Junkfood Science: Heart healthy diets — Part two of the American Heart Association’s new Guidelines for Women

June 17, 2007

Heart healthy diets — Part two of the American Heart Association’s new Guidelines for Women

Last month, the American Heart Association released its new Evidence-based Guidelines for Cardiovascular Disease Prevention in Women. In a dramatic move, they had made their secondary management recommendations, formerly used for high cardiac risk patients, applicable to all women, as being necessary for the primary prevention of heart disease and prevention of premature deaths.

In part one, we began an examination of the evidence used by the AHA to support its new Guidelines, specifically those for the clinical management of serum “cholesterol” levels in women. The evidence was not nearly the slam dunk most healthcare professionals, and certainly most consumers, believed.

As promised, we’re devoting an entire post to the healthy eating recommendations in the AHA Guidelines, which included recommendations for watching our weight and dieting. The idea that a healthy diet can prevent heart disease and other chronic diseases of aging, and premature death is perhaps the most strongly-held belief among consumers today. It’s become popular to feel that people are responsible for their health problems and have control over their health risk indices, especially by eating “right.”

As we look at each study listed in the summary of evidence used by the AHA Expert Committee to support its Heart Healthy Diet, alcohol and weight recommendations, readers will be better able to see for themselves if beliefs in an ideal way to eat and the benefits of a heart “healthy” diet are based on credible science.


The AHA Guidelines state that to prevent heart disease and premature death: Women should maintain or lose weight through an appropriate balance of physical activity, caloric intake, and formal behavioral programs when indicated to maintain/achieve a BMI between 18.5 and 24.9 and a waist circumference ²35 in.

The expert panel provided these studies to support its recommendations:

Harris 1993 — an observational study reporting only relative risks for developing cardiovascular disease; found that weight loss increased risks by about two-fold, regardless of the women’s BMI or if they lost more or less than 7.7% of their body weight; the heavier women had slightly higher risks from dieting than thinner women; no mortality data reported

Harris 1997 — observational study reporting only relative risks for developing cardiovascular disease; found risks doubled among those who lost 10% or more of body weight — with higher BMIs increasing risks by 2 1/2 times and BMIs under 24 seeing an increased risk of just under two-fold; those with high BMIs who gained 10% of their body weight saw the same increase in risks; no benefit seen in weight loss; no mortality data reported

Li 2006 — observational study using Nurse’s Health Study database; examined physical activity, not weight loss

Manson 1995 — observational study using a self-selected subgroup from the Nurses Health Study database looking at BMI - mortality correlations which have since been refuted; didn’t examine weight loss

Paffenbarger 1993 — No women in this cohort study; didn’t examine dieting or weight loss

Singh 1996 — A randomized clinical trial on 480 middle-age East Indian women with heart disease or at high risk — not a primary prevention trial and not relevant to these Guidelines.[It compared a diet high in produce and legumes and exercise with a group following a low-fat diet without exercise. Four more deaths (3 additional sudden cardiac deaths) in the nonexercising, low-fat group.]

Walker 1995 — No women in this cohort and no mortality data reported; but men who lost up to 10% of their weight increased their risks for cardiovascular disease by 49%, whereas those who gained up to 10% of their weight reduced their’s by more than half.

Conclusion: The AHA supplied no clinical evidence to support recommendations for women to watch their weight or lose weight to prevent heart disease or reduce their risks of dying prematurely.

Yet, if you’ve taken one of those Health Risk Assessments from your insurance company or employer, weight loss is encouraged for two-thirds of us, even though the highest mortality figures are consistently seen among those who’ve lost weight.


The AHA Guidelines to prevent heart disease and reduce risks for premature death recommends that women restrict alcohol consumption to no more than 1 drink per day.

To support its alcohol recommendations, the AHA supplied an impressive-looking list of 54 studies. This is an example of being careful not to be taken in by large bibliographies. A close look reveals that few studies were clinical trials and most were observational studies of selected cohorts. As we’ve learned, such epidemiological studies can only look for correlations, not ever make conclusions as to causation.

First, most only looked at men! Women were not included in 32 of the studies, and the AHA Expert Committee specifically noted on each of these that their results are “not applicable to women:”

Albert 1999

Blackwell 1980

Boffetta 1980

Camargo 1997

Doll 1994

Emburso 2005

Farchi 1992

Hart 1999

Hein 1996

Kauhanan 1997

Kitamura 1998

Kittner 1983

Kivela 1988

Kozarevic 1982

Mukamal 2003

Muntwyler 1998

Paunio 1996

Renaud 2004

Rimm 1992

Romelsjo 19991

Rosengren 1988

Salonen 1983

Shaper 2000

Suh 1992

Suhonen 1987

Tanasescu 2001

Tsugane 1999

Wannamethee 1996, 1997, 1999, 1999 and 2001

A number of other studies were not primary prevention trials, but secondary management of patients with heart disease or other risk factors, and irrelevant for these Guidelines on the primary prevention of heart disease among asymptomatic women, such as:

Mukamal 2001, 2005, 2006 — cohort studies of seniors hospitalized for heart attacks

Palmer 1995 — cohort study of hypertensive patients (although it reported lower mortality risks among those drinking the most and 25% higher relative risks among nondrinkers)

Solomon 2000 — observational study of diabetic women in the Nurses Health Study database (found lower mortalities among those with the highest alcohol consumptions)

Whiteman 1999 — a study of British women with angina (reported the lowest cardiovascular and stroke deaths among the heavy drinkers)

Yang 1999 — A heart watch cohort of diabetic and patients at high risk, but no mortality data was reported

So, we are left with 15 primary prevention studies that included women. See if you can find one that supports the AHA alcohol guidelines:

Ebbert 1995 — observational study using Iowa’s Women Health Study database; among nonsmokers (current or former) risks of cardiovascular deaths were less (nearly half) among those drinking more than 14 grams* of alcohol a day

Garg 1993 —no mortality data was reported in this cohort, but it noted that the greater the alcohol consumption the lower the relative risk for heart attacks, and risks only began creeping up at amounts over 25 grams* a day, but were still lower than those drinking less than 4.5 grams per day

Gronbaek 2000— a cohort of women around 57 years old; found cardiovascular deaths were lower the more alcohol consumed, with relative risks half among those drinking more than 35 drinks a week

Gronbaek 2004— reported that relative risk for all cause mortality decreased with increasing alcohol intake, with those drinking more than 13 drinks a week were still associated with lower risks than those drinking less than one drink a week

Keil 1997— cohort study which found lower mortality rates among men and women who drank as compared to those who didn’t; those drinking more than 80 grams* of alcohol per day having no increased relative risks; the highest risks seen among nondrinkers

Klatsky 1992— cohort study that found no increase in relative risks for mortality until women were drinking more than 6 drinks per day

Lazarus 1991 — observational cohort study which reported higher relative risks for mortality among drinkers who stopped drinking, with greatest risks (72% higher) seen among the heaviest drinkers who stopped drinking

Mukamal 2006— cohort study that reported no mortality figures; did find steadily lower risks for cardiac events the more alcohol that was consumed, with risks nearly half among those drinking more than 14 drinks per week

Mukamal 2005 — no mortality data reported on this cohort, but found fewer strokes among those who drank, with no increased relative risks even among those drinking more than 14 drinks per week

Rehm 1997— reported 2 1/3 times higher risks of mortality among nondrinkers, lowest relative risks seen among drinkers, only increasing above nondrinkers at more than 29 drinks per week

Simons 1996 — reported lower mortality risks associated with the greater the alcohol consumed; 34% lower relative risk among the highest consumptions (up to 28 drinks per week)

Stampfer 1988 — reported a 3-5 times higher relative risk among nondrinkers as compared to drinkers; the highest consumptions of alcohol (more than 25 grams* per day) associated with a 40% lower mortality risk

Trevisan 2001 — observational study of cohort of 15,168 men and women which found lower mortalities among drinkers of wine or spirits, with higher risks seen only among 38 people (0.2%) who guzzled alcohol unassociated with meals

Walsh 2002 — observational using Framingham Heart Study database; no mortality data reported

Wannamethee 2004 — cohort of German women reported lower mortality risks among drinkers, even those drinking more than 20 grams* per day

Conclusions: These findings are quite different from the AHA Guideline recommendations. While these were epidemiological studies rather than clinical intervention trials, objectively, the relative risks for mortality and heart disease consistently support the healthfulness of drinking. They clearly provide no support for limiting alcohol to one drink per day.

Yet, if you’ve taken one of those Health Risk Assessments from your insurance company or employer, any alcohol reported is given a black mark, even though the highest mortality figures are consistently seen among nondrinkers.


The AHA Guidelines state that to prevent heart disease and lower risks for premature death: Women should consume a diet rich in fruits and vegetables; choose whole-grain, high-fiber foods; consume fish, especially oily fish, at least twice a week; limit intake of saturated fat to <10% of energy, and if possible to <7%, cholesterol to <300 mg/d, and sodium intake to <2.3 g/d (approximately 1 tsp salt). Consumption of trans-fatty acids should be as low as possible (eg, <1% of energy).

To support its dietary recommendations, the AHA supplied an even more impressive-looking list of 94 studies under its Heart Healthy Diet evidence supplement. Again, we can’t be impressed simply by large bibliographies, as a close look reveals that only one study was a clinical trial, the rest were epidemiological studies dredging through databases of selected cohorts looking for correlations and reporting findings as relative risks.

Of the studies the AHA used, 36 included no women at all and the Expert Committee specifically noted their results were not applicable to women:

Abbott 2003 — a study of magnesium

Al-Delaimy 2003 — a study of calcium supplements

Appleby 1999

Ascherio 1994, 1995, 1996

Dauchet 2004 — reported “no evidence for any association between vegetable intake and total CHD events” among the men

Daviglus 1997

deLorgeril 1999

Ellingsen 2003

Elwood 1996

Fehily 1993

Gartside 1995, 1998

Hertog 1993

Hu 2000

Jensen 2004

Kromhout 1985, 1996

Kushi 1985

Leren 1989, 1989 — secondary treatment studies among male heart patients

Liu 2003

Mann 1997

Morris 1995

Oomen 2000

Pietinen 1997

Rimm 1996, 1996

Rissanen 2003

Shekelle 1989

Suh 1992

Tsugane 1999

Voutilainen 2001

Yano 1977

Yuan 2001

Another group of studies were secondary preventive treatment studies on cardiac patients or specific high risk cohorts, not primary prevention studies, and were not relevant to these Guidelines:

deLorgeril 1994 — a secondary prevention study, not applicable

Hu 2003 — observational study on diabetics in Nurses’ Health Study database of dietary questionnaires, reported untenable differences in relative risks for CHD associated with highest and lowest consumptions of fish

Lee 2004 — observational study of diabetics in the Iowa Women’s Health Study database looking at vitamin C; reported no associations between CHD mortality and dietary intakes of vitamin C and “vitamin C intake was unrelated to mortality from cardiovascular disease in nondiabetic subjects at baseline”

Mukamal 2004 — looked at coffee consumed by hospitalized male and female heart attack patients; not applicable

Nestel 2005 — observational study of secondary care of cardiac patients in Australia and New Zealand; not applicable

Singh 2002 — clinical trial of East Indian men, with small subset of women, with heart disease, history of heart attacks or angina or at high risk for heart disease, including diabetics; 2 year followup of diet low in fats, saturated fats, cholesterol compared to diet high in fruits and vegetables, nuts, legumes and whole grains; 14 more deaths in low-fat diet group; not primary prevention trial, not applicable

Trichopoulou 2005, 2003 — observational study of male and female cardiac patients in Greece; not applicable

Among the remaining observational studies are a hodgepodge of various cohorts (including both men and women, and some diabetic patients) looking for associations and reporting findings in terms of relative risks. But note that none of these studies reported tenable associations — values beyond what could be due to random chance, confounding factors, or minor statistical errors in their computer model — and would not be considered valid for these types of epidemiological studies. Or they found no links at all. Briefly, these observational studies included:

Arts 2001 — examination of catechin intake using dietary survey data from Iowa Women’s Health Study; reported no overall benefits from catechin foods (fruits, wine, teas); cardiovascular disease mortality risks were no different among different catechin intakes

Bazzano 2001 — using data from national dietary surveys of both men and women, and heart disease as reported on insurance claim data; reported untenable associations between legume intake and cardiovascular disease (relative risks differing only 0.08 among different intakes); no mortality data

Bazzano 2002 — using same dietary survey database, looked for correlations with fruit and vegetable intake; subanalysis found no benefits among women

Bazzano 2003 — using same dietary survey database of both men and women; reported no tenable differences between highest and lowest consumptions of dietary fiber and heart disease events; no mortality data reported even with 19 years of followup

Boniface 2002 — cohort of British men and women; found no association between congestive heart disease and amounts of total or saturated fat in men; untenable associations among women; no mortality data

Burr 1988 — cohort of British men and women; relative risks for all-cause mortality a mere 4% lower among vegetarians as compared to meat eaters

Ellsworth 2001 — using Iowa Women’s Health Study database; found nontenable differences in relative risk for all-cause mortality among those eating the most and fewest nuts (12%)

Erkkila 2004 — using Nurses Health Study database looking at associations between heart disease and calculated phylloquinone intake from dietary questionnaires; found no dose-related relationships

Esprey 1996 — using dietary questionnaire data, found no relationship between total fat or saturated fats and heart-related mortality

Fraser 1992 — using Seventh Day Adventist databank, found 11% lower cardiac deaths associated with highest “wholewheat” bread intakes, 48% lower relative risks between highest and lowest nut consumptions and heart disease deaths, all under tenable relative risk numbers; no correlation found among women between cardiac endpoints and intakes of red meat, fish, cheese, coffee, legumes or fruit

Gaziano 1995 — looking for correlations between carotene intakes estimated from mailed dietary questionnaires on men and women, difference of 5 actual cardiac deaths (0.3%) between highest and lowest consumptions, no overall mortality data

Geleijnse 2002 — used databank of dietary questionnaires gathered from Dutch seniors and diabetics; found a difference of 2 actual fatal heart attacks between highest and lowest tea consumptions at baseline

Gillum 2000 — used dietary survey databank (NHANES 1), found no correlations between fish intake and heart-related deaths

Hu 1997, 1998, 1999, 1999, 1999 — dredges from Nurses’ Health Study database; no tenable relative risk correlations reported between total fats or types of fat, between types of protein (animal or vegetable), nut consumption, or calculated alpha-linolenic acid intakes and CHD; no overall mortality data

Joshipura 2001 — using Nurses’ Health Study and Health Professionals’ databank of dietary questionnaires; no tenable differences in relative risks between highest and lowest intakes of fruits and vegetables and CHD; no overall mortality data

Key 1996, Thorogood 1994 — cohort of “health conscious” vegetarians in Britain in the Oxford Vegetarian Study, reported no tenable associations between highest and lowests intakes of wholegrain breads, bran cereals, nuts or dried fruits, fresh fruits, raw salads and risk factors for heart disease or deaths. [Interestingly, not included in the AHA bibliography was a 2003 follow-up study on this cohort which reported “mortality for major causes of death was not significantly different between vegetarians and nonvegetarians.”]

Klipstein-Grobusch 1999 — cohort of Rotterdam population of men and women, no correlation between consumptions of beta-carotene, vitamin E or vitamin C and heart attacks; no mortality data

Klipstein-Grobusch 1999 — same Rotterdam cohort; no tenable relative risks between highest and lowest intakes of iron and heart attacks; no mortality data

Knekt 1994, 1996, 2002 — used Finnish Mobile Clinic Health Examination Survey database of men and women compiled in 1967-72; reported untenable differences in relative risks between highest and lowest consumptions of fruits and vegetables, carotenoids, vitamin C, vitamin E, and margarine, or all flavonoids (RR .92 and .92) and CHD mortality

Knoops 2006 — used database of elderly Europeans aged 70-90, untenable lower relative risks between Mediterranean diet and mortality, but dietary associations were less than those between smoking or physical activity

Kushi 1996 — post-menopausal women cohort; no tenable relative risks between dietary supplements and CHD deaths; no overall mortality data

Lapidus 1986 — used database of dietary questionnaires on calories, protein, fat, carbohydrates, vitamin C, and fish; reported “none of the dietary factors were associated with stroke or total mortality”

Liu 2000, 2000 — used Women’s Health Study database; reported no tenable relative risks or dose-related risks associated with lowest and highest intakes of fruits and vegetables and cardiac incidents; no mortality data

Liu 1999, 2000, 2000, 2000, 2003 — used Nurses’ Health Study database of dietary questionnaires; found no tenable relative risks between highest and lowest consumptions of whole grains, whole grain breakfast cereals, glycemic load, refined or total carbohydrate intake and CHD, no mortality data

Malavia-rahchi 2002 — used Nova Scotia Nutrition Survey database of men and women; found no difference in total iron and MIs, no mortality data

Mozaf-farian 2003 — men and women cohort; no tenable relative risk between fish consumption and cardiac incidents

Osganian 2003 — data dredge of Nurses Health Study database looking for correlations between calculated intakes of carotenoids and cardiac events; no significant relative risks reported

Osler 2003 — meta-analysis of observational cohorts of men and women; reported no tenable correlations between highest and lowest fish intakes and CHD (RR 1.02 -0.86-1.00-0.93) or all cause mortality

Osler 2002 — cohort of men and women in Copenhagen; reported no tenable differences in CHD events associated with Western food pattern and “prudent” diet of wholegrains, fruits and vegetables

Reunanen 1995 — men and women cohort looking at iron binding capacity; no mortality data

Sesso 2003 — used database of U.S. female health professionals, including diabetics; no associations between highest and lowest consumptions of flavonoids and CHD

Stampfer 2000 — used Nurses Health Study database and found lowest relative risks for CHD among those who exercised, didn’t smoke, drank at least half a drink a day, and were in the highest 40% of women in eating cereal fibers, fish, and low-fat; no adjustments for socioeconomic factors; no mortality data

Steffen 2003 — men and women cohort; no tenable or dose-related correlations between whole grains and fruits and vegetables (RR 1.00-1.10-1.21-1.06-0.82) and cardiac incidents

Vander 2005 — cohort of women and diabetics around Utrecht; no tenable relative risks between highest and lowest consumptions of heme iron and newly diagnosed CHF, no cardiac incidents or mortality data reported

Vander Schouw 2005 — cohort of Dutch women and diabetics; concluded “results do not support the presence of a protective effect of higher intake of phytoestrogens”

Willet 1993 — used Nurses’ Health Study database; no tenable or dose-related differences in relative risks for CHD and highest and lowest consumptions of trans fatty acids

Only one observational study specifically examined the U.S. Dietary Guidelines and the health effects of following the government’s “Healthy Eating Index” developed by the USDA:

McCullough 2000 — used Nurses' Health Study database to specifically evaluate health effects of adherence to the Dietary Guidelines for Americans; reported: “After adjustment for smoking and other risk factors, the [Healthy Eating Index] score was not associated with risk of overall major chronic disease in women....These data suggest that adherence to the 1995 Dietary Guidelines for Americans, as measured by the [Healthy Eating Index] will have limited benefit in preventing major chronic disease in women.”

And there was only one clinical intervention trial in the AHA’s list of studies it cited in support of its Guidelines:

Howard 2006 —Women’s Health Initiative Dietary Modification Trial was a randomized controlled trial of 48,835 postmenopausal women (age most associated with risks for heart disease) of intense behavior modification to reduce fat to 20% of calories, increase fruits and vegetables to 6 or more servings a day, and increase whole grains to 6 or more servings a day; it concluded: “the diet had no significant effects on incidence of CHD, stroke, or CVD.”

Not one observational study was able to credibly support the AHA heart healthy eating recommendations for women to prevent heart disease or premature death. The only observational study specifically looking at Healthy Eating in accordance with our government’s dietary guidelines found no benefit. And finally, the strongest evidence — an actual clinical trial of the heart healthy diet on the primary prevention of heart disease in women, that went on for more than 8 years — found it had no effect on heart disease.

Reviews of clinical trials conducted on heart healthy programs to date have found them of doubtful effectiveness, with no effect on mortality. Our beliefs in healthy eating have gone far beyond well-founded advice to eat normally and enjoy a variety of foods in order to prevent deficiencies, fuel our bodies, and for pleasure; to beliefs in special powers of foods as medicines or poisons.

This review looked at the evidence being used to support “evidence-based” recommendations for a heart healthy diet. When we hear the term “evidence-based,” most of us probably had a very different picture in our minds.

While the AHA calls for rigorous public policies to implement its preventive guidelines population-wide in order to “combat the pandemic of heart disease in women,” how many politicians and healthcare professionals will have taken the time to look at the evidence behind these recommendations? But we will have, and can make a more informed choice about what we want to eat.

© 2007 Sandy Szwarc

* How many grams of alcohol is in a drink? To find out, multiply the number of ounces by the percentage of alcohol in your beverage, then multiply by 0.23. So, 1.5 ounces of 80-proof hard liquor (40% alcohol) has 14 grams of alcohol. A 5 ounce glass of wine has 15 grams. A 12-ounce beer has 11 grams.

In the UK, the alcohol content of drinks is measured in units. One UK unit contains eight grams of pure alcohol. The UK Government has less stringent recommendations, advising men to not regularly drink more than three to four units a day and women not more than two to three (up to 24 grams per day).

Bookmark and Share