Post-tuberculosis incidence of diabetes, myocardial infarction, and stroke: Retrospective cohort analysis of patients formerly treated for tuberculosis in Taiwan, 2002-2013.

OBJECTIVES: To estimate the incidence of diabetes, acute myocardial infarction (AMI), and stroke; and to determine factors associated with diabetes, AMI, and stroke incidence among patients previously treated for tuberculosis (TB) disease.
METHODS: A retrospective cohort study was conducted among non-pediatric TB patients registered in the Taiwan National Health Insurance Research Database (NHIRD) from 2002-2013. Diabetes, AMI, and stroke incidence were defined by International Classification of Diseases (ICD)-9 codes, drug prescriptions, and records of patient's clinic visits. Cox proportional hazard models were used to estimate the hazard rate ratio (HR) of incident diabetes, AMI, and stroke.
RESULTS: From 2002-2013, there were 157,444 patients treated for TB registered in NHIRD. Among 129,453 patients with no prior history of diabetes, the age-adjusted incidence rate (IR) of diabetes was 3.85 (95%CI 3.70-4.01) per 1000 person-years. Among 143,646 patients with no prior history of AMI, the age-adjusted IR of AMI as 3.26 (95%CI 3.13-3.40). Among 118,774 patients with no prior history of stroke, the age-adjusted IR of stroke was 16.08 (95%CI 15.76-16.32).
CONCLUSIONS: Chronic non-communicable disease risk factors like dyslipidemia, hypertension, and chronic kidney disease diagnosed before time of TB diagnosis were predictive of diabetes, AMI, and stroke incidence.

Introduction

Increasing evidence suggest that tuberculosis (TB) disease may disrupt host metabolism and contribute to subsequent risks of chronic non-communicable diseases (NCDs) (Huaman et al., 2017, Kamper-Jorgensen et al., 2015, Sheu et al., 2010). Importantly, the majority of TB patients reside in low- and middle-income countries where chronic NCDs such as type-2 diabetes (diabetes), acute myocardial infarction (AMI), and stroke are increasingly threatening public health achievements (Gaziano et al., 2010).

Diabetes is a well-established risk factor for TB infection (Lee et al., 2017), TB disease, and poor TB treatment outcomes (Riza et al., 2014). However, epidemiologic data to support the relationship between TB and risk of NCDs are scarce. Chlamydia pneumoniae, Helicobacter pylori, influenza, and HIV are associated with cardiovascular disease risk, likely due to chronic systemic inflammation that can lead to atherosclerotic lesions (Huaman et al., 2015). However, whether individual-level TB risk factors (e.g., comorbidities, treatment duration, site of infection) increase the risk of post-infection incident NCDs is unknown. Therefore, among a large cohort of patients post-TB treatment, we aimed to estimate the incidence rates of diabetes, AMI, and stroke and to determine characteristics associated with these incident NCDs.

Methods

We conducted a cohort study among patients formerly treated for TB disease using data from the Taiwan National Health Insurance Research Database(NHIRD)during2002–2013. NHIRD is maintained by the National Health Research Institute and contains original registration and insurance claim data from the single-payer national health insurance system which covers 99.9% of the Taiwanese population.

Eligible patients included those with non-pediatric active TB disease (≥15 years old for DM study; ≥25 years old for AMI and stroke studies) confirmed by ICD-9 codes (010–018) and prescriptions of anti-TB drugs for ≥28 days within a year. The earliest date of TB diagnosis indicated by ICD-9 codes was defined as the TB date. Patients with ICD-9 diagnoses of diabetes, AMI, or stroke on or before date of TB disease were excluded. Incident diabetes was defined as patients who received anti-diabetes drug prescriptions for ≥28 days within a year and a confirmed diabetes ICD-9 (250.X0 or 250.X2) first indicated ≥2 years after TB date. Incident AMI (ICD-9 410.XX) and stroke (ICD-9 430.XX - 438.XX) were defined by ICD-9 codes and ≥3 out-patient clinic visits or 1 hospitalization indicated ≥1 year after TB date. Other comorbidity characteristics including dyslipidemia, hypertension, overweight, HIV, and chronic kidney diseases (CKD) were determined by ICD-9 codes. We calculated incidence of diabetes, AMI, and stroke using Poisson regression, and proportional hazards regression was used to estimate relative hazards of NCDs incidence. Incidence rates (IR) were presented per 1000 person-years. Patients were censored if they left the insurance system or did not develop diabetes, AMI, or stroke by the end of follow-up (December 2013).

Results

During the study period, there were 157,444 patients treated for TB disease and 1315 were <15 years and excluded. Among adult TB cases, 82.9% (129,453/156,129) patients did not have prior diabetes diagnosis; 98.0% (143,678/146,646) did not have prior history of AMI; and 81.0% (118,774/146,646) did not have prior stroke diagnosis (Supplemental Material 1) and were included. Among those without preexisting diabetes, 2479 incident diabetes diagnoses occurred during 643,754 person-years (age-adjusted IR 3.85, 95%CI 3.70–4.01) (Table 1). The median time of incident diabetes post TB date was 7 years (IQR 5–9) (Supplemental Material 2). In the adjusted model, TB patients with pre-existing dyslipidemia (adjusted hazard ratio [aHR] 1.41; 95%CI 1.25–1.58), hypertension (aHR 1.78, 95%CI 1.62–1.94), and overweight (aHR 2.67, 95%CI 1.76–4.03) had increased risk of diabetes post-TB treatment (Table 2).

Table 1

Incidence rates of diabetes, acute myocardial infarction (AMI), and stroke among a cohort of patients formerly treated for tuberculosis, Taiwan National Health Insurance Research Database (NHIRD), 2002–2013.

Characteristics	Diabetes		AMI		Stroke
	N (%) = 2479 (1.97%)		N (%) = 2287 (1.59%)		N (%) = 9539 (8.08%)
	Total PTY = 643754		Total PTY = 700890		Total PTY = 596395
	Overall IR = 3.85 (3.70–4.01)		Overall IR = 3.26 (3.13–3.40)		Overall IR=15.99 (15.68–16.32)
	N/PTY	IR (95% CI)	N/PTY	IR (95% CI)	N/PTY	IR (95% CI)
Age group
15–24	14/55138	0.25 (0.15–0.43)	Excluded	N/A	Excluded	N/A
25–44	344/151792	2.27 (2.04–2.52)	68/162616	0.42 (0.33–0.53)	530/159684	3.32 (3.05–3.61)
45–64	959/179001	5.36 (5.03–5.71)	663/229965	2.88 (2.67–3.11)	2788/211591	13.18 (12.70–13.67)
≥65	1162/257823	4.51 (4.26–4.77)	1556/308309	5.05 (4.80–5.30)	6275/225120	27.87 (27.19–28.57)
Gender
Female	639/217826	2.93 (2.72–3.17)	490/223450	2.19 (2.01–2.40)	2447/194418	12.59 (12.10–13.09)
Male	1840/422458	4.36 (4.16–4.56)	1797/474570	3.79 (3.62–3.97)	7146/399157	17.90 (17.49–18.32)
Unknown	0/3470	N/A	0/2870	N/A	0/2820	N/A
TB site
Pulmonary	2145/547526	3.92 (3.76–4.09)	2001/600655	3.33 (3.19–3.48)	8331/512211	16.26 (15.92–16.62)
Extrapulmonary	320/91256	3.51 (3.14–3.91)	264/94947	2.78 (2.47–3.14)	1206/79812	15.11 (14.28–15.99)
Miliary	14/4972	2.82 (1.67–4.75)	22/5288	2.74 (2.74–6.32)	56/4372	12.81 (9.86–16.64)
TB treatment duration
≤6 months	1749/469072	3.73 (3.56–3.91)	1564/504205	3.10 (2.95–3.26)	6698/428522	15.63 (15.26–16.01)
7–12 months	668/159157	4.20 (3.89–4.53)	656/178856	3.67 (3.40–3.96)	2633/152429	17.27 (16.63–17.95)
>12 months	62/15525	3.99 (3.11–5.12)	67/17829	3.76 (2.96–4.78)	262/15444	16.96 (15.03–19.15)
Dyslipidemia[a]
No	2108/574547	3.67 (3.52–3.83)	1766/599014	2.95 (2.81–3.09)	7900/518986	15.22 (14.89–15.56)
Yes	371/69207	5.36 (4.84–5.94)	521/101876	5.11 (4.69–5.57)	1693/77409	21.87 (20.85–22.94)
Hypertension[a]
No	1410/448700	3.14 (2.98–3.31)	917/444348	2.06 (1.93–2.20)	4795/415207	11.55 (11.23–11.88)
Yes	1069/195054	5.48 (5.16–5.82)	1370/256542	5.34 (5.07–5.63)	4798/181188	26.48 (25.74–27.24)
Overweight[a]
No	2456/641308	3.83 (3.68–3.98)	2276/697646	3.26 (3.13–3.40)	9554/593576	16.10 (15.78–16.42)
Yes	23/2446	9.40 (6.25–14.15)	11/3244	3.39 (1.88–6.12)	39/2819	13.83 (10.11–18.94)
HIV[a]
No	2474/639526	3.87 (3.72–4.02)	2276/696710	3.27 (3.14–3.41)	9579/592428	16.17 (15.85–16.50)
Yes	5/4228	1.18 (0.49–2.84)	8/4180	1.91 (0.96–3.83)	14/3967	3.53 (2.09–5.96)
CKD[a]
No	2378/612063	3.89 (3.73–4.05)	1985/660103	3.01 (2.88–3.14)	8870/567777	15.62 (15.30–15.95)
Yes	101/31691	3.19 (2.62–3.87)	302/40787	7.40 (6.62–8.29)	723/28618	25.26 (23.49–27.17)
Diabetes[a]	Excluded
No			1512/582699	2.60 (2.47–2.73)	7183/502688	14.29 (13.96–14.62)
Yes			775/118191	6.56 (6.11–7.04)	2410/93707	25.72 (24.71–26.77)
Myocardial infarction[a]			Excluded
No	2457/637226	3.86 (3.71–4.01)			9453/590847	16.00 (15.68–16.32)
Yes	22/6528	3.37 (2.22–5.12)			140/5548	25.23 (21.38–29.78)
Stroke[a]
No	2133/563277	3.79 (3.63–3.95)	1707/598831	2.85 (2.72–2.99)	Excluded
Yes	346/80477	4.30 (3.87–4.78)	580/102059	5.68 (5.24–6.17)

Notes:

Abbreviation: AMI - Acute Myocardial Infarction; IR - incidence rate; CI - Confidence Interval; PTY - person time in years.

Comorbidities (i.e., dyslipidemia, hypertension, overweight, HIV, and chronic kidney disease) were diagnosed earlier or at the time of TB diagnosis.

Table 2

Adjusted Cox regression model[a] to predict post-tuberculosis incident diabetes, acute myocardial infarction (AMI), and stroke, Taiwan National Health Insurance Research Database (NHIRD), 2002–2013.

Characteristics	DiabetesaHR[c] (95% CI)	AMIaHR[c] (95% CI)	StrokeaHR[c] (95% CI)
Age group
15–24	0.11 (0.07-0.19)	Excluded	Excluded
25–44	Ref	Ref	Ref
45–64	1.96 (1.73-2.23)	4.72 (3.67-6.08)	3.11 (2.83-3.42)
≥65	1.66 (1.46-1.90)	8.01 (6.24-10.28)	6.57 (6.00-7.20)
Gender
Female	Ref	Ref	Ref
Male	1.36 (1.24-1.49)	1.57 (1.42-1.74)	1.28 (1.22-1.34)
TB site
Pulmonary	Ref	Ref	Ref
Extrapulmonary	0.93 (0.82–1.04)	0.90 (0.79–1.02)	1.00 (0.94–1.07)
Miliary	0.82 (0.48–1.38)	1.30 (0.86–1.98)	0.82 (0.63–1.06)
TB treatment duration
≤6 months	Ref	Ref	Ref
7–12 months	1.05 (0.96–1.14)	1.09 (0.99–1.19)	1.03 (0.99–1.08)
>12 months	0.78 (0.61 – 1.01)	1.03 (0.81 – 1.32)	0.89 (0.78 – 1.00)
Dyslipidemia[b]
No	Ref	Ref	Ref
Yes	1.41 (1.25-1.58)	1.29 (1.17-1.43)	1.20 (1.13-1.27)
Hypertension[b]
No	Ref	Ref	Ref
Yes	1.78 (1.62-1.94)	1.70 (1.54-1.86)	1.67 (1.60-1.75)
Overweight[b]
No	Ref	Ref	Ref
Yes	2.67 (1.76-4.03)	1.08 (0.60–1.95)	0.98 (0.72–1.35)
HIV[b]
No	Ref	Ref	Ref
Yes	0.44 (0.18–1.06)	1.81 (0.90–3.64)	0.62 (0.36–1.04)
CKD[b]
No	Ref	Ref	Ref
Yes	0.78 (0.64-0.96)	1.86 (1.65-2.11)	1.26 (1.17-1.36)
Diabetes[b]	Excluded
No		Ref	Ref
Yes		1.96 (1.79-2.14)	1.47 (1.40-1.54)
Myocardial infarction[b]		Excluded
No	Ref		Ref
Yes	0.84 (0.55–1.28)		1.05 (0.89–1.24)
Stroke[b]			Excluded
No	Ref	Ref
Yes	0.96 (0.85–1.08)	1.24 (1.12-1.37)

Abbreviation: AMI - Acute Myocardial Infarction; IQR - interquartile range; CHR - Crude hazard ratio; AHR - Adjusted Hazard Ratio; CI - Confidence Interval; TB - Tuberculosis.

Bold indicates statistical significance (two-sided P-value <0.05).

Proportional hazard assumption was assessed using Schoenfeld’s residual and log of negative log curve.

Comorbidities (i.e., dyslipidemia, hypertension, overweight, HIV, and chronic kidney disease) were diagnosed earlier or at the time of TB diagnosis.

Hazard Ratios after adjusting for age, gender, type of TB, anti TB treatment duration, dyslipidemia, hypertension, and chronic kidney disease status.

The age-adjusted IR of AMI was 3.26 (95%CI 3.13–3.40), and 16.08 (95%CI 15.76–16.32) for stroke. The median time of incident AMI and stroke post-TB date was 6 years (IQR 3–8) and 6 years (IQR 3–9), respectively. In the adjusted models, factors predictive of incident AMI and stroke included pre-existing dyslipidemia (aHR AMI 1.29, 95%CI 1.17–1.43; aHR stroke 1.20, 95%CI 1.13–1.27), hypertension (aHR AMI 1.70, 95%CI 1.54–1.86; aHR stroke 1.67, 95% CI 1.60–1.75), CKD (aHR AMI 1.86,95%CI 1.65–2.11; aHR stroke 1.26, 95%CI 1.17–1.36), and diabetes (aHR AMI 1.96,95%CI 1.79–2.14; aHR stroke 1.47, 95%CI 1.40–1.54).

Discussion

Our preliminary findings suggest that older age, male gender, and traditional pre-existing chronic NCDs are predictive of diabetes, AMI, and stroke incidence post-TB treatment. Although overall TB treatment duration was not predictive of diabetes, AMI, and stroke incidence, we found that patients treated for 7–12 months had higher incidence of NCDs compared to those treated for ≤6 months. Furthermore, our incidence estimation of diabetes, AMI, and stroke were higher when compared to previous Taiwanese national estimates using NHIRD database. For example, the annual age-standardized incidence rate of diabetes was <1% across Taiwanese population (Jiang et al., 2012). Additionally, the age and gender-adjusted incidence of AMI was 50.7/100,000 persons in 2015 (Lee et al., 2018) and the estimated stroke incidence was 12.3/1000 person-years (Chan et al., 2018). Our results highlight the need for studies to determine the impact of TB disease on risk of NCDs; such studies will require substantial follow-up time and a control group without TB disease.