Caffeine in Food and Dietary Supplements: Examining Safety: Workshop Summary (2014)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

2 Intake and Exposure to Caffeine1 Given the large and growing number of different sources of caffeine, assessing the level of caffeine exposure is an important and complex first step toward understanding the safety of such exposure. In the Day 1, Ses- sion 1, panel, moderated by Barbara J. Petersen, Ph.D., M.P.H., of Expo- nent, panelists considered different methods being used to assess levels of caffeine exposure in the U.S. general population. Petersen emphasized that different methods yield different results and that the results present- ed here should not be interpreted as final answers. For her, a key ques- tion to consider is how people behave when a source of caffeine disap- pears or a new source appears and whether people substitute one source for another. This chapter summarizes the panelists’ presentations and the discussion that followed. Key points made by each speaker are shown in Box 2-1. CAFFEINE INTAKE FROM BEVERAGES IN THE UNITED STATES Presented by Diane C. Mitchell, M.S., R.D., Pennsylvania State University The International Life Sciences Institute (ILSI) has supported re- search on caffeine since 1983, resulting in several publications (e.g., Knight et al., 2004, 2006). Diane Mitchell presented ILSI’s most recent 1 The words “exposure” and “intake” were used interchangeably throughout the workshop. 11 

12 CAFFEINE IN FOOD AND DIETARY SUPPLEMENTS caffeine study, which was conducted in collaboration with Kantar Worldpanel (Mitchell et al., 2013). The primary objective of the study was to estimate caffeine intakes in the United States from the consump- tion of caffeinated beverages using a current (2010–2011) population- based beverage survey. According to Mitchell, the study was conducted primarily because of the lack of comprehensive, current, and reliable population-based data on caffeine intakes. Most studies examining expo- sure to caffeine in the United States were surveys conducted more than a decade ago, including research sponsored by ILSI. The data provide a current perspective on patterns of beverage caffeine exposure in the United States in the total population as well as in demographic sub- groups. They not only update past ILSI-sponsored research, but they also provide an opportunity to compare trends in beverage caffeine intakes between then and now. BOX 2-1 Key Points Made by Individual Speakers According to Diane Mitchell, most studies examining exposure to caf- feine in the United States were conducted more than 10 years ago. Mitchell described the methods and results of a recent study on caf- feine intakes from beverages in the United States. The study, conducted by the International Life Sciences Institute, was the first population-based study of caffeine intakes from beverages in the United States in more than a decade. The survey included 7 days of online dietary records entered daily by each respondent. Data collection excluded children under 2 years of age. Data collected included type, brand, preparation, location, and amount of all beverages consumed. In sum, mean daily intake was 165 mg per day, compared to 120 mg per day in 1999. Most beverage caffeine intake in the United States comes from coffee and, to a lesser extent, tea and carbonated soft drinks, with energy drinks contributing very little. En- ergy drink consumption was not estimated in the 1999 survey. In Mitchell’s opinion, the data do not support the notion that the introduction of addi- tional caffeinated beverages into the marketplace has resulted in propor- tionately higher caffeine intakes by consumers. Victor Fulgoni III described four approaches to analyzing caffeine in- take data from the National Health and Nutrition Examination Survey (NHANES): usual intake, intake per consumption event, trends in intake over time, and trends in intake over time by food source. The data cover all foods and beverages, but not dietary supplements, with more than 90 per- cent of caffeine intake coming from beverages. Data were collected from one 24-hour in-person dietary recall, and for a subset of participants a second recall was conducted by phone. For trend data, Fulgoni used only the single in-person intake day. Survey participation was not limited by age. Data collected included all foods and beverages consumed. In sum, 

INTAKE AND EXPOSURE TO CAFFEINE 13 the data demonstrate stable intakes between 2001 and 2010. If anything, some age groups show slight decreases in caffeine intake. With respect to food sources, there has been a small but statistically significant increase in energy drink consumption among 18- to 35-year-olds. Fulgoni noted several limitations to the NHANES survey data, includ- ing the self-report nature of the data, the lag between trends in the mar- ketplace and data available to NHANES, and the small number of energy drink consumers. Both Mitchell and Fulgoni emphasized the age dependency of caf- feine exposure. Methods Mitchell described the study methodology and analysis. Data were collected from the Kantar Worldpanel Beverage Consumption Panel sur- vey. Kantar Worldpanel is a global consumer panel company focused on the continuous measurement and analysis of consumer behaviors. This company has been conducting a continuous annual survey for more than 30 years, targeting U.S. consumers of all ages. About 42,000 respondents were recruited from a panel of about 1 million people, with U.S. Census Bureau data used to ensure that the se- lection of participants was representative of the U.S. population. Sample selection criteria included age, gender, race, income, geographic region, household size, and presence of children in the household. Invitations were sent to English-speaking participants via e-mail and were staggered in batches sent weekly to ensure a balanced sample across all months of the year. Data were collected from the beginning of October 2010 through the end of September 2011. All panel participants were 1 year of age or older, with respondents consuming less than 21 beverages in 7 days excluded from the dataset. Mitchell reported that participants were asked to begin completing their online surveys on a specific day for a total of 7 days. The online surveys required dragging and dropping beverage categories (soda/pop, hot coffee, iced coffee, milk, bottled water, and so forth) into different drinking occasion slots and then entering product details, including the type of beverage, amount consumed, where it was consumed, brand name, and other descriptive information. Beverages consumed both in- side and outside the home were included. In addition, participants were asked to complete a “my info” questionnaire, which contained demo- graphic, lifestyle, attitude, health, and nutrition questions. Height and weight data were also collected, with weight data for children collected 

14 CAFFEINE IN FOOD AND DIETARY SUPPLEMENTS for only 9 months of the 12-month data collection period (i.e., January through September 2011). From this detailed beverage diary, Mitchell and colleagues were able to obtain a consolidated beverage list for each of the beverage categories (i.e., total beverages [i.e., all caffeinated beverages reported in the sur- vey], carbonated soft drinks, energy drinks, energy shots, chocolate drinks, total coffee, total tea) and then, using the lists, develop a data- base. Developing the database required identifying which beverage cate- gories contained beverages with caffeine. Information was obtained from manufacturers where possible, although the researchers also used a num- ber of other resources as well. Mitchell emphasized that no single re- source was comprehensive enough to capture information for all of the beverage and beverage types reported in the survey. Even the large na- tional and research databases do not contain brand-specific data for the variety of beverages reported in the survey. The researchers assigned default values for beverages with no data available, insufficient detail, or no specified brand name. For example, the default value for a regular, brewed, nonspecialty brand of caffeinated coffee was 11.9 mg per fluid ounce. According to Mitchell, the range of default values reflects the var- iability in caffeine among different coffee types, with values ranging from 4.1 to 20.0 mg/fluid ounce for ready-to-drink bottle or canned cof- fee to 46.7 to 62.8 mg/fluid ounce for specialty espresso. Once the database was developed, the researchers consolidated caf- feine beverages into a more manageable set of categories for detailed analysis. The researchers were consistent with previous ILSI work with respect to the types of beverages included in each beverage category with one notable exception: energy drinks and energy shots were included as separate beverage categories. Neither was considered in the 1999 ILSI survey; energy drinks were not introduced into the market until 1997. Although Mitchell and colleagues analyzed total caffeine intakes for all age groups across all categories, because coffee, carbonated soft drinks, tea, and energy drinks contributed approximately 98 percent or more of the caffeine consumed, Mitchell presented data only for those four categories. The analysis covered only caffeinated beverage consumers, that is, consumers who drank 1 or more caffeinated beverages in 7 days, and only consumers 2 years of age and older. The researchers relied on pa- rental reports for children, with two exceptions (i.e., data excluded from the analysis) to control for implausible survey entries: children with body weight data below the 3rd percentile or above the 97th percentile based 

INTAKE AND EXPOSURE TO CAFFEINE 15 on weight for age, using the Centers for Disease Control and Prevention (CDC) growth chart data as a reference, and children with total fluid in- takes greater than two standard deviations above the mean fluid intake within a specific age year. The small sample sizes for some subgroups in some beverage cate- gories created another challenge. This was of particular concern for en- ergy drinks for several age groups and for coffee for children between 2 and 5 years of age. In those cases, sample size was considered too small to provide reliable population estimates. Results Of the more than 42,000 survey respondents, 37,602 (85 percent) were caffeine consumers, that is, individuals who consumed at least one caffeinated beverage over a 7-day period of time (see Table 2-1). By age group, the proportion of the population consuming caffeine ranged from 43.0 percent in children 2 to 5 years of age to almost 100 percent of adults over the age of 65. This trend is consistent with previous reports in the literature, according to Mitchell. Mean daily intake, expressed as both milligrams per day and milli- grams per kilograms of body weight, steadily increased with age up to 65 years, when it fell slightly (see Table 2-1). The 50- to 64-year-old age group showed the highest intake. For the most part, the 90th percentile data follow the same trend. Again, the pattern is similar to what was re- ported almost a decade ago, although the intakes are higher. Mitchell noted that although there were some differences in mean daily intakes between men and women, the differences disappeared after adjusting for body weight. When examined by beverage category, coffee, carbonated soft drinks, tea, and energy drinks together accounted for nearly all caffeine intake (see Table 2-2). As far as which type of beverage contributed the most, it was clearly coffee for all ages combined and for adults. For chil- dren, intakes were distributed fairly equally across coffee, carbonated soft drinks, and tea. Energy drinks contributed very little to beverage caf- feine intakes. 

16 CAFFEINE IN FOOD AND DIETARY SUPPLEMENTS TABLE 2-1 Number of Users and Mean and 90th Percentile Daily Intakes of Total Caffeinated Beverages, by Age Caffeine Intake No. of % Users Age Users (based on total mg/day mg/kg/day Group (sample U.S. mg/day 90th mg/kg/day 90th (years) size) population) mean percentile mean percentile All ages 37,602 85.0 165±1 379.5 2.2±0.0 5.0 2–5 732 43.0 24±2 57.8 1.5±0.2 3.7 6–12 1,768 63.0 37±1 94 1.1±0.0 2.7 13–17 1,772 83.2 83±2 182.9 1.3±0.0 2.9 18–24 1,178 85.8 122±4 285.9 1.7±0.1 3.9 25–34 4,155 87.2 137±2 299.8 1.9±0.0 4.2 35–49 9,128 92.1 199±2 428.1 2.5±0.0 5.4 50–64 12,691 93.3 226±2 467.4 2.9±0.03 5.9 65 and 6,178 99.6 207±2 419.9 2.6±0.03 5.4 older SOURCE: Diane Mitchell. Presented to the Planning Committee for a Work- shop on Potential Health Hazards Associated with Consumption of Caffeine in Food and Dietary Supplements on August 4, 2013. Mitchell mentioned again that chocolate-containing beverages were not included in the data she presented because their contribution to total caf- feine intake was so small, even though about 14 percent of the population surveyed reported consuming them. In the case of energy shots, the propor- tion of consumers in the sample was too low to estimate even when com- bining all ages. Together, chocolate-containing beverages and energy shots contributed less than 1 to 4 percent to total caffeine intake depending on age group. The amount of caffeine intake from those two categories is included in the estimates of the total caffeinated beverage category, however. In order to understand more clearly what drives caffeine intake, Mitchell and colleagues examined caffeine intake within each beverage category for users only (i.e., users of that beverage category) (see Table 2-3). Among all ages, 55 percent of all caffeinated beverage consumers consumed coffee, 63 percent carbonated soft drinks, 53 percent tea, and 4 percent energy drinks. As an example of how to read Table 2-3, within the 50–64-year age group, 71 percent of caffeine users consumed coffee, with an estimated average daily intake among coffee users of 223 mg and a 90th percentile of 452 mg. Table 2-3 also shows where sample size was 

INTAKE AND EXPOSURE TO CAFFEINE 17 TABLE 2-2 Mean Caffeine Intake by Age and Beverage Category Caffeine Intake (mg/day) Total Total Age Group Caffeinated Carbonated Total Energy (Years) Beverages Total Coffee Drinks Total Tea Drinks All ages 165±0.9 105±0.8 28±0.2 28±0.3 2.6±0.1 2–5 24±1.8 6±1.4 7±0.5 9.7±0.8 0.3±0.1 6–12 37±1.2 8±0.8 15±0.5 12±0.6 1.9±0.3 13–17 83±2.2 24±1.6 28±0.8 24±1.1 6.1±0.8 18–24 122±4.2 60±3.6 31±1.2 23±1.3 6.2±0.8 25–34 137±2.2 80±2 32±0.7 21±0.7 3.6±0.3 35–49 199±2.1 126±2.0 38±0.6 32±0.7 2.5±0.2 50–64 226±1.8 159±1.7 28±0.4 37±0.7 0.9±0.2 65 and older 207±2.3 159±2.2 16±0.4 32±0.7 0.9±0.2 NOTE: Total caffeinated beverages include carbonated soft drinks, energy drinks, total coffee, total tea, and a small percentage (1–4 percent depending on age group) from other sources (cocoa and chocolate-containing beverages, ener- gy shots). SOURCE: Diane Mitchell. Presented to the Planning Committee for a Work- shop on Potential Health Hazards Associated with Consumption of Caffeine in Food and Dietary Supplements on August 4, 2013. too low to obtain reliable estimates, which includes coffee among the youngest children and energy drinks among several age groups. Summary of Caffeine Intake in the United States In summary, caffeine intake in the United States comes primarily from four beverage types: coffee, tea, carbonated soft drinks, and energy drinks. In this recent ILSI survey, of those who reported consuming caf- feinated beverages, more than half consumed each of three types of bev- erages (i.e., carbonated soft drinks [63 percent], coffee [55 percent], and tea [53 percent]), reflecting a significant number of caffeine consumers consuming more than one type of beverage. Although consumption of chocolate-containing beverages was high, the caffeine content of such beverages was low and contributed little to total caffeine intake. 

TABLE 2-3 Caffeine Intakes for Users Within Each Beverage Category (mg/day) 18 Carbonated Soft Drink Drinkers Energy Drink Users Coffee Drinkers (n=23,103) (n=22,415) Tea Drinkers (n=20,578) (n=1,202) % of % of % of % of users Mean 90th users Mean users Mean 90th users Mean Age within age daily % within age daily 90th % within daily % within age daily 90th Group group intake intake group intake intake age group intake intake group intake % intake All ages 55 191 394 63 45 102 53 53 126 4 60 144 2–5 7 — — 51 13 34 45 22 56 1 — — 6–12 10 76 179 70 21 51 39 29 69 4 — — 13–17 27 92 224 77 37 80 53 45 104 10 59 133 18–24 46 130 288 70 45 99 54 43 110 9 — — 25–34 53 153 316 68 46 108 52 41 98 6 60 160 35–49 61 209 418 66 57 127 52 61 145 4 63 149 50–64 71 223 452 57 49 109 57 65 153 2 58 126 65 and 82 194 394 45 35 75 56 56 125 1 — — older NOTE: Missing cells indicate where sample sizes were too low to obtain reliable estimates. SOURCE: Diane Mitchell. Presented to the Planning Committee for a Workshop on Potential Health Hazards Associated with Consumption of Caffeine in Food and Dietary Supplements on August 4, 2013. 

INTAKE AND EXPOSURE TO CAFFEINE 19 Mean daily caffeine intake was 165 mg. Intakes were higher than previously reported in the 1999 ILSI-sponsored beverage survey, when mean daily intake was 120 mg. The difference represents about half a cup of coffee or a can of carbonated soft drink. Mitchell cited several possible explanations for the increase. First, there were slightly more caffeine-consuming occasions reported in the more recent survey: 1.8 compared to 1.5. Second, there was a slight increase in the amount of coffee consumed (by fluid ounce) and a decrease in the amount of car- bonated soft drinks consumed, and carbonated soft drinks have less caf- feine. Third, the database values used for the more recent survey reflect higher caffeine values for specialty brand coffees, which may also have contributed. A notable finding of the survey was the low consumption of energy drinks. Energy drinks were relatively new to the marketplace in 1999, and thus their intake was not estimated in the previous survey. So even though some energy drinks contain levels of caffeine similar to those of coffee, consumption of caffeine from energy drinks contributed little to total caffeine intake. Caffeine intakes, in particular from coffee, for children between the ages of 2 and 12 years were higher than previously reported. But to keep that finding in perspective, Mitchell noted that only 9 percent of children between the ages of 2 and 12 years were reported to have consumed coffee. Mean daily caffeine intake at the 90th percentile for all caffeinated beverages and among all ages was 380 mg or 5 mg per kg of body weight, most of which came from coffee. Daily caffeine intake at the 90th percentile for adults 35 years of age and older was slightly higher (420–467 mg/day) than the recommended maximum of 400 mg/day. Mitchell noted that the 400 mg/day threshold is not an official recom- mendation for caffeine in the United States. It is a Health Canada rec- ommended level that is often used as a reference value. In addition, the FDA released two letters in 2012 stating that 400 mg per day was not associated with any adverse health effects and that the 400 mg per day value reflects recommendations set forth by Health Canada. For women of childbearing age (the 18- to 24-year-old and the 25- to 34-year-old age groups), both mean and 90th percentile caffeine intakes were below the <300 mg/day recommended levels for pregnancy. For children younger than 12 years of age and older children aged 13 to 17 years, 90th percen- tile intakes were slightly higher than the recommended 2.5 mg/kg/day. In conclusion, as far as Mitchell was aware, the survey she described was the first population-based study in more than a decade to estimate 

20 CAFFEINE IN FOOD AND DIETARY SUPPLEMENTS caffeine intakes from beverages. The increase in accuracy afforded by the caffeine database developed for this study was a major strength, in Mitchell’s opinion, although it may have contributed to the slightly high- er caffeine intakes than previously reported. Overall, caffeine intakes remain largely driven by coffee consumption and, to a lesser extent, tea and carbonated soft drinks. Energy drink intakes contribute very little. In Mitchell’s opinion, the data do little to support the notion that the intro- duction of additional caffeinated beverages into the marketplace has re- sulted in proportionately higher caffeine intakes by any of the various subpopulations of consumers. VARIOUS ASPECTS OF CAFFEINE INTAKE IN AMERICA: ANALYSIS OF NHANES Presented by Victor Fulgoni III, Ph.D., Nutrition Impact, LLC Victor Fulgoni III described four ways that data from the National Health and Nutrition Examination Survey (NHANES) have been used to assess caffeine intake among Americans: (1) current usual intake of caf- feine; (2) current usual intake of caffeine per consumption event; (3) trends in caffeine intake over the past decade; and (4) food sources of caffeine intake over the past decade. Data from multiple sets of surveys were used: 2001–2010 (N = 42,154) for the trend analyses and 2007–2010 (N = 17,387) for the intake analyses, with individuals under the age of 2 years and pregnant and/or lactating females excluded. Data include caffeine in- take from all foods and beverages, but not dietary supplements. To determine usual intake and usual intake per consumption event, Fulgoni and colleagues used the National Cancer Institute method (Tooze et al., 2010), which allows assessment of multiple days of intake and re- moval of intraperson variation. According to Fulgoni, the method pro- vides a reasonable estimate of habitual chronic intake, which is what usual intake is intended to reflect. For trend data, Fulgoni and colleagues used 1-day intakes regressed over time, with p < 0.01 deemed significant. Fulgoni reiterated what Mitchell had emphasized regarding evidence showing that more than 90 percent of the caffeine intake among Ameri- cans is from beverages, mostly coffee, tea, soda, and energy drinks. For coffee, caffeine levels included in the NHANES database range from 0.4 to 509 mg per reference amounts customarily consumed (RACCs). The 

INTAKE AND EXPOSURE TO CAFFEINE 21 RACC for coffee is usually around 8 ounces. So, the range of caffeine levels is quite high, with the most frequently consumed coffee, that is, a typical brewed coffee, having 95 mg per RACC. For tea, the range is 0 to 48 mg/RACC, which covers all hot and cold tea beverages including zero-caffeine herbal tea. The most commonly consumed tea, a standard cup of leaf tea, has 48 mg/RACC. For soda, the range is 0 to 65 mg/RACC, with the most commonly consumed soda being a cola-type soda containing about 20 mg/RACC. Finally, for energy drinks included in the NHANES database, the range is 45 to 86 mg/RACC, with the most frequently consumed energy drink containing 72 mg/RACC. Usual Intake For caffeine consumption per day, among the total sample popula- tion (N = 17,387), mean intakes ranged from about 25 mg among 2- to 11-year-old children to more than 200 mg in older adults (see Figure 2-1). Among consumers only (N = 13,923), daily consumption ranged from less than 50 mg in 2- to 11-year-old children to more than 250 mg in the 50- to 59-year-old age group (see Figure 2-1). Even at the 90th percentiles of intake, daily intake was only about 50 mg among young children (2 to 11 years) and barely more than 100 mg in adolescents (12 to 17 years) (see Figure 2-2). The highest-consuming age group, those 50 to 59 years of age, had a percentile intake of about 450 mg/day. Among consumers only, with the approximately 4,000 non- caffeine consumers removed from the dataset, the only age group that showed a noticeable difference from the total population 90th percentile intake was the 50- to 59-year-old age group, which jumps up to about 515 mg/day (see Figure 2-2). In sum, Fulgoni noted, for both mean usual intake and at the 90th percentile, caffeine intake is highly age-dependent. Both mean intake and the 90th percentile of intake were lowest in chil- dren 2 to 11 years and adolescents 12 to 17 years and highest in adults 50 to 59 years. 

22 CAFFEINE IN FOOD AND DIETARY SUPPLEMENTS 300 250 mg Caffeine/day 200 150 100 50 0 2–11 12–17 18–29 30–34 35–39 40–49 50–59 60+ Years 300 250 mg Caffeine/day 200 150 100 50 0 2–11 12–17 18–29 30–34 35–39 40–49 50–59 60+ Years FIGURE 2-1 Mean usual intake of caffeine for total sample population (top) and caffeine consumers only (bottom). SOURCE: NHANES. Usual Intake by Consumption Event A consumption event was defined as every time an individual con- sumed a food or beverage with caffeine, whether it was chocolate milk, coffee ice cream, coffee, an energy drink, cola, or something else. Again, both mean and the 90th percentile of usual intake per consumption event 

INTAKE AND EXPOSURE TO CAFFEINE 23 600 500 mg Caffeine/day 400 10th 300 25th 50th 200 75th 100 90th 0 2–11 12–17 18–29 30–34 35–39 40–49 50–59 60+ Years 600 500 mg Caffeine/day 400 10th 300 25th 50th 200 75th 100 90th 0 2–11 12–17 18–29 30–34 35–39 40–49 50–59 60+ Years FIGURE 2-2 The 90th percentiles of usual caffeine intake for total sample pop- ulation (top) and caffeine consumers only (bottom). SOURCE: NHANES. are highly age-dependent. The total population mean per consumption event was 65 mg. But among children 2 to 11 years, mean usual intake per consumption event was about 15 mg per event; among adolescents, it was about 35 mg; and among adults, it was 70 to 80 mg, depending on the age group (see Figure 2-3). At the 90th percentile, again, as with usual intake, among children the usual caffeine intake per consumption 

24 CAFFEINE IN FOOD AND DIETARY SUPPLEMENTS event was low: only about 25 mg, increasing to about 60 mg among ado- lescents and about 140 mg among adults (see Figure 2-3). Trends in Caffeine Intake Trends in caffeine intake data were assessed using single 24-hour recall information collected over time, using data from 2001 to 2010. Most age groups showed flat regression lines over time, meaning that 100 90 80 70 mg, Caffeine 60 50 40 30 20 10 0 2–11 12–17 18–29 30–34 35–39 40–49 50–59 60+ Years 200 180 160 10th mg, Caffeine 140 120 25th 100 50th 80 60 75th 40 90th 20 0 2–11 12–17 18–29 30–34 35–39 40–49 50–59 60+ Years FIGURE 2-3 Mean and percentiles of usual caffeine intake (mg) per consump- tion event. SOURCE: NHANES. 

INTAKE AND EXPOSURE TO CAFFEINE 25 caffeine intake has remained relatively stable over the past decade. The only statistically significant regression coefficients were for the 2- to 11- year-old age group (p < 0.01) and the 35- to 39-year-old age group (p < 0.01). Both of those age group’s regression coefficients were actually negative, suggesting slightly lower intake over time. For the 2- to 11- year-old age group, caffeine intake decreased about 2.5 mg on average per data release (i.e., data are released every 2 years); for the 35- to 39- year-old age group, it decreased about 19 mg on average per data release. In sum, intakes did not change much between 2001 and 2010. If any- thing, there were slight decreases in two age groups. Trends in Caffeine Intake by Food Sources Trends in caffeine intake by food sources were analyzed using the to- tal sample population (thus not just consumers). For the 2- to 11-year-old age group, sodas, the major source of caffeine in that age group, showed a statistically significant downward trend in consumption from 2001 to 2010 (p < 0.01). The other beverage categories (coffee, tea, and energy drinks) showed no change over time. Fulgoni emphasized that one of the limitations of this analysis was the very low sample size for energy drinks. In the 12- to 17-year-old age group, again soda was the major source of caffeine, and, again, it showed a statistically significant decline over time (p < 0.01). Caffeine intake from soda also decreased over time among the 18- to 35-year-old age group (p < 0.01). This is the only age group for which a statistically significant increase in energy drinks was observed (p < 0.01). Fulgoni noted that it was a small increase, but it was statistically significant. In the 51-years-and-older age group, despite be- ing the highest caffeine consumers, researchers saw no change in intake by food source over time. Coffee remains by far the number-one source of caffeine in adults. In sum, sodas are the largest source of caffeine in children 2 to 11 years and in adolescents (12 to 17 years). In both groups, caffeine intake from soda has actually decreased over the past decade. Coffee is the largest source of caffeine in adults. Energy drinks contribute little to the caffeine intake for any age group, although a small increase in intake (1 mg per 2-year data release) was detected for one age group. 

26 CAFFEINE IN FOOD AND DIETARY SUPPLEMENTS Conclusion from the NHANES Analysis With trends in caffeine intake remaining stable, or decreasing in some age groups, but with new sources of caffeine entering the market- place, Fulgoni suggested that new sources of caffeine may be replacing older sources. Studies specifically designed to assess trade-off of caffeine sources would need to be conducted to confirm this suggestion. Fulgoni concluded by remarking that, as is true of any study, there are strengths and limitations to these NHANES analyses. Their strengths are that they are based on a large nationally representative sample of children and adults and that usual intakes were analyzed using sophisti- cated statistical techniques. None of the data were adjusted for body size, although those data are available if necessary. Their limitations are that intake data were self-report; intake data were limited by what was avail- able through NHANES (i.e., 2001–2010, with 2009–2010 the last public- ly released data available); and some age groups had a small number of consumers of energy drinks. PANELIST DISCUSSION WITH THE AUDIENCE Following Fulgoni’s presentation, workshop participants were invit- ed to ask questions of the two panelists. This section summarizes the dis- cussion that occurred. Most questions revolved around data and clarifica- tion of how those data have been analyzed and whether the same data can be analyzed in other ways to address additional questions, such as questions about vulnerable populations. The 90th Percentile The panelists were asked whether the data they presented allow for an examination of exposure among even heavier caffeine consumers— for example, consumers at the 95th or even 99th percentile. If not, are there other data available that can be used to examine exposure among the heaviest caffeine consumers in the population? Fulgoni replied that some intake estimates can be calculated at both the 95th and 99th percen- tiles, depending on gender and age. 

INTAKE AND EXPOSURE TO CAFFEINE 27 Overweight and Obesity There was some discussion about whether recent increases in over- weight and obesity in the United States have implications with respect to the way caffeine is distributed in the body (e.g., in fat cells) and what the most appropriate metric is for evaluating or recommending dosage. Mitchell replied that estimates of caffeine exposure in terms of milligram per kilogram of body weight is probably not the best metric for heavier people if caffeine is not distributed in body fat the same way that it is distributed in lean mass. Fulgoni suggested that one possible solution might be to do what is done with protein recommendations, that is, use measure of height to determine ideal body weight (i.e., based on a 24.9 body mass index, or BMI) and then set a recommended exposure level on that basis. Use of Averages to Estimate Caffeine Intake Some participants expressed concern that “key facts” have been bur- ied under the weight of averages, especially with respect to potentially vulnerable populations. For example, how many exposures of 150 mg, 200 mg, or 400 mg or more per event are occurring among adolescents yearly? And how has the number of those exposures changed over the years as the use of caffeine-containing energy drinks has increased? Mitchell and Fulgoni agreed that such an analysis could be done, for ex- ample, by analyzing consumption events per single day. Still, there might not be enough people consuming energy drinks to examine energy drink exposures in particular, depending on the age group. Fulgoni further re- marked that the analyses he described were conducted before this work- shop was conceived, and thus they did not separately examine pregnant and lactating women. But that group could be analyzed separately, he said. The NHANES Database One audience member asked Fulgoni whether the analyses he pre- sented accounted for the fact that caffeine concentrations in products change over time. Fulgoni explained that all data releases are updated as 

28 CAFFEINE IN FOOD AND DIETARY SUPPLEMENTS necessary and that, in fact, one could use the NHANES database to ana- lyze change in composition over time. Another audience member asked about the accuracy of self-report data, especially with parents self-reporting consumption among their children. Fulgoni repeated that self-report is a limitation of the NHANES data. With respect to parental self-reporting of their children’s consump- tion, parents report food and beverage consumption for children up to the age of 6 years. Between 6 years and 12 years, analysts use a combination of child and parental data. He was not aware of any study validating the NHANES self-report data for those younger ages. Workshop planning committee chair Lynn Goldman commented on the number of new caffeinated products, such as candies and marshmal- lows, that are entering the marketplace and asked whether those types of products will be identifiable in the NHANES database in the future. Fulgoni replied that, although those specific products are not yet in the database, the product forms are. So, for example, marshmallows are in the database, making it possible to identify the age groups most likely to consume marshmallows and modeling intake on that basis. Finally, Fulgoni highlighted the public availability of the NHANES da- tabase and remarked that any of his analyses could be repeated by others. Dietary Supplement Exposure Among Fitness Enthusiasts There was some discussion about caffeine use among individuals consuming sports nutrition supplements and whether that consumption might be skewing results for the intake of caffeine. Regan Bailey ob- served that when she dug into the NHANES database, she found only 17 reports of a caffeine-containing dietary supplement between 2007 and 2010. Fulgoni agreed that there is very little consumption of caffeine- containing dietary supplements. When asked whether the data he pre- sented included Monster Energy consumption, given that Monster Ener- gy was classified as a dietary supplement until recently, he replied “yes.” If the interest is in sports and fitness, Fulgoni pointed out, it would be possible to analyze caffeine intake based on the activity metrics that are included in some of the NHANES datasets. 

INTAKE AND EXPOSURE TO CAFFEINE 29 Energy Drink Marketing Data Speaker John Higgins pointed out that marketing data indicate that 2.5 gallons of energy beverage per person, including babies and children, were consumed in the United States in 2009. He also observed that ener- gy beverages are marketed very heavily to the teenage through 35-year- old age range. He asked, who is drinking all the energy beverages? Ba- bies are clearly not drinking them, so who is? And is it possible that the data Mitchell and others are collecting are not capturing all energy bev- erage consumption events? He said he expected greater energy consump- tion events, given how marketing data indicate a high consumption of en- ergy beverages. Mitchell reiterated that she and her colleagues found very low consumption. She suggested that the limitations of self-report data may have been a factor. Fulgoni observed that 2.5 gallons per year, or 320 ounces, amounts to less than an ounce a day, implying that it is not much. Higgins reiterated that the 2.5 gallons per year is across the total popula- tion and that some people consume more than others. Goldman pointed out the time lag between the NHANES data release and marketing data, noting that the NHANES data are “always a few years behind.” Addition or Substitution of Caffeine Sources? The panelists were asked their thoughts on whether consumers are substituting new sources of caffeine for old sources or consuming multi- ple types of caffeine-containing foods and beverages. Mitchell observed that consumers are “obviously” consuming more than one type of caf- feinated beverage. Although she and her colleagues did not separately examine multiple users—for example, to assess what and when they were consuming—the data are robust and could be analyzed to answer those types of questions. Likewise, with the NHANES database, accord- ing to Fulgoni, its robustness would allow for analyzing consumption of multiple types of beverages. He cautioned, however, that a robust esti- mate of energy drink consumption in children would probably require future oversampling. 

30 CAFFEINE IN FOOD AND DIETARY SUPPLEMENTS REFERENCES Knight, C. A., I. Knight, D. C. Mitchell, and J. E. Zepp. 2004. Beverage caffeine intakes in U.S. consumers and subpopulations of interest: Estimates from the Share of Intake Panel survey. Food and Chemical Toxicology 42:1923– 1930. Knight, C. A., I. Knight, and D. C. Mitchell. 2006. Beverage caffeine intakes in young children in Canada and the U.S. Canadian Journal of Dietetic Prac- tice and Research 67:96–99. Mitchell, D. C., C. A. Knight, J. Hockenberry, R. Teplansky, and T. J. Hartman. 2013. Beverage caffeine intakes in the U.S. Food and Chemical Toxicology S0278-6915(13)00717-5 63:136–142. Tooze, J. A., V. Kipnis, D. W. Buckman, R. J. Carrolee, L. S. Freedman, P. M. Guenther, S. M. Krebs-Smith, A. F. Subar, and K. W. Dodd. 2010. A mixed-effects model approach for estimating the distribution of usual intake of nutrients: The NCI method. Statistical Methods 29(27):2857–2868.