High prevalence and heterogeneity of Dysglycemia in patients with tuberculosis from Peru: a prospective cohort study

Tuberculosis kills 1–2 million people per year, especially in low and middle-income countries, and despite recent efforts to improve control programs, its incidence is declining at slow rates [1]. High-risk groups undermine the success of TB programs in the reducing the disease burden notwithstanding targeted interventions. One relevant aspect is diabetes mellitus (DM), which increases the risk of developing active TB in approximately three times [2]. More recently, prediabetes (preDM) has been also described as a risk factor for developing TB [3, 4]. Both the World Health Organization (WHO) as the Peruvian National TB Program (NTP) recommend screening for DM in people with active TB and for TB between household contacts [5]. Despite those indications, most individuals are unaware of their DM or pre-DM status. In 2017, the Peruvian NTP communicated a DM incidence of 6.2% with a testing coverage of 77.9% of all TB patients whereas other instances of the national government reported in 2016 a DM incidence around 10.4% (government communication). Those differences may reflect several limitations in the DM screening such as the use of only fasting plasma glucose (FPG) as the screening approach [5]. It is widely known that sensitivity of DM tests (such as HbA1c, fasting glucose and oral glucose tolerance) is variable [6‐9]. It has been reported that HbA1c detects more people with DM [10] (or preDM) with higher sensitivity than other screening tests [11], so systematic use of FPG as screening tool, as observed in Peru, may result in incomplete record of TB-DM cases [12].

Due the discrepancies in sensitivity of DM screening tests and in order to determine the real prevalence of DM or preDM in TB patients and their household contact (HHC), we conducted a study in Lima, Peru, with the hypothesis that both DM and pre-DM are more frequent than previously officially reported. In addition, we compared the performance of different screening methods to detect DM and preDM in TB cases. Furthermore, we examined the associations between different other clinical and sociodemographic risk factors with the possible more extended comorbid TB-DM burden.

We initially screened 349 microbiologically-confirmed TB cases at the primary health care centers which were part of the present study, between February and November 2017. During the screening, 206 individuals were excluded for a number of reasons listed in Fig. 1a, and 143 patients with active TB were examined in the first study visit (Fig. 1a and Fig. 1b). At this stage, additional 7 persons were excluded due to HIV diagnosis (n = 3) and consent withdraw (n = 4), resulting in a cohort of 136 patients. During the search of HHCs of the TB index cases, 205 people were identified for screening and only 149 were effectively screened. Of those, 8 withdrew consent and 3 were excluded for having positive HIV status (n = 1) or pregnancy (n = 2) resulting in a total of 138 HHC participants (Fig. 1).

At enrollment, the DM prevalence was 13.97% (95% CI: 8.14, 19.80%) (n = 19) among TB patients (all of whom refereed DM diagnosed prior to study enrollment), while the prevalence of DM among HHC was 6.52% (95% CI: 2.40, 10.64%) (n = 9). The prevalence of preDM was 30.88% (95% CI: 23.81, 38.65%) in TB patients (n = 42) and 28.99% (95% CI: 21.42, 36.56%) among the HHC (n = 40). Diagnosis of preDM was performed at study enrollment. Comparisons at this timepoint revealed that TB and HHC groups exhibited similar frequencies of DM and preDM (Fig. 2a). After 2 months of antitubercular treatment (ATT) commencement, DM frequency was nearly double in TB vs. HHC whereas preDM was higher in the latter group (Fig. 2a). At month 6 of ATT, frequency of DM and preDM was once again not different between the study groups, although there was a remarkable 75% loss to follow-up in the HHC group at this timepoint (Fig. 2a).

Among the TB index cases at study baseline, DM individuals were on average older than preDM and normoglycemic individuals (median age: 46.41 yrs. IQR:33.5–54.8 vs. 39.8 yrs., IQR: 26.7–54.0 and 26.4 yrs. IQR: 22.3–34.7, respectively, p < 0.01) (Table 1). In addition, lower level of education (primary and secondary school years) was associated with presence of dysglycemia (DM or preDM) (p < 0.01; Table 1). Hypertension was more frequently in DM patients (21.1%) than that in normoglycemic TB cases (7.3%) (p < 0.01). The study groups were similar with regard to a number of other characteristics including sex, history of prior TB, asthma, renal disease and lifestyle habits (alcohol use, smoking and illicit drugs use) (Table 1).

When the cohort of HHC was stratified according to the glycemic status, we found once again that DM patients were on average older than those with preDM or normoglycemia (Table 2). In addition, the highest values of body mass index (BMI) and waist circumference were detected in the subgroup of preDM HHC (Table 2). Other characteristics were similar between the HHC subgroups.

The FPG and HbA1c tests were performed prospectively in all patients recruited (except OGTT for known DM cases). At the baseline visit, as expected, TB patients with coincident DM exhibited higher values of FPG and HbA1c than those diabetic but without TB (Fig. 2b). Such discrepancies in laboratory measurements were reduced at month 2 and 6 of ATT and no statistically significant differences were observed between diabetic patients with or without TB (Fig. 2b). The OGTT was performed only at baseline in patients without prior DM diagnosis. One individual was identified as having DM (from the HHC group). Once again, presence of TB was associated with higher values of OGTT results than in HHC individuals, suggesting that Mtb infection may drive inflammation-associated dysglycemia [19, 20] (Fig. 2b).

Moreover, the frequencies of TB-related symptoms between TB index cases with different degree of dysglycemia were compared and we found no statistically significant difference (Fig. 3a). Moreover, frequency patients presenting with increased acid-fast bacilli grades in sputum smears was higher in TB cases with DM compared to the other groups (chi-square p = 0.004; Fig. 3b). These findings indicate that TB clinical presentation is not worse in diabetics as we have previously demonstrated in different studies from Brazil [3, 19], but confirmed previously reported data showing increased mycobacterial loads in sputum smears [19].

Circulating hemoglobin levels were low in diabetic patients with coincident TB compared to those with preDM or normoglycemia (Fig. 4a and b). In fact, within the group of TB, anemia was detected in all diabetics but only in 73.8% of preDM (n = 31) and 66.2% of normoglycemic patients (n = 49) (p = 0.006). In HHC, diabetes was not associated with substantial changes in hemoglobin levels (Fig. 4a and b). Moreover, frequency of anemia was not different between the groups of TB contacts presenting with diverse degree of dysglycemia (p = 0.08). This finding suggested that a potential synergistic effect of DM and TB on the degree of anemia may exist. Indeed, results from FPG and HbA1c in TB patients demonstrated that values observed in DM were substantially higher than in preDM or normoglycemia, and similar trend was observed in OGTT between preDM and nondiabetics (Fig. 4c). Of note, in HHC, values of FPG and HbA1c were lower in anemic individuals compared to those with normal hemoglobin levels (Fig. 4c). Furthermore, OGTT levels were able to distinguish preDM from normoglycemia only in the subgroup of participants without anemia (Fig. 4c).

We next analyzed the overall accuracy of FPG and HbA1c in detecting DM or preDM cases in individuals stratified according to TB diagnosis (TB patients and HHC). FPG levels were able to detect more cases with preDM, but not with DM, than HbA1c did in the entire study population (Fig. 5a). When only TB cases were considered, again FPG was able to detect more preDM patients then HbA1c (Fig. 5b). Similar findings were observed in HHC (Fig. 5c). These results indicate that the laboratory tests examined here displayed different values of accuracy to detect DM or preDM in the study population. Summary of the accuracy and predictive values are shown in Additional file 1: Table S1. Concordance on DM diagnosis using either FPG or HbA1c levels was low, with lower performance of HbA1c (Fig. 6a). In addition, poor agreement was also observed between OGTT versus HbA1c or FPG tests to detect preDM in both TB patients and HHC (Fig. 6b).

Estimating the prevalence of dysglycemia among TB cases is important to understand the real burden of TBDM and drive changes to optimize detection and treatment of this comorbidity. In the present study, high prevalence of DM and preDM was found in both TB patients (14%) and HHC (6.5%), which is unexpectedly higher than previous reports from the Peruvian Minister of Health (6.2%) [21], but similar to results from a recent report [22]. In the last years, low and middle income countries, such as Peru, have experienced a nutritional transition [23]. These populations historically affected by hunger and malnutrition now face additional problems of obesity and other non-communicable diseases, such as diabetes and hypertension [24]. Because DM is a known risk factor of TB [2, 25, 26], systematic screening should be performed in high-risk populations. Nevertheless, maybe due to high cost, many countries do not perform systematic screening and for this reason the dual burden of TB and DM is likely still underestimated. Findings from this study and others [27, 28] argue that, to provide a more reliable picture of the TBDM/preDM burden and based on some other references, it is necessary to consider the joint use of FPG and HbA1c in all patients, while OGTT could be used, when it is safe, to rule out prediabetes.

In our prospective evaluation, we found that frequency of DM remained stable during the follow up period, including in the group of TB patients. This could be due to the fact that most of the DM patients investigated exhibited high levels of HbA1c, compatible with uncontrolled diabetes [29] and these values remained high during the follow-up, in disagreement to results from a previous report [30]. We hypothesize that patients with very high levels of HbA1c are not able to substantially decrease values after ATT-induced reduction of systemic inflammation. On the other hand, frequency of preDM significantly reduced in TB patients undergoing ATT but were again stable in the follow up timepoint of the HHC cohort. This observation argues that resolution of systemic inflammation during ATT may lead to reduction of transient dysglycemia detected in some patients with TB.

There is evidence that TB clinical presentation is exacerbated in individuals [25, 31, 32]. In addition, poor glycemic control has been described to negatively impact radiographic manifestation of pulmonary disease in patients with TBDM [25]. In agreement with this idea, we have recently described in two distinct studies from Brazil that DM and also preDM were associated with increased frequency of clinical symptoms associated with TB [19]. Nevertheless, the association between DM and worse TB clinical and/or radiographic presentation has not been found by other investigations, including a recent study which examined Peruvian individuals [33]. It is possible that genetic or social differences between countries may influence the perception of the symptoms. Moreover, differences in timing from disease onset to admission at a TB clinical center may diminish the differences in clinical symptoms between normoglycemic and dysglycemic patients. Further studies are warranted to directly test these hypotheses. Interestingly, we found that TBDM patients exhibited increased AFB grades in sputum smears, which has been described by independent investigations worldwide [3, 19]. These findings reinforce the idea that DM may impact capability to restrain mycobacterial growth and thus may be associated with increased risk of transmission.

Anemia is a common clinical condition associated with TB [34]. It has been recently reported that chronic anemia is linked to a distinct systemic inflammatory profile that persists after 2 months of ATT in a Brazilian cohort [26]. In addition, hemoglobin levels have been described to affect detection of glycated hemoglobin, which brings potential challenges in DM diagnosis in patients with severe anemia [26]. Herein, all the TB-DM patients were anemic (median Hb level: 9.9 g/dL IQR: 7.6–13.4 g/dL). However, anemic status apparently did not prevent detection of DM in TB patients as the screening tests exhibited similar performance. Whether anemia impacts TB clinical presentation in DM patients is still unknown and deserves future investigation.

In our study, we found that all TB-DM people and a good proportion of TB-preDM patients were anemic. There is evidence supporting the idea that performance of HbA1c in detecting dysglycemia is affected by occurrence of anemia [35]. Herein, we analyzed the relationship between HbA1c and anemia and found no clear influence in the test performance to diagnose DM. Furthermore, our results demonstrated a low accuracy of the HbA1c vs. other tests to diagnose preDM, irrespective of the TB status. FPG was a better predictor of preDM than HbA1c in every condition tested. These observations suggest that FPG is indeed a good marker to be used for screening of DM or preDM in Peru, in addition to OGTT.

Our study has some limitations. OGTT was not tested on individuals who were recruited and a prior diagnosis of DM. This led to incomplete testing in all the study participants. In addition, we had study dropouts, which although were not very high, resulted in reduced number of individuals who were investigated in the latter study visits. Finally, the HbA1c levels could have been affected by the degree of anemia, as demonstrated before [36, 37]. Our study was not designed to directly test the effect of anemia on changes in HbA1c levels and thus further research is warranted to address this question. Regardless of such limitations, our study results clearly show the high prevalence and heterogeneity of DM and preDM in a Peruvian population, corroborating with observations reported from other countries such as India [38].

The results presented here reveal a previously under notified high prevalence of DM and preDM in TB patients from Peru. Thus, screening strategies for both DM and preDM should be implemented in under routine program conditions for all patients. Using only HbA1c can lead to an error of sub-notification of preDM so the additional implementation of FPG and OGTT is needed in all cases where dysglycemia is suspected but not confirmed. Implementation of such screening tests in routine investigation at TB clinics could result in early detection of hyperglycemia and timely therapeutic interventions, which lead to improvements of clinical outcomes.