How do we know if the food we are eating is healthy? Has the American obsession with healthiness turned it into an unhealthy nation? Many of us love math. Why can’t we develop a nice little algorithm that will separate good foods from bad? Part of the problem is that nutrition is complicated. As hard as we try, we can’t find a simple formula that makes our task a simple one. Even so, there are scientists and mathematicians who are working on it. What can we learn from these efforts? In this post, Linn Steward and I look at the concept with a particular emphasis on the Food Compass (1).
Nutrient profiling systems
Rob: I have mixed feelings about these systems. I like them as they base their scores on real data not a mere supposition that may or may not be accurate. The systems take into account both positive and negative characteristics of a food. They judge each fresh item or processed product on its own merits, not as part of a large group of unrelated foods. We can then discuss why one food scores higher than another. Does the difference make sense? Why or why not? Are there underlying characteristics that we don’t understand.
And yet, I am not comfortable with some aspects of nutrient profiling. First, I evaluate meals and diets rather than individual foods. Second, algorithms form the basis for each profile. These mathematical constructs are not always useful. Sometimes the Netflix recommendations unveil a hidden gem. Too often they introduce me to a junk movie that wastes too much of my time. And what are the assumptions that contribute to the score? Are they based on solid science? Or, are they only a convenient way of assessing nutritional healthiness? Third, they may be too harsh on food additives. Criticism of food additives appears to be hype rather than based on actual data. Can we reduce such a complex maze of characteristics down to one number? I am not so sure.
Linn: Nutrient profiling systems are useful. Some are based on weight, others on kcal, still others on serving size. Nutri-Score uses 100 grams. The Food Compass uses 100 kcal. The Nutrition Facts Label uses serving size. My problem with all nutrient profiling systems however is the assumption that healthy patterns can be quantified and reduced to a nutrient count. This is not to say nutrients are not important, that malnutrition and nutrient deficiency diseases aren’t serious public health issues, or that nutrition quantification has no useful purpose.
Rob: The Food Compass makes a massive number of comparisons. I like that it not only stops at individual foods. It moves on to profiling combined popular dishes. The Compass uses the USDA data base as its transparent source. Mumbo-jumbo combinations of hidden equations defy any way to challenge the results. The rationale for its design is that “Once you get beyond ‘eat your veggies, avoid soda,’ the public is pretty confused about how to identify healthier choices.” By using the USDA database, I envision development of a mechanism to profile every meal we eat. Such a device would be like learning the nutritional composition of our diets using select programs. Processed foods get a fair shake with Food Compass unlike NOVA. Not bad.
Yet, are nutrients the sole determinants of healthiness? I am not so sure. My background in food science equates nutrients with healthiness. But nutrition is more than presence or absence of nutrients. The whole is greater than the sum of its parts. For example, certain nutrients enhance the absorption of other nutrients. Absence of one of these nutrients compromises effectiveness of the other. Some nutrients antagonize other nutrients. I doubt the scores allow for such enhancement or antagonism. Then there is a case of eating healthy foods/meals over and over again. Too much of a good thing might miss adverse health consequences. The algorithm could take these flaws into account. I also don’t like the negative treatment of popular food additives. Except for nitrates/nitrites, scientific data do not support negative scores for additives.
Linn: I do not like that ruminant meat (beef, lamb, goat) is penalized. I’m all for reducing the amount of meat on the plate and I’m all for more plants of every kind. However, I’m not okay with labeling freshly cut meat, which NOVA considers minimally processed, unhealthy. And no distinction is made between fortification and intrinsic micronutrients for calculation purposes. There is a structural reason for this oversight. Neither the FNDDS database or the USDA legacy database are coded to discriminate between intrinsic and fortified nutrients.
I am pleased that some additives are tagged and result in a negative point loss in the Food Compass. The additives include Added Sugars, Nitrite, artificial sweeteners, partially hydrogenated oils. Plans are underway to obtain data on additives which are not currently tracked in FNDDS. Examples are artificial flavors/colors, corn syrup, partially hydrogenated or inter-esterified oils. Total fat is common in other nutrient profiling systems and still appears in the recent Dietary Guidelines, but The Food Compass does not assess for total fat. This reflects more recent science on the health effects of fat. Saturated fat is also not assessed in isolation but as part of overall fat quality and assessed using the fatty acid ratio. The algorithm is flexible. Single products, market baskets, restaurant meals, or a fast-food menu boards can all be scored and compared. The database architecture enables updates based on evolving scientific literature.
Can these profiling systems be improved?
Rob: I envision tweaks that could take my objections into account. The algorithm could take on these flaws. Then we would have an algorithm that is no longer transparent. What information are they hiding? How do we know the basis for the assumption(s) that account for such a difference? Now we are back to Netflix and its hit-or-miss approach. The important thing to understand, these systems represent a nutrient profile. They do not represent nutritional value. Do they really describe healthiness as they claim? I would leave them alone. BTW, Linn, FDA eliminated the incorporation of partially-hydrogenated oils in 2018. If they were so dangerous, we should have seen a massive decrease in deaths by trans-fat. Banning them was not the magic bullet we were promised.
Linn: There should always be room for improvement built into the architecture of a nutrient profiling system. The current FDA Nutrition Facts Label was based on data current in the 1980s. Database architecture and nutrition science have both evolved over the last 40 years. The regulations however move remained set in regulatory cement. It can take years if not decades for FDA to make changes.
Rob: Long ago, Linn disabused my notion that nutrient profiles represent healthiness. Nutrients matter more than Linn suggests. Industrial formulations are healthier than she gives them credit for. If sugar and salt are bad in processed foods, they are also bad in homemade ones. I don’t buy the idea that we consume less sugar and salt in homemade foods than in processed ones. If taste and flavor are the criteria, then we will eat more homemade foods . Why? Because they taste so much better than the store-bought versions. Or so we are told!
Linn: Nutrients are in my opinion only one contributing factor to a healthy pattern. My biggest concern with all nutrient profiling systems is that nutrients overshadow other important components like regional preferences, seasonality, and personalized taste – all of which can in my opinion also contribute to a healthy eating pattern.
To be continued
Nutri-Score assigns a letter grade from A to E. Linn likes that system better than the Food Compass. We need to study it more in-depth before we discuss that one. Stay tuned.
Next week: What happens to a fruit or vegetable after separation from the plant?
* According to Food Compass, the tuna salad with light mayonnaise is 73 and the chicken breast is 61. Raw raspberries score 100; fat-free, ready-to eat vanilla pudding scores 1.
Mozaffarian, D., N.N. El-Abbadi, M. O’Hearn, J. Erndt-Marino, W.A. Masters, P. Jacques, P. Shi, J.B. Blumberg, and R. Micha, 2021. Food Compass is a nutrient profiling system using expanded characteristics for assessing healthfulness of foods. Nature Food 2:809-818. https://doi.org/10.1038/s43016-21-00381-y
4 thoughts on “Analyzing Food Compass and other nutrient-profiling schemes ”
Good move on your part to add intrinsic and fortified to your reading focus. Effectiveness is one important aspect. So is the reason for fortification in the first place. Looking forward to another go around.
My themes for the next few months are value chains for fresh fruits and vegetables in May, world hunger in in June, insects and other alternative proteins in July, and vitamins in August. There will be chances for further discussion.
Congratulations to us both. Together I think we did a credible job. Just one little clarification on my contribution. I’m okay with the Food Compass. And I admire NutriScore. Both can add value as FOP; both are incomplete. It’s our American version “Facts up Front” that makes me crazy. Given a choice between Facts up Front and NutriScore, I’ll take NutriScore any day. 👍🍾🥂
I am pleased and glad that you are pleased. I am not as up-to-date as you are on Facts Up Front. I will need to study both that and Nutri-Score, before I can make a judgement on either. What drives me up the wall is the Stoplight front of package label which is way too NOVA like for my tastes! Let’s try to have another discussion on such label designations later this year. My textbook on Vitamins is coming next week. We need to set some parameters before I start reading it. I plan to add the difference between the effectiveness of “intrinsic and fortified nutrients.”