There are two diseases that share the name “diabetes mellitus”. This is unfortunate, because the diseases have very little in common, except for the fact that both are associated with high levels of glucose in the blood stream. Type 1 diabetes is an auto-immune disease, in which the immune system destroys the insulin producing cells that reside in the pancreas. Type 1 diabetics quickly die if they aren’t treated with insulin. For them, it is immediately and dramatically life saving.
Type 2 diabetes, on the other hand, is a lifestyle disease, caused by excessive consumption of refined carbohydrates. This results in metabolic dysfunction and “insulin resistance” (a state in which muscle cells and fat cells stop responding normally to insulin, which causes glucose levels to rise in the blood stream). While type 1 diabetics literally produce no insulin, type 2 diabetics produce plenty of insulin.
So, when a person with type 1 diabetes takes insulin, they are replacing a substance that they are lacking, and which they need to survive. When a person with type 2 diabetes takes insulin, they are taking more of a substance that they’re already producing a ton of. There is no immediate survival benefit.
So why would anyone ever come up with the idea of giving insulin to people with type 2 diabetes, who already produce lots of insulin?
That is a very reasonable question. Here’s a long-winded answer: Insulin is a hormone, i.e. a signalling molecule that circulates in the blood stream. Among its many functions is telling muscle cells and fat cells to hoover up glucose from the blood stream. Like I said before, the one thing type 1 and type 2 diabetes have in common is that the levels of glucose in the blood stream are high. In type 2 diabetes, the muscle cells and fat cells don’t respond normally to insulin, but they can be “bludgeoned” in to doing what they’re supposed to if they are given a sufficiently high dose. Which is why doctors frequently give insulin to type 2 diabetics.
The extra insulin helps to lower the blood sugar. But does that actually matter?
Well, that’s where it gets complicated. Blood sugar is a surrogate marker, just like blood pressure and LDL-cholesterol. What really matters to people is whether they have a decreased risk of bad outcomes, like strokes and heart attacks, not what their specific blood sugar level happens to be. It has been assumed that the harms associated with type 2 diabetes are primarily due to the high blood sugar levels. Which is a reasonable hypothesis, but it needs to be tested.
We don’t care about lowering blood sugar if there is no beneficial effect on survival, or heart disease risk, or risk of blindness, or something else that patients actually care about. Additionally, lowering blood sugar with insulin isn’t a harm-free intervention. First of all, insulin dosing is hard to get right, and it’s quite common for people taking insulin to accidentally overdose and end up experiencing an episode of hypoglycaemia (which can occassionally be fatal). Second, insulin raises blood pressure. Third, insulin makes you fat. So any potential positive effects that you get from insulin use need to be weighed against the negative effects.
Proponents of insulin treatment for type 2 diabetes usually point to the UKPDS study, which was published in the Lancet in 1998. Since it forms the basis for the opinion that it’s a good idea to give insulin to type 2 diabetics, it’s worth looking at in some detail. So, here’s what happened.
3,867 people with newly diagnosed type 2 diabetes were randomized to either intensive blood sugar control with insulin, aiming at a fasting blood sugar of less than 6 mmol/l (which is quite low, actually within the range of what is considered a normal fasting blood sugar for a non-diabetic), or “relaxed” blood sugar control with diet, in which active measures with blood sugar lowering drugs were only taken if fasting blood sugar went above 15 mmol/l (a level that is high enough that it causes lots of doctors and nurses to panic).
In other words, the study was testing two quite extreme strategies, one in which blood sugar was lowered aggressively, and one in which nothing was done about blood sugar unless it went very high. The participants were followed for an average of ten years, which is a long time, and combined with the fact that there were almost 4,000 people in the study, this is definitely enough data to show a meaningful difference in outcomes if one exists. So, what happened?
Well, fasting glucose and HbA1c (a measure of average blood sugar over the last month or two) were significantly lower at all time points throughout the study in the intensive treatment group, so the insulin treatment was successfully keeping blood sugar levels down. But did this make a difference to any of the hard outcomes that people actually care about?
Let’s look at the hardest of hard outcomes first: death. 0.2% of participants in the insulin group had died after ten years. In the control group, 0.2% of participants had died. In other words, there was no difference in mortality.
Next, let’s look at heart attacks. 16.4% of participants in the insulin group had a heart attack, compared with 18.1% in the control group. That is a measly 1.7% reduction in risk over ten years. It wasn’t statistically significant.
Now, let’s look at strokes. 4.6% of participants in the insulin group had a stroke over ten years, compared with 5.2% in the control group. That is a 0.6% reduction in risk over ten years, and again, it wasn’t statistically significant.
What about outcomes that we worry about more specifically with type 2 diabetes, such as amputation?
Well, 1.9% of those treated with insulin experienced an amputation during the ten years, compared with 1.7% of those in the control group. Oops – There were slightly more amputations in the insulin treated group than in the control group.
So, why then are there doctors who think this study shows that it’s a good idea to give insulin to type 2 diabetics?
Well, it all comes down to one outcome – microvascular disease (the fancy name given to a set of diabetes complications that includes impaired vision, reduced sensation in the feet due to nerve damage, and loss of kidney function). 8.5% of participants in the insulin group suffered a microvascular complication, as compared with 11.6% of participants in the control group. In other words, the insulin treated group were 3% less likely to suffer microvascular complications after ten years.
That would mean that you could avoid a microvascular complication in one person for every 33 people treated with insulin for ten years. Which doesn’t feel like a very impressive effect to me, but the authors of the study spin this as a “25% reduced risk of microvascular complications” (yes, 8.5% is 25% less than 11.6%, when speaking in relative terms – read this if that makes no sense to you). It might be worth noting here that the study was part-funded by drug companies that make and sell insulin – not that that in any way influenced the authors of the study…
According to the authors of the study, the 3% (or “25%”, depending on how you choose to think about it) difference was statistically significant, but it wasn’t.
Warning – this next bit gets nerdy: The p-value for the difference was 0.015, which the authors consider statistically significant, because they’ve forgotten (as almost all medical researchers do), that the normal cut-off for statistical significance (p = 0.05) only applies when you’re looking at one endpoint. When you’re looking at 14 endpoints (as the authors of the study were), you’re supposed to correct for the fact that you’re looking at lots of relationships by dividing 0.05 by 14, which gives a p-value for statistical significance of 0.0036. The reason you have to make this correction is because, if you’re looking at lots of endpoints, then you’re going to get some that seem statistically significant just by chance, even though they’re not. Since 0.015 is more than 0.0036, the difference between the groups was in fact not statistically significant.
It’s also worth noting here that most of the 3% reduction in microvascular disease was actually a reduction in probability of getting laser therapy to the retina, i.e. was not connected with any noticeable symptoms but rather with a treatment that was given based on opthalmoscopic findings (people with diabetes regularly have opthalmoscopic examinations of their retinas to look for signs of disease progression). Note also that the study wasn’t blinded, so it’s not beyond the scope of possibility to think that the researchers might have influenced events in such a way that the insulin treated participants got less laser therapy than the control group, in order to improve their stats.
There was, in fact, no difference in visual acuity between the insulin treated group and the control group at ten years. Nor was there a difference in the proportion with absent ankle reflexes or knee reflexes between the groups (which is used as a sign of nerve damage in the lower extremities). So it’s highly questionable how meaningful the small reduction in “microvascular disease” was, even if it had been statistically significant (which it wasn’t).
So, the UKPDS study showed marginal reductions in heart attacks, strokes, and a questionable measure of microvascular disease among those treated with insulin after ten years of treatment, none of which was statistically significant. That is the entire basis on which millions of type 2 diabetics are currently being treated with insulin.
The rationale feels pretty weak to me, especially when you consider that there are downsides to insulin therapy – multiple daily injections, a need to monitor blood sugar continuously, weight gain, and an increased risk of hypoglycemia. In the UKPDS study, the annual risk of a serious hypoglycemic episode was 0.1% in the control group, but 2.3% in the group treated with insulin. That means that for every 45 type 2 diabetics treated with insulin for one year, one will suffer a serious episode of hypoglycemia (the definition of which is hypoglycemia so severe that it requires rescue by a third party or hospitalization). Note, that isn’t over ten years like all the previously discussed effects, that is over one year. And if we look at more minor episodes of hypoglycemia, 36.5% of participants in the insulin group suffered an episode of hypoglycemia per year of treatment (compared with 1.2% among those in the control group).
An even bigger trial was published in the New England Journal of Medicine in 2012. 12,537 people over the age of 50 with type 2 diabetes or pre-diabetes and cardiovascular risk factors were randomized to receive either long-acting insulin plus “standard care” or just standard care on its own. The goal was to see whether insulin would be useful as an adjunct in people who are at particularly high risk of cardiovascular disease. Among those getting insulin, the dose was raised until a target fasting glucose level of 5.3 mmol/l was reached. The participants were followed for an average of six years. So, what were the results?
Despite the fact that the insulin treated group managed to maintain lower blood glucose levels throughout the six year study period, there was no difference between the groups in terms of mortality, heart disease risk, stroke risk, amputation risk, or microvascular disease risk. Basically, six years of treatment with insulin provided no beneficial effect whatsoever, even though this was an older group at particularly high risk of cardiovascular complications (which should increase the odds of seeing a difference between the groups massively).
What can we conclude?
The evidence that exists really doesn’t support treating type 2 diabetics with insulin. It’s questionable if insulin provides any benefit whatsoever, and if it does, then the benefit is tiny and easily outweighed by the harms. As I’ve discussed previously on this blog, type 2 diabetes can be effectively treated with a carbohydrate restricted diet. In fact, a carbohydrate restricted diet is by far the most effective treatment in existence when it comes to type 2 diabetes, and can often reverse the disease completely. That is where physicians should focus their efforts.
Please support my work by becoming a patron. Your support is what allows me to dedicate time to producing new content. Regardless of whether you can contribute a lot or a little, it all makes a difference – the more patrons I have, the more of my time I am able to dedicate to this work. As a bonus, patrons get:
- Access to all the interesting discussions happening in my private forum.
- A private podcast feed with full-length versions of all the interviews I do.
- The ability to send me direct messages (I always respond to patrons).