Dr KM Appleton, School of Psychology, Queen's University, Belfast, 18-30 Malone Road, Belfast BT9 5BP, UK. E-mail: firstname.lastname@example.org
Several short-term studies have investigated the effects of a vegetable oil emulsion on subsequent food intake, although findings have been inconsistent. This work aimed to review all studies, and investigate differences in study outcomes based on methodology. All known studies were identified. Data were abstracted from published studies (n = 7). Details of unpublished studies were gained from investigators/sponsors (n = 5), or were unavailable for reasons of confidentiality (n = 4). Available data were combined using meta-analyses. A combined appetite suppressant effect of the emulsion compared with control was found for test meal intake at approximately 4, 12 and 36 h post-treatment: smallest combined mean difference (random effects model) = 0.53 MJ (95% confidence interval 0.20, 0.86), P < 0.01. However, considerable heterogeneity (variability) between study results was also found (smallest I2 = 94%, P < 0.01), questioning the predictive validity of the above findings. Meta-regression suggested this heterogeneity to be related to differences in the processed nature of the product, treatment dose and in particular year of study (smallest B = 0.54, 95% confidence interval 0.06, 1.03, P = 0.04), although again heterogeneity was found. The only consistent finding was a lack of effect on food intake 4 h post-preload in studies conducted after 2003. These results suggest a small but inconsistent appetite suppressant effect of the vegetable oil emulsion. However, due to the large heterogeneity, no definitive conclusions can be drawn.
The prevalence of overweight and obesity continues to increase throughout the world (1). Alongside this increase, the health costs and implications associated with obesity also continue to rise (2). These rising health costs and implications highlight the need for strategies and aids to treat and prevent weight gain.
Industry has responded to this growing need with the development and manufacture of various products aiming to aid weight loss and promote weight management. One such product is a novel vegetable oil emulsion (Fabuless™[previously Olibra®]), formulated from palm oil and oat oil fractions, demonstrated to prolong the passage of foods through the small intestine (3), potentially resulting in increased and extended post-prandial satiety.
The effects of this vegetable oil emulsion on appetite have been investigated repeatedly. Early studies report a suppression of appetite (reduction in food intake and hunger ratings) after a single dose of this oil emulsion compared with control. Burns et al. (4) found that the oil emulsion reduced test meal intake, Burns et al. (5) found that the emulsion reduced both test meal intake and rated hunger for up to 24 h, and Burns et al. (6) reported reduced test meal intake up to 36 h following the oil emulsion. Other studies, however, have failed to find effects. Logan et al. (7) found no effects compared with control on three separate test days during a 21-day consumption period, and Rogers and colleagues at the University of Bristol found no effects of the oil emulsion in several currently unpublished studies.
Short-term laboratory-based studies of acute appetite effects have the advantage of a high degree of control over extraneous variables and accurate measurement of food intake. However, in relation to the assessment of food ingredients as potential aids to weight management, they are limited by the short time frame over which effects are measured, the artificial nature of the situations involved, and the assumption that weight management is achieved only through changes in appetite (8,9).
Differences between short-term study outcomes may occur as a result of differences in study methodology. Explanation of these differences may increase understanding of the potential value of this oil emulsion to aid weight management. This paper reviews all studies investigating the effects of the vegetable oil emulsion on appetite and food intake, and investigates explanations for differences between study outcomes based on study methodology.
Review of studies
All studies investigating the effects of the vegetable oil emulsion (Fabuless™, DSM Nutritional Products, Switzerland, formerly also known as Olibra®) on appetite and food intake were identified. Published studies were obtained from relevant journals. Data on the methodology of all published studies were abstracted independently by two researchers, and discordances were resolved by a third researcher. Investigators and sponsors of all unpublished studies were contacted and data were requested. Where investigators and sponsors agreed to the use of data, required details were provided.
Analysis of studies
The methods and findings of all available studies were reviewed. All aspects of methods thought to be important in appetite research, as judged by the authors, contributors from DSM, and as published (8–10), were considered. Data from all available studies were subsequently combined using meta-analysis. Both random and fixed effects models were used to estimate combined mean differences, although random effects models are theoretically preferable when combining the results of studies when heterogeneity exists (11–13). Heterogeneity was investigated using Higgins' I2 statistic (14,15), which describes the proportion of total variation in study estimates that is due to heterogeneity. Meta-regressions were conducted to explore possible sources of heterogeneity. Sensitivity analyses were subsequently conducted where possible sources of heterogeneity were identified. For all sensitivity analyses, effects were investigated by dividing available studies into two approximately equal groups and comparing effects in each separate group. Analyses were performed in Stata Version 8.1 (StataCorp. Stata Statistical Software: Release 8.0. College Station, TX: Stata Corporation, 2003), using the ‘meta’ and ‘metareg’ commands.
Sixteen studies investigating effects on appetite and food intake were identified. Seven of these studies have been published (Burns et al. (4)– studies 1 and 2; Burns et al. (5); Burns et al. (6); Logan et al. (7); Diepvens et al. (16); Smit et al. (17)), and nine studies are unpublished. All published studies were obtained and are included in the review and analysis. Of those that are currently unpublished, full methods and data were obtained from investigators and sponsors for five studies (studies BF, U1, U3, U4, DSM), but neither methods nor data could be obtained for four studies, for reasons of confidentiality.
The methods of all studies for which methodological details were available are shown in Table 1. Studies were conducted between roughly 1998 and 2006 at the University of Ulster, the Unilever Health Institute, Vlaardingen, the University of Bristol and the University of Maastricht. All studies used a single or double blind, placebo-controlled within-subjects design, with a preloading procedure, in which participants experienced the treatment as a preload and appetite was subsequently measured over the short term. In all studies, appetite was measured using food intake and subjective ratings, excepting one (study U1) where only subjective ratings were used. All participants experienced each treatment only once, with the exception of the study by Logan et al. (7), where participants experienced the treatment up to 21 times. Male and female participants were involved in all studies, although in differing proportions, excepting in one study (16) where only female participants were used. The majority of studies involved participants with a mean age of approximately 25 years, although in two studies, mean age was approximately 30 years (5 [overweight group], 7), and in three studies, mean age was approximately 40 years (5 [obese group]; 16 [older group]; study U1). Studies involved lean (4, studies 1 and 2, 5–7, 16, 17, studies BF, U3, U4, DSM), overweight (4, study 2, 5, 16, study U1), and obese participants (5). The majority of studies also involved only non-smokers, non-vegetarians and participants low in dietary restraint (5, 16, 17, studies BF, U3, U4, DSM), although other selection criteria differed between studies.
Table 1. Review of methodology and results for all studies as published
Treatments were provided in a yoghurt, yoghurt drink, milk-based drink or soup vehicle, and provided different doses of Fabuless™ fats. Burns et al. (5) used 0.8 g Fabuless™ fats; Burns et al. (6) and study U3 used 2 g; study U1 used 2.5 g; study U3 used 3.2 g; Burns et al. (6) and Diepvens et al. (16) used 4 g and study BF is estimated to have used 4 g; Burns et al. (4, studies 1 and 2), Burns et al. (5), Logan et al. (7), Smit et al. (17), U4 and DSM used 5 g; and Burns et al. (6) used 6 g. In three studies, unprocessed Fabuless™ was used (6, 17, study BF), and in all others, excepting study U1, processed Fabuless™ was used, although type of processing varied between studies. Information on processing was unavailable for study U1. Preloads were given at 9:00 with the exception of those in study U3 which were given at 10:30, those in study U1 which were given at 12:00, and those in the study by Burns et al. (4) and study BF which were given at 13:00.
Appetite was assessed using buffet test meals, composed of 1 (16), 3 (17), 4 (studies U3, U4), 17 (6, DSM), 22 (study BF), 23 (5), 36 (4, studies 1 and 2) or 43 (7) foods, provided in either the dining area of a metabolic suite (4, studies 1 and 2; 5–7; study BF) or individual testing booths (17, studies U3, U4, DSM), 4 h after the preload, excepting in study U3 where the duration between preload and test meal was 3 h. Test meals lasted for an unlimited duration (4, studies 1 and 2, 5–7, study BF), 30 min maximum (17; studies U3; U4; DSM) or 20 min maximum (16). A second test meal was also provided in five studies (5; 17; studies U3; U4; DSM), composed of 3 (17), 4 (study U3), 5 (study U4), 14 (study DSM) and 36 (5) foods. Excepting the study by Diepvens et al. (16), all studies recorded subsequent intake for the rest of the day using either food diaries (4, studies 1 and 2, 5–7, study BF) or a snackbox plus food diaries (17; studies U3, U4, DSM). Three studies also recorded intake on the following day using food diaries (5–7). All studies measured appetite using subjective perceptions, and the majority of studies also measured ill effects/gastrointestinal complaints (5, 6, 17, studies BF, U1, U3, U4, DSM). Studies also used similar procedural controls with the exception of the Logan et al. (7) study, due to study design.
Studies for which data were available were combined, to provide an overall mean difference for Fabuless™ compared with control for test meal intake. Separate analyses were conducted for cumulative test meal intake at approximately 4 h, 8 h, 12 h and 36 h after preload, and studies with available data were included in each analysis. Due to the greater experimental control and the greater number of available studies, test meal intake at approximately 4 h post-preload was used as the primary outcome measure, and intake up to all other time points were considered as secondary outcome variables. Cumulative intake was analyzed as opposed to individual test meal intake at each time point to provide results relevant to everyday life. All studies with available data were included in each analysis. Studies with multiple treatments (6, 17, study U3) were included in each analysis, where each treatment was included separately. One study with multiple assessments (7) was included only once to abide by the assumptions of meta-analysis. In this case, data from the first day of testing were used, as this was comparable with all other studies. This resulted in the possible inclusion of 17 individual trials for analysis – Burns et al. (4) study 1; Burns et al. (4) study 2; Burns et al. (5); Burns et al. (6) 0.8 g dose; Burns et al. (6) 2 g dose; Burns et al. (6) 4 g dose; Burns et al. (6) 6 g dose; BF, unprocessed dose; BF, processed dose; Logan et al. (7) assessment day 1; Diepvens et al. (16); Smit et al. (17) unprocessed dose; Smit et al. (17) processed dose; U3, 2 g dose; U3, 3.2 g dose; U4; DSM.
Test meal intake at approximately 4 h post-preload
All studies available for analysis investigated test meal intake approximately 4 h post-preload. Initial combination of all studies using a random effects model showed a combined mean difference (Fabuless™ compared with control) of 0.53 MJ (95% confidence interval [CI] 0.20, 0.86), P < 0.01, although considerable heterogeneity between studies was also found (I2 = 97%, P < 0.01). Initial combination using a fixed effects model showed a combined mean difference (Fabuless™ compared with control) of 0.24 MJ (95% CI 0.14, 0.33), P < 0.01. The forest plot from the random effects model is given in Fig. 1. These results suggest a small appetite suppressant effect of Fabuless™, although caution must be exercised due to the considerable heterogeneity.
To investigate this heterogeneity, meta-regression was conducted on data from all 17 individual studies from the random effects model, where weighted study mean differences were predicted by all variables of potential importance in determining energy intake that were found to differ between studies. These variables were: gender of participants (proportion who were female), age of participants (mean), body mass index (BMI) of participants (mean), dose of treatment, processed nature of treatment (unprocessed vs. processed), time of treatment (time of preload), composition of the test meal (number of foods), test setting, and year of study. Study mean data for gender, age and BMI were used initially to ensure inclusion of all data for all studies. Test setting included location of study (Ulster/Bristol/Maastricht), test meal setting (dining area of a metabolic suite/individual test booths/unknown) and time allowed for consumption of the test meal (unlimited/30 min maximum/20 min maximum.). All these variables co-varied exactly, and were included in the analyses either as one dichotomous variable or as three dummy variables (Test setting [Ulster], Test setting [Bristol], Test Setting [Maastricht]). Year of study was considered to be 2 years prior to the date of publication for all published studies unless known.
The results of the meta-regression on data investigating test meal intake approximately 4 h post-preload are given in Table 2. These results suggest that the differences between studies are, at least in part, a result of differences in year of study, treatment dose, and the sex, mean age and mean BMI of participants.
Table 2. Coefficients for all variables in the meta-regression for test meal intake at approximate 4 h post-preload
BMI, body mass index; CI, confidence interval; SE, standard error; T, t-value; P, P-value.
Sex of participants (% female)
Age of participants
BMI of participants
Dose of treatment
Processed nature of treatment (unprocessed = 1, processed = 2)
Time of treatment
Number of foods in test meal
Test setting (setting Bristol = 1, other = 2)
Test setting (setting Maastricht = 1, others = 2)
Year of study
To investigate possible effects of year of study, sensitivity analyses were conducted on all studies conducted before 2003, and all studies conducted after 2003. For all studies conducted before 2003 (n = 9), the results from a random effects model showed a combined mean difference (Fabuless™ compared with control) of 1.07 MJ (95% CI 0.58, 1.55), P < 0.01, although considerable heterogeneity between studies was found (I2 = 86%, P < 0.01). For all studies conducted after 2003 (n = 8), the results from a random effects model showed a combined mean difference (Fabuless™ compared with control) of −0.11 MJ (95% CI −0.22, 0.01), P = 0.08, with no heterogeneity between studies (I2 = 4%, P = 0.47). These results suggest an appetite suppressant effect of Fabuless™ vs. control in short-term studies conducted before 2003, although this result should be treated with caution due to the considerable heterogeneity, and show no effect in short-term studies conducted after 2003.
To investigate possible effects of dose of treatment, sensitivity analyses were conducted on all studies using less than 5 g Fabuless™, and all studies using 5 g Fabuless™ or more. For all studies using a dose of less than 5 g (n = 8), the results from a random effects model showed a combined mean difference (Fabuless™ compared with control) of 0.47 MJ (95% CI 0.02, 0.93), P = 0.04, although considerable heterogeneity between studies was found (I2 = 87%, P < 0.01). For all studies using a dose of 5 g or more (n = 9), the results from a random effects model showed a combined mean difference (Fabuless™ compared with control) of 0.58 MJ (95% CI 0.08, 1.08), P = 0.02, although again considerable heterogeneity between studies was found (I2 = 98%, P < 0.01). These results suggest an appetite suppressant effect of Fabuless™ vs. control using both doses of less than 5 g Fabuless™ and doses of 5 g Fabuless™ or more, although the results of both meta-analyses should be treated with caution due to high heterogeneity.
To investigate possible effects of participant sex, age and BMI, individual data were used as opposed to study mean data. Individual data were available for all studies excepting two (studies BF, 16), although where only one individual existed in any one sex, age or BMI category in any study, data from this individual also could not be used in this analysis. Using these data, no effects of sex (B = 0.14, 95% CI −0.48, 0.76, P = 0.65), age (B = −0.19, 95% CI −0.77, 0.39, P = 0.51), or BMI (B = −0.14, 95% CI −0.77, 0.50, P = 0.67) were found.
Subsequent intake up to approximately 8, 12 and 36 h post-preload (Cumulative Intake)
Similar analyses were also conducted for cumulative intake up to approximately 8, 12 and 36 h post-preload, although intake was not measured using test meals in all studies at these time points. Instead, in some studies and at some time points intake was measured using snack boxes (studies U3, U4, DSM), and in others or at other time points intake was measured using food diaries (4, studies 1 and 2, 5–7, study BF).
At approximately 8 h post-preload (measured in nine studies), initial combination using a random effects model showed a combined mean difference (Fabuless™ compared with control) of 0.64 MJ (95% CI −0.02, 1.29), P = 0.06, although considerable heterogeneity between studies was found (I2 = 93%, P < 0.01). Two meta-regressions (one investigating personal factors, one investigating procedural factors) suggested that none of the variables included in the models accounted for the heterogeneity between studies (largest B = −0.54, 95% CI −1.30, 0.21, P = 0.09).
At approximately 12 h post-preload (12 studies), initial combination using a random effects model showed a combined mean difference (Fabuless™ compared with control) of 2.37 MJ (95% CI 0.83, 3.91), P < 0.01, although considerable heterogeneity between studies was found (I2 = 97%, P < 0.01). Two meta-regressions (one investigating personal factors, one investigating procedural factors) suggested that the heterogeneity between studies was, at least in part, related to the processed nature of the treatment (unprocessed vs. processed) (B = −3.74, 95% CI −6.25, −1.24, P = 0.02) and Fabuless™ dose (B = 0.54, 95% CI 0.06, 1.03, P = 0.04).
Investigation of possible effects due to the processed nature of Fabuless™ revealed a combined mean difference of 5.21 MJ (95% CI 3.78, 6.64), P < 0.01 in studies using unprocessed Fabuless™ (n = 4) (I2 = 88%, P < 0.01), and no effect in studies using processed Fabuless™ (n = 6) (combined mean difference = 0.51 MJ [95% CI −0.50, 1.52], P = 0.32; I2 = 94%, P < 0.01). Investigation of effects due to dose of Fabuless™ suggested an appetite suppressant effect (combined mean difference = 2.74 (95% CI 0.32, 5.16), P = 0.03) in studies using a dose of less than 5 g (n = 5) (I2 = 98%, P < 0.01), and no effect in studies using 5 g or more (n = 5) (combined mean difference = 2.01 (95% CI −0.30, 4.31), P = 0.09; I2 = 98%, P < 0.01).
At approximately 36 h post-preload (six studies), initial combination of all studies using a random effects model showed a combined mean difference (Fabuless™ compared with control) of 5.69 MJ (95% CI 3.32, 8.06), P < 0.01, although considerable heterogeneity between studies was also found (I2 = 94%, P < 0.01). Meta-regression suggested that none of the variables included in the model accounted for the heterogeneity between studies (largest B = 0.90, 95% CI – 0.11, 1.90, P = 0.06).
Effect of Fabuless™ on test meal intake
Initial combination of all studies available at each time point showed an appetite suppressant (reduced food intake) effect of Fabuless™ compared with control. Specifically, Fabuless™ reduced cumulative energy intake at 4, 12 and 36 h post-preload, with a marginal effect at 8 h post-preload. However, caution must be exercised in respect of any conclusion about the effects of Fabuless™ on food intake due to considerable heterogeneity between studies. Heterogeneity refers to the variability or differences between study results (11,12). The considerable heterogeneity here suggests marked variability between studies, making the combined mean differences difficult to interpret and of limited predictive validity.
The random effects model investigating effects at approximately 4 h post-preload showed a combined mean reduction in calorie intake of ∼0.53 MJ (equalling ∼125 kcal) relative to control. Although the magnitude of this effect appears small, using data from the National NHANES survey, it was calculated that affecting energy balance by approximately 100 kcal day−1 could influence weight gain in most of the US population (18). Although caution must be exercised in respect of any conclusion about the effects of this vegetable oil emulsion on food intake due to considerable heterogeneity between studies, it is of interest that small differences in calorie intake may be relevant in a weight maintenance context.
Sources of heterogeneity
Meta-regression suggested that the heterogeneity between studies is, at least in part, due to differences in year of study (4 h post-preload); the processed vs. unprocessed nature of the treatment (12 h post-preload), and treatment dose (4 and 12 h post-preload).
Effects related to year of study were found where an appetite suppressant effect of Fabuless™ compared with control was apparent 4 h post-preload in studies conducted before 2003, but no significant food intake effect was found in studies after 2003. Heterogeneity, however, is very high for the earlier studies, although the later studies show good agreement. These differences in findings are most plausibly explained as an effect due to the manufacture, processing or preparation of Fabuless™ during the time period over which these studies were conducted. The Fabuless™ product is a manufactured food ingredient, which is processed further when added to food products. Fabuless™ can be incorporated into fermented milk products, such as yoghurt, using processing procedures common in the dairy industry, with certain precautions to ensure the integrity of the Fabuless™ emulsion structure. Over the years it has become clear that Fabuless™ should be dispersed in fine droplets throughout a vehicle product (Sein, 2010, DSM, personal communication), and is susceptible to breakdown when exposed to high shear forces at low temperatures and an acid environment. Other fats or emulsions and dehydration may also affect emulsion integrity. A high salt content, such as that found in soups can also destabilize the emulsion, especially if sterilized in liquid form to achieve a long shelf-life (Viberg, 2008, DSM, personal communication). In addition, Fabuless™ is made from natural products with possible seasonal variation. This is particularly true for the oat oil and its molecular composition, but whether these changes affect the behavior of the emulsion in the gastro-intestinal track is difficult to predict or study (Sein, 2010, DSM, personal communication). All these factors influencing the nature of Fabuless™ during the processing, however, have only gradually been elucidated over time and were not recorded in detail in each of the studies.
The suggestion that effects of year may be a result of differences in the manufacturing or processing of Fabuless™ is also supported by the effects found based on the processed nature of the treatment at 12 h post-preload, where effects were found using unprocessed Fabuless™, while no effects were found using processed Fabuless™. This finding strongly suggests that activities that occur as part of the processing of Fabuless™ have in some studies rendered the treatment inactive. It is important however to stress that the high levels of heterogeneity in these analyses and the presence of effects only at 12 h post-preload, suggest that these results should again be treated with considerable caution.
Effects of treatment dose on food intake are unclear. Sensitivity analyses on effects after 4 h did not confirm an effect of dose, but after 12 h indicate a greater appetite suppressant effect of Fabuless™ using doses of less than 5 g and no effect using higher doses. These findings suggest that low doses of Fabuless™ may suppress appetite more effectively than high doses, and this is supported by data from studies in which 2 g doses of Fabuless™ were found to reduce food intake (e.g. study BF). Meta-regression coefficients, however, suggest greater effects using higher doses, as was also observed in the dose–response study by Burns et al. (6). The discrepancy between the results of the sensitivity analyses and the meta-regressions indicates that effects of treatment dose are highly influenced by other variables included in the meta-regression models. Given the additional effects of year of study and the processed nature of the treatment, it is possible that the effects of treatment dose revealed in the sensitivity analyses are, likewise, actually due to differences between studies in the manufacturing or processing of Fabuless™. These effects highlight also the difficulties of conducting meta-analyses on limited and highly variable data.
No relationships were found between the appetite suppressant effects of Fabuless™ and the sex, age or BMI of participants in all studies, the timing of the preload, test setting or number of foods in the test meal. In other words, Fabuless™ had similar effects, or lack of effect, compared with control in both males and females, in individuals of different ages, in individuals of different BMIs, when provided at 9:00, 10:30 and 13:00, and when tested in different settings using different test meals. Effects of age were previously reported by Burns et al. (5) and Diepvens et al. (16) where greater effects were found in younger compared with older participants, but in both studies, age was confounded by BMI. Effects of BMI have previously been reported, where effects were found to be inversely related to weight group (5,16). This was explained as a result of lower satiating properties of fat in obese groups (19), but BMI groups in these studies were confounded by age. No other studies have reported different effects dependent on body weight. These findings have implications for the potential value of Fabuless™ in the real world, and suggest equal likelihood of effect (or lack of effect) in a variety of individuals and a variety of situations.
The absence of clear findings in some analyses and the heterogeneity or variability between studies found in all but one sensitivity analysis highlight the difficulties of conducting and interpreting meta-analyses where only limited and highly varied studies are available. The meta-analyses here are clearly limited in this respect. It is unfortunate that data from all studies using Fabuless™ were not available. The studies that were unavailable to us for reasons of confidentiality are also likely to have found no differences between treatment and control, resulting in a possible bias in the combined mean differences given here (13). There also remains the possibility that the present analyses have failed to identify a key factor affecting the outcome of these studies on the appetite suppressant effects of Fabuless™. Considering the discussion on processing activities, it is unfortunate that information on the exact manufacturing and processing activities for all studies is not available. Numerous other differences in methodology are also likely to exist between studies. It was not possible to investigate these factors due to a lack of available information.
In investigating effects on appetite, the meta-analyses here are also limited by their focus on test meal intake. Previous studies have found that Fabuless™ reduced ratings of hunger as well as test meal intake (e.g. 5), although not all studies have found effects on hunger. Such results are relevant to the understanding of the usefulness of Fabuless™ for weight management, because, by decreasing hunger Fabuless™ may allow reduced intake to be subjectively more acceptable (16). Hunger ratings were not analyzed here due to limited availability of appropriate data.
With regard to weight management, the present analyses are also limited by their consideration only of short-term laboratory-based studies. Laboratory studies are highly controlled but conducted in artificial settings, so their relevance to the real world and naturalistic eating behaviours is questionable (8,9). Short-term studies also focus on fixed time periods, whereas human dietary intake may vary on an hourly and daily basis, and compensation or counter-compensation following short-term studies is well known (8,9). Short-term studies should ideally be conducted in controlled environments using the controlled measurement of food intake for as long as practical, but the limitations of this approach, particularly in relation to ecological validity must be accepted. Furthermore, while short-term studies are valuable in terms of accuracy and control, the assessment of efficacy in relation to weight control clearly requires longer-term studies.
In conclusion, the combination of these short-term studies suggests a small appetite suppressant (reduced food intake) effect of Fabuless™ compared with control, although considerable heterogeneity or variability between studies was also found. This heterogeneity suggests the combined effects found are variable, difficult to interpret and difficult to predict. This heterogeneity was partly explained by differences between studies in the processed nature of the treatment, treatment dose and in particular year of study, but caution should again be exercised in relation to these conclusions due to the considerable heterogeneity also found in the sensitivity analyses. The only consistent finding was the absence of food intake effects found in studies conducted after 2003.
Conflicts of Interest Statement
This work was funded by DSM, the Netherlands, http://www.dsm.com. There are no other conflicts of interest.
Grateful thanks are extended to the authors and sponsors of unpublished studies for the provision of methodological details and data. This work was funded by DSM, the Netherlands, http://www.dsm.com. DSM identified all known studies, aided with the discussions on the aspects of methodology to be considered, and aided with discussions on potential mechanisms for effects. All contributions from DSM have been acknowledged in the text. Data abstraction and analysis were undertaken solely by the researchers.