By continuing to browse this site you agree to us using cookies as described in About Cookies
Notice: Wiley Online Library will be unavailable on Saturday 7th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 08.00 EDT / 13.00 BST / 17:30 IST / 20.00 SGT and Sunday 8th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 06.00 EDT / 11.00 BST / 15:30 IST / 18.00 SGT for essential maintenance. Apologies for the inconvenience.
Bolus Formation and Disintegration during Digestion of Food Carbohydrates
Gail M. Bornhorst,
Author Bornhorstis is with Dept. of Biological and Agricultural Engineering, Univ. of California, Davis, 1 Shields Ave., Davis, CA 95618, U.S.A. Author Singh is with Dept. of Biological and Agricultural Engineering, Univ. of California, Davis, 1 Shields Ave., Davis, CA 95618, U.S.A., and Riddet Inst., Massey Univ., Palmerston North, New Zealand. Direct inquiries to author Singh (E-mail:firstname.lastname@example.org).
Author Bornhorstis is with Dept. of Biological and Agricultural Engineering, Univ. of California, Davis, 1 Shields Ave., Davis, CA 95618, U.S.A. Author Singh is with Dept. of Biological and Agricultural Engineering, Univ. of California, Davis, 1 Shields Ave., Davis, CA 95618, U.S.A., and Riddet Inst., Massey Univ., Palmerston North, New Zealand. Direct inquiries to author Singh (E-mail:email@example.com).
Abstract: The first step in the digestion process is mastication, or chewing, when food is broken down, lubricated with saliva, and formed into a cohesive mass known as the food bolus. Upon swallowing, the bolus moves to the stomach and undergoes further breakdown during gastric digestion. The subject of this review is the formation of the food bolus and its subsequent breakdown in the stomach. Bolus formation has been widely studied, especially in terms of food particle size and lubrication. However, information about bolus disintegration is limited, and this review focuses on the breakdown of bread and starch-based foods. Bolus formation and disintegration are key steps in the overall digestion process, as they control the rate at which ingested food components and nutrients are absorbed and released into the body. Information on the rate kinetics of bolus disintegration is necessary in developing a quantitative understanding of the food digestion process.
The first step in the digestion process is oral digestion, which begins with mastication, or chewing. During mastication, the simultaneous processes of food comminution and lubrication occur, the end product being the food bolus. After formation, the bolus will be swallowed for further digestion in the stomach. Although mastication seems like a simple process, there are many factors involved. Physiological characteristics of the individual performing the chewing action, such as facial anatomy, gender, age, personality type, time of day, dentition status, as well as properties of the food being chewed, such as hardness, moisture content, fat content, food portion size, and food structure (Yurkstas 1965; Gonzalez and others 2004; Rey and others 2007) all have an effect on the formation of the food bolus. After formation, the bolus will be swallowed, transported through the esophagus, and move into the stomach. As boluses enter the stomach, they will “stack up” in the curvature of the stomach according to the time they were ingested (Schulze 2006). Layers will begin to form according to density and solids content. During the gastric digestion process, the boluses are physically reduced in size while being chemically broken down due to the acidic and enzymatic conditions of the gastric secretions. The rate at which foods disintegrate will control the rate at which they are emptied from the stomach and move to the intestines where nutrients are absorbed.
The first topic of this review focuses on the importance of food mechanical properties during bolus formation. Food properties such as hardness, moisture content, and fat content will influence bolus formation as a result of variations in saliva secretion and differences in chewing forces and duration. Especially, in the case of starch-based foods, the amount of saliva incorporated in the bolus may play a significant role in the breakdown of starches, due to the α-amylase content of saliva. The second topic of this review focuses on the breakdown of bread and other starchy foods during gastric digestion. This breakdown has been studied in vivo in terms of gastric emptying rate and glucose response after consumption of a meal. Similarly to bolus formation, gastric emptying and glucose response of a meal have been correlated to both food macro- and microstructure in bread and starchy foods. Since many in vivo studies have been performed on humans using noninvasive methods to measure gastric emptying or glucose response after a meal, it has been difficult to quantify the precise relationship between food structure and its bolus breakdown during gastric digestion. However, in vitro systems can be used for this purpose and a brief description of a variety of in vitro gastric digestion systems is presented.
Stages of oral processing
After a food product is ingested, it is processed in the oral cavity. The oral process is comprised of 4 key steps: Stage I, processing, Stage II, and Stage III (Hiiemae and Palmer 1999). During Stage I, the ingested food is moved from the front of the mouth to the teeth so that it can be broken down. This stage occurs quite fast and takes about the same duration for most food types, around 280 ms. The processing phase occurs when the food particles are broken down via crushing/grinding with the teeth; this step takes longer for harder products. Stage II (transport) occurs gradually during the processing phase; as particles are broken down to the appropriate size, and they are transported to the back of the oral cavity to form a bolus, leaving larger particles to be further broken down (Hiiemae and Palmer 1999; Smith 2004). A bolus is formed by folding and manipulating food particles with the tongue (Prinz and Heath 2000). Stage III occurs after a bolus is formed, which is the preswallowing stage; the bolus is moved to the back of the tongue in preparation for swallowing (Hiiemae and Palmer 1999; Smith 2004).
Bite size and chewing time
To determine the time of the entire chewing sequence (from Stage I to Stage III) for various food products, Hiimae and others (1996) tested apple, banana, and biscuit (hard cookie) as 3 different foods with vastly different texture on 11 human subjects. They measured both the “natural” (self-selected) bite sizes as well as the chewing sequence period and number of swallows during chewing. They found that the size of the bite varied inversely with product hardness. For a soft food, such as banana, the mean bite size was 12.45 ± 3.45 g, although for a hard food, such as a biscuit, the mean bite size was 2.48 ± 0.88 g, and for an intermediate food, such as an apple without the peel, the mean bite size was 7.19 ± 2.28 g. The total chewing duration varied slightly for each of the foods; this variation was attributed to the type of food and the varying amount of food ingested. Although the total chewing duration was not affected by food texture, the number of chews required increased with food hardness. Soft foods (banana) required a lower number of chews, while harder foods (biscuit) required a larger number of chews. This study demonstrates the importance of food hardness in the overall chewing process, with both bite size and number of chews being influenced by food hardness.
Saliva plays an important role in the digestive process, especially in digestion that occurs in the upper digestive tract (mouth, pharynx, esophagus, stomach, and duodenum). There are many functions of saliva in the oral cavity, including, but not limited to, lubrication of the surfaces of the oral cavity, facilitation of speech and mastication, formation of the food bolus, cleaning food in the oral cavity, neutralizing oral acids, cooling hot food, acting as an antimicrobial agent, and beginning the digestion of starches and lipids by means of salivary α-amylase and lipase, respectively (Pedersen and others 2002).
About 90% of whole saliva is released by three major salivary glands in the oral cavity: the parotid, the submandibular, and the sublingual glands. The parotid releases a low-viscosity, amylase-rich saliva, while the submandibular and sublingual glands release a higher viscosity, mucin-rich saliva. The remaining 10% of whole saliva is released by minor glands, distributed throughout the oral cavity, that release the majority of salivary proteins. In normal healthy individuals, between 0.5 and 1.5 L of saliva is produced daily (Pedersen and others 2002). The specific production of saliva depends on many factors and shows high interindividual variation, with up to 40% average variation for daily saliva secretion over a period of 3 days (Richardson and Feldman 1986). Salivary secretion also varies with circadian rhythm throughout the day, in both flow rate and composition (Dawes 1972). Stimulated salivary secretion can be measured either by expectoration of saliva after a stimulus (subjects are given a piece of candy or gum and they are asked to spit out any saliva in the oral cavity), or the saliva can be removed using a catheter and a suction pump. Both methods have shown to give similar results and are highly correlated with one another (Richardson and Feldman 1986). Unstimulated saliva can be measured by a draining method, where saliva is allowed to drain through a funnel in the mouth; a spitting method, where subjects spit out any accumulated saliva at certain intervals; a suction method, where plastic tubing attached to a vacuum pump is inserted under the tongue to collect salivary secretion; or the swab method, where dental cotton rolls are weighed, placed into the mouth to absorb any secreted saliva, then removed and reweighed to determine the amount of saliva absorbed. All of these methods result in similar mean saliva flow rates; however, their variability and reproducibility differ. The swab method produces the highest variability and lowest reliability, making it the least preferred method. The suction method may actually stimulate salivary secretion, leading to erroneously increased secretion values. The spitting or draining methods are the preferred methods for collection of unstimulated saliva (Navazesh and Christensen 1982).
Gavião and others (2004) examined the salivary secretion in response to chewing toast, toast with margarine, various sizes of cake, and cheese. Although the results of the study showed no difference in saliva secretion in mL/min, the total amount of saliva secreted for the various products was different in terms of mL/g food, due to differences in the chewing time (longer chewing time results in more saliva/g food incorporated). Samples of unbuttered toast required the highest levels of saliva (mL/g) and the longest chewing time, but had the lowest moisture and fat contents. This substantiates previous studies that suggest dry, lower fat products that need more saliva in order to adequately moisten them and prepare a cohesive bolus to be swallowed. Chewing time increased and saliva secretion (mL/g) slightly decreased with an increasing sample size when small, medium, and large portions of cake were chewed. The chewing time for portions of cake ranged from 17.4 (small) to 30.7 s (large), with the salivary incorporation ranging from 0.40 mL/g (small) to 0.33 mL/g (medium). The chewing time was significantly different according to portion size for all 3 portions; the salivary incorporation was significantly different between the small cake portion and the medium and large cake portions.
In a similar study by Engelen and others (2005), chewing time and saliva flow rates were measured on 7 natural foods: cake, melba toast, bread, toast, carrot, peanut, and cheese. They found that chewing time for an equal volume of product varied with product moisture and hardness in the following order: cake and bread < toast < cheese < peanut < melba toast < carrot. This demonstrates the relationship between product hardness and number of chewing cycles until swallowing, with harder foods requiring a greater number of chewing cycles. This study also showed that when the cake, melba toast, and toast were chewed with butter, they required a lower number of chewing cycles, exhibiting an inverse relationship with fat content and chewing cycles. Assuming that the salivary secretion (mL/min) is constant, as shown by (Gavião and others 2004), this signifies that for a longer chewing time, more saliva will be incorporated into the food product.
Similarly, a study by Brudevold and others (1990) tested chewing times and saliva secretion for cookies with varying sucrose and fat contents. They found chewing times varied from 28 to 43 s. The addition of sucrose and fat both shortened the chewing time and the chewing times were significantly different according to type of cookie chewed. The percentage of water in the bolus ranged from 39.1%–49.3%; however, the amount of sucrose did not influence the final water content of the bolus. The cookies with added fat had a lower water percentage than those with no fat added. The saliva flow rates varied from 6.3 to 8.3 mL/min with increasing levels of sucrose, and from 6.9 to 7.0 mL/min with increasing levels of fat.
The main enzyme found in saliva is α-amylase, which breaks down carbohydrates into maltose, maltotriose, and α-limit dextrins by cleaving the α-1–4 glycosidic bond in the complex carbohydrates (Robyt and French 1970). These products are then further hydrolyzed by brush-border enzymes to produce the glucose that can be absorbed in the small intestine (Rosenblum and others 1988). Saliva, by some, is not considered to have a major impact on carbohydrate digestion, as its optimum pH level is 6.8 and becomes inactivated by the low pH of the gastric acid in the stomach (Pedersen and others 2002). However, although salivary amylase does become inactive by encountering very low concentrations of hydrochloric acid, it may still be present for up to 15–30 min into gastric digestion; and it accounts for an average digestion of 76% of the starch in mashed potatoes and 59% of the starch in bread (Bergeim 1926). Depending on the specific pH conditions of the stomach, salivary α-amylase may actually reach the small intestine without becoming inactive. Fried and others (1987) showed that salivary α-amylase became inactive in a pH range of 3.3–3.8 during in vitro testing, and accounted for about 14% of the total (active) amylase found in the small intestine (the remaining 86% being comprised pancreatic amylase), indicating that there is still significant starch digestion that may occur in the stomach due to the presence of salivary α-amylase. The inactivation of α-amylase may also be influenced by specific meal composition. Rosenblum and others (1988) showed that the presence of 0.1% and 1% starch will significantly decrease the α-amylase inactivation at pH 3.0. They demonstrated this same effect with increasing concentration of maltose and maltotriose (up to 5%), showing that α-amylase will be inactivated at a slower rate, even at pH 3.0, in the presence of polysaccharides and oligosaccharides. They hypothesized that this effect may be due to an interaction of the starch at the active site of the α-amylase that influences the amylase inactivation.
Another enzyme found in saliva is lingual lipase, which breaks down lipids. However, lingual lipase only breaks down a small fraction of the ingested lipids, as most of triglyceride digestion is caused by pancreatic lipase (Pedersen and others 2002).
Other effects of saliva on oral digestion
Other properties of saliva can also have an effect on oral digestion, such as the surface tension and viscosity of the saliva. The surface tension of saliva is partially responsible for the adhesive effects of saliva, both to the oral cavity and also for the food particles in the bolus. Saliva surface tension was measured to be 53.1 dynes/cm, averaged from 24 individuals (Glantz 1970). The viscosity of saliva can be affected by many factors, some of which may even be related to diet. It has been shown that dietary tannins (the astringent compounds found in foods such as coffee, tea, and red wine) significantly decrease the viscosity and increase the friction in saliva. These changes will affect the lubricating properties of saliva, and ultimately, the bolus formation (Prinz and Lucas 2000).
Modeling of Particle Breakdown During Mastication
The biological objective of chewing, when foods are ingested and broken down in the mouth, is to increase the surface area of the particles in order to release flavor compounds in the mouth and to facilitate further breakdown and enzymatic digestion later in the gastrointestinal tract. This increase in surface area of particles is caused by fracturing and deforming the ingested food by means of shear stresses applied by the teeth. Both the elastic behavior of food materials and their toughness are essential parameters to determine food fracturability. The elastic behavior of a food material, as long as it has an approximately linear stress–strain relationship, can be quantified as the stress–strain gradient, or Young's modulus (Lucas and others 2004). The toughness can be measured with a variety of methods, such as recording the force required to shear a certain material (Friedman and others 1963). Since the key food properties that affect food breakdown can be measured, these properties can be used in a variety of models to better understand the effects of food material properties on their breakdown during chewing. Models have been established for the selection and breakdown of particles during the mastication process, the effect of material properties on food breakdown, and the particle size distribution after chewing.
General breakdown theory
In researching the degradation of coal and other hard materials, Epstein (1947) introduced an asymptotic logarithmic-normal particle size distribution based on the probability of the material to fracture as well as the degree of fragmentation of the particles. Although this model was not originally used for food materials, it has been applied to the mastication process because essentially, the same steps are taking place: a solid material (food or coal) is being broken apart by the crushing and grinding of an outside tool (the teeth or coal crushing apparatus). The fundamental concept to this model is that any fracturing process can be broken down into various discrete steps, each step being a separate “breakage event” or one step in the degradation process. This degradation process can then be described after any finite number of steps.
Selection and breakage functions
Lucas and Luke (1983a,b) used the fundamental breakdown theory of Epstein (1947) to create a more complex model simulating the breakdown of solids during human mastication. They separate the mastication process into 2 distinct processes. The first process is the arrangement of the particles in the mouth so that they can be broken apart (selection). The second process concerns the actual fracture of the particles and their subsequent size distribution (breakage).
The first process, or selection function (S), is the proportion (by either weight or volume) of particles of a finite size range, x to δx, which is broken during one chew. The average selection function can be described as
where P1 is the percentage of particles of average size of after C1 number of chews and P2 is the percentage of these particles remaining after C2 number of chews. If the percentage of particles remaining after a certain number of chews is assumed to be a geometric series with 1 – as the common ratio of the series, P2 can then be described as
From these expressions, the selection function can be derived as defined in Equation 1. This equation assumes that does not vary with the values of C1, C2, and C2−C1 (Lucas and Luke 1983a,b).
The second step of the size reduction process, or the breakage function (B), is the proportion of broken particles that have a size below a given size y (where y > x). By definition, B= 1, when y/x= 1. The distribution of the breakage function can be described by these two equations:
where s and r are constants and y/x is the ratio of the particle size of interest over the initial particle size, and
where a and b are constants and y/x is the same as in Equation 3. These equations to describe the breakage function were derived empirically, using in vivo data from masticated carrot pieces, and were found to be a good representation for the data (Lucas and Luke 1983b). While determining these relationships, it was also noted that the selection did not vary with the number of chews for particle sizes above 4.8 mm, showing that the selection function has little dependence on the particle size. The breakage function took about 10 chews for the distribution to establish itself, as all of the particles were initially one size and needed to start the breakdown process.
The above-mentioned equations were developed using in vivo chewing data to describe the selection and breakage functions (Equations 1, 3, and 4), and then were used to theoretically examine particle breakage over a selected number of chews. To determine the total percentage of particles in each size range, the percentage of particles below size y that is produced during each chew can be calculated as
where xmax is the largest particle size present, y is the particle size of interest, δP(x) is the percentage of the total volume of particles that fall within a size range of x to x+δx, S(x) is the selection function (Equation 1), and B(y,x) is the breakage function (Equations 3 and 4), assuming that all particles have an equal probability of being broken, regardless of size or the number of chews. The use of these models had a good correlation with particle size distributions from in vivo chewing data, showing the possibility of using a simple two-step model to effectively describe particle breakdown during the mastication process.
Voon and others (1986) expanded upon this simple model by adding a power law relationship to the selection function (Equation 1) to further relate it to particle size. Baragar and others (1996) took a slightly different approach, deriving an analytical expression for the measures of the central tendencies of the particle size distribution and degree of particle fracture for both small and large volumes at the beginning of the chewing process. This model was shown to have a good correlation to in vivo chewing data (van der Bilt and others 1987).
Prediction of the particle size distribution after mastication may not be as straightforward as the simple selection and breakage functions (and modifications of these functions). As the particles begin to adhere to each other during bolus formation, their selection function will be changed, resulting in different selection functions for foods with higher or lower cohesive forces. However, these complex interactions have not been extensively modeled at this point. Also, none of the above models have taken into account any of the food properties, such as hardness or fracturability when determining the particle breakdown. In addition, these models have not taken into account the effect of ingesting a mixed solid–liquid meal and how the fluid dynamics of the stomach may later influence the particle breakdown kinetics.
Food Property Influence on Breakage
In an investigation on the influence of food material properties on food breakdown during chewing, Agrawal and others (1997; 1998) proposed the use of several mechanical property indices to relate food mechanical properties to their rate of breakdown. The response of a solid food particle to fracture will depend on 2 key mechanical properties: the fracture stress and Young's modulus E. The fracture stress will be represented by R, the toughness, or the energy required to form a crack in the material. By testing various food products, including cheeses, raw vegetables, and nuts, it was found that if stresses are limiting in a food, then (ER)1/2 is the fragmentation criterion, otherwise (R/E)1/2 is the appropriate index for the fragmentation criterion. These fragmentation criteria were correlated to changes in specific surface area of food products and also to surface electrical activity of the jaw muscles during chewing, showing that they can give an accurate representation of at least part of the role played by material properties in particle breakdown during chewing. As the fragmentation criterion demonstrates, there is a relationship between properties of food materials and their breakdown during chewing. Such relationship may be a useful addition to a model predicting particle breakdown, which has yet to be established.
Bolus Particle Size
The size of food particles in the bolus plays an important role in not only the swallowing and oral processing of the bolus, but also in the further digestion as the bolus reaches the stomach. Numerous studies have shown that different food types of varying physical properties will produce diverse particle size distributions before swallowing. In most in vivo chewing studies, the particle size of a test product is determined by allowing a subject to chew the product for either a certain number of chewing strokes or until it is felt that a swallow is about to be triggered. At this point, the test product is expectorated and the mouth is rinsed. The expectorated test product is then used for analysis. The traditional method of analyzing the particle size distribution after mastication has been sieving, where a set of sieves is used to determine the mass fraction of particles on each sieve. However, in recent years, laser diffraction and image analysis, where particle size is measured using a computer system to calculate the size and shape of particles from an image, are becoming increasingly powerful methods for particle size determination (Yurkstas and Manly 1950; Jiffry 1981; Lucas and Luke 1986; Prinz and Lucas 1995; Hoebler and others 1998; Fontijn-Tekamp and others 2004; Peyron and others 2004; Mishellany and others 2006; Jalabert-Malbos and others 2007).
Median particle size values
One of the key values used to represent the particle size distribution is the median particle size (d50) of the distribution. This value represents the aperture of a theoretical sieve through which 50% of the weight of the particles could pass. The d50 is also commonly used to represent masticatory performance and swallowing threshold (Fontijn-Tekamp and others 2004).
Particle size analysis and key factors affecting particle size
As can be seen in Table 1, different food types produce distinct particle sizes upon swallowing. Although physiological factors (age, gender, dental status, and so on) play a role in particle size manufacture, this interindividual variability has been shown to be quite low in comparison with the variability over different food types of varying structure and texture (Peyron and others 2004).
Table 1–. Values reported in the literature of median particle diameters from in vivo studies. All particle size distributions were determined by sieving methods.
By using both laser diffraction and sieving, Peyron and others (2004) showed that although the overall distribution shape was similar between nuts (peanuts, almonds, and pistachios) and vegetables (cauliflower, radishes, and carrots), particle sizes were much larger for all types of vegetables when compared to nuts. The differences in particle size were hypothesized to be caused by variation in bolus cohesion and plasticity between the vegetables and nuts. Another study by Mishellany and others (2006) used the same test foods of nuts and vegetables, but used image analysis to quantify the particle size as well as the shape index ((perimeter2/(4π·area), corresponding to the circularity of the particles, with a value of 1 being a perfect circle) of the particles. They also showed that nuts and vegetables resulted in distinct particle size distributions. Nuts resulted in a bolus having many particles smaller than 2 mm, and vegetables resulted in a bolus with more particles greater than 2 mm. The particle shape, as quantified by the shape index from the image analysis, was also influenced by the food type, with the greatest differences seen between the nut group and vegetable group instead of between individual food products. These particle size results were also observed with a very low interindividual variation, suggesting that most individuals will reduce food particles to a similar size before swallowing by different means, regardless of chewing efficiency.
A study by Lucas and Luke (1986) that examined the particle size distribution of carrots and Brazil nuts, with a varying number of chewing strokes, showed that Brazil nuts consistently broke down faster than carrots, although both products were swallowed after a similar number of chews (Brazil nuts had a smaller particle size upon swallowing). This study also indicated that Brazil nuts did form a cohesive bolus, while carrots did not. The differences in particle size were not fully explained, but hypothesized to be due to a combination of the overall state of the bolus (cohesive or not), the lubrication of the particles (amount of saliva and initial moisture content), and also could be due to differences in the amount of particles being swallowed in “intermediate swallows” (particles being swallowed in small quantities involuntarily).
In a related study, Prinz and Lucas (1995) fed test subjects Brazil nut particles of 4 different size categories suspended in plain yogurt in various concentrations to determine the number of chewing strokes and time needed to swallow each mixture. This study showed that both the particle size and the concentration of particles affected the number of chewing strokes before swallowing and the time needed for chewing. Interestingly, these results fit with the model that a certain particle size and lubrication must be met before swallowing can take place. This was shown by quantifying the “chewing frequency” or the number of chews/total chewing time. The chewing frequency was constant at concentrations of >20% Brazil nuts, but rose sharply with concentrations of <20%, showing that for low concentrations of Brazil nuts, the subjects took a number of “exploratory chews” in a very short time before swallowing the mixture. When combined with the data from various sizes of nut particles, they showed that particles of <2.0 mm in size are small enough to be swallowed, but if the concentration is greater than 20%, they still must be broken down to provide sufficient lubrication before swallowing can occur.
Jalabert-Malbos and others (2007) conducted a study examining the particle size distributions of 10 natural foods (for d50 values, see Table 1) with varying structural/textural properties. They concluded that, for optimal swallowing particles, foods must be broken down to a size less than 2 mm, unless they are very soft (such as boiled egg whites or cottage cheese) and will not injure the upper digestive mucosa, supporting the theory of a particle size threshold coupled with a factor for the food consistency/texture.
In a study designed to determine if the particle size distribution at swallowing would change based on the amount of food consumed, Lucas and Luke (1984) tested varying amounts of food (1, 5, 8, and 12 g size “mouthfuls”). They found that smaller particles were swallowed from a 1 g mouthful, but mean particle size did not differ between 5, 8, and 12 g mouthfuls. They suggested that since the salivary secretion rate does not increase linearly with the mass of the material being chewed, smaller mouthfuls might be subject to a saliva threshold, while larger mouthfuls might be subject to a size threshold before swallowing.
Hoebler and others (1998; 2000) examined the particle size distribution of bread and pasta after chewing. Bread particles could be analyzed using laser diffraction, due to the small size (<1000 μm) and general spherical shape. Due to their larger size (<1000 μm length) and nonhomogeneous shape, pasta particles had to be analyzed using image analysis to quantify the particle area. Their results showed that bread loses much of its initial structure during mastication and forms a bimodal particle size distribution with peaks at 30 and 500–620 μm. However, pasta was shown to break down into a unimodal distribution, with sizes ranging from 0.5–60 mm2 and peak values around 15–20 mm2. These differences in particle size were attributed to the initial food structure and texture.
Modeling particle size distribution
A commonly used equation to describe the particle size distribution after the mastication process is the Rosin–Rammler equation (Rosin and Rammler 1933):
where is the fraction of particles with a size smaller than X, X50 is the median particle size, and b is a variable representing the spread of the distribution. This distribution was fit to in vivo chewing data in a variety of studies (Olthoff and others 1984, 1986; van der Bilt and others 1987, 1993; van der Glas and others 1987) using an artificial test product, Optosil, as the chewing medium. This model may be appropriate for certain food products, but since a strong correlation has been shown between food hardness and subsequent breakdown during chewing, perhaps, a more suitable model should include an index to account for the food hardness.
Bolus texture may be one method to quantify both the saliva–food interactions in the bolus as well as the cohesive forces within the particles, as described above by Prinz and Lucas (1997). However, bolus texture is a property that has not been well investigated. A few studies, performed on boluses of cooked meat, sought to quantify the bolus texture and salivary secretion (Mioche and others 2002; 2003; Yven and others 2005). The studies used boluses of meat with “tender” texture and “tough” texture, varied by using different cooking processes. Individuals were instructed to chew the pieces of cooked meat for varying time intervals, and the subsequent meat boluses were analyzed for a variety of properties, including salivary impregnation and shear stress. The shear stress, an indication of the bolus texture, was measured by recording the maximum shear stress acquired by using a double-bladed shear cell on an Instron Universal Testing Machine with a displacement of 60 mm/min. These studies found that by the end of the chewing cycle, the meat with the tougher texture required more chewing strokes and incorporated more saliva than the more tender pieces of meat. However, this difference was not seen after only 7 s of chewing, when both boluses had the same amount of saliva incorporated. The shear stress of the boluses decreased over the chewing time for both meat textures. Before chewing and in the middle of the chewing cycle, the more tender meat had a lower shear stress. Before swallowing, 2 of the studies (Mioche and others 2002, 2003) showed a significant difference in the end shear stress values between the boluses of different texture. The third study showed a slight, but not significant, difference between the shear stress values from the 2 bolus initial texture values (Yven and others 2005). These studies demonstrate the importance of initial food texture in both the chewing cycle properties (duration, saliva secreted, and so on) and also in the texture of the final bolus. This is consistent with many of the fundamental chewing studies that have shown that the food material properties play an important role in their breakdown during oral digestion. Since these material properties influence the final bolus properties, they will also play a role in the digestion processes in the stomach and the rest of the gastrointestinal tract.
Given that the food material properties and initial texture have been shown to play a role in determining the bolus texture upon swallowing (Mioche and others 2002, 2003), measurements of the initial texture of food could prove to be important in the overall digestion process. Food texture plays a role in the perceived taste of the food (Izutsu and Wani 1985), which could lead to a varying degree of salivary secretion (Mese and Matsuo 2007). The sensory aspect of food texture, as perceived by the individual, has been correlated to laboratory compression measurements in many food products, such as nuts, chocolate, cheese, vegetables (Boyd and Sherman 1975), white pan bread (Gambaro and others 2002), and rye and French bread slices (Brady and Mayer 1985). As the sensory perception of food is important to the food oral processing (Wilkinson and others 2000), food texture is a variable that should be measured when considering the oral digestion of a food.
In vitro Chewing Models
Although many attempts have been made to model the particle breakdown during the chewing process, few have looked at creating an in vitro system to simulate this process due to its complex nature and the number of factors that vary considerably between individuals. The in vitro systems that have been determined are only suitable for a few food types, and research is lacking in a reliable in vitro chewing procedure that is appropriate for a wide variety of foods.
An in vitro system, designed by Olthoff and others (1986), involved compressing a cylindrical sample of Optosil (a tasteless, silicon–rubber artificial test food material (Sierpinska and others 2008)), turnip, carrot, Gouda cheese, and peanuts to 40% of the original sample size using a probe with a varying cusp angle of 90° (Figure 1). These compression tests were repeated 100 times in order to obtain a particle size distribution of the fragmented pieces of each product. The in vitro tests showed that the various food types presented distinct particle size distributions after chewing. However, the study was more concerned with the force applied during chewing as opposed to the particle size distribution. Results with Optosil showed considerable differences between the particle size, as modeled by the Rosin–Rammler equation, from the in vitro model and similar in vivo chewing tests.
Artificial masticatory advanced machine (AM2)
Another in vitro mastication simulator, the Artificial Masticatory Advanced Machine (AM2) was developed by Woda and others (2010) that attempted to reproduce the basic jaw movements of translation and rotation in a mechanical model. To do this, they used 2 identical disks that are compressed and slid against each other to recreate the action of chewing. The masticatory force was altered by the strength of the spring behind one of the disks. This model is versatile, as it allows for many of the chewing parameters to be varied, such as the masticatory force, the chewing time, the number of chewing strokes, and the amplitude of the chewing strokes. The model also allows for saliva addition (with a user-defined flow rate, depending on the food product), and temperature control. The AM2 was shown to accurately reproduce the particle size distributions determined in vivo for peanuts and carrots after 10, 20, and 30 chewing strokes by altering the spring strength of the apparatus (masticatory force). The AM2 seems to be a promising in vitro mastication simulator due to the ease of chewing parameter adjustment. However, a key disadvantage is that the spring strength needed to correctly reproduce in vivo particle breakdown must be correlated for each food being chewed, or at least a representative food. This need for calibration using an in vivo study may take away some of the utility of the simulator. The AM2 has only been tested with peanuts and carrots, both of which are quite hard, and there should be more testing done with the reproducibility of in vivo mastication for a wider range of food products.
A study by Hoebler and others (2000) examined the particle size distribution of starch-based products after both in vitro and in vivo chewing experiments. For the in vitro experiments, they used either fresh bread or pasta (spaghetti and tortiglioni). The foods were ground using a meat grinder outfitted with a 6-mm-hole plate. These “chewed” particles were then analyzed by using both laser diffraction (for bread particles <1 mm in size), and image analysis (for nonspherical pasta particles), and were compared with bread and pasta chewed by human subjects for an average chewing time. Their study of the particle size distribution showed an acceptable correlation between the in vitro and in vivo chewing of starchy products. This shows that a simple mincing procedure using a meat grinder can be used to simulate mastication for a variety of starch-based foods of varying textures.
Swallowing is the mechanism through which ingested food is transported from the oral cavity to the stomach in order to undergo further digestion. Although the mechanism of swallowing is clearly defined, the factors contributing to a swallow being triggered are under debate, with various theories present as to the driving force behind swallowing. In healthy individuals, swallowing occurs about 600 times each day and about 6 times/h during sleeping periods (Pedersen and others 2002). Swallowing is a key step in the beginning of the digestion process. It is important to consider under what conditions a swallow will be triggered, as this will influence what the food properties will be of the foods that are swallowed, and hence, the properties of the foods as they begin gastric digestion in the stomach.
Mechanism of swallowing
There are 3 phases in the process of swallowing consisting of 1 voluntary phase and 2 involuntary phases. The first phase, which is voluntary, is the oral phase; it involves moving the masticated food particles and saliva toward the back of the oral cavity using the tongue to “mold” the food–saliva mixture into a bolus. During this phase, the pressure generated by the tongue pushing against the back of the oral cavity can rise to a level of 10 kPa. The second phase, known as the pharyngeal phase, is an involuntary triggering of the swallowing reflex to open and close the upper esophageal sphincter that lasts about 0.7 s. During this phase, the bolus is pushed into the esophagus by a peristaltic wave, where pressures seen by the bolus can be up to 4 kPa in the pharynx. The third, or esophageal phase, also involuntary, involves successive peristaltic contractions of the esophagus in order to convey the bolus to the stomach (Pedersen and others 2002; Chen 2009).
Swallowing threshold theory 1: Degree of structure and lubrication
Hutchings and Lilliford (1988) suggested that a certain “degree of structure” (particle size) and a certain “degree of lubrication” (amount of saliva), as well as a certain “time” (time in the mouth), must be reached before a swallow can take place. The “degree of structure” factor accounts for the size of the food particles as they are broken down during the mastication process. The “degree of lubrication” is really more of the “perceived lubrication” of a food product in the mouth. This takes into account the combination of various factors, such as the moisture (saliva) initially present in the mouth, free liquids liberated from the food product, saliva released during mastication, as well as other factors, such as the presence of fat. The “time” factor takes into account the fact that the breakdown of food products during mastication is a progressive process with time being an obvious factor; the time needed for the breakdown of foods with varying properties has a wide range.
This theory suggests that a food product such as a juicy steak may have sufficient lubrication upon entering the mouth; however, the “degree of structure” must be significantly reduced before it can be swallowed. In contrast, for a food such as a dry piece of cake, although the structure will be quickly broken down, it will need more time in the mouth to sufficiently reduce the “degree of lubrication” before it can be swallowed.
Swallowing threshold theory 2: Optimal cohesive forces
An alternate theory suggests that as the food particles are reduced in size, they will be coated with saliva and begin to cohere to each other as the bolus begins to form. A swallow will be triggered when the cohesive forces in the bolus reach a maximum, optimizing the transport into the esophagus and minimizing the risk of aspirating food particles (Prinz and Lucas 1997). The mathematical description of this theory assumes that each of the food particles produced during mastication is spherical. Upon breaking, each particle will become coated with saliva, allowing it to either stick to the oral mucosa or to other particles. The force at which a food sticks to the walls of the oral cavity can be modeled as
where r is the radius of the food particle in meters and λ is the surface tension of the oral fluid in N/m (Bowden and Taylor 1950). The surface tension of saliva is, on average, 0.053 N/m (Glantz 1970). The attractive and cohesive forces between particles are more complex, but can still be modeled. If the food bolus is considered as a spherical ball of particles with 2 disc-like surfaces on each side, the attractive forces between the particles can be modeled by the viscous forces required to separate the 2 “discs” of the bolus as
where η is the viscosity of the saliva between the food particles in Pa·s, R is the radius of the “disc” of particles in meters, t is the time span over which the particles are separated in seconds, and d is the average distance between particles in meters (Cottrell 1964). Prediction of the particle size distribution has already been discussed (Modeling of Particle Breakdown during Mastication: Selection and Breakage Functions). Food particles will agglomerate and begin to form a bolus when FV > FA, or FV–FA > 0. The maximum FV−FA value is the point at which a swallow is triggered.
This theory was tested on sample food products, where in vivo mastication data were available regarding the particle size distribution and chewing parameters. The foods chosen were raw carrots and Brazil nuts, 2 foods with different breakdown rates and moisture contents, both of which will not absorb saliva during chewing. The results of the simulation showed that, initially, FV−FA is negative, implying that the particles do not cohere to each other and are more likely to stick to the walls of the oral cavity. Then FV−FA becomes positive, meaning that the particles begin to agglomerate together, and there is a rapid increase in cohesive force until the peak is reached, after which there is a slow decline as the particles begin to fall apart (Figure 2). In both carrots and Brazil nuts, the peak cohesive force occurs after 20–25 chews; however, this cohesive force is different for each food. The peak force for Brazil nuts is about 0.02 N, while the peak force for carrots is about 0.005 N. These results agree with the observations that carrots do not tend to form an exceptionally cohesive bolus upon swallowing, as the cohesive forces between particles are low. The number of chews to reach the peak cohesive force also agrees with the number of chews needed to trigger a swallow found in in vivo mastication studies (Lucas and Luke 1986).
The cohesive forces model can demonstrate the effect of other factors, such as salivary flow rate, salivary viscosity, and food particle size distribution on bolus formation. An increase in salivary flow rate will “flood” the bolus, causing it to break apart sooner, resulting in a lower peak cohesive force after fewer chews. In the foods used to demonstrate this model, both were chosen as to not absorb saliva during oral digestion. When foods absorb saliva (such as cereal-based systems or foods with lower moisture and fat content), the salivary flow rate will effectively be reduced, again changing the cohesive forces in the bolus. When the saliva viscosity is affected by certain food components, the bolus formation will also be altered. For example, tannins have been shown to effectively reduce the viscosity of saliva (Prinz and Lucas 2000), which will decrease the cohesive forces in the bolus and delay swallowing. However, upon an increase of saliva viscosity, cohesive forces may be increased, promoting earlier swallowing. The size distribution of the food particles will also affect bolus formation.
The transit time through the esophagus to the stomach has been shown to vary depending on the physical composition of the product swallowed as well as the physical conditions of the subject (supine or sitting). Although boluses of water will completely leave the esophagus in only one swallow, having a transit time of a few seconds, other foods may take much longer. It was shown that a gelatin capsule will pass through the esophagus in about 15 s, while a liver cube will remain in the esophagus up to 120 s. These times are increased when the subject is supine as compared to sitting; standing shows a faster passage of particles through the esophagus (Fisher and others 1982). It has also been shown that capsule shape plays a role in esophageal transit, with oval-shaped capsules having a shorter transit time than round tablets (Channer and Virjee 1986).
As boluses enter the stomach they will “stack up” in the curvature of the stomach according to the time they were ingested. Layers will begin to form according to density and solids content, with large particles settling in the antrum (Figure 3), watery substances in a layer above, and less dense high fat layers at the top (Schulze 2006). The stomach consists of 2 regions with distinct functions: the proximal and the distal region. The proximal region includes the fundus and about one-third of the stomach body (Figure 3). It plays a major role in the gastric emptying of liquids and acts as a reservoir for undigested materials. During swallowing, the proximal stomach relaxes, creating space for the swallowed boluses in the stomach (Sernka and Jacobson 1979). The distal region includes the remaining two-thirds of the stomach body, the antrum, and the pylorus (Figure 3), acting as a mixer, grinder, sieve, and pump to reduce solid particles in size before they can pass through the pyloric sphincter into the small intestines (Kelly 1980; Schulze 2006). Food will be broken down both physically and chemically, by the movement of the stomach walls (peristalsis) and from the acidic and enzymatic conditions in the stomach.
The mechanical size reduction of ingested solid particles increases the surface area for enzymatic attack, facilitating chemical digestion by the enzymes found in the gastric secretions. The mechanical size reduction is enhanced by the contractions of the stomach walls, known as peristalsis. Upon ingestion of a mixed meal, the stomach goes from fasted state (at rest) to a fed state, at which peristalsis occurs until the meal is emptied from the stomach. During the fed state, there are 2–3 peristaltic contractions occurring simultaneously, moving from the fundus to the pylorus, with duration of about 60 s each. These contractions generate 2 types of flow patterns; flow vortices (eddies) that circulate particles between successive waves, causing radial mixing and carrying gastric secretions throughout the stomach, and the retropulsive jet, caused by the terminal antral contraction. The retropulsive jet is essential in mixing and emulsifying the solid and liquid phases of the ingested meal. The solid particles are rubbed and ground against each other and the walls of the stomach, then forcefully retropelled from the pylorus toward the body, reaching velocities up to 13 mm/s (Schulze-Delrieu and others 1996; Schulze 2006). Although these flow phenomena have been clearly established, their importance in relation to solid particle breakdown during gastric digestion remains to be determined.
The chemical breakdown of foods will occur due to both the acidic and enzymatic conditions found in the stomach. In the human stomach, the proximal region (Figure 3) of the stomach is composed primarily of parietal (or oxyntic) cells that are responsible for secreting the gastric acid and pepsinogen. The composition of the gastric secretions from the parietal cells is correlated to the total amount of secretions (acid and salt). A low secretory rate will result in a higher concentration of salts (Na+ and K+) and a lower concentration of HCl, and a high secretory rate will result in a lower concentration of salts and a higher concentration of HCl. The distal region is composed primarily of chief cells that are responsible for secreting mucins and the hormone gastrin, and neck cells that produce mucus and proteases (Johnson 1991). The average fasting pH in the stomach is around 2, but will increase up to 5–6 after the consumption of a meal. The average pH may take between 2–4 h after meal consumption before it returns to its fasting level of 2 (Malagelada and others 1979). Gastric acid secretion begins at birth, with babies even a few days old already having a pH of about 2.0 in the stomach (Mason 1962). The gastric acid will act to hydrolyze both starches and proteins while they are in the stomach. The stomach also contains gastric lipases, to digest fats, and pepsin, to digest proteins. The gastric environment is a complex combination of physical and chemical factors that all must be taken into consideration when studying the gastric digestion of a food or developing a gastric in vitro model.
Numerical modeling of gastric fluid dynamics
The physiological complexity of the stomach has prevented a thorough understanding of the mechanical conditions and fluid dynamics that occur during gastric digestion. However, computational fluid dynamics (CFD) have been used to develop a three-dimensional model of the fluid movement and stomach motility during gastric digestion (Ferrua and Singh 2010). This model has been used to predict the velocity and pressure fields in the stomach after consumption of meals with different viscosities. Results have shown that by increasing the meal viscosity, from 1 cP (approximately the viscosity of water) to 1000 cP (similar to the viscosity of honey), the fluid movement throughout the stomach decreases and the pressure field increases (Figure 4). This work also indicates the presence of previously observed retropulsive jet, as well as flow recirculations that occur in fluids with low viscosities. The computational results obtained by this study were shown to be in good agreement with previously published experimental data (Ferrua and Singh 2010). Computational models that study the fluid dynamic conditions in the stomach can also provide a unique insight to the gastric conditions that will affect the breakdown of food particles during digestion.
Gastric emptying and disintegration models
Various models have been proposed to represent the rate of gastric emptying. One such model is a modified power exponential model, which accounts for the exponential decrease seen in the emptying of liquids, but also accounts for a lag phase due to the slower emptying of solids:
where y(t) is the percentage of meal retained at time t, k is the gastric emptying rate per minute, and b is the extrapolated y-intercept from the terminal part of the curve (Siegel and others 1988). This model has been shown to provide a good fit for the disintegration of individual particles of raw carrots, cooked carrots, and ham at different forces during in vitro digestion (Kong and Singh 2008b).
Another model used by Kong and Singh (2009) to better describe the lag phase seen in the disintegration of various food products is a linear exponential model. This model was adapted from a model originally used by Goetze and others (2005) to describe changes in stomach volume:
where y(t) is the mass retention of the food product at time t, k is a dimensionless constant representing the lag phase, and β (1/min) is a constant representing the curve concavity. This model has been applied to the disintegration of various food products, including ham, fried dough, carrots, beef jerky, almonds, and peanuts (Kong and Singh 2009). This model has also been used by Bornhorst and others (2011) to describe the disintegration of 6 bread boluses during in vitro gastric digestion. These models have all been shown to provide a suitable description of the gastric breakdown and emptying of a wide variety of food products, but could be modified to include other factors that will play a role in the gastric breakdown, such as the solid–liquid fractions of the meal and the initial food properties (hardness, moisture content, or fat content).
Blood Glucose Absorption
After the ingested food is physically broken down in the stomach, it passes through the pylorus and into the small intestine where further chemical breakdown occurs and many of the food compounds are absorbed into the blood. These food components, such as sugars, fatty acids, amino acids, and low-molecular-weight substances, are moved from the lumen of the small intestine across the intestinal membrane lining into the mucosal cells and then into the blood (Sernka and Jacobson 1979). It has been shown that all of the ingested glucose will be absorbed in the small intestine, the majority in the proximal half of the small intestine (Borgström and others 1957). Glucose is absorbed in the intestine via two mechanisms: active absorption by a sodium/glucose transport protein, and by a diffusive process, mediated by a glucose transporter protein (GLUT2) (Kellett and others 2008; Gibney and others 2009). The rate of glucose absorption into the blood is an indication of the extent to which the food has been digested. The factors affecting the rate of glucose absorption include those related both to the food and the consumer of the food. Food-related factors include food particle size, structure, cell wall integrity, amylose–amylopectin content of starch, and lipid content. Consumer-related factors include the extent of chewing (particle size at the point of swallow), gastric emptying rate and gastric conditions, and small intestine transit time (Gibney and others 2009). Specific metabolism of glucose is mediated by many different enzymes; however, details of these processes are out of the context of this review.
Blood glucose can easily be measured, as done by diabetic patients several times per day, using a capillary finger prick test where a small amount of blood is taken from the fingertip and analyzed for glucose content. This measurement is repeated over a number of hours to compare the rise in glucose after consumption of a certain test meal. The glycemic response in capillary blood is larger than that from venous blood or plasma, so the capillary finger prick test is able to detect smaller changes in glucose than other sampling methods (Wolever and others 1991).
Along with the glucose response, the plasma insulin concentration may also be measured. Insulin is a hormone secreted by the pancreas that regulates plasma glucose levels and stimulates glucose metabolism (Reaven 2005; Giugliano and others 2008). For normal subjects, the glucose and insulin responses are normally similar in relative magnitude. However, in diabetic subjects, the insulin response is reduced. Consequently, the glucose is not effectively utilized, resulting in higher blood glucose concentrations, both in a fasting state and after a meal is consumed (Cerasi and Luft 1967).
To easily compare glucose responses for a variety of different foods or test meals, the concept of glycemic index (GI) was developed. The GI is expressed as the area under the curve of the glucose response in comparison with the area under the curve (from the same subject) for a standard test food, normally glucose or white bread. The area under the curve is only calculated for the part of the response that is above the baseline glucose levels. Any drops of the blood glucose below the baseline value will not be included in the calculation of the GI value (Wolever and others 1991). The GI value will increase with meal carbohydrate content until reaching a plateau after a certain amount of carbohydrate is consumed, so most tests are performed with a standard 50 g total carbohydrates (Jenkins and others 1981). The GI has become a widely accepted measure that gives insight into the physiological effects of, especially, carbohydrate-rich foods. The GI demonstrates differences depending on food structure, starch type, moisture content, and other related factors (Read and others 1986; Foster-Powell and others 2002). International tables of GI values have been established for >150 different food products, and GI has been shown to be a better predictor of in vivo digestion rate than in vitro laboratory tests, as it accounts for other physiological factors, such as gastric emptying (Foster-Powell and others 2002).
Gastric emptying and blood glucose correlation
Many nutrition-based studies have examined the glycemic response of foods, particularly cereals and starch-based food products, as they are the main source of carbohydrates in the Western diet. Studies have also been conducted to relate a food's glycemic response to its gastric emptying rate. In general, these studies show an inverse correlation between the glycemic response and the gastric emptying rate.
In a study examining 4 starchy foods of varying structure and composition (spaghetti, rice, French bread, and mashed potatoes), Mourot and others (1988) found that there was a significant negative correlation between the gastric emptying half-time (determined by scintigraphy) and the maximum variation in plasma blood glucose. They hypothesized that the particle size of the ingested foods could have caused the variation in gastric emptying rate. Foods that already had a small particle size upon ingestion (mashed potatoes) required the shortest gastric emptying time and had the highest glycemic response as they did not require much digestion. In contrast, more structured foods, such as rice and pasta, which need to be considerably broken down before they can empty from the stomach showed a longer gastric emptying time and a lower glycemic response.
Similarly, Hlebowicz and others (2007) showed an inverse relationship between postprandial blood glucose and gastric emptying rate after consumption of rice pudding with and without added cinnamon. The addition of 6 g of cinnamon to the rice pudding was hypothesized to slow the gastric emptying, which, in turn, mediated the delivery of the glucose to the blood, thus reducing the glycemic response. The exact mechanisms by which cinnamon decreases gastric emptying rate are not clear, although other natural essentials oils, such as oregano, fennel, and ginger have been shown to lower blood glucose concentrations in rats (Talpur and others 2005), suggesting that natural spices may have some type of slowing effect on gastric emptying.
Other studies have also reported this inverse correlation between postprandial blood glucose and gastric emptying rate with pure glucose solutions (Horowitz and others 1993), as well as glucose solutions with added pectin and guar gum (Holt and others 1979). These studies suggest that the gastric emptying rate of the ingested food product is the key factor in controlling the absorption of glucose into the blood.
Two studies that have examined the gastric emptying rate and glycemic response of various bread products found no difference in the gastric emptying of whole meal rye bread and white wheat bread (Hlebowicz and others 2009), as well as white wheat bread, whole meal rye bread baked with 60% rye kernels, whole meal rye bread with added oat β-glucan (to increase fiber content), and dark durum wheat pasta (Juntunen and others 2002). However, these studies also showed no difference in the postprandial blood glucose response of the different products, which again demonstrates the correlation seen between gastric emptying and glucose response of a variety of food products.
Glycemic response to various cereal products
Numerous studies have been performed regarding the glycemic and insulemic responses of cereal-based products. Sample GI values can be found in Table 2. Although some of the specific responses and their mechanisms are not fully understood, a general conclusion from most studies is that the structural characteristics and the availability of starch in the bread or grain are key factors controlling the digestion of carbohydrate food products.
The effect of food structure on the gastric emptying of bread and flour products has been widely studied in a variety of different cereal grains. These studies have examined how the bread and grain macrostructure (addition of whole grains to bread products), and microstructure (protein–starch interactions in bread) have an effect on the carbohydrate digestion.
To elucidate some of the mechanisms of starch digestion in bread and to examine the potential protective effect of a protein–starch interaction, Jenkins and others (1987) examined the blood glucose response to white bread made with regular white wheat flour, gluten-free white wheat flour, and gluten-free flour with added gluten (after the initial removal process). They showed that the bread made from regular white wheat flour had a lower postprandial glucose response than both the gluten-free flour bread and also the gluten-free flour bread with added gluten. Both of the gluten-free flour breads showed a similar glucose response. They attributed this response to the starch–protein interaction found in wheat flour. The starch core in the flour is surrounded by a protein network, which may inhibit the rate of starch hydrolysis. When this protein network is disrupted by the removal of the gluten, the starch in the wheat flour is more available for hydrolysis and undergoes a faster digestion. Even if the gluten is added in after its initial removal, the natural structure that has been destroyed does not form again. This finding affirms the importance of particle structure in relation to starch hydrolysis and digestion.
Structural properties, as well as processing conditions, have been shown by numerous studies to play a major role in the digestibility of wheat, rye, and barley products. Jenkins and others (1986) examined the GI and blood glucose response after the consumption of white, wheat, rye, and pumpernickel (rye bread with 80% rye kernels) breads, bulgur (parboiled cracked wheat grains), wheat kernels, and rye kernels. They found that consumption of bulgur resulted in only a slight increase in the blood glucose when compared to the whole wheat kernels (not processed). This difference was attributed to be due to the processing conditions of the bulgur: the whole wheat grains are cooked with the endosperm prior to drying and milling, resulting in a physical restriction to water uptake that causes a reduction in the starch gelatinization, reducing the time of starch digestion. They also found that using 80% whole grains to make the pumpernickel bread reduced the blood glucose in comparison with the breads made without kernels, while the glucose response for whole meal rye bread was similar to that of whole meal wheat bread.
Similarly, in another study by Jenkins and others (1988), by comparing whole meal wheat and barley flour bread with increasing ratios of whole grains added, as well as barley kernels and bulgur wheat, it was found that the glycemic responses of each of the respective bread types decreased with increasing proportions of whole kernels added to the bread. This decrease in glucose response with an increase in kernels was attributed to the effect of food form/structure on the starch digestibility, and possibly also to an alteration in gastric emptying.
In another study involving digestion of grains with varying structures, Granfeldt and others (1994) examined the postprandial blood glucose concentration after consumption of both boiled intact barley kernels and barley flour from barley with varying amylase concentrations in comparison with white wheat bread. They found that all of the barley products had a lower glucose response than white wheat bread during the first 70 min of testing. The intact kernels from all barley types showed a significantly lower glucose response than their corresponding boiled flours. The amylose content of the kernels and flours did not affect the glucose response. They also measured the starch hydrolysis and showed a significant correlation between the amount of starch hydrolysis and the GI values. This reaffirmed the conclusion that increasing the starch availability for enzymatic hydrolysis by changing the inherent grain structure (in milling or other processing) is a key factor in starch in vivo digestion. This study also showed that boiled high-amylose starch produced much higher satiety scores than other flour products, hypothesized to be caused by clumping in the stomach as well as a reduced gastric emptying rate.
Effect of Processing
Food processing and food structure are inherently linked, as the food structure may change during processing. Refined and whole grains have been studied at the length to determine the differences (if any) between their digestive characteristics, as they represent a more processed (refined, or white flour) and a less processed (whole wheat, or simply wheat flour) version of the same product. The observations made on the effect of processing may be due to physical changes caused by the food processing (change in structure due to heating), or by nutritional changes caused as a result of the processing (removal of fiber during wheat milling).
Grimes and Goddard (1977), while studying the gastric emptying of white bread and whole wheat bread, found that the solid portion of the meal emptied at the same rate for both refined white and whole wheat flour. However, the liquid portion of the white bread meal emptied much faster than whole wheat bread. They hypothesized that the differences were due to a stimulation of gastric motility by the white bread and perhaps an increase in viscosity of the liquid component of the wheat bread gastric contents, slowing the liquid emptying.
Also studying the differences between refined wheat and whole grain wheat (not refined) flour products, Kristensen and others (2010) examined the differences between blood glucose and GI of refined and whole grain wheat bread and pasta. They found that the glycemic response between the whole grain and refined products (both bread and pasta) showed no significant difference during the 3 h test period. However, the bread meals showed a higher glucose response than the pasta products. The differences seen between bread and pasta were hypothesized to be due to structural differences between the bread and pasta as a consequence of processing. Also, due to its lower initial moisture content, bread requires a larger degree of mastication before swallowing, so it will have a larger incorporation of saliva and α-amylase. Since the glycemic response between the groups of refined and whole grain products was not significantly different, the authors concluded that in addition to fiber content, other factors play a role in the overall digestion, such as energy density, volume, botanical structure, and particle size.
Various investigations have been done on the glycemic response of rye bread, as it is an essential part of the daily diet in many north European countries. One such study examined the gastric emptying and blood glucose response to commercial white wheat and whole meal rye breads. They did not find a significant difference in either the gastric emptying or the blood glucose response between the bread types (Hlebowicz and others 2009). Similarly, 2 studies from Finland (Leinonen and others 1999; Juntunen and others 2003) examining the glucose and insulin responses of human subjects to wheat, rye, and whole kernel rye bread also showed no significant difference in the glucose response of the various bread products. However, the insulin responses of the rye breads were significantly less than those for the control white wheat bread. Their results indicate that neither the fiber amount nor the amount of whole kernels in the breads had an impact on the glucose response. However, they also indicated that the whole rye kernels may have been disrupted during the bread baking processes, leaving them with an increased sensitivity to enzymatic attack. The reduced insulin response was attributed to the differences in structure between the wheat and rye breads. Light microscopy revealed that the rye bread had a less porous, firmer structure with a continuous phase formed by densely packed starch granules. This is where the amylase had leached out during baking, forming a hydrolysis-resistant coat around the granule. On the other hand, the wheat bread also had starch granules trapped in a broad gluten network, where the starch remained inside the granule and became gelatinized, making it more available for hydrolysis. Overall, these structural differences, and not the fiber content of the breads, were used to explain the differences in insulin responses seen between wheat and rye bread, further stressing the importance of food structure as a controlling factor for its breakdown during digestion.
Further supplementing the evidence that food structure and processing is a key factor in the digestion of starchy foods is a study by Brand and others (1985) that examined the GI and starch digestion of both conventionally cooked and processed versions of rice, corn, and potato. They found that the processed forms of all food products showed a greater digestion of starch than the conventional forms of the respective products. The GI of all processed food was greater than that of the conventionally cooked product, except in the case of potato crisps, which had a GI similar to boiled potatoes. They attributed their results to the fact that many of the processed products underwent a processing operation, such as extrusion puffing. During this process, the starch in the food product was hydrated and gelatinized, and the structure of the grain was disrupted because of the high temperatures and pressures. On the contrary, conventional cooking processes (such as boiling) involve much less physical disruption of the inherent granule structure as well as less harsh temperature and pressure conditions, leaving the starch granules less vulnerable to hydrolysis.
In Vitro Starch Digestion
Starch is one of the crucial factors affecting the digestion and absorption of dietary carbohydrates; however, the state of the botanical grain structure, physical food texture, hydration of the starch granule, as well as the starch granule chemical structure all affect the specific interaction of starch during digestion. Many of these factors have been investigated using in vitro starch digestion studies to further elucidate the driving mechanisms of starch digestion in carbohydrate food products (Bjorck and others 1994).
Starch makes up about 80%–90% of the polysaccharides consumed in the normal human diet. It is a mixture of the α-glucan polysaccharides amylose and amylopectin. The ratio of amylose and amylopectin depends on the botanical origin of the starch grain and varies in different types of starch. These polysaccharides are broken down by the amylolytic enzymes of the digestive tract at their α-glycosidic bonds. However, the availability of these bonds to hydrolysis by α-amylase depends upon the physical form of the starch molecules. Starch can be classified into 3 categories: rapidly digestible starch (RDS), or starch digested within 20 min of enzymatic hydrolysis; slowly digestible starch (SDS), or starch digested between 20–120 min of enzymatic hydrolysis (completely digested in the small intestine); and resistant starch (RS), or starch products not digested in the small intestine that will pass into the large intestine undigested. Resistant starch can be initially present in a food product, such as the resistant-type starch granules found in bananas and raw potatoes, or physically inaccessible starch, due to being enclosed in a food structure such as an intact cell or granule, or the starch can become resistant during a processing step, such as cooking, when starch is gelatinized and then retrogrades to a form that is not easily hydrolyzed by α-amylase (Englyst and Hudson 1996; Elmstahl 2002). When whole grains, such as wheat or maize (corn), are ground into fine flours, they show a larger rate of starch digestion than the whole grains, due to the loss of structural integrity (Heaton and others 1988). In vitro starch digestion investigations for various food products have shown that the gelatinization of starch in foods such as bread, corn flakes, banana, and oats will facilitate the enzymatic attack of the granules and allow for easier and faster starch digestion (Muir and O'Dea 1992; Englyst and others 1999). However, allowing the starch to retrograde, such as in the process of boiling and cooling potatoes, significantly increases resistant starch content when the amylose and amylopection macromolecules reassociate into crystalline structures that are not easily susceptible to α-amylase (Muir and O'Dea 1992; Faisant and others 1995).
In Vitro Gastric Digestion Models
A limited number of in vitro systems are available for pharmacological, nutrition, safety, and food processing assessments during gastric digestion. Each system has its own advantages and limitations and a brief description of the various systems is given in the following sections.
Bucco-gastric digestion simulation
A variety of in vitro digestion systems have been developed specifically to investigate starch digestion and chemical changes during the oral (buccal) digestion as well as during gastric digestion. One such system is described by Hoebler and others (2002). The basic reactor of this system utilized an Erlenmeyer flask held at 37 °C with a paddle stirrer. The chewing phase of digestion was simulated by mincing products by using a meat grinder (Hoebler and others 2000) and mixing them with a saliva solution containing pancreatic amylase for 5 min. The gastric phase was simulated by adding a mixture of acid and salts to reproduce the gastric juices following a pH profile as measured in vivo with human subjects. Gastric emptying was mimicked by pumping out quantities of the remaining fluid in the beaker, also following an in vivo profile determined from gastric emptying in pigs. The contents emptied at various time intervals were subjected to further analysis.
A similar system was used by Bornhorst (2010) and Bornhorst and others (2011) to study gastric digestion of bread boluses. Oral digestion was simulated by artificial mastication using a meat grinder and the subsequent samples were mixed with simulated saliva using a mortar and pestle. Gastric digestion was studied by placing bolus samples in mesh metal baskets that were soaked in gastric juice in both static and agitated conditions. This study showed that soaking condition (static or agitated), saliva level, presence of α-amylase, and bread type affected the rate of disintegration (Figure 5). The changes in disintegration of the boluses of different bread types were hypothesized to be due to variations in bolus cohesive force, caused by the initial bread structure and moisture content.
Although in vitro digestion studies have various advantages (compared to in vivo digestion studies), namely, simplicity and ease of reproduction, as well as the use of pH and gastric emptying profiles found in vivo, there are some key disadvantages. Since structure has been shown to be very important in starch digestion, it is also vital to consider structural breakdown caused by mechanical forces during digestion, one factor that is not considered in these models. Although the structural breakdown during chewing is reproduced through a mincing procedure, the structural breakdown during the gastric digestion process is not accurately reproduced.
The United States Pharmacopeia (USP) disintegration test is commonly used in the pharmaceutical industry to assess drug erosion and dissolution. There are 2 commonly used apparatuses, as recommended for drug dissolution and disintegration tests by the Food and Drug Administration. One involves a 40-mesh wire basket rotated at a constant speed of 25–150 rpm in a dissolution medium. The second is similar to the first, with the basket replaced by a rotating paddle. Although these methods have been widely used for drug dissolution testing and modeling of the hydrodynamic forces encountered by solids (McCarthy and others 2003), they do not accurately imitate the forces encountered in vivo. Studies have shown that the agitation forces encountered by tablets are position-dependent and that these forces are much weaker than the physiological forces in the GI tract (Kamba and others 2003). The physiological forces, accounting for mechanical impact and destructive forces, are crucial to accurately mimic GI function (Shameem and others 1995). Thus, the applicability of the USP tests is limited (Siewert and others 2003).
Dynamic mechanical models
A number of dynamic mechanical models have been developed in recent years to better mimic both the physical and chemical conditions encountered during digestion. One such model is the TNO intestinal model (TIM), developed at the TNO Nutrition and Food Research Center (Zeist, The Netherlands). TIM is a computerized dynamic system designed to mimic the conditions found throughout the entire gastrointestinal (GI) tract, and has been used to assess nutrient bioavailability (Verwei and others 2003), antimutagenic activity (Krul and others 2001), and heterocyclic aromatic amine availability (Krul and others 2000). The portion of TIM responsible for gastric digestion is a glass compartment with inner flexible walls, with water pumped between the glass outer layer and the flexible inner layer. The contraction/expansion of the flexible inner layer and the material temperature can be controlled by the water pressure and water temperature, respectively. TIM is computer-controlled such that partially digested foods will be moved to the intestinal sections of the model once gastric digestion is finished (Dominy and others 2004).
Another system is the “model gut,” developed at Plant Bioscience Limited and the Institute of Food Research, Norwich, U.K. (Wickham and Faulks 2008). The model gut incorporates 3 stages of the digestive process, taking into account both the physical and chemical processes as observed in vivo. The 3 stages considered are: Stage 1, the main body (upper part) of the stomach, where the food is introduced and mixed in the stomach; Stage 2, the antrum (lower part) of the stomach, where the food is subjected to higher shear forces before emptying from the “stomach” portion of the model; and Stage 3, the duodenum (first part of the small intestine), where the food is subjected to certain chemical conditions before moving through the lower parts of the small intestine (Wickham and Faulks 2008).
A third in vitro system, the Human Gastric Simulator (HGS), was developed to account for the mechanical forces that occur during digestion using a simple moving-wall reactor (Kong and Singh 2010). The HGS consists of a flexible latex chamber (filled with gastric juice and the material to be tested) surrounded by 4 sets of rollers controlled by a motor (Figure 6). The rollers move along the sides of the chamber, simulating the peristaltic waves of the stomach. Gastric secretions can be added to the digesta at a controlled rate, and liquid can be emptied through the “pylorus” to mimic gastric emptying profiles observed in vivo. HGS has been used to study the disintegration of rice, apples, and refined white, whole wheat, rye, sourdough, barley, and almond–wheat bread during the gastric digestion process (Kong and others 2010; Kong and Singh 2010). It was shown that breads baked with distinctive flour types (refined white, whole wheat, rye, barley, and almond–wheat flours) exhibited different rates of breakdown, as quantified by the particle size distribution after in vitro digestion in the HGS, and that the relative rate of breakdown correlated well with observed GI values. This suggests that the HGS may be a useful system to study food carbohydrate breakdown during gastric digestion.
Food material properties are an important part of the entire digestion process, beginning with oral digestion. The breakage that occurs during the chewing process will be heavily influenced by food texture and structure. The food properties, such as moisture and fat contents, will determine how much saliva is secreted and the cohesive forces that will occur during the formation of the food bolus. After the food bolus reaches the stomach, some starches will have been digested by the salivary α-amylase, but with the low pH in the stomach, the amylase may become inactive while the gastric digestion process occurs. During the gastric digestion process, the food bolus will be physically broken apart by the crushing and grinding motions of the stomach, caused by the peristaltic waves of the stomach walls. The food structure is further broken down by pepsin and acid hydrolysis. The rate at which food leaves the stomach, or the gastric emptying rate is controlled by many factors, and the gastric emptying rate can be correlated to the glucose response of starchy food products. The glucose responses of starch-based food products have been extensively studied in the nutrition field and the varying glucose response and starch digestion rate has mainly been attributed to the food and starch granule structure. However, a mechanistic approach that applies the enzymatic conditions and mechanical forces found in vivo, as well as an investigation into the specific factors that most affect starch digestion in vitro has not been conducted.
A more thorough understanding of the driving forces in the breakdown of carbohydrate foods could be applied to the medical and food processing fields. Modifications in the processing of food to give desired digestive characteristics (altering the rate of glucose absorption) can be optimized if a strong linkage between food properties and their breakdown is determined.