Some downplayed and/or ignored revelations from Shai

link, click to enlarge
This will be a relatively short post, but due to my recent offering, I re-read some of the original Shai LF v. Mediterranean v. LC diet comparison study and had some commentary to add.  First, leaving aside reasons for withdrawal, there was a significant difference in compliance/completion between the three groups -- roughly 90% of LF, 85% of MDTN and 78% of LC subjects that completed the 2-year intervention.  

The analysis in Shai was "intent to treat" -- I so wish the raw data were available for such studies so that truly valuable information on what a truly compliant-successful-completer might hope to attain following a diet plan.  Isn't that what it is supposed to be all about??  But it is what it is, and in the results reported in the mass media, the ITT data was counted.  This generally means that all subjects data are included at all timepoints.  When they drop out, their last data points are commonly continued forward.  This is especially interesting, to me, for the LC group (but also MDTN).  You see 4 of the 109 randomized to LC dropped out before the diet, this means their initial data were carried throughout and at early time points the low carbers lost even more weight at six months than the average might imply!   But then it gets interesting, because 3 more withdrew at the 6 month mark, and 13 more at the 1 year mark.  This means that 3 early losses were likely carried through the calcs at all subsequent time points, and 13 subjects had 1 year data carried forward to the 2 year mark.  Still, they did give data for the completers which is favorable to the LC group as a whole.
All 322 participants at 2 years:
     LF: 2.9±4.2 kg     MDTN: 4.4±6.0 kg      LC: 4.7±6.5 kg 
Among 272 participants who completed 2 years:
     LF: 3.3±4.1 kg     MDTN: 4.6±6.0 kg      LC: 5.5±7.0 kg 
So... stick with it to some degree and LC seems to work nominally better.  Of course these are just self-reported adherents and there is little-to-no data available as to the degree of compliance.  Still, you have roughly 12% more of the initial group of low-fatters represented at two years than low-carbers, and I think this bears mentioning in terms of dedication to the plan.

But here's some other data we don't hear about.  This study was in mostly (86%) men, the entire cohort being middle-aged (avg. age 54).    
Among all 277 males at 2 years: (values in parentheses are 95% CI)
     LF:   3.4 kg (2.5-4.3)     MDTN: 4.0 kg (3.0-5.1)       LC: 4.9 kg (3.6-6.2) 
Among all 45 females at 2 years
     LF:   0.1 kg (1.9-2.2)     MDTN:  6.2 kg (1.9-10.2)     LC:  2.4 kg (2.2-6.9)

This is interesting folks!!!   Look at that.  Granted we're not talking about a lot of women, but the men did nominally better on LC vs the other approaches.  But the low carbing women did more poorly on both LF & LC diets, and more poorly on LC after two years than the men did on LF!!  Not that they did great on LF, essentially breaking even with little variability, but look at the MDTN!  With great variation (why they report normal std. deviation or std. error for raw data in the primary analysis and %CI (estimated variation, not raw data error/variation) for this data I'll never understand.  But it doesn't matter.  The weight loss superstars in all of this, if there are any to be lauded, were the women on the Mediterranean diet!!

And so, four years later, we're not given this breakdown.  Sigh.  But I think we can make a few inferences that are far more valid than guruspeak interpretations of same.  The MDTN group regained less on average.  Did the women who "ruled" at the 2 year mark carry the diet?  Perhaps.  


Puddleg said…
Very good, this is one of your better posts - always interesting when something unexpected like this shows up.
it's an argument for food quality being the overlooked factor; and perhaps for the mdtn diet being more "compliable" for women, because closer to the message they've already had about what's good for them, a good match with their expectations. Not so much men though.
Whereas perhaps some drop outs on VLC are due to pressure from outside. The diet was just too weird or extreme for friends, family, the voices in one's head, or other health providers.
John Yudkin used to say "a food that is not eaten has a nutritional value of zero". In this case, a diet that can't be followed, for whatever reason, has a slimming value of zero.
Samad Shahid said…
This comment has been removed by a blog administrator.
Puddleg said…
To my mind the most interesting NuSi experiment would measure the difference between quality and quantity.
Pit a quality-intense diet (highly nourishing, low GI, low in food toxins) like Perfect Health Diet against a low-carb diet off the usual menu.
At what point does the higher quality of carbs cease to offset the greater quantity? What is the "Carb Quality Differential"?
Sanjeev said…
Eades and Taubes are missing a small opportunity: as popularly perceived the Mediterranean diet's more rewarding than either low carb or low fat

; they're missing the opportunity to spin Shai as one disproving data point for reward theory
Sanjeev said…
which would require focusing on the often better results for the (self reported) MD;
CarbSane said…
Call me cynical George, but I don't think NuSI has any interest in testing various carbohydrates. Part of the problem is that the half they are blaming now (fructose) has never been pushed by the mainstream. Really, kids cereals were far worse when I was a kid, and I never heard a lunch lady or teacher criticize some kid's sugary snack, or disallow birthday cupcakes ... even for the rare chubby kid.

On a more serious note, I'm not sure we could establish a thresholdl. Perhaps no more than 5% of one's diet from dust.
CarbSane said…
Thanks George! I can't recall which interview, but Taubes once said a while back that he'd like to test LC in men because it seems to work better for them. I thought this was a telling comment, especially taken together with his penchant for showing the picture of the obese ovarectomized rats.

Of course there are female success stories on LC, but anecdotally men seem to do much much better, whether it be rapid initial losses and/or generally maintaining them.

The diet in Shai was only VLC for the first month, so we're talking a rather moderate LC diet.

I dunno, I can only go by my experiences IRL and since I found the LC community on the web. My docs have never bothered me about my diet when my health numbers are good. Years ago I had a doc who didn't like that I was doing that "dangerous" Atkins to lose around 40 lbs at the time, but that's about it. Yes, LC is more different from "normal" than other plans and this may factor into adherence. That's one reason I think the online community is so tight and popular. But it also fosters a rebel/martyr streak that can be counter-productive. Sometimes doctors and scientists get it right, even when funded by evil "Big Whatever".

I would LOVE to see data analysis for JUST the compliant. That would be actionable information, no? In this regard I'm ticked at Shai. If you're going to bother to follow-up, would it be so difficult to have broken the data down for the 67% who remained at least nominally on their plans?
Paul N said…
I know this was referenced in a comment to your previous post on this, but it seems to me a missing factor here is just what "vegetable fats" the LC group were eating. It seems unlikely they were eating lots of butter/lard/tallow/egg yoks, and if they were using canola, soy or "wesson" or (worse) "crisco", then there can be health complications from that. Eat those at the same time as what little sugar/carbs you have and there can be even more problems.

I am agreeing here with George about "quality" - it is certainly possible to eat a low quality diet, be it LF, LC or in between.

The WAPF style, and most "ancestral" diets looked for nutrient dense foods, and worked, even at wide varieties of carb/fat ratios.

An LC diet that has lots of omega six oils, and the carb fraction from dense, acellular carbs" (i.e. refined) is far less nourishing than the same caloric and carb ratio diet with whole food sources - as Paul Jaminets PHD shows us.

In this group, in looks like the MDTN got closest to a whole foods diet, so its no surprise that the women did better on it - there seems to be a lot of anecdotal evidence that they don't do well, long term, on LC/VLC.

Given what we know, and what people like George have experienced, to design a study like this and tell the LC cohort to favour "vegetarian" fats - without being specific about avoiding n-6 - sounds like the experiment is biased in favour of the MDTN from the start.

Seems to me, then, that doing more testing on the various fats is where NuSci should focus - there has been plenty of work done on carbs already.