Introduction
Poor nutrition is a cause of chronic diseases such as ischemic heart disease (IHD), stroke, obesity, and type 2 diabetes (Reference Bush, Blumberg and El-Sohemy1;Reference Jardim, Mozaffarian and Abrahams-Gessel2). In 2019, dietary risk factors contributed globally to approximately 7.94 million deaths and 188 million disability-adjusted life years among people aged 25 years and older (3). Moreover, dietary factors account for approximately 18.2 percent of the costs associated with IHD, stroke, and type 2 diabetes in the United States (Reference Jardim, Mozaffarian and Abrahams-Gessel2). Personalized nutrition has emerged as a promising field to address the limitations of current diet interventions and slow down the chronic disease pandemic (Reference Bush, Blumberg and El-Sohemy1). Since each individual has different nutrient needs and responses to diets, insights into these individual needs and responses can be leveraged to prevent, manage, and treat diseases and to improve health (Reference Ferguson, De Caterina and Görman4). Personalized nutrition has been defined by Ordovas et al. (Reference Ordovas, Ferguson, Tai and Mathers5) as an approach that utilizes individual characteristics to provide targeted nutritional advice, products, or services. To develop such advice, products, or services, clinical assessments, biomarkers of physiological function and pathological processes, genetic information, and other available data derived from advanced technologies are needed (Reference Bush, Blumberg and El-Sohemy1).
While information on lifestyle and personal goals is commonly used to formulate personalized nutrition advice, the same is not true for advanced technologies such as those involving metabolomics and genotypic data, despite their potential to improve health outcomes (Reference Adams, Anthony and Carvajal6;Reference Shyam, Lee and Tan7). One project that explored the potential of advanced technologies in people with normal weight, overweight, and obesity is PREVENTOMICS, a recently completed European Horizon 2020 project (8), which investigated the potential of omics (especially metabolomics) as an input for personalized nutrition advice (Reference Keijer, Escoté and Galmés9). By combining phenotypic characterization at the metabolomic level with a person’s genotype, lifestyle, health status, preferences, and physiological status, a novel platform was developed and integrated into third-party applications. This integration resulted in three PREVENTOMICS interventions (Reference Keijer, Escoté and Galmés9), which included the following: [1] integration of the platform for personalized food delivery, [2] integration of the platform at the retailer level for personalized recommendations when shopping, and [3] integration of the platform with a software to support healthcare professionals with formulating personalized dietary plans for consumers (Reference Bothos10).
Decisions regarding the implementation of new approaches in healthcare, such as PREVENTOMICS, are rarely simple (Reference Rouse and Serban11). Growing pressure on healthcare budgets has resulted in increased scrutiny of the overall value of new health technologies and programs (Reference Joore, Grimm and Boonen12). In this context, the importance of conducting a health technology assessment (HTA) is emphasized. HTA is a “multidisciplinary process that uses explicit methods to determine the value of a health technology at different points in its lifecycle” (13). “Value” includes different dimensions, such as clinical effectiveness, safety, costs, and ethical and legal issues. HTA promotes transparency and accountability in government performance, and it can also help developers of new technologies in understanding how their technology will be assessed (i.e., early HTA); by conducting such an “early HTA”, the time and financing required for their product to gain market entry or get reimbursed can potentially be reduced (Reference Garrido, Kristensen, Nielsen and Busse14;Reference Ananthakrishnan, Luz and Kc15).
Previous HTAs have often assessed only the costs, health effects, and cost-effectiveness of nutrition interventions and have not systematically examined a wider range of possible issues relating to health care and society (Reference Gutiérrez-Ibarluzea and Arana-Arri16). To overcome the variance in the extent and scope of HTA, and the differences in reporting of the results, the European Network for HTA (EUnetHTA) developed the HTA Core Model (Reference Lampe, Mäkelä and Garrido17). Conducting an (early) HTA with the HTA Core Model offers advantages such as the identification of key assessment components of interventions, the provision of a structured analysis of (early) scientific evidence, and the highlight of existing gaps from which the recommendations for subsequent decision-making steps can be formulated (18). Despite these benefits, only a limited number of studies utilizing the HTA Core Model for HTA have been published in scientific journals (Reference Bilekova, Gavurova and Rogalewicz19–Reference Mueller, Pattinson, Hlongwane, Busse and Panteli21), and none of them were conducted in the nutrition field. As we believe that assessing the PREVENTOMICS interventions with the HTA Core Model in the premarket phase can help to inform further development and potential implementation decisions, this study aimed to compare these interventions with non-personalized nutrition interventions for people with normal weight, overweight and obesity, on all of the domains found in the HTA Core Model.
Materials and methods
General information regarding the HTA Core Model
The PREVENTOMICS interventions were evaluated using the HTA Core model developed by EUnetHTA, which has nine domains covering all aspects of an HTA (see Table 1) (Reference Kristensen, Lampe and Wild22). This model was chosen because of its methodological framework for producing and sharing HTA information (Reference Kristensen, Lampe and Wild22). Alternative frameworks were evaluated but not selected for various reasons. For example, the ISPOR Value Flower, which offers a broader perspective on factors contributing to value in healthcare, was not chosen because it predominantly centers on the concept and measurement of value rather than on the process and execution of HTA (Reference Neumann, Garrison and Willke23). The methodological framework of the HTA Core Model includes three components: [1] an HTA ontology including standardized questions (i.e., assessment elements) organized within a framework featuring nine domains that encompass all aspects that may be relevant for HTA and thereby value assessment, [2] methodological guidance, and [3] a common reporting structure. We used the first two components of the framework wherever possible. We did not use the common reporting structure and instead provided a summary of the relevant information per domain related to the PREVENTOMICS interventions, which gives a streamlined and accessible documentation of essential information. We believe that this is sufficient for stakeholders who are interested in further development or in taking (decision-making) steps regarding the implementation of the interventions.
Different domains of the HTA Core Model, including the related methodology and sources used to address the domain
a All results were part of D6.5 (“Health Technology Assessment”).
b Published online: https://preventomics.eu/deliverables/#1593502709004-84c73ce5-2fe4.
c In this regard, “patient” and “individual” denotes those receiving a technology. This study focused on people without chronic diseases, and therefore the term “Individual” (or “participant”) was used in this HTA.
D, Deliverable; HTA, Health Technology Assessment; PREVENTOMICS, Empowering consumers to PREVENT diet-related diseases through OMICS sciences.
Domain specific methods
Table 1 gives an overview of all domains, the description of the domains, and the different sources used to gather information. A summary of domain-specific methods is given below. In general, information for the different domains was gathered via [1] different statistical analyses (i.e., analyses of health outcomes and questionnaires) and modeling studies (i.e., cost-effectiveness modeling and choice modeling); [2] questions asked via email to partners of the PREVENTOMICS project, who are experts in this field; or [3] other (un)published materials. Published materials included literature published in scientific journals, PREVENTOMICS blog posts, and presentations. Unpublished materials included project deliverables (D). These deliverables are also known as supplementary outcomes (such as information, specialized reports, or brochures) that were required to be generated at a specific time throughout the project (24). All published materials related to the PREVENTOMICS project can be accessed on the website (8), and information about the referenced deliverables is provided in Supplementary Table S1.
In most domains, (un)published materials were used as input, as well as the questions that were asked of the project partners (see Table 1). Additionally, clinical trial data were used as input for the “clinical effectiveness” and “cost and economic evaluation” domains and were analyzed using statistical methods (see footnote Table 3 for more details), with some results extrapolated over a lifetime. Although some of these results were already published elsewhere (Reference Aldubayan, Pigsborg and Gormsen26–Reference Galekop, Uyl-de Groot and Redekop29), we provided a summary of the trial-based effectiveness on dietary intake (i.e., Mediterranean Diet Adherence Score), anthropometrics (i.e., body fat, waist circumference, and body mass index (BMI)) and QoL (assessed with the EQ-5D-5L and the Obesity and Weight Loss Quality of Life (OWLQOL)) (30;Reference Patrick and Bushnel31).
The Markov obesity model with a 1-year cycle length was used to analyze data over a lifetime horizon and had different health states: diabetes, IHD, stroke, and death (see Figure 1 for the model structure) (Reference Hoogendoorn, Galekop and van Baal32). The model simulated the disease occurrence for an obese cohort based on various inputs (e.g., population demographics and trial-based effectiveness on BMI). The effectiveness measure was quality-adjusted life years (QALYs) and the cost-effectiveness was expressed in the incremental cost-utility ratio. More details about the model and inputs can be found elsewhere (Reference Hoogendoorn, Galekop and van Baal32). Detailed lifetime results were published elsewhere (Reference Galekop, Uyl-de Groot and Redekop27–Reference Galekop, Uyl-de Groot and Redekop29) and summarized in this study.
Structure of the Markov model for obesity as presented by Hoogendoorn et al. (Reference Hoogendoorn, Galekop and van Baal32). BMI, body mass index; IHD, ischemic heart disease.
Input for the “patients and social aspects” domain was supplemented with a validated diet satisfaction questionnaire (DSat-28 (© Laboratory for the Study of Human Ingestive Behavior, The Pennsylvania State University)), that assesses satisfaction with weight-management diets (Reference James, Loken and Roe33). The DSat-28 consists of 28 items with five response options ranging from “disagree strongly” to “agree strongly.” The total score was calculated by averaging the summed score; higher scores indicate greater diet satisfaction. Additionally, preferences regarding personalized nutrition interventions were obtained from results from two published discrete choice experiments (DCEs) (Reference Galekop, Veldwijk, Uyl-de Groot and Redekop34), that assessed preferences about [1] personalized nutrition advice and [2] personalized meals. More information about the methodology of these DCEs can be found elsewhere (Reference Galekop, Veldwijk, Uyl-de Groot and Redekop34).
Results
Health problem and current use of technology
The PREVENTOMICS interventions were used in four countries (Denmark, the United Kingdom (UK), Poland, and Spain) targeting overweight and obese populations (8). Spain also included individuals with normal weight (see Supplementary Table S2 for obesity classification by BMI). All interventions aimed to prevent diet-related diseases and improve health (8). More details can be found in Table 2.
Details on the PREVENTOMICS interventions, including information on the different intervention arms, study population, and target condition
a Behavioral change program: delivered via ONMI (https://www.onmi.design/preventomics). Participants received two to three Do’s (behavioral prompts) per week. In nature, participants were prompt to take a specific action. The Do’s in the PP group were based on participants’ reports from the behavioral questionnaire at baseline and inputs from nutritional recommendations.
ALDI, supermarket; BMI, body mass index; cm, centimeter; ICT, information, and communication technology; kg, kilogram; m, meter.
The burden of obesity is high; in 2016, over half of the population in OECD countries was overweight and nearly one in four had obesity (35). Poor diet significantly contributes to this obesity epidemic, with almost half of the population not meeting healthy diet guidelines and international standards. Overweight and related co-morbidities reduce average life expectancy in OECD countries by 2.7 years on average (35). Moreover, overweight and obesity result in an economic burden due to increased healthcare costs and reduced productivity. Over the next 30 years, OECD countries are projected to spend an average of 8.4 percent of their health budget on overweight-related problems, leading to a 3.3 percent reduction in gross domestic product due to obesity (35).
Although countries have implemented policies to tackle overweight and obesity, their success has been limited (35). Improvements in specific strategies such as mobile apps to promote healthier lifestyles could potentially tackle overweight and obesity. One study (D1.2 (“Consumers Report”)) and the literature (Reference Ghelani, Moran, Johnson, Mousa and Naderpoor36) found that many mobile apps for this purpose already exist. However, as far as we know, PREVENTOMICS uses a unique approach by applying new technologies (see “description and technical characteristics of the technology” domain) (Reference Keijer, Escoté and Galmés9).
Description and technical characteristics of the technology
The PREVENTOMICS interventions assessed in this HTA involved the use of a platform in different ways. In general, the platform used relevant algorithms and analytics services to analyze user data (genetic, biological, nutritional, psychological) and stored it for providing personalized nutrition recommendations (Reference Keijer, Escoté and Galmés9). These recommendations were transmitted through three different dietary apps: SimpleFeast, ALDI, and MetaDieta.
In more detail, the first PREVENTOMICS intervention integrated the platform with the SimpleFeast app for personalized meal delivery in Denmark (Reference Bothos10;Reference Aldubayan, Pigsborg and Gormsen26;Reference Aldubayan, Pigsborg and Gormsen37). The second intervention integrated the platform at the retailer level with an ALDI supermarket app in Spain (developed ad hoc), which enabled customers to read personalized food product recommendations while grocery shopping (Reference Bothos10;Reference Del Bas38). The third intervention integrated the platform with the MetaDieta app, designed for use by dieticians and study participants in the UK and Poland (Reference Bothos10;Reference Malczewska-Malec39;Reference Calder40). Dieticians used this app to prepare diet plans and share them with the participants. Moreover, all interventions included a behavioral change program (Reference van Berlo41) (see Table 2 for additional intervention details, Supplementary Figure S1 for the PREVENTOMICS user journey, Supplementary Figures S2a–d for the study designs, and Supplementary Table S3 for required training and tools).
Reimbursement policies for nutrition-related technologies vary both across and within countries. Generally, nutrition interventions or related areas such as digital health tools are not reimbursed (Reference Bush, Blumberg and El-Sohemy1;Reference Poley42). However, recent initiatives, such as the introduction of the Digital Healthcare Act (Digitale-Versorgung-Gesetz) in Germany, aim to improve healthcare through digitalization and innovation by reimbursing tools such as obesity apps (Reference Gerke, Stern and Minssen43) (see Supplementary material S1, for example, of reimbursement policies for different areas related to the PREVENTOMICS interventions in different countries).
Safety
PREVENTOMICS interventions are generally safe for individuals; no specific safety risks are related to the use of digital tools (a major component of the interventions). However, other activities related to the interventions may have safety hazards. For example, drawing blood (one to two times per year) may cause minor bruising at the puncture site. Moreover, there is a risk of contamination due to improper needle management. To address these concerns, alternatives such as skin monitors for blood glucose measurement (44) or finger pricks (for small blood volumes) (Reference Klingler and Koletzko45) can be used. In addition, there is a theoretical possibility that participants could receive the wrong type of personalized nutrition. However, manual checks minimize this risk. Moreover, since all dietary plans are based on the Mediterranean diet, recognized as a healthy diet, any potential error would have limited impact on health outcomes. The interventions do not pose risks to environmental or occupational safety.
Clinical effectiveness
To summarize the effectiveness of the PREVENTOMICS interventions, both short-term effectiveness (trial-based effectiveness) and long-term effectiveness (modeling trial-based effectiveness over lifetime) were studied (see Table 3) and varied by intervention and country. In both intervention groups (PP and PN: see Table 2 for description) and the control, we observed short-term changes in health outcomes, including shifts in BMI and utilities (i.e., quality of life score) from baseline to follow-up. These shifts were generally associated with improved health (i.e., decreased BMI and improved EQ-5D-5L utilities); BMI change ranged from −1.31 kg/m2 (PP group, UK) to 0.08 kg/m2 (control, Spain) and utility change ranged from −0.02 (control, Denmark and UK) to 0.06 (PN, UK). Additionally, these changes from baseline to follow-up in PP and PN groups were compared with those in the control group, providing estimates of the difference in effectiveness between interventions and control, accompanied with 95 percent confidence intervals. The highest (statistically significant) effect on BMI was measured when PN was compared with control in Spain (−0.53 kg/m2) and in utilities when PP was compared with control in Denmark (0.04). Notably, we observed contrasting effectiveness results in BMI in Poland when PN was compared with control; BMI in the control group decreased more than in the PN group, resulting in a +0.20 kg/m2 difference. Analysis of the OWLQOL indicated significant increases in QoL for all PP and PN interventions compared to baseline (e.g., PP in Denmark: +3.85 (SE: 1.67)). However, statistically significant differences in OWLQOL between interventions were generally not observed in most countries, except for PN versus control in Poland.
Trial and model outcomes related to (discounted) effects, costs, and cost-effectiveness
a Different statistical tests were performed. Generalized estimation equations were used to analyze the EQ-5D-5L utilities and linear mixed models were used to quantify the differences in effects between the PP/PN and control of all other health outcomes.
b p<0.01 significantly change from baseline.
c p<0.05 significant difference between groups.
d p<0.05 significantly change from baseline.
e p<0.01 significant difference between groups.
f Quality of life score.
g ©Laboratory for the Study of Human Ingestive Behavior, The Pennsylvania State University.
h Discounted results were presented.
i Base case: Point estimates of BMI as observed from the trials were used as input in the model.
j Scenario: The lower level of the 95% confidence intervals from the effect in BMI was used as input in the model.
k All costs were then converted from 2020 national currency to 2020 Euros using the following exchange rates: 1 DKK = 0.134 Euro, 1 Zloty = 0.225 Euro, 1 pound = 1.123 Euro.
l WTP thresholds: Denmark; €47,817 per QALY gained (357,100 DKK), Spain; €30,000 per QALY gained, UK; €22,461 per QALY gained (20,000 pounds), Poland; €38,430 per QALY gained (171,092 Zloty).
BMI, body mass index; CI, confidence interval; cm, centimeter; Cum, cumulative; DKK, Danish krone; DSAT; diet satisfaction questionnaire; EQ-5D, EuroQol five-dimension questionnaire; ICUR, incremental cost-utility ratio; IHD, ischemic heart disease; kg, kilogram; MEDAS, Mediterranean diet score; m, meter; OWLQOL, Obesity and Weight Loss Quality of Life; PN, personalized nutrition intervention; PP, personalized plan intervention; QALY, quality-adjusted life year; SE, standard error; UK, United Kingdom.
Predicting long-term outcomes based on short-term effects on BMI and utilities revealed that generally both PP and PN interventions led to improved lifetime health outcomes compared to the control group, translating into potential benefits such as fewer years with diabetes, increased life expectancy, and lifetime health (QALYs). However, as Poland showed contrasting effectiveness results over the trial period, PN also had worse lifetime health outcomes compared to control (e.g., −0.015 QALYs) in base-case scenario. Scenario analyses, using the lower 95 percent confidence limit of short-term effectiveness on BMI (i.e., −0.45 kg/m2), revealed increased QALYs for PN compared to control (+0.032), consistent with findings in other countries. More details on health outcomes can be found in Table 3 and in published materials (Reference Aldubayan, Pigsborg and Gormsen26–Reference Galekop, Uyl-de Groot and Redekop29;Reference Clamp and Baker46–Reference Rabassa, Bosch, Companys and Calderon48).
Costs and economic evaluation
The interventions (PP and PN) had higher costs compared to the control over the trial period, with Denmark showing the highest costs (see Table 3). Supplementary Tables S4a–d provide further details on the intervention costs. Over a lifetime horizon, costs were considered from an extended societal perspective, including obesity-related disease costs, unrelated medical costs, nonmedical costs, informal care costs, and productivity costs. In summary, lower costs related to diabetes, IHD, and stroke were offset by higher costs in other areas (i.e., unrelated medical costs, nonmedical costs, and informal care) due to increased life years resulting from the interventions. Depending on the chosen willingness-to-pay threshold and the specific intervention (PP or PN), some interventions were deemed cost-effective, such as PP and PN in the UK and PP in Poland. Scenario analyses revealed additional cost-effective interventions, including PN in Spain and PN in Poland (see Table 3 and published materials (Reference Galekop, Uyl-de Groot and Redekop27–Reference Galekop, Uyl-de Groot and Redekop29) for more details). Given the high prevalence of overweight and obesity, personalized nutrition interventions would have a substantial budget impact.
Ethical aspects
This HTA included an examination of ethical issues. The PREVENTOMICS interventions demonstrated a favorable benefit–harm balance, as they showed no significant harms (safety domain) but some improvements in clinical effectiveness (effectiveness domain). Moreover, the interventions respect individual autonomy, human dignity, human rights, and participants’ privacy and integrity. However, health inequality may arise if these interventions are not reimbursed by a third party and may thus be necessary to prevent disparities between wealthier and poorer individuals. More specifically, lower-income individuals generally have poorer diets and higher disease burdens, while higher-income individuals have better access to the interventions (Reference Mathers49). Additionally, older individuals may face challenges in using the interventions due to digital illiteracy or lack of suitable mobile phones (see Supplementary material S2 for more details).
Organizational aspects
In general, the PREVENTOMICS interventions were considered supplementary to the existing work processes of professionals such as nutritionists or dieticians. Professionals were likely to be familiar with the use of apps to document health behaviors but were asked to perform additional tasks related to genetic and metabolic sampling, which they usually do not do. Besides guidance on sampling for genetics and metabolomics, minimal training or education is expected (see Supplementary Table S3). However, besides the comparable study design in the UK and Poland, the (cost)-effectiveness results were not consistent. One possible explanation is that the UK utilized a more didactic approach for providing recommendations, resulting in better outcomes. Providing training to professionals on delivering information may therefore optimize results.
Personalized nutrition requires that participants undergo tests, which might decrease their enthusiasm. However, an app to document food habits and other information could help maintain their motivation. Overall, participants generally accepted the PREVENTOMICS interventions well, despite some difficulties in app usage, particularly in the UK and Poland. However, most problems were solved or had minimal impact. More details and examples can be found in Supplementary material S3.
Patients and social aspects
Understanding the experiences of overweight or obese individuals is crucial for the success of PREVENTOMICS interventions. Farrell et al. (Reference Farrell, Hollmann and le Roux50) found that people with obesity experience negative issues, such as emotions, traumas, restrictions in movements, stigma, and lack of respect. The DSat-28 results indicated slight increases in diet satisfaction for almost all intervention groups compared to baseline (see Table 3). Additionally, a DCE study revealed willingness to choose personalized nutrition interventions, with total expenditure being the most important factor influencing peoples’ preferences (Reference Galekop, Veldwijk, Uyl-de Groot and Redekop34). Behavioral reminders were not highly valued. The DCE study also showed participation rates for specific scenarios, including scenarios somehow similar to PREVENTOMICS interventions, and revealed rates varying from 26 percent to 49 percent across countries and interventions (Reference Galekop, Veldwijk, Uyl-de Groot and Redekop34). Moreover, a UK cohort study revealed substantial variations in genetic testing preferences, which tests are also needed in personalized nutrition interventions, between white and ethnic minority individuals, with the white cohort being twice as likely to undergo genetic testing (Reference Harris, McCabe and Shafique51).
Gaining user trust is crucial for intervention success, emphasizing the importance of transparent and simple explanations of interventions and their benefits (D1.2 (“Consumers Report”)). In the Danish trial, 50 percent of the participants were excited to be part of the study and inspired to eat more vegetarian-based food, but they also missed familiar meals and felt isolated (D5.3 (“Report on the outcome of each intervention study”). In the Spanish trial, participants criticized time-consuming shopping lists. In the UK and Poland, participants felt cared for by healthcare professionals, and some participants felt better during the dietary intervention than before. However, some mentioned that adhering to the diet was more time-consuming and expensive than their previous diet.
Legal aspects
Personalized nutrition lacks specific legal regulations due to its multifaceted nature (which includes aspects such as advice, testing and foods), making legislation fragmented (Reference Ahlgren, Nordgren and Perrudin52;Reference Röttger-Wirtz and De Boer53). In other words, personalized nutrition interventions can be categorized as “health” or “lifestyle” intervention or “food” or “medicine,” affecting the applicable rules and regulations (Reference Röttger-Wirtz and De Boer53). Röttger-Wirtz and De Boer (Reference Röttger-Wirtz and De Boer53) analyzed food laws and showed for example that, it is often unclear whether certain nutrigenomic or nutrigenetic effects should be classified as health optimizing, health maintaining, or disease preventive effects. Classifying it as disease preventive, results, for example, in regulating the intervention as a medicinal product, rather than governed by food laws.
There are legal requirements that apply to all personalized nutrition interventions, including the General Data Protection Regulation (GDPR) for personal data. GDPR guidelines were prioritized in the PREVENTOMICS interventions by ensuring anonymization. Moreover, CE marking is required under the current medical device regulation for the European market, as interventions like PREVENTOMICS are classified as in vitro diagnostic medical devices (54). For more details, see Supplementary material S4.
Discussion
This study aimed to assess the PREVENTOMICS interventions in a pre-market phase with the HTA Core Model to inform development and implementation decisions. Conducting an “early HTA” is an effective method to identify and address potential issues regarding market access and reimbursement (Reference Ijzerman, Koffijberg, Fenwick and Krahn55). The different domains showed that approaches like PREVENTOMICS to reduce overweight and obesity are needed. Moreover, people express willingness to use these interventions (Reference Galekop, Veldwijk, Uyl-de Groot and Redekop34), though certain groups (i.e., white individuals) exhibit a higher likelihood of genetic testing than others (i.e., ethnic minority individuals) (Reference Harris, McCabe and Shafique51). Furthermore, our findings indicate that PREVENTOMICS interventions entail low safety risks and require minimal training. While their implementation may require some challenges at the organizational level, the trials showed that they are resolvable.
PREVENTOMICS interventions could have favorable effectiveness results; small short-term effects observed during the trials could translate into long-term health benefits (Reference Bush, Blumberg and El-Sohemy1;Reference Jardim, Mozaffarian and Abrahams-Gessel2). Results align with other studies; see Aldubayan et al. (Reference Aldubayan, Pigsborg and Gormsen26) for comparison of PREVENTOMICS effectiveness results with other studies. Additionally, Galekop et al. (Reference Galekop, Uyl-de Groot and Redekop56) found that personalized nutrition interventions often led to incremental QALYs between 0 and 0.1, comparable with our study findings. While the effects observed are small, most effects are clinically meaningful (requiring a minimum 0.03 difference in utility score) (Reference Drummond57;Reference Bergmo58). However, in Spain, short-term effects resulted in minimal long-term benefits for both PP and PN interventions compared to control (incremental QALYs of 0.002 and 0.006, respectively), contrasting with other countries where incremental QALYs were at least 0.01. Between country differences may stem from the diverse interventions and populations, including cultural differences and targeted weight classifications. For example, Aune et al. (Reference Aune, Sen and Prasad59) demonstrated a J-shaped relationship between BMI and all-cause mortality, potentially explaining the lower effect observed in Spain, which encompasses the general population, including those with normal weight, unlike other countries where studies focused on people with overweight and obesity.
Although clinical trials on technology-based and personalized nutrition interventions often feature small sample sizes and short follow-ups (Reference Shyam, Lee and Tan7;Reference Raaijmakers, Pouwels, Berghuis and Nienhuijs60), leading to effectiveness and parameter uncertainties in cost-effectiveness analyses, Hogervorst et al. (Reference Hogervorst, Vreman, Mantel-Teeuwisse and Goettsch61) suggested improving data quality and quantity to reduce uncertainty, which for PREVENTOMICS interventions could be achieved by longer and larger trials. Our cost-effectiveness analyses explored the potential health benefits of the interventions in the scenario analyses and revealed promising cost-effectiveness results for the interventions in Spain, the UK, and Poland.
The use of PREVENTOMICS interventions would likely increase both short-term and lifetime costs, which raises various questions. First, our findings support the literature indicating that personalized nutrition is more often used by motivated and wealthier individuals (Reference Röttger-Wirtz and De Boer53), particularly when out-of-pocket payments are required. This raises ethical concerns, as personalized nutrition can exacerbate health inequality, given that individuals with lower socioeconomic status often have poorer diets and higher disease burdens but may struggle to afford these interventions (Reference Mathers49). Therefore, third-party reimbursement for effective personalized nutrition interventions is crucial. However, budget constraints may prevent decision-makers to reimburse interventions for the whole target population. It may therefore be advisable to consider reimbursing effective personalized nutrition interventions only for subpopulations with the highest health or economic burden (e.g., severely obese) (Reference Cawley, Meyerhoefer, Biener, Hammer and Wintfeld62). Alternatively, partial subsidies could be provided, covering specific components of the interventions, such as testing or mobile app costs.
Additionally, we recommend that stakeholders, such as policymakers, should collaborate to develop a cohesive legal framework that fosters consumer trust, engagement and enables personalized nutrition to reach its full potential (Reference Röttger-Wirtz and De Boer53;63). Furthermore, policymakers, together with developers, should focus on addressing the concerns of ethnic minority individuals, specifically regarding employment repercussions of genetic tests (Reference Harris, McCabe and Shafique51), ensuring inclusivity and avoiding exclusion due to information shortages. Moreover, despite the ending of the EUNetHTA Joint Actions by September 2023, collaboration on HTAs is recommended between countries to keep track of the fast-changing field of personalized nutrition and to produce timely HTA information for decision-makers. The new “regulation on HTA” is expected to support this future collaboration (64).
This HTA has several limitations. First, as the HTA Core Model was not designed for personalized nutrition interventions (25), additional domains or assessment elements may be needed. Becla et al. (Reference Becla, Lunshof and Gurwitz65) highlighted the importance of ethical, organizational, social, and legal aspects in personalized health care and suggested rethinking the “gold standard” of large trials and instead considering “personal evidence.” Moreover, Von Huben et al. (Reference Von Huben, Howell, Howard, Carrello and Norris66) identified inconsistencies in current HTA frameworks for digital health tools, suggesting the inclusion of digital-specific content in existing or new elements of the HTA Core Model. More specifically, potential additions to the HTA assessment of PREVENTOMICS interventions could be the consideration of device features like size, battery life, operating system, technical support, and connectivity (assessment element ID B0007 should be modified). Moreover, adding new assessment elements could be considered, for example, DHT08 in the safety domain (Reference Von Huben, Howell, Howard, Carrello and Norris66): “how well are updates/continuity of digital health technologies managed?” While we believe all essential aspects are covered in our HTA, future research should analyze more aspects for a more comprehensive overview of digital tools in personalized nutrition interventions.
Second, we obtained expert opinions in this HTA without a systematic approach and we did not fully follow the recommended EUnetHTA methodological framework. Nonetheless, we believe that our approach identified the most critical issues in personalized nutrition interventions.
Third, this HTA primarily focused on BMI as (short-term) outcome measures, but other health outcomes such as waist circumference, blood glucose, systolic blood pressure, or LDL cholesterol might even be more important (67;Reference Millstein68). However, there is limited literature on translating short-term changes in these outcomes into lifetime estimates of disease risk, health outcomes, and costs (Reference Hoogendoorn, Galekop and van Baal32).
In addition to previously mentioned future research suggestions, another recommendation is to extend this HTA by using multiple criteria decision analysis (MCDA) to systematically evaluate and rank ideas based on weighted criteria (Reference Baltussen, Marsh and Thokala69). Since MCDA can identify the relative importance of different criteria, this method can help to maximize societal value when resources are allocated (Reference Baltussen, Marsh and Thokala69).
Conclusion
In conclusion, our HTA emphasizes the relevance of evaluating personalized nutrition interventions beyond costs, effects, and economic aspects by addressing different (related) issues. While PREVENTOMICS interventions exhibit potential (cost)-effectiveness, developers should prioritize gathering additional evidence through longer and larger-scale trials. Addressing organizational issues and early discussions with third-party payers about reimbursement options are recommended for developers. Additionally, policymakers, together with developers, should work on collecting and providing accessible and comprehensive information (e.g., on genetic testing) for all ethnic groups. Moreover, a cohesive legal framework and a system-wide collaboration among stakeholders, including European HTA, are needed, prior to making implementation decisions.
Supplementary material
The supplementary material for this article can be found at http://doi.org/10.1017/S0266462324000060.
Funding statement
This study was funded by the European Union’s Horizon 2020 research and innovation program (grant number 818318). The HTA Core Model is a registered trademark, with use subject to a license that we acquired.
Competing interest
The authors declare none.
Open access
