Friday, July 14, 2017

Update On Dr. Faustman's BCG Research

There was some media buzz about Dr. Faustman's BCG research as reported on a poster at ADA 2017.  I can't link directly to it, but you can go to this page:
and search for Faustman to see the poster and the abstract.

Dr. Faustman's research has about 15 years of history, which I've blogged about before, here:
And in several other blogs.

But to summarize very quickly: Dr. Faustman's previous research was based on the idea that BCG (a Tuberculosis vaccine) increases the levels of TNF, and higher TNF levels could result in less autoreactive T cells, and this would lead to a cure for type-1 diabetes.  Unfortunately, this did not pan out.  Her phase-I trial showed that BCG did not change the levels of live autoreactive T cells in circulation.  However, using some non-standard data analysis, Dr. Faustman interpreted the phase-I study to show that there was an increase in T-reg immune cells.  T-reg cells get rid of the autoreactive T cells, so increasing the T-reg cell count could also lead to a cure for type-1 diabetes.

What is the New News?

The poster presented at ADA 2017 contained data on a possible mechanism whereby BCG could increase the level of T-regs in a type-1 diabetic.  This mechanism involved changes to how genes are used in the body (called "epigenetic changes").  These changes can be long lasting.  Obviously, a long lasting treatment is better than a short lasting one, so (if this finding is confirmed by more research) this is good news.

Also, although not required, it is nice to have a theoretical basis for why a treatment should work, if you are going to test that treatment. This is particularly true of the BCG research, since the phase-I trial showed pretty clearly that the previous theory was wrong, so having a new theory is good.

How Important Is This?

In my opinion, it is not very important. Why not?  In a nutshell:
1. It's a finding about a possible mechanism of how a cure works, not evidence of effectiveness, for a treatment where effectiveness is the important, unanswered question.
2. It's based on data from 3 people.
3. It's a poster, not a presentation (or a paper).

1. It's a mechanistic finding so it is aimed at answering the question "how does BCG work?", but the important question is "does BCG work?"  That needs to be answered before the "how" question is important.  The results from the phase-I trial were not successful (see discussion below).  Therefore, for me, research into the mechanism of how it might work is less important until we get some evidence that it does work.

2. Dr. Faustman's phase-I study gave BCG to only three people.  It is the smallest phase-I study I have ever covered in the field of type-1 diabetes.  This poster is based on data from those same 3 people, and that is not a large enough foundation to get me excited.

3. As a poster, this is a lesser form of publication, than a journal article or a presentation.  That is not fatal, of course, but it is a handicap.  For comparison, Gleevec and Oral Insulin both merited presentations, so the organizers of the ADA scientific sessions clearly thought that those studies were more important than this BCG result.  (And I agree with them.)

Some Discussion on The Results From the Phase-I Trial

Obviously, Dr. Faustman considers the phase-I trial to be a success, which is why she has started a phase-II trial, so why do I consider it unsuccessful?  For two reasons:

First, the normal definition of success for a clinical trial is successful results from the primary outcome.  Her phase-I trial's primary outcome was autoreactive T cells, and there was no difference between the the treated group and the control group.  So that's an unsuccessful result.

She did report two tiny successful outcomes in her secondary results.  But to get those, she had to do two different data manipulations.  I discussed those in detail in 2012 when the original results were published:
See the section called "Weaknesses in Data Analysis".

A very short summary is:
1. She shifted 1/3 of her control group to be analysed as though it were in the treatment group, after she had seen the results, and saw that would change the outcomes.
2. She did not use a single control group.  Rather, she used different control groups for different outcomes, including (in some cases) pulling data from other studies.

My older blog posting describes in detail how damaging these are to her analysis.  But suffice it to say, if the data shows success, there is no reason to do those manipulations.  And if the data is successful (even at a tiny level), only after those manipulations, then how real is it?

Now, if her phase-II study (which is expected to enroll about 150 people) is successful using standard data analysis techniques, then she can rerun this poster study with data from those 150 people.  She will then have more people and a successful treatment and at that point, a mechanistic study would be interesting. The phase-II study is expected to end around 2023.

Joshua Levy
publicjoshualevy at gmail dot com
All the views expressed here are those of Joshua Levy, and nothing here is official JDRF or JDCA news, views, policies or opinions. My daughter has type-1 diabetes and participates in clinical trials, which might be discussed here. My blog contains a more complete non-conflict of interest statement. Thanks to everyone who helps with the blog.

Phase-III Results from an Oral Insulin Clinical Trial In Presymptomatics

A Quick Introduction to Oral Insulin to Prevent T1D

Obviously, all type-1 diabetics need to take insulin in order to process carbohydrates. This insulin must be injected, because if it were taken orally it would be digested into smaller pieces and would not work as insulin [d1]. Injecting insulin in this way does not cure or prevent type-1 diabetes, it just treats it.

However, one of the autoantibodies that is associated with type-1 diabetes targets insulin molecules[d2]. Therefore, there is a theory that giving insulin to people with T1D might prevent or delay the onset of type-1 diabetes by training the body not to produce this autoantibody. The process is vaguely similar to giving small amounts of peanut proteins to people with peanut allergies[d3]. Insulin pills may work for this purpose even though they would not work as a treatment for type-1.

The Phase-III Study (Structure and Results)

These researchers started with 10,000s of TrialNet participants, and enrolled 560 people who were "presymptomatic". They showed two autoimmune markers, but no symptoms of type-1 diabetes. All of these people tested positive for one particular autoantibody associated with T1D (micro insulin autoantibody) [d2], but they were further subdivided into four groups based on the other autoantibodies they tested positive for, and how much insulin they were producing.

Each subgroup was split in half. One of these halves got oral insulin twice a day, and the other half got a placebo. They were followed for a year or longer to see how many people in each group developed type-1 diabetes as measured by "classic" symptoms.

If you look at the entire study, oral insulin did not have a statistically significant effect. However, if you looked at one subgroup specifically [d4], that subgroup did show a statistically significant effect. For that one subgroup, about 18% of the treated group came down with T1D, while 34% of the untreated group did. The researchers viewed this as delaying the onset of type-1 diabetes by 2.5 years (on average) for this subgroup.

Presentation Slides:

News Coverage:

Clinical Trial Record:

Discussion Of This Study

The results of this study are clearly "bad news on one hand, good news on the other". If you look at the study as applied to all presymptomatics, it was not successful. On the other hand, if you look at it for one specific subgroup, then it was successful. So the obvious thing to do is to try to replicate the results on the specific subgroup where it was previously successful. If so, this could turn into a delaying or preventative treatment for the 10% of patients who fall into this group. Since this subgroup had a specific combination of autoantibodies, it is straightforward to test ahead of time, and give oral insulin to people with this same combination, but not other combinations (where it did not work).

This trial was the size of a phase-III trial. However, the success was only seen in a subgroup, and that subgroup was the size of a phase-II trial. So I would not think of this as a successful phase-III trial, but rather as a successful phase-II trial (meaning that at least two phase-III trials should be expected before it becomes commonly available) [d5].

It is also important to remember that as group size gets smaller, the chance for accidental correlation gets bigger. With these sorts of subset analysis, it is always possible that the effect seen is cause by luck rather than effect. In this particular case, the results were statistically significant even for the smaller group, which is a good sign, but only larger studies will be conclusive.
The History of Oral Insulin

Oral Insulin has a long, complex history of clinical trials, and the results are very mixed (like this study). Just before I started my blog (so 10+ years ago), the results of an oral insulin for prevention trial were announced, and the trial was unsuccessful. However, the researchers analysed the data in more depth after the study concluded, and realized that it had worked for one subgroup (sound familiar?) That was the micro insulin autoantibody subgroup, and that's why everyone in this study had that antibody.

But the idea that oral insulin might prevent/delay type-1 is a popular one, and there are at least three clinical trials running right now. All three of these studies are similar, except for size: the first is 44 people, the second is 220, and the third is 92.

Oral Insulin Starts a Phase-II Trial In Germany (pre-POINT Early)
Forschergruppe Diabetes, Klinikum rechts der Isar, Technische Universität München, Lehrstuhl für Diabetes und Gestationsdiabetes der Technischen Universität München, München, Germany, 80804
Started in Aug 2015 and expected to finish in Aug 2017
News: Vaccination against type 1 diabetes may soon be available to young children:
Clinical Trial:

Phase-II Oral Insulin Trial In Germany (Fr1da)
Forschergruppe Diabetes, Klinikum rechts der Isar, Technische Universität München, Lehrstuhl für Diabetes und Gestationsdiabetes, der Technischen Universität München, München, Deutschland (DEU), Germany, 80804
Anette-G. Ziegler, Prof. Dr., MD +49 (0)800 464 ext 8835
Started in Dec 2015 and expected to finish in June 2021
Clinical Trial:

Phase-II Oral Insulin Trial In The US (TN20)
Not recruiting.
Started in Jan 2016 and expected to finish in Dec 2017.
Clinical Trial:
Extra Discussion

[d1] To complicate things, several researchers are working on creating a form of insulin which could be eaten, but which would avoid digestion, so that it could be used to treat type-1 diabetes. This is also called "oral insulin" research. In this blog posting, I'm talking about oral insulin as a cure or preventative for T1D, not as a treatment.

[d2] Autoantibodies are the malfunctioning antibodies which cause the immune system to attack beta cells. There are five autoantibodies associated with type-1 diabetes, and there may be more that we haven't discovered yet. The five we know about are:
* micro insulin autoantibodies (mIAA or just IAA)
* islet-cell antibodies (ICA)
* glutamic acid decarboxylase (GAD) antibodies
* islet antigen-2 (IA-2) antibodies
* zinc-transporter 8 (ZnT8) autoantibodies
The last one was not used in this study, possibly because it tends to show up later in the disease process.

[d3] It is important to realize that type-1 diabetes is NOT a conventional allergy to insulin. It is similar to allergies in that it is the body's immune system overreacting to something that it should not react to, but other than that, is quite different. Allergies involve the immune system overproducing histamines. These histamines attempt to get physical irritants, like pollen, out of your body. You can counter this histamine reaction by taking antihistamines. Type-1 diabetes involves the immune system overproducing malfunctioning killer T-cells (or perhaps under producing regulatory T-cells). These malfunctioning killer T-cells mistakenly kill beta cells, thinking they are foreign cells (ie. living creatures like viruses, that have invaded the body). So the mechanism is different (histamines vs. T-cells), and the mistaken target is different (physical things, like pollen or wheat vs. living organisms, like viruses).

[d4] The subgroup that showed the effect was the micro insulin autoantibody (which everyone in this study had), and either the ICA autoantibody or both the GAD and IA-2 antibodies, and also low insulin secretion at the start of the study.

[d5] As far as I can tell, oral insulin is not approved for the treatment of any disease in the US, and is not available either by prescription or "over the counter". Therefore, it will need to go through full US FDA approval, which requires two phase-III trials. I don't know if the FDA would consider this a phase-III trial for approval purposes.

Joshua Levy
publicjoshualevy at gmail dot com

All the views expressed here are those of Joshua Levy, and nothing here is official JDRF or JDCA news, views, policies or opinions. My daughter has type-1 diabetes and participates in clinical trials, which might be discussed here. My blog contains a more complete non-conflict of interest statement. Thanks to everyone who helps with the blog.

Wednesday, July 5, 2017

Presymptomatics and Two Clinical Trials for Victoza / Liraglutide

First, a new word: "Presymptomatic". This refers to people who have tested positive for two autoantibodies, but who have no other symptoms of type-1 diabetes. Their blood glucose levels are normal (not elevated), etc.

Presymptomatics are not yet diagnosed with type-1 diabetes in the classic way, but current theory is that all of these people will eventually be diagnosed. It is just a matter of time. So in the same way I might say "Drug X starts a phase-I trial in honeymooners" or "Treatment Y starts a prevention trial" or "Drug Z starts a trial in people with established type-1 diabetes", I will also start to report "Drug W starts a phase-I trial in presymptomatics".

You can think of presymptomatics as pre-honeymooners.  They are like honeymooners, but even earlier in the disease process.

I also want to stress that although the JDRF, the ADA, and the Endocrine Society agree that two autoantibodies is the earliest diagnostic for type-1 diabetes, this is not universal agreement, and what agreement there is, is only about 2 years old.  Here are some web sites which describe this view of the stages of type-1 diabetes:
This also changes when "diagnosis" occurs.  In the past, diagnosis occurred when symptoms were seen, and confirmed with a blood glucose measurement.  However, now diagnosis occurs when two autoantibodies are measured, and this is often years before symptoms are seen, or blood glucose levels are noticeably abnormal.

So, moving forward, I will use the term "classic diagnosis" or to refer to people who were diagnosed because they showed symptoms, as was done in the past, so it's obvious what kind of diagnosis I'm talking about.


I expect there will be more studies like the two described below, that specifically target presymptomatics. After all, any treatment that researchers thought might work for honeymooners (but did not), should now be retested on presymptomatics. This is especially true of treatments which change the immune system.

In the past, it's generally been understood that to cure type-1 diabetes, you needed to change the immune system (so it stopped generating autoantibodies and stopped attacking beta cells), but you also needed to regrow beta cells.  However, presymptomatics have enough beta cells so that they can regulate their own blood glucose levels.   To cure them (ie. to prevent symptoms from ever showing up), "all" you need to do is change the immune system.  No need to regrow any beta cells.

That sounds important, and it is, especially when you think about treatments that have already been shown to stop the destruction of beta cells.  In the last 5-10 years, several treatments have been shown to "preserve beta cells" meaning that once given, beta cells stop being killed off by the immune system.  Since these studies were typically done in honeymooners, this did not cure anybody, it just extended the honeymoon.

But if those same treatments showed the same results in presymptomatics, then it could be said that they prevented type-1 diabetes.  I very much hope that every treatment which has previously been found to preserve beta cells, will now be tested on presymptomatics. Some of the treatments which have preserved beta cells in honeymoon diabetics (at least to some degree) are: T-Rex (polyclonal Tregs), Abatacept (Orencia), Etanercept (ENBREL), and Teplizumab. 

Victoza / Liraglutide Starts A Phase-I Trial In Presymptomatics

About this trial: it's testing the theory that Liraglutide (sold as Victoza) might help people use less insulin or delay their use of insulin, when given to people before they are classically diagnosed with type-1 diabetes.  This is an early phase-I trial.  Only 10 people will be enrolled, and there is no control group.   This trial recruits people who have started to have trouble generating insulin in response to food that they've eaten.  They will be followed for one year.  The trial started in March 2016 and they hope to finish by July 2018.

They are recruiting only by invitation at several nordic hospitals:
  • University of Oulu and Oulu University Hospital, Dept of Children and Adolescents Oulu, Finland, 90029
  • University of Tampere and Tampere University Hospital Tampere, Finland, 33520
  • University of Turku and Turku University Hospital Turku, Finland, 20520
  • Lund University and Skåne University Hospital Malmö, Sweden, 205 02
Clinical Trial Record:


The Clinical Trial Record lists this as a phase-II trial, but with only 10 people included and no control group, I consider it a phase-I trial.

I don't see how this trial could ever prove any level of effectiveness.  We don't know how many people eligible to enroll in this study would "naturally" have type-1 diabetes symptoms within the one year study time.   And, this trial has no control group.  So there is no way to compare the results from this study to "normal" results to see if it worked or not.

It could show safety, but the drug being tested has been approved for use in overweight people and also people with type-2 diabetes for years, and is used "off label" by some people with type-1 diabetes, so safety is not really an issue.

Victoza / Liraglutide Starts A Phase-I Trial In Presymptomatics

This trial is similar to the one above, except that it is much larger, and recruiting a slightly different population.  This second trial recruits people who have two autoantibodies and one of several different glucose abnormalities, so it's a larger group of people, and also more in tune with the "two autoantibodies means type-1 diabetes" definition of Presymptomatic.  

This is an phase-II- trial which will enroll 82 people with half in a control group and half getting the treatment.  People will be followed for one year.  The trial started in 2016, and is expected to finish in mid 2019.  They are recruiting by invitation only at the same hospitals listed in the previous trial.


(As you read this remember that I'm not a statistician, and have never take a college level class in statistics.)
I'm a little worried about the statistical power of this study.  They are going to have 41 people in their treatment group, and will follow them for one year.  The data I've seen suggests that about 10% of the people with two more more autoantibodies will show classic type-1 diabetes systems each year.  So that means that we should expect about 4 from this group to show symptoms by the end of the study. If this treatment were perfect in preventing type-1 diabetes, then 0 in the treated group would show symptoms.   But the difference between 0 and 4 is not that large.   And the treatment is unlikely to be perfect the first time it is tested.  Let's say that 2 people in the treated group get symptoms, but 4 people in the untreated group get them.  Is that a 50% reduction in diagnosis (which would be huge) or is that just a little good luck involving two patients?  It's hard to tell, and that is what I'm worried about.

On the other hand, this is a phase-II study, so will not be the last word, in any case, and Victoza is already approved, so is not a particularly risky drug.  Also, if the results for the first year are good, then it would be relatively easy to extend this trial for another year (or longer) which would increase its statistical power.

Joshua Levy
publicjoshualevy at gmail dot com
All the views expressed here are those of Joshua Levy, and nothing here is official JDRF or JDCA news, views, policies or opinions. My daughter has type-1 diabetes and participates in clinical trials, which might be discussed here. My blog contains a more complete non-conflict of interest statement. Thanks to everyone who helps with the blog.