The association between tuberculosis and the development of insulin resistance in adults with pulmonary tuberculosis in the Western sub-district of the Cape Metropole region, South Africa: a combined cross-sectional, cohort study

The existence of a bi-directional relationship between tuberculosis (TB) and insulin resistance (IR)/diabetes has previously been alluded to in the literature. Although diabetes has been linked to an increased risk for TB development, the role of TB as a causative factor for development of IR remains unclear. Insulin resistance is often defined as a condition where the body cells become resistant to the effects of insulin [1–4], resulting in a larger than normal insulin release to maintain a normal glycaemic response in the body [5]. Insulin resistance has previously been postulated to have either a genetic or environmental causation [6] and has been implicated in the aetiology of several diseases/conditions, such as the metabolic syndrome, cardiovascular disease, polycystic ovarian syndrome, hypertension, type 2 diabetes mellitus and obesity [6–9]. Although accurate diagnosis and measurement of IR is currently challenging, the estimated global prevalence ranges from 20 to 40% in the general population [10, 11].

It has been hypothesised that stress resulting from a long-term infection, such as TB or HIV, could increase the occurrence of IR in the body [12–14]. Recent literature has also described occurrences of impaired glucose tolerance, distortions in carbohydrate metabolism and altered insulin action among newly diagnosed TB patients [15–17]. The pro-inflammatory response accompanying a period of infection is postulated to result in decreased insulin production, which leads to a hyperglycaemic state [12]. This process may also be accompanied by the release of certain stress hormones such as epinephrine, cortisol and glucagon, which further impair the action of insulin [13]. The phenomenon of ‘transient hyperglycaemia’ (subsiding of glucose intolerance upon diagnosis after active TB treatment) can also not be discounted [18, 19]. The effect of Rifampicin, one of the pharmaceutical agents used in the treatment of TB, has also been found to result in transient hyperglycaemia soon after treatment commencement due to its strengthening of intestinal glucose absorption [20]. Furthermore, according to Schwartz’s theories, it has been postulated that the pancreas could be assaulted by TB either via concomitant pancreatitis (resulting in heightened susceptibility to inflammation and/or amyloidosis) or via the forced habitation of the pancreas [21, 22]. Moreover, the persistence of TB bacteria in adipose tissue has been thought to be a possible causative factor for systemic IR [23].

To the author’s knowledge, there are no published studies documenting IR prevalence in pulmonary tuberculosis (PTB) patients using HOMA-IR and QUICKI tests. Given the high prevalence of both communicable and non-communicable diseases in South Africa, it was considered prudent to investigate the relationship between the two morbidities in a developing country setting. The study therefore aimed to determine if an association existed between TB and IR development in ambulatory adults with newly diagnosed PTB through the use of HOMA-IR and QUICKI. It was additionally aimed to document changes in IR status during follow-up. Once IR participants were identified, it was intended to document any differences between the IR and non-IR groups.

Despite the mean BMI at baseline being classified in the normal range for both males and females, there was an overall prevalence of 33.9% of undernutrition (BMI <18.5 kg/m²) among the study population. These rates of undernutrition are in agreement with previously described rates of between 20% and 71.6% [37].

A low BMI at diagnosis has been linked to an increased risk of relapse [38] and demonstrates an indirectly proportional relationship existing between BMI and mortality risk [39]. Previous studies have also indicated that the occurrence of undernutrition among TB patients cannot be solely attributed to the disease itself but rather to a multitude of contributing factors, such as extreme poverty, food insecurity and reduced health-seeking behaviour [40, 41].

The majority of anthropometrical measurements showed a significant increase whilst the patients were on treatment, with the greatest changes taking place during the intensive phase (first two months of treatment) [42]. However, some patients failed to gain more than 5% weight over the five-month period and/or remained in the underweight BMI category, thereby placing them at greater risk of treatment failure and/or relapse at a later stage [38]. Overall weight gain of TB patients whilst on treatment is a common phenomenon [43–47] although some researchers are hesitant to use weight gain as a marker of successful response to treatment [48, 49].

Findings verifying low levels of albumin and high CRP levels in TB patients have been well documented in literature [50–54]. Baseline study results yielded an elevated mean CRP, while the mean albumin level was in the normal range. It is perhaps the raised CRP level that is more indicative of active TB disease because it has been described as a non-specific measure of systemic inflammation in the body [51]. The degree of CRP escalation has also been linked to the presence of weight loss and to disease severity, and this may increase levels even further [51]. The albumin levels were possibly not as severely affected due to the patients being relatively ‘well’ TB patients (i.e. with predominantly normal BMI’s, not hospitalised, drug-resistant or HIV-co-infected).

A lowered total cholesterol level has previously been described in the literature [51, 55, 56], in which it was hypothesised that prolonged persistence of the TB bacterium may result in cholesterol breakdown, especially in persons who had been infected latently for a period of time. Reduced HDL levels have been reported in numerous studies with various infectious or inflammatory conditions [57–59]. A study by Deniz et al. in 2007 showed lowered levels of HDL-cholesterol in their PTB-specific study population [56], which could be linked to the activated acute phase response (APR) often seen in TB disease [60–63].

The current study saw an increase in albumin levels, as well as a decrease in CRP (both over time). This would suggest a ‘resolution’ of the APR or a suppression of the inflammatory response [52], which is a phenomenon well reported in literature, both for albumin [64] and CRP [52–54, 64, 65]. There were, however, some patients in the follow-up group who still presented with a raised CRP at the five-month mark, which could perhaps indicate a sub-optimal treatment response [53] or the presence of other infections, especially if the CRP did not show an overall downward trend. Total cholesterol, as well as LDL-and HDL-cholesterol, experienced a significant increase during the intensive treatment phase. There were however no significant differences between the two-and five-month periods. This is an unusual finding because one might expect the values to increase even more at the end of the follow-up period and warrants more investigation given the important role of adequate cholesterol levels in protection against the mycobacteria.

One individual in the current study was classified with IFG (1.7%) and two with diabetes (3.4%). A systematic review published by Jeon et al. in 2010 showed total DM prevalence in TB patients to fall between the range of 1.9% and 35% [66], but some of these were diagnosed only after development of TB. The two ‘suspected’ diabetic patients were, however, not confirmed as having diabetes. In a cross-sectional study conducted in West Africa, participants displayed a 5% IFG and a 1.9% DM rate, which is comparable with the current study despite differing inclusion criteria [67]. A retrospective study performed in Sri Lanka found 7.1% pre-existing DM among their study sample, as well as 20% IFG and 2% DM rates (although lower fasting glucose cut-off values were used) [68]. These raised levels could also be attributed to the “stress/transient hyperglycaemia” phenomenon [18, 19, 69, 70], as well as the pro-inflammatory response [12, 13] and medication effects [20].

The HOMA-IR is one of the so-called ‘fasting indices’, which utilises fasting measurements of both glucose and insulin. The HOMA-IR has been used extensively in epidemiological research and as a routine measurement in clinical practice [71] and together with the QUICKI [29] has proved to be a very popular fasting index [28]. Although the current gold-standard method of diagnosing IR, the hyperinsulinaemic-euglycemic clamp (HEC), is very often the preferred diagnostic tool, the inherent study design did not lend itself to this particular technique, largely due to the time-consuming nature of the test and the fact that study participants were ambulatory out-patients. A notable drawback of the HOMA-IR and QUICKI is, however, the lack of a standardised cut-off point to identify individuals with IR. Previous studies have suggested that IR occurs between the HOMA-IR levels of 2.1 and 3.8 [28, 72], which is a vast range but in which the value of 2.477 calculated in this study fits comfortably. Using this calculated cut-off point, the IR prevalence at baseline among this study population was 25.4%, which equates to one in four persons with newly diagnosed PTB having IR. These findings were echoed by the QUICKI. This prevalence rate falls within the proposed range of 20–40% IR in the general population, of whom the majority are healthy persons (i.e. do not have TB) [11, 71]. Recent studies have duly reported high levels of hyperglycaemia, impaired glucose tolerance and DM in TB patients [12, 15–19, 66, 69, 70, 73–78].

The follow-up profiles of both HOMA-IR and QUICKI seemed to indicate an improvement in IR over time. Although there were no significant differences noted between any of the time frames, there was an overall downward pattern with the HOMA-IR and an upward curve for the QUICKI, both signifying a ‘lessening’ of the IR state. This ties in closely with an improved and diminishing inflammatory response, as demonstrated with the improved CRP, albumin and white cell count. This could also be due to the well-described ‘stress or transient’ hyperglycaemia seen to occur in the early stages after TB diagnosis and which often resolves with progression of treatment [18, 19, 73]. Despite the largely improved IR status over the five-month follow-up period, two participants who had normal HOMA-IR and QUICKI values at baseline subsequently developed IR at two and five months, and two patients redeveloped IR at five months after originally presenting with it at baseline. This contests the stress hyperglyacemia hypothesis, as well as the improvement of an inflammatory state, and is an interesting phenomenon to be pursued.

The development of IR has often been linked to the process of aging [29, 79], since age progression is generally associated with greater gains in body weight and/or fat mass, especially in the central area of the body [80], as well as with an increased prevalence of chronic diseases of the lifestyle, including metabolic syndrome [81, 82]. As the younger patients in this study displayed a greater prevalence of IR, this could signal the need for increased awareness of greater disease potential among the more youthful TB population, despite contrasting results to date. Recent research has indicated the possibility of genetic influences on the development of IR, although at the present moment, minimal variants have been associated herewith [83].

There is generally a greater risk of IR with increasing anthropometrical measurements [84], and positive correlations have been documented between IR and BMI [85, 86], waist and hip circumference, body fat content and weight gain [87, 88]. Several of these were seen in the current study (i.e. percentage body fat) and could perhaps be indicative of the phenomenon of increasing anthropometric measurements. A study conducted in 2012 by Addo et al. among adolescents in the United States showed that skinfolds (tricep and subscapular) were able to identify individuals at risk of developing IR [89]. The sum of skinfolds could perhaps be used as a marker to identify those at risk of developing IR, which may be prudent in populations that do not exhibit typical metabolic syndrome-like characteristics.

Because IR is largely termed an ‘inflammatory’ condition, levels of CRP (a positive acute phase protein) are often raised [90]. The inflammation typically found in IR is more of a ‘low-grade’ inflammatory state, which is often the result of the production of cytokines from visceral adipose tissue (VAT) [91]. The CRP levels are often produced as a response to the release of various cytokines designed to enable the inflammatory response [92]. One might thus have expected the CRP levels of the IR participants to be higher than the non-IR group, given the inflammatory nature of IR itself, its relationship with the APR and the fact that TB patients generally have increased CRP levels. However, the converse was seen. Given the fact that patients included in this study did not fit the conventional ‘metabolic syndrome’ or obesity profile, it could be speculated that they may have slightly higher levels of subcutaneous fat, compared with visceral tissue, which would, therefore, render a slightly less inflammatory profile, and this could have an inhibitory effect on CRP levels in the blood.

Leptin, an adipokine, has recently come under scrutiny regarding its role as a regulator of the immune system, as well as its effect on appetite [93] and IR reduction [94]. Findings regarding leptin levels in TB have to date been conflicting, with some studies showing higher leptin levels in TB patients [95–97] and others showing lower levels [98–101]. Individuals with reduced leptin concentrations have experienced an increased body weight over a short time duration, [102] which may be prudent in further investigating the link between TB and IR.

Recommendations for clinical practice include the implementation of integrative bi-directional screening (TB and DM) in health care facilities, as well as close monitoring of TB patients presenting with hyperglycaemia upon treatment commencement. Vulnerable patients with an undesirable anthropometrical, biochemical or clinical presentation should also be referred timeously for nutritional support, preferably during the intensive phase of treatment. Regarding future research avenues, it would be prudent to assess IR prevalence in other susceptible TB populations (HIV-co-infected, extra-pulmonary patients and different life stages) as well as to utilise additional IR diagnostic tools. Re-assessment of patients who developed IR at the five-month mark may also yield valuable results once treatment is completed. Although it was not possible in this study, a desired outcome could be the determination of a specific biomarker cut-off (e.g. CRP) that would be able to identify IR in TB patients and largely act as a diagnostic tool.

Limitations of the study included the paucity of standardised reference values for HOMA-IR and QUICKI, as well as issues surrounding the reliability of the fasting insulin measurement. It was furthermore unclear whether patients presented with hyperglycaemia before study recruitment. Logistical issues such as the stipulated time of data collection (as IR is reportedly higher in the morning), patients already having commenced with TB treatment, use of a single recruitment site, compliance with overnight fasting guidelines and inherent aspects of the study design (including sample size) could also be viewed as limitations.

This study found an association between TB and IR development in newly diagnosed PTB patients, with one in four patients having IR. This high prevalence rate in the study population signals the need for early identification, especially in the younger and more vulnerable TB population, in order to facilitate a possible reversal of IR and prevent future IR-related complications.