In addition to the wide array of environmental chemicals that people are exposed to in their daily lives, emerging evidence suggests that a number of other exposures may possibly increase an individual’s chances of developing obesity or metabolic disease. These include sugar, artificial sweeteners, antibiotics, and viruses, and they were the subject of the workshop’s fourth session.
Could obesity be triggered by certain types of infections? There is growing evidence that this may be the case. Nik Dhurandhar, professor and chair of the Department of Nutritional Sciences at Texas Tech University, discussed what is known about “infectobesity,” a term he coined to describe obesity of infectious origin.
The Concept of Infectobesity
The question of whether obesity is an infectious disease, Dhurandhar said, has many similarities to a question that was asked many years ago: is gastric ulcer an infectious disease? At one time that seemed a ridiculous notion, and medical students were taught that the main causes of gastric ulcer were stress and spicy foods. Now, of course, it is widely recognized that ulcers are caused by the bacterium Helicobacter pylori. Obesity, too, may one day be recognized as—at least in some cases—the product of infection.
At present, Dhurandhar said, a number of pathogens have been shown to cause obesity in animal models. The first was reported in 1982 in Science magazine: canine distemper virus, which causes obesity in mice. A variety of other pathogens, including viruses, scrapie agents, bacteria, and even parasites, cause obesity in different animal models, he said.
Among these are several adenoviruses. Dhurandhar and colleagues were the first ones to describe the adipogenic—that is, fat-creating—properties of an adenovirus, in this case, the avian adenovirus SMAM-1, and they were also the first to describe the adipogenic properties of a human adenovirus, adenovirus type 36 (Ad36). Other research groups later discovered additional adipogenic adenoviruses, such as Ad37 and Ad5.
Dhurandhar said that for the purpose of his presentation he would be focusing on Ad36 to illustrate various attributes of infectobesity, mainly because that it is the microbe whose adipogenic properties have received the most study. Ad36 is one of more than 50 known human adenoviruses, and it is antigenically unique from the rest of them. It was first isolated in Germany in the late 1970s from a fecal sample from a girl who was suffering from diarrhea, a clue that the virus might be responsible for causing gastrointestinal disturbances.
Animal Models of Infectobesity
Dhurandhar and his colleagues have done a large number of experiments in which they infect animals with Ad36 and look for changes. Ad36 has been associated with a significantly greater prevalence of obesity in a large number of animal models, including chickens, mice, rats, and marmosets.
In one experiment with marmosets, for example, the infected animals gained fat about three times as much as the uninfected animals after 6 months. The infected animals also had significantly greater body fat at the end of those 6 months (Dhurandhar et al., 2002).
In an experiment done with mice on chow diets, Dhurandhar and his colleagues used three groups of mice: the uninfected controls; a group infected with Ad2, a nonadipogenic adenovirus used as a control for infection; and a group infected with Ad36. The group infected with Ad36 gained more body weight and more body fat, but what was perhaps even more interesting was the glucose levels of the three groups. The fasting glucose levels of the mice infected with Ad36 dropped steadily as they aged from birth to 12 weeks, whereas the levels stayed constant in the other two groups. The insulin levels also dropped in the mice infected with Ad36. The implication is that the Ad36-infected mice had much better glycemic control than the other two groups of mice.
Dhurandhar and colleagues did a similar experiment with the same three groups of mice but this time fed them high-fat diets. Such diets
usually increase the amount of fat stored in the liver and deteriorate glycemic control. The results showed that the animals infected with Ad36 seemed to be somewhat protected in terms of how much fat accumulated in the liver and had better glycemic control. A great deal of fat accumulated in the liver tissue of the control mice and the Ad2infected mice, while the Ad36-infected mice had much less fat accumulation; indeed, the levels of fat in the livers of Ad36-infected mice fed the high-fat diet were much closer to the fat levels seen in control mice fed the standard chow diet, which have very little fat in their livers (Krishnapuram et al., 2011).
A group led by Jae-Hwan Nam of South Korea found that Ad36 requires the presence of the cytokine monocyte chemotactic protein 1 (MCP-1) to produce its adipogenic effect in mice. In particular, they showed that knockout mice missing the gene for the manufacture of MCP-1 do not get obese when infected with Ad36 (Na and Nam, 2012).
In another experiment, Nam and colleagues showed that an anti-inflammatory agent, mulberry extract, attenuates Ad36-induced obesity. And in yet another experiment they showed that vaccination of mice against Ad36 prevented the development of obesity after they were infected with Ad36.
In general, Dhurandhar said, animals infected with Ad36 have no overt symptoms other than the obesity; there is no additional mortality, for example. Ad36-induced obesity can also be passed from one animal to another through the normal infection routes—both through direct contact with an infected animal or by injection of the blood of an infected animal into the veins of an uninfected animal. The animals that are infected with Ad36 through direct contact or injection of blood get obese as they age. In short, obesity can be transmitted from one animal to another like an infection. This fulfills a Koch’s postulate, which is used to determine the infectious nature of a disease.
Finally, Dhurandhar said, Ad36 infection does not have much effect on food intake, so it seems unlikely that the infection produces obesity through an increase in caloric intake. Dhurandhar said that he and his colleagues have not made objective measurements of the activity of Ad36-infected animals, but they are kept in cages, and no obvious differences in their activity levels have been observed.
Mechanism of Action
A large number of studies have examined the possible mechanisms of action for viral infections leading to obesity, Dhurandhar said, and he summarized their results in this way: it appears that the virus-induced expansion of adipose tissue is due to an increased proliferation, commitment, and differentiation of adipose tissue-derived stem cells as well as lipid accumulation in adipocytes.
This effect of Ad36 on lipid accumulation is dose dependent, Dhurandhar said. The greater that the viral load is, the more lipids accumulate—to a point. At a certain viral load, the effect levels off. Furthermore, an antiviral drug, cidofovir, that kills or blocks Ad36 reduces the level of lipid accumulation in the cells.
The lipid accumulation in adipocytes is very specific to the presence of the infection. Cells that have Ad36 accumulate lipids, while those that are free of the virus do not.
Dhurandhar then described a working model, in terms of cell signaling, of what the virus appears to be doing. On the one hand, it seems to increase the cyclic adenosine monophosphate (cAMP) pathway by acting on the cAMP-responsive element-binding protein (CREBP) pathway. On the other hand, the Akt pathway is activated. The combined effect is to increase the proliferation of adipocyte progenitors, leading to their differentiation and to lipid accumulation.
One interesting finding, Dhurandhar said, is that Ad36 increases glucose uptake by cells. This can be seen in experiments with adipose tissue biopsy specimens from humans. If the tissue is divided and cultured in two sections and if one of those two sections is infected with Ad36, the infected tissue shows significantly greater glucose uptake than the noninfected tissue. There is a similar effect in skeletal muscle tissue. Thus, Dhurandhar said, Ad36 appears to increase glucose uptake by both adipose tissue and human skeletal muscle cells.
Thus, the working model looks like this, Dhurandhar said: Ad36 infects adipocyte progenitors and increases their commitment and differentiation, resulting in increased adipogenesis; the virus also increases glucose uptake inside the cell (see Figure 5-1). In summary, he said, there are more adipocytes that have more lipids, and they are taking in more glucose. That may explain the increased adiposity that is also accompanied by enhanced glucose clearance from the systemic circulation.
Dhurandhar then moved to a discussion of human studies. “The million dollar question,” he said, “is, Do certain infections cause human obesity?”
There are a number of challenges to doing human studies, he said. One is that obesity has an insidious onset. An obese person may not remember what happened 3 or 5 years ago to trigger the obesity, for example, whether there was an infection around the time of the onset.
A second challenge is the presence of multiple etiological factors for obesity, such that it is very difficult to attribute obesity to any one factor. If you perform a study comparing people who have had a certain obesity-related infection with those who have not, the people in the control group with no history of the infection may still have become obese for some other reason.
A third challenge is that the infection by itself may not be enough to trigger obesity. Other factors may be needed. Thus, some people who get the infection become obese and others do not, and it may be difficult to determine what is going on.
Finally, ethical considerations preclude infecting people for an experiment, so, unlike in animal studies, it may be impossible to ever show a direct cause-and-effect relationship between an infection and obesity in humans. Dhurandhar compared the situation with that with smoking and lung cancer. There has been no direct experimental proof that smoking causes lung cancer in humans, so the proof relies upon strong indirect evidence of the connection.
So, Dhurandhar said, he and his colleagues have set out to uncover similar indirect evidence to answer the question of whether infections cause obesity. They began by looking for people whose blood held neutralizing antibodies against Ad36, which indicated that they had been exposed to the virus at some time in the past. In one study with more than 500 subjects, they found that 30 percent of the obese subjects had Ad36 antibodies, while only 11 percent of the nonobese subjects did. Furthermore, the subjects in the study who had antibodies to Ad36 had a significantly greater body mass index (BMI) than the subjects who did not have antibodies. Even within the nonobese group, those with the Ad36 antibodies were likely to be heavier than those who were negative for the antibodies. As a control, the researchers also looked at antibodies against two viruses that have not been associated with obesity, Ad2 and Ad31. For those viruses, there was no such relationship between the presence of antibodies and BMI or obesity.
Dhurandhar’s group also carried out a human twin study looking at the relationship between the presence of antibodies and obesity (Atkinson et al., 2005). They recruited 90 twin pairs and tested their blood for the presence of the antibodies. They retained only those twin pairs in which one twin was positive for the antibodies and the other was negative. This left them with 26 pairs of twins, 20 of which were identical twins and the other 6 of which were fraternal twins. Twins usually have very similar BMIs, but among these 26 pairs of twins, the twins who were antibody positive had significantly higher BMIs than those who had not been exposed to Ad36. The implication is clear: infection with Ad36 may make it more likely that a person will gain weight and become obese. “I think this is the closest one can come to determining the role of Ad36 in human obesity, short of infecting them,” Dhurandhar said.
Another study examined 1,500 Caucasian, Hispanic, and African-American men, women, and children who were screened for the presence of Ad36 (Krishnapuram et al., 2011). Analysis of the data showed that a previous Ad36 infection was associated with better glycemic control and lower hepatic lipid levels. This result is very similar, Dhurandhar noted, to the results of the experiments in mice that were infected with Ad36.
A 10-year prospective study of 1,400 men and women found similar results. The subjects were screened for exposure to Ad36 at the beginning of the study and followed for 10 years. Those who were positive for Ad36 antibodies at the beginning had a significantly greater increase in body fat over the next 10 years than those who did not have the Ad36 antibodies, and they exhibited significantly less decline in their glucose control than their antibody-negative counterparts. Compared with the findings of cross-sectional studies, this prospective study provides stronger evidence about the possible role of Ad36 in increasing fat.
There is evidence that, at least in some countries, the prevalence of Ad36 infections is rising. Dhurandhar showed some data from Sweden tracking the rates of Ad36 infections in lean Swedes over time, and the data showed that the prevalence increased from about 7 percent in the mid-1990s to nearly 20 percent in 2009. During that same period, Dhurandhar noted, the prevalence of obesity increased in Sweden in both men and women.
Summarizing the data, Dhurandhar said meta-analyses have found that most—although not all—of the studies report an association between exposure to Ad36 and an increased risk of obesity.
Significance and Implications
In closing, Dhurandhar spoke of the significance and implications of what is known about the connection between infections and obesity. Obesity is a complex disease with a multifactorial etiology, he said, and given that multifactorial etiology, it will be necessary to employ a multifactorial treatment and prevention approach. In particular, infection-related obesity will likely have its own unique prevention and treatment strategies, and it may well be the case that the development of a vaccine against Ad36 will be easier than dealing with obesity by inducing widespread behavioral change.
He cautioned that he did not mean to suggest that all obesity is due to infection. It is clear that infection with Ad36 causes obesity in animals
and is correlated with human obesity, and there may be many adipogenic pathogens that have yet to be discovered. The important question, he said, is how much have infections contributed to the increase in obesity since 1980? “I do not know the answer to that question,” he said, “but that is a question to ask.”
Finally, Dhurandhar showed a series of PowerPoint slides with data from the Centers for Disease Control and Prevention on the geographical prevalence of asthma, influenza, and obesity. Looking at the changes in the prevalence of asthma over time, it is clear that the pattern is random, which makes sense because asthma is a noninfectious disease. However, looking at the changes in the geographical prevalence of influenza over time, clear patterns seem to show focal points and the spread from state to state. Again, this makes sense because influenza is an infectious disease, and one would expect to see such changing geographical patterns over time. Finally, he showed a series of slides showing the prevalence of obesity over time and noted that the pattern was very similar to what was seen with influenza but not similar to that seen with asthma.
“I want to leave you all with this question,” he said. “Why does the spread of obesity in the United States resemble an infectious disease?… I truly do not know the answer to that question. To me, it does look like an infectious disease spreading through the United States.”
Lynn Goldman of the Milken Institute of Public Health at George Washington University opened the discussion session with a question about the modes of transmission of Ad36. Is there, for example, maternal–fetal transmission? That is, are some people born with the virus? She also asked what the disease looks like in its acute stage. Are people acutely ill when they have this infection?
Dhurandhar answered that no specific information about what Ad36 does in people is available. In general, there are several categories of adenovirus infections. Some of them are linked with upper respiratory tract infections, some are linked with conjunctivitis, and some are linked with gastrointestinal disturbances. Because it was first isolated from a girl suffering from enteritis, it may be one of the classes of adenoviruses that cause gastrointestinal disturbances, but nothing is known for sure, and it does not produce diarrhea or any other major symptoms in animals.
As for the issue of mother-to-child transmission, the answer is not known. Dhurandhar said he would like to do that experiment and wrote a proposal requesting a grant, but it did not get funded.
Linda Birnbaum of the National Institutes of Environmental Health Sciences (NIEHS) asked if it would be worthwhile to do studies looking at what happens when people are exposed both to infectious agents and to environmental chemicals, because it is known that environmental exposures can alter susceptibility to infectious agents. Dhurandhar said that this would be a good research question. Indeed, some studies have shown that it is only when a virus acts in collaboration with various other factors that the virus expresses its phenotype.
Several studies have suggested the possibility that the use of antibiotics during infancy may increase a child’s chances of later developing obesity. For example, experiments with mice have shown that if pregnant mice are given antibiotics at about the time of birth, the pups are more likely to grow into obese adults. Charles Bailey, an assistant professor of clinical pediatrics at the University of Pennsylvania and the lead investigator for the Data Coordinating Center of PEDSnet, a collaboration among several pediatric academic centers working to provide a standardized model for clinical care, provided an overview of what is known about the potential connection between antibiotic use and obesity in children. He spoke by phone.
The Role of the Microbiome
How might the use of antibiotics increase the chances of developing obesity? The most likely explanation, Bailey said, is that the antibiotics would differentially affect gut microbes. Several studies have suggested that a decreased diversity of gut microbes and, in particular, a shift to a small number of higher-risk species is associated with a higher prevalence of obesity. The mechanisms behind this are unclear, he said. One possibility is that the gut microflora has an effect on energy metabolism and how a person’s body uses calories. Yet another possibility is that changes in the microflora result in changes in a person’s digestive patterns, which in turn cause changes in food-seeking or calorie-consuming behavior. It is also possible, Bailey said, that antibiotics affect a person’s immune responses and, in particular, lead to the creation of chronic low-level inflammatory stimuli, which could in turn
predispose a person to the development of obesity via such mechanisms as increased endogenous steroid release and insulin resistance.
In his presentation, Bailey focused on the possibility that infancy and early toddlerhood is a critical period for the establishment and stabilization of the gut microbiome, making this a particularly sensitive period for the use of antibiotics. Studies with older children and adults have shown a tendency for a sort of “microbiomic homeostasis,” that is, a tendency for the gut microbiome to return to its original composition if some outside influences disturb it. The question, then, is, When exactly is this preferred composition of the gut microbiome set?
The establishment of the microbiome begins at birth or even before, Bailey said. Parents clearly have some effect on a child’s microbiome, because studies looking at the composition of the gut flora have shown that components of the microbiome in children are “inherited,” in the sense that they are concordant between a child and the child’s parents and are very similar between twins. Components of a child’s microbiome also seem to reflect the environment rather than the parents. So it is possible that the first few years of life are a critical period in establishing the composition of the gut microflora and that a disturbance during that time—such as a disturbance caused by antibiotic use—could change the composition of the microbiome and affect a child’s likelihood of becoming obese.
The only way to know for sure whether this happens is to examine the data, so Bailey reviewed a number of observational studies that have looked for a connection between antibiotic use in childhood and the development of obesity.
Before describing these studies, Bailey pointed out that the sort of population-level research that he was talking about depends on a large-scale collaboration among not just investigators but also patients. “In the era of electronic health records and health information exchanges,” he said, “we have a new kind of substrate to look at the way health is played out in the delivery of normal clinical care that incorporates the work of lots of unnamed collaborators seeing patients every day and lots of patients who have entrusted us with their data.”
Bailey began by describing two studies that laid the groundwork for the study of antibiotics and obesity. The first was a cohort study in the United Kingdom that followed more than 10,000 children from birth
through 9 years of age. The children in the cohort were mainly upper middle class and 93 percent Caucasian. Surveys and parental recall were used to estimate environmental exposures for the children, while growth parameters were measured directly and obesity rates could be estimated directly at several time points up through 7 years of age (Trasande et al., 2013).
A retrospective review of the data from this study found an increase in obesity at 3 years of age if antibiotics had been taken in the first 6 months after birth but not if they had been taken subsequently. The BMI among 7-year-old children in the study who had taken antibiotics between 15 and 23 months after birth was also a slightly higher. Thus, there was some connection between antibiotic use and extra weight or obesity, but it was not definitive (Trasande et al., 2013).
The second study followed a cohort of nearly 30,000 children in Denmark. It used parental recall to determine antibiotic usage and gathered growth data with a questionnaire administered at 7 years of age. Interestingly, the data showed a major effect of whether the mother was overweight. There was a 54 percent increase in the risk of being overweight among children whose mothers were not overweight and who had been given antibiotics in the first 6 months after birth; conversely, there was a 46 percent decrease in risk of being overweight among children whose mothers were overweight and who had been given antibiotics in the first 6 months after birth (Ajslev et al., 2011). This finding of a decrease in the risk of being overweight after antibiotic use is unique to this cohort, Bailey said. Still, both studies suggest that perturbing the microbiome that one inherits in part from one’s parents may affect the long-term risk of obesity.
A recent study in Canada found an association between antibiotic exposure in infancy and a later risk of obesity risk, but only in boys. “Whatever the mechanism behind the correlation is,” Bailey commented, “it is not going to be simple and straightforward.”
He also noted that the designs of these studies—the use of parental recall for information about exposure to antibiotics and a reliance on cross-sectional measurements of growth parameters at discrete time points—have both advantages and disadvantages. They can get very good information about covariates, including lifestyle and environmental factors, for example, but they are limited in their ability to get detailed longitudinal information on the degree of antibiotic exposure and the growth trajectories of the children.
The Current PEDSnet Study
Bailey then shifted to discussing the current study that he is involved with, beginning with a description of the PEDSnet Collaborative Network. This is a data-sharing partnership of eight large children’s hospitals that are committed to the development of learning health systems in pediatric clinical care. The network is interested in obesity because it is one of the more pressing public health problems among children, he said.
In 2010–2011, six of the eight hospitals in the network took part in a pilot study on the subject of the role of antibiotics in the development of obesity. The purpose of the study, which examined a 2-year block of outpatient data on about half a million children, was to see if routine clinical data could be effectively used to answer questions regarding obesity and the treatment of obesity.
The study focused on clinical data regarding BMI. Because pediatricians routinely record the heights, weights, and BMIs of their patients, if for no other reason than they are necessary to determine the doses of medications, the records provided a relatively rich source of data, Bailey said. “We have a lot of opportunity to offset noise in the data by looking at the behavior of the data in aggregate.”
Checking their data against results from the National Health and Nutrition Examination Survey (NHANES), the researchers found that the two agreed very well. In particular, their estimates of the prevalence of obesity among the children in the pilot study sample matched the obesity rates from NHANES quite closely, which gave them confidence that it was reasonable to use data from the clinical records to measure obesity in populations of children receiving both well-child care and acute care.
The pilot study also validated the use of prescription data as a measure of prescription drug use among children seen by doctors and other health care workers in the network. The prescription data do have some gaps, Bailey acknowledged—such as prescriptions from outside the network and drugs prescribed but not used—so the data have a particular set of systemic biases. However, he noted, other studies have taken different approaches as a way to balance out those biases.
Bailey also explained that the use of the height, weight, and BMI data provides a view of the prevalence of obesity among children very different from that obtained by the use of such administrative data as claims data or data from state-level data sets. The reason is that, according to their data, physicians recorded a diagnosis of obesity in only
about 18 percent of their juvenile patients who met the criteria for obesity according to their BMI. Indeed, he said, except for physicians in dedicated weight management clinics, no one is documenting obesity diagnoses for more than about one-third of their patients who meet the BMI criteria for obesity. Thus, any studies that rely on administrative data will miss about 80 percent of children who are obese according to BMI criteria.
Having validated the approach of using clinical data to study obesity, the group of researchers then took a close look at whether different exposures to antibiotics among children could explain at least some of the children’s long-term chances of developing obesity. To do that they examined 15 years of data from the records of children in one large health care system in the mid-Atlantic region of the United States. The records held information on both prescription medications and growth measurements, and for a majority of the children in the study, there were at least 8 or 9 years of records. This made it possible to assemble a longitudinal cohort.
The researchers looked only at children who had received primary care in the network within the first year after birth and who had also had a longitudinal follow-up in the network at least past their third birthday. Ultimately, after eliminating subjects for which there was not enough information, the researchers were left with a cohort of 65,000 children who reflected the structure of the network, with one-third of them living in urban neighborhoods in and around Philadelphia, Pennsylvania, and the remaining two-thirds living in suburban neighborhoods. Nearly half of them were identified as belonging to some racial or ethnic minority, and approximately 40 percent were covered by a public insurance provider at some point during the study period.
To assess antibiotic exposure, the researchers collected data on prescriptions for antibiotics that were written for the subjects before 2 years of age. Antibiotics were classified as either narrow spectrum or broad spectrum. The data showed that the vast majority of the antibiotic prescriptions were for common childhood infections, such as ear infections, pneumonia, or sinus infections.
To determine obesity rates, the researchers collected data on BMI for 3 years beginning at 24 months of age. Obesity was defined as being at or above the 95th percentile for BMI according to norms from the 2000 NHANES data, while the cutoff for overweight was the 85th percentile for BMI calculated from the NHANES data.
Among the study’s preliminary observations was that the use of antibiotics is very common in early childhood, Bailey said. Roughly two-thirds of the children in the study had some antibiotics prescribed to them before their second birthday. Furthermore, a substantial number of the children were given antibiotics multiple times before they were 2 years old. These repeat exposures were typically for the treatment of common childhood infections rather than for the treatment of chronic medical conditions that might result in growth trajectories that were different from those of the majority of the children.
A multivariate analysis of the data found a number of factors to be correlated with an increased risk of obesity at ages 2, 3, and 4 years. Boys were more likely than girls to be obese. Hispanic ethnicity but not membership in other racial and ethnic groups was associated with increased obesity. Public insurance coverage—a rough indicator of a lower socioeconomic status—was also associated with an increased risk of obesity, although there was no difference in obesity rates between children in urban versus suburban practices.
There was a strong association between obesity and a diagnosis of asthma. Some of that was probably due to steroid usage, Bailey noted, which is itself an independent predictor of increased obesity risk. The fact that children with asthma are less likely to be active may also play a role, he said, as may the inflammation associated with asthma.
Taking all of those factors into account, the researchers then calculated the increased risk of obesity—the “hazard ratio”—associated with taking antibiotics at various ages. Not only were children exposed to antibiotics more likely to be obese than children who were not, but there was also a trend toward a greater risk of obesity as the number of different antibiotics that a child had been exposed to increased.
Bailey broke the analysis down by broad-spectrum versus narrow-spectrum antibiotics. There is a dramatic difference there, he noted. For the broad-spectrum antibiotics, there was a trend toward a greater risk of obesity with exposure to a greater number of antibiotics, but that trend was not apparent for the narrow-spectrum antibiotics.
The data point to two conclusions, Bailey said. The first is that exposure to antibiotics early in childhood—in particular, in infancy—is associated with a small but persistent increase in obesity later in childhood. The second is that the majority of this effect appears to be associated with broader-spectrum antibiotics. “This is a significant clinical point,” he said. “There is substantial evidence that narrower-spectrum drugs are adequate therapy for most of the kinds of infections
that are being treated in this population. The decision to use broader-spectrum antibiotics will often hinge on physician preference or convenience of dosing or other nonantimicrobial parameters. It is important to underscore that there may be unintended effects at work here.”
It remains to be seen exactly what is mediating the association between antibiotics and obesity risk, if indeed the antibiotics are playing a causal role. Because the obesity occurs well after the antibiotic use, the association cannot explained by temporary deflections in a child’s growth trajectory or by perturbations in the gut microbiome at the time of exposure, he said. His group is particularly interested in the possibility that the use of antibiotics causes lasting changes in the composition of the microbiome that persist beyond infancy, and it is currently conducting a study looking at the direct effects of antibiotic exposures on the composition of gastrointestinal bacteria that may shed some light on the precise mechanism.
Bailey then offered a number of caveats concerning his study that are related to its design. For example, certain biases are caused by the fact that it relies on health care system data. Specifically, although there is information on diagnoses and antibiotic prescription, Bailey pointed out that there is much less information about longitudinal factors that affect the family, lifestyle, and diet. This is particularly important, he added, because the effect sizes that they are observing are relatively small, on the order of a 10 to 20 percent increase in the risk of obesity. In the future, he said, it will be important to combine this type of study with the sorts of birth cohort studies that he described earlier, which provided detailed insight into lifestyle factors. Another caveat is that it was a regional study, so there is the possibility that it reflects geographic or local practice effects.
In summing up, he said that it is clear that the problem of obesity will not be solved with a silver bullet and that instead it will require a large number of incremental changes, some of which will be useful to only certain subsets of patients. In particular, he said, it seems possible that the careful use of antibiotics—tailoring them to the appropriate indications and to the appropriate antimicrobial spectrum—could have an effect on the risk of obesity at the population level. Thus, it makes sense to do further research on such things as the establishment and maintenance of the gut microbiome and the determination of “obesity-friendly” health care practices.
In the brief discussion period that followed his presentation, Bailey first addressed a question from Linda Birnbaum of NIEHS about whether it might also be worth examining the effects of antiviral medication use in children. Bailey answered that while this is worth looking into, the rate of use of antiviral medications in children is much, much lower than the rate of use of antibacterial medications, so he did not believe that it would be possible to see any possible effects of antivirals in a study population of the size that he had used. It might also be worthwhile, he said, to look at the effects of antifungal use in children. To his knowledge, he said, no one has tested the effects of antifungals on the gut microflora.
Sheela Sathyanarayana from the University of Washington mentioned the experiment that Bailey had described, in which antibiotics were given to mice during the perinatal period and led to obesity later on, as the pups grew. Has anyone looked at whether antibiotics given to human mothers in the human perinatal period have an effect on their children’s risk of obesity? Bailey said he would love to see such a study and that it should be feasible in the right setting, but it has not been done.
It is no surprise to anyone that sugar plays a role in obesity—after all, sugar supplies calories in an appealing and easy-to-eat form, and the consumption of too many calories steadily over an extended period of time leads to obesity. But does sugar—and, in particular, a certain type of sugar, namely, fructose—play a different and more direct role in this disease? That was the subject addressed by Ayca Erkin-Cakmak, a clinical research associate at the University of California, San Francisco, who has been investigating the effects of a reduced amount of sugar in the diet on the metabolic health of obese children.
She began by showing a couple of videos, one made in 1970 and one made much more recently, of doctors saying that it is calories, not sugar per se, that people have to be careful of—that as long as you keep the total calories low enough you do not gain weight and that sugar itself poses no risks. Erkin-Cakmak said that this view is too simplistic and that science has something else to say.
According to science, she said, some calories are more likely to cause disease than others. Different types of calories are metabolized differently, she explained. For example, a person who consumes a cup of
almonds with 160 calories absorbs only 130 calories because of the fiber in the almonds. In the case of proteins, the body is required to supply a certain amount of energy to use their amino acids for energy, so the net calorie gain is less than the number of calories in the protein. She also noted that although some fats are healthier than others, 1 gram of fat counts as 9 kilocalories no matter what type of fat. “So a calorie is not a calorie,” she said.
Erkin-Cakmak then reminded the audience of the definition of toxicity. Toxicity refers to the degree that a substance can damage an organism. In that definition, there is no distinction between acute and chronic toxicity. If fructose is to be considered a toxic substance for the human body, it must pose a risk factor independent of the effects of its calories and independent of the obesity that those calories can cause. Furthermore, one must establish a causal relationship between the fructose and whatever negative consequences it is associated with.
One of the criticisms of the claim that fructose is toxic to humans is that the toxicity has been shown in animal models and not in humans and that it has been demonstrated at doses that are in excess of what humans normally ingest. But, Erkin-Cakmak said, she would be talking about human data and the doses routinely ingested by people.
The main problem with fructose is not obesity, she said. People do not die from obesity. Instead, people die from various components of the metabolic syndrome, particularly diabetes. Diabetes is a disease that has cost a fortune to treat and prevent, so that is the problem resulting from the ingestion of fructose that she has been investigating.
Thirty percent of the adult population in the United States is obese, Erkin-Cakmak noted, and 80 percent of the obese population is sick with various syndromes: type 2 diabetes, cardiovascular disease, hypertension, and hyperlipidemia. “They have all these components of metabolic syndrome because they exceeded their daily intake and they became fat,” she said. Or is that really it? Forty percent of people in the United States who are of normal weight also suffer metabolic dysfunction, she said—again, type 2 diabetes, hyperlipidemia, and hypertension. The prevalence is less among those who are of normal weight than among those who are obese, but the metabolic dysfunction is still a problem. Almost half of the population of the United States is metabolically unhealthy, she said. “It looks like there is an exposure which affects the [entire] population.”
What could that exposure be? Erkin-Cakmak showed a graph of U.S. sugar consumption from 1822 to 2005. That consumption grew steadily until the 1930s, when it was approximately stable until the 1970s, when
people became concerned about fats in their diets and replaced some of them with sugar, at which point sugar consumption began growing sharply again. The rise in sugar consumption is correlated with the appearance of various health issues, such as cardiovascular problems and type 2 diabetes. That does not prove causation, she noted, but there is clearly a relationship between the increasing amount of sugar consumed and the development of metabolic problems.
Whatever relationship there is between sugar consumption and diabetes is confounded by obesity, Erkin-Cakmak said, because obesity is strongly correlated both with sugar consumption and with diabetes. However, she continued, although there is a strong relationship between diabetes and obesity, the two conditions are not concordant. Some countries have a relatively high prevalence of obesity but a low prevalence of diabetes, while other countries have a high prevalence of diabetes but a relatively low prevalence of obesity. Thus, diabetes is not a subset of obesity, she concluded. Indeed, while the prevalence of obesity is increasing worldwide by 1 percent per year, the prevalence of diabetes is increasing at about 4 percent per year. If diabetes were a subset of obesity, they should be increasing at the same rate.
Plausibility of the Sugar–Diabetes Connection
How might fructose be leading to diabetes? One possible connection is through liver disease. Fatty liver disease can be caused by either alcohol or sugar, and histologically, the damage to the liver looks the same for the two types of fatty liver disease.
Nonalcoholic fatty liver disease has become an epidemic in the United States, Erkin-Cakmak said. Today, one-fourth of African Americans, one-third of Caucasian Americans, and almost half of Latinos have steatosis, or the abnormal retention of lipids within the cells of the liver, and 5.5 percent of the U.S. adult population is suffering from nonalcoholic fatty liver disease (Browning et al., 2004). Autopsies of children from the ages of 5 to 19 years who died from causes other than metabolic problems showed fatty liver disease in 13 percent of them, and the number jumps to 38 percent if the children were obese (Schwimmer et al., 2006).
There are three kinds of fat, Erkin-Cakmak noted: visceral fat, subcutaneous fat, and liver fat. While visceral fat is not good, liver fat is even worse and is associated with poor health. For example, a study done in South Korean adults found that nonalcoholic fatty liver disease was a
primary predictor of type 2 diabetes, after such factors as age, sex, BMI, and alcohol consumption were controlled for.
Another study tried to model changes in insulin dynamics as a function of either visceral fat or liver fat. In the models, when the researchers held the liver fat constant, they found that they could see no difference in insulin dynamics between those with low levels of visceral fat and those with high levels. However, when visceral fat was held constant, larger amounts of liver fat were correlated with greater insulin resistance. Thus, the liver fat seems to be more closely related to changes in insulin dynamics than visceral fat.
Mechanism of the Sugar–Diabetes Connection
An explanation of the mechanism behind the sugar–diabetes connection begins with the observation that fructose is not glucose, Erkin-Cakmak said. While glucose is essential for every single cell in the body, fructose is utilized by cells only if it is necessary. “The common wisdom is a calorie is a calorie and sugar is just empty calories,” she said, “but chronic fructose exposure promotes liver fat accumulation.” This in turn promotes the metabolic syndrome and also increases protein glycation, which promotes cellular and structural aging.
When a person consumes glucose, only 20 percent of that glucose enters the liver, while the rest is consumed by the body, particularly the muscles. Most of the glucose that enters the liver is stored as glycogen, while a very small amount of it enters the mitochondrial tricarboxylic acid (TCA) cycle and generates adenosine triphosphate (ATP). A tiny amount leaves the mitochondria and contributes to de novo lipogenesis; the newly synthesized lipids are transported as triglycerides out of the liver. Thus, glucose consumption does not add lipids to the liver cells.
Ethanol is quite different, with 80 percent of the ethanol consumed entering liver cells. The ethanol that enters the liver cells goes into their mitochondria and overruns the mitochondria, exceeds the TCA cycle’s capacity, and then leaves the mitochondria via the citrate shuttle and contributes to de novo lipogenesis. The newly synthesized lipids leave the liver cells as triglycerides. This is why alcoholics suffer from hypertriglyceridemia, high blood levels of triglycerides. Some of the newly synthesized lipids, very low density lipoproteins, participate in the generation of liquid droplets that are stored in the liver, which results in alcoholic fatty liver disease.
In the case of fructose, 100 percent enters the liver cells. It is not used in glycogen synthesis. Instead, it overruns the mitochondria in the same way that ethanol does, with lipids stored in the liver and triglycerides being exported into the bloodstream. This is how alcohol and fructose are metabolized in the same way.
The results of this can be seen in a study by Jean-Marc Schwartz at San Francisco General Hospital. Normal-weight adults were given either a fructose-based diet or a complex carbohydrate diet, where the two diets had exactly the same number of calories. Those who had the fructose-based diet stored 27 percent more lipids in their livers than those who were on the complex carbohydrate diet. Erkin-Cakmak said that she and her colleagues had recently finished a very similar study in African-American and Latino obese adolescents and saw very similar results.
A second problem with fructose is the browning reaction, also known as the Maillard reaction or nonenzymatic glycation. This browning reaction generates reactive oxygen species from the fructose. In the human body, this glycation happens seven times faster for fructose than for glucose, producing reactive oxygen species seven times as fast.
Turning to human studies, Erkin-Cakmak showed results from the EPIC-Interact study in Europe that looked at the relationship between the consumption of sugar-sweetened beverages and diabetes. The researchers found that those who consumed one or more cans of sugar-sweetened beverage per day had a 29 percent greater risk of developing diabetes, after adjusting for energy intake and BMI.
In an international study looking at diet and diabetes, researchers used data from the Food and Agriculture Organization on the total number of calories consumed and the consumption of fruits, oil, sugar, meat, cereals, roots, nuts, and vegetables. They combined those data with data on diabetes prevalence worldwide from the International Diabetes Federation as well as economic data from the World Bank World Development Indicators Database. An analysis of the data showed that of all the food types, only sugar was correlated with diabetes prevalence. Furthermore, each additional 150 calories consumed was associated with a 0.1 percent increase in diabetes prevalence; however, if those additional 150 calories were from a can of soda, the increase in diabetes prevalence was 1.1 percent; that is, the prevalence of diabetes was 11 times greater if the additional calories were from a can of soda than from other sources. The researchers estimated that 25 percent of the cases of diabetes worldwide is explained by sugar consumption (Basu et al., 2013).
This study addresses some of the Bradford Hill criteria for causal medical inference, Erkin-Cakmak said: dose, duration, directionality, and precedence. Higher doses of fructose consumption lead to greater increases in diabetes prevalence. The duration of exposure is related to the development of diabetes, and the direction of causality is clear. Precedence is also clear, because whenever the availability of sugar increased in a country, 3 years later the diabetes prevalence increased in the same country.
Given all of these different forms of evidence pointing to the role of fructose in metabolic syndromes and particularly in diabetes, Erkin-Cakmak concluded by saying that understanding of the role of diet in disease must change.
In the discussion session following the presentation, Linda Birnbaum of NIEHS asked if there is any information about what fructose consumption might do to the microbiome. Erkin-Cakmak responded that she believes that fructose can be toxic to the microbiome and that, conversely, alteration of the microbiome might lead to a more excessive amount of fructose absorption.
Barbara Corkey of Boston University asked about what role factors other than sugar consumption might play in the development of diabetes, given that a number of other factors have been changing in parallel with the increased consumption of sugar, such as exposure to the plasticizers in the soda containers. It is troubling to attribute causation to one of these factors and eliminate all of the others from consideration, she said. Erkin-Cakmak replied that this was a good point and that she did not believe that the authors of the paper had considered changes in environmental exposures over time.
Nik Dhurandhar of Texas Tech University asked Erkin-Cakmak about the role of sugar in obesity. She replied that obesity is not the main problem. “Being metabolically healthy or unhealthy is the problem.”
For several decades people have used sugar substitutes—noncaloric sweeteners—as a way of sweetening food without adding the calories that come with sugar, but could those noncaloric sweeteners themselves cause obesity? That was the question addressed by Kristina Rother, a clinical investigator in the Diabetes, Endocrinology, and Obesity Branch at the National Institute of Diabetes and Digestive and Kidney Diseases.
Rother broke her presentation into three sections: an overview of artificial sweeteners and how they convey sweetness, a look at studies that have reported an association between artificial sweetener use and obesity, and a review of the data and concepts that either support or rebut a causal role for artificial sweeteners in the development of obesity.
Artificial Sweeteners and How They Convey Sweetness
Six artificial sweeteners are currently regulated by the U.S. Food and Drug Administration (FDA). The first one to be put on the market was saccharin, Rother said, and it is about 300 times sweeter than sucrose. Then there are aspartame (200 times sweeter), acesulfame potassium (200 times sweeter), and sucralose (600 times sweeter). The ones that were developed the most recently are neotame and advantame. The last two are also the two sweetest, with each one being 10,000 and 20,000 times sweeter than sucrose. They are so sweet, Rother said, that they are difficult to work with because an incredibly small amount makes a product taste extremely sweet. No products have yet been made with advantame, although there will be some in the future, she predicted.
FDA also sets the accepted daily intake (ADI) for each of the sweeteners. This is the amount that a person can ingest each and every day throughout his or her lifetime without experiencing any adverse health consequences. For saccharin, this is 5 milligrams per kilogram of body weight, or about the amount in three sodas sweetened with saccharin. For sucralose, the ADI is equivalent to five sodas per day, and for acesulfame potassium the ADI is equivalent to about 30 sodas per day. Most adults will not reach the ADI, Rother observed, but a publication by South Korean scientists showed that children can easily reach the ADI because of their small body size. If they drink two sodas a day, children may already be beyond the ADI.
To explain how artificial sweeteners work, Rother began with a description of the sense of taste. The tongue’s taste buds are concentrated mainly on the sides and the back of the tongue. In each taste bud there is one cell that responds to only one taste—either salt, sour, sweet, bitter, or umami. When one of these cells is activated, it sends a signal via the cranial nerves to the insula in the brain, reporting what has been tasted.
Interestingly, these taste receptors are located not only on the tongue or the oral pharynx, Rother said. There are taste receptors all over the mouth as well as in the intestine, in the beta cells of the pancreas, and
even in the lungs. “What they do in the lungs, we really have no clue,” she said. “There are things that we really don’t understand yet.”
In the intestine, endocrine cells have sweet taste receptors that respond to sugars by increasing the level of incretin hormones. These hormones lead to an increase in insulin levels and thus a decrease in blood glucose levels. It makes no difference to these sweet taste receptors in the intestine whether you have ingested carbohydrates, sugars, or an artificial sweetener—the receptors will respond in the same way.
One of the important incretins produced by the cells in the intestines is glucagon-like peptide 1 (GLP1). This hormone acts not only to increase insulin but to slow gastric emptying, decrease appetite, and decrease the levels of glucagon.
Studies Reporting an Association Between Artificial Sweetener Use and Obesity
A number of studies have suggested an association between the use of artificial sweeteners and obesity. As an example, Rother described a study by Sharon Fowler and colleagues whose results were published in 2008. They examined data from the San Antonio Heart Study, which enrolled more than 2,000 people, mostly white Americans and Hispanic Americans. When the researchers running the study enrolled the participants, they assessed their food intake and asked them how much regular soda and diet soda they drank. These people were then followed for 7 to 8 years (Fowler et al., 2008).
Fowler and her colleagues used the data from the study to see how the BMI of the participants changed over the course of the study and to compare that change with the consumption of diet sodas. What they found was a clear relationship between increases in BMI and soda consumption, with those people who drank more diet sodas experiencing a greater increase in their BMI over the 7 to 8 years of the study than those who did not drink diet sodas or who drank fewer diet sodas (see Figure 5-2). For example, people who did not drink sodas at all had an average increase in BMI of 1.0—equivalent to a person who was 5 feet 6 inches tall and weighing 155 pounds (BMI = 25) gaining about 6 pounds. Because most people gain weight as they age, that 6-pound gain over 7 to 8 years is a reasonable weight gain. However, the BMI of people who drank an average of 3 to 10 sodas per week increased by about 1.5 over that same time period. For a given individual, Rother noted, the difference
in weight gain is not particularly noticeable—only an additional 3 pounds for that 5-foot-6-inch, 155-pound person. But, she said, at the population level, “this is very important.”
In short, there is evidence of a relationship between the use of artificial sweeteners and obesity, but studies such as the study of Fowler and colleagues identify only a correlation between the two. They say nothing about causality, that is, whether the use of the artificial sweeteners actually did something to cause the weight gain. Rother examined that question next.
Data and Concepts Supporting or Rebutting a Causal Role for Artificial Sweeteners in Obesity
To examine whether there is a causal relationship between artificial sweeteners and obesity, Rother began by discussing in vitro studies and then moved to animal studies and, finally, studies with human populations.
Researchers have used in vitro tests to study the effects of artificial sweeteners on a number of different types of cells, including preadipocytes, 3T3-L1 cells, mature adipocytes, and human mesenchymal stem cells. A variety of different effects have been observed, Rother said. For example, saccharin, sucralose, and acesulfame potassium have all been shown to cause preadipocytes to turn into adipocytes more quickly. In mature adipocytes, artificial sweeteners cause a decrease in lipolysis, or the breakdown of lipids, so that more lipids accumulate in the cells. There is also decreased lipolysis in human mesenchymal stem cells. That last result, Rother said, was unpublished data obtained in collaboration with her colleagues Sabyasachi Sen and Allison Sylvetsky at George Washington University that they were planning to present at a meeting of the Endocrine Society later during the same week that the workshop was held.
Referring back to Barbara Corkey’s workshop presentation on the previous day, Rother noted that Corkey had reported that various artificial sweeteners increase the amount of insulin secreted from rodent pancreatic beta cells in vitro. The same increase in insulin secretion has also been shown in MIN6 cells in response to sucralose and saccharin, she said.
Artificial sweeteners are also known to be bacteriostatic, that is, they inhibit the growth of bacteria. This is generally a useful function because artificial sweeteners are used in a variety of products, including lip balm and toothpaste. Dentists are happy about artificial sweeteners in toothpaste, Rother said, because they help prevent the growth of bacteria in the mouth and around the teeth and gums. Other research has shown that artificial sweeteners—sucralose in particular—suppress intestinal microflora, the complex ecosystem of bacteria that live in the intestines and help with the digestion of food.
An important paper published in Nature in October 2014 showed that this effect on the intestinal microflora could be linked to obesity, Rother said. A group of researchers led by Eran Elinav from Israel examined the effects of feeding saccharin to mice (Suez et al., 2014).
The saccharin-fed mice got significantly fatter than glucose-fed mice, which led to the question of why. The researchers found that there was no difference between the two groups of mice in either energy intake or energy expenditure, which led them to deduce that it was the changed microbiome in the saccharin-fed mice that made them fat. In particular, the mice became more efficient at digesting their food because certain pathways in the microbiome that help the mice use the calories that they digest are upregulated; that is, the response of those pathways to calories is increased.
To verify that this was really what was going on, the researchers transplanted the microbiomes from the saccharin-fed mice and the glucose-fed mice into other, germ-free mice and examined those mice. What they found was that the microbiomes of the mice into which the microbiomes had been transplanted ended up resembling those of the particular type of mice from which they had received the transplants: The mice with transplants from saccharin-fed mice had higher levels of glucose in their bloodstreams and got fatter than the mice with transplants from the glucose-fed mice.
The Israeli researchers also looked for differences between people who regularly used artificial sweeteners and those who did not consume artificial sweeteners. There were a variety of differences: the artificial sweetener users had higher BMIs and higher levels of hemoglobin A1C, for example. Everything was a little worse for that group, Rother said, but she added that she did not find the human data in that paper to be particularly convincing, other than showing that the microbiome in people who consume artificial sweeteners looks different from the microbiome in those who do not, which is not particularly surprising. It will be important to go beyond these sorts of associations in people and start examining causation, she said.
Some interesting data could appear from some clinical trials now under way in Sweden, in which researchers are testing human microbiome transplantation as a treatment for obesity. “We will see what the human studies show,” she said.
A different line of research has examined the effects of the consumption of artificial sweeteners by lactating mothers. Studies have shown, for example, that lactating rats concentrate acesulfame potassium sixfold in breast milk. Rother indicated that preliminary studies done in her lab have found higher concentrations of acesulfame potassium in human milk than in the mother’s blood serum. “I think that when we do
cleaner studies,” she said, “we will be able to show that acesulfame potassium also accumulates in human breast milk.”
Could that affect the baby? Studies in rats have shown that offspring that were exposed to artificial sweeteners during pregnancy and lactation had a higher sweet taste preference later in life. Thus, it is possible, Rother said, that children whose mothers consumed large amounts of artificial sweeteners during pregnancy or nursing might develop a preference for sweeter foods.
Rother’s research group has also examined the short-term effects that the consumption of artificial sweeteners has on the blood levels of glucose and different hormones. They carried out oral glucose tolerance tests with 22 healthy adolescents and young adults ranging in age from 12 to 25 years in which changes in the blood levels of glucose and certain hormones were tested at various points in time after the subject was given a large dose of glucose. Before giving them the glucose and starting the test, they gave the subjects either mineral water or a diet soda with sucralose and acesulfame potassium. Those who had been pretreated with the diet soda had significantly higher levels of GLP1 during the test. When they repeated the test with 11 subjects with type 1 diabetes, they saw the same effect.
This would seem to be a good result, because GLP1 decreases the rate of gastric emptying, so that drinking a diet soda before eating, say, a pizza would cause a person to get full sooner and not each as much. “In fact,” Rother said, “some nutraceuticals have been started to be developed that mix all kinds of artificial sweeteners based on the principal that maybe you can get your GLP1 up.”
But there is more to the story, she said. A study of 17 obese women (average BMI of 42 kilograms per square meter) found that giving them sucralose before an oral glucose tolerance test made their test results worse (Pepino et al., 2013). In particular, during the test the levels of both glucose and insulin were much higher in the women who had consumed sucralose than in those who had just had water.
Rother’s group did the same experiment with 31 middle-aged adults of normal weight (average BMI of 26 kilograms per square meter). They found the same thing: giving them artificial sweeteners before an oral glucose tolerance test led to higher insulin levels during the test. One problem, she said, is that there is a great deal of variability in the outcomes of oral glucose tolerance tests, so such tests will require a large group of subjects to produce results that are statistically significant. However, she said she believes that it is worth following up on this
experiment because the size of the difference that they observed—a 20 percent increase in the amount of insulin in the bloodstream over time—is clinically relevant. “That is as much increase as you get with metformin or with weight loss or with all kinds of things,” she said.
Those studies were all looking at acute effects; that is, they focused on changes that occurred in the minutes and hours after consumption of artificial sweeteners. Rother then described experiments that looked for long-term effects. There are actually very few of these, she said. Two of them were published in 2012 in the same issue of the New England Journal of Medicine. One was a randomized trial of the effects of diet beverages on the weight of adolescents, while the other was a similar trial but with children instead of adolescents.
In the first one, the researchers randomly split a group of 224 overweight and obese adolescents into two groups (Ebbeling et al., 2013). One got diet drinks delivered to their homes to replace their usual sugar-sweetened drinks. The others were given supermarket gift cards with no instructions so that they could buy what they wanted. Although the group receiving the diet drink gained less weight, on average, after 1 year, at the end of the 2-year study there was no difference in weight gain between the two groups. The replacement of sugary drinks with diet drinks had failed to help the adolescents gain less weight.
In the other study, performed in the Netherlands, 641 children who already drank sugar-sweetened sodas were divided into two groups, with one of them continuing to drink soda (one can per day) and the other given one can of an artificially sweetened drink per day. After 18 months the researchers looked for differences between the groups. All of them gained weight, of course, because they were growing children, but the sugar-sweetened group gained more weight than the children who received the drink with the artificial sweetener. The main problem with the study, Rother said, is that there was no group of children who drank just water, so it is impossible to know how children who drink diet sodas would fare versus those who drink just water, but the experiment did clarify one thing: “We know now that if you drink soda, you gain weight beyond what you should gain,” Rother said. “If you replace it with artificial sweeteners, you gain less weight.”
Rother ended her talk by commenting that the human brain is perfectly capable of differentiating between caloric sweetness and noncaloric sweetness. Research has shown that while artificial sweeteners activate the insula in much the same way that sucrose does, sucrose activates certain parts of the brain that artificial sweeteners do not, in particular, dopamine-
dependent areas. That raises the question of whether sweetness with calories provides a different sort of reward to the brain than sweetness without calories and whether that makes a difference, particularly with chronic use. It is an issue that needs to be investigated further, she said.
In summing up, Rother said that there is no convincing evidence that artificial sweeteners prevent or alleviate obesity in humans. On the other hand, there are clear data showing that, in vitro, artificial sweeteners lead to more adipogenesis, less lipolysis, and more insulin secretion than sucrose. It is also clear that artificial sweeteners can affect the intestinal microflora and, in lab animals at least, that they can lead to higher glucose levels and greater weight gain. In humans, however, it has been very difficult to establish causality. We know that artificial sweeteners change the microbiome and that it is possible, by looking at the microbiome, to tell the difference between someone who consumes artificial sweeteners and someone who does not. But do artificial sweeteners cause people to gain weight because of these changes? Rother believes that there are plausible reasons for the connection between artificial sweeteners and obesity but that we are far away from answering that question on the basis of rigorous clinical studies.
In the discussion session following Rother’s presentation, Barbara Corkey began by asking about the various doses of artificial sweeteners used in the various experiments. Rother answered that the in vitro experiments generally used levels higher than those that would be seen in humans consuming a reasonable amount of artificial sweeteners, but her research group is hoping to be able to do some in vitro experiments that reduce the doses to levels that are realistic for human users of artificial sweeteners. On the other hand, at least some of the experiments in lab animals—in particular, the mice experiment described in Nature—used doses that are comparable to the ADI in humans.
Linda Birnbaum commented that she believes that there are probably many people who drink large amounts of diet sodas every day and that there is probably a wide range of exposures in the population. In particular, there may be a subset of the population that routinely consumes levels of artificial sweeteners that are much higher than most health researchers would expect.
Corkey added that even educated people may not understand how prevalent artificial sweeteners are in food. Someone who picks up a
container of yogurt and sees that it contains acesulfame potassium, for example, is not likely to realize that this is an artificial sweetener. A related issue is that many people look for low-calorie and low-sugar foods in the belief that they are healthier and forget that this means that they contain artificial sweeteners. Rother mentioned one of her projects in which parents of children were asked if they would give their children artificial sweeteners and most of them said no, but when they were asked to choose products for their children, they chose a lot of products that said “No sugar added”—exactly the products that would have artificial sweeteners (Sylvetsky et al., 2014).
Lynn Goldman of George Washington University started off the panel discussion following the final presentation of the session by asking the four presenters to talk about what she saw as an emerging theme at the workshop: that it is not obesity so much as related metabolic changes that are the main health consequence to be concerned about.
Erkin-Cakmak agreed that the main issue is whether a person is metabolically healthy or unhealthy. She added that in a recent study conducted with African-American and Latino populations—both of which are generally metabolically unhealthy—she and colleagues found that they could improve the subjects’ metabolic health by giving them a healthier diet, even one that gives them the same number of calories that they had been consuming, for just 10 days.
Rother disagreed somewhat, commenting that, in general, adiposity is associated with increased inflammation and is clinically associated with more cardiovascular risk. Although not everybody needs to lose 50 pounds to be healthy, it is important to accept that, in general, obesity is associated with certain health conditions.
Dhurandhar offered two quick points. First, he said, while it is important to focus on the metabolic consequences of obesity, those are not the only consequences of obesity. “There are numerous other [nonmetabolic] comorbidities or adverse conditions that are linked with obesity that we may overlook if we only focus on diabetes or cardiovascular disease,” he said.
Second, he said, while research usually involves isolating one factor out of many and examining how it affects other things, in nature various factors work together and thus need to be considered in conjunction with one another. He mentioned as an example a study of animal models of celiac disease that found four different factors had to exist at the same time for the celiac disease to be expressed.
Dhurandhar asked whether, because antibiotic exposure is a marker for infections, there is a higher rate of obesity among children who have more infections. A related question came from a webcast audience member who asked if the association between antibiotic use and obesity can be confounded by a relationship between infections and obesity, assuming that antibiotics were being prescribed for obesity-causing pathogens.
Bailey replied that in the health services study that he and his colleagues did, they found no association between obesity and such infections as colds or upper respiratory infections. Furthermore, once the use of antibiotics was taken into account in the multivariate analysis, such infections as inner ear infections were not associated separately with obesity. Given that close to 90 percent of ear infections are viral in origin, he said, he thinks that it is unlikely that children who are more susceptible to clinical viral infections are also more susceptible to obesity.
Rother offered an anecdote to illustrate how various confounding factors leading to obesity may be at play. A case study trying to understand obesity in Kuwaiti children found that the strongest factor for childhood obesity was maternal employment. This could be due to several reasons, he said. He knows from experience that working mothers sometimes ask for antibiotics for their sick children because they do not want to miss work. On the other hand, a working mother is less able to supervise what her children are eating at various times of the day or how much television they are watching. It can be difficult to tease apart these various factors.
Janet Young, a webcast audience member, asked whether probiotic treatment might either complement an antibiotic treatment or take the place of an antibiotic treatment and how this might affect obesity. Rother said that it is extremely hard to get evidence-based answers to such questions and therefore little is known on the topic.
Corkey commented that despite the general consensus that obesity is not healthy, there is no healthy treatment for obesity, which raises the question of what clinicians should be doing about it. “You talk about people going to the grocery store,” she said. “What kind of advice should we give to people about how to conduct their eating lives?”
Erkin-Cakmak said she is raising her toddler with as little salt, sugar, or artificial sweeteners as possible and is feeding him mainly unprocessed food that is cooked from scratch. A recent birthday party indicated that the approach may be working, because her son, when he tried the birthday
cupcake with frosting pronounced it “yucky” and threw it in the trash can. She explained that cooking from scratch, avoiding processed food, and consuming real food would be the key to become metabolically healthy. That’s what she and her colleagues are encouraging in their patients who are seen in the clinic.
Dhurandhar said that the problem is that scientists do not yet know enough about the disease of obesity or its complexity and multifactorial etiology, similar to cancer. However, he added, the cause and the treatment of a disease are two separate issues. Celiac disease, for instance, can be treated by restricting gluten, but gluten does not cause it. Not knowing the best treatment for cancer does not mean that clinicians do not try to treat it. “I think that is where we are in terms of obesity treatment right now,” he said. “We have a blanket treatment to offer for obesity today,” which is to eat less and move more, regardless of the cause. “That is what underscores the need for research so that we can consider all of these causes, contributory factors, and come back with a better and really effective biologically meaningful treatment and prevention strategy that can take the weight off meaningfully and keep it off. We are not there yet.”
Bailey added that the best advice that he can offer people at this point in time is “moderation in all things.… I end up recommending a balance, a balance in diet, a balance in exercise, a balance in activity, and, to steal the line from Mark Twain, a balance in balance, too.”
Rother added that there is not enough evidence to answer certain questions with certainty, such as whether people should stop drinking artificially sweetened sodas. Thus, she said, she agrees with Dhurandhar that moderation is a good strategy.
Corkey elaborated that in the past people have been given advice for which there was no scientific evidence and that has occasionally been proven wrong. “For example,” she said, “our focus on lipids has been all wrong. Poor eggs suffered for many years.” She continued by commenting that scientists really do not know what causes obesity and that there is very little useful, outside of surgery and a few moderately effective medications, to treat it. Thus, she said, it is important to make the situation clear to the public and emphasize that instead of looking to researchers for answers right now, the public should recognize that what is needed is much more funding to do the research to find those answers.
Linda Birnbaum of NIEHS responded that it is not necessary to wait for 100 percent understanding and certainty before taking action. A lot of information is available from mechanistic studies, animal studies,
observational studies, and the few clinical intervention trials that have been done, she said. “You put it all together and there are certainly some warning flags that we need to look at.”
Responding to Corkey’s question about advice to individuals on what they should do differently, Birnbaum said that she believes that individual behavioral change is much less effective in the long term than public policy changes. A major issue that needs to be addressed, she said, it that while fresh foods are the healthiest option, they are not available to the entire population, and this situation particularly affects people from disadvantaged backgrounds, who are most at risk of obesity.
Young asked a question about the presence of small amounts of glyphosate—the pesticide sold as Round Up—in foods, particularly corn and high-fructose corn syrup. A paper published by Shehata and colleagues (Shehata et al., 2013) reported that many pathogenic gut bacteria are resistant to glyphosate, whereas many beneficial bacteria are susceptible to it. Could the presence of glyphosate in corn syrup be a confounding variable in the results showing a link between fructose and metabolic dysfunction? Goldman answered that she has not seen any studies addressing the issue, but that it is a reasonable question to ask.
Ajslev, T. A., C. S. Andersen, M. Gamborg, T. I. A. Sørensen, and T. Jess. 2011. Childhood overweight after establishment of the gut microbial: The role of delivery mode, pre-pregnancy weight and early administration of antibiotics. International Journal of Obesity 35:522–529.
Atkinson, R. L., N. V. Dhurandhar, D. B. Allison, R. L. Bowen, B. A. Israel, J. B. Albu, and A. S. Augustus. 2005. Human adenovirus-36 is associated with increased body weight and paradoxical reduction of serum lipids. International Journal of Obesity 29:281–286.
Basu, S., P. Yoffe, N. Hills, and R. H. Lustig. 2013. The relationship of sugar to population-level diabetes prevalence: An econometric analysis of repeated cross-sectional data. PLoS ONE 8(2):e57873.
Browning, J. D., L. S. Szczepaniak, R. Dobbins, P. Nuremberg, J. D. Horton, J. C. Cohen, S. M. Grundy, and H. H. Hobbs. 2004. Prevalence of hepatic steatosis in an urban population in the United States. Hepatology 40(6):1387–1395.
Dhurandhar, N. V. 2012. Insulin sparing action of adenovirus 36 and its E4orf1 protein. Journal of Diabetes and Its Complications 27(2):191–199.
Dhurandhar, N. V., L. D. Whigham, D. H. Abbott, N. J. Schultz-Darken, B. A. Israel, S. M. Bradley, J. W. Kemnitz, D. B. Allison, and R. L. Atkinson. 2002. Human adenovirus Ad-36 promotes weight gain in male rhesus and marmoset monekys. The Journal of Nutrition 132:3155–3160.
Ebbeling, C. B., H. A. Feldman, V. R. Chomitz, T. A. Antonelli, S. L. Gortmaker, S. K. Osganian, and D. S. Ludwig. 2013. A randomized trial of sugar-sweetened beverages and adolescent body weight. New England Journal of Medicine 367:1407–1416.
Fowler, S. P., K. Williams, R. G. Resendez, K. J. Hunt, H. P. Hazuda, and M. P. Stern. 2008. Fueling the obesity epidemic? Artificially sweetened beverage use and long-term weight gain. Obesity 16(8):1894–1900.
Krishnapuram, R., E. J. Dhurandhar, O. Dubuisson, H. Kirk-Ballard, S. Bajpeyi, N. Butte, M. S. Southern, E. Larsen-Meyer, S. Chaley, B. Bennett, A. K. Gupta, F. L. Greenway, W. Johnson, M. Brashear, G. Reinhart, T. Rankinen, C. Bouchard, W. T. Cefalu, J. Ye, R. Javier, A. Zuberi, and N. V. Dhurandhar. 2011. Template to improve glycemic control without reducing adiposity or dietary fat. American Journal of Physiology, Endocrinology, and Metabolism 300(5):E779–E789.
Na, H.-N., and J.-H. Nam. 2012. Adenovirus 36 as an obesity agent maintains the obesity state by increasing MCP-1 and inducing inflammation. Journal of Infectious Diseases 205(6):914–922.
Pepino, M. Y., C. D. Tiemann, B. W. Patterson, B. M. Wice, and S. Klein. 2013. Sucralose affects glycemic and hormonal responses to an OGTT. Diabetes Care 36(9):2530–2535.
Schwimmer, J. B., R. Deutsch, T. Kahen, J. E. Lavine, C. Stanley, and C. Behling. 2006. Prevalence of fatty liver in children and adolescents. Pediatrics 118(4):1388–1393.
Shehata A. A., W. Schrödl, A. A. Aldin, H. M. Hafez, M. Krüger. 2013. The effect of glyphosate on potential pathogens and beneficial members of poultry microbiota in vitro. Current Microbiology 66(4):350–358.
Suez, J., T. Korem, D. Zeevi, G. Zilberman-Schapira, C. A. Thaiss, O. Maza, D. Israeli, N. Zmora, S. Gilad, A. Weinberger, Y. Kuperman, A. Harmelin, I. Kolodkin-Gal, H. Shapiro, Z. Halpern, E. Segal, and E. Elinav. 2014. Artificial sweeteners induce glucose intolerance by altering the gut microbiota. Nature 514(7521):181–186.
Sylvetsky, A. C., M. Greenberg, X. Zhao, and K. I. Rother. 2014. What parents think about giving nonnutritive sweeteners to their children: A pilot study. International Journal of Pediatrics Article ID 819872, 5 pages.
Trasande, L., J. Blustein, M. Liu, E. Corwin, L. M. Cox, and M. J. Blaser. 2013. Infant antibiotic exposures and early-life body mass. International Journal of Obesity 37:16–23.