Portion Sizes and Obesity - The ResearchSave this online in Del.icio.us. [?] Vote For this Post
Here's the research on portion sizes and obesity my partner, Matt Rosenberg, and I did. What follows is an annotated bibliography of the studies we looked at. It was done in the context of a research methods course, in preparation for our own survey. Pardon the erratic spacing. Tomorrow I'll post the Literature Review Assignment itself, which goes more into depth on the variables, and discusses what we planned to do in our own survey. (I won't post our survey, because it was done halphazardly and has no scientific value. The studies we researched were done by professionals, however.)
Levitsky, D. A. & Youn, T. (2004). The more food young adults are served, the more they overeat. The Journal of Nutrition, 134 (10), 2546-2550.
Levitsky and Youn’s research focuses on nutrition, as evidenced by their affiliations. These are, respectively, Department of Nutrition and Psychology, and Department of Nutritional Sciences, both at Cornell University. The research in this study, to be specific, covers obesity and its potential causes, namely eating more. The research question therefore, is whether a causal relationship can be demonstrated between eating more and becoming obese.
In their research, Mr. Levitsky and Mrs. Youn sought to prove a hypothesis put forward by fellow nutritional researchers Young and Nestle. Young and Nestle, based on statistics showing a correlation between larger portion sizes and an increase in obesity, concluded that the former caused the latter. As proper research methodology dictates, the manner in which a causal link can be proven is an experiment, and this is just what Levitsky and Youn did. In other words, the researchers’ goal was to provide experimental evidence to support the hypothesis that larger portion sizes are a cause of obesity.
Thirteen subjects were employed. They consisted of undergraduates and staff at the University who were recruited by means of flyers and class announcements. In addition to this, candidates were filtered in the selection process by checking for food allergies and for having a standard level of self-restraint according to the Stunkard scale.
The variables which were examined were food intake, portion size and corresponding weight gain. The variables were operationally defined as follows (own words).
Consumption of food was defined as the average amount consumed by each subject of each food. To establish the baseline of consumption therefore, Levitsky and Youn first found the mean of each food eaten by each subject over the course of the first week of their experiment. This was measured by calculating the weight of their plate of food before (with the food in it) and after (with the food they ate in their stomachs) they were done eating.
Then, this information was used for establishing larger portion sizes in week 2 of the study. Portion sizes were therefore defined as the quantity of the food that was consumed when served in a buffet and eaten at participants’ own discretion.
Weight gain simply meant having a higher body weight in kilograms as a result of food consumption. The weight gain would then be compared between weeks 1 and 2 of the study to establish if subjects gained more weight with larger portions sizes than with regular portions.
Once subjects were selected and variables defined, the experiment proceeded as follows. In week one, subjects were served 4 foods (rigatoni pasta and tomato sauce, vegetable soup, breadsticks, and ice cream) in a buffet. They were allowed to choose whatever they wanted and in whatever quantities. Water was consumed without being measured. The solid foods were measured as stated above. At the end of the week, the average intake of food was used as a starting point for determining larger portion sizes. Once these were calculated, the study’s participants were served the larger portions in week 2.
They were weighed during what the author calls ‘the period of testing’ or alternatively ‘test day’. Given that the researchers calculated the average intake of food each week, that the study was carried out with 3 lunches a week (i.e. the subjects participated in the study three times a week – they didn’t skip lunch 4 days a week), and that ‘test day’ was referred to say that there was no interaction observed between portion sizes and ‘test day’, we can surmise that the weigh-in day was Friday (when they would calculate the weekly average and probably also weekly weight gain). It only makes sense to do the weighing at the end of the week, after subjects have done the eating.
As regards physical activity, participants were asked to record what they’d done before lunch. They were also asked to maintain the same level of physical activity throughout the experiment. Another test for confounding variables was applied in the form of an ‘hunger-rating’ test before and after they had lunch.
In complicated technical jargon which will not be repeated herein, the researchers noted that there were no anomalies in food consumption, as a result of boredom due to eating the same foods repeatedly, for example.
The jargon perseveres to explain that significant discrepancies were observed between the amounts of food consumed in the first week and the second week. In the one sentence that seemed clear of jargon, Levitsky and Youn stated their findings bluntly and perfectly: the ‘greater the amount of food subjects were served, the more they consumed…’
The hypothesis clearly was supported. Larger portion sizes obviously resulted in higher food consumption. One can assume that with a constant rate of activity, participants in the study would likely have gained weight as a result of larger portion sizes. Larger portion sizes can therefore be asserted to contribute to obesity.
There are, of course, caveats. One of these is the fact that the study had a tiny sample of people, who are not generally representative of the American obese population. It is incorrect to say that 100% of obese Americans went to University, let alone Cornell. Therefore, demographically, the sample is unrepresentative. It is incorrect to assert that there is a ratio of 9 men to each 4 women, in the United States (2.25:1). It is incorrect to say that the methods of recruitment used would obtain a representative sample of obese Americans, given that not all obese Americans might respond to flyers and class announcements.
Another major caveat is that only 1/7th of the participants’ meals in the two weeks (assuming 3 meals a day, 7 days a week) were taken into account. Perhaps participants ate less at home because they felt fuller. Perhaps they had big exams during the time-frame of the experiment and drank lots of coffee and soda to stay up late in order to study.
Nevertheless, the researchers, probably working with limited financial resources, made a reasonable attempt to control extraneous variables (asking to maintain constant activity, hunger tests, etc.) and should be commended for this. Their study did provide experimental evidence to support Young and Nestle’s suggestion that larger portion sizes were contributing to obesity, but this evidence should be taken with a grain of salt.
Smiciklas-Wright, H. & Mitchell, D. C. & Mickle, S. J. & Goldman, J. D. & Cook, Annetta (2003). Foods commonly eaten in the United States, 1989-1991 and 1994-1996: Are portion sizes changing? Journal of the American Dietetic Association. 103 (1), 41-47.
The research in this study is about the change over time of American portion sizes. The researchers wanted to determine, on a statistical basis, whether or not Americans were consuming different sized portions, and what foods this concerned, if any. Therefore, with the help of the US Department of Agriculture (USDA), the researchers compared food consumption over the years 1989-1991 to the years 1994-1996. There was no specific hypothesis stated, though one can surmise that they expected to find a difference in portion sizes, else they would not have spent their time carrying out the research.
The study looked at individuals ages 2 and older who answered all the requests for information in the USDA’s surveys (CSFII). The samples broke down into 10 age and sex groups. They are: males and females 2-5 years of age; males and females 6-11 years of age, males 12-19 years of age, females 12-19 years of age, males 20-39 years of age, females 20-39 years of age, males 40-59 years of age, females 40-59 years of age, males 60 and older, and finally females 60 and older. The group sizes varied between 672 and 2042, with numbers generally averaging in the low 1000s (e.g. 1200~). Overall, there were 11,488 people in CSFII 89-91’s sample, and 14,262 people in CSFII 94-96’s sample. Breastfed children were excluded. Participants were selected to be nationally representative.
The variables measured in this study were portion sizes, defined as “amounts consumed per eating occasion”, and the amounts of food consumed per day. An eating occasion was defined by the time of day, as opposed to its name.
The study involved surveys. The article describing the study only mentioned that they had been extensively described previously, however. This had been done by two of the researcher’s references, both subdivisions of the USDA.
What is most valuable to know about the study is that it was evaluated for reliability (i.e. its estimates/results). The criteria for evaluation were the guidelines the USDA put out. Given the source of the guidelines, we can assume that any evaluation is accurate. In this case, it is positive; the study’s estimates were seen to be reliable.
The study found significant differences in portion sizes. For most of the foods with significant differences between 89-91 and 94-96, portions were larger in the mid 90s. Many were grains/cereals and soft drinks.
The last paragraph in the Results section, on trends, seemed convoluted and, at times, self-contradictory. For example, no “food showed differences for all of the age and sex groups in this study.” This is followed later with “the significant differences for a given food were in the same direction for all age and sex groups.” Even after several re-readings, that section of the report did not appear any clearer. What was interesting was that portion sizes for beer were larger for all males 40 and over. This seems to suggest the effect of marketing on consumption, given that much beer is advertised towards a target market of older men.
The ‘hypothesis’ was supported, in that a variety of significant changes were found in portion sizes. Most of these tended upwards, but portions for some things, such as pizza, shrunk.
Overall, the researchers concluded that there were a variety of implications for the health of Americans in their findings. Firstly, the trend towards overweight seen in recent years seems to have found a possible cause, namely, larger portion sizes. These have been shown by some of the researchers references to be positively correlated to energy intake which is correlated to overweight. Secondly, growth in beer consumption seems to have been scientifically related to beer bellies, technically referred to as “larger waist-hip ratio”. This was shown for whites and blacks of both genders. Thirdly, a recently established relation between higher caffeine intake and higher bone loss in the spine for postmenopausal women is even more worrying now that larger portion sizes of beverages with caffeine in them have been reported.
A few limitations were reported by the researchers for their survey. There were some slight differences in methodology between 89-91 and 94-96, though these were stated to not impact the data on portion sizes. The reporting itself can be affected by inaccuracies. A few minor changes over the years for different foods including bananas, chicken, and macaroni and cheese may also have had an impact. Overall however, the problems with this study seem minor and almost negligible.
Our SurveyVariables we might consider include number of items in people’s lunches, how filling each of these is (on a scale of one to ten), what they eat (tick off from a list), whether or not they buy lunch or other meals, the frequency with which this occurs, and perhaps what are their three most frequented restaurants (including fast-food places). Based on these things, we might be able to learn how much people are eating in absolute terms (e.g. quantity, and how filling those things are), we can learn what sorts of food are being eaten (e.g. McTrio ... easily accessible, high-calorie…) and determine to what extent their food providers are determining how much they eat at any particular meal (based on what they’re served).