The epidemiology and association rules of concurrent pulmonary tuberculosis and extrapulmonary tuberculosis (PTB-EPTB) in China: a large-scale multi-center observational study

Background Tuberculosis (TB) is a multi-systemic disease with a protean presentation and remains a major global health problem. Concurrent pulmonary tuberculosis (PTB) and extrapulmonary tuberculosis (EPTB) are common in clinical practice. However, the information about concurrent PTB-EPTB is scarce. This study aimed to study the epidemiology of concurrent PTB-EPTB by summarizing the diagnostic types of TB and determine the association rules by a large-scale multi-center observational study in China. Methods The study was performed at 21 hospitals from 15 provinces in China. All the consecutive inpatient with conrmed TB diagnosis during the years from Jan 2011 to Dec 2017 was included in the study. The association rules of concurrent PTB-EPTB were analyzed by Apriori algorithm. Results Of 438,979 TB inpatients evaluated, the most common were PTB (82.05%), followed by tuberculous pleurisy (23.62%), etc. Concurrent PTB-EPTB occurred in 129,422 cases (29.48%). Concurrent PTB and tuberculous pleurisy was the most common concurrent PTB-EPTB types. In the fully adjusted multivariable logistic models, the odds ratio of concurrent PTB-EPTB was different by gender and age group. In PTB with concurrent EPTB, the strongest association rule was PTB with concurrent bronchial tuberculosis (lift=1.09). In EPTB with concurrent PTB, the strongest association rule was pharyngeal /laryngeal tuberculosis with concurrent PTB (lift=1.11). The condence and lift of concurrent PTB-EPTB varied with gender and age. Conclusions There were many types of concurrent PTB-EPTB. The condence and lift of concurrent PTB-EPTB varied with gender and age. The clinicians should be alert to the presence of concurrent PTB-EPTB and take effective treatment regimen. magnitude of association rules of concurrent PTB-EPTB varied with gender. In males in this study, tuberculous empyema with concurrent PTB was the strongest association rule (lift = 1.20). The proportions of tuberculous empyema with concurrent PTB in male were more than 70%. In females, bronchial tuberculosis with concurrent PTB was the strongest association rule (lift = 1.64).The proportion of bronchial tuberculosis with concurrent PTB in females was more than 55%. Cellular immunity, hormones, access to health care, socio-economic factors and cultural factors had been linked to these differences [24–26]. ZYL in the implementation of the investigation. KWL, YZY,MZX, data


Abstract
Background Tuberculosis (TB) is a multi-systemic disease with a protean presentation and remains a major global health problem. Concurrent pulmonary tuberculosis (PTB) and extrapulmonary tuberculosis (EPTB) are common in clinical practice. However, the information about concurrent PTB-EPTB is scarce. This study aimed to study the epidemiology of concurrent PTB-EPTB by summarizing the diagnostic types of TB and determine the association rules by a large-scale multi-center observational study in China.
Methods The study was performed at 21 hospitals from 15 provinces in China. All the consecutive inpatient with con rmed TB diagnosis during the years from Jan 2011 to Dec 2017 was included in the study. The association rules of concurrent PTB-EPTB were analyzed by Apriori algorithm.
Results Of 438,979 TB inpatients evaluated, the most common were PTB (82.05%), followed by tuberculous pleurisy (23.62%), etc. Concurrent PTB-EPTB occurred in 129,422 cases (29.48%). Concurrent PTB and tuberculous pleurisy was the most common concurrent PTB-EPTB types. In the fully adjusted multivariable logistic models, the odds ratio of concurrent PTB-EPTB was different by gender and age group. In PTB with concurrent EPTB, the strongest association rule was PTB with concurrent bronchial tuberculosis (lift=1.09). In EPTB with concurrent PTB, the strongest association rule was pharyngeal /laryngeal tuberculosis with concurrent PTB (lift=1.11). The con dence and lift of concurrent PTB-EPTB varied with gender and age.
Conclusions There were many types of concurrent PTB-EPTB. The con dence and lift of concurrent PTB-EPTB varied with gender and age. The clinicians should be alert to the presence of concurrent PTB-EPTB and take effective treatment regimen. Background Tuberculosis (TB) is an infectious disease caused by the bacillus Mycobacterium tuberculosis. TB remains a major global health problem. It causes ill-health among millions of people each year worldwide. According to the World Health Organization (WHO), the estimated global incidence of TB cases was 10.0 million in 2018 [1]. TB typically affects the lungs (pulmonary TB, PTB) but can also affect other sites (extrapulmonary TB, EPTB), such as pleura, lymph nodes, abdomen, genitourinary tract, skin, joints and bones, meninges, etc [2][3][4][5].
In recent years, considerable efforts have been made to gain a deeper understanding of TB [6][7][8].TB is a multi-systemic disease with a protean presentation. In clinical practice, PTB and EPTB may be present in the same patient [9,10].The treatment of PTB concurrent with EPTB is di cult, and the treatment regimen of some PTB concurrent with EPTB may be different from single PTB or EPTB. However, the information about concurrent PTB-EPTB is scarce. Consequently, summarizing the diagnostic types of TB patients, exploring the epidemiology and association rules of concurrent PTB-EPTB is important. The purpose of the study was to analyze the epidemiology, association rules of concurrent PTB-EPTB, alert clinicians to the presence of concurrent PTB-EPTB and forewarning treatment regimen by a large-scale multi-center observational study.

Study subjects
The study was performed at 21 Hospitals from 15 provinces in China. All the consecutive inpatient with con rmed TB diagnosis during the years from Jan 2011 to Dec 2017 was included in the study. TB was mainly categorized by the lesion site. Diagnosis of TB was made by WHO guidelines [11] and Clinical Diagnosis Standard of TB issued by Chinese Medical Association [12].In general, TB has generally been diagnosed by traditional and modern methods that rely on clinical symptoms, physical signs together with the results of bacteriological methods (including sputum smear microscopy, bacterial culture and molecular diagnostic methods), the tuberculin skin test (TST; puri ed protein derivative (PPD) skin test), X-ray examination results, T-SPOT.TB, Gene Xpert MTB/RIF assay, and successful outcome of treatment with a course of tuberculosis chemotherapy, etc.

Data Management And Statistical Analysis
Measures taken to guarantee the data quality included standardized study protocol and standardized training of research staff. Trained health workers collected medical information by use of a standardized questionnaire. From medical records, we obtained the clinical characteristics of TB inpatients such as age, gender, site of disease etc. The descriptive statistical analysis included frequencies and proportions with 95% con dence intervals (CIs) for categorical variables. Multivariable logistic regression was used to examine the association of gender and age group with the odds ratio of concurrent PTB-EPTB. P < 0.05 was the threshold for statistical signi cance.
Analysis by association rules is used for discovering relationships hidden in large databases. The technique was developed in computer science and has been used in a variety of elds [13][14][15].The Apriori algorithm provides a way of applying a set of association rules in data mining. The principle of Apriori is based on two steps. The rst step searches for item sets that exceed the minimum support, while in the second step, association rules are generated and ltered by selecting "con dence" item sets (based on a threshold) from those found in the rst step [16,17]. If the association rule of A concurrent with B, support, con dence and lift were de ned as: Support = P (A), Con dence = P(B|A), Lift = P(A∩B)/[P(A)*P(B)]. A is antecedent and B is consequent. Lift was used to evaluate the magnitude of association rules. Lift > 1 indicate a positive association rule. The association rules for concurrent PTB and the diverse types of TB patient characteristics A total of 438,979 TB inpatients were included from Jan 2011 to Dec 2017 at 21 hospitals from 15 provinces in China, most of which were specialized tuberculosis hospitals (Fig. 1). The ratio of male: female was 1.83. There were 83 kinds of tuberculous lesions involved in 604,114 sites in the 438,979 TB inpatients. On average each TB inpatient had 1.38 TB lesion types. Among the 438,979 TB inpatient cases, the most common types of TB were PTB (82.05%, 95%CI: 81.94%-82.16%), followed by tuberculous pleurisy (23.62%, 95%CI: 23.49%-23.74%), bronchial tuberculosis (7.01%, 95%CI: 6.94%-7.09%), etc. The types of TB proportion ≥ 0.1% cases were shown in Table 1.  Figure S1 in the Supplementary Appendix. In each age group, the number of concurrent PTB-EPTB cases in males was more than in females. In the fully adjusted multivariable logistic models, female (OR = 1.119, 95%CI:  The most common of concurrent PTB-EPTB types According to the association rules analysis of concurrent PTB-EPTB, the TOP 20 most common of concurrent PTB-EPTB were listed in Table S1 in the Supplementary Appendix, sort by cases. Concurrent PTB and tuberculous pleurisy (15.35%, 95%CI: 15.25%-15.46%), concurrent PTB and bronchial tuberculosis (6.28%, 95%CI: 6.20%-6.35%) were more than others concurrent PTB-EPTB types.
The association rules analysis of concurrent PTB-EPTB In order to nd most of the possible association rules with Antecedent = PTB, the minimum con dence degree was set as 1.00%. After executing the association model, six association rules were obtained. The association rules were shown in Table 3, sorted by con dence. The rst rule row (ID = 1) in Table 3 was interpreted as showing that with PTB totaling 360,187 cases (Instances), PTB accounted for 82.05% of all TB cases (Support), PTB with concurrent tuberculous pleurisy accounting for 18.71% of PTB cases (Con dence). The con dence of concurrent bronchial tuberculosis in PTB cases was the next highest (7.65%), followed by tuberculous meningitis (2.72%), etc. The strongest association rule in PTB with concurrent EPTB was PTB with concurrent bronchial tuberculosis (lift = 1.09). The lift value of 1.09 means PTB was positively associated with bronchial tuberculosis. In order to nd most of the possible association rules with Consequent = PTB, the minimum support degree was set as 0.1% and the minimum con dence degree was set as 40%. After executing the association model, 22 association rules were obtained, including ve rules with con dence above 70%. The association rules were shown in Table 4, sorted by con dence. The rst rule row (ID = 1) in Table 4 was interpreted as showing that 2,382 cases (Instances) of pharyngeal/laryngeal tuberculosis accounted for 0.54% of all TB cases (Support), pharyngeal /laryngeal tuberculosis with concurrent PTB accounted for 91.23% of pharyngeal/laryngeal tuberculosis cases (Con dence).The con dence of concurrent PTB in bronchial tuberculosis cases was the next highest (89.51%), followed by tuberculosis of mediastinal lymph nodes (77.57%), etc. The strongest association rule in EPTB with concurrent PTB were pharyngeal/laryngeal tuberculosis (lift = 1.11). That means pharyngeal/ laryngeal tuberculosis were positively associated with PTB. Notes: The rst column represents the consequents (the "then" part of the rule), while the next column represents the antecedents (the "if" part of the rule).
ID displays the sequence of the association rules.
Instances display the cases of TB.

*: lift > 1
The association rules of concurrent PTB-EPTB types with gender Most types of TB can be found both in males and females, with obvious exceptions such as tuberculosis of ovary, oviduct tuberculosis etc. We found association rules in males and females through setting the minimum support degree and the minimum con dence degree (Tables S2&3 in the Supplementary Appendix). In males, tuberculous empyema with concurrent PTB was the strongest association rule (lift = 1.20), followed by costal tuberculosis with concurrent PTB (lift = 1.16), etc. In females, bronchial tuberculosis with concurrent PTB was the strongest association rule (lift = 1.64), followed by supraclavicular lymph node tuberculosis with concurrent PTB (lift = 1.56), etc.
The association rules of concurrent PTB-EPTB types with age We found association rules in all age groups through setting the minimum support degree and the minimum con dence degree (Tables S4 ~ S10 in

Discussion
TB is spread when people who are sick with PTB expel the bacteria into the air. When a person breathes in TB bacteria, the bacteria can settle in the lungs and begin to grow. From there, they can disseminate through the lymphatic or hematogenous systems and subsequently affect single or multiple extrapulmonary sites, such as the pleura, lymph nodes, meninges, bones and joints, etc. PTB is the most common presentation of TB. However, EPTB contributes considerably to morbidity, lifelong sequelae, and mortality [18]. The mechanisms for EPTB dissemination are complicated [19]. PTB concurrent with EPTB are common in clinical practice, but the information about concurrent PTB-EPTB is scarce. Given the variety of clinical presentations and the nonspeci c systemic symptoms of TB, a more profound understanding of the site distribution of TB should be sought. In this study, we summarized the diagnostic types of TB and explored the association rules of concurrent PTB-EPTB so as to alert the clinicians to the presence of concurrent PTB-EPTB by a large TB sample.
Tuberculous pleurisy was one of the most common forms of EPTB [20].We found tuberculous pleurisy (23.62%) was the second most common types of TB.
Tuberculous pleurisy is thought to represent primarily a hypersensitivity reaction to tuberculous protein [21]. Previous studies have also noted that concurrent PTB-EPTB patients [9,10]. Boonsarngsuk et al [9]. demonstrated that 12.2% were of concurrent PTB-EPTB(120/986). In this study, concurrent PTB-EPTB occurred in about 30% in TB patients. We also found that the strongest association rule of PTB concurrent with EPTB was PTB concurrent with bronchial tuberculosis (lift = 1.09). In this study, 7.65% PTB had concurrent bronchial tuberculosis. Because bronchi are adjacent to the lung, PTB is prone to cause bronchial tuberculosis. But this is not inevitable, hence the proportion of EPTB with concurrent PTB was different depending on whether patients were primarily viewed as EPTB or PTB patients. Laryngeal tuberculosis was an infrequent manifestation of EPTB. Usually, it was seen as a complication of PTB [22]. The strongest association rule of EPTB concurrent with PTB was pharyngeal/ laryngeal tuberculosis concurrent with PTB (lift = 1.11). In this study, 91.23% of the pharyngeal/laryngeal tuberculosis patients had concurrent PTB. Expulsion of Mycobacterium tuberculosis by PTB patients is liable to affect the larynx. Consistent with this, most of laryngeal tuberculosis patients had concurrent PTB. These ndings suggested that once PTB was diagnosed, bronchial tuberculosis should be considered rst. Furthermore, once pharyngeal/laryngeal tuberculosis or bronchial tuberculosis was con rmed, attention should be paid to PTB. In essence, the diagnosis of TB must be carefully undertaken to avoid misdiagnosis.
Most types of tuberculous lesions can be found both in males and females. But female (OR = 1.119, 95%CI: 1.104-1.134) was more likely to have concurrent PTB-EPTB than male in this study. Jung et al. [23] also found female gender was an independent predictor of concomitant EBTB in patients with active PTB (OR = 4.35, 95%CI: 1.78-10.63). The magnitude of association rules of concurrent PTB-EPTB varied with gender. In males in this study, tuberculous empyema with concurrent PTB was the strongest association rule (lift = 1.20). The proportions of tuberculous empyema with concurrent PTB in male were more than 70%. In females, bronchial tuberculosis with concurrent PTB was the strongest association rule (lift = 1.64).The proportion of bronchial tuberculosis with concurrent PTB in females was more than 55%. Cellular immunity, hormones, access to health care, socio-economic factors and cultural factors had been linked to these differences [24][25][26].
TB affects all age groups, but overall the best estimate for 2018 was that about 90% of cases were adults (aged ≥ 15 years). The prevalence of TB is strongly associated with age [1]. Our research found that the strongest association rule in children and adolescents (< 15 years) were PTB with concurrent tuberculous meningitis (lift = 3.89). This may be also related to the particular physiological characteristics and the immunological mechanisms in children and adolescents. We also found that the strongest association rule in 15-24 year group of TB patients was splenic tuberculosis with concurrent PTB (lift = 2.23). This study had several strengths including its large-scale multi-center representative sample, and detailed analysis of the diagnostic types of TB and the con dence/lift of concurrent PTB-EPTB for the rst time in the world. There were several limitations to our study. First, our study may have been subject to Berkson bias. The study population was hospitalized TB patients. There is a high likelihood that the concurrent PTB-EPTB would have more possibility to be hospitalized, which may overestimate the proportions in the population at large. Therefore, data collected from whole population-based studies will be needed to clarify the associations. Secondly, most of the hospitals in our study were TB-specialized hospitals. Therefore, these ndings may not represent the general In conclusions, our present study found many types of concurrent PTB-EPTB and analyzed the association rules between PTB and EPTB for the rst time in the world by a large sample. The concurrent PTB and tuberculous pleurisy was the most common types of concurrent PTB-EPTB. The strongest association rule in PTB with concurrent EPTB was PTB with concurrent bronchial tuberculosis. The strongest association rule in EPTB with concurrent PTB was pharyngeal/laryngeal tuberculosis with concurrent PTB. The con dence and lift of concurrent PTB-EPTB varied with gender and age. The clinicians should be alert to the presence of concurrent PTB-EPTB and take effective treatment regimen to treat the patients.

Declarations
Ethical Approval and Consent to participate: Given that the medical information of inpatients was recorded anonymously by case history, which would not bring any risk to the participants, the Ethics Committee of Beijing Chest Hospital, Capital Medical University approved this study with a waiver of informed consent from the patients.
Consent for publication: Not applicable.
Availability of supporting data: Data are not publicly available. However, de-identi ed data can be obtained on a reasonable request to correspondence.
Competing interests: The authors declare that they have no competing interests.