management. Further, there remains a need for a transdermal formulation of an opioid analgesic, preferably, buprenorphine, that provides effective analgesic levels of buprenorphine for periods of time beyond that contemplated or practical in the prior art, while eliminating or minimizing dependence, tolerance, and side effects, thus providing a safe and effective method of pain management.”

Accordingly, the specification states at [0011] that the object of the invention is to:

“… provide a method and pharmaceutical formulation (medicament) which allows for reduced plasma concentrations of buprenorphine over a prolonged time period than possible according to prior art methods, while still providing effective pain management.”

After describing further objects of the invention, and setting out consistory clauses corresponding to the claims, the specification provides some general information about analgesics at [0023] to [0025] and information about the chemical structure of buprenorphine, its analgesic properties, its suitability for transdermal delivery, its safety and pharmacokinetic properties at [0024] to [0040]. It then goes on to describe the preferred plasma concentrations of buprenorphine that are to be achieved in humans at [0041] to [0042], and preferred release rates at [0043] to [0054].

At [0055] to [0056] the specification discusses the flux of transdermal delivery devices, and notes that it is well known that, in order to achieve a desired flux rate for a desired dosing period, it is necessary to include an overage of active ingredient.

At [0057] the specification states:

“Surprisingly, it has been found that it is possible to treat pain according to the present invention by providing a transdermal delivery device containing a sufficient amount of buprenorphine to provide a desired relative release rate for up to 3 days, and after single administration (application) of the transdermal dosage form, leaving the dosage form on the skin for approximately a 5 to 8 day time period, thereby resulting in the flux being maintained over the prolonged period and effective blood plasma levels and pain management being maintained over the prolonged period. Preferably, the desired flux is maintained at least about 5, preferably at least about 8 days after application of the transdermal delivery device. If the transdermal delivery device is removed 3 days after its administration, no analgesia is present a short time after removal (the remaining time period during which analgesia is provided being dependent upon, for example, the amount of drug in a skin depot at the site of the application of the transdermal delivery device). Surprisingly however, if the same transdermal delivery device is maintained in contact with the skin for an about 5 to about 8 day period, analgesia is maintained over the prolonged period of contact, but the patient continues to experience analgesia. In other words, inclusion of the aforementioned overage of buprenorphine provides analgesia for at least about twice the expected 3 day dosing interval.”

At [0060]-[0071] the specification discusses the other constituents of the transdermal delivery devices used in the invention. In particular, it states at [0065] that the devices may include a “permeation enhancing agent” and at [0068] that they may include a “softening agent”. It gives a number of examples of suitable softening agents, including dodecanol and levulinic acid.

At [0073] the specification states:

“In a preferred embodiment, the transdermal delivery device is prepared in accordance with Example 1 appended hereto. In this example, the transdermal delivery device was prepared in accordance with the disclosure of [Hille]. In this device, the buprenorphine transdermal delivery device contains resorption-promoting auxiliary substances. The resorption promoting auxiliary substance forms an undercooled mass. The delivery device contains 10% buprenorphine base, 10-15% acid (such as levulinic acid), about 10% softener (such as oleyloleate); 55-70% polyacrylate; and 0-10% polyvinylpyrollidone (PVP).”

It is common ground that the last sentence of this paragraph is an attempt to encapsulate Hille’s examples, but there is a dispute as to whether that includes Example 1a as opposed to Example 1b. I shall return to this point below.

The specification goes on at [0074] to explain that, where embodiments of the invention are “prepared in accordance with” Hille, it is contemplated that, in order to achieve particular nominal delivery rates, amounts of buprenorphine of “about 5 mg”, “about 10 mg”, “about 20 mg”, “about 30 mg” and “about 40 mg” are used.

The specification then states at [0075]:

“In accordance with a method of the invention, the above-described transdermal delivery device has been designed to be adhered to the patient for only three days and is expected to release analgetically effective doses of buprenorphine for only about 3 days. Instead, in accordance with the present invention, the transdermal delivery device is maintained in contact with the skin of the patient for a much longer time period, e.g., from about 5 to about 8 days, without any change in the formulation of the transdermal device itself. It has surprisingly been found that analgesia is maintained for this extended period of time (the time beyond the useful life designed for the transdermal formulation).”

After acknowledging five items of prior art at [0076] to [0080], the specification states at [0082] that the invention, by maintaining a lower blood level of drug over a 7 day dosing period while maintaining effective pain management, had a lower incidence of side effects.

At [0084] the specification attempts to explain the effects described at [0057] and [0075]:

“It is hypothesized that there is a skin reservoir which occurs through the use of transdermal buprenorphine delivery devices, including those demonstrated in the Examples. In cases where the transdermal delivery device is reapplied prior to complete depletion of the skin reservoir, it is preferred that the transdermal delivery device be reapplied in the same location on the patient's skin in order to replenish the skin reservoir (or ‘depot’). The plasma concentration curves obtained during, e.g., 7-day dosing may be explained by a small amount of buprenorphine still being released from the transdermal delivery device on day 4 through day 7 after initial application of the device onto the patient's skin (which replenishes the skin depot), along with a continued release and continued delivery of buprenorphine from the skin depot. The balance between continued release from the device and continued delivery from the skin depot would (among other things such as distribution) determine the plasma concentration seen in any particular individual. It would also explain why the buprenorphine plasma concentrations don't precipitiously drop on day 4 through, e.g., day 7 as they do when the device is removed from contact with the patient's skin. The skin depot hypothesis set forth herein would also explain why the elimination curve following patch removal is prolonged compared to what one skilled in the art would expect for the drug buprenorphine had the drug been administeredintravenously. The skin depot hypothesis is preferred for explanatory purposes only and is not meant to limit the claims in any manner whatsoever.”

In short, the Patent hypothesises that a depot of buprenorphine is formed in the skin which is replenished with small amounts of buprenorphine from the device over time and that buprenorphine is released from the skin depot into the systemic circulation. As discussed below, there is some disagreement as to the extent to which the existence of such skin depots would have formed part of the skilled person’s common general knowledge. It is common ground, however, that, either way, the skilled person would have regarded this hypothesis as a reasonable one, although the skilled person would also appreciate that occlusion of the skin by the device was likely to play a role.

At [0083] and [0085]-[0095] the specification provides definitions of the following terms: “effective analgesia”; “effective pain management”, “mean”; “breakthrough pain”; “rescue”; “first order”; “zero order”; “mean relative release rate”; “sustained release”; “steady state”; “minimum effective analgesic concentration”; “buprenorphine”; and “overage”.

Example 1 is described at [0098]-[0101]. The description of this Example begins as follows:

“[0098] A seven day pharmacokinetic/pharmacodynamic study was conducted on 24 healthy human patients. The subjects were comprised of approximately an equal number of male and female subjects. In this study, the buprenorphinewas administered via a transdermal patch which is described in [Hille].

[0099] The transdermal patch is prepared in accordance with the disclosure of [Hille] for Example 1 therein as follows:”.

The specification then sets out verbatim the text of Example 1a of Hille quoted in paragraph 43 above (except that the English translation of the original German is slightly different in ways that are not suggested to affect the sense of the passage).

The specification goes on at [0100]:

“The formulation utilized for Example 1 is substantially the same as that described in Example 3 of [Hille] which is prepared in accordance with Example 1 and is stated therein to include 10% buprenorphine, 10% levulinic acid, 10% polyvinylpyrollidone, 10% oleyloleate, and 60% polyacrylate. ”

The methodology of the study is described at [0101]-[0106]. The results of the study are set out in Tables 1, 2, 3 and 4 and discussed at [0107]-[0112]. Most of the details do not matter for present purposes, but it can be seen from [0112] that the majority of the buprenorphine remains in the patch. Of about 10 mg in the patch at the outset, on average 8.59 mg remains after 7 days.

The specification then describes comparative examples A-C (a single 3 day patch removed after 3 days, a single intravenous dose and three sequential 3 day patches) at [0113]-[0128] and comparative examples D-F (a 10 mg patch, two 5 mg patches and a 20 mg patch) at [0129]-[0138]. Again, the details do not matter for present purposes. The results are discussed at [0139]-[0146]. It is sufficient for present purposes to note the following passages:

“[0141] The data presented in Table 18 shows that, surprisingly, plasma levels effective to provide analgesia werepresent in Example 1 (patch remained on skin for 7 days) even 7 days after application of the patch; whereas in Comparative Example A (patch removed after 3 days), blood levels fell dramatically once the patch was removed, such that plasma levels which would be indicative of ineffective treatment for the dosage of buprenorphine occurred not long after patch removal. …

[0143] The results depicted in Table 19 confirm that the method according to the present invention provides effective plasma levels over the 7-day period; whereas if the patch (or patches) containing the same dose is removed after 3 days, the buprenorphine plasma levels fall precipitously over the next 24 hour interval to levels which would be indicative of ineffective treatment for the dosage of buprenorphine. This result is surprising in view of the fact that the patches are designed to provide effective analgetic levels of buprenorphine only for a three day period - these patches are not designed to provide effective plasma levels of buprenorphine over a substantially longer period of time. …”

It is common ground that it is only necessary to consider claim 1 of the Patent, which is as follows:

“A buprenorphine transdermal delivery device comprising a polymer matrix layer containing buprenorphine or a pharmaceutically acceptable salt thereof, for use in treating pain in humans for a dosing interval of at least 7 days, wherein the transdermal delivery device comprises 10 %-wt buprenorphine base, 10 to 15 %-wt levulinic acid, about 10 %-wt oleyloleate, 55 to 70 %-wt polyacrylate, and 0 to 10 %-wt polyvinylpyrrolidone.”

A patent specification is addressed to those likely to have a practical interest in the subject matter of the invention, and such persons are those with practical knowledge and experience of the kind of work in which the invention is intended to be used. The addressee comes to a reading of the specification with the common general knowledge of persons skilled in the relevant art, and he (or she) reads it knowing that its purpose is to describe and demarcate an invention. He is unimaginative and has no inventive capacity. In some cases, the patent may be addressed to a team of persons having different skills.

In the present case it is common ground that the Patent is addressed to a team of people interested in the transdermal administration of buprenorphine for the treatment of pain. It is also common ground that the skilled team would include a formulator with expertise in the formulation of transdermal delivery devices. Although there were minor disagreements between the parties as to the extent to which the team would be led by the formulator and as to the extent to which the team would be concerned with the commercialisation of such devices, no party suggested that these disagreements mattered for the purposes of resolving the issues in this case.

Although it is common ground that in practice there would be a team of people with different backgrounds, both sides were able to address the issues in this case by calling a single expert whose expertise was in the formulation of transdermal delivery devices. Accordingly, I shall refer for convenience to the skilled person rather than the skilled team

I reviewed the law as to common general knowledge in KCI Licensing Inc v Smith & Nephew plc [2010] EWHC 1487 (Pat), [2010] FSR 31 at [105]-[115]. That statement of the law was approved by the Court of Appeal in that case [2010] EWCA Civ 1260, [2011] FSR 8 at [6].

There is little, if any, dispute that everything I have set out in the technical background section of this judgment formed part of the skilled person’s common general knowledge, and in any event that is my finding. There were two topics in relation to which there was some disagreement, or at least uncertainty, as to the skilled person’s common general knowledge.

The Defendants contend that it was common general knowledge that (i) some drugs could form a depot or reservoir in the skin, or more specifically the stratum corneum, in appropriate circumstances and (ii) the formation of such a depot or reservoir depended partly on the physiochemical properties of the drug and partly on other components of the formulation. Napp disputes this, and in particular point (ii).

The Defendants’ contention was supported by the evidence of Prof Williams, who exhibited two extracts from Prof Barry’s textbook Dermatological Formulations: Percutaneous Absorption (1983) at pages 116-118 and 127-129 as confirming his opinion. Dr Miller’s evidence was that corticosteroids were well known to form such depots, particularly when applied topically, but not other drugs. Significantly, he exhibited exactly the same extracts from Prof Barry’s book as Prof Williams.

In my judgment the evidence establishes that the Defendants are correct. Thus Prof Barry states at 117:

“The mechanism of reservoir formation most probably arises simply from the physicochemical nature of drug solubility and diffusion within the stratum corneum. It is not necessary to invoke special pharmacological reactions to explain the effect but simply to realise that the depot is prominent with a certain class of material i.e. those substances with small diffusivities and low solubilities in the stratum corneum. In general those factors that promote percutaneous absorption also potentiate reservoir formation. …. In particular, certain solvents (known as accelerants, promoters, or penetration enhancers) induce superior reservoirs. …”

Moreover Prof Barry states explicitly at 118 that “Substances other than steroids may form reservoirs” and gives some examples.

Further support for this conclusion is provided by an extract from a book by Hans Schafer and Thomas Redelmeir, Skin Barrier: Principles of Percutaneous Absorption (1996), at 162-163 which was put to Dr Miller in cross-examination by counsel for Dr Reddy’s and which contains similar statements.

It is common ground that it was well known that the drying process used to remove solvents from the drug adhesive blend when making a TDS had the potential to drive off a proportion of some of the other components of the formulation, and in particular semi-volatile excipients. What is less clear is whether the skilled reader of the Patent would either know from his common general knowledge, or ascertain as a matter of routine, that levulinic acid was such a semi-volatile component.

Dr Miller gave evidence in his main report that the skilled person would be unlikely to be familiar with levulinic acid in the context of transdermal patch formulations. He nevertheless said that the skilled person “would know or would easily find out” its physicochemical properties, and specifically that it was a small organic acid also known as 4-oxopentanoic acid with a molecular weight of 116 g/mol and thus was likely to evaporate to some degree during the drying process. He did not state its boiling point, however, nor did he specifically state that the skilled person would look this up. I have ascertained that the boiling point is in fact 245-246 ^oC. In his reply report Prof Williams appeared to agree that the skilled person would know or find out that levulinic acid was liable to evaporate during the drying process.

I have been troubled by this aspect of the case, because in the light of Dr Miller’s evidence I am sceptical that this property of levulinic acid would have been common general knowledge and I am not clear why it is said that the skilled person would ascertain it as a matter of routine. Counsel for Napp submitted in his closing submissions and counsel for Dr Reddy’s expressly conceded in his closing submissions, however, that this would have been known to the skilled person by one route or the other, and counsel for Sandoz did not dissent from this. Accordingly, I shall proceed on that basis.

The general principles applicable to the construction of patent claims were reviewed by Lord Hoffmann in Kirin Amgen Inc v Hoechst Marion Roussel Ltd [2004] UKHL 46, [2005] RPC 9 at [18]-[52] and summarised by Jacob LJ in Virgin Atlantic Airways Ltd v Premium Aircraft Interiors UK Ltd [2009] EWCA Civ 1062, [2010] RPC 8 at [5]. In short, the question is always what the person skilled in art would have understood the patentee to be using the language of the claim to mean.

Numerical ranges. The construction of numerical ranges in a claim was recently considered by the Court of Appeal in Smith & Nephew plc v ConvaTec Technologies Inc [2015] EWCA Civ 607, [2015] RPC 32. In that case the patent concerned a gel-forming wound dressing including an antimicrobial agent. The question of the meaning of numerals arose because the claims of the patent called for the antimicrobial agent to be “between 1% and 25% of the total volume of treatment”. The issue was what the range 1% to 25% covered, and in particular whether it covered the alleged infringer’s formulation, which contained 0.77% of antimicrobial agent.

Kitchin LJ summarised the parties’ arguments at [7]:

“On the numerical boundaries issue, ConvaTec contended that the expression ‘between 1% and 25%’ did not define precise values but would be seen in terms of whole numbers and so, applying the basic rounding convention, the claim encompassed all concentrations greater than or equal to 0.5% and less than 25.5%. By contrast, Smith & Nephew's primary contention was that the expression meant what it said and that it defined a range of concentrations, with a lower limit of precisely 1% and an upper limit of precisely 25%. But Smith & Nephew also had a fall back position, namely that the 1% and 25% limits would be read in terms of significant figures, and that 0.95% and above would be understood to round up to 1%, and that 25.49% and below would be understood to round down to 25%, and so the claim encompassed all concentrations greater than or equal to 0.95% and less than 25.5%.”

Having referred to Kirin Amgen and quoted Jacob LJ’s summary in Virgin Atlantic, Kitchin LJ said:

“16.

I would add the following two principles which are also drawn from Lord Hoffmann's speech and which have a particular bearing on this appeal. First, the reader comes to the specification with the benefit of the common general knowledge and on the assumption that its purpose is to describe and demarcate an invention. Second, the patentee is likely to have chosen the words appearing in the claim with the benefit of skilled advice and, in so far as he has cast his claim in specific rather than general terms, is likely to have done so deliberately.

17.

It seems to me that all of these principles are just as applicable to a claim containing a numerical range as they are to a claim containing descriptive words or phrases. In both cases the critical question is what the skilled person would understand the author to have been using the words or numerals appearing in the claim to mean.”

Kitchin LJ then considered previous authorities, both from the UK and the EPO, dealing with numerical ranges. During the course of this review, he said:

“21.

… I would endorse the reasoning of Mr Peter Prescott QC, sitting as a deputy judge of the High Court, in Auchincloss v Agricultural & Veterinary Supplies Ltd [1997] RPC 649. He said this at page 664:

‘… It is a descriptive word or phrase to which the concept of an immaterial variant applies. In the cited case [Catnic Components v Hill & Smith Ltd [1982] RPC 183 (HL)] the “descriptive” phrase was “extending vertically”. If, instead, the claim had said “extending at an angle between 87.0º and 93.0º”, this would not have been a descriptive phrase, but a specification of a precise range. It would be a unilateral statement in words of the patentee's own choosing informing the reader what he claimed to be an essential feature of his invention. If a product fell outside that range it would not be a question of a “variant” at all, minor or otherwise. It would be a failure to adopt an essential feature. Even if an angle falling somewhere outside that range would work just as well, and this were obvious to all concerned, to hold non-infringement would not be to deny “a fair protection for the patentee” in terms of the Protocol to the Convention; for the range he chose to specify was one which he himself must have considered to be fair at the relevant date. On the contrary, to hold otherwise would be to deny “a reasonable degree of certainty for third parties”…’

23.

The deputy judge developed his reasoning at pages 689-690 in these terms:

‘… Where the patentee has expressed himself in terms of a descriptive word or phrase there may be room for supposing that he was using language figuratively, and did not intend to restrict himself to the purely literal meaning. But where the patentee has defined an integer of his claim in terms of a range with specified numerical limits at each end, his purpose must be taken to have been to claim thus far and no further. His reason for doing so may not be apparent, but it may exist all the same, for instance it may lie “buried in the prior art”. Further, in this case I believe that there are evident reasons of convenience and certainty which would have led him to claim in this way, as I have observed.’”

Following his review of the authorities, Kitchin LJ summarised the position at [37] as follows:

“As I have said, the approach to be adopted to the interpretation of claims containing a numerical range is no different from that to be adopted in relation to any other claim. But certain points of particular relevance to claims of this kind do emerge from the authorities to which I have referred and which are worth emphasising. First, the scope of any such claim must be exactly the same whether one is considering infringement or validity. Secondly, there can be no justification for using rounding or any other kind of approximation to change the disclosure of the prior art or to modify the alleged infringement. Thirdly, the meaning and scope of a numerical range in a patent claim must be ascertained in light of the common general knowledge and in the context of the specification as a whole. Fourthly, it may be the case that, in light of the common general knowledge and the teaching of the specification, the skilled person would understand that the patentee has chosen to express the numerals in the claim to a particular but limited degree of precision and so intends the claim to include all values which fall within the claimed range when stated with the same degree of precision. Fifthly, whether that is so or not will depend upon all the circumstances including the number of decimal places or significant figures to which the numerals in the claim appear to have been expressed.”

Following that review of the law Kitchin LJ went on to apply his analysis to the claim in question. He first rejected an exact value understanding of the figures. The skilled addressee would not understand the patentee to have intended the limits to be read in that way. The question was what degree of precision was required. In that case, the question was whether the skilled addressee would apply a whole number approach or a significant figure approach to the construction of 1% to 25%. The difference between the two was that the whole number approach would cover a range ≥0.5% to <25.5, whereas the significant figure approach would cover a range ≥0.95% to ≤25.5%. He held that the whole number approach was correct, for the reasons he expressed at [60] as follows:

“In my judgment there can be no logical basis for preferring the significant numbers approach over the whole number (or zero decimal places) approach in construing the claim in issue. The purpose of expressing numbers to a particular degree of precision may be to convey to the reader the degree of accuracy with which he needs to make a particular measurement or carry out a calculation. In the context of the claimed method, it is to convey to the reader the range of permissible binding agent concentrations and the accuracy with which those concentrations need to be determined. There is no reason to suppose that this can vary depending upon whether the bottom of the range is 1%, 2% or 5%, or whether 10% is at the top or bottom of the range. It seems to me that Professor Kennedy therefore put it entirely correctly in saying as he did in his first expert report that it is not the number of significant figures that is important in this context, and instead it is the precision with which a number is written. I consider that Professor Kennedy was also right to say that the skilled person would understand the 1% and 25% limits to have been expressed to the nearest whole number.”

“About” and similar expressions. Part F Chapter IV paragraph 4.7 of the Guidelines for Examination in the European Patent Office (November 2015 edition) states:

“Terms like ‘about’ and ‘approximately’

Particular attention is required whenever the word ‘about’ or similar terms such as ‘approximately’ or ‘substantially’ are used. Such a word may be applied, for example, to a particular value (e.g. ‘about 200°C’), to a range (e.g. ‘about x to about y’) or to a structural unit of an apparatus (e.g. ‘a tray plate with a substantially circular circumference’). In each case, the examiner should use his judgment as to whether the meaning is sufficiently clear in the context of the application read as a whole, and having regard to the meaning a particular technical term qualified by such a word usually has in the field concerned. If such words mean that a certain effect or result can be obtained within a certain tolerance and the skilled person knows how to obtain the tolerance, then the use of such words may be acceptable. If, however, the use of such a word suggests that deviations are included which are larger than those accepted tolerances, then the wording becomes vague and undefined. …”

Before turning to the construction of the claim, it is first necessary to resolve two other disputes between the parties. The first concerns the interpretation of Example 1 in the Patent. Napp contends that the patch used in Example 1 of the Patent is made in accordance with the method and proportions of Example 1a of Hille. The Defendants contend that the patch used in Example 1 of the Patent is made in accordance with the method of Example 1a which is quoted in the Patent at [0099] and the proportions of Example 3 of Hille which are set out in the Patent at [0100].

This dispute largely turns on the correct interpretation of [0100] (quoted in paragraph 73 above). Napp contends that the skilled person would read this as saying that the “formulation utilized for Example 1 [of the Patent, which is Example 1a of Hille] is substantially the same as described in Example 3 of [Hille] which is prepared in accordance with Example 1[a] [of Hille]”. The Defendants contend that the skilled person would read this as saying that the “formulation utilized for Example 1 [of the Patent] is substantially the same as described in Example 3 of [Hille] which is prepared in accordance with Example 1[a] [of Hille]”.

Counsel for Napp advanced five arguments in support of Napp’s interpretation:

Counsel for the Defendants answered these arguments as follows:

Counsel for the Defendants advanced three main arguments in support of the Defendants’ interpretation:

Counsel for Napp had no convincing answers to these arguments. In my judgment, the Defendants’ interpretation is the correct one for the reasons given by counsel for the Defendants.

There is also a dispute between the parties as to how the skilled person would interpret the passage in the Patent at [0084] which sets out the skin depot hypothesis (quoted in paragraph 68 above). Napp contends that the skilled reader would understand that the formation of a depot was purely due to the properties of buprenorphine. The Defendants dispute this and contend that the skilled reader would appreciate that the formation of a depot could be affected by some of the components of the TDS, and in particular levulinic acid and oleyl oleate. In my judgment the Defendants are right about this. The skilled person would know from his common general knowledge that penetration enhancers and the like could influence the formation of a depot, and there is nothing in [0084] to suggest that it is solely attributable to the properties of buprenorphine.

The first main issue of construction of the claim is whether the composition figures refer to the ingredients in the recipe for the matrix (i.e. the inputs into the manufacturing process) or the composition of the finished product (i.e. the outputs from the manufacturing process). Napp’s primary case is that that they refer to the inputs. The Defendants contend that they refer to the outputs.

The Defendants make the forensic point that it was only at a relatively late stage of the proceedings, namely on 15 April 2016, that Napp adopted its primary construction of the claim. If that construction is the correct one, however, the fact that Napp was slow to espouse it is immaterial.

It is convenient to note two points before proceeding further. First, Napp accepts that the percentage weights are to be calculated excluding the solvent(s) used to make the drug adhesive blend. Secondly, it will be appreciated that, that being so, there will only be a difference between the two constructions in circumstances where there is some loss of one or more of the specified components during the course of the manufacturing process, and in particular during the drying stage. For the reasons explained above, the skilled person would appreciate, if he thought about it, that some of the levulinic acid could be lost during drying. The skilled person would note, however, that there is no discussion of this possibility in either Hille or the Patent.

Counsel for the Defendants submitted that the wording of the claim is perfectly clear: it refers to a “transdermal delivery device comprising a … matrix layer … where the device comprises …”. This language is clearly and unambiguously referring to the finished product, and hence the figures refer to the outputs of the manufacturing process. As the skilled reader would appreciate, the figures must refer to the composition of the product after removal of the solvent. Moreover, the skilled person’s appreciation that semi-volatile components, and in particular levulinic acid, could be lost during drying reinforces this reading, since it is the composition of the finished product that matters in terms of its performance. This is particularly so given the skin depot hypothesis advanced in the Patent. Furthermore, as Dr Miller accepted, it would be feasible for the skilled person to carry out assays to determine the composition of the finished product. Indeed, that is what is done when validating a process for commercial production, at least with respect to components which are regarded as important by the regulator. Finally, if the figures referred to the inputs, the skilled reader would face the difficulty that the Patent does not tell him how to make a patch in accordance with the invention, because Hille does not specify the drying conditions and the Patent does not supply the missing information.

Counsel for Napp advanced seven arguments in support of Napp’s primary construction:

Counsel for the Defendants answered these arguments as follows:

In my judgment, the Defendants’ interpretation is the correct one for the reasons given by counsel for the Defendants.

The second main issue of construction of the claim concerns the interpretation of the numerical limits of the ranges in percentage weight of buprenorphine, levulinic acid and oleyl oleate. (There is no issue as to PVP or polyacrylate). The Defendants contend that they would be understood to be expressed in terms of whole numbers, and applying the conventional rounding approach extend to ≥9.5 to <10.5 %-wt buprenorphine, ≥9.5 to <15.5 %-wt levulinic acid and ≥9.5 to <10.5 %-wt oleyl oleate. Napp contends that the numerals in the buprenorphine and levulinic acid ranges are expressed to the nearest 5 %, and hence extend to ≥7.5 to <12.5 %-wt buprenorphine and ≥7.5 to <17.5 %-wt levulinic acid, while “about 10 %-wt oleyloleate” extends to 15 %-wt.

Counsel for Napp relied on a number of general points as showing that the claim should be interpreted broadly.

First, the invention lies in the use that is made of the TDS, not in the formulation of the patch. Against this, counsel for the Defendants pointed out that the specification states several times that it consists of a method and a pharmaceutical formulation. Nevertheless, the skilled reader would appreciate that the formulation of the TDS was itself old, in that it had been disclosed in Hille. It does not follow, however, that the skilled person would consider that the formulation was unimportant. On the contrary, as discussed above, the skilled reader would note that the patentee had selected the best performing formulation from Hille.

Secondly, as discussed above, the skilled person would appreciate that there is inherent manufacturing variability in TDSs and would usually anticipate that reasonably broad tolerances would be acceptable. I accept that these factors mean that the skilled person would not expect that great precision was necessary. But the question remains as to how the skilled person would interpret the limits in the claim.

Thirdly, the claims are expressed in terms of percentage weights, not absolute weights. In my view this is immaterial. A difference in percentage weight will mean a difference in concentration, which in turn will affect the flux. The fact that the claim covers a range of absolute weights does not alter this.

Fourthly, the skilled person would appreciate that the TDS disclosed in Hille and in the Patent was in the early stages of development. I accept this, and I also accept Dr Miller’s evidence that in practice the exact composition of a TDS is likely to change during scale up, but it does not follow the skilled reader would think that compliance with the limits specified was unimportant.

Napp relies upon two points in support of its construction of this integer, in addition to the general points above. First, Napp contends that the concentration of buprenorphine is not critical, because only a small amount of the buprenorphine is administered to the patient. This is a non sequitur, since what matters is the flux. As I have said, however, I accept that the skilled person would not think that great precision was required.

Secondly, Napp relies upon the last sentence of [0073]. As discussed above, it is common ground that this is an attempt to encapsulate Hille’s Examples. Napp contends that the skilled person would note that the Examples in Hille (including Example 1a), and [0073], specify the components in increments of 5 %. Accordingly, Napp argues, the skilled person would conclude that the limit in the claim was expressed to the nearest 5 %-wt.

I do not accept this argument. There is nothing in the specification of the Patent which sheds any light on the precision of this numerical limit. On its face, it appears to be expressed to the nearest whole number. As the Defendants accept, applying the usual scientific convention, it would follow that “10 %-wt” extends to ≥9.5 to <10.5 %-wt. The skilled person would appreciate that this provided some margin for error due to variability in the manufacturing process and/or analysis of the finished product. Even assuming that the skilled reader noticed that the Examples in Hille can be described as being specified in increments of 5 %, I consider that [0073] would lead the skilled person away from interpreting the claim in that way, since it refers to ranges of “10-15% acid”, “55-70% polyacrylate” and “0-10%” PVP.

To the extent that this issue is one for expert evidence, I found Prof Williams’ evidence more persuasive than that of Dr Miller. Dr Miller did not really support the 5% interpretation in cross-examination, but rather treated the limits in the claim as no more than guidelines.

Even assuming that it makes no difference to the way in which the invention works, I consider that the reasoning of Mr Prescott QC in Auchincloss which Kitchin LJ approved in Smith & Nephew is applicable here. To interpret the claim in the way contended for by Napp would be to deny a reasonable degree of certainty for third parties. To interpret the claim in the manner contended for by the Defendants would not be to deny the patentee fair protection, but to give the patentee the protection it chose when drafting the claim.

The arguments in relation to this integer are essentially the same as in relation to the preceding integer, and my conclusion is the same: the numbers are expressed to the nearest whole number.

As noted above, Napp contends that “about 10 %-wt” should be interpreted as extending to 15 %-wt (which, applying Napp’s case on the precision of the numbers, would appear to mean that this integer covers ≥7.5 and <17.5 %-wt). The basis for this contention is that [0073] would be understood by the skilled person as encapsulating Hille’s Examples, including Example 1a, which has 15 %-wt oleyl oleate. The Defendants contend that “about” would be disregarded by the skilled person as being meaningless.

I do not accept either of these contentions. In my judgment, given that both [0073] and the claim specify 10-15 %-wt for the acid, the skilled person would think that, if the patentee had intended to cover a range of 10-15 %-wt for the oleyl oleate, the claim would have said so. So far as Example 1a of Hille is concerned, the skilled reader would consider it explicable that this was not covered by [0073] or by the claim, because no results are given for it. Accordingly, I do not accept Napp’s interpretation. I therefore disagree with the decision of the Opposition Division on this point.

As for the Defendants’ interpretation, this amounts to treating the word “about” as not being present, which is contrary to principle. That said, it is very difficult indeed to ascertain from the specification what the patentee intended to signify by its inclusion of this word. In my judgment, the inclusion of this word is contrary to paragraph 4.7 of the Guidelines, and the claim lacks clarity. In those circumstances it is tempting to conclude that it is also insufficient. But I have concluded that the better view is that the claim is not invalid, but rather that the word “about” should be taken to connote a small degree of permitted imprecision over and above that implied by the usual rounding convention. I do not consider that it would provide reasonable certainty for third parties if this integer were interpreted as extending beyond ≥9.0 to <11.0 %-wt, however.

Sandoz’s PPD describes in some detail the process used by Hexal to manufacture Reletrans and of the testing undertaken by Hexal to validate that process in order to meet regulatory requirements. It is not necessary for me to set these matters out in any detail for the purposes of this judgment, and they are in any event confidential. It suffices to say that the PPD sets out (among other things) the quantities of the ingredients used to produce 150 m² of drug adhesive laminate (although larger batches may be manufactured); the accuracy to which those quantities are weighed; the time for which, and the manner in which, the drug adhesive blend is stirred to produce a homogeneous coating solution; the assays carried out to demonstrate homogeneity of the coating solution for regulatory purposes; the manner in which the drug adhesive laminate is produced (although the PPD does not condescend to detail with respect to the drying process); assays carried out to ascertain the buprenorphine and levulinic acid content of the drug adhesive laminate both during process validation for regulatory purposes and during commercial production; the production of the finished dosage form (i.e. the patches); and assays which have been carried out to ascertain the buprenorphine and levulinic acid content of sample patches.

As counsel for Sandoz pointed out, Napp did not challenge Mr Kohr’s evidence. As he submitted, it follows that it is not open to Napp to cast doubt on the accuracy of Sandoz’s PPD, as counsel for Napp tried to do at points.

Annex 1 to the Sandoz PPD states that the precision of the analytical technique (high-performance liquid chromatography or HPLC) employed by Hexal to assay the buprenorphine and levulinic acid content of the drug adhesive laminate is “RSD (n 6) <2.0%”. My understanding is that this means that the relative standard deviation upon six repetitions is less than or equal to 2%, but whether this is correct was not explored in evidence, nor were the consequences of that degree of precision.

I summarise the input composition of Sandoz’s product and the available data with respect to its output composition, so far as relevant to the issues on infringement, in Confidential Annex A.

The issues on infringement are whether the Sandoz product has (i) 10 %-wt buprenorphine, (ii) 10-15 %-wt levulinic acid and (iii) about 10 %-wt oleyl oleate. For the reasons given above, I conclude that these questions fall to be answered on the bases that (a) the percentage weights are of the final product and (b) the numerical limits are to be interpreted as being expressed to the nearest whole number, except that “about 10 %-wt” is to be interpreted a little more broadly. Napp accepts that, on those assumptions, the product will not fall within the claim regardless of statistical considerations.

In case this case goes further, it is necessary for me to consider what the position would be if (a) the percentage weights are of the final product, but (b) the numerical limits are to be interpreted as contended for by Napp. On those bases, I understand it to be common ground that the sole issue would be the probability of Sandoz’s product having a levulinic acid level of 7.5 %-wt or greater, because the product will have a buprenorphine level ≥7.5 and <12.5 %-wt and an oleyl oleate level of ≥7.5 and <17.5 %-wt.

So far, as explained in more detail in Confidential Annex A, the levulinic acid level has been tested in 105 samples taken from drug adhesive laminate during process validation and in 20 samples taken from drug adhesive laminate during production of the commercial product. There is no dispute that, in each set of samples tested, the mean %-wt was less than 7.5 %-wt. I do not understand Napp seriously to contend that any of the patches made from any of the commercial laminates produced to date would fall within the claim. Napp nevertheless contends that the variability in the product is such that there is a threat by Sandoz to sell an appreciable quantity of infringing patches in the period until 24 February 2018 given that, as discussed above, it is not unrealistic to suppose that Sandoz could sell 2-3 million patches in that period.

It is important to appreciate that, as noted above, Napp’s claim against Sandoz is primarily a quia timet one. Napp could have waited until the Patent had expired, and then brought a claim against Sandoz for past infringements. If Napp had chosen to do that, then, at least in theory, the court would have had available to it data in respect of all the patches sold by Sandoz. One would thus know the composition of all of those patches, subject to any limitations in the available data. Napp chose not to wait, however. That is because Napp is not content with obtaining a financial remedy for any infringements by Sandoz. Rather, it wants to obtain an injunction to stop Sandoz coming onto the market before the Patent expires.

Napp’s claim gives rise to issues both of law and statistics.

There are five aspects to the legal analysis. The first concerns the application of the de minimis principle in the context of patent infringement. The second concerns the test for infringement where the claim is made quia timet. The third concerns the burden of proof to be applied to allegations of future infringing acts. The fourth concerns the standard of proof to be applied to allegations of future infringing acts. The fifth concerns the remedies for infringement, and in particular the remedy of an injunction.

The de minimis principle. There is a general principle of English law (and indeed of most developed legal systems) which is encapsulated in the Latin maxim de minimis non curat lex (the law does not concern itself with trifling matters). More specifically, there is a well-established principle of statutory construction which is expressed in Halsbury’s Laws of England (5^th ed), vol. 96 at §1143 as follows (footnotes omitted):

“Unless the contrary intention appears, an enactment by implication imports the principle of legal policy expressed in the maxim de minimis non curat …; so if an enactment is expressed to apply to matters of a certain description it will not apply where the description is satisfied only to a very small extent.”

As Lord Dyson remarked in R (Alvi) v Secretary of State for the Home Department [2012] UKSC 33, [2012] 1 WLR 2208, an immigration case, at [88], “[i]t goes without saying that the principle de minimis non curat lex (the law is not concerned with very small things) applies in the present context as in most others.” As Lord Phillips of Matravers stated in Sienkiewicz v Greif (UK) Ltd [2011] UKSC 10, [2011] 2 AC 229 at [108]:

“I doubt whether it is ever possible to define, in quantitative terms, what for the purposes of the application of any principle of law is de minimis. This must be a question for the judge on the facts of the particular case.”

The application of the de minimis principle to patent law has proved surprisingly controversial. Counsel for Dr Reddy’s helpfully referred me to a number of authorities relevant to this question, which I will consider in chronological order.

In Hoechst Celanese Corp v BP Chemicals Ltd [1998] FSR 586 the patent in suit was for a process of making acetic acid by carbonylation of methanol. The claim required that “there are maintained in the reaction medium during the course of the reaction” specified ranges of reactants. Jacob J construed this as follows:

“25.

Hoechst say that this means that the process must be operated to produce HOAc for a period of time within the required parameters. Fluctuation from outside into the forbidden area is not excluded. Maintained means essentially running a continuous (rather than batch) process. If there is any significant production of HOAc within the specified parameters, there is infringement. So if on one day the parameters are satisfied, there is infringement on that day, even if on other days (even many other days) the parameters are not met. If there were a momentary ‘spike’ that might not fall within the claim but no one is concerned with that. Even a day's production amounts to 800 tonnes.

26.

BP say the words mean that the parameters must be kept within the limits of the claim for a significant period of time in the context of the reaction as a whole. A short term fluctuation into the range will not alter the fact that it is being ‘maintained’ outside the range. …

27.

The patent gives little explicit guidance on the point. …

28.

I think there are several difficulties with the BP approach. First, over what period is one to take an ‘average’? Suppose the reactor is run for a year with one or two days of excursion. On the BP approach there is no infringement. But suppose those two days are at the beginning of the year. If one looks just at the first two days, all the operation is within the specified parameters so there is infringement. How (and when) can it cease to be an infringement because for a later period there is operation outside the parameters? Second, the approach is close to introducing a subjective element: at what level did the alleged infringer intend to operate? That cannot be a relevant matter. Third I cannot think of any reason why the patentee would want to exclude any significant commercial production within his specified parameters. Doubtless he would not be interested in truly transient ‘spikes’ but anything as much as a day's production he would surely wish to catch.

29.

Accordingly I think Hoechst are right on this point. Since the measurements were taken daily, I propose to approach the question of infringement on the basis that, if, on any particular day, BP were within the parameters, they infringed on that day.”

The first point to note about this passage is that Jacob J did not, at least explicitly, consider the application of the de minimis principle, presumably because BP did not rely on that principle. The second point is that this is understandable given that the available data consisted of daily measurements, and that a single day’s production amounted to 800 tonnes, a substantial quantity on any view. The third point is that Jacob J was concerned with an allegation of historic infringement which Hoechst accepted had come to an end (see [1]), and both sides had adduced extensive evidence about the BP process conditions over the whole period of alleged infringement (see [59]). Thus the claim was the exact opposite of a quia timet case.

In Monsanto Technology LLC v Cargill International SA [2007] EWHC 2257 (Pat), [2008] FSR 7 the patent in suit was concerned with enzymes called EPSPS enzymes which, if expressed in a plant, conferred resistance to the herbicide round up. Claim 1 was to an isolated DNA sequence encoding an EPSPS enzyme, claim 4 (using the numbering after amendment) was to a recombinant double-stranded DNA molecule having particular characteristics, claim 5 was a method claim corresponding to claim 4 and claim 6 was to a plant cell comprising a DNA molecule of claim 4. The allegation of infringement concerned 5000 tonnes of soya bean meal shipped on the MV Podhale to the United Kingdom. The meal had been made from soya beans grown from seed carrying the gene for one of the EPSPSs disclosed in the patent. Pumfrey J found as a fact that there was present in the meal some genomic DNA which included the EPSPS gene and that some or all of that DNA was double-stranded. He nevertheless held that the allegations of infringement failed for various reasons. He added at [89]:

“I should also deal with a de minimis point. The DNA present in the meal, such as it is, is entirely irrelevant to the meal as an animal feedstuff, is present in small, variable, quantities and may not be present at all if processing conditions are changed. It is not in any serious sense genetic material. It is just the remains of the material which was in the soybeans from which the meal was extracted. This, it seems to me, is irrelevant. It may raise a question on damages, that there is no causative relationship between acts of infringement, as opposed to acts which are not infringing by English law, and the loss suffered by Monsanto, but this was not argued. There is, generally, no authority in favour of trace quantities of infringing material being held not to infringe, and some authority against it. In any event, I had no proper estimate of the quantity of DNA in the Podhale meal that survived: Professor Lichtenstein put it at 5 per cent of the DNA in cross-examination but the issue was not properly pleaded.

It is not clear what Pumfrey J had in mind when he said there was “some authority against” “trace quantities of infringing material being held not to infringe”, although it may have been Hoechst v BP. Be that as it may, if the 5% estimate was accurate, that would have meant that 250 tonnes of meal contained potentially infringing DNA.

In Napp Pharmaceutical Holdings Ltd v Ratiopharm GmbH [2009] EWCA Civ 252, [2009] RPC 18 one of the patents in suit was to “a controlled release oxycodone dosage form ... wherein said dosage form has an in vitro dissolution rate, when measured by the USP Paddle method” of specified amounts. The USP method was described in the US Pharmacopeia. This described acceptance tests which involved testing at least 6 tablets and possibly up to 24 tablets. There was an issue of construction as to whether the claim called for the whole procedure described in the USP to be carried out, including the acceptance test, so that the claim only covered what passed the acceptance test, or whether it was sufficient that an individual dosage form passed the test. As Jacob LJ explained at [53], the point mattered because “only a modest (7%, nevertheless significant) proportion” of the defendants’ tablets had dissolution rates within the claim. The Court of Appeal agreed with the judge that the claim on its true construction only required an individual tablet to have the specified rates of dissolution. Jacob LJ addressed one of the contrary arguments at [56] as follows:

“Mr. Silverleaf sought to avoid that construction because, necessarily, a single dosage form would be destroyed in testing it for infringement. He suggested it would be more rational to conclude that the skilled reader would be concerned with a batch, which is what the USP acceptance criteria are aimed at. We do not agree. If a single tablet is tested and is shown to be within the claim, it follows that the defendant has infringed. If some of his tablets are outside the claim, experiments can clearly be run to estimate the scale of infringement. Moreover, Mr. Silverleaf's construction also involves destruction of tablets—and perhaps quite a lot of them. So the destruction point is essentially neutral.”

Again Jacob LJ did not consider the application of the de minimis principle.Moreover, his comments must be seen against the context that it was not disputed that 7% of the tablets fell within the claim. As Jacob LJ pointed out,that was a significant proportion. Jacob LJ was not considering what the position would be if, say, a single tablet out of a million fell within the claim. Still less was he considering a claim that, statistically speaking, it could be predicted that one tablet out of a million produced over an extended period would fall within the claim.

In H. Lundbeck A/S v Norpharma Spa [2011] EWHC 907 (Pat), [2011] RPC 23 the patent in suit claimed a process for preparing a chemical referred to as 5-cbx which involved “heating the mixture at 120-145^oC”. The period of heating was not specified. One of the issues was what happened if the temperature went into and out of the claimed range. Floyd J held that the skilled person would understand that, provided he made some 5-cbx within the specified temperature range, then he would infringe, even if the rest of the 5-cbx was made at a temperature outside the claimed range. He went on at [49]:

“It is arguable that there may come a point where the amount of time that the process is within the range is so small that it can be ignored. Obviously if the time is so short that no 5-cbx is made during that time, then this does not infringe. I did not hear argument on whether there should be a ‘de minimis’ limit and, if so, what it might be. The point is not free of authority: see the observations of Pumfrey J. in Monsanto Technology LLC v Cargill International SA [2007] EWHC 2257 (Pat), [2008] F.S.R. 7 in connection with a product claim. As I do not think that infringement by any of the processes with which I am concerned really turn on this point, I say no more about it here.”

In Generics (UK) Ltd v Warner-Lambert LLC [2015] EWHC 2458 (Pat) I held that Actavis had not infringed the Patent pursuant to section 60(1)(c) of the Patents Act 1977 because it was not foreseeable that the Lecaent would be administered for the treatment of pain “save in a small number of exceptional cases which I consider that it is proper to regard as de minimis.” I did not consider the application of the de minimis principle in any detail, however, and I did not have the benefit of the citation of authority which I have had in the present case. Furthermore, this decision is presently under appeal.

In the present case counsel for Napp accepted that the de minimis principle applied to claims for patent infringement. Moreover, he accepted that, to take an extreme example, if only one patch in a trillion (10¹²) fell within the claim, then the Defendants would not infringe the claim. He nevertheless submitted that the Defendants would infringe if larger, but still very small, proportions of their patches infringed. By way of illustration, he postulated a scenario in which it could be shown that, out of 2 million patches sold by a defendant between now and the expiry of the Patent, 200 fell within the claim (i.e. 1 in 10,000 or 0.01%). He argued that a defendant who sold just 200 patches which fell within the claim would undoubtedly infringe, and that it made no difference if the 200 patches constituted a single hour’s production out of a year’s worth of production, nor even if the 200 patches were randomly distributed amongst 1,999,800 non-infringing patches.

I can only say that I disagree. It seems to me that most people, and specifically the skilled person, would be very surprised by the proposition that selling products only 0.01% of which fall within the claim constitutes patent infringement, particularly where the 0.01% are randomly distributed among the remainder. I consider that this is precisely the kind of situation covered by the de minimis principle.

Furthermore, while I accept the force of Lord Phillips’ warning about the dangers of trying to define a quantitative limit, it seems to me that, for reasons which will become clear, in the present case the court is forced, as a matter of practical reality, to draw a line somewhere. Where that line should be I will consider below.

Quia timet claims. It is well established that in the context of patent infringement, as in the context of many other wrongs, it is not necessary for the patentee to establish that the defendant has already committed a wrongful act in order to commence proceedings. Rather, the patentee can commence proceedings on the basis that the defendant threatens to do an act which will infringe the patent. Such claims are known as quia timet (Latin for “because he fears”) claims. In many cases, by the time the case gets to trial, the defendant has started doing the acts which were said to be threatened, in which case the fact that the claim began as a quia timet claim is usually academic. Even if the defendant has not started doing the acts, frequently there is no dispute that the defendant intends to do such acts, and the issue is whether they will amount to infringements. Occasionally, however, there is a dispute as to whether the defendant has in fact threatened to do the acts which the patentee alleges the defendant has threatened to do. It may be wondered what the point of such a dispute is. Generally speaking, there are two possible reasons for seeking the determination of such a dispute. The first is because the patentee wants to obtain an injunction. Of course, if there is no threat to do the acts, then the patentee does not need an injunction; but the patentee may distrust the defendant’s protestations that it does not intend to do such acts. The second reason is because of a dispute over the costs of the litigation. What has happened in a number of cases is that, by the time of the trial, it is clear that there is no threat by the defendant, but the claimant says that there was a threat when it started the proceedings and so the claimant is entitled to its costs.

As counsel for Sandoz pointed out, Birss J considered the law on this question in Merck Sharp Dohme Corp v Teva Pharma BV [2013] EWHC 1958 (Pat), [2014] FSR 13 at [39]-[59]. Having reviewed the authorities, he stated the applicable principle at [56] in the following terms:

“The principle I derive from these authorities is that the question the court is asking in every case is whether, viewed in all the relevant circumstances, there was a sufficiently strong probability that an injunction would be required to prevent the harm to the claimant to justify bringing the proceedings. In adding the word sufficiently to the word strong … I am seeking to encapsulate the idea that the degree of probability required will vary from case to case depending on all the circumstances but that mere possibilities are never enough. To justify coming to court requires there to be a concrete, strong and tangible risk that an injunction is required in order to do justice in all the circumstances.”

The question which arises in this case is how this principle relates to the de minimis principle. I think it is clear that, if what the defendant threatens to do would only involve infringement on a de minimis scale, then that threat does not justify the commencement of proceedings by the patentee (whether in order to seek an injunction or in order to seek a financial remedy). But counsel for Sandoz went further: he submitted that, even if the scale of infringement which was threatened was such that, viewed after the event, it could not be discounted as de minimis even though it was very small, it did not follow that, viewed prospectively, that justified the commencement of proceedings on a quia timet basis. Rather, what was required was a threat to infringe on a scale which gave rise to a sufficiently strong probability of harm to the patentee.

Although I was initially attracted by this submission, upon reflection I do not accept it. In my view it confuses two different concepts. One is whether there is a sufficiently strong threat to justify the commencement of proceedings. The other is whether the acts threatened will fall within the claim sufficiently often to amount to an infringement. If there is a clear threat to do acts which will fall within the claim sufficiently often that they cannot be discounted as de minimis, then in my judgment that is sufficient to justify the bringing of proceedings on a quia timet basis. Whether it is sufficient to justify the grant of an injunction is a separate question, which I shall consider below.

The burden of proof for future acts. It is common ground between Napp and Sandoz that Napp bears the burden of proving that Sandoz has infringed the Patent, in so far as the claim concerns acts which have already been committed, and that Sandoz will infringe the Patent, in so far as the claim concerns acts which Sandoz threatens to do in the future. There is no dispute as to what acts Sandoz proposes to do. The dispute is as to whether those acts will involve patches which fall within the claim. In my judgment it follows from the discussion above that Napp bears the burden of proving that the acts which Sandoz proposes to do in the period before expiry of the Patent will result in infringement on a scale which is more than de minimis.

The standard of proof for future acts. The standard of proof applied by English courts to civil claims is proof on the balance of probabilities. Although it can be difficult to apply in practice due to evidential difficulties, this standard is conceptually reasonably straightforward when the court is concerned with events which have already occurred (or are alleged to have occurred). But does the same standard apply when the court is concerned with future events, and in particular future events only a small proportion of which satisfy the relevant legal criterion? If so, how is it to be applied? If not, what standard should be applied? I received very little assistance from counsel on these questions, but in fairness to counsel, they appear to have received very little consideration prior to the present case. I have not found any authorities or commentaries which shed any light on them.

In order to explain the problem, I must set out some of the statistical dimension to it. At this stage, I shall try and do so in as non-technical manner as possible.

When discussing the de minimis principle, I considered the example postulated by counsel for Napp of a defendant who sells 2 million patches before expiry of the Patent, of which 200 fall within the claim. This example assumes that it is known that 1 in 10,000 patches will fall within the claim. But what if it is not known what proportion of patches will fall within the claim?

As explained above, the manufacturing processes used to produce TDSs are inherently variable, at least to some extent. It follows that, even if the specification for Sandoz’s product is intended to ensure that the levulinic acid content of the patch is (to take a hypothetical figure) 6.5%, it is inevitable that some patches will have more and some less. To put it in statistical terms, there will be a distribution of levulinic acid values. It may be assumed for present purposes that the distribution is a normal distribution. (Dr Carroll pointed out in relation to Napp’s case against Dr Reddy’s that this is not necessarily the case, but he himself assumed that it was for the purposes of his analyses for Napp’s case against Sandoz, and I consider that it is clear from the evidence that Dr Carroll’s concerns in this respect were theoretical ones. Although counsel for Sandoz himself suggested a reason why the values might not be normally distributed, I do not consider that it was a sound one. Moreover, Prof Sasieni’s evidence was that there was nothing in the Dr Reddy’s data which was inconsistent with it being normally distributed.)

In addition, most analytical techniques are not 100% precise. Rather, there is some degree of random error in the results. (There can also be a problem with the accuracy of the technique i.e. a systematic error which introduces a bias into the results, but there is no suggestion that the assays in the present case suffered from this problem.) Again, it may be assumed for present purposes that repeated measurements of the levulinic acid content of the same patch would produce results which are normally distributed.

If all 2 million patches could be tested for levulinic acid content, one could determine the mean levulinic acid content and the standard deviation of the measurements. The standard deviation would reflect the two sources of variability in the measurements: the underlying variability in the levulinic acid content and the random error in the measurements. If one wanted to know what proportion of patches contained 7.5 %-wt levulinic acid or more, one would not be able to give a precise answer to that question. One could only give an answer which was as precise as the precision of the analytical technique permitted. The less precise the analytical technique, the less precise the answer would be, that is to say, the greater the uncertainty as to the correctness of the answer.

In the present case, it is not even possible to do this. This is for two reasons. First and most fundamentally, most of the 2 million patches have not yet been manufactured. To date, only a relatively small number have been made. Secondly, the testing is destructive, so it would in any event be necessary to test a representative sample of patches rather than all of them. As matters stand, therefore, the tests that have been carried out to date can be used to determine an observed mean and standard deviation for levulinic acid content, but one cannot be certain how close the observed mean and standard deviation of the sample patches which have been tested to date are to the true mean and standard deviation of the whole population of 2 million patches which will be manufactured. This is because one cannot be certain how representative the tested samples are of the whole population. Assuming that the samples can be, and are, randomly selected, this is essentially a function of sample size: the larger the number of samples tested, the more likely the tested samples are to be representative of the whole population.

So far as the legal question of what standard of proof to apply is concerned, I see no reason why Napp should be subjected to a more stringent burden of proof than the balance of probabilities standard. Equally, to anticipate my consideration of Dr Reddy’s claim for a declaration of non-infringement, in respect of which the burden of proof lies on Dr Reddy’s, I see no reason why Dr Reddy’s should be subjected to a more stringent burden of proof than the balance of probabilities standard. The question, therefore, is how the balance of probabilities standard should be applied in these circumstances.

As I have already said, Napp bears the burden of proving that Sandoz will commit infringing acts before the expiry of the Patent on a scale which is more than de minimis. Napp must prove this on the balance of probabilities. This inquiry can be divided into two questions: (1) what proportion of patches Napp has proved on the balance of probabilities will fall within the claim?; and (2) is that proportion more than de minimis?

So far as question (1) is concerned, the answer will inevitably be affected by the uncertainties caused by (a) the lack of precision in the analytical data and (b) the relatively small number of samples that have been tested so far. In my view this must involve the application of a statistical test which takes these matters into account. But in considering what statistical test is appropriate, the following important points must be borne in mind.

First, Napp has chosen to accept the analytical data relied upon by Sandoz. Napp could have challenged the analytical technique employed by Sandoz or its precision. It could also, without challenging the appropriateness or the claimed precision of the technique, have investigated the consequences of the claimed precision. Napp did none of those things. Instead, Napp elected to proceed by taking the analytical data at face value.

Secondly, Napp has chosen to rely upon data from the relatively small number of samples that have been tested so far (or, to be more accurate, the Annex 7 subset of the samples which have been tested) in support of its claim. As noted above, Napp could have waited and relied upon a larger quantity of data. Instead, Napp has elected to proceed upon the basis of a very small sample of data.

In my judgment Napp is not entitled to take advantage of the uncertainties which flow from its own choices when it comes to the selection of the appropriate statistical test. On the contrary, those uncertainties are matters that Sandoz is entitled to rely upon. To the extent that those uncertainties make it more difficult for Napp to establish on the balance of probabilities what proportion of patches will fall within the claim, Napp has only itself to blame.

Remedies. What remedy or remedies should the court grant if it finds that the defendant is threating to do acts which will fall within the claim sufficiently often that they cannot be discounted as de minimis, but which nevertheless amount to infringement on a very small scale? Suppose, for example, it is concluded on the balance of probabilities that the defendant will sell 2000 infringing patches randomly distributed among 1,998,000 non-infringing patches, and it is also concluded that that quantity cannot be discounted as de minimis?

As counsel for Napp accepted, remedies for patent infringement, and in particular the remedy of an injunction, must be effective, proportionate and dissuasive and must be applied in such a manner as to avoid the creation of barriers to legitimate trade and to provide for safeguards against their abuse: see Article 3(2) of European Parliament and Council Directive 2004/48/EC of 29 April 2004 on the enforcement of intellectual property rights and HTC Corp v Nokia Corp [2013] EWHC 3778 (Pat), [2014] Bus LR 217 at [19]-[28].

It seems to me that these principles provide the answer to the question posed in paragraph 168 above. Even if the level of infringement cannot be discounted as de minimis in such a case, I consider that an injunction would be both disproportionate and a barrier to legitimate trade. It would be disproportionate because the harm to the patentee from infringement on such a small scale would be indistinguishable from the harm caused by wholly non-infringing acts. It would be a barrier to legitimate trade because the practical effect of such an injunction would be to require the defendant to operate even further outside the boundaries of the claim, and thus would effectively extend the scope of the patentee’s monopoly. In such a case, the appropriate remedy would be a financial one.

On the face of their respective reports, there was a startling, and indeed alarming, difference between Dr Carroll’s and Prof Sasieni’s analyses of the levulinic acid data in the Sandoz PPD.

Dr Carroll expressed his conclusion as follows:

“On the basis of [the data in Annex 7 to the Sandoz PPD], the probability that a Sandoz Patch would contain more than or equal to 7.5 %-wt levulinic acid is 1/1560 (with 95% confidence). This means that, on average, I would expect 1 out of 1560 Sandoz Patches to contain more than or equal to 7.5 %-wt levulinic acid.”

Dr Carroll went on to say that this analysis assumed that each of the data points in Annex 7 was an individual value, and did not take into account the fact that each of the data points was the mean of two samples. Allowing for that, the figure was 1/1280 rather than 1/1560.

Dr Sasieni expressed his conclusion as follows:

“Thus in the three batches [in Annex 7 to the Sandoz PPD], the proportions of patches that will, with 95% confidence, have less than 7.5 %-wt levulinic acid is all but 23 in a trillion. This does not mean that 23 per trillion patches will have a levulinic acid level of 7.5% or more, it means that with 95% confidence, the proportion of patches with less than 7.5 %-wt levulinic acid will be 1 – 2.3x10^-11. In lay language, with 95% confidence, all but 23 patches in any given trillion patches will have less than 7.5 %-wt levulinic acid.

Further, the proportion of patches that will, on the balance of probabilities have less than 7.5 %-wt is all but 19 per 10²⁰ (that is 1 followed by 20 zeros). This does not mean that this proportion of patches will have a levulinic acid level of 7.5% levulinic acid or more; it means that, with 95% confidence [sic], all but 19 patches of 10²⁰ patches will have less than 7.5 %-wt levulinic acid.”

Before proceeding further, I should explain that Dr Carroll analysed the Sandoz data before Prof Sasieni did and that Dr Carroll only analysed the data in Annex 7 to the Sandoz PPD. He did not analyse the data in Annex 6 because they derived from a small number of samples from a single batch, and given the shortage of time he did not consider that the data were likely to be informative over and above the data contained in Annex 7. Prof Sasieni did not agree that the data would be uninformative, but for consistency with Dr Carroll’s analysis, Prof Sasieni confined his analysis to the data in Annex 7 as well.

More importantly, neither expert analysed the data in Annexes 8 and 10 to the Sandoz PPD. It is fair to note that this only became available on 2 June 2016. Nevertheless, it should have been possible for Dr Carroll to have repeated his analysis incorporating this data. In the case of Prof Sasieni, from 30 May to 5 June 2016 he was heavily engaged in analysing the data which had just became available relating to the Dr Reddy’s product (as to which, see below).

In my view the most important question which arises out of the evidence of Dr Carroll and Prof Sasieni is as to the nature of the statistical test which is to be employed, rather than as to the sometimes rather subtle differences in their methods and calculations. I have discussed the legal dimension to this question above.

As Prof Sasieni explained, in statistics there are three related concepts that are easily confused. The first is the confidence interval (1-α). This gives an interval which will include the true value of a population parameter, such as the mean, a certain percentage of the time. The second is the prediction interval. This gives an interval which it can be expected will include a certain percentage of the population. The third is the tolerance interval (p, 1-α). The tolerance interval gives an interval that will contain a portion of the population (p) to a degree of certainty (1-α).

All these intervals require the selection of an appropriate level of confidence. As Prof Sasieni acknowledged, it is conventional in many statistical contexts to select an interval of 95%. In particular, it is commonplace to conduct an experiment (e.g. a clinical trial) to test a hypothesis (e.g. that a particular substance is more efficacious to treat a particular condition than a placebo). In that context it is conventional to take a significance level (α) of 0.05 or 5%, which corresponds to a confidence interval of 95%, for the threshold of statistical significance.

It is worth noting that confidence intervals are frequently misunderstood, even by professional scientists. For example, a 95% confidence interval does not mean that for a given interval calculated from sample data there is a 95% probability the population parameter lies within the interval, nor that there is a 95% probability that the interval covers the population parameter, as is sometimes supposed.

In his first report, which was prepared on behalf of Dr Reddy’s, Prof Sasieni said that he had been instructed to devise a testing regime and statistical protocol to show with “suitable statistical certainty” that batches of patches fell within or without the relevant ranges of %-wt of levulinic acid and oleyl oleate. He derived an equation one of the variables of which was (1-α), and worked examples using confidence intervals of 50%, representing the balance of probabilities, and 95%. He did not express an opinion as to the appropriate confidence level, on the basis that this was a matter for the Court.

Prof Sasieni explained in his second report that the equation he had derived was that of a tolerance interval. He explained in his third report, and during cross-examination, that his equation was similar to (but more conservative than) that set out in section 7.2.6.3 of the Engineering Statistics Handbook, which is a reference manual published by the US National Institute of Standards and Technology (NIST).

Dr Carroll made a number of criticisms in Annex A to his first report of the approach adopted by Prof Sasieni in his first report. Although Dr Carroll did not make the point explicitly in Annex A to his first report, Prof Sasieni noted in Annex A to his second report that Dr Carroll appeared to favour an approach based on prediction intervals. Dr Carroll confirmed in cross-examination that he considered that the use of a prediction interval was more appropriate. This was because Dr Carroll understood that Prof Sasieni was estimating the probability that the mean of all the patches in a batch exceeded a certain limit, whereas Dr Carroll considered that the right question was whether an individual patch selected at random exceeded that limit. As counsel for Dr Reddy’s submitted, this is in part a misapprehension. As Prof Sasieni explained in cross-examination, his approach involved considering the interval for a percentile (or rather, a fraction-of-a-percentile). But that is not the whole story, as I shall explain shortly.

A separate point which Dr Carroll made in Annex A to his first report was that a 50% confidence level was very rarely used in statistical testing because it was unreliable. He enlarged on this point in cross-examination, saying that it was inappropriate because it failed to take into account the size of the data set from which the probability was being estimated. No doubt for this reason, in Annex B to his first report Dr Carroll analysed the data in the Sandoz PPD using a 95% confidence level.

Dr Carroll’s approach in Annex B was to use the data in Annex 7 to calculate the probability that a randomly selected patch from a future batch would have a levulinic acid level of 7.5% or more. I have quoted the key part of his conclusion in paragraph 172 above.

In Annex B to his third report Prof Sasieni both analysed the levulinic acid data in Annex 7 to the Sandoz PPD and commented on Dr Carroll’s analysis. As Prof Sasieni explained, he analysed the data in Annex 7 to estimate the number of patches from the three batches that would have 7.5% or more levulinic acid. I have quoted Prof Sasieni’s conclusion with respect to this analysis in paragraph 173 above. For good measure, he also set out the results of using the NIST formula, which produced even smaller figures. He then went on to set out calculations about future batches based on various numbers of samples tested in each batch, on both 95% and 50% confidence bases.

Prof Sasieni commented on Dr Carroll’s conclusion which I have quoted in paragraph 172 above as follows:

“Irrespective of what I think of Dr Carroll’s Approach, I disagree with the way in which Dr Carroll characterises the probability he is seeking to calculate. For instance in paragraph 27 he writes ‘The probability of a Sandoz Patch containing more than or equal to 7.5%-wt levulinic acid has therefore been calculated, with 95% confidence to be 1/1560’. In fact Dr Carroll's calculations represent with 95% confidence the chance that all but one of the 1560 Sandoz patches in any particular sample size fall below the relevant threshold level of levulinic acid (in this case 7.5 %-wt). The chance that the remaining patch is greater than the threshold is only 5%.”

As counsel for Sandoz pointed out, Prof Sasieni was not challenged on this paragraph in cross-examination. Rather, what was put to Prof Sasieni, and he agreed, was that it would be correct to say, on the basis of Dr Carroll’s calculations: “We can say with 95% confidence the probability that a patch has levulinic acid greater than or equal to 7.5% is at most 1/1560.” This is a statement about the upper limit of the confidence interval. This suggests, and Dr Carroll’s cross-examination appeared to me to confirm, that Dr Carroll did not really take issue with Prof Sasieni’s criticism of his characterisation of the probability.

Counsel for Napp attempted to deal with this point in closing submissions by submitting that, even on this basis, the probability that a Sandoz patch would have levulinic acid greater than or equal to 7.5 %-wt was 1 in 25,600 (1/20 x 1/1560). I agree with counsel for Sandoz that this submission was not open to Napp, since the correctness of the assumption that the two probabilities can be simply multiplied in this way was not put to Prof Sasieni in cross-examination.

But, as counsel for Sandoz also submitted, there is a more fundamental objection to this calculation, which is that it is not appropriate to use the upper limit of the 95% confidence interval in this way. On any view, the lower limit of the 95% confidence interval is a very small number indeed. The whole point of the 95% confidence interval is that it encompasses both extremes and everything in between. Thus Napp’s approach amounts to saying that the lack of precision in the analytical technique and the small sample size creates statistical uncertainty, and then exploiting that uncertainty by basing the claim for infringement upon the most favourable extreme of the statistical possibilities. If a 99% confidence were selected instead of a 95% confidence interval, and the upper limit taken, that would magically improve Napp’s case even further.

In these circumstances the appropriate confidence interval to apply is the 50% confidence interval. As is common ground, the effect of using a 50% confidence is to equate the observed mean and standard deviation with the true mean and standard deviation. This may seem counter-intuitive when one knows that, for the reasons explained above, the observed mean and standard deviation may not accurately reflect the true mean and standard deviation. But, as I have endeavoured to explain, it is Napp which has chosen to base its case upon a limited and imprecise data set. There is no unfairness to Napp in saying, in effect, “given that you have chosen to base your claim upon this limited and imprecise data set, what does that data show, on the balance of probabilities, about the proportion of Sandoz patches which have a levulinic acid level equal to or greater than 7.5%?”.

Prof Sasieni pointed out in Annex B to his third report that Dr Carrollhad not attempted to determine the probability of a Sandoz patch having a levulinic acid level equal to or above 7.5 %-wt on the balance of probabilities. He calculated that, using the approach adopted by Dr Carroll, on the balance of probabilities only 1 in 153 million patches from future batches would exceed 7.5% levulinic acid (corresponding to 5.685 standard deviations away from the observed mean). When cross-examined about this, Dr Carroll disagreed with one aspect of the way in which Prof Sasieni had carried out this calculation. According to Dr Carroll’s method, the correct answer is 1 in 69 million patches (corresponding to 5.548 standard deviations away from the observed mean). If I have understood correctly, this difference relates to Dr Carroll’s preference for using a prediction level and Prof Sasieni’s preference for using a tolerance interval.

Be that as it may, I do not think it matters who is right with respect to the method of calculation. Even assuming that Dr Carroll is right, 1 in 69 million patches is not sufficient to lead to a probability of infringement even by a single patch before the expiry of the Patent. In any event, it is plainly de minimis. Accordingly, on the assumptions stated in paragraph 131 above, there is no threat by Sandoz to infringe the Patent.

I would add that, even if Napp was correct in using the upper limit of the 95% confidence interval in the way that it tried to, I consider that 1 in 25,600 patches would be de minimis. Accordingly, even on that basis, there is no threat by Sandoz to infringe the Patent.

I should record that Sandoz offered if necessary to undertake that it would not import into the UK any patches made from a batch with a mean levulinic acid content greater than or equal to 7.0 %-wt. In the light of my conclusions, it is not necessary for Sandoz to give this undertaking.

Although it does not affect the conclusions expressed above, it remains necessary finally to consider a separate point about Dr Carroll’s and Prof Sasieni’s analyses. In his first report, Prof Sasieni expressly assumed that there would be no systematic variation along the length of the Dr Reddy’s production line. As he explained in Annex A to his second report, he did not assume that there would be no batch-to-batch variability, but he did assume that the variability within a batch would be the same from one batch to the next. As noted in paragraph 21 above, he stated that he considered that this was an entirely reasonable assumption. At that stage, he had not analysed any of the Sandoz data.

In Dr Carroll’s analyses of both the buprenorphine data in Annex 4 and the levulinic acid data in Annex 7 to the Sandoz PPD in Annex B to his first report, he found that there was strong evidence, statistically speaking, that there was batch-to-batch variability between the three batches. In his analyses of the buprenorphine data in Annex 4 to the Sandoz PPD, he also found that there was strong evidence, statistically speaking, that there was systematic variation across the width of the laminate with the %-wt of buprenorphine increasing from left to right. In his analysis of the levulinic acid data in Annex 7 to the Sandoz PPD, he also found that there was evidence of systematic variation across the width of the laminate with the %-wt of levulinic acid being higher at the sides than in the middle, although there was not enough data to assess the statistical significance of this.

As noted above, Prof Sasieni analysed the levulinic acid data in Annex B to his third report. He was not instructed to analyse the buprenorphine data, because Sandoz’s lawyers appreciated that this was unnecessary.

Subsequently Dr Carroll set out in his second report the means and standard deviations he had calculated from the buprenorphine data in Annex 4 and the levulinic acid data in Annex 7. In each case, there is a striking difference between the standard deviation of one batch and the standard deviation of the other two (although which batch is the odd one out differs). Prof Sasieni gave evidence that, when he noticed the difference in the buprenorphine standard deviations, he asked those instructing him about it and was told that there was reason to believe that the calibration of the buprenorphine assay was wrong for one batch, but anyway the buprenorphine level was not critical. Sandoz did not adduce any evidence about the calibration, but nevertheless it is understandable that in these circumstances Prof Sasieni did not concern himself greatly with the buprenorphine data.

In any event, however, Prof Sasieni had expressly noted in Annex B to his third report that, based on his own analysis of the levulinic acid data which showed that a mean for one batch that was noticeably different to the means for the other two batches, the estimated within-batch variation was considerably less than the between-batch variation.

Leaving aside the question of Prof Sasieni’s credibility, what the evidence shows is that the assumption which Prof Sasieni originally made is not a valid one, as he accepted. As I have already said, this does not affect the conclusions I have reached in relation to Napp’s claim against Sandoz. I will address its relevance to Napp’s claim against Dr Reddy’s below.

Dr Reddy’s PPD describes the process which it is proposed will be used by Dr Reddy’s manufacturer to manufacture its product and of the assay method for determining the %-wt of levulinic acid and oleyl oleate in the patches. In some respects, it contains less detail about the process than the Sandoz PPD. For example, it states that the drug adhesive blend is “stirred for at least 60 minutes until homogenous”. Dr Reddy’s resisted giving any further details of the process. The explanation for this is that, for the reasons indicated in paragraph 6 above, Dr Reddy’s approached this dispute on the basis that, in order to avoid infringement, it would implement a testing regime based on a statistical protocol devised by Prof Sasieni to ensure that no more than a vanishingly small proportion of Dr Reddy’s patches fell within the claim once it had been construed by the Court. Moreover, Dr Reddy’s is willing to undertake not to infringe the Patent.

Although Dr Reddy’s counterclaimed for a declaration of non-infringement, it only sought a declaration that transdermal patches having specified alternative ranges of (i) levulinic acid and (ii) oleyl oleate would fall outside the claim. It did not counterclaim for a declaration that a product manufactured in a specified manner would fall outside claim (and it is debatable whether its PPD would have enabled it to do so without further elaboration). Nor did it counterclaim for a declaration that its proposed testing regime was fit for purpose to avoid infringement.

Despite this, Napp joined issue with Dr Reddy’s as to the suitability of Prof Sasieni’s statistical protocol. Thus Napp’s quia timet claim against Dr Reddy’s, on the assumption that the claim specifies output %-wts rather than inputs, is based purely on the proposition that, because the proposed statistical protocol is flawed, there is a threat by Dr Reddy’s to market patches between the date Dr Reddy’s is ready to launch and the expiry of the Patent an appreciable number of which will fall within the claim.

Dr Reddy’s made it clear, however, that the final statistical protocol which would be adopted was conditional upon the court’s decision in a number of respects: first, the Court’s construction of the claims; secondly, the Court’s view as to the appropriate confidence level; and thirdly, the Court’s view to what would constitute a de minimis level of infringement.

In order to explain my reasoning and conclusions below, I need to explain that the Dr Reddy’s process involves cutting the drug adhesive laminate into “bolts” approximately 250 m long. In between each bolt, a shorter 50 cm strip is cut for quality control purposes (a “QC strip”). There are typically 3-5 bolts in a production run. (This information, it may be noted, is not contained in Dr Reddy’s PPD, but in Prof Sasieni’s first report.) It is the QC strips which are proposed to be assayed for levulinic acid and oleyl oleate content.

Like the Sandoz PPD, the Dr Reddy’s PPD contains information about the precision (and accuracy) of the analytical methods which will be used. Again, Napp did not raise any issue as to the precision (or accuracy) of these analytical methods.

Until shortly before trial, there was no data relating to Dr Reddy’s proposed product. As mentioned in paragraph 10 above, however, data relating to Dr Reddy’s proposed product became available shortly before trial. This data was exhibited by Prof Sasieni to, and analysed by him in Annex A to, his fourth report. Three points should be noted about this data. First, Dr Reddy’s did not adduce any factual evidence to prove this data. Nevertheless, Napp with commendable pragmatism did not require Dr Reddy’s formally to prove the data. Counsel for Napp did point out that there was no evidence that the patches in question were made by the process described in the PPD either, but there is no reason to doubt this.

Secondly, the data showed that the assumption which Prof Sasieni had made in his first report that there would be no variation along the length of the production line was invalid and that there was (as with the Sandoz product) evidence of variation with respect to levulinic acid content across the width as well. Accordingly, Dr Reddy’s revised its proposed protocol and testing regime. In particular, whereas Dr Reddy’s had originally proposed a protocol based on a two-stage process of benchmarking and release testing, the revised proposal for is for a single-stage process of testing each batch of product. This is an important change, since it addresses one of Napp’s key criticisms of the original protocol.

Thirdly, in my view the data provide a better basis for assessing whether Napp has a viable quia timet claim than the battle over the statistical protocol. I bear in mind, however, that, because the data became available so late, Dr Carroll did not have time to carry out his own analysis of it, as opposed to commenting on Prof Sasieni’s analysis. Accordingly, although I shall express some provisional views based on the data, I shall nevertheless proceed to consider the protocol.

I summarise the input composition of Dr Reddy’s product and the available data with respect to its output composition, so far as is relevant to the issues on infringement, in Confidential Annex B.

The issues on infringement are whether the Dr Reddy’s product will have (i) 10-15 %-wt levulinic acid and (ii) about 10 %-wt oleyl oleate. For the reasons given above, I conclude that these questions fall to be answered on the bases that (a) the percentage weights are of the final product and (b) the numerical limits are to be interpreted as being expressed to the nearest whole number, except that “about 10 %-wt” is to be interpreted a little more broadly. Although Napp made no concession equivalent to the concession it made in respect of Sandoz, it seems to me that the analysis of the data relating to the Dr Reddy’s product in Annex A to Prof Sasieni’s fourth report shows that, if the data is representative, then the Dr Reddy’s product will not fall within the claim.

If, on the other hand, (a) the percentage weights are of the final product, but (b) the numerical limits are to be interpreted as contended for by Napp, then Prof Sasieni’s analysis of the data shows that, if the data is representative, then a substantial proportion of every batch of the Dr Reddy’s product will fall within the claim. Since Dr Reddy’s position is that, unless a batch can be shown with suitable statistical certainty to fall entirely outside the claim, the batch will not be imported into or sold in the UK, it does not appear that Dr Reddy’s will be able to sell its product within the UK. As I understand counsel for Dr Reddy’s closing submissions, Dr Reddy’s accepts this.

As to whether the data is representative, Napp contends that it may well not be. There is no dispute that, in principle, samples should preferably be taken randomly. Napp points out that Dr Reddy’s procedure involves samples that are not taken at random, but are taken at fixed intervals along the production line. Thus they do not exclude the possibility of cyclical systematic variation with a period the same or similar to that of the period between QC strips. Dr Reddy’s accepts that that is so, but it contends that it does not follow that they are unrepresentative.

Prof Sasieni was very clear that this was really a question for someone who was an expert in the manufacturing process. Moreover, he accepted that this was a much more serious issue than some of the other points raised on behalf of Napp. Nevertheless, he said that he had not seen anything to suggest that the samples were not representative.

There is no evidence suggesting that a cyclical variation in the process with a period which coincides with that of the QC strips could arise. Moreover, the evidence from the data analysed in Prof Sasieni in his fourth report strongly suggests that there is no cyclical variation of the kind postulated by Napp. On the contrary, it indicates that, in the case of levulinic acid, there is a steady decrease in the %-wt along the length of the production run (see paragraph 232 in Confidential Annex B below). Accordingly, I conclude that samples taken from the QC strips are representative of the bolts between them.

A related point made by Napp concerns the number of samples which are to be taken. The data analysed by Prof Sasieni in his fourth report consists of nine samples from each of six QC strips i.e. 54 samples per production run of five bolts. While I agree that this is on the low side, particularly given that the samples will only come from six points along the length of the run, it is not significantly lower than the numbers of samples tested by Sandoz. In my view it is an adequate number.

Turning to the proposed statistical protocol, a number of criticisms were put to Prof Sasieni in cross-examination. He was adamant, however, that his approach was a conservative one which, if anything, would tend to overestimate the probability of patches having a levulinic acid or oleyl oleate content greater than the specified limit. In my view it is sufficient for me to address the two main criticisms which were pursued by counsel for Napp in his closing submissions.

The first concerns variability within each batch. As noted above, Dr Reddy’s revised protocol involves separate testing of each batch, and therefore variation between batches is no longer a concern. The revised protocol takes into account Prof Sasieni’s finding that the levulinic acid data showed both systematic variation along the length of the run (a steady decline in the levulinic acid content) and a systematic variation across the width of the run (levulinic acid content being lower in the middle than at the sides) while the oleyl oleate data showed a systematic variation along the length of the run (oleyl oleate content being higher at the beginning and end of the run than in the middle) and no systematic variation across the width of the run. Accordingly, the revised protocol provides for the data to be analysed according to either (i) a function for V-shaped systematic variability with length or width or (ii) a linear function of systematic variability with length or width or (iii) no systematic variability.

Dr Carroll did not agree with this approach. He considered that the correct approach would be to assess all the data in order to detect any source of variability. Prof Sasieni’s response was to reiterate that his approach was a conservative one and that, in any event, the intra-batch variability was relatively small and so there would be little difference in the results whichever approach was applied. I would add that, on the evidence, I see little realistic likelihood of any form of systematic variation other than the two possibilities he allowed for anyway. Accordingly, I do not accept this criticism of the protocol.

The second criticism goes back to Dr Carroll’s view that a prediction interval was more appropriate than a tolerance interval and the (as I understand it) related disagreement as to the method of calculation which I touched on in paragraph 191 above. So far as this point is concerned, I found Prof Sasieni’s evidence more persuasive; but in any event, the difference between the two approaches is unlikely to be material, as can be seen from paragraph 191.

Finally, I must address the fact that Dr Reddy’s protocol depends upon the Court’s view as to what proportion of infringing patches amounts to de minimis. This is because it involves using the equation derived by Prof Sasieni to determine the maximum level of levulinic acid and minimum level of oleyl oleate for each batch for a given level of confidence (1-α) and a given number of N specified by the Court. I have already concluded that the appropriate level of confidence is 50%, reflecting the balance of probabilities. N represents the number of patches such that no more than 1 in N patches falls within the claim. In other words, 1 in N represents a threshold for the application of the de minimis principle. In my judgment the appropriate value of N is 10,000.

I therefore conclude that the testing regime and statistical protocol proposed by Dr Reddy’s is suitable, with those values of (1-α) and N, to ensure that the patches marketed by Dr Reddy’s in the UK do not fall within the claim other than to a de minimis extent. Accordingly, there is no threat by Dr Reddy’s to infringe the Patent.

Neither Sandoz’s nor Dr Reddy’s product will fall within the claim, and accordingly there is no threat by either Defendant to infringe the Patent.