The impact on incident tuberculosis by kidney function impairment status: analysis of severity relationship

This retrospective study was conducted in a tertiary-care referral center in northern Taiwan under the approval of the Institutional Review Board of Research Ethics Committee of National Taiwan University Hospital (NO. 201510009RINB). From January 2008 to December 2013, patients aged ≥20 years were identified. We included patients with data of serum creatinine and excluded those having follow-up periods at the study hospital of less than 6 months and 3 months, respectively, before and after the index data of initial kidney function. In addition, we excluded those who had active tuberculosis before recruitment and within 3 months right after the index data of the kidney function.

We retrieved the participants’ clinical information such as age, gender, body mass index, and underlying disease from the hospital’s electronic record database within a total 12 months (6 months before and after the index date). We used an abbreviated Modification of Diet in Renal Disease (MDRD) equation to estimate the glomerular filtration rate (eGFR) [16] and classified the patient’s initial kidney function according to the guidelines of the National Kidney Foundation [17]. The eGFR of ≥90, < 90~ ≥ 60, < 60~ ≥ 45, < 45~ ≥ 30, < 30~ ≥ 15, and < 15 ml/min/1.73m² stood for no CKD or stage 1, stage 2, stage 3a, stage 3b, stage 4, and stage 5, respectively. We defined active TB by positive culture result for Mycobacterium tuberculosis. For the diagnosis without microbiological evidence, we could define active TB by the unique diagnosis code in electronic chart record, because active TB is a certifiable infectious disease that needs to report to the Taiwan Centers for Disease Control. We used one time of admission diagnosis, or at least two outpatient clinic diagnoses plus prescription of anti-tuberculous agents for TB diagnosis. Co-morbidity of long-term dialysis, transplantation, pneumoconiosis, idiopathic pulmonary fibrosis, cirrhosis of the liver, cancer, lupus erythematous, rheumatoid arthritis, polymyositis, and dermatomyositis were defined by a record of a catastrophic illness card. Other underlying diseases were categorized by one admission diagnosis or more than 2 outpatient clinic diagnoses.

We followed up all enrolled participants till December 2016 and defined occurrence of active TB as the primary outcome. The incidence of TB was calculated as the average number per 100,000 person-years. Inter-group differences were compared using t test or one-way ANOVA for continuous variables, where appropriate, and chi squared test for categorical variables. We used Cox proportional hazard regression for time to TB event analysis. Once the patients with CKD at the initial received long-term dialysis or kidney transplantation in the follow-up or were lost to follow-up, they were censored in the time dependent analysis. We combined no CKD or stage 1 and stage 2 as a reference group (≥ 60 min/ml per 1.73m²) in Cox regression analysis because their TB incidence were similar [18]. We calculated the correlations between the kidney function of eGFR and the univariate hazard ratio for TB by Spline Cox proportional regression. We investigated the risk factors for active TB using multivariate Cox proportional hazard regression analysis with possible associated variables from univariate analysis. Only variables with a two-sided p < 0.05 were retained in the final model. All analyses were performed in SAS version 9.4 (SAS Institute Inc., Cary, NC, USA).

During the study period, a total of 485,190 patients in the study hospital were eligible, and 289,579 were enrolled for analysis after excluding 124,274 for short follow-up (< 90 days), 69,912 for short previous data (< 6 months) and 1425 for having TB within 90 days (Fig. 1). Among them, 1012 patients had development of active TB events during the overall follow up of 1,196,206 person-years, that was averagely 4.13 years/person. Among those with different stages of kidney function impairment (Table 1), old age and male predominance were generally found for those with advanced CKD stages. The mean age of all participants was 5.18 years old (SD: 17.3). The mean age was 43.8 (15.2) in no CKD or stage 1, 55.5 (15.7) in stage 2, 69.5 (12.4) in stage 3a, 71.6 (12.8) in stage 3b, 70.1 (14.2) in stage 4, 62.7 (15.7) in stage 5, 56.9 (14.4) in dialysis, and 54.8 (11.7) in transplant group (p < 0.0001, Table 1). Male gender was around 45.3% in all participants and distributed as 39.7, 49.4, 51.9, 50.3, 50.4, and 52.2% from no CKD or stage 1, stage 2, 3a, 3b, 4 and 5, respectively. Many underlying comorbidities significantly increased in advanced CKD stage included CHF, stroke, DM and SLE. By contrast, obesity was less frequently in in patients with worse kidney function than those with no CKD or stage 1 and 2.

In comparison with patients without TB (Additional file 1: Table S1), those who developed TB during follow-up were older and predominantly male. In addition, they had lower eGFR and more advanced renal disease, dialysis, and kidney transplant. The proportion of patients with active TB development was lower in no CKD or stage 1 and 2, but it became higher in advanced CKD stage 3 to 5 in comparing with those without TB (p < 0.0001). The underlying diseases of pulmonary disease, gastroesophageal reflux disease, cancer, heart failure, cirrhosis, diabetes mellitus and autoimmune disease were higher in patients with TB development. By contrast, obesity was higher in patients without TB development.

According to kidney function, the incidence rate of TB was similar in no CKD or stage 1 and stage 2 (70.06 and 71.08 per 100,000 person-years, respectively), and increased apparently from stage 3a (167.68 per 100,000 person-years, Hazard ratio [HR]: 2.404 [1.956–2.956], p < 0.0001) to stage 3b, stage 4 and stage 5 (incidence: 229.25, 304.95 and 349.29 per 100,000 person-years, and HR: 3.273 [2.509–4.270], 4.317 [3.078–6.054], and 4.852 [3.154–7.646], respectively) (Table 2). Those with long-term dialysis or kidney transplant before case enrollment had TB incidences of 175.60 and 1429.33 per 100,000 person-years, respectively and HR of 2.493 [1.330–4.673] and 20.575 [5.147–82.239], respectively.

For confirming that the case exclusion criteria of follow-up period ≤90 days did not influence the results of TB incidence, we used different follow-up period as exclusion criteria, including ≤30 days or ≤ 180 days and analyzed the TB incidence by different kidney function status. We found that the trends of TB incidence in the different CKD subgroups were similar (Additional file 1: Table S2 and S3).

In a Cox proportional hazard regression model (Table 3), we analyzed the hazard ratio (HR) of TB development for kidney function and other clinically relevant factors. In univariate analysis, patients had similar TB risk in no CKD or stages 1 vs stage 2 (HR: 1.024 [95% C.I.: 0.888–1.182], p = 0.7408, compared with no CKD or stage 1), so we used them as the reference group for those with deteriorated kidney function in Cox regression analysis. In univariate analysis, the dose response of TB risk among different kidney functions increased significantly from CKD stage 3a to stage 5. When we evaluated the univariate HR of TB risk at every eGFR below 90 ml/min/1.73m² using spline Cox regression, the correlation actually increased slowly in stage 2 and obviously from stage 3a. A plateau was noticed around the beginning of stage 3b and a mild decline at late stage 5. However, the outcome number was small in stage 5, and the 95% confidence interval became large (Fig. 2).

In multivariate Cox regression (Table 3), we adjusted all factors found to be significant in the univariate analysis, and the final model showed that old age (HR: 4.439 [3.539–5.583] for age > 65 years old; HR: 2.040 [1.592–2.614] for age 55–65 years old; HR: 1.392 [1.079–2.797] for age 40–55 years old, compared with age < 40 years old), male (HR: 2.558 [1.977–2.578], and poor kidney function since CKD stage 3b were correlated with higher TB risk. In patients with deteriorated kidney function, the risk of TB development increased significantly from CKD stage 3b to stage 5, similarly to the results of univariate analysis but with the risk adjusted downward. The HR of CKD stage 3a was 1.215 (95% C.I.: 0.995–1.484, p = 0.0563), HR: 1.538 in stage 3b ([1.182–2.002], p = 0.0014), 2.065 in stage 4 ([1.473–2.894]), and 2.877 in stage 5 ([1.872–4.421]) in comparison with stages 1 and 2. The risk in patients with long-term dialysis was HR of 2.041 (1.092–3.815, p = 0.0254), which is similar to that of stage 4 CKD. The HR was 15.269 in those with kidney transplant, and it was much higher than those of other subgroups. The proportion of ever corticosteroid use (n = 24 [63.2%] vs n = 87,922 [30.4%], p < 0.001) and immune-suppressants use (n = 21 [55.3%] vs n = 2239 [0.8%], p < 0.001) during the follow up were higher in patients receiving kidney transplant than other patients. In addition, the underlying diseases of COPD, bronchiectasis, pneumoconiosis, cancer, congestive heart failure, diabetes mellitus, SLE and RA were independent conditions for favoring TB development. Among them, pneumoconiosis (HR: 10.149), SLE (HR: 6.323), bronchiectasis (HR: 2.959), RA (HR: 2.305), and cancer (HR: 2.006) were the most important (HR >2.0) clinical diseases predisposing TB development.