Are the Seed Oil Trials Confounded by Trans Fats?

Part 1: The LA Veterans Administration Hospital Study

The randomized controlled trials of seed oils done in the mid-20th century in net showed that seed oils are not good at preventing heart disease and probably raise the risk of cancer.

The idea behind these trials was that the polyunsaturated fatty acids (PUFAs) in seed oils will lower cholesterol levels in the blood and thereby prevent heart disease.

Detractors from this idea focus on PUFAs making our tissues highly vulnerable to oxidative damage — very similar to the rust that accumulates on metal objects kept outdoors over time — and maintain that these fats could worsen the risk of heart disease by making the lipoproteins that carry cholesterol in the blood, such as LDL, more vulnerable to oxidative damage.

The failure of the seed oil trials to show a convincing reduction in heart disease and their worrisome finding that they may increase the risk of cancer suggests that this vulnerability to tissue damage is the mechanism that wins out in the long run.

One of the criticisms of these trials is that they are confounded by the use of trans fats in the seed oil groups.

While there are some trans fatty acids that occur in nature, such as in beef or dairy fat, there are specific trans fatty acids that occur in massive amounts in partially hydrogenated vegetable oils that do not occur in nature.

Hydrogenation itself is done to make liquid oils harder. When carried to completion, it turns all of the monounsaturated and polyunsaturated fatty acids to saturated fatty acids.

However, the path from a polyunsaturated fatty acid to a saturated fatty acid involves a monounsaturated fatty acid as an intermediate.

The definitional chemical characteristic of a saturated fatty acid is that it contains only single bonds and no double bonds, whereas monounsaturated fatty acids contain one double bond and polyunsaturated fatty acids contain more than one. The path from a natural double bond found in an unsaturated fatty acid to the single bond of a saturated fatty acid involves the formation of an unnatural trans double bond.

Thus, partial hydrogenation causes a loss of PUFA, but an increase in both monounsaturated and saturated fats, and the unique production of unnatural trans fats that are not found in any other foods.

Partially hydrogenated oils have been effectively banned in the United States and various other countries, but were widespread during the time of the classic seed oil trials.

Were these trials truly confounded by these unnatural trans fats?

We start our analysis with the longest seed oil trial ever conducted, the eight-year LA Veterans Administration Hospital Study.

The claim that it is confounded by trans fats relates to the use of a margarine in the seed oil group. There were actually two margarines used, one which was free of trans fats and one which contained trans fats. Here, we dig into historical sources such as advertising materials, preserved boxes with ingredient labels, patent infringement lawsuits, and scientific papers of the era. We blend this with direct analyses of the margarines and foods made with them and physiological biomarkers of trans fat consumption to narrow in on the best answer.

But first, why was this trial so incredibly important?

Download This Article

Masterpass members can download this article to print it out and read it in natural lighting as recommended in What Everyone Should Be Doing For Their Health here:

What the LA Vet Trial Tested

The LA Veterans Administration Hospital Study was just over eight years long. It enrolled its subjects between 1959 and 1967, completed in 1968, and had its major publication published in 1969.

Most of the subjects were enrolled in the first two years of the trial, so most of the subjects who survived were in the trial for 6-8 years.

856 male veterans ages 55 or older staying at the Los Angeles home for disabled veterans were randomized to one of two diets that each provided 40% of calories from fat: either a control diet or a diet involving “a replacement of [two thirds of] the saturated animal fats and hydrogenated shortenings of the conventional diet by equal quantities of unsaturated fat in the form of vegetable oils.”

Apart from this change, the diets were nearly identical.

Men under 55 years old were excluded because they tended to have short stays at the home and heart disease takes a long time to develop.

The major importance of this trial lies in these three points:

1. It is one of only two double-blind trials, the other being the Minnesota Coronary Survey, that made a specific substitution of seed oils for traditional fats.

2. It is the longest of any of these trials, most of which were only five years long.

3. It is the only trial with a mean age above 65, which allows us to see the impact of seed oils on cancer.

There were two dining halls, one providing the control diet and one providing the seed oil diet.

The subjects were randomized to receive punch cards that only worked at one or the other dining hall.

The punch cards were used to track how many meals they ate, as they were allowed to leave the home and eat elsewhere.

In order to best preserve blinding, the institutional diet was changed from its historical diet and everyone was told they were on a special diet. Using this new, altered diet as a base, the two dining halls fed the exact same foods except that seed oils replaced traditional fats in one of the halls and only one egg per day was allowed in that hall.

In the first two years of the trial, over 130 people withdrew and there were 53 more withdrawals in the seed oil group than the control group. However, in the last six years of the trial there were only a few withdrawals per year and they were evenly distributed between groups.

The chief complaints in the excess of people withdrawing from the seed oil group in the first two years were disliking the flavor of the milk and complaining about the restriction of eggs. The milk had its fat removed and replaced with seed oils. The investigators improved this process after the first two years to generate a better flavor and people stopped complaining about it.

It appears the people who hated the restriction of eggs most all left during this time and were effectively weeded out. Either the same thing happened to the people who disliked the flavor of the milk or the investigators genuinely closed the gap between the taste of the seed oil-filled milk and regular milk.

Overall, this suggests there was a minor problem with the blinding of the subjects in the first two years but it was subsequently resolved.

The physicians were surveyed in the seventh year to see if they could guess which group the subjects were in. From 229 attempts, they got the control group right 49% of the time and the seed oil group 54% of the time, which is indistinguishable from a coin flip. So, the physicians were well blinded.

While the subjects were living in the home, they ate on average 80% of their meals there.

Because many of the subjects had stopped living there before they died or the study terminated, the total number of meals eaten in the halls was 56% for the control group and 49% for the seed oil group.

Since roughly half of the usual diet was replaced by the seed oil diet in the seed oil group, and since the seed oil diet replaced two-thirds of the traditional fat with seed oils, this is best described as a highly effective randomization to replace one third of the total traditional fat the person ate with seed oils.

At each meal, they took one representative tray and saved it for analysis. Once a week they pooled together the saved trays, homogenized them, and performed chemical analysis of the nutrition. This allowed exquisite characterization of the nutritional content of the diets.

They had originally planned to weigh the food left over, but they found that the amount thrown out was trivial so they stopped doing this. When people asked for second helpings, they gave them foods that were low in fat. This helped preserve the rigor of the quantitative difference in fats between the two groups.

What the LA Vet Trial Found

In the graphs below, “experimental” or “EXP.” refers to the seed oil group.

This graph shows survival from fatal atherosclerotic events over the eight years of the study:

Because the graph shows survival, the line being higher means more people survived and is good.

There are fewer fatal atherosclerotic events in the seed oil group, but the difference is not statistically significant.

The difference begins emerging after the first year, gets wider over time, and spreads apart most in the last two years of the trial.

Statistical significance could be achieved by grouping together the incidence of all “hard endpoints” including definite heart attack, sudden death due to coronary heart disease, stroke, ruptured aneurysm, and amputation due to peripheral artery disease. This brought the p value down to 0.02.

They also achieved statistical significance for fatal atherosclerotic events by separating those between the ages of 54 and 65 from those older than 65.

The benefit was largely confined to the younger group:

Statistical significance is a completely arbitrary cutoff. The p value is best understood as a continuous variable that tells us the probability we would see a difference this large or larger produced by random chance alone. A p value of 0.02 means this probability is 2%.

If we take the finding that seed oils prevent heart disease in young people but not old people at face value, this suggests one of two things are true:

• You need to intervene with seed oils early enough in the course of disease to impact it.
  
  OR
  

• As people get older, everyone will become vulnerable to heart disease and seed oils can’t do anything about it. However, they can prevent someone from getting heart disease much younger than they otherwise would.

The other double-blind trial, the Minnesota Coronary Survey, shows that seed oils make heart disease worse, not better. We will cover that trial in the next article.

The incidence of fatal cancers was higher in the seed oil group:

Because this graph shows the incidence of fatal cancer, the line going up further means more people died of cancer and is bad.

The difference didn’t even start to emerge until year two and really took off in years 5-8. It just barely missed statistical significance at p=0.06. This means the chance of observing a difference this large or larger if seed oils don’t really raise the risk of cancer is 6%.

While there were more deaths from atherosclerosis than from cancer (70 in the control group and 48 in the seed oil group), the relative difference was almost twice as large for cancer deaths: 31 in the seed oil group and 17 in the control group. That’s 46% more atherosclerotic death in the control group but 82% more cancer deaths in the seed oil group.

Taking both findings at face value, seed oils saved 22 people from dying of atherosclerosis but killed 14 people with cancer.

Critics point out that people in the seed oil group who died of cancer had lower adherence than others, but “adherence” was defined as meals eaten, and cancer decreases appetite.

Examining adherence is also an observational study nested inside the trial and leaves the realm of strong causal inferences made from randomization because people weren’t randomized to adhere more or less strongly.

The greater cancer deaths in the seed oil group is especially remarkable because randomization failed to distribute smokers evenly. The control group had twice as many heavy smokers and 60% more moderate smokers!

In fact, the excess atherosclerotic death was found exclusively among those in the control diet smoking more than ten cigarettes per day:

Smoking raises the need for vitamin E, and the control diet was remarkably deficient in vitamin E.

The only clear role of vitamin E in the body is to protect delicate polyunsaturated fats (PUFA) in our cell membranes from being damaged by metabolism or from toxic exposures such as cigarette smoke. It is best evaluated by its ratio to the PUFA it is protecting.

Seed oils have much more PUFA than other fats, so have more vitamin E. However, since the seeds that seed oils come from are not metabolically active, the vitamin E does not have to work very hard. Once these fats accumulate in human tissues, the vitamin E gets used up faster due to the damaging byproducts of human metabolism.

The control diet had barely more than a third of what is found in run-of-the-mill commercial butter and far less than what is found in grass-fed butter, while the seed oil diet was just below the optimal ratio of 0.6 milligrams alpha-tocopherol per gram of PUFA.

Smoking raises the requirement for vitamin E, so the excess of atherosclerosis is best attributed to an interaction between cigarette smoking and vitamin E deficiency.

These should have also biased toward greater cancer, yet the fatal cancer was almost doubled in the seed oil group.

There was no difference at all in total mortality:

However, it seems like the total survival might be starting to diverge after the eighth year, with a higher death rate in the seed oil group emerging.

The beginnings of divergence after the 8-year mark represents a three-month period during which everyone was on the control diet while undergoing final examinations. If this is a real divergence, it is likely a result of 6-8 years of seed oil consumption increasing tissue PUFA content and thereby increasing the vitamin E requirement, while the vitamin E intake was suddenly dropped to the deficient level of the control diet.

However, the excess of non-atherosclerotic death starts diverging in year 7, not after the trial was done:

The reason this doesn’t show up in the total survival graph is because the seed oil group’s advantage for atherosclerotic death is increasing at the same time. Thus, one for one, every person that the seed oil diet saved from an atherosclerotic death was balanced by someone who died of something else.

The most rigorous causal inference from the randomization is that seed oils decreased death from atherosclerosis, increased death from cancer, and had no impact on total mortality.

However, other important observations include that the difference in fatal cancer and non-atherosclerotic death takes years to show up and gets worse as time goes on, and that the putative beneficial effect on heart disease is probably a mirage created by the interaction between smoking and vitamin E deficiency.

The interaction between smoking and vitamin E deficiency should have biased the trial toward worse outcomes in the control group. That total mortality simply broke even suggests the PUFA load in the seed oil group killed people to the same extent as smoking and vitamin E deficiency killed people in the control group.

While not fatal, the seed oil diet also caused a 2.4-fold increase in the the risk of gallstones.

The height of bias would be to reject the cancer finding and accept the heart disease finding at face value.

The level of bias required to do this is extraordinary, for the following reasons:

• First, statistical significance was only found for atherosclerotic death by p-hacking and data mining, either by grouping “hard endpoints” that weren’t fatal with actual fatalities or by subgrouping the atherosclerosis fatalities by age. The non-hacked p value for atherosclerosis fatalities was 0.26, 4.3-fold worse than the non-hacked p value for cancer. The effect size for cancer is almost twice as big.

• Second, the increase in non-atherosclerotic death is, while not straightforwardly “statistically significant,” every bit as statistically powerful as needed to completely abolish any benefit toward atherosclerotic death.

• If the total mortality is identical, you cannot rationally claim a net benefit.

One of the major conclusions of the authors was that they had shown that these trials need to be even longer than eight years, not shorter:

This small excess of nonatherosclerotic mortality in the late years of the study raises the very important and difficult question of whether future clinical trials of diets rich in unsaturated fat must be planned for periods well in excess of eight years rather than for the five-year periods that have been the usual goal.

They considered it “unresolved” whether seed oils have “toxicity” and took the lack of an effect on total mortality to leave their usefulness for heart disease prevention a question without a definitive answer:

Total longevity was not affected favorably in any measurable or significant degree. . . . For this reason, and because of the unresolved question concerning toxicity, we consider our own trial, with or without the support of other published data, to have fallen short of providing a definitive and final answer concerning dietary prevention of heart disease.

Was the LA Veterans Administration Hospital Study Confounded by Trans Fats?

According to the authors, what they were testing was “a replacement of the saturated animal fats and hydrogenated shortenings of the conventional diet by equal quantities of unsaturated fat in the form of vegetable oils in the experimental diet.”

Taken at face value, this suggests the trans fats would be higher in the control diet due to the hydrogenated shortening.

However, there was a margarine used in the seed oil diet that may have increased trans fat consumption in this group.

The following quote from the primary 1969 paper shows the authors themselves considered both diets to be low in trans fats:

We did not analyze our diets or tissue lipids for trans fatty acids. Presumably both diets contained some trans fatty acids from partially hydrogenated fats, but such fats were only minor components of both diets. Therefore, the cholesterol-elevating⁵⁹ and triglyceride-elevating⁶⁰ effects of unnatural fatty acids cannot have been an important factor in this trial.

But is there more to the story?

As they detailed in a 1962 paper, approximately 32 grams of fat per day in the seed oil diet — 28.5% of the total fat — came from an “unsaturated margarine.” 75% of this was used as a table spread, while the rest was used to bake deserts or cook potatoes.

This margarine is the putative source of trans fats in the seed oil group.

But did it contain trans fats?

If it did, how much?

Was it enough to exceed the partially hydrogenated shortening that was explicitly used in the control group?

Let’s dig in.

What Was the “Margarine Based on Unhydrogenated Corn Oil”?

They reported about this table spread that “during the major part of this study, a margarine based on unhydrogenated corn oil has been used.”

According to the footnotes, “this term is used to describe a margarine in which the major component is unaltered vegetable oil, with a small amount of hydrogenated fat added as a hardening agent” and in “the major part” of the study up through at least 1962 the margarine used was Mazola by Corn Products Company.

The funding and gifts statement includes both Mazola Margarine provided by Corn Products Company and Emdee Margarine donated by the Pitman-Moore Company.

Corn Products Company had recently been formed by a merger between the Corn Products Refining Company and Best Foods, while Pitman-Moore was a pharmaceutical company that was a division of Allied Labs, which became part of Dow Chemical in 1961.

The authors wrote two 1962 papers, one on diet and one on “preliminary observations.” They also published a study on the vitamin E status of subjects in the trial in 1965. All three of these papers as well as the primary paper from 1969 refer to both of these margarines, but none of them clarify how much of each was used, whether they were used at the same time, or whether they transitioned from one to the other and when.

We will try to cut at this from several angles later, but first we look at how much trans fat we can expect to be in each of these margarines.

Emdee Margarine Was Free of Trans Fats

It is easiest to show that Emdee margarine did NOT contain any meaningful trans fats, so we start with Emdee.

According to an early advertisement in the Archives of Internal Medicine, the patent-pending margarine was produced by a “special process” that does “not alter the natural characteristics of corn oil.” It consisted of 80% “nonhydrogenated corn oil which has been specially processed to preserve its original content of unsaturated fatty acids” and was fortified with vitamins A and D. It was sold over-the-counter, but was only available in pharmacies. It was marketed as an “effective substitute for animal fats and hydrogenated vegetable oils.”

According to a documents from a patent lawsuit that mentioned Emdee’s composition by way of evaluating a dispute between Mazola and Fleichman’s, the other 20% of Emdee was “fully hydrogenated coconut oil.” This matches the listing of “hardened coconut oil” in the ingredients list of the last ad picture above.

Emdee was associated with US Patent 2,890,959, granted to Robert A Phillips and assigned to Allied Laboratories.

This Emdee can being sold on eBay explicitly references this patent:

Emdee ads started referencing the patent as soon as it was granted in 1959.

The patent itself does not explicitly require full hydrogenation, but it does require melting points so high that they can only be achieved with full hydrogenation, so it effectively requires full hydrogenation of the coconut oil.

While the patent prefers coconut oil, it does not require it. It explicitly gives the example of using fully hydrogenated soybean oil as an alternative and states that any other vegetable oil with a melting point in the correct range could be used.

Fully hydrogenated oil by definition contains no trans fats, because all of the unsaturated fats are converted to saturated fats. It is only partially hydrogenated oil that contains trans fats.

Clearly, the Emdee margarine used in the LA Vet trial was free of any meaningful trans fat.

There is one piece of major uncertainty here: while there is no evidence that coconut oil was ever removed from their ingredients, there is also no evidence that its use persisted past 1962. The ad showing the can with the ingredients cannot be found after 1958. No date can be safely attributed to the can above being sold on eBay, which shows coconut oil in the ingredients. The materials in the Mazola/Fleichman’s suit were submitted in 1962, despite being referenced in the final appeal decision in 1965.

The reason this is important is that coconut oil contains a highly unique fatty acid, lauric acid, that is not found in many other foods and is largely limited to coconut and tropical foods highly related to coconut. This can act as a “fingerprint” of Emdee consumption in the LA Vet trial, but until and unless evidence for the Emdee ingredients between 1963 and 1968 show up we cannot assume the fingerprint will work after 1962.

Mazola Margarine Was Consistently Marketed as Unhydrogenated

Now let’s tackle the slightly less straightforward question of whether Mazola margarine had trans fat.

Mazola consistently emphasized that lack of hydrogenation was one of the major virtues of its margarine. Listen to the woman in this 1961 Mazola Margarine TV commercial explain it:

“I used to buy a corn oil margarine which I thought gave me the same benefits as Mazola, but at a recent food demonstration I learned that the ordinary corn oil margarine I had put such faith in is hydrogenated. So most of its natural liquid corn oil nutrition is lost! But the corn oil in Mazola margarine is never hydrogenated, so you get the full nutritional benefit of Mazola’s liquid corn oil.”

Or take it from the man in this earlier 1960 TV commercial:

“Mazola Margarine — because its major ingredient, pure corn oil, has less saturated fat than the hydrogenated corn oils used in other leading margarines. Less saturated fat because its corn oil is never hydrogenated.”

In a 1960 ad in the Journal of the Medical Society of New Jersey, they had bolded “non-hydrogenated” and used all-caps to say again that its corn oil was “NOT hydrogenated.” This version was also repeated in 1960 issues of the Rhode Island Medical Journal, the Bulletin of the New York Academy of Medicine, and the Illinois Medical Journal; and in 1961 issues of the Journal of the American Osteopathic Association, Industrial Medicine and Surgery, Geriatrics, The Modern Hospital, and Hospital Management.

They used italics to emphasize that the corn oil is never hydrogenated in 1962 issues of Time, the New Yorker, Good Housekeeping, and Ladies’ Home Journal; and in 1963 issues of Better Homes and Gardens, American Home, and Ladies Home Journal.

A 1963 ad in an Ohio newspaper and a California newspaper used italics to emphasize that the “major ingredient” in Mazola margarine is never hydrogenated, and even offered a raffle to win $20,000 that required the person to use their knowledge from the first page to complete this sentence on the second page: “The corn oil in Mazola Margarine has ________ ________ ________ than the hydrogenated oil most other margarines use.”

A brief “History of the Corn Products Company” in a 1961 issue of the Journal of the American Oil Chemists Society says that “Mazola margarine” was “developed in response to demands for new foods containing unhydrogenated corn oil, as a source of the linoleic acid now known to be so important in human nutrition.”

Despite Its Marketing, Mazola Was Made with Partially Hydrogenated Seed Oils

According to a 1961 article about the upcoming “Cholesterol War on TV” in the programming magazine Sponsor, the composition of Mazola margarine was as follows:

This Corn Products spread is 47% corn oil and 53% combination cotton seed and soybean oil. The latter two are hydrogenated until they form irregular walls inside which the liquid corn oil is trapped.

This echoes an earlier 1960 blurb in the journal Food Processing:

Seeming paradox of a firm liquid-oil-containing product is result of revolutionary process whereby a matrix of partially hydrogenated oils holds sub-microscopic droplets of liquid Mazola Corn Oil, much as a honeycomb holds liquid honey.

A holder of the patent connected to Mazola Margarine, Daniel Melnick, published a paper in 1964 that described how it is made. While he didn’t name it as Mazola, he directly tied it to the Mazola patent and his description about the “major ingredient” strongly echoes the marketing of Mazola:

The first, hereafter called blended corn oil margarine fat (BCOMF) is a liquid oil margarine (27) made with non-hydrogenated corn oil as its major ingredient; other components of the blend are a lightly hydrogenated cottonseed oil and a partially hydrogenated soybean oil. The corn oil in this product provides about 90% of the total linoleic acid; the remaining 10% is contributed by lightly hydrogenated cottonseed oil.

The Main Mazola Patent Implies It Has Trans Fats

The main patent that we know is tied to Mazola Margarine, the one Melnick cited, is consistent with all of the above evidence that it contained trans fats.

A 1961 article in Advertising Age gives important background to what we know about the main patent.

Mazola Margarine was originally branded as Cornette in 1959 when it was tested on six markets but then released to the broad market in late 1960 as Mazola.

Fleichman’s beat them to the punch with a corn oil-based margarine, which they released several months before Corn Products started testing Cornette. By 1961, Fleichman’s had double the market share as Mazola.

Fleichman’s made the claim to be 100% corn oil, but all of its corn oil was partially hydrogenated, making it lower in natural polyunsaturated fatty acids.

Mazola made the claim that their corn oil was pure and unaltered, making it higher in these fatty acids.

Fleichman’s second product, however, was just as high in polyunsaturates as Mazola. Hence Mazola had lost their differentiator.

The dilemma at Corn Products is: How can they reach the consumer with the message that Mazola is a superior product, from the polyunsaturate point of view, to the salted Fleichman’s brand?

By 1962, the answer had become clear: sue them for patent infringement.

The lawsuit claimed that both Mazola Margarine and a new reformulation of the older Nucoa Margarine, also made by Corn Products, were covered by the Melnick and Luckmann Patent, No. 2,955,039, which was filed in February of 1959 and granted in October of 1960. The Fleichman’s Regular and Fleichman’s Unsalted margarines were claimed to infringe the patent.

Mazola Margarine ads began in October of 1960 emphasizing this patent:

That ad is from the October 11, 1960 Stamford, CT newspaper the Stamford Advocate.

Ads bearing the patent were repeated throughout 1961, including in a full-page ad in the Time magazine issue that featured Ancel Keys on the cover, in issues of McCall’s, the Saturday Evening Post and in a bulk mailer, and in at least 50 newspaper ads.

Mazola advertisements after 1961 never reference this patent, most likely because the litigation started with pretrial conferences in January of 1962.

The patent infringement claim was ultimately rejected by an appeals court in 1965 on the basis that the technology wasn’t actually novel and that the claims were far too broad.

Good Luck margarine had used a blend of liquid and hydrogenated oil from 1937-1942, and Emdee had made a margarine based on unhydrogenated corn oil in 1958, prior to both Fleichman’s and Cornette in 1959 and Mazola in 1960.

The court ruled that it was clear that Emdee was “the spark” that drove the subsequent development of both Fleichman’s and Mazola.

While the patent was ruled invalid, the patent itself still tells us how the margarine was made and how Melnick, Luckmann, and Corn Products Company viewed the importance of the technology they used.

According to the patent, the novelty of the covered margarine was that it was high in essential fatty acids and had a high ratio of essential fatty acids to saturated fatty acids.

It doesn’t mention anything about trans fats.

The process described involves using 30-70% unhydrogenated liquid oil with the remainder a hydrogenated oil. The hydrogenation used settings that would cause it to yield “substantially complete” hydrogenation of linoleic acid and “selectivity” that would bias it to hydrogenate polyunsaturated fats like linoleic acid rather than monounsaturated fats like oleic acid.

The fact that the linoleic acid was expected to completely convert to saturated fatty acids while the oleic acid was not suggests that the oil would be particularly high in elaidic acid, the trans isomer of oleic acid.

The hydrogenation described is indeed partial hydrogenation, which generates trans fats, consistent with early descriptions in the press and Melnick’s description in the science literature.

Mazola Ingredient Labels of the Era Consistently List “Partially Hardened” Oils

Mazola ads almost never show the ingredients and never show the ingredients clearly. Usually, the Mazola box appears as stylized art that readjusts the logo to occupy the space where the ingredients would be so they could be removed without showing any dead space.

Nevertheless, with a little craftiness we can get our hands on clear ingredients lists from that era.

This is the label of the original 1960 release printed in the journal Modern Packaging:

The second ingredient is “partially hardened soybean and cottonseed oils.”

This Instagram post shows this image, bearing the exact same ingredients list, from a set of grocery store posters from 1964. The account holder messaged me a picture of the box the posters came in, which reads “Brand Names Week May 21-31, 1964 Display Ads.”

This is almost certainly the actual box in circulation in 1964. The original 1960 box showed the corn cob facing up-and-to-the-left, not straight-up. The up-and-to-the-left position persists through 1963, but the straight-up position is found at least as early as 1965.

However, beginning in 1965, the Mazola margarine box read “New! See the Difference! Visibly Better! Definitely More Delicious” when it was “now made by a special process so that it doesn’t burn, blacken, or smoke at normal frying temperatures.”

The package remained like this through 1967.

Thus, 1964 is the only year where the corn cob could have been turned straight up before the new formula was announced on the box.

Jason Liebig (Instagram, LinkedIn) the lead storyteller for the History Channel’s “The Food That Built America,” “The Mega-Brands That Built America” and “Hazardous History with Henry Winkler,” is a collector of vintage brand memorabilia and sent me this image of a 1966 Mazola carton:

The ingredients are identical, except that instead of just using isopropyl citrate to protect the flavor, they had then started using calcium disodium EDTA to protect the flavor as well.

The use of isopropyl citrate and EDTA together to protect the flavor was submitted as a Corn Products Company/Daniel Melnick patent in 1959 and granted in 1961 as US Patent 2,983,615.

An eBay seller claims this is a 1960s-era box:

This is, however, most likely a 1970s-era box.

When I asked for a more precise year, the seller told me they had no idea except that it was from before the advent of bar codes. But the first bar code was scanned on June 26, 1974, so the absence of a bar code does not necessarily isolate it to the 60s.

The corporate office of Best Foods is listed on this box as Englewood Cliffs, NJ, whereas the earlier boxes all list it in New York, NY. An article from Long Beach, CA’s Press Telegram from November 6, 1968 refers to an open house for the new corporate office in Englewood Cliffs. This dates the box to late 1968 at earliest.

I cannot find the phrase “contains golden corn oil” in Mazola Margarine ads prior to 1973.

Another differentiator of this box from earlier ones is that it has a picture that zooms in to show a section of the corn rather than the whole cob and there is a stick of margarine sliced into pats with one of the pats melting on the corn.

Yet another is that the box is a rectangle instead of a square.

All of this imagery can be found in the 1970s-era “We Call It Maize” television campaign.

According to the November 5, 1979 issue of the Oakland Tribune, the New Mexican Apache Tenaya Torres struck a deal for this series in 1976.

All of the boxes in all of these commercials have the sliced margarine stick with one pat melting on the zoomed-in section of corn, and all have the “contains golden corn oil” droplet. I cannot find the rectangular box in any extant videos, but this Facebook post has images of Torres using a rectangular one in some of the ads.

Thus, this box almost certainly comes from the 1970s, not the 1960s.

And yet, its ingredients are exactly the same as they were in 1966, which were the same as from 1960-1965 with the sole exception of added EDTA.

This confirms that throughout the entire LA Veterans Administration Hospital Study, Mazola Margarine listed “partially hardened soybean and cottonseed oils” as its second ingredient.

While “partially hardened” is not a chemically specific term and would no longer be allowed on ingredient labels, FDA guidance at the time was wishy washy. In 1971 FDA increased the specificity required on labels above the regime that had prevailed since 1940 by making the industry be specific about what oil was used and always expressing some type of processing word for hydrogenated oils rather than simply calling them “shortening.” Nevertheless, it stated that the choice between “hydrogenated,” “partially hydrogenated” and “hardened” could be made based on whatever was “factual and desirable” without defining those qualities in any way.

“Partially hardened” should be interpreted in the light of the evidence from the patent, early press, and scientific literature, combined with common sense, to mean “partially hydrogenated.” This suggests that Mazola margarine contained trans fats throughout the entire length of the LA Vet trial.

How Misleading Was the Mazola Advertising?

All of the Mazola marketing focuses on the positive characteristic of having natural polyunsaturated fatty acids rather than the absence of the negative characteristic of having trans fats. It consistently prides itself on being higher in “polyunsaturates” or “linoleates,” and having a high ratio of these to saturated fat.

This can be seen visually in the early television ads, where the cups of corn oil are shown disappearing as the hydrogenation is said to deplete their nutrition.

The 1961 history of the company said the demand for unhydrogenated oil was due to its richness in “the linoleic acid now know to be so important to human nutrition.”

In its late 1960s marketing, it starts comparing itself to protein and vitamins. The “plus” of lobster or steak is “protein,” orange juice’s is “vitamin C,” and that of carrots is “vitamin A,” while Mazola’s “plus” is “polyunsaturates.”

The Mazola brand was so obsessed with increasing intake of polyunsaturated fat that it tried turning “polyunsaturating” into a sexy verb.

1969:

1969:

1970:

They even implied Ringo would be polyunsaturating Beatles tunes in the family den:

It was the embarrassing thing you catch your husband doing with himself in the middle of the night:

What you do with your girlfriends on girls’ nights:

And what you vow to the man you love and honor:

This can explain how Mazola must have internally justified its quiet sleight of hand, where it constantly emphasized the fact that its corn oil or its major ingredient was non-hydrogenated, never even mentioning in the ad copy that it had cottonseed and soybean oil that was hydrogenated, and never showing the ingredients list that bore the incriminating evidence in those ads.

If the whole point was, in the Mazola view, that hydrogenation destroys the “plus” of the natural polyunsaturates, then there’s nothing particularly misleading about this.

Once we become concerned about the “minus” of trans fats, however, this suddenly becomes deeply misleading.

Mazola Patented A Trans Fat-Free Margarine in 1960 But Never Used It

While Mazola as a brand was exclusively interested in getting more polyunsaturated fatty acids into people’s diets, their oil chemist product developers were indeed interested in reducing people’s intakes of trans fats.

Daniel Melnick and Chester Gooding filed a Corn Products Company patent on June 19, 1957 (granted Jan 19, 1960) aimed at using interesterification to produce a margarine with “relatively low trans fatty acid content.”

Interesterification rearranges the three fatty acids on the glycerol backbone of a triglyceride to make them pack together better or give them some other desirable property.

“Hydrogenated fats,” they wrote, “have been subject to frequent criticism in both the scientific and lay press since there is a decrease in essential fatty acid content and a development of isomers of fatty acids, e.g. trans oleic acid, as a result of hydrogenation of oil.”

The specific goal was to get trans fat down below 10%.

The account Melnick submitted to the court in the Mazola/Fleichman’s patent suit was that this patent was never commercialized because it wasn’t “successful,” which presumably means it didn’t produce a margarine that they thought would sell well.

Melnick told the court that his “three prongs” were “reduce saturates, eliminate all isomers of fatty acids, and increase the polyunsaturates.”

“Eliminate all isomers” means to produce a margarine that had zero trans fats.

Chester Gooding filed a second Corn Products Company patent later that year in August of 1960 (granted in mid-1963) with the explicit rationale to replace partially hydrogenated margarines and shortenings with healthier versions that are rich in essential fatty acids and have no trans fats.

This “invention” was a “novel fat” that could be used to produce “margarines and shortenings” that are “characterized by having a high essential fatty acid content” and have “the physical characteristics of fats prepared by partial hydrogenation but which, unlike the latter, contain no partially hydrogenated fats and therefore no glycerides containing isomers of natural unsaturated fatty acids.”

The rationale from the Melnick-Gooding patent was repeated: “Hydrogenated fats . . . have been subject to frequent criticism in both the scientific and lay press since there is a decrease in essential fatty acid content and a development of isomers of fatty acids, e.g., trans oleic acid, as a result of hydrogenation of the vegetable oil.”

The patent covers eight blends. Seven of them use 60, 70, or 80% cottonseed stearine, which is a largely saturated product derived from the refining of cottonseed oil. Blend eight is the exception, and is over 74% corn oil.

The remainder was one of eight different mixtures of fully hydrogenated “interesterified base fat” that could be mixed and matched with the main blend.

The various base fat combinations were made from coconut and cottonseed, palm kernel (very similar to coconut oil) alone, coconut and palm kernel, palm kernel and cottonseed, or coconut and soybean.

Blend 8 using 74% corn oil with the fully hydrogenated and interesterified coconut oil and/or palm kernel oil base fat would have produced a margarine that would be both remarkably similar in its composition to Emdee and consistent with the Mazola marketing that the major ingredient is liquid golden pure unhydrogenated Mazola corn oil, yet would be free of trans fats.

Nevertheless, this patent seems to have languished unused.

No reference to it occurs in the scientific literature until one paper in 1974 and two papers in the 1990s reference it purely to note that it is possible to produce a trans fat-free margarine and that methods to do so have already been patented.

The one time we know they changed their formula, they acted on a five-year-old patent, listed the change in the ingredients list, and introduced it with fanfare on the box label. The absence of any of this for a margarine free of trans fats strongly supports that they never acted on the Gooding patent.

Despite Gooding and Melnick being aware of “frequent criticism in both the scientific and lay press,” public awareness and interest in trans fats was subdued until the mid-1990s. The overwhelming push during the 1960s was to increase intake of PUFA and decrease intake of saturated fat to lower plasma cholesterol levels, and this was where Mazola staked their brand reputation.

Direct Analysis of Mazola Margarine

A 1962 study analyzed seven commercially available margarines. Their seventh was 20.8% linoleic acid and 50.2% trans fat. Since partial hydrogenation destroys linoleic acid and creates trans fat, and since most margarines at the time were low in linoleic acid and high in trans fat, the high linoleic acid content made them speculate that it was based on US Patent 2,983,615. This is the Melnick/Luckmann patent tied to Mazola and Nucoa. This margarine was 50% trans fat.

Daniel Melnick, named on US Patent 2,983,615, co-authored a 1964 paper analyzing a margarine that they directly stated was based on this patent. They strongly echoed the Mazola margarine advertising by saying it was made with “non-hydrogenated corn oil as its major ingredient.” The “other components of the blend,” they wrote, were “lightly hydrogenated cottonseed oil and a partially hydrogenated soybean oil.” This matches the marketing, ingredients, and descriptions of Mazola margarine and it was 25% trans fat.

These two groups used the same methodology to measure trans fats but the first margarine had double the trans fat content of the second, so it is very unlikely these are the same margarine.

The first paper probably refers to Nucoa. It was published in June of 1962, so probably submitted a few months earlier. This was after the 1961 Mazola ads explicitly tying US Patent 2,983,615 to Mazola, but before anything about the 1962 patent suit was made public that would have tied Nucoa to the patent. This would explain why they would use speculative language even though they knew what they bought in the grocery store: Nucoa was made by Corn Products Company and plausibly made according to the Corn Products patent, but wasn’t known to be tied to the patent yet the way Mazola was.

Thus, our best estimate is that Melnick measured Mazola margarine in 1964 as being 25% trans fat.

Which Margarine Was Used In Most of the Trial? Emdee or Mazola?

The LA Vet Trial investigators acknowledged using both Emdee and Mazola margarines in at least four of their papers, none of which explicitly state how much of each was used and whether they transitioned at any point to choosing one or the other.

This is critical, because Emdee Margarine did not have trans fats and Mazola Margarine did.

As such, this question makes or breaks the case of whether trans fats were in the diet of the seed oil group.

We now look at several forms of evidence: textual evidence, historical availability, dietary analysis, and biomarkers.

What the Trial Investigators Actually Said

We start first with the LA Vet trial authors comment that trans fats were “minor components of both diets.”

We then turn to their statements about the use of Mazola rather than Emdee “during the major part of the study” up through 1962.

What Does Trans Fats Being “Minor Components” Mean?

The following quote from the primary 1969 paper shows the authors themselves considered both diets to be low in trans fats:

We did not analyze our diets or tissue lipids for trans fatty acids. Presumably both diets contained some trans fatty acids from partially hydrogenated fats, but such fats were only minor components of both diets. Therefore, the cholesterol-elevating⁵⁹ and triglyceride-elevating⁶⁰ effects of unnatural fatty acids cannot have been an important factor in this trial.

What is a reasonable definition of “minor” here?

Since the next sentence situates this statement with respect to whether trans fats could have had impacts on the trial participants discussed in references 59 and 60, our minimum definition of “minor” has to be “well below what increased cholesterol and triglycerides in those studies.”

Reference 60 is an abstract of an experiment where 1.6-3.9-fold increases in plasma triglycerides were produced by feeding five people diets that were 20-40% of total calories as trans fats. As such, it is not useful for drawing a lower bound beneath which we could consider something “minor.”

On the other hand, reference 59 is very useful for this purpose.

This is a 1961 paper by Ancel Keys and colleagues reporting three controlled experiments, each using 23-27 institutionalized psychotic men, showing the impact of trans fats on serum cholesterol. These experiments used either 30 grams or 100 grams per day of partially hydrogenated oil and compared them to the same amount of the same oil in its natural form, or compared them to blends of other fats and oils that were richer in monounsaturated or saturated fats.

The partially hydrogenated oils were 33-37% trans fat, so the lower bound is 9.9 grams of trans fat per day and the upper bound is 37 grams of trans fat per day.

30 grams of partially hydrogenated safflower oil containing between 9.9 and 11.1 grams of trans fat per day increased serum cholesterol by 10 mg/dL compared to unhydrogenated safflower oil, but this was similar to what would have been predicted by the increased saturated fat in the partially hydrogenated oil. This made it impossible to attribute a specific rise to the trans fats.

By contrast, 100 grams of partially hydrogenated safflower or corn oil providing 33-37 grams per day of trans fats raised serum cholesterol by 25 mg/dL, and this was unambiguously more than what could be attributed to their saturated fat.

These experiments led to the preliminary conclusion that each increase of 1% of total calories in the diet from saturated fats with 12-18 carbons (later revised to exclude the 18-carbon stearic acid) predicts an increase of 2.68 mg/dL serum cholesterol, whereas the same increase in trans monounsaturated fat would predict a rise of 2.1 mg/dL.

The 30 grams of partially hydrogenated safflower oil provided trans fat at 2.7-3% of calories, which the prediction equation derived from the second experiment suggests should have caused a 5.7-6.4-mg/dL rise in serum cholesterol.

This indicates that the trans fats were not failing to increase the serum cholesterol in the first experiment. Rather, the fact that the partially hydrogenated oil had similar amounts of saturated and trans fats made it hard to mathematically tease their impact apart.

“Minor” could mean many things to many people, but here it is meant to dismiss a potential confounder, so it means “we can ignore this because its impact, if any, would be trivial.”

In the context of a randomized controlled trial, it also means “we can assume this was sufficiently distributed across the two groups that it could not have created a meaningful inaccuracy in the estimation of the difference in outcomes between the two groups.”

To satisfy the first definition, the trans fats should have been under 1% of total calories. It is below this point that it would be difficult to detect any impact even under very rigorously controlled conditions. In the LA Vet trial, where the average energy intake was 2496 Calories per day, this would amount to 2.8 grams of trans fat per day.

However, I could see them rationalizing the sloppier definition of less than 30 grams per day of partially hydrogenated oil or less than 10 grams per day of trans fat because this is the lower dose fed in the Keys paper that on the surface was claimed to not show an impact.

In the context of the caloric intake in the LA Vet trial, according to the Keys paper, 10 grams of trans fat per day would translate to a 7.6-mg/dL difference in cholesterol levels. This could only be considered “minor” if it was thought to be spread equally across the two groups.

The average difference in serum cholesterol between the two groups in the LA Vet trial was 29.5 mg/dL. A 7.6-mg/dL theoretical difference is 26% of the actual difference, which is not a “minor” portion of it.

On the other hand, if a 7.6-mg/dL increase was distributed into both groups evenly, to one through shortening and to the other through margarine, I could see them saying that it didn’t matter because the seed oils still created a 29.5-mg/dL difference between groups.

In other words, the confounder would blend into the sameness of the two groups, and the difference that rises above this sameness is what is important.

Therefore, I could see two possible definitions of “minor” here:

• Trans fats were less than 2.8 grams per day in both groups, in which case the difference between the two groups doesn’t matter.

• Trans fats are close to 10 grams per day, but they are evenly distributed between the two groups.

What Does Using Mazola “During The Major Part of the Study” Up Through 1962 Really Mean?

At least four papers from the LA Vet trial reference the use of both Emdee and Mazola margarines.

The only statement indicating the dominance of Mazola is from the 1962 diet paper:

During the major part of this study, a margarine⁷ based on unhydrogenated corn oil has been used.

Footnote 7 then says “Mazola Margarine, Corn Products Company.”

“The major part” implies that Emdee was used in “the minor part.”

However, papers published in the mid-1960s and 1969 simply list both margarines without making any reference to one being more dominant than the other.

The 1962 paper does not tell us whether they started with Emdee and switched to Mazola rather early on or started with Mazola and switched to Emdee rather recently.

By the time the trial was complete, whatever they used from 1963 onward was “the major part” and whatever was “major” in 1962 could have been rendered irrelevant.

The Historical Availability of Mazola and Emdee Margarines

Recruitment for the LA Vet trial began in the summer of 1959, and the first man was assigned to a diet in September of 1959. Everyone was switched onto the control diet eight years later in September of 1967 during a 3-month final observation period.

Thus, the seed oil diet had a margarine demand from September of 1959 to September of 1967.

Emdee was available at least as early as April of 1958.

Emdee almost immediately was used for scientific research. A major early study they boasted about in their advertising was published in the May 16, 1959 edition of JAMA. Emdee was fed to 300 institutionalized psychotics for five months, during which it reduced plasma cholesterol by 23% from an average of 250 mg/dL to 193 mg/dL.

Ads for Emdee continued running in newspapers through late 1968, as shown by this drug store ad from the November 18, 1968 edition of the New Orleans-based The Times-Picayune:

Emdee Margarine completely disappears from newspaper ads after November of 1968. It lingers into the 1970s in catalogs of drugs, manufacturers, and trade names, and it makes its way into a 1990 book about companies and their brands and a 2013 book and its 2021 edition about the politics of nutrition. There is no evidence it was ever sold after November of 1968.

Mazola was released into six markets as “Cornette Margarine” in 1959. The first newspaper ad I can find for this margarine is in the October 13, 1959 issue of The Times Record (Brunswick, Maine).

Cornette had its name changed to Mazola as it was released into the broad market in September of 1960.

Mazola was available throughout the rest of the trial period and is still available today.

In 1970, the year after the LA Vet trial had its main paper published, Daniel Melnick, who was named on the main Mazola Margarine patent, co-wrote an article titled Dynamic utilization of recent nutritional findings: Diet and cardiovascular disease that takes a dig at Emdee:

The first so-called “liquid oil” margarine on the market⁸ was a blend of corn oil with hydrogenated coconut oil. This was a soft, shortening-like product packaged in a tin can for sale in drugstores. Although rich in polyunsaturated fatty acids, it lacked the form, texture, and other sensory characteristics of margarine and, in addition, contained an undesirably high level of the saturated fatty acids from hydrogenated coconut oil.

The next sentence references Mazola:

The first margarine in stick form having a relatively high polyunsaturated and low saturated fatty acid content was introduced in 1959.⁹ Other brands soon became available, giving the consumer considerable choice among products meeting the criteria of the American Medical Association.

The paper has a long accounting of the LA Vet trial itself. It does cover their use of margarine, but doesn’t take the opportunity to tie it to Emdee or Mazola.

While Melnick could be taken to imply that Emdee went out of business because consumers didn’t like it, a more plausible reason is found in the writings of Myra Waldo, a food writer of the era.

Waldo was an early advocate of these margarines. Her cake recipes appeared in the March 12, 1959 New York Post and the March 18, 1959 Dallas Morning News using exclusively Emdee margarine. Her 1961 cookbook, Cooking for Your Heart Health, however, had abandoned Emdee in favor of Mazola because Emdee was “much more expensive and available only in drug stores.”

Since Emdee was available before Mazola, this should lean us toward Emdee being used during the first year until Mazola became available in late 1960.

Analysis of the LA Vet Trial Diets

As a blend of 80% corn oil and 20% fully hydrogenated coconut oil, values for these foods in the USDA database predict that Emdee should be 31% saturated, 24% monounsaturated, and 45% polyunsaturated.

The Mazola/Fleichman’s patent lawsuit materials include a reference to a direct chemical analysis that Melnick had done on Emdee showing it to be 28% saturated fat, a little lower than predicted from the USDA database.

Approximately half of the fully hydrogenated coconut oil should be lauric acid, a 12-carbon saturated fatty acid that is found almost exclusively in coconut and some closely related tropical fats. That should make Emdee 10% lauric acid.

Mazola advertised their margarine as being 19% saturated, 53% monounsaturated, and 28% polyunsaturated.

Based on the production method in the Mazola patent, nearly all of the trans fat should be monounsaturated, and according to Melnick’s analysis it is 25% of the total. Thus, the Mazola breakdown is more precisely 19% saturated, 25% trans fat, 28% natural monounsaturated, and 28% polyunsaturated.

Based on 32 grams of margarine being consumed per day:

• Emdee would provide 3.2 grams of lauric acid per day, while Mazola would provide none.

• Mazola would provide 8 grams of trans fat, while Emdee would provide none.

• Emdee would provide 9-10 grams of saturated fat per day, while Mazola would provide 6.

• Emdee would provide 14 grams of polyunsaturated fat per day, while Mazola would provide 9.

• Emdee would provide 8 grams of monounsaturated fats per day, while Mazola would provide 17 grams, but 47% of Mazola’s monos would be trans fat.

The trial investigators aimed to keep the “iodine value” of the diet constant. The iodine value is a measure of the number of fatty acid double bonds and is based on the ability of these double bonds to absorb iodine. If we calculate a simplified version of this as (monounsaturated fat x 1)+(polyunsaturated fat x 2), Emdee is 36 and Mazola is 35, nearly identical.

This means a change from one to the other would not necessitate a meaningful adjustment to the amount of margarine or to the amounts of other oils to maintain a constant iodine value, so should show up as a change in the fatty acid subtypes of the overall diet:

• If they changed from Emdee to Mazola, saturated fat should have gone down by 3-4 grams, polyunsaturated fat down by 5 grams, and monounsaturated fat up by 9 grams.

• If they changed from Mazola to Emdee, the reverse changes should have happened: saturated fat up 3-4 grams, polyunsaturated fat up 5 grams, and monounsaturated fat down 9 grams.

The LA Vet trial published these distributions of dietary fatty acids.

1962:

1964:

1965:

1966:

1969:

In no case where the 12:0 (lauric acid) is reported does it show up beyond a trace level, and a little blip is consistently shown in the control diet but not in the seed oil diet. There is thus no sign of coconut oil consumption in the seed oil group from 1962 through the end of the trial.

This would seem to rule out the use of Emdee after 1962.

It is uncertain that Emdee used coconut oil 1963-1968, but Emdee was definitely using coconut oil from 1959-1962, so it would seem from the 1962 papers that Emdee had already been cut loose.

Now, suppose also that Emdee did change their formula to use a different fully hydrogenated fat, or that Mazola acted on the Chester Gooding patent to make a margarine free of trans fat. Even if neither were marketing these fats publicly it is at least conceivable that they could have issued a special product for the trial. The resulting margarine would be relatively close to Emdee in its composition, so we would expect to see the same types of changes, just without any lauric acid.

A change toward this type of margarine would mostly show up as an increase of 18:0 and 18:2 at the expense of 18:1.

However, there is no indication that there is any sudden change or any drift in the fatty acid composition at all. This makes it incredibly unlikely that any kind of formula change was made after the publication of the 1962 diet paper.

This suggests that Emdee had been cut loose prior to 1962 and Mazola was used exclusively thereafter.

Biomarker Analysis

In 1962, the “preliminary observations” paper reported that lauric acid was 1% of subcutaneous fat prior to randomization; it dropped to “trace” in the control group after one year and to zero by the second year, but in the seed oil group it stayed at 1% after one year and didn’t drop to zero until the second year.

This suggests that the pre-study “regular” diets of the people contained small amounts of coconut, while the trial diets contained little or no coconut, but that something was sustaining the lauric acid content of the subjects in the seed oil group for the first year of the study that had stopped sustaining it by the second year.

Serum and adipose fatty acids were reported for each year up through year five in 1966, but lauric acid was not included.

In the 1969 main paper, lauric acid is described as “identified inconstantly” both at baseline and at the end of the trial in both groups.

The change from 1% to “identified inconstantly” in the baseline measurements likely reflects the expansion of tissue biopsies from 75 people to 240 people and may reflect the fact that prior to the study some people ate coconut and others didn’t.

The 1969 paper does not report an expanded selection of year-one biopsies and therefore never claims that that first-year reporting of 1% lauric acid in the subcutaneous fat of the seed oil group should be revised.

Nevertheless, the year-one biopsies were only reported for six people in each group in 1962 and therefore could have been strongly influenced by sampling error.

Still, they are consistent with a persistent presence of lauric acid in the seed oil diet but not the control diet for the first year of the study.

Together with the dietary analysis suggesting Emdee had been cut loose by 1962, this suggests that Emdee was the margarine of choice during the first year of the trial but was not used thereafter.

Together with the data on historical availability, this suggests that Emdee was used in the first year because it was the only margarine based on unhydrogenated corn oil on the market in September of 1959, while they switched to Mazola in September of 1960 when it became available.

As to why they made the switch, cost does not make sense because the margarine was provided as a gift.

Either they thought the Mazola Margarine had better sensory characteristics and was better for blinding the subjects, Mazola was simply better at supplying the volume they needed, or they did it because they valued their relationship with Mazola, who was also providing the liquid corn oil that constituted the top oil ingredient of the seed oil diet.

Which Group Had More Trans Fat? Seed Oil or Control?

Since the margarine in the seed oil group appears to have been 25% trans fat after Emdee was abandoned one year into the study, we can quantify the trans fat in the seed oil diet as zero during year one and as eight grams per day years two through eight.

So it was confounded?

Hold your horses.

They do not tell us how much partially hydrogenated shortening was used in the control group, but we need to estimate this before we can even speculate whether there was more trans fat in the seed oil group or the control group.

We now take two approaches:

• textual analysis: what makes sense given how the diets were described, and what comports with their definition of “minor”?

• biomarker analysis: what do the changes in serum cholesterol indicate could have been the “trans fat gap” between the two groups?

Taking a Look at the Diet

Here is a breakdown of the seed oil diet, where the “added vegetable oil” and “added unsaturated margarine” would be replaced by natural animal fats and some portion of partially hydrogenated shortening in the control group.

The main foods I could see them using partially hydrogenated shortening in would be deserts, potatoes, and vegetables.

Deserts accounted for 9% of the fat, potatoes and their substitutes 7%, and vegetables 11%.

If all of these foods were cooked in shortening, shortening would have supplied 27% of the fat. The total fat was 111.2 grams per day, and 27% of this is about 30 grams per day.

The deserts definitely contain some dairy fat, especially in ice cream, so the total shortening should be less than this.

I would say some one-third to two-thirds of deserts were probably foods like pie crusts and pastries where you would expect shortening to be used, so the lower bound is about 3-4 grams of fat from shortening per day (just a third of the desert fat) and the upper bound is about 27 grams per day (these three categories minus a third of the desert fat).

What Fits the Definition of “Minor”?

Because now we are forced to reject the definition of “minor” that no more than 2.8 grams of trans fat were used in each group, we now need to shift to the more liberal definition of “minor” that up to ten grams of trans fats could have been used in each group, but they would have to be distributed equally between the two groups.

This definition is not consistent with 3-4 grams of shortening per day being used in the control group, but it is consistent with around 27 grams per day being used.

We do not have an analysis of the shortening available, but it could easily have been as low as 20% trans fat or as high as 40% trans fat, so there are quite a range of amounts of shortening that could plausibly supply similar amounts of trans fats as used in the control group.

Analyzing the “Trans Fat Gap” Implied by Serum Cholesterol

We now turn to the best biomarker we can use: the difference between groups in serum cholesterol.

This cannot tell us the absolute amount of trans fat consumed by either group, but it can give us an indication of the difference between the two groups.

The preliminary suggestion by Keys that each 1% of calories increase in trans monounsaturated fat would raise serum cholesterol by 2.1 mg/dL has been revised by subsequent research meta-analyzed in 2001 and 2003. The revised prediction is that a 1% increase of calories from trans monounsaturated fat raises cholesterol by 0.31 mmol/L, which is 1.19863 mg/dL.

According to the main 1969 LA Vet trial paper, the Keys equation would have predicted a difference of 36.5 mg/dL between groups, while the observed difference was 29.5 mg/dL.

The authors pointed out that they observed difference is 81% of the predicted difference, and that this is very close to the 78% “adherence” to the diet during the period in which people were living in the home. They therefore suggested that the 22% of the meals that were not eaten at the home (which may or may not have been replaced by a meal outside the home) could have been the reason for the discrepancy.

This makes no sense at all, because in the preceding paragraphs they describe at length how they could not model “adherence” as an explanation of serum cholesterol that achieved a correlation coefficient higher than 0.31. The square of this coefficient represents the explanatory power of one variable for the other. This means that “adherence” to the diet explained 9.6% of the serum cholesterol.

If “adherence” explained less than 10% of serum cholesterol, how could the 78% “adherence” explain why the difference in cholesterol was only 81% of what was predicted by the Keys equation?

The closeness of 78% to 81% is obviously a coincidence.

What we rather have, here, is a 7-mg/dL gap between the difference predicted by the Keys equation and that observed in the study.

Before we consider whether trans fat could account for this, we should also take into account that the LA Vet trial investigators used the equation Keys published in 1965, which attributed the cholesterol-raising effect of saturated fat to be primarily due to lauric (C12:0), myristic (C14:0), and palmitic (C16:0) acids, but did not differentiate between the ability of each of those three fatty acids to raise cholesterol.

A more modern version of the Keys equation by Müller, 2001 attributes a much stronger cholesterol-raising effect to myristic acid than to palmitic acid and attributes a modest lowering effect to oleic acid (C18:1) rather than dismissing it. In this version, the change in total cholesterol is expressed as the following, where each fatty acid is in units of a 1% change in total calories:

∆ total cholesterol = 0.386654 ∆(12:0) + 4.63985 ∆(14:0) + 2.20393 ∆(16:0) + 1.19863 ∆(trans monounsaturated) – 0.170128 ∆(18:1) – 0.65731 ∆(18:2,18:3)

My reanalysis of the LA Vet trial data using this equation leads to a prediction of a 25.7 mg/dL difference.

This is actually closer to the observed effect than the 1965 Keys equation, but, critically, the difference is in the opposite direction.

Specifically, “my” 25.7 mg/dL prediction (by simply plugging in numbers into the Müller 2001 equation) is 87% of the observed 29.5 mg/dL difference, whereas the Keys equation is 123% of it. “My” prediction falling 13% below the observed difference is barely more than half the difference of the Keys 1965 prediction rising 23% above it.

This supports the improvements made to the Keys equation over time.

None of these equations are perfect, however, because in the context of a full diet, there could be many other things impacting the cholesterol, so there could be specific food effects in the trial diet that biased the net impact in one direction or another.

For example, cheese consistently lowers total cholesterol relative to butter providing the same amount of milkfat, which might be an effect of the food matrix or mineral content.

Nevertheless, the gap between the predicted and observed difference in serum cholesterol could easily be explained by the difference in trans fat consumption.

Since this gap could be explained by other effects such as food matrix and micronutrients, and since there is certainly some component of random error, this should be viewed as the upper bound of the “trans fat gap.”

The prediction being lower than observed suggests there was more trans fat in the control group.

It means that the seed oils were more effective than expected at reducing serum cholesterol because the control group was more cholesterol-raising than expected due to some of its “oleic acid” actually being the trans isomer, known as elaidic acid.

According to the Müller, 2001 equation, switching from oleic acid to elaidic acid switches the coefficient from -0.170128 to +1.19863 mg/dL, resulting in a 1.368758-mg/dL rise in cholesterol for each 1% of calories making the shift.

Therefore, the 6.2-mg/dL difference in total cholesterol observed versus predicted could be explained by the control group consuming an extra 4.5% of calories as trans fat, which would be an extra 13 grams per day.

There is definitely uncertainty around these calculations and we should not put too much trust in this as an exact estimate. However, I looked through the papers published since Müller, 2001 and I do not see any that challenged the equation as inadequate or provided a better one. The fact that it predicts the observed cholesterol difference with hardly more than half the gap as the Keys, 1965 equation supports its improved utility, which in turn supports the flipped direction toward greater trans fat in the control group.

However, using the Keys papers from 1961 and 1965, we would flip to attributing 3.3% more calories from trans fat to the seed oil group and conclude that they were eating 9.2 grams of extra trans fat per day.

The case to use Müller, 2001 is uncertain but solidly rooted in objective arguments: improvement over time with a larger base of research, unchallenged since, and tighter agreement to the observed results.

The best conclusion is that we should not rule out either difference but we should consider it more likely there was more trans fat in the control group than in the seed oil group.

One way to put this is that our range of possibilities is 9 extra grams in the seed oil group to 13 extra grams in the control group, with probability being toward more in the control group and the exact amount being very uncertain but almost definitely less than these boundary amounts.

Since we should value the Müller, 2001 equation over the Keys 1961/1965 equations, we should be biased toward the right side of the estimate: more trans fat in the control group.

Since the total meals eaten in the home were approximately half of the meals eaten during the study period, we should cut the outer boundaries in half: more likely that there were 6-7 extra grams in the control group than that there were 4-5 extra grams in the seed oil group.

If we give weight to the investigators’ comment that these were “minor” constituents of both diets, this should bias us toward the middle of this estimate rather than the boundary.

This leads me to the tentative conclusion that there was more trans fat in the control group, but not that much more.

Does the Trans Fat Matter?

If I am right that there was more trans fat in the control group, but not that much more, then the trans fats largely fade into the “sameness” across the two groups and the few grams per day pales in comparison to the deliberate randomization of one-third of the fat, to the accidental distribution of twice as many heavy smokers to the control group, and to the poor selection of fats in the control group that rendered them having ten times less vitamin E and almost three times less vitamin E per gram of PUFA.

Another way to look at this would be to ask the question, how much extra trans fat would we expect to need to consume to generate the outcomes observed?

The problem with this is that the outcomes need to be analyzed in the context of the other trials because the benefit to heart disease in this trial is an anomaly.

We therefore turn next to whether trans fats confounded the Minnesota Coronary Survey, the only other double-blind trial, and the one that is distinguished for being largest in size despite not being as long as the LA Vet trial.

Our conclusion here, for now, is that there was probably more trans fat in the control group of the LA Vet trial, but not that much more, and as a result it probably didn’t matter.