In episode 342 of Sigma Nutrition Radio, we discussed the evidence relating to vegan diets and health, with emphasis on available intervention studies. However, there are a number of additional studies which are frequently cited as evidence demonstrating that plant-based diets may reverse coronary heart disease (CHD). This article is intended to provide a complementary discussion of three specific studies which were not discussed at length in episode 342.
In this article the focus will be on three studies specifically:
- The Lifestyle Heart Trial (i.e. “The Ornish Study”)
- “The Esselstyn Study”
- Mount Abu Open Heart Trial
In discussing these studies, specific reference will be made to the type of diet used by reference to the some specific definitions that are important to use to distinguish between plant-based diets.
In the first instance, it is necessary to provide some clarity on terminology with regard to diet, as often the most general term of “plant-based” is used as a descriptor, which does not by definition necessarily mean exclusion of all animal produce (although it is frequently implied in this regard). Additionally, within the umbrella term of a “plant-based diet” there has been particular attention placed recently on ‘whole-food plant-based’ (WFPB) diets.
For the purposes of this article, the following definitions will be used:
- Vegan: Diet exclusive of all animal produce. No specific prescriptions otherwise in relation to fat intake, carbohydrate intake, or otherwise.
- Whole-food plant-based: Diet with specific recommendations for total fat (generally <10%, range of 7-15%) and exclusion of all animal produce. Only whole-food form is consumed, i.e., olive oil would be excluded, but whole olives consumed in lieu.
- Vegetarian: Excludes animal meats (flesh), but may include certain animal products such as dairy, eggs, or fish, depending on the specific subtype of the diet (e.g. lacto-vegetarian, lacto-ovo vegetarian, pescatarian, etc.)
- Low-fat vegetarian: This is with specific reference to the Ornish et al. studies referenced below, which were similar to WFPB diet in macronutrient prescription (~10% fat, 70-75% carbohydrate from complex sources, 10-15% protein), but allowed egg whites and non-fat milk or yogurt.
1. The Ornish Study
The intervention began with a one-week residential retreat to a hotel to teach the lifestyle intervention. The intervention encompassed individually prescribed aerobic exercise levels based on target heart rates, and training to use rate of perceived exertion (RPE), with a weekly target of three hours total exercise and a minimum of 30 minutes in a target heart rate zone. Stress management techniques included stretching, progressive relaxation, breathing techniques, meditation, and imagery. Participants in the intervention group attended twice-weekly group support sessions, each four hours in duration, to boost adherence.
The diet was a low-fat vegetarian diet, which included fruits, vegetables, grains, legumes, and soybean products, egg whites and 250ml/d of non-fat milk or yoghurt (no other animal products). The diet contained ~10% fat, ~15-20% protein, and ~70-75% predominantly complex carbohydrates, cholesterol intake was limited to 5mg/day or less. No restrictions were placed on total energy intake. Caffeine was eliminated, and alcohol limited to no more than 2 units per day (no one was encouraged to drink). Vitamin B12 was supplemented.
Twenty participants completed the intervention and another nineteen were in the control group. Diet was assessed with a three-day diet diary at baseline and at the end of the one-year intervention. The only data presented in relation to diet following the intervention period at one year was for total fat, cholesterol, and total energy intake:
- Total fat intake decreased from 67.4 g/d to 14 g/d (14g = 6.8% of total calories)
- Dietary cholesterol decreased from 213 mg/d to 12.4 mg/d
- Total energy decreased from 1959 kcal/d to 1816 kcal/d
Exercise increased from 11 minutes to 38 minutes per day. While stress reduction techniques went from about five minutes per day to 82 minutes per day. Smoking cessation occurred in the whole study group.
Notable results in the intervention group at one year include:
- Bodyweight decreased from 91.1kg to 81.0kg
- Total cholesterol decreased from 5.88mmol/L (227mg/dL) to 4.45mmol/L (171mg/dL)
- LDL-C decreased from 3.92mmol/L (151mg/dL) to 2.46mmol/L (95mg/dL)
- Average diameter in arterial stenosis [which is a narrowing of the aortic valve that restricts blood flow from the heart] regressed from 40.0% to 37.8%.
- Angina frequency decreased from 5.10 to 0.45 (on a 1-7 rating scale)
- Chest pain duration decreased from 2.73 to 1.58
- Chest pain severity decreased from 2.3 to 1.7
There were some substantial differences between intervention and control group at baseline with key risk factors, in particular weight, which was 91.1kg vs. 80.4kg. This is relevant insofar as the magnitude of weight loss over the course of the intervention may have had a significant influence on the results, particularly for outcomes like angina, having regard to the modest regression of arterial stenosis.
The initial paper stated that participants in the control group “were not asked to make lifestyle changes, although they were free to do so.” Perhaps this explains why at the end of the one-year intervention, 8 of the 19 participants in the control group exhibited average changes in arterial lesion scores in the direction of regression. This is stated in the paper, but a limitation of the initial one-year publication was that results were only presented as two-sided p-values without confidence intervals, which was flagged in response to the publication. In the absence of confidence intervals, it is difficult to see beyond the mean differences per se.
Side bar: Confidence Intervals
A confidence interval (CI) is a range used of values that help describe the reliability of research data. A narrow range of values shows high precision and credibility, whereas a wide interval would suggest the opposite. CIs are calculated at a confidence percentage level, with 95% used most often.
Example: Consider a reported relative risk of 1.31 (i.e. 31% increased risk). And in this hypothetical case, let’s say the CI is between 1.27 and 1.35 (or an increased risk of 27-35%). This range is where there is 95% certainty that the true relative risk would lie.
A limitation of the trial methodology was the method of randomisation, , which ultimately confined it to a small sample size. It was stated as a “randomised invitational design”, meaning that participants were randomised before being formally invited to participate in the study.
Fifty-three had originally been randomised to the intervention group and 40 to the controls, however, ultimately only 20 participants completed the intervention and were available for five-year follow-up angiograms (15 from the control group were available for the five-year follow-up). Half of the losses were refusals to participate after randomisation, resulting in a potential for selection bias, and a small sample size - although the relevance of this is difficult to ascertain, as no power calculations were presented in either the one-year or five-year papers.
In relation to selection bias, gaining consent after randomisation means that the investigators in effect implicitly left participation to people who may have been highly motivated and not necessarily representative of the wider population, thus negating a key advantage of randomisation.
With regard to the five-year follow-up, the data indicated continued high adherence to the overall lifestyle protocol:
- Exercise averaged 4.34 times per week for 3.56hrs
- 4.93 stress management sessions averaging 48.53mins per day
- 17g/d of total dietary fat (8.5% energy), 18.6mg/d dietary cholesterol, and mean energy intake of 1845kcal/d.
Of the 10.1kg of weight lost after one-year (91.1kg to 81kg), at five-years follow-up 5.8kg of weight loss was maintained (85.3kg bodyweight). With regard to the primary outcome of arterial stenosis, at five-years mean stenosis diameter had regressed by 3.1% compared to baseline (a further 0.9-1.2% from the one-year follow-up).
Ultimately, in both the one-year and five-year follow-up, the main finding was a linear relationship between adherence to the lifestyle program and changes in diameter stenosis (Figures below).
Adherence relative to change in arterial stenosis from the original one-year intervention (top image) and the five-year follow-up (bottom image).
Adherence was measured as a continuous score which encompassed intake specifically in:
- fat and cholesterol
- frequency and duration of exercise
- frequency and duration of stress management
- smoking cessation
Other than fat and cholesterol, no additional data is presented on diet for either actual food intake or quantification of important variables, in particular dietary fibre. The only statement in the 5-year follow-up is as follows:
"The percentage of daily energy (calories) provided by fruits, vegetables, whole grains, soy, other legumes, nonfat dairy, and alcohol was comparable at 1 year and at 5 years.”
It is therefore difficult to deduce exactly the extent of the dietary modifications made over the course of the follow-up period, foods consumed, and any other nutrient-related factors that could correlate with the outcome. Consequently, adherence in the study was assessed as a composite of all of the above variables combined. No multivariate linear regression analysis was conducted to determine the effects of these several independent variables for their associations with the outcome variable of arterial stenosis.
As can be seen from the graphs above, highest adherence to the overall lifestyle changes displayed a dose-response, with the greatest percentage change in arterial stenosis correlating with the highest degree of adherence to the lifestyle intervention. This was evident at both one-year and five-years follow-up.
Therefore, the most that may be stated in relation to the findings of the Ornish et al. study is that regression of arterial stenosis clearly tied the highest levels of adherence to the total lifestyle program, not merely dietary adherence. Yet, this study is often cited as evidence for an independent effect of a very-low fat vegetarian diet on reversing heart disease, which it does not demonstrate from the data.
The most encouraging finding of this study has always been the effects of an intensive lifestyle intervention on heart disease risk without using lipid-lowering medications. However, the study does not demonstrate any independent effect of diet, but this is often the context in which it continues to be cited. More particularly from a scientific standpoint, to date the intervention has never been replicated. As such, it is a single study, 30 years from the date of initial publication, with 20 completed participants.
In the intervening period, the evidence for the use of high-intensity statin regimes and PCSK9-inhibitors has indicated that reductions in arterial stenosis of >50% are possible with reductions of LDL-C to ~1.5mmol/L (60mg/dL), which is arguably unlikely through dietary intervention.
2. The Esselstyn Study
Similar to the study by Dr. Ornish, the study by Dr. Caldwell Esselstyn conducted from his physical practice in Cleveland, Ohio, is reified as evidence for the effects of a low-fat, plant-based diet reversing coronary heart disease. In this study, the diet was a low-fat vegetarian diet identical in prescription to the dietary recommendations used in the Dr. Ornish study, and included 22 patients with angiography-documented coronary artery disease. The primary hypothesis was that total cholesterol levels <3.88 mmol/L (<150 mg/dL) could arrest or reverse the progression of coronary artery disease, measured by angiography.
The study was an uncontrolled longitudinal investigation recruiting patients from the Cleveland Clinic, with some patients requesting to join the study. Twenty-two patients began the study, and the interventions involved dietary change to a diet with less than 10% calories from fat, combined with individualised cholesterol-lowering drugs. In particular, patients were commonly prescribed a combination of 4g cholestyramine (a bile acid sequestrant which inhibits intestinal cholesterol uptake) taken twice daily together with 40-60mg lovastatin. To promote dietary adherence, each participant was seen bi-weekly for five-years and then monthly for the following five-years.
The initial publication in 1995 reported on both the five-year and nine-year follow-up periods. A 12yr follow-up was published in 1999, which is discussed, below. Eleven participants completed the initial five-years follow-up, 5 participants dropped out within the first 2 years, while 6 participants did not have data collected but reportedly continued the diet.
In 25 arterial lesions identified by angiography, 11 lesions showed regression while 14 lesions remained unchanged. The mean average percent stenosis reduction was 7% (95% CI 3.3-10.7). Mean minimal lumen diameter (MLD; the diameter length of a lesion in the artery) increased by 0.08mm, indicating no significant difference in disease progression for MLD.
In all 11 participants at five-years follow-up:
- 5 displayed evidence of arterial stenosis regression
- 4 displayed progression
- 1 remained stable
- 1 both regressed and progressed
In relation to blood cholesterol, mean total cholesterol levels in the 11 participants was 6.36 mmol/L (246 mg/dL), which decreased to a mean of 3.42 mmol/L (132 mg/dL) over the duration of follow-up, with a range of 2.84mmol/L (109.9 mg/dL) to 3.88mmol/L (149.9 mg/dL). For LDL-C, baseline measures were only available for 5 of 11 participants, thus distorting the baseline mean (which for 5 participants was 4.4mmol/L [173mg/dL]) Mean LDL-C was 1.8mmol/L (71mg/dL), with a range of 1.2mmol/L (48.5mg/dL) to 2.3mmol/L (90mg/dL), over the 5.5yr follow-up.
At nine-years follow-up, five participants continued to maintain the diet, and two had managed to maintain total cholesterol to <3.88mmol/L (150mg/dL), while one maintained levels <4.53mmol/L (175mg/dL). None of these five participants reported new cardiac events, while the five participants who discontinued diet (one was lost to follow-up) all reported new cardiac events.
With regard to the 12-year follow-up, it is difficult to ascertain the veracity of the data. The 12-year publication states:
"The remaining 18 patients adhered to the study diet and medication for 5 years….During the 7 years since the conclusion of the 5-year study, all but 1 patient have continued to adhere to the prescribed diet and medication."
This would mean there is data on 17 compliant patients who adhered to both diet and medication. Recall the numbers in the previous studies of adherent patients:
- At five years: 11 participants completed follow up (6 reportedly maintained diet but did not complete data collection).
- At nine years: 5 participants (of original 11) maintained the diet.
- At twelve years: 17 participants maintained diet and have lipid data available.
Even if we assume the six participants who reportedly continued diet, but did not have follow-up data collections, did continue, then this would only make 11 participants (i.e. the 5 who continued diet at nine-years follow-up, plus the reported 6 who maintained the diet). This is difficult to assume, given that the 6 patients who reportedly maintained the diet did not complete the data-collection, which makes it difficult to ascertain how they had lipid data available at 12 years follow-up.
These numerical discrepancies reflect the difficulty with an uncontrolled longitudinal investigation. There are three fundamental issues with this research as it relates to diet:
- Lack of control group
- Lack of dietary assessment
- The drug regimen
The absence of a control group, which could have for example continued with the same drug regimen but no dietary change, renders it impossible to state that diet was responsible for the outcomes.
The lack of dietary assessment means that we have no data at all on total energy intake, macronutrient intake, and intakes of potential mediating nutrients, nor could any analysis be undertaken to elucidate the respective contributions of diet vs. drugs on the outcomes of interest.
Finally, the importance of the drug therapies in this context cannot be understated. Ranges of 20-80mg lovastatin, the statin therapy used by Esselstyn et al., have been shown to reduce progression of atherosclerosis and formation of new lesions. In the Monitored Atherosclerosis Regression (MARS) Study, lovastatin resulted in a decrease of total cholesterol averaged over 2.2 years from 5.97mmol/L (231mg/dL) to 4.03mmol/L (156mg/dL), LDL-C decreased from 3.91mmol/L (151mg/dL) to 2.41mmol/L (93mg/dL), and mean percent diameter stenosis decreased by 4.1%. These results are relatively comparable to those attained over five-years in the Esselstyn et al. study. Moreover, the MARS study used lovastatin monotherapy as the intervention. The addition of 4-8g cholestyramine, the dosage used in by Esselstyn et al., to lovastatin has been shown to reduce total cholesterol and LDL-C by 43% and 61%, respectively, further to reductions of 36% and 45% from lovastatin alone. Thus, the drug regimen alone is sufficient to explain the reductions in blood lipids and effects on arterial stenosis observed in the Esselstyn et al. study. The absence of a control group, or any dietary data presented and analysed, precludes any other conclusion.
Ultimately, an uncontrolled longitudinal study, where all subjects were on individualised cholesterol-lowering medications with sufficient potential to explain the results, can not be taken to show that diet has any causal role in the outcome.
3. Mount Abu Open Heart Trial
This study is the final study we will consider for oft-cited research purporting to support regression of coronary heart disease through diet. The Mount Abu Open Heart Trial was conducted in Mount Abu, India, with patients recruited with angiography-documented coronary artery disease. Angiography was repeated for up to 2 years follow-up, while cardiac events were monitored for up to six-years follow-up. The study was unblinded and had no control group.
Similar to the Ornish et al. study, this was a highly intensive lifestyle intervention program encompassing a range of psychosocial interventions and dietary modifications. The diet was a low-fat vegetarian diet, with:
- ~15% energy from fat (predominantly monounsaturated and polyunsaturated fats)
- <50 mg/d dietary cholesterol
- ~50 g/d soluble fibre
- ~65% complex carbohydrates
- 1 g/kg bodyweight of protein (from soybean and plant sources)
Daily 30-45 minute support sessions were conducted at local Brahma Kumaris Rajyoga centres, and the study contained a number of unique aspects to the intervention, in particular the emphasis on principles of self-awareness, self-empowerment, and spiritual focus. While no patients were on statin therapy, they were taking a range of cardiovascular medications, including nitrates, anti-platelet agents, beta blockers, calcium channel blockers, and ACE inhibitors.
Adherence was the primary determinant of the outcomes, and patients were divided into high, medium, and least adherence. While the mean percent diameter stenosis decreased by 6.1% after two years, results differed by adherence level:
- Those with highest adherence exhibited an 18.23% regression
- Those with medium adherence resulted in an 11.85% regression
- Those with the least adherence showed progression of 10.56%
So 63% of overall patients showed trends toward regression, with 90.9% of high adherence participants displaying trends toward regression. The investigators conducted univariate and multivariate analysis to elucidate the determinants of arterial stenosis regression. Univariate analysis (examining each independent variable as a single variable influencing the outcome) indicated that regression was statistically significant for decreased body weight, decreased triglycerides, diet adherence, exercise adherence, and meditation. In multivariate analysis (examining two or more variables for their relationship with the outcome), however, only psychosocial stress management through Rajyoga was found to be an independent predictor of regression.
The take-home message for the Mount Abu Trial is similar to the Ornish et al. study insofar as it demonstrates that intensive, multi-disciplinary practice interventions may arrest or regress arterial stenosis in the absence of lipid-lowering medications.
However, it is also similar insofar as independent effects of diet were not demonstrated, in fact the variable independently associated with stenosis regression was psychosocial stress interventions. Although this intervention did increase adherence to other aspects of the overall lifestyle program, it again indicates that the total, overall lifestyle intervention is required. It should be noted that the magnitude of effect in the Mount Abu Trial was significantly greater than the Dr. Ornish Study, but with no control group again, we are limited in the inferences of the evidence.
Concluding Remarks: Evidence and Enthusiasm
- Ultimately, on the basis of the scientific evidence available, it cannot be said that either a WFPB diet or low-fat vegetarian diet reverses heart disease.
- None of the commonly cited studies examined above demonstrated independent effects of diet alone, and only the Ornish et al. study had a control group.
- The quality of the evidence in support of these dietary interventions for regression of heart disease is disconnected from the enthusiasm with which these studies are referenced.
- Beyond our current understanding of dietary patterns and specific factors related to diet, like replacement effects of dietary fats (which you can read more on here), it is currently not possible to say that highly specific dietary interventions are superior. To reach this point, better quality trials with appropriate control groups, and better dietary assessment and analysis, are required.
- Nevertheless, the available evidence demonstrates arrest and reversal of atheroma is possible through achieved levels of LDL-C <1.8mmol/L (70mg/dL) from pharmacotherapy.