More on the Mechanisms of the Glycemic Index: A Fatty Acid Roller Coaster?

A Continuation of The Mechanisms of the Glycemic Index: A Fatty Acid Roller Coaster? ....

Quick summary of Ludwig's mechanism for high-GI making us fat v. 2002:

High GI carb causes glucose and insulin to spike and fatty acids to plummet early on.  Then glucose plummets resulting in hypoglycemia and counterregulatory hormones kick in.  These bring glucose back or slightly elevated and cause fatty acid levels to rebound to levels reminiscent of a long fast making the person hungry (hypoglycemia) and hungrier (feeling fasted) so they eat more.    

The Mechanism of High-GI ~ 2012

by Davis S. Ludwig

In 2012, with colleague (often listed as a co-lead investigator) Cara Ebbeling and others, Ludwig published:  Effects of Dietary Composition on Energy Expenditure During Weight-Loss Maintenance.  In JAMA.  I only mention this study here for two reasons.  First, to demonstrate that the 2002 review paper was not by then forgotten.  Second, as will be discussed towards the end of this post, the mechanism changed despite similar results that should have lead an unbiased researcher in a different direction ... or at least caused him or her pause.  From the paper:

Obesity treatment should emphasize behavioral methods to foster and maintain decreased energy intake. Several recent clinical trials indicate a direct relationship between dietary adherence and weight loss, regardless of dietary treatment group assignment.5-7

The diet comparison studies cited are Dansinger, Foster, and Sacks . This has pretty much been the conclusion of just about every dietary clinical trial.  Adherence predicts success.  And yet we still get treated to further waste-of-time comparisons.  

However, because metabolic pathways vary in energetic efficiency, dietary composition could affect energy expenditure directly by virtue of macronutrient differences or indirectly through hormonal responses to diet that regulate metabolic pathways.8,9

Reference 8?  Well, who else but Feinman and Fine.  I suppose one should be thankful it's not to the abominable First Law Violates the Second Law paper, rather Thermodynamics and Metabolic Advantage of Weight Loss Diets.   Reference 9 is, as you've probably guessed, Ludwig's 2002 Glycemic Index paper.  Although the focus had shifted to energy expenditure (for reasons I'm not exactly sure why), it is important that THIS review was cited.    Out of that larger study in 21 subjects, 8 were chosen for a sub-study resulting in the following paper:  Effects of Diet Composition on Postprandial Energy Availability during Weight Loss Maintenance.  In PLOS One.

I've mentioned this study before, because its outcome puts yet another nail in the coffin of the intellectually dishonest argument of "internal starvation".  See here (though I'm repeating pertinent points in this post).   Does anyone reading this care to guess where I first learned of Ludwig's 2002 Glycemic Index review?  Anyone?  Bueller?  Bueller?  

Yep!  This "Energy Availability" paper.    

Circulating levels of the major fuels the body used for metabolic processes, including glucose, free fatty acids (FFA), and ketones, are tightly regulated by hormonal mechanisms. When circulating levels are high, insulin promotes deposition of glucose and fatty acids into muscle, liver and adipose and suppresses their production and release from storage sites. Conversely, when circulating metabolic fuels are low, counter-regulatory hormones (especially glucagon, and also cortisol, epinephrine and growth hormone) stimulate lipolysis, glycogenolysis, gluconeogensis, and ketone formation [1].

1. Ludwig DS (2002) The glycemic index: physiological mechanisms relating to obesity, diabetes, and cardiovascular disease. JAMA 287: 2414–2423.

Many studies have shown that circulating levels of the individual fuels affect appetite [2,3,4,5]. In both rodent [2,3]

2. Friedman MI, Granneman J (1983) Food intake and peripheral factors after recovery from insulin-induced hypoglycemia. Am J Physiol 244: R374–382.3. Scharrer E, Langhans W (1986) Control of food intake by fatty acid oxidation. Am J Physiol 250: R1003–1006.

and human [4] models, hypoglycemia and decreased FFA levels lead to increased food intake.

4. Ludwig DS, Majzoub JA, Al-Zahrani A, Dallal GE, Blanco I, et al. (1999) High glycemic index foods, overeating, and obesity. Pediatrics 103: E26.

Limited data also suggest that ketones, such as betahydroxybutyrate (BHB), decrease appetite [5].

5. Chearskul S, Delbridge E, Shulkes A, Proietto J, Kriketos A (2008) Effect of weight loss and ketosis on postprandial cholecystokinin and free fatty acid concentrations. Am J Clin Nutr 87: 1238–1246.  

However, these studies of individual fuels may not provide a comprehensive view of the metabolic regulation of hunger, because the body can utilize a varying mix of fuels under varying dietary conditions. Indeed, numerous popular weight loss diets have advocated specific macronutrient prescriptions, in part because of their intrinsic effects on metabolism and hunger [6,7,8,9].

We previously proposed that high glycemic load diets reduce availability of metabolic fuels in the postprandial period by eliciting a high insulin to glucagon ratio, leading to excessive hunger and overeating [4].

4. Ludwig DS, Majzoub JA, Al-Zahrani A, Dallal GE, Blanco I, et al. (1999) High glycemic index foods, overeating, and obesity. Pediatrics 103: E26.

The purpose of this study is to examine a novel measure of total circulating energy availability (EA), defined as the combined relative energy density (circulating level6relative energy content) of each of the major metabolic fuels, in overweight and obese young adults consuming diets ranging widely in macronutrients. Specifically, we propose that EA will be lower after a high glycemic load (low-fat, LF) diet in the postprandial period, compared to diets with a moderate glycemic load (moderate fat, low-glycemic index, LGI) or low glycemic load (very low carbohydrate, VLC).

For 8 of the 21 subjects from the original study, which were fed the calorie matched weight maintenance diets for 4 weeks each in random cross-over fashion, postprandial levels of various metabolic fuels and hormones were measured at the end of the fourth week of each phase.  After overnight hospitalization during which diet-appropriate dinner was fed, subjects were fed a test meal of 25% TDEE (intake in this phase averaged ~2600 cal/day), approximately 650 Calories.   Here are the test meals standardized to 500 Calories each.

As an aside, I wonder how much isolated fructose the LGI group got during the course of this study.  Despite citing Ludwig's 2002 Glycemic Index review at the start, somehow the postprandial period tested in studies from his research group seem to end at the 5 hour mark.   

Blood Glucose Roller Coaster?

dot = LF , square = LGI , triangle = LC

Just in case it matters, despite the high glycemic index, and liquid nature of the LF diet (aka the HGI diet), there is no "blood sugar roller coaster", and glucose simply returned to baseline by the 5 hour mark.  By contrast, the LC dieters started at a nominally higher  glucose level, and go "hypoglycemic" at 90 minutes after the meal.  It's not anything that significant, but in this study, it was the LC diet that caused a metabolic fuel crisis on the blood glucose front.

From the paper:  

• In the early postprandial period the three diets differed significantly (p = 0.0002), with glucose higher with the LF diet than the VLC diet (p = 0.0001).
• There was no effect of diet on the level of the glucose curves in the late postprandial period (p = 0.66).

Fatty Acid Roller Coaster?

Here are the fatty acid profiles for the three diets.  The FFA/NEFA levels are roughly 50% higher at baseline.  Good?  Would you prefer the glucose levels be 50% higher?   The difference may not be clinically significant (or it might be), but if anything, metabolic health is associated with lower NEFA levels, not higher.  

On the right hand side I "translated" the LF line upwards so that all three plots are "standardized" to the same baseline.  By 90 minutes, it is the LGI diet that sees NEFA plunge by around 50% more vs. baseline while the LF and LC NEFA are suppressed similarly (somewhat surprising).  Now from there, the fatty acids make a come back quicker after the LC meal, but there they go rebounding above baseline -- just as Ludwig posited happening, only for a high GI meal. (Don't look now but the LGI meal also rebounded above baseline.)   So again, the mechanisms put forth by Ludwig in 2002,  by which High-GI meals elicit hunger, seemed to occur following the LC and LGI meals in this study.   Meanwhile look at that nice smooth fatty acid profile for the LF diet.  {grin}

Only the absolute trends were discussed in the results:

• ... there was a significant effect of diet over the whole curve (p = 0.0001), and a diet x time interaction (p = 0.0002).   Over the whole curve, the FFA level was lower with the LF diet than with the LGI (p = 0.0001) and VLC (p = 0.0001) diets.

That Novel Energy Availability (EA) Measure

For the full effect, I used the same color coding for Early, Mid and Late postprandial period.  Stacked vertically are the major metabolic fuels on the left, insulin and glucagon on the right, and total energy and hunger ratings in the middle.  I'm going to ignore the ketone (BHB) data as the concentrations are quite low across the board and unlikely to influence the EA.  

From the article:

• Insulin level was affected by diet in the early postprandial period (p = 0.0001), being higher with the LF diet than the LGI (p = 0.0001) and VLC (p = 0.0001) diets, and higher with the LGI diet than the VLC diet (p = 0.0001). 
• There was no such effect [on insulin] in the late postprandial period (p = 0.84). 
• Over the entire curve, glucagon was higher with the VLC diet than with the LF (p = 0.0001) and LGI (p = 0.0001) diets. 
• For cortisol, there was ... no effect of diet (p = 0.37) and no diet x time interaction (p = 0.72). 
• For epinephrine, there was no significant effect of time (p = 0.27) or diet (p = 0.20), and no diet x time interaction (p = 0.39). 

• Hunger  did not differ  significantly among diets in either the early (p = 0.87) or late (p = 0.86) postprandial period.

There are a few things that jumped out at me when I first saw this.

• There is NO DIFFERENCE in the hunger ratings between the diets at any time point.  Yes, I just highlighted that quote from the text above.  It bears repeating.  With emphasis!
• The trajectories with which the hunger rating profiles most closely track are those for NEFA.  If anything, reduced hunger is associated with suppressed NEFA in the postprandial period. 

From the Discussion:

In this study, we examined the postprandial effects of three common dietary patterns in young adults during a period of weight maintenance after weight loss. We proposed a novel outcome measure, EA, which combines the relative energy available from three major metabolic fuels (glucose, FFA, and BHB) and hypothesized that an LF meal would lower total metabolic fuel availability and energy expenditure in the late postprandial period by causing a high insulin to glucagon ratio secondary to a high dietary glycemic load. 

As hypothesized, the LF meal led to lower EA in the late postprandial period compared with the LGI and VLC meals.

This is not technically true.  The EA was not higher in the LF group after the meal, but this is because it was lower before the meal.  It is misleading to use the phrase "the LF meal led to" this observation/result.  Instead, it should have been noted that the baseline EA was lower for the LF diet, and this difference was essentially sustained throughout the postprandial period.

I did my best to translate the LF line upwards to meet the others at baseline and plot "change in EA". 

It is important to keep the units in mind as well.  They report a concentration in kcal/L ... ONE kcal = ONE Calorie!!!   An average adult has about 5 liters of blood.  So:

• The max increase in total EA is for the LF diet, let's call that 1.4 x 5 = 7 Calories.
• The max suppression in total EA is for the LGI diet, let's call that -0.7 to -0.8, at most - 4 Calories
• What was the ultimate change in EA "led to" by the meals?  One more graphic for you where I snipped the 300 min marks and copied them to the left of the plot of absolute EA.  

Continuing with the discussion ...

Consistent with previous studies [4,23,24,25], the early postprandial period after an LF meal was characterized by a high glucose level and concomitant exaggerated insulin response.

4. Ludwig DS, Majzoub JA, Al-Zahrani A, Dallal GE, Blanco I, et al. (1999) High glycemic index foods, overeating, and obesity. Pediatrics 103: E26.

This is not at all unexpected, and the authors could have probably listed ten times as many references in support.  But I stress that Ludwig's study in teens is cited again.

Despite the high glucose peak 30 minutes after the LF meal, EA did not differ in the early postprandial period due to suppression of FFA and BHB after this meal.

L-to-R:  1st hour absolute EA, change in EA,
and absolute change in FFA

No!!  The EA is not higher after the LF meal due to the fact that EA was lower before the LF meal.  At right I have cropped the early EA, change in EA and NEFA/FFA levels.  This finding clearly does NOT justify the statements made by the researchers here.

It is important to keep in mind that blood glucose makes up at least 70% of the Energy Availability regardless of dietary composition (with the exception of an truly ketogenic diet) at all times.

Continuing from the Discussion:

In the late postprandial period, the expected reactive hypoglycemia [23,24,25] did not occur; differences in late postprandial EA were instead due to significantly lower levels of FFA after the LF meal as compared with the LGI and VLC meals...

These patterns are similar to those observed with testing after a single meal, suggesting that the difference in EA is likely not due to habituation to a given dietary pattern [4].

4. Ludwig DS, Majzoub JA, Al-Zahrani A, Dallal GE, Blanco I, et al. (1999) High glycemic index foods, overeating, and obesity. Pediatrics 103: E26.

Similar?  Well, OK .... But not similar to what is being claimed.  At the risk of being overly redundant, here are the postprandial profiles, Reference 4 -- the oatmeal vs. omelet study -- on the left, the 2012 paper on the right.  Since baseline glucose, insulin and glucagon were roughly similar in 2012, the comparison to change from baseline works.  Since fatty acids were not, I substituted the change from baseline plot created previously.

In the end, the results are similar, but NOT what is predicted or proposed as a mechanism for how "the hormonal effects" and glycemic index contribute to overeating (hunger) and obesity.   And still, in 2012, Ludwig hasn't thought to extend measurements out to 6 hours ...

Aside ...

From the paper:

Respiratory quotient: Respiratory quotient [RQ] values ranged from 0.88–0.94 after the LF meal, 0.80–0.88 after the LGI meal, and 0.77–0.80 after the VLC meal. All pairwise comparisons for respiratory quotient indicated significant differences over the entire curve, with respiratory quotient higher with the LF diet vs. the LGI (p = 0.0001) or VLC (p = 0.0001) diet, and with the LGI vs. VLC diet (p = 0.0002).

Conclusion #1:  Fat burning beasts are no more or less hungry than the sugar burning savages.  

Conclusion #2:   Fatty acid oxidation is increased when glucose is lowered, and is not related to


The LF group would, as expected, increase carbohydrate oxidation after a meal, while the LGI group would see less of a stimulus, and the VLC group would expect to see none.

Concluding Remarks

I think it is instructive, after all of that, to look at the totality of the abstract in the 2012 paper.

Note the following (my commentary):

• Background:  It is stated in the affirmative that circulating metabolic fuels regulate hunger.  Their EA may be useful in informing dietary recommendations after weight loss.   
• Aim:  Let's look at EA for different diets. (Let's not mention that "we" examined hunger concurrently with the aim of associating lower EA with increased hunger.)
• Methods:  Oh we mention hunger ratings here as a secondary measure.
• Results:  EA was higher for VLC vs. LF in the postprandial period. (Just pay no attention to the magnitudes, oh and let's forget about any interesting results in the secondary outcomes.

And finally, their conclusion:

These findings suggest that an LF diet may adversely affect postprandial EA and risk for weight regain during weight loss maintenance.

This is appalling.  How do they reach this conclusion?  In ... what ... the 7th or 8th grade, we learned how to write a science lab report.  Perhaps Dr. Kelly even taught us in the 6th grade.  Science labs (at all academic levels) generally are performed to demonstrate known concepts and relationships.  It always seemed absurd to have to write it out AGAIN, even though it was in the lab book, but we were required to do this in HS and college, to write a full lab report for our experiments.  We had special notebooks with carbon paper (remember that?!) that could be placed between the pages, and eventually the special paper that made a "carbon copy" for us so that our results were set in stone and couldn't later be fudged.

Now sometimes the parts had different names or were combined (e.g. Purpose often replaced Problem & Hypothesis in the above template), but overall this is pretty standard form.   It was drilled into my head at an early age that the Conclusion MUST relate back to the Purpose, and if your Hypothesis was not supported by the results, it was imperative that this be addressed in the Discussion/Analysis section.  It didn't matter if it was a physics experiment to measure the acceleration due to gravity which is irrefutably 9.81 m/s^2.  If you got a wildly inconsistent result, you would not conclude that g=9.81 m/s^2 if your result was more like 5.33 m/s^2.   Then you better come up with something plausible to explain this in your Discussion, or you were toast!  Usually it was that danged spark timer when I taught this basic physics lab ... those ancient things were touchy at best ...

Bottom line, you cannot conclude that which your results do not support.  This is Good Science 101.   In their introduction they wrote:

Many studies have shown that circulating levels of the individual fuels affect appetite [2,3,4,5]. In both rodent [2,3] and human [4] models, hypoglycemia and decreased FFA levels lead to increased food intake. 

There's that oatmeal-and-omelet study misrepresented once again!!

However, these studies of individual fuels may not provide a comprehensive view of the metabolic regulation of hunger, because the body can utilize a varying mix of fuels under varying dietary conditions ... we propose that EA will be lower after a high glycemic load (low-fat, LF) diet in the postprandial period, compared to diets with a moderate glycemic load (moderate fat, low-glycemic index, LGI) or low glycemic load (very lowcarbohydrate, VLC).

What happened to the part about hunger/appetite and increased food intake?  Clever tact there -- imply the hypothesis, but omit that part from the formal hypothesis so you can ignore that in your discussion and conclusions.  Peer-reviewers apparently let them get away with that :( .

The conclusion here should have mentioned:

• Change in postprandial EA was not different between diets.  Any differences in the total are explained by the lower baseline EA due to habitual (3+ weeks) consumption of the different diets.
• Neither total EA, nor change in EA was associated with differences in hunger at any time point for 5 hours after the meal.

Therefore:  The results of their study do not support their hypothesis.

Ahh, but ... further studies.  NuSI millions to the rescue!  The energy expenditure study Popular Diets Study) will be largely repeated (although with some key changes particularly in protein intake) for a longer time period and side-by-side instead of cross-over fashion.  It is even getting a cute name: Framingham State Food Study  (FS)².  But wait!  There will also be a Metabolic Fuels Study.  They will look at more things in search of something they can use to support the internal starvation meme.  But the primary outcome, of groundbreaking solve-the-obesity-crisis research funded by NuSI?  It will again be postprandial EA.

Science!