Large effect sizes can be expected in drug studies, intended to show superiority or comparability, as there is a no-intake control group or a true placebo. There is a distinction in the reduction of the symptoms of disease, which is present in the enrolled population and the absence of the intervention does not cause disease (23, 28).

In contrast, true placebos are not possible with nutraceutical studies and the interventions involve the intake of nutrients, the absence of which may cause disease. In the absence of a true placebo group, and the requirement of ‘healthy' participants, nutrients have a smaller effect size and a longer response time to detect health outcomes. Essential nutrients are necessary for health and thus the underlying hypothesis that low or inadequate intake of nutrients causes or contributes to disease (29). Heaney stated, “with nutrients, the question is always not ‘whether' but ‘how much'?” (29). As such, EBN is a complex and intricate puzzle. Only one function of a nutrient (the first to appear and lead to disease or death) is used to define the disease (i.e., deficiency syndromes like beriberi and scurvy). But our understanding of all the other functions (beyond deficiency) is where we find the challenge of demonstrating the many benefits of a single nutrient when the proof of efficacy is limited to EBM (29).

The regulatory requirements of enrollment of populations “defensible as healthy” in studies intended for substantiation of structure/function claims limit the purpose of nutraceutical investigations. This accentuates the small effect size, decreasing the gap between the placebo/control and intervention (30). In the absence of an acceptable definition for ‘healthy,' a regulation that uses such a guideline to reject a study for claims is questionable. Changing medical treatment algorithms (31) that continuously redefine healthy populations while moving the dichotomy in the definition of health vs. disease and advocating earlier pharmacological treatment narrows the population that may be called ‘healthy.' Not only are these restrictions impractical for participant recruitment, but they are also not conducive for moving markers of respective outcomes and highlights flaws in logical application.

Persistent advocacy and requirements result in findings of clinical trials being limited to narrow and unrepresentative populations. This is a disservice to consumers seeking alternative solutions to improve their health. More recently, the concept of ‘healthy' has been upended by investigations showing that those previously considered ‘healthy' may in fact be metabolically unhealthy (32). A paradigm change is needed in defining the term ‘healthy' when applied to the nutraceutical industry.

The tendency of scientists and researchers to rely and focus on p values as the “gold standard of statistical validity,” above and beyond other frameworks for data analysis, which include statistical power, false positives and negatives have been challenged (33). The p value does not address the question if the probability of the study hypothesis is correct in the first place, and furthermore does not account for the actual size of the effect. Nuzzo argued that “researchers need to realize the limits of conventional statistics” and “bring into their analysis elements of scientific judgment about the plausibility of a hypothesis and study limitation” (33).

Recent statements by the American Statistical Association (ASA), EFSA, and New England Journal of Medicine (New Engl J Med) point away from such requirements that are founded on misinformation and urge movement toward providing clinical significance. “The p value was never intended to be a substitute for scientific reasoning,” said Wasserstein, stating “Well-reasoned statistical arguments contain much more than the value of a single number and whether that number exceeds an arbitrary threshold” and is a directional move away from and toward a “post p < 0.05 era” (34).

The EFSA Scientific Committee suggested that statistical significance should not be the primary objective in analysis and focus should be directed on confidence intervals and biological relevance (35). New Engl J Med followed with a statement regarding publications and p values stating that nutrient studies should be directed to solve a health indication, not for statistical convenience (36).

The multifunctional nature of nutraceuticals and their multi-targeted outcomes are not readily determined within the scope of the current RCT model. These multifunctional effects could be considered a complex intervention (37) and are not often captured in the setting of a single primary endpoint, resulting in the intervention being found ineffective and leading to higher proportions of false negatives. The Medical Research Council of the United Kingdom stated that complex interventions require a different approach. While a single primary outcome is usually the most straightforward for statistical analysis, it may not provide adequate assessment of the success of an intervention that has effects across a range of indications (37).

A global outcome that assesses and scores the effects of an investigational product across systems corresponds more closely to the multifunctional outcome proposition of nutrients in the human body. When designed to reflect health rather than disease, a global outcome captures systemic effects (37). Previously, such a concept has been proposed as a composite of several outcomes providing a higher statistical power to the study to detect small changes across multiple organ systems (38).

Global indices of health are appropriate in the nutritional sciences realm since functional foods, nutraceuticals, and other NHPs do not fit into the pharmaceutical model. A global index is different from a composite of several outcomes. A composite index cannot identify the variable that contributes to the majority of the effects, and when combined as a primary outcome can also mask a real benefit of a single constituent endpoint by being diluted from multiple null effects within the cluster (39, 40). A global index weights each outcome allowing for a better and more informed assessment of the outcome(s) contributing to the results. Health-related quality of life questionnaires such as the 36-item Short Form Survey (SF-36) or Profile of Mood States (POMS), and objective biomarkers might be examined as an approach for creating global health indices that better capture the efficacy of the intervention. This has been previously reported where Antony et al. suggest combining outcomes of cognition with physiological biomarkers of immunity and metabolism to arrive at a global index for cognitive health (41).

Furthermore, the index should evaluate and quantify positive movement of a negative health marker. Such concepts have previously been evaluated in quantifying disease activity in patients with rheumatoid arthritis and includes patient history, laboratory tests, physical examination, and imaging studies (42). Multiple common indices, were used to assess adiposity, such as body mass index (BMI), bioelectric impedance analysis (BIA), and anthropometric measures like skinfold thickness and waist: hip ratio as well as computed tomography imaging (42). Thus, a global index as the primary outcome comprising multiple single endpoints could improve statistical precision, increase efficiency, reduce trial size and cost, and may provide trial results earlier (43).

Heaney questioned “whether we need as much proof of efficacy for a nutrition policy decision as we do for approval of powerful, expensive and potentially dangerous pharmaceutical agents” (29).

Reductions in intake of a nutrient may lead to an increase in the risk of disease or may result in disease. It is unethical to lower nutrient levels to study efficacy endpoints. Heaney et al. proposed a shift from examining proof of efficacy to that of probable harm, a calculus of benefit compared to harm, to be evaluated on a nutrient-by-nutrient basis (29). Investigating pre-diseased populations where the placebo or control group presents either low or harmful levels of surrogate biomarkers may be beneficial in the application to nutraceuticals. Using the calculus of benefit vs. harm, the proof of harm is established in a pre-diseased population since they will eventually progress to a diseased state in the absence of an intervention. If the safety of nutraceutical interventions can be demonstrated through high-quality and comprehensive studies, then the calculus of benefit vs. harm of the intervention shifts toward benefit.

A recent study examined this concept in an RCT model and found that following the progression of the placebo group from baseline to the end of the study was valuable. Left untreated, an at-risk population progressed to disease, which was observed in the placebo group (44) confirming Heaney's concept (29). On further investigation and application of the Framingham risk score, it was found that left untreated, those with cardiovascular risk factors may progress to a more hypertensive and hypercholesterolemic state (44).

A small effect size and study populations, as well, the flawed expectations that nutraceuticals should behave like drugs often result in RCTs generating null results. If one were to stop at this point, most RCTs would suggest that the ingredient and or the formulation was not efficacious.

Therefore, application of EBM principles to EBN solely for meeting the standards of acceptable evidence for policy makers is contentious. Applying the principles of EBM to EBN in its entirety is troublesome and unsuitable for the assessment of nutrients, dietary supplements, and foods. A paradigm shift is required in trial design for evaluating nutrient principles and examining RCT designs that are better suited for nutraceuticals. Blanket application of the RCT concept to the nutraceutical model is not sensitive to establish health promotion and optimization and, in many instances, has led to an exercise in futility.

Thus, it would be reasonable to expect that the evidence required to prove efficacy for drugs and nutrients would need to be different. Both drugs and nutrients have different attributes in contributing to the welfare of an individual. Importantly, drugs are therapeutic in nature, whereas nutrients may not only reduce the risk for disease but may help promote optimal health and performance.

There are various considerations for designing high-quality studies for nutrition and nutraceutical research. Considerations have been previously presented by Lichtenstein et al. and the Agriculture and Agri-Food Canada “Best Practices for Food-Based Clinical Trials” (46, 47). Following rigorous design considerations, there have been large-scale, long-term, double-blind, placebo-controlled trials of dietary supplements including but not limited to the Age-Related Eye Disease Study (AREDS) and AREDS 2 (48), Alpha-Tocopherol, Beta-Carotene Cancer Prevention (ATBC) Study (49), Carotene and Retinol Efficacy Trial (CARET) (50), Cocoa Supplement and Multivitamin Outcomes Study (COSMOS) (51), Selenium and Vitamin E Cancer Prevention Trial (SELECT) (52), Supplementation en Vitamines et Mineraux Antioxydants (SU.VI.MAX) (53), VITamin D and OmegA-3 TriaL (VITAL) (54), Women's Antioxidant and Folic Acid Cardiovascular Study (WAFACS), Physicians' Health Study-I and II, Women's Health Initiative (WHI), and Women's Health Study (WHS). However, these large-scale trials are extraordinarily expensive, not always feasible, and mostly unnecessary in the context of evaluating the health benefits of nutraceuticals.

Previous research in nutraceuticals validates the importance of measuring responsiveness to treatment at the individual level (24). About 25% of participants responded positively to an improvement in blood pressure and approximately 33% to insulin sensitivity (24). An estimated 66-75% of participants were non-responders to the intervention, which indicated that the intervention worked in a particular subset of participants, a group that could not be identified prior to the study. Examination of group variability allows researchers to make informed decisions of the inter-individual differences of clinical relevance (55). Population-wide recommendations, currently in place for reduction in dietary sodium, were based on the benefits in hypertensive subjects and implementation of folic acid fortification programs were based on neural tube defect prevention (56). The evidence demonstrates a precedence exists for nutrition study results in a responder population that can be generalized to the greater population when positioned correctly. Studies on probiotic interventions for antibiotic-induced diarrhea have indicated similar results (57).

The history of N-of-1 show this type of study was developed due to the need for determining “optimal therapy” for a patient and proved to be “spectacularly helpful” in patient treatment (58). Since its publication, clinicians have reported on thousands of N-of-1 trials. These studies improved patient outcomes and drug development, expanded to encompass many indications (59), and have reported their value for conditions where conventional RCTs are available (60). N-of-1 trials are defined in part by randomizing multiple crossover trials in a single participant (61–64). This N-of-1 study design has been utilized or proposed in nutrition and nutraceutical study designs (65–69). For example, the Westlake N-of-1 Trials for Macronutrient Intake (WE-MACNUTR) examined the use of personalized dietary recommendations on glucose metabolism and gut microbiota in the prevention of chronic disease (66, 68). Further, Soldevila-Domenech et al. reviewed the use of nutraceutical interventions, including phenolic compounds, omega-3 polyunsaturated fatty acids, vitamin D, vitamin C and other micronutrients, aimed at improving cognitive function in older adults with the potential application to N-of-1 designs specifically for dementia prevention (69).

Recently, N-of-1 designs, as part of the umbrella of single-case experimental designs (SCDs), has been discussed in detail by a Nutrition Research Task Force of the American Society for Nutrition (70). They state that “to determine whether, and to what extent, people respond differently to interventions, different designs needs to be used” (70). A general example of an N-of-1 trial design used for comparing responses to two interventions is shown in Figure 1. The intention of this design is to utilize multiple N-of-1 studies in informing decision making.

When compared to the traditional RCT, N-of-1 provides greater strength and more reliable data in clinical decision-making (60). Further, this design may provide specificity and inherently address the heterogeneity of treatment effects when compared to RCTs (70). Models have been expanded to accommodate crossover features including carry-over effects (71) and maintains randomization, blinding, and formal outcome assessment. The Oxford Centre for Evidence-Based Medicine has classified N-of-1 trials as Level 1 evidence, the same level of evidence for systematic reviews of RCTs (72). N-of-1 studies are a unique method identifying optimal intervention approaches for individual cases. Accumulation of multiple N-of-1 studies utilized in meta-analysis can be generalized across investigational populations (60, 70).

The N-of-1 trial allows for rapidly identifying responders vs. non-responders, requires enquiry into clinical significance, and provides a venue for the use of a global index outcome that is more generalizable to the nutraceutical industry (73). Furthermore, an N-of-1 at a Level 1 evidence stage allows for the measurement of probable harm in the absence of an intervention, provides clinical relevance, allows for the accuracy of detection of efficacy, and can function within the guidelines of the traditional RCT. Such a model would be of value in providing reliable and reasonable information regarding the intervention.