Fasting Insulin & Diet ~ Time for some Cognitive Dissonance?

Reader Josh brings this study to our attention:

Huh?  They compared dietary intake (24 hour recall) for 1069 non-diabetics (assessed by administering a glucose tolerance test) with fasting insulin levels.  
Subjects were seen for up to three visits from 1984 to 1992. A 24-h diet recall and fasting insulin concentrations were collected at all visits....
... High total and saturated fat intake were associated with higher fasting insulin concentrations after adjusting for age, sex, ethnicity, body mass index, waist circumference, total energy intake and physical activity. Dietary fibre and starch intake were inversely associated with fasting insulin concentrations....
... Associations were similar in men and women and for active and inactive subjects, though associations of fibre and starch intake with insulin concentrations were strongest in lean subjects. 

I'd like to see studies of this nature involve better assessment of dietary intake (weigh/measure/log vs. estimate/recall), but we can't have everything.  The description of the subjects is in the methods section,  the average followup for the study was just over four years.  All subjects had a baseline fasting insulin and dietary recall, the majority had three, and some had two or only one.  They eliminated those who were diabetic at any followup.  This analysis did not report trends over time.  The subjects averaged just over 50 y.o.a. and BMI around 25 at baseline, 47% males/53% female.

In any case, this study is really pretty simple -- look at what a bunch of people habitually eat, break down the diet composition by carbohydrate and fat type, and see how this impacts a biomarker for insulin resistance, that being fasting insulin concentration.  First, the dietary compositions of the subjects are at right.  Granted this was in the mid-to-late 80's, but this is when the obesity epidemic was taking off, right?  I'm always struck by looking at actual data, however flawed, rather than things like NHANES estimates, etc.  Considering these were roughly 50/50 male/female, the daily intake of 8024 kJ works out to just over 1900 cal/day.  And check out the carbs there.  Not the crazy 300-400 g/day we're told, although they were 53g into insidious weight gain terroir by Grokian standards.  In any case, although accompanied by roughly 100g non-starch (aka sugar) carbs, the starch intake here is right in line with PHD, at roughly 100g = 400 cal.  No indication of the nature of said starch.  I would also note that the O6:O3 ratio is ~ 9:1 -- not ideal, but far less than the 15-20:1 and such we're often told is going on.  While we're on the fats, look at that fat percentage:  38%!!  Here I thought we all went low fat....

So, I think the next graphic will -- or at least should -- cause a little bit of cognitive dissonance for the disciples of L.Ron Rosedale and assorted denominations of TWICHOOB's.

Adjusted percent change in fasting insulin concentrations for a given difference in nutrient intake (shown in parentheses under each column). Calculated from the regression model including sex, ethnicity, physical activity, age, body mass index, waist circumference, total energy intake and the nutrient.
Now, I'm sure we'll hear that this would all go away if people ate even more fat and that fat was eaten without the carbs, but one of the knocks against blogs like this one, and scientific research in this realm in general, is that these studies produce no actionable information or relevant information.  I would say otherwise in this case.  It appears that "on average", these subjects were eating the SAD:  47% carb / 15% protein / 38% fat.  It's important to look at those p-values along the bottom row.  Usually statistical significance begins with p<0.05 which is why the authors describe the correlations as follows {I added the p's}:
High saturated fat intake (p=0.02) and low starch (p=0.0007) and fibre (p=0.008) intake, in separate models, were significantly associated with higher fasting insulin concentrations after adjusting for age, gender, ethnicity, vigorous activity, BMI, waist circumference and total energy (Fig. 1). High total fat (p=0.07) and linolenic acid (p=0.07) intakes were marginally associated with higher fasting insulin concentrations.

Within this SAD-eating population, fasting insulin levels decreased with 10g increments in starch intake and 5g increments in fiber intake, and increased with 5g increments in saturated fat intake.  Now, it's important to interpret this properly, this doesn't say that if an individual increases their sat fat intake by 5g, or decreases their starch intake by 10g they'll see their fasting insulin rise, and vice-versa.  But it does show, at least in this population, that diets higher in starch and lower in sat fat resulted in lower fasting insulin levels.    The linolenic (omega 3) acid bar is a bit shocking, even with the marginal p=0.07.  Also the trend for MUFA and oleic acid in particular trended towards increasing fasting insulin.  Here's what they had to say about this:

MUFA:  ... to the extent that people reporting high saturated fat intakes also eat more monounsaturated fats, the apparent adverse association of monounsaturated fat with insulin concentrations may be due only to its association with saturated fat intake (Table 3). When saturated fat was held constant by adding it as a term in the regression model with oleic acid (the primary monounsaturated fatty acid in the diet and more strongly related to insulin concentrations in Figure 1 than total monounsaturated fat), the oleic acid effect was reduced by 73 % to near zero (p = 0.74).
Omega 3 PUFA:   The effect of linolenic acid was only reduced by 18 % (p = 0.15) after further adjustment for saturated fat intake. Animal studies have suggested that linolenic acid in the absence of linoleic acid [O6] may improve insulin sensitivity. However, a scatter plot of linoleic acid intake by linolenic acid intake revealed that there were no subjects with high linolenic acid intakes in the absence of linoleic acid intakes to allow us to test this hypothesis.
I've actually seen some studies that finger O3's with decreasing insulin sensitivity.  Interestingly while paleo diets are lowER in O6's, many are not so much so as they might think if they're eating a lot of nuts and avocado without eating a lot of oily fish or popping fish oils.  But there's lots showing the opposite. 

Anyway ... also from the discussion:
There was no apparent interaction between saturated fat and fibre (or starch) intake, suggesting that the positive association between saturated fat and fasting insulin concentrations was present at all levels of fibre intake (data not shown). 
Perhaps this is why, in SAD-styled dietary interventions we don't see fiber doing its thing.  Next up, sucrose:
It was also of interest to explore the apparent lack of effect of sucrose on insulin concentrations, since sucrose has been implicated in the development of insulin resistance. When both carbohydrate and energy were held constant in the model, high sucrose intake was significantly associated with higher insulin concentrations. This suggests that eating sucrose instead of other sources of carbohydrate may be detrimental to insulin sensitivity.
This is in agreement with what most studies seem to show because the 50% fructose content of sucrose does not elicit an insulin response.  Those fructose and glucose bars off on the right of the figure above are a bit misleading, p=0.74-0.75!  
To further determine if this was due to beneficial effects of starch or adverse effects of sucrose, terms for starch, sucrose, fructose and glucose were simultaneously included in the model predicting fasting insulin concentrations. In this model, starch remained significantly and inversely related to fasting insulin concentrations, and sucrose, glucose and fructose were not associated with fasting insulin.

Let's go back to the combined sugar intake of this population.  Combining 1/2 sucrose + fructose, they averaged around 40g fructose per day.  Now this was the 80's, but to flirt with the 100g/day dose so often used in studies to produce some degree of IR, you would be talking a doubling of this intake.  Not unlikely if one is guzzling sodas, but many MANY of those who've gotten hyperinsulinemic/IR through the path of obesity, sugar sweetened beverages are not part of the equation.  More importantly, the beneficial effect of starch consumption remains.

OK, lastly, how about these associations taking BMI into account?  The figure at right shows the fat and carb effects for different BMI's.  The black bars = all BMI, the center-slashed bar = BMI of 22, right-dotted bar = BMI of 28.  For total and sat fats, the differences are not significant, but they are for all carb, starch and fiber (the p-values are for comparison between BMI levels).  It seems that starch and fiber are insulin sensitizing (lowering fasting insulin levels) over the range of BMI's, that persists, if significantly less in the highest BMI group.   What we do not see, that is important to point out, is where starch consumption negatively impacted insulin in the high BMI group.  

The evidence mounts folks.  Indeed, I've yet to see evidence supporting the notion that postprandial carb-induced insulin has any negative bearing on insulin sensitivity and basal insulin production.  Wait, we did have the Grey & Kipnis metabolic ward study where fasting insulin levels plummeted with an essentially ZC diet and skyrocketed with an essentially ZF diet (high sugar) ... but even those levels seemed to be returning to baseline after 3-4 weeks following the abrupt and extreme changes.  

The  "carbs causes insulin/hyperinsulinemia causes insulin resistance" is not supported by the research, or in an analysis of "real people" in this case.  It's time for those who propagate this mistaken notion to revisit the science.


Swede said…
"...although they were 53g into insidious weight gain terroir by Grokian standards."

Ha that is hilarious!

Tom Naughton would poo poo this study for its use of the recall method. While I agree broad recall questionnaires (how many servings per week do you consume of low fat dairy, for example) can be very inaccurate, a 24 hr recall is not that hard for most folks to complete with a reasonable degree of accuracy.
CarbSane said…
Yeah, recall sucks ... I really don't know why they can't have subjects weigh/measure and log for 3 days or something like that. Although, come to think of it, doing that may influence intake so it's kind of damned if you do, damned if you don't. Having a third party measure food provided and consumed would be better, but cost prohibitive for this number of subjects I would imagine. Here's one example of a better recall protocol. Not perfect, but when you consider over 1000 people doing this, for up to 3 visits, it would be difficult to explain away how that starch correlation persists controlling for all manner of other variables, for example. Or that they would preferentially under report carbs and overreport fats -- mid-late 80's I would think the opposite if anything.
Sanjeev said…
Evelyn: Jimmy Moore wimps out on Oz invite!
... Maybe he was afraid they would put Jessica Biel up to oppose him! ;-P
JM: At least I know I'm on their radar screen. (HUGE SMILEY)

So he's happy to be on their radar screen but refuses to get on the scale ... afraid to break the scale Jimmy?
Sanjeev said…
Tom Naughton would poo poo this study for its use of the recall method
but ... Naughton's GOD never heard of doubly labeled water studies ... no such thing in GT's vocabulary,

and the man that GT worships for his facility with English nastiness is forever telling everyone that if you write "doubly labeled water" you are committing malicious slander against the obese and accusing them all of lying, lying, lying.

These folks are quite the collection.
Josh said…
Nice summary. It would have been interesting if the study had recorded the sources of starch. I couldn't find any stats on starch consumption in USA in the 80's but I'm guessing that the majority of starch consumed in the study was likely from grains, most likely refined grains - wheat, corn, rice? If that was the case then when it comes to starches perhaps it's better to be unsafe than sorry.
MM said…
Funny, this looks a lot like the diet I lost weight on. I don't know the detailed breakdown, but I think I ate slightly more protein, less sugar and less polyunsat oil. The only veg oil I have in the house is olive oil. However, the macro-nutrient breakdown is very similar. These were my usual numbers: 45% carb, 17% pro, 38% fat. I was really surprised when I plugged it into fitday and saw this breakdown just about every time I did it. I don't know why it worked for me, because it seems very much like it shouldn't have. I didn't meticulously count calories or anything, but I was eating less calories on what looked like a SAD diet than I was on VLC without really trying too hard. It wasn't very fast weight loss. It took me a year to lose 20 pounds, but it was relatively painless. It still boggles me, and I guess I don't really have a point. Just sharing.
CarbSane said…
@Sanjeev: I tweeted in response to Beth that I misread his reasoning. I have no idea why Oz would contact Jimmy Moore to be the "expert" arguing against the notion of food addiction. Why not an Emily Deans, or Stephan, or even Andreas? How would Oz describe Jimmy to his audience? I don't geddit!

@MM: Thanks for sharing that. It's funny, because Zone and other middle of the road macro diets don't seem to perform as well as extremes in head to head comparisons, and yet you are by far not the only person who has reported success eating this way. I think Stephan's diet broke down somewhat similar as well.

My PHD-ish diet works out just slightly lower carb, higher protein than they advocate, and not too far from these SAD proportions as well.

Which kinda goes to show that macro percents are not really what causes obesity (Andreas Eenfeldt blaming "replacing" fat with carb is nonsense, IMO) or affects weight loss either.

For those who haven't been here that long, MM has shared that she was stuck with weight loss for almost a year doing VLC. Sound familiar?
CarbSane said…
Hi Josh, Yeah, a breakdown of various foods might have been nice. Having "come of age" around that time, it does not appear to me that there was any major shift in starch consumption from the 70's to the 80's ... except perhaps more in favor of potatoes over wheat. I suspect that because fast food french fries, the supersize and improvements in frozen varieties, really took off in the 80's, and the lion's share of increased veggie intake in America is attributable to potato consumption - mostly fries. And yet the O6PUFA + starch combo that is the modern veggie-oil deep fried french fry doesn't seem to be so bad??
P2ZR said…
Wow, I am still scratching my head over the huge difference in response to fiber between BMI cohorts. Forget p-val; just eyeballing it, even having the gigantickest error bars would still result in significance. So much for groks who aim for weight loss nibbling on fibrous veggies for the most favorable (or least nefarious?) effect on insulin. Or is that notion supposed to be wrt. postprandial instead of fasting levels? But I do wonder what'd account for the difference. I mean, any 'adaptation' or lack thereof to fiber would mostly concern the GI tract, no?

Also, that SFA/MUFA thing was quite the revelation. Never thought about how even tallow is ~50% sat fat, and butter is 65%ish; negligible PUFA's means the remainder is an astounding amount of MUFA. Funny how the whole HFLC in-yo'-face 'lookit all the sat fat I'm pounding!' is really um, a whole lot of heart-healthy mono fats ;)

Also, thanks for adding the blogroll! Very very minor quibble (others are of course free to disagree)--while the auto-scrolling looks nifty, and will stop if hovered over, it feels a bit awkward to me. I'd prefer a stationary list for the purposes of navigation, though a manual scrollbar would be a-ok if you want to avoid a mushrooming list. (But please, please, whatever you do, do not let it grow to the size of Paul's! Love PHD blog, but the size [and variation in quality] of that blogroll is just plain intimidating.)
P2ZR said…
^I'm also assuming that most of the fiber wasn't soluble, so not too much of the soluble fibers => butyrate (yippee!) thing going on. Somehow I don't think this study involved an oatmeal-fest. Whatever, I'm still trying to figure out what my 'goldilocks' amount of fiber is....
P2ZR said…
^argh, that should read 'the' goldilocks amount, not 'my'. Inadvertent TMI-sounding awkwardness.
CarbSane said…
Hi Sarah, yeah, the fiber thing is rather surprising. I'm going to be blogging a bit on MUFA (probably will have to wait to 2012). Looking into the Rosedale stuff led to some interesting discoveries. Lower sat fat (total and percent), higher carb, higher MUFA percentage, higher fiber (I suspect butyrate) = better glucose metabolism. Which is not to say sat fat is bad, but I don't think it's "bring it on" good for us either.

As to the blogroll, it's been a long time coming. I didn't want to take up a lot of space for it and didn't want to put all blogs I read (a LOT more than are there). Still, I wanted to be more "social" and acknowledge where others acknowledge me and "give back". A scroll bar means either only the low alphabet blogs show all the time, or somesuch. That's why I liked the rolling idea ... it's not too distracting I don't think.
Anonymous said…

Wish I could read the actual study!
P2ZR said…
@Evelyn - "scroll bar means either only the low alphabet blogs show all the time". Oh yeah, that's definitely true and I didn't think of that, sorry. By distracting, I meant that I wanted to be able to see the whole blogroll on my (not its) terms--but I'll deal with it (especially because apparently it's far from exhaustive, anyway!) :)

@euler - Yay, thanks! (But why is quick poking around not yielding institutional access?) RS hmm...reminds me of that epic commenter on PHD blog whose page was or something. I still don't understand where massive amounts of soluble fibers/RS would come from in any mildly carb-conscious (<200g/day?) diet, but hmmm!
LeonRover said…
Well, Evelyn, I guess I've got Statistical Dissonance from the analysis provided by Marshall and Hamman - they being the Biometricians on this study.

I fail to see to see why consumption changes - 20 gms Starch or 20 gms Fat - which predict a change of 2% Fasting Insulin otherwise an absolute amount of 0.2 mU/l are relevant to Diabetic Endos, the target audience of Diabetologica.

My understanding is that levels of 3 to 6 are "good", levels > 13 are hyper, while levels 6 to 13 are pre-hyper.

Thus to go from 7 mU/l to 6 mU/l one must change one's average diet by reducing 100 gm fat or increasing by 100 gram starch.

Have I got this wrong?
bentleyj74 said…
The actual break down of peoples comfortable macro combinations continues to be interesting to me. I for one could have fallen over after discovering that those raw vegans I so enjoy disdaining are only off my ratios by about 5-10% on average.
MM said…
Actually, I was stuck for over a year. It was more like a year and a half. I don't know why I kept at VLC so long. I Believed it should work I guess.
CarbSane said…
@Leon: I don't think the study has much bearing on health and such, because as you point out, we're talking small differences between normal. However, if the point is to minimize basal insulin levels (I don't think that should be the point per se, but lower basal insulin is indicative of better insulin sensitivity), this study shows that eating starch apparently does just that.

I think the "carbs made me IR/diabetic" meme is more dangerous than anything else that was birthed from GCBC ... oh wait ... L.Ron's been saying that for 20 years, I should give credit for misinformation where credit is due.
Galina L. said…
Nowadays I try not to comment about things that involve counting calories because I don't count, and my opinion about the subject would be a guess. Similarly, I don't know how my insulin level was affected by my diet because it was never measured, unlike lipid panel, for example. However, I based my weight-loss strategy on the insulin theory trying to lower insulin - restrictive carbs, limiting the amount and the size of my meals, practicing IF. What could be the non-direct indicators of insulin going down? Lowering blood pressure, loosing water, raising FBS? I thought the leg edema was gone
(unless I eat some pasta) after I started LC because of the level of insulin became lower.. Before middle age, I lost several times 30 - 50 lb of weight just by counting calories, after 45 I couldn't do it anymore, my guess it was because I became pathalogicaly IR. Also, one of strategies to lower FBS is based on increasing insulin level by eating something before going to bed for healthy person, or getting some insulin shot for D2, if I remember it right. Such practices suppose to contradict results of the study. So, again my question the people on the blog " What could be the non-direct indicators of insulin going down or up?"
CarbSane said…
@Galina: I think your post indicates what's wrong with the insulin theory. Lots of misinformation out there. This was sort of the point of this post because it is one of numerous studies showing that carb consumption -- e.g. postprandial insulin "spikes" -- are not directly related to basal insulin levels. Indeed, when the T2 diabetic progresses, they have an insufficient pp-insulin response yet remain chronically hyperinsulinemic.

TWICHOO would predict that, however small the differences, the starch eaters would have higher fasting insulin and the fat eaters, lower.

I have a theory that I'm going to blog on as to why LC seems to work better (at least initially) for the IR vs. a HC diet.

Fasting blood glucose levels are largely a manifestation of hepatic insulin resistance and/or insulin production capacity, mostly the former as I understand it. Personally I think all metabolic panels should include NEFA and insulin along with BG. I'm also coming across a wave of studies indicating that the postprandial metabolism may be better to assess (both BG and lipids) to catch dysfunction sooner. As I understand from what folks report on the internet, an OGTT is not always accompanied by measurement of insulin levels. It should be!

Elevated fasting insulin is a measure of IR, absent something like insulinoma. While one can alter this, I'm not aware of something overnight that would work. FBG on the other hand can be manipulated, an insulin injection would do the trick. I think glucose metabolism should be assessed by the trifecta HbA1c, FBG and an OGTT AND fasting and pp insulin.
Frugal Jen said…
So maybe it's not the insulin, it's the orexin system?

(Would love to see your take on this one?)
Galina L. said…
I experimented with FBS , and it looks like it gets 15 - 20 poins down if I snack very close to my bed time. I don't think everybody will get the same result, as usual. I know there are more things influence FBS than insulin, so it is difficult to say something very conclusive. Body works with great complisity, there are always things to overlook.
I agree with the necessaty to improve the assessment of metabolism, it just I don't see right now any indicators it is going to be changed soon, at least on GP level. So far people on their own.