Abbott Diabetes Care Inc & Ors v Dexcom Inc & Ors

This is a patent case about continuous glucose monitoring (“CGM”) technology, which plays an important role in the management of type 1 diabetes. The appellants, Abbott, and the respondents, Dexcom, are major players in the CGM market. Until recently they were suing each other in patent cases in several jurisdictions. In the UK, the dispute was case managed into a series of technical trials. This appeal relates to the first one.

Abbott alleged that three of Dexcom's systems infringed two of its patents: European Patent (UK) 2 146 627 (‘EP 627’), relating to aspects of the user interface; and European Patent (UK) 2 476 223 (‘EP 223’), concerning the validation of safety-critical software on devices such as smartphones. Conversely Dexcom alleged two of its patents were infringed by Abbott's Freestyle Libre 2 and 3 systems, the dispute there focussing on the way users are provided with alarms when high or low blood glucose is detected. Each side also counterclaimed for revocation of the other’s patents for invalidity.

The trial took place over six days in November-December 2022 before Mellor J. In a judgment delivered over a year later on 15 January 2024, all four patents were held invalid. The consequential order was made another year later, on 13 January 2025. Abbott sought permission to appeal the order revoking EP 627 and EP 223, which Arnold LJ granted on 25 March 2025. For each of those two patents, the judge had found the patent invalid for lack of novelty. For each patent that is the sole issue on this appeal.

The overall dispute between Abbott and Dexcom has now settled and so Dexcom have not opposed the appeal. Accordingly, consistent with the procedure laid down in PD52D, para 14.1, the Comptroller-General of Patents, Designs and Trade Marks (‘the Comptroller’) was asked to attend the hearing and assist the Court.

Section 1(1) of the Patents Act 1977 provides as follows

1.

Patentable inventions

(1)

A patent may be granted only for an invention in respect of which the following conditions are satisfied, that is to say—

(a)

the invention is new;

(b)

it involves an inventive step;

(c)

it is capable of industrial application;

(d)

the grant of a patent is not excluded by subsections (2) and (3) or section 4A below;

and references in this Act to a patentable invention shall be construed accordingly.

This section sets out the requirements which must be satisfied by the invention for a patent to be valid. The requirements for novelty and inventive step are defined further in sections 2 and 3, as follows:

2.

Novelty

(1)

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

(2)

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

[ … ]

3.

Inventive Step

(1)

An invention shall be taken to involve an inventive step if it is not obvious to a person skilled in the art having regard to any matter which forms part of the state of the art by virtue only of section 2(2) above […]

These provisions correspond to Art 52, 54 and 56 of the European Patent Convention (EPC). The provisions on novelty provide that in order to be patentable an invention must be new, and an invention is new if it does not form part of the state of the art. The rest of s2 explains what is meant by the state of the art. The only relevant provision in this case is s2(2) which provides that the state of the art comprises all matter which has at any time before the priority date of an invention been made available to the public. In UK law the term anticipation is used to refer to a situation in which an invention lacks novelty.

In order to apply these provisions one needs to know what the invention is and this is where s125(1) of the 1977 Act comes in. It provides that for the purposes of the Act an invention is that which is claimed in a patent claim, as interpreted by the description and any drawings. This provision corresponds to Art 69 EPC and, by s125(2) of the 1977 Act, an instrument called the Protocol on the Interpretation of Art 69 EPC applies for the purposes of s125(1) as it does to Art 69. The essence of the modern approach to the construction of patents is set out by the Supreme Court in Actavis v Eli Lilly [2017] UKSC 48. Briefly, the Protocol represents a compromise between approaches to construction which might have been seen as either too literal or too liberal. The correct approach, which is in between those extremes, is one from the perspective of the person skilled in the art imbued with the common general knowledge of such a person at the priority date and combines a fair protection for the patentee with reasonable certainty for third parties.

A more detailed overview of the law on patent construction is in Saab Seaeye v Atlas Elektronik [2017] EWCA Civ 2175 at [18]-[19], where Floyd LJ, in light of Actavis v Lilly, updated the list of principles elaborated by Jacob LJ in Virgin Atlantic v Premium Aircraft [2010] RPC 8. Counsel for the Comptroller provided a helpful summary in his skeleton argument, on which the following is based. In short, the construction exercise must be:

It follows that where a claim uses general non-exhaustive language, it is not normally permissible to read that language ‘down’ so as to construe the patent narrowly if the patentee has chosen not to do so (Nokia v IPCom [2009] EWHC 3482 (Pat) at [41], per Floyd J). Conversely, where a claim uses specific exhaustive language, it is equally inappropriate to read that language ‘up’ so as to construe the patent broadly (Fisher and Paykel Healthcare Ltd v Flexicare Medical Ltd [2020] EWHC 3282 (Pat) at [150], per Meade J). The correct approach is simply to interpret the claim as the objective skilled person would understand the author to have intended, i.e., no more or less restrictedly than is necessary to give effect to his purpose.

Coming back to novelty under s2 of the 1977 Act, the relevant question is: “Was the claimed invention anticipated by the state of the art?” As Lord Hoffmann made clear in Synthon BV v SmithKline Beecham plc [2005] UKHL 59, [2006] RPC 10 at [19]-[33], this question may only be answered in the affirmative where two requirements are met. The first is that a prior art document discloses clear and unmistakeable directions to do something covered by the claim. This is usually referred to as “disclosure” and in the metaphorical language of General Tire & Rubber Co v Firestone Tyre & Rubber Co Ltd [1971] FSR 417 at 444) it occurs when the prior document ‘plants the flag’ on the invention before the patentee does. The second requirement in Synthon is that the skilled person reading the document would be able to perform the claimed invention if he attempted to do so by using the matter disclosed, read and understood together with his common general knowledge. This is called “enablement”. So together, an item of prior art is said to anticipate if it makes an enabling disclosure of the invention.

The appeal in this case relates to disclosure, enablement is not in issue. Disclosure may be explicit but there are also two senses in which, even though the disclosure is not explicit, the prior art has nevertheless anticipated the invention. One case is where there is some unstated element in the prior disclosure which is nevertheless necessarily and unambiguously implicit in what has been stated. The other case is about circumstances in which carrying out the directions in the prior art will inevitably result in something falling within the patent claim, even though the instructions did not say so. These two cases may overlap but they are not identical. In both cases a skilled person given the prior art would necessarily make or do something which falls within the claim (Synthon at [19]-[22], applying General Tire at 443-444) but for somewhat different reasons.

Crucially, either kind of non-explicit disclosure is a strict test. It will not do that the document contains a direction which it is possible or obvious to carry out in a manner which falls within the claim—the invention must necessarily follow (ModernaTX Inc v Pfizer Ltd and ors [2025] EWCA Civ 1030 at [40], per Arnold LJ). Evidence that something is possible or obvious may have a role to play in a case about inventive step under s3 of the 1977 Act, but it is not relevant to novelty under s2.

One theme which is raised on this appeal is the proper role of expert evidence in determining implicit disclosure. To be clear, as Arnold LJ noted in Moderna at [58], the true construction of a prior art document is a question of law “once the Court has been properly instructed by the expert evidence as to the common general knowledge with which the skilled person reads the document, and hence the technical considerations which bear upon its interpretation”. In other words, while it is for the Court to determine what the document ultimately discloses by applying the correct legal standard, before it does so, it must “don the mantle” of the skilled person (Technip France SA’s Patent [2004] EWCA Civ 381, [2004] RPC 46 at [6], per Jacob LJ).

The point is that expert evidence informs, but does not usurp, the Court’s role in construing of the document. That is why, as is well known, it is inappropriate to invite a witness to pontificate on the meaning of ordinary English words or to place reliance on any such evidence. By contrast it is appropriate and usually necessary to seek evidence on the skilled person’s attributes, including what they think, know, and are likely to glean from the document.

An appellate court will defer to a trial judge’s finding of fact or evaluative determination unless there is a clear error of approach or the conclusion is rationally insupportable (FAGE UK Ltd v Chobani UK Ltd [2014] EWCA Civ 5, [2014] ETMR 26 at [114], applied in Byers v Saudi National Bank [2022] EWCA Civ 43, [2022] 4 WLR 22 at [99]). However the inordinate delay in producing the judgment means that the judge’s findings must be reviewed with special care, as Arnold LJ observed in his reasons for granting permission in the present case.

Diabetes is a disease in which the body’s ability to produce or respond to insulin is impaired, resulting in abnormal metabolism of carbohydrates and elevated levels of glucose in the blood. Many diabetics rely on self-administered insulin injections to manage their blood glucose levels.

In recent years, CGM devices have been developed, which enable patients to monitor their blood glucose levels continually with a view to avoiding adverse glycaemic events—either hypoglycaemia (when their blood glucose levels become too low) or hyperglycaemia (when they become too high), both of which can be life-threatening if untreated. As the judge noted at [26], the primary objective of insulin therapy is in effect to walk a tightrope constantly between these two states. CGM devices are therefore of importance in supporting patients to help them manage their condition and ensure they remain within a safe ‘target’ range.

At the priority dates (14 April 2007 for EP 627 and 8 September 2009 for EP 223) a typical CGM device consisted of an on-body receiver unit, which could be affixed to an item of clothing, and a subcutaneous analyte sensor and transmitter, which communicated wirelessly with the receiver unit. The unit would output blood glucose readings and provide audible and/or visual alarms and alerts. An illustration from the judgment at [46] is this:

Common to all CGM devices at the priority dates was some form of alert or alarm function, which notified users of an impending hypo- or hyperglycaemic event. Also common at the priority dates was a desire to enable patients to use their smartphone for blood glucose readings and to receive alarms and alerts, although this desire appears largely to have been unrealised at the time ([248]-[249]).

EP 627 concerns a method for detecting and notifying a user of what is referred to as a “predefined alarm condition” associated with an analyte monitoring device during and after the execution of a so-called ‘predetermined routine’ on the receiver unit. A non-limiting example of a predefined alarm condition would be a monitor reading indicating an impending hypo- or hyperglycaemic event.

The background to this is the fact that, in addition to their primary function of monitoring a user’s blood glucose, CGM devices also support other ancillary functions. The invention is intended to allow a user to be notified of a detected predefined alarm condition even though they are executing one of these ancillary functions on the receiver unit. The term “predetermined routine” is used in the patent. It encompasses these ancillary functions while not being limited in that way. Examples of what a predetermined routine might be are given at [0080] of the patent. They include performing a finger stick blood glucose test for periodically calibrating the sensor unit, configuration of device settings, review of historical data, data communication management, data transfer to a terminal, or viewing alarm conditions with a different priority in a preprogrammed hierarchy.

Claim 1 of the patent (broken down for convenience) is as follows:

Analysing a patent claim by breaking it down into its constituent integers in this way is a conventional and necessary exercise. For one thing it makes it easier to refer to particular aspects of the wording. Nevertheless it is also a truism that patent claims, like any other text, need to be read and understood as a whole. A trap inherent in breaking down a claim into its component parts in this way is that one can sometimes fail to see the claim as a whole, missing the wood by focusing on the trees.

In the claim the expression “predetermined routine” includes any process which:

Similarly, the term “predefined alarm condition” encompasses any condition which:

As the judge noted at [170], the above is little more than a restatement of the express wording of the claim. However it is useful to set the scene for the issues to be addressed.

The prior art which was found to deprive EP 627 of novelty was called the STS Guide. It is a user guide for a Dexcom CGM device. There is no dispute that the guide was available to the public before the priority date. The sensor for this device had a finite working life, in this case 72 hours (i.e. 3 days). After that period the user had to replace it with a new sensor. The device provided the user with a kind of countdown to the expiry of the sensor. This is explained at page 65, which sets out the process whereby the user is notified of impending sensor expiry:

As this explains, the user receives four notifications. The first three are prospective (in the sense of alerting the user to the fact that the sensor is going to expire in the future) and the last of which is retrospective (alerting them to the fact that the sensor has now expired). On screen the notifications are in the form of an hourglass. The four notifications are:

As the guide notes, the notifications can be dismissed by the user pressing any button on the device.

Novelty as compared to the STS Guide is dealt with in the judgment at [203] to [215], ending with a conclusion that what was disclosed in the guide anticipated EP 627. The conclusion is based on the following essential findings:

Amongst other things Abbott had submitted (i) that the 3 day sensor expiry session was not a predetermined routine, and (ii) that the alerts had to relate to the same predetermined alarm condition whereas in the scheme in the STS Guide the two indications related to different conditions. These arguments failed at [212] and [214] respectively.

On appeal Abbott submits that:

As put in the grounds of appeal, these two points were the other way around, the first being ground 1(b) and the second ground 1(a). I will deal with ground 1(b) first because it is more convenient to do so.

The anticipation case is based on the proposition that looking at the process described in the guide, the countdown routine is the predetermined routine while at the same time the two indications arise from that self-same countdown routine. The proposition strikes me intuitively as a surprising result, given that the patent at first sight is effectively about messages which pop up without warning, caused by an alert arising from the detection of something such as the risk of hypo- or hyperglycaemia, in the middle of a different process, such as a process configuring the device settings. The patent claim may have been drawn so widely as to cover such a situation but that would be a long way from what the patent is about.

To start with I will say that I cannot see how it makes any difference whether one characterises the alleged predetermined routine as the countdown or as the 3 day session itself. They are the same thing.

Abbott submits that expiry of the three day sensor is not a predetermined routine at all, for two reasons. The first is that, by integer 1.6, a predetermined routine must be something that is capable of interruption by the provision of the first indication. Therefore since this process simply measures the passage of time, it is not a routine capable of interruption in the sense contemplated by the claim. Time continues to run even though a notification is displayed to the user. The second point is that in any case this routine does not ‘include one or more processes that interface with the user interface of the receiver unit’, as required by integer 1.9. The notifications themselves cannot be both (a) the processes which are part of the predetermined routine that interface with the user interface required by 1.9 and also (b) the indications separately required by 1.3 - 1.6. That would be circular and wrong.

The Comptroller’s main argument in relation to this ground is the importance of taking account of the broad language of the claim and the need (e.g. Nokia v IPCom) to avoid reading limitations from the examples into the claim language. The patent at [0080] gives non-exhaustive examples of what a predetermined routine could be and there is no basis for artificially restricting the language in this way. Counsel also observes that while the flow of time itself cannot be interrupted, the mechanism for telling the time can (a clock can be stopped). In any case, the language of 1.6 does not actually require that the predetermined routine be capable of interruption, only that it was not in fact interrupted by the first indication, which the countdown was not. Finally, counsel contends that the countdown mechanism does interact with the user interface via the alarms and, since there was no interruption, this does not involve any circularity.

While I fully accept the imperative to avoid artificially reading in limits to wide words of the patentee’s own choosing, nevertheless in my judgment, the routine counting down to sensor expiry is not what the claim means by a predetermined routine. Although the Comptroller’s counsel is correct that the language of 1.6 does not explicitly state that the predetermined routine has to be capable of interruption, considering the purpose of that feature reveals that the claim language presupposes that the indications arise from something which would be capable of interrupting the predetermined routine, absent a requirement not to do it. The question whether a predetermined routine can be said either to be capable of interrupting itself or not to be capable of doing that, does not make sense either way. Consider a routine consisting of steps A then B then C then D. It is unreal to ask of step B, which is an integral part of the routine, whether the routine is executed without interruption by step B. In my judgment a person skilled in the art reading this claim as a whole and in the context of the patent as a whole, would see that the indications referred to must be caused by something distinct from the predetermined routine. Only if the indication is caused by something distinct from the predetermined routine, can it then be meaningful to contemplate that it could interrupt the predetermined routine, and therefore also be meaningful to require it not to do so.

The other point is about integer 1.9 which requires that the predetermined routine has a process which interfaces with the user interface. Again, reading the claim as a whole, it is apparent that as a matter of drafting the requirement for a process which interfaces with the user interface at integer 1.9 was intended by the author to be a different thing from the references at 1.3 and 1.4 to outputting indications to the user interface. If indications satisfying 1.3 and 1.4 also satisfied 1.9, then the latter requirement would be surplusage. Sometimes language is truly redundant but in this case it is an indicator that something has gone awry with the suggested interpretation. While the claim could have been better drafted, in order to make sense of it a person skilled in the art would see that the indications are simply not intended to be aspects of the predetermined routine. In other words the one or more processes referred to in integer 1.9 which must interface with the user interface cannot be the indications.

Standing back, the position is fairly simple. The STS Guide describes a singular process with alerts built into it. To read the claim onto that process contorts the claim out of shape in a manner which no skilled person would think the author had intended. I would allow the appeal on ground 1(b).

The issue here is about whether the two indications relate to the same predefined alarm condition as required by integers 1.3 and 1.4. Abbott contends that since the 30-minute notification relates to the time remaining until sensor expiry, whereas the 0-hour notification relates to the fact of the sensor having expired, they do not.

The Comptroller concedes that the two notifications are not identical in a literal sense, but submits that they nevertheless relate to the same unchanging condition—namely, sensor expiry at a particular point in time. The point is also made that integer 1.4 merely requires that the alert be ‘associated with’ the predefined alarm condition. Both the 30-minute and 0-hour notifications, the Comptroller suggests, are ‘associated with’ sensor expiry.

The first point to make, which is not in dispute, is that the claim does refer to a single “detected predefined alarm condition”. Patents are construed purposefully and the purpose of the invention is clear. The idea is to have a scheme which operates when the user is doing something (such as reviewing old data). This is the predetermined routine. The system then provides the user with a temporary indication that the device has detected a possible impending important event – such as a hypo- or hyperglycaemic event. This is the predefined alarm condition. The purpose of the fact that the first indication must include a temporary indication, is that this first indication is in effect a nudge which does not interrupt the predetermined routine. Then later, when the predetermined routine is over, the scheme provides a second indication of the same detected occurrence, which includes an alarm (1.8).

There is no factual dispute about how the process described in the STS Guide works or what happens. The issue is whether that process is something falling within the claim, properly construed. In my judgment is it not. I accept that the actual expiration of the sensor – which triggers the second notification, output after completion of the routine – is a condition. I also accept that the fact that there are 30 minutes to go before the sensor expires is a condition too, but these two are distinct conditions; and the claim would not be satisfied based on them. The judgment at [214] holds that the condition is “sensor expiry” but I fail to see how that can be said to be a condition at all. I would hold the two notifications do not relate to the same condition and allow the appeal on ground 1(a).

EP 223 relates to a method for hosting a safety critical application on an uncontrolled data processing device (“UDPD”), such as a smartphone. At the priority date, both Abbott and Dexcom harboured ambitions to enable patients to use their smartphone for blood glucose readings and to receive alarms and alerts. One obstacle had been that whereas the manufacturer of a CGM system could control both the hardware and software of a separate on-body unit to ensure it functioned properly, with the result that it could be relied on by users to guide their self-treatment, the manufacturer could not control the workings of a smartphone, particularly where such a device was configured to permit users to make their own software and/or hardware changes. Consequently, there was a risk that the smartphone could cause the manufacturer’s software to malfunction, with potentially catastrophic consequences for the user.

The invention seeks to solve this problem by performing a series of checks to ensure that a safety critical application, once loaded onto a UDPD, has been both installed and is functioning correctly. If either of the checks fail, the invention would disable safety critical features while continuing to enable non-safety critical features. Abbott describes this as ‘selective enablement’.

Claim 1, broken down into integers at [336], is in this form:

Step 1.2(a) provides for the checking of installation and functioning of the safety critical application on the UDPD. Then 1.2(b) to 1.2(d) specify what happens if there is a problem with the installation or functioning. In that case selective enablement takes place with the safety critical features disabled but the non-safety critical features enabled. By contrast 1.2(e) specifies what happens if the testing is satisfactory. The safety critical application operates free of restrictions.

Features 1.3 to 1.4 essentially limit the generality of the claim to a method where the UDPD has a display, the safety critical application is a medical application, and the device (which could be a phone) communicates with an analyte monitoring device (which could be a CGM sensor) and/or a drug administration device.

The prior art which was found to anticipate EP 223 is a US Patent Application called Gejdos. It relates to a database incorporated into a healthcare management software system used by healthcare professionals to review patient data, such as historic blood glucose levels. It might be used by healthcare professionals to review patient data such as diaries of blood glucose values (e.g. [0002] and [0028] of Gejdos). One of the problems with such a system is that data may be uploaded into it from different sources and in different units. For example, as the judgment explains at [28], the units used to measure blood glucose values are either mg/dL or mmol/L. The unit used usually depends on the country. While the preferred unit in the UK is mmol/L, in the US it is mg/dL. One can therefore see why it would be worthwhile to be sure that measurements were stored using the correct units ([0045] Gejdos).

Gejdos discloses a method for checking the integrity of a patient database by carrying out a database integrity check. This involves checking “integrity information”. One example of this is the units ([0033] of Gejdos). The way this works is that the information in the database is checked against integrity information provided external to the database, which is known to be accurate. The integrity check is performed upon installing or launching the healthcare management software system.

At [429] to [449] the judgment finds Gejdos anticipates EP 223 on the basis that:

Abbott contends that conclusions are in error for the following reasons:

I will take these in turn.

The reasoning in the judgment is straightforward. At [433] the point is made that an inaccuracy in the database in Gejdos could lead to significantly detrimental consequences for patients. Therefore what is being described there is a safety critical application. The issue on appeal is about the ambit of the term safety critical application. It is notable that the words are ordinary English words. This is not a debate in which the expert evidence had a role to play at trial or on appeal.

Safety critical applications are defined at [0001] of EP 223 as being systems “whose failures or malfunctions may result in significantly detrimental consequences such as death or injury to persons, severe damage or loss to equipment or to environment”. A skilled person reading this, Abbott contends, would understand that for a system to qualify as a safety critical application, its failure or malfunction must directly result in a risk of significantly detrimental consequences. Abbott argues that a healthcare database of the type disclosed by Gejdos is not a safety critical application because its contents will inform a medical professional’s treatment decisions, but does not directly cause them. Therefore any problem with the accuracy of the database, even if such an inaccuracy did ultimately play a part in a medical professional’s decision which led to harm to the patient does not straightforwardly cause such a decision.

The distinction Abbott is seeking to draw is essentially between a system relied on by non-medically trained patients making decisions in real-time about self-administered treatment, where there will necessarily be a high degree of deference to the readings provided by the system (and thus a high degree of ‘criticality’), and one used by trained medical professionals reviewing historic data in a clinical setting to inform non-critical treatment decisions, where there is likely to be more capacity and scope to interrogate inaccuracies (and hence less ‘criticality’). Without this distinction, Abbott submits, the term “safety critical application” would effectively cover all systems which carry even a remote risk of indirect harm, such as a pocket calculator used by a physician to calculate an insulin bolus, whereas this is not what the claim contemplates.

The Comptroller argues that Abbott is wrong to read a requirement for ‘directness’ into the definition of a safety critical application. The claim language includes no such requirement. On the contrary, the language at [0001] is deliberately broad—all that is necessary is that the failure or malfunction of a system ‘may result in’ deleterious consequences. Similarly, at [0003], the patent points to medical systems being ‘an example of safety critical systems that require a certain level of confidence that the system will operate and continue to operate properly’. In other words, there is no requirement that the system be medical or that its failure or malfunction entail harm to a patient. Rather, as [0001] makes clear, the harm may simply be to equipment or the environment.

I agree with the Comptroller. There is no basis for reading a limitation into the degree of safety and/or criticality necessary to satisfy the expression ‘safety critical’. The patent is widely drawn. Although Abbott suggests that a system is only a safety critical application if it is used in isolation, i.e., as the sole piece of evidence guiding treatment decisions, no such limitation is present in the claim.

Accordingly, a healthcare database of the kind described in Gejdos is a safety critical application within the meaning of EP 223 and the appeal in respect of ground 1 is refused.

Abbott’s second ground of appeal relating to EP 223 is that the judge erred in finding that the database integrity check disclosed by Gejdos was (i) an installation check, (ii) a functional check, and (iii) that it could be both an installation check and a functional check within the meaning of the claim (at [438], [441] and [444], respectively).

Abbott contends that this process of checking the integrity of the database contents in Gejdos is a different exercise from checking whether the database itself (or the healthcare management software system as a whole) is fully installed or fully functional. Only the latter will do, Abbott submits. Abbott also argues that although it is conceivable that the integrity of a database may have become compromised due to a failure of installation or of functionality, so that it follows that checking for integrity might, in such circumstances, indirectly reveal an installation or functional problem, nevertheless that is not the same as directly checking for a failed installation / lack of functionality. Finally, it says that there is no basis for holding that a single check can separately satisfy the requirement for both an installation and a functional check. The claim is said to require two distinct processes.

The Comptroller acknowledges that the two checks are conceptually distinct, but submits there is nothing in the claim that precludes a single check from serving as both. Indeed, as the Comptroller points out, the claim does not refer to two distinct ‘checks’ at all. What the claim requires is that the invention determine that a safety critical application ‘is installed properly and functions properly’ on the device (integer 1.2(a)). The claim is general in its wording, so that a check that the safety critical application functions properly, which would be failed if it had not installed properly, would satisfy the claim.

As to whether the integrity check disclosed by Gejdos is a functional check, the Comptroller submits that since the healthcare management software system will not function properly if the database has lost integrity (as the judgment finds at [440]), checking for the latter is one way of checking for the former. The Comptroller also observes that the language at [0043] of Gejdos states in terms that one way of indicating that the database integrity test has successfully verified database integrity is the launching of the healthcare management software system itself, which would not happen if installation had not been successful.

In my judgment, the judge was entitled to find that the database integrity check disclosed by Gejdos satisfied both integers 1.2(a) and (b) of the claim. I cannot see any basis for Abbott’s assertion that the integrity check does not inform the user about the installation or functionality of the system. There is nothing in the claim to suggest that the check must do so directly rather than indirectly. Equally, the skilled person reading the claim would not, in my view, understand it to mean that the determination required by integer 1.2(a) must be of every last aspect of functionality or that every individual component has installed correctly. Rather, they would see that all that is required for the process envisaged by 1.2(a) and (b) to be satisfied is that there is some sort of test which evaluates whether the safety critical application has been installed and is functional.

Accordingly, the database integrity check disclosed by Gejdos was both an installation and a functional check within the meaning of EP 223 and the appeal in respect of ground 2 is refused.

Abbott’s final ground of appeal relating to EP 223 is that the judge erred in holding that Gejdos disclosed ‘selective enablement’, i.e. disabling safety critical features of an SCA while enabling non-safety critical features in accordance with integers 1.2(c) and (d) of the claim.

Starting with the Gejdos document itself, as Abbott submits, Gejdos discloses only one example of what takes place when there is a failed integrity test. The example is in a passage at [0043] of Gejdos. What that passage states is simply that an error message is presented by an output device. Nothing else is said. It is clear therefore that this does not amount to an explicit disclosure of selective enablement. There is no explicit mention in Gejdos of turning off some features and leaving others enabled.

However in his expert’s report Dexcom’s witness Dr Stirbu said this:

“246.

Paragraph [0043] of Gejdos states that the failure of any of the checks described above is indicated to the user with an error message, e.g. “an exemplary indication of a non-verified database integrity test is an error message presented by one of the output devices”. The skilled person would understand that this includes disabling database related functions when the database integrity check fails. These features are safety critical.”

There was no cross-examination on this passage of evidence, no doubt because in a complex case with numerous issues, the focus was elsewhere.

The judgment accurately summarises this evidence at [445], notes evidence from Abbott’s expert Dr Palerm at [446] without comment, notes Abbott’s submission that what was described in Gejdos was “nothing like the selective enablement of the non-safety critical features while disabling safety critical features of an application, as required by claim 1”, and then concludes at [448] that “[o]n this point, I accept Dr Stirbu’s evidence. Accordingly, Gejdos discloses ‘selective enablement’ within integers 1.2(c) and (d).”

The difficulty with this approach is that the judgment does not grapple with the question whether accepting Dr Stirbu’s evidence would be a sufficient basis for a finding of lack of novelty. In a situation in which a document does not state some proposition explicitly (as Gejdos does not) an opinion by an expert that a skilled person would “understand” that proposition from reading the document is most likely based on that the expert’s view that the proposition would be obvious to the skilled person reading the document. While it might be relevant to inventive step, that would no basis for a finding of lack of novelty. The opinion might be based on a view that the proposition was necessarily implicit, but if that is what the expert thought, one might have expected them to say so and say why. Since Dr Stirbu was not cross-examined on this point, there is no evidence either way. In some cases a judge having heard an expert might feel able to draw a conclusion about what the expert meant in order to resolve a conundrum like this, but this judgment does not do that. Nor, given the delay in producing the judgment, would that have necessarily carried significant weight.

Counsel for the Comptroller regarded this ground of appeal as the one which, as he put it, is closest to having real force. The Comptroller’s submission in support of the judgment was that the correct way to look at this is to treat the evidence as relevant to implicit disclosure. Since construction of documents is approached through the eyes of the person skilled in the art, expert evidence is relevant to assist the court in adopting that approach in order to help perform a legal task, which is to interpret the document.

I accept the point made by the Comptroller’s counsel as a general matter, however when looking at what happened here, I do not see how the conclusion of anticipation can stand. For Gejdos to deprive EP 223 of novelty, the invention in the claim must necessarily follow from it (see Moderna above.) The evidence of Dr Stirbu does not demonstrate that and there was no other basis for the conclusion. Accordingly, I would allow the appeal in respect of ground 3.

For the above reasons, the appeal in respect of grounds 1 and 2 relating to EP 223 are dismissed; the appeal in respect of grounds 3 relating to EP 223 and 1(a) and 1(b) relating to EP 627 are allowed.

I agree.

I also agree.