Other than certain nutrients, probably no other food ingredient has been more heavily studied than caffeine. Yet, as illustrated throughout this report, a wealth of unanswered questions remains about exposure to caffeine in food and dietary supplements and the health consequences of that exposure, especially in certain potentially vulnerable populations (e.g., children, adolescents, individuals with underlying genetic susceptibilities, pregnant women). Indeed, this workshop was convened in part to identify major data gaps.
Throughout the workshop, individual speakers mentioned relevant data gaps. Jennifer Temple, in Chapter 6, thought the current data on energy drink usage among children and adolescents was out of date. Furthermore, she identified a need for prospective research relating the early use of caffeine to subsequent substance abuse. She also asked for research on the long-term effects of caffeine use, particularly among vulnerable groups. Also in Chapter 6, Amelia Arria noted a lack of valid assessment methods for energy drink consumption as a data gap, and she reinforced her comments in the final workshop session.
In the final session of the workshop, panelists identified what they each considered the most important data gaps and considered ways to fill those gaps. This chapter summarizes the panelists’ presentations and the discussion that followed. Note that here, as throughout this report, the observations and opinions expressed are those of individual workshop participants.
Before the panelists spoke, moderator Joseph V. Rodricks, Ph.D., Environ, categorized potential data gaps into five major categories: (1) acute and chronic adverse health effects of caffeine; (2) dose–response issues requiring further study; (3) population variability and how dose–response relationships vary among populations; (4) interactions among ingredients
in products that contain caffeine; and (5) intake and exposure issues requiring further study. Box 9-1 describes key points made by the speakers.
• Amelia Arria suggested that more systematic data collection is needed to better understand patterns of caffeine use over time. Arria suggested that the variable methods used to assess caffeine exposure may partially explain some surveys’ disparate findings (e.g., different findings regarding substitution versus addition of different sources of caffeine). She also suggested developing a more proactive approach to evaluating the magnitude of health consequences associated with caffeine use and developing screening tools that busy physicians can use to estimate the proportion of cardiovascular cases potentially attributable to caffeinated energy drink use.
• According to Alvin Bronstein, the national poison call center database lends itself to more systematic data collection and analysis. Acknowledging the nuanced nature of the data, Bronstein encouraged workshop participants to consider the many untapped ways that the data could be used. He also suggested that educating the public about the existence of poison call centers and that changing the name of the centers by replacing the word “poison” with a less alarming word might help to alleviate underreporting.
• Regan Bailey raised another semantics issue, that is, the use of the term “energy drink.” She noted that many energy products in the dietary supplement label database do not contain caffeine. She urged that the words “caffeine” and “energy” not be used synonymously when discussing energy drinks or other energy products.
• Bailey identified the lack of data on the amount of caffeine in caffeine-containing foods and dietary supplements as the most critical data gap. Not having those data make it difficult to assess exposure.
• Christina Chambers added that among pregnant women, a critical data gap is exposure levels in the period of time prior to pregnancy recognition. Chambers advocated for cohort studies to help fill that and other data gaps.
• Steven Lipshultz said that when safety signals emerge, it is difficult to know who is at risk, the percentage of individuals at risk, or the severity of the risk. Lipshultz explained how this is true of any safety signal and emphasized the importance of the well-designed clinical study as a means to addressing these questions.
• On the basis of what has been reported thus far in the scientific literature, Stephen Schaffer called for more data on both the acute and chronic effects of combinations of ingredients in caffeinated energy drinks.
Amelia M. Arria, Ph.D.
University of Maryland, College Park
Although caffeine has been widely studied and is well understood, Amelia Arria emphasized the need for more data on new ways that new caffeine-containing products are being consumed. Specifically, she called for estimates of the prevalence of use and possible health hazards associated with that use.
Arria suggested that the varying methods used to assess caffeine exposure may partially explain some surveys’ disparate findings. She observed that data presented during the workshop from the International Life Sciences Institute (ILSI) and National Health and Nutrition Examination Survey (NHANES) analyses did not inform about the proportion of the population who used caffeine in the past month or past year, which are standard items that would be captured in other federally sponsored surveillance systems (see Chapter 2 for a summary of the ILSI and NHANES analyses presented at this workshop). She urged leveraging opportunities from what she identified as the most widely used survey of American school children, the National Institutes of Health (NIH)-sponsored “Monitoring the Future” survey, which began asking about energy drinks in 2010. As far as she was aware, the only publicly reported results from that survey were in a 2011 press release stating that the results indicated recent use among 35 percent of 8th graders and 29 percent of 10th and 12th graders.
In addition, Arria called for a better understanding of the proportion of children with underlying cardiovascular medical conditions who are being exposed to high levels of caffeine. Such an understanding will help to gauge the public health response. In response to physicians in the workshop audience who had indicated at various times during the discussion that none of their patients were attributing their medical complaints to energy drinks, Arria asked in return, “How systematically are you asking about recent consumption of energy drinks when a patient, especially an adolescent patient, presents with cardiovascular symptoms?” Better assessment and screening tools for busy physicians are needed to estimate the true proportion of cases that can be attributable to energy drink use. For Arria, the “bottom line” with respect to data methods is the need to start asking and reporting more systematically by putting forms into the hands of physicians who see patients with cardiovascular complications. It is a classic problem in epidemiologic surveillance, one analo-
gous to parents’ reports of children’s alcohol consumption. Arria said, “They don’t ask. Children don’t tell.”
Regarding the substitution of different sources of caffeine, Arria mentioned data suggesting that, contrary to what was suggested earlier in the workshop, energy drinks in young adults are being consumed in addition to, rather than instead of, traditional caffeinated beverages. Specifically, her own data from approximately one thousand young adults in their fourth year of college showed that 3.4 percent did not use any form of caffeine; 2.6 percent used energy drinks but no coffee, tea, or soda; 31 percent used coffee, soda, or tea but no energy drinks; and 63 percent used both energy drinks and another type of caffeinated beverage (Arria, 2013).
Arria commented on evaluations of the magnitude of health consequences associated with caffeine use. Many evaluations rely on what she described as “tip of the iceberg” datasets that measure only the most severe consequences. Federally sponsored surveillance of emergency department admissions, poison control, and the U.S. Food and Drug Administration’s (FDA’s) Adverse Event Reporting System are some examples, although many people are not even aware that they can call poison control centers or that the FDA has this reporting system. Although these datasets are useful for providing safety signals of what might be below the surface, they hugely underestimate the true proportion of medical complications that might be associated with the consumption of any one substance. A preferable approach to estimating the proportion of individuals experiencing health problems would be to ask consumers proactively about their experiences, an approach akin to the required and rigorous methods for adverse event reporting for pharmaceuticals in clinical trials.
Finally, said Arria, other data gaps include the safety among adolescents of current product formulations involving higher doses; interactions between caffeine and other ingredients, including medications; the use of highly caffeinated beverages before and during exercise, especially with young people; and basic data on the pharmacokinetics of caffeine under different circumstances of consumption, such as hot versus cold, and among individuals with varying sensitivities or genetic differences.
Alvin C. Bronstein, M.D., FACEP
Rocky Mountain Poison Center, Denver, Colorado
Alvin Bronstein discussed how poison centers can help to address some of the many unanswered questions about how these caffeine-
containing products are used, whether any health issues are associated with their use, and which populations are most at risk. He proposed that researchers look more closely at poison center data. He and his colleagues have focused on single exposures, but the poison center database lends itself to mixed exposures as well—for example, when people use a product in combination with ethanol or another drug. Bronstein recognized the nuances of poison center data and the need to interpret them carefully, but he emphasized the systematized way that those data can be gathered.
With respect to underreporting and ways to encourage more people to call poison centers to report issues with products, Bronstein suggested that perhaps the word “poison” needs to be replaced. The public also needs to be educated that such centers are places where one can call to talk to a health care professional.
In addition to a data resource, poison centers also function as a surveillance system for identifying index cases and cases that meet certain syndromic surveillance criteria.
Regan L. Bailey, Ph.D., R.D.
Office of Dietary Supplements, National Institutes of Health, Bethesda, Maryland
The first gap that needs to be filled, in Regan Bailey’s opinion, is to define “energy product.” She observed that “caffeine” and “energy” seem to be used synonymously, but they are not the same. Many energy products in the dietary supplement label database contain no caffeine at all and are simply high-dose B vitamins. Approximately 1,500 products in the dietary supplement label database have energy listed somewhere on the label, and 157 products have “energy” in the product name. Some of those products have caffeine, and others do not.
Another important need pointed out by Bailey is a database on the amount of caffeine in caffeine-containing foods and dietary supplements. Although caffeine must be listed on a label if a product contains caffeine, the amount of caffeine present is not required. Not having that information makes it difficult to assess usual intakes of caffeine and to assess usage patterns.
It is not even clear at this point how best to assess usage, Bailey said. Social desirability issues may impact reporting, for example, with children not telling their parents that they are using caffeinated products or
with people feeling embarrassed that they are using them. Also, beverages are often forgotten foods in reporting on 24-hour recalls or other types of dietary assessment methods. Although NHANES serves as a good tool for monitoring, it might not be the best method for answering some questions. That said, Bailey has used NHANES data to examine energy drink use from two 24-hour recalls. People who had used a product on either of those two occasions were considered “users.” Given that definition, from 2007 to 2010, 2 percent of the U.S. population were users. Energy drink use was most common in males from 19 to 30 years of age, with 5 percent of that population defined as users. The most common product used was Red Bull, providing an average of 154 mg caffeine per day.
Christina Chambers, Ph.D., M.D.
University of California, San Diego
Christina Chambers reiterated that although there is quite a bit of data on low to moderate caffeine consumption in pregnant women, there are less data on high-dose exposure to traditional caffeine-containing products (i.e., coffee, tea, and cola) and no data on newer caffeine-containing products. An important period of time to consider measuring or better measuring the effects of these exposure levels is prior to pregnancy recognition.
Chambers called for improved methods to estimate both acute and chronic exposure. She agreed with Bailey that dietary recalls may not capture everything and wondered about the potential to develop a technology that might be more continuous or at least repeated through the course of pregnancy, or any period of time, to capture other than a snapshot of information.
Another knowledge gap is in the area of health behaviors surrounding caffeine consumption, particularly high levels of consumption, and whether those behaviors are associated with any other risky behaviors. Chambers opined that a cohort approach would be the optimal way to collect that type of information. One of the major benefits of the survey design of the National Children’s Study, as it was originally proposed, is its ability to collect this kind of information on a very large sample of women and their children over the course of a pregnancy.
Steven E. Lipshultz, M.D.
University of Miami, Florida
After spending 35 years looking at first-generation survivors of many formerly fatal illnesses of early childhood and seeing late effects after transient early exposures to chemotherapy, Steven Lipshultz has learned that he and his colleagues do not really know when safety signals come up, who is at risk, what percentage of their patients are at risk, or the degree of severity. For example, about 1 percent of childhood cancer survivors experience acute heart failure, which is the leading late effect for early exposure, especially for those who were exposed early in life. Physicians are seeing the same effect with early-in-life exposure to antiretroviral therapy to block transmission. Radiation exposure is another type of exposure with long-term consequences. Lipshultz and colleagues recently published a paper on the follow-up of 18 million patients who had received radiation and found a strong signal for heart disease.
It is a recurring theme, Lipshultz observed, that exposures have long-term consequences. Still, even when a safety signal, or something that might be a safety signal, is detected in a vulnerable population, such as in children, it may be only the tip of the iceberg. Lipshultz said, “It might be true. It might not. It might be important. It might not.”
An appropriate study design is essential for this type of long-term study, given that one cannot think a priori of all the questions that might come up. His approach is to let the patients direct what is necessary. He usually approaches the design in three ways. First, he gains an understanding of the clinical course. In order to gain continuous funding for a lifetime-exposure-effects study, one needs unambiguous quantitative subclinical data. Second, he identifies risk factors, whether those are whole exosome sequences, biomarkers, or something else. Third, as part of the longitudinal study, he evaluates whether any of what he had initially identified as theoretically vulnerable populations are truly of concern.
Lipshultz observed that many of his colleagues make recommendations on the basis of clinical impression. For him, the question is, “How do we go from clinical impression to understanding where the truth is?” The answer: only by conducting a longitudinal study and determining whether any of the risk biomarkers are truly predictive of outcome based on a higher incidence of subclinical or, later, clinical issues.
Stephen Schaffer, Ph.D.
University of South Alabama, Mobile
When looking through the scientific literature on ingredients in caffeinated energy drinks, Stephen Schaffer was struck by how little literature there was. When attempting to define the interactions of the various ingredients in such drinks, he found basically no information at either the animal or human level about the effects of such interactions on either cardiovascular or central nervous system physiology. The clinical studies that have been reported were not conducted properly, in his opinion, with very few randomized, double-blind, well-controlled studies.
In addition, he emphasized the need to study both acute and chronic exposure, given that chronic exposure can alter gene regulation and that responses to chronic exposure can be very different from responses to acute exposure. He also urged consideration of how energy drinks affect energy metabolism, blood glucose levels, and insulin secretion. Finally, many energy drink ingredients are antioxidants. Although antioxidants are generally perceived as good, some scientists now argue that too many antioxidants can be problematic. In sum, Schaffer said, “There’s a lot of work to be done.”
In addition to better defining the science, Schaffer called for defining what a vulnerable population is. What is the evidence that a vulnerable group is vulnerable? “If you can’t provide the scientific evidence [that] they are vulnerable groups,” he said, “you don’t have vulnerable groups.”
In the final panelist discussion with the audience, most questions asked of the panelists revolved around the differences between a longitudinal follow-up study and a registry, both of which were suggested at various times over the course of the workshop; the pros and cons of taking the sort of approach that Health Canada has taken with respect to setting safety standards for potentially vulnerable populations; other types of potential safety signals worth evaluating (i.e., besides sudden cardiac death); the potential value of data from military studies on caffeine exposure; and the potential value of industry data and the availability of such data. This section summarizes the discussion that took place.
What Next?: Longitudinal Follow-Up Study, Registry, or Something Else?
Lipshultz and Bronstein were asked to clarify the different roles that a longitudinal study and registry would serve with respect to tracking and analyzing potential safety signals associated with caffeine-containing foods and dietary supplements. Lipshultz replied that the two types of research are very different.
Generally, registries provide a means to understand the course of exposure and characterize the exposed population. That information can then be used to formulate any of a number of hypotheses about that exposure. Lipshultz noted that, in addition to designing and conducting longitudinal studies, he has been the principal investigator of the National Heart, Lung, and Blood Institute (NHLBI) Pediatric Cardiomyopathy Registry for more than three decades. The registry, which was set up as a way to collect more consistent outcomes for children with cardiomyopathies, helped researchers better understand the course of outcomes and identify potential risk factors. Lipshultz described it as “a very useful platform.” Later, NHLBI decided to also fund a biological specimen laboratory, which the researchers have been using to see whether some patients are more genetically susceptibility to poor outcomes. He and his colleagues recently published a paper demonstrating a ninefold increased rate of dead heart muscle with certain exposures. Lynn Goldman noted that an additional value of registries is that their large size allows for rare effects to be studied.
Lipshultz explained that longitudinal cohort studies, on the other hand, are more hypothesis-driven. They are more limited in scope, with very select inclusion and exclusion criteria based on what is known about likely risk. They are the type of study conducted when a safety signal exists and needs to be tested.
Either way, according to Lipshultz, whether one embarks with a registry or longitudinal study, 5 years from now the field is going to be in a much better position to answer many of the questions being put forth at this workshop. With a cohort study, some questions will be answered. With a registry, exposed patients will be better characterized.
He emphasized the lack of consensus around safety concerns associated with caffeine-containing products. He said, “Your degree of concern varies based on who you are in this room.” Nonetheless, there are concerns. Given those concerns, in his opinion the next step is to evaluate those concerns, whether through a registry or a longitudinal cohort study. They are both potentially valuable designs.
Bronstein clarified that he had not recommended a registry per se. Rather, he was recommending that poison center data be used to answer many of the questions being put forth at this workshop—for example, questions about caffeine consumption in combination with other ingredients or caffeinated product consumption in combination with other products. In addition to reporting exposures, the poison center’s current surveillance system could be expanded from tracking exposures in real time to evaluating exposures on a weekly or even daily basis. Trends could be identified, and cases that meet certain public health criteria could be investigated. As an example of how poison center data have been used in the past, he mentioned a product called Total Body Formula, a dietary supplement product that contained excess selenium, which caused, among other effects, nail changes and hair loss. The FDA removed the product from the market. Poison centers worked with the FDA and the CDC to locate the approximately 160 cases called to poison centers so that public health departments could investigate the cases. He reiterated that the public needs to be better educated about poison centers.
Goldman noted that many people for many years have believed that the poison center reporting/surveillance system could be improved with more complete case follow-up, not just for caffeine- and energy drink–related calls but for many other types of calls as well. Goldman noted the difficulty in interpreting a lot of poison center data because of the lack of data on outcomes.
Health Canada’s Approach to Setting Safety Standards for Caffeine Exposure
Given so many uncertainties in the science of the safety of caffeine-containing foods and beverages, Health Canada, in Goldman’s opinion, seems to have taken what she described as “kind of a simple approach” by advising healthy people to consume no more than 400 mg per day, pregnant women to consume no more than 300 mg per day, and children to consume no more than 2.5 mg per kg per day. Although this advice is pragmatic in terms of what is known and what is not known about safe levels of caffeine exposure, is it an appropriate approach?
Steven Lipshultz replied that the levels are based on the best available population data. In his opinion, although population data and population-based recommendations are important, this approach may not be an appropriate one for specific vulnerable groups, whether those groups are
defined as such on the basis of research or clinician concern. When potentially vulnerable populations are identified, the typical pediatric approach is to determine safe levels in those populations and, at the same time, avoid exposure if there is no real therapeutic efficacy associated with such exposure. Rodricks clarified that the Health Canada recommendation is not a single standard and that it includes two recognized vulnerable populations, women of childbearing age and children. Even so, Lipshultz said, among children there are potentially vulnerable subpopulations—for example, children with cardiomyopathies and arrhythmias. For those children, is the standard for all children appropriate? He said, “I bet you would find almost no pediatric cardiologist that would buy into that concept. If you think there are vulnerable populations, it behooves you to either protect them, test it, or do both.”
While agreeing with Lipshultz, Arria opined that recommendations are among the weakest public health responses and that more effective risk management will require a stronger response. People do things other than what is recommended. Chambers agreed, noting that most women interpret “reproductive age” as “when I am pregnant.” Not only does the evidence need to support that all women of reproductive age, pregnant or not, are at increased risk for adverse outcomes; another issue to consider is whether the adverse outcome(s) of concern even matters to individual patients. For example, some women and health care providers may not consider the risk of delivering a baby 50 grams lighter than the infant would be otherwise to be an important risk.
For Schaffer, the problem (with making recommendations for vulnerable populations) is that “the science isn’t there.” When someone dies a sudden death or develops an arrhythmia, one can look back in history and determine that the patient consumed some energy drinks. But there are no double-blind, randomized studies where different groups of individuals consume different combinations of energy drinks or other products and an end point is examined. “That’s the problem,” he said. “Until you define clearly what the vulnerable groups are and under what conditions, then I think it’s very difficult to make recommendations.” Rodricks replied that the Health Canada recommendations are in fact based on a large number of studies. Schaffer opined that more studies need to be done. He said, “Carefully controlled studies need to be done to define what a vulnerable group is—if there is in fact a vulnerable group and under what conditions.”
Arria agreed that if the ability existed to do those randomized controlled trials, questions would be answered. But conducting such trials
would be difficult given ethical concerns about administering doses similar to what are being consumed in the natural population, particularly doses being consumed by adolescents and young adults. She said, “I don’t think you could get it through an investigative review board (IRB) to randomly assign children and adolescents and even young adults to high doses of caffeine.” The type of naturalistic cohort study described by Lipshultz, on the other hand, where high doses are not administered but rather natural levels of consumption are measured, would be advantageous in that regard.
Other Potential Safety Signals
In Roland Griffiths’s opinion, one of the safety signals for which the most data have been produced is physical dependence. While many other adverse effect signals are difficult to study, with ethical constraints limiting prospective research, physical dependence has not been similarly constrained. Researchers have learned quite a bit about physical dependence from prospective studies. Although it is not as serious as sudden death, neither is it a trivial issue. An exposure level as low as 100 mg per day is enough to produce headache. As a parent, Griffiths said that he would not want his children to be erratically going in and out of caffeine withdrawal while attending school and trying to learn. It is known that a withdrawal signal exists in doses that are relatively low, low enough to be detectable in a single can of many energy drinks. He suggested that further consideration be given to restraining the promotion of those products to school-age children.
The Value and Availability of Military Data
A member of the audience asked whether any data being collected by the military might be helpful, given that both energy drink and caffeine consumption in the military are quite high and that an Institute of Medicine Committee on Military Nutrition Research has issued a few reports on caffeine consumption over the years. There was some agreement among the panelists that this was a good suggestion. Chambers noted that the U.S. Millennium Cohort Study is following families with active-duty military family members and that it would be a good way to study exposure to caffeine-containing products.
The Value and Availability of Industry Data
Although some physicians have not seen the signals that others are seeing, John Higgins observed that, nonetheless, signals are being seen and should not be ignored. At the very least, more information should be gathered. Are the signals real, or are they an artifact? He called for more research, specifically research on caffeine-containing energy drinks. He said, “The signals we are hearing are from these beverages, not from caffeine, not from coffee.” He recognized the challenge and questioned whether any of the manufacturers might be able to help by providing marketing and testing data. Marketing data could help researchers to understand exposure (who is being exposed), and testing data could help them to understand the effects of consumption.
Bob Arnot, representing Monster Energy, responded by clarifying that the company decided not to market their product to children because of a lack of safety data. He observed that the label on the can states the product is not recommended for children, pregnant women, or sensitive populations. In addition, all cans are labeled with the amount of caffeine. He acknowledged the contentiousness of the issue with regard to adolescents and observed that the company does not market to adolescents. For example, they do not advertise on top-40 radio or through television. Arnot observed that company leadership is as concerned about the issue as the scientific community is and is looking as actively. “They haven’t seen anything,” he said. He remarked that when the American Medical Association issued its warning about energy drinks, they did so under the assumption that energy drinks have 50 times as much caffeine as coffee. One of the surprises about mainstream energy drinks is how little caffeine they contain. An 8-ounce serving contains roughly 80 mg of caffeine, whereas a commercial coffee product has 320 mg of caffeine. On a milligram-per-ounce basis, energy drinks have half the caffeine of coffee.
Later during the discussion, Arnot communicated that he had been in touch with company leadership and that they were happy to enter a dialogue.
Lipshultz applauded the idea of partnering among industry, the FDA, and others. He observed that about 17 years ago, when he was a voting member on the FDA Oncologic Drugs Advisory Committee, some terrible cardiac signals associated with a prescription product emerged but that there was inadequate data to study those signals. The company involved agreed to participate with the FDA and to conduct studies to gather the necessary data, so a black box warning label was issued. Several years later, postmarket surveillance indicated that the signals were
still present, but the promised research was never delivered. Over the past several years, similar issues have come up with over-the-counter products for children. Although he applauded industry’s willingness to at least consider providing data, he questioned what to do in the meantime given that there is no life-saving therapeutic reason to consume these products and that multiple independent databases have yielded safety signals. In his opinion, until safety is established, it is important to consider how to move forward in the meantime.
Arria, A. M. 2013 (unpublished). The college life study. http://www.cls.umd.edu (accessed December 5, 2013).