Nina Tei¢holz, Shai'ster ~ Part II: Applicability of RCTs to the General Population

Part I: Discussed the reporting and makeup of the study diets in Shai et.al.  Bottom line, if intake reports are to be believed, the "high fat" group in this study DEcreased absolute fat intake by a few grams to 70% of baseline intake depending on the reporting used.

The study:   Weight Loss with a Low-Carbohydrate, Mediterranean, or Low-Fat Diet

Nina Teicholz on Nightline, link to full story

Will the Real Applicable Clinical Trial Please Stand Up!

In this installment I am going to discuss how Nina Teicholz couches the Shai study and then dismisses another study for reasons that apply equally, if not moreso, to her favored example.  Keep in mind that both in The Big Fat Surprise, and in her media blitz promoting the book, Teicholz makes reference to Keys' bad science contrasted by the definitive, conclusive, rigorously controlled clinical studies from the past decade.


These studieS, boil down mostly to one:  Shai.  In Chapter 7 (of 10, they're mercifully short despite the length of the "book") entitled Selling the Mediterranean Diet:  What is the Science?, after pictures of Keys and other famous luminaries associated with the diet, and after being introduced to Big Olive, in a section entitled A Test of the Real Mediterranean Diet, Teicholz introduces us to Shai:

The first such trial [of the Mediterranean diet], in 2008, was conducted in Israel. It was well designed and rigorous, with an international group of professors on board, including the epidemiologist Meir Stampfer of the Harvard School of Public Health. These researchers selected 322 moderately obese middle-aged people, mostly men, and fed them one of three diets: one low in carbohydrates, one low in fat, and the third, Mediterranean. ^XXVIII Specially prepared meals were served at a workplace cafeteria, allowing for a high degree of control over what and how much foods were eaten. And the experiment lasted two years, a long time for a trial that involves overseeing the preparation and service of food.   {3652}

** all numbers in {} are the starting Kindle location out of 12033 total, this should give those with hard copies a general idea to where they can find these quotes in a hard copy.

The Roman numerals there are Teicholz's footnotes.  These sometimes link to additional information that would not flow in the text, but tucked at the end of the chapter, they are quite often used to "hide information in plain sight".  By that I mean that many reading a physical book will not leaf to the end of the chapter to see what's up, and by the time they get there forget what the footnote pertained to anyway.   Here, XXVIII informs the reader that the Mediterranean diet is Willett's interpretation.  This is important because she sets this up as a test of the "real" Mediterranean diet and this is a point of contention in nutritional circles.  After the above paragraph, Teicholz goes on to describe the 2013 PREDIMED study.  Teicholz apparently donned her Holmes hat and grabbed her magnifying glass with that study, reading every last word of the appendix (or so the impression is given).   It would appear that she did not do the same diligence with Shai.  But the crux of this section of the book comes in the next paragraph:

If the Israeli trial had never existed, everyone could have assumed that the Mediterranean option in PREDIMED was the best possible regime for health. But that third, low-carb arm in Israel had revealed that an even better option was possible. (Previous shorter trials had found the same thing, as we will see in Chapter 10.)^XXXI       {3687}

Now here is a prime example of Teicholz hiding information in plain site.  Read the short paragraph above again, please, and tell me there's any inkling there that another two-year study on the Mediterranean diet even existed.  She only implies that there were shorter trials including low carb as an option.    Here's the footnote:

XXXI . There was one other long-term (two-year) trial comparing a low-fat to a Mediterranean diet, with results in 2004. It showed that the Mediterranean diet performed better. But the study involved men and women with metabolic syndrome, not a normal population, and could not be generalized (Esposito et al. 2004).  {3493}

First, I submit that this is the footnote that should have been placed at the end of the red bolded statement regarding the 2008 Shai study being the *first* long term trial of the Mediterranean diet.   But it's that blue bolded statement that prompted this post.

Devilish Details

As an aside here, the first paragraph is a prime example of how Teicholz weaves her overt bias into much of this book.   Prominent names and prestigious institutions are stressed often -- Harvard appears 52 times -- especially, as in this case, to give the appearance of gravitas or higher standards.  The Shai study was an international effort with a Harvard bigwig on board.  The Esposito study she dismisses?   A bunch of no names I guess.   It's also interesting what she leaves out or saves for (much) later, such as the partial funding of Shai by The Atkins Foundation.

The description of the control in this study will be the subject of my next installment as it contrasts wildly with how she describes the LA Veterans study that she maligns and outright mocks in various interviews.  Suffice it to say, that although the subjects in Shai only had access to their cafeteria at lunch time (during work days I might add), there is not even any indication that they were required to make appropriate selections, how much, etc.  These little "details" in how similar situations are described depending on whether the study supports or refutes Teicholz's views are quite telling throughout the book.

Representative Samples

In her footnote about Esposito, Teicholz inadvertently makes an excellent point.  I say inadvertently because, per what I just discussed, if she applied this standard equally to the Shai study (or for that matter her various fat-eating cultures), there'd be no book.  But it's an excellent point nonetheless, and it's one that should be discussed more when interpreting clinical nutritional research.

All clinical trials are done, necessarily, on samples drawn from the population.  It is important that the sample reflect the population in composition as closely as possible so that the results translate to inferences about that population.  All of these clinical trials do some initial screening so essentially these samples are not representative of the general population.  And that's OK ... it makes no sense to test the effect of an asthma medication on non-asthmatics for example.  But it's not OK to extrapolate the results of such studies to the general population without acknowledging these limitations.

With the obesity and diabetes epidemics in full bloom, a goodly portion of clinical nutrition trials are aimed at one or both of these populations in terms of potential management, treatment or prevention of the conditions.  But if a diet reduces HbA1c from diabetic levels down into non-diabetic levels and other health markers improve, this does not mean that this diet will have any impact on someone who is not diabetic to begin with.  Further, even if it has a similar effect in the non-diabetic, there is no indication that the result is even clinically meaningful.  I used HbA1c as an example because risk factors really "take off" at around the 8 mark, so that lowering HbA1c even one half of one percent from 8.5 to 8.0 is potentially meaningful.  That same reduction from 7.5 to 7.0 is much less so, and in the normal range from 5.5 to 5.0?  The inferences don't translate.   We so often hear that the dietary rules change for diabetics, which may or may not be as significant as claimed, but we don't often hear the admonition in reverse.  That is to say, what applies to diabetics does not necessarily apply to the majority that are not.

It's important to keep in mind the makeup of study groups when considering the implications of the results.   One knock against Big Pharma is that it used to test drugs predominantly in men -- either young healthy males or middle-aged sick men and seemingly nothing in between.  This is an important point (great strides have been made in this regard), let's apply it evenhandedly.

The Other Two-Year Mediterranean Diet RCT:

So let's take a look at this 2004 Esposito study that Teicholz dismisses because it "could not be generalized":  Effect of a Mediterranean-Style Diet on Endothelial Dysfunction and Markers of Vascular Inflammation in the Metabolic Syndrome. A Randomized Trial.  While this study was smaller over all, it only had two groups so at 90 vs. 104-9 per group in Shai it was roughly similar.   But more importantly, Esposito was conducted in a MORE generalizable (is that a word?!) sample than was Shai -- yet another of Teicholz' mounting "errors" in the book.   Here are the criteria for Esposito and Shai, (apologies, green = inclusions, red = exclusions):


[Esposito] involved men and women with metabolic syndrome, not a normal population, and could not be generalized.        ~Teicholz

In the table below I've provided relevant study subject characteristics.  I've noted the 5 inclusion criteria from Esposito in the Met Syn column, and the * denote disease or medications associated with biomarkers.

This is inexcusable.  Just inexcusable.   Firstly, the gender issue alone should be a red-flag as to the applicability of Shai to the general population.  The Esposito study came far closer to a 1:1 ratio, the Shai study at 82 to 91% male in each group?  Not even in the same ballpark!  This was a study in middle aged men, over a third of whom had CVD or diabetes and whom on average would qualify as having metabolic syndrome.  Further, in the 2009 Adherence report, we learn that women in Shai had a higher drop-out rate than the men, something I hope to address in a future installment.  But Shai specifically included those with active diabetes or heart disease regardless of age or weight status, and those currently taking medications to manage those diseases.  How can one possibly attribute any of the results to diet alone?  Further, Shai included smokers which  I'll address when I take up the LA Veterans study.

The Esposito study, which excluded those with active disease and/or anyone taking any medications that would alter biomarkers being assessed, was truly a test of impact of diet on those headed in the wrong direction from a metabolic health point of view.   Could the results be generalized?  Well, considering that Katie Couric tells us 95% of us will be in this boat or worse in another few decades, and what are the statistics on obesity and metabolic syndrome now?  

Dat Mediterranean Diet:

I've seen Willett's Mediterranean diet described as your standard low fat diet with olive oil poured on top.  The sample menu from Shai at right would make it appear somewhat so.  

How about the diet used in Esposito?

The dietary advice was tailored to each patient on the basis of 3-day food records. The recommended composition of the dietary regimen was as follows: carbohydrates, 50% to 60%; proteins, 15% to 20%; total fat, less than 30%; saturated fat, less than 10%; and cholesterol consumption, less than 300 mg per day. Moreover, patients were advised to consume at least 250 to 300 g of fruits, 125 to 150 g of vegetables, and 25 to 50 g of walnuts per day; in addition, they were also encouraged to consume 400 g of whole grains (legumes, rice, maize, and wheat) daily and to increase their consumption of olive oil.

I note that intake is still self reported, but the dietary recommendations were individualized for caloric intake and the reporting was a 3-day diary instead of a single 24 hour recall or FFQ.  We also get a little more information in the study of what they consumed in terms of foods encouraged by the dietary recommendations.  The few sample menus from Shai raise more questions than they do provide meaningful information.

Oops!  They included legumes with the whole grains in the recs, yet itemize them with fruits, veggies and nuts in the intake table.  We see that in this intervention, the subjects actually made what appear to be meaningful changes in their diet, more than doubling up on the FVN&L and the whole grains, and almost doubling olive oil consumption.   We also get some information on "complex carbs" that is absent from the reporting in Shai.  It is doubtful the Shai Med dieters followed a remotely similar diet to the one here as these Med dieters increased fiber by 18 g/day (vs. an increase of 1 gram per day in Shai).  There is also a fairly significant reduction in saturated fat intake of 5% total energy from 13 to 8%.  In Shai, there was an absolute reduction of 4g of saturated fat resulting in a 0.1% change from 9.7 to 9.6%.   Increases in MUFA are obscured by the change in MUFA:SaFA ratio reported in Shai ... it becomes more apparent with each further look into this study why the reporting is so ... shall we say ... cloudy.

I hope to revisit the Esposito study at some point to give it its full due.

Surprise or Disguise?

So now I ask you:  Which study was a better test of the Mediterranean diet?  Of the impact of lowering SaFA while raising MUFA?  Of increasing fiber?  Of these "complex" carbohydrates?  Of a high carbohydrate, low fat diet?  

Which study is the *real* test of the diet?

Which study is the more applicable to the general population of males and females for whom we are trying to recommend a healthy diet?

Teicholz, after dismissing Esposito as a study that "could not be generalized" goes on to say this in the final chapter of her book (which is basically propaganda for Atkins funded or affiliated researchers Eric Westman, Jeff Volek and Stephen Phinney and NuSI affiliated Gary Taubes and Ronald Krauss.) 

The one statement that seems safe to make is that the refined carbohydrates and sugars that we were recommended to eat by the AHA as part of a healthy, fat-avoiding diet , are not merely indifferent, “empty calories,” as we’ve long been told, but are actively bad for health in a variety of ways. ^XXV Moreover, the clinical trials [sic] in recent years imply that any kind of carbohydrate, including those in whole grains, fruits, and starchy vegetables, are also unhealthy in large amounts. Remember that the Shai study in Israel found that the Mediterranean diet group, eating a high proportion of calories as these “complex” carbohydrates, turned out to be less healthy and fatter than the group on the Atkins diet, although they were healthier than the low-fat alternative.    {5356}

Those statements (blue and red), folks, are outright lies.  Although I suppose by the definition of a lie, it might not be as Teicholz summarily dismissed Esposito as being remotely relevant.  So she could, I suppose, claim ignorance ... or scholarly incompetence as her mentor Gary Taubes is wont to do.

This book is such a joke.

---

More on Nina Teicholz and The Big Fat Surprise:

• Ladies and Gentlemen I present .... The new Gary Taubes. Nina Tiecholz
• Book Review ~ Big Fat Surprise by Nina Teicholz
• No Big Surprise ... A Compendium of Errata Etc. from The Big Fat Surprise by Nina Teicholz