Establishing causality in nutritional science is a challenge. Complexity of human diet and behaviors, substantial interindividual variability in metabolic responses to dietary exposures, and variations in foods and food products are just a few reasons for this challenge (Vitolins & Case, Reference Vitolins and Case2020). Most nutrition literature relies on observational studies, but these studies are not at the top of the hierarchy of evidence. Cumulative biases, with extensive residual confounding, unclear cause-and-effect relationships, and measurement errors in dietary data, which are often self-reported and imperfectly measured, are among their key limitations (Ioannidis, Reference Ioannidis2018; Jacobs Jr., 2023). Randomized controlled trials (RCTs) are considered the gold standard for causal inference in clinical and public health research (Collins et al., Reference Collins, Bowman, Landray and Peto2020; Okeh & Ugwu, Reference Okeh and Ugwu2009). However, their application in nutrition is often limited due to high costs, logistical burdens, and ethical constraints. The effect sizes in nutrition RCTs are commonly modest. Therefore, they frequently require large sample sizes to achieve adequate statistical power, which complicates trial feasibility (Jadad, Reference Jadad1998; Reeves et al., Reference Reeves, Van Binsbergen and Van Weel2005; Stolberg et al., Reference Stolberg, Norman and Trop2004).
Combinations of bioactive compounds in human diet, influenced by cultural, behavioral, and environmental factors as part of human variations, make nutritional research even more complicated (Drewnowski & Kawachi, Reference Drewnowski and Kawachi2015; González, Reference González2020). In addition, causal interpretation in nutrition is often hindered by a lack of formal training in causal inference, leading to overreliance on correlational findings from large datasets (Chiolero, Reference Chiolero2018). Another obstacle in nutrition research is the heterogeneity in individual responses to dietary interventions. For instance, data from the POSITIVe network, a European research consortium investigating plant bioactives, demonstrated up to a 37-fold difference in metabolic responses to orange juice among participants. These differences were explained by the variation in absorption, digestion, and excretion pathways (Manach et al., Reference Manach, Milenkovic, Van de Wiele, Rodriguez-Mateos, de Roos, Garcia-Conesa, Landberg, Gibney, Heinonen, Tomás-Barberán and Morand2017). Also, the findings in the PREDICT 1 study, a cohort of 1002 twins and unrelated adults in the United Kingdom, showed substantial interindividual variability in postprandial triglyceride, glucose, and insulin responses following standardized meals (Berry et al., Reference Berry, Valdes, Drew, Asnicar, Mazidi, Wolf, Capdevila, Hadjigeorgiou, Davies, Al Khatib, Bonnett, Ganesh, Bakker, Hart, Mangino, Merino, Linenberg, Wyatt, Ordovas and Spector2020). Factors that may influence this variability include sex, age, BMI, habitual diet, microbiome composition, and genetic polymorphisms (de Roos et al., Reference de Roos, Aura, Bronze, Cassidy, Conesa, Gibney, Greyling, Kaput, Kerem, Knežević, Kroon, Landberg, Manach, Milenkovic, Rodriguez-Mateos, Tomás-Barberán, van de Wiele and Morand2019; Favari et al., Reference Favari, de Alvarenga, Sánchez-Martínez, Tosi, Mignogna, Cremonini, Manach, Bresciani, Del Rio and Mena2024; Head & Buckley, Reference Head and Buckley2020).
Twin studies are a potential methodological tool for addressing these challenges. Monozygotic (MZ) twins are genetically identical, while dizygotic (DZ) twins share, on average, 50% of their genes. This characteristic provides the opportunity for researchers to distinguish the genetic influences from environmental exposures (Boomsma et al., Reference Boomsma, Busjahn and Peltonen2002; Pallister et al., Reference Pallister, Spector and Menni2014; van Dongen et al., Reference van Dongen, Slagboom, Draisma, Martin and Boomsma2012). Utilizing international twin registries such as the CODATwins Project has been a step toward conducting large-scale, diverse, and longitudinal investigations into health outcomes and their determinants in twin populations (Jelenkovic et al., Reference Jelenkovic, Sund, Hur, Yokoyama, Hjelmborg, Möller, Honda, Magnusson, Pedersen, Ooki, Aaltonen, Stazi, Fagnani, D’Ippolito, Freitas, Maia, Ji, Ning, Pang and Silventoinen2016; Jelenkovic et al., Reference Jelenkovic, Yokoyama, Sund, Pietiläinen, Hur, Willemsen, Bartels, Aaltonen, Ji, Ning, Pang, Rebato, Busjahn, Kandler, Saudino, Jang, Cozen, Hwang, Swan and Silventoinen2017; Silventoinen et al., Reference Silventoinen, Jelenkovic, Sund, Honda, Aaltonen, Yokoyama, Magnusson, Pedersen, Ooki, Ji, Ning, Pang, Rebato, Busjahn, Kandler, Saudino, Jang, Cozen, Hwang and Kaprio2015; Silventoinen et al., Reference Silventoinen, Jelenkovic, Sund, Hur, Yokoyama, Honda, Magnusson, Pedersen, Ooki, Aaltonen, Ji, Ning, Pang, Rebato, Busjahn, Kandler, Saudino, Jang, Cozen and Kaprio2016). Among the most valuable applications of the twin design is the co-twin control method. In this design, one twin is randomized to an intervention while the other serves as a genetically matched control. With this approach, both measured and unmeasured confounding decrease, and internal validity and statistical power increase even in small samples (Martin et al., Reference Martin, Carr, Oakeshott and Clark1982; McAdams et al., Reference McAdams, Rijsdijk, Zavos and Pingault2021; McGue et al., Reference McGue, Osler and Christensen2010; Plomin & Haworth, Reference Plomin and Haworth2010).
Advantages of the twin study design in nutrition RCTs are clear, and particularly beneficial for studies on genetic or early-life environment. However, twin designs remain underutilized in nutrition research (Craig et al., Reference Craig, Calais-Ferreira, Umstad and Buchwald2020). Existing twin studies predominantly use observational designs or assign twin pairs to the same group rather than using discordant co-twin interventions (Sumathipala et al., Reference Sumathipala, Yelland, Green, Shepherd, Jayaweera, Ferreira and Craig2018; Trepanowski & Ioannidis, Reference Trepanowski and Ioannidis2018). With the shift in nutrition science toward personalization and precision, co-twin randomized trials can be an efficient design to determine whether measured outcomes are caused by the intervention itself or by biological differences (de Roos & Brennan, Reference de Roos and Brennan2017).
This systematic review aims to evaluate randomized controlled trials using a co-twin design in nutrition science. In this review, we will synthesize methodological approaches and study characteristics of existing twin-based RCTs, with attention to their strengths, challenges, and implications for future research designs in nutrition science.
Methods
Study Design
This scoping review was conducted and reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews (PRISMA-ScR; Tricco et al., Reference Tricco, Lillie, Zarin, O’Brien, Colquhoun, Levac, Moher, Peters, Horsley, Weeks, Hempel, Akl, Chang, McGowan, Stewart, Hartling, Aldcroft, Wilson, Garritty, Lewin, Godfrey, Macdonald, Langlois, Soares-Weiser, Moriarty, Clifford, Tunçalp and Straus2018).
Eligibility Criteria
Studies were eligible for inclusion if they met all of the following criteria: (1) the study employed a randomized controlled trial design; (2) participants included twin pairs, with co-twins allocated to different intervention arms (i.e., a co-twin control design); (3) the intervention involved a dietary or nutritional component, including whole-diet approaches, macronutrient manipulation, or dietary supplementation; and (4) at least one health-related outcome was reported. No restrictions were placed on participant age, sex, zygosity, health status, or geographic location. Studies were excluded if: (1) they did not use an interventional design (e.g., observational, cross-sectional, or cohort studies); (2) both twins within a pair were assigned to the same treatment arm; (3) the intervention was not dietary or nutritional in nature; (4) the study was published only as a conference abstract without sufficient methodological detail; (5) full-text was unavailable; (6) the study was conducted in animals; (7) the study examined the effects of breastfeeding as the primary dietary exposure; or (8) the study focused on the effects of maternal diet on the birth outcomes of twins (i.e., the twins were the offspring rather than the participants receiving the intervention). There were no language or date restrictions applied.
Information Sources and Search Strategy
A systematic search was conducted on July 1, 2024, across four electronic databases: PubMed (MEDLINE), CINAHL, Web of Science, and Scopus. To ensure currency of the evidence, a search update was conducted in March 2026 using the same search strategies across all four databases. Search strategies were developed using a combination of controlled vocabulary terms and free-text keywords related to two core concepts: twin study designs and dietary interventions. The PubMed strategy combined MeSH terms (Twins[Mesh]; Diet[Mesh]) with title and abstract keyword searches (twin[tiab]; diet[tiab]). Analogous strategies were applied in CINAHL using subject headings (MH Twins+; MH Diet+) and title/abstract fields, and in Web of Science and Scopus using topic and title-abstract-keyword fields respectively. Boolean operators (AND, OR) were used to combine search terms across concepts. Full search strings for all databases are provided in Supplementary File 1. Reference lists of included studies were also hand-searched to identify any additional eligible records not captured by the database searches.
Study Selection
All records retrieved from the database searches were imported into Covidence (Veritas Health Innovation, Melbourne, Australia), a web-based systematic review management platform, where duplicate records were automatically identified and removed. Study selection proceeded in two stages. In the first stage, two reviewers independently screened all titles and abstracts against the eligibility criteria. In the second stage, full texts of potentially eligible records were retrieved and independently assessed by the same two reviewers. Disagreements at either stage were resolved through discussion and consensus between the reviewers. Reasons for exclusion at the full-text stage were recorded. The study selection process is summarised in a PRISMA-ScR flow diagram (Figure 1).
PRISMA diagram to illustrate the literature search process and the resulting number of reviewed articles.
Figure 1. Long description
The flowchart illustrates the literature search process and the resulting number of reviewed articles. It starts with the identification phase, where studies from databases/ registers totaling 5405 are identified. References removed include 1930 duplicates identified manually, by Covidence, and by automation tools. The screening phase begins with 3475 studies screened, of which 3181 are excluded. 294 studies are sought for retrieval, with none not retrieved. These 294 studies are assessed for eligibility, with 243 studies excluded for various reasons such as being animal studies, not randomized, wrong study design, conference abstracts, and not being nutrition/diet-related. 51 full-text articles are assessed in detail for final eligibility, with 38 studies excluded for reasons like both twins receiving the same intervention, duplicates, and not having a twin population. Finally, 13 studies are included in the review.
Data Extraction
Data were extracted from all included studies by two reviewers independently using a prespecified data extraction form developed a priori. The following data items were extracted from each included study: study number; study name; year of publication; study purpose; diets or dietary exposures compared; intervention description; intervention duration; setting (e.g., free-living, clinical, metabolic ward); study design; outcome measures; number of twin pairs enrolled; zygosity (MZ, DZ, or both); recruitment source; health status of participants; age group (child, 0–18 years; adult; or older adult); participant age; study location; sex ratio (male:female); race and ethnicities; exclusion criteria; outcomes reported; study limitations; considerations specific to the use of a twin design; future research directions noted by the authors; and any additional notes. Any discrepancies in extracted data were resolved by discussion between the two reviewers. A formal quality assessment or risk of bias appraisal of included studies was not performed, consistent with the scoping review methodology, which aims to map the breadth of available evidence rather than appraise its quality (Munn et al., Reference Munn, Peters, Stern, Tufanaru, McArthur and Aromataris2018).
Data Synthesis
Given the heterogeneity of study designs, intervention types, participant populations, and outcome measures across included studies, a narrative synthesis approach was adopted. Findings were summarized descriptively and organized thematically according to intervention type (e.g., whole-diet interventions, macronutrient manipulation, dietary supplementation, behavioral interventions) and primary outcome domain (e.g., cardiometabolic health, bone mineral density, sensory perception, epigenetic aging, cognition). Study characteristics were tabulated to facilitate comparison across studies. No quantitative pooling or meta-analysis was conducted.
Results
A total of 5405 records were identified and imported for screening. Following the removal of duplicates, 3475 studies were screened by title and abstract. After excluding clearly irrelevant studies, 294 full-text articles were assessed for eligibility. The majority of full-text exclusions were due to the intervention not being diet- or nutrition-related, being conference abstracts, or not involving twin participants. After full-text screening, an additional 38 studies were excluded for reasons such as identical interventions administered to both twins, observational design, or ineligible population (Figure 1).
A total of 13 studies met the inclusion criteria for this scoping review, published between 1987 and 2025. These studies were conducted across various geographic locations, primarily the United States and Australia, with two based in Europe. The interventions varied in duration, from two study visits to longer periods lasting up to three years. In terms of the type of intervention, we can categorize them in four groups, including dietary pattern interventions, nutrient/supplement-based interventions, dietary component modification interventions, and behavioral or multimodal interventions. A detailed overview of the included studies, covering study design, intervention details, population demographics, and outcome measures, is presented in Table 1.
Summary of twin studies utilizing a co-twin randomized controlled trial design
Note: Key characteristics of included studies, including intervention details, study design, participant characteristics, and outcome measurements.
MZ, monozygotic; DZ, dizygotic; RCT, randomized controlled trial; BMI, body mass index; LDL-C, low-density lipoprotein cholesterol; HDL-C, high-density lipoprotein cholesterol; DXA, dual-energy X-ray absorptiometry; HEI-2015, Healthy Eating Index–2015.
Participant Characteristics and Recruitment
Participants’ characteristics and recruitment strategies varied across the studies. Twin pairs were primarily recruited from national twin registries, such as Twins Research Australia (TRA), the TwinsUK Registry, and the Stanford Twin Registry (STR), as well as through local advertisements, community outreach, university records, and school-based initiatives. Healthy populations were targeted, and most studies had explicit exclusion criteria to screen out individuals with chronic conditions or factors that could interfere with adherence to the intervention or protocol (e.g., recent antibiotic use, pregnancy, severe allergies, or metabolic diseases).
Participants ranged in age from as young as 5 years to over 60. Some studies focused on specific life stages such as premenarcheal girls (Cameron et al., Reference Cameron, Paton, Nowson, Margerison, Frame and Wark2004) or older adults with low protein intake (Ní Lochlainn et al., Reference Ni Lochlainn, Bowyer, Moll, García, Wadge, Baleanu, Nessa, Sheedy, Akdag, Hart, Raffaele, Seed, Murphy, Harridge, Welch, Greig, Whelan and Steves2024). In most included studies, both sexes were recruited, except for two, which focused exclusively on female participants due to sex-specific research aims (Cameron et al., Reference Cameron, Paton, Nowson, Margerison, Frame and Wark2004; Nowson et al., Reference Nowson, Green, Hopper, Sherwin, Young, Kaymakci, Guest, Smid, Larkins and Wark1997) and one study that included only male participants (Faith et al., Reference Faith, Rose, Matz, Pietrobelli and Epstein2006). Reporting of sex ratios and racial/ethnic backgrounds varied across studies. Zygosity composition also differed: seven studies enrolled only MZ twins, while six included both MZ and DZ pairs. Sample sizes ranged from 3 to 70 twin pairs.
Interventions
The 13 included trials used four broad types of dietary interventions. Dietary pattern interventions, defined as assignment to distinct overall dietary patterns (e.g., vegan vs. omnivorous diets; Carter et al., Reference Carter, Zeng, Ward, Landry, Perelman, Hennings, Meng, Weakley, Cabrera, Robinson, Nguyen, Higginbottom, Maecker, Sonnenburg, Fischbach, Gardner and Sonnenburg2025; Dwaraka et al., Reference Dwaraka, Aronica, Carreras-Gallo, Robinson, Hennings, Carter, Corley, Lin, Turner, Smith, Mendez, Went, Ebel, Sonnenburg, Sonnenburg and Gardner2024; Landry et al., Reference Landry, Ward, Cunanan, Durand, Perelman, Robinson, Hennings, Koh, Dant, Zeitlin, Ebel, Sonnenburg, Sonnenburg and Gardner2023; Zeitlin et al., Reference Zeitlin, Ward, Cooper, Landry, Krenek, Durand, Cunanan, Robinson, Dant and Gardner2025); nutrient- or supplement-based interventions, involving the addition of specific nutrients or bioactive compounds (e.g., calcium or botanical extracts; Cameron et al., Reference Cameron, Paton, Nowson, Margerison, Frame and Wark2004; Johnston et al., Reference Johnston, Miller, Slemenda, Reister, Hui, Christian and Peacock1992; Nowson et al., Reference Nowson, Green, Hopper, Sherwin, Young, Kaymakci, Guest, Smid, Larkins and Wark1997, Perrinjaquet-Moccetti et al., Reference Perrinjaquet-Moccetti, Busjahn, Schmidlin, Schmidt, Bradl and Aydogan2008); dietary component modification, in which a single dietary component such as fat or sodium intake was systematically altered while maintaining overall dietary structure (Costanzo et al., Reference Costanzo, Liu, Nowson, Duesing, Archer, Bowe and Keast2019; Costanzo et al., Reference Costanzo, Nowson, Orellana, Bolhuis, Duesing and Keast2018, Luft et al. Reference Luft, Miller, Weinberger, Grim, Daugherty and Christian1987); and behavioral or multimodal interventions, combining dietary components with behavioral strategies or additional modalities such as exercise (Faith et al., Reference Faith, Rose, Matz, Pietrobelli and Epstein2006, Ní Lochlainn et al., Reference Ni Lochlainn, Bowyer, Moll, García, Wadge, Baleanu, Nessa, Sheedy, Akdag, Hart, Raffaele, Seed, Murphy, Harridge, Welch, Greig, Whelan and Steves2024).
The most recent studies in this review were dietary pattern interventions that used randomized controlled trials to compare vegan and omnivorous dietary patterns over 8 weeks. In the initial phase, participants typically started with structured meal delivery and then self-prepared meals. This approach supports dietary adherence while also reflecting how these diets might realistically be adopted in everyday life (Carter et al., Reference Carter, Zeng, Ward, Landry, Perelman, Hennings, Meng, Weakley, Cabrera, Robinson, Nguyen, Higginbottom, Maecker, Sonnenburg, Fischbach, Gardner and Sonnenburg2025; Dwaraka et al., Reference Dwaraka, Aronica, Carreras-Gallo, Robinson, Hennings, Carter, Corley, Lin, Turner, Smith, Mendez, Went, Ebel, Sonnenburg, Sonnenburg and Gardner2024; Landry et al., Reference Landry, Ward, Cunanan, Durand, Perelman, Robinson, Hennings, Koh, Dant, Zeitlin, Ebel, Sonnenburg, Sonnenburg and Gardner2023; Zeitlin et al., Reference Zeitlin, Ward, Cooper, Landry, Krenek, Durand, Cunanan, Robinson, Dant and Gardner2025). Dietary intake was assessed using repeated 24-hour dietary recalls to evaluate adherence to assigned dietary patterns (Landry et al., Reference Landry, Ward, Cunanan, Durand, Perelman, Robinson, Hennings, Koh, Dant, Zeitlin, Ebel, Sonnenburg, Sonnenburg and Gardner2023; Zeitlin et al., Reference Zeitlin, Ward, Cooper, Landry, Krenek, Durand, Cunanan, Robinson, Dant and Gardner2025). Outcomes in the primary analyses were primarily related to cardiometabolic risk factors (Landry et al., Reference Landry, Ward, Cunanan, Durand, Perelman, Robinson, Hennings, Koh, Dant, Zeitlin, Ebel, Sonnenburg, Sonnenburg and Gardner2023) and a subpaper provided detailed information about dietary quality (Zeitlin et al., Reference Zeitlin, Ward, Cooper, Landry, Krenek, Durand, Cunanan, Robinson, Dant and Gardner2025). Multiple subanalyses from this trial extended the primary findings; for example, one integrated detailed biological profiling, including metabolomic, cytokine, and gut microbiome analyses, to examine potential mechanistic pathways linking dietary patterns to metabolic regulation (Carter et al., Reference Carter, Zeng, Ward, Landry, Perelman, Hennings, Meng, Weakley, Cabrera, Robinson, Nguyen, Higginbottom, Maecker, Sonnenburg, Fischbach, Gardner and Sonnenburg2025), while the other focused on DNA methylation-based epigenetic aging markers to evaluate changes in biological age following the dietary intervention (Dwaraka et al., Reference Dwaraka, Aronica, Carreras-Gallo, Robinson, Hennings, Carter, Corley, Lin, Turner, Smith, Mendez, Went, Ebel, Sonnenburg, Sonnenburg and Gardner2024).
Nutrient- and supplement-based interventions were another major methodological category. These interventions typically used placebo-controlled or dose-comparison designs. Calcium supplementation trials used placebo-controlled co-twin designs over extended durations ranging from 6 months to 3 years in the pediatric population. These studies showed greater gains in bone mineral density in supplemented twin compared with the placebo-treated co-twin (Cameron et al., Reference Cameron, Paton, Nowson, Margerison, Frame and Wark2004; Johnston et al., Reference Johnston, Miller, Slemenda, Reister, Hui, Christian and Peacock1992; Nowson et al., Reference Nowson, Green, Hopper, Sherwin, Young, Kaymakci, Guest, Smid, Larkins and Wark1997). Perrinjaquet-Moccetti et al. (Reference Perrinjaquet-Moccetti, Busjahn, Schmidlin, Schmidt, Bradl and Aydogan2008) evaluated the effects of two doses of olive leaf extract in MZ twins with borderline hypertension using a dose-comparison design.
Another group of studies used dietary component modification methods, such as controlled low-fat versus high-fat diets. These studies were designed to maintain weight stability while assessing molecular and sensory adaptations, including taste perception and gene expression responses. Based on the results, dietary fat modification influenced gene expression and taste sensitivity independent of genetic background (Costanzo et al., Reference Costanzo, Liu, Nowson, Duesing, Archer, Bowe and Keast2019; Costanzo et al., Reference Costanzo, Nowson, Orellana, Bolhuis, Duesing and Keast2018). Another example is the sodium restriction protocol implemented by Luft et al. (Reference Luft, Miller, Weinberger, Grim, Daugherty and Christian1987), in which dietary sodium intake was reduced and subsequently reintroduced to examine blood pressure responses within twin pairs. The co-twin design showed that differences in blood pressure response to salt were likely due to genetic differences, not just environmental factors.
Behavioral and multimodal interventions represented a smaller but distinct category of twin-based trials. One of these interventions used reward systems to modify dietary behaviors, to increase fruit and vegetable intake. This approach increased fruit and vegetable intake by 2.0 servings/day (Faith et al., Reference Faith, Rose, Matz, Pietrobelli and Epstein2006). Multimodal interventions combined dietary supplementation with an additional lifestyle component, resistance exercise. This intervention was implemented using remotely monitored or home-based protocols. In this study with twins sharing lifestyle conditions, the effect of prebiotic supplementation was isolated, resulting in improvements in cognitive function and gut microbiota composition (Ní Lochlainn et al., Reference Ni Lochlainn, Bowyer, Moll, García, Wadge, Baleanu, Nessa, Sheedy, Akdag, Hart, Raffaele, Seed, Murphy, Harridge, Welch, Greig, Whelan and Steves2024).
Discussion
Following a comprehensive search of clinical trial registries and multiple databases, only 13 studies met our inclusion criteria, suggesting that co-twin randomized dietary interventions are rare, especially when compared with standard studies using nontwins or twins assigned to the same group. Twins can be incorporated into clinical trials using three different approaches. They may be randomized to the same treatment group, independently assigned without consideration of their twin pair, or intentionally allocated to different groups, as in the co-twin control design. Each approach offers unique methodological advantages and limitations. Randomizing both twins to the same group can help them follow the study protocol better and minimize variability in treatment delivery, but it limits direct comparison within pairs (Segal, Reference Segal2017). Independent randomization treats each twin as an individual. This approach increases generalizability but potentially neglects shared genetic and environmental factors that are unique characteristics in the twin population (van Dongen et al., Reference van Dongen, Slagboom, Draisma, Martin and Boomsma2012). The co-twin control design, in contrast, enhances internal validity, effectively controls for many confounding variables, and allows more precise causal inference (Polderman et al., Reference Polderman, Benyamin, De Leeuw, Sullivan, Van Bochoven, Visscher and Posthuma2015). However, this design also faces challenges, such as limited generalizability, potential carryover effects between twins, and ethical concerns about deliberately assigning different treatments within closely bonded pairs (Tan et al., Reference Tan, Christiansen, von Bornemann Hjelmborg and Christensen2015).
Zygosity and whether the sample includes MZ-only or mixed MZ/DZ twin pairs is another important factor in co-twin designs, because each design may offer different inferential value. MZ twins, who share nearly all of their segregating genetic variation, provide the strongest control for genetic confounding among twin designs. Therefore, MZ-only samples can provide a unique opportunity to reduce genetic confounding and more clearly estimate the effect of dietary interventions by minimizing the influence of inherited biological differences (Boomsma et al., Reference Boomsma, Busjahn and Peltonen2002; McAdams et al., Reference McAdams, Rijsdijk, Zavos and Pingault2021; van Dongen et al., Reference van Dongen, Slagboom, Draisma, Martin and Boomsma2012). However, MZ-only trials may be more difficult to recruit and may have limited generalizability because they represent a narrower subgroup of the population. In contrast, mixed MZ/DZ samples may improve feasibility, increase sample size, and better represent the broader twin population. However, because DZ twins share only about 50% of their segregating genes on average, mixed samples provide less complete genetic control than MZ-only designs and may require stratified or sensitivity analyses by zygosity when sample size allows (Boomsma et al., Reference Boomsma, Busjahn and Peltonen2002; Polderman et al., Reference Polderman, Benyamin, De Leeuw, Sullivan, Van Bochoven, Visscher and Posthuma2015; van Dongen et al., Reference van Dongen, Slagboom, Draisma, Martin and Boomsma2012). The choice between MZ-only and mixed MZ/DZ designs should be aligned with the study objective.
Because of the shared genetic and environmental factors among twins, when they are utilized in nutritional RCTs, many confounding factors are controlled by nature. In addition, comparing co-twins can reduce between-person variability and strengthen internal validity (McAdams et al., Reference McAdams, Rijsdijk, Zavos and Pingault2021; McGue et al., Reference McGue, Osler and Christensen2010). Several studies in this review had relatively small sample sizes, but they were able to detect meaningful intervention-related differences. For example, Landry et al. (Reference Landry, Ward, Cunanan, Durand, Perelman, Robinson, Hennings, Koh, Dant, Zeitlin, Ebel, Sonnenburg, Sonnenburg and Gardner2023) compared healthy vegan and omnivorous diets in 22 identical twin pairs. They reported a greater reduction in LDL-C in the vegan group compared with the omnivorous group after 8 weeks (−13.9 mg/dL). In comparison, the NEW Soul trial, as a conventional nontwin RCT with a similar exposure and outcome, randomized 159 African-American adults with overweight or obesity to a vegan or low-fat omnivorous diet. They found no significant between-group differences in weight or lipid outcomes, including LDL-C, at 12 months (−2.56 mg/dL in the vegan group vs. −0.79 mg/dL in the omnivorous group; Turner-McGrievy et al., Reference Turner-McGrievy, Wilcox, Frongillo, Murphy, Hutto, Wilson, Davey, Bernhart, Okpara, Bailey and Hu2023). It should be considered that these two studies differed in population, intervention design, and outcome measurements, and this is just an indirect comparison.
The current evidence does not prove that co-twin RCTs produce larger effect sizes with smaller sample sizes compared with nontwin studies. There are examples that show almost the same power using these two designs. Nowson et al. (Reference Nowson, Green, Hopper, Sherwin, Young, Kaymakci, Guest, Smid, Larkins and Wark1997) reported that calcium supplementation in 42 female twin pairs produced a 1.62% greater increase in spine BMD at 18 months compared with placebo. In a conventional RCT, Lloyd et al. (Reference Lloyd, Andon, Rollings, Martel, Landis, Demers, Eggli, Kieselhorst and Kulin1993) randomized 94 adolescent girls and found greater increases in lumbar spine BMD in the calcium group compared with placebo (18.7% vs. 15.8%). Even with twin designs, intervention effects may still be small or inconsistent, as shown in some included studies (Ní Lochlainn et al., Reference Ni Lochlainn, Bowyer, Moll, García, Wadge, Baleanu, Nessa, Sheedy, Akdag, Hart, Raffaele, Seed, Murphy, Harridge, Welch, Greig, Whelan and Steves2024). Twin studies do not guarantee larger effects, but they may help detect intervention-related effects with less confounding variability.
Across the 13 included studies, the co-twin RCT design appeared to offer methodological advantages in three main contexts. First, for outcomes with strong biological or genetic influences, this design helped reduce genetic and shared environmental variability. Therefore, intervention-related differences could be interpreted with greater internal control. Examples include lipid response, blood pressure response, bone mineral density, taste perception, gene expression, and epigenetic markers (Costanzo et al., Reference Costanzo, Liu, Nowson, Duesing, Archer, Bowe and Keast2019; Costanzo et al., Reference Costanzo, Nowson, Orellana, Bolhuis, Duesing and Keast2018; Dwaraka et al., Reference Dwaraka, Aronica, Carreras-Gallo, Robinson, Hennings, Carter, Corley, Lin, Turner, Smith, Mendez, Went, Ebel, Sonnenburg, Sonnenburg and Gardner2024; Johnston et al., Reference Johnston, Miller, Slemenda, Reister, Hui, Christian and Peacock1992; Landry et al., Reference Landry, Ward, Cunanan, Durand, Perelman, Robinson, Hennings, Koh, Dant, Zeitlin, Ebel, Sonnenburg, Sonnenburg and Gardner2023; Nowson et al., Reference Nowson, Green, Hopper, Sherwin, Young, Kaymakci, Guest, Smid, Larkins and Wark1997). Second, when studies use more controlled or structured interventions, the co-twin design adds another layer of control. The calcium supplementation trials are examples of this pattern. These studies used placebo-controlled co-twin designs, standardized calcium doses, objective measurements of bone outcome, and relatively long follow-up periods to examine bone mineral density in children and adolescents (Cameron et al., Reference Cameron, Paton, Nowson, Margerison, Frame and Wark2004; Johnston et al., Reference Johnston, Miller, Slemenda, Reister, Hui, Christian and Peacock1992; Nowson et al., Reference Nowson, Green, Hopper, Sherwin, Young, Kaymakci, Guest, Smid, Larkins and Wark1997). More recently, the vegan versus omnivorous diet trial provided meals during the first 4 weeks before participants transitioned to self-prepared meals, allowed dietary exposure to be more controlled during the early phase of the intervention (Landry et al., Reference Landry, Ward, Cunanan, Durand, Perelman, Robinson, Hennings, Koh, Dant, Zeitlin, Ebel, Sonnenburg, Sonnenburg and Gardner2023). Third, co-twin designs may also be especially useful for mechanistic or exploratory outcomes. In these studies, the goal is not only to test whether a dietary intervention changes a clinical outcome, but also to understand the biological pathways that may explain the effect. For example, Carter et al. (Reference Carter, Zeng, Ward, Landry, Perelman, Hennings, Meng, Weakley, Cabrera, Robinson, Nguyen, Higginbottom, Maecker, Sonnenburg, Fischbach, Gardner and Sonnenburg2025) examined microbiome, metabolomic, and immune markers in twins assigned to vegan or omnivorous diets, while Dwaraka et al. (Reference Dwaraka, Aronica, Carreras-Gallo, Robinson, Hennings, Carter, Corley, Lin, Turner, Smith, Mendez, Went, Ebel, Sonnenburg, Sonnenburg and Gardner2024) examined DNA methylation and epigenetic aging markers from the same intervention. Because these outcomes may be influenced by inherited biological differences, the co-twin design can help reduce genetic variability and make it easier to interpret whether observed biological changes are related to the dietary intervention. Co-twin RCTs may be particularly useful for smaller mechanistic nutrition trials, while large nontwin pragmatic trials may be more appropriate when the main goal is population-level generalizability.
Despite the discussed advantages, twin studies face several methodological challenges. The potential for behavioral contamination between cohabitating twins (when twins influence each other’s behaviors, such as sharing diet habit) is a concern to internal validity, especially in behavioral or dietary interventions (Sumathipala et al., Reference Sumathipala, Yelland, Green, Shepherd, Jayaweera, Ferreira and Craig2018). Ethical concerns and participant preferences may complicate recruitment, adherence, and acceptability in co-twin RCTs. For example, some twins may be less willing to enroll or may find it more difficult to adhere to the protocol when co-twins are assigned to different interventions, although allocation itself should remain determined by the randomized study design (Craig et al., Reference Craig, Calais-Ferreira, Umstad and Buchwald2020). Small sample sizes and recruitment challenges may also reduce statistical power and generalizability (Collins et al., Reference Collins, Bowman, Landray and Peto2020; Sumathipala et al., Reference Sumathipala, Yelland, Green, Shepherd, Jayaweera, Ferreira and Craig2018). Twin populations represent only a small subset of the general population — approximately 2.4% of all live births (J. A. Martin et al., Reference Martin, Hamilton, Osterman, Driscoll and Drake2015), and this makes the identification of eligible twin pairs challenging. Twin registries are valuable resources for efficiently locating and enrolling twins for research. However, not all countries have well-developed twin registries. Without access to these established resources, researchers often rely on less systematic recruitment methods such as public advertising or referrals, which may lead to slower enrollment, selection bias, and less representative samples. Consequently, these recruitment difficulties can hinder the feasibility of adequately powered twin RCTs and limit the ability to conduct large-scale, multicenter studies. International collaborations that leverage existing twin registries across countries could help overcome some of these barriers by expanding the available twin participant pool and enhancing study generalizability. Integration of twin registries into omics technologies (e.g., nutrigenomics and metabolomics) can help researchers better understand cause-and-effect relationships and create more personalized diet recommendations (Boomsma et al., Reference Boomsma, Busjahn and Peltonen2002; de Roos & Brennan, Reference de Roos and Brennan2017; van Dongen et al., Reference van Dongen, Slagboom, Draisma, Martin and Boomsma2012).
Twin-based RCT designs can be compared with the other causal inference approaches used in nutrition research. Mendelian randomization can support causal inference by using genetic variants related to an exposure, which may reduce confounding and reverse causation. However, unlike co-twin RCTs, it does not directly test the effect of assigning participants to a specific dietary intervention, such as a supplement or dietary pattern, under trial conditions (Davey Smith & Ebrahim, Reference Davey Smith and Ebrahim2003; Davies et al., Reference Davies, Holmes and Davey Smith2018). Other within-family designs, including sibling or family-based comparisons, can reduce confounding from shared family background, but they generally provide less complete genetic control than co-twin designs (Brumpton et al., Reference Brumpton, Sanderson, Heilbron, Hartwig, Harrison, Vie, Cho, Howe, Hughes, Boomsma, Havdahl, Hopper, Neale, Nivard, Pedersen, Reynolds, Tucker-Drob, Grotzinger, Howe and Davies2020; McGue et al., Reference McGue, Osler and Christensen2010). Crossover trials also reduce between-person variability because each participant serves as their own control. However, they may be limited by carryover effects, washout requirements, and the need for short-term or reversible interventions (Lim & In, Reference Lim and In2021).
Future directions should include consistent and standardized methods for conducting twin trials, find ways to reduce contamination bias, and compare twin RCTs with traditional trials involving nontwin individuals to statistically optimize power and feasibility in nutritional studies (Ioannidis, Reference Ioannidis2018; Trepanowski & Ioannidis, Reference Trepanowski and Ioannidis2018). Incorporating modern technologies, such as omics analyses, microbiome profiling and digital dietary tracking tools, into co-twin studies can help researchers better understand the many interacting factors that influence how dietary exposures affect health outcomes (McAdams et al., Reference McAdams, Rijsdijk, Zavos and Pingault2021).
This review was conducted as a descriptive scoping review without a meta-analysis and did not include a quantitative synthesis of study outcomes. Our primary focus was on the methodological aspects of the included studies, such as design, sample characteristics, and use of twin models, rather than on evaluating the efficacy of the interventions or synthesizing their results. A major strength of this review is its focused evaluation of co-twin randomized controlled trials, a relatively underutilized design in nutrition research. By emphasizing methodological features, this review provided an overview of how twin designs have been implemented to strengthen causal inference in nutrition science.
Conclusion
This scoping review highlights the unique value and methodological advantages of using twin designs, particularly co-twin control RCTs, in nutritional intervention research. Few studies with this design were identified in the nutrition literature. Co-twin control designs strengthen causal inference and allow for more precise effect estimates by controlling for genetic and shared environmental confounding. Twin studies also present unique challenges, including limited sample sizes, recruitment barriers, and ethical concerns. As nutrition science moves toward precision and personalized approaches, co-twin randomized trials may play an important role in clarifying complex diet–health relationships and interpretability of dietary intervention studies.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/thg.2026.10074.
Data availability statement
All data generated or analyzed during this study are included in this published article and its supplementary materials.
Financial support
Partial funding for publishing this article open access was provided by the University of California Libraries under a transformative open access agreement with the publisher.
Competing interests
None.
Ethical standards
This scoping review was exempt from Institutional Review Board (IRB) review, as it involved the synthesis of previously published, publicly available data and did not include the collection of primary data from human participants.
Open access
