Prevalence and risk factors of active tuberculosis in patients with rheumatic diseases: a multi-center, cross-sectional study in China

ABSTRACT Evidence of active tuberculosis (ATB) in patients with rheumatic diseases are research priorities but limited data from China have been reported. Research targeting patients not taking anti-TNF biologics are especially insufficient. We aimed to investigate the prevalence and risk factors of ATB in this at-risk population. We conducted a tertiary hospital-based, multi-center, cross-sectional study by using stratified multi-stage cluster sampling strategy to screen ATB in patients with rheumatic diseases. We estimated the prevalence of ATB in patients with rheumatic diseases and identified risk factors among those who were not taking anti-TNF biologic. A total of 13,550 eligible patients were enrolled, and the result showed the standardized prevalence of ATB according to the composition ratio of various types of rheumatic disease was 882/100000 (95% confidence interval (CI): 706-1057). Multivariable logistic regression analysis in patients not taking anti-TNF biologics showed that the independent risk factors of ATB were having systemic lupus erythematosus (SLE) (OR=2.722, 95% CI: 1.437-5.159, p=0.002), having Behcet’s disease (BD) (OR= 5.261, 95% CI: 2.071-13.365, p<0.001), taking azathioprine(AZA) within the past two years (OR=2.095, 95% CI: 0.986-4.450, p=0.054), exposing to glucocorticoids ≥30mg/d for more than four weeks within the past two years (OR=2.031, 95% CI: 1.247-3.309, p=0.004) and having evidences of previous TB (OR= 6.185, 95% CI: 3.487-10.969, p<0.001). The prevalence of ATB was higher in patients with rheumatic diseases compared to the general population. Patients with SLE or BD, prolonged exposure to moderate to high dose of glucocorticoids and previous TB were independent risk factors for ATB.


Background
China has the third largest number of tuberculosis (TB) cases in the world [1]. The prevalence of active pulmonary TB was 459/100,000 in population aged over 15 in 2010 [2]. In 2019, 833,000 people were newly diagnosed with TB in China, accounting for nearly 8.4% of all TB cases worldwide [1]. Rheumatic disease is also a serious problem in China. It was estimated that China has approximately 20 million patients with rheumatic disease [3,4]. Due to the primary disease as well as the use of immunosuppressants, patients with rheumatic disease were at high risk for developing ATB.
The coexistence of TB and rheumatic diseases requires many complex considerations. TB infection can make the diagnosis and treatment of rheumatic diseases complicated. On the other hand, rheumatic patients are usually under immunocompromised conditions, leading to prolonged infection duration and higher mortality rate [5,6]. Therefore, prevention of new infections of Mycobacterium tuberculosis (MTB) and the progression to ATB is especially crucial for patients with rheumatic diseases.
WHO strongly recommends that patients initiating anti-TNF treatment should be systematically tested and treated for latent TB infection (LTBI). However, there is no recommendation of TB preventive treatment for patients with rheumatic conditions and receiving steroid treatment [7][8][9]. In China, the expert consensus on treatment for LTBI was reached for rheumatic patients receiving TNF antagonists in 2014, but no recommendation of TB prevention has been made when anti-TNF drugs are not administered until now [10].
Tuberculosis prevention and control in patients with rheumatic diseases faces two difficulties. Firstly, the prevalence of ATB in this population is unknown. This is the first step for TB control, but national epidemiological data is still lacking. Secondly, the risk factors of ATB in rheumatic patients not receiving anti-TNF treatment are unclear. Given the medical resource constraint and the fact that anti-TB therapy may induce severe adverse drug reactions when combined with rheumatic diseases [11], TB preventive treatment for all rheumatic patients may not be an optimal option. Identifying risk factors offer doctors reliable information in assessing the risk of ATB in patients with rheumatic diseases, which can greatly facilitate the generation of efficient prevention strategies based on clinical individualized risk assessment. Previous case-control and cohort studies have shown that the use of glucocorticoids and some immunosuppressants can increase the risk of ATB in patients with rheumatic disorders [12][13][14][15][16]. However, the findings are inconsistent, and most of the data come from countries or regions with low or medium TB burden.
In China, due to the complexity of the diagnosis and treatment of rheumatic diseases, only tertiary general hospitals are required by the government to set up departments of rheumatology. Under this circumstance, a multi-center, cross-sectional study (ETHERTB) based on tertiary general hospitals was conducted to investigate the prevalence of ATB among patients with rheumatic diseases and to determine risk factors associated with ATB in patients not taking anti-TNF medications.
This study was approved by the Ethics Committees of Peking Union Medical College Hospital (No.S-715) and 12 participating hospitals. Written informed consents were obtained from all patients and their legal guardians if necessary.

Sampling
TB epidemic is closely related to local economic condition. The regional economic disparity in China leads to the distinct burden of TB among different regions [2]. Therefore, we adopted a stratified multi-stage cluster sampling strategy in order to recruit a representative sample of patients for this study. Firstly, China was divided into three regions based on geographic location and economic status, namely, eastern, central, and western areas. Then, we used simple random sampling method to select the province in each region. We estimated that there was no significant difference in the prevalence of rheumatic diseases among different regions [4,17], the number of cases with rheumatic diseases would be proportional to the total population. So, we determined the number of provinces according to the population size of different regions and selected 6 out of 9, 3 out of 10 and 4 out of 12 provinces, municipalities or autonomous regions were selected from the eastern, central, and western region, respectively. Secondly, for each selected province, municipality, or autonomous region, one tertiary general hospital was randomly chosen as study site (Supplementary Figure 1). Finally, the required sample size for each region was equally allocated to the sites in that region and eligible outpatients and inpatients with rheumatic diseases were consecutively screened and enrolled.
Firstly, the required sample size of each region was calculated using the formula of n = Z 2 · P · (1 − P) d 2 where n is the sample size, Z is the statistic corresponding to 95% confidence interval, P is expected prevalence of ATB in patients with rheumatic diseases and d is precision. We assumed there was no significant difference in the prevalence of rhematic disease among different regions [4,17] and the prevalence of ATB in rheumatic patients was a 10-fold of prevalence of ATB in the general population. So we set the expected prevalence of ATB in rheumatic patients to be 3%, 5% and 7% in the eastern, central and western regions, respectively [2]. And d was set as one-fifth of the prevalence. Confirming the diagnosis can sometimes be complicated and may require regular follow-up. Then, the sample size was amplified by considering the loss-to-follow-up rate and design effect. We assumed approximately 10% loss to follow-up. Also, we set the design effect as 2 by considering the stratified multi-stage cluster sampling method. Accordingly, the sample sizes for the eastern, central and western regions were estimated to be 6900, 4056, and 2836, respectively.

Screening for ATB
Chest X-rays or CT scans were performed on all patients for ATB screening. When ATB was suspected, specimens from lesions including sputum, other respiratory secretions, cerebrospinal fluid, pleural effusion, ascites and pericardial fluid were collected for acidfast staining and TB culture. Histopathological examinations were conducted when necessary. The screening algorithm of ATB is summarized in Figure 1. The diagnosis of ATB was independently made by two respiratory or infectious disease specialists. Disagreement was resolved by consulting a third senior physician. The diagnostic criteria of ATB are listed in Table 1.

Potential risk factors for ATB
Age, gender, type of rheumatic disease, duration of rheumatic disease, dosage and duration of glucocorticoids within the past two years, use of immunosuppressants within the past two years and evidence of previous TB were collected as variables for the risk factor analysis.

Statistical analysis
Normality of the numerical variables were tested by the Kolmogorov-Smirnov test. Numerical variables with normal distributions were expressed as mean ± standard deviation (SD), non-normal variables were described as median and interquartile range [IQR]. Categorical variables were presented as frequencies and proportions. Comparisons of continuous variables between patients with and without ATB were performed using t-test and Mann-Whitney U test for normal and non-normal data respectively. Categorical data were compared using Chi-squared or Fisher's exact test as appropriate.
The prevalence of ATB in each rheumatic disease was calculated and a crude overall prevalence was estimated as well. The risk of ATB varied among different rheumatic diseases. In order to avoid bias due to differences in the disease composition between the included patients and the general rheumatic patients, we collected the disease type data of all patients with rheumatic disease in 13 sub-centers throughout 2014 and standardized the overall ATB prevalence according to the weight of the disease types. We estimated the prevalence of ATB based on Possion distribution. 95% confidence interval (CI) of the prevalence was presented.
We used univariate analysis and multivariable logistic regressions to identify the risk factors of ATB. Patients who were on anti-TNF biologics and TB preventive treatment were excluded from the analysis. Age, gender, the type of rheumatic disease, the duration of rheumatic disease, use of immunosuppressants within past two years, dosage and duration of glucocorticoids within the past two years and evidence of previous TB were included in the univariate analysis. Variables that were considered clinically relevant or that showed a significant association with ATB in the univariate analysis were entered into the multivariable logistic regressions model. Given the number of events available, variables included in the multivariate logistic model were carefully chosen to ensure parsimony of the final model (Backward LR, entry 0.05, removal 0.10).
All statistical analyses were performed with SPSS (IBM SPSS Statistics for Windows, Version 22.0. Armonk, NY: IBM Corp). A p < 0.05 was considered statistically significant.

Results
A total of 13,550 patients were finally included for the estimation of ATB prevalence. After excluding participants with history of anti-TNF biologics or TB preventive treatment, 11,649 rheumatic patients were included in the analysis for identifying risk factors ( Figure 2). The median age of 13,550 patients was 45 [IQR 33-56] years old and 10,286 patients (75.9%) were female. Patients with SLE (2959, 21.8%) and RA (5116, 37.8%) had the largest population, accounting for nearly 60%. General characteristics of participants are shown in Table 2.

Risk factors for ATB
Risk factors for ATB were analyzed in 11,649 subjects who had no history of anti-TNF biologics or TB preventive treatment. Eighty-four of these patients were diagnosed with ATB. The results of the univariate analysis are shown in Table 5    dose ≥30mg/d for more than four weeks within the past two years (OR=2.031, 95% CI: 1.247-3.309, p=0.004) and history of previous TB (OR= 6.185, 95% CI: 3.487-10.969, p<0.001) were independent risk factors of ATB (Table 6).

Discussion
This is the first study conducted using stratified multistage cluster sampling design to estimate the prevalence of ATB and to identify risk factors in patients with rheumatic diseases in China. This will contribute to TB epidemiological data and provide information on TB risk stratification in this patient population, which supports the development of precise TB prevention and management strategies for this vulnerable population. This study revealed that the standardized prevalence of ATB was 882/100,000 in patients with rheumatic diseases, nearly twice as much as the prevalence in the general population (459/100,000) [2]. Patients with different type of rheumatic diseases are varied in the risk of ATB. SLE and BD patients were most likely to have ATB with a prevalence of 3.0 and 4.9 times higher than the general population, respectively. Previous studies reported the prevalence of ATB in patients with SLE ranging from 1.3% to 5.3% [19][20][21][22][23], which were based on retrospective analysis of medical records rather than epidemiological survey. The prevalence of ATB in patients with SLE in Mexico (1.3%) and Colombia (1.4%) was similar to 1386/100,000 in our study. However, the prevalence from Asia (2.0% to 5.3%) was higher than that in this study, and these studies only included hospitalized patients with more severe conditions, which may result in an overestimation of the prevalence due to selection bias. Few studies reported the prevalence of ATB in patients with BD. Our previous study showed that 5.4% of the hospitalized patients with BD were diagnosed with ATB [24]. Similarly, the prevalence of ATB in hospitalized patients with BD was higher than that obtained in this epidemiological survey (2251/100,000). In this study, the prevalence of ATB in RA patients was similar to that in the general population, which may be related to the prophylaxis of tuberculosis in patients initiating anti-TNF treatment [25].
This study explored risk factors associated with ATB in rheumatic patients not taking anti-TNF medications. The independent risk factors for ATB included BD and SLE. BD itself may produce a defect in cell-mediated immunity, which increased the risk of developing ATB [26,27]. Additionally, TB infection may also be involved in inducing BD in individuals with genetic susceptibility [28]. According to a retrospective study from Singapore, SLE was an independent risk factor for ATB (OR = 4.6, p <0.001) [19]. Similarly, a record-linkage study in the United Kingdom found an RR of 9.4 (p <0.001) in SLE patients [29]. The above studies did not consider confounding factors such as medication, but the results are generally in agreement with our analysis.
Current literature suggests that glucocorticoids can suppress cellular immune responses through a variety of mechanisms, including inhibition of lymphokine effect and monocyte chemotaxis, depression of IL-1 and TNF production, and impairment of T cell activation [30]. Some case-control and cohort studies conducted in regions with low to medium burden of TB found a higher risk for TB reactivation in patients receiving corticosteroids (OR/RR of 1.7-4.9) [12][13][14]. Another study from Canada found that the risk of TB reactivation was only significantly increased when taking medium to high doses (≥10mg per day) of glucocorticoids [16]. A retrospective study from China showed that the accumulated doses of glucocorticoid (GC) (OR = 2.32, 95% CI 1.69-3.20, P < 0.001) were associated with TB [20]. Consistent with previous findings, history of long-term and high-dose steroid therapy (≥30mg/d and >4 weeks) was an independent risk factor for ATB in our study, suggesting a similar condition in regions with high burden of TB.  Our study showed that exposure to AZA within the past two years was a risk factor of borderline significant for ATB (OR = 2.095, p =0.054). AZA was found to significantly increase the risk of developing TB in patients with inflammatory bowel disease (RR = 6.27, p =0.046) [31]. A retrospective cohort study showed that the use of AZA was an independent risk factor for ATB after lung transplantation (OR = 10.6, p =0.038) [32]. Few studies evaluated the risk of ATB in patients with rheumatic diseases under treatment of AZA, a systematic review of randomized controlled trials found that the risk of ATB reactivation was higher when anti-TNF agents were combined with AZA as compared with anti-TNF monotherapy (OR = 13.3, p <0.001) [33]. Our study provides more evidence on this issue in patients with rheumatic diseases.
This study also found a significant association between ATB and previous TB infection (OR = 6.185, p <0.001). With the advent of molecular epidemiology, more and more evidence from genotyping studies suggest that recurrent TB is mainly caused by endogenous reactivation [34][35][36]. In our study, it was not possible to identify whether the recurrence of ATB was due to reactivation of latent infection or exogenous reinfection, but undoubtedly more attention associating TB prevention should be paid to patients with previous infection as they are more vulnerable to secondary ATB.
There are some limitations in this study. Firstly, pathogenic evidence was not obtained from two thirds of ATB cases. The molecular techniques for MTB diagnosis had not been approved by the National Medical Products Administration during the study period, so the proportion of TB patients confirmed by microbiology might be underestimated. Although only a third of ATB patients were diagnosed microbiologically, we established a strict diagnostic algorithm, including clinical manifestations, laboratory test results and imaging findings. Most importantly, we followed up patients for 12 weeks to determine whether the diagnostic anti-TB treatment was effective. Moreover, the diagnosis of ATB was independently made by two physicians and a senior doctor was consulted when there was disagreement. These measures increased the accuracy of ATB diagnosis. Secondly, the range of confidence interval was quite wide in certain types of rheumatic diseases due to the small number of patients. Although the point estimates are higher than those of the general population, it needs to be confirmed by further studies in a specific population. Thirdly, several immunosuppressant variables with few observations in ATB group may make the results less robust, it also needs to be confirmed by future studies with large-scale ATB patients using immunosuppressants.
In conclusion, the prevalence of ATB among patients with rheumatic diseases is higher than that of the general population in China. The risk factors including having SLE or BD, taking AZA within the past two years, exposing to long-term medium to high dosage of glucocorticoids and having evidence of previous TB should be considered in the individualized clinical assessment of the risk of ATB in patients with rheumatic diseases not taking anti-TNF medications.