I just got back from the 2012 Low-Carb Cruise, where I gave a very well received 45-minute talk called “Why Animal Fats Are Good For You,” in which I pointed out the important nutritional value of animal fat, but also pointed out the wide diversity of fat and carbohydrate intakes among populations free of heart disease. I concluded that we should simply “take the fear out of fat.” Rather than advocating specific levels of carbohydrate or animal fat, I suggested, we should select the mix of foods that works best for us, choosing these foods from a broad menu that compiles the many different smaller menus found in traditional diets associated with vibrant health.
The best part about the cruise was all of the phenomenal people I was able to meet and spend time with, and I’m very grateful to Jimmy Moore and the other organizers for putting the whole thing together and providing such an honor to little insignificant me by inviting me to be an important part of it. Thanks Jimmy!
I had one particular encounter with a delightful gentleman that is especially germane to this post. I’ll provide him with a respectful cloak of anonymity for now since the conversation was private, but in short he suggested that my analysis of randomized controlled trials substituting vegetable oil for animal fats and testing the effect of this substitution on heart disease incidence focused too much on old and thus necessarily outdated research at the expense of newer and better studies that could take into account all of the advances we’ve made in our understanding of physiology. This gentleman thus provided me with the perfect reminder to blog about the latest vegetable oil substitution trial published last month in the American Journal of Clinical Nutrition (AJCN) (1), which claimed to show that substitution of margarine and sunflower oil for butter lowers the amount of fat in the liver and may protect against heart disease by increasing the activity of the LDL receptor.
“Updated” is only better than “outdated” when new studies either provide revolutionary new discoveries or methodically fill in the gaps in our knowledge left open by older studies. When the latest studies address questions that had been rendered irrelevant decades earlier, however, we are left wondering why we should even bother reading them. As we will see below, when viewed in the context of existing knowledge, the new AJCN study should leave us scratching our heads and asking just one question: “So what?”
What We Do and Don’t Know About Vegetable Oils and CHD
The Hahhh-vahd Version
To be fair, the authors acknowledged in their justification of their statistical methods that their “primary aim was academic” and that they were not attempting to “evaluate a clinical treatment and its adverse events in patients.” The study is indeed interesting from the standpoint of academic curiosity, but few scientists would deliberately design a study to answer a research question with no practical relevance at all, even if their primary aim were academic, and this group of investigators is no exception. Part of the problem with this study is that the authors uncritically accepted the conventional view that substitution of polyunsaturated vegetable oils for saturated animal fats prevents heart disease when they set out to justify the study’s design. The gap in knowledge about this effect, as they saw it, was simply how vegetable oils prevent heart disease. Here’s that “knowledge gap” in their own words:
Solid evidence indicates that replacing SFAs [saturated fatty acids] with PUFAs [polyunsaturated fatty acids] reduces coronary artery disease (CAD) events (11, 12) and possibly prevents diabetes (13). The mechanism for the CAD risk reduction involves the LDL-cholesterol lowering of n-6 PUFAs (14), but other potential metabolic effects need further study. The mechanism behind SFA-induced increases in serum cholesterol may involve both lowered LDL-receptor numbers and PGC-1b induction (15, 16) but the LDL-cholesterol lowering of PUFAs is unclear.
In other words, we already have “solid evidence” that vegetable oils prevent heart disease, and we can be quite certain that part of the way they do this is by lowering cholesterol, but we are still wondering just how they lower cholesterol and whether they do any other cool stuff to boot.
The “solid evidence” these investigators cited is an analysis that Dariush Mozaffarian of Harvard Medical School and two of his colleagues from the Harvard School of Public Health published in 2010 (2), and a review that cites this same analysis (3). Mozaffarian and his colleagues pooled together the “randomized controlled trials” that lacked “other major concomitant interventions (e.g., blood pressure or smoking control, other multiple dietary interventions, etc.)” and concluded that substitution of vegetable oils for saturated fats does indeed prevent heart disease. The catch is that their odd definition of “randomized” and their tricky little definition of “other multiple dietary interventions” led them to pool together a — let’s just say “convenient” — set of trials. Here are the trials they did include, clickable-to-enlarge:
The Rose Trial Lies in the Grave
The first ever vegetable oil substitution trial published by Rose and colleagues in 1965 (4) is conspicuously absent. The reason given in the “Supporting Information Table” is that it involved “multiple interventions.” This is rather strange considering the presence of the Oslo Diet-Heart Study, which included fruits, vegetables, and cod liver oil as part of the treatment; or the STARS study, which included fruits, vegetables, and nutritional counseling as part of the treatment. What were those pesky confounders in the Rose trial — confounders so confounding that the trial had to be banished from the analysis? Let’s take a look:
Apparently the inclusion of an olive oil group constitutes “other multiple dietary interventions” and justifies deep sixing the trial by burying it alive in a supplementary table, even though we can easily ignore the olive oil group if we want to compare the corn oil group to the control group. On the other hand, including fruits, vegetables, cod liver oil, and vegetable oil all as parts of the same treatment constitutes not even the flimsiest barrier or the lowest hurdle to chalking up any effect of such a multifaceted treatment solely to the vegetable oil.
Phew! Without the assistance of the analytical acrobatics taught at Harvard, we’d be stuck with some silly notion that vegetable oils might increase the risk of heart disease as happened in the Rose trial instead of decreasing the risk of heart disease as happened in the Oslo Diet-Heart and STARS trials. And we might come to the crazy conclusion that fruits, vegetables, or even cod liver oil prevented heart disease in these latter trials.
Wherefore Art Thou, O Sydney Diet-Heart?
The Sydney Diet-Heart Study (5) was excluded because it measured a “non-CHD [coronary heart disease] endpoint.” Of course the title of the Sydney study was “Low Fat, Low Cholesterol Diet in Secondary Prevention of Coronary Heart Disease,” but the fact that only 90 percent of the men in the study died of CHD apparently led to its exclusion. Sixty out of 67 deaths resulted from heart disease in this study. The others included three deaths from cerebral vascular disease, two from cancer, and two from car accidents. The primary figure (shown below) only includes 65 deaths, perhaps because it excludes the two car accidents. Thus 92 percent of the deaths represented in the figure apparently resulted from heart disease. While it is certainly a shame that the authors didn’t provide us with a figure that only showed the effect on heart disease mortality, with 92 percent “purity” it would obviously be pretty close to the effect on total mortality, shown here as a decrease in survival:
Although the Harvard investigators could legitimately complain that a small handful of these deaths resulted from something other than heart disease, it seems like a stroke of phenomenal luck that they were able to relegate this inconvenient trial into the rusty ruins of an unread supplementary table where it could keep the Rose trial company instead of peskily piquing our suspicions with that statistically significant increase in mortality it showcases so clearly as the result of vegetable oil.
An Odd Definition of “Randomization”
The Harvard investigators excluded the Saint Vincent’s Hospital Study (6) because it was “non-randomized.” While it’s true that these authors had been so unsatisfied with their inability to show any benefit of replacing relatively saturated coconut and peanut oils with highly polyunsaturated corn and safflower oils that they eventually collapsed the two groups into one and compared the resultant “study group” to a new “control group” that they pulled out of thin air (6, 7), the authors had indeed randomly assigned the initial set of subjects to one or the other diet and after five years they reported no difference in total mortality, with four deaths in the group consuming coconut and peanut oils and five in the group consuming corn and safflower oils (8). Seven of these nine deaths resulted from heart attacks (7). Even through the ten-year mark the authors admitted that “the degree of unsaturation of the diet did not influence cardiovascular disease mortality” (7), although they only reported actual data for the five-year mark.
If there were a legitimate reason to exclude the trial, it would have been not that the trial wasn’t randomized but that the authors never reported the precise number of cardiovascular deaths in each group — with only seven of nine deaths owing to fatal heart attacks the “purity” of the total mortality figure was only 78 percent. And while we might excuse the Harvard authors for excluding it from their pooled analysis for this reason, it’s starting to seem more and more difficult to describe the evidence this analysis provides as “solid” when it excludes three of the six randomized, controlled trials substituting polyunsaturated oils for saturated fats as a single-factor treatment, while including trials like Oslo and STARS where it is impossible to tease out the effects of vegetable oil from the effects of other aspects of the interventions like fruits, vegetables, nutritional counseling, or cod liver oil.
The most bizarre part of the Harvard analysis, however, is its inclusion of the Finnish Mental Hospitals Study. If any study should have been excluded for being “non-randomized” it should have been this one! And as we can see from the lines marked “Finnish” in that clickable-to-enlarge list of studies, which I’ll reproduce below for convenience, the inclusion of this study was a major determinant of the final conclusion. It contributed over 15 percent of the data, but since it showed the largest “effect” out of any of the included studies, it accounted for over a third of the supposed 19 percent reduction in cardiovascular risk attributed to the use of vegetable oils.
As I described in a previous blog post, “When Standing at the Brink of the Abyss, Staring Into the Great Unknown, We Randomize,” the purpose of randomly allocating subjects into two or more groups is to ensure that each group is a random sample of the initial pool of subjects. By ensuring this, we ensure that all confounders, whether known or unknown, are as similar in distribution as possible between the two groups. We can expect the randomization to fail to distribute some confounders evenly, especially if the total number of subjects is small, but randomization brings us as close to an even distribution as we can come.
It should go without saying that in order to have an even distribution of confounders in each group we need to have a distribution of confounders in each group, and that if we only have one member of each group we cannot possibly have a distribution of anything in either group. But this is exactly what the investigators of the Finnish Mental Hospitals Study did! The investigators “randomized” two different hospitals to one of two dietary groups so that each group had an “n” of one hospital and then proceeded to study the incidence of heart disease among the patients attending the hospital who had not been randomly allocated to anything.
To consider this study a “randomized, controlled trial” is to make a mockery of the very concept of randomization and to flout the scientific method and every philosophical, logical, and statistical concept incorporated therein. The Finnish study is no more “randomized” than a comparison of the heart disease incidence in two countries would be after one of them “randomly” passed a tax on saturated fat. It is essentially one of the least useful ecological studies ever published, akin to the China Study or Ancel Keys’ famous six-country graph, just with far fewer data points than either of these studies had and thus providing far fewer opportunities to mine the data for interesting associations.
The three-fold difference in margarine consumption between hospitals, the two-fold difference in the use of the heart-disrupting, sudden death-inducing antipsychotic medication thioridazine between hospitals, and innumerable other known or unknown confounders could have accounted for the supposed benefit of vegetable oils “shown” by this trial (9). It is precisely this multitude of known and unknown confounders that provides the rationale for randomizing, a technique whose basic mechanics clearly eluded the Finnish investigators and have continued to conveniently elude several generations of investigators that have ever since been making uncannily crazy claims about this study, an editorial in AJCN once even claiming that it was “one of the earliest and most convincing studies of the better efficacy of unsaturated than of saturated fat in reducing cholesterol and heart disease” (10).
Swigging Soybean Oil Prevents “Possible” Angina?
Eliminating the Finnish Mental Hospitals Study from this analysis alone would render the pooled results of the vegetable oil trials null, but the problems don’t end there. Certainly the Medical Research Council Study (11) was rightfully included. Its provision that members of the treatment group use half of the soybean oil unheated, often drinking it with fruit juice, while allowing members of the control group to use traditional fats for ordinary cooking without reservation made it likely that differences in heat damage (or placebo effect!) between the two diets would confound the results, but given the quality of these trials overall that should hardly justify its exclusion.
The problem here, however, is which set of data we should use from this study. If we look at “definite” cardiovascular events or “probable” cardiovascular events that occurred without being preceded by any “possible” events, there were no differences between the two groups. If we look at cardiovascular deaths, there were no differences. There was, however, an excess of “possible” events in the control group. These “possible” events occurred when subjects reported angina pain to unblinded personnel and the unblinded personnel then decided the reports warranted passing along to a blinded review panel. If the blinded review panel couldn’t find any objective evidence of an event, such as an abnormality on the patient’s electrocardiogram, they would declare the whole affair a “possible” event.
After such an “event,” the physicians were allowed to switch the person onto a new diet, and no doubt some of these worried-sick patients, who for all we know may have been hypochondriacs, could have decided to start swigging the soybean oil themselves. And indeed, there were more than twice as many subjects in the control group — twelve versus five in the soybean oil group — who suffered a definite heart attack after having a “possible” attack of angina, and it was by including these seventeen secondary heart attacks that the Harvard investigators were able to count 51 total events in the control group and 45 total events in the soybean oil group, calculating from this that soybean oil produced a 14 percent reduction in the risk of cardiovascular events. But it is precisely these seventeen secondary heart attacks that are smothered in gooey coats of iffiness, while any of the more reliable numbers such as definite or probable “first” events (those not preceded by any “possible” events) or actual cardiovascular deaths show no effect of soybean oil at all.
Even the authors suggested that the excess in possible angina could have occurred because “the men in the control group themselves may have felt that they were having less active treatment, and might have been psychologically motivated both to complain more and more readily of cardiovascular symptoms.” Of course if drinking soybean oil with fruit juice is so psychologically powerful that it subdues complaining, then there is no reason to think such a technique can’t similarly subdue psychogenic pain or the tendency of a high-anxiety hypochondriac to worry himself into a real heart attack. No doubt many of us would develop a little bit of anxiety or hypochondria if we had established heart disease as these patients did, and no doubt if the investigators had blessed an oily elixir by telling us they were testing its ability to prevent a second heart attack many of us would find its psychological power greatly enhanced. This is one of the reasons I place special emphasis on the only two double-blind studies of this nature ever conducted, which we’ll discuss in more detail below.
Conclusions From the Hahhh-vahd Version
I wouldn’t recommend standing atop the Harvard version of what we do and don’t know about vegetable oils and heart disease because this “solid evidence” may just begin to waver beneath your feet, and sinking into the Netherworld of Nebulous Nonsense may just be the fate of anyone who tries this statistical equivalent of walking on water.
A More Nuanced Version
In my own analysis of these trials as presented in Precious Yet Perilous, Good Fats Bad Fats, or my recent lectures at Wise Traditions and the Low-Carb Cruise, I’ve placed special emphasis on the LA Veterans Administration Hospital Study because it was one of only two double-blind studies, it was the only study whose subjects had a mean age over 60 and were therefore more vulnerable to cancer, and it lasted for over eight years. It was, moreover, the only randomized, controlled trial that substituted vegetable oils for saturated fats as its only intervention that actually showed this intervention to reduce the risk of heart disease. The Rose trial and the Sydney Diet-Heart trial showed the opposite. The Saint Vincent Hospital and Medical Research Council studies showed no effect at all. The Minnesota Coronary Survey was the only other double-blind trial. It showed no effect of vegetable oil on cardiovascular disease, and showed that animal fat slightly but insignificantly favored better total survival. It was in any case far too short to warrant any serious conclusions: most of the subjects were only on the diet for a year. The LA VA study is thus unique in many ways.
This study revealed a number of disturbing trends. Cancer didn’t begin increasing in the vegetable oil group until after two years, and really took off only in the last few years of the study (12).
The increase in total non-cardiovascular mortality, shown below as a decrease in survival, didn’t even begin emerging until the fourth year and really took off only in the seventh year (13).
Overall, vegetable oil appeared to decrease the risk of death from cardiovascular disease but increase the risk of death from other diseases, so that total survival stayed pretty similar between groups throughout the eight years (13).
From these results we arrive at several burning questions:
- What happens to total mortality after ten or fifteen years of consuming vegetable oils, once the increase in non-cardiovascular mortality becomes dominant?
- Why did the negative effects take so long to emerge?
- If the negative effects take so long to emerge, what on earth is the point of conducting short-term PUFA studies?
The authors themselves arrived at similar questions and concluded as follows:
This small excess 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.
The answer to the second question, why the negative effects of vegetable oils take so long to emerge, may relate to their slow accumulation in fat tissue and the resultant demand imposed on our requirement for antioxidants. The LA VA study showed that it takes at least two or three years for adipose tissue to fill up most of its stores of linoleic acid, the primary polyunsaturated fatty acid found in vegetable oil, and that adipose levels of this fatty acid never quite reach a perfect plateau even after eight years:
Back in the 1960s, around the same time the LA Veterans Administration Hospital Study was being conducted, researchers at the Elgin State Hospital in Elgin, Illinois and Chicago’s University of Illinois College of Medicine provided evidence that the accumulation of linoleic acid in adipose tissue causes vitamin E to accumulate along with it in order to protect it from oxidation. Here’s a figure from one of their studies where they fed rats a diet of safflower oil for four months and then switched them over to a diet of beef fat (14):
We can see that the amount of vitamin E in adipose tissue traces quite closely the amount of PUFA. PUFA-rich vegetable oils are generally very rich in vitamin E. The question is whether they contain just enough vitamin E to protect the PUFA from oxidation, more than enough so that they wind up being an excellent source of extra vitamin E, or not quite enough so that they actual cause a vitamin E deficit over time. The Illinois investigators suggested that if the turnover of vitamin E in adipose tissue was higher than the turnover of PUFAs, then the accumulation of PUFA in adipose tissue would eventually cause a vitamin E deficit, especially if that extra supply of vitamin E were at some point cut off: if, for example, someone were to switch from a vegetable oil back to an animal fat, leaving their adipose tissue stuffed to the brim with PUFA but suddenly taking in much less vitamin E in their daily diet.
They conducted a study on thirty schizophrenic patients kept in a metabolic ward from 1954 to 1967 that generated some support for these ideas, but unfortunately they only appear to have published the results in an abstract (15) and in reviews (16, 17), so it is impossible for me to critically analyze the strengths and limitations of this study. Nevertheless, here is a graph of adipose linoleic acid (circles) and plasma vitamin E (x’s) from one of their subjects:
Even after five years of consuming safflower oil, having already come to constitute about half of this person’s fat stores, linoleic acid continued to accumulate in a linear fashion. Plasma vitamin E increased by 75 percent over the first year and a half, but then it began plummeting, finally leveling off towards the end of the third year after having lost two thirds of its initial increase. Although we would prefer to see aggregate data from all of the subjects, these changes suggest that vegetable oils initially provide an excess of vitamin E, but over time the continued accumulation of PUFA in adipose tissue imposes a tax on the supply of vitamin E as our fat stores demand more and more of it to protect their ever-growing supplies of linoleic acid from oxidation. Of course the graph above leaves open the question of whether the person is better off after three to five years consuming safflower oil since his plasma vitamin E is still higher than it was before he began consuming the oil, but it makes it clear that the long-term effect is quite different from — and apparently worse than — the short-term effect.
The authors of the LA VA study disputed the claim that vegetable oils raise the vitamin E requirement (18). The vegetable oil diet in this study provided ten times more vitamin E than the control diet, and after consuming these diets for two to three years the plasma vitamin E concentrations of the subjects consuming the vegetable oil diet were twice as high as those consuming the control diet.
There are two serious problems with this interpretation. First of all, two to three years is before the rise in cancer took off and long before the rise in total non-cardiovascular mortality took off. Did these events correspond to deteriorations in vitamin E status? To my knowledge the investigators never published data that could answer this question.
Second, the control diet was mysteriously deficient in vitamin E. Animal experiments suggest that we should obtain 0.6 milligrams of vitamin E for every gram of PUFA in our diet. As shown below, the vegetable oil diet approached this ratio (0.52) but the control diet fell miserably short of it (0.18). The vitamin E content of the vegetable oil diet was thus ten times higher than that of the control diet and the vitamin E-to-PUFA ratio was three times higher, but plasma vitamin E was still only twice as high. The investigators gave us very little detail about the control diet except that it was based primarily on animal fat and that it contained at least some butter and at least some partially hydrogenated vegetable oil. Had it been based primarily on butter, the vitamin E-to-PUFA ratio would likely have ranged from 0.5 to 1.3 depending on whether the cows were fed grain, hay, or grass (19). It is thus very likely that if the control diet had been based on high-quality animal fats, vitamin E status would have been even better than on the vegetable oil diet.
Despite randomization, moreover, there were twice as many heavy smokers and 60 percent more moderate smokers in the control group and more light smokers and non-smokers in the vegetable oil group. We now know that smoking increases the demand for vitamin E (20), so the combination of more smoking and a low-quality, vitamin E-deficient diet may have produced the higher incidence of heart disease in the control group rather than the “saturated fat” in that group’s diet. Consistent with this, members of the control group who smoked more cigarettes had a greater risk of heart disease, but members of the vegetable oil group did not (21).
It may be, then, that a certain level of vitamin E deficiency is needed to “activate” some of the harmful effects of smoking, and that poor-quality animal fats and a few helpings of partially hydrogenated vegetable oils are the best way to achieve this, but that fats that weren’t tested in this study such as high-quality butter or relatively saturated but vitamin E-rich virgin palm oil may provide the most superior protection.
It thus appears that something in the diet of the control group — perhaps the saturated fat or the lack of PUFA — protected these vitamin E-deficient smokers against cancer but wasn’t quite able to protect them against heart disease. Whether it was the saturated fat, the smoking, or the vitamin E deficiency that predisposed the control group to heart disease is impossible to tell for certain, but the graph above suggests it may have been a combination of smoking and vitamin E deficiency, and the inability of any of the other vegetable oil substitution trials to generate positive results argues against a simple effect of “saturated fat.”
Were the Illinois investigators right? When it comes to vitamin E, do vegetable oils bring in the boom or the bust? My guess is it depends on a person’s exposure to oxidants. When an oxidant encounters a lipid membrane, it’s about a thousand times more likely to attack the little bit of vitamin E therein than it is to attack the abundant amount of PUFA because vitamin E is the loyal and self-sacrificial guardian of that PUFA. If doubling the amount of PUFA in adipose tissue doubles the amount of vitamin E required in that tissue, this may exert little or no harm when the levels of oxidants are low and the vitamin E remains comparatively stable. But the scenario could change dramatically if the levels of oxidants increase, destroying the vitamin E faster than it can be supplied through the consumption of vegetable oil.
It may well be, then, that vegetable oils exert little harm in the absence of other stressors, but that in the presence of copious oxidants they ultimately impose a serious tax on the supply of vitamin E and other antioxidants. If this is in fact the case, long-term consumption of PUFA-rich vegetable oils would always be unwise, because whether through aging or through our seemingly random but nevertheless inevitable encounters with disease, we are bound at some point to face stresses that could potentially call upon us to settle our PUFA accounts, the record of our indulgences written indelibly into our adipose.
Vegetable Oil and Heart Disease: The Real Knowledge Gaps
The real knowledge gaps in my view, then, are not the questions posed by the latest AJCN paper: that is, just how vegetable oils work their well characterized magic shown repeatedly by “solid evidence” wherein they reduce the risk of heart disease by lowering cholesterol and perhaps by pulling off other metabolic tricks. They are, instead, at least as I see them, the following:
- What is the effect of consuming vegetable oil for “well in excess of eight years,” as the LA Veterans Administration Hospital Study investigators had suggested future trials should address?
- What is the relationship between vegetable oils and the vitamin E requirement, and how does this change as we consume these oils over the course of years?
- How do oils with different proportions of omega-3 and omega-6 fatty acids, or minor constituents such as phytosterols and antioxidants, stack up against one another when compared head to head? How do any potential differences interact with oxidative stress and the vitamin E requirement?
- How do these oils interact with other dietary and lifestyle factors and do they act differently in different populations? To what extent are any of these interactions dependent on oxidative stress and the vitamin E requirement?
What We Do and Don’t Know About Veggie Oils and Fatty Liver
The AJCN paper also tested the effect of vegetable oils on the amount of fat in the liver, so it’s important to properly frame this research question in the context of what is and isn’t already known as well. To my knowledge, the AJCN paper is the first randomized controlled trial conducted in humans testing the effect of substituting vegetable oil for animal fat on the amount of fat that accumulates in the liver, but it is worth reviewing some of the data from animal experiments to put the finding in context.
In my 2008 Special Report, “How Essential Are the Essential Fatty Acids? The PUFA Report Part I: A Critical Review of the Requirement for Polyunsaturated Fatty Acids,” I described how the mammalian body appears to do everything in its power to get rid of excess PUFAs: in rats, for example, 60 percent are burned for energy, 20 to 30 percent are converted into saturated and monounsaturated fats or cholesterol, and most of the rest are secreted into the fur (22).
In my 2011 Wise Traditions article, “Nonalcoholic Fatty Liver Disease: A Silent Epidemic of Nutritional Imbalance,” I noted that, when investigators add different types of fat to an otherwise standard rodent chow, saturated fats are more likely to accumulate in the liver than polyunsaturated fats. I suggested this could be because “the body is quick to burn polyunsaturated fats for energy, since having them hang around is so dangerous. After all, if they hang around, they are likely to contribute to oxidative damage.” Speculating “why” something works the way it does is always, well, speculative, but the take-home point is that saturated fats for whatever reason raise the choline requirement a little bit higher than PUFAs do. Providing extra choline in the diet clears all this fat from the liver. Extra protein works too, because the sulfur amino acid methionine is a precursor to choline.
I cited the study whose results are shown below, for example, in my Wise Traditions article and my blog post, “The Sweet Truth About Liver and Egg Yolks: Choline Matters More to Fatty Liver Than Sugar, Alcohol, or Fat.” This study found that long-chain, saturated fats produced the greatest accumulation of fat in the livers of rats fed a low-protein diet while highly polyunsaturated cod liver oil (CLO) produced the least accumulation, but it also found that these fats had no effect in rats fed sufficient protein (23, 24):
I also cited the study whose results are shown below, which showed that the amount of choline required to minimize liver fat is 30 percent higher in rats fed butter than in rats fed corn oil (25):
There are other contexts, however, in which feeding rodents PUFAs produces greater accumulation of fat in their livers than feeding them saturated fats. This is the case in some (26) but not all (27) studies using methionine- and choline-deficient diets, which involve severe nutrient deficiencies that we wouldn’t expect to find in humans. Replacing polyunsaturated corn oil with saturated beef fat or cocoa butter proves protective in models where animals consume almost a third of their calories as alcohol, and replacing polyunsaturated fish oil with saturated and monounsaturated palm oil has the same effect (28, 29 30). This could be because alcohol and extreme nutrient deficiencies make PUFAs more likely to oxidize. When PUFAs oxidize in the liver, they contribute to the oxidative destruction of lipoproteins before the lipoproteins are able to leave the cells, causing fat to become trapped inside the cells. Researchers can produce this effect simply by incubating liver cells with PUFAs, and can abolish the effect by removing excess iron from the cells (31).
A simple way to synthesize all this evidence would be to suggest that in the absence of a major oxidative burden, PUFAs are less likely than saturated fats to cause the accumulation of fat in the liver because they are more readily burned for energy and thus demand less choline to support their export from the liver. In the presence of a major oxidative burden like excess alcohol or iron, they are more likely than saturated fats to cause fat to accumulate in the liver because they are more prone to oxidation and thus contribute to the destruction of lipoproteins and their entrapment in the liver. Future research would best clarify this by comparing some of these different models head to head in the same study. And of course we need to determine to what extent we can extend these principles from animal models to humans.
The animal models would further suggest that all of this is moot once a sufficient amount of fat has actually accumulated in the liver. Fatty liver is probably always a bad thing, but it is the progression from simple fatty liver to an inflammatory disorder called nonalcoholic steatohepatitis or NASH that paves the way for fibrosis, cirrhosis, and in some cases even death from liver cancer or acute liver failure. The infamous high-fat, lard-based rodent diet we always hear about in the press produces fatty liver in eight weeks but there is little sign of any inflammation (31), whereas corn oil will produce overt NASH in just three weeks (32). Corn oil is also the special ingredient in commercial methionine- and choline-deficient rodent diets that allows them to produce NASH, and replacing the corn oil in these diets with carbohydrate, coconut oil, or beef tallow abolishes this nasty little trait (27).
All of this leaves us, in my view, with the following questions, especially when we consider the gaps in our knowledge left open by these animal experiments in the context of what we have learned from the long-term, randomized, controlled vegetable oil trials conducted in humans that we reviewed above:
- If saturated fat increases the amount of fat in the liver in the absence of a major oxidative burden, does a little extra choline abolish this effect in humans just as it does in the animal models? How much is needed, and what other nutrients prove protective?
- Does replacing saturated animal fat with polyunsaturated vegetable oil have different effects in humans depending on whether excessive alcohol, iron, or other oxidative burdens are present, as the animal models suggest?
- Does long-term consumption of vegetable oils over the course of five to eight or more years pose a sufficient oxidative burden in and of itself to promote the accumulation of liver fat?
- Do polyunsaturated vegetable oils promote the progression of fatty liver to steatohepatitis (NASH) when the vitamin E content of the diet is controlled, as it is in animal models?
- If the vitamin E present in vegetable oils is sufficient to prevent the progression of fatty liver to NASH in short-term studies, does vitamin E status deteriorate over the course of five to eight or more years, and does long-term consumption of vegetable oils in this way set the groundwork for progression to NASH?
Okay! To the three of you left reading, congrats! And thanks! Now that we’ve properly set the context to evaluate the findings of this new study, let’s take a look at how many of these interesting questions it can answer. Buckle up!
AJCN: Mystical Implications of the Ten-Week Trend
The authors of the AJCN paper randomly assigned just over sixty abdominally obese adults to eat either a diet rich in PUFAs, centered on sunflower seeds, sunflower oil, margarine, and scones baked in sunflower oil, or a diet rich in saturated fat, centered on butter and scones baked in butter. The subjects ate their assigned diets for ten weeks, precluding the study from shedding any light on any of the unresolved questions about the long-term effects of vegetable oils. The investigators didn’t bother to track intakes of vitamin E, choline, or any other micronutrients, precluding the study from shedding any light on any of the unresolved questions about the interactions between PUFAs and these nutrients, including whether the switch to butter may have left some of the people in the position that the Illinois investigators had feared: their adipose tissue stuffed to the brim with PUFA, having accumulated a vitamin E debt that rivals the cash debt imposed by Harvard Medical School student loans, suddenly switching to a an income of vitamin E that doesn’t even allow them to meet their monthly payments. The investigators nevertheless came up with a few interesting findings, so let’s take a look.
They measured liver fat by magnetic resonance imaging (MRI) and magnetic resonance spectroscopy (MRS). Here are my own graphs of the results:
These data convince me that the ten-week effects of the diets are real for two reasons: first, the two measurements corroborate one another; second, the effect is most impressive in the second graph where there is no difference whatsoever between the groups at baseline. This makes it essentially impossible that we’ve been tricked by regression to the mean, a phenomenon I described in detail in a previous post, “How a Study Can Show Something to Be True When It’s Completely False: Regression to the Mean.”
On the other hand, it’s difficult to conjure up any confidence that these changes are meaningful. Neither diet produced fatty liver. After ten weeks the butter group wound up with an average liver fat of 3.9% as measured by MRS, but they’d have had to bump that up closer to 5.6% to reach the bare minimum needed to qualify for fatty liver using this measurement (32). If continuing to eat butter were to keep pushing liver fat higher and higher at the same rate, it would take the butter buddies another six months to “qualify,” and by that time the sunflower suckers would have all the fat sucked out of their livers and be left with none at all. In fact, given a few extra weeks, the folks eating the sunflower oil diet would have negative liver fat! If that seems hard to believe, just take a look at the 40-week projection, assuming the trends stay the same over time:
And what happens then? Does the liver implode? Does it suck the pancreas into an alternate dimension? Of course if the LA Veterans Administration Hospital Study has taught us anything I hope it’s that the trends don’t stay the same over time. Just in case the late-blooming rise in cancer in that study hasn’t convince some of us, just in case the increase in total non-cardiovascular mortality that suddenly struck in the seventh year didn’t turn some of our heads, and since fatty liver is so closely tied to obesity, let’s take a look at the body weight data from that most important of clinical trials:
If the LA Veterans Administration Hospital Study had only lasted three years, we might have concluded that vegetable oils make us just a little bit fatter than animal fats. But the difference between the groups began closing in by the fourth year and disappeared by the sixth. Of course if it took the trends in liver fat shown in the AJCN paper six years to reach an equilibrium the poor subjects eating butter would literally become walking tubs-o-lard and the liver fat of the lucky subjects eating sunflower oil would become so negative that their livers would probably suck their entire bodies into an alternate dimension. They might just become the first human beings to form black holes.
AJCN: Trade In Your Liver Fat for Cholesterol In Ten Weeks!
Another interesting finding from this study is that subjects consuming sunflower oil had lower blood levels both of PCSK9, an enzyme that degrades the LDL receptor, and lathosterol, an intermediate in cholesterol synthesis. Both of these normally follow the natural rhythm of cholesterol synthesis, falling in the evening when cholesterol synthesis is lowest and rising to a peak in the early morning while we are still in the depths of sleep, when cholesterol synthesis is highest (33). PCSK9 probably traces cholesterol synthesis so closely because when our cells make more cholesterol, they need to take in less from the blood. PCSK9 degrades the LDL receptor, decreasing the amount of cholesterol cells take in from the blood. The authors considered the lower PCSK9 and lathosterol on the vegetable oil diet to be “in line with the notion that PUFA increases hepatic cholesterol,” since the accumulation of cholesterol in the liver would shut down cholesterol synthesis.
Unfortunately the investigators didn’t measure the accumulation of cholesterol in the liver, but if they are correct, it would mean that when we trade butter for sunflower oil our livers trade fat for cholesterol. If there really were “solid evidence” that vegetable oils reduce the risk of heart disease, we might consider this finding to be valuable insight into the mechanism, but since this solid evidence doesn’t exist the clinical implications of this trade are unclear. It’s a good thing, then, that the “primary aim” of this study “was academic” and that the authors weren’t trying to “evaluate a clinical treatment and its adverse effects in patients.” Thankfully, a quick perusal of the “Statistical Analyses” section will disabuse the careful reader of any notion that the authors recommend substituting sunflower oil for butter as a safe and effective way of preventing fatty liver or heart disease.
AJCN: Insulin and Inflammation
The investigators also looked at insulin sensitivity, glucose tolerance, fatty acid metabolism, oxidative stress, and inflammation. The two diets had no effect on fasting glucose, but fasting insulin was higher in the group consuming butter.
Although the investigators were unable to assess fatty acid metabolism directly, they were able to look at various fatty acids in the blood, and based on the relative abundance of these fatty acids they suggested that the anabolic pathways that build up more complex fats rather than breaking them down predominated in the butter group. Insulin normally serves to increase the liver’s uptake of glucose from the blood, prevent its release of glucose into the blood, and increase its production of fats. Since fasting glucose remained unchanged but fasting insulin and fat production rose, the butter-based diet appeared to produce a divergence between the glucose-handling and fat-handling functions of insulin. In the scientific literature we call this divergence selective hepatic insulin resistance (34, 35), and it is probably a strategy the liver uses to cope with energy overload.
By becoming resistant to insulin’s stimulation of glucose uptake and suppression of glucose release, the liver helps minimize its accumulation of energy-rich glucose. By staying sensitive to insulin’s stimulation of fat production, the liver protects itself from the toxicity of free fatty acids by sequestering them as triglycerides, and if it has sufficient choline it exports those triglycerides into the bloodstream. The downside is that if this phenomenon persists, more and more insulin is required to keep blood glucose levels normal, taxing the pancreas until it is finally exhausted and elevated fasting glucose ensues.
Out of eight markers of inflammation and lipid peroxidation, two inflammatory markers were increased on the butter diet but the others remained unchanged. These “inflammatory” chemicals regulate insulin signaling (36), and are probably part of the liver’s strategy to cope with energy overload. The authors pointed out that “these markers are elevated in individuals long before the onset of type 2 diabetes,” increasing incrementally for six years prior to the onset of this disease, but there remains no clinical evidence that butter contributes to diabetes or that sunflower oil prevents it, and whether any of the changes observed in this study actually persist for six years or even six months is exactly the type of question this study is unable to answer.
The Bottom Line: Longing For the Old Days
Studies like this make me miss the old days when scientists seemed to realize that changes in diet take years to affect the development of disease. How far back should we go? Should we long for the 1960s when the LA Veterans Administration Hospital Study investigators suggested that “future clinical trials of diets rich in unsaturated fat must be planned for well in excess of eight years”? Should we long for the 1930s and 1940s when Weston Price emphasized the inter-generational effects of nutrition? Certainly there are researchers who recognize these issues in our own day, but in the race to publish no one has time to conduct trials “well in excess of eight years” unless they have time to perish.
We have known for well over a half century that choline protects against all dietary models of fatty liver. We have known for over forty years that PUFAs alter the vitamin E requirement in a potentially time-dependent manner and that trials lasting fewer than eight years are insufficient to demonstrate the long-term effects of consuming vegetable oils. What, then, can we learn from a ten-week study testing the effect of vegetable oils on fatty liver without paying any attention to micronutrients like vitamin E and choline? It seems to me that if studies completed decades ago render the latest studies irrelevant, it is the latest studies that we should consider “outdated” rather than the older studies whose relevance persists through our own day.
Some have suggested that recognizing the importance of randomized trials places too much power in the hands of the pharmaceutical companies that have the resources to conduct these trials. This is like saying that making gasoline out of oil puts too much power in the hands of the oil companies that own all the drills and wells. In the particular context of our economic structure, these and many other things are unequally distributed, but we can’t create rules of logical inference out of thin air to be more “fair” to the underdog any more than we can create gasoline out of water. We should instead admit what we do and do not know, and in the absence of clinical trials conducted for well in excess of eight years that test the interactions between vegetable oils and various other dietary and lifestyle factors that could alter their effects, there’s a whole lot we don’t know about vegetable oils.
One thing we do know is that before the advent of modern vegetable oils, pretty much everyone was thin. Given the close relationship between obesity and fatty liver, those lean folks probably had lean livers. And we know that populations free of heart disease, where even the most vulnerable members of the population appear to be protected, have consumed high-fat diets and low-fat diets, butterfat, meat fat, tropical oils, and lots of other foods, but none of them ate scones with baked-on sunflower oil. This hardly proves cause-and-effect, but it does show that a broad spectrum of foods are at least consistent with population-level freedom from heart disease, and thus consistent with protection of even the most vulnerable members of the group, something that can be said for butterfat but not for sunflower oil. And it was precisely the society that decided it had to “prove” everything that became riddled with heart disease, obesity, and fatty liver — perhaps because there is something protective about faith or intuition, or perhaps just because proof is so darn elusive.
Read more about the author, Chris Masterjohn, PhD, here.
1. Bjermo H, Iggman D, Kullberg J, Dahlman I, Johansson L, Persson L, Berglund J, Pulkki k, BASU S, Uusitupa M, Rudling M, Arner P, Cederholm T, Ahlstrom H, Riserus U. Effects of n-6 PUFAs compared with SFAs on liver fat, lipoproteins, and inflammation in abdominal obesity: a randomized controlled trial. Am J Clin Nutr. 2012;95(5):1003-12.
2. Mozaffarian D, Micha R, Wallace S. Effects on coronary heart disease of increasing polyunsaturated fat in place of saturated fat: a systematic review and meta-analysis of randomized controlled trials. PLoS Med. 2010;7(3):e1000252.
3. Astrup A, Dyerberg J, Elwood P, Hermansen K, Hu FB, Jakobsen MU, Kok FJ, Krauss RM, Lecerf JM, LeGrand P, et al. The role of reducing intakes of saturated fat in the prevention of cardiovascular disease: where does the evidence stand in 2010? Am J Clin Nutr 2011;93:684–8.
4. Rose GA, Thomson WB, Williams RT. Corn Oil in Treatment of Ischaemic Heart Disease. Br Med J. 1965;1(5449):1531-3.
5. Woodhill JM, Palmer AJ, Leelarthaepin B, McGilchrist C, Blacket RB. Low fat, low cholesterol diet in secondary prevention of coronary heart disease. Adv Exp Med Biol. 1978;109:317-30.
6. Bierenbaum ML, Fleischman AI, Green DP, Raichelson RI, Hayton T, Watson PB, Caldwell AB. The 5-year experience of modified fat diets on younger men with coronary heart disease. Circulation. 1970;42(5):943-52.
7. Bierenbaum ML, Fleischman AI, Raichelson RI, Hayton T, Watson PB. Ten-year experience of modified-fat diets on younger men with coronary heart-disease. Lancet. 1973;1(7817):1404-7.
8. Bierenbaum ML, Green DP, Florin A, Fleischman AI, Caldwell AB. Modified-fat dietary management of the young male with coronary disease. A five-year report. JAMA. 1967;202(13):1119-23.
9. Ramsden CE, Hibbeln JR, Majchrzak SF, Davis JM. n-6 fatty acid-specific and mixed polyunsaturate dietary interventions have different effects on CHD risk: a meta-analysis of randomized controlled trials. Br J Nutr. 2010;104(11):1586-600.
10. Knopp RH, Retzlaff BM. Saturated fat prevents coronary artery disease? An American paradox. Am J Clin Nutr. 2004;80(5):1102-3.
11. Report of a Research Committee to the Medical Research Council. Controlled trial of soya-bean oil in myocardial infarction. Lancet. 1968;2(7570):693-9.
12. Pearce ML, Dayton S. Incidence of cancer in men on a diet high in polyunsaturated fat. Lancet. 1971;1(7697):464-7.
13. Dayton S, Pearce ML, Hashimoto S, Dixon WJ, Tomiyasu U. A Controlled Clinical Trial of a Diet High in Unsaturated Fat in Preventing Complications of Atherosclerosis. Circulation. 1969;40:II-1-II-63.
14. Harmon EM, Witting LA, Horwitt MK. Relative rates of depletion of alpha-tocopherol and linoleic acid after feeding polyunsaturated fats. Am J Clin Nutr. 1966;18(4):243-8.
15. AJCN, 1967;20:366
16. Witting LA. The interrelationship of polyunsaturated fatty acids and antioxidants in vivo. Progress in the Chemistry of Fats and Other Lipids. 1971;9:517-53.
17. Witting LA. The role of polyunsaturated fatty acids in determining the vitamin E requirement. Ann NY Acad Sci. 1972;203:192-8.
18. Dayton S, Hashimoto S, Rosenblum D, Pearce ML. Vitamin E Status of Humans During Prolonged Feeding of Unsaturated Fats. J Lab Clin Med. 1965;65:739-47.
19. Searles, Sharon K. Vitamin E, Vitamin A, and Carotene Contents of Alberta Butter. J Dairy Sci. 1970;50(2):150.
20. Bruno RS, Traber MG. Vitamin E biokinetics, oxidative stress and cigarette smoking. Pathophysiology. 2006;13(3):143-9.
21. Dayton S, Pearce ML. Diet and atherosclerosis. Lancet. 1970;1(7644):473-4.
22. Sinclair AJ, Attar-Bashi NM, Li D. What Is the Role of α-Linolenic Acid for Mammals? Lipids 2002;37:1113-23
23. Channon HJ, Wilkinson H. The effect of various fats in the production of dietary fatty livers. Biochem J. 1936;30(6):1033-1039.
24. Channon HJ, Wilkinson H. Protein and the dietary production of fatty livers. Biochem J. 1935;29(2):350-356.
25. Benton DA, Elvehejem CA, Harper AE, Quiros-Perez F, Spivey HE. Effect of different dietary fats on choline requirement of rats. Proc Soc Exp Biol Med. 1957;94(1):100-3.
26. Rivera CA, Gaskin L, Allman M, Pang J, Brady K, Adegboyega P, Pruitt K. Toll-like receptor-2 deficiency enhnaces non-alcoholic steatohepatitis. BMC Gastroenterol. 2010;10:52.
27. Lee GS, Yan JS, Ng RK, Kakar S, Maher JJ. Polyunsaturated fat in the methionine-choline-deficient diet influences hepatic inflammation but not hepatocellular injury. J Lipid Res. 2007;48(8):1885-96.
28. Ronis MJ, Korourian S, Zipperman M, Hakkak R, Badger TM. Dietary saturated fat reduces alcoholic hepatotoxicity in rats by altering fatty acid metabolism and membrane composition. J Nutr. Apr 2004;134(4):904-912.
29. You M, Considine RV, Leone TC, Kelly DP, Crabb DW. Role of adiponectin in the protective action of dietary saturated fat against alcoholic fatty liver in mice. Hepatology. Sep 2005;42(3):568-577.
30. Nanji AA, Sadrzadeh SM, Yang EK, Fogt F, Meydani M, Dannenberg AJ. Dietary saturated fatty acids: a novel treatment for alcoholic liver disease. Gastroenterology. Aug 1995;109(2):547-554.
31. Pan M, Cederbaum AI, Zhang YL, Ginsberg HN, Williams KJ, Fisher EA. Lipid peroxidation and oxidant stress regulate hepatic apolipoprotein B degradation and VLDL production. J Clin Invest. May 2004;113(9):1277-1287.
32. Szczepaniak LS, Nurenberg P, Leonard D, Browning JD, Reingold JS, Grundy S, Hobbs HH, Dobbins RL. Magnetic resonance spectroscopy to measure heaptic triglyceride content: prevalence of hepatic steatosis in the general population. Am J Physiol Endocrinol Metab. 2005;288(2):E462-8.
33. Calandra S, Tarugi P, Speedy HE, Dean AF, Bertolini S, Shoulders CC. Mechanisms and genetic determinants regulating sterol absorption, circulating LDL levels, and sterol elimination: implications for classification and disease risk. J Lipid Res. 2011;52(11):1885-926.
34. Brown MS, Goldstein JL. Selective versus total insulin resistance: a pathogenic paradox. Cell Metab. 2008;7(2):95-6.
35. Li S, Brown MS, Goldstein JL. Bifurcation of insulin signaling pathway in rat liver: mTORC1 required for stimulation of lipogenesis, but not inhibition of gluconeogenesis. Proc Natl Acad Sci USA. 2010;107(8):3441-6.
36. Miao H, Zhang Y, Lu Z, Liu Q, Gan L. FOXO1 involvement in insulin resistance-related pro-inflammatory cytokine production in hepatocytes. Inflamm Res. 2012;61(4):349-58.