Do diabetics who take Metformin really live longer than non-diabetics?

I’m not a doctor; I’m just a guy who doesn’t want to die at age 80. These are some thoughts based on my own research. They are absolutely not medical advice.

I’ve heard this claim bandied about quite a bit — that diabetics who take metformin, a Type II diabetes drug, live longer than non-diabetics. This is an extraordinary claim so requires extraordinary evidence to back it up, right? Just hearing that makes me want to rush out and take metformin every day for the rest of my life (not so fast for those of you who exercise). But is there enough evidence to do so? This article will hash that out a bit.

David Sinclair takes metformin[citation needed]. The drug is old, discovered in 1922, and surely we would know by now if there were detrimental effects, right? There are tens of millions of people in the USA alone that take metformin. Although sitting here typing this I’m thinking that maybe there are detrimental effects for non-diabetics who take it and we just don’t know about them yet because metformin hasn’t been sufficiently studied in non-diabetics yet. Not enough of us have taken metformin for a long enough period to really know.

If diabetics do indeed live longer than non-diabetics this would be a super compelling reason to take metformin, maybe even to give oneself diabetes first (LOL jk). Metformin is super cheap and is a generic drug (but does require a prescription in most countries so is not always super easy to get) so it’s worth spending quite a bit of time to understand exactly what the effects of metformin are, on non-diabetics and diabetics alike.

So, here is the study that initially reported that diabetics who take metformin live 15% longer. The study is an observational one. This means that the study’s authors did not do a prospective study that recruits patients and has a control group to determine causality of a single variable. Observational studies look at data that was already collected, often by large governmental health agencies and observe correlations between variables.

The study is (pseudo-eponymously) entitled Can people with type 2 diabetes live longer than those without? A comparison of mortality in people initiated with metformin or sulphonylurea monotherapy and matched, non-diabetic controls and was published by Diabetes, Obesity and Metabolism in 2014. The study looked at 78,241 people who were treated with metformin, and 12,222 treated with sulphonylurea. They matched these two groups of diabetes patients with 90,463 subjects who did not have diabetes and who were of a similar age, and had similar health traits such as whether or not they smoke and their BMI (although there were significant pre-observation health factors like whether or not they were taking cardiovascular drugs and how many times they had seen a physician in the preceding period). In total they observed 503,384 human-years with 7,498 total deaths. I think it should be noted that the metformin and metformin-control groups averaged 61.2 years of age at the beginning of the observation period.

The metformin diabetics had an 1.44% annual mortality rate during the observation period. The metformin control group, and we will dig deeper into what this means below, had a 1.52% mortality rate. This means that diabetics who were taking metformin died 5.5% less per year.

Therefore the study did indeed find that “patients with type 2 diabetes initiated with metformin monotherapy had longer survival than did matched, non-diabetic controls”.

When people talk about this study, they often say “diabetics who take metformin live 15% longer than non-diabetics”. I think it’s important to point out exactly what this means — or where this number comes from. Let’s look at that.

Here’s where this number comes from: “in diabetic patients initiated with metformin monotherapy … adjusted median survival time was 15% lower (STR = 0.85, 95% CI 0.81–0.90) in matched individuals without diabetes“. But does this mean that the diabetics lived to 92 and the non-diabetics lived to 80 or so, which is roughly the Western/developed average life expectancy? No. It means that during the study’s observation period, which averaged 2.8 years, the metformin-diabetics lived 15% longer. The 15% is a percentage of the observation period, not of a whole person’s lifespan. Big difference. So the metformin-diabetics lived about 5 months longer. This is something, but nothing like have a lifelong health span of 15% longer. Though perhaps if the diabetics had taken metformin their whole lives they would have actually lived 15% longer. We don’t know this, and likely won’t know the answer to this question for a very long time.

Funding Information

This study lists Astra-Zeneca and Bristol Myers Squibb as funding contributors.

So my skepticism about this study just massively increased because of the financial backers. Let’s dig into more details of this study and see if maybe we shouldn’t immediately throw it out because there was funding from drug companies.

Ok, so first off, Astrazeneca was releasing a diabetes drug around the time of this study’s publication, so yes they did have a financial incentive to boost the attractiveness of metformin. But metformin is an old drug, originally discovered in 1922. I looked recently at a list of metformin suppliers and found 44 different ones on every continent except Antarctica.

What this study was originally looking for is whether metformin or sulfonylurea was the superior Type II diabetes drug. The study’s goal was not to answer the question “do diabetics who take metformin live longer than non-diabetics”. The goal of the study was to compare all-cause mortality between the two most commonly prescribed Type II diabetes drugs. So the fact that the study showed that metformin diabetics lived longer was a surprise to the authors and they say so in the paper.

I think it is fair for these companies to want to highlight the superiority of metformin over sulfonylurea, especially if there had already been concerns about increased mortality in sulfonylurea subjects.

So they found that diabetics who took metformin did live slightly longer during the period in which they observed the mortality of all the groups in the study. In the figure below, the green line is the diabetics who took metformin and the black line is the non-diabetic controls.

So yes, the metformin diabetics lived longer but only slightly more. I can barely make out a difference in those lines.

Now one criticism of this article could be “don’t diabetics diet and exercise” to fight their diabetes. And I think the answer is sure, some of them likely did diet and exercise. But we don’t know how much dieting and exercise these diabetics were doing in this study’s observational period. We also know that Type II diabetes is largely reversible, but the vast majority don’t actually reverse it; they just live with it and take metformin. But we do know that the metformin-diabetics lived longer (unless the study’s authors cherry picked the non-diabetic controls). So you could argue that whatever these diabetics were doing caused them to live longer than non-diabetics. That is a compelling take away. Something they were doing reduced their mortality albeit only by five months.

So assume the diabetics were “dieting” (whatever that means), exercising and taking metformin. The takeaway from this study is that if you become diabetic there is still hope for you. Diet, exercise and take metformin. This is a testament to these three actions, all of which are very cheap (arguably they have zero cost in the sense that if you eat less you could probably easily save the cost of metformin).

We also know that the diabetics were objectively less healthy than the controls to start with.

Metformin Diabetics

Controls

Body Mass Index

32.4

27.4

Charlson Index

1.9

0.7

Prior lipid-lowering therapy, n (%)

39,407 (50)

15,913 (20)

Prior antihypertensive therapy, n (%)

52,016 (66)

Prior antiplatelet therapy, n (%)

28,285 (36)

One interesting figure from this study is how the possible benefits of metformin were more pronounced as the observed participants got older.  The following figure is for the 71-75 year old cohort in the study. The green line are the metformin-diabetics; the red and black lines are the two control groups. The blue line are the poor sulphonylurea patients 😩. The above chart showed barely any difference between all age cohorts. But the below image clearly shows a difference between the metformin-diabetics and the control groups. Could this extended lifespan be due to the fact that cancer risk increases exponentially with age?

One question I have about this study is why have two different control groups to compare the metformin and sulfonylurea against? Why not just have a single control group, that way you know you are comparing both drugs to the same group. I am not an expert in study design, but having two control groups seems a bit suspicious to me. Wouldn’t you want to compare the two drug groups with a larger control group. Wouldn’t this lead to a smoother control group chart and more statistically significant results? I suppose we can just sort of mentally average the control groups in the image above to accomplish this goal. Still I wish there was one control group. If you are more versed in this sort of thing, feel free to email me about this and edify me. Does having two control groups lead to the possibility of being able to cherry pick data or have an easier comparison to make? Or am I making a mountain out of a mole hill.

Another suspicious fact about this study is how the Sulphonylurea group and the matched control groups were almost 7 years older than the metformin and its control group. Sort of seems like they were setting metformin up for success by doing that.

Here is my non-doctor, non-medical advice conclusion:

It would not be wise to base a large investment in, and the daily taking of, metformin, solely on this study, if you are non-diabetic (whew, there is a long sentence with lots of commas that I do not feel like rewriting). There are too many unanswered questions. It is not an RCT; some undiagnosed diabetics were surely in the control groups. The quote that diabetics taking metformin live 15% longer is really taken out of context. They lived 15% longer in the observation period, which was a mean of 2.8 years. So they lived 5 months longer (15% * 33.6 month long median observation period). That is not as hugely positive as “they lived 15% longer”. Though as I am finishing writing this after having slept on this, what if you take metformin and you live 15% longer for any length of observation period?

I think metformin is something to pay close attention to and research further: it is very cheap, seems to be quite safe and has only one, I believe, long-term negative side effect (vitamin b12 deficiency). It also may interfere with some benefits of exercise so don’t take it close to when you exercise. Here’s a great video about metformin and some discussion of the half-life of metformin and how you should probably avoid taking it around exercise.

As with most epidemiological pursuits, the devil is in the details. I plan to do more research about metformin for non-diabetics.