You are on page 1of 7

Tuberculosis 116 (2019) S59–S65

Contents lists available at ScienceDirect

Tuberculosis
journal homepage: www.elsevier.com/locate/tube

Diabetic trends and associated mortality in tuberculosis patients in Texas, a T


large population-based analysis
Duc T. Nguyen, Edward A. Graviss∗
Houston Methodist Research Institute, 6670 Bertner Ave, Houston, TX 77030, USA

ARTICLE INFO ABSTRACT

Keywords: Background: Tuberculosis (TB) and diabetes mellitus (DM) comorbidity (TB-DM) is a major public health
Tuberculosis challenge worldwide. This analysis aimed to determine the risk factors and trends associated with TB-DM
Diabetes morbidity and mortality.
Risk score Methods: Risk factors for TB-DM morbidity and mortality were identified by logistic regression using de-iden-
Mortality
tified surveillance data of all TB patients from Texas, USA. reported between 01/2010–12/2016. Non-parametric
TB-Diabetes
testing was used for the morbidity and mortality trends.
Trend
Prevalence Results: From 2010 to 2016, 1400/9002 (15.6%) TB patients were diabetic with an annual prevalence increase
from 12.5% to 18.7% (p = 0.005). Reported TB-DM patients had a higher mortality (10.3%) than non-DM
patients (7.6%, p = 0.001) with nearly a 3-fold increase in the odds of death (overall and during treatment).
Older age, being Hispanic, chronic kidney failure, pulmonary cavitation and positive TB culture or smear were
associated with TB-DM. Age ≥45, US-birth, resident of long-term care facility, injecting-drug user, chronic
kidney disease, TB meningitis, abnormal chest radiograph, non-conversion of culture, and HIV(+) were in-
dependently associated with a higher mortality.
Conclusions: TB-DM is an increasing public health problem in Texas with significantly high mortality. Risk
factors for mortality determined by multivariate modeling will provide a foundation for the development of
more effective strategies for TB-DM management.

1. Introduction sputum conversion, and treatment failure or death [6–9].


In the United States (US), TB-DM comorbidity remains a major
Tuberculosis (TB) is one of the leading causes of death due to an public health problem. In 2016, 16.4% (1524) of TB patients also re-
infectious diseases with an estimation of 10.4 million new TB cases and ported having diabetes, nearly double the DM prevalence of 9.4% in the
nearly 1.7 million deaths worldwide in 2016 [1]. Although the TB in- US general population [10]. One of the states with the highest TB
cidence declined by about 1.4% per year between 2000 and 2016, this prevalence in the US, Texas also has a DM prevalence of 11.2%, which
decreasing rate is still far below the goal of 4.5% per year by 2020 set is higher than the national average [11]. The prevalence of TB-DM
by the World Health Organization (WHO)'s End TB Strategy [1]. Among comorbidity is also alarmingly high (up to 39%) in certain Texas sub-
the potential challenges that prevent TB programs from achieving the populations [12]. However, information regarding the association be-
goal of eliminating TB by the end of this century, the global steady tween TB-DM comorbidity and mortality were inconsistently reflected
increase of obesity and diabetes (DM) cases may play a major role with in literature. For different populations in the US, some analyses re-
an estimated frequency of 424.9 million DM patients in 2017 and ported no significant difference in the mortality or even lower mortality
projected to be 628.6 million in 2045 [2,3]. Compared with non-DM during treatment in TB-DM patients compared with non-diabetic TB
patients, diabetic patients may have up to 4-fold increase in the risk of patients [13,14]. Little in-depth and updated state-wide information is
active TB [4]. Meta-analyses have suggested a high prevalence of DM available regarding the trends of TB-DM and the risk factors associated
among TB patients, which ranges from 1.9% to 45% [5]. Many studies with the morbidity and mortality in TB-DM individuals in settings of
have also reported significantly higher adverse TB treatment outcomes low TB prevalence. Additionally, the lack of a standardized prognostic
in TB-DM patients such as severe clinical manifestations, delayed system to predict the mortality risk during TB treatment remains a


Corresponding author. Department of Pathology and Genomic Medicine, Houston Methodist Research Institute, Mail Station: R6-414, 6670 Bertner, Houston, TX
77030, USA.
E-mail address: eagraviss@houstonmethodist.org (E.A. Graviss).

https://doi.org/10.1016/j.tube.2019.04.011
Accepted 9 October 2018
1472-9792/ © 2019 Elsevier Ltd. All rights reserved.
D.T. Nguyen and E.A. Graviss Tuberculosis 116 (2019) S59–S65

challenge for health care providers in managing TB-DM patients.


Therefore, the present study aimed to determine the trends and risk
factors associated with the morbidity and mortality in TB-DM patients
using state-wide surveillance data.

2. Methods

De-identified surveillance data from the Centers for Disease Control


(CDC)'s Tuberculosis Genotyping Information Management System
(TBGIMS) of all TB confirmed patients from the state of Texas reported
between 01/2010 through 12/2016 were included in this retrospective
Fig. 1. Flowchart of the study population.
data analysis. According to the CDC's reference guide, TB-DM in the
TBGIMS data set was defined as TB patients who had a documented
diagnosis of diabetes mellitus (type I or type II) either before or at the shown).
time of TB diagnosis. Diabetes diagnosis was obtained from the patient's The prevalence of diabetes in TB patients in Texas has consistently
medical chart or from a reliable source such as the health care provider. increased from 12.5% in 2010 to 18.7% in 2016 with a positive overall
Undocumented reporting from patient or persons other than health care trend z = 4.50, p < 0.001 (Fig. 3). Compared with non-diabetic TB
providers were not accepted [15].The conversion of specimen TB cul- patients, TB-DM patients were significantly older with 78% of patients
tures was defined as a patient who had an initial positive TB culture age ≥45 years (versus 44.3%, p < 0.001), more likely to be Hispanic
that converted to a documented negative TB culture without converting ethnicity (adjusted odds ratio [aOR] 2.61, 95% confidence interval (CI)
back to positive TB culture during the entire treatment course [15,16]. 2.23, 3.05, p < 0.001) and diagnosed with chronic kidney failure (aOR
Patient characteristics were reported as frequency and proportion. 5.20, 95% CI 3.34, 8.08, p < 0.001). TB-DM patients were more likely
The difference between diabetic and non-diabetic patients was com- to have pulmonary cavitation (aOR 1.61, 95% CI 1.39, 1.87,
pared using the Chi-squared or Fisher's exact tests as appropriate. p < 0.001) and a positive TB culture, nucleic acid amplification (NAA)
Spatial distribution of TB-DM cases was presented by Stata's Geographic or acid-fast bacilli (AFB) smear testing (aOR 1.77, 95% CI 1.43, 2.19,
Information Systems (GIS) mapping function. A non-parametric trend p < 0.001) than non-diabetic TB patients. Non-diabetic TB patients
test was used for the morbidity and overall mortality trends. had a higher proportion of homelessness, being an inmate of a correc-
Logistic regression analyses were used to determine prognostic tional facility, non-injecting drug user or excessive alcohol consumption
factors associated with (1) overall mortality (death at diagnosis plus compared with TB-DM patients (Table 1).
death during TB treatment) and (2) mortality during TB treatment. All From 2010 through 2016, non-diabetic TB patients had a significant
patients were used in the overall mortality analyses while only reported decrease in the mortality with an overall negative trend z = −3.33,
patients who either died during treatment or completed the treatment p = 0.001. Meanwhile, the mortality was not significantly different
were used in the analyses regarding mortality during TB treatment. over time in TB-DM patients (z = 0.50, p = 0.62) and this mortality
Variables with a p-value < 0.2 in the univariate analysis or considered was always significantly higher than that of non-diabetic TB patients
as clinically significant including age and race/ethnicity were further (z = 3.05, p = 0.002) (Fig. 4). Findings regarding the crude and ad-
evaluated in the multiple logistic regression modeling. Variable selec- justed associations between characteristics and overall mortality are
tion for the multiple logistic regression was conducted using the reported in Table 2. The multiple logistic regression modeling indicated
Bayesian model averaging (BMA) method [17]. Model discrimination that TB-DM patients had nearly a 3-fold increase in the odds for overall
was assessed using the area under the receiver operating characteristic death in TB-DM compared with non-diabetic TB patients (aOR 2.75,
(ROC) curve (AUC). Model calibration was performed using the 95% CI 1.40, 5.39, p = 0.003). Age ≥45 years, being born in the US,
Hosmer-Lemeshow goodness-of-fit test. Stratification by TB-DM status being a resident of a long-term care facility, injecting drug user (IDU),
was done to identify if there is any difference in the risk factors af- having chronic kidney failure, abnormalities on chest radiograph con-
fecting the patient mortality. All analyses were performed with Stata sistent with tuberculosis (TB-CXR), TB meningitis, miliary TB, being an
MP15.1 (StataCorp LLC, College Station, TX). A p value < 0.05 was organ transplant recipient, having a non-conversion of specimen TB
considered statistically significant. cultures, and having a positive or unknown HIV status were in-
dependently associated with a higher overall mortality (Table 2). After
3. Results stratification by the diabetic status, age ≥45, being US-born, TB me-
ningitis, TB-CXR and positive or unknown HIV status remained sig-
In 9002 confirmed TB patients in Texas who were reported to the nificant in the association with a higher overall mortality (Table 3).
National TB Surveillance System (NTSS)'s TBGIMS database from 01/ The analysis for the mortality during treatment was conducted on
2010 to 12/2016, 1400 (15.6%) were also diagnosed as having dia- 7740/9002 (86%) patients who received TB treatment and had an
betes. In the TB-DM group, 32 (2.3%) patients died at time of diagnosis outcome as either “died during treatment” (n = 530, 6.9%) or “com-
and 112 (8.0%) died during TB treatment. In non-DM TB patients, 157 pleted the treatment” (n = 7210, 93.2%). TB-DM was significantly as-
(2.1%) died at time of diagnosis and 418 died during TB treatment sociated with the mortality during TB treatment in multivariate analysis
(5.5%). The overall mortality in TB-DM and non-DM TB patients was (aOR 2.43; 95% CI 1.13, 5.23; p = 0.02). Age ≥45, being US-born,
10.3% and 7.6%, respectively (Fig. 1). resident of long-term care facility, organ transplant recipient, having
Among 1400 TB-DM patients, Hispanic patients were dominant chronic kidney failure, TB meningitis, miliary TB, TB-CXR, cultures not
(67.6%, Table 1) over other ethnic groups such as Asian (13.8%), Black converted from positive to negative, and positive or unknown HIV
(10.7%), white (7.1%) and Other (0.7%) (data not shown). GIS map- status were independently associated with mortality during TB treat-
ping on the distribution of TB-DM cases suggested a higher TB-DM ment (Table 4).
burden in urban counties and in counties along the Mexican border
(Fig. 2). The ten counties with the highest number of TB-DM patients 4. Discussion
were Harris (345/2141, 16.1%), Dallas (153/1290, 11.9%), Hidalgo
(143/509, 28.1%), Bexar (88/565, 15.6%), Tarrant (67/538, 12.5%), Using the population-based surveillance data of the state of Texas
Cameron (65/383, 17.0%), El Paso (52/295, 17.6%), Travis (52/356, during a 7-year period (2010–2016), our analyses found a significantly
14.6%), Webb (49/241, 20.3%) and Starr (14/53, 26.4%) (data not increasing trend of TB-DM comorbidity in this state. This finding is even

S60
D.T. Nguyen and E.A. Graviss Tuberculosis 116 (2019) S59–S65

Table 1
Characteristics associated with diabetes in TB patients in Texas, 2010–2016.
Characteristics Total (N = 9002) Non-diabetes Diabetes Unadjusted OR (95% Adjusted OR (95% Adjusted p-value
(n = 7602) (n = 1400) CI) CI)

Age (years)
0–14 580 (6.4) 578 (7.6) 2 (0.1) 0.21 (0.05, 0.91) 0.47 (0.10, 2.17) 0.33
15–24 1041 (11.6) 1024 (13.5) 17 (1.2) (Reference) (Reference)
25–44 2935 (32.6) 2646 (34.8) 289 (20.6) 6.58 (4.01, 10.79) 11.71 (6.15, 22.27) < 0.001
45–64 2984 (33.1) 2265 (29.8) 719 (51.4) 19.12 (11.75, 31.10) 34.41 (18.21, 65.03) < 0.001
≥65 1462 (16.2) 1089 (14.3) 373 (26.6) 20.63 (12.60, 33.80) 27.02 (14.18, 51.48) < 0.001
Male gender 5872 (65.3) 4966 (65.4) 906 (64.7) 0.97 (0.86, 1.10)
Hispanic 4701 (52.2) 3754 (49.4) 947 (67.6) 2.14 (1.90, 2.42) 2.61 (2.23, 3.05) < 0.001
US-born 3952 (43.9) 3441 (45.3) 511 (36.5) 0.70 (0.62, 0.78) 1.02 (0.87, 1.20) 0.79
Homeless 502 (5.6) 457 (6.0) 45 (3.2) 0.52 (0.38, 0.71) 0.48 (0.32, 0.71) < 0.001
Inmate of a correctional facility 928 (11.5) 885 (12.9) 43 (3.6) 0.25 (0.18, 0.34) 0.30 (0.21, 0.42) < 0.001
Resident of long-term care facility 112 (1.2) 87 (1.1) 25 (1.8) 1.57 (1.00, 2.46)
IDU 221 (2.5) 199 (2.6) 22 (1.6) 0.59 (0.38, 0.93)
Non-IDU 886 (9.8) 808 (10.6) 78 (5.6) 0.50 (0.39, 0.63) 0.64 (0.47, 0.88) 0.01
Excessive alcohol use 1584 (17.6) 1375 (18.1) 209 (14.9) 0.80 (0.68, 0.93) 0.62 (0.51, 0.76) < 0.001
Organ transplant recipient 31 (0.3) 23 (0.3) 8 (0.6) 1.89 (0.85, 4.24)
Chronic kidney failure 131 (1.5) 63 (0.8) 68 (4.9) 6.11 (4.32, 8.65) 5.20 (3.34, 8.08) < 0.001
Immunosuppression (medical condition or 180 (2.0) 141 (1.9) 39 (2.8) 1.52 (1.06, 2.17)
medication)
Pulmonary TB 7727 (85.8) 6464 (85.0) 1263 (90.2) 1.62 (1.35, 1.96)
TB site, cervical lymph 435 (4.8) 413 (5.4) 22 (1.6) 0.28 (0.18, 0.43)
TB meningitis 175 (1.9) 161 (2.1) 14 (1.0) 0.47 (0.27, 0.81)
Miliary TB 238 (2.7) 200 (2.7) 38 (2.7) 1.02 (0.72, 1.46)
TB-CXR 7501 (83.3) 6273 (82.5) 1228 (87.7) 1.66 (1.35, 2.04)
Cavitation on CXR 2462 (32.9) 1894 (30.2) 568 (46.3) 1.99 (1.75, 2.25) 1.61 (1.39, 1.87) < 0.001
AFB smear (+) 3488 (46.7) 2728 (43.8) 760 (61.2) 2.03 (1.79, 2.29)
Mtb culture (+) 5090 (68.5) 4132 (66.7) 958 (77.6) 1.73 (1.50, 1.99)
Culture not converted (from positive to 923 (10.3) 761 (10.0) 162 (11.6) 0.90 (0.75, 1.09)
negative)
Positive culture, NAA, or AFB smear 6938 (77.1) 5722 (75.3) 1216 (86.9) 2.17 (1.84, 2.56) 1.77 (1.43, 2.19) < 0.001
HIV status
Negative 7192 (79.9) 5996 (78.9) 1196 (85.4) (Reference) (Reference)
Positive 570 (6.3) 543 (7.1) 27 (1.9) 0.25 (0.17, 0.37) 0.33 (0.21, 0.51) < 0.001
Unknown 1240 (13.8) 1063 (14.0) 177 (12.6) 0.83 (0.70, 0.99) 0.83 (0.67, 1.04) 0.10
MDR-TB 65 (0.7) 53 (0.7) 12 (0.9) 1.55 (0.60, 3.97) – –
East Asian 1078 (17.4) 934 (18.3) 144 (13.1) 1.01 (0.78, 1.30) – –

AUC = 0.80. Values are in number (%) unless otherwise indicated; AUC, area under the receiver operating characteristic (ROC) curve; CXR, chest radiograph; TB-
CXR, abnormalities on CXR consistent with tuberculosis; IDU, injecting-drug user; MDR-TB, Multi-drug resistant TB; NAA, Nucleic Acid Amplification; AFB, acid-fast
bacilli.

more concerning when the mortality trend in TB-DM patients was have been conducted in the U.S with inconsistent outcomes [13,25]. In
consistently high compared with non-DM TB patients without a sign of a retrospective cohort study on 139 adult TB patients, Oursler et al.
decrease over the past 7 years despite the fact that TB control strategies reported that TB-DM patients had more than 6-fold increase in the risk
have been deployed and well operated by local and state TB control for death during treatment (adjusted hazard ratio HR = 6.7; 95% CI
programs [18]. Our results from the multivariable modeling suggested 1.6, 29.3; p = 0.01) [25]. Meanwhile, Magee et al. found no significant
that TB-DM patients have a poorer outcome with nearly a 3-fold in- association between TB-DM and all-cause mortality in the multivariable
crease in the odds for both overall mortality and mortality during modeling although they observed a higher mortality during treatment
treatment compared with non-diabetic TB patients. These findings in the univariate analysis for TB-DM patients in a larger study using
confirm the enormous challenge in managing the combination of TB data of adult TB patients reported from the state of Georgia from Jan-
and DM, and therefore, urgently suggests for more effective strategies in uary 2009 to September 2012 [13]. Using a much larger sample size
managing TB, diabetes, and TB-DM patients to reduce the diseases' and including variables that are routinely collected by most TB pro-
burden and improve patient survival. Early identification of TB-DM grams, we found a significant difference in the mortality (both overall
patients who are at high risk of mortality and prompt provision of and during treatment) in TB-DM patients compared with non-DM TB
appropriate medical support are the most important strategies [19]. patients. Having a robust area under the ROC curves, which ranged
Multivariate analysis with stratification by TB-DM status suggested from 0.86 (for overall mortality) to 0.89 (for mortality during treat-
with an exception of IDU, there was no different in the risk factors. Risk ment), our models have enough power to identify additional risk factors
factors associated with a higher mortality identified by multivariable associated with the TB-DM morbidity and mortality, which were not
modeling (such as older age, being US-born, IDU, chronic kidney considered significant in the analyses by Magee et al. such as TB me-
failure, TB meningitis, military TB, TB-CXR, non-conversion of culture, ningitis, organ transplant recipient or non-conversion of the culture
and positive or unknown HIV status) could be used in the development results [13].
of a specific prognostic scoring system for TB-DM patients as suggested Our multivariable modeling also identified subgroups that are more
by recent mortality risk score modeling for TB and TB/HIV [19,20]. In likely to be associated DM development. Although screening for DM
the TB-DM population, TB meningitis, chronic kidney failure, and non- have been done routinely for all TB patients, more attention should be
conversion of sputum were the strongest risk factors for mortality. extensively targeted in TB patients of older age, Hispanic ethnicity, or
These findings are consistent with the current literature [21–23]. having chronic kidney failure, pulmonary cavitation and positive cul-
Although the association with higher mortality of TB-DM patients ture, smear or NAA.
have been recognized in high TB burden settings [5,6,24], few studies In our study, Hispanic patients had more than a 2.6 times higher

S61
D.T. Nguyen and E.A. Graviss Tuberculosis 116 (2019) S59–S65

Fig. 2. Spatial distribution of TB-DM cases in Texas, 2010–2016 TB- DM, tuberculosis-diabetes comorbidity.

Fig. 3. Trend of diabetes prevalence in TB confirmed patients in Texas, Fig. 4. Trend of mortality in TB confirmed patients in Texas, 2010–2016.
2010–2016.
(aOR 2.95; 95% CI 2.61, 3.33) compared with non-Hispanic whites
odds of having comorbid diabetes compared with patients of other [26]. The National Health Interview Survey data from 200 to 2005
ethnic groups. This finding can be explained by the dominance of indicated that US-born Hispanic patients had more twice the odds of
Hispanic patients (52.2%) in the overall TB case count. Additionally, TB-DM [27].Our finding is also consistent with the recent report from
the Hispanic ethnic group accounts for 40% of the general population in the CDC which reported that the prevalence of TB cases in the Hispanic
Texas (second after non-Hispanic whites [41.9%] and much higher than population was eight time higher than that in non-Hispanic whites [28].
Blacks [11.5%]) [11]. The dominant of Hispanics, especially Mexican Being born in countries with high TB prevalence, having poor living
Americans in the prevalence of TB and diabetes has been observed for a conditions and having obesity and diabetes are probably major con-
long time with nearly a 3-fold increase in the odds of having TB-DM tributors to this fact [28].

S62
D.T. Nguyen and E.A. Graviss Tuberculosis 116 (2019) S59–S65

Table 2
Characteristics associated with overall mortality in TB patients in Texas, 2010–2016.
Characteristics Alive or Unknown Dead, overall Unadjusted OR Adjusted OR Adjusted

(n = 8291) (n = 712) (95% CI) (95% CI) p-value

Diabetes 1256 (15.2) 144 (20.0) 1.40 (1.16, 1.70) 2.75 (1.40, 5.39) 0.003
Age ≥45 (years) 3830 (46.2) 616 (85.7) 6.95 (5.62, 8.60) 2.92 (1.50, 5.67) 0.002
Male gender 5369 (64.9) 503 (70.0) 1.26 (1.07, 1.49)
White 914 (11.0) 134 (18.6) 1.85 (1.51, 2.26)
US-born 3533 (42.7) 419 (58.3) 1.88 (1.61, 2.19) 1.51 (1.26, 1.82) < 0.001
Homeless 437 (5.3) 65 (9.1) 1.79 (1.36, 2.35)
Inmate in a correctional institution 905 (12.3) 23 (3.5) 0.26 (0.17, 0.39)
Resident of long-term care facility 77 (0.9) 35 (4.9) 5.45 (3.63, 8.19) 2.39 (1.45, 3.93) 0.001
IDU 189 (2.3) 32 (4.5) 1.99 (1.36, 2.92) 1.73 (1.08, 2.78) 0.02
Non-IDU 810 (9.8) 76 (10.6) 1.09 (0.85, 1.40)
Excessive alcohol use 1430 (17.3) 154 (21.4) 1.31 (1.09, 1.58)
Organ transplant recipient 22 (0.3) 9 (1.3) 4.76 (2.18, 10.38) 1.66 (0.59, 4.68) 0.34
Chronic kidney failure 81 (1.0) 50 (7.0) 7.57 (5.27, 10.86) 4.43 (2.84, 6.92) < 0.001
Immunosuppression (medical condition or medication) 140 (1.7) 40 (5.6) 3.43 (2.39, 4.91)
Pulmonary TB 7100 (85.7) 627 (87.2) 1.14 (0.90, 1.43)
TB site, cervical lymph 428 (5.2) 7 (1.0) 0.18 (0.09, 0.38)
TB meningitis 132 (1.6) 43 (6.0) 3.93 (2.76, 5.59) 6.42 (4.11, 10.02) < 0.001
Miliary TB 191 (2.3) 47 (7.0) 3.12 (2.25, 4.34) 3.03 (2.02, 4.54) < 0.001
TB-CXR 6897 (83.3) 604 (84.0) 1.87 (1.38, 2.52) 2.43 (1.72, 3.42) < 0.001
Cavitation on CXR 2290 (33.2) 172 (28.5) 0.80 (0.67, 0.96)
AFB smear (+) 3238 (46.1) 250 (56.3) 1.51 (1.24, 1.83)
Mtb culture (+) 4726 (67.6) 364 (82.5) 2.26 (1.76, 2.91)
Culture not converted (from positive to negative) 713 (8.6) 210 (29.2) 12.54 (9.71, 16.20) 13.55 (10.28, 17.86) < 0.001
Positive culture, NAA, or AFB smear 6269 (75.7) 669 (93.0) 4.30 (3.21, 5.75)
HIV status
Negative 6805 (82.2) 387 (53.8) (Reference) (Reference)
Positive 500 (6.0) 70 (9.7) 2.46 (1.88, 3.23) 2.07 (1.50, 2.87) < 0.001
Unknown 978 (11.8) 262 (36.4) 4.71 (3.97, 5.59) 3.46 (2.82, 4.25) < 0.001
MDR-TB 59 (0.7) 6 (0.8) 1.00 (0.43, 2.33)
East Asian 980 (17.4) 98 (17.3) 0.99 (0.79, 1.25)
Diabetes × age (≥/ < 45) (interaction term) – – 3.12 (1.53, 6.35) 0.002

AUC = 0.86. Values are in number (%) unless otherwise indicated; AUC, area under the receiver operating characteristic (ROC) curve; CXR, chest radiograph; TB-
CXR, abnormalities on CXR consistent with tuberculosis; IDU, injecting-drug user; MDR-TB, Multi-drug resistant TB; NAA, Nucleic Acid Amplification; AFB, acid-fast
bacilli.

On the GIS mapping of TB-DM distribution across Texas counties, study [12].
Harris County and Dallas County had the highest frequencies of TB-DM Our study has several limitations. First, given the nature of sur-
cases. This finding is not surprising as these two counties are also the veillance data from the National Tuberculosis Surveillance System
most populous counties in the state and have the highest number of TB where there was some lag time so that the data from local health TB
cases [29]. Having a high proportion of the Hispanic population, the programs could be submitted, validated and then uploaded into the
counties along the Mexican border were also among the counties with TBGIMS database, the data used in our analysis might reflect under-
the highest frequencies of TB-DM patients as described in a previous reporting and as the result, our findings might underestimate the actual

Table 3
Adjusted association of characteristics and overall mortality in TB patients in Texas 2010–2016, stratified by diabetes status.
Non-diabetic patients (n = 7261) TB-DM patients (n = 1299)

Adjusted OR Adjusted Adjusted OR Adjusted

(95% CI) p-value (95% CI) p-value

Age ≥45 (years) 8.96 (6.92, 11.60) < 0.001 2.97 (1.49, 5.93) 0.002
US-born 1.49 (1.21, 1.82) < 0.001 1.65 (1.09, 2.51) 0.02
Resident of long-term care facility 2.81 (1.59, 4.96) < 0.001 1.43 (0.52, 3.97) 0.49
IDU 1.52 (0.89, 2.57) 0.12 3.20 (1.02, 10.03) 0.046
Organ transplant recipient 2.00 (0.63, 6.37) 0.24 1.06 (0.10, 11.29) 0.96
Chronic kidney failure 4.06 (2.12, 7.81) < 0.001 4.99 (2.64, 9.44) < 0.001
TB meningitis 6.85 (4.28, 10.98) < 0.001 4.34 (1.12, 16.90) 0.03
Miliary TB 2.92 (1.86, 4.59) < 0.001 3.32 (1.30, 8.44) 0.01
TB-CXR 2.33 (1.60, 3.40) < 0.001 2.74 (1.19, 6.30) 0.02
Culture not converted (from positive to negative) 11.64 (8.56, 15.83) < 0.001 24.21 (12.53, 46.75) < 0.001
HIV status
Negative (Reference) (Reference)
Positive 1.94 (1.37, 2.74) < 0.001 3.24 (1.19, 8.79) 0.02
Unknown 3.35 (2.66, 4.21) < 0.001 4.05 (2.52, 6.51) < 0.001

AUC = 0.86 for non-diabetic patients and AUC = 0.87 for TB-DM patients. AUC, area under the receiver operating characteristic (ROC) curve; CXR: chest radiograph;
TB-CXR, abnormalities on CXR consistent with tuberculosis; IDU, injecting-drug user; MDR-TB, Multi-drug resistant TB; NAA, Nucleic Acid Amplification; Multiple
logistic regression modeling was conducted on patients having complete data for all the variables in the model.

S63
D.T. Nguyen and E.A. Graviss Tuberculosis 116 (2019) S59–S65

Table 4
Characteristics associated with mortality during TB treatment in Texas, 2010–2016.
Characteristics Total (N = 7740) Completed treatment Died during Unadjusted OR Adjusted OR Adjusted
treatment

(n = 7210) (n = 530) (95% CI) (95% CI) p-value

Diabetes 1227 (15.9) 1115 (15.5) 112 (21.1) 1.46 (1.18, 1.82) 2.43 (1.13, 5.23) 0.02
Age ≥45 (years) 3801 (49.1) 3345 (46.4) 456 (86.0) 7.12 (5.55, 9.14) 2.46 (1.16, 5.23) 0.02
Male gender 4967 (64.2) 4593 (63.7) 374 (70.6) 1.37 (1.13, 1.66)
White 906 (11.7) 805 (11.2) 101 (19.1) 1.87 (1.49, 2.35)
US-born 3450 (44.6) 3153 (43.7) 297 (56.0) 1.64 (1.37, 1.96) 1.28 (1.04, 1.59) 0.02
Homeless 409 (5.3) 353 (4.9) 56 (10.6) 2.30 (1.71, 3.09)
Inmate in a correctional institution 643 (9.2) 627 (9.7) 16 (3.3) 0.32 (0.19, 0.53)
Resident of long-term care facility 98 (1.3) 68 (0.9) 30 (5.7) 6.30 (4.06, 9.78) 3.62 (2.06, 6.36) < 0.001
IDU 179 (2.3) 158 (2.2) 21 (4.0) 1.84 (1.16, 2.93) 1.64 (0.91, 2.94) 0.10
Non-IDU 736 (9.5) 675 (9.4) 61 (11.5) 1.26 (0.95, 1.66)
Excessive alcohol use 1348 (17.4) 1232 (17.1) 116 (21.9) 1.36 (1.10, 1.69)
Organ transplant recipient 27 (0.3) 18 (0.2) 9 (1.7) 6.90 (3.09, 15.44) 2.17 (0.73, 6.43) 0.16
Chronic kidney failure 109 (1.4) 68 (0.9) 41 (7.7) 8.81 (5.91, 13.11) 5.45 (3.29, 9.04) < 0.001
Immunosuppression (medical condition or 161 (2.1) 132 (1.8) 29 (5.5) 3.10 (2.06, 4.69)
medication)
Pulmonary TB 6657 (86.0) 6182 (85.7) 475 (89.6) 1.44 (1.08, 1.91)
TB site, cervical lymph 391 (5.1) 385 (5.3) 6 (1.1) 0.20 (0.09, 0.46)
TB meningitis 131 (1.7) 105 (1.5) 26 (4.9) 3.49 (2.25, 5.41) 6.22 (3.56, 10.89) < 0.001
Miliary TB 197 (2.6) 158 (2.2) 39 (7.5) 3.56 (2.48, 5.11) 2.82 (1.79, 4.46) < 0.001
TB-CXR 6489 (83.8) 6017 (83.5) 472 (89.1) 2.26 (1.56, 3.27) 2.74 (1.80, 4.16) < 0.001
Cavitation on CXR 2176 (33.6) 2030 (33.8) 146 (30.9) 0.88 (0.72, 1.08)
AFB smear (+) 3082 (47.2) 2855 (46.4) 227 (59.6) 1.70 (1.38, 2.10)
Mtb culture (+) 4416 (68.0) 4108 (67.2) 308 (81.1) 2.09 (1.60, 2.71)
Culture not converted (from positive to negative) 538 (7.0) 328 (4.5) 210 (39.6) 25.74 (19.69, 33.65) 26.66 (19.90, < 0.001
35.71)
Positive culture, NAA, or AFB smear 5930 (76.6) 5435 (75.4) 495 (93.4) 4.62 (3.26, 6.53) 6.17 (4.16, 9.15) < 0.001
HIV status
Negative 6261 (80.9) 5934 (82.3) 327 (61.7)
Positive 478 (6.2) 422 (5.9) 56 (10.6) 2.41 (1.78, 3.25) 2.33 (1.61, 3.36) < 0.001
Unknown 1001 (12.9) 854 (11.8) 147 (27.7) 3.12 (2.54, 3.84) 2.19 (1.68, 2.84) < 0.001
MDR-TB 42 (0.5) 37 (0.5) 5 (0.9) 1.53 (0.60, 3.91)
East Asian 958 (17.9) 877 (17.8) 81 (18.6) 1.06 (0.82, 1.36)
Diabetes × age (≥/ < 45) (interaction term) – – – – 3.12 (1.39, 7.02) 0.01

AUC = 0.89. Values are in number (%) unless otherwise indicated; AUC, area under the receiver operating characteristic (ROC) curve; CXR, chest radiograph; TB-
CXR, abnormalities on CXR consistent with tuberculosis; IDU, injecting-drug user; MDR-TB, Multi-drug resistant TB; NAA, Nucleic Acid Amplification; AFB, acid-fast
bacilli.

burden of TB-DM in Texas. Secondly, as certain historical information the mortality significantly decreased from 2010 through 2016 in the
in the surveillance data was obtained from patient interviews, recall population of non-DM TB patients, no improvement was found in the
bias cannot be completely ruled out. Thirdly, information regarding the mortality of TB-DM patients. These findings raise the urgent need for
body mass index, DM treatment, hemoglobin A1c assay (HbA1c) and developing of more effective strategies in managing this high-risk
lipid profile was not available. Although lacking these DM-specific subgroup of TB patients. Risk factors for mortality determined by
variables, our multivariable models still have a good discrimination multivariable modeling may lay a foundation for the development of a
power with the C-statistic values ranging between 0.86 (for overall prognostic scoring system, which can be used as a practical tool for
mortality) and 0.89 (for mortality during treatment), i.e. the models can health professionals to quickly identify TB-DM patients who have high
predict the mortality risk in most of the cases. Lastly, as Texas is one of risk of mortality and eventually, to develop more effective strategies to
the states having the highest TB prevalence and a more diverse popu- manage TB-DM patients.
lation with 40% being Hispanic and one-third being immigrants, find-
ings from our multivariable modeling might need to be validated in Funding
other populations.
Despite these limitations, our study has notable strengths. Using None.
population-based surveillance data of the entire state of Texas over a
period of 7 years, our study has provided the state-wide concerning
Conflicts of interest
situation of the TB-DM comorbidity and its associated mortality. With
the multivariable models having robust discrimination power and
The authors declare no conflicts of interest.
having large study sample size, we have been able to discriminate the
significantly higher risk of mortality of the TB-DM compared with non-
DM TB patients as well as to identify the independent risk factors as- Financial disclosure
sociated with the mortality in this high-risk population.
Publication of this supplement was supported by The University of
Texas Health Science Center at Houston.
5. Conclusions
Acknowledgments
TB-DM remains a major public health problem in Texas with sig-
nificantly higher mortality compared with non-DM TB patients. While This work was presented in part at the Texas Tuberculosis Research

S64
D.T. Nguyen and E.A. Graviss Tuberculosis 116 (2019) S59–S65

Symposium (TTRS) 2018, El Paso, Texas, USA, sponsored by the Texas [13] Magee MJ, Foote M, Maggio DM, Howards PP, Narayan KM, Blumberg HM, et al.
Tech University Health Sciences Center El Paso. The authors acknowl- Diabetes mellitus and risk of all-cause mortality among patients with tuberculosis in
the state of Georgia, 2009-2012. Ann Epidemiol 2014;24:369–75.
edge the selfless work of public health officials and staff at the City of [14] Abdelbary BE, Garcia-Viveros M, Ramirez-Oropesa H, Rahbar MH, Restrepo BI.
Houston Bureau of Tuberculosis Control, Houston Department of Tuberculosis-diabetes epidemiology in the border and non-border regions of
Health & Human Services, Harris County Public Health, TB Elimination Tamaulipas, Mexico. Tuberculosis (Edinb) 2016;101S:S124–34.
[15] Centers for Disease Control and Prevention (CDC). Report of verified case of tu-
Program, Texas Department of State Health Services and the US Centers berculosis (RVCT) instruction manual. 2009.
for Disease Control that made the data available for use in this analysis. [16] Centers for Disease Control and Prevention. Department of health and human
Services. National tuberculosis surveillance system (NTSS), data set reference guide.
2015 Available at: ftp://ftp.cdc.gov/pub/Software/TIMS/2009%20RVCT%
References 20Documentation/Vocabulary%20Information/NTSS%20Data%20Reference
%20Guide.pdf, Accessed date: 5 April 2018.
[1] World Health Organization (WHO). Global tuberculosis report 2017 2018 Available [17] Wasserman L. Bayesian model selection and model averaging. J Math Psychol
at: http://www.who.int/tb/publications/global_report/en/, Accessed date: 19 2000;44:92–107.
March 2018. [18] Texas Department of State Health Services. TB Control Standards. Texas
[2] International Diabetes Federation. IDF diabetes atlas. Eight Edition 2017 Available Tuberculosis Work Plan. Available at: https://www.dshs.texas.gov/IDCU/disease/
at: http://www.diabetesatlas.org/resources/2017-atlas.html, Accessed date: 20 tb/policies/TBWorkPlan.pdf. Accessed on 05/02/2018.
April 2018. [19] Nguyen DT, Graviss EA. Development and validation of a prognostic score to pre-
[3] Pizzol D, Di Gennaro F, Chhaganlal KD, Fabrizio C, Monno L, Putoto G, et al. dict tuberculosis mortality. J Infect 2018 Oct;77(4):283–90. 2018.02.009. Epub
Tuberculosis and diabetes: current state and future perspectives. Trop Med Int 2018 Apr 9.
Health 2016;21:694–702. [20] Nguyen DT, Jenkins HE, Graviss EA. Prognostic score to predict mortality during TB
[4] Al-Rifai RH, Pearson F, Critchley JA, Abu-Raddad LJ. Association between diabetes treatment in TB/HIV co-infected patients. PLoS One 2018;13:e0196022.
mellitus and active tuberculosis: a systematic review and meta-analysis. PLoS One [21] Reis-Santos B, Gomes T, Horta BL, Maciel EL. The outcome of tuberculosis treat-
2017;12:e0187967. ment in subjects with chronic kidney disease in Brazil: a multinomial analysis. J
[5] Workneh MH, Bjune GA, Yimer SA. Prevalence and associated factors of tubercu- Bras Pneumol 2013;39:585–94.
losis and diabetes mellitus comorbidity: a systematic review. PLoS One [22] Underwood J, Cresswell F, Salam AP, Keeley AJ, Cleland C, John L, et al.
2017;12:e0175925. Complications of miliary tuberculosis: low mortality and predictive biomarkers
[6] Jimenez-Corona ME, Cruz-Hervert LP, Garcia-Garcia L, Ferreyra-Reyes L, Delgado- from a UK cohort. BMC Infect Dis 2017;17(295). 017-2397-6.
Sanchez G, Bobadilla-Del-Valle M, et al. Association of diabetes and tuberculosis: [23] Lee HG, William T, Menon J, Ralph AP, Ooi EE, Hou Y, et al. Tuberculous me-
impact on treatment and post-treatment outcomes. Thorax 2013;68:214–20. ningitis is a major cause of mortality and morbidity in adults with central nervous
[7] Huangfu P, Pearson F, Ugarte-Gil C, Critchley J. Diabetes and poor tuberculosis system infections in Kota Kinabalu, Sabah, Malaysia: an observational study. BMC
treatment outcomes: issues and implications in data interpretation and analysis. Int Infect Dis 2016;16. 296,016-1640-x.
J Tuberc Lung Dis 2017;21:1214–9. [24] Reed GW, Choi H, Lee SY, Lee M, Kim Y, Park H, et al. Impact of diabetes and
[8] Degner NR, Wang JY, Golub JE, Karakousis PC. Metformin use reverses the in- smoking on mortality in tuberculosis. PLoS One 2013;8:e58044.
creased mortality associated with diabetes mellitus during tuberculosis treatment. [25] Oursler KK, Moore RD, Bishai WR, Harrington SM, Pope DS, Chaisson RE. Survival
Clin Infect Dis 2018;66:198–205. of patients with pulmonary tuberculosis: clinical and molecular epidemiologic
[9] Centers for Disease Control and Prevention (CDC). Reported tuberculosis in the factors. Clin Infect Dis 2002;34:752–9.
United States. 2016 Available at: https://www.cdc.gov/tb/statistics/reports/2016/ [26] Pablos-Mendez A, Blustein J, Knirsch CA. The role of diabetes mellitus in the higher
pdfs/2016_Surveillance_FullReport.pdf, Accessed date: 3 October 2018 2017. prevalence of tuberculosis among Hispanics. Am J Public Health 1997;87:574–9.
[10] Centers for Disease Control and Prevention (CDC). National diabetes statistics re- [27] Marks SM. Diabetes and tuberculosis, US national health interview survey, 2000-
port Estimates of Diabetes and Its Burden in the United States; 2017 Available at: 2005. Int J Tuberc Lung Dis 2011;15:982–4.
https://www.cdc.gov/diabetes/pdfs/data/statistics/national-diabetes-statistics- [28] Centers for Disease Control and Prevention (CDC). Tuberculosis in Hispanics/
report.pdf, Accessed date: 3 June 2018. Latinos. Available at: https://www.cdc.gov/tb/publications/pdf/tbhispanics_rev.
[11] Texas Department of State Health Services. Texas Diabetes and Prediabetes Fact pdf. Accessed on 05/04/2018.
Sheet. Available at: https://www.dshs.texas.gov/diabetes/PDF/data/ [29] Texas Department of State Health Services. reportEpidemiology and Supplemental
2017TXprediabetesDiabetesFactSheet.pdf. Accessed on 05/04/2018. Projects Group. Texas TB Surveillance Annual Report 2016. Austin, TX; 2017.
[12] Restrepo BI, Camerlin AJ, Rahbar MH, Wang W, Restrepo MA, Zarate I, et al. Cross- Available at: https://www.dshs.texas.gov/IDCU/disease/tb/statistics/
sectional assessment reveals high diabetes prevalence among newly-diagnosed tu- TBSurveillanceReport.pdf. Accessed on 05/04/2018.
berculosis cases. Bull World Health Organ 2011;89:352–9.

S65

You might also like