Challenges in Nutrient Profiling

The Science of Nutrient Profiling

Nutrient profiling is defined as the science of categorizing foods according to their nutritional composition. Historically, foods have been categorized based on how they’re grown or produced, creating just five overbroad food groups: vegetables, fruit, grains, dairy and protein foods.

Simply considering the food group a food belongs to doesn’t tell us very much about its nutritional value. For example, the nutritional composition of a potato is very different from kale, yet these both belong to the vegetable food group. While they’re both nutritionally valuable whole foods, they aren’t interchangeable on our plates nor do they benefit our health comparably. As another example, the nutritional composition of lentils is very different from salmon, yet these both belong to the protein food group. Again, both are nutritionally valuable foods: per serving, they provide similar amounts of protein, copper, biotin, and vitamin B1; but lentils provide a large amount of fiber and have much more folate and manganese whereas salmon has much more vitamin B2, vitamin B3, vitamin B6, vitamin B12, selenium, potassium, and zinc while being one of our best sources of omega-3 fatty acids.

In order to understand how to choose foods with complimentary nutrition in order to meet our nutritional needs, we must necessarily begin with a more sophisticated method to understand the nutritional value of different foods. And, that’s what nutrient profiling is all about.

Previous Nutrient Profiling Methods

Nutrient profiling research began in the early 2000s with the development of several similar methods to quantify the nutritional value of foods, including (but not limited to): Nutrient for Calorie (NFC), Calorie for Nutrient (CFN), Nutritious Food Index (NFI), Naturally Nutrient Rich (NNR) Score, Nutrient-Rich Foods Index (NRF), Nutrient Adequacy Score (NAS), and Nutrient Density Score (NDS). These scores/indices differ from each other in small but meaningful ways.

Let’s examine these differences in more detail, to understand the algorithmic decisions that were made for the Nutrivore Score, and why none of these previous nutrient profiling methods have been broadly adopted for everyday use.

The Problem with Capping a Nutrient’s Contribution

Some of these scores (like the NDS) cap a nutrient’s contribution at 100% DV, whereas others recognize that a food having more than a 100% DV of a specific nutrient per serving makes that food a very valuable source of that nutrient (and we won’t always eat that food in large enough quantities to actually get 100% DV of whatever nutrient it’s particularly rich in)!

There’s no such thing as a nutritionally complete food, so the goal is nutrient sufficiency of the entire diet, not of any one particular food. Thus, it’s important to understand how to combine foods that are good sources of different nutrients in order to achieve dietary nutrient sufficiency. Capping a nutrient’s contribution to the nutrient density score at 100% hinders our ability to understand the importance of food combining for nutrient sufficiency.

For example, Brazil nuts are the most nutrient-dense nut attributable to their very high selenium content—a 1-ounce serving delivers nearly 1000% DV of selenium! But, if you capped the contribution of selenium to the calculation of Brazil nut nutrient density, they would erroneously appear to be one of the lowest nutrient-density nuts. It just makes sense to fully value the awesome selenium content of Brazil nuts and have that reflected in its nutrient density score!

The Most Useful Denominator: Calories

Some scores calculate nutrient density as a function of food weight, which is influenced by nonnutritive compounds like water content, but most represent the nutrient density as a function of energy. Given that our food supply is overabundant in calories while being overall depleted in nutrients, the more important information for the average consumer is how to maximize nutrients for each calorie, rather than for each gram or serving of food.

The Problem with Weighting Scores

Some of the scores are normalized (so for example, the range is 0 to 100 or 0 to 1000), whereas others are simply totals. The problem with normalization is communication: while it may seem simpler to represent all foods on a scale from 0 to 100, the immense difference between the nutrient density of vegetables versus, say, fast food, is minimized when the scale is smaller.

Some scores also include a weighting by food group, so the most nutrient-dense grain gets allocated the score of 100 in the grains group and the most nutrient-dense vegetable gets allocated the score of 100 in the vegetables group. This food group weighting system completely undermines the entire concept of nutrient profiling. For example, the Nutrivore Score of kale is 4233 whereas the Nutrivore Score of oats is 208, but when you apply a correction for food groups, their nutrient density scores are very similar. Studies show their health benefits are not equivalent, why would be normalize their nutrient-density scores to make them appear to be equally nutrient-dense foods when they clearly aren’t?

The Problem with Penalizing for Certain Nutrients

Some scores penalize for the presence of nutrients whose excess consumption have been linked to health problems, like sodium, saturated fat, added sugars, and cholesterol.

Dietary cholesterol does not increase serum lipids for most people and is the backbone of vitamin D and other steroid hormones. Dietary saturated fat is only problematic when intake exceeds about 15% of total calories (although this does depend on genetic predisposition). Sodium is only problematic when intake exceeds about 7 grams per day, and there are studies showing that even this level is only worrisome when potassium intake is concomitantly low. Added sugars become problematic above about 10% of total calories (and 25% of total carbohydrates). Most importantly, all of these nutrients are healthy in moderate amounts, and only unhealthy when the whole diet includes excess.

All of these potentially problematic nutrients are abundant in fast food, junk food and other hyperpalatable manufactured foods that are also low in essential nutrients and very high in calories—this is why scores that penalize for these nutrients give these types of foods scores less than zero! Penalizing individual foods for the presence of these nutrients does not reflect the diet as a whole, and has the capacity to undervalue otherwise nutrient-dense whole foods which can fit into a health-promoting diet, while not being necessary to show that fast food and junk food are unhealthy.

The Problem with Emphasizing Certain Nutrients

Another topic for discussion is whether to weight certain nutrients more in nutrient profiling. It makes sense on the surface to count those nutrients for which a larger proportion of the population are deficient more than those nutrients for which very few people are deficient.

But, here’s the challenge with that logic: the potential for overcorrection and simply shifting towards different common nutrient deficiencies. If those foods that are particularly good sources of the nutrients that, for example, an estimated 70% or more of Americans routinely don’t consume enough of (vitamin B9, vitamin D, vitamin E, choline, calcium, potassium, zinc, omega-3 fats, and polyphenols) had inflated scores as a result of weighting these nutrients more heavily in nutrient profiling, food sources of other nutrients end up being undervalued.

This approach has the capacity to shift food choices in a way that could help address some population level nutrient deficiencies over the short term, but that’s not the same thing as moving towards nutrient sufficiency, especially over the long term.

Selecting Nutrients to Include in the Calculation

The other way these scores differ is in the nutrients used to make the calculations. Most include protein and fiber, but the CFN only includes protein and the NFI only includes fiber. Which vitamins are utilized in the calculations ranges from only vitamin C (in the NDS5) all the way to including all of the vitamins except vitamin B7 (biotin) and vitamin K (in the NDS15). And which minerals are utilized ranges from just calcium and iron (in the NQI, NRF6, and NDS5) all the way to calcium, iron, zinc, magnesium, copper, iodine, and selenium (in the NDS23).

For the NRF and NDS, multiple variations were created, incorporating anywhere from 5 to 23 nutrients into their calculations (hence the number after the acronym, for example NDS5 or NRF23) and either with or without penalizing foods for their sodium, added sugars and saturated fat content (in which case, the acronym has a “.3” added at the end, for example the NRF15.3 includes 15 nutrients in its calculation and penalizes for 3). None of these scores incorporate phytonutrients into their calculations.

The Problem with Retrofitting to Dietary Guidelines

There has been much discussion among scientists over which nutrients to include in nutrient profiling calculations. This concern originates from the fact that certain nutrients are more strongly correlated with health outcomes than others (typically those that we’re most likely to be deficient in rather than the nutrient itself being less important), such as high consumption of omega-3 fats, fiber and vitamin D.

It makes vastly more sense to devise a nutrient profiling method that simply reflects the nutritive value of a food, and then to study how eating more nutrient-dense foods impacts disease risk.

To attempt to hone in on the best sampling of nutrients to include in a nutrient-density score, a couple of studies have compared the NRF calculated with 5 to 16 nutrients (with or without limits for sodium, sugars and saturated fats, and with our without weighting for various food groups) to the Healthy Eating Index (HEI). The HEI is designed a way to quantify compliance with the USDA dietary guidelines, by assigning an energy adjusted score for servings from 9 food groups or nutrients to encourage (total fruits, whole fruits, total vegetables, greens and beans, whole grains, dairy, total protein foods, seafood and plant protein, and fatty acids ratio) and subtracting servings from 4 food groups or nutrients to discourage (refined grains, sodium, added sugars, and saturated fat).

Interestingly, these studies found that an NRF with fewer nutrients in the calculation better aligned with the HEI, with 9 nutrients (protein, fiber, vitamin A, vitamin C, vitamin E, calcium, iron, magnesium and potassium) being optimal.

But, here’s where this line of reasoning is fundamentally flawed. Why would we retrofit a nutrient density score to align with dietary guidelines that were crafted without nutrient density or nutrient sufficiency in mind? This especially makes no sense when you consider that there is also little understanding of how individual nutrient-dense foods fit into healthful dietary patterns. Given how incredibly prevalent nutrient deficiencies are, even when people follow the USDA dietary guidelines, and how these deficiencies increase risk of chronic and infectious disease.

It makes vastly more sense to devise a nutrient profiling method that simply reflects the nutritive value of a food, and then to study how eating more nutrient-dense foods impacts disease risk. In fact, there was a 2104 study of people over the age of 55 that showed that the higher the NRF9 of their diet as a whole, the lower their risk of all-cause mortality—the highest NRF9 quartile had a 16% lower chance of dying than the lowest NRF9 quartile. This study helps to prove that potentially huge health benefit of a diet replete with nutrient-dense foods, but more studies like this that incorporate even more nutrients into the calculus are necessary to advance this field of research.  

To advance the public’s understanding of what constitutes a nutrient-dense food, nutrient profiling must necessarily be algorithmically independent from the Healthy Eating Index and USDA dietary guidelines. Only then can we use nutrient profiling to improve dietary guidelines.