Between-meal snacks represent an important source of solid fats, added sugars, and sodium in the American diet [1–6]. The USDA has identified solid fats and added sugars (SoFAs) as the main sources of “empty” calories, contributing ample energy but few nutrients [7, 8]. Snacking between meals is consistently listed among potential causes of obesity and weight gain [9–11] and is an issue of public health concern.

Healthier snacking patterns could improve diet quality by reducing empty calories and sodium, and increasing intakes of some insufficiently consumed nutrients [12, 13], including fiber and potassium [14]. Healthier snack patterns may also shift the fatty acid ratio away from solid fats towards oils, by increasing mono- and polyunsaturated fats (MUFAs and PUFAs), and plant-based omega-3 fatty acids, and reducing saturated and trans fat consumption [15, 16]. In the United States, consumption of added empty calories (e.g., added sugars and solid fats) and sodium is much higher than recommended, while consumption of some food groups, including whole grains, plant and seafood protein, whole fruits and vegetables lags well below recommended levels [17–19].

Tree nuts, including almonds, walnuts, and cashews, have been featured prominently in multiple dietary recommendations and guidelines [14], as well as consumer-facing recommendations regarding healthy snacking from numerous profesional and advocacy groups, including the American Heart Association, the American Diabetes Association and Academy of Nutrition and Dietetics [20–22]. In the nationally representative cross-sectional National Health and Nutrition Examination Surveys (NHANES), tree nut consumption was associated with better nutrient adequacy, higher diet quality, and improved health risk markers [15, 16, 23]. Prospective cohorts have also observed that consumption of tree nuts reduces the risk of all-cause mortality, deaths due to heart disease, cancer and respiratory disease, the incidence of heart disease, and may reduce the likelihood of weight gain [24–28]. Analyses of dietary trends suggest that the consumption of nuts/seeds among US adults has risen from 0.43 to 0.69 oz-equivalents per day from 1999–2000 to 2011–2012. In 2011–2012, just over 30% of adults consumed ≥5 servings of nuts/seeds per week, the recommended amount per 2000 kcal/d from the 2015–2020 Dietary Guidelines for Americans [14]. While no specific recommendation exists for tree nuts, the trend in their consumption has been particularly strong, increasing 145% from 0.11 to 0.27 oz-equivalents per day from 1999–2000 to 2011–2012 [17]. Among the reasons for the substantial growth in tree nut consumption may be their well-documented health benefits as well as the certification of a qualified health claim for nuts and heart disease in 2003 [29].

Following earlier studies on food pattern modeling [30, 31], we sought to quantify the nutritional impact of replacing between-meal snacks with tree nuts or almonds. First, a tree nut composite was created based on the relative frequency of consumption within the population (e.g., a weighted average of almonds, walnuts, pecans and other tree nuts). In substitution models, between-meal snacks (excluding beverages) were replaced, on a per calorie basis, with this weighted tree nut composite. One model replaced all solid snack foods and a second model replaced all solid snacks, with the exception of whole fruits, non-starchy vegetables and whole grain foods. Based on the well-documented high nutrient density of tree nuts [32], we expected to observe measurable improvements in the quality of food patterns generated. The primary goal was to quantify the impact of such a replacement on key dietary constituents of public health interest, including added sugars, solid fats, empty calories, fatty acids, sodium, potassium and magnesium.

Among the 17,444 study participants, 76.9% consumed a snack on their first recall day (defined as any energy from any food or beverage identified as a snack). Overall, mean energy from all snacks was 303 kcal/d, whereas the median was 204 kcal/d. Among snack-consumers, mean energy from snacks was 394 kcal/d, whereas the median was 293 kcal/d. Mean energy from solid snacks eligible for substitution was 252 kcal/d and the median was 148 kcal/d. For Model 1, the amount of solid snack energy in kcal/d that was eligible for substitution depended on age: 250 (ages 1–3y), 341(4–8y), 374 (9–13y), 350 (14–19y), 316 (20–30y), 309 (31–50y), 285 (51–70y), and 206 (71 + y). Median energy from snacks (293 kcal/d) corresponds to 51 g of almonds or 45 g of walnuts.

Results of primary models

The nutrient content of modeled food patterns is presented in Figs. 2, 3, 4 and 5 for the entire population and by age group. Model 1 and Model 2 showed significant declines in consumption of empty calories (−20.1% and −18.7%), and a significant drop in solid fats (−21.0% and −19.3%) and in added sugars (−17.8% and −16.9%). Consistent with the decline in solid fats, consumption of saturated fatty acids (SFA) declined modestly in both models (−6.6% and −7.1%). By contrast, the consumption of oils (+65.3% and +55.2%), total polyunsaturated fatty acid (PUFA, +42.0% and +35.7%), alpha-linolenic acid (ALA, +53.1% and +44.7%) and monounsaturated fatty acid (MUFA, 35.4% and 29.6%) all increased. The significant increase in total fat (+20.5% and +16.8%) was accompanied by a greatly improved mono- and polyunsaturated to saturated fat ratio. Dietary carbohydrates dropped significantly (−13% and −10%), whereas a small but statistically significant increase was observed for protein (+2.6% and +1.7%).

Sodium declined from 3,518 mg/d in observed diets to 3,086 mg/d (−12.3%) and 3,124 mg/d (−11.2%) in Model 1 and 2, respectively. For dietary fiber, an 11.1% and 14.8% increase was observed compared to observed diets. Small changes were observed for potassium (not shown, +1.5% and +2.9%) and strong increases were observed for magnesium (+29.9% and +27.0%).

Impact of models on healthy eating index-2010

To measure the total impact of the substitution models on diet quality, the HEI-2010 was examined (see Fig. 6). Based on observed diets, the population mean HEI-2010 value was 58.5 (95% CI 57.4, 59.7). For both models, the HEI-2010 values were significantly higher, 67.8 for Model 1 and 69.7 for Model 2 (p < 0.001 for each). For all age groups, the HEI-2010 score was significantly higher in both Model 1 and Model 2 as compared to observed diets. The effect was particularly profound among children and adolescents due to low HEI-2010 baseline values and higher frequency of consuming lower quality snacks. As expected, the results of Model 2 tended to be stronger than for Model 1. When evaluating individual HEI-2010 components (see Fig. 7), modest but statistically significant declines were observed for total vegetables, total fruit, whole fruit, whole grains and dairy comparing observed diets to Model 1. However, component values for seafood and plant protein, fatty acid ratio, sodium, refined grains, and energy from solid fat and added sugars all significantly improved. Similar results by component were observed for Model 2, but with no change in whole fruit, total fruit and whole grains, and modest declines in dairy and total vegetables.

Faced with the need to improve US diets, the 2015–2020 Dietary Guidelines for Americans stressed that making small shifts in eating patterns can make a big difference in diet quality overall [14]. One recommendation was to improve the nutrient quality of between-meal snacks, replacing high-calorie snacks with more nutrient-rich options [14]. In a parallel effort to improve diets of children, new federal rules require healthier snacks in schools [47]. Quantifying the impact of replacing typically consumed solid food snacks with tree nuts was the primary purpose of the present study. For the same amount of calories the modeled food patterns replacing currently consumed solid food snacks with tree nuts were more nutrient dense. First, there was a significant reduction in empty calories. Second, there was a shift in dietary fat profiles. The modeled diets had much less saturated fat and a more favorable fatty acid profile, characterized by more oils, more MUFAs and PUFAs, and more plant omega-3 s (alpha-linolenic acid). The ratio of PUFA and MUFA to saturated fatty acids was much more favorable than in observed diets. Total fat content was higher, as might be expected after replacing many refined grains or sugar-sweetened grains with nuts containing fat, fiber, and protein. Overall, Healthy Eating Index-2010 scores improved in the modeled compared to observed diets.

Frequent between-meal snacking has long been an issue of public health concern [48, 49]. In several past studies, increased snacking frequency has been linked to rising obesity rates [9, 10]. First, more frequent snacking has been associated with consuming more calories overall [8]. Second, some of the commonly eaten snacks were of low, if not minimal, nutritional value [8]. Foods and beverages consumed as snacks provided higher proportions of alcohol, carbohydrates, and total sugars relative to calories, while providing lower proportions of many other nutrients. On average, US adults consumed 24% of daily calories as snacks [49]. Consistent with prior reports [3, 50, 51], the most commonly eaten snacks in the US were sweet desserts high in carbohydrates, sugars, and saturated fats. The top items in the 2009–2012 NHANES were cookies, brownies, cakes and pies, ice cream and frozen dairy desserts, chocolate candy and doughnuts, pastries and sweet rolls. Also on the list were corn chips, crackers and potato chips. Between-meal snacks consisting of apples, bananas, yogurts and other fruit were eaten less often.

Substitution modeling of dietary patterns offers a compelling way to test the nutritional impact of dietary guidance [30]. When based on nationally representative dietary data, such models can provide insight into the impact of federal dietary guidelines on the diet of the entire population or specific population sub-groups [30, 31]. The modeled food patterns can be used to evaluate, compare, and rank the impact of following dietary advice on measures of diet quality at the population level. Results of modeling studies can be used to inform effective nutrition communications and the sub-populations that may benefit most from such efforts. For example, we observed the greatest impact of replacing snacks with tree nuts among children and adolescents, as they tended to consume proportionally more nutrient sparse and less healthful snacks as compared to adults. In addition, children/adolescents are less likely to consume nuts/seeds both overall and as a snack [14].

With that said, such models also have limitations. First, for any substitution model, the units of the substitution model are critical (e.g., whether foods are replaced on per-calorie or per-serving). Given the diverse number of foods in the present study and the arbitrary nature of serving sizes, we opted for an isocaloric substitution model. In simple terms, such a model would result in no change in energy intakes, and therefore no change in body weight. However, if applied in practice, given the relatively high satiety of nuts compared to commonly consumed solid snacks, we might expect fewer calories to be consumed [52, 53]. However, this assumption is not verifiable, so relying on an isocaloric model seems most appropriate. Beyond impacts on weight, substantial improvements to the diet independent of energy changes is likely to have considerable population health benefits. Second, as evidenced by the impact of the models on Healthy Eating Index-2010 component scores, replacement of all solid snacks with tree nuts resulted in modest declines in whole fruit, vegetable, whole grain and dairy intakes. Model 1 was agnostic as to the foods being replaced; both less healthful and more healthful snacks were replaced. As expected Model 2, which did not replace whole fruits, non-starchy vegetables or whole grain snacks with tree nuts, resulted in better improvements and no significant declines in HEI-2010 component scores, with the exception of total dairy. However, because it is important to quantify both the positive and potentially negative impact of any model, we opted to use two models, one that universally replaced solid snack foods and another, which more judiciously replaced snack foods. Despite major improvements observed for many of the modeled dietary constituents evaluated, in many cases population mean intakes remained well below recommended levels (e.g., for sodium, fiber and potassium), suggesting that multiple other changes would need to be implemented simultaneously. For example, increasing whole grain, whole fruit or vegetable consumption, specifically through replacing refined grains and foods high in empty calories would also have tremendous population health impact. As such, these models should be viewed as an approach for quantifying and comparing the impact of various strategies, not as a prescription for specific changes an individual should make.

Substitution modeling complements and builds upon past studies comparing the diets of tree nut consumers and non-consumers. In past studies, based on 2005–2010 NHANES data, usual diets of tree nut consumers were also determined using two 24-h dietary recalls and the NCI Method [16]. That study assessed diet quality based on percentages of tree nut consumers below the estimated average requirement (EAR) or above the adequate intakes (AI) relative to non-consumers. Diets of tree nut consumers showed more favorable intakes for vitamins A and C, folate, calcium, iron, magnesium and zinc as well as potassium and fiber. The Healthy Eating Index 2005 score (HEI 2005) was higher for tree nut consumers as compared to non-consumers.

In cross-sectional studies, tree nut consumers and non-consumers were of necessity different people, who may differ by both observed (e.g., age) and unobserved or difficult to measure factors (e.g., health-seeking behavior). By contrast, substitution modeling offers some insight into how diets of individual consumers can be improved by making some small changes in food choices that are based, in part, on the existing eating patterns of the broader population as the tree nut composite was tailored based on population consumption patterns.

Past studies of tree nut consumers and non-consumers found that the two groups did not differ in their usual sodium intakes and both exceeded recommended adequate intake [16]. Given that tree nuts are generally very low in sodium, nut consumers, typical of the American diet consumed excessive amounts of sodium from other sources. In some past studies, tree nut consumers had lower sodium intakes than did non-consumers [15, 54]. Consistent with those data, the present modeling study showed that the sodium content in modeled diets was much lower than in observed diets. Given that a small increase was observed for potassium, there was a decline in the dietary sodium to potassium ratio, another index of diet quality. Comparable patterns were observed for almonds only modeling. These modeled data counter the perceptions that tree nut snacks may act as a vehicle for increasing dietary sodium.

The present study quantifies the impact of replacement of solid food snacks with tree nuts, suggesting that such a replacement may have a profound effect on population diet quality. However, the critical question moving forward is what policies and interventions may be most effective at increasing consumption of tree nuts at the population-level. Schools and childcare settings may be one environment where tree nut consumption could be increased. A number of tools were recently made available online by the US Department of Agriculture to help schools identify those food items that meet Smart Snacks Standards, which were implemented in 2014–15 [55]. Based on these Standards, almonds, dry roasted without salt in 1 oz packages (167 kcal) are compliant with the school guidelines, providing <200 kcal per serving [56], as nuts/seeds are exempt from the total fat threshold. Products consisting of only dried fruit with nuts and/or seeds with no added sugars or fats are also exempt from the saturated fat standard [57]. Additional policies that may promote nut consumption include the adoption of healthy vending policies, though the contribution of vending machines to snack intake at the population-level remains limited. Additional strategies to increase nut consumption may include social marketing and/or mass communication efforts to increase awareness regarding the health benefits of nuts or address potential misconceptions (e.g., that nuts consumption may lead to weight gain) [58]. In addition, targeted subsidies to reduce their cost in comparison to less healthful snacks might be an additional approach to increase their intake [59]. The efficacy of these approaches specific to increasing nut consumption has not yet been established.

The present study had a number of strengths. First, the data are nationally representative and we used the NCI method to estimate usual intake distributions for the entire population and by age group. Second, same-person replacement modeling complements past studies based on consumers and non-consumers of a specific food. Third, we used a composite tree nut, based on existing nut consumption patterns, which allows for a more precise assessment of the potential nutritional effects of behavioral changes. Further, the tree nut composite included both salted and unsalted nuts, weighted by consumption frequency. Therefore, the modeled results are representative of what would be observed if current consumption patterns continued.

Numerous limitations are also worth noting. First, all NHANES data were based on self-reports. Proxy reports were used for young children [60] and some foods may be under-reported, either through omission or error in recording portion sizes [61]. It is also possible that some participants reported tree nuts, while eating peanuts [15]. In addition, the results presented here are based on models and do not reflect actual behavior of individuals. Nonetheless, substitution modeling in the context of a specific meal or eating occasion, offers one way to test the potential nutritional impact of dietary guidance.

Based on present modeling analyses, widespread replacement of solid food snacks with tree nuts or almonds would improve the quality of the diet. Even a partial replacement, exempting already nutrient rich snacks, had a significant positive effect on diet quality. Substitution modeling can provide added insight into the likely impact of following 2015–2020 DGAs on diet quality in the US.