CORDIS - EU research results
CORDIS

Concurrent Tuberculosis and Diabetes Mellitus;<br/>unraveling the causal link, and improving care

Final Report Summary - TANDEM (Concurrent Tuberculosis and Diabetes Mellitus;unraveling the causal link, and improving care)

Executive Summary:
Tuberculosis (TB) is the most important cause of death from an infectious disease, and has enormous social and economic impact. Diabetes Mellitus (DM) is associated with a higher TB risk and worse TB treatment outcome. With around 10% of the world population living with DM, and DM dramatically increasing in prevalence in TB endemic countries, DM seriously threatens global TB control. Therefore, there is an urgent need for optimal and cost-effective screening and management of combined TB and DM, and to understand the underlying cellular and molecular mechanisms. The TANDEM consortium has addressed these key issues, combining field studies in countries that have a high TB incidence and growing DM prevalence with basic laboratory sciences.

Among over 2000 TB patients, 15% were found to have Type 2 DM. DM was either previously known (yet often poorly managed), or newly diagnosed. To screen TB patients for DM, a two-step screening process (using non-fasting blood glucose followed by a POC HbA1c for those above a threshold on the initial test) was found most cost-effective and feasible. Poor management of DM was associated with higher TB risk, providing yet another reason to step up global efforts to improve proper management of DM. Among over 2000 DM patients, only 14 cases of TB were newly identified, questioning the cost-effectiveness of screening for TB in DM clinics. A higher ratio of active TB: latent TB infection among individuals with DM suggests that prioritising individuals with DM for TB prophylaxis might be a component of intensified TB control in high TB incidence countries.

With regard to clinical management, TANDEM found that DM was associated with higher TB treatment failure, relapse and mortality, especially among those with poor glycemic control. Reassuringly, more intensive glucose monitoring with standardised adjustment of glucose-lowering treatment led to significantly better glycaemic control. Metformin as the first choice drug for DM showed no relevant interaction with rifampicin. Under routine conditions, glycaemic control deteriorated after completion of TB treatment, and significant and unmet needs in terms of prevention of cardiovascular disease in these individuals.

From a pathophysiological point of view, TANDEM found that DM was associated with enhanced and prolonged inflammatory gene expression and altered immune cell phenotypes during TB treatment, especially in those with poor glycaemic control. In vitro, high glucose had no major effect on Mycobacterium tuberculosis intracellular growth or immune responses, but altered lipids – also found in diabetic TB patients - and diabetes drugs had. DM is linked to obesity, and fat cells were found to facilitate M. tuberculosis growth. Finally, TANDEM identified genetic markers associated with the interaction of TB and DM, that can be further analysed in a systems biology approach using the vast repository of TANDEM samples and data.

TANDEM had significant scientific and societal impact, reaching a wide audience with 39 scientific publications and 8 conference proceedings so far, as well as 66 oral and 10 poster presentations at international conferences. TANDEM led to significant strengthening of clinical care and academic development of its partners, through completion of 23 MSc and PhD theses. TANDEM also disseminated its concepts and findings to policy makers and a wider public through important stake-holder meetings (organising 2 conferences and workshops) and 4 presentations to the wider public, and involvement in drafting new global guidelines for TB and DM. Finally, TANDEM led to the development of lasting partnerships and new project applications.
Project Context and Objectives:
The TANDEM Consortium was formed to address the alarming global growth of type 2 diabetes mellitus (DM), which was posing a threat to TB control worldwide, as individuals with DM have a roughly 3-fold higher risk of developing TB, with higher rates of death, treatment failure and relapse.

TANDEM addressed an urgent need for basic knowledge to help understand and control the intertwined epidemics of TB and DM. The International Union Against TB and Lung Disease and the World Health Organization published a ‘Collaborative Framework for Care and Control of Tuberculosis and Diabetes’ in 2011. This increased awareness of the potential public health and clinical importance of the relationship between these two diseases, but had not been translated into action. From a clinical-operational point of view important data were lacking to improve care for patients with concomitant DM and TB. Firstly, screening of TB patients for DM could improve case detection, early treatment and secondary prevention of DM. However, such screening was not routinely performed in most settings, and the optimal and most cost-effective approach had yet to be defined. Similarly, some studies had found that diabetes was associated with increased TB treatment failure, death and relapse but it is uncertain if optimal glucose control could reduce these effects and how optimal glucose control is best achieved as both TB and anti-TB medications hamper glucose control in DM (the TB drug rifampicin increases metabolism of most oral anti-diabetic drugs). Finally, no study had examined the treatment needs of TB patients with newly diagnosed DM once TB treatment was completed, in terms of lifestyle interventions and anti-diabetic, anti-hypertensive, lipid-lowering and other medications for TB-associated DM need to be defined.

From a pathophysiological point of view, data on the interaction of TB and DM were scarce. Very few studies have examined the cellular and molecular basis of the increased susceptibility of DM patients for TB. It was not known if DM exerts its effect, for example, through altered function of macrophages, adipocytes, T-cells or other cell types; whether this is genetically determined, influenced by hyperglycemia, or both; and whether TB-DM patients differ in their gene expression signatures during active TB disease and in response to TB treatment compared to non-DM patients. Also, identification of diagnostic biomarker signatures could help early identification of high risk groups, and monitoring of treatment success. Therefore, there was also an urgent need for involvement of basic sciences to help unravel the causal relationship between DM and TB, in order to design more effective strategies for control of both diseases.

In conclusion, at time of writing the application it was clear that the alarming growth of type 2 DM was posing a serious threat to TB control worldwide as well as in Europe. However, there was a lack of (1) evidence to support many of the recently advocated guidelines for care and control of TB and diabetes, and (2) understanding of the mechanisms underlying the effect of diabetes on TB susceptibility and treatment outcome.

The TANDEM consortium addressed both these key issues. Our central hypotheses were that:
• Screening and management of DM among TB patients can be greatly improved and simplified with a major impact on control of TB-DM co-morbidity.
• The effect of DM on TB susceptibility and outcome is regulated, at least partly, by hyperglycemia, which dysregulates the cellular response to M. tuberculosis, and by genetic variation common to both diseases.

To address these hypotheses TANDEM used a comprehensive and integrated approach combining clinical, epidemiological and cutting edge expertise in laboratory sciences, bringing together a multi-disciplinary consortium. Based on multiple previous partnerships (‘TANDEMs’) between researchers from Germany, Indonesia, Netherlands, New Zealand, Peru, Romania, South Africa and the United Kingdom, a strong consortium was formed. Field sites in four countries with high yet variable TB incidence rates and increasing DM prevalence rates – Indonesia, Peru, Romania and South Africa provided an excellent setting for the field studies, and collectively TANDEM partners brought together huge expertise in epidemiology, public health, clinical medicine, microbiology, health economics, molecular biology, immunology and genetics. Protocols were synchronised across field sites and a central electronic database was set up to connect all collected data and combine data analysis.

TANDEM addressed the following four objectives in 4 work packages (WPs):
1. To identify feasible, accurate and cost-effective ways of screening TB patients for diabetes, and determine the prevalence of DM among TB patients and of TB in DM patients in different geographic areas (WP1).
2. To determine the level of DM management required during and after TB treatment, and the effect of glucose control on TB treatment outcome (WP2).
3. To identify key pathways which may account for enhanced susceptibility to, and poorer treatment outcomes of, combined TB and DM by comparing gene expression and biomarker profiles in TB patients with and without DM through treatment (WP3).
4. To establish the cellular and molecular basis responsible for the causal link between diabetes and TB, and in particular to determine the effect of hyperglycemia and genetic variation on the host protective response to M. tuberculosis (WP4).

Each of the 4 scientific work packages had a number of sub-objectives, as outlined below:

Work package 1 addressed screening of TB patients for DM and vice versa, aiming to:
• evaluate the most accurate and feasible strategies for screening pulmonary TB patients for DM
• determine the prevalence of DM in newly diagnosed TB patients, in distinct locations, and stratified by gender and age group
• determine the prevalence of active and latent TB in DM patients, in distinct locations, and stratified by gender and age group
• investigate the proportions of DM-associated TB in TB-endemic settings that are attributable to reactivation, or to recently acquired infection

Work package 2 addressed clinical management of diabetes in TB patients, aiming to:
• define the requirements for glycaemic control during TB treatment, as well as the long-term requirements for diabetes management in TB patients diagnosed with diabetes
• determine safety, tolerability, effectiveness and pharmacokinetics of metformin as a hypoglycaemic agent when combined with TB treatment
• examine the cost-effectiveness of different strategies for screening and diabetes management during TB treatment
• determine the relation between glycaemic control and clinical-microbiological response to TB treatment.

Work package 3 address differences in biomarker profiles between TB and TB-DM aiming to:
• determine blood biomarker profiles which discriminate TB-DM from uncomplicated TB patients
• characterise TB treatment-response biomarker profiles in TB-DM compared to TB patients
• establish whether enhanced glycaemic control affects normalization of host biomarker profiles
• investigate the possible prolonged mycobacterial persistence in TB-DM by analysis of memory T cell phenotype

Work package 4 addressed the cellular basis of the link of DM and TB, aiming to:
• evaluate the effect of hyperglycemia on the innate human macrophage response to M. tuberculosis, including intracellular fate of M. tuberculosis
• determine the response of adipocytes to M. tuberculosis infection, immune regulation and insulin stimulation
• determine whether the association of TB and DM is at least partly caused by genetic variation common or unique to both diseases and examine the influence of genetic variants associated with TB-DM or DM on gene function and cellular response to M. tuberculosis

Work package 5 addressed the dissemination, and Work package 6 the management structure of TANDEM.

TANDEM was led by Prof Hazel Dockrell as PI from LSHTM, UK and Prof Reinout van Crevel, scientific coordinator from Radboud UMC – Netherlands. Each work package had a WP leader, and each consortium member a PI. TANDEM was assisted by Eurice, Germany who provided project management as partner 12, and led work packages 5 and 6, on dissemination and management. TANDEM also had an advisory board and an ethics review group.
Consortium meetings were held in London – UK (2013), Cape Town – South Africa (2014), Bandung - Indonesia (2015), Craiova - Romania (2016) and Lima - Peru (2017) with additional laboratory meetings in Leiden - Netherlands, complemented by regular teleconferences.
Project Results:
TANDEM ran for 4.5 years and has amassed a large body of data that inform both its original objectives as well as the questions highlighted by the 2011 IUATLD/WHO framework.
In total, TANDEM recruited 2185 TB patients and 2096 individuals with DM. A centralised database containing anonymised data was set up at St George’s University of London in the first 6 months of the project, using the open access RedCap system. In the field sites electronic data capture used tablets or laptops and pc computers as appropriate from which the data could be uploaded thus removing the errors associated with manual data entry. Laboratory analyses and experiments were also performed in London, Leiden, Groningen, Nijmegen, Berlin and Cape Town.

Workpackage 1. Screening TB patients for DM and vice versa

• Screening TB patients for diabetes
In all, 2185 TB patients were screened for diabetes. We found that the cases of diabetes identified were often previously diagnosed, although 3.0 to 7.1% of those screened had newly diagnosed diabetes. The prevalence of newly-diagnosed diabetes was 7.5% (95% CI 5.8-9.6) in Indonesia, 6.9% (4.4-10.7) in South Africa, 6.6% (4.7-9.3) in Romania, and 3.0% (1.9-4.8) in Peru. The overall crude prevalence of diabetes among newly-diagnosed TB patients was 15.4% (95% CI 13.9-16.9).
TB patients with previously identified DM often presented with very high HbA1c levels (median HbA1c between approximately 10% and 11% in each site), suggesting poor DM control prior to TB onset or substantial stress hyperglycaemia. There was heterogeneity in HbA1c values among newly diagnosed DM patients by site. In Indonesia and Peru, the diabetes identified was more advanced, with higher median HbA1c values of 10.5% and 7.9% respectively; the diabetes cases in Romania and South Africa tended to have early onset or borderline diabetes phenotypes with median HbA1c values of 6.7% in Romania and 6.8% in South Africa.
Due to this heterogeneity, no single strategy could be clearly identified as superior, and local context needs to be considered in designing screening programmes. However, diabetes screening was best diagnosed using a 2-step combination of random (non-fasting) blood glucose followed by point-of-care HbA1c test for those above a specific threshold (6.1 mmol/l), as this reduced the need for the more expensive POC HbA1c test by about 70% in our study population. For example in Indonesia, the 2 step combination had a sensitivity of 90.7 (79.7 to 96.9) and specificity of 95.8 (94.0 to 97.2) but the sensitivity was substantially lower in other sites where HbA1c levels were only marginally raised.
Using POC HbA1c plus age worked well, but would be more expensive if HbA1c testing was performed on the whole population. The optimal diagnostic cut-point for a single POC HbA1c measurement was ≥6.2%, which achieved higher sensitivity (71.8% vs. 65.5%), but lower specificity (83.0% vs. 88.4%) than the standard ≥6.5% cut-point. The combination of using random plasma glucose with age also worked well in Indonesia, and is very inexpensive. We tested three recognised diabetes risk scores but these did not perform as well. We also derived our own TANDEM risk scores which performed well, but requires validation in further study populations.
These results indicate that diabetes can be diagnosed in TB patients using simple point of care tests. Screening TB patients for DM is recommended, as it is likely to be feasible and cost-effective in most places, particularly for TB patients above the age of approximately 35-40. However it was clear that in some of the TANDEM settings, diabetes was being very poorly managed and controlled, putting patients at greater risk of subsequent complications.
As noted in the detailed summary of the TANDEM findings, the fact that 15.4% of TB patients were found to have diabetes confirms that screening of TB patients for diabetes using a suitable DM test should be performed. As almost no diabetes patients were below 35 years of age, it is advisable for screening to be reserved to those older than 35 years old.

• Screening diabetes patients for active TB
A total of 2096 diabetes patients were screened for TB in the four TANDEM field sites. Definite TB was defined as having a positive sputum culture for M. tuberculosis; probable TB as having a positive sputum smear for M. tuberculosis as well as a possible/typical chest x ray, and possible TB as having a positive M. tuberculosis smear together with symptoms suggestive of TB. Very few diabetes patients were found to have definite or probable TB (0.67%); of the 14 patients identified, 12 were from Indonesia. A greater number of diabetes patients had previously been diagnosed with TB, suggesting that lifetime TB risk may still be significant for DM patients (n=85; 4.05%).
As TB incidence is much lower than DM prevalence, even in the highest burden TB countries, the number of TB cases that will be newly identified in diabetes clinics is much lower than the number of people with diabetes identified by screening TB patients. In the four countries where TANDEM recruited patients, the incidence of TB ranges from 159 – 800/100,000 population. From over 2000 diabetes patients screened, only 14 definite or probable cases of TB in DM patients were identified. Logistic and budgetary restrictions limited TANDEM screening of greater numbers of DM patients. Thus although staff in diabetes clinics should be aware of the possibility of TB, particularly in settings where TB incidence is high, the yield from screening DM patients for TB will always be much lower than that from screening TB patients for DM. As the prevalence of TB is significantly higher in individuals with diabetes, we recommend that health care workers involved in DM care are trained to recognise the signs and symptoms of TB, in order to identify TB cases and refer them for appropriate treatment which will also reduce TB transmission within DM clinics. At present, and based on TANDEM data, screening of all DM patients for TB is unlikely to be cost- effective in most low-middle income countries. However, use of chest X-ray (frequently done anyway at time of diagnosing diabetes), with automated reading for TB might change this situation.
One exciting result from TANDEM came from the screening of DM patients for latent TB infection (LTBI). Compared to TB household contacts, a known high risk group for TB disease, the ratio of active TB versus LTBI was much higher in individuals with DM. This suggests a higher rate of disease breakthrough in DM, and an even greater benefit of TB prophylaxis in this group. TANDEM partners have submitted a phase 3 clinical trial of preventive treatment in DM patients in high TB incidence countries to be conducted in 3 countries for funding.

• Observations regarding patient recruitment in TANDEM.
TANDEM identified a lower number of participants with combined TB and DM than expected. This was based on lower than expected DM prevalence rates among TB patients. DM prevalence among TB patients is highly variable between countries and settings, and the published data may be biased towards some ‘hotspot’ countries like Mexico and India.
The predictions of the numbers of cases of combined TB and DM that could be recruited within TANDEM were based on available estimates of prevalence in the four countries where recruitment occurred. An increased number of participants had to be recruited overall (2185 instead of 1500; 145%), and that recruitment had to be extended over a longer period, as well as adding additional hospitals and clinics as recruitment sites in Romania, Indonesia and Peru. Another contributory factor was the rigorous case definitions for TB-DM and the need for patients to be available for follow up. Our conclusion is that future studies should estimate more conservatively possible recruitment of the most critical clinical group. The TANDEM Consortium was grateful for the no-cost extension awarded by the EU which allowed it to reach its initial target of 350-400 diabetic TB patients.
It should be noted that recruitment has proved a problem for other research consortia, for example when predictions of how many TB contacts will progress to develop TB are used to set sample sizes for recruitment. This is partly as a result of sample size analyses based on the only available epidemiological data being used to predict the number of individuals recruited to research studies that comply with the more rigorous definitions used in a research study. For example, although figures for TB prevalence and incidence are usually available, for research studies such as those conducted in TANDEM, particularly when analysing gene expression profiles and genetic heterogeneity, it is important to exclude patients whose TB diagnosis has not been microbiologically confirmed; who have TB confounded with HIV co-infection which has a major impact on immune status; or who have multi-drug resistant tuberculosis. Most previous studies of DM prevalence in TB patients had not carried out repeated DM testing, although strictly this is required to confirm DM diagnosis unless a patient presents with classical DM symptoms. As TB patients present with stress hyperglycaemia, lack of repeated testing may have significantly overestimated DM prevalence among TB patients in some earlier studies. TANDEM also found that some initially recruited patients were lost to follow-up if they moved away, were in prison, or started a new job which interfered with study participation, a few did not wish to continue in the study and withdrew consent and a small number died during the study.
The Consortium combined clinical and epidemiological studies with laboratory studies. Although some of these were free-standing and did not rely on the collection of the clinical samples from the field sites, other parts of the work required all the sample collection to be completed in order to analyse the samples with an appropriate experimental design. For example, as all samples should be processed for RNA extraction and RNA-Seq and analysed together to avoid batch variations, this required all the longitudinal samples to be collected before the samples could be processed and analysed. Secondly, the extracted RNA is less stable than when stored pre-extraction in the PAXgene tubes. Thirdly, having the full dataset and completed database allowed the samples for particular analyses to be selected and randomised. Although major analyses of the RNA-Seq dataset have been completed, the dataset and sample collection provides a resource for both TANDEM members and the community at large for much further work that was not possible within the timescale of TANDEM.
In reality, studies such as those performed by TANDEM require 5-10 years funding which is not possible with current grant funding streams. We are very grateful to all those who participated in TANDEM, whose hard work enabled us to achieve so much within 4.5 years.

Work package 2: Clinical management

• Glucose monitoring and glycemic control of DM in TB patients
The WHO / IUATLD framework on TB and DM advocates screening of TB patients for DM. However, data regarding the management of newly diagnosed DM in patients with active TB are lacking. Tight glycaemic control may improve TB treatment outcome and help reduce symptoms. However, active TB and TB treatment hamper glycaemic control. Patients starting TB treatment experience rapid changes in appetite, body composition, and inflammation (which increases insulin resistance); inflammation is a feature of untreated TB and following an increase as a result of initial bacterial killing, inflammation subsides with successful treatment. In addition, TB medication (rifampicin) increases the metabolism of oral anti-diabetic drugs including the widely used sulphonylureas, though a possible interaction with the anti-diabetic drug metformin had not been previously examined. Frequent monitoring of blood glucose during the course of TB treatment may therefore be needed. However, frequent monitoring is associated with additional costs, and tools and skills for glucose monitoring and diabetes treatment may be lacking in TB or pulmonary clinics, creating a need to refer patients to other health providers. As such, a less intense schedule, preferably following the established decision points in TB treatment after 2 and 6 months, would offer significant advantage. None of these issues have been addressed systematically so far.
TANDEM examined the requirements, feasibility and cost-effectiveness of different strategies of glucose monitoring, using a pragmatic trial comparing ‘standard’ and ‘intensified’ glucose monitoring. TANDEM randomised a total of 150 patients with TB and DM to receive either more intensive monitoring at 2 weeks, 4 weeks and monthly thereafter (3, 4, 5 and 6 months) compared to the normal standard of care with essential evaluations for the trial at 2 and 6 months. Seventy-six patients received the intervention and 74 patients formed the control group. The trial was performed in Bandung, Indonesia, where most of the diabetes in these patients was poorly controlled with most patients having HbA1c values of >10, and approximately a third of the TB-DM patients in both groups had had diabetes for 1-5 years. The primary outcome of the trial was how well glucose concentrations were controlled, measured using the HbA1c test that measures glycated haemoglobin.
We found that the more intensive monitoring greatly improved glucose control, with the mean HbA1c values falling from 11.1% at baseline to 7.7% at 3 months in the intervention group compared to from 11.3% to 9.3% in the control group, a result that was highly significant. We also found that patients in the intervention group had significantly lower HbA1c values at 6 months compared to the controls. This was the first randomised trial comparing more intensive monitoring of glucose control compared to the normal standard of care in patients with both TB and diabetes. Our results show that patients with TB and diabetes can be managed in routine clinics using a formal algorithm and regular scheduled glucose monitoring and that this results in better glucose control.

• Long-term requirements for DM management in TB patients diagnosed with diabetes
Very often, DM is newly diagnosed in TB patients in low-resource (TB-endemic) settings, but data are lacking regarding the prognosis and treatment needs of DM in these patients once TB treatment is completed. Firstly, patients with TB-associated DM are likely to be different from DM patients in the general population: they may be younger, and more likely to be smokers. Secondly, TB-associated inflammation may lead to a transient increase in insulin resistance and DM might reverse following the decrease of inflammation during TB treatment. However, to date no studies have followed patients with TB-associated DM after completion of TB-treatment to establish their health status and risk profile and to see if DM may spontaneously resolve in some. In conclusion, screening TB patients for diabetes may increase diabetes case finding and access to care for diabetes, and this may reduce secondary (e.g. cardiovascular) complications, but the impact of screening on relevant health outcomes will depend on the DM phenotype of patients with TB-associated DM.
TANDEM for the first time defined long term requirements for DM management after completion of TB treatment in terms of life-style interventions (e.g. smoking cessation), and anti-diabetes, anti-hypertensive and lipid-lowering drugs in line with international guidelines.
At 6 months, at time of TB treatment completion, 37% of TB-DM patients were overweight, with central obesity present in 53% of women and 22% of men. Twenty percent of TB-DM patients were smoking; 82% had elevated LDL, yet only 3% were using a statin; 56% of patients had hypertension, yet only 20% were using anti-hypertensive drugs. Only 13% were using insulin, even though 56% had an HbA1c > 10%. At 12 months (6 months after completion of TB treatment), 48% of TB-DM patients were overweight, with 5% central obesity in women, and 29% in men. The proportion of individuals who were smoking had increased to 30%.
These unique data showed that the long-term requirements of DM in patients with combined TB and DM are substantial. Our data clearly indicate that complications (diabetic nephropathy, micro- and macro-vascular complications) are common and were often not managed or only poorly managed at the time of recruitment to TANDEM. In addition, long-term glycaemic control after completion of TB treatment is challenging. There is a substantial proportion of patients requiring long-term insulin, but few who are actually using insulin. Finally, the cardiovascular risk profile is moderately severe, with more patients smoking at the end of TB treatment, and more individuals picking up smoking in the months following completion of treatment. As such there is a clear need for better cardiovascular risk management.

• Metformin as a hypoglycaemic agent when combined with TB treatment
Rifampicin, as a very potent cytochrome (CYP) P450 enzymes inducer, decreases plasma levels of antidiabetic drugs, including sulphonylureas, metiglinides and thiazolidinediones by 30-70%. As a result, glycaemic control is often not reached during TB treatment, while increasing the dose of hypoglycemic agents magnifies the risk of hypoglyaemia, especially if patients do not take their TB treatment consistently. Another antidiabetic drug, metformin, may not have these drawbacks. Metformin is a first choice drug for type 2 DM. It does not lead to hypoglycaemia, is relatively cheap, widely available and associated with less weight increase compared to other oral antidiabetic drugs. Another advantage is that metformin is not metabolised by P450 enzymes and thus its levels should not be affected by rifampicin. Possible disadvantages when combined with anti-tuberculous drugs include gastrointestinal side effects, possibly leading to non-adherence and poor TB treatment outcome, but this had not been studied. Similarly, no study had examined possible pharmacokinetic interactions with rifampicin and isoniazid, although these seemed unlikely as mentioned above.
TANDEM examined the impact of rifampicin on metformin pharmacokinetics and its glucose-lowering effect in 31 Indonesian TB patients with DM, by comparing metformin blood concentrations when combined with rifampicin, and without rifampicin (shortly after completion of TB treatment). Metformin drug concentrations were roughly 1.3 fold higher when co-administered with rifampicin. The higher metformin total and peak exposure is probably the result of altered metformin absorption kinetics under influence of rifampicin exposure, because metformin urinary clearance and tubular secretion was not affected by rifampicin. The observed pharmacokinetic interaction did not result in a statistically significant change in the glucose-lowering effect of metformin. However, (mild) gastro-intestinal adverse effects (29%) were common. The addition of metoclopramide to the protocol and the separation in time of metformin and TB drug intake lowered the risk of side effects.
These data show that metformin can be used in combination with rifampicin in its standard dose, and that side effects occur, especially if drug intake is done simultaneously with TB drugs.

• Cost-effectiveness of different strategies for screening and DM management during TB treatment
No data have been published on the economic impact of the recommended screening and management of DM in TB patients. The anticipated value of early DM diagnosis and more intense monitoring (as opposed to standard monitoring) derives from better, earlier glycaemic control, improving health outcomes for patients. These may increase or decrease overall treatment costs of DM and associated co-morbidities.
TANDEM assessed the cost of carrying out various tests for TB and diabetes in the different country settings. For diagnosing TB the gold standard is to use sputum smears followed by sputum culture. A total of 10 TB testing algorithms were assessed based on combinations of the four screening and diagnostic tests (symptom screen, chest x-ray (CXR), sputum smear and culture). We found that the cheapest algorithm per accurate diagnosis in Indonesia was the TB symptom screen followed by two Ziehl-Neelsen stained sputum smears, while the complete algorithm of symptom screen, chest X-ray, two sputum smears followed by two microscopic-observation drug susceptibility (MODS) culture tests had the highest cost per accurate diagnosis.
For diagnosing diabetes in people with TB, 24 screening and diagnosis algorithms were assessed. The DM testing algorithm with the lowest cost per accurate diagnosis was age screen followed by point of care random plasma glucose tests. One of the highest costs per accurate diagnosis in the two countries was the point of care HbA1c test followed by two laboratory HbA1c tests.
The burden of out-of-pocket payments and lost productivity was 2.5 times greater for patients in the intensive DM monitoring arm than those in standard DM care, comprising 2% and less than 1% of the average monthly income, respectively. These proportions are considered to be well below the threshold for catastrophic costs, suggesting that there has been success in bringing TB services close to the patient in Indonesia.
People with TB only reported a better health related quality of life (HRQoL) at baseline than people with combined TB and DM in Peru and Romania. No difference in HRQoL between these two groups was found in Indonesia, which also had the largest sample size. There is weak evidence in Indonesia that people undergoing intensive DM monitoring have better HRQoL than those in standard care after six months of TB treatment and DM management. The association was further attenuated after adjustment for clinical and demographic confounders. Socio-economic quintiles were also used to assess the differential effects of socio-economic status on HRQoL of patients with TB and TB-DM at baseline. Patient HRQoL measured at baseline improved from the poorest to the wealthiest quintiles for Indonesia only.
Patients in the intensive arm of the RCT had a better HRQoL but also incurred greater costs during treatment. Therefore the size of the intervention effect and the additional provider costs and cost per accurate diagnosis needs to be incorporated into an economic model to fully assess the cost-effectiveness of bi-directional screening and treatment of concurrent disease from a societal perspective.

• The relation between glycaemic control and response to TB treatment.
Although previous studies have found that TB patients who also have DM have an increased risk of TB treatment failure and relapse, there is still a need for more evidence for this. Also, the literature is sparse with respect to the relationship between glucose control and outcome of TB treatment. The TANDEM project, with standard data collection and monitoring in different countries, has the potential to address these questions.
Based on data from 2281 TB patients, including 412 with combined TB and DM, the relative risk of treatment failure associated with DM was 2.62 (95% confidence interval 1.83-3.75). After adjustment for additional factors including age, HIV and smoking this was 2.09 (95% CI 1.39-3.15). TB patients with diabetes also had a roughly 2-fold higher risk of relapse (adjusted RR 1.94; 95% CI 1.03-3.66 p<0.01).
Patients with combined TB and DM also had a higher mortality. Six-month mortality among TB patients with DM was 3.4% compared to 1.2% among those without DM. As such, DM was associated with a nearly 3-fold increase in the crude odds of mortality (OR of 2.80; 95% CI 1.43 - 5.5; p = 0.0017].
These data show that DM significantly increases the risk of TB treatment failure, relapse and death. Ongoing analysis should determine if this due to higher rates of disease severity at baseline, lower exposure to antibiotic drug levels, increased rates of drug resistance or other factors. Also, the role of glycemic control will be determined. Future studies should determine if higher drug concentrations should be used or if TB treatment should be extended in case of co-prevalent diabetes.

Diabetes and Tuberculosis from a pathophysiological point of view

It is largely unknown why DM patients have an increased susceptibility to TB. Some studies have compared cytokine production in TB patients with and without diabetes, or possible perturbations of DM patients in immune response to M. tuberculosis patients in vitro. In addition, there are no published data regarding possible genetic determinants of concomitant TB and DM, or the role of hyperglycemia and glucose control in the immune response to M. tuberculosis.
TANDEM conducted a comprehensive and integrated mechanistic analysis to unravel the cellular and/or genetic basis of the increased TB risk in individuals with DM. This included a combination of gene expression, biomarker, and immune profiling studies in patients with TB, diabetes or combined TB and diabetes comorbidity in direct ex vivo samples, and following in vitro M. tuberculosis infection using macrophages, T-cells and adipocytes under normo- and hyper-glycemic conditions; genotyping of patients with TB, DM and patients with combined TB and DM; and functional genomics to integrate these different components and establish causality and underlying mechanisms. These different approaches will be discussed in more detail below. We also analysed how these parameters were affected by treatment of TB, to see whether there were signatures that might indicate a poor treatment response in diabetic TB patients.

Workpackage 3 addressed whether there are differences in biomarker profiles between patients with uncomplicated TB and patients with TB and DM comorbidity, and how these are affected by drug treatment of TB.

• Cross-sectional gene expression studies
TANDEM investigated how gene expression was different in the blood of patients with TB and diabetes, compared to TB or diabetes alone. We used RNA-Seq, which produces a large amount of data on how the cells are responding at the time the blood is collected. Gene expression was examined in a total of 249 individuals. At the time of TB diagnosis, patients with TB without diabetes show changes in the expression of many genes, including those associated with the “innate” cells including polymorphonuclear leukocytes that often phagocytose the M. tuberculosis bacilli, as well as in a number of markers of inflammation. We found that these changes were altered in patients with both TB and diabetes. There were more genes that were upregulated, with greater expression, as well as more that were down regulated, with less expression, compared to the patients with TB alone. Patients with combined TB and DM had an enhanced inflammatory response but blunted type 1 interferon response compared to the TB-only patient group. This uncoupling of the responses might be indicative of the enhanced risk of TB development in people living with diabetes.
TANDEM has also shown that modulation of ex vivo gene expression in the blood, similar to that found in patients with combined TB and DM, was also present in individuals with TB with HbA1c values of 5.7 - 6.4. This finding has a major public health implication and may be one of the important findings in TANDEM. HbA1c values in this range are variously called intermediate hyperglycaemia, pre-diabetes (though many are uncomfortable with this term), or “high risk” of diabetes. If inflammation and dysregulation is present in such individuals, this may mean they might be silently developing pathology in the pancreas and elsewhere. This group may require active intervention and we recommend further studies should target this group to monitor rates of progression to DM, and to assess whether life style interventions and/or changes to the cut-offs for active DM treatment are required. A recent study in Kenya reported very high rates (37.5%) of pre-diabetes in a cohort of 454 TB patients (Owiti et al, Public Health Action 2017).

• TB Treatment-response biomarker profiles in diabetic TB patients compared to uncomplicated TB patients
TANDEM also investigated how these gene expression profiles were modulated through TB treatment. Looking longitudinally, through TB treatment, the patients with both TB and diabetes showed greater inflammatory changes and lower expression of genes associated with Type 1 interferons than patients with TB alone. Furthermore, there was a significant delay in the TB treatment-induced resolution of Type 1 interferon-related genes in diabetic TB patients, which was temporally separated from the resolution of the inflammatory response.
Strikingly, people with Pre-DM, and only moderately enhanced glycaemia, also exhibited this altered transcriptomic profile though treatment, again indicating that predisposition to TB may occur earlier along the diabetic spectrum than anticipated. A Nanostring analysis of QuantiFERON cell pellets, in which blood was stimulated with antigens from M. tuberculosis, also revealed that some of the differences evident between TB-DM and uncomplicated TB also exist in individuals with Pre-diabetes and TB. The dominant upregulation of CCL2 (MCP1) in antigen-stimulated cells from TB-DM patients may provide a mechanistic insight as CCL2 is known to impair insulin signalling and reduce insulin-stimulated glucose uptake in skeletal myocytes. The SOCS1 and SOCS3 genes have also been implicated previously in insulin resistance, so the enhanced TB antigen-induced expression of these genes in these people living with diabetes may also drive co-morbidity.

• Serum biomarkers distinguishing uncomplicated TB and diabetic TB patients
In addition to gene expression, TANDEM found differences in serum components between diabetic and non-diabetic TB patients. Prior to the start of treatment, diabetic TB patients generally had higher concentrations of several cytokines, which are proteins which transmit signals between cells involved in the immune response. Some cytokines, particularly inflammatory markers, decreased in response to TB treatment in both uncomplicated and diabetic TB patients, but remained elevated in diabetic TB patients at all time points up to the end of TB treatment. A different group of cytokines, including known protective cytokines such as IFNγ and TNFα, were reduced during the intensive phase of TB treatment in both uncomplicated and diabetic TB patients, but significantly increased again at the end of TB treatment only in the diabetic TB patients, indicating enhanced immunological activity, possibly linked to increased risk of poor TB treatment outcome. In contrast, acute phase proteins, which are upregulated during inflammatory events including bacterial infection, were similarly down-regulated in both uncomplicated and diabetic TB patients.
Some metabolites, such as glutamine, cholines, acetylated glycoproteins and sphingomyelins, were altered in diabetic TB patients compared to uncomplicated TB patients and compared to both uncomplicated DM and healthy control individuals: these metabolites might be interesting biomarkers of the concomitant occurrence of TB-DM, but this will need additional research.

• Effect of enhanced glycaemic control on normalization of host biomarker profiles during TB treatment
We also investigated whether patients in whom better glycaemic control was achieved, with lower HbA1c values at the end of treatment, would also show improved normalisation of gene expression profiles. The laboratory HbA1c measurement at 6 months was used to quantify the glycaemic control throughout treatment, and the laboratory HbA1c values from the diagnosis and month 6 time points were used to classify the patients into subgroups. This in-depth analysis of the TANDEM RNA-Seq data set showed that there was a significant change in the expression of some gene modules, especially the Type 1-interferon related gene module, at month 2 in uncomplicated TB patients. There was not a significant change overall in the diabetic TB patients because of the varied responses (heterogeneity) in this group: however, we did find a weak correlation between the degree of change in gene expression and the degree of glycaemic control, indicating that good diabetes control leads to a positive improvement in the host immune response in TB.
Similarly, the concentration of many cytokines correlated with HbA1c values at diagnosis in the analysis of serum, reinforcing the finding that the degree of abnormal immunity is related to the extent of diabetes.
These gene expression studies will be extended to analyse further samples which have been collected from uncomplicated TB and diabetic TB patients through treatment in TANDEM. This enhanced analysis, using a different methodology, will allow validation of our RNA-Seq results, as well as testing whether biomarker profiles are predictive of TB treatment outcome.

• Possible prolonged mycobacterial persistence in diabetic TB patients analysed by measurement of memory T cell phenotype
Combined TB and DM is associated with increased TB-relapse and TB-treatment failure, which could in part be due to lower bioavailability of rifampicin, leading to delayed sputum conversion. There may also be delayed clearance of the persistent mycobacteria, which are below the sensitivity level of sputum culture. Evidence for such very low bacterial numbers might be detected in the phenotype of circulating T cells, as they can indicate continued infection and immune activation.
TANDEM found that there were differences in cell phenotypes between diabetic TB patients and uncomplicated TB patients at the end of treatment, with a delay in the increase in activated naïve CD8+ T cells in diabetic TB patients, suggesting a delay in bacterial clearance. However, we also found significant changes in subpopulations of CD4+ T cells in diabetic TB patients through treatment which were not seen in uncomplicated TB, revealing that there is a more complex dynamic abnormal T cell response occurring in patients with TB and diabetes co-morbidity. Differences were also seen in the expression of phagocytic receptors on monocytes, which would affect the ability of cells to ingest M. tuberculosis.
Taken together, these differences in patients with TB and DM comorbidity compared to those with TB alone yield insights into pathogenesis mechanisms underlying susceptibility to TB in individuals living with DM, and provide indications of avenues of research for host-directed therapy development, targeting the inflammatory and interferon pathways, as well as development of biomarkers of diabetes-associated TB. They also strongly indicate that further studies of patients with HbA1c values below those used to diagnose diabetes (5.7
Work package 4 addressed the cellular basis of the link of DM and TB.

Here further laboratory analyses were used to investigate the link between diabetes and TB.

• Effect of hyperglycaemia on the innate human macrophage response to M. tuberculosis, including intracellular fate of M. tuberculosis
The interplay between M. tuberculosis and human macrophages helps determine the outcome of infection with M. tuberculosis: killing of the bacteria and subsequent antigen presentation to the adaptive immune system; ongoing bacterial replication; or latency. No studies have specifically examined perturbations in macrophage function, and response to M. tuberculosis in the presence of hyperglycemia.
Macrophages are the most abundant cells with phagocytic and antigen presenting functions in humans and play critical roles in innate as well as adaptive immunity by orchestrating T cell differentiation and activation. Macrophages are targeted by M. tuberculosis bacilli. We were therefore interested to know if increased glucose concentrations would make macrophages, cells in which the M. tuberculosis bacteria can hide and survive within the body, less able to kill the M. tuberculosis bacteria. Macrophages were placed in solutions containing normal or increased glucose concentrations and their ability to phagocytose the bacteria or to inhibit their growth them (measured by counting M. tuberculosis colony forming units) measured. However we found that simply increasing the glucose concentration did not affect how macrophages took up or killed the M. tuberculosis. This result shows that the changes that occur in the body during diabetes are complex, and that although measuring glucose concentrations gives a reliable test for diabetes and how well it is controlled, diabetes has other more complex effects on the body.
Diabetes also induces marked changes in concentrations of circulating lipids, with some lipids also being modified (for example by oxidation of low density lipoproteins). We found that when lipids were modified in this way, their ability to control the growth of the M. tuberculosis bacteria was reduced. Other lipid changes include alterations in the types of lipids (ratios of saturated and unsaturated fatty acids); it seems that it is the saturated fatty acids that have the greatest effect on the survival of the M. tuberculosis bacteria within the macrophages. Together these results confirm that the effect of diabetes on lipids may have more direct effects on the body’s ability to control TB than the concentration of glucose itself. We have therefore identified lipid metabolism and the changes that result in this in diabetes, as an important research area for the future.

• Response of adipocytes to M. tuberculosis infection, immune regulation and insulin stimulation
Although TB affects many individuals, a much larger group – 2 billion people – harbour latent TB infection (LTBI) without clinical symptoms, but at lifelong risk of reactivation. The physiological niches where M. tuberculosis persists remain incompletely defined and both pulmonary and extrapulmonary sites have been proposed. Adipose tissue constitutes 15–25 % of total body mass and is an active production site for hormones and inflammatory mediators. The increasing prevalence of obesity, has led to greater incidence of type 2 diabetes, and weight gain is usually (although not always) associated with type 2 diabetes. These patients have a two-three times higher risk of developing TB, pointing to a potential link between adipose tissue and disease pathogenesis.
Besides regulating energy homeostasis, adipose tissue is rich in macrophages and involved in cytokine production, with a high fat diet promoting a switch from anti- to pro-inflammatory phenotype, and prolonged inflammation can lead to insulin-resistance and type 2 DM. Adipose tissue contains adipocytes, monocytes and macrophages, all of which accumulate in obesity. We wondered whether increased body fat might provide a niche where the M. tuberculosis bacilli might hide and escape killing by the cells of the immune system. We studied whether the M. tuberculosis bacteria could be taken up by and killed by fat cells (adipocytes) using cell lines in the laboratory. We found that the M. tuberculosis bacteria can get inside the fat cells, and survive there. This suggested that they might provide a “safe haven” for the M. tuberculosis. We also showed, using experimental models of infection, that M. tuberculosis could be found in fat tissues and fat cells in the body; higher infecting doses of M. tuberculosis led to greater numbers of the bacilli in adipose tissue.
In individuals with LTBI, we found that M. tuberculosis survives in a stressed, non-replicating state with low metabolic activity and that resting macrophages serve as a preferred habitat and become effector cells capable of killing the bacteria after appropriate stimulation. We demonstrated that M. tuberculosis can infect and persist within adipocytes where it upregulates stress-related genes. In vivo, relative proportions of leukocyte subsets infiltrating adipose tissue varied under different conditions of infection. During natural aerosol M. tuberculosis infection, distinct leukocyte subsets, including mononuclear phagocytes, M. tuberculosis-specific CD8+ T cells and NK cells infiltrated adipose tissue and became activated. This shows that adipose tissue undergoes significant alterations during M. tuberculosis infection and is a potential reservoir for this pathogen.
These findings highlight that it is too simplistic to think of TB as a disease that only attacks the lungs. TB disease can affect most parts of the body and this may include adipose tissue. When the M. tuberculosis bacilli are present in adipose tissue in a mouse model of infection this induces cytokines and chemokines that attract other cells of the immune system tothe site of infection. However so far there has been very little research on how M. tuberculosis interacts with adipose tissue. We identify this as another important area for future research.

• Testing whether the association of TB and DM is at least partly caused by genetic variation common or unique to both diseases

Genetic association studies have proven the involvement of a strong genetic component both in DM and TB. These studies have provided novel insights into TB and DM disease mechanisms. However, so far, no study has examined a possible genetic basis for the increased risk of TB among DM patients. Thus, we are missing the critical genetic information to reveal necessary pathways deregulated among DM patients that make them highly susceptible to M. tuberculosis infection. We hypothesized that common gene polymorphisms associated with type 2 DM and/or unknown rare genetic variants – with a distinct effect on gene function - are associated with increased susceptibility to TB. By combining the associated variants with the gene-expression data, we aimed to determine the specific effects of the variants (cis and trans-eQTLs).

In TANDEM we looked for genetic variations in people with TB, TB with diabetes or diabetes alone. Samples from over 1500 TB patients and over 500 healthy controls were obtained from the Indonesian study site in Bandung, as well as from 341 TB-DM and over 700 patients with diabetes alone. These samples were transferred to Groningen in The Netherlands where the samples were extracted and processed using RNA-Seq. The sequencing data obtained were used to identify variable genes with variable expression (quantitative trait loci or cis-eQTL). Further analysis looked for associations between particular genetic variants and disease, using genome wide case-control association analysis. The variable genes identified were in human leucocyte antigens (HLA), that play critical roles in how the cells of the immune system interact. Analysis showed that patients with both TB and DM were more likely to have a HLA risk allele that was associated with susceptibility to TB. Samples from patients in Romania were used to confirm this result.

By comparing the genomes of TB patients to those of healthy people we confirmed the significant role of human HLA alleles in determining susceptibility to TB. In addition, we also found evidence for enrichment of TB-risk HLA alleles in diabetic TB patients in both Indonesian and Romanian populations. These results suggested that individuals carrying the TB-risk HLA alleles are more prone to develop diabetes-associated TB. Further studies are needed to investigate this possible genetic interaction between tuberculosis and type 2 DM mediated by the HLA locus.
Primary cells (monocytes, adipocytes, T-cells) isolated from individuals with relevant gene polymorphisms can also be stimulated in vitro with purified mycobacterial components and live M. tuberculosis, so that the function of these cells (in terms of cytokine production, antigen presentation, killing capacity etc) can be assessed. Monocyte-derived macrophages prepared from the blood of normal healthy individuals with and without the particular SNP of interest were tested in vitro for cytokine production after stimulation with M. tuberculosis. The SNPs associated with TB-DM susceptibility were also associated with enhanced production of proinflammatory cytokines.
Although identifying such variants in genes does not directly lead to better treatments or interventions, an improved understanding of the particular genetic variants and how they result in susceptibility to TB will give us a greater understanding of TB, and may result in the design of better drugs and diagnostic tests in due course.

Potential Impact:
Besides its scientific impact, TANDEM had a clear public health focus. It helped (1) produce the necessary evidence supporting the policy on TB-DM which the collaborative framework of WHO / IUATLD identified in 2011 as missing; (2) define economic impact of new policy recommendations to control and treat TB-DM; (3) health providers in various settings with practical implementation of these recommendations. Finally, TANDEM strengthened existing, and created new, networks and strengthened capacity of individual partners

Awareness of the interactions between TB and diabetes
TANDEM increased awareness of the intertwined epidemics of TB and DM wherever possible, including through its website, published reviews, and through symposia and high-profile scientific and policy meetings. We highlight a few activities here:
• TANDEM members joined a policy meeting on TB-DM hosted by the Indonesian Ministry of Health.
• TANDEM organised larger symposia linked to the annual TANDEM meetings in Indonesia, Romania and Peru. Attendance at these open symposia exceeded 200 attendants in each case. The most recent symposium held in Lima in May 2017 is also available in Spanish on YouTube (see https://www.youtube.com/watch?v=cHkWvwlK2O0xxxx). These events provided an opportunity for TANDEM participants and other external speakers to highlight the importance of raising awareness of the TB-DM co-morbidity, and to discuss the situation of TB and diabetes locally, with physicians and nurses who work with TB patients, as well as researchers and undergraduate and postgraduate students interested in this topic
• Awareness of the interaction between TB and DM among diabetologists/endocrinologists is low, but TANDEM was involved in a series of high-profile papers in Lancet Diabetes, the most important scientific journal about DM.
• Two TANDEM partners have been invited to be part of a working group (consisting of 7 international members) of the International Union Against Tuberculosis (IUATLD), updating TB-DM programmatic guidelines for treatment and management of this comorbidity. As members of the committee they will be writing a 1st edition of a global framework “Management of Diabetes-Tuberculosis; a Guide to Essential Practice (1st edition, 2018)” an initiative of the IUATLD and the World Diabetes Foundation. Hence initial TANDEM findings are informing the development of national and international guidance, widely used by TB programme managers globally.
• In 2017 the team in Lima presented some preliminary results of the TANDEM project at the National TB meeting and at the RePORT International Meeting 2017 (http://www.redetb.org/attachments/article/293/RePORT_International_Meeting_Agenda_Rio_Sept%202017.pdf) which has helped draw the issues related to TB and diabetes comorbidity to the notice of a wider audience in Peru.
• TANDEM was featured in the World Diabetes Foundation/IUATLD call for action on TB and DM that was published in November 2014 and that addresses the threat of DM-TB comorbidity at global level. The report calls for policies that address both diseases and claims that a clear strategy is needed for screening both DM patients for TB and TB patients for DM, as well as for providing treatments for both diseases. TANDEM is featured in the context of the TB-DM Research Agenda that, according to the report, is needed to “address this threat before it takes a larger toll in death and disability as well as economic impact – and before we see the gains made against TB in the past decade rolled back by diabetes”.
• TANDEM was presented at the world’s first International Summit on Tuberculosis and Diabetes, which was held in Bali, Indonesia, from 2-3 November 2015.
• TANDEM members gave presentations at both the 46th and 47th Union World Conference on Lung Health, and will also contribute to the programme in 2017 at the 48th Union World Conference on Lung Health.
• In March 2016 during the Keystone Symposium on Tuberculosis Comorbidity in Denver, TANDEM was presented as part of a an invited keynote talk to this major meeting on TB comorbidity.
• The first presentation of the gene expression work in TANDEM was given to a scientific audience at the Acid Fast Club summer meeting in Leicester 2017. The Acid Fast Club represents the UK’s main mycobacterial research society, and their meetings with limited spaces are attended by over 100 participants twice a year.
• One of the TANDEM member from Peru’s team was invited in 2017 to be part of the National TB Expert Board at the National TB Program and to lead the discussion for a guideline in TB/DM care in Peru. Based on the results and the interaction of TANDEM with the National TB Program, gaps in health systems, training and education in TB/DM were identified, leading to submit a proposal to the World Diabetes foundation in collaboration with LSHTM. This proposal was funded in 2016 and has developed several activities in TB-DM training, using the experience learned in TANDEM, including on-line courses, meetings in health centres and developing educational materials (posters, booklet for TB-DM patients and a guide for health workers who worked with TB-DM patients). Now, also under this grant and based on the lessons learned in TANDEM, a mixed-methods study has been launched to evaluate the cascade-of-care of TB-DM in Lima, evaluating screening, diagnostic and care gaps and bottlenecks in the health system.

Disease prevention, management and prognosis
The results from WP1 have direct impact on disease prevention.
• TANDEM identified the best and most cost-effective way to screen TB patients for DM. This will enable policy makers to plan for, and incorporate, a definitive DM screening plan in National TB Control Programmes. Early diagnosis of DM in TB patients enables those with DM to be managed appropriately to reduce their symptoms and lengthen the time to development of DM complications, and control of hyperglycaemia may improve the outcome of TB treatment in those with DM.
• TANDEM provided precise estimates of the burden of DM in TB patients across the sites, stratified by age and gender, which are important for policymakers. These figures will enable policymakers to incorporate a properly justified budget for the diagnosis and management of DM in TB patients into annual planning processes.
• TANDEM concluded that the yield from the active case finding for TB in DM patients across our study sites was relatively low, and that screening is unlikely to be cost-effective.
• TANDEM’s findings on latent TB infection (LTBI) suggest that individuals with DM in high TB incidence settings should be prioritised as a high risk group for LTBI screening and prophylactic management
TANDEM provided key data related to therapeutic management of TB and DM:
• Characterization of individuals with DM, both those with and without TB, in four countries highlighted the enormous challenges in DM management in resource-limited settings. A large proportion of DM in TB patients was previously known yet not properly treated or poorly managed.
• It was found that intensive glucose monitoring can significantly improve poor glycaemic control, yet at 2.5 fold higher costs. Long-term cost-utility analysis will help define the public health and societal impact of this intervention
• TANDEM established that metformin, the 1st choice oral DM drug is safe, and not affected by concomitant use of rifampicin, unlike other oral DM drugs, providing guidance for guidelines for DM treatment during TB treatment.
• TANDEM was the first to identify high and unmet requirements for DM management after completion of TB treatment, in terms of life-style counselling, anti-diabetic, anti-hypertensive, lipid-lowering and other medication.
TANDEM also provided key data related to prognosis of TB-DM patients in several ways.
• TANDEM roughly doubled the evidence base for the current consensus that diabetic TB patients experience more TB treatment failure and are more likely to die and to have TB relapse.
• TANDEM also confirmed the hypothesis that poor glucose control in DM patients with TB is associated with worse health outcomes, underlining the need to implement better glucose control in diabetic TB patients
• TANDEM has provided the first evidence to help define long-term requirements for DM management. The short term prognosis of TB-DM is largely dictated by TB, but DM is more important for the long term prognosis of TB-DM patients. TANDEM showed that DM control under routine conditions worsened after completion of TB treatment.

Policy
All these data will help set the next research agenda of questions related to screening and treatment in relation to TB and DM (e.g. prolonged TB treatment, secondary prophylaxis, health systems interventions to secure long-term DM management).
The lack of firm evidence on which to base policy decisions was highlighted in the Collaborative WHO / IUATLD Framework for Care and Control of Diabetes and Tuberculosis drafted in 2011. TANDEM has provided data from settings where the prevalence of both TB and DM vary, and that are typical of settings where economic improvement and changes in lifestyles are likely to increase the risk of DM substantially. For example, the Framework was unable to make recommendations on the diagnosis or treatment of latent TB infection, because of the paucity of evidence. TANDEM has provided key data to address these issues and develop more specific guidelines. As outlined above, TANDEM is providing key data for many of the questions included in the 2011 Framework.
Importantly, TANDEM partners will help draft the 1st global guide to management of TB and DM, an initiative of the Global TB Union (IUATLD) and the World Diabetes Foundation (WDF).

Economic impact
TANDEM conducted economic evaluations as part of the comparison of three screening methods in WP1 and the two DM management strategies in WP2. Through further economic modelling, TANDEM will be able to estimate incremental costs per treatment failure averted and per life years gained. In addition, the economic impact on individual patients, on clinics and public health practice, and on the populations overall can be estimated alongside the guidelines that will be proposed. Finally, in the long term, new insights into the underlying mechanisms of the association between the two diseases may lead to new interventions that reduce the morbidity and mortality associated with both diseases, and in turn the cost to society that they currently incur.
While collecting economic data as part of TANDEM, there was interest and engagement from financial directors of health facilities from the participating institutes who were keen to see the final analyses, which they felt would not only be useful for decision making at the central level but also for their own facilities. The economic data collection was facilitated by the country PI as well as TANDEM team members (clinicians and laboratory personnel) who now have an appreciation of the process for conducting a costing study. Though economic studies are not uncommon in Indonesia, the level of detail that was performed in Bandung had not been done before for either TB or DM and this was seen as important for more accurate priority setting. Indonesian health economists participated and developed skills in some aspects of the treatment costs data collection and were crucial in facilitating access to some data.

European networks and capacity strengthening for individual partners
Globally, both TB and DM are major public health issues. Besides research infrastructure TANDEM addressed strengthening and integration of health systems across three sites outside Europe, for instance by connecting local DM and TB physicians. We have witnessed a significant impact of patient-oriented research in terms of staff development, implementation of existing guidelines, and quality and patient care in general.
As expected, the TANDEM project had significant impact for individual partners at several levels of healthcare. For all partners TANDEM helped strengthen academic development, eg by training MSc and PhD students and by having many staff members involved in scientific writing of papers etc.
Below we summarise some specific aspects of ‘impact’ for participating institutes or the wider community.

• University of Padjadjaran, Bandung, Indonesia:
TANDEM has had a major impact on the capacity building of its staff. Ten nurses and 8 medical doctors were trained. Three research managers had learned to carry out a multi-center study and to write scientific reports and other training activities for members of the laboratory, database managers and supporting staff members. The project has increased the capacity for large studies like TANDEM. Scientific papers on the TANDEM data are being written by UNPAD staff.
TANDEM has led to collaboration with more than 30 Community Health Centers (Puskesmas), which serve 50% of the total 2.5 million population in Bandung. TANDEM has raised their awareness of TB-DM co-morbidity and increased the capacity for TB diagnosis and DM screening. We have presented our preliminary result to the head of the Municipal Health Office and were asked to present the result in the meeting with all of the Puskesmas.
At a national level, data were presented at the Ministry of Health Diabetes Care meeting in 2015, and a staff member was included in the Expert committee of the National TB Program. The National TB program and the Diabetes Control Program acknowledge our work and are waiting for the formal results of the TANDEM project, which we are invited to present at National TB and DM seminars.

• University of Medicine and Pharmacy, Craiova, Romania:
At a public health level, The TANDEM research team interacted with hospital personnel from two TB and two DM clinics, increasing awareness of the comorbidity and identifying potential obstacles to the management of the TB-DM patient in Romania. TANDEM’s findings can help to sort out the logistic issues of screening and management of the TB-DM patient.
TANDEM helped increase laboratory capacity through 2 workshops in 2014, and exchange visits with the TANDEM RUMC. In Jan 2015, 3 members of the UMFCV lab personnel visited RUMC to learn about new laboratory techniques (GeneXpert, liquid media culturing) as well as exchanging good laboratory practices for personnel safety.
In terms of academic capacity, members from our team have worked in centers of other TANDEM partners learning molecular techniques and advanced data analysis. TANDEM also gave the opportunity for higher visibility in the scientific community for our centre.

• Universidad Peruana Cayetano Heredia, Lima, Peru:
TANDEM provided many advantages and opportunities in Peru, not only through the research results, but also by providing evidence to develop other initiatives:
TANDEM participated in the 2014, 2015, 2016 and 2017 National TB Meetings, leading round tables on TB-DM and increasing the discussion about the importance of DM in TB control in Peru. Dr Ugarte, the TANDEM co-investigator in Peru, was invited in 2017 to be part of the National TB Expert Board at the National TB Program and to lead the discussion for developing a guideline for TB-DM care in Peru.
The TANDEM team also gave lectures on TB-DM at TANDEM clinical sites and other health centers as requested.
Based on the results and the interaction of TANDEM with the National TB Program, we identified gaps in health systems, in training & education in TB-DM, leading to the submission of a proposal to the World Diabetes Foundation in collaboration with LSHTM (https://www.worlddiabetesfoundation.org/projects/peru-wdf15-1224). This proposal was funded in 2016 and has developed several activities in TB-DM training, using the experience learned in TANDEM, including on-line courses, meetings in health centres and developing educational materials (posters, a booklet for TB-DM patients and a guide for health workers who work with TB-DM patients). Now, also under this grant and based on the lessons learned in TANDEM, we are running a mixed-methods study to evaluate the cascade-of-care of TB-DM in Lima, evaluating screening, diagnostic and care gaps and bottlenecks in the health system.
Finally, Dr Ugarte and his team will disseminate the TANDEM final report through a presentation to the National TB Program board, National Diabetes Program which will be uploaded to the UPCH website and shared in social media (Facebook, Twitter). The synthesis of the results will be translated in Spanish.

• Stellenbosch University, Cape Town, South Africa:
The TANDEM project had a beneficial effect on employment, and partially or fully supported the salary of multiple staff members at various levels. This included drivers, clinical research workers, registered nurses and doctors. While some of these staff members have moved on to embrace other opportunities, most of them were retained and work on new projects of our growing clinical research team.
Four post-graduate students have already graduated successfully, because of the opportunity they had to work on projects related to the TANDEM study. This includes one student taking a BSc.-honours degree and three taking MSc. degrees. An additional two postgraduate students – one MSc. and one PhD. – will complete their TANDEM-related projects next year.
The TB-DM bio-profiling project contributed to strengthen the group’s expertise in the field of endocrine – immune interactions, especially in relation to diabetes and tuberculosis. One of the senior scientists that worked on the project has since become one of the lead applicants on a successful NIH-funded grant to research the endocrine and immune profiles of DM patients and the association with an increased risk of TB infection and disease.
Data and stored samples collected during the TANDEM project are already being used as an invaluable source for investigating the causes, mechanisms and biomarkers associated with unfavourable outcomes during and after TB treatment, and can also be used for future projects.
Partnerships with local healthcare were facilitated by TANDEM, as the recruitment of diabetes patients necessitated forging new partnerships with healthcare facilities. We again called upon these partnerships during further recent studies related to diabetes and HIV respectively. The group has also recently started conducting clinical trials and the team have established a community advisory board to assist with communication to the local community.

• University of Otago, Otago, New Zealand:
The University of Otago has been part of the clinical and epidemiological aspects of TANDEM, and for the Indonesian site in particular. More recently, the University of Otago has played a key role in preliminary analyses and final analyses of datasets from TANDEM, and in writing up results for publication. In addition, the University of Otago has been able to lead the development of interlocking projects in Bandung, Indonesia, and human capacity building. Interlocking projects with separate funding that include:
- The establishment of a cohort of household contacts of TANDEM TB cases to study early clearance of M. tuberculosis, placing a PhD student on the ground from Otago University to do this project.
- The introduction of the QuantiFeron test to evaluate the prevalence of latent TB infection in TANDEM DM patients in Indonesia.
- A follow-up study of TANDEM DM patients to estimate the incidence rate of TB.
- A study on knowledge in TB-DM patients who are in the RCT.
Human capacity building with separate funding has included an MSc in epidemiology for a candidate from Bandung, funded by the University of Otago, by distance learning at the London School of Hygiene and Tropical Medicine, and a PhD in epidemiology, hosted and funded by Otago University. TANDEM has been one of the key projects that the new international health initiative at Otago University has been involved in. International health in New Zealand has only recently been formally established as a discipline and involvement in TANDEM has been a part of its foundational period.

• London School of Hygiene & Tropical Medicine, London, UK:
The TANDEM project allowed integration of work across three departments and two faculties, on clinical epidemiology, health economics and more basic gene expression profiling.

TANDEM has supported two PhD students, who have submitted theses on “The effects of comorbidities on blood transcriptomes in tuberculosis patients before and during treatment” (Clare Eckold) and “Strategies to detect and treat concurrent Tuberculosis and diabetes mellitus in Indonesia, Peru and Romania: Costs, operational feasibility and impact on health-related quality of life” (Yoko Laurence).

A key benefit was the development of new collaborations with all the field sites, allowing continuation and broadening of our research. In partnership with UNPAD, we have been awarded a Newton Fund award (~£550K for 2 years), from UK MRC and Indonesian DIPI, for the project “The Role of Complement Component C1q in Tuberculosis and Diabetes Co-morbidity”. This allows the continued collaboration between Jackie Cliff and Hazel Dockrell with Rovina Ruslami and Bachti Alisjahbana. The project will extend the immunological characterisation of the interaction between TB and diabetes, by conducting in-depth molecular analysis of the role of C1q in host-mediated protection at LSHTM, combined with phenotypic assessment of C1q and its regulatory factors in patients with TB and/or diabetes in Indonesia. The project involves a strong capacity development element, with a new Immunology PI in UNPAD, Dr Afiat Berbudi, and also an Indonesian research assistant will spend time in the UK to help to transfer assays and knowledge.

• St George’s, University of London, London, UK:
TANDEM has enabled the development of new international partnerships between SGUL and other partners in the project, many of whom had not worked together previously, but are now drafting proposals for successor projects together. TANDEM has also cemented links between SGUL and LSHTM, particularly with population health at SGUL and health economists working at LSHTM. SGUL and LSHTM are currently developing joint proposals (for PhD studentship funding) to develop modelling approaches to estimate the longer term population health impacts of better management of DM during TB treatment (Julia Critchley and Yoko Laurence).
In addition, TANDEM funding enabled St George’s, University of London to lead the development of interlocking projects to develop mathematical models of the interactions between TB and DM, with partners outside of TANDEM (separate funding of US$850k). Results from these projects suggest that the impact of DM on TB is much greater than previously thought, since the indirect effects of diabetes (increasing onward transmission of TB because people with TB and DM have more infectious smear positive forms of TB) are very substantial (manuscript under review). This project is also evaluating the potential population health benefits and costs of interventions to screen and manage DM better among TB patients, and will use data from the TANDEM project (and elsewhere) to parameterise key models, thus maximising use of the data collected during TANDEM
The TANDEM project has also been particularly helpful in promoting links across different sections of SGUL which did not have a strong history of collaboration previously, particularly between the Population Health Research Institute, Infection and Immunity Institute, and the St George’s Hospital NHS trust. TANDEM results will be presented at a key Inter-TB consortium meeting in the near future, further enhancing understanding about the importance of the association with key trialists in the TB world. Julia Critchley and Prof Tom Harrison (Infection & Immunity Institute at St George’s) are also investigators on a recent outline proposal to investigate preventative treatment of latent TB in DM patients led by the University of Otago submitted to the DFID/NIHR/MRC/Wellcome Trust Global Health Trials fund, together with RUNMC Nijmegen, UNPAD, Bandung and other partners.

Joint pump priming proposals have been written with the St George’s Hospital Trust to investigate associations between HbA1c and drug resistant infections, for which Trust has a large regional database, as we identified there is a lack of high quality data about the role of DM in promoting drug resistance with infections generally, not just for TB.

TANDEM has also been particularly helpful in enhancing the experience and employability of the key data manager employed on the project who completed an MSc in Data Science for Research in Health and Biomedicine at, UCL, London, using TANDEM data examining patterns of missingness for her dissertation. She was successfully awarded this degree (with distinction) in 2016 from University College London. TANDEM experience has substantially cemented her experience as a data manager, providing her with a new and unique opportunity to work on an international multi-centre country project, providing a focal point in each country and providing her with experience in managing the data requirement needs for a large and very diverse research project. This has cemented her position at SGUL, and she has since worked as a trial manager on a cluster randomized controlled trial.

Benefits of EU-funded research consortia
Working together in multi-partner, multi-country consortia also beings a number of benefits to the clinicians and health care workers in the countries where patients are recruited, as well as to the scientists who work on the more basic and mechanistic studies in the laboratory.
• without these partnerships, access to research funding is often limited in LMIC.
• participation in such consortia has a beneficial effect on clinical practice, even though further efforts are needed to ensure that changes in the management of patients are embedded into local practice in a sustainable way. For example, we found that despite the known association between TB and diabetes, there are presently few links between the clinicians and clinics treating these diseases in our four clinical sites.
• A large number of staff members and researchers received training through the TANDEM consortium, which will contribute to their Masters and PhD theses.
• It is very valuable for basic scientists to visit the field sites and clinics and appreciate the challenges involved with patient recruitment and follow-up, as well as the complexities associated with the diagnosis of both TB and diabetes.

Implications of participation in TANDEM for Consortium members
“TANDEM has confirmed for me the value of multi-partner consortia that combine both basic and clinical studies. It is very valuable for basic laboratory scientists to learn at first hand the complexities and challenges of patient recruitment and diagnosis, and how valuable such clinical samples are. However such research requires significant investment and a sufficient time scale.” (LSHTM)
“Refining the coordination of multi-country data collection as well as understanding the peculiarities of both TB and diabetes have been invaluable in the work that I now do as a Research Fellow in Health Economics. With these skills, I can confidently assist NTPs with the economic data collection for another multi-country study that started in 2017 as well as engage in developing several grant applications that seek to further understand the relationship between TB and diabetes, in different settings.” (SGUL)
“TANDEM has strengthened our collaborations with all partners involved in the project, some partners were already long-term collaborators however, TANDEM also expanded our network. TANDEM strongly boosted the basic science work on the interaction between the immune system and host metabolism in our department. This resulted in an important new research area in our department, which we aim to continue beyond TANDEM.” (LUMC)
“The Otago group worked closely with the partners in Bandung. The TANDEM project was been successfully conducted in Bandung and has enabled collaborative interlocking projects to be developed with other funding streams, and human capacity building. The Otago University-Padjadjaran collaboration has been enhanced, as have other collaborations that we have across the consortium. We will continue to work with TANDEM partners. Indeed, we are taking a leadership role on a submission to the DFID/NIHR/MRC/Wellcome Trust global health trials fund together with St George’s, Radboud and Padjadjaran and other partners. A University of Otago PhD student was hosted by Radboud University for several months during TANDEM and UO hosted several Indonesian collaborators to their annual global health institute conference. UO hope that Prof van Crevel will be funded to have a study visit at Otago in 2018.” (Otago)
“The TANDEM project has been beneficial for UMCG site both at the level of conducting fundamental research as well as in training young researchers. More specifically, the data generated for TANDEM project helped to train a Romanian PhD student in genetic analysis as well as a Master’s student to learn more specialized genomics research. The financial support provided by the TANDEM for attending international meetings helped the team to not only disseminate TANDEM findings but also created an opportunity to connect with a network of international scientists. The availability of genome-wide genetic data in two different populations of TB patients for the first time also allowed UMCG to collaborate with the main researchers in tuberculosis genetics research, which will hopefully lead to bigger studies and potential discoveries that may ultimately help TB patients”. (UMCG)
List of Websites:
www.tandem-fp7.eu