Page 89 of 114 European Respiratory Journal
1 2 3 4 Trilateral overlap between tuberculosis, diabetes and HIV-1 in a high burden African 5 6 setting: Implications for TB control. 7 8 9 10 11 Tolu Oni 1,2 MD(Res), Natacha Berkowitz 1,2 MBChB, Mmamapudi Kubjane 1 MPH, Rene 12 13 Goliath 2 BSc(Hons), Naomi S Levitt**3 MD, Robert J Wilkinson**2,4,5 PhD. 14 15 16 1. Division of Public Health Medicine, School of Public Health & Family Medicine, 17 18 19 University of Cape Town, Cape Town 7925, South Africa. 20 21 2. Wellcome Centre for Infectious Disease Research in Africa, Institute of Infectious Disease 22 23 24 and Molecular Medicine, University of Cape Town, Cape Town 7925, South Africa. 25 26 27 3. Division of Diabetes and Endocrinology, Department of Medicine, Groote Schuur 28 29 Hospital, Cape Town, 7925, South Africa and Chronic Disease Initiative for Africa. 30 31 32 4. The Francis Crick Institute, London NW1 2AT, United Kingdom. 33 34 5. Department of Medicine, Impreial College London W2 1PG, United Kingdom. 35 36 37 38 39 40 *Corresponding author: Associate Professor Tolu Oni, Room 4.41, Entrance 5, Level 4, 41 42 Falmouth building, Faculty of Health Sciences, University of Cape Town, Anzio road, 43 44 45 Observatory, Cape Town 7925. Tel: +27 21 650 1299. Email: [email protected] 46 47 48 **Joint Senior Authors 49 50 Key words: tuberculosis (TB), diabetes mellitus (DM), impaired glucose regulation (IGR), 51 52 53 determinants, HIV 1, sub Saharan Africa, South Africa. 54 55 56 [Word count, Body:] 3000 words 57 58 59 60 1 European Respiratory Journal Page 90 of 114
1 2 3 4 5 6 Take home message 7 8 9 DM/TB association is significant in HIV infected, but more accurate glycaemia markers at 10 11 TB diagnosis are needed. 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 2 Page 91 of 114 European Respiratory Journal
1 2 3 Abstract 4 5 Background 6 7 8 The diabetes (DM) burden is growing in countries where tuberculosis and HIV 1 remain 9 10 major challenges, threatening tuberculosis control efforts. This study determined the 11 12 13 association between tuberculosis and diabetes / impaired glucose regulation (IGR) in the 14 15 context of HIV 1. 16 17 18 Methods 19 20 A cross sectional study was conducted at a TB clinic in Cape Town. Participants were 21 22 23 screened for DM and IGR, using fasting plasma glucose, oral glucose tolerance test and 24 25 HbA1c. 26 27 28 Results 29 30 31 414 TB and 438 non TB participants were enrolled. In multivariable analysis, diabetes was 32 33 associated with tuberculosis (Odds ratio: 2.4; 95% CI 1.3 – 4.3 p=0.005); with 14% 34 35 population attributable risk fraction. but this association varied by diagnostic test (driven by 36 37 HbA1c). The association remained significant in HIV 1 infected (Odds ratio: 2.4; 95% CI 38 39 40 1.1 – 5.2; p=0.030). A high prevalence of IGR (65.2% amongst tuberculosis cases) and a 41 42 significant association with tuberculosis (Odds ratio: 2.3; 95% CI: 1.6 – 3.3; p<0.001) was 43 44 also found. 45 46 Conclusions 47 48 49 Diabetes and IGR prevalence was high and associated with tuberculosis, particularly in HIV 50 51 1 infected persons, highlighting the importance of DM screening. The variation in findings 52 53 54 by diagnostic test highlights the need for better glycaemia markers to inform screening in the 55 56 context of tuberculosis and HIV 1. 57 58 59 60 3 European Respiratory Journal Page 92 of 114
1 2 3 Introduction 4 5 Tuberculosis (TB) remains a health challenge in most low and middle income countries 6 7 (LMIC) including sub Saharan Africa (SSA). Poverty, smoking, alcoholism, human 8 9 10 immunodeficiency virus (HIV 1) and diabetes mellitus (DM) are known drivers of the TB 11 12 epidemic.[1] DM triples the risk of TB;[2] and is associated with adverse outcomes.[3,4] The 13 14 TB/DM association is increasingly important in LMIC experiencing a growing burden of 15 16 non communicable disease. According to the 2013 Global Burden of Disease Study, HIV 1, 17 18 19 TB and DM rank first, third and ninth among the top ten causes of life years lost in South 20 21 Africa (SA).[5] 22 23 24 The prevalence of DM among TB patients varies by geographic region and background DM 25 26 population in the general population, with a higher prevalence reported in South Asian 27 28 (range: 25·3 44%)[6 10] and Latin American (range: 14 39%)[11,12] populations. In SSA, 29 30 DM prevalence in TB patients ranges from 8·5% to 16·4%,[13 16] although a large 31 32 33 proportion of DM in SSA remains undiagnosed.[17] Of note, most studies have used fasting 34 35 or random blood glucose to diagnose DM, which may underestimate the true prevalence. A 36 37 study in SA estimated the general population prevalence of DM and impaired glucose 38 39 tolerance to be 12·3% and 11·2% respectively.[18] 40 41 42 The increased risk of TB associated with DM is well documented. However, this risk is 43 44 variable dependent on the population studied. A systematic review that reported a 3 fold 45 46 47 increased risk of TB in DM patients[2] did not include studies from SSA where there are 48 49 competing TB risk factors such as HIV 1. Recent studies in African settings showed 50 51 increased odds of TB among DM patients with the exception of one study,[19] possibly due 52 53 to the low background prevalence of DM. A Ugandan study found HIV 1 infected TB 54 55 patients to have reduced odds of DM compared to HIV 1 uninfected,[20] and a Tanzanian 56 57 58 study showed a stronger TB/DM association observed among HIV 1 uninfected (OR = 4.23, 59 60 4 Page 93 of 114 European Respiratory Journal
1 2 3 95% CI 1.54 – 11.57) compared to HIV 1 infected (OR = 0.14, 95% CI 0.01 – 1.81) 4 5 individuals.[21] The reasons for this apparent incongruity in the associations between TB, 6 7 DM and HIV 1 are unclear as there are only a limited number of published studies.[22] 8 9 10 11 12 13 The only SA study that assessed the prevalence of DM, using the oral glucose tolerance test, 14 15 among TB patients was conducted in 1980 when SA was 42% urbanised (compared to 62% 16 17 18 urbanised today);[16] and reported a 2·1% prevalence. SA has the largest antiretroviral 19 th 20 therapy (ART) program globally,[23] and is ranked as having the 6 highest TB incidence 21 22 globally.[24] Given this HIV 1/TB burden, reducing the TB burden in South Africa is a 23 24 priority in the global fight against TB. The strong association between HIV 1 and TB, 25 26 27 dysglycaemic effects of some ART drugs, especially protease inhibitors and nonnucleoside 28 29 reverse transcriptase inhibitors ,[25, 26] and the emerging DM epidemic highlight the 30 31 importance of investigating the association between TB, DM, and HIV 1 in this setting, the 32 33 results of which will have implications for TB control strategies. This study investigated the 34 35 prevalence of DM and impaired glucose regulation (IGR), the association between TB and 36 37 38 DM/IGR; and the population attributable risk of TB due to DM in South Africa. 39 40 41 42 43 Methods 44 45 Study setting and population 46 47 In South Africa, TB patients in the public sector are largely screened and treated in TB 48 49 50 clinics. This study was conducted at the largest TB clinic in Khayelitsha, a peri urban 51 52 township of 390 000 predominantly black Africans, in Cape Town, Western Cape province. 53 54 In this province, DM, HIV 1 and TB rank first, second and third leading causes of death, 55 56 respectively.[27] The 2012 HIV 1 antenatal prevalence in Khayelitsha was 34% (95% CI 57 58 59 60 5 European Respiratory Journal Page 94 of 114
1 2 3 31.0 36.6%) (unpublished data from the 2012 Western Cape Department of Health Antenatal 4 5 Survey) and the 2015 TB case notification rate was 917 per 100 000 population (Virginia de 6 7 Azevedo, personal communication), with a 60% HIV 1/TB co infection rate.[28] 8 9 10 Study design and sampling 11 12 13 We conducted a cross sectional study on consecutive patients with respiratory symptoms 14 15 presenting to the clinic from July 2013 August 2015. Patients were eligible if they provided 16 17 consent, were 18 years or older, and had not received more than 48 hours of TB 18 19 chemotherapy. Those who were critically ill and in need of emergency clinical care were 20 21 ineligible as they were too physically unwell to give informed consent. The study was 22 23 24 approved by the University of Cape Town Human Research Ethics Committee (HREC REF: 25 26 403/2011). Assuming a 3 fold higher DM prevalence in HIV 1 uninfected TB cases, 7% DM 27 28 prevalence in HIV 1 uninfected, 70% of TB cases HIV 1 co infected and 80% power, the 29 30 study aimed to recruit 400 TB and 400 non TB participants. 31 32 33 Study procedures 34 35 Case-definitions: TB cases were diagnosed according to SA guidelines,[29] with the Gene 36 37 38 Xpert analysed in a centralised national health laboratory. Non TB participants were those 39 40 with a negative TB diagnosis and resolution of respiratory symptoms. All participants were 41 42 encouraged to undertake an HIV 1 test, and were tested for DM using all 3 of the following: a 43 44 fasting plasma glucose (FPG), a 2 hour oral glucose tolerance test (OGTT) and glycosylated 45 46 47 haemoglobin (HbA1c). DM diagnosis was defined as:[30,31] self reported DM, FPG ≥7·0 48 49 mmol/L; OGTT ≥ 11·1 mmol/L; or HbA1c ≥ 6·5%. IGR was defined as 5·7% ≤ HbA1c < 50 51 6·5%; 5·5 mmol/L ≤ FPG < 7·0 mmol/L; or 7·7 mmol/L ≤ OGTT < 11·1 mmol/L. 52 53 54 55 Measurements: Venous blood was drawn from the antecubital vein at 0 (after an overnight 56 57 58 fast) and 120 minutes in evacuated fluoride (glucose) and EDTA (HbA1c) tubes. All blood 59 60 6 Page 95 of 114 European Respiratory Journal
1 2 3 samples were processed on the day of collection at a centralised national health laboratory 4 5 using standardised operating procedures of the Roche/Hitachi cobas c311 system analyser 6 7 assay.[32] Weight, height, and waist circumference were measured using standardised 8 9 2 10 techniques.[33] The body mass index (BMI kg/m ) was categorised: underweight: < 18·5; 11 12 normal: 18·5 ≤ BMI < 25; overweight: 25 ≤ BMI < 30; obese: ≥ 30).[33] The cutpoint for 13 14 high waist circumference was ≥ 94 cm (males) and ≥ 88 cm (females).[33] Hypertension was 15 16 defined as a single measured blood pressure (BP) of systolic BP >140 mmHg or diastolic BP 17 18 19 >90 mmHg,[31] or a pre existing diagnosis. 20 21 22 23 24 Questionnaire : Chronic disease risk factors were ascertained using the STEPs 25 26 instrument.[34] Information on socio economic, demographic, and medical history were 27 28 collected using a researcher administered questionnaire. 29 30 31 32 33 Statistical analysis 34 35 Medians and inter quartile ranges (IQR), and proportions were used to summarise continuous 36 37 and categorical variables. Chi squared and Fisher’s exact tests assessed associations between 38 39 40 categorical variables. The Mann Whitney test was used to compare medians between two 41 42 groups and the Kruskal Wallis test for more than two groups. A multivariable logistic 43 44 regression model for the association between TB and DM was manually built, using forward 45 46 selection, controlling for potential confounding variables. In order to retain statistical power 47 48 and reduce potential biases, in the regression analysis involving the key variable HIV sero 49 50 51 status, multiple imputation was used to impute HIV sero status for 50 participants with 52 53 unknown HIV 1 status. Imputation using the chained equations method was implemented in 54 55 STATA using the ice command. Five imputed datasets were created. The mi STATA 56 57 command was used to perform logistic regression analysis on the combined imputed datasets. 58 59 60 7 European Respiratory Journal Page 96 of 114
1 2 3 The base model included a priori confounding variables of age and sex. Variables associated 4 5 with TB (p<0.10) on univariable analysis were then added, sequentially including variables 6 7 that improved the model based on the significant lowering of the Akaike Information 8 9 10 Criterion (AIC) in non nested model comparisons. Potential effect modification between 11 12 variables was determined by exploring the statistical significance of interaction variables. The 13 14 fit of the model was assessed using Pearson’s goodness of fit test with a p value >0.05 15 16 indicating a good fit. The population attributable risk was calculated:[35] 17 18 19 ARpopDM=p(DM)(OR 1)/[1+p(DM)(OR 1)] where p(DM)= general population diabetes 20 21 prevalence and OR= adjusted odds ratio. Statistical significance was set at p value < 0·05. 22 23 All data were analysed using STATA 13.0 (Stata Corp, College Station, TX, USA). 24 25 26 Results 27 28 Baseline characteristics 29 30 31 A total of 986 participants were recruited. Forty eight participants (4.9%) were excluded as 32 33 they had an undetermined TB status (TB could not be confirmed or excluded). A further 86 34 35 36 participants did not complete DM screening at baseline (did not return for fasting bloods). A 37 38 total of 852 participants were therefore included in the final analyses: 414 TB cases and 438 39 40 non TB participants. The overall median age of participants was 38 years (IQR 31 – 47 41 42 years), with 53% being male. The overall prevalence of HIV 1 1 was 61.2% (95% CI 29.9 – 43 44 45 36.2%) and was significantly higher in participants with TB (versus non TB) and was in 46 47 females (Table 1). Compared to those without TB, TB cases were younger, with a lower 48 49 prevalence of obesity, and a greater proportion of men. Of the 414 TB cases, 9 had rifampicin 50 51 resistance (2.2%), all of whom did not have DM. Table 1 summarises baseline characteristics 52 53 54 of participants, stratified by TB status. 55 56 57 58 59 60 8 Page 97 of 114 European Respiratory Journal
1 2 3 Prevalence of DM and IGR 4 5 6 The overall prevalence of DM (using any of the 3 diagnostic criteria) was 11.3% (95% CI 9.3 7 8 – 13.6%); 12 6% among TB cases (95% CI 9 7–16 1%) and 10.1% (95% CI 7.6 – 13.2%) in 9 10 non TB participants (p=0.246). Among DM participants, 59.4% (57 of 96) were not 11 12 13 previously diagnosed with DM (61.5% in TB cases and 56.8% in non TB), and all previously 14 15 diagnosed DM patients were on DM treatment. Despite treatment, participants with a prior 16 17 DM diagnosis had higher median FPG (7.45 mmol/L (IQR 5.2 – 11.3 mmol/L) versus 5.45 18 19 mmol/L (IQR 5.1 – 7.1 mmol/L); p<0.001) and HbA1c (9.7% (IQR 7.0 – 11.4%) versus 20 21 6.5% (IQR 6.4 – 6.9%); p<0.001) levels, compared to newly diagnosed DM patients. 22 23 24 The overall statistically significant difference in glycaemic status between participants with 25 26 27 and without TB was driven by the prevalence of IGR which was significantly higher in TB 28 29 cases than non TB (65.2% vs 50.0%; p<0.001). The prevalence of DM and IGR varied by 30 31 diagnostic test (Table 2). The majority of DM diagnoses were made based on the FPG and 32 33 HbA1c tests, whilst IGR was largely driven by a positive HbA1c test. 34 35 36 The overall prevalence of DM was lower in HIV 1 infected patients compared to those 37 38 uninfected or those with HIV 1 status unknown (8.9% vs 16.0% or 10.0% respectively) 39 40 (Table 3). Whilst there was no association between TB and DM overall, when stratified by 41 42 43 HIV 1 status, the prevalence of DM was found to be higher in HIV 1 infected TB cases 44 45 versus HIV 1 infected participants without TB (11.1% versus 6.2%; p=0.049). This 46 47 difference was not observed in HIV 1 uninfected participants (Table 3). 48 49 50 51 52 53 TB/DM association and population attributable risk 54 55 56 On univariable analysis, age, sex, marital status, household size, income, employment status, 57 58 previous TB, HIV 1, smoking, hypertension, waist circumference, a history of time in prison, 59 60 9 European Respiratory Journal Page 98 of 114
1 2 3 education, BMI and a history of being a miner were all associated with TB at the 10% level 4 5 of significance. These variables were used in turn to build the multivariable model. After 6 7 adjusting for confounding variables, DM was associated with a 2.4 fold higher odds of TB 8 9 10 (95% CI 1.1 5.2), with this association remaining significant in HIV 1 infected but not HIV 11 12 1 uninfected persons (Table 4). Further analysis by DM diagnostic test revealed that this 13 14 significant association is only noted using the HbA1c test. IGR was also associated with a 2.3 15 16 fold higher odds of TB (95% CI 1.6 3.3); with the strongest association shown when OGTT 17 18 19 was used as a diagnostic test compared to other tests (Table 4). Unlike the DM analysis, the 20 21 association between TB and IGR (using any test) was significant in both HIV 1 infected and 22 23 uninfected persons, albeit driven by different tests: a significant association using the OGTT 24 25 test in HIV 1 infected persons, and the HbA1c test in HIV 1 uninfected persons. Based on a 26 27 12% prevalence of DM in the general population and the adjusted OR of 2.4, the population 28 29 30 attributable risk of TB due to DM is 14%. 31 32 33 34 35 Discussion 36 37 38 The prevalence of DM, using any 3 criteria, was 13% in TB cases in Cape Town, South 39 40 Africa, largely contributed to by HbA1c. There was also an alarmingly high prevalence of 41 42 43 IGR, with 65% all TB cases having IGR. There is good evidence that TB is associated with 44 45 transient hyperglycaemia, best measured using FPG and 2 hr OGTT, markers of short term 46 47 glucose status, unlike the HbA1c which is a marker of glycaemic control over 2 3 months. 48 49 The finding of the prevalence of DM being highest using the HbA1c is therefore surprising. 50 51 52 One possible explanation is the choice of cutpoint. The ADA/WHO cutpoint is 6.5%, but a 53 54 recent study investigating the diagnostic accuracy of HbA1c against OGTT in a population 55 56 survey of black South Africans, suggested that the optimal cutpoint for detection of DM was 57 58 59 60 10 Page 99 of 114 European Respiratory Journal
1 2 3 6.0%.[36] However, using this lower cutpoint would have resulted in a higher, not lower, 4 5 prevalence. Similarly, anaemia, common in TB patients, may have resulted in lower HbA1c 6 7 in TB cases, unless there was iron deficiency with or without anaemia which may result in an 8 9 10 increase in HbA1c compared with non TB participants, without a concomitant rise in glucose 11 12 indices.[37] It may be that use of HbA1c in the context of TB and/or HIV 1 is flawed, or that 13 14 the cutpoint should be revised. Further research is required to better understand factors, such 15 16 as altered red cell survival, that may influence HbA1c in TB/HIV 1 patients. 17 18 19 We noted a significant association between DM and TB, in agreement with the majority of 20 21 the published literature from other settings, with a 2.4 and 2.3 higher odds of TB in DM and 22 23 24 IGR patients respectively; and 14% of TB cases attributed to DM. Only two of these studies 25 26 were conducted in a setting with a high HIV 1 burden[21,38] with odds ratios ranging from 27 28 2.2 – 10.7. The co existence of a high prevalence of HIV 1, TB, and DM has significant 29 30 implications for optimal control of each condition, highlighting the importance of targeting 31 32 33 TB control interventions, such as intensified TB screening for DM and DM/HIV 1 patients, 34 35 to achieve TB elimination goals. 36 37 38 There was also a significant TB/IGR association (aOR 2.3), suggesting that a finding of IGR 39 40 should also prompt TB screening. Whilst the high prevalence of IGR would support the 41 42 theory of these cases being transient hyperglycaemia, the finding of the high prevalence 43 44 strongly driven by HbA1c is a conundrum as it would suggest a more sustained state of 45 46 47 hyperglyacemia. Negative outcomes have been associated with TBDM.[3] HIV 1 uninfected 48 49 TBDM patients are at higher risk of death than TBHIV patients,[39,40] but there is a paucity 50 51 of research on the association between IGR and TB outcomes to inform the need for 52 53 interventions to control blood glucose in IGR TB patients. 54 55 56 57 58 59 60 11 European Respiratory Journal Page 100 of 114
1 2 3 The association between DM and TB also varied depending on the diagnostic test used. We 4 5 found that the association was largely driven by the HbA1c test, as did Boillat Blanco et 6 7 al[38] in Tanzania. Using either FPG or OGTT alone as a DM diagnostic test did not reveal a 8 9 10 statistically significant association. Other studies from Tanzania and Guinea Bissau[19,21] 11 12 also found that IGR/DM diagnoses varied when using random blood glucose, FPG and 13 14 OGTT. Further research is required to adequately define the most appropriate test to identify 15 16 and monitor DM in TB patients, inlcuding in HIV 1 coinfected persons. 17 18 19 On stratification by HIV-1 serostatus, the association between TB and DM remained 20 21 statistically significant only in participants with HIV-1 infection. The reported effect of HIV 22 23 24 1 co infection on the association between TB and DM has been variable. A study using 25 26 random blood glucose reported reduced odds of DM in HIV 1 infected.[20] Another study 27 28 using FPG and OGTT reported similar findings of a stronger TBDM association in HIV 1 29 30 uninfected.[21] However, a recent study that used all 3 DM diagnostic tests showed a slightly 31 32 33 stronger TB/DM association among HIV 1 infected individuals, as measured by FPG and 34 35 OGTT, but not by HbA1c.[38] In our study, the significant TBDM association in HIV 1 36 37 infected participants was noted when DM defined using any of the 3 tests; largely driven by 38 39 HbA1c. There are limited data on the effect on HIV 1 infection on HbA1c. One study that 40 41 compared HbA1c values in HIV 1 infected and uninfected women found slightly lower 42 43 44 HbA1c values in HIV 1 infected women after adjustment for fasting glucose values.[41] 45 46 However, this association was nullified on multivariable analysis, with the differences 47 48 accounted for by higher red cell mean corpuscular volume in HIV 1 infected. Another study 49 50 also suggests that HbA1c may underestimate glycemia in HIV infected individuals, 51 52 53 particularly those with higher mean corpuscular volume, nucleoside reverse transcriptase 54 55 inhibitor use, and lower CD4 count.[42] 56 57 58 59 60 12 Page 101 of 114 European Respiratory Journal
1 2 3 The association between TB and DM was stronger in participants with a prior DM diagnosis, 4 5 with a 3.9 fold higher odds of TB. These participants had a higher median FPG and HbA1c 6 7 and so the stronger association may reflect the greater degree of hyperglycaemia in this 8 9 10 group. This is in line with published studies that have shown that poorly controlled DM is 11 12 more strongly associated with TB.[43] Of note, HIV 1 infected patients with a previous DM 13 14 diagnosis were at an even greater risk of TB with the aOR increasing from 2.4 to 6.3. In this 15 16 and other studies[13,20] a high percentage of DM amongst TB patients was previously 17 18 19 undiagnosed. Data from chronic disease audits across >100 primary care facilities in the 20 21 Western Cape Province of South Africa, within which Cape Town is located, have reported 22 23 that <15% of DM patients attending clinics have HbA1c levels under 7% (unpublished data 24 25 from the Western Cape Department of Health). Our results therefore strongly support the 26 27 need to prioritise improved DM management as a TB control interventions. 28 29 30 31 32 33 Strengths and Limitations 34 35 36 A significant strength of this study was the performance of 3 different tests for DM using 37 38 venous blood samples with all tests performed in an accredited national laboratory. In 39 40 addition, multivariable analysis of the association between TB and DM/IGR adjusted for the 41 42 43 most prevalent and significant known confounders including a previous history of time in 44 45 prison, mine work and household income, as well as HIV 1 infection. 46 47 48 Our study had a number of limitations. Glycaemia was measured at baseline, and it is 49 50 possible that the hyperglycaemia was transient. We attempted to limit the effect of transient 51 52 hyperglyacemia on the reported TB/DM association, by selecting non TB participants from 53 54 patients with respiratory symptoms that later resolved. Studies have shown an association 55 56 57 between hyperglyacemia at TB diagnosis (even when transient) and TB treatment failure and 58 59 60 13 European Respiratory Journal Page 102 of 114
1 2 3 death [38] highlighting the clinical importance of intervening in patients with baseline 4 5 hyperglycaemia to improve TB outcomes. Small numbers after stratification by HIV 1 status 6 7 meant we were unable to investigate the effect of ART or CD4 count. In addition, given that 8 9 10 the association between TB and DM was significant overall, and that the study was powered 11 12 to detect this association, the lack of statistical significance noted in HIV 1 uninfected 13 14 participants may be due to a reduced statistical power after stratification; and this study is not 15 16 able to exclude the possibility of an association in HIV uninfected persons. Finally, given the 17 18 19 association between DM and poorer TB outcomes, the exclusion of critically ill patients due 20 21 to their inability to give consent may have introduced selection bias with possible exclusion 22 23 of TB patients with DM. Similarly, whilst the MDR TB prevalence found in this study is 24 25 similar to the national prevalence (2.8%),[44] given the poor outcomes associated with MDR 26 27 TB, the exclusion of critically ill patients may also have resulted in a greater proportion of 28 29 30 excluded patients with MDR TB. 31 32 33 In conclusion, DM is a significant contributor to the TB burden in this high TB/HIV 1/DM 34 35 burden setting. Our study utilised international guideline approved DM tests and found a high 36 37 prevalence of DM in TB cases and a significant association between TB and DM, even in the 38 39 context of HIV 1 co infection; with greater odds of TB in participants who were previously 40 41 diagnosed with DM. These findings point to the importance of screening for DM in those 42 43 44 with TB and those with HIV 1, who are already at high risk for developing TB. A 45 46 surprisingly high prevalence of IGR was also found in this study, emphasizing the need for 47 48 further investigation into the TB outcomes and management of this group. The interpretation 49 50 of our findings is complex because of the variations in DM prevalence and TB/DM 51 52 53 association by different diagnostic tests. Existing glycaemia measures can be affected by 54 55 HIV 1, ART, anaemia, acute infection, red cell survival and iron status. This highlights the 56 57 need for more accurate markers of glycaemia that are independent of these factors in order to 58 59 60 14 Page 103 of 114 European Respiratory Journal
1 2 3 inform policy on how best to screen for DM at the time of TB diagnosis, in (often low 4 5 resource and primary care) settings with a high burden of HIV 1 and TB. 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 15 European Respiratory Journal Page 104 of 114
1 2 3 Table 1: Baseline socio-demographic, anthropometric, and co-morbidity characteristics by TB status. 4 Key: ART antiretroviral therapy; BMI body mass index 5 6 Characteristic Non-TB TB Total p-value 7 (n=438) (n=414) (n=852) 8 Age (years) N (%) 9 18-24 16 (3.7) 43 (10.4) 59 (6.9) <0.001 10 25-34 126 (28.8) 151 (36.5) 277 (32.5) 11 35-44 126 (28.8) 138 (33.3) 264 (31.0) 12 45-54 95 (21.7) 53 (12.8) 148 (17.4) 13 55 and older 75 (17.1) 29 (7.0) 104 (12.2) 14 852 15 Age (years) 16 (median (IQR)) 41 (32 – 50) 36 (30 43) 38 (31 – 47) <0.001 17 Range: 20 – 80 Range: 18 80 Range: 18 – 80 18 852 19 Sex N (%) 20 Female 224 (51.3) 175 (42.6) 399 (47.0) 0.011 21 848 22 Education level N (%) 23 Up to primary 130 (30.4) 129 (32.3) 259 (31.3) 0.268 24 Up to secondary 289 (67.7) 257 (64.3) 456 (66.0) 25 Higher education 8 (1.9) 14 (3.5) 22 (2.7) 26 827 27 Marital status N (%) Single 276 (64.8) 299 (74.8) 575 (69.6) 0.002 28 826 29 Work N (%) 30 Unemployed 235 (55.0) 213 (53.5) 448 (54.3) 0.662 31 825 32 Household size N (%) 33 0 – 2 individuals 214 (51.7) 230 (59.3) 444 (55.4) 0.031 34 2 and more individuals 200 (48.3) 158 (40.7) 358 (44.6) 35 802 36 Income categories N (%) 37 R0.0 21 (5.5) 9 (2.6) 30 (4.1) <0.001 38 R 1 – R 1 600 248 (64.8) 160 (46.1) 408 (55.9) 39 R 1 601 – R 3 200 71 (18.5) 103 (29.7) 174 (23.8) 34 (8.9) 62 (17.9) 96 (13.2) 40 R 3 201 – R 6 400 R 6 401 – R 12 800 8 (2.1) 12 (3.5) 20 (2.7) 41 R 12 801 or more 1 (0.3) 1 (0.3) 2 (0.3) 42 730 43 Binge drinking amongst drinkers N (%) 44 Yes 425 (97.0) 403 (97.3) 828 (97.8) 0.784 45 852 46 Current smoker N (%) 47 Yes 123 (28.7) 90 (22.5) 213 (25.7) 0.04 48 828 49 Prison history N (%) 50 Yes 21 (4.9) 42 (10.3) 63 (7.5) 0.003 51 837 52 Miner (past or present) N (%) 53 Yes 17 (4.0) 7 (1.7) 24 (2.9) 0.053 833 54 Health care worker N (%) 55 Yes 8 (1.9) 7 (1.7) 15 (1.8) 0.889 56 838 57 58 59 60 16 Page 105 of 114 European Respiratory Journal
1 2 3 TB contact N (%) 4 Yes 54 (12.6) 51 (12.6) 105 (12.6) 0.999 5 836 6 Previous TB N (%) 7 Yes 196 (46.0) 129 (31.9) 325 (39.2) <0.001 8 830 9 Previous DM N (%) 10 Yes 19 (4.4) 20 (4.9) 39 (4.7) 0.717 11 838 12 HIV-1 status N (%) 13 HIV-1-uninfected 160 (36.5) 121 (29.2) 281 (33.0) <0.001 14 HIV-1-infected 242 (55.3) 279 (67.4) 521 (61.2) 15 Unknown 36 (8.2) 14 (3.4) 50 (5.9) 16 852 17 ART status (among HIV-1-infected, n=521) N (%) 18 Yes 166 (68.6) 89 (31.9) 255 (48.9) <0.001 19 521 20 Previous gestational DM (among females, n=399) N (%) 21 Yes 4 (1.9) 8 (4.8) 12 (3.2) 0.105 22 379 23 Hypertension N (%) Yes 154 (35.2) 75 (18.1) 229 (26.9) <0.001 24 852 25 BMI N (%) 26 Less than 18 kg/m 2 (underweight) 27 (6.6) 41 (10.3) 68 (8.4) <0.001 27 18 – 25 kg/m 2 (normal) 224 (54.4) 277 (69.6) 501 (61.9) 28 26 – 30 kg/m 2 (overweight) 69 (16.8) 59 (14.8) 128 (15.8) 29 More than 30 kg/m 2 (obese) 92 (22.3) 21 (5.3) 113 (14.0) 30 810 31 Waist circumference N (%) 32 Wide (>94 cm males; > 88cm 144 (28.9) 53 (14.3) 167 (21.8) <0.001 33 females) 34 765 35 36 37 38 39 40 Table 2: Prevalence of DM and IGR, including previously diagnosed DM, by diagnostic test. 41 Key: FPG fasting plasma glucose; OGTT oral glucose tolerance test; HbA1c glycated haemoglobin; IGR 42 impaired glucose regulation.
43 44 Diagnosis based on: Overall Non-TB TB p-value % (95% CI) % (95% CI) % (95% CI) (chi square) 45 46 Diabetes 47 Any test 11.3 (9.3 – 13.6) 10.1 (7.6 – 13.2) 12.6 (9.7 – 16.1) 0.246 48 FPG 4.1 (3.0 – 5.7) 3.9 (2.4 – 6.2) 4.4 (2.8 6.9) 0.752 49 OGTT 3.3 (2.3 – 4.8) 3.5 (2.1 – 5.8) 3.1 (1.7 – 5.3) 0.704 50 HbA1c 8.2 (6.5 – 10.2) 6.2 (4.3 – 8.9) 10.2 (7.7 – 13. 6) 0.032 51 52 IGR (excluding DM) 53 Any test 57.3 (53.7 – 60.8) 50.0 (45.1 – 54.9) 65.2 (60.0 – 70.0) < 0.001 54 55 FPG 10.6 (8.6 – 13.1) 12.5 (9.5 – 16.1) 8.6 (6.1 – 12.2) 0.089 56 OGTT 10.6 (8.6 – 13.0) 4.9 (3.1 – 7.5) 16.9 (13.3 – 21.2) <0.001 57 HbA1c 39.5 (36.1 – 43.0) 34.1 (29.6 – 38.9) 45.4 (40.3 – 50.6) 0.002 58 59 60 17 European Respiratory Journal Page 106 of 114
1 2 3 4 5 6 Table 3: DM prevalence in TB and non-TB participants; stratified by HIV-1 status 7 8 Overall Non-TB TB p-value % (95% CI) % (95% CI) % (95% CI) 9 (N=852) (N=438) 10 (N=414) 11 12 HIV-1-uninfected 16.0 (12.2 – 20.8) 16.9 (11.8 – 23.6) 14.9 (9.5 – 22.5) 0.651 13 (n=281) (n=45) (n=27) (n=18) 14 HIV-1-infected 8.9 (6.7 – 11.6) 6.2 (3.8 – 10.1) 11.1 (7.9 – 15.4) 0.049 15 (n=521) (n=46) (n=15) (n=31) 16 HIV-1 status unknown 10.0 (4.1 – 22.4) 5.6 (1.3 – 20.8) 21.4 (6.0 – 54.0) 0.093 17 (n=50) (n=10) (n=2) (n=3) 18 19 20 21 22 Table 4: Univariable and multivariable analysis of the association between TB and DM/IGR 23 Key: FPG fasting plasma glucose; OGTT oral glucose tolerance test; HbA1c glycated haemoglobin; IGR 24 impaired glucose regulation. 25 26 27 Variable Overall HIV-1-infected HIV-1-uninfected 28 Odds Ratio; (95% CI) Odds Ratio; (95% CI) Odds Ratio; (95% CI) 29 Crude Adjusted* Crude Adjusted* Crude Adjusted* 30 DM 31 Overall 1.3 (0.8 – 2.0); 2.4 (1.3 – 4.3); 2.0 (1.0 – 3.8); 2.4 (1.1 – 5.2); 1.0 (0.5 – 1.9); 2.4 (0.9 – 6.7); 32 p=0.247 p=0.005 p=0.033 p=0.030 p= 0.995 p=0.081 33 HbA1c 1.7 (1.0 – 2.9); 2.4 (1.2 – 4.6); 2.2 (1.1 – 4.6); 2.4 (1.0 – 5.9); 1.5 (0.7 – 3.2); 2.2 (0.7 – 6.4); 34 p=0.034 p=0.012 p=0.032 p=0.05 p=0.268 p=0.160 35 FPG 1.1 (0.6 – 2.2); 2.3 (0.9 – 5.5); 1.8 (0.5 – 6.0); 2.9 (0.7 – 12.2); 1.1 (0.5 – 2.6); 1.9 (0.6 – 6.4); 36 p=0.725 p=0.068 p=0.336 p=0.148 p=0.800 p=0.295 37 OGTT 0.9 (0.4 – 1.9); 1.2 (0.5 – 3.3); 1.8 (0.5 – 6.2); 2.5 (0.6 – 10.3); 0.5 (0.2 – 1.7); 0.5 (0.1 – 3.1); 38 p=0.704 p=0.690 p=0.322 p=0.218 p=0.285 p=0.491 39 Previously 1.1 (0.6 – 2.1); 3.7 (1.5 – 9.1); 2.1 (0.5 – 8.3); 6.3 (1.3 – 30.8); 1.2 (0.6 2.6); 3.1 (0.9 10.1); 40 diagnosed p=0.717 p=0.004 p=0.283 p=0.022 p=0.652 p=0.066 41 DM only 42 IGR (excludes DM cases) 43 Overall 1.9 (1.4 2.5); 2.3 (1.6 – 3.3); 2.4 (1.7 – 3.4); 2.4 (1.5 – 3.8); 1.2 (0.7 – 2.0); 2.3 (1.1 – 4.7); 44 p<0.001 p<0.001 p<0.001 p<0.001 p=0.467 p=0.024 45 HbA1c 1.4 (1.0 – 1.8); 1.6 (1.1 – 2.3); 1.6 (1.1 2.3); 1.5 (1.0 – 2.3); 1.1 (0.7 – 1.8); 2.2 (1.1 – 4.1); 46 p=0.026 p=0.009 p=0.012 p=0.084 p=0.569 p=0.017 47 FPG 0.8 (0.5 – 1.1); 0.9 (0.5 – 1.5); 1.0 (0.6 – 1.7); 1.2 (0.6 2.2); 0.4 (0.2 0.8); 0.4 (0.1 – 1.2); 48 p=0.182 p=0.695 p=0.867 p=0.588 p=0.018 p=0.116 49 OGTT 3.0 (1.9 – 4.8); 3.9 (2.1 – 7.0); 5.0 (2.6 – 9.5); 5.4 (2.4 – 12.0); 1.4 (0.6 – 2.9); 2.8 (1.0 – 8.0); 50 p<0.001 p<0.001 p<0.001 p<0.001 p=0.432 p=0.058 51 *Adjusted for: sex, age, household size, income, hypertension (baseline), previous miner, previous prisoner, marital status, work status, 52 HIV 1 status. 53 54 55 56 57 58 59 60 18 Page 107 of 114 European Respiratory Journal
1 2 3 Acknowledgements 4 5 6 Author Contributions. TO designed the study and was Principal Investigator. TO, NL, and 7 8 RJW were involved in the design of the study. NB led participant recruitment, retention, 9 10 phenotypic characterization of all participants, and data management. RG led study co 11 12 13 ordination, and contributed to patient recruitment, follow up and collection of data. Data 14 15 analysis and interpretation were conducted by MK and TO. All authors were involved in the 16 17 drafting of the manuscript, led by TO. The final manuscript draft was approved by all 18 19 authors. TO had full access to all the data in the study and takes responsibility for the 20 21 integrity of the data and the accuracy of the data analysis. 22 23 24 Conflicts of interest : The authors have no conflicts of interests to declare. 25 26 27 Funding. This study was funded by the Clinical Infectious Disease Research Initiative, as 28 29 part of a Wellcome Trust Strategic Grant (WT 084323, 104873), a Carnegie Corporation 30 31 32 Postdoctoral Fellowship; and a Harry Crossley Senior Clinical Fellowship . Francis Crick 33 34 Institute received support from Cancer Research UK, RC (UK) and the Wellcome Trust. 35 36 RJW receives support from the National Research Foundation of South Africa (96841). The 37 38 funders played no role in the design, conduct, or analysis of the study. 39 40 41
42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 19 European Respiratory Journal Page 108 of 114
1 2 3 References 4 5 6 1. Patra J, Jha P, Rehm J, Suraweera W. Tobacco smoking, alcohol drinking, diabetes, 7 8 low body mass index and the risk of self reported symptoms of active tuberculosis: 9 10 individual participant data (IPD) meta analyses of 72,684 individuals in 14 high 11 12 13 tuberculosis burden countries. PLOS One . 2014; 9(5): e96433. 14 15 16 2. Jeon CY, Murray MB. Diabetes mellitus increases the risk of active tuberculosis: a 17 18 systematic review of 13 observational studies. PLOS Med. 2008; 5(7): e152. 19 20 21 3. Baker MA, Harries AD, Jeon CY, Hart JE, Kapur A, Lönnroth K, Ottmani 22 23 SE, Goonesekera SD, Murray MB. The impact of diabetes on tuberculosis treatment 24 25 26 outcomes: a systematic review. BMC Med . 2011; 9:81. 27 28 4. Magee MJ, Kempker RR, Kipiani M, Gandhi NR, Darchia L, Tukvadze N, Howards 29 30 31 PP, Narayan KM, Blumberg HM. Diabetes mellitus is associated with cavities, smear 32 33 grade, and multidrug resistant tuberculosis in Georgia. Int J Tuberc Lung Dis . 2015; 34 35 19(6): 37–54. 36 37 38 5. GBD 2013 Mortality and Causes of Death Collaborators. Global, regional, and national 39 40 age–sex specific all cause and cause specific mortality for 240 causes of death, 1990– 41 42 2013: a systematic analysis for the Global Burden of Disease Study 2013. Lancet . 43 44 45 2015; 385(9663): 117–71. 46 47 48 6. Raghuraman S, Vasudevan KP, Govindarajan S, Chinnakali P, Panigrahi KC. 49 50 Prevalence of diabetes mellitus among tuberculosis patients in urban Puducherry. N 51 52 Am J Med Sci. 2014; 6(1): 30–4. 53 54 55 7. Padmalatha P, Hema K. Study on prevalence of Diabetes Mellitus in Tuberculosis 56 57 patients attending a tertiary care hospital in Guntur, Andhra Pradesh. Indian J Basic 58 59 60 20 Page 109 of 114 European Respiratory Journal
1 2 3 Appl Med Res . 2014; 4AD: 494–8. 4 5 6 8. Balakrishnan S, Vijayan S, Nair S, Subramoniapillai J, Mrithyunjayan S, Wilson N, 7 8 Satyanarayana S, Dewan PK, Kumar AM, Karthickeyan D, Willis M, Harries AD, 9 10 11 Nair SA. High diabetes prevalence among tuberculosis cases in Kerala, India. PLOS 12 13 One . 2012; 7(10): e46502. 14 15 16 9. Viswanathan V, Kumpatla S, Aravindalochanan V, Rajan R, Chinnasamy C, Srinivasan 17 18 R, Selvam JM, Kapur A. Prevalence of diabetes and pre diabetes and associated risk 19 20 factors among tuberculosis patients in India. PLOS One. 2012; 7(7): 1–9. 21 22 23 10. Gupta S, Shenoy VP, Bairy I, Srinivasa H, Mukhopadhyay C. Diabetes mellitus and 24 25 HIV as co morbidities in tuberculosis patients of rural South India. J Infect Public 26 27 Health . 2011; 4(3): 140–4. 28 29 30 11. Restrepo BI, Camerlin AJ, Rahbar MH, Wang W, Restrepo MA, Zarate I, Mora 31 32 33 Guzmán F, Crespo Solis JG, Briggs J, McCormick JB, Fisher Hoch SP. Cross 34 35 sectional assessment reveals high diabetes prevalence among newly diagnosed 36 37 tuberculosis cases. Bull World Health Organ . 2011; 89(5): 352–9. 38 39 40 12. Alladin B, Mack S, Singh A, Smith B, Cummings E, Hershfield E, Mohanlall J, 41 42 Ramotar K, La Fleur C. Tuberculosis and diabetes in Guyana. Int J Infect Dis. 43 44 2011;15(12):e818–e821. 45 46 47 13. Olayinka AO, Anthonia O, Yetunde K. Prevalence of diabetes mellitus in persons with 48 49 tuberculosis in a tertiary health centre in Lagos, Nigeria. Indian J Endocrinol Metab. 50 51 52 2013;17(3):486. 53 54 55 14. Faurholt Jepsen D, Range N, PrayGod G, Jeremiah K, Faurholt Jepsen M, Aabye 56 57 MG, Changalucha J, Christensen DL, Witte DR, Andersen AB, Friis H. The role of 58 59 60 21 European Respiratory Journal Page 110 of 114
1 2 3 anthropometric and other predictors for diabetes among urban Tanzanians with 4 5 tuberculosis. Int J Tuberc Lung Dis . 2012;16(12):1680–1685. 6 7 8 15. Baldé NM, Camara A, Camara LM, Diallo MM, Kaké A, Bah Sow OY. Associated 9 10 tuberculosis and diabetes in Conakry, Guinea: Prevalence and clinical characteristics. 11 12 13 Int J Tuberc Lung Dis . 2006;10(9):1036–1040. 14 15 16. Marais RM. Diabetes mellitus in black and coloured tuberculosis patients. S Afr Med J . 16 17 18 1980;57(13):483–484. 19 20 21 17. International Diabetes Federation. Diabetes Atlas, 6th edition. 22 23 http://www.diabetesatlas.org/. Accessed December 20, 2016. 24 25 26 18. Peer N, Steyn K, Lombard C, Lambert EV, Vythilingum B, Levitt NS. Rising diabetes 27 28 prevalence among urban dwelling black South Africans. PLOS One. 29 30 31 2012;7(9):e43336. 32 33 34 19. Haraldsdottir TL, Rudolf F, Bjerregaard Andersen M, C, Joaquím LC, Stochholm 35 36 K, Gomes VF, Beck Nielsen H, Ostergaard L, Aaby P, Wejse C. Diabetes mellitus 37 38 prevalence in tuberculosis patients and the background population in Guinea Bissau: a 39 40 disease burden study from the capital Bissau. Trans R Soc Trop Med Hyg . 41 42 2015;109(6):400–407. 43 44 45 20. Kibirige D, Ssekitoleko R, Mutebi E, Worodria W. Overt diabetes mellitus among 46 47 48 newly diagnosed Ugandan tuberculosis patients: a cross sectional study. BMC Infect 49 50 Dis .2013;13:122. 51 52 53 21. Faurholt Jepsen D, Range N, PrayGod G, Jeremiah K, Faurholt Jepsen M, Aabye 54 55 MG, Changalucha J, Christensen DL, Pipper CB, Krarup H, Witte DR, Andersen 56 57 58 59 60 22 Page 111 of 114 European Respiratory Journal
1 2 3 AB, Friis H. Diabetes is a risk factor for pulmonary tuberculosis: a case control study 4 5 from Mwanza, Tanzania. PLOS One . 2011;6(8):e24215. 6 7 8 22. Bailey SL, Ayles H. Association between diabetes mellitus and active tuberculosis in 9 10 Africa and the effect of HIV. Trop Med Int Health . January 2017. [Epub ahead of print]. 11 12 13 23. Evans D. Ten years on ART – where to now? S Afr Med J . 2013;103(4):229–231. 14 15 16 24. TBfacts.org. http://www.tbfacts.org/tb statistics/. Accessed December 20, 2016. 17 18 19 20 25. Levitt NS, Peer N, Steyn K, et al. Increased risk of dysglycaemia in South Africans 21 22 with HIV; especially those on protease inhibitors. Diabetes Res Clin Pract . 23 24 2016;119:41–47. 25 26 27 26. Karamchand, S., Leisegang, R., Schomaker, M., Maartens, G., Walters, L., Hislop, M., 28 29 Dave, J.A., Levitt, N.S. and Cohen, K.. Risk factors for incident diabetes in a cohort 30 31 32 taking first line nonnucleoside reverse transcriptase inhibitor based antiretroviral 33 34 therapy. Medicine. 2016;95 (9). 35 36 37 27. Statistics South Africa. Mortality and causes of death in South Africa, 2010: Findings 38 39 from death notification. 40 41 http://www.statssa.gov.za/publications/P03093/P030932013.pdf. Accessed December 42 43 44 20, 2016. 45 46 47 28. World Health Organisation (2014). Global Tuberculosis Report 2014. 48 49 http://apps.who.int/iris/bitstream/10665/137094/1/9789241564809_eng.pdf. Accessed 50 51 December 20, 2016. 52 53 54 55 29. National Tuberculosis Management Guidelines 2014. Department of Health, Republic 56 57 of South Africa. http://www.nicd.ac.za/assets/files/Acrobat%20Document2.pdf. 58 59 60 23 European Respiratory Journal Page 112 of 114
1 2 3 Accessed December 20, 2016. 4 5 6 30. World Health Organisation (2006). Definition and Diagnosis of Diabetes Mellitus and 7 8 Intermediate Hyperglycemia. 2006. 9 10 11 http://www.who.int/diabetes/publications/diagnosis_diabetes2006/en/index.html 12 13 (accessed Feb 3, 2015). 14 15 16 31. American Diabetes Association. Standards of Medical Care in Diabetes 2013. 17 18 Diabetes Care . 2013;36(Supplement_1):S11–S66. 19 20 21 32. ROCHE cobas c311 user manual. 22 23 24 http://www.roche.hu/content/dam/roche_hungary/hu_HU/docs/cobas_c_311_en.pdf. 25 26 Accessed December 20, 2016. 27 28 29 33. Alberti KGMM, Zimmet P, Shaw J. Metabolic syndrome a new world wide definition. 30 31 A Consensus Statement from the International Diabetes Federation. Diabet Med . 32 33 34 2006;23(5):469–480. 35 36 37 34. World Health Organization. STEPS Instrument for Chronic Disease Risk Factor 38 39 Surveillance 2009. http://www.who.int/chp/steps/instrument/STEPS_Instrument_V3.1. 40 41 Accessed December 20, 2016. 42 43 44 35. Greenland S, Robins JM. Conceptual problems in the definition and interpretation of 45 46 47 attributable fractions. Am J Epidemiol . 1988;128(6):1185–1197. 48 49 50 36. Hird TR, Pirie FJ, Esterhuizen TM, O'Leary B, McCarthy MI, Young EH, Sandhu 51 52 MS, Motala AA. Burden of Diabetes and First Evidence for the Utility of HbA1c for 53 54 Diagnosis and Detection of Diabetes in Urban Black South Africans: The Durban 55 56 Diabetes Study. PLoS ONE . 2016;11(8):e0161966. 57 58 59 60 24 Page 113 of 114 European Respiratory Journal
1 2 3 37. English E, Idris I, Smith G, Dhatariya K, Kilpatrick ES, John WG. The effect of 4 5 anaemia and abnormalities of erythrocyte indices on HbA1c analysis: a systematic 6 7 review. Diabetologia . 2015;58(7):1409–1421. 8 9 10 38. Boillat Blanco N, Ramaiya KL, Mganga M, Minja LT, Bovet P, Schindler C, Von 11 12 13 Eckardstein A, Gagneux S, Daubenberger C, Reither K, Probst Hensch N. Transient 14 15 hyperglycemia in patients with tuberculosis in Tanzania: Implications for diabetes 16 17 screening algorithms. J Infect Dis . 2016;213(7):1163–1172. 18 19 20 39. Faurholt Jepsen D, Range N, Praygod G, Jeremiah K, Faurholt Jepsen M, Aabye 21 22 MG, Changalucha J, Christensen DL, Grewal HM, Martinussen T, Krarup H, Witte 23 24 DR, Andersen AB, Friis H. Diabetes is a strong predictor of mortality during 25 26 27 tuberculosis treatment: a prospective cohort study among tuberculosis patients from 28 29 Mwanza, Tanzania. Trop Med Int Heal. 2013;18(7):822–829. 30 31 32 40. Workneh MH, Bjune GA, Yimer SA. Diabetes mellitus is associated with increased 33 34 mortality during tuberculosis treatment: a prospective cohort study among tuberculosis 35 36 patients in South Eastern Amahra Region, Ethiopia. Infect Dis Poverty. 2016;5(1):22. 37 38 39 41. Glesby MJ, Hoover DR, Shi Q, Danoff A, Howard A, Tien P, Merenstein D, Cohen 40 41 M, Golub E, Dehovitz J, Nowicki M, Anastos K. Glycated haemoglobin in diabetic 42 43 44 women with and without HIV infection: data from the Women's Interagency HIV 45 46 Study. Antivir Ther . 2010;15(4):571–577. 47 48 49 42. Monroe, A.K., Glesby, M.J. and Brown, T.T. Diagnosing and managing diabetes in 50 51 HIV infected patients: current concepts. Clin Infect Dis. 2015;60(3):453 462. 52 53 54 55 56 57 58 59 60 25 European Respiratory Journal Page 114 of 114
1 2 3 43. Yoon YS, Jung J W, Jeon EJ, Seo H, Ryu YJ, Yim JJ, Kim YH, Lee BH, Park YB, Lee 4 5 BJ, Kang H, Choi JC. The effect of diabetes control status on treatment response in 6 7 pulmonary tuberculosis: a prospective study. Thorax . 2016 Aug 23. 8 9 10 44. National Institute for Communicable Diseases. South African tuberculosis drug 11 12 13 resistance survey 2012 14. http://www.nicd.ac.za/assets/files/K 14 15 12750%20NICD%20National%20Survey%20Report_Dev_V11 LR.pdf . Accessed 31 16 17 March 2017. 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 26