This article was first published in 听
While the relationship between diet and health is complicated, there is scientific consensus that a Mediterranean diet can reduce the risk of cardiovascular disease and that excessive intake of ultra-processed food increases it. But 鈥渞educe鈥� and 鈥渋ncrease鈥� are vague terms until numbers are applied to them. Those numbers, though, can be confusing and are often open to misinterpretation. The measure commonly used in nutritional studies to evaluate the effects of some dietary intervention is the 鈥渉azard ratio,鈥� or 鈥淗R.鈥� It is important to understand what this means 鈥� and what it does not mean.
听
Let鈥檚 start with a specific example: Primary Prevention of Cardiovascular Disease with a Mediterranean Diet, published in the New England Journal of Medicine.
The study involved separating 7,447 subjects known to have risk factors for cardiovascular disease into three groups. One group followed a Mediterranean diet with added extra-virgin olive oil, another followed the Mediterranean diet with extra nuts, and the third, the 鈥渃ontrol group鈥� continued to eat in their usual way with some advice to watch their fat intake. A Mediterranean diet involves eating vegetables, fruit, legumes, nuts, fish and olive oil, while reducing sweets and red and processed meats, according to听
After a median followup of 4.8 years, there were 96, 83 and 109 cardiovascular events in the three groups, respectively. This resulted in a hazard ratio calculation of 0.7. Some media reported, 鈥淢editerranean diet cuts heart disease by 30 per cent,鈥� leading people to believe that if 100 individuals with cardiovascular risk factors switched to a Mediterranean diet, 30 would be saved from a heart attack, stroke or death over the next five years. This is not so.
The hazard ratio is an expression of the likelihood that at any moment during the study one group experiences an event compared with another. If the HR is equal to 1, then there is no difference between the groups. If it is less than 1, then the treatment group, in this case the Mediterranean dieters, has a lower event rate, and if greater than 1, the treatment group has a higher event rate. An HR of 0.7 means that at any point during the period of the study the treatment group is experiencing an event rate that is 70 per cent that of the control group. It says nothing about the absolute number of events 鈥� which is what we are really interested in.
Time to crunch the numbers
When we look at events per year, we find that for every 1,000 people following the Mediterranean diet, there were eight events, while the control group experienced 11. Dividing eight by 11 gives us 0.7, the reported hazard ratio. Events are happening in the treatment group at a rate that is 70 per cent of the rate of the control group. But in absolute numbers, only three events are avoided every year by 1,000 people following the Mediterranean diet. Over five years, 15 events would be avoided for every 1,000 people.
听
In other words, if a person with cardiovascular risk factors adopts a Mediterranean diet, the chance of avoiding a heart attack, stroke or death within five years is 15 in 1,000, or 1.5 per cent. That sounds less significant than 鈥渃utting heart disease by 30 per cent,鈥� although technically that is a correct statement. While the potential benefit for an individual is small, the potential benefit for a population can be huge. For every million people adopting a Mediterranean diet, 15,000 events would be saved over five years.
There is yet another factor to consider. To what extent is the calculated HR reliable? That is determined by the 鈥渃onfidence interval.鈥� In the Mediterranean diet study, the HR of 0.7 is reported with a confidence interval of 0.54 to 0.92, meaning that if the study were repeated many times, results could be anywhere between 0.54 and 0.92, so that the Mediterranean diet group could be experiencing an event rate that is anywhere from 46 per cent to eight per cent of the control group. That is quite a range, meaning that the data is not very reliable.
Let鈥檚 now turn to a paper about processed foods published in the British Medical Journal with the title 鈥淯ltra-processed food intake and risk of cardiovascular disease: a prospective cohort study.鈥� In this case, French researchers followed more than 100,000 participants who over five years regularly filled out food surveys designed to evaluate the consumption of 3,300 food items. They reported any incidence of cardiovascular disease. The data was evaluated in terms of the NOVA system, which classifies foods in four categories ranging from unprocessed to ultra-processed and a hazard ratio was determined based on comparing high to low ultra-processed food consumers. An HR for each 10-per-cent increase in ultra-processed food intake was calculated as 1.12, leading to headlines that screamed, 鈥淯ltra-processed foods increase heart-disease risk.鈥� Yes, the study did indeed show that the rate of heart disease was 12 per cent higher in the ultra-processed food consumers.
听
鈥楬igher than what?鈥�
Digging into the data reveals that of 1,000 people consuming a low ultra-processed food diet for a year, about five will suffer a cardiovascular event. The high consumers risk is increased by 12 per cent, meaning an absolute increase of about 0.6 events a year for every 1,000 people. Roughly 2000 people would have to follow a diet low in ultra-processed foods to save one cardiovascular event annually. That sounds a lot less scary than telling consumers that eating ultra-processed foods increases your risk of cardiovascular disease by 12 per cent. Once again, although an individual has a very small chance of benefiting, the result for the population at large could be significant.
The confidence interval reported for this study is 1.05 to 1.20, which is large. Instead of 12 per cent, the difference could be as little as five per cent or as much as 20 per cent. In general, confidence intervals in nutritional studies are large because such studies have many confounders that can skew results. For example, people who consume fewer ultra-processed foods may also be more physically active or may have altered their diet due to a diagnosis of high blood pressure, high cholesterol or diabetes. In nutritional studies that attempt to evaluate some risk, an HR of less than 1.2 usually comes with a wide confidence interval and represents a small absolute risk. Truly causal exposures, like smoking and lung cancer, have HRs in the 10-20 range with small confidence intervals.
I am in no way suggesting that there is no point in following a Mediterranean diet or in cutting back on ultra-processed foods to reduce cardiovascular risk. Even if the data shows that only one in 100 individuals may benefit, that still translates to a huge benefit for the population. Roughly 12 million of the 30 million adults in Canada have at least three risk factors for heart disease, so even at the rate of one in a hundred, we are looking at preventing 120,000 from experiencing a cardiovascular event.
Even though only one in 100 may benefit, you may be that one. So go for the extra virgin olive oil, the nuts, fruits and veggies and cut back on the chips, hot dogs and sodas.
