Impact of energy density on energy intake in children and adults: a systematic review and meta-analysis of randomized controlled trials

Study selection and categorization

The literature search process for identification of eligible studies is shown in Fig. 1. Out of 1188 identified studies, 38 RCTs remained for analysis.

Fig. 1figure 1

PRISMA flowchart for study inclusion. *Hand-search via database Ovid representative for Cochrane Library Search Strategy

Summary of study characteristics

A detailed overview of the characteristics for the single trials is presented in Table 1. The characteristics across the studies are given in the text and summarized in Fig. 2.

Table 1 Summarized study characteristics for crossover trialsFig. 2figure 2

Changes in energy intake and food intake after lower energy density (ED) in comparison to higher ED diet across studies. Energy intake, food intake: ↑ intake is higher with lower ED than with higher ED intervention; ↓ intake is lower with lower ED than with higher ED intervention, ↔ no significant differences between lower ED and higher ED intervention; NR not reported

Among the 38 trials, most were conducted in America (n = 25; 65.8%), followed by Europe (n = 12; 31.6%) and Asia (n = 1; 2.6%). These were RCTs published between 1988 and 2020.

Total participants from the eligible trials for quantitative analysis of RCTs were 1831 participants; of which 874 were children or adolescents, whereas 957 were adults.

For the children studies (n = 11, [35,36,37,38,39,40,41,42,43,44,45]), the median age was 4.6 [4.3–8.3] years, covering the ages between 2 and 12 years. Girls represented 53% of the participants. The median BMI percentile was 59 [56.1–68.8], with a range of 42.5–94.5. The majority of the trials investigated the effects on children of normal weight and only two trials included children with overweight in their research. Two studies offered the children a manipulated preload and analyzed the subsequent ad libitum meal, whereas nine studies provided children with a manipulated entrée. All of the preload studies manipulated only one meal per day, in contrast to 33% (n = 3) of entrée studies lasting longer than one meal.

In the studies with focus on adults (n = 27, [46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72]), the median age was 25 [22.5–26.9] years, covering the ages between 21 and 84 years. Women represented 52% of the participants. The general median BMI for adults was 23.2 [22.4–24.0] kg/m2, with a range of 21–34.7 kg/m2. One study examined the effects of ED manipulation on adults with overweight, whereas the remaining studies focused on the effects on participants of normal weight. Five of the 27 studies offered a preload and analysed the subsequent ad libitum meal, whereas 78% of the studies (n = 21) served a manipulated entrée. One study applied both preload and entrée design [52].

As a result, a total of eight studies [39, 40, 48, 51, 52, 59, 60, 72] served a compulsory manipulated preload and measured the food intake of the following unmanipulated ad libitum test meal. A total of 31 studies modified an entrée and measured the ad libitum intake of this entrée. The length of all interventions ranged from 2 to 48 days with a median length of 6 [4–8] days. A washout period was performed between the dietary interventions in each of the 38 crossover studies. The washout periods lasted minimum 2 and maximum 28 days with a median length of 6 [6–7] days. When the effects of the ED manipulation were investigated for only a single and not multiple meals (n = 15), lunch was mostly used as the intervention meal (n = 14), followed by breakfast (n = 8) and dinner (n = 1). There were test foods with a median ED of 1.1 kcal/g [0.80–1.2], ranging from 0.1 to 2.0 kcal/g in the lower ED intervention and a median ED of 1.5 kcal/g [1.10–1.80] with a range of 0.5–4.6 kcal/g in the higher ED intervention. The ED consumed correlates linearly with the ED served (R2lowerED = 0.9181, R2higherED = 0.9494). In 22 studies, in addition to ED manipulation, portion size (n = 12), sensory quality (e.g. viscosity, taste, color or palatability, (n = 7)), other macronutrient compositions (n = 2) or information regarding a manipulation (n = 1) were varied. This resulted in 2 × 2 or 3 × 2 factorial crossover designs, with the ED manipulation supplemented by one or two of the aforementioned manipulations in each case. In most trials, energy intake was the primary endpoint, only 2 studies (5%) considered energy intake as secondary endpoint.

Summary of study outcomes

Overall, the heterogeneity of studies was high with respect to study design, sample size and research question.

Energy intake

Energy intake was compared between the lower ED and higher ED interventions at qualitative and quantitative levels for all 38 studies. The results of the qualitative analysis are presented as an overview in Table 1 and across studies in Fig. 2.

Thirty-seven studies (97%) indicated that energy intake was lower with lower ED than with higher ED intervention. Only one study [38] showed no change in energy intake in children after the lower ED intervention, indicating that the same amount of energy was consumed via both the lower ED and higher ED diets. There were also no differences between participants with normal-weight or overweight/obesity.

For quantitative analysis, the 38 multifactorial crossover studies were split according to their study conditions [32] resulting in 71 effects. The result of the quantitative analysis is presented as a forest plot in Fig. 3. Energy intake was reduced in the lower ED relative to higher ED conditions (mean energy intake difference – 223 kcal (95% CI: – 259.7, – 186.0); p < 0.001). However, the heterogeneity was high with I2 = 97% despite the applied random effect model.

Fig. 3figure 3

Quantitative analysis of energy intake of all randomized controlled trials receiving either lower energy density (ED) or higher ED meals. The forest plot displays effect estimates and 95% confidence intervals (CI) for individual studies and the summary of findings. Additionally, for each study mean energy intake [kcal], standard deviation (SD) [kcal] and the number of total participants of both lower ED and higher ED conditions are presented. IV inverse-variance

To investigate the sources of heterogeneity, subgroup analyses were performed according to participants’ age (subgroup analysis 1), meal type (subgroup analysis 2) and intervention length per day (subgroup analysis 3).

Subgroup analysis 1: effects of participants’ age (children versus adults) on energy intake

This analysis aimed at reducing heterogeneity by dividing the studies according to the age of the participants (Figure S1). Although heterogeneity of adult studies remained high (I2 = 94%), was still reduced in the lower ED relative to higher ED interventions (mean energy intake difference – 302 kcal (95% CI: – 358.9, – 246.4); p < 0.001). In contrast, for trials with children heterogeneity was reduced (I2 = 80%) and accompanied by a drop in efficacy (mean energy intake difference – 65 kcal (95% CI: – 83.5, – 47.0); p < 0.001).

Subgroup analysis 2: effects of meal type (preload versus entrée design) on energy intake

In the analysis examining preload versus entrée studies (Figure S2), lower ED conditions were associated with a reduction in energy intake (mean energy intake difference – 111 kcal (95% CI: – 159.2, – 62.5); p < 0.001) and manipulated entrées (mean energy intake difference – 261 kcal (95% CI: – 304.6, – 217.9); p < 0.001), although treatment effects were significantly greater for manipulated entrées than manipulated preloads (p < 0.001). Heterogeneity decreased when analyzing preload studies (I2 = 66%), but remained high for entrée studies (I2 = 98%).

Subgroup analysis 3: effects of intervention length (1 meal versus > 1 meal) on energy intake

Lastly, this analysis distinguished according to length of intervention period (Figure S3). Effectiveness of the multiple meal interventions was superior to single meal interventions (p < 0.001), indicating a persistence effect of the ED manipulation. However, heterogeneity remained high for both single interventions (I2 = 97%) and multiple interventions (I2 = 92%).

Combination of subgroup analysis 1 + 2: effects of age and meal type on energy intake

Heterogeneity decreased when subgroups 1 and 2 were combined for analysis (Fig. 4). Here, heterogeneity decreased slightly when analyzing only children/entrée studies (I2 = 83%), but dropped strongly for children/preload interventions (I2 = 0%). Nevertheless, no significant subgroup effect for preloads was found (p = 0.07). A reduced heterogeneity was found when analyzing adult/preload studies (I2 = 42%) but adult/entrée studies remained high in their heterogeneity (I2 = 95%). Subgroup differences were significant (p < 0.001). Energy intake was lower in all subgroups in the lower ED relative to the higher ED intervention with mean energy intake differences of – 69, – 37, – 374 and – 139 kcal for children/entrée, children/preload, adults/entrée, and adults/preload studies, respectively. No further subgroup analysis could reduce heterogeneity, which is why no further analyses are mentioned.

Fig. 4figure 4

Quantitative analysis of age (children versus adults) and meal type (preload versus entrée) on energy intake of randomized controlled crossover trials in humans receiving either lower energy density (ED) or higher ED diets. The forest plot displays effect estimates and 95% confidence intervals (CI) for individual studies and the summary of findings. Additionally, for each study mean energy intake [kcal], standard deviation (SD) [kcal] and the number of total participants of both lower ED and higher ED conditions are presented. IV inverse-variance

Overall, the subgroup analyses were able to explain the heterogeneity of studies in the full analyses. The data of the RCTs clearly demonstrated that the lower ED intervention reduced the energy intake compared to the higher ED intervention.

Food intake

To improve understanding of the findings regarding the outcome of energy intake in lower ED versus higher ED interventions, the amount of food intake in both conditions is presented in the following. The results regarding food intake for the single studies are presented at qualitative level for 38 studies as an overview in Table 1 and across studies in Fig. 2.

Twenty-five studies showed no significant difference in food intake between the two interventions (66%), nine studies reported an increase in food intake after a lower ED meal in comparison to a higher ED meal (24%) and the remaining studies (n = 4, 10%) did not report on food intake. Except for one study [28], all of the studies with higher food intake in the lower ED than in the higher ED diet were studies in which adult study participants received a manipulated entrée.

For quantitative analysis, 26 studies were included and the results are presented as a forest plot in Fig. 5. Independent of ED manipulation, the amount of food consumed was rather similar between the intervention groups, although in some cases food intake was slightly increased in the lower ED test meals (mean food intake difference 20 g (95% CI: 8.5, 30.6); p < 0.001), meaning marginally more food was eaten in lower ED interventions. The heterogeneity from the trials (I2 = 65%) required no further exploration.

Fig. 5figure 5

Quantitative analysis of food intake of randomized controlled trials receiving either lower energy density (ED) or higher ED meals. The forest plot displays effect estimates and 95% confidence intervals (CI) for individual studies and the summary of findings. Additionally, for each study mean energy intake [kcal], standard deviation (SD) [kcal] and the number of total participants of both lower ED and higher ED conditions are presented. IV inverse-variance

Relationship between delta ED consumed and delta energy intake

Substantial linear relationships between △ consumed ED (lower versus higher ED condition) and △ energy intake were found across different meal types and age (Fig. 6). The linear relationship in children with entrée design and 1 meal per day (A) was stronger (R2 = 0.90) than in adults (R2 = 0.71, B). In adults, the linear relationship became very strong when analyzing the entrée design including more than 1 intervention meal per day (R2 = 0.93; C). Studies with preload design (D) also showed a clear linear relationship (R2 = 0.68; separation between adults and children was not possible due to the small sample size).

Fig. 6figure 6

Relationship between △ energy density (ED) and △ energy intake. Data of △ ED (lower versus higher ED condition of each single study; △ kcal/g) with the corresponding △ energy intake (lower versus higher ED condition of each single study, kcal) are displayed. A: Entrée studies in children, 1 meal interventions. B: Entrée studies in adults, 1 meal interventions. C: Entrée studies in adults, > 1 meal interventions. D: Preload studies in children and adults

Risk of bias

Figure 7 is a risk of bias summary showing the review authors judgements about each risk of bias item for each included study. The overall risk of bias was low in 37 studies and with some concerns in 1 study. None of the studies were identified with a high risk. All of the 38 trials were analyzed per protocol rather than intention-to-treat.

Fig. 7figure 7

Risk of bias. D1 Randomization process, D2 Deviations from the intended interventions, D3 Missing outcome data, D4 Measurement of the outcome, D5 Selection of the reported results. + : Low risk of bias, ! : Some concerns in risk of bias, − : High risk of bias

Only in 4 out of 38 studies (11%) the ED condition was evident to the participant. All other studies tried to conceal the ED condition to the highest possible extent. However, it is unclear if this goal was achieved in the single studies. Excluding the 4 studies (n = 430, [49, 55, 62, 63]) with overt manipulation did not influence the findings and no sub-group differences were observed between overt and covert manipulation (data not shown). Hence, all studies were included in this meta-analysis.

留言 (0)

沒有登入
gif