Food Technology Magazine | Issues and Insights

Trust the Process: A Food Scientist’s Perspective on Nova

This controversial food classification system raises reasonable questions about ultra-processed foods, but we need to think critically about what it fails to factor in.

By Matt Teegarden
Photo Illustration by Chris cummings

Photo Illustration by Chris Cummings; Images from stock.adobe.com, includes images generated by AI

Over the course of my lifetime, the proportion of the world’s adult population that has obesity or is overweight has almost doubled. It is honestly staggering to think that this increase happened in such a short time. (OK, I am not that young. I am in my mid-30s, but that’s a large change in 30-plus years.) The increases are even more dramatic for children and teenagers, and the statistics for diet-related diseases, such as cardiovascular disease and diabetes, are equally grim.

For years, processed foods have been a primary focus when assigning blame for these concerning health trends. Food scientists like me are quick to point out that essentially all foods are processed to some degree, so blanket statements such as this are, scientifically speaking, not accurate. But the term “processed food” has never really been just about processing as a food scientist would define it. Instead, this term is loaded with implications about formulation, industrialized production, nutritional content, and marketing tactics to name a few. This fuzziness precluded a lot of systematic research on the role of processed foods in health until 2009, when the Nova (sometimes represented as NOVA, even though it’s not an acronym) food classification system was published.

Boy shopping in a grocery store

The Nova food classification system attempts to distinguish foods based on the extent and purpose of processing. © kali9/E+/Getty Images

Boy shopping in a grocery store

The Nova food classification system attempts to distinguish foods based on the extent and purpose of processing. © kali9/E+/Getty Images

In its current form, Nova attempts to distinguish foods based on the extent and purpose of processing, ranging from minimally processed (1) to ultra-processed (4). This simple scale has facilitated an explosion of research that associates ultra-processed food (UPF) intake with diet-related chronic diseases. Some scientists advocate that the data are convincing enough to justify urgent public health policies limiting UPF consumption. In fact, Brazil has already updated its dietary guidelines to reflect the Nova scale, and the U.S. 2025 Dietary Guidelines Advisory Committee considered the topic of UPFs in its latest literature review.

Widespread adoption of Nova would have a huge impact on the global food system, so it is important to understand its strengths and weaknesses from multiple points of view. My intent with this article is to discuss some of the major points of Nova and offer a food scientist’s interpretation of them.

UPFs and Health Risks

The evidence supporting Nova thus far is mostly observational, meaning epidemiologists have established that there are associations between higher consumption of UPFs and elevated risks of negative health outcomes. New publications are constantly being added to an already expansive body of literature, so I won’t attempt a comprehensive review here. I will offer the main points from a 2024 umbrella review as a salient example of what the research shows (Lane et al. 2024). This study, which systematically summarized research on 45 conditions across 14 meta-analyses, found that the risks for most of the health conditions analyzed were increased with exposure to UPFs. The authors also concluded that the associations with cardiometabolic diseases, common mental disorders, and overall mortality were particularly convincing based on the available data.

While these findings can be concerning, they need to be interpreted carefully. Nutritional epidemiology is inherently challenging for many reasons. It’s difficult to accurately measure what people eat, and the relationship between dietary patterns and health is riddled with confounding variables. Scientists fastidiously apply mathematical corrections to account for these things, but it is impossible to correct for everything. One study attempted to demonstrate residual confounding by measuring the association between UPFs and accidental death (Morales-Berstein et al. 2023). They found a positive association, so be careful out there.

We have studies that compare apples to orange juice, but we also need to compare apples to applesauce.

Nova also brings its own unique challenges to epidemiology, which have recently been summarized by Lauren O’Connor and colleagues (O’Connor et al. 2024). The main issue here is that most, if not all, of the data that epidemiologists rely on do not directly measure the Nova categories of foods that people eat. The Nova scale must be retroactively applied and processing levels are inferred based on limited information. I do not bring this up to question the legitimacy of this very challenging work, and I am not trying to say that the associations between UPFs and disease are not real or worth investigating further. However, I do think that these are major limitations that should be considered in discussions around UPFs and health. We also need to remember that association does not equal causation. Direct experimental evidence from human clinical trials is needed to validate these findings.

By now, many are familiar with the clinical study done by Kevin Hall’s group at the National Institutes of Health (NIH) (Hall et al. 2019). In this seminal study, participants lived in the NIH clinic and were fed diets containing either UPFs or unprocessed foods for two weeks at a time. The key finding was that, on average, study participants consumed more calories, ate faster, and gained weight when consuming UPFs. This was recently somewhat replicated in a clinical study out of Japan (Hamano et al. 2024). In that study, participants seemed to gain weight no matter which food they were given (Hey, free food, right?), but people still tended to eat more calories, eat faster, and gain more weight on the UPF diet. These studies are interpreted by some as proving the key hypothesis posed by Nova, that the processing levels of foods, and dietary patterns by extension, dictate their healthfulness.

While I agree these results provide important evidence on UPFs, I think they missed the mark on proving this crucial point. The menus in these studies were designed to differ in overall processing level while keeping macronutrient values and calories reasonably constant. In other words, the UPF menus were designed to provide most if not all calories from UPFs and vice versa for the minimally processed menus. However, the menus did not control for the types of foods that were offered to participants. Take, for example, the dinners from Day 1 of the Japanese study. Those consuming the UPF diet were given a Kentucky Fried Chicken platter, while the non-UPF meal was grilled pork with fresh vegetables, rice, and walnuts. The meals clearly differ in their level of Nova processing, but the foods that make up these meals are also completely different. This makes it hard to conclude if it was the processing level or something else about the foods that led to negative outcomes.

garbanzo beans

Where garbanzo beans are categorized on the Nova scale depends on whether they are dried, canned with salt, or canned with salt and EDTA. ©FreshSplash/E+/Getty Images

garbanzo beans

Where garbanzo beans are categorized on the Nova scale depends on whether they are dried, canned with salt, or canned with salt and EDTA. ©FreshSplash/E+/Getty Images

Research Limitations

The current studies also fail to appreciate the variety of foods that fall within the Nova categories, with a tendency to test the most “extreme” versions of foods within each processing level. Evidence-based dietary guidelines already recommend limiting the consumption of many foods that would fall into Nova group 4 and would be considered UPFs—things like fast food, salty fried chips, and sugar-sweetened soda. But the logic of Nova also allows for similar foods to be classified across different processing levels. Garbanzo beans, for example, could be Nova group 1, 3, or likely even group 4, depending on whether they are dried, canned with salt, or canned with salt and EDTA. Mass-produced whole grain breads are considered group 4, while whole grain bread purchased from a local bakery would be group 3. A cake made at home, completely from scratch, could potentially be classified as group 3, while a cake made from a boxed mix would definitely be classified as group 4. Meanwhile, lard and olive oil would both be classified as group 2, and thus considered similarly healthy. I bring these examples up to demonstrate that Nova is often at odds with current nutritional science that is built upon decades of evidence. The UPF category is also extremely broad and contains many foods that would conventionally be considered part of a healthy diet. The intricacies of Nova need to be thoroughly tested to understand whether it is actually a useful tool to improve nutrition.

The clinical research we have suggests that people eating high UPF diets tend to eat excess calories, and it is reasonable to wonder why.

To put this all another way, we have studies that compare apples to orange juice, but we also need to compare apples to applesauce. Julie Hess at the U.S. Department of Agriculture Grand Forks Human Nutrition Research Center has developed test menus that make this comparison (Hess et al. 2024). These menus were designed to differ in overall processing level, while carefully controlling not only for nutrient content but also for food types. As an example, the main dish for the more processed lunch on Day 2 is a breaded chicken patty on a white hamburger bun with Miracle Whip and romaine lettuce. Meanwhile, the less processed lunch on this day has a similar sandwich made with chicken strips on a homemade hamburger bun with mayonnaise and iceberg lettuce. This menu tests some of the nuances of Nova that seem antithetical to current evidence-based dietary guidelines. Of course, human studies still need to be done, but Hess’s work also makes a few important, theoretical points: It is possible to have a diet, composed of mostly UPFs, that is of high nutritional quality, according to current nutritional science (Hess et al. 2023). Conversely, it is also possible to have a relatively unhealthy diet composed of mostly unprocessed or minimally processed foods.

Shopping in a local bakery.

Whole grain bread’s Nova classification depends on whether it was mass produced or purchased from a local bakery. © alvarez/E+/Getty Images

Shopping in a local bakery.

Whole grain bread’s Nova classification depends on whether it was mass produced or purchased from a local bakery. © alvarez/E+/Getty Images

Nova Lacks Precision

If I sound critical of Nova, it’s because I am. The food and nutrition communities are moving toward an exciting future where dietary recommendations are more precise, but Nova works against this with its lack of precision. Nova categories are based on an odd combination of formulation and processing. Both of these factors can certainly impact the healthfulness of foods, but they need to be considered as two independent variables. However, it’s important to acknowledge that Nova goes beyond how a food scientist may think about processing; it also considers the ways that processing may be used to create widespread availability of foods that are rich in calories but poor in nutritional value. It is additionally worried about making these types of food so convenient and enticing that people opt to eat them instead of more nutritionally balanced foods. While these concerns aren’t necessarily unfair to a point, it’s also likely wrong to say that they apply to all UPFs because, again, the Nova definition of UPFs is entirely too broad. Interestingly, epidemiologists have started separating UPFs into subcategories in their analyses—things like breads and cold cereals, sugar-sweetened beverages, and savory snacks. (See Mendoza et al. 2024 as an example.) When this is done, only certain categories of UPFs are associated with disease outcomes, while others appear to be benign at worst.

Do the shortcomings of Nova mean that we shouldn’t be researching UPFs? Not necessarily. They are a huge part of our food system, after all, and there are some real questions that need to be answered. The clinical research we have suggests that people eating high UPF diets tend to eat excess calories, and it is reasonable to wonder why. Ciaran Forde at Wageningen University studies ingestive behaviors, and his group recently published results from a trial where participants were fed minimally processed foods and UPFs that varied in hardness (Teo et al. 2022). They found that participants ate more of the soft foods, regardless of processing level. At the same time, the processing level did affect the total calories consumed because the UPFs in this study were more calorically dense than the minimally processed foods.

If we can understand which specific attributes of UPFs may have negative health impacts, we can work toward tailored solutions that preserve the universal benefits of these products.

There are other theories that should also be carefully tested, but the key question for Nova will be if they apply exclusively and uniformly to all UPFs. Certain concerns, like the increased bioaccessibilty of carbohydrates due to the breakdown of plant cell walls, may apply to groups 2, 3, and 4, and therefore be a result of processing overall, rather than ultra-processing per se. Many of the potential issues with UPFs, whether they are exclusive to the category or not, could also be addressed with new advances in food science and technology, as recently reviewed by Julian McClements (McClements 2024). For instance, optimizing the microscopic structure of foods may help improve how macronutrients are released during digestion, but it is unclear if these solutions would be acceptable under Nova guidelines.

It is also important to consider how results from UPF research will translate into real life. There are many factors that influence what and how people eat, like their knowledge of cooking, available time to prepare meals, hedonic preferences, and ability to purchase food, to name a few. Removal of UPFs from the diet without also addressing these things may lead to null results or other unintended consequences. If we can understand which specific attributes of UPFs may have negative health impacts, we can work toward tailored solutions that preserve the universal benefits of these products, like convenience, abundance, and shelf stability.

Burger and Fries.

Evidence-based dietary guidelines recommend limiting the consumption of many foods that would be included in Nova group 4. © rzoze19/iStock/Getty Images Plus

Burger and Fries.

Evidence-based dietary guidelines recommend limiting the consumption of many foods that would be included in Nova group 4. © rzoze19/iStock/Getty Images Plus

Necessary Nuance

It would be difficult to look at global health statistics and believe that absolutely nothing about our food system needs to change. And while research on so-called UPFs could be an important part of this very complex puzzle, I am just not convinced that the definition offered by Nova is the answer. The “one-size-fits-all” approach that Nova takes to define UPFs ignores the nuance that needs to be understood. There are too many factors that influence how a food can impact health; it is impractical to think that this can be captured in a four-point scale. We need a better, science-based system to advance research that will improve health (Trumbo et al. 2024).

Taking a step back, I think the issues with Nova illustrate the overall need for more interdisciplinary research in food and nutrition. For too long, scientists have worked in their own discipline-aligned silos on something that affects literally everyone on the planet. (Everybody eats!) We need to build bridges that allow us to capitalize on our collective expertise, and I think food scientists can play a key role in this. After all, we are trained across several disciplines to provide the world with safe, affordable, delicious, and nutritious foods. There is a huge incentive for scientists to be working toward a common goal here. We just need the right tools that enable us to do the work.ft

The opinions expressed in this article are those of the author.

About the Author

Matt Teegarden, PhD, is a food scientist who has worked in academia and industry ([email protected]).