Table 3.

Description of included restriction/elimination studies

Reference	Study aims	Intervention type, comparator and duration	Outcome measurement	Main results
Acute studies
Rolls BJ, 2004 USA	To determine whether the meal energy intake was minimized by consuming one of the salads as a first course, or by omitting the first course completely. Ages: 19-45 yrs (mean 26 yrs) Total n completed = 42 women	Within-subjects crossover design with seven experimental conditions. Once per week for 7 weeks, subjects came to the laboratory to eat lunch (salad and pasta). At six of the lunches, the first course was one of six different versions of a salad that varied in energy density and portion size; subjects were required to consume the salad in full. 1) Energy density 0.33 kcal/g (a- 150 g [equivalent to 1.5 cups] and b- 300 g [equivalent to 3 cups]) 2) Energy density 0.67 kcal/g (a- 150 g and b- 300 g) 3) Energy density 1.33 kcal/g (a- 150 g and b- 300 g) In the control condition, no food was served for the first course.	Weighed food before and after eating	1) Doubling the portion size of the salad (300 g vs. 150 g) reduced intake of the pasta (98 ± 30 kcal, P < 0.0001). 2) Eating the low-energy-dense salad decreased meal intake by 7 % (64 ± 26 kcal) for the small portion and 12 % (107 ± 29 kcal) for the large portion. 3) Consuming the high-energy-dense salad increased meal intake by 8 % (71 ± 27 kcal) for the small portion and 17 % (145 ± 22 kcal) for the large portion). 4) In comparison with having no first course, eating the low-energy-dense salad decreased meal intake by 7 % (64 ± 26 kcal) for the small portion and 12 % (107 ± 29 kcal) for the large portion (P < 0.0001).
Savage BJ, 2012 USA	To examine the effect of varying entrée portions on children’s ad-libitum energy intake of macaroni and cheese and fixed portions of unsweetened applesauce, green beans, and whole-wheat roll served with the entrée Ages: 3-6 yrs Total n completed = 17	Within-subject design with repeated measures. Macaroni and cheese meal and fixed portions of unsweetened applesauce, green beans, and whole-wheat roll served with the entrée (different entrée portions: 100 g, 160 g, 220 g, 280 g, 340 g, 400 g)	Weighed food before and after eating	1) The percentage of the macaroni and cheese eaten decreased significantly as portion size increased (i.e. 95 % of 100 g portion eaten down to 64 % of 400 g portion eaten, P < 0.001). 2) Increasing portion size (i.e. 100-400 g) increased children’s intake of the macaroni and cheese (~100 g up to ~ 250 g, P < 0.01) 3) Increasing portion size decreased the intake of other foods served with the macaroni and cheese such as fruit and vegetables (~180 g other foods consumed with 100 g macaroni and cheese portion, down to ~100 g of other foods consumed with 400 g macaroni and cheese portion, P < 0.0001). 4) There was a 61 % higher energy intake at lunch as portion size increased (P < 0.0001).
Fisher J, 2007 USA	To test the effects of portion size and energy density on children’s food and energy intakes at a meal. Ages: 5-6 yrs Total n completed = 53	2 × 2 within-subject factorial design. Each child was seen in 4 conditions differing only in the portion size (329 kcal/250 g vs. 658 kcal/500 g in the 1.3 kcal/g entrée and 460/250 g vs. 920 kcal/500 g in the 1.8 kcal/g) and energy density (1.3 kcal/g vs. 1.8 kcal/g) of a macaroni and cheese entrée served at a dinner meal.	Weighed food intake	1) Children consumed 33 % more of the entrée in the large portion conditions vs. reference conditions (210 ± 11 g vs. 158 ± 11 g, P < 0.0001). 2) Children consumed 33 % more energy (332 ± 19 kcal vs. 249 ± 19 kcal; P < 0.0001) from the entrées when served either the larger or the more energy-dense entrées vs. reference versions. 3) Total energy intake consumed at the meal was ~15 % higher when the large vs. reference portion entrées were served (548 ± 19 kcal vs. 478 ± 19 kcal, P < 0.001).
Ebeling CB, 2007 USA	To determine whether reducing portion sizes and slowing eating rate, to attenuate gorging, would decrease energy intake, during a fast food meal. Ages: 13-17 yrs (mean 15 yrs) Total n completed = 20	Feeding study with cross-over design. 1) Meal (chicken nuggets, French fries, and cola) presented as 1 large serving at a single time point: (Control A). 2) Same meal as above portioned into 4 smaller servings presented at a single time point (Condition B). 3) Same meal as above portioned into 4 smaller servings presented at 15-minute intervals (Condition C).	Weighed food before and after eating	1) NS in mean ± SEM energy intake between conditions (A: 7350 ± 496 kJ; B: 7287 ± 491 kJ; C: 7333 ± 487 kJ).
Looney SM, 2011 USA	To investigate the impact of portion size and energy density on intake, both grams and kilocalories, of snacks in preschool-aged children. Ages: 2-5 yrs (mean 3.9 yrs) Total n completed = 35	A 2 × 2 crossover design (within-subject factors of portion size and energy density). Small portion size was 150 g (lower-energy dense = 64.5 kcal of apple sauce; higher-energy-dense = 178.5 kcal of chocolate pudding). Large portion size was 300 g (lower-energy-dense = 129 kcal of apple sauce; higher-energy-dense = 357 kcal of chocolate pudding). The same lunch menu used for all intervention days.	1) Weighed food before and after eating 2) Anthropometry	1) More calories were consumed with increasing portion size (small vs. large portion size: 84.2 ± 30.8 kcal vs. 99.0 ± 52.5 kcal, P < 0.05).
Marchiori D, 2011 Belgium	To examine the effect of modifying food-item size of snack foods on subsequent portion and energy intake in an individualized and free-consumption setting. Ages: 18-27 yrs (mean 20 yrs) Total n completed = 33	Randomized between-subjects design with two experimental conditions. 1) Candies left unchanged (10 normal-sized red candies and 10 normal-sized cherry candies). 2) All candies were cut in half (20 half-sized red ribbon candies (2 g each) and 20 half-sized cherry shaped candies (2.5 g each).		1) Participants with the smaller candies consumed the same number of candies vs. those with larger candies (6.2 candies vs. 6.9 candies, P > 0.7). 2) Participants offered the larger candies consumed twice as much in gram weight of candies (30.7 ± 18.2 g vs 16.3 ± 20.3 g, P = 0.04), equivalent to a an increase in nearly 60 kcal (109.04 ± 64.5 kcal vs 49.22 ± 57.2 kcal, respectively).
Marchiori D, 2012 Belgium	To examine the effect of modifying small vs. large cookies on children’s food and caloric intake in a typical and familiar eating environment. Ages: Mean age 9 yrs Total n completed = 77	Between-subjects randomized design with 2 experimental conditions: 1) Half the children received 36 half-sized cookies (3.5 g each) 2) Half the children received 18 normal-sized cookies (7 g each).		1) Children offered the smaller cookies consumed more cookies than children offered larger cookies (14.6 cookies vs. 9.2 cookies, P < 0.001). 2) Mean energy intake from the large cookies was higher than in children consuming the smaller cookies (342 kcal vs. 274 kcal, P < 0.05). 3) Children in the large item condition consumed 25 % more gram weight of cookies vs. children in the small item condition, resulting in an increase of 68 kcal (i.e. 64 kcal vs. 51 kcal, P < 0.05).
Stroebele N, 2009 USA	To determine whether or not the portion-controlled packages of snack foods result in less consumption as compared to larger packages when the amount of food provided was held constant. Ages: Mean age 38 yrs Total n completed = 59	Randomized two-period cross-over design for two 7-day study period, with a 1-week washout period. 1) Standard size packages of 10 different snack choices. 2) 100 kcal packages of 10 different snack choices. Participants asked to take the box home and to consume as much and whenever they would like over a 7-day period. The 100 kcal snack package units ranged from 19-26 g per package whereas the standard size package units ranged from 187-369 g.	Weighed food record	Participants consumed ~187 fewer grams of snacks per week when receiving 100 kcal snack packs vs. standard size packages of snacks (P < 0.0001).

n number of participants, NS not significant, SEM standard error of the mean, yrs years of age