**The Potential of Modulating the Reducing Sugar Released (and the Potential Glycemic Response) of Mu**ffi**ns Using a Combination of a Stevia Sweetener and Cocoa Powder** †

**Jingrong Gao 1,2,3,\*, Xinbo Guo 1,4, Margaret A. Brennan 2,3, Susan L. Mason 2, Xin-An Zeng 1,4 and Charles S. Brennan 1,2,3,4,\***


Received: 16 October 2019; Accepted: 20 November 2019; Published: 5 December 2019

**Abstract:** Muffins are popular bakery products. However, they generally contain high amounts of sugar. The over-consumption of muffins may therefore result in a high calorie intake and could lead to increased health risks. For this reason, muffins were prepared substituting sucrose with two levels of a base of stevia (Stevianna®). In addition, cocoa powder and vanilla were added to the muffin formulation with and without Stevianna® to mask any potential off flavors. Results illustrate that muffins with 50% Stevianna® replacement of sucrose were similar to the control samples in terms of volume, density and texture. However, replacement of sugar with 100% Stevianna® resulted in reductions in height (from 41 to 28 mm), volume (from 63 to 51 mL), and increased firmness (by four-fold) compared to the control sample. Sugar replacement significantly reduced the in vitro predictive glycemic response of muffins (by up to 55% of the control sample). This work illustrates the importance of sugar in maintaining muffin structure as well as controlling the rate of glucose release during simulated digestions.

**Keywords:** muffin; in vitro starch digestibility; glycemic index; stevia; sugar replacement

#### **1. Introduction**

In recent years, consumers have gained an increasing awareness regarding the effect of dietary carbohydrates on the nutritional quality of foods. In particular, attention has been focused on the relationship between the various types of carbohydrate containing foods and the different postprandial glucose responses by these foods post ingestion [1–7]. The glycemic index (GI) is a physiological classification widely accepted for carbohydrate foods based on their ability to raise the concentration of glucose in the blood [7–9]. Bakery foods, muffins for example, are regarded as a high glycemic impact food due to the high concentration of sugar contained in the muffins. Previous research [10,11] has shown that the over-consumption of sucrose can lead to a number of metabolic complications including hyperinsulinemia, hyperglycemia, hypertension and insulin resistance, as well as being related to dyslipidemia and ectopic lipid deposition in healthy subjects with diabetes [12]. Indeed, high GI food products are quickly digested and their carbohydrate is

rapidly absorbed, resulting in higher blood glucose levels [13]. On the contrary, the health benefits of the low GI products are thought to be derived from the slower the rate of carbohydrate absorption, consequently leading to a gradual rise in blood glucose level and better glycemic control [14].

The food industry has focused on reducing the calorific content of food to promote a healthier diet. Therefore, different natural sweeteners have been used in sugar-reduced or sugar-free products based on their multiple potential health benefits and functional properties, including maintaining sweetness and acceptable texture [15–18].

Steviol glycosides have been extracted and purified from the leaves of *Stevia rebaudiana* Bertoni, commonly known as stevia; they are naturally sweet-tasting, have good solubility in water, good temperature and pH stability [19–21] as well as having no calorific value [22], allowing them to be used as a sugar substitute or natural sweetener. Stevioside and rebaudioside A are the major glycoside constituents responsible for sweetness and are the most abundant glycosides in the *Stevia rebaudiana* Bertoni plant [23–25]. They are very useful as a food additive due to their relative sweetness being 250–300 times sweeter than table sugar [26].

Extracts from stevia have broad health-promoting properties for blood glucose and insulin levels in human studies [27]. Steviol glycosides are not hydrolyzed by human digestive enzymes of the mouth, stomach, and small intestine [28]. However, rebaudioside A and stevioside are hydrolyzed (in vitro and in vivo) to aglycone steviol by colon microflora through the successive removal of glucose units [29]. Chang et al. [27] reported that insulin sensitivity is increased due to stevia consumption in rodent models, and thus does not increase blood glucose and insulin levels [22]. Furthermore, previous work has found that a reduction in the predicted glycemic response was observed due to 50% or 100% replacement of sucrose with Stevianna® in muffins during in vitro digestion experiments [30]. Therefore, stevia has the potential to be a low-cost natural sweetener due to important pro-health properties, such as being non-calorific, non-fermentable and non-toxic as well as having a high-intensity sweetness [31], and it is also recommended as a treatment for diabetics and obese persons [23].

However, several studies have shown that the utilization of stevia as a sugar replacer in baking leads to a negative effect on appearance, compactness, moisture and texture of the bakery products structure [17,32,33]. These results have indicated that stevia is not acceptable to replace sucrose completely in bakery products as stevia exhibits high-intensity sweetness but does not possess the necessary bulking characteristics [34]. That is why Stevianna® (product code ST001 SE supplied by Stevianna® NZ) is used for our study, as it incorporates rebaudioside A (98% steviol glycoside; 1%) with erythritol (99%).

Erythritol is a four-carbon sugar alcohol or polyol with approximately 60% to 80% of the sweetness of sucrose [35]. It is not only a sweetener but also a bulking agent, and thus can be used as a sugar replacer in bakery products. Partial replacement of sucrose with erythritol had no negative influence on physical quality characteristics in a baked product [34,36]. In addition, previous studies reported that erythritol is useful as a non-glycemic and low-calorie sweetener that is safe for diabetics [37,38]. Erythritol has been demonstrated to have a small molecular size, thus it is rapidly absorbed by the small intestine and does not undergo systemic metabolism by the human body [37,39]. Some research has shown that the combination of a high-intensity sweetener with bulking agents or fibers in sugar-reduced formulations of food resulted in bakery products with acceptable physical quality [26,29,40,41].

None of these previous studies assessed a complex food sweetener to replace traditional sugar in bakery products. The aim of the study was to evaluate the replacement of sugar with Stevianna® (1 × sweetness of sucrose) and the addition of cocoa powder and/or vanilla to muffins for their physical properties and glycemic response, compared with a control muffin formulation with no added Stevianna®, cocoa powder, or vanilla.

#### **2. Materials and Methods**

#### *2.1. Raw Materials*

Wheat flour (Medal Premium baker flour, Champion, Auckland, New Zealand), white sugar (Chelsea, Auckland, New Zealand), baking powder (Edmonds, Christchurch, New Zealand), iodized table salt (Cerebos, Auckland, New Zealand), skim milk powder (Pams, Auckland, New Zealand), 100% cocoa powder (Cadbury, Dunedin, New Zealand), vanilla (Hansells, Sydney, Australia), canola oil (Pams, Auckland, New Zealand), and fresh eggs were purchased from a local supermarket and tap water was used. Muffins were prepared containing 0%, 50% and 100% Stevianna® (produce code ST001\_SE; Stevianna®, Auckland, New Zealand) as a replacement for sucrose. Stevianna® utilizes Reb-A 98% steviol glycoside as the main sugar substitute along with erythritol.

#### *2.2. Mu*ffi*n Preparation*

The muffin recipe was adapted from a previous study [30] and is given in Table 1. The Stevianna® was dissolved in the water and mixed with liquid whole egg and oil. After that, the dry ingredients were added into the liquid components and mixed for 5 min. The batter was poured into a paper baking case in a muffin pan. The muffins were baked for 18 min in a preheated Simpson Gemini Atlas series oven at 180 ◦C set to fan bake. Baked muffins were cooled at room temperature for 1 h, then packed in plastic resealable bags and stored in a refrigerator at 4 ◦C until physical analysis.

#### *2.3. Mu*ffi*n Height*

The muffin product was taken out from the paper baking case, and the muffin height was measured with an electronic caliper (INSIZE) from the highest point of the muffin to the bottom of the muffin.

#### *2.4. Moisture Content*

A domestic kitchen food chopper (Zyliss®) was used to crush and homogenize the muffin (crust and crumb) of each formulation. Approximately 4 g was dried in an air oven at 105 ◦C for 16 h, until no further weight change.

The moisture content (MC) was calculated using the following equation:

$$\text{MCC (\%)} = (\text{W}\_{\text{before drying}} - \text{W}\_{\text{after drying}} / \text{W}\_{\text{before drying}}) \times 100 \tag{1}$$

where W denotes weight (g).

#### *2.5. Mu*ffi*n Volume*

The volume of the muffins was measured by the rapeseed displacement method. Each muffin was placed in a plastic beaker of known volume (total volume, Vt), and the remaining space in the plastic beaker was then filled with rapeseed; the volume of the rapeseed required (Vs) was then determined by graduated cylinder. Muffin volume was calculated as the difference between the total volume and volume of rapeseed—the muffin volume = Vt − Vs [36].

#### *2.6. Mu*ffi*n Texture*

A texture analyzer (TA.XT. Plus, Stable Microsystems, Surrey, UK) was used to measure the texture profile of muffins in terms of the firmness and springiness of the samples. The samples were compressed to a strain of 25% of the original height using a 75 mm cylindrical probe and a 50 kg load cell, and a test speed of 1.0 mm/s was used. Data was obtained from the Texture expert software (Stable Microsystems, Surrey, UK). Firmness and springiness values were calculated as the overall force of compression required and the resistance post compression.



60

Stevianna + Cococa + Vanilla (100S + CP + V).

#### *2.7. Mu*ffi*n Total Starch*

Total starch analysis was carried out according to the official American Association of Cereal Chemists method 76.13 [42], using Megazyme (Bray, Dublin, Ireland) total starch kit.

#### *2.8. In Vitro Predictive Glycemic Response Digestion Analysis*

The procedure used for the determination of potential glycemic response is the same as that reported previously by [30]. This procedure measures the breakdown of carbohydrates to sugars by the action of amylase enzymes added to the baked muffin. Whole muffins were chopped with a domestic kitchen food chopper (Zyliss®) to stimulate particle size reduction which occurs during natural mastication for at least one minute of steady chopping until a fine crumb was achieved. A 3.5 g sample was used to determine the predictive glycemic response.

Triplicate samples of product (approximate 1 g of cooked muffin) were each placed into the 60 mL plastic pots and 30 mL of distilled water added, and duplicate blank samples. These pots were inserted to a pre-heated 15 place magnetic heated stirring block (IKAMAG® RT15, IKA®-WERKE Gmblt & Co., Staufen, Germany) preheated to 37 ◦C, on each pot one magnetic stirrer, followed by 0.8 mL of 1 M aqueous HCl. Then, 1 mL of a 10% pepsin (Acros Organics, New Jersey, NJ, USA CAS: 901-75-6) solution in 0.05 M HCl was added in order to replicate gastric digestion. The sample was incubated at 37 ◦C for 30 min with slow constant stirring (130 rpm) to simulate gastric digestion conditions. In vitro stomach digestion was halted by the addition of 2 mL NaHCO3. Small intestine digestion was mimicked by the addition of 5 mL 0.1 M Na maleate buffer pH 6. An aliquot (1 mL) was withdrawn (Time 0) and added to 4 mL absolute ethanol to stop any further enzyme reaction. A 0.1 mL dose of amyloglucosidase (A.niger, Megazyme, E-AMGDF; 3260 U/mL) was added to prevent end-product inhibition of pancreatic amylase. A 5 mL 2.5% pancreatin (EC: 232-468-9, CAS: 8049-47-6, activity: 42362 FIP-U/g, Applichem GmbH, Darmstadt, Germany) in 0.1 M Na maleate buffer pH 6 followed by the volume being made to 53 mL with continued stirring and heat maintained at 37 ◦C for 120 min. Triplicate 1 mL aliquots were withdrawn at 0, 20, 60, 120 min and added to 4 mL absolute ethanol. Reducing sugar content was analyzed by dinitrosalicyclic (DNS) colorimetry, and the area under the curve (AUC) was calculated by dividing the graph into trapezoids as described elsewhere [30]. The reducing sugar content was regarded as an indicator for the predictive glycemic response.

#### *2.9. Statistical Analyses*

All analyses were conducted in triplicate. Analysis of variance (one-way ANOVA) was performed on the data, and Tukey's comparison test (*p* < 0.05) was used to determine the significance. These analyses were performed using Minitab (Minitab Pty Ltd., Sydney, Australia).

#### **3. Results and Discussion**

#### *3.1. Moisture Content*

Table 2 shows that the moisture content of muffin samples ranged from 19% to 27%. The moisture content of the muffin samples produced was higher when cocoa powder or/and vanilla was used. In addition, Figure 1 shows that moisture content values increased significantly (*p* < 0.05) when sucrose was replaced by Stevianna®—in particular the moisture content of 100% Stevianna® samples were higher than the full-sucrose muffin samples. Sucrose plays an important role in water retention that results in reduced moisture loss during the baking of the muffins [43]. However, the moisture content increased when sucrose was replaced because the Stevianna® acted as a humectant and prevented water from escaping during baking. Research using other types of sugar replacers has shown similar results. Martínez-Cervera et al. [44] used erythritol in muffins for its water retention properties. Ghosh and Sudha [45] showed that the use of the polyol sorbitol was reflected in a significantly higher moisture content (*p* < 0.05). Due to the high water-binding capacity of formulations with carbohydrate-based sugar replacers, a greater amount of water is required in cereal products.


**Table 2.** Effect of Stevianna on texture profile analysis and total starch in muffins with or without cocoa powder and/or vanilla.

Control (C); Vanilla (V); Cocoa Powder (CP); Cocoa+Vanilla (CP + V); 50% Stevianna (50S); 50% Stevianna + Vanilla (50S + V); 50% Stevianna + Cocoa (50S + CP); 50% Stevianna + Cocoa + Vanilla (50S + CP + V); 100% Stevianna (100S); 100% Stevianna + Vanilla (100S + V); 100% Stevianna + Cocoa (100S + CP); 100% Stevianna + Cococa + Vanilla (100S + CP + V). All measurements are the mean values ± SD of triplicate determinations. Means in the same column with different letters are significantly different (*p* < 0.05).

**Figure 1.** Moisture content for muffins of formulation made from two levels of Stevianna without/with cocoa powder and/or vanilla. Control (C); Vanilla (V); Cocoa Powder (CP); Cocoa + Vanilla (CP + V); 50% Stevianna (50S); 50% Stevianna + Vanilla (50S + V); 50% Stevianna + Cocoa (50S + CP); 50% Stevianna + Cocoa + Vanilla (50S + CP + V); 100% Stevianna (100S); 100% Stevianna + Vanilla (100S + V); 100% Stevianna + Cocoa (100S + CP); 100% Stevianna + Cococa + Vanilla (100S + CP + V). Values with different letters are significantly different to one another *p* < 0.05.

Moisture content in bakery products is an important factor as it has a direct impact on the texture attributes and a strong correlation has been found between moisture content and firmness [46]. As can be seen from the Table 2, muffin firmness increased as moisture content increased. As reported by Rößle et al. [47], this must be related to the replacement of the sugar by Stevianna®, affecting the formation of muffin structure.

#### *3.2. The Impact of Sugar Replacement on Product Physico-Chemical Characteristics*

The height of the muffins prepared with the different levels of Stevianna® with/without cocoa powder and/or vanilla is shown in Figure 2. The full-sucrose muffin was significantly higher (*p* < 0.05) than the muffins that were prepared using Stevianna®. The lowest height was found in the 100% Stevianna® muffin samples. The full-sucrose muffin with cocoa powder and/or vanilla group had a greater height than the control and other samples (Figure 2). These results indicate that the decrease in muffin height was associated with an absence of interconnectivity of a more compact structure and with a low number of air cells for levels of sucrose replacement higher than 50% (Figure 3).

**Figure 3.** Effect of two levels of Stevianna with/without cocoa powder and/or vanilla in muffins: Control (C); Vanilla (V); Cocoa Powder (CP); Cocoa + Vanilla (CP + V); 50% Stevianna (50S); 50% Stevianna + Vanilla (50S + V); 50% Stevianna + Cocoa (50S + CP); 50% Stevianna + Cocoa + Vanilla (50S + CP + V); 100% Stevianna (100S); 100% Stevianna + Vanilla (100S + V); 100% Stevianna + Cocoa (100S + CP); 100% Stevianna + Cococa + Vanilla (100S + CP + V).

Photographs of vertical cross-sections of the different muffin formulations are shown in Figure 3. As the Stevianna® content increased, in the formulations, the air bubbles became smaller and the air channels gradually diminished. This could be due to the fact that muffins with a full sucrose content gained an increased number of air bubbles during the beating of the batter, and these air bubbles are then expanded by carbon dioxide and water vapor pressure generated during baking, resulting in the formation of air channels, which influence the texture of the finished muffin product. The lack of air channels as the sucrose was replaced may also be associated with earlier thermosetting of the batter during the heating process in the oven, therefore, not allowing enough time for bubble expansion and formation of air channels [43,44]. Martínez-Cervera et al. [44] also found that the number of small air bubbles increased, air channels diminished, and circular bubbles increased with an increase in sucrose replacement by polydextrose and sucralose in a muffin product.

The volume of the muffin is an important indicator of air bubble expansion during baking and consequently also of the porous structure of the product. The volumes of muffins prepared with different levels of Stevianna® with/without and/or vanilla along with the control muffin are presented in Figure 4A. The samples with 100% Stevianna® muffin group had significantly lower volumes (*p* < 0.05) compared to those of the full-sucrose muffin products. Muffin density appeared to be negatively correlated with muffin volume (Figure 4B). The density of the muffins was calculated from mass and volume after baking. Table 2 illustrates that when sugar was completely substituted with Stevianna®, there was a significant increase (*p* < 0.05) in muffin density. Additionally, product quality characteristics such as springiness and firmness were greatly affected (Table 2). These results indicate that an increase in the level of Stevianna® had an adverse effect on volume, density and texture of the muffin. Manisha et al. [26] also reported that replacement of sucrose with 100% stevioside and liquid sorbitol caused a significant deterioration in quality which decreased volume and resulted in a firmer texture in cake properties.

A function of sugar during cake baking is that it delays starch gelatinization, thus contributing to the aeration of the batter and the optimum quality of sugar will affect formation of the cake structure and improve crumb texture and tenderness [26]. The decrease in sugar-free muffin expansion is the result of less air bubble incorporation and reduced air holding capacity during baking [48]. In addition, starch gelatinization temperature seems to contribute to volume development due to different interactions between the Stevianna® and starch and proteins of the batter, and these interactions affect starch gelatinization and protein denaturation temperatures. These results are in agreement with Ronda et al. [49]'s findings which showed that a decrease in starch gelatinization and protein denaturation temperatures in sorbitol cakes is expected to cause a premature thermosetting of protein or starch matrix—this process will start at the crust due to direct contact with the heating medium. Therefore, this lowers the heat transfer rate, and produces a vapor pressure build-up, resulting in inadequate expansion of individual bubbles. Additionally, Ronda et al. [49] found that high-fructose corn syrup (HFCS) mainly contributed to the early gelatinization of starch during the baking process and restricted the volume of baked products compared to sucrose.

However, the 50% Stevianna® used had no significant effect on the volume and density of muffin compared to the full-sucrose muffin samples (Figure 4). These results suggest that muffin samples containing half the amount of Stevianna® have a similar ability, compared with muffins with full sucrose, to retain air. These results are consistent with those of Lin et al. [38], who found no significant differences among the volume estimates for 50% erythritol cakes. Furthermore, the addition of the 50% Stevianna® in muffin samples exhibited a texture close to that of the full-sucrose muffin samples (Table 2), which conferred an appearance of firmness and springiness. The results were consistent with previous research [30].

#### *3.3. The Impact of Sugar Replacement on the In Vitro Predictive Glycemic Response*

The total starch of modified muffins was measured and compared with the control sample (Table 2). Compared to the control muffin, 50% or 100% sucrose replacement with Stevianna® with added cocoa powder samples had significantly lower amounts of total starch. Similar levels of total starch were observed in control and full-sucrose muffin samples—50% and 100% Stevianna® with/without cocoa powder and/or vanilla muffin samples. Thus, the presence of cocoa powder with Stevianna® in muffin had a significant effect on total starch contents.

The effects of Stevianna® on in vitro starch digestion in muffin and chocolate muffin products were investigated by measuring the glucose released during starch digestion. Figure 5 shows the reducing sugars curves of two levels of Stevianna® with/without cocoa powder and/or vanilla muffin samples that were compared with full-sucrose with/without cocoa powder and/or vanilla samples, respectively. These two levels of Stevianna® used in this study were found to decrease reducing sugars released by digestive enzymes, compared with the full-sucrose muffin samples. The rate and extent of reducing sugars released were the highest in the control muffin, followed by 50% Stevianna® with/without cocoa powder and/or vanilla muffin products, and 100% Stevianna® with/without cocoa powder and/or vanilla muffins (Figure 5). In particular, muffins with Stevianna® showed a significant decrease in terms of reducing sugars released throughout the 120 min starch digestion process.

The total area under the hydrolysis curve (AUC) relates the total glucose release to the digestion time of 120 min. The concentration of the Stevianna® had a significant effect on the AUC values (*p* < 0.05), which demonstrated that the replacement of sucrose with 100% Stevianna® resulted in the lowest AUC value of muffin samples in a dose response (Figure 6). It is of interest that the additions of vanilla and/or cocoa powder with muffin production did not lead to a significant reduction of in vitro digestion values compared to the full-sucrose—50% Stevianna®, and 100% Stevianna® samples, respectively. These results are consistent with the previous report by Gao et al. [30].

**Figure 5.** Amount of reducing sugars released per g of food material during in vitro digestion. Control (C); Vanilla (V); Cocoa Powder (CP); Cocoa + Vanilla (CP + V); 50% Stevianna (50S); 50% Stevianna + Vnilla (50S + V); 50% Stevianna + Cocoa (50S + CP); 50% Stevianna + Cocoa + Vanilla (50S + CP + V); 100% Stevianna (100S); 100% Stevianna + Vanilla (100S + V); 100% Stevianna + Cocoa (100S + CP); 100% Stevianna + Cococa + Vanilla (100S + CP + V).

**Figure 6.** Values for area under the curve (AUC) comparing the control and other low-sugar muffins made with two levels of Stevianna with/without cocoa powder and/or vanilla. Control (C); Vanilla (V); Cocoa Powder (CP); Cocoa + Vanilla (CP + V); 50% Stevianna (50S); 50% Stevianna + Vanilla (50S + V); 50% Stevianna + Cocoa (50S + CP); 50% Stevianna + Cocoa + Vanilla (50S + CP + V); 100% Stevianna (100S); 100% Stevianna + Vanilla (100S + V); 100% Stevianna + Cocoa (100S + CP); 100% Stevianna + Cococa + Vanilla (100S + CP + V). Values with different letters are significantly different to one another *p* < 0.05.

This study did not focus on the impact of sweeteners on in vitro starch digestion analysis of bakery products. However, several research projects have been designed to test the effects of the stevia or erythritol on postprandial glucose and insulin levels in vivo and in vitro digestion methods as compared to sucrose [50,51].

The breakdown or disruption of starch granules that results from salivary amylase causes a greater susceptibility of the granule to further enzyme degradation. This process will lead to more readily digestible starch, and hence create a higher blood glucose response [52]. The level of postprandial blood glucose is a major factor in predicting the profile of insulin resistance. Alizadeh et al. [50] found that there were differing effects on postprandial blood insulin levels that were dependent on the type and amount of sweetener consumed. The effect of the consumption of beverages containing stevia has been tested by measuring the in vivo glycemic impact [53], and it was found that postprandial glucose and insulin levels were significantly reduced in the stevia beverages compared to the sucrose beverages. These effects on postprandial glucose levels are mainly due to the lack of calories and carbohydrate content of Stevianna®, and thus there are no reducing sugars released. A similar trend has been observed in that the postprandial insulin levels were reduced in stevia ice cream samples compared to full-sucrose ice cream samples [50], and this is most likely due to the functional properties of stevia that results in no contribution to the available carbohydrate and glycemic response in food products. In addition, Roberts and Renwick [54] illustrated that steviol glycosides are not readily absorbed by the upper small intestine when it is administered orally to normal rat or human subjects. There are no human digestive enzymes present in the small intestine to hydrolyze the β-glycosidic linkages, resulting in limited small intestine digestion.

Lin et al. [36] illustrated that 0%–100% sugar replacement with erythritol in cookies decreased the carbohydrate contents by in vivo digestion. Since the calorie value of erythritol is approximately 0.4 kcal/g [39], it provides no energy to the body and thus it is not systemically metabolized nor fermented in the colon [37]. It has been suggested that the consumption of erythritol does not raise postprandial glycemic and insulin levels by oral ingestion in healthy human subjects [28]. In a previous study [39], more than 90% of erythritol is rapidly absorbed by the small intestine when eaten and is excreted unchanged in the urine.

The Stevianna® used in our study was composed of rebaudioside A (stevia) and erythritol and, therefore, the observations made are consistent with those made by the above studies. Our experiment results showed that under in vitro conditions a lower reducing sugar liberation took place when sucrose was replaced by Stevianna® in muffins, and consequently this can be beneficial to as it will decrease the postprandial blood glucose. Additionally, it is probable that the intake of these muffins decreases the rate of intestine absorption of glucose and delays gastric emptying.

#### **4. Conclusions**

The stevia-containing product, Stevianna®, has been shown to be a suitable sucrose replacement for a low-sucrose formulation of muffins. The results showed that 50% sugar replacement with Stevianna® had similar physical quality characteristics in terms of volume, density and texture to a control muffin. However, when the sugar was replaced by 100% Stevianna®, the muffin quality showed a reduction in volume, an increase in textural firmness and a correspondingly high density of the product when compared to the control muffin samples. Furthermore, Stevianna® was able to simulate sucrose functionality in muffins, producing an increase in moisture content in comparison with the full-sucrose muffins. The negative effect of Stevianna® on muffin properties can be associated with the fact that as the Stevianna® level was raised, it led to a reduction of air bubble expansion during the heating process (possibly due to the weakening of the starch–protein–sugar interface of the muffin, allowing for greater structural collapse) and thus a corresponding reduction in height. This research illustrates that Stevianna® is a major factor impacting on the physical characteristics of muffins. The addition of cocoa powder and/or vanilla did not affect the quality of muffins significantly.

In relation to the nutritional quality of the muffin products, the effect of Stevianna® inclusion on the predicted glycemic impact as determined by in vitro digestion illustrated the role of sugar in elevating the glycemic response during digestion. The replacement of sugar with increasing levels of Stevianna® was found to significantly decrease the potential glycemic response values, and this is most likely to be attributed to the fact that Stevianna® was not degraded into glucose units and acted as an inert filler within the muffin samples. Therefore the inclusion of cocoa powder and/or vanilla powder did not have a significant change to the predicted glycemic response values of the muffins.

The breakdown or disruption of starch granules that results from salivary amylase causes a greater susceptibility of the granule to further enzyme degradation. This process will lead to more readily digestible starch, and hence create a higher blood glucose response [52]. The level of postprandial blood glucose is a major factor in predicting the profile of insulin resistance. Alizadeh et al. [50] found that there were differing effects on postprandial blood insulin levels that were dependent on the type and amount of sweetener consumed. The effect of the consumption of beverages containing stevia has been tested by measuring the in vivo glycemic impact [53], and it was found that postprandial glucose and insulin levels were significantly reduced in the stevia beverages compared to the sucrose beverages. These effects on postprandial glucose levels are mainly due to the lack of calories and carbohydrate content of Stevianna®, thus there are no reducing sugars released. A similar trend has been observed in that the postprandial insulin levels were reduced in stevia ice cream samples compared to full-sucrose ice cream samples [50], and this is most likely due to the functional properties of stevia that results in no contribution to the available carbohydrate and glycemic response in food products. In addition, Roberts and Renwick [54] illustrated that steviol glycosides are not readily absorbed by the upper small intestine when it is administered orally to normal rat or human subjects. There are no human digestive enzymes present in the small intestine to hydrolyze the β-glycosidic linkages, resulting in limited small intestine digestion.

Lin et al. [36] illustrated that 0%–100% sugar replacement with erythritol in cookies decreased the carbohydrate contents by in vivo digestion. Since the calorie value of erythritol is approximately 0.4 kcal/g [39], it provides no energy to the body and thus it is not systemically metabolized nor fermented in the colon [37]. It has been suggested that the consumption of erythritol does not raise postprandial glycemic and insulin levels by oral ingestion in healthy human subjects [28]. In a previous study [39], more than 90% of erythritol is rapidly absorbed by the small intestine when eaten and is excreted unchanged in the urine.

Finally, it can be seen that a partial replacement of Stevianna® for sucrose with/without cocoa powder and/or vanilla in muffins gave a product with quality characteristics close to that of the full-sucrose muffin sample. At the same time, the reduction in potential glycemic response values was greater than would have been expected with 50% sucrose reduction and consequently providing a quality muffin that produces a lowered postprandial response with the potential associated health benefits.

**Author Contributions:** J.G., M.A.B., C.S.B., X.G. and X.-A.Z. conceived and designed the experiments; J.G. and X.G. performed the experiments; J.G., M.A.B., S.L.M. and C.S.B. analyzed the data; J.G., C.S.B. and M.A.B. were responsible for writing the manuscript.

**Funding:** The research was supported by Lincoln University postgraduate funding. This research was also supported by the 111 Project (B17018) as well as S&T projects of Guangdong Province (2015A030312001 and 2013B020203001).

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **Comprehensive Nutrition Review of Grain-Based Muesli Bars in Australia: An Audit of Supermarket Products**

#### **Felicity Curtain 1,\* and Sara Grafenauer 1,2**


#### Received: 25 July 2019; Accepted: 23 August 2019; Published: 28 August 2019

**Abstract:** Muesli bars are consumed by 16% of children, and 7.5% of adults, and are classified as discretionary in Australian Dietary Guidelines, containing "higher fat and added sugars" compared with core food choices. This study aimed to provide a nutritional overview of grain-based muesli bars, comparing data from 2019 with 2015. An audit of muesli bars, grain-based bars, and oat slices was undertaken in January 2019 (excluding fruit, nut, nutritional supplement, and breakfast bars) from the four major supermarkets in metropolitan Sydney. Mean and standard deviation was calculated for all nutrients on-pack, including whole grain per serve and per 100g. Health Star Rating (HSR) was calculated if not included on-pack. Of all bars (*n* = 165), 63% were ≤ 600 kJ (268–1958 kJ), 12% were low in saturated fat, 56% were a source of dietary fibre, and none were low in sugar. Two-thirds (66%) were whole grain (≥8 g/serve), with an average of 10 g/serve, 16% of the 48 g Daily Target Intake. HSR featured on 63% of bars (average 3.2), with an overall HSR of 2.7. Compared to 2015, mean sugars declined (26.6 g to 23.7 g/100 g; *p* < 0.001), and 31% more bars were whole grain (109 up from 60 bars). Although categorised as discretionary, there were significant nutrient differences across grain-based muesli bars. Clearer classification within policy initiatives, including HSR, may assist consumers in choosing products high in whole grain and fibre at the supermarket shelf.

**Keywords:** muesli bars; grains; whole grain; dietary fibre; snack foods; nutrition

#### **1. Introduction**

'Muesli bar' is a generic term that refers to baked or cold-formed cereal-based snack bars, and may contain other ingredients such as fruit, nuts, seeds, chocolate, yoghurt, and a variety of other fillings and/or toppings [1]. They are a popular food in Australia, with consumption per capita considered the third highest worldwide, behind Canada and the USA [2]. An estimated 7.5% of Australian adults ate muesli bars the day prior to the 2011–12 Australian Health Survey, with consumption more common in younger age groups (16% of 4–13 year olds, compared to 12.8% of 14–18 year olds, and less than 8% of those aged 19–50 years) [3]. Their popularity with children was noted in a 2005 paper reviewing the lunchbox content of Australian school children, which found an estimated 41.8% of lunchboxes included a muesli/fruit bar, though this also included non-grain-based bars, excluded from this research [4].

Data from the 2011–12 Australian Health Survey found muesli bars contributed overall less than 1% of total energy, protein, fats, sugars, and dietary fibre to Australians aged 2 years and older [5]. However for females aged between 2–18 years, these figures were slightly higher; 1.1% energy, 1.2% total sugars, and 1.5% dietary fibre, and for males 2–18 years; 1.2% energy, 1% saturated fat, 1.4% total sugars, and 1.6% dietary fibre [5]. There is a lack of consensus on what constitutes a 'snack food', with definitions ranging from foods consumed between main meals or at specific times of day, food-type,

or participant-described. Based on 'time of day' consumption, bars can be considered a snack food, generally eaten between main meals, and snacking of this kind has been linked with concern around increased risk of obesity and related chronic disease [6], though importantly, these health outcomes are multifactorial, with food choice and energy balance key in determining whether snacking is a healthful or harmful food behaviour [7,8].

Between 1995 and 2012, the prevalence and frequency of children snacking (defined as a single eating occasion between main meals) rose in Australia, with more than double the number of children snacking four or more times per day in 2012 [9]. Subsequently, the contribution of snacks to total energy intake significantly increased, from 24–30.5%. Foods consumed as snacks were a mix of traditional 'snack' foods such as sweet biscuits, cakes, fresh fruit, and 'meal' foods, such as bread and milk. Fruit and vegetable juice was the top contributor to energy from snacks in 1995, but did not appear in 2012, with pome fruit moving up as the top contributor. Muesli bars did not feature in the top snacks in 1995, but were number seven in 2007, and number nine in 2012, where they contributed an estimated 12.5% of total energy to snacks [9]. In Australian adults, cakes, muffins, scones, breads, and dairy milk were the three greatest contributors to energy from snacks, with 22% of total energy derived from snacking occasions [10]. While no data has reviewed changes in snacking habits among Australian adults, steady increases from 1977–2006 amongst adults in the USA mirror Australian children's results, contributing more kilojoules, mainly from discretionary foods like desserts, sugar sweetened beverages, and salty snacks [11].

The popularity of muesli bars, and increasing levels of consumption [9] have attracted attention from public health groups, government, and the media, not least since they are considered a 'discretionary' food in the Australian Dietary Guidelines, where their consumption is discouraged based on having "higher fat and added sugars" [12]. Importantly, they are not depicted in the accompanying Australian Guide to Healthy eating, which visually represents core and discretionary foods. Instead, muesli bars are listed in the longer form supplementary text, and are therefore hidden from view, so it is unclear how well understood their classification as discretionary is among consumers. Similarly, the New Zealand Eating and Activity Guidelines present muesli bars as an example of a 'highly processed' food that may be refined and contain added saturated fat, sugar, and salt [13], and the United Kingdom's Eat Well Guide cautions that cereal bars may have high levels of added sugars [14].

In 2018, proposed sugar reformulation targets for muesli bars were developed by The Healthy Food Partnership, an initiative established by the Australian Government in 2015, which aims to improve public health nutrition through several policy areas, including food reformulation [1]. Their inclusion was noteworthy, as they did not comply with the initial criteria (contributing significantly (≥1%) to sodium, sugars, and/or saturated fat in the Australian population's intake), instead being included based on their high level of consumption among children [1]. The proposed targets call for a "10% reduction in sugar across defined products containing over 28 g sugar/100 g, and a reduction in sugar to 25 g/100 g for products between 25–28 g sugar/100 g by the end of 2022". It is important to recognise that many companies have their own nutrition policies and commitments, as outlined in a 2018 Australian report, which found 16 of the 19 food companies surveyed included nutrition in their corporate strategy and had a commitment to product reformulation, while 11 out of 19 had committed to implementing the voluntary Health Star Rating (HSR) system [15].

The HSR is an interpretive Front of Pack Labelling system, first introduced in Australia and New Zealand in 2014, as a joint initiative between Government, public health, industry, and consumer groups. The system uses an algorithm to assign a star rating between 0.5–5 stars, and is intended to aid consumers in making healthier choices within categories [16,17]. The HSR algorithm rates foods on a per 100 g basis, considering both 'negative nutrients' (kilojoules, saturated fat, total sugars, and sodium), and 'positive' elements (fruit, vegetables, nuts and legumes, as well as protein and dietary fibre in some cases), which is then converted to a star rating [18]. Muesli bars were a key category of consideration in the ongoing HSR 5-year review, which noted they had received negative media

attention based on products scoring "inappropriately high scores", despite their categorisation as discretionary foods [19].

However, grain-based muesli bars may also be a potential source of positive ingredients and nutrients within the diet pattern, particularly considering whole grain and dietary fibre content, which are promoted within Australian Dietary Guidelines [12]. Widespread evidence supports whole grains and whole grain foods for their protective health benefits, including lower total and cause-specific mortality, type 2 diabetes [20–24], weight gain [25], and colorectal cancer [26]. Globally, low whole grain intake has been recognised as the second greatest dietary risk factor for mortality (behind sodium), and the greatest dietary risk factor for morbidity, responsible for more than 80 million Disability-Adjusted-Life-Years [27]. Irrespective of its well-documented health benefits, whole grain intake in Australia is low, with follow up data from the Australian Health Survey recording median intake for children at 16.5 g per day, and adults at 21.2 g/day—both less than half of the established Daily Target Intake (DTI) of 48 g per day for adults, and between 32 and 40 g per day for children [28–30]. Equally, a large body of evidence points to the benefits of dietary fibre and its role in reducing chronic disease risk, yet most Australians fall short, with more than half of children, and more than 70% of adults not meeting their respective targets [31].

Due to their popularity and increasing consumption in Australia, muesli bars are often criticised and met with confusion regarding their nutritional value, with a particular focus on sugar content. This study aimed to provide an overview of the nutritional status of grain-based muesli bars on shelf including muesli bars, grain-based bars, and oat slices in Australian supermarkets, and provide a comparison of 2019 with 2015 data.

#### **2. Materials and Methods**

An audit of grain-based muesli bars was conducted January 2019, in four major supermarkets in metropolitan Sydney (Aldi, Coles, IGA, and Woolworths). Collectively, these supermarket chains make up more than 80% of total Australian market share, and were chosen in preference to smaller, independent grocery stores in an attempt to reflect food choices that the majority of Australians are faced with during food shopping [32]. This recognised process has been outlined in previously published research [33] and the same process was utilised in the data collection from 2015. Smartphones were used to capture all information on food packaging, including ingredient lists, Nutrition Information Panels (NIP), health and nutrition claims, HSR, and any additional logos and endorsements. Outlined in Table 1 below, products accounted for in the audit included muesli bars, grain-based bars, (including fruit-filled bars and twists, and those made from wheat, puffed rice, or other grains), and oat slices. Products were further categorised to determine whether they were specifically marketed towards children, by the presence of cartoons, promotions, or sporting figures, as described in previous research [34,35]. Products excluded were fruit-based bars, fruit leather/straps, nutritional supplement bars (e.g. protein/'low-carb' bars), nut/seed based bars, and breakfast bars/biscuits (e.g. those designed as a meal replacement, indicated in the product name), in line with exclusions within the Healthy Food Partnership proposed reformulation targets [1]. A supplementary internet search was conducted through supermarket websites and identified manufacturer websites using key words such as "snack bars", "muesli bars", "grain-based bars", "oat slices", and "snack bars", to ensure all products were captured.

Data from photographs taken at both timeframes (2015 and 2019) were transcribed into a Microsoft® Excel® spreadsheet (Version 2013, Redmond, Washington, DC, USA) for analysis. Information for the data entry included the NIP per serve and per 100 g, ingredients, percentage of whole grains, nutrition and health related claims, including whole grain, protein, dietary fibre, saturated fat, sugars, and sodium. Eligibility for products to make nutrition content claims was also assessed, in line with Food Standards Australia New Zealand and GLNCs Code of Practice for Whole Grain Ingredient Content Claims (The Code) [30], as well as proportion of products meeting the Healthy Food Partnership proposed reformulation targets for sugar reduction. HSR was not collected in 2015 as this was not on

pack at this time. Where HSR was not featured on packaging, it was calculated for all products using the HSR website calculator [36]. A second, independent reviewer checked data for any inconsistencies and errors, and results were compared with 2015 data that followed the same process, to assess changes.



#### *Statistics*

All data were checked for normality using Shapiro–Wilk test (IBM SPSS®, version 25.0, IBM Corp., Chicago, IL, USA) and mean and standard deviation were presented. As expected, there were missing values for dietary fibre and whole grain as these are often not presented unless a claim is being made on-pack, therefore dietary fibre and whole grain were analysed separately.

One-way ANOVA with post hoc Tukey analysis (IBM SPSS®, version 25.0, IBM Corp., Chicago, IL, USA) was used to compare differences per serve and per 100 g between (1) muesli bars, (2) grain-based bars, (including fruit-filled bars and twists, and those made from wheat, puffed rice, or other grains), and (3) oat slices for all available nutrients reported on-pack, including where relevant, dietary fibre, whole grain (g and %) and HSR (per 100 g). Independent samples *t*-test (IBM SPSS®, version 25.0, IBM Corp., Chicago, IL, USA) was used to compare whole grain and refined grain bars, which was defined according to each product's eligibility for registration with The Code (≥8 g whole grain per manufacturer serve), a method that has been described in previously published research [37]. *T*-tests were also used to determine difference in HSR for all products /100 g, between whole grain and refined grain categories and for data per 100 g from 2015 compared with 2019.

#### **3. Results**

Data from 165 bars were collected, including 96 muesli bars, 46 grain-based bars, and 23 oat slices from 18 manufacturers where the top three (Nestle Ltd., Kellogg (Aust) Pty. Ltd. and Carman's Fine Foods Pty. Ltd.), hold more than 60% market share (Retail World, December 2018) and have national distribution. Of these, 28 bars (17%) were identified as being specifically marketed towards children; these were predominantly grain-based bars (71%), with the remaining 8% muesli bars. Overall, mean serve size varied substantially between categories, with grain-based bars the smallest (27 g), followed by 35 g for muesli bars, and 55 g for oat slices.

There was a significant difference in nutrients including whole grain across all categories per serve and per 100 g (Tables 2 and 3). Post hoc Tukey analysis (per serve) comparing muesli bars and grain-based bars revealed no significant differences in saturated fat (*p* = 0.181), carbohydrate (*p* = 0.365), sugars (*p* = 0.274), and sodium (*p* = 0.869). Grain-based bars and oat slices were significantly different across all nutrients and whole grain content. Conversely, muesli bars and oat slices were the closest in composition for dietary fibre and whole grain (*p* = 0.273 and *p* = 0.238 respectively) with grain-based bars significantly lower (*p* < 0.001). Almost all (95%) grain-based bars met the Australian Dietary Guidelines recommendations of 600 kJ or less as a 'serve' of discretionary food, as well as 61% of muesli bars, but only 8% of oat slices.


**Table 2.** Nutrients per serve (mean & SD): muesli bars, grain-based bars, and oat slices including whole grain.

One Way ANOVA 95% CI.

**Table 3.** Nutrients, whole grain, and HSR/100 g (mean & SD) in muesli bars, grain-based bars, and oat slices.


One Way ANOVA 95% CI.

Comparing per 100 g, post hoc Tukey analysis revealed no difference in saturated fat (*p* = 0.558) between muesli bars and grain-based bars although all other nutrients and HSR were significantly different (*p* < 0.001). Similarly, all nutrients were significantly different between grain-based bars and oat slices except sodium (*p* = 0.952) and although muesli bars are most similar to oat slices in terms of dietary fibre and whole grain content as noted earlier, there were significant differences in fat (*p* = 0.001), saturated fat (*p* < 0.001), sodium (*p* = 0.009), and HSR (*p* = 0.001). Muesli bars were highest in dietary fibre, contributing an average of 9.4 g/100 g, the lowest in sodium (112.2 mg/100 g), and had a significantly higher HSR (3.0). They also contained the highest percentage of whole grain ingredients (40.7%) compared with grain-based bars and oat slices. The average HSR for all products was 2.7, but was higher for the 63% of products that displayed it on-pack (3.2 stars) compared to those that did not (1.8 stars).

The overall results for bars specifically targeted towards children were similar to the averages for grain-based bars, with an average of 1659 kJ ± 120 per 100 g, 6.1 ± 3.3g protein, 9.8 ± 3.5 g total fat, 4.3 ± 2.8 g saturated fat, 67 ± 7.9 g carbohydrate, 26 ±8.1 g sugars, 6.1 ± 4 g dietary fibre, and 161 ± 78.1 mg sodium. Children's bars contained 19 ± 23.8% whole grain ingredients (contributing an average of 4.6 g to the 32–40 g Daily Intake Target for the 4–13 year old age group), and had an average HSR of 2.7 ± 1.1 stars, in line with the mean for the total snack bar category.

The percentage of products meeting nutrition claim criteria are presented in Table 4. More than half of muesli bars and oat slices were eligible for a 'contains whole grain' claim (compared to only 4% of grain-based bars), and 17% of oat slices were considered very high in whole grain. Six products did not report their percentage of whole grain ingredients, required to determine claim eligibility, so these were assumed as ineligible. Similar results were obtained for fibre claim eligibility, with 56% of the

total category at least a source of fibre, mostly represented by muesli bars (69%), and oat slices (61%). The greatest proportion of grain-based bars were low in saturated fat (30%), compared to only 5% of muesli bars, and no oat slices. While none of the investigated bars were considered low in sugar, 48% overall met the most stringent proposed sugar reformulation target for muesli bars, (<25 g/100 g), and an additional 13% met the lower level proposed target of between 25–28 g sugar/100 g, with 29% falling outside the criteria.

**Table 4.** Percentage of products meeting claim criteria and proposed reformulation targets \*.


\* Healthy Food Partnership proposed reformulation targets (September 2018).

As outlined in Table 5, bars categorised as whole grain (≥8 g per manufacturer serve) were significantly higher in energy, total fat, and dietary fibre, and lower in sugars and sodium than refined grain bars. Interestingly, there was no significant difference noted in HSR between whole grain and refined grain bars, with 0.7 star between those categorised as whole grain and the remaining 'non-whole grain bars' which were categorised as refined grain bars.


**Table 5.** Whole grain versus refined grain nutrients (per 100 g) (mean and SD).

Independent samples *t*-test 95% CI. \* Based on eligibility for registration with GLNCs Code of Practice for Whole Grain Ingredient Content Claims (≥8 g per manufacturer serve). \*\* Includes six bars that did not report percentage of whole grain ingredients.

In regards to other on-pack claims, 'No artificial colours/flavours/preservatives' was the most common claim made on packaging, featuring on almost three-quarters (73%) of the total category, and on 91% of oat slices, 80% of grain-based bars, and 66% of muesli bars. More than half made a dietary fibre claim (56%), including 60% of both oat slices and muesli bars, and 30% of grain-based bars. Similarly, 49% made a whole grain claim on-pack, mainly seen on oat slices (70%), and muesli bars (68%), with only 9% of grain-based bars making this claim. An additional 28 products were eligible, but did not make a whole grain claim.

Compared with 2015 (Table 6), 3.5% fewer bars were captured (171 versus 165), with apparent growth in the number of muesli bars (82 to 96 products), and oat slices (18 to 23 products), but a decline in grain-based bars (71 to 46 products), these being the most nutritionally poor products within the category. Over time, there was a significant decrease in total sugars from 26.6 g/100 g to 23.7 g/100 g (*p* < 0.001) across the total category in the four years since 2015, largely attributed to muesli bars, containing 4.2 g/100 g less sugars, while grain-based bars remained stable, and oat slices decreased by 1.1 g/100 g. The proportion of whole grain bars within the category increased, from 35 to 66% in four years (60/171 up to 109/165 bars). HSR data was not captured in 2015 due to the system being newly introduced, so no comparison of this metric over time was possible.

**Table 6.** Comparison of nutrients and whole grain in total bars between 2015 and 2019 per 100g (mean and SD).


Independent samples *t*-test 95% CI.

#### **4. Discussion**

Despite their widespread popularity, consumption of grain-based muesli bars are discouraged by the Australian Dietary Guidelines based on their classification as a discretionary food. This study aimed to provide a comprehensive overview of the nutritional status of grain-based muesli bars on shelf in Australian supermarkets, compared to data collected in 2015.

Overall, wide nutrient ranges were demonstrated between and within the categories examined although muesli bars are treated as a homogenous category in food policy and in advice to consumers. A major factor influencing these differences was the range in average serve sizes, with oat slices more than double that of grain-based bars. Serve size discrepancy may be a point of confusion for shoppers, as nutrient content of the smaller sized grain-based bars may appear more favourable, yet these were the highest in some nutrients of concern on a per 100 g basis. Conversely, oat slices are larger and appear the highest in some positive nutrients per serve, but not when compared per 100 g. This may suggest that the nutrition features of bars may be difficult to compare using the per serve nutrition information at the supermarket shelf. This has been previously described as 'health framing', whereby the impression of a healthier product may lead to overconsumption, however as all bars examined were individually wrapped and therefore portion controlled, this may be less of a concern than in other snack food categories such as cakes and biscuits. These findings are consistent with prior research in Australia which found significant variability in manufacturer serve size within both discretionary [38,39], and core food groups [40,41], and are partly explained by the lack of regulation around standard serving sizes in Australia, which is determined by food manufacturers [40].

Differing ingredients were also a major factor influencing variations in nutrition profile and serve size. Many grain-based bars consisted of puffed or flaked grains (such as corn or rice), which were likely lighter in weight than whole grains, more commonly found in muesli bars and oat slices. Oat slices often contained butter and coconut, both known for their high levels of saturated fat. Additionally, muesli bars and oat slices were all based on oats, which are unique among grains for their higher fat content (6–8%, compared to 2–3% in other grains [42]). The difference in ingredients provides basis for considering further differentiation within this category and at the same time, questions the broad categorisation of 'muesli bars' within the discretionary food group.

Almost one in five bars (17%) in 2019 were specifically marketed towards children, and these were mainly within the grain-based bars category (which are smaller and often made with puffed grains). Generally, these were less nutritious options, being lower in protein, dietary fibre, and whole grain, and higher in sugar than the category on average. Previous research has echoed this finding, with the products designed to appeal to children generally higher in some negative nutrients [34]. Encouragingly, their nutritional value was reflected in the average HSR of less than 3 stars, which has been determined as a cut off point for consumers identifying a food as unhealthy [43].

'Snackification', or the demand for convenience foods to suit modern lifestyles may drive continued innovation and reformulation. New Nutrition Business identified snacking as a key driver of food choice in 2018 and 2019, pointing to examples of manufacturers reinventing foods that were once impossible to eat on-the-go, such as peanut butter in portioned sachets and microwave porridge in individual pots, possibly increasing market competition for muesli bars as traditional snack foods [44]. When considering the top three contributors to adults (19–70+ years) discretionary food intake, the Australian Institute of Health & Welfare's 2018 Nutrition Across the Life Stages report listed alcohol, cakes/muffins/pastries, and soft drinks [45]. Similarly, a 2017 review analysing Australian children's discretionary food intake identified cakes/muffins/slices (4.2%), sweet biscuits (2.9%), and potato crisps/similar snacks (2.7%) as the top contributors to total energy, and the greatest contributors to added sugar were sugar-sweetened soft drinks (18.6%), cakes, muffins, and slices (10.6%), and cordials (6.7%). Conversely, 'sweet snack bars' (which included muesli/cereal bars, and fruit/nut/seed bars) contributed only 1.2% to total energy, and 1.6% added sugars [46]. When this is considered in the context of a typical Australian school lunchbox, including "about one sandwich, two biscuits, a piece of fruit, a snack of either a muesli/fruit bar or some other packaged snack, and a drink of fruit juice/cordial or water" [4], the particular focus on muesli bars as a food of concern may need to be reassessed against the full range of options that could be included in this meal occasion. Discretionary foods such as biscuits, cakes, potato chips, and cordial offer minimal nutritional benefits, so encouraging healthier options within the muesli bar category, alongside core foods in preference to these may be more beneficial advice to consumers and parents who are already under pressure to provide convenient, nutritious snacks.

Comparisons with 2015 data (in Table 6) are suggestive of improvements in terms of added sugars and whole grain content made by food industry. Reformulation aims to improve the nutritional content of manufactured foods, either by increasing beneficial nutrients, or reducing risk-associated nutrients. Often, manufacturers make modest nutritional changes over a period of time to allow consumers' tastes to adjust accordingly, referred to as "health by stealth" [47], but in recent years Australian muesli bar manufacturers have openly shared efforts to reduce salt, fat, sugar, and increase dietary fibre [48]. There is evidence to show reduction targets are effective, with a 2018 review of voluntary sodium reduction targets in soup demonstrating a 6% reduction in sodium levels in soup products between 2011 and 2014, with 67–74% of products compliant with targets [49]. Similarly, Australia's National Heart Foundation has reported significant reductions in line with targets set by the Food and Health Dialogue, such as 10% less sodium in bread and processed meats, and 32% less sodium in breakfast cereals [50], indicating that proposed targets set by the Healthy Food Partnership may encourage further improvements in the added sugars content of muesli bars.

Authors of the 2017 Global Burden of Disease study speculated that dietary policies focused on promoting consumption of whole grains, fruits. and vegetables, and other core food groups may have a greater effect than policies targeting excess consumption of sugar and fat [27]. Within the current study, whole grain bars were clearly identified as a healthier option overall, providing more protective nutrients, and fewer negative nutrients than refined grain bars. Across categories, the majority of oat slices and muesli bars were whole grain (≥8 g per manufacturer serve), and provided the equivalent of at least 30% of an adult's 48 g Daily Target Intake for whole grain, and up to half of a child's daily whole grain requirement (32–40 g/day) [30]. In light of this, whole grain bars may present a convenient, portion controlled, and accepted vehicle for whole grain, and their consumption over refined grain bars could aid in bridging the significant gap in consumption. Unlike other nutrients, whole grain claims are not regulated by Food Standards Australia New Zealand, but are instead encouraged through

GLNCs voluntary Code of Practice for Whole Grain Ingredient Content Claims (The Code), introduced in Australia and New Zealand in 2013 to encourage evidence-based promotion of whole grain foods. GLNC utilises audits of grain-based foods to monitor the operation of The Code and provide feedback to industry as necessary. While 60% of eligible bars were registered with The Code, its voluntary nature, and the fact that the percentage of whole grain ingredients is not mandatory in the ingredients list means deciphering which are whole grain options is not always clear to consumers. This was highlighted by the six bars identified that contained whole grain ingredients (such as rolled oats, and whole grain wheat), but did not report their percentages, so it was unclear whether they met The Code's whole grain criteria. Encouragingly, the number of whole grain bars have increased by 31% since 2015, suggesting positive changes have been made by manufacturers to existing products, new whole grain products have been added to the market due to consumer demand, or that labelling has been updated to more clearly communicate whole grain content.

The variability in nutrients supplied within the grain-based muesli bar category, combined with their popularity, may point towards education as the more powerful tool in supporting consumers to choose healthier products, in preference to discouraging consumption. The concept of 'knowledge-is-power' has been explored in previous research, with a review from the USA determining consumers with greater nutrition knowledge were more likely to consult nutrition labels, which may lead to healthier food choices [51]. The HSR attempts to clarify complex nutrition information and arm consumers with the knowledge to make healthier choices within food categories, and has been shown to perform well in directing consumers towards healthier, higher-scoring foods [43,52,53]. HSR scores for the bars category ranged from 1–5 stars, yet there was no significant difference between refined and whole grain varieties, with only 0.7 of a star between products. This finding highlights a shortcoming of the algorithm used to assign products a star rating, and builds on previous research that demonstrated an inability to differentiate whole grain and refined grain breads, breakfast cereals, rice, and flour products, as it does not directly account for, or reward foods for whole grain content [37]. There is a clear opportunity to refine the HSR by recognising whole grain as a positive food component, which could play a role in discerning healthier food choices across numerous categories, including muesli bars. However, to meet its objective of simple nutrition comparisons within categories, widespread uptake of a voluntary front-of-pack labelling system such as the HSR is required. Almost two-thirds (63%) of bars examined displayed a HSR, comparatively higher than overall uptake, which is estimated at 28% [16]. Consistent with existing literature, bars displaying a HSR tended to have higher scores, suggesting the system may be used strategically within and across brands [16,54]. Conversely, industry appear to be using the HSR as an incentive to improve a product's nutritional value, with recent studies in Australia and New Zealand identifying upwards of 83% of products displaying a HSR had been reformulated to increase their score [54,55].

Strengths of this study include its comprehensive nature, and to our knowledge, it is the first study that has reviewed muesli bars on shelf in Australia, with a comparison made to previously collected data. Also, where HSR was not provided, we calculated this for a more accurate representation of HSR across the category. However, there were some limitations. The research was focused only on grain-based bars, excluding others—such as nut bars and protein and low-carb bars—which may also be consumed as snacks though to a lesser extent than muesli bars [56]. While all efforts were made to capture the category in its entirety, differences may exist between geographic areas. As previously stated, reporting of dietary fibre and whole grain within the ingredients and Nutrition Information Panel is not mandatory in the absence of an on-pack claim, so was not always declared, and thus there was some missing data. Finally, we did not conduct an independent nutrition analysis, and were reliant on manufacturer information.

#### **5. Conclusions**

Although categorised as discretionary, there are significant nutrient differences across grain-based muesli bars, with well-chosen bars providing valuable amounts of whole grain and dietary fibre. Muesli bars are a widely consumed snack food, particularly among younger age groups in Australia, yet their contribution and role in the diet is controversial, based on their classification at discretionary by Australian Dietary Guidelines. This study demonstrated significant variation between and within the category, with the whole grain options emerging as more nutritious compared to refined grain bars, and an indication of sugar reduction since 2015. Within a balanced diet, it is clear that some muesli bars can offer a convenient and nutritious snack, with many bars providing around 30% of an adult's, and up to half of a child's daily requirement for whole grain, and more than half of all products are at least a source of fibre. Both whole grains and dietary fibre are encouraged within Dietary Guidelines yet intakes across age groups tend to fall short of dietary targets. The current HSR algorithm does not appear to be overly favouring muesli bars (with an overall score of 2.7), and instead, could be improved to capture and differentiate whole grain options. Ongoing promotion of the higher HSR scoring bars, alongside proposed voluntary sugar reformulation targets and trends such as snackification, may be suggestive of opportunities and incentives for manufacturers to further improve the current range of products. Clearer classification within policy initiatives utilising evidence-based assessment of available products may help refine advice from healthcare professionals, and may be key in providing better direction for consumers to make healthier and acceptable snack food and lunchbox choices.

**Author Contributions:** Conceptualization, F.C. and S.G.; Methodology, F.C.; Formal analysis, S.G.; Original draft preparation, F.C.; Review and editing, S.G.

**Funding:** This research received no external funding but was supported by the Grains & Legumes Nutrition Council, a not-for-profit charity.

**Acknowledgments:** Thanks to James Sze, Student Dietitian from the University of Wollongong, NSW, who was involved in data collection as part of his university studies, and to Joanna Russell for statistical advice.

**Conflicts of Interest:** S.G. and F.C. are employed by the Grains & Legumes Nutrition Council, a not-for-profit charity.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **Gluten-Free Alternative Grains: Nutritional Evaluation and Bioactive Compounds**

#### **Serena Niro 1, Annacristina D'Agostino 1, Alessandra Fratianni 1,\*, Luciano Cinquanta <sup>2</sup> and Gianfranco Panfili <sup>1</sup>**


Received: 7 May 2019; Accepted: 10 June 2019; Published: 12 June 2019

**Abstract:** Interest in gluten-free grains is increasing, together with major incidences of celiac disease in the last years. Since to date, knowledge of the nutritional and bioactive compounds profile of alternative gluten-free grains is limited, we evaluated the content of water-soluble (thiamine and riboflavin) and liposoluble vitamins, such as carotenoids and tocols (tocopherols and tocotrienols), of gluten-free minor cereals and also of pseudocereals. The analysed samples showed a high content of bioactive compounds; in particular, amaranth, cañihua and quinoa are good sources of vitamin E, while millet, sorghum and teff (*Eragrostis tef*, or William's Lovegrass) are good sources of thiamine. Moreover, millet provides a fair amount of carotenoids, and in particular of lutein. These data can provide more information on bioactive compounds in gluten-free grains. The use of these grains can improve the nutritional quality of gluten-free cereal-based products, and could avoid the monotony of the celiac diet.

**Keywords:** minor cereal; pseudocereal; bioactive compound; gluten-free grain; tocols; carotenoids

#### **1. Introduction**

Celiac disease is a chronic systemic, autoimmune disorder in genetically-predisposed individuals, triggered by exposure to dietary gluten, and resulting in mucosal inflammation, villous atrophy and crypt hyperplasia [1]. It is characterised by an abnormal immune reaction consisting of an excessive response of the immune system to a group of cereal proteins, called prolamines (gliadin, hordein, sekalina, avenin), which are found in wheat, barley, rye and oats. Celiac disease affects approximately 1% of the world population, and it has significantly increased due to an underestimation, since it is often left undiagnosed [2]. The only treatment for people with the celiac problem is the adherence to gluten-free foods for their whole lifetime.

Several studies demonstrated that sticking to a gluten-free diet for a lifetime can lead to a nutritional imbalance in celiac subjects, such as a malabsorption of nutrients, and deficiencies of several vitamins and minerals. These deficiencies are due both to the phenomena of malabsorption at the intestinal level, and to the monotony of a diet based mainly on rice and maize [3–6].

Recently, more attention has been given to gluten-free minor cereals and pseudocereals as alternatives to those conventionally used for celiacs. Many of them have been defined as "orphan crops" or "underutilised crops"; they are indigenous crops scarcely documented and rarely used by food industries [7]. Many underutilised crops are relatively more drought-tolerant than most major cereals; they play a significant role in many developing countries, providing food security and income to resource-poor farmers [8].

Gluten-free alternative sources studied in this work include minor cereals (sorghum, teff, millet and wild rice), and pseudocereals (quinoa, cañihua, chia, and amaranth). These grains are mainly consumed as flours and seeds, which can be added to preparations such as soups, yogurt, cakes, breads and others cereal-based products; nevertheless, any commercialisation of these products is still quite limited in the Italian market. Some of these are a source of nutrients and bioactive compounds that could improve the nutritional quality of gluten-free products.

Carotenoids are a significant group of bioactive compounds with health promoting properties [9,10] and are responsible for the colour of a wide variety of grains [11]. Some carotenoids are the precursors of retinol (vitamin A), and are very strong natural antioxidants. Carotenoids are known to be efficient physical and chemical quenchers of singlet oxygen, as well as potent scavengers of other reactive oxygen species [9]. Vitamin E is a natural antioxidant comprising two groups of vitamers, tocopherols and tocotrienols, occurring in eight forms: α-tocopherol (α-T), β-tocopherol (β-T), γ-tocopherol (γ-T), and δ-tocopherol (δ-T) and α-tocotrienol (α-T3), β-tocotrienol (β-T3), γ-tocotrienol (γ-T3), and δ-tocotrienol (δ-T3). Vegetable oils are the main tocol sources, however, substantial amounts of these compounds are also reported in most cereal grains [12–14]. The potential health benefits of tocols include the prevention of certain types of cancer, heart diseases and other chronic diseases [15,16]. Thiamine (B1) is one of the major water-soluble vitamins, as it plays an important role as a co-factor of several key enzymes involved in the carbohydrate metabolism and defence mechanism [17]. It can be found in moderate amounts in all foods: Nuts and seeds, legumes, wholegrain/enriched cereals and breads, as well as pork [18]. Thiamine deficiency is rare in healthy individuals in food-secure settings, where access to thiamine-rich foods ensures adequate intakes [19]. Riboflavin (B2) is a precursor of the co-enzymes flavin mononucleotide (FMN; riboflavin phosphate) and flavin adenine dinucleotide (FAD), which are components of oxidases and dehydrogenases. It is also important for skin health and normal vision, and can be found in whole cereals, breads, leafy green vegetables and milk products [18].

To date, the evaluation of nutritional and bioactive compound profiles of alternative gluten-free grains is limited, if not lacking [20–23]. These researches are of a great importance in order to formulate gluten-free cereal-based products with a higher nutritional value. Thus, in this work, samples of minor cereals and pseudocereals commercialised in Italy have been characterised for their nutritional value, with a particular focus on some bioactive compounds, such as carotenoids, tocols, thiamine and riboflavin, in order to increase the awareness of their nutritional profile. Moreover, data coming from this study may be included in food nutrient databases.

#### **2. Material and Methods**

#### *2.1. Sample Collection and Preparation*

Thirty one different minor cereals and pseudocereals were bought in Italian specialised shops (Table 1). Different brands were considered for each grain. Grains were grounded with a refrigerated IKA A10 laboratory mill (Staufen, Germany), then carefully mixed and stored at −20 ◦C until analysis. Each sample was analysed in triplicate.

**Table 1.** List of analysed gluten-free grains.


**Table 1.** *Cont.*


#### *2.2. Chemical Analysis*

#### 2.2.1. Proximate Analysis

Moisture, ash, fat, and protein contents were determined using an ICC standard procedure [24]. Briefly, moisture was determined using an oven set at 130 ◦C, and ash was quantified using a muffle furnace set at 525 ◦C. The protein content was determined though the Kjeldhal method (N × 6.25), and lipids were determined by the Soxhlet method. Carbohydrates plus fibre were calculated as a difference, using the following equation: (100 − (% moisture + % lipids + % proteins + % ash)).

#### 2.2.2. Carotenoid Analysis

Carotenoid extraction was carried out using the saponification method reported by Panfili et al. [14]. About 0.2 g of milled sample was weighed and placed in a screw-capped tube. Then, 5 mL of ethanolic pyrogallol (60 g/L) was added as an antioxidant, followed by 2 ml of absolute ethanol, 2 mL of sodium chloride (10 g/L) and 2 mL of potassium hydroxide (600 g/L). The tubes were placed in a 70 ◦C water bath and mixed every 5–10 min during saponification. After alkaline digestion at 70 ◦C for 45 min, the tubes were cooled in an ice bath, and 15 mL of sodium chloride (10 g/L) were added. The suspension was then extracted twice with 15 mL portions of n-hexane/ethyl acetate (9:1, *v*/*v*). The organic layers, containing carotenoids, were collected and evaporated to dryness; the dry residue was dissolved in 2 mL of isopropyl alcohol (10%) in *n*-hexane. A HPLC Dionex (Sunnyvale, CA) analytical system, consisting of a U6000 pump system and a 50 μL injector loop (Rheodyne, Cotati) was used. The chromatographic separation of the compounds was achieved by means of a 250 mm × 4.6 mm i.d., 5 μm particle size, Kromasil Phenomenex Si column (Torrance, CA, USA). The mobile phase was *n*-hexane/isopropyl alcohol (5%) at a flow rate of 1.5 mL/min. Spectrophotometric detection was achieved by means of a diode array detector set in the range of 350–500 nm. Peaks were detected at 450 nm. Carotenoids were identified through their spectral characteristic, and comparison of their retention times with known standard solutions. Data were stored and processed by a Dionex Chromeleon Version 6.6 chromatography system (Sunnyvale, CA, USA). All-trans-β-carotene and lutein were obtained from Sigma Chemicals (St. Louis, MO, USA); zeaxanthin and β-cryptoxanthin were obtained from Extrasynthese (Z.I. Lyon-Nord, Genay, France).

#### 2.2.3. Tocol Analysis

Tocols were determined after the same saponification method described for carotenoids. An aliquot of the carotenoid extract was collected and evaporated to dryness, and the dry residue was dissolved in 2 mL of isopropyl alcohol (1%) in *n*-hexane, and was analysed by HPLC, under normal phase conditions, using a 250 × 4.6 mm i.d., 5 mm particle size Kromasil Phenomenex Si column (Torrance, CA, USA) [14]. Fluorometric detection of all compounds was performed at an excitation wavelength of 290 nm and an emission wavelength of 330 nm by means of an RF 2000 spectrofluorimeter (Dionex, Sunnyvale, CA, USA). The mobile phase was *n*-hexane/ethylacetate/acetic acid (97.3:1.8:0.9 *v*/*v*/*v*), at a flow rate of 1.6 mL/min [14,25]. Compounds were identified by a comparison of their retention times with those of known available standard solutions, and quantified through the calibration curves of the standard solutions. The concentration range was 5–25 μg/mL for every tocol standard. Vitamin E

activity was expressed as Tocopherol Equivalent (T.E.) (mg/100 g of fresh weight f.w.), calculated as reported by Sheppard et al. [26].

#### 2.2.4. Thiamine and Riboflavin Analysis

Thiamine and riboflavin were extracted as in Hasselman et al. [27]. Briefly, samples were placed in 100 mL volumetric flasks containing 20 mL of 0.1 N HCl and heated in a water bath at 100 ◦C for 30 min. After cooling at room temperature, the pH of the samples was adjusted to 4.5 with 2.5 M NaOAc. Following the addition of 0.2 mL of aqueous Clara-Diastase (50 mg/mL), these samples were incubated for 3 h at 37 ◦C. After cooling, the samples were brought up to 25 mL with distilled water. Then these same samples were centrifuged and filtered through a 0.45 μm filter. Thiamine was converted to thiochrome by adding 1.25 mL of 1% potassium ferricyanide in 15% aqueous NaOH to 2.5 mL of filtered extract. After 1 min for oxidation, 0.25 mL of 85% H3PO4 was added. The extract was purified on a Sep-Pak C18 cartridge. The cartridge was washed with 5 mL MeOH, followed by 5 ml of 0.05 M NH4OAc (adjusted to pH 5.0 (acidic) with HOAc). The sample (5 mL) was loaded into a Sep-Pak C18 cartridge, and then the cartridge was washed with 0.05 M NH4OAc and, finally, the vitamins were eluted with 5 mL mobile phase. Extracts were separated by a HPLC Dionex (Sunnyvale, CA, USA), with a U3000 pump and an injector loop (Rheodyne, Cotati). Separation was made at a flow rate of 0.8 mL/min with Methanol: NaOAc (40:60 *v*/*v*) as a mobile phase, by using a 5 μm C18 Luna, Phenomenex (Torrance, CA, USA) stainless steel column (250 × 4.6 mm i.d.). Fluorometric detection was performed at an excitation wavelength of 366 nm and an emission wavelength of 453 nm for thiamine, and an excitation wavelength of 453 nm and an emission wavelength of 580 nm for riboflavin, by means of an RF 2000 spectrofluorimeter (Dionex, Sunnyvale, CA, USA). Data were processed by a Dionex Chromeleon Version 6.6 chromatography system (Sunnyvale, CA, USA). Thiamine and riboflavin were compared with known available standards, and identified considering their retention times and relative elution order. Thiamine and riboflavin standards were obtained from Sigma Chemicals (St. Louis, MO, USA).

#### **3. Results and Discussion**

#### *3.1. Nutritional Composition*

The nutritional composition of analysed minor cereals and pseudocereals is shown below in Table 2.


**Table 2.** Nutritional composition of gluten-free grains (g/100 g).

\* Calculated by difference; a: coefficient of variability.

The composition of the chia seeds notably differs from all the other cereal and pseudocereal samples, showing high concentrations of fats (35.4 g/100 g), proteins (21.5 g/100 g) and ash (4.5 g/100 g). These values are similar to those observed by other authors for the chia seeds [28]. In general, wild rice and pseudocereals are a good source of protein. Taking European law into account [29], wild rice, all quinoa seeds, cañihua and amaranth can be declared in a label with the claim "source of protein", since they contain at least 12 g of protein per 100 g. Chia seeds can be declared with a "high protein

content", since they contain at least 20 g of protein per 100 g. The fat content was significantly higher for pseudocereals, if compared to minor cereals. Wild rice shows the lower fat content (1.2 g/100 g).

#### *3.2. Carotenoids*

Table 3 shows the carotenoid amounts of analysed samples. Carotenoids content (μg/100 g dry weight d.w.) varied significantly from 22 μg/100 g in amaranth to 763 μg/100 g in millet. In all gluten-free grains the main compounds are lutein and zeaxanthin. A comparison with the literature related to the HPLC analysis of carotenoids is very difficult, since the available few data are obtained by different methods, and these pigments may vary depending on genotype and location. The total carotenoid content of millet, wild rice, quinoas and cañihua is comparable with that of wheat (about 305 μg/100 g for durum and about 150 μg/100 g for soft wheat) [12,30], and of pigmented rice (460–50 μg/100 g) [31], but it is significantly lower than that of maize (about 1110 μg/100 g) [30,32]. Among minor cereals, literature data are reported only for sorghum [33], where the authors found an average amount of 20 μg/100 g as the sum of lutein and zeaxanthin, with a high variability among the different genotypes.


**Table 3.** Carotenoid composition in gluten-free grains (μg/100 g d.w.).

a: Coefficient of variability; nd: not detectable; tr: traces.

In the present study, the variability of the total carotenoid content within the same cereal (expressed by the coefficient of variability, CV%), is from 4% in millet to 26% in pigmented quinoa. This variability may be due to genetic, pedoclimatic and varietal factors [34]. Regarding pseudocereals, results for chia are similar to those obtained in the work of da Silva et al. [28]. Significant differences between white and pigmented quinoas were found for total carotenoids, due to the different lutein amounts, as also observed by Tang et al. [35], who indicate a direct correlation between the higher total carotenoid content and the darkness of the seed coat.

#### *3.3. Tocols*

The characterisation of tocols in minor cereals and pseudocereals is reported in Table 4. Except for wild rice, which shows a minor content of total tocols (TC) (about 0.4 mg/100 g), the TC of minor cereals and amaranth are comparable with that of wheat, maize and rice (about 3.5–7.0, 6.0–7.0 and 2.3–2.7 mg/100g, respectively) [12,14,36] while, for the remaining pseudocereals, these values are significantly higher. Among minor cereals, teff shows the highest amount of total tocols (6 mg/100g d.w.), followed by millet and sorghum with about 4 and 3 mg/100g respectively.


**Table 4.** Tocol composition in gluten-free grains (mg/100g d.w.)

a: Coefficient of variability; b: Not detectable; tr: traces; T.E.: Tocopherol equivalent (mg/100g f.w.).

Except for wild rice, where α-tocopherol is the prevalent isomer, the main tocopherol isomer is γ-tocopherol, which represents the 92%, 72% and 75% of the total content in teff, millet and sorghum, respectively. For pseudocereals, the highest content of total tocols was found in cañihua (about 18 mg/100 g), followed by chia seeds (about 14 mg/100 g d.w.) and quinoas, with an average of 9.1 mg/100 g d.w. Contrarily to carotenoids, among all analysed quinoa seeds, all of the found vitamers did not show significant qualitative and quantitative differences. Amaranth is the pseudocereal with the lowest total tocol amounts (about 6 mg/100g). For chia, cañihua and quinoa the predominant isomer is γ-tocopherol (94%, 69% and 64% of total tocols), while for amaranth the prevalent isomer is β-tocopherol, which represents 54% of the total tocols.

γ-Tocopherol has also been found as the main vitamer in quinoa and chia in other works [28,35,37]. References for tocols are not available for all analysed gluten-free grains and, where present, they show similar results in millet and sorghum [3,23]. Moreover a comparison with the literature data related to tocol analysis is very difficult, for the same reasons already explained for carotenoids.

Table 4 also reports values of vitamin E activity provided by 100 g of product, expressed as Tocopherol Equivalent (T.E.) (mg/100 g product) [26]. Taking into account the Recommended Daily Allowance (RDA) for vitamin E, which is of 12 mg/day [38], 100 g of amaranth contribute to 22% of the RDA, while quinoas and cañihua approximately to 35% of the RDA, so as to be declared in a label as a "source of vitamin E". A portion of these pseudocereals (70 g) contributes approximately to 15% of the RDA for amaranth and to 25% of the RDA for quinoas and cañihua.

#### *3.4. Thiamine and Riboflavin*

Table 5 reports the values of the thiamine and riboflavin of analysed grains. The concentrations of thiamine are different between minor cereals and pseudocereals, except for wild rice. In whole wheat grains about 0.40 mg/100g are found in the literature [39,40]. Low values of riboflavin were found for all samples, except for wild rice, with values comparable to those of whole wheat grains and maize (0.15 and 0.20 mg/100g, respectively) [39,40].


**Table 5.** Thiamine and riboflavin content in gluten-free grains (mg/100g d.w.).

Taking into account the Recommended Daily Allowance (RDA) for thiamine, which is of 1.1 mg/day [38], 100 g of teff would contribute to approximately 17% of the RDA, while 100 g of millet and sorghum to 23% of the RDA, so as to be declared in a label as a "source of thiamine". A portion of 80 g contributes approximately to 16% of the RDA for teff and to 20% of the RDA for millet and sorghum.

#### **4. Conclusion**

Naturally gluten-free products are corn, rice, potatoes, soybean, millet, buckwheat, tapioca, amaranth, cassava, lentils, beans, sago, sorghum, nuts, as well as meat, fruit and vegetables. Among these, cereals and pseudocereals are becoming increasingly important. This work confirms that minor cereals and pseudocereals are an important source of bioactive compounds. In particular, wild rice and all analysed pseudocereals are good sources of protein. Taking into account the Recommended Daily Allowance (RDA) for vitamins established by the Commission of the European Communities, amaranth, cañihua and quinoa can be declared on the label as a source of vitamin E, the main antioxidant found in cells involved in the prevention of several diseases. Moreover, millet, sorghum and teff can be declared on the label as a potential source of thiamine. Millet also provides a fair amount of lutein. In the light of these results, it is possible to use the combined mix of these flours in order to improve the nutritional value of cereal-based gluten-free products.

**Author Contributions:** Conceptualization, writing-review and editing, G.P; investigation and writing-original draft preparation, S.N. and A.D.A.; data curation, writing-review and editing A.F; writing-review and editing, L.C.

**Funding:** This research received no external funding.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## **Micronutrient Analysis of Gluten-Free Products: Their Low Content Is Not Involved in Gluten-Free Diet Imbalance in a Cohort of Celiac Children and Adolescent**

#### **Jonatan Miranda 1,2**


Received: 19 July 2019; Accepted: 5 August 2019; Published: 7 August 2019

**Abstract:** Data about the nutritional composition of gluten-free products (GFP) are still limited. Most studies are based on ingredient and nutrition information described on the food label. However, analytical determination is considered the gold standard for compositional analysis of food. Micronutrient analytical content differences were observed in a selection of GF breads, flakes and pasta, when compared with their respective gluten-containing counterparts. In general terms, lower iron, piridoxin, riboflavin, thiamin, niacin, folate, manganese and vitamin B5 can be underlined. Variations in biotin and vitamin E content differed among groups. In order to clarify the potential contribution of the GFP to the gluten-free diet's (GFD) micronutrient shortages, analytical data were used to evaluate GFD in a cohort of celiac children and adolescent. Participants did not reach recommendations for vitamin A, vitamin E, folic acid, vitamin D, biotin, iodine, and copper. It does not seem that the lower micronutrient content of the analyzed GFP groups contributed to the micronutrient deficits detected in GFD in this cohort, whose diet was not balanced. Nevertheless, GFP fortification for folate and biotin is proposed to prevent the deficiencies observed in GFD, at least in the case of pediatric celiac disease.

**Keywords:** celiac disease; gluten-free diet; gluten-free product; micronutrient; vitamin and minerals; dietary recommendation

#### **1. Introduction**

Celiac disease (CD) is a chronic immune-mediated inflammatory pathology triggered by the gluten in the diet of genetically predisposed individuals. The need to avoid this protein in the diet of celiac people brought about some years ago the development of specific cereal-based gluten-free products (GFP). Despite the fact that these GFP allowed them to include a wide variety of foods in their diets, in recent years researchers have highlighted differences in the nutrient composition of GFP with respect to gluten containing counterparts [1,2], leading to a minor health rating in some food-groups [3,4].

It is important to note that most of the studies about the nutrient composition of the GFP are based on ingredients and the nutrition information described on the food label [2–4]. To improve these data, some works, such as that carried out by Mazzeo et al. (2015) [5], take advantage of the retention factors for each nutrient, including losses due to heating or other food preparation steps. However,

analytical determination is considered the gold standard for composition analysis of food. Accurate analysis could also provide detailed information about vitamins and minerals, which is not totally or commonly available on label [6]. Therefore, access to micronutrient data is already restricted to hardly any research [7–9].

Furthermore, a gluten-free diet (GFD) often implies some nutritional imbalances, as recognized in the literature [10,11]. Not only have inadequate fat, protein, sugar and fiber consumption been observed in GFD, but also a poor intake of micronutrients such as iron, zinc, magnesium, calcium, folate, vitamin D and B12 [12]. Similarly, celiac people seem to have lower blood values for hemoglobin, ferritin, vitamin D, and copper than the rest of the population [13,14]. There has been speculation about whether the characteristic composition of GFP is responsible for GFD inadequacy. A potential correlation between both facts has been proposed by others [15].

In the case of GFP, the use of raw material such as unenriched rice or maize refined flours, gums or enzymes in their formulation could lead to a different composition compared to their gluten containing homologues [16]. Moreover, as the micronutrient content of gluten-free pseudocereals and legumes is higher than that of the gluten free cereals [15,17], some authors proposed to promote their use in GFP formulation [12,18,19].

In view of the above, the aim of this study was to assess analytically the macronutrient and micronutrient content of a selection of GF breads, flakes and pasta, and to compare it with their respective gluten-containing counterparts. Additionally, in order to clarify the potential contribution of the GFP to the GFD's micronutrient shortages, vitamin and mineral analytical data were used to evaluate GFD in a cohort of celiac children and adolescents.

#### **2. Materials and Methods**

#### *2.1. Analytical Nutrient Content of GF Bread, Breakfast Cereals and Pasta*

The measured samples were thirty-seven selected GFP signed with the Crossed Grain symbol: 13 breakfast cereals, 12 breads and 12 pasta products (Supplementary Table S1). All the food items were purchased from the local market (Vitoria, Spain) and they were stored frozen (−20 ◦C) until analyzed. The analytically determined composition of GF foodstuffs was compared with the data of equivalent gluten-containing breads (*n* = 19), breakfast cereals (*n* = 18), and pasta products (*n* = 8), analyzed in the same way and at the same time for macronutrients, and with micronutrient data obtained from the Spanish Food Composition Database—BEDCA database [20]. These results were also compared with the data described in the food label of GFPs.

Analysis of the nutrient content of foodstuff has been carried out using official methods. Crude protein content was determined by the Kjeldahl method (AOAC, 960.52A) [21] in a Foss Kjeltec™ distillation unit (Höganäs, Sweden). Fat content was analyzed by the Soxhlet extraction method based on the official method (AOAC, 2003.05) [21], using a Soxtherm extraction system (Gerhardt, Bonn, Germany). Determinations were performed in duplicate.

For mineral determination, microwave-assisted digestion was carried out in a closed microwave device Mars 5 (CEM, Vertex, Barcelona, Spain) equipped with 8–24 teflon vessels and temperature controllers. The quantitative analysis of selenium, manganese and cooper was performed by using ICP-MS (7700x, Agilent Technologies, Palo Alto, CA, USA) and MicroMist micro-uptake glass concentric nebulizer (Glass Expansion, West Melbourne, Victoria, Australia). ICP-OES (Horiba Jobin Yvon Activa, Kyoto, Japan) was used with a quartz Meinhard concentric nebulizer, a Scott-type spray chamber and a standard quartz sheath connection between the spray chamber and the torch in the case of calcium, sodium, zinc and iron quantification. Working standard solutions of Ca, Na (0–20 mg/L), Fe, Zn and Se (0–100 μg/L) were prepared immediately prior to their use, by stepwise dilution of certified standard multi-element solution (100 mg/L) (Merck, Darmstadt, Germany) with HNO3 1.0 % *v*/*v* (Merck, Darmstadt, Germany). Additionally, a 10 mg/L multi-element standard solution (Y, Rh) from Inorganic Ventures (Equilab, Madrid, Spain) was also used as the internal standard in direct ICP-MS analysis.

As a step prior to vitamin quantification, samples were extracted by liquid-liquid extraction using an aqueous acidic mixture, centrifuged and filtrated, except for vitamin E. Biotin, Folate, Niacin, Pyridoxine, Riboflavin, Thiamine, vitamin B5 and B12 were measured by liquid chromatography (LC) with triple quadrupole mass spectrometry detection. High purity (>95%) standards (Merck, Darmstadt, Germany) were used for the identification of each vitamin by positive ionization of the electrospray and multiple reaction monitoring. Quantification was developed using the standard addition method. Vitamin E determination was carried out by previous saponification of the samples, followed by a liquid-liquid extraction and purification of the extracts. Afterwards, high performance LC with the fluorescence detector method was used to analyze vitamin E in each extract. Quantification was performed by an external calibration method using the calibration curve of the tocopherol standard (Merck, Darmstadt, Germany). Analytical determinations of micronutrients were carried out once in each sample, but it was verified before the analysis that the reproducibility of the methods was less than 5%.

As mentioned, the micronutrient content of GF foodstuffs was compared with that of gluten-containing counterparts, obtained from the Spanish Food Composition Database—BEDCA database- [20]. Data for biotin in all studied food groups and copper in cereals were obtained from McCance and Widdowson's "composition of foods integrated dataset" from the United Kingdom [22]. No available data were found with regard to the manganese content of cereal flakes in food composition databases from the UK, Australia, the USA or Spain [20,22–24].

#### *2.2. Dietary Assessment: Participants and Procedure*

Eighty-three minor celiac (age: 3 to 18 years; 53 girls and 30 boys) from the Basque Country took part in the study. The age of the participants was selected due to their higher consumption of GFP compared to adults [25,26]. All participants received oral and written information about the nature and purpose of the survey, and all of them gave written consent for involvement in the study. This study was approved by the Ethical Committee in University of Basque Country (CEISH/76/2011 and CEISH/194M/2013).

The dietary assessment followed in the research was described elsewhere [26]: three days food records (two weekdays and one weekend day) were selected for each patient, 24-h food recalls (24HRs) were filled in by each celiac patient. Micronutrient intake was calculated by a computerized nutrition program system (AyS, Software, Tandem Innova, Inc., Huesca, Spain). The analytically measured vitamin and mineral content of tested GF products was added into the food composition database of the program before calculations. Dietary reference intakes (DRI) for the Spanish population issued by the Spanish Societies of Nutrition, Feeding and Dietetics (FESNAD) in 2010 were taken as references for the interpretation of the 24HRs [27].

#### *2.3. Statistical Analysis*

Results are presented as mean ± standard deviation (SD) of the mean. Statistical analysis was performed using SPSS 24.0 (SPSS Inc., Chicago, IL, USA). After confirming the normal distribution of lipid and protein content variables using Shapiro-Wilks normality, paired-samples student's t test was used for comparison. Due to their skewed distribution, micronutrients variables for analytical and database information were analyzed by Mann–Whitney *U*. The level of significance was set to *p* < 0.05.

#### **3. Results and Discussion**

#### *3.1. Macronutrient Content of GF Rendered Foods*

With the aim of assessing representative products of a GFD, GFP from the three main cereal food-types contributing to a balanced diet, such as flakes, pasta and bread, were selected. Protein and lipid contents of the three GFP groups analyzed are shown in Table 1. Results were compared to the nutritional composition of their gluten-containing counterparts. With regard to breads, lipid content was higher and the protein content was lower than that of gluten containing products. Similarly, GF bread has been described as poor in proteins and rich in fat content by others [28]. GF pasta provided a lower protein amount, although the comparison to gluten containing pasta did not reach statistical significance. In general terms, lower protein content in GFP than in their counterpart has been proposed by previous research [2–4]. Nevertheless, and in good accordance with our data, Missbach et al. did not observe this pattern in flakes [2].

**Table 1.** Analytical protein and lipid content in gluten-free rendered foodstuffs divided by food groups, compared to gluten-containing products, expressed by 100 g of foodstuffs.


Values are means ± SD. SD, standard deviation; GFP, gluten-free product; GCP, gluten-containing product; *p*, statistical significance; NS, not significant.

Some clues for justifying the results could be extracted from the list of ingredients of GFP (Supplementary Table S1). Rice and maize flours are extensively used in GFP, especially in breads, and according to composition databases, their protein content is lower than that of wheat. Moreover, maize and rice starches, usually added as a substitute, are especially poor in this macronutrient. For pasta and flakes, other ingredients could hinder the protein deficit, such as cocoa or eggs, soy protein or meat from the filled pasta. For lipids, the use of additives like mono and diglycerides of fatty acids (E-471) in GFP, especially in breads, could affect the final composition. However, this study did not consider the label information of ingredients of GCP, thus making conclusive statements is not possible.

It is important to point out that the comparative study between GFP and their homologues with gluten in the present work was performed as suggested by Staudacher and Gibson [6], by direct analytical methods and in paired form. As stated in the introduction, most of the studies evaluating the differences between both foodstuffs are based on nutrition information taken from the food label. For this reason, the analytical results obtained were compared to those reported in the nutritional panel information and some interesting data were collected. With regard to bread, experimental data reported a lower lipid (23%, *p* = 0.07) and higher protein (37%; *p* = 0.03) content than that supplied by the label. Similarly, in the case of cereal flakes, the measured protein amount was higher (19%; *p* = 0.04). No differences were observed between analyzed and labelled data in GF pasta.

In view of Regulation (EU) No 1169/2011 [29], the declared values on labels shall be average values based on (a) the manufacturer's analysis of the food; (b) a calculation from the known or actual average values of the ingredients used; (c) a calculation from generally established and accepted data. It is not possible for us to determine how each manufacturer calculated label information. However, it must be highlighted that nutrient variations observed in bread types are not within the tolerance ranges between label information and our direct food analysis (tolerance ranges: ±1.5 g for lipids and ±2 g for proteins, when its content in food is <10 g per 100 g). This information brings to light that previous studies about bread described in the literature could be reconsidered, and additionally, it validates, in part, others about pasta and cereals.

#### *3.2. Micronutrient Content of GFP, Compared to Gluten-Containing Products*

Despite the growing market of the GFP [30], data about their vitamin and mineral contribution remain scarce. Moreover, the data found in the literature are usually calculated from ingredients and their composition databases, which has been proposed to lead to overestimation [5]. Table 2 shows analytical micronutrient content of GF bread, flakes and pasta, compared to that of their gluten-containing counterparts. Lower iron, piridoxin, riboflavin and thiamin content was found in the three GFP groups analyzed. Niacin reduction was observed in GF flakes and breads. With regard to iron, similar results were found by Rybicka [8], who described that 273 of 408 GFP analyzed fulfilled less than 10% of recommended nutrient intake per portion and only 23 products were major contributors to daily intake (over 25% of recommendation intake per portion). In a study performed with 368 GFP, including flours, breads, pasta and cold cereals, overall it was observed that these kinds of products contained lower amounts of thiamin, riboflavin and niacin than the wheat product they were intended to replace [31]. These results are in line with the results obtained in the present study.

Folate content was lower in GF flakes and pasta types; manganese amount was lower only in GF pasta, and that of vitamin B5 in GF flakes. As stated before, commonly used ingredients for GFP are maize and rice flours as well as a variety of starches (potato, corn), among others. It seems that removal of protein-rich fractions from flours may result in dramatic depletion of folates. Additionally, rice flours are not very rich in this vitamin [9]. In fact, we calculated a reduction of almost 80% of folate content in rice flour with respect to wheat flour (*p* = 0.05) comparing the nutrient composition of both flours obtained from food composition databases from the UK, Australia, the USA or Spain [20,22–24].

Several studies have claimed lower zinc and copper and higher sodium content for GFP [4,32]. However, no significant differences in those minerals were found in our data.

Finally, biotin content differed widely among groups, being higher in cereal flakes and lower in pasta GFP than in their counterparts. Moreover, we found that some GF cereals were fortified with biotin, thus explaining its higher content in this GF food group. Similarly, although vitamin E contribution from GFP was lower in flakes, no differences were observed in pasta and bread. Moreover, it is worth mentioning that half of the analyzed bread types showed a formulation with sunflower oil (Supplementary Table S1), which led to higher vitamin E content in those specific stuffs.

It is important to point out that food technology interventions to improve the shelf life and rheological properties of GFP have influenced their nutritional profile [12]. In order to avoid the absence of the mentioned micronutrients without fortifying foodstuffs, different strategies can be proposed: avoiding starch as a major ingredient, sourdough fermentation, and using less popular grain GF flour such as that from pseudocereals (buckwheat, quinoa, amaranth and teff) or legumes, including wholemeal forms of gluten-free cereals [18,19,33,34]. In our samples, only one out of twelve foodstuffs analyzed in each group contained pseudocereals in their ingredients list (4 to 5 g in 100 g), reflecting the need of more research on the properties and technological characteristics of these raw materials, and promotion of their use.

#### *3.3. Micronutrient Intake in Celiac Children and Adolescents*

It is known that GFD can lead to imbalanced macronutrient distribution. Our previous work [26] reported that celiac children and adolescents consumed more fat and less carbohydrate than recommended and pointed at GF rendered foods as one of the culprits. Thus, taking into account directly analyzed micronutrient content, their intake on that pediatric cohort was calculated considering their age group and gender, and compared to FESNAD recommendations (Supplementary Figure S1).

More than 1/4 of participants did not reach recommendations for vitamin A and vitamin E. Four out of ten children and adolescents with CD showed low intake of folic acid, which was even less than 66% of the recommendation for 25% of participants. Sixty percent of participants did not get that for vitamin D, and moreover, about 40% of them did not reach 25% of the recommendation. Most participants showed very low intakes of biotin, iodine and copper. Slightly over half the participants did not fulfil 50% of iodine recommendation and more than 40% were not able to achieve 25% of that of biotine. The intake of the rest of micronutrient was appropriate. With the exception of vitamin D, the results obtained differ from those obtained in similar pediatric research on celiac children, where low intake of iron, calcium, selenium and magnesium was observed [10,35,36].


#### *Foods* **2019**, *8*, 321

gluten-containing

 products, expressed by 100 g

**Table 2.**

Analytical [4]

micronutrient

 content in gluten-free rendered foodstuffs divided by food groups, compared to

Considering all the above mentioned, it does not seem that the GFP groups analyzed contribute to the micronutrient deficits detected in young celiac people's diets. In fact, cereals have only a modest role as source of these micronutrients. It is important to highlight that in our previous study [26] we reported unhealthy dietary habits in these celiac children and adolescents: very low cereal and vegetable consumption, low fruit and nut intake and excessive meat consumption. Thus, general recommendations to promote healthy GFD should be given to amend the observed wrong habits. It is worth mentioning that this conclusion refers to our cohort, and that in other dietary patterns, GFPs role could be different.

It must be pointed out that, in the case of folic acid, we observed a lower content of this vitamin in GFP than in their gluten containing equivalents. In this regard, in Canada and USA [37,38] the fortification of wheat flour with folic acid is mandatory, but not for other alternative flours, such as the ones used in GFP. Taking into account the folate deficiency observed in GFD, its fortification in GFP or ingredients could be of interest for celiac children. Folate fortification measures could also be extended to biotin, whose widespread diet-deficiency in celiac population was alarming. In fact, some of the GF cereals analyzed were supplemented with this vitamin (Supplementary Table S1).

It is of interest to point out that some deficiency diseases found in celiac people, such as anemia, low bone density or zinc depletion [39] are not only justified by nutritional shortages. Other pathological situations such as systemic inflammation or intestinal microbiota alteration appear to contribute to the persistence of those deficiencies in some celiac individuals [12,40,41].

It has to be highlighted that this paper presents wide-ranging high-quality nutritional information about GF bread, pasta and cereal micronutrient content. This remains limited in the literature and even more so in food panels or in databases used for GFD design and evaluation, where it is crucial. Moreover, it has assessed not only GFP composition but also its dietetic role, discussing, in general terms, its involvement in micronutrient deficiencies of the GFD of children and adolescents. Nevertheless, extrapolation to celiac adults is limited and needs further research. Moreover, as proposed elsewhere [42], the bioavailability of GFP is a matter of concern that should also be taken into account in further studies. Finally, it is also of great interest to analyze the nutritional composition of GFPs considering their ingredients list to define the role of ingredients such as gluten free cereals or pseudocereals, starches and additives in the final composition of the product.

The practical outcomes of the present study are relevant in improving the universal guidelines for food fortification in CD [43,44]. Some individualized supplementation is usually proposed for celiac people based on micronutrient related blood monitoring. Nevertheless, GFP fortification for folate and biotin could contribute to preventing the deficiencies observed in GFD, at least in the case of celiac children and youngsters.

#### **4. Conclusions**

Even if lower micronutrient content was found in the analyzed GFP groups, this fact was not related with the micronutrient deficits detected in GFD in a cohort of celiac children and adolescent. Nevertheless, according to the obtained results, GFP fortification for folate and biotin seems to be a suitable proposal in order to prevent the deficiencies observed in GFD.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2304-8158/8/8/321/s1, Figure S1: Percentage of celiac children and adolescents who accomplished or did not achieve 2/3 of the dietary reference intake of the vitamins and minerals (proposed by the Federation of Spanish Societies of Nutrition and Dietetics, FESNAD), Table S1: Analyzed gluten free products and ingredients declared on the package label.

**Author Contributions:** E.S. carried out the experimental design. I.L. and I.T. analyzed lipid, protein and micronutrient content. V.N. and A.L. performed the analysis of diet. M.Á.B., M.d.P.F.-G. and J.M. analyzed all data and contributed to statistical analysis. I.L., I.T., V.N. and J.M. wrote the manuscript.

**Funding:** Idoia Larretxi is a fellowship of the University of the Basque Country, UPV/EHU. This research was supported by a grant from the Basque Government (Proyectos de Investigación Focalizada Agricultura PA15/01) and a grant from the University of the Basque Country, UPV/EHU, (University-Society US15/06 and US18/15). The authors thank for technical and human support provided by SCAB from SGIker of UPV/EHU.

**Conflicts of Interest:** The authors declare that they have no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **Lipids and Fatty Acids in Italian Durum Wheat (***Triticum durum* **Desf.) Cultivars**

#### **Valentina Narducci, Enrico Finotti, Vincenzo Galli and Marina Carcea \***

Research Centre for Food and Nutrition, Council for Agricultural Research and Economics (CREA), Via Ardeatina 546, 00178 Rome, Italy; valentina.narducci@crea.gov.it (V.N.); enrico.finotti@crea.gov.it (E.F.); vincenzo.galli@crea.gov.it (V.G.)

**\*** Correspondence: marina.carcea@crea.gov.it; Tel.: +30-06-5149-4638

Received: 22 May 2019; Accepted: 19 June 2019; Published: 21 June 2019

**Abstract:** The level of variation in lipids and their fatty acids was determined in the grains of 10 popular durum wheat cultivars commercially grown in Central and Southern Italy. Samples were harvested for two consecutive years to account for differences due to changes in climatic conditions. Total fat content was determined by means of the International Association of Cereal Science and Technology (ICC) Standard Method No. 136, whereas the fatty acid profile was determined by gas chromatography. Total lipid content ranged from 2.97% to 3.54% dry basis (d.b.) in the year 2010 and from 3.10% to 3.50% d.b. in the year 2011, and the average value was 3.22% d.b. considering both years together. Six main fatty acids were detected in all samples in order of decreasing amounts: linoleic (C18:2) > palmitic (C16:0) ≈ oleic (C18:1) > linolenic (C18:3) > stearic (C18:0) > palmitoleic (C16:1). Significant variations in the levels of single acids between two years were observed for three samples. These results will be very useful in the updating of food composition databases in general and will help authorities to set proper quality standards for wholegrain flours and products where the germ should be preserved, considering also the recent interest of industry and consumers for these kinds of products.

**Keywords:** durum wheat; fatty acids; grain; kernel; lipids

#### **1. Introduction**

Durum wheat (*Triticum durum* Desf.) kernels contain about 2.4–3.8% dry basis (d.b.) of lipids [1]. Roughly two thirds (66%) of them are contained in the germ, 15% are in the bran (particularly in the aleuronic layer), and about 20% are distributed in the endosperm, partly within the starch granules. From a chemical point of view, the most abundant fraction is composed by nonpolar lipids, which are mainly storage acylglycerols. Phospholipids, glycolipids and other classes are present in lesser amounts. The fatty acids of wheat lipids are mostly unsaturated (C18:2, C18:1, C18:3 and C16:1) and two of them are essential (linoleic and linolenic). This increases the value of wheat lipids for human nutrition, because essential fatty acids are precursors of important classes of biomolecules in the human body (like prostaglandins and membrane phospholipids) and are involved in metabolic processes like regulation of blood lipid levels, particularly cholesterol [1–3].

Lipid content, lipid classes and fatty acid levels in wheat kernels depend on a set of factors, some of which are genetic, such as species and variety [4], whereas others depend on the environment and are related to pedoclimatic conditions, agronomic practices and maturity level [1,4,5]. For example, durum wheat and hard red wheat generally have a higher lipid content than soft white wheat and the levels of fatty acids are different in durum and in soft wheat. In regard to climatic conditions, it has been seen that cold weather favors an increase of lipid content in wheat and a higher degree of unsaturation in fatty acids due to the need for membrane fluidification [6]. Other kinds of biotic and abiotic stresses can influence the level of saturated and unsaturated fatty acids in plants [7]. Moreover, different extraction and analytical methods can also account for the differences found in the literature [1,8]. Notwithstanding the number of samples analyzed, we can assume that data about fatty acid levels in durum wheat are abundant in the literature, but it is difficult to have a clear idea of their content and to make comparisons for a number of reasons: (i) different authors report fatty acids as percentage, alternatively referring to: (1) total lipids, (2) total fatty acids, or (3) kernel weight (in addition, some authors analyze germ oil and others analyze whole kernels); (ii) authors interested in statistic elaborations (e.g., in order to investigate variation factors or to look for discriminating parameters) often report charts and graphs rather than tables of data; (iii) cultivars are different in different countries and new ones are constantly bred; and (iv) databases do not always report the sample numerosity and the standard variation of the means.

In this work, the content and level of variation in lipids and of their fatty acids in the durum wheat kernels commercially grown in Italy (where durum wheat is an important cereal crop mainly used for pasta manufacturing) were assessed. For this reason, we selected 10 cultivars amongst the most commonly grown for pasta making. Samples were collected in several locations of Central and Southern Italy to account, at least partially, for differences due to different pedoclimatic environments; Southern Italy is characterized by milder winters and warmer springs and summers with respect to Central Italy, however both areas are considered highly suitable for durum wheat cultivation. Moreover, crops from two consecutive years were collected from the same fields.

The knowledge generated by this research will be very useful in the updating of food composition databases in general and will help authorities in setting proper quality standards for wholegrain flours and products where the germ should be preserved, considering also the recent interest of industry and consumers for these kinds of products and the lack, in several cases, of specific legislation.

#### **2. Materials and Methods**

#### *2.1. Samples and Sample Preparation*

Representative samples of durum wheat grains, belonging to 10 cultivars selected amongst the most frequently grown in Italy, were collected at harvest for two consecutive years (2010–2011) in 10 different locations of Central and Southern Italy (Table 1). Eight samples came from the Central regions of Italy (Tuscany and Marche) whereas twelve were from different locations in the Sicilian region, in the South. All locations belong to the area traditionally dedicated to durum wheat cultivation in Italy.


**Table 1.** Durum wheat sample specifications: cultivar, region and location.

Durum wheat in Italy is grown under rain-fed production: it is planted in late autumn or early winter and harvested in early summer, which often leads to limited rainfall and high temperatures, resulting in water stress during grain filling. Crop rotation and balanced nutrient management (mainly nitrogen and phosphorus, pre-sowing and topdressing fertilization) are practiced to ensure that the crop produces the greatest possible high-quality yield with the moisture that is available. The main

climate factors influencing durum wheat crop quality are rainfall and temperature during the growing season. Data on these two factors of the years 2009–2011 in Central and Southern Italy can be found in the reports by the Italian High Institute for Environmental Protection and Research (ISPRA,) [9–11].

Fifty grams of each cleaned sample were milled by means of a Cyclotec laboratory mill (Foss-Tecator, Hillerød, Denmark) equipped with a 0.5 mm screen, to obtain wholemeal flours that were used for the subsequent analyses.

#### *2.2. Chemicals*

Chloroform, ethyl alcohol (96% *w*/*w*), methanol, *n*-hexane, formic acid (99% *w*/*w*) hydrochloric acid (37% *w*/*w*) and anhydrous sodium sulphate were of analytical grade and were purchased from Carlo Erba (Milan, Italy). Boron trifluoride (approximately 10% *w*/*w* in methanol for gas chromatography (GC) derivatization) was purchased from Sigma-Aldrich (St. Louis, MO, USA). Fatty acid standards (C16:0, C16:1, C17:0, C18:0, C18:1, C18:2, C18:3) were also purchased from Sigma-Aldrich.

#### *2.3. Analyses*

Moisture of wholemeal flours was determined by oven drying at 130 ◦C according to the ICC Standard No. 110/1 [12].

Total fat was determined by hydrolysis in formic acid and hydrochloric acid at 75 ◦C reflux for 20 min followed by extraction in hexane and evaporation, according to the ICC Standard No. 136 [12].

The fatty acid profile was determined by gas chromatography (GC). About 5 g of wholemeal flour (in duplicate) was introduced in a Corning tube and suspended in 10 mL of chloroform–methanol 2:1 acidulated with 6 N HCl. A magnetic bar was added, and the tube was left to extract overnight at room temperature on a magnetic stirrer. The mixture was filtered through Whatman Grade 1 (1–11 μm) filter paper into an oven dried flask, then the solvent was evaporated by nitrogen flux followed by oven drying at 30 ◦C. The contents of the flask were re-dissolved in chloroform–methanol 2:1 to a volume of exactly 10 mL, then an aliquot was derivatized according to Zweig and Sherma [13] as follows: 100 μL of this solution was introduced into a Corning tube containing 3 mL of methanol and a few boiling stones, then 0.5 mL of BF3–methanol (10% *w*/*w*) was added and the tube caps were loosely screwed. The tubes were put onto a heating plate in a water bath and left to gently reflux at 72 ◦C for 30 min. Following this, the reaction was quenched with 2 mL of water, then the mixture was cooled to room temperature and extracted three times with 3 mL of *n*-hexane. The hexane extracts were reunited into a vial and finally the hexane was evaporated by nitrogen flux. The vial was stored under nitrogen at −18 ◦C for a few days. Immediately prior to GC analysis, the contents of the vial were re-dissolved in 300 μL of hexane and 2 μL were injected. The GC instrument was an HP 5890 equipped with a Supelco (Sigma-Aldrich, St. Louis, MO, USA) SPB®-PUFA (poly unsaturated fatty acids) column of 30 m length and a flame ionization detector (F.I.D.). The instrumental analysis was run according to Finotti et al. [14]: 50 ◦C for 1 min, ramp of 10 ◦C/min until 160 ◦C, stay at 160 ◦C for 1 min, ramp of 2 ◦C/min until 240 ◦C. The detected peaks were individuated by comparison with chromatograms of standards (C16:0, C16:1, C18:0, C18:1, C18:2, C18:3) and quantified by using C17:0 as an internal standard.

#### *2.4. Statistics*

The Shapiro–Wilks normality test, *F*-test for homogeneity of variance, Student's *t*-test and Friedman test followed by Wilcoxon pairwise comparisons were performed by means of the PAleontological STatistics (PAST) statistical package [15]. Two-way ANOVA followed by Tukey's test (only in cases with a normal variable and homogeneous variances) and box-plots were performed by means of StatSoft Statistica 8.0 (TIBCO Software, Palo Alto, CA, USA). Calculations were performed by means of Microsoft Excel (Redmond, Washington State, USA).

#### **3. Results**

#### *3.1. Total Lipids*

Total lipids ranged from 2.97% to 3.54% d.b. in the year 2010 and from 3.10% to 3.50% d.b. in the year 2011, and the average value was 3.22% d.b. considering both years together (Table 2). The moisture content of grains ranged between 10.5% and 12.3% and the average was 11.4% (Table 2). Total lipid content was strongly dependent on the combination of cultivar (cv)/growing site (*p* < 0.01) and to a minor extent on the growing year (*p* < 0.05), whereas the interaction cv/site × year was not a statistically significant factor of variation. In any case, differences were very small: up to 0.57 between samples of different cultivars and up to 0.18 between years for samples of a same cv/site (Table 2). Differences between years for samples of the same cv/site were not significant. The total lipid values found in this study are in line with those reported by the USDA National Nutrient Database (2.8 g/100 g d.b. for product N. 20076 "wheat, durum", mean of 18 samples, standard error 0.060) and by the Italian food composition tables (3.3 g/100 g d.b. for "durum wheat") compiled by the Italian National Institute for Research on Food and Nutrition (INRAN) [16,17]. If we take into account the geographical separation into Central and Southern Italy, we can say that the average total lipid values for all samples were 3.24% and 3.21% d.b. respectively, whereas the range of values was 2.97–3.54% for Central Italy and 3.09–3.41% d.b. for Southern Italy.


**Table 2.** Moisture and total lipids in the grains of 10 Italian durum wheat cultivars grown in different locations for two consecutive years.

abcdef: different letters correspond to significant differences (*p* < 0.05) according to 2-way ANOVA and Tukey's test. ns: not significant.

#### *3.2. Fatty Acid Profile*

Six main fatty acids were detected in all samples, as expected. In order of decreasing amounts, they are: linoleic (C18:2) > palmitic (C16:0) ≈ oleic (C18:1) > linolenic (C18:3) > stearic (C18:0) > palmitoleic (C16:1). This can be clearly seen from the box plot elaboration reported for each separate year and for the two years together (Figure 1). This distribution did not change whether considering both years separately or together. Detailed data of fatty acids in all samples are reported in Table 3.

Linoleic acid (C18:2) was present in amounts ranging from 0.50–1.14 g/100 g d.b. throughout all samples, with a mean of 0.68 and a standard deviation (SD) of 0.16 (Table 3). For comparison, the USDA National Nutrient Database reports 1.04 g/100 g d.b. for product N. 20076 "wheat, durum" and the INRAN food composition tables report 1.36 g/100 g d.b. for durum wheat. Neither database reports any information on standard errors for all acids.

Palmitic (C16:0) and oleic (C18:1) acids were detected in equal amounts. Palmitic acid ranged from 0.17–0.36 g/100 g d.b., mean 0.24 (SD 0.04) and oleic acid ranged from 0.17–0.43 g/100 g d.b., mean 0.24 (SD 0.07). The USDA reports 0.51 g/100 g d.b. for palmitic acid and 0.40 g/100 g d.b. for oleic acid, whereas the INRAN database reports 0.47 g/100 g d.b. and 0.38 g/100 g d.b., respectively (Table 3).

**Figure 1.** Box plot (percentiles) of fatty acids in samples of Italian durum wheat (10 cultivars, grown in the same location for two consecutive years).

Linolenic acid (C18:3) ranged from 0.06–0.14 g/100 g d.b., mean 0.08 (SD 0.02). The USDA and the INRAN databases report 0.05 g/100 g d.b. and 0.11 g/100 g d.b., respectively. Stearic acid (C18:0) ranged from 0.01–0.03 g/100 g d.b., mean 0.02 (SD 0.005). The USDA and the INRAN databases report, for this acid, 0.03 g/100 g d.b. and 0.02 g/100g d.b. respectively. Finally, palmitoleic acid (C16:1) was detected in very small amounts, ranging from 0.004–0.007 g/100 g d.b., mean 0.005 (SD 0.001). Both the USDA and INRAN databases report 0.01 g/100 g d.m. for this acid.

A series of *t*-tests, performed for each fatty acid on each pair of samples from the same cv/site between the two growing years, showed a significant difference between the years 2010 and 2011 in a few cases only, namely: all acids except C16:1 varied in Ancomarzio SI and Iride AG; only the acids C18:1, C18:2 and C18:3 varied in Ciccio EN (Table 3).


**Table 3.** Fatty acids in 10 Italian durum wheat cultivars, grown in different locations for two consecutive years (g/100 g sample, d.m.).

Asterisks indicate significant difference between the year 2010 and 2011, \* *p* < 0.05, \*\* *p* < 0.01 (*t*-test). ‡: values on dry basis, calculated by the authors from original data in USDA and INRAN databases that are expressed on as-is basis.

#### *3.3. Saturated and Unsaturated Fatty Acids*

As expected, polyunsaturated fatty acids were preponderant over saturated and monounsaturated fatty acids in all samples (*p* < 0.01, Friedman test; see Figure 1), ranging from 0.57–1.28 g/100 g d.b. (Table 3). Total monounsaturated and total saturated, whose levels were roughly similar (*p* < 0.01), covered from 0.18–0.44 g/100 g d.b. and from 0.19–0.39 g/100 g d.b., respectively. The unsaturated/saturated ratio ranged from 3.15–4.44 g/100 g d.b. considering all samples, with a mean of 3.83 (SD 0.38). This mean is higher than that reported by USDA (3.0) and INRAN (3.5). A series of *t*-tests, performed on each pair of samples from the same cv/site grown in different years, showed a significant difference for the unsaturated/saturated ratio between years in only three cases (Ancomarzio SI, Iride AG and K26 EN) (Table 3).

#### **4. Discussion**

Total lipids were in line with the values reported by the USDA and the INRAN databases (nearer to the Italian value) and it was not possible to detect any difference between the geographical areas of Central and Northern Italy.

In regard to fatty acid composition, even if Bottari et al. in 1999 [18] observed the presence of more than 60 peaks by gas chromatography and mass spectrometry (GC-MS) and identified fatty acids with even numbers of carbon atoms from C12 to C30 as well as C15 and C17, the major fat components were saturated and unsaturated C16 and C18 and particularly C16:0, C18:1 and C18:2, which together represented around 90% of the total.

Actually, the USDA database (but not the INRAN one) and other works also report small amounts of C14:0 in durum wheat kernels (USDA 0.003 g/100 g fresh matter, corresponding to 0.0035 g/100 g d.b.). We did not detect this acid, as it was at the limit of detection of our method. There are publications reporting other fatty acids as well (i.e., C17, C20, C22 and C24), some in kernels (Beleggia et al. [5] who uses a GC-MS instrument) and others in germ oil [19,20]. However, only C16:0, C18:0, C18:1, C18:2 and C18:3 are constantly reported by all published works and are regarded as the most important ones in durum wheat, with others amounting to about 1–2% in total [1].

For all fatty acids except C18:3 and for total saturated, total monounsaturated and total polyunsaturated acids, the mean calculated for our samples was lower than the values reported by USDA and INRAN (roughly two thirds–half, *p* < 0.01 against a hypothetical value; see Table 3). However, the range of the detected values contained the reference values, except for C16:0 and for total saturated acids, for which the detected range extended entirely below the USDA and INRAN means. Neither database reports the standard deviations for fatty acids in durum wheat and only the USDA one reports sample numerosity (that is, 29); in this latter case, a certain width around the reported value can be supposed, but it is not quantified. On the contrary, for the unsaturated/saturated ratio, the range of the detected values extends entirely above the mean reported by USDA and contains that reported by INRAN. As a matter of fact, there are notable differences between the two references used. The INRAN values are equal to or higher than the USDA values for the considered variables, in particular for C18:0 (+50%), C18:2 (+31%), C18:3 (+120%), total polyunsaturated acids (+34%) and unsaturated/saturated ratio (+17%).

All the reported differences can be explained by the differences in genetic characteristics, pedoclimatic conditions, agronomical treatments and analytical procedure, as stated in the Introduction. In particular, Beleggia et al. [5] identified the interaction genotype × year × treatment as the main contributor to the variability of the fatty acid levels observed in 24 durum wheat samples, especially for linoleic, oleic and stearic acids. Armanino et al. [4] linked the fatty acid profile of 135 samples of durum wheat to the cultivar, the geographic origin and the harvest year. The variation in saturated and unsaturated fatty acids within the same variety is also associated with various kinds of biotic and abiotic stresses, like low or high temperature, salt, drought, pathogens and others [6,7].

Also, in our study, different conditions related to location and climatic factors can account for some of the observed variability in lipid parameters. In fact, from the ISPRA reports [9–11], we can briefly say that in both areas of Italy (Central and South), temperatures were similar in the first part of the two growing seasons (October–December 2009 and 2010), except the month of December which was warmer in 2009 than in 2010. In the second part of the growing season (January–June, particularly April–June), the Central area showed warmer temperatures in 2011 than in 2010. In regard to precipitation, the first growing season (2009) started with a lesser amount of rain in October–December with respect to the second one (2010) and continued with a higher amount of rain in the January–June period. This happened both in Central and Southern Italy.

#### **5. Conclusions**

This work contributes to the knowledge on the content and variability of total fats and of the main fatty acids in durum wheat kernels. The values obtained in this study are also compared with reference values from national and international databases. In this paper, the use of standard methods of analysis, statistical data (numerosity of samples, mean, standard errors) and the specification of all the elements that allow for conversion of results into different units of measure (g/100 g dry or wet sample, g/100 g fat matter) make this data very useful in the compilation of databases and easy to compare with other data. Moreover, updated data on lipids are needed to set proper quality standards for products such as wheat wholegrain flours and foods where the presence of germ is desirable.

**Author Contributions:** Conceptualization and supervision, M.C. and E.F.; investigation, E.F., V.N. and V.G.; data curation and writing—original draft preparation, V.N.; funding acquisition and writing—review and editing, M.C.

**Funding:** This research was supported by the Project PASTA-COUS financed by the Italian Ministry of Education, University and Research (MIUR) (Grant Number CUP H36J16000730001).

**Acknowledgments:** The authors wish to acknowledge Luigi Bartoli's technical assistance in sample preparation.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

MDPI St. Alban-Anlage 66 4052 Basel Switzerland Tel. +41 61 683 77 34 Fax +41 61 302 89 18 www.mdpi.com

*Foods* Editorial Office E-mail: foods@mdpi.com www.mdpi.com/journal/foods

MDPI St. Alban-Anlage 66 4052 Basel Switzerland

Tel: +41 61 683 77 34 Fax: +41 61 302 89 18