Dietary variety has historically been a component of national and international dietary guidelines because of its association with improved nutritional quality(Reference Kant, Block and Schatzkin1, Reference Kennedy2). Dietary patterns characterised by the consumption of a diverse selection of nutritionally distinct and wholesome foods encourage nutrient adequacy and improved health outcomes(Reference Remick, Polivy and Pliner3). However, more recently, with increasing globalisation and greater availability of highly processed foods, emerging evidence suggests that individuals who consume varied diets are at greater risk of being overweight(Reference Remick, Polivy and Pliner3). Dietary variety improves the palatability of the overall diet, which may enhance total energy consumption and be a contributing factor to the present global obesity epidemic(Reference Remick, Polivy and Pliner3). Although most studies suggest that dietary variety increases energy intake(Reference Raynor4), it is not possible to infer whether dietary variety also has an adverse impact on body weight and other measures of adiposity based on these findings. Physically active individuals must consume more energy, and plausibly, a greater variety of foods to maintain energy balance. Consequently, greater dietary variety may merely be a marker for a more physically active or healthful lifestyle(Reference Willett, Howe and Kushi5).
Most evidence addressing dietary variety and the risk of overweight comes from controlled, short-term feeding studies conducted in both animals and human subjects(Reference McCrory, Suen and Roberts6). These studies consistently demonstrated that the presence of multiple food items is associated with approximately 25 % greater energy intake(Reference McCrory, Suen and Roberts6). Additionally, across animal studies, rats consuming a varied diet consistently gain more weight and accrue more fat mass compared with rats consuming a standard diet(Reference Raynor and Epstein7).
In epidemiological studies, the association between dietary variety and excess adiposity is inconsistent, and varies based on the study design, the definition and measurement of variety, and the study population(Reference Kennedy2, Reference Raynor and Epstein7). Further research investigation into the associations between dietary variety and excess adiposity began after a seminal study revealed a positive correlation between dietary variety in some food groups and body fatness(Reference McCrory, Fuss and McCallum8). This finding generated concern about US national recommendations encouraging dietary variety because of uncertainty whether promoting dietary variety could unintentionally support overconsumption(Reference Dixon, Cronin and Krebs-Smith9).
Despite ambiguity surrounding the relationship between dietary variety and excess adiposity, many dieters choose to restrict broad categories of food when trying to lose weight(Reference Hill10). Diets restricting specific food groups produce initial success because dieters unconsciously reduce energy intake as variety in their diet decreases and eating becomes less pleasurable(Reference Strychar11). However, over time, this monotony can encourage food cravings, which may partly explain the high rate of attrition associated with restrictive diets(Reference Hill10). The potential risks of restricting or limiting dietary variety should be comprehensively evaluated, particularly in the context of renewed scientific interest in manipulating dietary variety for weight loss and weight maintenance(Reference Epstein, Carr and Cavanaugh12, Reference Avena and Gold13).
At present, it is unclear whether diets comprising a variety of lower-energy, nutrient-dense options can increase the likelihood of initiating and maintaining a reduced-energy lifestyle. Given the inconsistencies in the epidemiological literature and the potential negative consequences of limiting variety from both an overall health and a weight maintenance perspective, the purpose of the present systematic review is to examine the evidence examining the associations between dietary variety and measures of adiposity and its consistency across epidemiological studies.
Methods
Protocol
In accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses statement, we systematically searched for cross-sectional, case–control, cohort and experimental studies examining the associations between any measure of dietary variety and measures of body adiposity. The population, intervention, comparator and outcomes method was used to narrow the focus of the research question(Reference Spahn, Lyon and Altman14). The present review focused on healthy adults aged 18 years and older. The interventions and comparators that were examined were exposure to a high- v. low-variety diet. Outcome measures were measures of body adiposity including BMI, waist circumference (WC), waist:hip ratio (WHR) and percentage of body fat. MEDLINE and Thomson Reuters' Web of Science databases were searched for the following key terms: (‘dietary variety’ OR ‘dietary diversity’) AND (‘obesity’ OR ‘body mass index’ OR ‘body fat’ OR ‘waist circumference’ OR ‘body adiposity’ OR ‘body weight’ OR ‘overweight’). No exclusion criteria were used in the search strategy to prevent potential misclassification of relevant articles. This review includes all studies published from January 1999 to June 2012.
Inclusion and exclusion criteria
Studies eligible to be included in the present review were published in English and evaluated the associations between dietary variety and at least one of the following outcomes: (1) body weight or BMI or (2) measures of body adiposity (WC, WHR and percentage of body fat). These markers were selected based on their clinical and epidemiological relevance. Studies were conducted in human subjects and utilised quantitative methods to examine the associations between dietary variety and the aforementioned outcome measures. Because the relationship between dietary variety is highly associated with overall health in developing countries with greater food insecurity, only studies conducted in medium-to-very-high Human Development Index countries were included in the present review. The Human Development Index is a newer and more comprehensive measure of the development status of a country and includes factors such as life expectancy, educational attainment and income(15). Additionally, only studies published from 1999 onwards were included in the present review as a concern regarding the associations between dietary variety and excess energy consumption became a greater research focus after McCrory et al. (1999)(Reference McCrory, Fuss and McCallum8) noted positive correlations between dietary variety in some food groups, energy intake and measures of body adiposity. Observational studies were excluded if they were insufficiently powered (n< 200) and if the measurement of dietary variety was not selected a priori (i.e. cluster analyses were excluded). Finally, studies were excluded if they were conducted in unhealthy populations or among adults younger than 18 years of age.
Database searches generated a total of 2229 abstract titles of which 1681 were eliminated from the title alone because they did not pertain to dietary patterns, dietary variety or body weight. The review of the remaining 548 abstracts resulted in eighty-three unique references requiring full-text review. Of the remaining eighty-three articles, sixty were excluded after full-text review because they did not measure the exposure or outcome(s) of interest or they were conducted among unhealthy populations or within countries ranked as ‘Low Human Development’ by means of the Human Development Index. A total of twenty-three articles met the eligibility criteria described above and were included in the present review. A further three articles were added from the bibliographies of the review articles, resulting in twenty-six eligible studies, which examined BMI or other measures of body adiposity (Fig. 1). Each included study was evaluated for quality using a component approach based on the Preferred Reporting Items for Systematic Reviews and Meta-Analyses recommendations(Reference Liberati, Altman and Tetzlaff16). The selection of the components was based on the criteria published by Sanderson et al. in the International Journal of Epidemiology (Reference Sanderson, Tatt and Higgins17). The key components examined for assessing study quality included participant selection, measurement of exposure and outcome variables, study design-related bias, confounding and appropriateness of statistical methods. Studies with minimal bias, appropriate measurement of exposure and outcome variables, and adjustment for confounders were considered to be of high quality.
Results
Dietary variety and measures of body adiposity
In the present qualitative synthesis, twenty-six studies were included that examined the associations between various dietary variety indices and measures of body adiposity including BMI, waist and hip circumference and percentage of body fat. The majority of the studies were cross-sectional (77 %) and the results varied among populations, dietary variety assessment methods and outcome measures. For ease of comparison, the results of the studies have been organised by the dietary variety assessment method utilised, which have been divided into the following categories: (1) overall variety (Table 1); (2) variety in recommended foods (Table 2); (3) variety in non-recommended foods (Table 3). When multiple dietary variety assessments were used in a single study, each result from that study was reported in the appropriate table.
DDS, Dietary Diversity Score; WHR, waist:hip ratio; WC, waist circumference; Q4, quartile 4; Q1, quartile 1; FDSK-11, Food Diversity Score Kyoto; INCAP, Institute of Nutrition of Central America and Panama; FVS, Food Variety Score; RFS, Recommended Foods Score; NRFS, Non-Recommended Foods Score; TLGS, Tehran Lipid and Glucose Study; BNFCS, Belgian National Food Consumption Survey; NHANES III, National Health and Nutrition Examination Survey III; HEI, Healthy Eating Index; USDA, US Department of Agriculture; CSFII, Continuing Survey of Food Intake by Individuals.
*+ Denotes a positive association; − denotes an inverse association; NS denotes a non-significant association; NS − denotes a non-significant but inverse trend; NS+ denotes a non-significant but positive trend.
DDS, Dietary Diversity Score; Q3, quartile 3; Q1, quartile 1; BPRHS, Boston Puerto Rican Health Study; FV, fruits and vegetables; WC, waist circumference; INCAP, Institute of Nutrition of Central America and Panama; FVS, Food Variety Score; RFS, Recommended Foods Score; NRFS, Non-Recommended Foods Score; N/A, not available; DVQ, Diet Variety Questionnaire; DEXA, dual-energy X-ray absorptiometry; NHANES III, National Health and Nutrition Examination Survey III; DDS-R, Dietary Diversity Score for Recommended Foods; RFBS, Recommended Foods and Behavior Score, Q4, quartile 4.
*+ Denotes a positive association; − denotes an inverse association; NS denotes a non-significant association; NS − denotes a non-significant but inverse trend; NS+ denotes a non-significant but positive trend.
INCAP, Institute of Nutrition of Central America and Panama; FVS, Food Variety Score; RFS, Recommended Foods Score; NRFS, Non-Recommended Foods Score; WC, waist circumference; N/A, not available; DVQ, Diet Variety Questionnaire; DEXA, dual-energy X-ray absorptiometry; CSFII, Continuing Survey of Food Intake by Individuals.
*+ Denotes a positive association; − denotes an inverse association; NS denotes a non-significant association; NS − denotes a non-significant but inverse trend; NS+ denotes a non-significant but positive trend.
The decision to organise the results based on the type of dietary variety assessment method utilised is informed by the landmark study conducted in 1999 by McCrory et al. (Reference McCrory, Fuss and McCallum8). This analysis examined seventy-one healthy men and women to evaluate the associations between dietary variety within ten food groups and obesity. Greater variety within snack foods, carbohydrates, sweets and entrées was positively correlated with the percentage of body fat (r 0·38, P= 0·001), while variety within vegetables was inversely correlated with the percentage of body fat (r − 0·31, P= 0·008). A variety ratio that measured the ratio of vegetable variety:variety within snack foods, carbohydrates, sweets and entrées was also negatively correlated with the percentage of body fat (r − 0·27, P= 0·02). Following the McCrory study, the role of dietary variety in body weight management became more controversial and many studies began examining the associations between dietary variety, body weight and excess adiposity.
Overall dietary variety and measures of body adiposity
There were fourteen cross-sectional studies (Table 1) that examined the associations between dietary variety within the overall diet and its associations with various body adiposity measures. Of these fourteen studies, six assessed dietary variety using diet assessment methods considered valid for the measurement of usual intake, such as a FFQ(Reference Azadbakht and Esmaillzadeh18–Reference Kimura, Okumiya and Sakamoto23). These studies were inconsistent in their conclusions with two studies reporting inverse associations(Reference Azadbakht and Esmaillzadeh18, Reference Kent and Worsley22), two with non-significant associations(Reference Gregory, McCullough and Ramirez-Zea21, Reference Kimura, Okumiya and Sakamoto23) and two with positive associations(Reference Azadbakht, Mirmiran and Azizi19, Reference Azadbakht, Mirmiran and Esmaillzadeh20). The remaining eight studies assessed dietary variety using one or two 24 h recalls, which are considered less valid estimates of usual dietary intake compared with other self-report measures that assess a longer period of time(Reference Dodd, Guenther and Freedman24). Overall, five of the eight studies reported non-significant associations between dietary variety and measures of body adiposity(Reference Ervin25–Reference Vandevijvere, De Vriese and Huybrechts29), one study reported a positive association(Reference Ponce, Ramirez and Delisle30) and two studies reported an inverse association(Reference Tande, Magel and Strand31, Reference Kennedy, Bowman and Spence32).
Among the studies reporting positive associations, the most consistent association emerged with BMI or obesity status (determined using BMI cut-offs) as outcome measures. Among Iranian adults, greater diversity (quartile 4 v. quartile 1) within five main food groups and twenty-three subgroups was associated with a higher BMI and a greater percentage of obese adults(Reference Azadbakht, Mirmiran and Azizi19, Reference Azadbakht, Mirmiran and Esmaillzadeh20). Ponce et al. (Reference Ponce, Ramirez and Delisle30) also noted that obese and overweight Mexican adults had higher levels of dietary variety within twenty-four food groups compared with normal-weight and underweight individuals (P= 0·007). However, no studies noted a positive association for other measures of adiposity such as WC or WHR(Reference Azadbakht, Mirmiran and Azizi19, Reference Azadbakht, Mirmiran and Esmaillzadeh20).
There were four studies that reported inverse associations between dietary variety and measures of adiposity(Reference Azadbakht and Esmaillzadeh18, Reference Kent and Worsley22, Reference Tande, Magel and Strand31, Reference Kennedy, Bowman and Spence32). Young Iranian females with a higher Dietary Diversity Score had lower WC and WHR as well as a reduced percentage of obesity(Reference Azadbakht and Esmaillzadeh18). Greater dietary variety was inversely associated with BMI among Australian men, but no association was noted among women(Reference Kent and Worsley22). Using the Healthy Eating Index measure of dietary variety, male National Health and Nutrition Examination Survey III participants had a modest reduction in the odds of abdominal adiposity as dietary variety increased, while obese women had higher variety scores when compared with their normal-weight counterparts (P< 0·03)(Reference Tande, Magel and Strand31). Using the same measure of variety, participants from the Continuing Survey of Food Intake by Individuals had a lower BMI with greater variety if >55 % of total energy in the diet was consumed from carbohydrate (P< 0·05); no difference in BMI was seen among individuals consuming a lower percentage of energy from carbohydrate(Reference Kennedy, Bowman and Spence32).
There were five studies that reported no significant association between overall dietary variety and adiposity measures(Reference Ervin25–Reference Vandevijvere, De Vriese and Huybrechts29). Adults in Louisiana(Reference Mohindra, Nicklas and O'Neil27) with higher Dietary Diversity Scores within five major food groups had a non-significant 11 % reduced risk of overweight and a 19 % non-significant reduced risk of obesity. In a study of Belgian adults, no significant association was detected between variety among and within food groups (data not shown)(Reference Vandevijvere, De Vriese and Huybrechts29). Also, two US studies(Reference Ervin25, Reference McCabe-Sellers, Bowman and Stuff26) did not detect a significant association between the Healthy Eating Index measure of dietary variety and BMI among older (60+year) US men and women. Overall, total dietary variety is inconsistently associated with measures of body adiposity and varies based on the type of dietary assessment method utilised and the population studied.
Dietary variety within recommended foods and measures of body adiposity
We identified ten studies that examined the associations between variety of recommended foods and measures of body adiposity (Table 2). Recommended food groups varied between studies, but included fruits, vegetables, low-fat dairy products and other lower-energy foods. Of these studies, seven studies reported inverse or mixed associations(Reference Roberts, Hajduk and Howarth28, Reference Brunt and Rhee33–Reference Lyles, Desmond and Faulk38), where intake of vegetables and grains and other foods recommended by the dietary guidelines was inversely associated with BMI and body weight. Among the National Health Interview Survey participants, the proportion of adults classified as overweight by BMI standards was lower among adults with higher Recommended Foods and Behavior scores(Reference Kant, Graubard and Schatzkin35). Similarly, National Health and Nutrition Examination Survey participants with greater variety in recommended foods had a lower average BMI than individuals with less variety (β = − 0·18 and − 0·38; P< 0·00001 using the Recommended Foods Score and Dietary Diversity Score for Recommended Foods)(Reference Kant and Graubard34). Brunt & Rhee(Reference Brunt and Rhee33) noted that grain and vegetable variety was higher among normal v. obese participants. Importantly, dietary variety measured via the Recommended Foods Score was found to be protective against obesity in men over time (BMI: 25·4 v. 25·6 kg/m2 in quartile 5 v. quartile 1; P for trend < 0·001)(Reference McCullough, Feskanich and Stampfer36). However, the reverse was true among female nurses (BMI: 25·0 v. 24·7 kg/m2 in quartile 5 v. quartile 1; P for trend < 0·001)(Reference McCullough, Feskanich and Stampfer36). Similarly, Lyles et al. (Reference Lyles, Desmond and Faulk38) found that variety in carbohydrates was inversely associated with BMI in women and positively associated with BMI in men, but it did not correlate with other measures of adiposity.
The remaining three cross-sectional studies noted no significant trends between variety in recommended foods and measures of adiposity(Reference Gregory, McCullough and Ramirez-Zea21, Reference Bezerra and Sichieri39, Reference Bhupathiraju and Tucker40). Among Puerto Rican adults, there was no association between tertile of fruit and vegetable variety and BMI or percentage of abdominal adiposity(Reference Bhupathiraju and Tucker40). Similarly, there was no significant relationship between Recommended Foods Score and BMI, or WC among male and female Guatemalan adults(Reference Gregory, McCullough and Ramirez-Zea21). Among Brazilian adults, there was a borderline significant difference in the prevalence of obesity by tertile of healthy Dietary Diversity Score (10·6 v. 9·0 foods in quartile 3 v. quartile 1; P= 0·05). However, it is important to note that these analyses examined approximately ten households as the average sampling unit rather than the individual. Overall, dietary variety in recommended foods is not adversely associated with body adiposity and, in some cases, may be protective.
Dietary variety among non-recommended foods and measures of body adiposity
Of the included studies, nine (four cross-sectional, one longitudinal and four intervention studies) examined the associations between the intake of non-recommended or energy-dense foods and measures of body adiposity. Of these, three of the four cross-sectional studies(Reference Roberts, Hajduk and Howarth28, Reference Brunt and Rhee33, Reference Lyles, Desmond and Faulk38), one longitudinal study(Reference Woo, Cheung and Ho41) and two of the four intervention studies noted positive associations(Reference Raynor, Jeffery and Tate37, Reference Mattes42), while the remaining three did not detect a significant association between dietary variety in non-recommended foods and measures of body adiposity(Reference Gregory, McCullough and Ramirez-Zea21, Reference Raynor, Niemeier and Wing43, Reference Raynor, Steeves and Hecht44).
Among the studies that detected a positive association, Woo et al. (Reference Woo, Cheung and Ho41) found that variety in snack foods as well as the food variety ratio (i.e. the ratio of snack foods:grains and meats) increased the risk of overweight by 36–45 %. The food variety ratio was developed from a previous case–control study within Hong Kong Chinese adults that noted that variety within snack foods was positively correlated with the percentage of body fat, and waist and hip circumference, whereas variety in meats and grains was inversely correlated with these measures(Reference Sea, Woo and Tong45). Variety among energy-dense foods, defined as carbohydrate-based foods with more than 10 % energy from fat, was associated with a higher BMI among US older adults (β = 0·12; P< 0·001)(Reference Roberts, Hajduk and Howarth28). In younger adults, obese individuals reported greater variety in meats, pork, lamb, veal and other game when compared with their normal-weight counterparts (P< 0·05)(Reference Brunt and Rhee33). Similarly, Lyles et al. (Reference Lyles, Desmond and Faulk38) noted that a 1-unit increase in fat and protein variety was associated with a 0·96 and a 1·98 kg/m2 increase in BMI.
Also, two intervention studies support a relationship between dietary variety in some food categories and excess weight. In an 18-month weight-loss trial, Raynor et al. (Reference Raynor, Jeffery and Tate37) noted that variety in high-fat foods was positively associated with body weight between 0 and 6 months (β = 0·26; P< 0·01) and between 6 and 18 months (β = 0·24; P< 0·01). Another intervention noted greater weight loss over a 7-week period among women randomised to a single cereal v. variety of cereal group(Reference Mattes42).
However, three studies did not support an association between dietary variety in non-recommended foods and measures of body adiposity(Reference Gregory, McCullough and Ramirez-Zea21, Reference Raynor, Niemeier and Wing43, Reference Raynor, Steeves and Hecht44). In one study among Guatemalan adults, the Non-recommended Foods Score was not significantly associated with BMI in men (β = 0·03, 95 % CI − 0·08, 0·14) or women (β = 0·004, 95 % CI − 0·05, 0·06). Additionally, the Non-recommended Foods Score was not significantly associated with WC in either sex(Reference Gregory, McCullough and Ramirez-Zea21). There was no significant difference in body weight reduction (lifestyle+low variety: − 9·9 (sd 7·6) %); lifestyle: − 9·6 (sd 9·2 (%) in a recent trial where participants were randomised to standard weight-loss counselling v. standard counselling plus a low-variety condition, where they were asked to limit non-nutrient-dense, energy-dense variety to two foods over the 18-month follow-up period(Reference Raynor, Steeves and Hecht44). These results are consistent with a previous intervention, which did not detect a significant difference in weight loss among women choosing between a single snack v. a variety of snacks(Reference Raynor, Niemeier and Wing43). However, while women in the single snack condition did reduce their snack food variety, they did not limit snacks to a single food choice per the instructions. Incomplete compliance with intervention instructions may have attenuated their findings. Overall, greater variety in non-recommended food categories may adversely affect body adiposity measures. Fig. 2 summarises the key findings of the twenty-six studies included in this review.
Evaluation of study quality
There was some variation between the studies with respect to the five quality components considered (aforementioned in the Methods section). All twenty-six studies provided detailed information regarding their participant selection criteria and limited their populations to healthy adults. Target populations ranged in age from approximately 20 to 85 years and were drawn from both US and international populations, and more than 50 % of studies utilised random sampling techniques. Although measurement error exists with all self-reported dietary assessments, validated tools were used to measure dietary exposures in all but one included study(Reference Bezerra and Sichieri39). Further, most studies utilised interviewer-administered assessments, tools developed for their target population and standardised protocols to enhance quality control of dietary data. The included studies incorporated techniques to minimise study design-related bias. For example, most cohort and intervention studies described rates of attrition and any observed differences between included and excluded participants. Furthermore, all intervention studies utilised appropriate randomisation techniques and blinded both researchers and participants when possible. Self-report bias was also reduced in most studies as trained interviewers measured anthropometric measures that were used to define outcome variables used in the present review.
Some studies were primarily descriptive in nature, and consequently did not adjust for relevant confounding variables such as age, sex and smoking status(Reference Kimura, Okumiya and Sakamoto23, Reference Ervin25, Reference Brunt and Rhee33). Additionally, studies examining the associations between dietary variety and body adiposity as secondary outcomes did not sufficiently control for confounding(Reference Kant, Graubard and Schatzkin35, Reference McCullough, Feskanich and Stampfer36); however, these studies met other quality criteria. Finally, all studies utilised appropriate statistical methods such as multivariate regression.
Discussion
At present, there is no consensus in the understanding of the associations between dietary variety and measures of body adiposity. Dietary variety is hypothesised to influence body adiposity through increasing energy intake. Existing cross-sectional, longitudinal and experimental literature consistently support an association between dietary variety and energy intake(Reference Raynor4). Although definitions of dietary variety vary, the correlation between various measures of dietary variety and energy intake ranges from r 0·25 to 0·72, depending on the measure of dietary variety used, sex and race. Variety within recommended foods or within the broader five food groups is more weakly or non-significantly correlated with energy intake (r 0·25–0·37)(Reference Gregory, McCullough and Ramirez-Zea21, Reference Kant and Graubard34, Reference McCullough, Feskanich and Stampfer36, Reference Murphy, Foote and Wilkens46, Reference Kant, Schatzkin and Graubard47) compared with overall variety or variety within energy-dense or non-recommended foods (r 0·30–0·72)(Reference Gregory, McCullough and Ramirez-Zea21, Reference Murphy, Foote and Wilkens46, Reference Foote, Murphy and Wilkens48–Reference Ahn, Engelhardt and Joung51). Importantly, weight-loss participants with greater variety in energy-dense foods consume more energy each day, and those who restrict variety in some food groups may be more effective at reducing energy intake compared with those who only practise portion control(Reference Raynor, Jeffery and Tate37, Reference Raynor, Niemeier and Wing43). Given the strength of the association between dietary variety and energy intake, it is important to examine whether dietary variety also influences body adiposity.
We reviewed a total of twenty-six studies examining the associations between varying definitions of dietary variety and measures of body adiposity. We noted that the relationship between dietary variety and body adiposity is multifaceted and moderated by the foods or food groups used to measure dietary variety. Furthermore, total food variety has been multiply defined and measured across studies, limiting comparisons between studies. Taken together, studies that assess usual total dietary variety (i.e. total variety over more than a 2 d period) found that dietary variety is either protective against overweight and obesity or does not influence body adiposity(Reference Azadbakht and Esmaillzadeh18, Reference Gregory, McCullough and Ramirez-Zea21–Reference Kimura, Okumiya and Sakamoto23). This is consistent with the hypothesis that a total diet approach is necessary to achieve nutrient adequacy within the diet(Reference Tucker52). However, this association was more volatile when studies assessed usual dietary variety using one or two 24 h recalls; five of the nine studies examined reported non-significant findings, though many approached significance and varied in whether they were positively or inversely associated with adiposity measures. When variety measurements are limited to only healthful or recommended foods, we consistently observe that dietary variety is either inversely associated with measures of body adiposity or does not substantively influence adiposity in most populations. Conversely, when variety measures only include non-recommended or energy-dense foods, greater variety is reliably associated with increased measures of adiposity, though associations become less consistent among intervention studies.
Although the relationship between dietary variety and excess adiposity is complex, it is important to consider the factors that contribute to a limited understanding of the relationship between dietary variety and body adiposity. In part, relying on self-reported weight and height to calculate BMI may introduce bias, as underestimation of weight and overestimation of height are common in some populations(Reference Merrill and Richardson53). Additionally, the clinical relevance of adiposity outcome measures should be taken into consideration. While the present review noted inconsistencies between dietary variety and obesity using all measures of adiposity (BMI, WC, WHR and percentage of body fat), some of these measures (e.g. WHR) are more predictive of chronic disease risk, and therefore may be more appropriate for assessing the diet–disease relationship(Reference Leitzmann, Moore and Koster54). For example, Iranian adults who consumed a more varied diet had a higher BMI than adults who consumed a less varied diet; however, there was no difference in measures of abdominal adiposity, and adults consuming varied diets had better metabolic profiles(Reference Azadbakht, Mirmiran and Azizi19). However, only eight of the twenty-six studies used a method other than BMI to determine excess adiposity, which limits our understanding of the association between dietary variety and obesity.
Consistent and comprehensive definitions of dietary variety are also needed to clarify the relationship between dietary variety and excess adiposity. Importantly, the definition and measurement of total variety appear to be inconsistently assessed across studies, which limits comparison across studies. The number and healthfulness of the food groups included in the variety scores differ considerably across studies, which also limits accurate assessment. When dietary variety scores only measure variety within recommended foods or non-recommended foods, they exclude other food groups that may affect measures of body adiposity in the context of the total diet. Similarly, pure count methods introduce bias because individuals may be misclassified or insufficiently distinguished from one another. When the diet is divided into food groups that individuals either consume or do not consume in the time period examined, an individual who eats a single serving of fruit is categorised identically to the person who consumes three servings, for example(Reference Dixon, Cronin and Krebs-Smith9). These techniques have the effect of constraining variety scores within food groups, which also have the effect of attenuating risk associations by failing to distinguish between individuals who consume a minimally varied diet v. a highly varied diet.
Dietary proportionality and quality components are also missing from existing scores, which are important confounders in the relationship between dietary variety and excess adiposity. National recommendations provide guidance on the proper balance of each food group in the diet (i.e. proportionality), and it is possible that variety in energy-dense sources is less detrimental to body weight when consumed in appropriate portions. Moreover, most dietary variety scoring methods used in epidemiological studies fail to account for dietary quality. This means that an individual who consumes a large number of nutrient-dense foods will receive the same score as an individual who consumes the same number of nutrient-poor foods. Consequently, the failure to include (1) all foods consumed, (2) a measure of proportionality and (3) a dietary quality component in dietary variety scores probably contributes to the mixed findings between dietary variety and measures of body adiposity within the literature.
Measurement error associated with diet assessment methods may also adversely affect conclusions(Reference Tucker55). Dietary variety as assessed from one or two 24 h recalls may not capture true usual long-term intake(Reference Dodd, Guenther and Freedman24), and, without appropriate statistical adjustment, is insufficient for assessing the association between dietary variety and body adiposity. FFQ are designed to measure usual intake, but are prone to random and systematic error(Reference Subar, Kipnis and Troiano56), which may be especially problematic in the context of assessing dietary variety. There is high participant recall burden when trying to report usual intake over an extended period of time, which can be especially difficult with episodically consumed foods. Additionally, predefined food lists artificially constrain variety measurement and may exclude a number of foods consumed by various cultural groups. These biases are amplified in the context of assessing causal relationships from primarily cross-sectional studies.
Further, study populations require careful consideration, as small or non-representative populations can introduce bias and make conclusions non-generalisable to the larger population. Notably, many of the experimental studies are small (n< 200) and use successful weight-loss participants, who differ substantively from the general US population. Similarly, generalisability may be problematic when extrapolating data from populations undergoing nutrition transitions or with a relatively low socio-economic status. Dietary variety has been shown to increase proportionally with income, and foods that are selected are typically highly palatable, energy-dense items(Reference Hoddinot and Yohannes57, Reference Rivera, Barquera and Campirano58). Consequently, this may confound associations when analyses do not adjust for socio-economic status.
In summary, we are limited in our ability to draw strong inferences from the existing research for a number of reasons. First, study outcomes vary between studies and may have limited clinical relevance. Definitions of total variety, variety in recommended foods and variety in non-recommended foods are inconsistent. Further, measurement error in dietary assessment instruments is not sufficiently addressed statistically, which results in biased estimates. Additional bias is introduced because important confounding variables are not accounted for when count-based scores are employed. Finally, the use of non-representative populations limits the generalisability of conclusions.
The present systematic review addresses a number of gaps in the existing literature. By thoroughly examining all relevant studies and methods for assessing dietary variety, we are able to provide a more specific understanding of the ambiguities in the existing literature. We concur with Kennedy(Reference Kennedy2) that observed inconsistencies are partly a consequence of the varying definitions and measures of dietary variety used within nutrition studies. However, this review also detected important gaps that limit the robustness of existing conclusions including: (1) insufficient use of clinically relevant outcomes; (2) measurement error and bias in the assessment of dietary variety; (3) inconsistent definitions of dietary variety; (4) excessive generalisation based on non-representative samples.
In conclusion, the present review is an important first step in clarifying the associations between dietary variety and excess adiposity, which is currently limited by the methods used to assess variety. In the present review, we found that (1) dietary variety within recommended and low-energy foods alone do not increase the odds of overweight and obesity, (2) greater variety within less healthful, energy-dense foods increases the odds of overweight and obesity and (3) the association between total dietary variety and adiposity is mixed and accurate evaluation of this association requires a consistent and theoretically valid measurement tool.
The clinical implications of these conclusions are noteworthy. Dietary variety recommendations have largely been removed from the US Dietary Guidelines secondary to the concern about its association with excess energy intake and adiposity. While it may be prudent to avoid broadly recommending dietary variety, limitations in the measurement and interpretation of these constructs may be generating excessive caution. Dietary variety must be understood utilising a total diet approach in order to make concrete recommendations, and the methods used to measure dietary variety must move beyond simple count methods.
Future research requires a dietary variety scoring method that captures all distinct foods consumed in the diet along with a measure of proportionality and a measure of quality in order to accurately assess the associations between dietary variety and excess adiposity. Using a comprehensive dietary variety score to evaluate these associations in a large, nationally representative US sample will help clarify this relationship and provide researchers with a tool that can be used cross-culturally in order to more effectively and consistently evaluate the associations between dietary variety and adiposity. A clear understanding of the role of dietary variety in obesity management is important in the context of the current obesity epidemic and may aid in the development of simple, actionable national nutrition recommendations to facilitate healthful weight management.
Acknowledgements
The present review was supported by the American Heart Association's Founders Affiliate Predoctoral Fellowship (12PRE9320023, awarded to M. V.). M. V. conceptualised, designed and conducted the literature review and was the lead author of the paper. L. B. D. was involved in reviewing the manuscript and organisation of the content. N. P. supervised the research and critically reviewed the manuscript for important intellectual content. All authors read and approved the final manuscript and took full responsibility for the final content. The authors have no conflicts of interest to disclose.