Friday, December 2, 2011

Dead in Bed: What is the Chance?

Note: this posting was edited for clarity on Dec-24 .
Warning: this posting deals entirely with the worst side effect of type-1 diabetes: death.
Please do not read this posting if discussion of death upsets you.
Also, if you are a brittle type-1 diabetic or the relative of one, then [r12] will be particularly shocking.  Please read the entire discussion with that reference, if you read any at all.
Some people may find [r13] particularly upsetting, as it deals with a specific death, rather than statistical deaths in general.

The r-numbers in square brackets [r1] refer to references which are discussed throughout the post, the d-numbers [d1] to extra discussion at the bottom of the post. 

Recently, there was a lot of shock and horror when JDRF published an ad which said that for a type-1 diabetic the chance of dying of low blood sugar was about 5% over a lifetime.  So in this blog posting I'll examine the data on "dead in bed" to see if 5% is the correct rate.  Obviously, this posting is not about my regular topic: clinical trials aimed at curing type-1 diabetes.  But it is a subject important to everyone near to type-1 diabetes, and I wanted to see if that 5% number is true.

I don't know if anyone has said this before, but if not, I'm saying it now:
"There is never enough data to convince someone of something that they don't want to believe."
My one paragraph summary: there is no doubt that for type-1 diabetics who die young (ie under about 40 years old), over 5% of these deaths are due to hypoglycemia (low BG).   All the recent studies show this.  I could not find any data at all to come to a conclusion about the death rate for type-1 diabetics older than that. Since the data we do have is over 5%, I think 5% is a conservative estimate, although the lack of data for older diabetics does leave room for speculation that it is lower, there is no data to suggest that it is lower.

This posting is in four sections:
1. Some background information and discussion about how to measure death.
2. A review of the studies that JDRF referred to in their follow-up email as supporting the 5% number.
3. A review of other available studies, from my own research.
4. Some discussion on the social and political importance of this data.
    Background Information

    Measuring Death is Harder than you Think

    Measuring how type-1 diabetics die is a lot harder than you might think, with unconnected medical records, like the US [d1]. The obvious thing to do, is to select a group of people, wait for them to die, and record how they died. However, you need to wait for them to die, so the data is available an entire generation after you selected the people. You could also do this "in reverse", research everyone who dies in a given place, and find out which ones have type-1 diabetes, and then record how they die. To do this for 300 type-1 diabetics, you'll need to research about 90,000 people who die just to find the 300 who have type-1 (remember only 1 in 300 will have type-1). That's a problem, too. A third way to do it is to follow many people of different ages, and then splice the data together grouped by age, to get a chance of death over an entire lifespan. But that requires following a lot of people, in several different groups, and it's not easy, either.

    But none of these techniques are going to give you quick, up-to-date, and easy-to-get information on how type-1 diabetics die.

    To make matters worse, not all "dead in bed" cases are hypoglycemia[d2,r10], and in many cases, especially in the past, these were tracked as sudden, unexplained death (or similar) but not generally considered a side effect of diabetes.

    Many of the studies done in the past reported on "chronic complications" of type-1 diabetes (things like heart attack, loss of limbs, etc.) and "acute complications" (either low BG or high BG / ketoacidosis).  But they did not provide data on the number of low BG related deaths, just on all acute complications combined.

    Finally, and perhaps most horribly, some researchers have referred to "dead in bed" or hypoglycemia as "insulin overdoses" or "drug misuse".  This has the effect of blaming the type-1 diabetic for their own death, or maybe blaming their doctor for prescribing too much insulin.  In any case, if a researcher had the choice of listing death as "insulin overdose" or "unknown cause of death" which did you think they did?  But then the true cause of death is lost from later analysis.

    These issues have in the past lead to an under counting of deaths caused by hypoglycemia, but they provide little help in determining what the rate actually is.

    A Review of JDRF's Sources

    A quick summary of the data is as follows: two of JDRFs sources were very similar, and written by the same person, and used a total of 5 studies to estimate the 5% number.  See the quote under [r2] below.  Basically they showed that older studies had 2%-4% numbers and newer studies had 6%-10% numbers.  For reasons described above, I agree with JDRF that the new studies should be given more weight.

    Another study that JDRF cited was the DCCT trial.  This is a large, recent study on the complications of type-1 diabetes.  This well respected study is commonly cited when researchers need data on rates of complications.  I suspect it has been used dozens, if not hundreds of times in the years since it was published.  It found a rate of 6%.

    Below are listed the 9 sources that JDRF referred to in their email as supporting their 5% number:

    [r1] Cryer PE. The barrier of hypoglycemia in diabetes. Diabetes 2008;57(12):3169?76.
    full paper:
    This paper came to very similar conclusions to the one below, based on the same underlying research, and done by the same person, so see [r2] for details.

    [r2] Cryer, PE. Hypoglycemia in Type 1 Diabetes Mellitus. Endocrinol Metab Clin North Am. 2010. 39(3): 641-654.
    full paper:
    Early reports suggested that 2% to 4% of deaths of people with diabetes are the result of hypoglycemia [r1][r16]. More recent reports indicate that 6% to 10% of deaths of people with T1DM are caused by hypoglycemia [r7][r8][r9]. Regardless of the exact rates, the existence of iatrogenic mortality is alarming.

    [r3] Cryer PE. Death during Intensive Glycemic Therapy of Diabetes: Mechanisms and Implications. Am J Med 2011 124(11):993-996.
    No abstract or paper available, still in process of being printed.

    [r4] Deckert T, Poulsen JE, Larsen M. Prognosis of diabetics with diabetes onset before the age of thirty-one. I. Survival, causes of death, and complications. Diabetologia. 1978;14:363-370.
    No abstract or paper available to me.

    [r5] Tunbridge WMG. Factors contributing to deaths of diabetics under fifty years of age. Lancet. 1981;2:569-572.
    No abstract or paper available to me.

    [r6] Laing SP, Swerdlow AJ, Slater SD, et al. The British Diabetic Association Cohort Study, I: all-cause mortality in patients with insulin treated diabetes mellitus. Diabet Med. 1999;16:459-465.
    Neither abstract had data on low BG deaths, and paper was not available to me.

    [r7] Diabetes Control and Complications Trial/Epidemiology of Diabetes Interventions and Complications Study Research Group. Long-term effect of diabetes and its treatment on cognitive function. N Engl J Med 2007;356(18):1842?52.
    Full paper:
    A total of 1144 patients with type 1 diabetes enrolled in the Diabetes Control and Complications Trial (DCCT) and its follow-up Epidemiology of Diabetes Interventions and Complications (EDIC) study were examined on entry to the DCCT (at mean age 27 years) and a mean of 18 years later with the same comprehensive battery of cognitive tests.
    Of the 53 deaths during the DCCT and the EDIC study, 3 were attributed to hypoglycemia ...[So a 6% rate.]

    [r8] Feltbower RG, Bodansky HJ, Patterson CC, et al. Acute complications and drug misuse are important causes of death for children and young adults with type 1 diabetes: results from the Yorkshire Register of diabetes in children and young adults. Diabetes Care 2008;31(5):922?6.
    A total of 4,246 individuals were followed up, providing 50,471 person-years of follow-up. Mean follow-up length was 12.8 years for individuals aged 0-14 years and 8.3 for those aged 15-29 years.  ...   A total of 47 of 108 deaths (44%) occurred from diabetes complications, 32 of which were acute and 15 chronic. [
    The [r9] study below found that about 30% were acute and about 10% were low BG, so if that ratio is true for this study as well, then this study would also find about 10% death rate from low BG.

    [r9] Skrivarhaug T, Bangstad HJ, Stene LC, et al. Long-term mortality in a nationwide cohort of childhood-onset type 1 diabetic patients in Norway. Diabetologia 2006;49(2):298?305.
    Full paper:
    All Norwegian type 1 diabetic patients who were diagnosed between 1973 and 1982 and were under 15 years of age at diagnosis were included [1,906 people]. Mortality was recorded from diabetes onset until 31 December 2002 and represented 46,147 person-years. The greatest age attained among deceased subjects was 40 years and the maximum diabetes duration was 30 years.
    This paper found that about 10% of the people who died, died of low BG.

    A Review of Other Sources

    When summarizing research papers, the biggest single source of bias is to only include papers which support your position in the list of papers summarized.  So, to see if that happened, I did my own search of the literature, using Pubmed, and Google Scholar as my primary sources.
    My summary of these other sources, is that most of them do not provide directly useful data, but that the data they do provide does not conflict with the 5% number from the JDRF ad.

    [r10] Abstract:
    Diabetes Care. 2005 Oct;28(10):2384-7.
    Mortality in childhood-onset type 1 diabetes: a population-based study.
    Dahlquist G, Källén B.
    Mean age at death was 15.2 years (range 1.2-27.3) and mean duration 8.2 years (0-20.7).

    Seventeen diabetic case subjects were found deceased in bed without any cause of death found at forensic autopsy. Only two of the control subjects died of similar unexplained deaths.  In my opinion, this shows two things: first, that most "dead-in-bed" cases are acute complications of type-1 diabetes, but also that a few are not.  This makes the accounting harder to do.

    [r11] Full paper:
    BMJ. 2011 Sep 8;343:d5364. doi: 10.1136/bmj.d5364.
    Time trends in mortality in patients with type 1 diabetes: nationwide population based cohort study.
    Harjutsalo V, Forsblom C, Groop PH.
    Key table:

    This paper found about 19% (very roughly) died of acute complications.  If the same 1/3 ratio seen in [r9] is also true here, that would result in about 6% of deaths caused by low BG.  However, this paper separated alcohol/drug related acute events and those unrelated.  About 40% of the acute deaths were related to alcohol or drugs.  I think that is important to remember.

    The outcome of brittle type 1 diabetes--a 20 year study.
    Cartwright A, Wallymahmed M, Macfarlane IA, Wallymahmed A, Williams G, Gill GV.
    Department of Diabetes/Endocrinology, University of Liverpool, Liverpool L9 1AE, UK.

    This was the most emotionally horrifying paper I came across.  It was a small (33 person) study focused on brittle diabetics, the ones most likely to die from low BG.  They found that 20% of the deaths were caused by low BG, and that the type-1 diabetics who started out brittle (by their definition) had a death rate of 50% (!) over a 20 year period.  At the end of the 20 year study, none of the surviving patients remained brittle.  A very depressing result, but I don't think the data applies to most diabetics.  But it certainly makes me understand why brittle diabetics would be willing to have transplantation surgery including rest-of-their life drug treatments.  According to this study, the alternative is a 50% chance of death, and many chronic complications. 

    However,  I later came across this follow on paper:
    which suggested that some of the type-1 diabetics in the previous study were brittle because of psychological issues or a lack of training.  The exact quote was this:
    Most [surviving type-1 diabetics from the previous study] attributed their previous instability to life stresses and/or inadequate diabetes-related education. Two (20%) admitted to inducing dysglycaemia by therapeutic interference. ... None of the survivors was actively brittle, and most attributed resolution of brittleness to positive life changes.
    [r13] Abstract:
    Confirmation of hypoglycemia in the "dead-in-bed" syndrome, as captured by a retrospective continuous glucose monitoring system.
    Endocr Pract. 2010 Mar-Apr;16(2):244-8.
    Tanenberg RJ, Newton CA, Drake AJ.

    These researchers were recording CGM data (not monitoring it in real time!), from a patient who died of hypoglycemia with the monitor attached.  They were able to provide absolute proof that, at least some, "dead in bed" cases were directly caused by low BG.

    For me, the most interesting data from this case, was that this person did not "spike low".  It is not that he suddenly dropped to a very low BG, and then died.  Nor is it that he dropped low, and then quickly died.  He was low for hours before death.  I suspect that his body was doing everything it could (glucagon, etc.) to try to keep the BG up.  The CGM was alarming repeatedly.  But after hours of keeping BG levels above fatal levels, the body simply could not do this any more, and the person died.

    Abstract did not mention % of deaths caused by low BG, but more than 90% of the people in this study were type-2 diabetics who were treated with insulin.

    H. Fishbein and P. Palumbo, "Acute Metabolic Complications in Diabetes," in Diabetes in America (Bethesda, Maryland: National Diabetes Data Group, 2nd ed. (1995) ch. 13, p. 283

    Chapter in a book, but no specific information on prevalence of low BG as cause of death.

    [r16]  Cryer PE. Pathophysiology, Prevalence and Prevention. American Diabetes Association; Alexandria, VA: 2009. Hypoglycemia in Diabetes.

    Some Discussion

    JDRF seems to have relied on several papers published by Dr. Cryer to develop their 5%.  His published record of research on the causes of death of type-1 diabetics goes back at least as far as 1990, and he has published a wide range of papers on this subject.

    Why Publicize the 5% Number

    To be blunt, because it is the only way to get the FDA to do their job. As I describe the situation, please remember that I'm speaking only for myself, and these are my opinions based on the (indirect) information available to me: The FDA is supposed to approve devices because they are scientifically shown to be safe and effective.  However, in this case they are simply refusing to do so [d4].  Insulin pumps with automatic shut offs for low BG conditions were approved in Europe years ago, and have been actively used there (by large numbers of people!) for at least two years.  There is no question about the safety or effectiveness of these systems.  Never the less, the FDA refuses to approve them here.  Luckily, the FDA, being a political agency, is subject to political pressure, and I assume that is why JDRF is publicly pointing out that the result of the FDA's lack of approval is death.  I very much wish that the FDA would do it's job based on the scientific data showing safety and effectiveness, but they aren't.  So this sort of pressure is the only other option available.

    BTW: If anyone who works for the FDA or has first hand knowledge of the approval process for automatic shut off insulin pumps: I would very much like to talk to you about what IS happening.  Send me email, and I'll send you my phone number.

    And remember, refusing to approve a safety cut off for low BG levels in a pump, has the effect (at least short term) of stopping all movement on a commercial artificial pancreas in the US.  Every pump manufacturer in the world knows that if the FDA won't approve a low BG cut off, they surely won't approve anything more advanced either.  So the best treatment likely to be available in the next decade or so, is being held hostage by FDA unreasonableness.  There is a lot at stake here.

    Late breaking news: in the last day or two, the FDA has announced new guidelines for testing closed loop / artificial pancreas systems.  I'll see if I can put together a blog on that news in the near future.  Better late then never, I guess.  Hopefully better guidelines rather than worse ones.

    Non-Data Based Arguments That the 5% Number is Wrong

    I was a little surprised (but I shouldn't have been) about some of the arguments that people made that the 5% number was wrong, that was not based on data at all.  I discuss two of those arguments -- very briefly! -- below.

    I never heard that number before, so it must be wrong. [d5]    Many people are uneasy with discussion about the possible death of themselves or their children.  There is a lot of pressure not to talk about death as a side effect of type-1 diabetes.  So it is not at all surprising that there is not as much talk about it as it deserves, and hence, many people have not heard about it before.  But that is no reason to assume, when it is talked about, that the data is wrong.

    Also, as long as doctors thought of low BG as insulin overdose or as drug misuse, then they also may choose not to talk about it, since they will end up blaming the dead, or the dead person's doctor.  So, both doctors and patients (including relatives of patients) had good reasons not to talk about "dead in bed", so some people are a little surprised to hear about something they are not used to hearing about.

    That number only applies to relatively young type-1 diabetes (under 40, for example), and doesn't apply to all because most die when they are older than that.  This is not a groundless argument, because the data we have is for people younger than 40, but most people die when they are over 40.  However, it requires a lot of speculation.  We have data for people under 40 and the more recent data shows a higher rate than 5%.  We have no data for people older than 40, so some people hope that the over 40 number death rate might be very different from the under 40 death rate, and therefore that the entire-life chance of dying from low BG might be lower than 5%. For me, that's not reasonable doubt, that's just speculation. Maybe "wishful thinking" is a better phrase to describe it. I do think that running a study focusing on older type-1 diabetics would be a good thing, and would fill an important hole in the data. But I do not think it is reasonable to speculate that the data we don't have is different from the data we do have.

    Why Talk about Scary Data? / Why Present the Data so Strongly? 

    When this data was presented several people felt it should be muted or toned down.  I think that is largely a matter of personal taste.  Do you get more from being quiet and polite or being loud and scary?  Different people will disagree and this is reasonable.  For my part, I think the JDRF and the the pump/CGM industry has been taking the quiet and polite tack for years, and it doesn't seem to be working.  So I'm cool with the loud and scary tactic at this point.

    Obviously death is the worst possible side effect of type-1 diabetes.  If we are not prepared to get loud and scary about that, then what? There is no question in my mind that we are looking at the cause of about 10% of the type-1 diabetics who die young (under 40).  That's huge all by itself, without even starting to discuss 5% over a lifetime.

    More Discussion and References

    [d1] I mean records which are scattered and hard to review or use for large scale studies.  In the US, we have death records, but no way to link them to health records.  If a patient changes doctors or health plans, their records become separated, and so on.  There is no place to look for a person's entire health history.

    [d2] In the 1990s co-worker of mine (in his 20s) was found dead-in-bed, and he did not have type-1 diabetes.  It was very mysterious and ominous.  See [r10] for a little more data.

    [d4] I don't follow FDA process closely myself, but my understanding is that the FDA started out saying that in order to be approved, an automatic cut-off system had to show that it reduced low BG events by 10% compared to MDI.  Obviously, this is NOT showing safe and effective, this is showing better than the competition, so already a groundless requirement for the FDA to make.  However, when the company actually presented the studies to the FDA, to get them reviewed prior to starting, the FDA changed it's mind, and decided that the company had to show 30% decrease!

    The European safety agency actually did what the FDA was supposed to do, they required tests that the shut off feature did not cause any problems (safety), and that it worked at least as well as current pump technology (effectiveness) in terms of low BG issues.

    [d5] The speaker thinks that because they themselves have never heard something, then therefore it must be false. Or the speaker thinks that something doesn't make sense to them, so therefore it must be false. The first is very dangerous because it assume a person is all-knowledgeable, so if they haven't heard it, it's not true. The second is dangerous because it assumes that the truth always makes sense (or is logical), and it doesn't.

    The soundtrack for this blog entry is Juke Box Hero (Any Live Version) by Foreigner as found on

    Joshua Levy
    All the views expressed here are those of Joshua Levy, and nothing here is official JDRF or JDCA news, views, policies or opinions. My blog contains a more complete non-conflict of interest statement. 


    Wendy said...

    Thank you for this most in depth post. I really appreciate the time you took to research the topic...and then provide a synopsis.

    Great post...definitely one I'd like to point my DOC circle towards.

    Wendy said...

    BTW, I'm very curious about an official definition of what exactly a "brittle diabetic" is. I've heard that term tossed around for decades, but really have no idea what criteria makes one PWD brittle and another just a plain old PWD.


    Joshua Levy said...

    I'm not sure that there is one "official definition" of what is meant by "brittle diabetic". I wanted to know the exact definition that the r12 study used, but was not able to find it. I don't pay for journal access, and the journal that published that study did not have a free version on-line. Your doctor (or a friend who works at a University) may be able to get it.

    And thank you for the kind words about my blog.

    Joshua Levy

    WendyP said...

    Really appreciate this post. We are in desperate need of some hard research on what is happening with DIB and why it seems to be happening to more often our teens. It was very interesting what you said about the one case where the CGM indicated that the BG was low for hours. That's the kid of info we need. I really wish JDRF would take a more serious look at this and try to give us some more answers. Thanks for your post. This is the first time I've seen anyone try to really look into this horrible situation.

    Charlotte Rossmann said...

    Thank you for this post and your research.
    Finding all the facts is the only way we can move out of the dark into finding a cure.

    Scott S said...

    I agree; besides, the conversation about acute deaths from type 1 diabetes never seems to be discussed, all while the FDA delays technology to prevent these deaths is just wrong. I would say this is a recent example where the JDRF moved to aggressively influence the FDA and it is a good thing. I question why the FDA would worry endlessly about anything that might marginally increase HbA1c's, which is itself a "surrogate endpoint" and not a guarantee of improved patient outcomes, to be completely illogical. I am glad the JDRF is pushing this issue. The FDA is the wrong party here, not the organization for prompting people with unfortunate but sad facts. I would have liked to see this response a while ago, but my assessment is "better late than never".

    Kristin said...

    Great analysis - thank you! I heard Dr. Buckingham speak about this recently (at CWD conference), and his conclusions were similar. He showed a couple slides of dead-in-bed instances caught on CGM (double-blind study), and while they were chilling (felt inappropriate just to be looking at them), he made the point as well that BG had been at 50 or lower for 3-5 hours before siezure or further decline. Cold comfort, maybe, but we need to know the details of what we're dealing with. Sadly timely post, yet again - you and Moira (Despite Diabetes) have helped process a tough couple weeks.

    Anonymous said...

    Hi- As a first time poster I first want to thank you profusely for your blog and the tremendous amount of work and thought that clearly goes into it. I have a 4 year old that has had type 1 since she was 13 months old, and recently had my 7 year old diagnosed a few months ago. I can't express enough the comfort hearing unbiased logical analysis of the current state of the world.

    On this post, I would say that any awareness and mobilization around the issue (both in terms of treatment and cure) are extremely important and I'm supportive, I have a specific question referring to the statement (and in general the headline number of 5% that's being sent around), can you please clarify:
    Early reports suggested that 2% to 4% of deaths of people with diabetes are the result of hypoglycemia [r1][r16]. More recent reports indicate that 6% to 10% of deaths of people with T1DM are caused by hypoglycemia [r7][r8][r9]. Regardless of the exact rates, the existence of iatrogenic mortality is alarming.

    There is a huge mathematical difference between 5% of the people with Type 1 dying of low BG and 5% of the deaths of people with Type 1 being attributed to low BG.

    Using the unscientific mortality rate of 0.87% (I simply took the results of a study last year that showed 279 people out of 1100 died over at least a 29 year period) that were diagnosed with type 1 when they were under 18. My 0.87% annual rate comes from 279/1100/29. If 5% of those deaths are due to low BG, that would represent an annual risk of 0.04% mortality due to low BG. Again, still very high and something to work very very hard to lower, but not what most people are thinking when they hear the 5% headline.

    Thanks again for everything you do!

    Link for mortality rate: