Monday, September 06, 2021

Modelling energy intake: Canagliflozin

It is quite possible to make a very reasonable estimate of how many calories a given person has consumed over the previous few weeks by estimating their total energy expenditure (TEE) using doubly labelled water (DLW), estimating the calories supplied from fat in adipose stores using the changes measured by DEXA scanning and applying a little arithmetic:

TEE (by DLW) - Fat mass change (by DEXA) = Food derived calories

Nice and simple. And very, very expensive.

Quite a few years ago Kevin Hall's group had the idea that you might be able to reverse engineer the intake of food calories simply from the change in weight over time using a mathematical model. They validated this against a two year conventional diet study where weight, TEE by DLW and fat mass changes by DEXA were repeatedly measured. They produced this paper:

Their model is pretty good within certain limits. You could trip it flat on its face pretty easily but that's not today's post. Just assume it works in the above study and also in this one:

The second study piggybacked on a diabetes study using canagliflozin, a sodium glucose co-transporter inhibitor which increases urinary glucose excretion. Canagliflozin produces the loss of around 90g/d of glucose, ie around 400kcal/d. This loss is insensible, other than via counting the number of trips to the bathroom. There was no specification within the study protocol to lose weight or to restrict calories.

Long-term efficacy and safety of canagliflozin monotherapy in patients with type 2 diabetes inadequately controlled with diet and exercise: findings from the 52-week CANTATA-M study

The interesting questions are whether this silent caloric loss produces weight reduction, what does it do to caloric intake and what mechanisms might be at work.

Here are the weight changes:

So. Obviously losing 400kcal/day does produce weight loss. Or is that genuinely obvious? Surely, if the hypothalamus wants to "see" a certain number of calories to run metabolism, shouldn't it immediately increase calories eaten to counter that 400kcal deficit? Yes, it should. Immediately. Except...

Here is what happened to the energy intake. The solid black line is Hall's model which does not include the starting point at time zero with weight change zero. I've added the red curve to include this and roughed in the rest of the data points as well as a curve in powerpoint can manage:

It's quite clear from the data points that there was an initial drop in total energy intake to a nadir, somewhere within the first three weeks. DLW is an averaging technique so the location of the nadir is an unknown but it must have happened, to explain the data points generated where week three is below time zero.  The effect is more marked in the placebo group and probably represents simply being in the trial and tidying up, in both groups, the worst of the normally execrable diabetic diet prior to starting the study.

The effect is blunted in the canagliflozin group, presumably because of those 400kcal/d down from day one and their hypothalamus will have noticed this and have kicked their cortex in to doing something about it (hunger). By 15 weeks the extra calorie intake estimate (around +350kcal/d) is getting pretty close to the urinary calorie loss estimate (around -400kcal/d). 

But for the first 15 weeks calorie intake was estimated to be well below urinary calorie loss. Food was ad libitum. Why any weight loss?

That's interesting.

Also, despite increasing food calories to match urinary losses, weight remained stable at over three kilograms below baseline, with no suggestion of weight regain at the end of a year.

That's interesting too.

Hall goes on to treat the changes in weight as an engineering control system, a bit like a black box, without any attempt at integrating any basic physiology. A quick search of the text shows no mention of insulin in the whole paper. Not surprising, given the stance taken by Hall over the CIM of obesity.

But even the most basic, strawman-facilitating version of the CIM of obesity has no problem explaining the results in some depth. It takes about 30 seconds on PubMed to ascertain what canagliflozin does to the insulin requirement of people with DMT2.

It drops the requirement.

For patients still using their own pancreas for insulin this seems very likely to simply be reflected in a spontaneous fall in plasma insulin, triggered by the loss of 90g/d of glucose which exits through the bladder rather than requiring insulin to stuff it in to storage within the body.

If we assume insulin drops by a fixed amount in proportion to 90g less of glucose, and stays at this reduced level for as long the canagliflozin is given, there will be an acute rise in lipolysis which will supply adipocyte derived calories to partially make up for the urinary loss.

As the hypothalamus monitors energy status it will see 90g/d of glucose as absent but being replaced by, initially, roughly a kilo of fat from adipocytes over three weeks. More arithmetic:

400kcal glucose x 21 days = 8400kcal deficit from glycosuria.

Weight loss of 1kg over three weeks = 9000kcal of fat from adipocytes.

I would suggest that fat loss comes as a direct response to lowered insulin levels and will easily at least partially replace the glucose loss, certainly initially. The fat loss can be described as "calories-in" without actually eating them. So people with an acutely lowered insulin level eat less than you would expect.

Let's look at this the correct way round. An all-glucose caloric deficit of 400kcal/d was acutely established which directly resulted in rapid drop in plasma insulin levels. Lipolysis was acutely increased which largely offset the glycosuric calorie deficit. Because over several weeks lipolysis gradually slowed to an appropriate level determined by the the new insulin levels, food calories had to increase in proportion, to maintain an adequate energy flux to keep the hypothalamus happy. Eventually extra food-in will equal urinary glucose-out giving stable weight. But with lower insulin levels this will occur at a lower total fat mass.

The weight loss/calorie intake deficit were both caused, directly, by a fall in insulin levels. Utterly simplistic CIM.

Kevin Hall is a great source of data. Of insight?

Not so much.



Passthecream said...

Peter, v. interesting, Ta! Is canaglifozin one of those Scgt inhibitors which sometimes make the private parts wither?

Peter said...

Fungal necrotising fasciitis sounds really fun. Not.

But obviously preferable to skipping 90g/d of carbs from your diet of course...


Bob Minty said...

Peter: Very keen observation, the obvious neglect of serum insulin by mainstream researchers.
I was able to find a single study that accidentally measured insulin as it was tracking SGLT2 inhibition association with NLRP3 inflamazone.
As you would expect there was a measurable reduction in endogenous insulin production, and also increased BHB levels.
Good work.

Unknown said...

"Effect of canagliflozin on N-terminal pro-brain natriuretic peptide in patients with type 2 diabetes and chronic heart failure according to baseline use of glucose-lowering agents"

raphi said...


Stephan Guyenet captures what the mainstream of obesity research thinks of insulin:

"[...] insulin does impact body fatness, even if only modestly. It is consistent with a model whereby multiple factors, including post-meal insulin release, determine body fatness"

Peter said...

Hi Unknown and Bob,

It seems quite respectable to look at what dropping 90d/g of glucose out of the metabolic milieux does. Whether via insulin, glucose levels of the PUFA enhancing effects of the previous two, taking out glucose seems to work. People can pick at the details of metabolism under canagliflozin without admitting they are studying LC...


Peter said...


You have to feel for The Good Doctor. He really lacks a framework to understand anything. Yet some of my most important insights stem from him, he introduced me to the concept that insulin resistance might be a physiological adaptive mechanism. I've no idea what he thinks nowadays...


raphi said...


Right! His choice of topics and papers are good enough to make me want to keep up with him every once in a while, despite the insight-free analysis he provides.

cavenewt said...

One of the daily newsletters I get is STAT News, a pretty mainstream pharma-centric source. So this morning I was interested to see a guest article by Gary Taubes announcing a new review article in the American Journal of Clinical Nutrition,in which he's a co-author.

"How a ‘fatally, tragically flawed’ paradigm has derailed the science of obesity"

The review article has 16 co-authors, with some familiar names including Ludwig and Volek…and, surprisingly (to me at least), Walter Willett.

"The carbohydrate-insulin model: a physiological perspective on the obesity pandemic"

Passthecream said...

Thanks for the links Cavenewt, that's an interesting essay in the hyperphagian mode. But it neatly completely sidesteps the question of " heart healthy " polyunsatured fatty acids so it could risk being lumped in with the other dogmatic thinking he argues against!

I'll read the paper though, maybe I'll have to eat my words. Fingers crossed that they aren't fattening.


cavenewt said...

In one of my comments on the STAT News article I did express my disappointment that Gary does not seem to have had time to look into polyunsaturated fats yet. I heard him say so in an interview once, and later, he's said he doesn't think there a problem, IIRC.

It isn't all one or the other; a lot of us who went low-carb years ago, but didn't bother to avoid seed oils, still saw vast improvement, so I still think Gary's message would be very helpful for the majority of the population that is subsisting on Twinkies and Coke.

raphi said...


I wouldn't go so far as to say Gary thinks seed oils aren't a problem "Refined grains and fructose-rich sugars would be my guess. Some add vegetable oils. Some argue it's glycemic index effect. I think the sugars/sugary beverages dominate"

'Dominate' being the operative word.

He's also mentioned elsewhere (can't find the tweet unfortunately) saying that he's uncomfortable with the level of certainty regarding the obesogenic effects of seed oils displayed by people like Paul Saladino. I've never heard him directly say it's a bad hypothesis or that there's 'no evidence'. I think he's simply being cautious and knows the carbs angle way better.

totally fine as far as I'm concerned..

Peter said...

cave and raphi,

I've yet to get the time to read either article so no comment as yet beyond that I really, really struggle with fructose.


cavenewt said...

raphi—I remember him mentioning it twice, and you've summed it up pretty much as I remember. Perhaps "not a problem" is putting words in his mouth.

raphi said...


Here's Gabor's report "Physiology and limitations of fructose absorption: considerations for sweetener formulations in beverages"


He hasn't spoken about LA publicly at length, so who knows really.. maybe one of us can get the question posed to him in an upcoming podcast!

Passthecream said...

I always have at the back of my mind the carbosis essay of Denise Minger wherein she referenced Walter Kempner's therapeutic diet of white rice flavoured with sugar and the occasional fruit juice for light relief, which seems to have worked for many people. They lost weight, got better. There were some other extreme diets she wrote about then, just before Peter wrote Protons (38). Peter also spoke/wrote recently about the possibility of a predominantly carb based diet --- assuming you have a functional set of beta cells etc.

DM doesn't seem to have written much publicly since then, not on her blog anyway. Was there a hate storm from the Ketone Cops? I'm still waiting for part 2 of that essay ...

raphi said...


Carbosis was clever and it forced (some) low-carbers to face their paradoxes. I have no idea what she's up to.

cavenewt said...

@Passthecream—Apropos of not much, "...forced (some) low-carbers to face their paradoxes". Things only seem paradoxical in light of one's preconceptions. Many (most?) of these paradoxes have to do with the metabolism dealing with extremes, at least some of which are evolutionarily novel.

In the context of Peter's next post, "Modelling energy intake (2): Corn oil", conducting in vitro or in vivo experiments that involve extremes can be academically enlightening but not necessarily applicable to normal life. That 1973 study is a great example. Superficially, the average person would think olive oil is fattening and corn oil will make them lose weight. (I have trouble imagining chugging 20 ounces of corn oil a day without feeling nauseated.)

Passthecream said...

Cavenewt that was Raphi with the paradoxes. Personally, I mostly have confusion.


LA_Bob said...

I can understand Taubes's "seed-oil hesitancy". He wrote about diet-induced corpulence and disease from carbohydrate during a time in history before seed oils were important. Think Western diets and the natives in African and Asian colonies in the late 19th and early 20th centuries. I'm sure linoleic acid played very little role then, but the natives who ate the refined carbs of the colonists gradually developed European disease profiles.

So, if Taubes (and my reading of Taubes) is right, what do seed oils do if not cause obesity? Turbocharge it?

And, yes, I realize that's separate from their peroxidation, carcinogenic, and ARDS problems.

cavenewt said...

LA_Bob—you're right, and I'm sure they're both big factors. "Turbocharge" is exactly the impression I have, it's like throwing gasoline on a fire.

You're probably aware that lard started being adulterated with cottonseed oil (although to what extent I can't remember) in the 1860s. Native Americans stuck on their new reservations were given flour, lard, coffee, and sugar, and not much else to eat.

A lot of us here have become fixated on seed oils, and since we've been told to switch to PUFAs for decades now the effect will certainly be not insignificant. I just wistfully hope that Gary will eventually get around to it with his usual reasonable thoroughness.

Passthecream said...

LA_Bob, Cavenewt,

Tucker Goodrich wrote a good piece about the history of lard vs cottonseed oil but I can't track down the link atm.

This is an interesting collection of writings ( propaganda + history) about flax seed that I stumbled on a while ago. Flax seed and meal has been a part of the animal feed chain for a long time so pufa have been finding their way into the human food supply ditto.


(Interesting btw to read in the Eastman chapter how vast quantities of the oil were shipped to the Soviets as food during/after WW2. That could explain a few things.)

So it's not just cottonseed which was an ' adulterant' in the 19C and that probably wasn't even thought of as such merely cheap calories. Aside from the gossypol problem. Not to be confused with the glossy coats due to fkaxmeal.

LA_Bob said...

Hi, Pass,

Just to be clear, cave and I were talking about omega-6 fatty acids (linoleate) finding their way into our diets in quantities beyond our ancestral exposure. I'm not sure omega-3 laden linseed meal fits that category.

Also, I'm not sure it's exactly fair to consider linseed meal fed to the animals in the same way cottonseed oil was added directly to our own diets. Or cooking oil and margarine today.

My point, and cave seems to agree, is that Gary Taubes doesn't need linoleic acid to explain "bad calories". History shows refined and highly digestible carbs work just fine.

Interesting to wonder, though, if the linseed meal has any adverse consequences to us beyond allowing producers to claim they're selling a better o-6 / o-3 ratio.

JR said...

Hi Pass, link you were looking for. O-6 were present earlier than we realize. Knobbe has also been investigating the history of o-6 use, coincidencing with his speciality maculatory decease.

It would be high time for taubes to get interested in seed oils...

Passthecream said...

Thanks JR!

LA_bob, the early mechanical pressing methods were not efficient and left a lot of oil in the seed cake. LSO is up to 60% triply unsaturated ALA and 20% la which is quite respectable, plus some mono OA. After reading G. Spitteller's conclusions I don't trust any of the multiply unsaturated fatty acids. From my point of view as an amateur varnish maker ALA is the best thing since sliced bread. It is much more potent as a drying oil than linoleic acid.

cavenewt said...

Thanks JR. That Tucker post's where I gleaned the info in my earlier comment.

Passthecream, in the straw bale house I'm building the floors are poured adobe (sand plus clay plus straw). After drying, the adobe is saturated with many coats of boiled* linseed oil. Over a period of months it polymerizes into something resembling linoleum, which is, I believe, where LINoleum originally came from. So it's good for more than just varnish — it makes great floors!

* They don't actually boil it anymore but treat it with some combination of chemicals that has the same effect.