Challenges in defining and classifying UPFs

The NOVA system demonstrates particular strength in identifying UPFs in Latin America, where it was developed and the packaged food supply is more uniform. However, countries with more diverse packaged food supplies face challenges in distinguishing nutritional quality within the UPF category^{(Reference Dicken, Batterham and Brown21)}. This challenge comes especially apparent with products that differ in processing and nutritional composition, particularly as food manufacturers develop nutritionally improved options that still rely on industrial formulations and processes.

The classification of breakfast cereals provides a compelling illustration of this challenge (Figure 2). Consider an Australian fruity loop breakfast cereal that contains 60 % whole grain ingredients (maize flour, wheat flour, oat flour), supplemented with sugar, vegetable oil, salt, and natural colours derived from paprika extract, carmine, and curcumin, along with essential micronutrients. Despite its substantial whole grain content and legitimate nutritional benefits from the fortified micronutrients, this product is classified as a UPF primarily due to its industrial processing methods. This classification becomes even more problematic when examining products like wheat bran breakfast cereal, containing 85 % wheat bran with minimal additions of sugar, barley malt extract, and fortificants. Despite its minimal processing and high whole grain content, extrusion technology (exclusive to the food industry and involving forcing, shaping, and cooking under pressure) alone classifies it as a UPF, potentially misleading consumers about its nutritional value.

Figure 2. UPF-characterising attributes, as defined by Monteiro et al. ^{(Reference Monteiro, Steele and Cannon22)}, of selected items. *cosmetic additives.

The limitations extend to seemingly simple products like commercial popcorn, where arbitrary distinctions emerge based on minor ingredient differences rather than meaningful nutritional variations. Two variants of popcorn - one containing only corn, oil, and salt (NOVA-3) and another adding natural flavours and colours (NOVA-4) - receive different classifications despite nearly identical nutritional profiles. The classification shifts based solely on the presence of nature-identical flavours and colours, which are chemically identical to compounds found in natural foods but are somehow considered as ‘cosmetic’ additives that define UPF.

The carbonated beverage category illustrates the classification challenges inherent in the NOVA system. All carbonated beverages receive a blanket NOVA-4 classification regardless of their ingredients, creating potential inconsistencies in how we assess food processing complexity. The widespread availability of home carbonation technology has fundamentally altered the traditional boundaries between commercial and domestic food processing. A consumer can now produce carbonated lemonade at home using methods and ingredients virtually identical to commercial production: lemon juice, sugar, water, and carbonation. The primary distinction lies in commercial products sometimes containing additional preservatives, artificial colours, or artificial flavours. This technological accessibility creates a significant classification paradox. Two beverages with substantially similar primary ingredients and processing methods receive different NOVA classifications based solely on their production location rather than their fundamental processing complexity or nutritional profile. A home-carbonated lemonade made with natural ingredients receives a lower NOVA classification than an identical commercially produced version, despite comparable processing methods and potentially similar health implications. While carbonated beverages may carry distinct health implications compared to their non-carbonated counterparts, the current classification approach fails to account for the convergence of home and commercial food processing capabilities, potentially undermining the system’s utility as a tool for assessing food processing complexity.

These classification inconsistencies become particularly problematic when applied across different cultural contexts, revealing the Western-centric limitations of the NOVA system. In parts of Asia, for example, canned meat products—explicitly categorised by Monteiro et al. ^{(Reference Monteiro, Steele and Cannon22)} as NOVA-3 foods—often contain monosodium glutamate, a ‘cosmetic’ additive that defines UPFs under the NOVA system criteria. This raises the question of whether such foods should be considered UPF simply due to the presence of an additive, despite their widespread consumption and cultural significance.

Tofu and gluten expose contradictions in the NOVA system, revealing its failure to account for traditional food preparation methods. Tofu, a nutritionally rich protein central to many Asian cuisines for generations, has been labelled a UPF by some food scientists^{(Reference Estévez, Arjona and Sánchez-Terrón23)}. This classification disregards cultural and dietary significance, reducing a time-honoured staple to a generic processed product.

Similarly, gluten further illustrates the system’s inconsistencies. Monteiro et al. ^{(Reference Monteiro, Steele and Cannon22)} categorise it as an ‘ingredient of no culinary use’ despite its long-standing role as a key protein source in various cultures^{(Reference Barnes24)} without the health concerns that are now linked to UPFs. Gluten could be classified as a NOVA-2 food, as it can be produced simply by washing wheat dough to remove starches, although it could also be considered a NOVA-3 food, given that NOVA-2 foods are not for direct consumption. However, under the NOVA system, it is considered harmless in traditional preparations but is suspected when included in industrial formulations. This contradiction raises a fundamental question: Is the issue gluten itself or how the NOVA system assesses industrial processing?

These cultural inconsistencies highlight a more fundamental issue: the binary nature of the NOVA system. This approach overlooks the nuanced processing spectrum and its relationship to nutritional quality. Products with varying levels of whole food ingredients, different processing methods, and diverse nutritional profiles are grouped under the UPF category, potentially oversimplifying complex nutritional relationships^{(Reference Dicken, Batterham and Brown21,Reference Barrett, Gaines and Coyle25)} . This limitation becomes particularly significant when considering fortified products or those reformulated to meet specific nutritional guidelines while retaining their UPF classification.

Heterogeneous health associations across UPF subtypes

The relationship between UPF consumption and health outcomes presents a complex landscape characterised by significant heterogeneity across different food categories and contexts. While numerous studies have documented associations between UPF consumption and adverse health outcomes^{(Reference Lane, Gamage and Du26–Reference Wang and Sun29)}, careful examination reveals important nuances that challenge the assumption of uniform health effects across all UPFs.

Vitale et al. ^{(Reference Vitale, Costabile and Testa28)} conducted a meta-analysis of prospective cohort studies, consistently finding higher risks of type 2 diabetes mellitus (T2DM) (37 %), hypertension (32 %), hypertriglyceridaemia (47 %), low HDL cholesterol (43 %), and obesity (32 %) linked to high UPF intake. However, the review emphasised that these risks shifted significantly depending on how UPF consumption was measured, whether through food frequency questionnaires (FFQs) or 24-hour recalls, resulting in risk estimates that differed by more than 50 % between methods. While this variation reflects the broader, well-established limitations of dietary assessment methods^{(Reference Bailey30,Reference Bingham31)} , including the constraints of single 24-hour recalls and the broad-spectrum nature of FFQs that affect all dietary intake measurements, it also highlights specific challenges in UPF assessment. Traditional dietary assessment tools were not originally designed to capture the processing characteristics that define UPF classification, creating additional complexity beyond standard dietary measurement limitations^{(Reference O’Connor, Herrick and Papier32)}.

Wang et al. ^{(Reference Wang and Sun29)} explored the interplay between UPFs, cardiometabolic health, and diet quality, pointing out that not all UPFs have equal health impacts. Some, like fibre-fortified cereals or plant-based milks, may offer nutritional benefits, while others, such as sugary drinks and processed meats, are linked to negative outcomes. They reported increased risks of CVD (17 %), T2DM (15 % per 10 % rise in UPF consumption), and hypertension (23 %) associated with higher UPF intake. The authors stressed that future research should go beyond classifying foods solely by their processing levels, integrating measures of overall diet quality, metabolomics, and gut microbiome influences to understand better how UPFs contribute to health risks.

Two landmark studies published in 2024 have provided comprehensive analyses of UPF consumption in relation to major chronic diseases, revealing striking patterns that challenge simplistic categorisations. Mendoza et al. ^{(Reference Mendoza, Smith-Warner and Rossato33)} analysed data from three large U.S. prospective cohorts – the Nurses’ Health Study (NHS), Nurses’ Health Study II (NHSII), and Health Professionals Follow-Up Study – comprising over 206 000 participants followed for nearly three decades. While overall UPF consumption was associated with increased CVD risk (HR: 1·11, 95 % CI: 1·06, 1·16), individual UPF categories exhibited strikingly divergent patterns. SSBs and processed meats consistently showed adverse associations with CVD outcomes, with hazard ratios reaching 1·19, 1·25 for these specific subtypes. In stark contrast, bread products, cold cereals, yoghurt/dairy desserts, and certain savoury snacks demonstrated inverse associations with CVD risk, with hazard ratios as low as 0·87, 0·96.

Similarly, Dicken et al. ^{(Reference Dicken, Dahm and Ibsen34)} examined data from the European Prospective Investigation into Cancer and Nutrition (EPIC), involving 311 892 individuals followed for approximately 10·9 years. Their investigation of T2DM revealed that despite an overall 17 % higher T2DM incidence per 10 % increment in total UPF intake, specific UPF categories showed opposite associations. Savoury snacks, animal-based UPF products, ready-to-eat mixed dishes, and sweetened beverages increased T2DM risk (HR range: 1·16–2·77), while bread products, biscuits, breakfast cereals, and plant-based alternatives were associated with reduced risk (HR range: 0·46–0·89).

However, these observational findings must be interpreted cautiously given the potential for residual confounding that characterises epidemiological research^{(Reference Fewell, Davey Smith and Sterne35–Reference Feng37)}. The complexity of dietary patterns, lifestyle factors, and socioeconomic variables makes it challenging to isolate the specific effects of individual food products or processing categories. The magnitude and direction of effect estimates may be influenced by unmeasured confounders that systematically differ between individuals consuming various UPF categories, potentially explaining some of the observed protective associations with certain processed foods.

Emerging research on gastrointestinal health outcomes adds another dimension to understanding UPF impacts. Studies examining gut microbiota composition suggest that different UPF categories may have varying effects on intestinal health, with some UPFs potentially disrupting beneficial bacterial populations while others may provide prebiotic benefits through fortification or specific ingredient profiles^{(Reference Brichacek, Florkowski and Abiona27,Reference Whelan, Bancil and Lindsay38–Reference Rondinella, Raoul and Valeriani42)} . These gastrointestinal effects represent an important mechanistic pathway through which food processing may influence broader health outcomes, including metabolic and CVD risk.

These consistent divergent risk profiles across different cohorts, geographic regions, and health outcomes strongly suggest that the relationship between food processing and health is considerably more nuanced than previously portrayed. Rather than recommending blanket reductions in all UPFs, a more targeted approach focused on specific high-risk UPF categories –particularly SSBs and processed meats – may prove more effective for chronic disease prevention⁽⁴³⁾. The beneficial associations observed with some UPF categories also suggest opportunities for food manufacturing innovation that prioritises nutritional quality while maintaining convenience and accessibility^{(Reference Harlan, Gow and Kornstadt41)}.

Criticising the NOVA system does not mean endorsing all UPFs as harmless. Existing nutrient-based classification systems, such as the high fat, sugar, and salt (HFSS) model, already highlight overlaps between the NOVA-3 and NOVA-4 categories (Figure 3). The main concern raised by critics is that grouping diverse products into a single NOVA-4 UPF category oversimplifies the issue, ignoring key differences in their health effects^{(Reference Gibney, Forde and Mullally3,Reference Gibney8,Reference Gibney9,Reference Visioli, Marangoni and Fogliano13,Reference Visioli, Del Rio and Fogliano44–Reference Dicken47)} . By collapsing items that other nutrient profiling systems classify separately, the NOVA system risks obscuring important distinctions that could inform more effective dietary guidance.

Figure 3. Overlap between high fat, sugar and salt classification with the NOVA-3 and NOVA-4 groups. LCBs, low-calorie beverages; SSB, sugar-sweetened beverages.

Methodological challenges in research

The assessment of relationships between UPF consumption and health outcomes faces several significant methodological challenges^{(Reference O’Connor, Herrick and Papier32)}, which affect both the quality of data collection and the reliability of research conclusions.

FFQs, while fundamental to dietary research, demonstrate inherent limitations in capturing the complexities of food processing. FFQs typically lack the precision to distinguish between processed and home-prepared versions of similar foods, especially since many were developed before the NOVA system and the term UPF existed^{(Reference O’Connor, Herrick and Papier32)}. For example, when assessing bread consumption, standard FFQs such as the Harvard FFQ⁽⁴⁸⁾ cannot differentiate between mass-produced commercial bread containing emulsifiers and preservatives (NOVA-4), artisanal bakery products (NOVA-3), and home-baked alternatives using traditional ingredients (NOVA-3), as the consumption of these were often lumped together into a single question (e.g. ‘How often do you consume white bread (slice), including pita bread’).

Classifying composite foods and mixed dishes presents an additional layer of complexity that reveals fundamental inconsistencies in the NOVA system’s application. In a study by Dicken et al. ^{(Reference Dicken, Batterham and Brown21)}, homemade cottage pie was categorised as NOVA-1 (minimally processed), even though typical recipes include ingredients from all four NOVA categories (Figure 4). This highlights a common concern among researchers and critics of the NOVA system: Should all homemade foods be classified as NOVA-1, regardless of their ingredients, simply because NOVA-4 emphasises ‘industrial formulation’? If so, what makes an ingredient or process acceptable at home but harmful in an industrial setting?

Figure 4. A common recipe of cottage pie which includes the use of ingredients from all 4 NOVA categories. Picture obtained from artist ‘moonkin’ via Adobe Stock under Education License from the author’s institution.

Contemporary dietary studies face significant obstacles in collecting and managing brand-specific information. Most dietary assessment methods do not capture detailed brand information, partly due to respondents’ limited ability to recall such details accurately and partly due to the practical constraints of dietary assessment methods^{(Reference Shim, Oh and Kim49)}. This limitation becomes particularly evident when examining seemingly identical products with different formulations, such as the popcorn variants discussed earlier. Additionally, many food composition databases in nutrition research reference generic food items and often lack detailed information about processing methods and complete ingredient lists^{(Reference Pennington50)}. This absence of comprehensive data makes accurate NOVA classification challenging or impossible^{(Reference Braesco, Souchon and Sauvant2)}.

Selective reporting and media amplification of UPF findings

While recent systematic reviews and meta-analyses have generated attention-grabbing headlines, e.g.^{(Reference Gregory51)}, a closer examination of the underlying research reveals critical gaps often overlooked in academic and public discussions.

Studies linking UPFs to poor health outcomes tend to have substantially higher Altmetric scores – an indicator of media exposure and public impact – than those reporting neutral or protective effects. While this pattern appears consistent with general tendencies in health research communication^{(Reference Wilson, Bonevski and Jones52–Reference Schwartz, Woloshin and Andrews55)}, empirical analysis of this potential publication and media bias and its impact on public perceptions represents an important area for future research.

The widely cited umbrella review by Lane et al. ^{(Reference Lane, Gamage and Du26)} provides a compelling example of this phenomenon. While the review identified direct associations between UPF exposure and 32 adverse health outcomes, including CVD, T2DM, anxiety, and common mental disorders, the strength of evidence varied considerably across these outcomes. Notably, the authors’ infographic acknowledged that 11 of these 32 harmful effects had ‘no evidence’ supporting such associations, while another 7 effects were supported only by ‘weak evidence’. Only 4 harmful effects were determined to have ‘convincing evidence’ – a crucial distinction rarely communicated in subsequent citations or media coverage.

Issues with experimental UPF studies

The experimental evidence regarding UPF health effects demonstrates several significant methodological challenges. While groundbreaking, the frequently cited controlled feeding study by Hall et al. ^{(Reference Hall, Ayuketah and Brychta56)} illustrates fundamental limitations in study design that complicate the interpretation of results. This study’s ultra-processed and unprocessed diets differed substantially in multiple characteristics beyond processing status alone. For instance, the study’s day one breakfast comparison exemplifies this issue: the ultra-processed meal (Honey Nut Cheerios, whole milk, and a packaged muffin) differed markedly from its unprocessed counterpart (Greek yoghurt, fresh berries, apple slices, and whole nuts) not only in processing level but also in fibre content and source, food form, palatability, and presentation.

The control of portion size and food presentation in intervention studies presents additional methodological challenges that may confound results with palatability differences rather than processing effects. For example, in the study by Hall et al. ^{(Reference Hall, Ayuketah and Brychta56)}, six out of seven dinners during the unprocessed diet period were salads (less palatable) compared with more palatable ‘UPFs’, which may have influenced consumption patterns through palatability differences alone rather than processing status. Similar design issues were present in the more recent Japanese study by Hamano et al. ^{(Reference Hamano, Sawada and Aihara57–Reference Louie59)}. While it is acknowledged that palatability is highly subjective and varies between individuals^{(Reference Johnson and Wardle60)}, it is theoretically possible to design a study showing higher ad libitum energy intake from NOVA-3 foods that many people find palatable (e.g. ham and cheese toasties and freshly squeezed orange juice with added fibre) compared to NOVA-4 foods that many find unpalatable (e.g. wheat bran sticks, soy milk with emulsifier, unsweetened Greek style yoghurt with golden syrup). This possibility highlights the importance of controlling for palatability when designing dietary intervention studies that examine the effect of UPFs on health outcomes.

Socioeconomic and environmental implications of UPFs reduction

While well-intentioned, the growing advocacy for reduced UPF consumption raises significant concerns regarding both socioeconomic accessibility and environmental sustainability, with implications that extend far beyond immediate health considerations.

The relationship between UPF consumption and socioeconomic factors reveals complex challenges in implementing NOVA system-based dietary recommendations. Research by Weaver et al. ^{(Reference Weaver, Dwyer and Fulgoni61)} demonstrates that processed foods in the U.S., many of which would fall under the NOVA-4 category, contribute significantly to essential nutrients in the American diet, providing 55 % of dietary fibre, 48 % of calcium, 43 % of potassium, and substantial proportions of vitamins and minerals including vitamin D, iron, folate, and vitamin B₁₂. While generally higher in energy density, some UPFs represent cost-effective nutrient sources for economically disadvantaged populations. Their analysis indicates that adhering to a diet composed primarily of minimally processed foods could increase household food costs substantially, potentially exacerbating food insecurity among vulnerable populations. Even prominent critics of UPFs acknowledge that wholesale restrictions on UPFs could harm food security^{(Reference Dénos, Vandevijvere and Boone62)}

The environmental implications of NOVA-based dietary recommendations present additional complexity in how the classification system characterises sustainability impacts. Research by Dénos et al. ^{(Reference Dénos, Vandevijvere and Boone62)} examining the Belgian food system reveals that NOVA categories demonstrate distinct environmental footprints that challenge assumptions about processing and sustainability. Despite UPFs (NOVA-4) providing approximately 50 % of daily calorie intake in Belgium, they contribute proportionally less to overall resource use than less processed alternatives, accounting for 50 % of climate change and land use impacts, 41 % of water use, and 38 % of fossil resource scarcity. In contrast, unprocessed and minimally processed foods (NOVA-1) demonstrate disproportionate environmental pressures relative to their caloric contribution, with red meat, beverages, and fruits and vegetables generating the largest impacts on greenhouse gas emissions, land use, and water consumption.

This suggests that NOVA’s processing-focused classification may not align directly with environmental sustainability metrics^{(Reference Ivens63)}. UPFs often incorporate industrial processes that maximise efficiency, reduce food waste, and optimise shelf life, factors that can contribute to lower overall environmental burdens through supply chain optimisation. Balancing health goals with sustainability efforts requires a careful, targeted approach. This may involve reducing intakes of UPFs with high environmental and health impacts, such as ultra-processed meats and sugary snacks^{(Reference Fardet and Rock64)} while considering how the food system can adapt without further burdening natural resources.

Policy implementation challenges of the NOVA system

While the NOVA system has gained significant traction in shaping dietary guidelines and public health policies, its practical implementation faces substantial challenges, primarily from its inherent complexity and the difficulty of translating its theoretical framework into actionable policy measures.

The integration of the NOVA system into dietary guidelines represents both its potential and its limitations. The Brazilian 2014 Dietary Guidelines are a prominent example of integrating the NOVA system into national nutrition policy, representing a significant shift in approach by incorporating nutritional quality considerations and socio-cultural aspects of food consumption^{(Reference Monteiro, Cannon and Moubarac1)}. However, comparative research suggests varying effectiveness across different populations. Studies comparing the NOVA system with traditional guidelines, such as the Australian Dietary Guidelines^{(Reference Grech, Rangan and Allman-Farinelli65)}, have revealed that traditional nutrient-based approaches might better predict certain health outcomes in specific populations, highlighting the need for context-specific adaptation rather than universal application.

Translating the NOVA system into enforceable regulations proves challenging due to their reliance on complex and often subjective criteria. Unlike clearly quantifiable measures such as sugar or sodium content, the concept of ‘ultra-processing’ proves difficult to codify in legislation. Food scientists and academics have repeatedly challenged the current NOVA definition for lacking objective precision^{(Reference Visioli, Del Rio and Fogliano44)}. The NOVA system contains numerous edge cases and exceptions^{(Reference Astrup and Monteiro7)}. When a definition requires an extensive list of exceptions far exceeding the original definition, it indicates a fundamental need for redrafting before it can serve as an effective legislative and regulatory tool.

Research into implementing policies targeting UPFs across various jurisdictions has revealed several strategies aimed at reducing their consumption. In Chile, comprehensive measures include mandatory front-of-package warning labels, restrictions on marketing to children, and bans on selling products high in calories, sodium, sugar, or saturated fat, many of which are UPFs, in schools^{(Reference Taillie, Bercholz and Popkin66)}. Mexico has banned the sale of junk food in schools, prohibiting items like sugary drinks and processed snacks to combat child obesity^{(Reference Basto-Abreu, Carnalla and Reyes-Sánchez67)}. Brazil’s National School Feeding Program mandates that at least 75 % of school meal funds be allocated to unprocessed or minimally processed foods, with no more than 20 % spent on processed foods, aligning procurement with dietary guidelines^{(Reference Azevedo, Bandoni and Amorim68)}. In the United States, while direct regulation of UPFs is limited^{(Reference Pomeranz, Mande and Mozaffarian69)}, the Dietary Guidelines Advisory Committee for 2025–2030 has been tasked with evaluating research related to UPF consumption^{(Reference Hoelscher, Anderson and Booth70)}, indicating a growing recognition of the issue. These varied approaches highlight the complexity of developing effective policies to address UPF consumption globally.

Future directions

The limitations of the NOVA system, particularly its broad categorisation of UPFs, suggest the need for more refined approaches to understanding the relationship between food processing and health outcomes. While the NOVA system has contributed significantly to discussions about food processing in public health, comparative analyses with other classification systems reveal opportunities for improvement and integration of alternative approaches that better serve public health objectives.

Comparisons between the NOVA system and systems like the Health Star Rating in Australia have revealed both areas of agreement and significant discordance, highlighting the complexity of food classification^{(Reference Barrett, Gaines and Coyle25)}. Similarly, studies comparing the NOVA system with the Nutri-Score system have demonstrated correlations between ultra-processing and lower nutritional quality while revealing the limitations of single-dimensional classification approaches^{(Reference Romero Ferreiro, Lora Pablos and Gomez de la Camara71,Reference Sarda, Kesse-Guyot and Deschamps72)} . These comparisons underscore the potential value of hybrid systems considering both processing methods and nutritional composition.

Research on the NOVA system is rapidly expanding^{(Reference Wang, Lu and Liu73)}, with advances such as automated classification using language models and validated questionnaires^{(Reference Dinu, Bonaccio and Martini18,Reference Hu, Flexner and Tiscornia74,Reference Oliveira, Frade and Gabe75)} , along with a refined classification system based on the NOVA system^{(Reference Bricher76)}. However, rather than contributing to these developments, the system’s original creator and supporters have called for a boycott, arguing that industry funding – though sourced from a non-profit foundation linked to a pharmaceutical company rather than the food industry – could introduce bias^{(Reference Monteiro77,Reference Rivera, van Tulleken and Baker78)} . Given the critiques from various academics^{(Reference Braesco, Souchon and Sauvant2,Reference Gibney, Forde and Mullally3,Reference Astrup and Monteiro7–Reference Gibney9,Reference Visioli, Marangoni and Fogliano13,Reference Visioli, Del Rio and Fogliano44–Reference Dwyer, Fulgoni and Clemens46,Reference Petrus, do Amaral Sobral and Tadini79–Reference Levine and Ubbink88)} and as outlined in this review, it remains an open question whether the system requires refinement.

Cultural and regional considerations must be central in refining food classification systems^{(Reference Katidi, Vlassopoulos and Noutsos20)}. Integrating the NOVA system with other food classification systems and developing hybrid approaches that combine the strengths of different systems offer promising avenues for future research^{(Reference Phulkerd, Dickie and Thongcharoenchupong89)}. This holistic approach could provide a more comprehensive assessment of food quality and its impact on health outcomes^{(Reference Phulkerd, Dickie and Thongcharoenchupong89)}, particularly when examining traditional food processing methods across different cultures, where blanket categorisations may inappropriately classify culturally significant foods as UPFs despite their established role in healthy traditional diets.

The evolution of food classification systems should focus on developing more granular distinctions within the UPF category, incorporating objective measures of processing impacts on nutritional quality, and establishing clearer links between specific processing methods and health outcomes. Rather than categorical judgments based on processing status alone, future efforts should focus on specific, measurable characteristics that influence health outcomes supported by robust scientific evidence.