The literature search yielded 29 studies that met the inclusion criteria of this systematic review. All studies measured subjective appetite ratings such as hunger, fullness, desire to eat and/or prospective consumption (i.e. how much food participants thought they could eat). Of these, 19 measured subsequent food intake and eight measured gut peptide responses.
The study selection was conducted in several phases following the checklist and flowchart of the PRISMA (Preferred Reporting for Systematic Reviews and Meta-Analyses) guidelines49 as shown in Fig. 2. Initially, a total number of 8,530 articles were identified using literature search in the afore-mentioned six electronic databases.
After removing the duplicates (2,602), the remaining 5,928 titles were screened by the first author (ES) based on their relevance to this review. Firstly, 5,661 studies were excluded based on the PICOS (Population, Intervention, Comparison, Outcome, and Setting) criteria i.e. articles involving animal studies (55), or clinical studies involving patients and/or children or elderly population (141) were excluded. Additionally, articles not addressing the topic of interest were excluded (5,465).
The articles were taken to the next phase where 267 abstracts were screened by ES and AS, resulting in the exclusion of an additional 173 articles (67 articles had no relevance to the topic (s) of systematic review, 56 had non-relevant outcome measures, 23 were new or validation of existing protocols, 11 were non-human studies with additional 7 being non-eligible population and 9 were reviews without any original data). A total of 103 full-text articles, including 9 articles that have been identified through supplementary approaches (e.g. manual searches of reference list of pre-screened articles) were equally divided and screened independently by ES, AS and CG. After a mutual agreement, articles with inappropriate interventions and designs (e.g. eating rate, chewing, free-living design, that included any cognitive manipulation, effect of sugar and fats on satiety and appetite) (n = 29) were excluded. In addition, studies not addressing the topic of interest (n = 16) or having non-relevant outcomes (n = 12) were not considered. Articles where the effects of fibres or dosage of fibre were studied without any direct relevance to textural manipulation (n = 11), articles with no full-text (n = 3) or non-relevant population (n = 3) were also eliminated. To sum it up, a total of 29 articles were included for qualitative synthesis.
Relevant information such as study design, participant gender, type of intervention on texture manipulation, methods of analysing/measuring the texture of food as well as study outcome on appetite ratings, gut peptides and food intake was extracted from the 29 included studies (Table 2).
A total of 817 participants were included in the qualitative synthesis with age ranging from 18 to 50 years (mean age 24.7 years), with the exception of two studies not reporting the participants’ age18,27. Ideally, studies should have an equal ratio of men and women, however, in five studies more women were included than men17,29,52,53,54. On the other hand, a number of studies included more men than women26,55,56. Moreover, in twelve studies men only were included19,21,22,24,25,51,57,58,59,60,61,62. No study included only females and two studies did not mention gender ratio20,27. Only five studies had an equal male/female ratio50,63,64,65,66. All studies selected participants within a healthy BMI range. Mourao et al.50 included both lean and obese participants. However, for this systematic review, the results of lean subjects only were included. In most studies, participants with dietary restrictions or dramatic weight change were specifically excluded as well as those who reported high levels of dietary restraint (11 out of 29) as assessed by either the Dutch Eating Behaviour Questionnaire (DEBQ) or the Three Factor Eating Questionnaire were excluded. Only one study was double-blinded55 and 14 studies used cover stories to distract participants from the real purpose of the study. In only twelve of the studies, a power calculation was used to determine the number of participants needed to find a significance difference21,50,52,54,55,56,57,58,59,64,65,66.
In 16 studies17,18,25,50,52,53,55,56,57,58,61,62,63,64,65,66, manipulations of food forms that were included consisted of liquid versus solid or liquid versus semi-solid or semi-solid versus solid, and included chunky and pureed food. Food consisted mainly of vegetables, fruit, meat and beverage (fruit juices) and texture was manipulated by blending the food. Eight studies20,21,24,51,59,60,61,67 investigated the effect of viscosity, such as low viscosity/ (sensorially termed as ‘thin’) versus high viscosity/ (sensorially termed as ‘thick’), and the texture was manipulated by adding fibres such as, starch, tara-gum, locust-beam gum, alginate, guar-gum, casein and pectin to food products, such as milk products or fruit juices. Two studies26,27 examined the effect of structural complexity, such as low complexity versus high complexity, and the intervention consisted of model foods i.e. hydrogels enclosing various layers and particulate inclusions such as poppy and sunflower seeds. One study19 looked at the homogenization of food, one at the aeration of food incorporating N2O into a liquid drink54 and one study assessed the effect of gels with different lubricity (low vs medium vs high lubricity) using κ-carrageenan and alginate to manipulate the texture29.
Food texture measurements
Nineteen studies measured food texture instrumentally, of which 14 assessed viscosity17,20,21,22,24,29,55,56,59,60,61,62,64,67, two measured lubricity indirectly by measuring friction coefficients29,60, one measured foam volume as a function of time54 and one used aperture sieving19. Eleven studies assessed food texture using sensory evaluation, of which three studies have used a trained panel (11, 29, 33 panellists)20,29,52, four studies untrained (20, 20, 24, 32 panellists)26,27,51,62, and in two studies it is unclear whether it was a trained or untrained panel (20 panellists)21,53. Two studies did not publish or did not show the data25,66. The sensory evaluation was carried out by using Quantitative Descriptive Analysis (QDA)27,29, modified Texture Profile (TP) and Temporal Dominance of Sensations (TDS)27.
Additional information with regards to objective textural manipulation that is characterized by instrumental and sensorial techniques and information on weight and energy density of the intervention, and time to next meal can be found in Supplementary Table S2.
Appetite ratings method
All studies used 100 mm visual analogue scale (VAS) or categorical rating scales to assess appetite ratings. The majority of studies that assessed appetite control measured hunger (n = 27), fullness (n = 25), and desire to eat ratings (n = 21). Two studies19,67 referred only generally to satiety instead of specifying exactly which appetite ratings were being measured.
Food intake measurements
From the limited number of studies that measured gut peptides (n = 8) using blood plasma samples drawn at baseline and different time points after the intervention, two measured CCK alone19,60, two measured ghrelin alone57,58 and four studies measured more than one gut peptides GLP-1 and PYY24, GLP-1 and CCK62, CCK and ghrelin61, ghrelin, CCK, GLP-1 and PYY67. The gut peptides were mainly assayed using commercial plate-based immunoassay test kits.
To assess the quality of studies (n = 29) included in this systematic review, Cochrane’s tool of risk of bias was used68 with regards to random sequence generation, allocation concealment and blinding of participants and personnel and it is reported in Supplementary Table S3. One study55 reported on all three criteria (random sequence generation, allocation concealment and blinding of participants and personnel), and therefore was included in the low risk-of-bias category. Twenty-five studies reported on one or two criteria and were considered as in the medium risk-of-bias category. And three trials17,18,25 did not report clearly on the assessment criteria, therefore were judged within the high risk-of-bias category.
Effect of food texture on appetite control
Of the total studies that measured appetite control (n = 29), 16 found a significant effect of food texture on reducing hunger and increasing fullness ratings. The textural manipulation within these studies ranged from the manipulation of solid-like characteristics to viscosity and to the design of well-characterized model gels with structural complexity (Table 2). For instance, it was noticed that the consumption of solid and/or semi-solid food more strongly suppressed appetite ratings as compared to ratings of liquid food. Flood-Obbagy and Rolls65 found that whole apples led to decreased hunger ratings and increased fullness when compared with their liquid counterparts (i.e. apple sauce and juice). These authors argued that the effect of food on satiety was due to the structural form of food itself and the larger volume in case of whole fruit as compared to the liquid versions, even when matched for energy content and weight. Interestingly, these findings were not associated with the amount of fibre as the fibre content was similar across liquid and solid conditions. Similar findings by Hogenkamp et al.52 indicated that hunger decreased, and fullness increased in the semi-solid condition compared to the liquid condition. They found that the semi-solid product (comparable with firm pudding) suppressed appetite greater than the liquid product (comparable with very thin custard). The authors related their findings to the triggering of the early stages of the satiety cascade10 through cognitive factors and sensory attributes such as visual and oral cues; whereas food forms might not affect the later processes in satiety cascade that are postulated to be governed by post-ingestive and post-absorptive factors52.
Foods with high viscosity also appeared to play a key role in appetite suppression compared to food with low viscosity20,22,51,60. Aiming to determine the effect of viscosity on satiety, Solah et al.20 used low and high viscous alginate-based breakfast drinks, on 33 subjects. It was found that hunger was lower after participants consumed the high viscous alginate drink as compared to those who consumed low viscous ones. The authors speculated that such findings were related to the gastric distention as a result of the ingested gel-forming fibre, although they did not measure the rheological properties of these foods in the gastric situation. In a rather long-term (7 non-consecutive days over a month) study, Yeomans et al.51 investigated low (thin) and high (thick) viscous drinks with both low and high energy content, respectively. They found that initially, appetite was suppressed after consuming high viscous foods as compared with those who consumed low viscous foods, corroborating the afore-mentioned effect of viscosity on satiety. They related their findings to a slower gastric emptying rate in the high viscous food. However, after repeated consumption of the drinks with seven non-consecutive days over a month, there were no noticeable differences in satiety between the low and high viscous conditions51. Expected satiation was higher for both high energy drinks and lower for both low energy drinks irrespective of the viscosity of the foods. This suggests that in a repeated consumption setting, the effect of viscosity can be negligible.
It is noteworthy that some of the authors relate their findings of increased satiety after consuming high viscous foods to a slower gastric emptying rate, which should be interpreted with some caution. For instance, Camps et al.21 directly measured the effect of viscosity on gastric emptying in their study using magnetic resonance imaging (MRI) abdominal scans and found that not only the viscosity of food but also the energy load led to a slow gastric emptying. The preloads in their case were four shakes differing in viscosity (low and high viscosity) measured in perceived thickness using 100-mm VAS scale and also differing in energy content (low/100 kcal and high/500 kcal) consumed within 2 min. The increase in the energy load led to slower gastric emptying over time; it only significantly slowed the emptying under the low-energy-load condition. Therefore, they suggested that viscosity loses its reducing effect on hunger if energy load is increased to a meal size of 500 kcal indicating that viscosity may not always affect the later parts of satiety cascade through delayed gastric emptying route, but contributes to the early parts of satiety cascade via mouth feel and oral residence time.
In addition to form and viscosity, textural complexity has also shown some significant effects on appetite control. However, the term textural complexity is rather poorly defined in the literature. Often it refers to the degree of heterogeneity or inhomogeneity in a food where the preload includes some inclusions, which distinguishes it from a control; the latter having a homogenous texture i.e. without inclusions. This research domain of studying the effects of so-called textural complexity on satiety is still in its early infancy. Tang et al.26 conducted the first trial on textural complexity (of the preload) (Fig. 1) and demonstrated that hunger ratings decreased when model food gels with higher complexity (i.e. gels layered with particulate inclusions) were served. The authors noticed that higher inhomogeneity in the gels with particle inclusions led to a decrease in hunger and desire to eat, and an increase in fullness ratings, suggesting that levels of textural complexity may have an impact on post-ingestion or post-absorption processes leading to a slowing effect on feelings of hunger.
The technique of aeration, (i.e. incorporation of bubbles in a food) has been also used as a textural manipulation and been shown to have an influence on satiety. Melnikov et al.54 found that hunger was lower, and fullness was higher in aerated drinks as compared to the non-aerated counterparts. Although these drinks differed in energy content (low/high energy non-aerated and low/high energy aerated), they demonstrated that such aeration independent of energy content was a promising textural manipulation to suppress appetite. The authors attributed the findings to the effect of the air bubbles on gastric volume leading to the feelings of fullness.
In thirteen studies out of the 29 studies, food texture was reported to have no effect on appetite ratings. This disparity in the results may be associated with the methodology employed. For instance, in several studies27,50,58 participants were instructed to eat their usual breakfast at home. Therefore, the appetite level before the preload was not controlled and this might have influenced the appetite rating results. Furthermore, some studies did not conceal the purpose of the study from the participants18,25. Thus, participants’ responses might have been biased and could have led to less reliable results69. Moreover, Mourao et al.50 firstly served an ad libitum meal to participants and then immediately the preload with different textural attributes. As such, the time interval between ad libitum intake and preload may have accounted for variation in outcomes70. All these factors may explain the disparities with regards to the effects of food texture on subjective appetite ratings.
Effect of food texture on gut peptides
Out of the limited number of studies (n = 8) that included gut peptides measurements; only two61,67 studies found an effect of food texture. Contrary to our expectations, Juvonen et al.67 found that CCK, GLP-1, PYY increased and ghrelin decreased in low viscous condition compared to high viscous one. The authors speculate that after consuming a high viscous drink, viscosity of the product may delay and prevent the close interaction between the nutrients and gastrointestinal mucosa required for efficient stimulation of enteroendocrine cells and peptide release. The same results were found in regard to food form. Zhu et al.61 found that liquid food (pureed liquid–solid soup) resulted in a higher postprandial response of CCK comparing with solid food (whole pieces of vegetables in a chicken broth). They related it to the capacity of CCK to be secreted in the duodenum in response to the presence of nutrients. As such, they suggest that the increase in the surface area of the nutrients due to the smaller particle sizes resulted from the pureeing could stimulate secretion of CCK more potently.
The rest of the studies found no significant effect of food texture (form, viscosity or complexity) on triggering relevant gut peptides. This may be due to the type of macronutrients used in such intervention. For example, intervention in Martens’ et al.57 study was high in protein and it is known that proteins are less effective in suppressing ghrelin13. Therefore, one may argue that the effect of food texture is only restricted to early stages of satiety cascade rather than later stages, where the type and content of macronutrient might play a decisive role. However, such interpretations might be misleading owing to the limited number of studies in this field. Also, in the majority of studies conducted so far, the biomarkers were limited to one gut peptide, such as CKK19,60,61 or ghrelin57,58, which provides a selective impression of the effects on gut peptides. Measuring more than one gut peptide could provide richer data and wider understanding of the relationship between food texture and gut peptides, which has yet to be fully evaluated69.
Effect of food texture on energy intake
Seven out of the total 29 studies found a significant effect of texture on food intake. Food form, such as solid, appeared to play a role in the subsequent food/energy intake. For example, in the study by Flood and Rolls65, 58 participants consumed apple segments (solid food) on one day and then apple sauce (liquid food) made from the same batch of apples used in the whole fruit conditions on another day. The preload was controlled for the energy density and consumed within 10 min and the ad libitum meal was served after a total of 15 min. As a result, they found that apple pieces reduced total energy intake at lunch as compared to the apple sauce, therefore suggesting that consuming whole fruits before a meal can enhance satiety and reduce subsequent food intake. Mourao et al.50 also confirmed such findings where participants consumed less energy after ingesting solid food form (cheese/watermelon fruit/coconut meat) as compared to the beverage form (milk/watermelon juice/coconut milk). However, it is worth noting that they had a different experimental approach in contrast to the rest of the studies in this systematic review. First, an ad libitum meal was served and then followed by a fixed preload consisting of solid and beverage form with one predominant macronutrient (milk-protein, watermelon-carbohydrate and coconut-fat). The time between ad libitum meal and the preload was not stated; it is only clear that it was served at lunch time. Food records were kept on each test day (for 24 h) to determine energy intake. Despite this different approach, it was demonstrated that solid food led to a lower subsequent energy intake compared with liquid food counterparts. Consequently, this study supports an independent effect of texture on energy intake.
In terms of viscosity, it has been found that higher viscous food can also lead to a reduced subsequent energy intake. This was noted in Juvonen’s et al.67 study, where participants consumed two identical, isoenergetic and isovolumic oat bran beverages that differed only in their viscosity (low, < 250 mPa; high, > 3,000 mPas) which was measured instrumentally. Authors reported that the beverage with high-viscosity led to a lower energy intake compared to the low-viscous beverage when energy consumption during the meal consumed ad libitum and during the rest of the test day was combined. Although authors attribute their findings to a slower gastric emptying rate, they did not measure it directly, nor was the effect of viscosity on mouth feel or oral residence time affecting early stages of satiety cascade investigated.
Even with a limited number of studies, textural complexity has been demonstrated to have a clear impact on subsequent food intake. For instance, in the studies of Tang et al.26 and Larsen et al.27, gels mixed with poppy and sunflower seeds reduced subsequent food intake independently of the oral transit time and energy density, suggesting a sole impact of food texture on food intake.
Interestingly, Krop et al.29 also showed a clear effect of texture on reducing subsequent snack intake by using hydrogels (having no energy content or micronutrients) that differed in their textural complexity in terms of their lubricating properties, which was measured both instrumentally and sensorially71. These authors related their findings to hydrating and mouth-coating effects after ingesting the high lubricating carrageenan-alginate hydrogels that in turn led to a lower snack intake. Moreover, they demonstrated that it was not the intrinsic chewing properties of hydrogels but the externally manipulated lubricity of those gel boli i.e. gel and simulated saliva mixture that influenced the snack intake. All these reports suggest that there is a growing interest in assessing food texture from a textural complexity perspective. This means introducing heterogeneity such as tribological/ lubrication alternation in food to have enhanced satiety and satiation consequences. This strategy needs attention in future satiety trials as well as longer-term repeated exposure studies.
The energy density of the preload across the studies varied from zero kcal29 or a modest energy density 40 kcal26,27 up to a higher value of – 600–700 kcal18,19 (see the Supplementary Table S2). It is noteworthy that the lower the energy density of the preload, the shorter the time interval between the intervention (preload) and the next meal (ad libitum meal). Some of these studies showed an effect of texture on appetite ratings and food intake, with food higher in heterogeneity leading to a suppression of appetite and reduction in subsequent food intake26,27. Also, gels with no calories but high in their lubrication properties showed a reduction in snack intake29. Contrary to those textures with zero (or modest levels of) calories, those textures high in calories tended to have a larger time gap between the intervention (preload) and the next meal. An interesting pattern observed across these studies employing high calorie-dense studies, is that an effect of texture on appetite ratings was found but no effect on food intake22,61,66. Therefore, in addition to the high energy density of the preload, it appears that time allowed between the preload and the next meal is an important methodological parameter.
A total of 23 articles were included in the meta-analysis. Two articles were excluded as data on a number of outcomes were missing19,50. Meta-analysis on structural complexity26,27, lubrication29, aeration54 and gut peptides could not be performed due to the limited number of studies that addressed this issue, and therefore a further four articles were excluded. Finally, meta-analysis was performed on the effect of form and viscosity of food on three outcomes: hunger, fullness and food intake. Data from 22 within-subjects and 1 between-subjects trials reporting comparable outcome measures were synthesised in the meta-analyses. These articles were expanded into 35 groups as some studies provided more than one comparison group. In most of the studies (n = 18), appetite was measured on 100 mm visual analogue scale (VAS).
Meta-analyses presenting combined estimates and levels of heterogeneity were carried out on studies investigating form (total of 20 subgroups, 651 participants) and viscosity (total of 15 subgroups, 281 participants) for the three outcomes hunger, fullness and food intake (see data included in the meta-analysis in Supplementary Tables S4a–c). There was an insufficient number of studies to carry out meta-analyses for the ones investigating complexity (n = 2)26,27, lubrication (n = 1)29, aeration (n = 1)54 (total of 4 studies, 103 participants) and gut peptides (total of 8 studies, 130 participants (e.g. 3 studies assessed GLP-1 with available data on 2 studies24,62, and 2 studies assessed PYY with available data on 1 study only24).
A meta-analysis of 556 participants, from 16 subgroups based on food form (13 comparing solid with liquid food and 3 comparing semi-solid with liquid food) revealed an overall significant decrease in hunger with the intervention (solid or semi-solid) group of − 5.00 mm (95% confidence interval (CI) − 8.27 to − 1.73, p = 0.003, I2 = 71%). There was a significant decrease in hunger with solid food of − 6.58 units (95% CI − 9.61 to − 3.54, p < 0.001, I2 = 39%) however no difference in hunger was seen for comparisons of semi-solid with liquid food (see Fig. 3a). A meta-analysis of 191 participants from 11 subgroups based on viscosity revealed a borderline significant decrease in hunger with higher viscosity food of − 2.10 mm (95%CI − 4.38 to 0.18, p = 0.071, I2 = 59%) (see Fig. 3b).
A meta-analysis of 263 participants, from 12 subgroups based on form (9 comparing solid with liquid food and 3 comparing semi-solid with liquid food) revealed no overall difference in fullness between the intervention (solid or semi-solid) group and control group (− 0.75 units, 95% CI − 3.93 to 2.43, p = 0.644, I2 = 91%). There was no difference in fullness between groups for either of the two subgroups (see Fig. 4a).
A meta-analysis of 155 participants from 11 subgroups based on viscosity revealed an overall significant increase in fullness for higher viscosity food of 5.20 mm (95%CI 2.43 to 7.97, p < 0.001, I2 = 76%) (see Fig. 4b).
A meta-analysis of 458 participants, from 12 subgroups based on form (9 comparing solid with liquid food and 3 comparing semi-solid with liquid food) revealed no overall difference in food intake with the intervention (solid or semi-solid) group compared with the control group (− 26.2 kcal, 95% CI − 61.7 to 9.4 kcal, p = 0.149, I2 = 0%) (see Fig. 5a).
There was a borderline significant reduction in food intake for studies comparing solid with liquid food of − 55.5 kcal (95% CI − 111.1 to − 0.1 kcal, p = 0.05, I2 = 0%) however no difference in food intake was seen for comparisons of semi-solid and liquid food. A meta-analysis of 191 participants from 9 subgroups based on viscosity revealed a non-significant decrease in food intake with higher viscosity food of − 66.7 kcal (95%CI − 144.2 to 10.9 kcal, p = 0.092, I2 = 84%) (see Fig. 5b). Funnel plots (see Supplementary Figure S1a–c) reveal that there was some evidence of asymmetry and therefore publication bias may be present, particularly for the meta-analyses for hunger.