Good Calories, Bad Calories Part 6

Since the late 1970s, investigators have demonstrated the existence of other hormonal mechanisms by which insulin raises blood pressure-in particular, by stimulating the nervous system and the same flight-or-fight response incited by adrenaline. This was first reported by Lewis Landsberg, an endocrinologist who was then at Harvard Medical School and would later become dean of the Northwestern University School of Medicine. Landsberg showed that, by stimulating the activity of the nervous system, insulin increases heart rate and constricts blood vessels, thereby raising blood pressure. The higher the insulin level, the greater the stimulation of the nervous system, Landsberg noted. If insulin levels remained high, so Landsberg's research suggested, then the sympathetic nervous system would be constantly working to raise blood pressure. The heart-disease research community has paid attention to Landsberg's work, but has considered it relevant only for the obese. Because obesity is a.s.sociated with higher insulin levels, and because it's now believed that obesity causes higher insulin levels (whereas obesity itself is al egedly caused by the consumption of excess calories of al types), any possible link to carbohydrate consumption or "carbohydrate overfeeding" is overlooked. Even Landsberg has concentrated almost exclusively on the obesity-insulin-hypertension connection and ignored the idea that the increase in insulin levels due to excessive carbohydrate consumption, or due to the consumption of refined and easily digestible carbohydrates, might have a similar effect.

One question that wil be addressed in the coming chapters is why medical investigators and public-health authorities, like Landsberg, wil accept the effects of insulin on chronic diseases as real and potential y of great significance, and yet inevitably interpret their evidence in ways that say nothing about the unique ability of refined and easily digestible carbohydrates to chronical y elevate insulin levels. This is the dilemma that haunts the past fifty years of nutrition research, and it is critical to the evolution of the science of metabolic syndrome. As we wil discuss, the observation of diseases of civilization was hardly the only evidence implicating sugar and refined carbohydrates in these diseases. The laboratory research inevitably did, too. Yet the straightforward interpretation of the evidence-from carbohydrates to the chronic elevation of insulin to disease-was consistently downplayed or ignored in light of the overwhelming belief that Keys's dietary-fat hypothesis had been proved correct, which was not the case.

The coming chapters wil discuss the history of the science of metabolic syndrome both in the context of how the research was interpreted at the time, in a universe dominated by Keys's hypothesis, and then how it arguably should have been interpreted if the research community had approached this science without bias and preconceptions. The next five chapters describe the science that was pushed aside as investigators and public-health authorities tried to convince first themselves and then the rest of us that dietary fat was the root of al nutritional evils. These chapters divide the science of metabolic syndrome and the carbohydrate hypothesis into five threads, to simplify the tel ing (although by doing so, they admittedly oversimplify).

The first (Chapter 9) covers the research that directly chal enged the fundamental premise of Keys's hypothesis that cholesterol itself is the critical component in heart disease, and instead implicated triglycerides and the kinds of molecules known as lipoproteins that carry cholesterol through the blood, both of which are effectively regulated by the carbohydrate content of the diet rather than saturated fat. The chapter then explains how this research, despite its refutation of the fat-cholesterol hypothesis, has been a.s.similated into it nonetheless.

The second thread (Chapter 10) fol ows the evolution of the science of insulin resistance and hyperinsulinemia, the condition of having chronical y elevated insulin levels, and how that emerged out of attempts to understand the intimate relations.h.i.+p of obesity, heart disease, and diabetes and led to the understanding of metabolic syndrome and the entire cl.u.s.ter of metabolic and hormonal abnormalities that it entails.

The third (Chapter 11) discusses the implications of metabolic syndrome in relation to diabetes and the entire spectrum of diabetic complications.

The fourth (Chapter 12) discusses table sugar and high-fructose corn syrup, in particular, and the research suggesting that they have negative health effects that are unique among refined carbohydrate foods.

The last section of this history (Chapter 13) discusses how metabolic syndrome, and particularly high blood sugar, hyperinsulinemia, and insulin resistance, have physiological repercussions that can conceivably explain the appearance of even Alzheimer's disease and cancer.

Throughout these five chapters, the science wil be more technical than has typical y been the case in popular discussions of what we should eat and what we shouldn't. I believe it is impossible, though, to make the argument that nutritionists for a half century oversimplified the science to the point of generating false ideas and erroneous deductions, without discussing the science at the level of complexity that it deserves.

Chapter Nine.

TRIGLYCERIDES AND THE COMPLICATIONS OF CHOLESTEROL.

Oversimplification has been the characteristic weakness of scientists of every generation.

ELMER MCCOLLUM, A History of Nutrition, 1957 THE DANGER OF SIMPLIFYING A MEDICAL ISSUE for public consumption is that we may come to believe that our simplification is an appropriate representation of the biological reality. We may forget that the science is not adequately described, or ambiguous, even if the public-health policy seems to be set in stone. In the case of diet and heart disease, Ancel Keys's hypothesis that cholesterol is the agent of atherosclerosis was considered the simplest possible hypothesis, because cholesterol is found in atherosclerotic plaques and because cholesterol was relatively easy to measure. But as the measurement technology became increasingly more sophisticated, every one of the complications that arose has implicated carbohydrates rather than fat as the dietary agent of heart disease.

In 1950, the University of California medical physicist John Gofman wrote an article in Science that would be credited, albeit belatedly, with launching the modern era of cholesterol research. Gofman pointed out that cholesterol is only one of several fatlike substances that circulate through the blood and are known col ectively as lipids or blood lipids. These include free fatty acids and triglycerides,*41 the molecular forms in which fat is found circulating in the bloodstream. These could also be players in the heart-disease process, Gofman noted, and the fact that there was no easy way to measure their concentrations in the circulation didn't change that. Both cholesterol and triglycerides are shuttled through the circulation in particles cal ed lipoproteins, and these could also be players. The amount of cholesterol and triglycerides varies in each type of lipoprotein. So, when physicians measure total cholesterol levels, they have no way of knowing how the cholesterol itself is apportioned in individual lipoproteins. It is possible, Gofman noted, that in heart disease the problem may be caused not by cholesterol but by a defect in one of these lipoproteins, or an abnormal concentration of the lipoproteins themselves.

Eventual y, researchers came to identify these different cla.s.ses of lipoproteins by their density. Of those that appeared to play obvious roles in heart disease, three in particular stood out even in the early 1950s. Two of these are familiar today: the low-density lipoproteins, known as LDL, the bad cholesterol, and the high-density variety, known as HDL, the good cholesterol. (This is an oversimplification, as I wil explain shortly.) The third cla.s.s is known as VLDL, which stands for "very low-density lipoproteins," and these play a critical role in heart disease. Most of the triglycerides in the blood are carried in VLDL; much of the cholesterol is found in LDL. That LDL and HDL are the two species of lipoproteins that physicians now measure when we get a checkup is a result of the oversimplification of the science, not the physiological importance of the particles themselves.

In 1950, the only instrument capable of measuring the density of lipoproteins was an ultracentrifuge, and the only ultracentrifuge available for this work in America was being used by Gofman at the University of California, Berkeley. Gofman was both a physician and a physical chemist by training. During World War I , he worked for the Manhattan Project, and developed a process to separate plutonium that would later be used to produce H-bombs. After the war, Gofman set out to use the Berkeley ultracentrifuge to study how cholesterol and fat are transported through the blood and how this might be affected by diet and perhaps cause atherosclerosis and heart disease.

This was the research Gofman first reported in Science in 1950. He described how his ultracentrifuge "fractionated" lipoproteins into different cla.s.ses depending on their density, and he noted that one particular cla.s.s of lipoproteins, which would later be identified as LDL,*42 is more numerous in patients with atherosclerosis than in healthy subjects, in men than in women, in older individuals than in younger, and particularly conspicuous in diabetics, al of which suggested a possible role in heart disease. What these low-density lipoproteins did not do, Gofman reported, was to reflect consistently the amount of cholesterol in the blood, even though they carry cholesterol within them. Sometimes total cholesterol levels would be low in his subjects, he noted, and yet the concentration of these low-density lipoproteins would be abnormal y high. Sometimes total cholesterol would be high while the cholesterol contained in the low-density lipoproteins was low. "At a particular cholesterol level one person may show 25 percent of the total serum cholesterol in the form of [low-density lipoproteins], whereas another person may show essential y none in this form," Gofman wrote.

After Science published Gofman's article, and after aggressive lobbying on Gofman's part, the National Advisory Heart Council agreed to fund a test of his hypothesis that lipoproteins are the important factor in heart disease and that cholesterol itself is not. The test would be carried out by four research groups-led by Gofman at Berkeley, Irving Page at the Cleveland Clinic, Fred Stare and Paul Dudley White at Harvard, and Max Lauffer of the University of Pittsburgh-that col ectively identified five thousand men who were free of heart disease. When heart disease eventual y appeared, they would determine whether total cholesterol or Gofman's lipoproteins was the more accurate predictor.

While the three Eastern laboratories took three years to learn how to use an ultracentifruge for fractionating lipoproteins, Gofman proceeded with his own research, refined his understanding of how these lipoproteins predicted heart disease, and he then insisted that the a.n.a.lysis techniques be updated accordingly. The other investigators, however, were having considerable trouble duplicating Gofman's original a.n.a.lysis, and so they refused to accept any further modifications.

In 1956, the four groups published a report in the American Heart a.s.sociation journal Circulation, with a minority opinion written by Gofman and his Berkeley col eagues and a majority opinion auth.o.r.ed by everyone else. As the majority saw it, based on the state of Gofman's research in 1952, cholesterol was indeed a questionable predictor of heart disease risk, but the measurements of lipoproteins added little predictive power. "The lipoprotein measurements are so complex," the majority report declared, "that it cannot be reasonably expected that they could be done reliably in hospital laboratories." Gofman's minority opinion, based on the state of his research in 1955, was that LDL and VLDL, the very low-density lipoproteins, were good predictors of heart disease, but that the single best predictor of risk was an atherogenic index, which took into account these two lipoprotein cla.s.ses measured individual y and added them together. The greater the atherogenic index, the greater the risk of atherosclerosis and heart disease.

Gofman would later be vindicated, but the majority opinion prevailed at the time: studying lipoproteins held no value in the clinical management of heart disease. Gofman and his Berkeley col aborators continued the research alone through 1963, when Gofman left to establish a biomedical-research division at the Lawrence Livermore National Laboratory and spent the rest of his career working on the health effects of radiation.

Lost entirely in the contretemps were the dietary implications of Gofman's research. "While it is true that, for certain individuals, the amount of dietary fat is an important factor," Gofman explained, "it turns out that there are other more significant factors that need to be considered. Human metabolism is so regulated that factors other than the actual dietary intake of one of these const.i.tuents may determine the amount of that const.i.tuent that wil circulate in the bloodstream. Indeed, important observations have been made which indicate that certain substances in the diet that are not fatty at al may stil have the effect of increasing the concentration of the fat-bearing lipoprotein substances in the blood."

Though Gofman's studies had demonstrated that the amount of LDL in the blood can indeed be elevated by the consumption of saturated fats, it was carbohydrates, he reported, that elevated VLDL-containing some cholesterol and most of the triglycerides in the blood-and only by restricting carbohydrates could VLDL be lowered.

This fact was absolutely critical to the dietary prevention of heart disease, Gofman said. If a physician put a patient with high cholesterol on a low-fat diet, that might lower the patient's LDL, but it would raise VLDL. If LDL was abnormal y elevated, then this low-fat diet might help, but what Gofman cal ed the "carbohydrate factor" in these low-fat diets might raise VLDL so much that the diet would do more harm than good. Indeed, in Gofman's experience, when LDL decreased, VLDL tended to rise disproportionately. And if VLDL was abnormal y elevated to begin with, then prescribing a low-fat, high-carbohydrate diet would certainly increase the patient's risk of heart disease.

This was why Gofman described the measurement of total cholesterol as a "false and highly dangerous guide" to the effect of diet on heart disease.

Total-cholesterol measurements tel us nothing about the status of VLDL and LDL. Prescribing low-fat diets indiscriminately to anyone whose cholesterol appears to be elevated, or bombarding us with "generalizations such as 'we al eat too much fat,' or 'we al eat too much animal fat,'" would increase heart-disease risk for a large proportion of the population. "Neglect of [the carbohydrate] factor can lead to rather serious consequences," wrote Gofman in 1958, "first, in the failure to correct the diet in some individuals who are very sensitive to the carbohydrate action; and second, by al owing certain individuals sensitive to the carbohydrate action to take too much carbohydrate as a replacement for some of their animal fats."

By 1955, Pete Ahrens at Rockefel er University had come to this same conclusion, although Ahrens was specifical y studying triglycerides, rather than the VLDL particles that carry the triglycerides. Ahrens was considered by many investigators to be the single best scientist in the field of lipid metabolism. He had observed how the triglycerides of some patients shoot up on low-fat diets and fal on high-fat diets. This led Ahrens to describe a phenomenon that he called carbohydrate-induced lipemia (an excessive concentration of fat in the blood). When he gave lectures, Ahrens would show photos of two test tubes of blood serum obtained from the same patient-one when the patient was eating a high-carbohydrate diet and one on a high-fat diet. One test tube would be milky white, indicating the lipemia. The other would be absolutely clear. The surprising thing, Ahrens would explain, was "that the lipemic plasma was obtained during the high-carbohydrate period, and the clear plasma during the high-fat regimen." (Joslin had reported the same phenomenon in diabetics thirty years earlier. "The percent of fat" in the blood, he wrote, "rises with the severity of the disease...and is especial y related to the quant.i.ty of carbohydrate, which is being oxidized, rather than with the fat administered.") Over the course of a decade, Ahrens had seen only two patients whose blood serum became cloudy with triglycerides after eating high-fat meals. He had thirteen in whom carbohydrates caused the lipemia. Six of those thirteen had such high triglycerides that they had original y been referred to Ahrens from physicians who had misdiagnosed them as having a genetic form of high cholesterol. Since the VLDL particles that transport triglycerides, as Gofman had noted, also carry cholesterol and so contribute to the total cholesterol in the circulation, an elevated triglyceride level can elevate total cholesterol along with it. Ahrens believed that the fat-induced lipemia was a rare genetic disorder but the carbohydrate-induced lipemia was probably "an exaggerated form of the normal biochemical process which occurs in al people on high-carbohydrate diets." In both cases, the fat in the blood would clear up when the subjects went on a low-calorie diet. To Ahrens, this explained why the carbohydrate-induced increase in triglycerides was absent in Asian populations living primarily on rice. As long as they were eating relatively low-calorie diets compared with their level of physical activity, which was inevitably the case in such impoverished populations, the combination would counteract the triglyceride-raising effect of the carbohydrates.

The critical question was whether prolonged exposure to an abnormal y high triglyceride level increased the risk of atherosclerosis. If carbohydrate-induced lipemia was as common as Ahrens believed, "especial y in the areas of the world distinguished by caloric abundance and obesity," then it was important to know. If so, then having patients with high triglycerides eat less fat would only make the condition worse. By 1957, Ahrens was also warning about the dangers of oversimplifying the diet-heart science: maybe fat and cholesterol caused heart disease, or maybe it was the carbohydrates and triglycerides. "We know of no solid evidence on this point," wrote Ahrens, "and until the question is further explored we question the wisdom of prescribing low-fat diets for the general population."

The evidence that Ahrens was looking for came first from Margaret Albrink, who was then a young physician working with John Peters, chief of the metabolic division in the Department of Medicine at Yale University. Once again, the available technology drove the research. Peters was renowned in the medical community for his measurements of the chemical const.i.tuents of body fluids. For this purpose he had a device cal ed an a.n.a.lytical centrifuge, a less sophisticated version of Gofman's ultracentrifuge, which could quantify the triglyceride concentration of the blood. Peters's lab also a.n.a.lyzed blood samples for New Haven Hospital (now YaleNew Haven Hospital), so Peters suggested to Albrink that they use the a.n.a.lytical centrifuge to measure the triglycerides in those blood samples and test the hypothesis that high triglycerides are a.s.sociated with an increased risk of heart disease. Peters was a "contrarian," Albrink says; he didn't believe the cholesterol hypothesis. Nor did Evelyn Man, Peters's longtime col aborator. Albrink also worked with Wister Meigs, a Yale professor of preventive medicine who also served as company physician for the nearby American Steel and Wire Company. Meigs had been recording cholesterol levels in the plant employees, along with their family history of heart disease, diabetes, and other ailments. By 1960, Albrink, Man, and Meigs (Peters died in 1955) were comparing triglyceride and cholesterol levels of heart-disease patients from New Haven Hospital with the levels among the healthy employees of American Steel and Wire. Elevated triglyceride levels, they concluded, were far more common in coronary-heart-disease patients than high cholesterol: only 5 percent of healthy young men had elevated triglycerides, compared with 38 percent of healthy middle-aged men and 82 percent of coronary patients.

In May 1961, just a few months after the American Heart a.s.sociation publicly embraced Keys's hypothesis, both Ahrens and Albrink presented their research at a meeting of the a.s.sociation of American Physicians in Atlantic City, New Jersey. Both reported that elevated triglycerides were a.s.sociated with an increased risk of heart disease, and that low-fat, high-carbohydrate diets raised triglycerides. The New York Times covered Ahrens's talk-"Rockefel er Inst.i.tute Report Chal enges Belief that Fat Is Major Factor"-in a story buried deep in the paper. Ahrens's data suggested that "dietary carbohydrate, not fat, is the thing to watch in guarding against [atherosclerosis and heart disease]," the Times reported, and this "came as something of a surprise to many of the scientists and physicians attending the meeting." Albrink's talk did not make the newspaper, but she later told a similar story about her presentation. "It just about brought the house down," she recal ed. "People were so angry; they said they didn't believe it." This remained the case for much of the next decade. Albrink continued to work out the connection between carbohydrates, triglycerides, and heart disease and would present her results at conferences, where she would inevitably be attacked by proponents of Keys's hypothesis.

By the early 1970s, Albrink's interpretation of the evidence had been confirmed independently, first by Peter Kuo of the University of Pennsylvania, then by Lars Carlson of the Karolinska Inst.i.tute in Stockholm, and by the future n.o.bel laureate Joseph Goldstein and his col eagues from the University of Was.h.i.+ngton. Al three reported that high triglycerides were considerably more common in heart-disease victims than was high cholesterol. In 1967, Kuo reported in The Journal of the American Medical a.s.sociation that he had studied 286 atherosclerosis patients, of whom 246 had been referred to him by physicians who thought their patients had the genetic form of high cholesterol. This turned out to be the case for fewer than 10 percent. The other 90 percent had carbohydrate-induced lipemia, and, for most of these patients, their sensitivity to carbohydrates had elevated both their triglyceride levels and their cholesterol. When Kuo put his patients on a sugar-free diet, he reported, with only five to six hundred calories of starches a day, both their triglyceride levels and their cholesterol lowered. Two months later, JAMA published an editorial in response to Kuo's article, suggesting that the "almost embarra.s.singly high number of researchers [who had] boarded the 'cholesterol bandwagon'" had done a disservice to the field. "This fervent embrace of cholesterol to the exclusion of other biochemical alterations resulted in a narrow scope of study," the editorial said. "Fortunately, other fruitful approaches have been made possible in the past few years by identification of the fundamental role of such factors as triglycerides and carbohydrate metabolism in atherogenesis."

By then, however, the science had already become secondary to more practical issues. Despite JAMA's optimism that a new era was dawning, it was no longer a question of whether it was cholesterol or triglycerides that caused atherosclerosis and heart disease, whether saturated fat or carbohydrates were to blame, but which of the two hypotheses dominated the research. Here Keys's hypothesis had precedence. A generation of clinical investigators -the "cholesterol bandwagon"-had gathered an enormous amount of data, however ambiguous, on cholesterol levels and heart disease; only Albrink, Kuo, and a handful of other researchers had studied triglycerides. Only Gofman had studied the VLDL particles that transport triglycerides through the circulation.

Moreover, measuring triglycerides was stil much more difficult than measuring cholesterol, and so only the rare laboratory had the facilities to do it. The National Inst.i.tutes of Health, which was effectively the only source of funding for this research in the United States, had already committed its resources to three enormous studies-the Framingham Heart Study, Keys's Seven Countries Study, and the pilot programs of the National Diet-Heart Study. These studies would measure only cholesterol and so test only Keys's hypothesis. No consideration was given to any alternative hypothesis. By 1961, Keys and his col aborators in the Seven Countries Study had measured cholesterol in over ten thousand men. By 1963, they had completed the exams on another eighteen hundred men. Even had it been technical y possible to include triglycerides in the measurements, or to return to the original locales and retest for triglycerides, the cost would have been astronomical. The result, as we've seen, was considered a resounding victory for Keys's fat-cholesterol hypothesis.

The research that would final y lead to a large-scale test of the carbohydrate/ triglyceride/heart-disease hypothesis emerged from the National Inst.i.tutes of Health in early 1967. This was a col aboration between Donald Fredrickson and Robert Levy, who would become directors of the National Inst.i.tutes of Health and the National Heart, Lung, and Blood Inst.i.tute respectively, and Robert Lees, then of Rockefel er University. It was published in a fifty-page, five-part series in The New England Journal of Medicine. First Fredrickson, Levy, and Lees proposed a simplified cla.s.sification of lipoproteins (perhaps an oversimplification, they acknowledged), which divided the lipoproteins in the bloodstream into four categories: LDL, which typical y carried most of the cholesterol; VLDL, which carried most of the triglycerides; the high-density lipoproteins, HDL; and chylomicrons, which carry dietary fat from the intestine to the fat tissue. Then they proposed a cla.s.sification scheme for disorders of lipoprotein metabolism, each delineated by a roman numeral, that included both those of abnormal y high amounts of LDL cholesterol, which they suggested might be ameliorated by low-fat diets, as wel as those characterized by abnormal y high triglycerides carried in VLDL, which would be ameliorated by low-carbohydrate diets.

Four of the five lipoprotein disorders described in this series were characterized by abnormal y elevated levels of triglycerides in the very low-density lipoproteins. For this reason, Fredrickson, Levy, and Lees also warned against the dangers of advocating low-fat diets for al patients, because these diets increased carbohydrate consumption and so would elevate triglycerides and VLDL even further. By far the most common of the five lipoprotein disorders was the one designated Type IV, characterized by elevated VLDL triglycerides-"sometimes considered synonymous with 'carbohydrate-induced hyperlipemia,'" they wrote-and it had to be treated with a low-carbohydrate diet. "Patients with this syndrome," Lees later wrote, "form a sizable fraction of the population suffering from coronary heart disease."*43 Because Fredrickson, Levy, and Lees had also described an innovative and inexpensive technique for measuring the triglycerides and cholesterol carried in these different lipoproteins, the NIH provided the necessary funding for five studies-in Framingham, Puerto Rico, Honolulu, Albany, and San Francisco-to measure LDL cholesterol and VLDL triglycerides in these populations and determine their significance as risk factors for heart disease.

This research would take almost a decade to complete, and would const.i.tute the first time that NIH-funded research projects would measure anything other than total cholesterol in large populations.

The new research would also mark the first time that HDL was measured in large populations, and this would further confuse the diet/heart-disease relations.h.i.+p. The hypothesis that HDL particles or the cholesterol in HDL protects against heart disease had first been proposed in 1951 by David Barr and Howard Eder of New York HospitalCornel Medical Center. It had been confirmed in a handful of smal studies through the 1950s, and by Gofman in the last paper he published on lipoproteins and heart disease, as had the observation that when HDL was low triglycerides tended to be high, and vice versa, which suggested some underlying mechanism linking the two. Nonetheless, heart-disease researchers had paid little attention to HDL, as the NIH biostatistician Tavia Gordon later explained, because the idea of a "negative relation" between cholesterol and heart disease-high HDL cholesterol implies a low risk of heart disease-"simply ran against the grain." "It was easy to believe that too much cholesterol in the blood could 'overload' the system and hence increase the risk of disease," Gordon wrote, "but how could 'too much' of one part of the total cholesterol reduce the risk of disease?

To admit that fact chal enged the whole way of thinking about the problem." Now HDL, too, would be measured in these populations.*44 The results from the five studies were released in 1977 and divided into two publications, although Gordon had done the a.n.a.lyses for both. One reported on a comparison of nine hundred heart-disease cases with healthy controls from al five of the populations. The other addressed the prospective evidence from Framingham alone-measuring triglyceride, lipoprotein, and cholesterol levels in twenty-eight hundred subjects and then waiting four years to see how wel these levels predicted the appearance of heart disease. The findings were consistent. Both a.n.a.lyses confirmed Gofman's argument that total cholesterol said little about the risk of heart disease, and that the measurement of the triglycerides and cholesterol in the different lipoproteins was considerably more revealing. In men and women fifty and older, Gordon and his col aborators wrote in the Framingham paper, "total cholesterol per se is not a risk factor for coronary heart disease at al ." LDL cholesterol was a "marginal" risk factor, they reported. Triglycerides predicted heart disease in men and women in the a.n.a.lysis of cases from al five studies, but only in women in the Framingham a.n.a.lysis.

HDL was the "striking" revelation. Both a.n.a.lyses confirmed that the higher the HDL cholesterol the lower the triglycerides and the risk of heart disease.

The inverse relations.h.i.+p between HDL and heart disease held true for every age group from forty-year-olds to octogenarians, in both men and women, and in every ethnic group from Framingham, Ma.s.sachusetts, to Honolulu. "Of al the lipoproteins and lipids measured HDL had the largest impact on risk,"

Gordon and his col eagues wrote. For those fifty and over, which is the age at which heart disease ceases to be a rare condition, HDL was the only reliable predictor of risk.

The finding that high HDL cholesterol was a.s.sociated with a low risk of heart disease did not mean that raising HDL would lower risk, as Gordon and his col eagues noted, but it certainly suggested the possibility. Only a few studies had ever looked at the relations.h.i.+p of diet and lifestyle to HDL, and the results had suggested, not surprisingly, that anything that raised triglycerides would lower HDL, and vice versa. The "fragmentary information on what maneuvers wil lead to an increase in HDL cholesterol levels," Gordon and his col aborators wrote, "suggests that physical activity, weight loss and a low carbohydrate intake may be beneficial" (my italics).

This is where the story now takes some peculiar turns. One immediate effect of the revelation about HDL, paradoxical y, was to direct attention away from triglycerides, and with them the conspicuous link, until then, to the carbohydrate hypothesis. Gordon and his col eagues had demonstrated that when both HDL and triglycerides were incorporated into the risk equations of heart disease, or when obesity and the prediabetic condition of glucose intolerance were included in the equations along with triglycerides, the apparent effect of triglycerides diminished considerably. This result wasn't surprising, considering that low HDL, high triglycerides, obesity, and glucose intolerance al seemed to be related, but that wasn't the point. The relevant question for physicians was whether high triglycerides by themselves caused heart disease. If so, then patients should be advised to lower their triglycerides, however that might be accomplished, just as they were being told already to lower cholesterol. These risk-factor equations (known as multivariate equations) suggested that triglycerides were not particularly important when these other factors were taken into account, and this was how they would be perceived for another decade. Not until the late 1980s would the intimate a.s.sociation of low HDL, high triglycerides, obesity, and diabetes be considered significant-in the context of Gerald Reaven's Syndrome X hypothesis-but by then the heart-disease researchers would be committed to the recommendations of a national low-fat, high-carbohydrate diet.

Heart-disease researchers would also avoid the most obvious implication of the two a.n.a.lyses-that raising HDL offers considerably more promise to prevent heart disease than lowering either LDL or total cholesterol-on the basis that this hadn't been tested in clinical trials. Here the immediate obstacle, once again, was the inst.i.tutional investment in Keys's hypothesis. The National Inst.i.tutes of Health had committed its heart-disease research budget to two ongoing studies, MRFIT and the Lipid Research Clinics Trial, which together would cost over $250 mil ion. These studies were dedicated solely to the proposition that lowering total cholesterol would prevent heart disease. There was little money or interest in testing an alternative approach.

Gordon later recal ed that, when he presented the HDL evidence to the team of investigators overseeing MRFIT, "it was greeted with a silence that was very, how should I say it, expressive. One of them spoke up indicating he suspected this was a bunch of s.h.i.+t. They didn't know how to deal with it."

Indeed, the timing of the HDL revelations could not have been less convenient. The results were first revealed to the public in an American Heart a.s.sociation seminar in New York on January 17, 1977. This was just three days after George McGovern had announced the publication of the Dietary Goals for the United States, advocating low-fat, high-carbohydrate diets for al Americans, based exclusively on Keys's hypothesis that coronary heart disease was caused by the effect of saturated fat on total cholesterol. If the New York Times account of the proceedings is accurate, the AHA and the a.s.sembled investigators went out of their way to ensure that the new evidence would not cast doubt on Keys's hypothesis or the new dietary goals. Rather than chal enge the theory that excess cholesterol can cause heart disease, the Times reported, "the findings re-emphasize the importance of a fatty diet in precipitating life-threatening hardening of the arteries in most Americans," which is precisely what they did not do. According to the Times, saturated fat was now indicted not just for increasing LDL cholesterol, which it does, but for elevating VLDL triglycerides and lowering HDL, which it does not, and certainly not compared with the carbohydrates that McGovern's Dietary Goals were recommending al Americans eat instead.

In a more rational world, which means a research establishment not already committed to Keys's hypothesis and not whol y reliant on funding from the inst.i.tutions that had embraced the theory, the results would have immediately prompted smal clinical trials of the hypothesis that raising HDL prevented heart disease, just like those smal trials that had begun in the 1950s to test Keys's hypothesis. If those confirmed the hypothesis, then longer, larger trials would be needed to establish whether the short-term benefits translated to a longer, healthier life. But the NIH administrators decided that HDL studies would have to wait. Once the Lipid Research Clinics Trial results were published in 1984, they were presented to the world as proof that lowering cholesterol by eating less fat and more carbohydrates was the dietary answer to heart disease. There was simply no room now in the dogma for a hypothesis that suggested that raising HDL (and lowering triglycerides) by eating more fat and less carbohydrates might be the correct approach. No clinical trials of the HDL hypothesis would begin in the U.S. until 1991, when the Veterans Administration funded a twenty-center drug trial. The results, published in 1999, supported the hypothesis that heart disease could be prevented by raising HDL. The drug used in the study, gemfibrozil, also lowered triglyceride levels and VLDL, suggesting that a diet that did the same by restricting carbohydrates might have a similarly beneficial effect. As of 2006, no such dietary trials had been funded.

Through the 1980s and 1990s, as our belief in the low-fat heart-healthy diet solidified, the official reports on nutrition and health would inevitably discuss the apparent benefits of raising HDL-the "good cholesterol"-and would then observe correctly that no studies existed to demonstrate this would prevent heart disease and lengthen life. By 2000, wel over $1 bil ion had been spent on trials of cholesterol-lowering, and a tiny fraction of that amount on testing the benefits of raising HDL. Thus, any discussions about the relative significance of raising HDL versus lowering total cholesterol would always be filtered through this enormous imbalance in the research efforts. Lowering LDL cholesterol would always have the appearance of being more important.

It was the revelations that emerged from the two HDL publications in 1977 that led to the conventional wisdom about LDL, triglycerides, and HDL that we live with today. The National Heart, Lung, and Blood Inst.i.tute and the American Heart a.s.sociation responded to the new research by focusing on two pragmatic concerns: first, to keep the science sufficiently simple that it could be translated into equal y simple guidelines for patient care, and, second, to reconcile these new observations with Keys's hypothesis and the $250 mil ion worth of studies that were putting it to the test. If total cholesterol was not a risk factor for heart disease above the age of fifty, as Gordon's Framingham a.n.a.lysis noted, then that seemed to refute Keys's hypothesis. One immediate goal, therefore, was to make sure that those aspects of the hypothesis that had seemed reasonably certain were not discarded prematurely on the basis of findings that might also someday turn out to be erroneous.

Since both of the new a.n.a.lyses had concluded that LDL cholesterol was a.s.sociated with a slightly increased risk of heart disease, and since up to 70 percent of the total cholesterol in the circulation may be found in LDL, the American Heart a.s.sociation and the proponents of Keys's hypothesis now s.h.i.+fted the focus of scientific discussions from the benefits of lowering total cholesterol to the benefits of lowering LDL cholesterol. "Whatever the underlying disorder," noted the Framingham investigators in 1979, "much of what has been learned in the past about the il effects of a high serum total cholesterol can be attributed to the a.s.sociated elevated levels of LD lipoprotein...."

Making LDL the "bad cholesterol" oversimplified the science considerably, but it managed to salvage two decades' worth of research, and to justify why physicians had bothered to measure total cholesterol in their patients. One consequence of this effort was an upgrading of the adjectives used to describe the predictive ability of LDL. In 1977, Gordon and his col aborators had described LDL cholesterol as a "marginal risk factor" for heart disease.

Within two years, the same authors were using the identical data to describe LDL as a "powerful predictor of risk in subjects younger than the age of 50"

and as showing "a significant contribution...to coronary heart disease in persons older than the age of 50 and practical y up into the eighties." This practice has continued unabated.*45 Another s.h.i.+ft in emphasis was to incorporate HDL and some combination of triglycerides, LDL, and total cholesterol into the calculation of a "lipid profile" of heart-disease risk, a process that was initiated with the very first articles by Gordon and his col aborators. These lipid profiles al owed for the continued use of LDL or total cholesterol in the calculation of heart-disease risk, even though they added little or no predictive power to the use of HDL alone.

Ironical y, these lipid profiles also provided the rationale for physicians to keep measuring total cholesterol in their patients, even though it had now been confirmed, as Gofman had noted a quarter-century earlier, that it was a dangerously unreliable predictor of risk. The reason is that LDL cholesterol itself happened to be particularly difficult to measure.46 It was not the kind of measurement that physicians could easily order up for their patients. And since it didn't seem to matter in these lipid profiles whether it was total or LDL cholesterol that was included along with HDL-either way, HDL was the dominant predictor of risk-then, "from a practical point of view," as Gordon and his col eagues noted, "total cholesterol can subst.i.tute for LDL cholesterol" in calculating risk. Total cholesterol could be measured easily in the clinic, so physicians would continue to measure it. The evidence had dictated a complete turnabout in the science, and then pragmatic considerations had turned it about again, until the clinical management of patients and the public perception were back exactly where they had started.

The revelations about HDL had equal y little influence on the inst.i.tution of a national low-fat, high-carbohydrate diet. Whether or not triglycerides were an independent risk factor, once the protective nature of HDL was confirmed, then Gofman's argument of 1950 was also reaffirmed: there were at least two potential diet-related ways of preventing heart disease, and any treatment that improved the situation with one risk factor had to avoid exacerbating the situation with the other. In the 1960s, Gofman, Ahrens, Albrink, and Fredrickson, Levy, and Lees had al discussed the dangers of replacing the fat in the diet with carbohydrates because this would elevate triglycerides. Now the dangers of lowering HDL became the issue. "In the search for an optimal therapy for avoiding or correcting atherosclerosis," as the Framingham investigators noted in 1979, "the ideal lipid response would appear to be the one that raises HD lipoprotein as it lowers LD lipoprotein. Therapeutic maneuvers that affect only one of these lipoprotein particle systems in a favorable way, while adversely affecting the other, may be less promising...."

Diets that lowered cholesterol by replacing saturated fat with polyunsaturated fats would have accomplished such a balancing act, but there was legitimate concern that polyunsaturated fats were carcinogenic, and so the AHA had simply recommended fat reduction in general. This meant replacing the fat calories with carbohydrates. But the "good cholesterol" in HDL would be diminished by eating more carbohydrates. By the 1980s, discussions of heart-disease prevention typical y avoided this dilemma by neglecting to mention the effect of carbohydrates on HDL.*47 Instead, people were told to raise their HDL through exercise and weight loss, and then prescribed, as the American Heart a.s.sociation did, low-fat, high-carbohydrate diets as the means to lose that weight.

In 1985, Scott Grundy and his col eague Fred Mattson provided what appeared to be the ideal compromise-a dietary means both to lower LDL cholesterol and to raise HDL cholesterol without consuming more carbohydrates or saturated fats. This was monounsaturated fats, such as the oleic acid found in olive oil, and it served to keep the focus on the fat in the diet, rather than the carbohydrates. In the 1950s, Keys had a.s.sumed that monounsaturated fats were neutral, because they had no effect on total cholesterol. But this apparent neutrality, as Grundy reported, was due to the ability of these fats simultaneously to raise HDL cholesterol and lower LDL cholesterol. Saturated fats raise both HDL and LDL cholesterol. Carbohydrates lower LDL cholesterol but also lower HDL. Grundy and Mattson's discovery of the double-barreled effect of monounsaturated fats, and particularly oleic acid, reignited the popular interest in the Mediterranean diet as the ideal heart-healthy diet, though it seemed to be heart-healthy only in some Mediterranean regions and not in others, and such diets, as even Grundy conceded, had never been tested. When they final y were tested in two clinical trials in the 1990s-the Lyon Diet Heart Trial and an Italian study known as GISSI-Prevenzione-both supported the contention that the diet prevented heart attacks, but neither provided evidence that it did so by either raising HDL or lowering LDL, which was how it was now al eged to work.

The observation that monounsaturated fats both lower LDL cholesterol and raise HDL also came with an ironic twist: the princ.i.p.al fat in red meat, eggs, and bacon is not saturated fat, but the very same monounsaturated fat as in olive oil. The implications are almost impossible to believe after three decades of public-health recommendations suggesting that any red meat consumed should at least be lean, with any excess fat removed.

Consider a porterhouse steak with a quarter-inch layer of fat. After broiling, this steak wil reduce to almost equal parts fat and protein.*48 Fifty-one percent of the fat is monounsaturated, of which 90 percent is oleic acid. Saturated fat const.i.tutes 45 percent of the total fat, but a third of that is stearic acid, which wil increase HDL cholesterol while having no effect on LDL. (Stearic acid is metabolized in the body to oleic acid, according to Grundy's research.) The remaining 4 percent of the fat is polyunsaturated, which lowers LDL cholesterol but has no meaningful effect on HDL. In sum, perhaps as much as 70 percent of the fat content of a porterhouse steak wil improve the relative levels of LDL and HDL cholesterol, compared with what they would be if carbohydrates such as bread, potatoes, or pasta were consumed. The remaining 30 percent wil raise LDL cholesterol but wil also raise HDL cholesterol and wil have an insignificant effect, if any, on the ratio of total cholesterol to HDL. Al of this suggests that eating a porterhouse steak in lieu of bread or potatoes would actual y reduce heart-disease risk, although virtual y no nutritional authority wil say so publicly. The same is true for lard and bacon.

"Everything should be made as simple as possible," Albert Einstein once supposedly said, "but no simpler." Our understanding of the nutritional causes of heart disease started with Keys's original oversimplification that heart disease is caused by the effect of al dietary fat on total serum cholesterol. Total cholesterol gave way to HDL and LDL cholesterol and even triglycerides. Al fat gave way to animal and vegetable fat, which gave way to saturated, monounsaturated, and polyunsaturated fat, and then polyunsaturated fats branched into omega-three and omega-six polyunsaturated fats. By the mid-1980s, these new levels of complexity had stil not deterred the AHA and NIH from promoting carbohydrates as effectively the antidote to heart disease, and either al fats or just saturated fats as the dietary cause.

What would now become apparent was that LDL cholesterol is little more than an arbitrary concept that oversimplifies its own complex diversity. The fact that LDL and LDL cholesterol are not synonymous complicates the science. Just as Gofman had reported in 1950 that cholesterol itself was divided up among different lipoproteins, and those lipoproteins had different atherogenic properties and responded differently to diet, a lipid metabolism specialist named Ronald Krauss, using Gofman's ultracentrifuge, began reporting in 1980 that low-density lipoproteins were in turn composed of different, distinct subcla.s.ses, each containing differing amounts of cholesterol, and each, once again, with different atherogenic properties and different behavior in response to the carbohydrates and fats in our diet. Although Krauss has long been considered one of the most thoughtful researchers in nutrition and heart disease-the American Heart a.s.sociation has treated him as such-it's worth noting in advance that his dietary research has been almost universal y ignored, precisely because of its ultimate implications for what const.i.tutes a healthy diet and what does not.

LDL cholesterol is only a "marginal risk factor," Tavia Gordon and his col eagues had observed in 1977. In other words, little difference can be observed between the average LDL cholesterol of those with and without heart disease. Only by comparing the LDL-cholesterol and heart-disease rates between nations (with al the attendant complications of such comparisons) can conspicuous differences be seen. In the a.n.a.lysis from Framingham, San Francisco, Albany, Honolulu, and Puerto Rico published by Gordon and his col aborators, the average LDL cholesterol of heart-disease sufferers was only a few percentage points higher than the average of those who remained healthy. "If you look in the literature and just look at the average coronary patients," Krauss says, "their LDL-cholesterol levels are often barely discernibly elevated compared to patients who do not have coronary disease."

In the late 1940s, Gofman and his col aborators began asking why the same level of LDL cholesterol wil cause heart disease in some people but not in others. Krauss and his col aborators began asking this question again, thirty years later.

Krauss himself is an idiosyncratic figure in this world. He has produced a dozen years of research suggesting that high-carbohydrate diets, for the great proportion of the population, are the nutritional cause of heart disease, and yet he has also chaired the nutrition committee of the American Heart a.s.sociation and was the primary author of the 1996 and 2000 AHA nutrition guidelines. In the process, he eased the AHA away from its thirty-year-old position that the maximum fat content of a heart-healthy diet should be 30 percent of calories. Or, as Krauss remarked, he managed to put the "30-percent-fat recommendation in smal print." Krauss trained as a physician in the late 1960s and then worked with Fredrickson and Levy at the NIH, where he discovered a protein known as hepatic lipase that regulates how the liver metabolizes lipoproteins. He then moved to Berkeley to practice internal medicine, and it was there, in 1976, that he began working with Gofman's ultracentrifuge and with Alex Nichols and Frank Lindgren, both of whom had col aborated with Gofman in the 1950s.

When Krauss began his research at Berkeley, he had what he cal s "this conventional notion, which many people stil have, that LDL is just one thing, a single ent.i.ty." But that turned out not to be the case. Using data from the ultracentrifuge dating back to the early 1960s, Krauss discovered that LDL actual y comes in distinct subspecies, al characterized by stil finer gradations in density and size. "It was blazingly obvious. Unignorable," says Krauss.*49 Eventual y, Krauss identified seven discrete subcla.s.ses of LDL. He also noted that the smal est and densest of the low-density lipoproteins had two significant properties: it had a strong negative correlation with HDL, and it was the subspecies that was elevated in patients with heart disease.

In the early 1980s, Krauss published three papers on what he cal s the "remarkable heterogeneity of LDL," al of which, he says, were met with indifference mixed with occasional hostility. Acceptance of Krauss's research was also constrained by the fact that Gofman's ultracentrifuge had been necessary to differentiate these LDL subcla.s.ses, which meant that this, too, was not the kind of measurement that could be ordered up easily by physicians. In his later publications, Krauss described a simpler, inexpensive measurement technique, but the research was stil perceived as an esoteric endeavor.

To understand the implications of this a.s.sociation between smal , dense LDL and heart disease, it helps to picture the configuration of the low-density lipoprotein itself. Imagine it as a bal oon. It has a single protein-known as apo B, for short-that serves as the structural foundation of the bal oon and holds it together. It has an outer membrane that is composed of cholesterol and fats of yet another type, cal ed phospholipids. And then, inside the bal oon, inflating it, are triglycerides and more cholesterol. The size of the LDL bal oon itself can vary, depending on the amount of triglycerides and cholesterol it contains. Thus, as Krauss reported, some people have mostly large, fluffy LDL, with a lot of cholesterol and triglycerides inflating the bal oon, and some people have mostly smal er, denser LDL particles, with less cholesterol and triglycerides.

In the 1970s, investigators had developed yet another way to quantify the concentration of these circulating lipoproteins, in this case by counting only the number of apo B proteins that provide the structural foundation to the LDL bal oon. Because there's only one protein per LDL particle, and because VLDL is also composed of identical apo B proteins, this technique measured the number of LDL and VLDL particles in a blood sample, rather than the cholesterol or triglycerides they contained. As it turned out, the number of apo B proteins, and so the total number of LDL and VLDL particles combined, is also abnormal y elevated in heart-disease patients. This was first reported in 1980 by Peter Kwiterovich, a lipid-metabolism specialist from Johns Hopkins, together with Al an Sniderman, a cardiologist from McGil University. Kwiterovich and Sniderman then col aborated with Krauss on the last of his three papers on the heterogeneity of LDL. In 1983, they reported that the disproportionate elevation in the apo B protein in heart-disease patients was due to a disproportionate elevation in the amount of the smal est and densest of the low-density lipoproteins.

This explained what Krauss had set out to understand: why two people can have identical LDL-cholesterol levels and yet one develops atherosclerosis and coronary heart disease and the other doesn't-why LDL cholesterol is only a marginal risk factor for heart disease. If we have low LDL cholesterol, but it's packaged almost exclusively in smal , dense LDL particles-the smal er bal oons-that translates to a higher risk of heart disease. If we have high LDL cholesterol, but it's packaged in a smal er number of large, fluffy LDL particles-the larger bal oons-then our heart-disease risk is significantly lower. Smal , dense LDL, simply because it is smal and dense, appears to be more atherogenic, more likely to cause atherosclerosis. Smal , dense LDL can squeeze more easily through damaged areas of the artery wal to form incipient atherosclerotic plaques. Sniderman describes smal , dense LDL as the equivalent of "little bits of sand" that get in everywhere and stick more avidly. The relative dearth of cholesterol in these particles may also cause structural changes in the protein that make it easier for it to adhere to the artery wal to begin with. And because smal , dense LDL apparently remains in the bloodstream longer than larger and fluffier LDL, it has more time and greater opportunities to do its damage. Final y, it's possible that LDL has to be oxidized-the biological equivalent, literal y, of rusting-before it can play a role in atherosclerosis, and the existing evidence suggests that smal , dense LDL oxidizes more easily than the larger, fluffier variety.

Through the 1980s, Krauss continued to refine this understanding of how LDL subspecies affect heart disease. He discovered that the appearance of LDL in the population fal s into two distinct patterns or traits, which he cal ed pattern A and pattern B. Pattern A is dominated by large, fluffy LDL and implies a low risk of heart disease; pattern B is the dangerous one, with predominantly smal , dense LDL. Pattern B is invariably accompanied by high triglycerides and low HDL. Pattern A is not. In 1988, Krauss and his col aborators reported in JAMA that heart-disease patients were three times more likely to have pattern B than pattern A. Krauss cal ed pattern B the atherogenic profile. Diabetics have the identical pattern.

The effect of diet on this atherogenic profile now became the pivotal issue. In the 1960s and most of the 1970s, the dietary goal was to lower total cholesterol. After the 1977 revelations about HDL, the best diet became the one that lowered LDL cholesterol and maybe raised HDL in the process. But if Krauss and his col aborators were right, a diet that lowers total cholesterol or LDL cholesterol can conceivably do so in a way that actual y increases the proportion of smal , dense LDL in the blood turning the healthy pattern A trait into the atherogenic pattern B. If we focus on LDL cholesterol alone, such a diet might appear to prevent heart disease. But if the size, density, and number of the LDL subspecies are indeed the important variables, the diet could in fact increase heart-disease risk.

Though pattern A and B traits appear to be strongly influenced by genetics, diet and other lifestyle factors play a critical role. In the late 1980s, Krauss began a series of clinical trials to explore the a.s.sociation between diet and the dangerous smal , dense LDL. The results of his seven trials have been consistent: the lower the fat in the diet and the higher the carbohydrates, the smal er and denser the LDL and the more likely the atherogenic pattern B appears; that is, the more carbohydrates and the less fat, the greater the risk of heart disease.

On a diet that Krauss cal s the "average American diet," with 35 percent of the calories from fat, one in three men wil have the atherogenic pattern B profile. On a diet of 46 percent fat, this proportion drops: only one man in every five manifests the atherogenic profile. On a diet of only 10 percent fat, of the kind advocated by diet doctors Nathan Pritikin and Dean Ornish, two out of every three men wil have smal , dense LDL and, as a result, a predicted threefold higher risk of heart disease. The same pattern holds true in women and in children, but the percentages with smal , dense LDL are lower. Krauss and his col eagues even tested the effect of types of fat on these lipoproteins, and reported that, the more saturated fat in the diet, the larger and fluffier the LDL-a beneficial effect.*50 Though the concept of smal , dense LDL as a risk factor for heart disease has been accepted into the orthodox wisdom, as has Krauss's atherogenic profile (although now renamed atherogenic dyslipidemia), his dietary research has had no perceptible influence on discussions of the dietary prevention of heart disease. The implications are so provocative that many investigators simply ignore them. Even those clinical investigators who firmly believe that smal , dense LDL is indeed the atherogenic form of LDL often refuse to comment on the dietary implications. "Wel , I would rather not get into that," said the University of Was.h.i.+ngton epidemiologist Melissa Austin, who studies triglycerides and heart disease and has col aborated with Krauss on studies of the smal , dense LDL.

Goran Wal dius, a cardiologist at the Karolinska Inst.i.tute in Stockholm, had the same response. Wal dius is the princ.i.p.al investigator of an enormous Swedish study to ascertain heart-disease risk factors. The 175,000 subjects include every patient who received a health checkup in the Stockholm area in 1985. Blood samples were taken at the time, and Wal dius and his col eagues have been fol owing the subjects ever since, to see which measures of cholesterol, triglycerides, or lipoproteins are most closely a.s.sociated with heart disease. Far and away, the best predictor of risk, as Wal dius reported in 2001, was the concentration of apo B proteins, reflecting the dominance of smal , dense LDL particles. Half of the patients who died of heart attacks, he reported, had normal LDL-cholesterol levels but high apo B numbers. Apo B is a much better predictor of heart disease than LDL cholesterol, Wal dius said, because LDL cholesterol "doesn't tel you anything about the quality of the LDL." But when asked in an interview to comment on Krauss's research and the subject of dietary interventions that might increase the size of LDL particles, Wal dius said, "I'l have to pa.s.s on that one."

The notion that carbohydrates determine the ultimate atherogenicity of lipoproteins is surprisingly easy to explain by the current understanding of fat-and-cholesterol transport. This model also accounts neatly for the observed relations.h.i.+p between heart disease, triglycerides, and cholesterol, and so const.i.tutes another level of the physiological mechanisms underlying the carbohydrate hypothesis. The details are relatively straightforward, but, not surprisingly, they represent a radical s.h.i.+ft from the mechanisms envisioned by Keys and others, in which coronary artery disease is caused by the simple process of saturated fat raising total-cholesterol or LDL-cholesterol levels. This is another way in which the subspecialization of medical researchers works against progress. For most epidemiologists, cardiologists, internists, nutritionists, and dieticians, their knowledge of lipoprotein metabolism dates to their medical or graduate-school training. Short of reading the latest biochemistry textbooks or the specialized journals devoted to this research, they have few available avenues (and little reason, as they see it) for keeping up-to-date, and so the current understanding of these metabolic processes escapes them. The details of lipoprotein metabolism circa 2007 remain a mystery to the great proportion of clinicians and investigators involved in the prevention of heart disease.

One key fact to remember in this discussion is that LDL and LDL cholesterol are not one and the same. The LDL carries cholesterol, but the amount of cholesterol in each LDL particle wil vary. Increasing the LDL cholesterol is not the same as increasing the number of LDL particles.

There are two ways to increase the amount of cholesterol in LDL. One is to increase the amount of cholesterol secreted to begin with; the other is to decrease the rate of disposal of cholesterol once it's been created (which is apparently what happens when we eat saturated fat). Either method wil eventual y result in elevated LDL cholesterol. Joseph Goldstein and Michael Brown worked out the details of the clearance-and-disposal mechanism in the 1970s, and this work won them the n.o.bel Prize.

As for secretion, the key point is that most low-density lipoproteins, LDL, begin their lives as very low-density lipoproteins, VLDL. (This was one implication of the observation that both LDL and VLDL are composed of the same apo B protein, and it was established beyond reasonable doubt in the 1970s.) This is why VLDL is now commonly referred to as a precursor of LDL, and LDL as a remnant of VLDL. If the liver synthesizes more cholesterol, we end up with more total cholesterol and so more LDL cholesterol, although apparently not more LDL particles. If the liver synthesizes and secretes more VLDL, we wil also end up with more LDL cholesterol but we have more LDL particles as wel , and they'l be smal er and denser.

This process is easier to understand if we picture what's actual y happening in the liver. After we eat a carbohydrate-rich meal, the bloodstream is flooded with glucose, and the liver takes some of this glucose and transforms it into fat-i.e., triglycerides-for temporary storage. These triglycerides are no more than droplets of oil. I