Ornish Diet Worsens Heart Disease Risk: Part I

Dr. Dean Ornish has come under a lot of criticism lately for his misleading statements about diet and heart disease. See, for example: Critique of Dean Ornish Op-ed, by Nina Teicholz, and Why Almost Everything Dean Ornish Says about Nutrition Is Wrong, from Scientific American.

Ornish made his name with a study that claimed to actually reverse heart disease [1]. There are at least three problems with the study.

First, it included several confounders to the dietary regimen. For example, the intervention groups spent an hour a day on stress management techniques, such as meditation, and three hours a week exercising.

Second, although it was touted as the first study to look at "actual" heart disease results, it made no measurements of cardiac events! Instead, it was based on measuring stenosis — the degree of narrowing of coronary arteries. Considering that stenosis is only a predictor of cardiac events, it seems disingenuous to call it a direct measure of heart disease.

Stenosis is used to predict heart disease (though it is often not the previously found blockages that are ultimate culprits [2]). However, the measurement has a lot of variability. Because of this, differences in measurements over time need to be quite large to be showing a true progression or regression, and not just error. We found three studies attempting to pinpoint the minimum difference in measurements to make such a claim. They respectively recommended 15%, 9.3%, and 7.8% as a basis for this judgment [3], [4], [5].

So how much reduction of stenosis was there in Ornish's study?

"The average percentage diameter stenosis decreased from 40.0 (SD 16.9)% to 37.8 (16.5)% in the experimental group yet progressed from 42.7 (15.5)% to 46.11 (18.5)% in the control group (p = 0.001, two-tailed)."

That's the extent of the success in a year: a -2.2% change for the claim of "regression" vs. a 3.4% change for the claim of "progression". It does not reach a level of significance given the measurement tool.

Fortunately, there were other measurements taken that are also predictors of cardiac events: blood lipids. Even the AHA, an association that changes its mind slowly in response to evidence, considers triglycerides above 100 to be higher than optimal [6]. Low HDL is a strong marker of heart disease, with HDL below 40 considered by the AHA a "major heart disease risk factor" [7]. The intervention group went from an average triglyceride level of 211 to 258, and their HDL from 39 to 38. This shows that the intervention actually worsened the participants' risk factors!

Moreover, although not acknowledged by the AHA, we know that the ratio of triglycerides to HDL is a very strong predictor of heart disease; among the best [8]. A triglyceride-to-HDL level of less than 2 is considered ideal. Over 4 is considered risky. Over 6 is considered very high risk. The intervention group's average triglycerides-to-HDL ratio leapt from 5.4 to 6.8! It went from bad to worse. Thus, the third problem with the study is that it actually showed a worsening of heart disease by other important measures.

The bottom line is that Ornish's study never showed what it claimed to show.

After a year of intervention, even with other lifestyle changes incorporated, the subjects on his diet had a higher risk of heart disease than before they started.


[1]Ornish, Dean, et al. "Can lifestyle changes reverse coronary heart disease?: The Lifestyle Heart Trial." The Lancet 336.8708 (1990): 129-133.

Eveidence type: experiment

Little WC, Constantinescu M, Applegate RJ, Kutcher MA, Burrows MT, Kahl FR, Santamore WP.
Circulation. 1988 Nov;78(5 Pt 1):1157-66.


To help determine if coronary angiography can predict the site of a future coronary occlusion that will produce a myocardial infarction, the coronary angiograms of 42 consecutive patients who had undergone coronary angiography both before and up to a month after suffering an acute myocardial infarction were evaluated. Twenty-nine patients had a newly occluded coronary artery. Twenty-five of these 29 patients had at least one artery with a greater than 50% stenosis on the initial angiogram. However, in 19 of 29 (66%) patients, the artery that subsequently occluded had less than a 50% stenosis on the first angiogram, and in 28 of 29 (97%), the stenosis was less than 70%. In every patient, at least some irregularity of the coronary wall was present on the first angiogram at the site of the subsequent coronary obstruction. In only 10 of the 29 (34%) did the infarction occur due to occlusion of the artery that previously contained the most severe stenosis. Furthermore, no correlation existed between the severity of the initial coronary stenosis and the time from the first catheterization until the infarction (r2 = 0.0005, p = NS). These data suggest that assessment of the angiographic severity of coronary stenosis may be inadequate to accurately predict the time or location of a subsequent coronary occlusion that will produce a myocardial infarction.


Evidence type: experiment



Clinical trials with angiographic end points have been used to assess whether interventions influence the evolution of coronary atherosclerosis because sample size requirements are much smaller than for trials with hard clinical end points. Further studies of the variability of the computer-assisted quantitative measurement techniques used in such studies would be useful to establish better standardized criteria for defining significant change.


In 21 patients who had two arteriograms 3-189 days apart, we assessed the reproducibility of repeat quantitative measurements of 54 target lesions under four conditions: 1) same film, same frame; 2) same film, different frame; 3) same view from films obtained within 1 month; and 4) same view from films 1-6 months apart. Quantitative measurements of 2,544 stenoses were also compared with an experienced radiologist's interpretation. The standard deviation of repeat measurements of minimum diameter from the same frame was very low (0.088 mm) but increased to 0.141 mm for measurements from different frames. It did not increase further for films within 1 month but increased to 0.197 mm for films 1-6 months apart. Diameter stenosis measurements were somewhat more variable. Measurement variability for minimum diameter was independent of vessel size and stenosis severity. Experienced radiologists did not systematically overestimate or underestimate lesion severity except for mild overestimation (mean 3.3%) for stenoses > or = 70%. However, the variability between visual and quantitative measurements was two to three times higher than the variability of paired quantitative measurements from the same frame.


Changes of 0.4 mm or more for minimum diameter and 15% or more for stenosis diameter (e.g., 30-45%), measured quantitatively, are recommended as criteria to define progression and regression. Approaches to data analysis for coronary arteriographic trials are discussed.


Evidence type: experiment

Brown BG1, Hillger LA, Lewis C, Zhao XQ, Sacco D, Bisson B, Fisher L.
Circulation. 1993 Mar;87(3 Suppl):II66-73.



Imaging trials using arteriography have been shown to be effective alternatives to clinical end point studies of atherosclerotic vascular disease progression and the effect of therapy on it. However, lack of consensus on what end point measures constitute meaningful change presents a problem for quantitative coronary arteriographic (QCA) approaches. Furthermore, standardized approaches to QCA studies have yet to be established. To address these issues, two different arteriographic approaches were compared in a clinical trial, and the degree of concordance between disease change measured by these two approaches and clinical outcomes was assessed.


In the Familial Atherosclerosis Treatment Study (FATS) of three different lipid-lowering strategies in 120 patients, disease progression/regression was assessed by two arteriographic approaches: QCA and a semiquantitative visual approach (SQ-VIS). Lesions classified with SQ-VIS as "not," "possibly," or "definitely" changed were measured by QCA to change by 10% stenosis in 0.3%, 11%, and 81% of cases, respectively. The "best" measured value for distinguishing definite from no change was identified as 9.3% stenosis by logistic regression analysis. The primary outcome analysis of the FATS trial, using a continuous variable estimate of percent stenosis change, gave almost the same favorable result whether by QCA or SQ-VIS.


The excellent agreement between these two fundamentally different methods of disease change assessment and the concordance between disease change and clinical outcomes greatly strengthens confidence both in these measurement techniques and in the overall findings of the study. These observations have important implications for the design of clinical trials with arteriographic end points.


Evidence type: experiment

Gibson CM1, Sandor T, Stone PH, Pasternak RC, Rosner B, Sacks FM.
Am J Cardiol. 1992 May 15;69(16):1286-90.


The purpose of this study was (1) to determine a threshold for categorizing individual coronary lesions as either significantly progressing or regressing, (2) to determine whether multiple lesions within individual patients progress at independent rates, and (3) to calculate sample sizes for atherosclerosis regression trials. Seventeen patients with 46 significant lesions (2.7 lesions/patient) underwent repeat coronary arteriography 3.0 years apart. With use of the standard error of the mean change in diameter from initial to repeat catheterization across 5 pairs of consecutive end-diastolic frames, individual lesions were categorized as either significantly (p less than 0.01) progressing or regressing if there was a 0.27 mm change in minimum diameter or a 7.8 percent point change in percent stenosis. The mean diameter change of a sample of lesions can also be analyzed as a continuous variable using either the lesions or the patient as the primary unit of analysis. A lesion-specific analysis can be accomplished using a multiple regression model that accounts for the intraclass correlation (rho) in the degree of change among multiple lesions within individual patients. The intraclass correlations in percent stenosis (rho = 0.01) and minimum diameter (rho = -0.24) were low, indicating that disease progression in different lesions within individual patients is nearly independent. With use of this model, 50 patients per treatment group would permit the detection of a 5.5% difference between treatment group means in the change in minimum diameter and a 2.7% percentage point (not percent) difference in the change in percent stenosis.(ABSTRACT TRUNCATED AT 250 WORDS)


From The American Heart Association's "Scientific Statement"

"New clinical recommendations include reducing the optimal triglyceride level from <150 mg/dL to <100 mg/dL, and performing non-fasting triglyceride testing as an initial screen."


From Levels of Cholesterol

Less than 40 mg/dL for men; less than 50 mg/dL for women: Major heart disease risk factor

60 mg/dL or higher Gives some protection against heart disease


Evidence type: observational

Gaziano JM1, Hennekens CH, O'Donnell CJ, Breslow JL, Buring JE.
Circulation. 1997 Oct 21;96(8):2520-5.



Recent data suggest that triglyceride-rich lipoproteins may play a role in atherogenesis. However, whether triglycerides, as a marker for these lipoproteins, represent an independent risk factor for coronary heart disease remains unclear, despite extensive research. Several methodological issues have limited the interpretability of the existing data.


We examined the interrelationships of fasting triglycerides, other lipid parameters, and nonlipid risk factors with risk of myocardial infarction among 340 cases and an equal number of age-, sex-, and community-matched control subjects. Cases were men or women of <76 years of age with no prior history of coronary disease who were discharged from one of six Boston area hospitals with the diagnosis of a confirmed myocardial infarction. In crude analyses, we observed a significant association of elevated fasting triglycerides with risk of myocardial infarction (relative risk [RR] in the highest compared with the lowest quartile=6.8; 95% confidence interval [CI]=3.8 to 12.1; P for trend <.001). Results were not materially altered after control for nonlipid coronary risk factors. As expected, the relationship was attenuated after adjustment for HDL but remained statistically significant (RR in the highest quartile=2.7; 95% confidence interval [CI]=1.4 to 5.5; P for trend=.016). Furthermore, the ratio of triglycerides to HDL was a strong predictor of myocardial infarction (RR in the highest compared with the lowest quartile=16.0; 95% CI=7.7 to 33.1; P for trend <.001).


Our data indicate that fasting triglycerides, as a marker for triglyceride-rich lipoproteins, may provide valuable information about the atherogenic potential of the lipoprotein profile, particularly when considered in context of HDL levels.

1 comment:

  1. http://ses.library.usyd.edu.au/bitstream/2123/11945/2/Bell_KJ_thesis_2.pdf

    What do you make of the idea of the Food Insulin Index? This is the wiki entry which is concise and helpful:

    "The Insulin Index is a measure used to quantify the typical insulin response to various foods. The index is similar to the Glycemic Index and Glycemic Load, but rather than relying on blood glucose levels, the Insulin Index is based upon blood insulin levels. This measure can be more useful than either the Glycemic Index or the Glycemic Load because certain foods (e.g., lean meats and proteins) cause an insulin response despite there being no carbohydrates present, and some foods cause a disproportionate insulin response relative to their carbohydrate load.

    Holt et al. have noted that the glucose and insulin scores of most foods are highly correlated,[1] but high-protein foods and bakery products that are rich in fat and refined carbohydrates "elicit insulin responses that were disproportionately higher than their glycemic responses." They also conclude that insulin indices may be useful for dietary management and avoidance of non-insulin-dependent diabetes mellitus and hyperlipidemia."

    I have also read that glucose levels can appear normal whilst blood insulin is running high......


Note: Only a member of this blog may post a comment.