Patients and treatment
This observational study was conducted in two tertiary care hospitals in Dhaka from June 2014 to May 2017. TB patients (n=40) were recruited from the National Institute of the Diseases of the Chest and Hospital (NIDCH) and TB-DM patients (n=40) from the Bangladesh Institute of Research and Rehabilitation for Diabetes, Endocrine and Metabolic Disorders (BIRDEM). Age- and sex-matched healthy individuals (n=20) were randomly recruited as controls. Diabetes status was defined (according to WHO criteria) as patients with either a fasting blood glucose concentration ≥7 mmol/liter or HbA1c ≥ 6.5%. Inclusion criteria: adult males and females, age 18-60 years, newly diagnosed TB using sputum-smear microscopy and/or GeneXpert MTB/RIF and a history of diabetes for ≤5 years (TB-DM group). Exclusion criteria comprise of previous history of TB, systemic or miliary TB, >1 week of anti-TB treatment, pregnancy, history of type 2 DM for >5 years, and concomitant illnesses such as cardiovascular, liver or kidney diseases, cancer or HIV infection. It was decided to exclude patients with DM disease >5 years in order to avoid the multiple accompanying vascular complications that are associated with long-term DM such as cardiovascular disease, retinopathy, neuropathy and nephropathy [29].
Standard anti-TB treatment involved directly observed therapy short-course (DOTS) regimen that consists of isoniazid, rifampicin, pyrazinamide and ethambutol for 2 months followed by isoniazid and rifampicin for a subsequent 4 months. At enrollment, DM patients received different types of anti-diabetic medications including metformin hydrochloride (n=3), insulin (n=29), a combination of metformin and insulin (n=7), or other drugs (n=1).
Clinical samples and procedures
After enrolment, socio-economic status (SES), body weight, height, clinical history including duration of illness, history of contact with active TB cases, Bacillus Calmette–Guérin (BCG) vaccination, smoking habits and medication history (TB-DM patients) were recorded. SES was estimated utilizing a wealth index generated through principal component analysis of household assets [30]. At baseline, and months-1, -2 and -6 after the initiation of standard anti-TB treatment, clinical evaluation (mentioned in Table 1 and Table S1) and radiological examinations were performed.
Fasting blood and sputum samples were collected from the patients at baseline and during follow up visits to the hospitals and specimens were brought to the Laboratory at the International Centre for Diarrhoeal Disease Research, Bangladesh (icddr,b). Healthy controls provided a fasting blood sample on one occasion. Whole blood was routinely analyzed for erythrocyte sedimentation rate (ESR) using an ESR analyser (SRS 100/II, Greiner Bio-One GmbH, Kremsmunster, Austria) and complete blood count (CBC) using an automated hematology analyzer (XN-1000, Sysmex Corporation, Kobe, Japan). CBC assessment included total and differential blood cell counts (Table S2) and hemoglobin concentration. Peripheral blood mononuclear cells (PBMCs) and plasma were isolated from heparinized blood using Ficoll-Paque PLUS (GE Healthcare, Uppsala, Sweden) density gradient centrifugation. Blood glucose and HbA1c level were measured using a Clinical Chemistry Analyzer (Cobas C311, Roch Diagnostics GmbH, Mannheim, Germany). Plasma levels of the metabolic hormones insulin and C-peptide were measured by Luminex assay using a Bio-Plex Diabetes kit (Bio-Rad Laboratories, Inc. USA). The homeostatic model assessment of insulin resistance (HOMA-IR) was calculated based on the following formula: HOMA-IR = (fasting insulin [pmol/L] × fasting plasma glucose [mmol/L]) / 135 [31]. C-peptide reactivity-insulin resistance (CPR-IR) was calculated using the following formula: CPR-IR = 20 / (fasting C-peptide [nmol/L] x fasting plasma glucose [mmol/L]) [32]. Expectorated sputum samples were used for sputum-smear microscopy and culture. PBMCs and an aliquot of sputum were stored at -80°C in 1 ml of RNAlater (Qiagen GmbH, Hilden, Germany) for mRNA analysis.
Clinical composite TBscore
The Bandim TBscore [33] was applied for assessment of the following signs of TB and TB symptoms: cough, hemoptysis, dyspnoea, chest pain, night sweats, anemia, fever, body mass index (BMI) and mid upper arm circumference (MUAC). These signs of TB and typical symptoms were recorded and scored as present (1) or absent (0) [34]. In contrast to the Bandim TBscore, data on tachycardia and lung auscultation were not included and consequently this TBscore had a maximum value of 11 points.
Mtb sputum-smear microscopy, sputum-culture and the Xpert MTB/RIF assay
Sputum-smear microscopy was conducted at the NIDCH or BIRDEM, while Mtb culture was performed in the Mycobacteriology laboratory at icddr,b. Ziehl-Neelsen staining was performed to detect and grade acid-fast bacilli by smear microscopy. Decontaminated sputum pellets were cultured on Lowenstein-Jensen (L-J) slants and were examined weekly until Mtb colonies were detected to confirm the TB diagnosis. The Xpert MTB/RIF assay was performed using sputum collected from all TB patients at NIDCH or BIRDEM to rule out Mtb resistance to rifampicin according to the national TB program (NTP) guidelines in Bangladesh. Drug susceptibility testing was not performed in the isolated Mtb strains as this is not recommended by the NTP.
Chest X-ray analysis
For lung pathology analysis, a semi-quantitative visual scoring system of 2-dimensional chest X-rays was used as previously reported [35, 36], although with some modifications. For each patient, the two lungs were divided into three zones: upper, middle and lower. Presence of nodules, patchy or confluent consolidations, and areas of cavitation were recorded in each of the three zones. The effusion volume, extent of opacification, cavitation or additional pathology were graded as the percentage (%) of the affected lung. Each of the three zones in the two lungs could be graded as having a maximal pathological involvement of 100%, and therefore the total % lung involvement in the upper, middle or lower zones, respectively, could be a maximum of 100 + 100 = 200%. Accordingly, in each patient, the total % lung involvement comprising all three zones in both lungs could be a maximum of 3 x 100 x 2 = 600%.
mRNA extraction and real-time PCR
mRNA analysis of PBMCs and sputum cells was conducted to assess inflammation in the peripheral circulation compared to the local site of Mtb infection. This analysis was performed on a sub-set of the study patients from whom longitudinal and matched blood and sputum samples could be collected. This selection was random and representative of median HbA1C levels and chest X-ray scores in the total patient cohort. mRNA was extracted using Ambion RiboPure RNA extraction kit (Life Technologies, Vilnius, Lithuania). cDNA conversion was performed using a SuperScript cDNA synthesis kit (Invitrogen, Carlsbad, CA). The target genes (CD4, CD8, IL-10, IL-1b, TNF-a and matrix metalloproteinase (MMP)-9), and the house-keeping gene (18srRNA) were amplified using Taqman gene expression master mix and primers in the QuantStudio 5 Real-Time PCR System (Applied Biosystems, Foster City, CA). The cycle threshold (Ct) values for the target genes were normalized to 18srRNA and relative mRNA expression was determined using the 2−ΔΔCT method. mRNA analysis was performed using sputum samples collected at baseline, month-1 and month-2, as by 6 months, most TB patients were unable to produce sputum.
Sample size calculation
This was an exploratory pilot study where our aim was to investigate immunological responses in patients with TB-DM compared to patients with TB alone. In general, simulation studies have identified a minimum sample size requirement of >20 for conducting powerful parametric analyses, even when the data is non-normally distributed. However, human clinical data tends to be variable when involving more than 30 patients or volunteers in their respective groups. Based on our previous experience with TB studies in high-endemic countries [37-41], inclusion of >20 patients in each sub-group enabled parametric analyses, while smaller sub-group analyses were performed using non-parametric methods. Assuming an initial sample size of 30 and accounting for a loss to follow-up of 15% and low quality or insufficient amount of clinical material in another 10% of the patients, we used a sample size of n=40 in the TB-DM and TB groups.
Statistical analysis
Data is presented as mean and standard deviation (SD) or median and interquartile range (IQR) for continuous data, and numbers with percentages for categorical variables. Statistical significance was determined using non-parametric Kruskal-Wallis test with Dunn’s post-test (comparing more than two unmatched groups), Mann-Whitney test (comparing two unmatched groups) and Friedman test with Dunn’s multiple comparison test (comparing non-parametric one-way repeated measurements). The Chi-square test was used to compare the proportion of patients with TB symptoms between the TB and TB-DM groups at baseline. Multivariable regression model and GEE (related to lung pathology only) analysis were used to estimate the difference in outcome variables between the TB-DM and TB groups, and data was adjusted for covariates (age, sex, BCG vaccination status, baseline BMI and SES score). The GEE model (beta (β) value and 95% confidence interval (CI)) were created with an interaction (follow-up month × groups) to analyze changes in blood hemogram markers and percent lung involvement across time by group status. This analysis was performed at baseline, and during three follow up visits with the following predictors: Time (months 0, 1, 2 and 6), group, and an interaction term of these two (visit month × group). Covariates that influenced the model R2 by 5% or more were selected to avoid collinearity. To evaluate changes within groups in the TBscore, BMI, radiological feature, blood glucose, HbA1c and CBC, analyses were performed using 2-way repeated measure Analysis of variance (ANOVA). Spearman´s correlation test was used for the correlation analyses. Stata/IC (v.13, Stata Corp., LP, College Station, Texas, USA) and GraphPad Prism 7.05 were used for statistical analysis. A P-value < 0.05 was considered as significant.