As previously highlighted, diabetes both increases susceptibility to tuberculosis and leads to worse treatment outcomes, with more deaths, tuberculosis treatment failures and recurrent disease. This raises the question if and how clinical management of combined diabetes and tuberculosis can be optimised. There is relatively little evidence to guide clinicians in clinical management of combined disease. We separately address tuberculosis and treatment and diabetes management in people with combined disease. Some key issues related to treatment of combined TB and diabetes are included in
Table 2, and recommendations have been included in the recent guideline on TB and diabetes [
10].
4.1. Tuberculosis Treatment in Patients with Comorbid Diabetes
Currently recommended TB treatment is similar for patients with combined TB and diabetes compared to those with TB only. However, this may have to be reconsidered as diabetes is associated with TB drug resistance, [
51] slower treatment response and higher rates of toxicity, failure and recurrent TB. First, TB treatment might have to be adjusted in length, as seems common practice already in some countries, including China [
52]. Indeed, in a large retrospective cohort study from Taiwan, nine-month treatment was associated with a lower rate of recurrent TB than six-month treatment (hazard ratio 0.76; 95% CI 0.59–0.97) [
53].
Besides length of treatment, higher dose TB treatment may also help improve treatment outcomes. Some studies have reported associations between diabetes and lower concentrations of TB drugs, [
54,
55] although this may partly be explained by inappropriate correction for body weight in TB patients with diabetes who are generally significantly heavier than those without [
56]. However, in an observational study in the USA, therapeutic drug monitoring for INH and rifampicin after two weeks treatment was associated with significantly shorter time to sputum culture conversion among patients with combined TB and diabetes (42 versus 62 days;
p = 0.01) [
57].
Another factor to consider is the apparent association between diabetes and TB drug resistance. A recent meta-analysis included 9289 patients from 13 studies found a significant association between diabetes and multi-drug resistant (MDR)-TB (OR 1.71, 95% CI 1.32–2.22) [
27,
51]. Possible factors contributing to this association include higher rates of acquisition of drug-resistance with higher initial bacterial load or slower response to treatment or nosocomial acquisition of drug-resistance with higher rates of hospital admission. However, diabetes is also associated with
primary drug resistance. TANDEM performed the first study to compare genotypic drug resistance between TB patients with (n = 159) and without diabetes (n = 737) [
58]. This study used whole genome sequencing on an unselected cohort of patients from Peru and all TB–diabetes plus age-matched non-diabetic TB patients from an Indonesian cohort. Drug resistance mutations were found in isolates of 18% of diabetic and 15% of non-diabetic patients in Indonesia and 39% of diabetic and 20% of non-diabetic patients in Peru, with
rpoB (rifampicin)
, fabG1 (INH) and
gyrA (fluoroquinolone) mutations associated with diabetes. In multilevel multivariable logistic regression, diabetes was the only factor significantly associated with genotypic drug resistance against at least one drug (OR 1.8, 95% CI 1.1–2.9). Interestingly, the association between diabetes and drug resistance was similar for patients with new (adjusted OR 2.0, 95% CI 0.9–4.4) and previously diagnosed diabetes (OR 1.7, 95% CI 0.98–2.9). This suggests that individuals with diabetes may be more susceptible to drug resistant
M. tuberculosis isolates; some drug resistance mutations lead to “loss of fitness” of
M. tuberculosis, but such “less fit” strains might still cause active tuberculosis among individuals with diabetes because of their decreased host immune function against tuberculosis [
59]. Whatever the underlying cause, TB–diabetes patients should probably be prioritised for drug susceptibility testing (DST) in those settings where DST is not routinely done for all patients, both at baseline and during follow-up. The Xpert MTB/RIF assay (Cepheid Inc, Sunnyvale, CA, USA) is being scaled up globally and within 2 h allows the confirmation of
Mycobacterium tuberculosis and the detection of rifampicin resistance—which is equivalent to MDR-TB. In 2013, WHO recommended that the assay be considered as the initial diagnostic test for all people requiring investigation for TB. While this is yet to happen for all patients, those with TB–diabetes should be prioritised.
There are several other things to consider when treating patients with combined diabetes and TB. For instance, there is a higher risk of drug–drug interactions. Rifampicin increases the metabolism of many drugs commonly used by diabetes patients including statins, all oral diabetes drugs except metformin, calcium-channel blockers, ACE-inhibitors, digoxin and warfarin. some antihypertensive drugs (van Crevel, Tuberculosis, in Cohen and Powderly, 2017). In addition, TB–diabetes patients may have more symptoms or symptoms that are more difficult to interpret. For instance, abdominal symptoms in an elderly TB patient with diabetes may be a side effect of TB drugs, hepatotoxicity, side effects of metformin, inferior myocardial infarction or bowel ischemia. In addition, drug toxicity may be worse; INH can aggravate diabetic neuropathy, and TB–diabetes patients have a higher risk of liver and kidney toxicity. This is also relevant for management of MDR-TB; patients with diabetes probably have a higher risk of renal toxicity with aminoglycosides and a higher risk of neuropathy with linezolid. Furthermore, a high pill-load when patients are treated for both diseases may lead to missed doses, incorrect drug intake, treatment interruptions or default. These and other issues necessitate more careful assessment prior to and during combined TB and diabetes treatment. It should be stated that there is a scarcity of data regarding all these aspects and that more study is needed.
A final consideration may be HIV co-infection; Sub-Saharan countries with a high TB–HIV burden are witnessing the most rapid growth in diabetes prevalence, while some other countries where TB–diabetes is common (e.g., India or Indonesia) show significant growth of the HIV epidemic. HIV is associated with diabetes; in a study in Tanzania, glucose metabolism disorders were six-fold more common among HIV-infected individuals compared to age- and weight-based controls [
60]. In addition, in Ethiopia and South Africa, longer duration of HIV treatment was associated with increased incidence of diabetes [
61,
62]. Finally, in part due to the successful roll-out of antiretroviral treatment, increasing numbers of HIV-infected patients are surviving to older ages and become at increased risk of developing diabetes [
63]. Therefore, a growing number of people may be affected by diabetes, HIV and active TB at the same time. In such cases, there is an even higher risk of drug–drug interactions, toxicity and overlapping side-effects, for which clinicians will need guidance.
4.2. Optimising Diabetes Management in Patients with Combined Tuberculosis
Management of diabetes is aimed at reducing short- and long-term complications such as cardiovascular disease, eye and kidney problems and foot amputations. Diabetes management consists of: lifestyle counselling (diet, weight loss, physical activity, smoking cessation and avoiding excess alcohol); treatment with blood glucose lowering drugs; measures to reduce the risk of cardiovascular disease and associated complications that include anti-hypertensive medications, lipid-lowering drugs and anti-platelet drugs if indicated; and management of specific complications such as diabetic feet and eye problems. The first priority for patients with combined diabetes and TB is to successfully initiate TB treatment, but diabetes management certainly deserves attention. For instance, severe hyperglycaemia, which can be symptomatic and is likely to affect TB outcomes, should be treated. TB itself may also increase risks of cardiovascular disease, as previously noted [
35].
As described above, TB–diabetes patients form a heterogeneous group, consisting of those with “known” (previously diagnosed) and newly diagnosed diabetes, with hyperglycaemia ranging from mild to severe, variable duration of disease and highly variable comorbidity, disease complications and treatment needs. Around 74% of TB–diabetes patients in the TANDEM cohort in Indonesia, Peru, Romania and South Africa had previously diagnosed diabetes, while 26% were newly detected as a result of diabetes screening. Few patients with “known” diabetes were still under diabetes care when their TB was diagnosed, some had never even been started on diabetes treatment [
64] and those supposedly taking medication showed very poor glycaemic control and often high cardiovascular risk. TB–diabetes patients also vary in terms of the number of diabetes complications and cardiovascular or other co-morbidities. Heterogeneity of diabetes in TB patients is especially notable between different countries, most likely due to genetic, environmental, nutritional and behavioural factors and differences in accessibility and quality of health services. Clearly, this heterogeneity has important implications for diabetes management.
4.3. Glycaemic Control
An accepted target for glucose control in diabetes is an HbA1c < 7% (53 mmol/mol), as recommended by the American Diabetes Association [
65], although the American College of Physicians recently recommended HbA1c levels of 7–8% [
66]. However, this may be hard to achieve during anti-TB treatment, due to drug interactions with rifampicin and altered patterns of food intake and energy expenditure during TB disease and treatment recovery [
25]. Advanced or long-standing diabetes may also render better control more complex. Some randomised controlled trials of “tight control” (target < 6% HbA1c) among older people with long-standing type 2 diabetes in high-income countries had unexpected, unfavourable outcomes [
67]. Therefore, although this is not entirely evidence-based, a more realistic and cautious treatment target—especially under programmatic conditions—may therefore be HbA1c < 8% and a target of RBG/FBG < 11.1 mmol/L (200 mg/dL) during the treatment of TB disease, which is in line with those for diabetes management in persons with significant co-morbidity [
65]. It is also important to realise that TB-associated inflammation can lead to temporary hyperglycaemia, which often spontaneously improves with anti-tuberculosis treatment [
68]. For instance, a Tanzanian study found that associations between hyperglycaemia and TB present at baseline were substantially attenuated after TB treatment; however baseline hyperglycaemia was associated with substantially increased risk of death or TB treatment failure) [
68]. Not all studies have shown such significant reversions of hyperglycaemia; this occurred in only 3.7% of patients in Indonesia [
69] and only 14% of TB patients with newly identified hyperglycaemia in the TANDEM study, which unlike many other studies used repeated tests to confirm hyperglycaemia increasing the certainty of the initial diagnosis. Hyperglycaemia in patients with “known diabetes” proved much more difficult to reduce in a small trial of improved diabetes management during TB treatment than among those with “new diabetes” (Koesoemadinata et al., under review).
The next question is how optimal glycaemic control can be achieved. Thus far, there is no guiding evidence as to what drugs should be used, how treatment should be monitored, who should best deliver diabetes treatment and how this should be adjusted for patients with new or previously known diabetes and for those with mild or severe disease. We first address the choice of glucose-lowering drugs, discussing three classes of drugs: biguanides, sulphonyl urea derivates and insulin. Although there are other drug classes to treat diabetes, including thiazolidinediones (TZD), DPP-4 inhibitors, SGLT2 inhibitors and GLP-1 receptor agonists, these medicines are more expensive and there is limited evidence of superior effectiveness.
Metformin is the first choice glucose-lowering agent recommended in type 2 diabetes, and there is no reason this should be different for patients with active TB disease. The advantages of metformin include extensive experience with the drug, extremely low risk of hypoglycaemia, effectiveness, low cost, beneficial effects on cardiovascular disease, [
67] lack of clinically relevant interaction with rifampicin [
70] and a potential benefit on TB itself [
71]. In a recent retrospective analysis from Taiwan, those with diabetes (30%) had a 1.9 times higher mortality, but among this group, metformin use was associated with lower mortality (HR 0.56, 95% CI 0.39–0.82) [
72]. However, all data on metformin’s effect on TB is from observational datasets and might be affected by selection biases, as metformin is the first line treatment for diabetes in many parts of the world. Diabetes patients taking other drug treatments may thus be substantially different in ways that affect TB prognosis and are difficult to adjust for. Metformin’s two main disadvantages are gastro-intestinal side effects, which may be worse when taken together with TB drugs (TANDEM, unpublished) and increased toxicity, including the development of lactic acidosis in patients with decreased kidney function. Lactic acidosis, usually presenting as vomiting, abdominal pain, signs of hypovolemia followed by Kussmaul breathing, neurological signs, cardio-respiratory, kidney or liver failure and finally death is extremely rare but it may be fatal if unrecognised and untreated [
73]. The dose of metformin needs adjustment with a renal clearance (eGFR) < 50 mL/min.
Sulphonyl urea derivates can be used as second choice oral glucose-lowering agents, probably as “add-on” to metformin if metformin alone is ineffective or if there is intolerance or a contraindication to metformin. The most widely used sulphonyl urea derivates are gliclazide, glibenclamide, glimepiride and glipizide. The two main disadvantages are the risk of hypoglycaemia and strong drug interactions with rifampicin that show wide inter-individual variation but result in their efficacy being reduced by 30–80% [
74].
Some argue that insulin is the preferred glucose-lowering treatment for TB–diabetes patients, but especially under programmatic conditions insulin is probably the third choice, except for sick and hospitalised patients, or patients who were already using insulin prior to a TB diagnosis. Insulin is indicated in cases of severe hyperglycaemia (e.g., HbA1c > 10% or blood glucose > 18 mmol/L). It has unlimited efficacy, but it is more expensive, requires refrigeration and subcutaneous injection and is associated with a risk of hypoglycaemia. In well-resourced settings, the use of insulin is usually accompanied by the need for self-monitoring of blood glucose through glucometers. Unavailability, insecure supply or high costs of diabetes medication, as well as issues related to storage and use of insulin, all compromise glycaemic control in many resource-constrained settings [
75,
76].
4.4. Adjustment of Treatment According to Patient Characteristics and Local Circumstances
Disease phenotype and severity (both with regard to TB and diabetes) and local circumstances will guide choice, timing and dosing of glucose-lowering drugs. Patients who are under diabetes treatment at the time of TB diagnosis can mostly continue their medication, although worsening hyperglycaemia as a result of TB may require intensification (e.g., increased insulin dose). Oral glucose-lowering drugs may have to be substituted or adjusted as, except for metformin, their metabolism is increased and therefore their glucose-lowering effect is decreased with concurrent use of rifampicin.
In TB patients with newly or known yet untreated diabetes, choice and timing of glucose-lowering drugs depends on the level of hyperglycaemia and local circumstances. If hyperglycaemia is mild (e.g., a HbA1c < 8%), initiation of glucose-lowering drugs can probably be postponed for at least 2–8 weeks. Mild hyperglycaemia often disappears with TB treatment only, and early start of glucose-lowering treatment may jeopardise successful TB treatment because of side-effects, drug toxicity, pill burden or difficulty for practitioners and patients to focus on two diseases at the same time.
Most TB patients are managed in ambulatory care and monitored at weekly or longer intervals in community or hospital outpatient clinics. In the early phase of TB treatment, TB patients should preferably not be referred to specialised diabetes services because of the risk of transmission of
Mycobacterium tuberculosis to those working in or attending these clinics. In addition, separate management of TB and diabetes has the risk of unrecognised drug interactions, medication errors and decreased retention to either TB or diabetes treatment. Thus, preferably, diabetes treatment is delivered at the TB clinic, as has been reported previously [
77].
One question is how diabetes can best be managed in TB clinics. To achieve optimal glycaemic control during TB treatment, monitoring of blood glucose during the course of TB treatment may have to be more frequent. 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. As such, a less intense schedule, preferably following the established decision points in TB treatment after two and six months, would offer significant advantage. To address this dilemma, the TANDEM project has conducted a randomised study in Indonesia to evaluate the effect of regular scheduled glucose monitoring and algorithm driven adjustment of diabetes medication on glycaemic control of diabetes in TB patients (NCT02106039). Among 150 TB–diabetes patients with a median baseline HbA1c > 11%, compared to standard care, use of a structured clinical algorithm led to much steeper decline of HbA1c (with a difference of 1.82% HbA1c; p < 0.001) and doubling the proportion of patients with an HbA1c < 8% at six months (Koesoemadinata, under review). This suggests good glycaemic control in TB–diabetes can be attained through a package of education and use of simple treatment algorithms.