Diabetes developments – by Simon O’Neill
In a regular blog series, Simon O’Neill, Diabetes UK’s Director of Health Intelligence and Professional Liaison, rounds up the latest diabetes research. This week he focuses on the latest technologies, medicines and treatments.
Encapsulated Beta Cells
A company called ViaCyte have pioneered the potential treatment of Type 1 Diabetes using lab-grown pancreas cells, derived from human embryonic stem cells which can be cultivated in the lab and have the ability to differentiate into any cell in the body. The research is funded by JDRF.
The protocol to implant islet cells has been around for several years but has not been as successful as originally hoped. Islet cells are in short supply, as they can only be taken from a deceased donor, and consequently patients also have to take immunosuppressive drugs to stop their bodies attacking the new cells. Although they have been effective, especially in preventing severe hypogycaemia and and getting more stable blood sugars, few recipients have remained totally insulin free for more than a few years and most have had to go back on to insulin injections.
This new approach uses stem cells, which overcomes the problem of availability as they can be cultured in the lab to produce as many as are needed. However, there is still the problem of turning them into real, functioning pancreas cells, especially the insulin-secreting beta cells, and the issue of how to evade the individual’s immune system, which will attack any transplanted cell and is particularly geared up to attacking beta cells.
ViaCyte’s solution is a biocompatible, plastic mesh capsule and the use of immature pancreas cells, which are easier to create from stem cells and rely on the body to transform them into beta cells. The capsule’s mesh can screen out the immune system’s killer T cells, which are too big to get through it, but still allows the 40 million transplanted pancreas cells to receive nourishment from the bloodstream, as well as to sense blood glucose and respond.
Initial trials in mice showed that the transplanted cells were able to produce insulin, glucagon and somatostatin (a growth hormone) and were able to regulate blood glucose levels.
The current human trial is mainly designed to test for safety of the procedure but there is also the hope that people may also see some reduction in their need for injected insulin. It isn’t clear how long the implanted cells will survive, but it is certain that people would need to have implants replaced periodically.
At least two other groups are also seeing outcomes in rodent studies using immature and mature islet cells in a protective environment and this may well be the way forward for a longer term treatment for Type 1 Diabetes.
Intensive Insulin Treatment in Type 2 Diabetes?
T2D develops when the insulin produced is not able to function effectively, eventually leading to a fall in the production of insulin as the beta cells fail. This failure is directly affected by high glucose levels themselves – glucotoxicity – and it is known that if you improve glycemic control you protect beta cell function.
Studies carried out as long ago as 2008 demonstrated that early intensive insulin therapy in patients with newly diagnosed Type 2 Diabetes can have favourable outcomes on the recovery and maintenance of beta-cell function and long term improved glycaemic control compared with treatment with oral hypoglycaemic agents alone.
The OpT2mise trial therefore looked at the use of pump therapy (CSII) compared with multiple daily injections of insulin (MDI) to see if the use of CSII could give even better results in those with poorly controlled, insulin treated T2D on MDI. The study of 331 people saw that those on CSII had a drop in HbA1c from 9% to 7.9% – much better than those on MDI, who only lowered HbA1c to 8.6%. Those using CSII also had no greater time in hypo and used less insulin than those on MDI. The patients that did best had the highest HbA1c to start with. Also, to improve adherence to CSII, the trial recommended the use of a behavioural contract which includes each person’s agreement to fulfil a detailed list of responsibilities.
So, rather than just transferring people with poorly controlled Type 2 Diabetes on to insulin, should we consider using CSII in this population as well?
Next steps with the Artificial Pancreas
As is usual at this meeting there were many presentations on the journey towards the artificial pancreas (AP).
To develop a real AP would be to technically produce something that is the equivalent of a human pancreas working normally. In people without diabetes the pancreas acts as a glucose sensor, closely monitoring any rise or fall in blood glucose; it then produces both insulin and glucagon, which work in harmony to keep blood glucose levels within very tight limits regardless of what we eat or the activity we undertake.
The AP uses already available components, such as an insulin pump and CGM blood glucose monitoring, and then applies an algorithm to this CGM data to increase or decrease the insulin given, to try and mimic the action of a real pancreas. However, although this is beginning to work in practice, there are still issues to overcome before we have a perfect system.
Potential issues include the time and duration of action of insulin once administered – even the fastest analogue insulins can still be working several hours after they are given; glucagon is currently unstable in liquid form (though several companies are working to try and overcome this) so it isn’t currently a part of many AP approaches; CGM, although good and getting better in terms of accuracy, still lags behind blood glucose results and sensors still need to be regularly replaced and often calibrated twice a day. Most AP solutions are now having quite impressive results overnight – but are still struggling to get control during the day, when we are eating and being active and so blood glucose levels are fluctuating more.
Despite this, over 40 studies have been published since 2010 and the move towards AP is definitely in the right direction, although is still often in the research lab (though spreading out toward real life situations more and more).
The earliest commercial successes have been with Sensor Augmented Pumps. These are a pump and CGM combination which initially were designed to simply stop administering insulin if CGM readings fell below a set level, to prevent hypos, particularly overnight. In 2015 Medtronic brought out their Minimed 640G with SmartGuard which enables the user to set a target level at which the pump will be stopped, in order to prevent going on to hypo. SmartGuard does this using data from the CGM and an algorithm which helps it predict when blood glucose levels are falling and where they are likely to fall to, based on insulin on board. It then restarts the insulin infusion as soon as it can, once the person is back in their ideal range. Of the first 40 patients using this technology, they managed to avoid hypos over 80% of the time with less rebound hyperglycaemia as well.
Interestingly people still chose to turn off this automatic suspend for some of the time when wearing the system. It isn’t clear why, but it may be due to ‘alarm fatigue’, where you are constantly being alerted that blood glucose levels are falling. In further studies on 4919 users it was found that glucose returned to normal within a 2 hour period in almost 90% of cases. In addition 73.9% never reached the pre-set low limit during the day – and 76.2% during the night and two thirds of people were happy to let the system automatically correct their glucose levels and didn’t intervene themselves at all.
The latest trials are on the next version of this system, the Minimed 670G, using the system day and night to manage blood glucose, with the exception of meal times, when the user will predict their carbohydrate intake and bolus insulin accordingly. This is being described as a hybrid closed loop system. The algorithm that makes this possible has been tested in real world situations, in terms of exercise, unannounced meals, false sensor calibrations, lost transmission between the CGM and pump etc and although not perfect, does appear to manage those issues safely. An initial pivotal study is being finalized in March on 120 people between the ages of 14-75 and a larger study on 1000 people for a year is being planned. Interestingly the FDA have allowed people on the trial to continue using the system off trial, which suggests that they may be open to approval of a hybrid closed loop system and this could be the first form of AP to market, perhaps as soon as 2018.
However, that doesn’t mean that work on a complete closed loop AP isn’t still the dream and target of many but this raises other issues. AP has really improved. It used to be a wired system – now it is wireless; it used to run the algorithm on a lap top – but now often uses a Smart phone; it used to only be used in a clinical system with the clinical team closely involved – but now there is remote monitoring; it was originally limited to night time use – but now it is being used in 24/7 external situations – even if still requiring user input during the day to achieve the best outcomes.
So what needs to happen next? We are seeing consolidation of devices, such as CGM combined with a pump but can we also embed the control algorithm in the same device? Medtronic and Animas are working toward that – perhaps with an app for additional use – but will that make the device rather unwieldy? Can we really have a closed loop AP with only insulin or do we need a dual hormone approach? Can we really get to a fully closed loop system, where everything is automated, or will people still have to deal with carb counting and boluses for food? Can we really get much better outcomes compared with Sensor Augmented Pumps?
And, perhaps most importantly, there is also the issue that is regularly discussed at the ATTD meeting – what do people with diabetes really want and what will they feel safe with? Although there is very positive feedback from users of these systems in trials, these are often keen, early adopters of new technology and we don’t yet know how many people in the real world would want to take up such advances or whether they would use them all the time (most people using CGM use it less than 80% of the time).
There is then the even bigger question – can health services afford an AP? Currently only about 5 countries in Europe routinely reimburse CGM. Many countries limit it to certain groups of people on a case by case assessment. For example, in France, discussions around different CGM systems may limit them to people who have had one episode of severe hypoglycaemia in the last 12 months – which would be about 3% of the Type 1 population. In the UK, even if you currently meet the new NICE criteria for both insulin pumps and CGM use, it is still often very difficult to get funding for both. In the UK just a pump and CGM system cost about £4,500 pa to run. If 10% of people with T1D used an insulin pump and CGM system it could cost £150M a year.
IBM Watson and the Internet of Me – can we get better at predicting hypos?
With a move to modern technology, we are all, every day, creating vast amounts of data about ourselves. Whether this is information about our search preferences online or information stored by our FitBit or other apps and technology, the quantity is only going to grow and grow. Some people are seeing this as having an enormous potential for innovation and have called it “The Internet of Me”
On the technology front we have moved from Tabulating Systems (such as punch card systems) in the early 20th Century to the programmable systems which we use today, but IBM believes that the next step will be “Cognitive Computing”, dealing with unstructured data. In this scenario a machine continually learns from continuous interaction with data and how humans react to that data, using natural language processing. This might seem futuristic, but IBM Watson (their super computer) used this way of interacting to learn and to beat two former champions at the game of Jeopardy! (The show features a quiz competition in which contestants are presented with general knowledge clues in the form of answers, and must phrase their responses in the form of questions.)
By 2017 the amount of data in the world of medicine will almost double and much of it could impact on our health. 10% is clinical information (that is held on records); 30% will be genomic factors and 60% will be exogenous factors (lifestyle, information from personal tests, clinical imaging, social care factors, FitBit information etc) Therefore any computer system will have to be able to learn at scale; they will need to be able to reason with a purpose; and they need to interact with humans naturally.
IBM Watson health uses things like predictive modelling from observing patterns across a patient population, so could it be used to predict a hypo after a bolus dose of insulin two to four hours ahead of time?
Programmers have taken 3-9 months’ worth of data from 2000 users of CareLink, which is a secure, online therapy management software that downloads information from Medtronic insulin pumps and CGM systems and then creates reports, allowing the user to track patterns of their diabetes management and therapy. IBM Watson looks at demographic data; both short term trends, such as sensor glucose values and carbs eaten; and long term behaviours, such as pump settings and the frequency of blood glucose trends. IBM Watson then takes that data and looks to see trends, clustering events to provide insights into the statistical relationships between user behaviour and outcomes.
The system was ‘trained’ on 80 % of the available data, so that it could start to learn these trends. It was then tested against the remaining 20% of data and tried to predict hypo events 2 and 4 hours after any given bolus dose. It split patients into 10 groups such as children; people in their early 30’s with adult onset, older people etc – and managed to get over 80% prediction accuracy for hypos in each group.
Obviously this is only at a population level so the next step is to see whether you can break that down to smaller and smaller specific groups until you can start to predict at an individual level. If this was possible, a device – once you had given it the various necessary parameters – could feasibly suggest what a patient could do to achieve improved glycaemic control. They even think it could advise that a hypo might happen in 4 hours’ time, because of current actions taken, so that the individual could think about what actions to take (whether just monitoring for the next few hours or to take action now).
Time for ‘Time in Target’?
We are all familiar with the use of HbA1c as the marker of long term glucose control. Since the Diabetes Control and Complications Trial reported in 1993, showing that improved HbA1c lowered the risk of microvascular complications in T1D, the HbA1c has been the test of choice. However, although a good biomarker, it does have its flaws. This is particularly true for people whose blood glucose roller-coasters between highs and lows, who can end up having a similar HbA1c to someone who is keeping in target range most of the time. A great HbA1c in clinic doesn’t actually tell anyone what is going on with your diabetes in the day to day – such as severe hypos, depression, family issues and quality of life. So is it time to move away from HbA1c and consider Time In Target instead?
One problem with moving away from HbA1c is that that is the marker used in the vast majority of studies looking at how we reduce the risk of complications, so it is a useful tool, but perhaps it could be combined with other measures to give a more rounded view. One suggestion has been to also add Time In Target – but how do you define what that target is? What is the ideal range to both prevent complications and to provide the best quality of life? And how much time in target do you need to get good control? Should we also be widening this to Quality of Life or Diabetes Burden to get a full picture? And who is this information for – the individual with diabetes, the health care professional, the people paying for healthcare? There are still a lot of questions to answer.
One issue considered at the meeting was how we define hypoglycaemia. Currently we class hypos as severe or mild – severe meaning that someone else has had to intervene to stop the hypo. However, in practice we know that different people experience hypos differently and would it be more helpful to look at the levels of hypo people are hitting – 3.9 (low), 3.3 (very low) and 2.8 (dangerously low) or, indeed, the amount of time each day that people spend in one of these hypo states (which, of course, would require CGM for most people).
It may be that we need a new composite score to assess overall diabetes management. One that was proposed in 2008 was the Glucose Pentagram by Andreas Thomas (Diabetes Technology & Therapeutics : 07/2009; 11(6):399-409). This had five axes – including HbA1c, Mean Glucose, and length of time over 8.9 mmol/l, and produced a result that showed someone’s glycaemic range during one day compared against someone without diabetes. It also highlighted problem areas that may need to be addressed. Could you take something like this and add other factors, such as time spent in hypo or Quality of Life, to come up with a composite number that reflects more aspects of a person’s life with diabetes? Although this might be seen as just another number like HbA1c, it would give more information about the day to day and may be a better clinical marker, allowing HCPs and people with diabetes to concentrate on problem areas that might need better input or support.
More research obviously needs to be done to solve these issues.
Abbott Freestyle Libre Update
Two new developments with the Abbott Freestyle Libre. Since its launch in late 2014, the company had struggled to keep up with production to meet need. Those who registered early, and were the first users, were always able to get stock – but, in order to enable this to happen, new customers had to go on to a waiting list, often for several months.
Production has now been stepped up considerably to meet need and now people registering on the site for the first time, can immediately make an order without having to wait.
In addition the device has now also obtained a CE Mark for its use in children and young people with diabetes aged 4-17 years old. This comes on completion of an accuracy study, which demonstrated that the system was clinically proven to be accurate, stable and consistent for up to 14 days without the need for finger prick calibration, for this age group. The study also showed that more than 97 percent of these CYPs found it easier to use than finger prick testing.