The New Low Carb v. Low Fat Study ~ Much Adieu About Not a Helluvalot!

So the LC Hills are alive with the sound of music to the ears of low carbers once again.  A 2-year NIH funded study pitted a classic lowfat CRD against a DANDR stye low carb diet.  DANDR fared equally with low fat on most measures and even posted a modest sustained improvement in HDL.

Here's the summary:

Background: Previous studies comparing low-carbohydrate and low-fat diets have not included a comprehensive behavioral treatment, resulting in suboptimal weight loss.
Objective: To evaluate the effects of 2-year treatment with a low-carbohydrate or low-fat diet, each of which was combined with a comprehensive lifestyle modification program.
Design: Randomized parallel-group trial. ( registration number: NCT00143936)
Setting: 3 academic medical centers.
Patients: 307 participants with a mean age of 45.5 years (SD, 9.7years) and mean body mass index of 36.1 kg/m2 (SD, 3.5 kg/m2).
Intervention: A low-carbohydrate diet, which consisted of limited carbohydrate intake (20 g/d for 3 months) in the form of low–glycemic index vegetables with unrestricted consumption of fat and protein. After 3 months, participants in the low-carbohydrate diet group increased their carbohydrate intake (5 g/d per wk) until a stable and desired weight was achieved. A low-fat diet consisted of limited energy intake (1200 to 1800 kcal/d; 30% calories from fat). Both diets were combined with comprehensive behavioral treatment.
Measurements: Weight at 2 years was the primary outcome. Secondary measures included weight at 3, 6, and 12 months and serum lipid concentrations, blood pressure, urinary ketones, symptoms, bone mineral density, and body composition throughout the study.  
Results: Weight loss was approximately 11 kg (11%) at 1 year and 7 kg (7%) at 2 years. There were no differences in weight, body composition, or bone mineral density between the groups at any time point. During the first 6 months, the low-carbohydrate diet group had greater reductions in diastolic blood pressure, triglyceride levels, and very-low-density lipoprotein cholesterol levels, lesser reductions in low-density lipoprotein cholesterol levels, and more adverse symptoms than did the low-fat diet group. The lowcarbohydrate diet group had greater increases in high-density lipoprotein cholesterol levels at all time points, approximating a 23% increase at 2 years.
Limitation: Intensive behavioral treatment was provided, patients with dyslipidemia and diabetes were excluded, and attrition at 2 years was high.
Conclusion: Successful weight loss can be achieved with either a low-fat or low-carbohydrate diet when coupled with behavioral treatment. A low-carbohydrate diet is associated with favorable changes in cardiovascular disease risk factors at 2 years.
Primary Funding Source: National Institutes of Health.

OK.  So let's start with the good here:  Following Atkins FULL plan (increasing carbs out of Induction), even with a bit of an extended induction period (12 weeks vs. 2 weeks) results in nominally improved biomarkers (HDL in particular) in the short and longer term vs. low fat.  IOW, Atkins as prescribed won't kill you and might actually be a bit more beneficial to the cardiovascular system.  This is GOOD news, but I think some are taking it to mean more than it does.  They studied DANDR as prescribed (IOW carbs were increased up the 5g rungs) in obese patients (BMI avg. ~36), excluding any with dyslipidemia (highly correlated with insulin resistance/pre-diabetes) and overt diabetics. 

Less than 10% of the original respondents eventually made the study, and 25% were excluded in the last step.  Granted only less than 15% who made it to the eligibility assessment were excluded for "cause", but these would have been obese persons essentially unaware of their dyslipidemia and/or diabetes.  It is not unreasonable to assume that a large portion of the 3140 who didn't "pass" the prelim phone screening were excluded because of these criteria, and it would also not be unreasonable to assume that a good portion of the 355 who missed their in-person screenings may have decided they wouldn't pass the next step because of missing these criteria.  Of course nobody can say for sure, but abnormal lipids are prevalent in the obese.  So, although I can see wanting to exclude overt diabetics,  I have reservations as to the exclusion of those with lipid abnormalities.  After all, the "risks" associated with obesity are associated with those very biomarkers, and improvements in health (CVD risk among them) associated with weight loss were what was being studied after all.  So I'm left to wonder if the sample studied by Foster is even representative of the obese population as a whole.  I'll update with a citation when I find it, but I recently read an article stating that up to 20% of the obese are   "metabolically normal" -- so basically this study, whatever its implications, only pertains to this subset of the obese population.  But what of the more-likely obese person with deranged lipid metabolism to begin with??  Hopefully that will be another study on another day.

In any case, here are the starting stats:


Two thirds of the participants were women and the average starting weight was ~103.5 kg for each group which equates to just over 225 lbs.  In one year the average weight loss is not encouraging and would still have the average weight of the participants over 200 lbs.  Here is the breakdown of the weight losses:

If the numbers in ( ) are +/- 1 standard deviation we see SD's of around 1 to just less than 2 kg -- IOW, weight loss was not highly variable amongst the groups.  Weight losses are hardly all that significant in this study (depressing, really).  Now although not statistically significant, the weight losses of the LC group follow a familiar pattern.  Slightly more in the first 6 months followed by a slightly greater regain for low carb vs. low fat.  

When one looks at the "crown jewel" finding of this study, HDL, we see that both groups started with levels in the mid-40's.  Here's the stats (remember low fat on left, low carb on right)

Yes, the low fat diet had the predictable (and not desirable) effect of actually lowering HDL initially, while the LC group only showed increases.  But we're talking an absolute difference in level of around 3mg/dL  at the 1 and two year marks.  I don't find that all that remarkable.  

OK, so some general comments on the study itself.  I have a copy of the full study (thanks J) complete with Appendix and nowhere does it list reported dietary intake for the two groups.  The dietary prescriptions provided are also vague.

Low-Carbohydrate Diet
Approximately half of the participants (n 153) were assigned to a low-carbohydrate diet, which limited carbohydrate intake but allowed unrestricted consumption of fat and protein. During the first 12 weeks of treatment, participants were instructed to limit carbohydrate intake to 20 g/d in the form of low– glycemic index vegetables. After the first 12 weeks, participants gradually increased carbohydrate intake (5 g/d per week) by consuming more vegetables, a limited amount of fruits, and eventually small quantities of whole grains and dairy products, until a stable and desired weight was achieved. They followed guidelines described in Dr. Atkins’ New Diet Revolution (15) but were not provided with a copy of the book. Participants were instructed to focus on limiting carbohydrate intake and to eat foods rich in fat and protein until they were satisfied.  The primary behavioral target was to limit carbohydrate intake.

Ummm ... stable and desired weight?  If the whole point was just to lose around 10% bodyweight I guess they achieved this in the first year.  But I would venture a guess that most 225 lb'ers desire to lose more like 30-40%.  In any case, a few days of food diary analysis at each assessment point should have been included so that the actual carb (and caloric) intake is known.  Did all participants procede up the rungs?  What was the final carb count after 6 months when this group began to regain?  Even fewer details are given for the low fat group:

Low-Fat Diet
The remaining 154 participants were assigned to consume a low-fat diet, which consisted of limiting energy intake to 1200 to 1500 kcal/d for women and 1500 to 1800 kcal/d for men, with approximately 55% of calories from carbohydrate, 30% from fat, and 15% from protein.  Participants were instructed to limit calorie intake, with a focus on decreasing fat intake. However, limiting overall energy intake (kcal/d) was the primary behavioral target.
Here food diaries should have been analyzed to measure compliance, no?  One is left also to wonder if at some point this group didn't also increase caloric intake once a "stable and desired weight" was achieved.  Otherwise we are to believe that two years of maintained caloric restriction resulted in regaining some of the weight after one year?

This is a serious criticism here.  When the comparison of outcomes between two dietary interventions is the primary focus of a study, an accounting of the actual dietary intake of the participants is paramount ... don't you think?  The authors felt compelled to provide details of the behavior modification program that was given to both groups in the Appendix, but failed to provide any details on the actual diet consumed.  In previous comparison studies -- cited by this study -- those being Gardner,  Shai, and Dansinger -- dietary intake and compliance was monitored and reported in far greater detail.  Self-reported food diaries may not be perfect, but should be a minimum requirement in order to draw any conclusions on dietary differences.  The authors further muddy the comparison waters by prescribing an exercise program (walking 4X 20min beginning week 4 ramped up to 4X 50min by week 19) and behavior modification program (SAFE - 75-90 min counseling sessions every other week for first 20 weeks then every other month for the duration of the study) but, there appears to be no effort to assess compliance/attendance with either program.

All of this leaves me with serious doubts over drawing any conclusions from this study.  It was either poorly planned, executed and/or reported.  Reminds me in a way about this NEJM study from 2009 -- not much differences in weight loss (and even more pathetic losses) probably due to poor compliance.  If you cannot at least try to demonstrate compliance to one plan or another, how are comparisons even meaningful?   One would hope that if the assessments missing in the article were indeed carried out that this data/results would be disclosed in a revised publication.  In any case, I don't know about y'all here, but when I read results like this -- all this to lose a measly 25 lbs in a year only to put 1/3rd of it back on in the next year -- it is rather discouraging.

Now many in the LC community are pointing to the improved triglycerides during the <g phase and I'm hearing a lot of "if only they hadn't ramped up the carbs".   If only they had used a more rigorous LC diet rather than this "less potent" form as one blogger referred to the diet as.   Fair enough, but that is NOT what Atkins laid out in DANDR ... and that is the point, right?  If DANDR "by the book" isn't the quintessential low carb diet, what is?  Neither the original nor DANDR even encourages staying at Induction (biologically zero) level carbs.  At least the New Atkins cautions that "for some" lifetime maintenance means staying <50g or even lower.  If true low carb living is at these levels shouldn't that be a disclaimer going in??  Or is this the unspoken dirty little secret of the low carb community -- wink wink??  I'm beginning to think it is.  Most of the outspoken long term low carbers seem to go lower and lower as time goes by.  Indeed my own method was <20g most days except for planned cheats, but when I stalled out the advice was almost uniformally about cutting out what few carbs I still ate (not a consumer of "product" here except for the occasional LC wrap).

Now, were the "headlines" I'm seeing heralding this study supportive of low fat dogma, the study authors would surely be derided as idiots, accused of lapses in scientific integrity, bias, etc.  But the authors do at least address some of the issues I've mentioned here, so I'll quote those here:
Our study also has several limitations.  First, the comprehensive behavioral therapy program used in this study makes it difficult to extrapolate our results to  general weight management in the community. However, the clinically significant weight losses achieved at 24 months underscore the need for providing patients with long-term behavioral support, whether by registered dietitians or other allied health professionals (32, 33). Our protocol was based on an Atkins version of a low-carbohydrate plan, which prescribes an increase in carbohydrate intake over time; thus, the effects of longer than 12 weeks of severe (20 g/d) carbohydrate restriction could not be assessed.  Finally, our findings should not be generalized to obese persons who have obesity-related diseases that were excluded from our study population, such as diabetes and hypercholesterolemia.

Bottom line: this study did not look at the long term efficacy/ramifications of the VLC WOE.  We can only postulate that the lipid profiles would be any better (or worse).  So this study offers absolutely no "vindication" as to the healthfulness of long term VLC eating.  Indeed it doesn't seem to tell us much more than the three studies done previously, Shai in particular, except perhaps the age/gender makeup was different.  But speaking of the gender makeup of the study group -- ~1/3 male, 2/3 female:   I wonder (a) why the gender split wasn't more reflective of 50/50 of the population, and  (b) no attempt was made to assess gender differences (if any).

Attrition:    At the 1 year mark, attrition was similar between groups - about 1 in 4, but at the 2 year mark, attrition may not have been statistically different, but while LF had ~1 in 3 drop out, LC had ~2 in 5  (32% and 42% respectively).  I would like to address attrition in this study in more detail, especially since this study refutes (further) erroneous claims made by Westman, Volek & Phinney in The New Atkins that LC diets are easier to stick to and losses are easier to maintain.  Here's what is reported in the study:
There were no statistically significant differences between the 2 groups in attrition, defined as not undergoing an assessment at a specific time point, independent of the reason. Attrition included participants who withdrew and intermittent missingness at each time point. In the low-fat group, 6%, 12%, 25%, and 32% of participants did not participate in assessments at 3, 6, 12, and 24 months, respectively.  Values for the low-carbohydrate participants were 9%, 16%, 26%, and 42%, respectively.
The differences may not be statistically significant according to their methods used, and I'm not privy to the raw data so certainly can't do my own analysis.  BUT there are two types of error in statistics:  sampling error and non-sampling error.  The attrition is consistently greater for the LC group at all time points.  If we're going to play the percent game with HDL (23% improvement sounds better than the absolute value for example), then one can say that 50% more dropped out/failed to report in the LC group vs. LF in the first 3 months, and the drop-out/ftr rates at 6 months and 2 years are 25% and 33% higher respectively in LC vs. LF.    This difference in attrition at least implies LC was more difficult to stick with.  Therefore attrition differentials could contribute a significant amount of "non-sampling error" to the results.  (See this post for a more complete discussion of sampling and non-sampling error.)

I've crunched down the attrition graphic from the study (left out reasons for discontinuing):

Note that the "attrition rate" is combined for those that dropped out of the study AND those who remained "in treatment" but missed their assessment.   The numbers are small, but the differences are noteworthy.  You have more than 3X as many participants not reporting for "duty" -- yet remaining in the study -- in the LC group vs. the LF group at 3 months and twice as many not assessed in the LC vs. LF group at 6 months.   Obviously the values reported can only be reported for those active and present for assessment.  How many of the 10 no shows in LC/3mo showed up and were "counted" in the 6 mo assessment?  Were the no shows all different participants or were any subjects "kicked out" for missing more than one assessment and not included in the data analysis.  At two years, the % attrition may not be a statistically significant difference, but the endpoint data is missing for 23/153 =  15% of the LC group that is considered "In Treatment".  This value is just under 8.5% for the LF group.    What effect might this have on the outcomes?  We cannot quantify it, and the authors do acknowledge the high attrition rate for both groups as a limitation of the study.  But where small differences are seen, it is quite possible that attrition changed the "makeup" of the samples.  On the positive side, those persisting on LC are likely the better adherents so this makes the HDL outcome promising.  On the other hand, we can only speculate about what the 10 additional 2yr no-shows and 8 additional 1yr-to-2yr dropouts in LC vs. LF group had on the ultimate outcome and the touted persistence of HDL improvements irrespective of weight regain.  

This study, in the end, illustrates mostly how difficult it is to reverse obesity.  Even with a significant amount of support the weight losses are not all that impressive for either group, and they regained on average 1/3rd of the weight loss.  It would be nice to know -- in raw numbers -- the number of participants that lost/maintained/gained weight from 3mo-to-6mo, 6mo-to-1yr, and 1yr-to-2yr.  This would have no consequence on any of us individually, but combined with an analysis of compliance with program between subgroups may elucidate a best (or better) strategy for weight loss and maintenance.

I'll leave you with the actual conclusions of the authors:
In conclusion, this 2-year, multicenter study of more than 300 participants revealed that neither dietary fat nor carbohydrate intake influenced weight loss when combined with a comprehensive lifestyle intervention. Both diet groups achieved clinically significant and nearly identical weight loss (11% at 6 months and 7% at 24 months), and persons who received the low-carbohydrate diet had greater 24-month increases in HDL-cholesterol concentrations than persons who received the low-fat diet. We found no differences between the groups for changes in bone or body composition. These long-term data suggest that a low carbohydrate approach is a viable option for obesity treatment for obese adults.


jamie hale said…
Thanks for posting, and thanks James for mentioning on your blog

jamie hale
CarbSane said…
Thanks guys, and I really appreciate your linking to this James.

The more I think about it, the more it bothers me that a dietary comparison study did not appear to include routine assessments and reporting of the actual dietary intakes!
LynMarie Daye said…
Great analysis!

The lack of dietary compliance and intake assessment is puzzling. Like you said, it's doubtful food diaries and such are 100% accurate, but it's hard to attempt to analyze diet studies properly without them.

I'd like to point out that while the study participants weren't "officially" dyslipidemic, their HDL levels were rather on the low side, IMO. Just from personal experience from having worked in a medical lab for many years, I'd say that the average HDL level for the average American would be in the low to mid 50s. It's possible that more than a few of the participants were mildly insulin resistant at the start of the study. Would you agree?
LynMarie Daye said…
Oops, I hit the Post Comment button too soon!

I just wanted to add that since women tend to have higher HDL levels than men, a mean HDL of 45 or 46 in a group of people in which 2/3 of them are women indicates to me that some degree of disordered lipid metabolism may be taking place. Nothing major but perhaps the start of a problem.
CarbSane said…
Yes, I would agree with you LynMarie that these subjects had mildly "concerning" lipids (low HDL as you say, especially given the gender ratio weighting of this marker). IR is not an all or nothing thing, so mild insulin resistance could certainly have been the case here. But this study excluded the vast majority of the obese based on this criteria and I'm not sure I agree with this exclusion. Excluding diabetics makes sense (especially if taking medications), but let's see what LC vs. LF does for improving lipids in the obese with dislipidemia. In the end, both groups saw modest improvements in HDL. I really don't know what to make of LDL anymore, only that some low carbers have VERY elevated LDL particle numbers, and it seems too soon for my satisfaction to write these off entirely if most of them are larger. Out of the ordinary numbers indicate something is awry.
LynMarie Daye said…
"Excluding diabetics makes sense (especially if taking medications), but let's see what LC vs. LF does for improving lipids in the obese with dislipidemia."

I agree 100%. I would love to see a well-designed study addressing this. As far as I can remember, there hasn't been, although it is hard to keep up with everything that comes out.

"I really don't know what to make of LDL anymore, only that some low carbers have VERY elevated LDL particle numbers, and it seems too soon for my satisfaction to write these off entirely if most of them are larger. Out of the ordinary numbers indicate something is awry."

LDL levels of 300+ are very rare. Weren't the traditional Masai found to have low LDL? It's all very confusing at this point.
CarbSane said…
LDL does seem to correlate somewhat with weight or, perhaps, someone's "caloric status". By that I mean if you are in just a small caloric surplus and gaining -- however slowly -- LDL seems to elevate. But, for some reason it doesn't go down while you are in a deficit while still being overweight. Anecdotally, rapid weight loss seems to spike the LDL.

I remember when I first topped the 200 mark and had a cholesterol screening for employment, it was before the whole LDL, HDL, etc. breakdown was routine. My cholesterol was apparently shockingly high for a 19-20 year old. Several years later when I had another test done, my doc (ahead of his time) did note a slightly elevated LDL, but also that my HDL was beneficially high.

Since then I've always skirted the statins with my docs. LDL goes up a bit when the weight is up and down when the weight is lower, but I have been blessed with stellar HDL numbers so that mostly keeps the statinators off my back!

It's too bad they excluded all that they did from this study. :(
CarbSane said…
p.s. 200 mark was pounds, not cholesterol.
James Krieger said…
Just a note on LDL particle number. A lot of people think the "large fluffy" buoyant LDL is less harmful. However, primate data indicates that it's more harmful given the same particle number. Usually what happens is that there's an inverse relationship between particle number and LDL size. So it's not the bigger size that's's the fact that you have less particles. Unfortunately VAP doesn't detect particle number...only particle size. There is a test that does particle number but it's not as widespread as VAP. But if low carbers have greater particle numbers, that's not a good thing, even if the particles are larger and more fluffy.
James Krieger said…
I saw a zero-carber on the low carb forum just recently claiming to have an LDL of 304
CarbSane said…

This came to my attention when Jimmy Moore posted some recent lipid profiles. He (and others) have gone a step further than describing large LDL as less atherogenic and calling it "protective" and/or using the descriptions like "good".

As I said, I'm not at all certain of anything regarding cholesterol levels -- correlation and causality being different animals and all that -- but any serious deviation from the norm is likely an indicator that something is amok in one's metabolism.
LynMarie Daye said…
Very interesting about LDL particle number vs. size, James. I found the following info on Wikipedia detailing a potential mechanism for the harmful effects of a high LDL particle number:

"...large numbers of LDL particles lead to competition at the APOB100 receptor (i.e. LDL receptor) of peripheral cells. Since such a competition will prolong the residence time of LDL particles in the circulation, it may lead to greater opportunity for them to undergo oxidation and/or other chemical modifications. Such modifications may lessen the particles' ability to be cleared by the classic LDL receptor and/or increase their ability to interact with so-called "scavenger" receptors. The net result is shunting of LDL particles to these scavenger receptors. Scavenger receptors typically are found on macrophages, with cholesterol laden macrophages being better known as "foam cells". Foam cells characterize atherosclerotic lesions. In addition to this possible mechanism of foam cell generation, an increase in the levels of chemically modified LDL particles may also lead to an increase in endothelial damage. This occurs as a result of modified-LDL's toxic effect on vascular endothelium as well its ability both to recruit immune effector cells and to promote platelet activation."
from Wikipedia Article on Apolipoprotein B

This is me just hypothesizing, but I think it's reasonable to say that someone with an LDL cholesterol concentration of 300+ may have to have a high number of LDL particles even if the particles are large. A 300+ LDL cholesterol is uncommonly high; for particle number to be low, I imagine the particle size would have to be VERY large (to house such a large amount of cholesterol). Assuming there is a limit to how large LDL particles can be, it stands to reason that at some point a high enough LDL cholesterol level would be indicative of an elevated particle number. That's my thought on this anyway! :`) And it's certainly subject to change!
LynMarie Daye said…
Have you seen some of the criticisms of this study? I came across a couple of blog articles suggesting that the researchers are biased against low carb diets and manipulated the data to get results more inline with their beliefs. I would love to hear your thoughts on this.

Another Biased Study? Maybe...

The Diet Wars: The Saga Continues
Alan said…
I am pretty skeptical about "lifestyle counseling" etc.