Diabetes is actually five separate diseases, research suggests | Hacker News Reader

Diabetes is actually five separate diseases, research suggests

bbc.co.uk

133 pointssjcsjc56 comments

"We did data-driven cluster analysis (k-means and hierarchical clustering) in patients with newly diagnosed diabetes (n=8980) from the Swedish All New Diabetics in Scania cohort. Clusters were based on six variables (glutamate decarboxylase antibodies, age at diagnosis, BMI, HbA1c, and homoeostatic model assessment 2 estimates of β-cell function and insulin resistance), and were related to prospective data from patient records on development of complications and prescription of medication. Replication was done in three independent cohorts: the Scania Diabetes Registry (n=1466), All New Diabetics in Uppsala (n=844), and Diabetes Registry Vaasa (n=3485). Cox regression and logistic regression were used to compare time to medication, time to reaching the treatment goal, and risk of diabetic complications and genetic associations"

k-means and hierarchical clustering are notoriously prone to producing false positive clusters.

k-means is just garbage, but hierarchical clustering can be good. The problem with hierarchical clustering is that you can produce whatever set of clusters you're biased towards by fiddling with the linkage function and the cut criterion.

I was surprised to learn on reading this that the Type 1 classification always means "autoimmune disease". I was under the impression that the two type classifications were "1: unable to produce enough insulin", and "2: producing what ought to be normal amounts of insulin, but for some reason resistant to it"; naively, those seemed like much more natural classes for classifying a syndrome. Having a finer-grained analysis can't be a bad thing; for example, the paper identifies some increased risks for the individual clusters.

I had in my head that Type 1 was "unable to produce enough insulin, because of autoimmune problems with beta cells". I was surprised to read about Cluster 2 where they don't produce enough insulin and nobody knows why.

I would have also intuitively put Cluster 2 under Type 1 diabetes though, because it's the same effect.

AFAIK in type 1, most of the beta cells get destroyed, and it happens because of autoimmune processes. To me this means that immune-modulating medication or perhaps stem cell injections could help a great deal.

The problem with those are that they're more complicated (even when perfected) than current treatment. Immune modulation would require either suppressing the entire system (or very large chunks of it) which is done with some other diseases but generally has super high costs, as it leaves you quite vulnerable. Or you're targetting only the handful of cells responsible (my DM I knowledge is rusty but I'm guessing b-cells?) in which case you're on your way to solving a bunch of different diseases.

Stem cells has a similar issue, in that once you made new beta cells your immune system will just react to those.

Current treatment is quite alright. There's ground to be gained in earlier prediction and treatment, and a ton (this should excite people here on HN) in bettering the treatment method—create a proper feedback loop with insulin/glucose readers that automatically push out insulin when needed instead of a huge shot of it when you're guessing you'll need it.

My doctor wants me to start thinking about a closed loop system on this year so I don't need to wake up every night for CGM alarms. One interesting project would be the open source pancreas, I've heard lots of good about it, of you have the right insulin pump and CGM.

https://openaps.org/

Until there’s a dual (glucagon, insulin) pump released, you’ll still be waking up to deal with those low BG alarms. You may be dealing with fewer alarms because of the predictive suspend feature, sure.

Our 6 year old has had a Medtronic 530g which (you probably already know) suspends basal delivery when it reaches a low threshold. I’m pretty sure it’s saved his life a few times (at night) over the past 4 years. He’s due for a new pump and we were hoping the Beta Bionics dusk chambered pump would be ready by this time, but sadly it still looks to be a couple of years away. The Medtronic 670g is likely his next pump, which will add predictive suspend and IIRC will also auto increase his basal rates to account for highs.

OpenAPS looks interesting but for now I’m not willing to experiment with it on our son.

My doctor's colleague in US uses the 670G and when they met he showed her how now every night he has a straight 6.5 mmol/l line and no wake-ups anymore. My need for basal changes quite often and this feature would be nice.

I use xdrip which wakes me up before I got hypo or hyper. I'd prefer to have eight hours of sleep every now and then.

Yeah we use a Medtronic MySentry (basically think of a baby monitor with a screen that displays his BG level...and relays CGM/pump alarms, loud enough to wake the dead.)

xDrip / parakeet looks awesome, but that looks to be Dexcom-only as far as I can tell.

Our son carries a medtronic Minimed Connect (pump<->bluetooth adapter) and an iPhone (yeah a kindergartener with an iPhone) so it can relay his BG when he's in school using Nightscout.

Unfortunately neither the MySentry, nor the Minimed Connect are supported with the 670g so we're a little concerned that we won't be able to monitor him remotely (while he's sleeping or when he's at school...the two times he's most vulnerable.)

I'd love to switch him to a Dexcom, but there are currently no pumps that support predictive / suspend functionality at this time (though the T:Slim X2 is getting close to getting this type of update.)

Thanks for the heads up on xDrip!

You could use https://github.com/pazaan/600SeriesAndroidUploader with your medtronic.

Yeah, I wrote the InfluxDB module to xDrip. Be sure to check the Nightscout version from github. The best tool for my personal diabetes care.

There's an enormous amount of research into both understanding the autoimmune reaction and into safe insulin producing cells that could be introduced.

Turns out they are both really hard problems.

It's a reason to be pretty excited about medicine over the coming years. Doctors are starting to tease apart some of the specifics of the immune system.

Immune-modulating medication maybe, but unless this is identified very early, they won't help. Once the beta cell destruction is complete, you are left without any. Even if the immune system stops attacking, at some point it no longer helps.

In fact, a long time ago some experimental treatment was tried, where active beta cells were transplanted into type 1 diabetes patients. They were very quickly destroyed by the patient's immune system.

Too bad this is just five "clusters" of characteristics that people with not-type-I diabetes can be grouped into. No clear underlying mechanism for the cause of each of the groups diabetes. Maybe more in the journal paper? Or now focused studies can be done on each group to find out the root cause? I hope so.

The quality of this article is rather poor. It's almost beyond belief that they didn't consult with Prof. Andrew Hattersley, who is regarded as one of the world leading experts in this.

We have known for many many years about several other types of diabetes. It is what I research as my job.

Type 1 is caused by the autoimmune destruction of the beta cells in the pancreas that generate insulin, and the patient becomes dependent on injected insulin.

Type 2 is caused by a progression from the body becoming resistant to insulin due to being overweight, the beta cells generating more insulin to compensate, and then eventually the beta cells burning out. Along the way, treatment progresses from diet, to tablets such as metformin, and eventually to insulin injections. Many people think that you progress to type 1 when you become dependent on injected insulin, but that is not the case - it is still type 2.

There is gestational diabetes, where the huge changes in the body caused by pregnancy can trigger temporary diabetes (although sometimes it persists for a while after pregnancy). For this reason, we tend to test the blood sugar level of pregnant women.

Then we have MODY (Maturity Onset Diabetes of the Young), which is a form of monogenic diabetes. There are at least 10 types, caused by single gene defects in various genes. The treatment and prognosis depends on the gene which is defective. For instance, if you have one of your two copies of the GCK gene defective, then you get GCK MODY, which typically has few symptoms, but is usually detected when a woman is pregnant and has their blood sugar levels tested. The danger with GCK MODY is over-treatment - they do not need any treatment at all, but doctors may try to lower the blood sugar levels by giving insulin injections, which can be dangerous. Other forms of MODY do require insulin.

There is also Neonatal Diabetes (NDM), which has two broad types, being permanent and transient, and there are several different sub-types depending on the gene defect. In this case, you are generally diagnosed with diabetes under the age of six months, and sometimes as young as hours after birth. If both of your copies of the GCK gene are defective, you will get permanent neonatal diabetes, and require insulin injections for life, from birth. However, if you have neonatal diabetes caused by defects in the ABCC8 or KCNJ11 genes, treatment with sulfonylurea tablets is highly effective. It is important to obtain a genetic diagnosis in order to get the correct treatment.

There are other types of diabetes other than these. It is very weird for this article to be claiming the fact that there is more than just type 1 and type 2 to be news - it has been widely known for many years.

A lot of the recent work has been on trying to distinguish the different types of diabetes. Neonatal diabetes is easy to distinguish, because it has a severe early onset. It is possible to determine the genetic cause in about 80% of patients that have this condition, and we are confident that further searches will reveal a large proportion of the remainder. MODY is however too similar to type 1 diabetes, so of patients sent for genetic testing for MODY, a proportion of them will have type 1 instead, so the genetic test success rate is lower, around 30%. Genetic risk scores for the autoimmune type 1 diabetes response are used to try and distinguish the two, in order to try and perform gene discovery research on as pure a MODY cohort as possible.

>We have known for many many years about several other types of diabetes.

So are you saying the research project didn't contribute any new knowledge at all? In that case, why the quotes from the experts that it did?

On the other hand, if you do think it contributed some new knowledge, you should have said that and mentioned some specifics.

You can add a sixth class. I have been told that Cystic Fibrosis Related Diabetes (CFRD) is neither Type 1 nor Type 2.

Alzheimer’s disease is informally referred to as Type 3 diabetes because of similarity; this informal name Is prevalent enough that when a different form of type I was discovered (which may or may not be cluster 2 in this article) it was called type 4

What about pregnancy-related diabetes? Some textbooks have it as a 3rd independent type of diabetes.

It seems a remarkable number of comments seem to think that people with Type II diabetes resulting from Obesity should be compelled not to eat and should be denied healthcare in the interim, such as this one [dead] comment:

> Why spend money on a population who has a free, safe, and natural remedy (eating less) available to them?

Please study food addictions¹²³⁴. Addiction is a disease⁵. Going cold turkey may work for some, but for many, this option is infeasible with drugs. With food, it's infeasible for everyone.

Food addiction is especially difficult to combat without medical intervention⁶. Denying victims access to said care will only exacerbate the problem.

¹ https://www.webmd.com/mental-health/eating-disorders/binge-e...

² http://onlinelibrary.wiley.com/doi/10.1002/eat.20957/full

³ https://www.sciencedirect.com/science/article/pii/S019566631...

⁴ https://www.sciencedirect.com/science/article/pii/S019566631...

⁵ https://scholar.google.com/scholar?hl=en&as_sdt=0%2C47&q=add...

⁶ https://www.sciencedirect.com/science/article/pii/S019566631...

What happens to people who refuse to submit to medical intervention?

That's up to them. I'm addressing the prospect of refusing medical intervention to people generally, especially to those who want it.

Reply on news.ycombinator.com