During the second half of the first session, two presenters explored the current knowledge on the known and suspected risk that chronic diseases pose to the development and severity of infectious diseases. Through the presentation of two case studies, they also discussed the knowledge gaps and implications of the cases on public health policy and practice. The first case was presented by Christoph Thaiss, assistant professor of microbiology at the University of Pennsylvania Perelman School of Medicine. He discussed the association between metabolic syndrome and the risk of enteric infection. The second case was presented by Julia Critchley, professor of epidemiology at the Population Health Research Institute, St. George’s University of London, who examined the epidemiology and public health consequences of the converging epidemics of tuberculosis (TB) and diabetes. This part of the session was moderated by Kent Kester, vice president and head of translation science and biomarkers at Sanofi Pasteur.
Christoph Thaiss, assistant professor of microbiology at the University of Pennsylvania Perelman School of Medicine, explained that environmental factors influence susceptibility for many common diseases, including metabolic diseases like obesity and diabetes as well as general inflammatory diseases. A 2015 study of tens of thousands of people in Denmark concluded that a high body mass index (BMI) is associated with an increased risk for
infections overall (Kaspersen et al., 2015). The investigators subdivided the infections that were studied into different organ systems and infectious agents, determining that this association holds true across various diseases—respiratory infections, skin infections, mucosal infections, and gastrointestinal infections. From the perspective of scientists, the lack of scientific mechanistic data to explain the association between BMI and infections was an important one to address.
Association Between Obesity and Intestinal Barrier Dysfunction
To address this gap, Thaiss and his group worked with the db/db mouse model, which has a mutation in the leptin receptor that prevents the mouse from feeling satiety (Thaiss et al., 2018). Even when the mouse is fed a normal diet, it becomes morbidly obese. Compared with healthy wild-type control mice, they found that db/db mice experience a leaky gut phenomenon. They observed that, in these mice, the microbial molecules have triggered some of the receptors of the innate immune system (such as the toll-like receptors and the nod-like receptors) in their serum, spleen, and livers. The signal created by triggering these receptors was always much higher in the db/db mice, which indicates translocation of microbial molecules across the gut barrier to systemic sites (Thaiss et al., 2018). To look in more detail at the intestinal barrier, the researchers used RNA sequencing to quantify the expression of ZO-1, one of the main proteins involved in creating a tight barrier in the gastrointestinal tract. This confirmed that in addition to being morbidly obese, the db/db mice have poor barrier function in the gastrointestinal tract (Thaiss et al., 2018).
To explore the relationship between intestinal barrier dysfunction and enteric infection, they infected the mice with a bioluminescent version of a mouse pathogen called Citrobacter rodentium, which is the equivalent of pathogenic Escherichia coli in humans. The db/db mice were massively colonized and, over time, there was a huge outgrowth of these bacteria in the morbidly obese host (Thaiss et al., 2018). In addition to bacterial growth, the microbial molecules translocated to systemic tissues like the mesenteric lymph nodes, the spleen, and the liver. Thaiss explained that this type of massive colonization never occurs in wild-type mice because Citrobacter rodentium is a self-limiting infection that stays in the gastrointestinal tract unless there is a barrier problem.
Hyperglycemia Drives Susceptibility to Enteric Infection
Thaiss explained that to test the presumed association between obesity and enteric infection, the next step was a paired feeding experiment. They used this experiment to prevent obesity by limiting the amount of food that
the db/db mice could eat and pairing it to the amount of food the wild-type mice had eaten. Even though the db/db mice did not become obese, they were still susceptible to the infection, which meant that obesity was not a sufficient explanation (Thaiss et al., 2018). To determine that hyperglycemia—rather than obesity per se—was driving the susceptibility to enteric infection, they used a type 1 diabetes model to generate hyperglycemia in wild-type mice. They injected the mice with the toxin streptozotocin, which kills all of the insulin-producing pancreatic cells. In this model, despite not being obese, hyperglycemia rendered these mice highly susceptible to the same enteric infection (Thaiss et al., 2018).
The results suggested that hyperglycemia drives intestinal barrier dysfunction, said Thaiss. As observed in the db/db mice, the streptozotocin-treated mice had systemic colonization by the bacteria in their spleens and livers, indicating a massive intestinal barrier problem and abrogated levels of E-cadherin, a cell–cell adhesion molecule that is a component of intestinal barrier stability (Thaiss et al., 2018).
Epithelial Glucose Transporter 2 Deficiency Restores Barrier Function
These mice studies also provide insight into the mechanisms underlying glucose transport. Typically, glucose enters into the body co-transported with sodium (Navale and Paranjape, 2016). Then, it is transported along a passive concentration gradient—called glucose transporter 2 (GLUT2)—into the bloodstream by exiting the intestinal epithelial cells (Navale and Paranjape, 2016). The hypothesis was that in the case of metabolic disease, in which the host is hyperglycemic and has high concentrations of glucose in the blood, the concentration gradient is flipped (Thaiss et al., 2018). Instead of leaving the intestinal epithelial cells, glucose is now accumulating in the intestinal epithelial cells and causing damage to the barrier. To test this hypothesis, Thaiss and colleagues generated mice lacking GLUT2, which would prevent glucose flux into intestinal epithelial cells. GLUT2-lacking mice were then treated with streptozotocin to render them hyperglycemic; the researchers noted that, at least partially, this prevented the barrier dysfunction problems. The microbial molecule assay revealed that the accumulation of these molecules was no longer occurring. The mice were also protected from systemic spread of bacteria in the spleen and liver when compared to controls.
Epithelial Glucose Metabolism Influences Barrier Function
Thaiss ended his presentation by explaining how epithelial glucose metabolism may influence barrier function in a person with diabetes. As glucose follows a retrograde flow into intestinal epithelial cells, it leads to glucose metabolism in the cell and transcriptional and epigenetic
reprogramming. This ultimately alters barrier function, which allows microbial molecules to enter the systemic circulation (Thaiss et al., 2018).
To begin exploring the mechanisms in humans, Thaiss’s group assessed a small cohort of healthy volunteers. They found that having high levels of hemoglobin A1c (HbA1c), an indicator of long-time glycemic control, was strongly correlated with having high abundance of microbial molecules, which is indicative of a leaky barrier in the gut. Interestingly, BMI was not correlated with an increased systemic presence of microbial molecules. In humans as in mice, it appears that it is specifically hyperglycemia—not obesity or metabolic disease per se—that is causing this phenotype (Thaiss et al., 2018). Expanding this human analysis will shed further light on the importance of glycemic control and the prevention of enteric infection, he said.
Julia Critchley, professor of epidemiology at the Population Health Research Institute, St. George’s University of London, examined the epidemiology and public health consequences of the converging epidemics of TB and diabetes. Critchley began by describing the global burden of both conditions. TB is the leading cause of mortality from a single infectious agent worldwide, causing approximately 1.6 million deaths in 2017, 95 percent of which occur in low- and middle-income countries (WHO, 2018a). More than 10 million active TB disease cases are diagnosed annually, and it is likely that about one-quarter of the global population is infected with latent TB infection (WHO, 2018a). Around 425 million people worldwide are estimated to have diabetes, primarily type 2 (International Diabetes Federation, 2017). However, many of those people remain undiagnosed, particularly in low- and middle-income countries where the most rapid increases of diabetes burden are occurring (Zhou et al., 2016). The number of people living with diabetes is predicted to increase to 629 million over the next few decades, affecting about 1 in 10 adults (International Diabetes Federation, 2017).
Critchley explained that the prevalence of diabetes is increasing because of known risk factors that have been increasing in most parts of the world, including (1) changes in urbanization, (2) change in lifestyles, (3) reductions in physical activity, (4) less healthy diets, and (5) increases in obesity (Zimmet, 2017). She noted that population aging worldwide, especially in Asia where diabetes is common, is an important factor hypothesized to lead to the predicted rise in diabetes over the next few decades, but this is often not acknowledged. Therefore, simply reducing obesity would not have a substantial effect on the huge burden of diabetes that will emerge in the coming decades, she said. The regions of the world with the largest predicted increases in diabetes burden over the next few decades—such as sub-Saharan
Africa and Southeast Asia—are also the regions where TB is endemic (International Diabetes Federation, 2017).
Critchley noted that the population impact of the converging TB and diabetes epidemics could be large (Awad et al., 2019a), as people with diabetes can also transmit TB to other people in their families and communities who do not have diabetes. Working with mathematical modelers, Critchley developed two conceptual frameworks taking into account all of the pathways by which diabetes might be affecting TB. Based on these frameworks, she observed that in 1990 approximately 14 percent of TB-related mortality and 11 percent of TB incidence were statistically attributable to diabetes in India (Awad et al., 2019b). However, as diabetes prevalence rises and TB incidence starts to decline, the model suggests that the proportion of TB that is statistically attributable to diabetes is likely to increase dramatically over the next few decades (Awad et al., 2019a). This model predicts that more than 40 percent of TB deaths and about one-third of TB incidence could be attributable to diabetes in India by the year 2050 (Awad et al., 2019b).
Association Between Diabetes and Tuberculosis
Critchley emphasized that examining the associations between diabetes and TB is a matter of urgent concern. Historically, Richard Morton was the first to highlight this association in the seventeenth century (Morton, 1694; Olayinka et al., 2013). In the 1950s, joint treatment for TB and diabetes was delivered in special clinics in the United Kingdom (Luntz, 1954, 1957; Walker and Unwin, 2010). Attention dissipated when new drugs for both conditions were introduced and substantially reduced their death rates. Renewed interest in the association has been driven by the sharply increasing burden in low- and middle-income countries, leading to updated reviews synthesizing the evidence about the strength of the association between diabetes and risk of developing active TB disease (Stevenson et al., 2007; Jeon and Murray, 2008; Al-Rifai et al., 2017). Evidence from observational studies, prospective cohorts, and case control studies shows that people with diabetes have roughly double the risk of developing active TB (Restrepo et al., 2011). This association is not as powerful as the one between HIV and TB, Critchley noted, but so many people are living with diabetes that it is likely to be meaningful at the population level. Prospective cohort studies suggest a slightly stronger risk in low-income settings with high incidence as well as a higher risk in Asia compared with Europe or the United States (Suwanpimolkul et al., 2014). Studies with better case definitions through microbiological confirmation of TB and blood glucose or HbA1c testing for diabetes also suggest higher risks (Jimenez-Corona et al., 2012).
Glycemic Control, Tuberculosis Risk, and Tuberculosis Outcomes
Critchley said that evidence is limited about the associations between glycemic control, TB risk, and TB outcomes (Leung et al., 2008; Leegaard et al., 2011; Pealing et al., 2015; Lee et al., 2016). Observational data suggest a higher TB risk among people with poorly controlled diabetes, but no randomized controlled trials have looked into this (Lee et al., 2017). Although glycemic control may be helpful in reducing the risk of infections, direct evidence is not available (Lee et al., 2017). Primary care data collected in the United Kingdom from 2010 to 2015, including more than 85,000 patients with diabetes and more than 150,000 controls, show a nonlinear association between glycemic control (measured by HbA1c levels) and the risk of serious infections (Critchley et al., 2018). This is particularly observed once HbA1c levels reach about 9 millimeters per liter (Critchley et al., 2018). In the United Kingdom, which does not have a large burden of TB (except in London), she reported an estimated one-quarter of TB patients with diabetes could be statistically attributable to poor glycemic control. However, she noted that this does not prove that glycemic control would help reduce the risk.
Multiple systematic reviews have looked at TB treatment outcomes among people with diabetes. A 2011 review suggested that diabetes worsens treatment outcomes among TB patients, but the included studies were relatively poor in quality and relied on observational data (Baker et al., 2011). A more recent review of 102 studies included 44 studies that reported on mortality, including 56,122 individuals with TB-diabetes and 243,035 with TB (Huangfu et al., 2019). This meta-analysis suggests that the risk of poor treatment outcomes (i.e., death and relapse) are roughly doubled for someone with diabetes compared to a TB patient without diabetes. The better-designed and analyzed studies that adjusted for key epidemiological confounders showed a greater risk of poor outcomes.
Risk of Diabetes Among Tuberculosis Patients
Diabetes is known to be common in TB patients, Critchley explained. A recent meta-analysis of 200 studies reported that overall about 15 percent of patients with active TB also have diabetes—this varied from 0.1 percent in Latvia to 45.2 percent in the Marshall Islands (Noubiap et al., 2019). Certain hotspots with high prevalence of diabetes in TB patients were South India (54 percent), Kiribati in the Pacific Islands (37 percent), and the southern Texas–Mexican border (37 percent) (Restrepo et al., 2011; Viney et al., 2015; Kornfeld et al., 2016). She added that studies in sub-Saharan Africa have not yet identified a high prevalence of diabetes in TB patients, however.
Critchley said that the TANDEM consortium looked at diagnosing diabetes in TB patients and found substantial heterogeneity across the four study
populations in Indonesia, Peru, Romania, and South Africa (Grint et al., 2018). Although it is possible to develop screening strategies that can help detect diabetes in TB patients, it is more difficult in some populations than in others. In some populations, researchers seemed to be identifying many people with stress hypoglycemia, increases in glucose levels, or increases in HbA1c associated with TB infection. This is not necessarily diabetes, she stated, although it may develop into diabetes in the future, but longitudinal follow-up data are not available.
Screening for Tuberculosis Disease in Patients with Diabetes
Although the need to screen TB patients for diabetes has been broadly accepted, said Critchley, there is less evidence and more uncertainty about the benefits of screening people with diabetes for active TB disease. She noted that there is a need for simple and cost-effective screening algorithms, but for these patients screening may need to be repeated owing to infection-related stress hyperglycemia (Lin et al., 2019). A major concern is what happens to patients identified with diabetes at the end of TB treatment, she added. TB treatment is relatively more accessible, and it is possible to identify undiagnosed diabetes during TB treatment. However, diabetes services are much harder to access in many settings, owing to cost or poor quality of public services; longitudinal follow-up is rarely available in these settings. She said, “We are only really helping people if we can manage their diabetes better after the end of TB treatment.”
Critchley summarized her presentation making the following main points:
- Diabetes increases the risk of TB disease (even when the mechanism is not fully understood), and it increases the risk of poor TB treatment outcomes, particularly mortality.
- Diabetes is common in TB patients and often goes undiagnosed.
- Higher glucose and HbA1c levels may increase the risk of developing TB and are related to negative TB treatment outcomes, but there is no evidence to support that this can be reversed.
- Screening for and managing diabetes during TB care would not be hypothesized to have a substantial effect on the overall burden—the estimated effect is less than 1 percent overall. However, strategies would have a larger effect if they were designed further upstream to reach people with diabetes before they develop active TB, or even better, to reduce the risk of developing diabetes and more severe hypoglycemia.
She said that confronting the converging epidemics will require addressing many heterogeneities with respect to diabetes, TB-diabetes, interactions with HIV, and other multimorbidities.
After both presentations, Kent Kester, vice president and head of translation science and biomarkers at Sanofi Pasteur, started the discussion by asking the presenters about the effect of sustained hyperglycemia on the immune system. Kester noted that hyperglycemia is known to have a negative effect on certain elements of the immune system, such as the inhibition of certain leukocytes (Graves and Kayal, 2008). Given that TB is related to impaired T-cell function and elevated HbA1c levels are typically a marker of sustained elevations in glucose over time, he asked about any insights into the association between diabetes and increased susceptibility to TB or worsening of TB disease.
Critchley replied that diabetes is a state of relative immune impairment that is associated with increased risks of infection, which warrants investigation into the immune system responses in both conditions. Immunologists in the TANDEM consortium have been looking at gene expression and have found different signatures for people with TB and diabetes compared to people with TB alone. An unexpected finding was that the signatures were also different for people with impaired HbA1c levels that were above normal, but below the diagnostic cutoffs for diabetes. Another group of researchers is looking at the benefits of host-directed therapies using statins or metformin in reducing TB risk or improving TB treatment outcomes; no randomized evidence is available as of yet, but there are animal and observational studies that suggest metformin may be associated with a reduced risk of TB (Lee et al., 2018; Tseng, 2018). However, Critchley cautioned about drawing conclusions from the observational evidence because of the potential for selection bias, as metformin is the first-line drug treatment for diabetes almost everywhere in the world. People who do not have access to metformin have different circumstances than people who do, whether the barrier is poverty, lack of insurance, socioeconomic status, severity of diabetes, or type 1 diabetes that is misdiagnosed as type 2. Randomized controlled trial evidence is needed, but getting this evidence is challenging owing to the ethics and pragmatics of running such a trial.
Kester also asked Thaiss if they were able to restore some of the barrier effect in their mice studies, while still controlling for the impaired leukocyte function when hyperglycemia is present. Thaiss replied that the restoration achieved by preventing epithelial dysfunction is not complete, so some of the residual problems may be accounted for by impaired immune function. When looking at neutrophils in a study, they observed differences that may result from a combination of both immune and nonimmune functions.
Correlation Between HbA1c and the Microbial Signal
Tolu Oni, clinical senior research associate in the MRC Epidemiology Unit at the University of Cambridge, asked Thaiss to elaborate on the strong
correlation between HbA1c (rather than glucose itself) and the microbial signal, along with potential implications for screening and management, as well as potential effects on the clinical outcomes. Thaiss replied that they did not expect to see both factors uncoupled, with a signal for HbA1c but not for glucose itself. This suggested that the association observed was with long-term glycemic control, not postprandial glucose responses, which is a meaningful distinction, considering the biology that underlies their differential regulation. His study included only healthy individuals without diabetes. Nonetheless, even within a small healthy cohort with healthy HbA1c ranges, they still saw a correlation with microbial molecules in the circulation system, which is indicative of barrier dysfunction (Thaiss et al., 2018). The implications for treatment are that the current approaches to long-term glycemic control, like metformin, might be more applicable than those for short-term glycemic control.
Critchley added that the TANDEM study included blood glucose and HbA1c testing, and the results were similar in terms of screening accuracy, although their data mainly identified people who had already been diagnosed with diabetes. She added that data from studies in Africa have found associations of different strength with HbA1c versus blood glucose markers of diabetes. She speculated that this difference may be caused by two different populations in these studies. One group includes people with TB who have undiagnosed diabetes that could be picked up with any test because their levels are high. The other group may include people for whom HbA1c is actually a marker of TB severity, rather than a marker of true diabetes. This could account for why HbA1c levels are correlated with poor outcomes. Her immunologist colleagues are looking at differences in gene expression that could provide a way to understand this biologically, although this may be less useful on the ground in clinics.
Converging Tuberculosis and Diabetes
The discussion turned to the socioeconomic dimension of the TB-diabetes convergence. Gene Bukhman, director of the Program in Global Noncommunicable Disease and Social Change at Harvard Medical School, asked whether any studies are collecting socioeconomic information about patients with both TB and diabetes, noting that in India, there are usually reverse gradients with diabetes and socioeconomic status. He wondered if the story with TB is more complicated or in line with expectations. Critchley replied that most of the studies she presented were based on clinical data from hospital or national TB database registries; they rarely collected any socioeconomic data. However, the TANDEM study that recruited 2,000 patients across Indonesia, Peru, Romania, and South Africa did collect detailed socioeconomic data (Grint et al., 2018). As expected, TB was
shown to be a disease of poverty. In Indonesia, for example, asking people if they had a bank account was almost as good a predictor as microbiological testing for TB. The situation with diabetes is relatively more complex as it becomes more common among all socioeconomic gradients—or even, as data from the population group studied by TANDEM suggest, that it is beginning to reverse and may even become more common in low- to middle-socioeconomic groups.
Patricia García, professor at the Cayetano Heredia University School of Public Health, moved the discussion to real-world practicalities of screening for TB-diabetes. She used as an example settings without the laboratory capacity to provide reliable HbA1c results, or where the only test available requires fasting on a person with TB receiving treatment. Most low-resource TB patients in Peru receive a food basket with high-carbohydrate foods to supplement their diets, she noted. It may be more complicated or expensive to provide appropriate food baskets for low-resource people with TB-diabetes. Critchley said that from a practical point of view, TB should generally be treated as the first priority, in order to reduce the risk of death from the infection or the transmission of drug resistance. Evidence is limited about the use of point-of-care tests for diabetes in TB clinics, but blood glucose testing could be a starting point because it is relatively inexpensive. Thinking about improving health rather than focusing only on disease outcomes requires ensuring accessibility and continuity of care for people being treated for TB who are identified or suspected as having diabetes. Once their TB has been cured, she said, having the knowledge and resources to effectively manage their diabetes contributes to long-term health outcomes. García asked about any associations between nonpulmonary TB and diabetes. Critchley replied that most of the studies have only looked at pulmonary TB and diabetes, which have the strongest evidence for an association, but the association between diabetes and less common extrapulmonary TB is unknown.
Building on García’s comments, Marcos Espinal, director of communicable diseases and health analysis at the Pan American Health Organization, noted that there is a need for simple initiatives to strengthen health services that are accessible to the entire population. He asked Critchley about her experience with any efforts to synergize diabetes and TB treatment that could be potentially replicated on a broader scale to improve health system responses and patients’ outcomes, as well as be more attractive to policy makers. Critchley noted efforts to integrate screening in places like India, but they are too recent to assess long-term outcomes. Blood glucose testing can be easily collected for TB patients, but longitudinal data are not available to assess patient experiences and outcomes. Initial steps around integration have been made in a limited number of places, but it is not yet clear how much they improve the experience for the individual.
Mosa Moshabela, dean and head of the School of Nursing and Public Health at the University of KwaZulu-Natal, South Africa, asked Critchley about the relative difficulty of studying the epidemiological patterns of patients with TB-diabetes and, given that TB is more immediately life-threatening than diabetes, about the opportunities to integrate care for those patients. Critchley said that from an epidemiological point of view, it is relatively straightforward to study TB-diabetes using evidence from registries and cross-sectional data. Data would improve if funders were prepared to invest in longitudinal studies. Reorganizing health systems to deliver integrated care could be supported by the recent sets of guidelines that provide simple frameworks for primary care clinics in rural settings to deal with suspected diabetes. A historical example comes from the joint TB and diabetes clinics in the United Kingdom that were apparently successful in reducing the mortality rate from those two joint conditions in the 1950s. Economic evidence is needed on the cost-effectiveness of integrating or reorganizing the health services to address the joint epidemics, she said. She estimated that only half of existing TB clinics would have access to blood glucose testing, so it would be helpful to look at the cost-effectiveness of putting point-of-care tests in place in primary care, for example.
Ultimately, however, cost-effectiveness comes down to providing long-term follow-up care for patients—identifying someone who had undiagnosed diabetes during TB care may or may not actually benefit the person’s long-term health, depending on whether the person can later access high-quality diabetes services to support good glycemic control, she stated. Evaluation of cost-effectiveness should take into account the patients’ journeys or health outcomes. For instance, some have speculated that the TANDEM study may have picked up people who were not actually undiagnosed—they knew they had diabetes, but they did not have access to care. The reasons why people drop out of care may be economically driven or related to lack of awareness of the importance of long-term care. She was optimistic that TB services could potentially provide a way for some patients to reintegrate with care, but only if the services are there to engage them at the end of TB treatment. She reflected:
That is where the synergy most concerns me.… I can pick up and evaluate the epidemiological evidence, but the carrying on into the future and what happens to those patients is really a big gap.
This page intentionally left blank.