
Dietary Supplements and HealthRisk Behaviors
Research conducted by: WenBin Chiou, ChaoChin Yang, and ChinSheng Wan
Case study prepared by: Robert F. Houser and Georgette Baghdady Overview
Although the dietarysupplement market in the U.S. is enormous, there is no apparent association between the use of dietary supplements and improved public health. The researchers of this study explored this paradox under the hypothesis that taking dietary supplements triggers a phenomenon called the “licensing effect,” namely, the tendency for positive choices to license subsequent selfindulgent, risky or unhealthful choices. The researchers hypothesized that supplement use confers “perceived health credentials,” leading people to feel invulnerable to health hazards and thus more likely to engage in risky, healthrelated behaviors.
The study involved two experiments. In the first experiment, 82 participants were randomly assigned to either a vitaminpill (multivitamin) group or control (placebo) group and were told the kind of pill they would be taking. However, only the control group was given correct information. In actuality, both groups received the placebo pill. After taking the pills, the participants completed a survey on leisuretime activities, rating the desirability of nine hedonic (pleasurable) activities, such as excessive drinking and wild parties, and nine exercise activities, such as yoga and running, on 7point scales. The survey also included a general invulnerability scale to assess a participant’s perceived invulnerability to harm and disease. After completing the survey, the participants were offered a free lunch, choosing freely between a buffet and a healthful, organic meal.
The second experiment involved different participants. The vitaminpill (multivitamin) group again unknowingly took placebo pills. After completing a questionnaire that included the general invulnerability scale and reading a medical report on the health benefits of walking, the distance participants walked in one hour was measured with a pedometer.
Questions to Answer
Does taking dietary supplements disinhibit unhealthy behaviors, such as eating unhealthful meals? Is the study sufficiently powered to detect significant differences between males and females?
Design Issues
The research was conducted in Taiwan, where cultural attitudes and behaviors related to dietary supplements may differ from those in the U.S. It is possible that the results might not generalize to other countries, so more research is needed. Participants in Experiment 1 had a wide range in age, from 18 to 46 years, with a mean (SD) of 30.9 (7.8) years. It would be helpful to consider age in the analysis, especially if age is associated with invulnerability scores. Leisuretime activities and invulnerability were assessed only postintervention; future studies should also measure these variables before the intervention to see if the two groups had similar scores at the start of the study. The general invulnerability scale used to assess perceived invulnerability to harm and disease has been validated only for adolescents.
Descriptions of Variables
VARIABLE 
DESCRIPTION 
Experimental condition 
Vitaminpill (multivitamin) condition or Control (placebo) condition 
Meal choice 
Either a buffet meal or a healthful, organic meal 
Gender 
The sex of participants 
References 
Chiou, WB, Yang, CC, Wan, CS. (2011). Ironic effects of dietary supplementation: Illusory invulnerability created by taking dietary supplements licenses healthrisk behaviors. Psychological Science, 22, 10811086.
Trout, A. T., Kaufmann, T. J., Kallmes, D. F. (2007). No significant difference … Says who? Editorial. American Journal of Neuroradiology, 28, 195197.

Links Chiou et al. article
The licensing effect
No Significant Difference … Says Who?
Exercises 
Please read the Chiou et al. article before performing the exercises. For all analyses, statistical significance is based on p<0.05.
 In a prior study, 106 participants were asked to rate their perception of the health value of buffet meals versus organic meals on a scale ranging from 1, “a buffet meal is clearly less healthful than an organic meal,” to 7, “a buffet meal is clearly more healthful than an organic meal.” The midpoint of the scale, 4, indicated the perception that buffet and organic meals are equally healthful.
 Perform a t test for a single mean comparing the mean perception rating (M = 2.44, SD = 1.04) to the midpoint of the scale (4), as performed by the authors.
 State the null hypothesis and alternative hypothesis.
 Interpret your findings and report the t and p values.
 For Experiment 1 of this study, Table 1 presents the proportions of buffettype meals chosen in the vitaminpill and control conditions. The sample size of each condition was 41.
 For each condition, calculate by hand the 95% confidence interval for the proportion of buffettype meals chosen. (Don’t forget to apply the correction for continuity.)
 Interpret the 95% confidence intervals that you calculated.
 Construct a 2 X 2 contingency table that examines the relationship between experimental condition (vitaminpill versus control) and meal choice (buffet versus organic). Here N = 82 (41 in each condition).
 Perform a Pearson Chi Square test. Is the relationship statistically significant?
 Interpret your findings and report the relevant percentages, Chi Square value, and p value.
 In Experiment 1, 49% of the female participants and 64% of the male participants chose to eat the buffet meal over the organic meal. Construct a 2 X 2 contingency table that examines the relationship between gender and meal choice. Here N = 82 (45 women and 37 men).
 Perform a Pearson Chi Square test. State the null hypothesis and alternative hypothesis.
 Interpret your findings and report the relevant percentages, Chi Square value, and p value.
 Statistical power is an important consideration in the design of a study.
 Define power.
 What are five ways to increase the power of a study?
 Which of these ways might be appropriate for increasing the likelihood of detecting significant gender differences in this study, if indeed gender differences really do exist?
 Experiment with larger sample sizes in Exercise #4 while keeping the proportions of females and males who chose the buffet meal unchanged; for example, you could double, triple, etc. the cell counts.
 Approximately what sample size (N) is needed to reject the null hypothesis at the 0.05 level?
 Approximately what sample size (N) is needed to reject the null hypothesis at the 0.01 level?
 Suggest a possible explanation other than the licensing effect and perceived invulnerability that could account for a lack of association between the use of dietary supplements and improved public health.

