STATS ARTICLES 2010
Mediterranean diet and heart disease
Rebecca Goldin Ph.D and & Jing Peng, July 7, 2010
Benefits of moderate alcohol consumption go unreported, but failure to note collinearity leads news orgs to misstate study’s findings.
Diet is critical for heart patients, and a recent study in the June issue of the American Journal of Clinical Nutrition gave a statistical fillip to the widely discussed benefits of a Mediterranean diet.
The Technical Details
The study involved a thousand heart patients in Athens, Greece, and attempted to correlate survival rates for after heart attacks with a Mediterranean diet, among other measurements of heart health. From May 2006 to March 2009, 1,000 consecutive patients with acute coronary syndrome (ACS) were enrolled in the study. Of 1,000 patients, 788 were men with an average age of 63 and 212 women with an average age of 69. Patients were diagnosed with ACS if they had had a heart attack or repeated heart attacks or unstable angina (heart palpitations) leading to hospitalization in a particular clinic in Athens, Greece.
The purpose of this study was to estimate the link between adherence to the Mediterranean diet and and several variables. The first variable is the development of left ventricular systolic dysfunction (LVSD) at hospitalization. The second one is left ventricular remodeling over a three-month period. The last variable is death or rehospitalization due to another cardiovascular event in the short period of 30 days and the longer period of time of two years.
Adherence to the Mediterranean diet was assessed by the validated Mediterranean Diet Score, whose range is 0-55. Researchers conducted the dietary evaluation after the third day of hospitalization. All patients were asked repeatable and validated food-frequency questions about 75 foods and beverages commonly consumed in a Greek diet. In the study the diet score was calculated as follows:
“For consumptions of items presumed to be close to the Mediterranean pattern (unrefined cereals and products, fruit and nuts, vegetables, olive oil, nonfat or low-fat dairy, fish, potatoes, and pulp), a score of 0 was assigned when a patient reported no consumption, 1 when reported consumption was 1-4times/mo, 2 for 5-8 times/mo, 3 for 9-12times/mo, 4 for 13-18 times/mo, and 5 for >18 times/mo.
In contrast, for consumption of foods presumed not to be consumed on a daily or weekly basis in a Mediterranean diet (such as meat or meat products, eggs, poultry, and dairy), reverse scores were assigned. Regarding alcohol intake, a score of 5 was assigned for consumption of <3 glasses/d; 0 for none or consumption >7 glasses/d; and scores of 4, 3, 2 and 1 for the consumption of 3, 4-5, 6 and 7 glasses/d, respectively. Higher values of this diet score indicate greater adherence to the Mediterranean diet (theoretical range: 0-55).”
The results of the study showed a clear correlation between better heart health and scoring high on the Mediterranean diet. None of those scoring high on the diet (above 21 points) died in the hospital, while 2 percent of those who were low (less than 17) did. In addition, those scoring in the highest range had 31 percent lower risk of another heart attack in the first month, though this was not statistically significant (p = 0.09) and 47 percent and 37 percent lower risk of cardiovascular events during the one and two-year follow up, which was highly statistically significant (p< 0.001). However, we should be cautious about causality: physical activity was much higher among those with a high score (76 percent of patients exercised) compared to the lowest scoring group (50 percent exercised). This last correlation was also highly significant (p<0.001).
The researchers also looked into the benefits of individual components of the diet. As seen in the scoring in the Mediterranean diet, alcohol is considered beneficial at the level of 1-7 glasses per day, and most beneficial at 1-2 glasses per day. However, the results were robust and the benefit of the Mediterranean diet was maintained even with the alcohol score removed.
News Coverage: the contrast between what was reported in the press and in the scientific article
Reuters news account had some trouble reporting the findings accurately. Of the Mediterranean diet, they noted that “This eating pattern, which includes lots of fruits and vegetables, nuts, vegetable oils, low-fat dairy products, legumes, whole grains, and fish, has been shown to help shield people from heart disease.”
Yet toward the end of their article, Reuters stated, “when researchers looked at different components of the Mediterranean diet separately, they found that vegetables and salad and nuts were the only foods that cut risk.”
However, this simplification does not speak to the whole study. While individual foods may correlate with the observed benefit, they are highly correlated with other foods in the diet, making it difficult to separate the individual from the whole. Controlling for weight, people who eat a lot of nuts are less likely to eat a lot of meat.
Indeed, a diet containing the favorable characteristics of the Mediterranean diet helps with reducing mortality and illness. What is more newsworthy was the finding that Mediterranean-style diets were associated with a 47 and 37 percent lower risk of heart disease over one year and two-year follow-up, compared with a 31 percent lower risk over the first month follow-up.
Reuters, however, reported that “people who ate vegetables and salad or nuts daily or weekly were at 20 percent lower risk of repeat heart problems within two years of their initial hospitalization compared to people who are these foods monthly or less often”.
But why would the only foods that offer a benefit offer one that is a 20 percent reduction in heart attack rates two years hence, but the overall results be more like a 40 percent reduction? The answer lies in the cumulative benefit of the diet taken as a whole. Individual items may not reach statistical significance, but combined with others they may. The authors of the article also note that people who eat salad, for example, are also more likely to eat legumes. This is called “collinearity” when individual variables that are inputs to your statistical analysis are highly dependent on one another.
Reuters also missed the impact of alcohol. If anything, the study result encourages people who want to be heart-healthy to consume a small amount of alcohol, such as 1-2 glasses per day. As a whole, media accounts underestimated the impact of the Mediterranean diet on cardiovascular events, and poorly described the diet as a whole.