The diet-heart hypothesis states three things:
- Dietary saturated fat increases blood cholesterol
- Elevated blood cholesterol increases the risk of having a heart attack
- Therefore, dietary saturated fat increases the risk of having a heart attack
To evaluate the second contention, investigators have examined the relationship between blood cholesterol and heart attack risk. Many studies including MRFIT have shown that the two are related (
1):
The relationship becomes much more complex when you consider lipoprotein subtypes, density and oxidation level, among other factors, but at the very least there is an association between habitual blood cholesterol level and heart attack risk. This is what you would want to see if your hypothesis states that high blood cholesterol causes heart attacks.
Now let's turn to the first contention, the hypothesis that dietary saturated fat increases serum cholesterol. This idea is so deeply ingrained in the scientific literature that many authors don't even bother providing references for it anymore. When references are provided, they nearly always point to the same type of study: short-term controlled diet trials, in which volunteers are fed different fats for 2-13 weeks and their blood cholesterol measured (
2)*. These studies show that saturated fat increases both LDL cholesterol ("bad cholesterol") and HDL cholesterol ("good cholesterol"), but typically the former more than the latter. These are the studies on which the diet-heart hypothesis was built.
But now we have a problem. Nearly every high-quality (prospective) observational study ever conducted found that saturated fat intake is not associated with heart attack risk (
3). So if saturated fat increases blood cholesterol, and higher blood cholesterol is associated with an increased risk of having a heart attack, then why don't people who eat more saturated fat have more heart attacks?
I'll begin to answer that question with another question: why do researchers almost never cite observational studies to support the idea that dietary saturated fat increases blood cholesterol? Surely if the hypothesis is correct, then people who habitually eat a lot of saturated fat should have high cholesterol, right? One reason may be that in most instances, when researchers have looked for a relationship between habitual saturated fat intake and blood cholesterol, it has been very small or nonexistent. Those findings are rarely cited, but let's have a look...
The StudiesIt's difficult to do a complete accounting of these studies, but I've done my best to round them up. I can't claim this post is comprehensive, but I doubt I missed very many, and I certainly didn't exclude any that I came across. If you know of any I missed, please add them to the comments. [UPDATE 4-2012: I did miss several studies, although they're basically consistent with the conclusion I came to here. I plan to update this post with the new references at some point.]
The earliest and perhaps most interesting study I found was published in the
British Medical Journal in 1963 and is titled "Diet and Plasma Cholesterol in 99 Bank Men" (
4). Investigators asked volunteers to weigh all food consumed at home for 1-2 weeks, and describe in detail all food consumed away from home. Compliance was good. This dietary accounting method is much more accurate than in most observational studies today**. Animal fat intake ranged from 55 to 173 grams per day, and blood cholesterol ranged from 154 to 324 mg/dL, yet there was no relationship whatsoever between the two. I'm looking at a graph of animal fat intake vs. blood cholesterol as I write this, and it looks like someone shot it with a shotgun at 50 yards. They analyzed the data every which way, but were never able to squeeze even a hint of an association out of it:
Making the most out of the data in other ways- for example, by analysis of the men very stable in their diets, or in whom weighing of food intake was maximal, or where blood was taken close to the diet [measurement]- did not increase the correlation. Because the correlation coefficient is almost as often negative as positive, moreover, what is being discussed mostly is the absence of association, not merely association that is unexpectedly small.
The next study to discuss is the 1976 Tecumseh study (
5). This was a large cardiovascular observational study conducted in Tecumseh, Michigan, which is often used as the basis for comparison for other cardiovascular studies in the literature. Using the 24 hour dietary recall method, including an analysis of saturated fat, the investigators found that:
Cholesterol and triglyceride levels were unrelated to quality, quantity, or proportions of fat, carbohydrate or protein consumed in the 24-hr recall period.
They also noted that the result was consistent with what had been reported in other previously published studies, including the Evans county study (
6), the massive Israel Ischemic Heart Disease Study (
7) and the Framingham study. One of the longest-running, most comprehensive and most highly cited observational studies, the Framingham study was organized by Harvard investigators and continues to this day. When investigators analyzed the relationship between saturated fat intake, serum cholesterol and heart attack risk, they were so disappointed that they never formally published the results. We know from multiple sources that they found no significant relationship between saturated fat intake and blood cholesterol or heart attack risk***.
The next study is the Bogalusa Heart Study, published in 1978, which studied the diet and health of 10 year old American children (
8). This study found an association by one statistical method, and none by a second method****. They found that the dietary factors they analyzed explained no more than 4% of the variation in blood cholesterol. Overall, I think this study lends very little support to the hypothesis.
Next is the Western Electric study, published in 1981 (
9). This study found an association between saturated fat intake and blood cholesterol in middle-aged men in Chicago. However, the correlation was small, and there was no association between saturated fat intake and heart attack deaths. They cited two other studies that found an association between dietary saturated fat and blood cholesterol (and did not cite any of the numerous studies that found no association). One was a very small study conducted in young men doing research in Antarctica, which did not measure saturated fat but found an association between total fat intake and blood cholesterol (
10). The other studied Japanese (Nagasaki and Hiroshima) and Japanese Americans in Japan, Hawai'i and California respectively (
11).
This study requires some discussion. Published in 1973, it found a correlation between saturated fat intake and blood cholesterol in Japan, Hawai'i but not in California. The strongest association was in Japan, where going from 5 to 75 g/day of saturated fat (a 15-fold change!) was associated with an increase in blood cholesterol from about 175 to 200 mg/dL. However, I don't think this study offers much support to the hypothesis upon closer examination. Food intake in Japan was collected by 24-hour recall in 1965-1967, when the diet was roughly 3/4 white rice by calories. The lower limit of saturated fat intake in Japan was 5g/day, 1/12th what was typically eaten in Hawai'i and California, and the Japanese average was 16g, with most people falling below 10g. That is an extraordinarily low saturated fat intake. I think a significant portion of the Japanese in this study, living in the war-ravaged cities of Nagasaki and Hiroshima, were over-reliant on white rice and had a very peculiar and perhaps deficient diet. Also, what is the difference between a diet with 5 and 75 grams of saturated fat per day? Those diets are probably very different, in many other ways than their saturated fat content.
In Japanese-Americans living in Hawai'i, over a range of saturated fat intakes between 5 and 110 g/day, cholesterol went from 210 to 220 mg/dL. That was statistically significant but it's not exactly knocking my socks off, considering it's a 22-fold difference in saturated fat intake. In California, going from 15 to 110 g/day of saturated fat (7.3-fold change) was not associated with a change in blood cholesterol. Blood cholesterol was 20-30 mg/dL lower in Japan than in Hawai'i or California at any given level of saturated fat intake (e.g., Japanese eating 30g per day vs. Hawai'ians eating 30g per day). I think it's probable that saturated fat is not the relevant factor here, or at least it's much less influential than other factors. An equally plausible explanation is that people in the very low range of saturated fat intake are the rural poor who eat a diet that differs in many ways from the diets at the upper end of the range, and other aspects of lifestyle such as physical activity also differ.
The most recent study was the Health Professional Follow-up study, published in 1996 (
12). This was a massive, well funded study that found no relationship between saturated fat intake and blood cholesterol.
ConclusionOf all the studies I came across, only the Western Electric study found a clear association between habitual saturated fat intake and blood cholesterol, and even that association was weak. The Bogalusa Heart study and the Japanese study provided inconsistent evidence for a weak association. The other studies I cited, including the bank workers' study, the Tecumseh study, the Evans county study, the Israel Ischemic Heart study, the Framingham study and the Health Professionals Follow-up study, found no association between the two factors.
Overall, the literature does not offer much support for the idea that long term saturated fat intake has a significant effect on the concentration of blood cholesterol in humans. If it's a factor at all, it must be rather weak. It may be that the diet-heart hypothesis rests in part on an over-reliance on the results of short-term controlled feeding studies. It would be nice to see this discussed more often (or at all) in the scientific literature. It is worth pointing out that the method used to collect diet information in most of these studies, the food frequency questionnaire, is not particularly accurate, so it's possible that there is a lot of variability inherent to the measurement that is partially masking an association. In any case, these controlled studies have typically shown that saturated fat increases both LDL and HDL, so even if saturated fat did have a modest long-term effect on blood cholesterol, as hinted at by some of the observational studies, its effect on heart attack risk would still be difficult to predict.
The Diet-heart Hypothesis: Stuck at the Starting GateAnimal Models of Atherosclerosis: LDL* As a side note, many of these studies were of poor quality, and were designed in ways that artificially inflated the effects of saturated fat on blood lipids. For example, using a run-in period high in linoleic acid, or comparing a saturated fat-rich diet to a linoleic acid-rich diet, and attributing the differences in blood cholesterol to the saturated fat. Some of them used hydrogenated seed oils as the saturated fat. Although not always consistent, I do think that overall these studies support the idea that saturated fat does have a modest ability to increase blood cholesterol in the short term.
** Although I would love to hear comments from anyone who has done controlled diet trials. I'm sure this method had flaws, as it was applied in the 1960s.
*** Reference cited in the Tecumseh paper: Kannel, W et al. The Framingham Study. An epidemiological Investigation of Cardiovascular Diseases. Section 24: The Framingham Diet Study: Diet and the Regulation of Serum Cholesterol. US Government Printing Office, 1970.
**** Table 5 shows that the Pearson correlation coefficient for saturated fat intake vs. blood cholesterol is not significant; table 6 shows that children in the two highest tertiles of blood cholesterol have a significantly higher intake of saturated fat, unsaturated fat, total fat and sodium than the lowest tertile. The relationship between saturated fat and blood cholesterol shows no evidence of dose-dependence (cholesterol tertiles= 15.6g, 18.4g, 18.5g saturated fat). The investigators did not attempt to adjust for confounding factors.
